xref: /freebsd/sys/netinet6/ip6_output.c (revision bc7512cc58af2e8bbe5bbf5ca0059b1daa1da897)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_kern_tls.h"
72 #include "opt_ratelimit.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 #include "opt_sctp.h"
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/ktls.h>
80 #include <sys/malloc.h>
81 #include <sys/mbuf.h>
82 #include <sys/errno.h>
83 #include <sys/priv.h>
84 #include <sys/proc.h>
85 #include <sys/protosw.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/syslog.h>
89 #include <sys/ucred.h>
90 
91 #include <machine/in_cksum.h>
92 
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_vlan_var.h>
96 #include <net/if_llatbl.h>
97 #include <net/ethernet.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/route/nhop.h>
101 #include <net/pfil.h>
102 #include <net/rss_config.h>
103 #include <net/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip_var.h>
108 #include <netinet6/in6_fib.h>
109 #include <netinet6/in6_var.h>
110 #include <netinet/ip6.h>
111 #include <netinet/icmp6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/tcp_var.h>
115 #include <netinet6/nd6.h>
116 #include <netinet6/in6_rss.h>
117 
118 #include <netipsec/ipsec_support.h>
119 #if defined(SCTP) || defined(SCTP_SUPPORT)
120 #include <netinet/sctp.h>
121 #include <netinet/sctp_crc32.h>
122 #endif
123 
124 #include <netinet6/ip6protosw.h>
125 #include <netinet6/scope6_var.h>
126 
127 extern int in6_mcast_loop;
128 
129 struct ip6_exthdrs {
130 	struct mbuf *ip6e_ip6;
131 	struct mbuf *ip6e_hbh;
132 	struct mbuf *ip6e_dest1;
133 	struct mbuf *ip6e_rthdr;
134 	struct mbuf *ip6e_dest2;
135 };
136 
137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
138 
139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
140 			   struct ucred *, int);
141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
142 	struct socket *, struct sockopt *);
143 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
145 	struct ucred *, int, int, int);
146 
147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
149 	struct ip6_frag **);
150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
152 static int ip6_getpmtu(struct route_in6 *, int,
153 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
154 	u_int);
155 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
156 	u_long *, int *, u_int);
157 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
158 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
159 
160 /*
161  * Make an extension header from option data.  hp is the source,
162  * mp is the destination, and _ol is the optlen.
163  */
164 #define	MAKE_EXTHDR(hp, mp, _ol)					\
165     do {								\
166 	if (hp) {							\
167 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
168 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
169 		    ((eh)->ip6e_len + 1) << 3);				\
170 		if (error)						\
171 			goto freehdrs;					\
172 		(_ol) += (*(mp))->m_len;				\
173 	}								\
174     } while (/*CONSTCOND*/ 0)
175 
176 /*
177  * Form a chain of extension headers.
178  * m is the extension header mbuf
179  * mp is the previous mbuf in the chain
180  * p is the next header
181  * i is the type of option.
182  */
183 #define MAKE_CHAIN(m, mp, p, i)\
184     do {\
185 	if (m) {\
186 		if (!hdrsplit) \
187 			panic("%s:%d: assumption failed: "\
188 			    "hdr not split: hdrsplit %d exthdrs %p",\
189 			    __func__, __LINE__, hdrsplit, &exthdrs);\
190 		*mtod((m), u_char *) = *(p);\
191 		*(p) = (i);\
192 		p = mtod((m), u_char *);\
193 		(m)->m_next = (mp)->m_next;\
194 		(mp)->m_next = (m);\
195 		(mp) = (m);\
196 	}\
197     } while (/*CONSTCOND*/ 0)
198 
199 void
200 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
201 {
202 	u_short csum;
203 
204 	csum = in_cksum_skip(m, offset + plen, offset);
205 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
206 		csum = 0xffff;
207 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
208 
209 	if (offset + sizeof(csum) > m->m_len)
210 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
211 	else
212 		*(u_short *)mtodo(m, offset) = csum;
213 }
214 
215 static void
216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
217     int plen, int optlen)
218 {
219 
220 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
221 	    "csum_flags %#x",
222 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
223 
224 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
225 		in6_delayed_cksum(m, plen - optlen,
226 		    sizeof(struct ip6_hdr) + optlen);
227 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
228 	}
229 #if defined(SCTP) || defined(SCTP_SUPPORT)
230 	if (csum_flags & CSUM_SCTP_IPV6) {
231 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
232 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
233 	}
234 #endif
235 }
236 
237 int
238 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
239     int fraglen , uint32_t id)
240 {
241 	struct mbuf *m, **mnext, *m_frgpart;
242 	struct ip6_hdr *ip6, *mhip6;
243 	struct ip6_frag *ip6f;
244 	int off;
245 	int error;
246 	int tlen = m0->m_pkthdr.len;
247 
248 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
249 
250 	m = m0;
251 	ip6 = mtod(m, struct ip6_hdr *);
252 	mnext = &m->m_nextpkt;
253 
254 	for (off = hlen; off < tlen; off += fraglen) {
255 		m = m_gethdr(M_NOWAIT, MT_DATA);
256 		if (!m) {
257 			IP6STAT_INC(ip6s_odropped);
258 			return (ENOBUFS);
259 		}
260 
261 		/*
262 		 * Make sure the complete packet header gets copied
263 		 * from the originating mbuf to the newly created
264 		 * mbuf. This also ensures that existing firewall
265 		 * classification(s), VLAN tags and so on get copied
266 		 * to the resulting fragmented packet(s):
267 		 */
268 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
269 			m_free(m);
270 			IP6STAT_INC(ip6s_odropped);
271 			return (ENOBUFS);
272 		}
273 
274 		*mnext = m;
275 		mnext = &m->m_nextpkt;
276 		m->m_data += max_linkhdr;
277 		mhip6 = mtod(m, struct ip6_hdr *);
278 		*mhip6 = *ip6;
279 		m->m_len = sizeof(*mhip6);
280 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
281 		if (error) {
282 			IP6STAT_INC(ip6s_odropped);
283 			return (error);
284 		}
285 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
286 		if (off + fraglen >= tlen)
287 			fraglen = tlen - off;
288 		else
289 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
290 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
291 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
292 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
293 			IP6STAT_INC(ip6s_odropped);
294 			return (ENOBUFS);
295 		}
296 		m_cat(m, m_frgpart);
297 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
298 		ip6f->ip6f_reserved = 0;
299 		ip6f->ip6f_ident = id;
300 		ip6f->ip6f_nxt = nextproto;
301 		IP6STAT_INC(ip6s_ofragments);
302 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
303 	}
304 
305 	return (0);
306 }
307 
308 static int
309 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
310     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
311     bool stamp_tag)
312 {
313 #ifdef KERN_TLS
314 	struct ktls_session *tls = NULL;
315 #endif
316 	struct m_snd_tag *mst;
317 	int error;
318 
319 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
320 	mst = NULL;
321 
322 #ifdef KERN_TLS
323 	/*
324 	 * If this is an unencrypted TLS record, save a reference to
325 	 * the record.  This local reference is used to call
326 	 * ktls_output_eagain after the mbuf has been freed (thus
327 	 * dropping the mbuf's reference) in if_output.
328 	 */
329 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
330 		tls = ktls_hold(m->m_next->m_epg_tls);
331 		mst = tls->snd_tag;
332 
333 		/*
334 		 * If a TLS session doesn't have a valid tag, it must
335 		 * have had an earlier ifp mismatch, so drop this
336 		 * packet.
337 		 */
338 		if (mst == NULL) {
339 			m_freem(m);
340 			error = EAGAIN;
341 			goto done;
342 		}
343 		/*
344 		 * Always stamp tags that include NIC ktls.
345 		 */
346 		stamp_tag = true;
347 	}
348 #endif
349 #ifdef RATELIMIT
350 	if (inp != NULL && mst == NULL) {
351 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
352 		    (inp->inp_snd_tag != NULL &&
353 		    inp->inp_snd_tag->ifp != ifp))
354 			in_pcboutput_txrtlmt(inp, ifp, m);
355 
356 		if (inp->inp_snd_tag != NULL)
357 			mst = inp->inp_snd_tag;
358 	}
359 #endif
360 	if (stamp_tag && mst != NULL) {
361 		KASSERT(m->m_pkthdr.rcvif == NULL,
362 		    ("trying to add a send tag to a forwarded packet"));
363 		if (mst->ifp != ifp) {
364 			m_freem(m);
365 			error = EAGAIN;
366 			goto done;
367 		}
368 
369 		/* stamp send tag on mbuf */
370 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
371 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
372 	}
373 
374 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
375 
376 done:
377 	/* Check for route change invalidating send tags. */
378 #ifdef KERN_TLS
379 	if (tls != NULL) {
380 		if (error == EAGAIN)
381 			error = ktls_output_eagain(inp, tls);
382 		ktls_free(tls);
383 	}
384 #endif
385 #ifdef RATELIMIT
386 	if (error == EAGAIN)
387 		in_pcboutput_eagain(inp);
388 #endif
389 	return (error);
390 }
391 
392 /*
393  * IP6 output.
394  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
395  * nxt, hlim, src, dst).
396  * This function may modify ver and hlim only.
397  * The mbuf chain containing the packet will be freed.
398  * The mbuf opt, if present, will not be freed.
399  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
400  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
401  * then result of route lookup is stored in ro->ro_nh.
402  *
403  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
404  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
405  *
406  * ifpp - XXX: just for statistics
407  */
408 int
409 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
410     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
411     struct ifnet **ifpp, struct inpcb *inp)
412 {
413 	struct ip6_hdr *ip6;
414 	struct ifnet *ifp, *origifp;
415 	struct mbuf *m = m0;
416 	struct mbuf *mprev;
417 	struct route_in6 *ro_pmtu;
418 	struct nhop_object *nh;
419 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
420 	struct in6_addr odst;
421 	u_char *nexthdrp;
422 	int tlen, len;
423 	int error = 0;
424 	int vlan_pcp = -1;
425 	struct in6_ifaddr *ia = NULL;
426 	u_long mtu;
427 	int alwaysfrag, dontfrag;
428 	u_int32_t optlen, plen = 0, unfragpartlen;
429 	struct ip6_exthdrs exthdrs;
430 	struct in6_addr src0, dst0;
431 	u_int32_t zone;
432 	bool hdrsplit;
433 	int sw_csum, tso;
434 	int needfiblookup;
435 	uint32_t fibnum;
436 	struct m_tag *fwd_tag = NULL;
437 	uint32_t id;
438 
439 	NET_EPOCH_ASSERT();
440 
441 	if (inp != NULL) {
442 		INP_LOCK_ASSERT(inp);
443 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
444 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
445 			/* Unconditionally set flowid. */
446 			m->m_pkthdr.flowid = inp->inp_flowid;
447 			M_HASHTYPE_SET(m, inp->inp_flowtype);
448 		}
449 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
450 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
451 			    INP_2PCP_SHIFT;
452 #ifdef NUMA
453 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
454 #endif
455 	}
456 
457 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
458 	/*
459 	 * IPSec checking which handles several cases.
460 	 * FAST IPSEC: We re-injected the packet.
461 	 * XXX: need scope argument.
462 	 */
463 	if (IPSEC_ENABLED(ipv6)) {
464 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
465 			if (error == EINPROGRESS)
466 				error = 0;
467 			goto done;
468 		}
469 	}
470 #endif /* IPSEC */
471 
472 	/* Source address validation. */
473 	ip6 = mtod(m, struct ip6_hdr *);
474 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
475 	    (flags & IPV6_UNSPECSRC) == 0) {
476 		error = EOPNOTSUPP;
477 		IP6STAT_INC(ip6s_badscope);
478 		goto bad;
479 	}
480 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
481 		error = EOPNOTSUPP;
482 		IP6STAT_INC(ip6s_badscope);
483 		goto bad;
484 	}
485 
486 	/*
487 	 * If we are given packet options to add extension headers prepare them.
488 	 * Calculate the total length of the extension header chain.
489 	 * Keep the length of the unfragmentable part for fragmentation.
490 	 */
491 	bzero(&exthdrs, sizeof(exthdrs));
492 	optlen = 0;
493 	unfragpartlen = sizeof(struct ip6_hdr);
494 	if (opt) {
495 		/* Hop-by-Hop options header. */
496 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
497 
498 		/* Destination options header (1st part). */
499 		if (opt->ip6po_rthdr) {
500 #ifndef RTHDR_SUPPORT_IMPLEMENTED
501 			/*
502 			 * If there is a routing header, discard the packet
503 			 * right away here. RH0/1 are obsolete and we do not
504 			 * currently support RH2/3/4.
505 			 * People trying to use RH253/254 may want to disable
506 			 * this check.
507 			 * The moment we do support any routing header (again)
508 			 * this block should check the routing type more
509 			 * selectively.
510 			 */
511 			error = EINVAL;
512 			goto bad;
513 #endif
514 
515 			/*
516 			 * Destination options header (1st part).
517 			 * This only makes sense with a routing header.
518 			 * See Section 9.2 of RFC 3542.
519 			 * Disabling this part just for MIP6 convenience is
520 			 * a bad idea.  We need to think carefully about a
521 			 * way to make the advanced API coexist with MIP6
522 			 * options, which might automatically be inserted in
523 			 * the kernel.
524 			 */
525 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
526 			    optlen);
527 		}
528 		/* Routing header. */
529 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
530 
531 		unfragpartlen += optlen;
532 
533 		/*
534 		 * NOTE: we don't add AH/ESP length here (done in
535 		 * ip6_ipsec_output()).
536 		 */
537 
538 		/* Destination options header (2nd part). */
539 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
540 	}
541 
542 	/*
543 	 * If there is at least one extension header,
544 	 * separate IP6 header from the payload.
545 	 */
546 	hdrsplit = false;
547 	if (optlen) {
548 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
549 			m = NULL;
550 			goto freehdrs;
551 		}
552 		m = exthdrs.ip6e_ip6;
553 		ip6 = mtod(m, struct ip6_hdr *);
554 		hdrsplit = true;
555 	}
556 
557 	/* Adjust mbuf packet header length. */
558 	m->m_pkthdr.len += optlen;
559 	plen = m->m_pkthdr.len - sizeof(*ip6);
560 
561 	/* If this is a jumbo payload, insert a jumbo payload option. */
562 	if (plen > IPV6_MAXPACKET) {
563 		if (!hdrsplit) {
564 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
565 				m = NULL;
566 				goto freehdrs;
567 			}
568 			m = exthdrs.ip6e_ip6;
569 			ip6 = mtod(m, struct ip6_hdr *);
570 			hdrsplit = true;
571 		}
572 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
573 			goto freehdrs;
574 		ip6->ip6_plen = 0;
575 	} else
576 		ip6->ip6_plen = htons(plen);
577 	nexthdrp = &ip6->ip6_nxt;
578 
579 	if (optlen) {
580 		/*
581 		 * Concatenate headers and fill in next header fields.
582 		 * Here we have, on "m"
583 		 *	IPv6 payload
584 		 * and we insert headers accordingly.
585 		 * Finally, we should be getting:
586 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
587 		 *
588 		 * During the header composing process "m" points to IPv6
589 		 * header.  "mprev" points to an extension header prior to esp.
590 		 */
591 		mprev = m;
592 
593 		/*
594 		 * We treat dest2 specially.  This makes IPsec processing
595 		 * much easier.  The goal here is to make mprev point the
596 		 * mbuf prior to dest2.
597 		 *
598 		 * Result: IPv6 dest2 payload.
599 		 * m and mprev will point to IPv6 header.
600 		 */
601 		if (exthdrs.ip6e_dest2) {
602 			if (!hdrsplit)
603 				panic("%s:%d: assumption failed: "
604 				    "hdr not split: hdrsplit %d exthdrs %p",
605 				    __func__, __LINE__, hdrsplit, &exthdrs);
606 			exthdrs.ip6e_dest2->m_next = m->m_next;
607 			m->m_next = exthdrs.ip6e_dest2;
608 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
609 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
610 		}
611 
612 		/*
613 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
614 		 * m will point to IPv6 header.  mprev will point to the
615 		 * extension header prior to dest2 (rthdr in the above case).
616 		 */
617 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
618 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
619 			   IPPROTO_DSTOPTS);
620 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
621 			   IPPROTO_ROUTING);
622 	}
623 
624 	IP6STAT_INC(ip6s_localout);
625 
626 	/* Route packet. */
627 	ro_pmtu = ro;
628 	if (opt && opt->ip6po_rthdr)
629 		ro = &opt->ip6po_route;
630 	if (ro != NULL)
631 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
632 	else
633 		dst = &sin6;
634 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
635 
636 again:
637 	/*
638 	 * If specified, try to fill in the traffic class field.
639 	 * Do not override if a non-zero value is already set.
640 	 * We check the diffserv field and the ECN field separately.
641 	 */
642 	if (opt && opt->ip6po_tclass >= 0) {
643 		int mask = 0;
644 
645 		if (IPV6_DSCP(ip6) == 0)
646 			mask |= 0xfc;
647 		if (IPV6_ECN(ip6) == 0)
648 			mask |= 0x03;
649 		if (mask != 0)
650 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
651 	}
652 
653 	/* Fill in or override the hop limit field, if necessary. */
654 	if (opt && opt->ip6po_hlim != -1)
655 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
656 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
657 		if (im6o != NULL)
658 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
659 		else
660 			ip6->ip6_hlim = V_ip6_defmcasthlim;
661 	}
662 
663 	if (ro == NULL || ro->ro_nh == NULL) {
664 		bzero(dst, sizeof(*dst));
665 		dst->sin6_family = AF_INET6;
666 		dst->sin6_len = sizeof(*dst);
667 		dst->sin6_addr = ip6->ip6_dst;
668 	}
669 	/*
670 	 * Validate route against routing table changes.
671 	 * Make sure that the address family is set in route.
672 	 */
673 	nh = NULL;
674 	ifp = NULL;
675 	mtu = 0;
676 	if (ro != NULL) {
677 		if (ro->ro_nh != NULL && inp != NULL) {
678 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
679 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
680 			    fibnum);
681 		}
682 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
683 		    (!NH_IS_VALID(ro->ro_nh) ||
684 		    ro->ro_dst.sin6_family != AF_INET6 ||
685 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
686 			RO_INVALIDATE_CACHE(ro);
687 
688 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
689 		    ro->ro_dst.sin6_family == AF_INET6 &&
690 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
691 			nh = ro->ro_nh;
692 			ifp = nh->nh_ifp;
693 		} else {
694 			if (ro->ro_lle)
695 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
696 			ro->ro_lle = NULL;
697 			if (fwd_tag == NULL) {
698 				bzero(&dst_sa, sizeof(dst_sa));
699 				dst_sa.sin6_family = AF_INET6;
700 				dst_sa.sin6_len = sizeof(dst_sa);
701 				dst_sa.sin6_addr = ip6->ip6_dst;
702 			}
703 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
704 			    &nh, fibnum, m->m_pkthdr.flowid);
705 			if (error != 0) {
706 				IP6STAT_INC(ip6s_noroute);
707 				if (ifp != NULL)
708 					in6_ifstat_inc(ifp, ifs6_out_discard);
709 				goto bad;
710 			}
711 			if (ifp != NULL)
712 			    mtu = ifp->if_mtu;
713 		}
714 		if (nh == NULL) {
715 			/*
716 			 * If in6_selectroute() does not return a nexthop
717 			 * dst may not have been updated.
718 			 */
719 			*dst = dst_sa;	/* XXX */
720 		} else {
721 			if (nh->nh_flags & NHF_HOST)
722 			    mtu = nh->nh_mtu;
723 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
724 			counter_u64_add(nh->nh_pksent, 1);
725 		}
726 	} else {
727 		struct nhop_object *nh;
728 		struct in6_addr kdst;
729 		uint32_t scopeid;
730 
731 		if (fwd_tag == NULL) {
732 			bzero(&dst_sa, sizeof(dst_sa));
733 			dst_sa.sin6_family = AF_INET6;
734 			dst_sa.sin6_len = sizeof(dst_sa);
735 			dst_sa.sin6_addr = ip6->ip6_dst;
736 		}
737 
738 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
739 		    im6o != NULL &&
740 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
741 			/* We do not need a route lookup. */
742 			*dst = dst_sa;	/* XXX */
743 			goto nonh6lookup;
744 		}
745 
746 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
747 
748 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
749 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
750 			if (scopeid > 0) {
751 				ifp = in6_getlinkifnet(scopeid);
752 				if (ifp == NULL) {
753 					error = EHOSTUNREACH;
754 					goto bad;
755 				}
756 				*dst = dst_sa;	/* XXX */
757 				goto nonh6lookup;
758 			}
759 		}
760 
761 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
762 		    m->m_pkthdr.flowid);
763 		if (nh == NULL) {
764 			IP6STAT_INC(ip6s_noroute);
765 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
766 			error = EHOSTUNREACH;;
767 			goto bad;
768 		}
769 
770 		ifp = nh->nh_ifp;
771 		mtu = nh->nh_mtu;
772 		ia = ifatoia6(nh->nh_ifa);
773 		if (nh->nh_flags & NHF_GATEWAY)
774 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
775 		else if (fwd_tag != NULL)
776 			dst->sin6_addr = dst_sa.sin6_addr;
777 nonh6lookup:
778 		;
779 	}
780 
781 	/* Then nh (for unicast) and ifp must be non-NULL valid values. */
782 	if ((flags & IPV6_FORWARDING) == 0) {
783 		/* XXX: the FORWARDING flag can be set for mrouting. */
784 		in6_ifstat_inc(ifp, ifs6_out_request);
785 	}
786 
787 	/* Setup data structures for scope ID checks. */
788 	src0 = ip6->ip6_src;
789 	bzero(&src_sa, sizeof(src_sa));
790 	src_sa.sin6_family = AF_INET6;
791 	src_sa.sin6_len = sizeof(src_sa);
792 	src_sa.sin6_addr = ip6->ip6_src;
793 
794 	dst0 = ip6->ip6_dst;
795 	/* Re-initialize to be sure. */
796 	bzero(&dst_sa, sizeof(dst_sa));
797 	dst_sa.sin6_family = AF_INET6;
798 	dst_sa.sin6_len = sizeof(dst_sa);
799 	dst_sa.sin6_addr = ip6->ip6_dst;
800 
801 	/* Check for valid scope ID. */
802 	if (in6_setscope(&src0, ifp, &zone) == 0 &&
803 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
804 	    in6_setscope(&dst0, ifp, &zone) == 0 &&
805 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
806 		/*
807 		 * The outgoing interface is in the zone of the source
808 		 * and destination addresses.
809 		 *
810 		 * Because the loopback interface cannot receive
811 		 * packets with a different scope ID than its own,
812 		 * there is a trick to pretend the outgoing packet
813 		 * was received by the real network interface, by
814 		 * setting "origifp" different from "ifp". This is
815 		 * only allowed when "ifp" is a loopback network
816 		 * interface. Refer to code in nd6_output_ifp() for
817 		 * more details.
818 		 */
819 		origifp = ifp;
820 
821 		/*
822 		 * We should use ia_ifp to support the case of sending
823 		 * packets to an address of our own.
824 		 */
825 		if (ia != NULL && ia->ia_ifp)
826 			ifp = ia->ia_ifp;
827 
828 	} else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
829 	    sa6_recoverscope(&src_sa) != 0 ||
830 	    sa6_recoverscope(&dst_sa) != 0 ||
831 	    dst_sa.sin6_scope_id == 0 ||
832 	    (src_sa.sin6_scope_id != 0 &&
833 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
834 	    (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
835 		/*
836 		 * If the destination network interface is not a
837 		 * loopback interface, or the destination network
838 		 * address has no scope ID, or the source address has
839 		 * a scope ID set which is different from the
840 		 * destination address one, or there is no network
841 		 * interface representing this scope ID, the address
842 		 * pair is considered invalid.
843 		 */
844 		IP6STAT_INC(ip6s_badscope);
845 		in6_ifstat_inc(ifp, ifs6_out_discard);
846 		if (error == 0)
847 			error = EHOSTUNREACH; /* XXX */
848 		goto bad;
849 	}
850 	/* All scope ID checks are successful. */
851 
852 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
853 		if (opt && opt->ip6po_nextroute.ro_nh) {
854 			/*
855 			 * The nexthop is explicitly specified by the
856 			 * application.  We assume the next hop is an IPv6
857 			 * address.
858 			 */
859 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
860 		}
861 		else if ((nh->nh_flags & NHF_GATEWAY))
862 			dst = &nh->gw6_sa;
863 	}
864 
865 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
866 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
867 	} else {
868 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
869 		in6_ifstat_inc(ifp, ifs6_out_mcast);
870 
871 		/* Confirm that the outgoing interface supports multicast. */
872 		if (!(ifp->if_flags & IFF_MULTICAST)) {
873 			IP6STAT_INC(ip6s_noroute);
874 			in6_ifstat_inc(ifp, ifs6_out_discard);
875 			error = ENETUNREACH;
876 			goto bad;
877 		}
878 		if ((im6o == NULL && in6_mcast_loop) ||
879 		    (im6o && im6o->im6o_multicast_loop)) {
880 			/*
881 			 * Loop back multicast datagram if not expressly
882 			 * forbidden to do so, even if we have not joined
883 			 * the address; protocols will filter it later,
884 			 * thus deferring a hash lookup and lock acquisition
885 			 * at the expense of an m_copym().
886 			 */
887 			ip6_mloopback(ifp, m);
888 		} else {
889 			/*
890 			 * If we are acting as a multicast router, perform
891 			 * multicast forwarding as if the packet had just
892 			 * arrived on the interface to which we are about
893 			 * to send.  The multicast forwarding function
894 			 * recursively calls this function, using the
895 			 * IPV6_FORWARDING flag to prevent infinite recursion.
896 			 *
897 			 * Multicasts that are looped back by ip6_mloopback(),
898 			 * above, will be forwarded by the ip6_input() routine,
899 			 * if necessary.
900 			 */
901 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
902 				/*
903 				 * XXX: ip6_mforward expects that rcvif is NULL
904 				 * when it is called from the originating path.
905 				 * However, it may not always be the case.
906 				 */
907 				m->m_pkthdr.rcvif = NULL;
908 				if (ip6_mforward(ip6, ifp, m) != 0) {
909 					m_freem(m);
910 					goto done;
911 				}
912 			}
913 		}
914 		/*
915 		 * Multicasts with a hoplimit of zero may be looped back,
916 		 * above, but must not be transmitted on a network.
917 		 * Also, multicasts addressed to the loopback interface
918 		 * are not sent -- the above call to ip6_mloopback() will
919 		 * loop back a copy if this host actually belongs to the
920 		 * destination group on the loopback interface.
921 		 */
922 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
923 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
924 			m_freem(m);
925 			goto done;
926 		}
927 	}
928 
929 	/*
930 	 * Fill the outgoing inteface to tell the upper layer
931 	 * to increment per-interface statistics.
932 	 */
933 	if (ifpp)
934 		*ifpp = ifp;
935 
936 	/* Determine path MTU. */
937 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
938 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
939 		goto bad;
940 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
941 	    "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
942 	    ifp, alwaysfrag, fibnum));
943 
944 	/*
945 	 * The caller of this function may specify to use the minimum MTU
946 	 * in some cases.
947 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
948 	 * setting.  The logic is a bit complicated; by default, unicast
949 	 * packets will follow path MTU while multicast packets will be sent at
950 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
951 	 * including unicast ones will be sent at the minimum MTU.  Multicast
952 	 * packets will always be sent at the minimum MTU unless
953 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
954 	 * See RFC 3542 for more details.
955 	 */
956 	if (mtu > IPV6_MMTU) {
957 		if ((flags & IPV6_MINMTU))
958 			mtu = IPV6_MMTU;
959 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
960 			mtu = IPV6_MMTU;
961 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
962 			 (opt == NULL ||
963 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
964 			mtu = IPV6_MMTU;
965 		}
966 	}
967 
968 	/*
969 	 * Clear embedded scope identifiers if necessary.
970 	 * in6_clearscope() will touch the addresses only when necessary.
971 	 */
972 	in6_clearscope(&ip6->ip6_src);
973 	in6_clearscope(&ip6->ip6_dst);
974 
975 	/*
976 	 * If the outgoing packet contains a hop-by-hop options header,
977 	 * it must be examined and processed even by the source node.
978 	 * (RFC 2460, section 4.)
979 	 */
980 	if (exthdrs.ip6e_hbh) {
981 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
982 		u_int32_t dummy; /* XXX unused */
983 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
984 
985 #ifdef DIAGNOSTIC
986 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
987 			panic("ip6e_hbh is not contiguous");
988 #endif
989 		/*
990 		 *  XXX: if we have to send an ICMPv6 error to the sender,
991 		 *       we need the M_LOOP flag since icmp6_error() expects
992 		 *       the IPv6 and the hop-by-hop options header are
993 		 *       contiguous unless the flag is set.
994 		 */
995 		m->m_flags |= M_LOOP;
996 		m->m_pkthdr.rcvif = ifp;
997 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
998 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
999 		    &dummy, &plen) < 0) {
1000 			/* m was already freed at this point. */
1001 			error = EINVAL;/* better error? */
1002 			goto done;
1003 		}
1004 		m->m_flags &= ~M_LOOP; /* XXX */
1005 		m->m_pkthdr.rcvif = NULL;
1006 	}
1007 
1008 	/* Jump over all PFIL processing if hooks are not active. */
1009 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1010 		goto passout;
1011 
1012 	odst = ip6->ip6_dst;
1013 	/* Run through list of hooks for output packets. */
1014 	switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1015 	case PFIL_PASS:
1016 		ip6 = mtod(m, struct ip6_hdr *);
1017 		break;
1018 	case PFIL_DROPPED:
1019 		error = EACCES;
1020 		/* FALLTHROUGH */
1021 	case PFIL_CONSUMED:
1022 		goto done;
1023 	}
1024 
1025 	needfiblookup = 0;
1026 	/* See if destination IP address was changed by packet filter. */
1027 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1028 		m->m_flags |= M_SKIP_FIREWALL;
1029 		/* If destination is now ourself drop to ip6_input(). */
1030 		if (in6_localip(&ip6->ip6_dst)) {
1031 			m->m_flags |= M_FASTFWD_OURS;
1032 			if (m->m_pkthdr.rcvif == NULL)
1033 				m->m_pkthdr.rcvif = V_loif;
1034 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1035 				m->m_pkthdr.csum_flags |=
1036 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1037 				m->m_pkthdr.csum_data = 0xffff;
1038 			}
1039 #if defined(SCTP) || defined(SCTP_SUPPORT)
1040 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1041 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1042 #endif
1043 			error = netisr_queue(NETISR_IPV6, m);
1044 			goto done;
1045 		} else {
1046 			if (ro != NULL)
1047 				RO_INVALIDATE_CACHE(ro);
1048 			needfiblookup = 1; /* Redo the routing table lookup. */
1049 		}
1050 	}
1051 	/* See if fib was changed by packet filter. */
1052 	if (fibnum != M_GETFIB(m)) {
1053 		m->m_flags |= M_SKIP_FIREWALL;
1054 		fibnum = M_GETFIB(m);
1055 		if (ro != NULL)
1056 			RO_INVALIDATE_CACHE(ro);
1057 		needfiblookup = 1;
1058 	}
1059 	if (needfiblookup)
1060 		goto again;
1061 
1062 	/* See if local, if yes, send it to netisr. */
1063 	if (m->m_flags & M_FASTFWD_OURS) {
1064 		if (m->m_pkthdr.rcvif == NULL)
1065 			m->m_pkthdr.rcvif = V_loif;
1066 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1067 			m->m_pkthdr.csum_flags |=
1068 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1069 			m->m_pkthdr.csum_data = 0xffff;
1070 		}
1071 #if defined(SCTP) || defined(SCTP_SUPPORT)
1072 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1073 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1074 #endif
1075 		error = netisr_queue(NETISR_IPV6, m);
1076 		goto done;
1077 	}
1078 	/* Or forward to some other address? */
1079 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1080 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1081 		if (ro != NULL)
1082 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1083 		else
1084 			dst = &sin6;
1085 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1086 		m->m_flags |= M_SKIP_FIREWALL;
1087 		m->m_flags &= ~M_IP6_NEXTHOP;
1088 		m_tag_delete(m, fwd_tag);
1089 		goto again;
1090 	}
1091 
1092 passout:
1093 	if (vlan_pcp > -1)
1094 		EVL_APPLY_PRI(m, vlan_pcp);
1095 
1096 	/* Ensure the packet data is mapped if the interface requires it. */
1097 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1098 		m = mb_unmapped_to_ext(m);
1099 		if (m == NULL) {
1100 			IP6STAT_INC(ip6s_odropped);
1101 			return (ENOBUFS);
1102 		}
1103 	}
1104 
1105 	/*
1106 	 * Send the packet to the outgoing interface.
1107 	 * If necessary, do IPv6 fragmentation before sending.
1108 	 *
1109 	 * The logic here is rather complex:
1110 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1111 	 * 1-a:	send as is if tlen <= path mtu
1112 	 * 1-b:	fragment if tlen > path mtu
1113 	 *
1114 	 * 2: if user asks us not to fragment (dontfrag == 1)
1115 	 * 2-a:	send as is if tlen <= interface mtu
1116 	 * 2-b:	error if tlen > interface mtu
1117 	 *
1118 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1119 	 *	always fragment
1120 	 *
1121 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1122 	 *	error, as we cannot handle this conflicting request.
1123 	 */
1124 	sw_csum = m->m_pkthdr.csum_flags;
1125 	if (!hdrsplit) {
1126 		tso = ((sw_csum & ifp->if_hwassist &
1127 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1128 		sw_csum &= ~ifp->if_hwassist;
1129 	} else
1130 		tso = 0;
1131 	/*
1132 	 * If we added extension headers, we will not do TSO and calculate the
1133 	 * checksums ourselves for now.
1134 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1135 	 * with ext. hdrs.
1136 	 */
1137 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1138 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1139 	tlen = m->m_pkthdr.len;
1140 
1141 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1142 		dontfrag = 1;
1143 	else
1144 		dontfrag = 0;
1145 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1146 		/* Conflicting request - can't transmit. */
1147 		error = EMSGSIZE;
1148 		goto bad;
1149 	}
1150 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1151 		/*
1152 		 * Even if the DONTFRAG option is specified, we cannot send the
1153 		 * packet when the data length is larger than the MTU of the
1154 		 * outgoing interface.
1155 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1156 		 * application wanted to know the MTU value. Also return an
1157 		 * error code (this is not described in the API spec).
1158 		 */
1159 		if (inp != NULL)
1160 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1161 		error = EMSGSIZE;
1162 		goto bad;
1163 	}
1164 
1165 	/* Transmit packet without fragmentation. */
1166 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1167 		struct in6_ifaddr *ia6;
1168 
1169 		ip6 = mtod(m, struct ip6_hdr *);
1170 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1171 		if (ia6) {
1172 			/* Record statistics for this interface address. */
1173 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1174 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1175 			    m->m_pkthdr.len);
1176 		}
1177 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1178 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1179 		goto done;
1180 	}
1181 
1182 	/* Try to fragment the packet.  Cases 1-b and 3. */
1183 	if (mtu < IPV6_MMTU) {
1184 		/* Path MTU cannot be less than IPV6_MMTU. */
1185 		error = EMSGSIZE;
1186 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1187 		goto bad;
1188 	} else if (ip6->ip6_plen == 0) {
1189 		/* Jumbo payload cannot be fragmented. */
1190 		error = EMSGSIZE;
1191 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1192 		goto bad;
1193 	} else {
1194 		u_char nextproto;
1195 
1196 		/*
1197 		 * Too large for the destination or interface;
1198 		 * fragment if possible.
1199 		 * Must be able to put at least 8 bytes per fragment.
1200 		 */
1201 		if (mtu > IPV6_MAXPACKET)
1202 			mtu = IPV6_MAXPACKET;
1203 
1204 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1205 		if (len < 8) {
1206 			error = EMSGSIZE;
1207 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1208 			goto bad;
1209 		}
1210 
1211 		/*
1212 		 * If the interface will not calculate checksums on
1213 		 * fragmented packets, then do it here.
1214 		 * XXX-BZ handle the hw offloading case.  Need flags.
1215 		 */
1216 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1217 		    optlen);
1218 
1219 		/*
1220 		 * Change the next header field of the last header in the
1221 		 * unfragmentable part.
1222 		 */
1223 		if (exthdrs.ip6e_rthdr) {
1224 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1225 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1226 		} else if (exthdrs.ip6e_dest1) {
1227 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1228 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1229 		} else if (exthdrs.ip6e_hbh) {
1230 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1231 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1232 		} else {
1233 			ip6 = mtod(m, struct ip6_hdr *);
1234 			nextproto = ip6->ip6_nxt;
1235 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1236 		}
1237 
1238 		/*
1239 		 * Loop through length of segment after first fragment,
1240 		 * make new header and copy data of each part and link onto
1241 		 * chain.
1242 		 */
1243 		m0 = m;
1244 		id = htonl(ip6_randomid());
1245 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1246 		if (error != 0)
1247 			goto sendorfree;
1248 
1249 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1250 	}
1251 
1252 	/* Remove leading garbage. */
1253 sendorfree:
1254 	m = m0->m_nextpkt;
1255 	m0->m_nextpkt = 0;
1256 	m_freem(m0);
1257 	for (; m; m = m0) {
1258 		m0 = m->m_nextpkt;
1259 		m->m_nextpkt = 0;
1260 		if (error == 0) {
1261 			/* Record statistics for this interface address. */
1262 			if (ia) {
1263 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1264 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1265 				    m->m_pkthdr.len);
1266 			}
1267 			if (vlan_pcp > -1)
1268 				EVL_APPLY_PRI(m, vlan_pcp);
1269 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1270 			    true);
1271 		} else
1272 			m_freem(m);
1273 	}
1274 
1275 	if (error == 0)
1276 		IP6STAT_INC(ip6s_fragmented);
1277 
1278 done:
1279 	return (error);
1280 
1281 freehdrs:
1282 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1283 	m_freem(exthdrs.ip6e_dest1);
1284 	m_freem(exthdrs.ip6e_rthdr);
1285 	m_freem(exthdrs.ip6e_dest2);
1286 	/* FALLTHROUGH */
1287 bad:
1288 	if (m)
1289 		m_freem(m);
1290 	goto done;
1291 }
1292 
1293 static int
1294 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1295 {
1296 	struct mbuf *m;
1297 
1298 	if (hlen > MCLBYTES)
1299 		return (ENOBUFS); /* XXX */
1300 
1301 	if (hlen > MLEN)
1302 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1303 	else
1304 		m = m_get(M_NOWAIT, MT_DATA);
1305 	if (m == NULL)
1306 		return (ENOBUFS);
1307 	m->m_len = hlen;
1308 	if (hdr)
1309 		bcopy(hdr, mtod(m, caddr_t), hlen);
1310 
1311 	*mp = m;
1312 	return (0);
1313 }
1314 
1315 /*
1316  * Insert jumbo payload option.
1317  */
1318 static int
1319 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1320 {
1321 	struct mbuf *mopt;
1322 	u_char *optbuf;
1323 	u_int32_t v;
1324 
1325 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1326 
1327 	/*
1328 	 * If there is no hop-by-hop options header, allocate new one.
1329 	 * If there is one but it doesn't have enough space to store the
1330 	 * jumbo payload option, allocate a cluster to store the whole options.
1331 	 * Otherwise, use it to store the options.
1332 	 */
1333 	if (exthdrs->ip6e_hbh == NULL) {
1334 		mopt = m_get(M_NOWAIT, MT_DATA);
1335 		if (mopt == NULL)
1336 			return (ENOBUFS);
1337 		mopt->m_len = JUMBOOPTLEN;
1338 		optbuf = mtod(mopt, u_char *);
1339 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1340 		exthdrs->ip6e_hbh = mopt;
1341 	} else {
1342 		struct ip6_hbh *hbh;
1343 
1344 		mopt = exthdrs->ip6e_hbh;
1345 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1346 			/*
1347 			 * XXX assumption:
1348 			 * - exthdrs->ip6e_hbh is not referenced from places
1349 			 *   other than exthdrs.
1350 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1351 			 */
1352 			int oldoptlen = mopt->m_len;
1353 			struct mbuf *n;
1354 
1355 			/*
1356 			 * XXX: give up if the whole (new) hbh header does
1357 			 * not fit even in an mbuf cluster.
1358 			 */
1359 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1360 				return (ENOBUFS);
1361 
1362 			/*
1363 			 * As a consequence, we must always prepare a cluster
1364 			 * at this point.
1365 			 */
1366 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1367 			if (n == NULL)
1368 				return (ENOBUFS);
1369 			n->m_len = oldoptlen + JUMBOOPTLEN;
1370 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1371 			    oldoptlen);
1372 			optbuf = mtod(n, caddr_t) + oldoptlen;
1373 			m_freem(mopt);
1374 			mopt = exthdrs->ip6e_hbh = n;
1375 		} else {
1376 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1377 			mopt->m_len += JUMBOOPTLEN;
1378 		}
1379 		optbuf[0] = IP6OPT_PADN;
1380 		optbuf[1] = 1;
1381 
1382 		/*
1383 		 * Adjust the header length according to the pad and
1384 		 * the jumbo payload option.
1385 		 */
1386 		hbh = mtod(mopt, struct ip6_hbh *);
1387 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1388 	}
1389 
1390 	/* fill in the option. */
1391 	optbuf[2] = IP6OPT_JUMBO;
1392 	optbuf[3] = 4;
1393 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1394 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1395 
1396 	/* finally, adjust the packet header length */
1397 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1398 
1399 	return (0);
1400 #undef JUMBOOPTLEN
1401 }
1402 
1403 /*
1404  * Insert fragment header and copy unfragmentable header portions.
1405  */
1406 static int
1407 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1408     struct ip6_frag **frghdrp)
1409 {
1410 	struct mbuf *n, *mlast;
1411 
1412 	if (hlen > sizeof(struct ip6_hdr)) {
1413 		n = m_copym(m0, sizeof(struct ip6_hdr),
1414 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1415 		if (n == NULL)
1416 			return (ENOBUFS);
1417 		m->m_next = n;
1418 	} else
1419 		n = m;
1420 
1421 	/* Search for the last mbuf of unfragmentable part. */
1422 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1423 		;
1424 
1425 	if (M_WRITABLE(mlast) &&
1426 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1427 		/* use the trailing space of the last mbuf for the fragment hdr */
1428 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1429 		    mlast->m_len);
1430 		mlast->m_len += sizeof(struct ip6_frag);
1431 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1432 	} else {
1433 		/* allocate a new mbuf for the fragment header */
1434 		struct mbuf *mfrg;
1435 
1436 		mfrg = m_get(M_NOWAIT, MT_DATA);
1437 		if (mfrg == NULL)
1438 			return (ENOBUFS);
1439 		mfrg->m_len = sizeof(struct ip6_frag);
1440 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1441 		mlast->m_next = mfrg;
1442 	}
1443 
1444 	return (0);
1445 }
1446 
1447 /*
1448  * Calculates IPv6 path mtu for destination @dst.
1449  * Resulting MTU is stored in @mtup.
1450  *
1451  * Returns 0 on success.
1452  */
1453 static int
1454 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1455 {
1456 	struct epoch_tracker et;
1457 	struct nhop_object *nh;
1458 	struct in6_addr kdst;
1459 	uint32_t scopeid;
1460 	int error;
1461 
1462 	in6_splitscope(dst, &kdst, &scopeid);
1463 
1464 	NET_EPOCH_ENTER(et);
1465 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1466 	if (nh != NULL)
1467 		error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1468 	else
1469 		error = EHOSTUNREACH;
1470 	NET_EPOCH_EXIT(et);
1471 
1472 	return (error);
1473 }
1474 
1475 /*
1476  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1477  * and cached data in @ro_pmtu.
1478  * MTU from (successful) route lookup is saved (along with dst)
1479  * inside @ro_pmtu to avoid subsequent route lookups after packet
1480  * filter processing.
1481  *
1482  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1483  * Returns 0 on success.
1484  */
1485 static int
1486 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1487     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1488     int *alwaysfragp, u_int fibnum, u_int proto)
1489 {
1490 	struct nhop_object *nh;
1491 	struct in6_addr kdst;
1492 	uint32_t scopeid;
1493 	struct sockaddr_in6 *sa6_dst, sin6;
1494 	u_long mtu;
1495 
1496 	NET_EPOCH_ASSERT();
1497 
1498 	mtu = 0;
1499 	if (ro_pmtu == NULL || do_lookup) {
1500 		/*
1501 		 * Here ro_pmtu has final destination address, while
1502 		 * ro might represent immediate destination.
1503 		 * Use ro_pmtu destination since mtu might differ.
1504 		 */
1505 		if (ro_pmtu != NULL) {
1506 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1507 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1508 				ro_pmtu->ro_mtu = 0;
1509 		} else
1510 			sa6_dst = &sin6;
1511 
1512 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1513 			bzero(sa6_dst, sizeof(*sa6_dst));
1514 			sa6_dst->sin6_family = AF_INET6;
1515 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1516 			sa6_dst->sin6_addr = *dst;
1517 
1518 			in6_splitscope(dst, &kdst, &scopeid);
1519 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1520 			if (nh != NULL) {
1521 				mtu = nh->nh_mtu;
1522 				if (ro_pmtu != NULL)
1523 					ro_pmtu->ro_mtu = mtu;
1524 			}
1525 		} else
1526 			mtu = ro_pmtu->ro_mtu;
1527 	}
1528 
1529 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1530 		mtu = ro_pmtu->ro_nh->nh_mtu;
1531 
1532 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1533 }
1534 
1535 /*
1536  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1537  * hostcache data for @dst.
1538  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1539  *
1540  * Returns 0 on success.
1541  */
1542 static int
1543 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1544     u_long *mtup, int *alwaysfragp, u_int proto)
1545 {
1546 	u_long mtu = 0;
1547 	int alwaysfrag = 0;
1548 	int error = 0;
1549 
1550 	if (rt_mtu > 0) {
1551 		u_int32_t ifmtu;
1552 		struct in_conninfo inc;
1553 
1554 		bzero(&inc, sizeof(inc));
1555 		inc.inc_flags |= INC_ISIPV6;
1556 		inc.inc6_faddr = *dst;
1557 
1558 		ifmtu = IN6_LINKMTU(ifp);
1559 
1560 		/* TCP is known to react to pmtu changes so skip hc */
1561 		if (proto != IPPROTO_TCP)
1562 			mtu = tcp_hc_getmtu(&inc);
1563 
1564 		if (mtu)
1565 			mtu = min(mtu, rt_mtu);
1566 		else
1567 			mtu = rt_mtu;
1568 		if (mtu == 0)
1569 			mtu = ifmtu;
1570 		else if (mtu < IPV6_MMTU) {
1571 			/*
1572 			 * RFC2460 section 5, last paragraph:
1573 			 * if we record ICMPv6 too big message with
1574 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1575 			 * or smaller, with framgent header attached.
1576 			 * (fragment header is needed regardless from the
1577 			 * packet size, for translators to identify packets)
1578 			 */
1579 			alwaysfrag = 1;
1580 			mtu = IPV6_MMTU;
1581 		}
1582 	} else if (ifp) {
1583 		mtu = IN6_LINKMTU(ifp);
1584 	} else
1585 		error = EHOSTUNREACH; /* XXX */
1586 
1587 	*mtup = mtu;
1588 	if (alwaysfragp)
1589 		*alwaysfragp = alwaysfrag;
1590 	return (error);
1591 }
1592 
1593 /*
1594  * IP6 socket option processing.
1595  */
1596 int
1597 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1598 {
1599 	int optdatalen, uproto;
1600 	void *optdata;
1601 	struct inpcb *inp = sotoinpcb(so);
1602 	int error, optval;
1603 	int level, op, optname;
1604 	int optlen;
1605 	struct thread *td;
1606 #ifdef	RSS
1607 	uint32_t rss_bucket;
1608 	int retval;
1609 #endif
1610 
1611 /*
1612  * Don't use more than a quarter of mbuf clusters.  N.B.:
1613  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1614  * on LP64 architectures, so cast to u_long to avoid undefined
1615  * behavior.  ILP32 architectures cannot have nmbclusters
1616  * large enough to overflow for other reasons.
1617  */
1618 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1619 
1620 	level = sopt->sopt_level;
1621 	op = sopt->sopt_dir;
1622 	optname = sopt->sopt_name;
1623 	optlen = sopt->sopt_valsize;
1624 	td = sopt->sopt_td;
1625 	error = 0;
1626 	optval = 0;
1627 	uproto = (int)so->so_proto->pr_protocol;
1628 
1629 	if (level != IPPROTO_IPV6) {
1630 		error = EINVAL;
1631 
1632 		if (sopt->sopt_level == SOL_SOCKET &&
1633 		    sopt->sopt_dir == SOPT_SET) {
1634 			switch (sopt->sopt_name) {
1635 			case SO_REUSEADDR:
1636 				INP_WLOCK(inp);
1637 				if ((so->so_options & SO_REUSEADDR) != 0)
1638 					inp->inp_flags2 |= INP_REUSEADDR;
1639 				else
1640 					inp->inp_flags2 &= ~INP_REUSEADDR;
1641 				INP_WUNLOCK(inp);
1642 				error = 0;
1643 				break;
1644 			case SO_REUSEPORT:
1645 				INP_WLOCK(inp);
1646 				if ((so->so_options & SO_REUSEPORT) != 0)
1647 					inp->inp_flags2 |= INP_REUSEPORT;
1648 				else
1649 					inp->inp_flags2 &= ~INP_REUSEPORT;
1650 				INP_WUNLOCK(inp);
1651 				error = 0;
1652 				break;
1653 			case SO_REUSEPORT_LB:
1654 				INP_WLOCK(inp);
1655 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1656 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1657 				else
1658 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1659 				INP_WUNLOCK(inp);
1660 				error = 0;
1661 				break;
1662 			case SO_SETFIB:
1663 				INP_WLOCK(inp);
1664 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1665 				INP_WUNLOCK(inp);
1666 				error = 0;
1667 				break;
1668 			case SO_MAX_PACING_RATE:
1669 #ifdef RATELIMIT
1670 				INP_WLOCK(inp);
1671 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1672 				INP_WUNLOCK(inp);
1673 				error = 0;
1674 #else
1675 				error = EOPNOTSUPP;
1676 #endif
1677 				break;
1678 			default:
1679 				break;
1680 			}
1681 		}
1682 	} else {		/* level == IPPROTO_IPV6 */
1683 		switch (op) {
1684 		case SOPT_SET:
1685 			switch (optname) {
1686 			case IPV6_2292PKTOPTIONS:
1687 #ifdef IPV6_PKTOPTIONS
1688 			case IPV6_PKTOPTIONS:
1689 #endif
1690 			{
1691 				struct mbuf *m;
1692 
1693 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1694 					printf("ip6_ctloutput: mbuf limit hit\n");
1695 					error = ENOBUFS;
1696 					break;
1697 				}
1698 
1699 				error = soopt_getm(sopt, &m); /* XXX */
1700 				if (error != 0)
1701 					break;
1702 				error = soopt_mcopyin(sopt, m); /* XXX */
1703 				if (error != 0)
1704 					break;
1705 				INP_WLOCK(inp);
1706 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1707 				    so, sopt);
1708 				INP_WUNLOCK(inp);
1709 				m_freem(m); /* XXX */
1710 				break;
1711 			}
1712 
1713 			/*
1714 			 * Use of some Hop-by-Hop options or some
1715 			 * Destination options, might require special
1716 			 * privilege.  That is, normal applications
1717 			 * (without special privilege) might be forbidden
1718 			 * from setting certain options in outgoing packets,
1719 			 * and might never see certain options in received
1720 			 * packets. [RFC 2292 Section 6]
1721 			 * KAME specific note:
1722 			 *  KAME prevents non-privileged users from sending or
1723 			 *  receiving ANY hbh/dst options in order to avoid
1724 			 *  overhead of parsing options in the kernel.
1725 			 */
1726 			case IPV6_RECVHOPOPTS:
1727 			case IPV6_RECVDSTOPTS:
1728 			case IPV6_RECVRTHDRDSTOPTS:
1729 				if (td != NULL) {
1730 					error = priv_check(td,
1731 					    PRIV_NETINET_SETHDROPTS);
1732 					if (error)
1733 						break;
1734 				}
1735 				/* FALLTHROUGH */
1736 			case IPV6_UNICAST_HOPS:
1737 			case IPV6_HOPLIMIT:
1738 
1739 			case IPV6_RECVPKTINFO:
1740 			case IPV6_RECVHOPLIMIT:
1741 			case IPV6_RECVRTHDR:
1742 			case IPV6_RECVPATHMTU:
1743 			case IPV6_RECVTCLASS:
1744 			case IPV6_RECVFLOWID:
1745 #ifdef	RSS
1746 			case IPV6_RECVRSSBUCKETID:
1747 #endif
1748 			case IPV6_V6ONLY:
1749 			case IPV6_AUTOFLOWLABEL:
1750 			case IPV6_ORIGDSTADDR:
1751 			case IPV6_BINDANY:
1752 			case IPV6_BINDMULTI:
1753 #ifdef	RSS
1754 			case IPV6_RSS_LISTEN_BUCKET:
1755 #endif
1756 			case IPV6_VLAN_PCP:
1757 				if (optname == IPV6_BINDANY && td != NULL) {
1758 					error = priv_check(td,
1759 					    PRIV_NETINET_BINDANY);
1760 					if (error)
1761 						break;
1762 				}
1763 
1764 				if (optlen != sizeof(int)) {
1765 					error = EINVAL;
1766 					break;
1767 				}
1768 				error = sooptcopyin(sopt, &optval,
1769 					sizeof optval, sizeof optval);
1770 				if (error)
1771 					break;
1772 				switch (optname) {
1773 				case IPV6_UNICAST_HOPS:
1774 					if (optval < -1 || optval >= 256)
1775 						error = EINVAL;
1776 					else {
1777 						/* -1 = kernel default */
1778 						inp->in6p_hops = optval;
1779 						if ((inp->inp_vflag &
1780 						     INP_IPV4) != 0)
1781 							inp->inp_ip_ttl = optval;
1782 					}
1783 					break;
1784 #define OPTSET(bit) \
1785 do { \
1786 	INP_WLOCK(inp); \
1787 	if (optval) \
1788 		inp->inp_flags |= (bit); \
1789 	else \
1790 		inp->inp_flags &= ~(bit); \
1791 	INP_WUNLOCK(inp); \
1792 } while (/*CONSTCOND*/ 0)
1793 #define OPTSET2292(bit) \
1794 do { \
1795 	INP_WLOCK(inp); \
1796 	inp->inp_flags |= IN6P_RFC2292; \
1797 	if (optval) \
1798 		inp->inp_flags |= (bit); \
1799 	else \
1800 		inp->inp_flags &= ~(bit); \
1801 	INP_WUNLOCK(inp); \
1802 } while (/*CONSTCOND*/ 0)
1803 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1804 
1805 #define OPTSET2_N(bit, val) do {					\
1806 	if (val)							\
1807 		inp->inp_flags2 |= bit;					\
1808 	else								\
1809 		inp->inp_flags2 &= ~bit;				\
1810 } while (0)
1811 #define OPTSET2(bit, val) do {						\
1812 	INP_WLOCK(inp);							\
1813 	OPTSET2_N(bit, val);						\
1814 	INP_WUNLOCK(inp);						\
1815 } while (0)
1816 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1817 #define OPTSET2292_EXCLUSIVE(bit)					\
1818 do {									\
1819 	INP_WLOCK(inp);							\
1820 	if (OPTBIT(IN6P_RFC2292)) {					\
1821 		error = EINVAL;						\
1822 	} else {							\
1823 		if (optval)						\
1824 			inp->inp_flags |= (bit);			\
1825 		else							\
1826 			inp->inp_flags &= ~(bit);			\
1827 	}								\
1828 	INP_WUNLOCK(inp);						\
1829 } while (/*CONSTCOND*/ 0)
1830 
1831 				case IPV6_RECVPKTINFO:
1832 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1833 					break;
1834 
1835 				case IPV6_HOPLIMIT:
1836 				{
1837 					struct ip6_pktopts **optp;
1838 
1839 					/* cannot mix with RFC2292 */
1840 					if (OPTBIT(IN6P_RFC2292)) {
1841 						error = EINVAL;
1842 						break;
1843 					}
1844 					INP_WLOCK(inp);
1845 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1846 						INP_WUNLOCK(inp);
1847 						return (ECONNRESET);
1848 					}
1849 					optp = &inp->in6p_outputopts;
1850 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1851 					    (u_char *)&optval, sizeof(optval),
1852 					    optp, (td != NULL) ? td->td_ucred :
1853 					    NULL, uproto);
1854 					INP_WUNLOCK(inp);
1855 					break;
1856 				}
1857 
1858 				case IPV6_RECVHOPLIMIT:
1859 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1860 					break;
1861 
1862 				case IPV6_RECVHOPOPTS:
1863 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1864 					break;
1865 
1866 				case IPV6_RECVDSTOPTS:
1867 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1868 					break;
1869 
1870 				case IPV6_RECVRTHDRDSTOPTS:
1871 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1872 					break;
1873 
1874 				case IPV6_RECVRTHDR:
1875 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1876 					break;
1877 
1878 				case IPV6_RECVPATHMTU:
1879 					/*
1880 					 * We ignore this option for TCP
1881 					 * sockets.
1882 					 * (RFC3542 leaves this case
1883 					 * unspecified.)
1884 					 */
1885 					if (uproto != IPPROTO_TCP)
1886 						OPTSET(IN6P_MTU);
1887 					break;
1888 
1889 				case IPV6_RECVFLOWID:
1890 					OPTSET2(INP_RECVFLOWID, optval);
1891 					break;
1892 
1893 #ifdef	RSS
1894 				case IPV6_RECVRSSBUCKETID:
1895 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1896 					break;
1897 #endif
1898 
1899 				case IPV6_V6ONLY:
1900 					INP_WLOCK(inp);
1901 					if (inp->inp_lport ||
1902 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1903 						/*
1904 						 * The socket is already bound.
1905 						 */
1906 						INP_WUNLOCK(inp);
1907 						error = EINVAL;
1908 						break;
1909 					}
1910 					if (optval) {
1911 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1912 						inp->inp_vflag &= ~INP_IPV4;
1913 					} else {
1914 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1915 						inp->inp_vflag |= INP_IPV4;
1916 					}
1917 					INP_WUNLOCK(inp);
1918 					break;
1919 				case IPV6_RECVTCLASS:
1920 					/* cannot mix with RFC2292 XXX */
1921 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1922 					break;
1923 				case IPV6_AUTOFLOWLABEL:
1924 					OPTSET(IN6P_AUTOFLOWLABEL);
1925 					break;
1926 
1927 				case IPV6_ORIGDSTADDR:
1928 					OPTSET2(INP_ORIGDSTADDR, optval);
1929 					break;
1930 				case IPV6_BINDANY:
1931 					OPTSET(INP_BINDANY);
1932 					break;
1933 
1934 				case IPV6_BINDMULTI:
1935 					OPTSET2(INP_BINDMULTI, optval);
1936 					break;
1937 #ifdef	RSS
1938 				case IPV6_RSS_LISTEN_BUCKET:
1939 					if ((optval >= 0) &&
1940 					    (optval < rss_getnumbuckets())) {
1941 						INP_WLOCK(inp);
1942 						inp->inp_rss_listen_bucket = optval;
1943 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1944 						INP_WUNLOCK(inp);
1945 					} else {
1946 						error = EINVAL;
1947 					}
1948 					break;
1949 #endif
1950 				case IPV6_VLAN_PCP:
1951 					if ((optval >= -1) && (optval <=
1952 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1953 						if (optval == -1) {
1954 							INP_WLOCK(inp);
1955 							inp->inp_flags2 &=
1956 							    ~(INP_2PCP_SET |
1957 							    INP_2PCP_MASK);
1958 							INP_WUNLOCK(inp);
1959 						} else {
1960 							INP_WLOCK(inp);
1961 							inp->inp_flags2 |=
1962 							    INP_2PCP_SET;
1963 							inp->inp_flags2 &=
1964 							    ~INP_2PCP_MASK;
1965 							inp->inp_flags2 |=
1966 							    optval <<
1967 							    INP_2PCP_SHIFT;
1968 							INP_WUNLOCK(inp);
1969 						}
1970 					} else
1971 						error = EINVAL;
1972 					break;
1973 				}
1974 				break;
1975 
1976 			case IPV6_TCLASS:
1977 			case IPV6_DONTFRAG:
1978 			case IPV6_USE_MIN_MTU:
1979 			case IPV6_PREFER_TEMPADDR:
1980 				if (optlen != sizeof(optval)) {
1981 					error = EINVAL;
1982 					break;
1983 				}
1984 				error = sooptcopyin(sopt, &optval,
1985 					sizeof optval, sizeof optval);
1986 				if (error)
1987 					break;
1988 				{
1989 					struct ip6_pktopts **optp;
1990 					INP_WLOCK(inp);
1991 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1992 						INP_WUNLOCK(inp);
1993 						return (ECONNRESET);
1994 					}
1995 					optp = &inp->in6p_outputopts;
1996 					error = ip6_pcbopt(optname,
1997 					    (u_char *)&optval, sizeof(optval),
1998 					    optp, (td != NULL) ? td->td_ucred :
1999 					    NULL, uproto);
2000 					INP_WUNLOCK(inp);
2001 					break;
2002 				}
2003 
2004 			case IPV6_2292PKTINFO:
2005 			case IPV6_2292HOPLIMIT:
2006 			case IPV6_2292HOPOPTS:
2007 			case IPV6_2292DSTOPTS:
2008 			case IPV6_2292RTHDR:
2009 				/* RFC 2292 */
2010 				if (optlen != sizeof(int)) {
2011 					error = EINVAL;
2012 					break;
2013 				}
2014 				error = sooptcopyin(sopt, &optval,
2015 					sizeof optval, sizeof optval);
2016 				if (error)
2017 					break;
2018 				switch (optname) {
2019 				case IPV6_2292PKTINFO:
2020 					OPTSET2292(IN6P_PKTINFO);
2021 					break;
2022 				case IPV6_2292HOPLIMIT:
2023 					OPTSET2292(IN6P_HOPLIMIT);
2024 					break;
2025 				case IPV6_2292HOPOPTS:
2026 					/*
2027 					 * Check super-user privilege.
2028 					 * See comments for IPV6_RECVHOPOPTS.
2029 					 */
2030 					if (td != NULL) {
2031 						error = priv_check(td,
2032 						    PRIV_NETINET_SETHDROPTS);
2033 						if (error)
2034 							return (error);
2035 					}
2036 					OPTSET2292(IN6P_HOPOPTS);
2037 					break;
2038 				case IPV6_2292DSTOPTS:
2039 					if (td != NULL) {
2040 						error = priv_check(td,
2041 						    PRIV_NETINET_SETHDROPTS);
2042 						if (error)
2043 							return (error);
2044 					}
2045 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2046 					break;
2047 				case IPV6_2292RTHDR:
2048 					OPTSET2292(IN6P_RTHDR);
2049 					break;
2050 				}
2051 				break;
2052 			case IPV6_PKTINFO:
2053 			case IPV6_HOPOPTS:
2054 			case IPV6_RTHDR:
2055 			case IPV6_DSTOPTS:
2056 			case IPV6_RTHDRDSTOPTS:
2057 			case IPV6_NEXTHOP:
2058 			{
2059 				/* new advanced API (RFC3542) */
2060 				u_char *optbuf;
2061 				u_char optbuf_storage[MCLBYTES];
2062 				int optlen;
2063 				struct ip6_pktopts **optp;
2064 
2065 				/* cannot mix with RFC2292 */
2066 				if (OPTBIT(IN6P_RFC2292)) {
2067 					error = EINVAL;
2068 					break;
2069 				}
2070 
2071 				/*
2072 				 * We only ensure valsize is not too large
2073 				 * here.  Further validation will be done
2074 				 * later.
2075 				 */
2076 				error = sooptcopyin(sopt, optbuf_storage,
2077 				    sizeof(optbuf_storage), 0);
2078 				if (error)
2079 					break;
2080 				optlen = sopt->sopt_valsize;
2081 				optbuf = optbuf_storage;
2082 				INP_WLOCK(inp);
2083 				if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2084 					INP_WUNLOCK(inp);
2085 					return (ECONNRESET);
2086 				}
2087 				optp = &inp->in6p_outputopts;
2088 				error = ip6_pcbopt(optname, optbuf, optlen,
2089 				    optp, (td != NULL) ? td->td_ucred : NULL,
2090 				    uproto);
2091 				INP_WUNLOCK(inp);
2092 				break;
2093 			}
2094 #undef OPTSET
2095 
2096 			case IPV6_MULTICAST_IF:
2097 			case IPV6_MULTICAST_HOPS:
2098 			case IPV6_MULTICAST_LOOP:
2099 			case IPV6_JOIN_GROUP:
2100 			case IPV6_LEAVE_GROUP:
2101 			case IPV6_MSFILTER:
2102 			case MCAST_BLOCK_SOURCE:
2103 			case MCAST_UNBLOCK_SOURCE:
2104 			case MCAST_JOIN_GROUP:
2105 			case MCAST_LEAVE_GROUP:
2106 			case MCAST_JOIN_SOURCE_GROUP:
2107 			case MCAST_LEAVE_SOURCE_GROUP:
2108 				error = ip6_setmoptions(inp, sopt);
2109 				break;
2110 
2111 			case IPV6_PORTRANGE:
2112 				error = sooptcopyin(sopt, &optval,
2113 				    sizeof optval, sizeof optval);
2114 				if (error)
2115 					break;
2116 
2117 				INP_WLOCK(inp);
2118 				switch (optval) {
2119 				case IPV6_PORTRANGE_DEFAULT:
2120 					inp->inp_flags &= ~(INP_LOWPORT);
2121 					inp->inp_flags &= ~(INP_HIGHPORT);
2122 					break;
2123 
2124 				case IPV6_PORTRANGE_HIGH:
2125 					inp->inp_flags &= ~(INP_LOWPORT);
2126 					inp->inp_flags |= INP_HIGHPORT;
2127 					break;
2128 
2129 				case IPV6_PORTRANGE_LOW:
2130 					inp->inp_flags &= ~(INP_HIGHPORT);
2131 					inp->inp_flags |= INP_LOWPORT;
2132 					break;
2133 
2134 				default:
2135 					error = EINVAL;
2136 					break;
2137 				}
2138 				INP_WUNLOCK(inp);
2139 				break;
2140 
2141 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2142 			case IPV6_IPSEC_POLICY:
2143 				if (IPSEC_ENABLED(ipv6)) {
2144 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2145 					break;
2146 				}
2147 				/* FALLTHROUGH */
2148 #endif /* IPSEC */
2149 
2150 			default:
2151 				error = ENOPROTOOPT;
2152 				break;
2153 			}
2154 			break;
2155 
2156 		case SOPT_GET:
2157 			switch (optname) {
2158 			case IPV6_2292PKTOPTIONS:
2159 #ifdef IPV6_PKTOPTIONS
2160 			case IPV6_PKTOPTIONS:
2161 #endif
2162 				/*
2163 				 * RFC3542 (effectively) deprecated the
2164 				 * semantics of the 2292-style pktoptions.
2165 				 * Since it was not reliable in nature (i.e.,
2166 				 * applications had to expect the lack of some
2167 				 * information after all), it would make sense
2168 				 * to simplify this part by always returning
2169 				 * empty data.
2170 				 */
2171 				sopt->sopt_valsize = 0;
2172 				break;
2173 
2174 			case IPV6_RECVHOPOPTS:
2175 			case IPV6_RECVDSTOPTS:
2176 			case IPV6_RECVRTHDRDSTOPTS:
2177 			case IPV6_UNICAST_HOPS:
2178 			case IPV6_RECVPKTINFO:
2179 			case IPV6_RECVHOPLIMIT:
2180 			case IPV6_RECVRTHDR:
2181 			case IPV6_RECVPATHMTU:
2182 
2183 			case IPV6_V6ONLY:
2184 			case IPV6_PORTRANGE:
2185 			case IPV6_RECVTCLASS:
2186 			case IPV6_AUTOFLOWLABEL:
2187 			case IPV6_BINDANY:
2188 			case IPV6_FLOWID:
2189 			case IPV6_FLOWTYPE:
2190 			case IPV6_RECVFLOWID:
2191 #ifdef	RSS
2192 			case IPV6_RSSBUCKETID:
2193 			case IPV6_RECVRSSBUCKETID:
2194 #endif
2195 			case IPV6_BINDMULTI:
2196 			case IPV6_VLAN_PCP:
2197 				switch (optname) {
2198 				case IPV6_RECVHOPOPTS:
2199 					optval = OPTBIT(IN6P_HOPOPTS);
2200 					break;
2201 
2202 				case IPV6_RECVDSTOPTS:
2203 					optval = OPTBIT(IN6P_DSTOPTS);
2204 					break;
2205 
2206 				case IPV6_RECVRTHDRDSTOPTS:
2207 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2208 					break;
2209 
2210 				case IPV6_UNICAST_HOPS:
2211 					optval = inp->in6p_hops;
2212 					break;
2213 
2214 				case IPV6_RECVPKTINFO:
2215 					optval = OPTBIT(IN6P_PKTINFO);
2216 					break;
2217 
2218 				case IPV6_RECVHOPLIMIT:
2219 					optval = OPTBIT(IN6P_HOPLIMIT);
2220 					break;
2221 
2222 				case IPV6_RECVRTHDR:
2223 					optval = OPTBIT(IN6P_RTHDR);
2224 					break;
2225 
2226 				case IPV6_RECVPATHMTU:
2227 					optval = OPTBIT(IN6P_MTU);
2228 					break;
2229 
2230 				case IPV6_V6ONLY:
2231 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2232 					break;
2233 
2234 				case IPV6_PORTRANGE:
2235 				    {
2236 					int flags;
2237 					flags = inp->inp_flags;
2238 					if (flags & INP_HIGHPORT)
2239 						optval = IPV6_PORTRANGE_HIGH;
2240 					else if (flags & INP_LOWPORT)
2241 						optval = IPV6_PORTRANGE_LOW;
2242 					else
2243 						optval = 0;
2244 					break;
2245 				    }
2246 				case IPV6_RECVTCLASS:
2247 					optval = OPTBIT(IN6P_TCLASS);
2248 					break;
2249 
2250 				case IPV6_AUTOFLOWLABEL:
2251 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2252 					break;
2253 
2254 				case IPV6_ORIGDSTADDR:
2255 					optval = OPTBIT2(INP_ORIGDSTADDR);
2256 					break;
2257 
2258 				case IPV6_BINDANY:
2259 					optval = OPTBIT(INP_BINDANY);
2260 					break;
2261 
2262 				case IPV6_FLOWID:
2263 					optval = inp->inp_flowid;
2264 					break;
2265 
2266 				case IPV6_FLOWTYPE:
2267 					optval = inp->inp_flowtype;
2268 					break;
2269 
2270 				case IPV6_RECVFLOWID:
2271 					optval = OPTBIT2(INP_RECVFLOWID);
2272 					break;
2273 #ifdef	RSS
2274 				case IPV6_RSSBUCKETID:
2275 					retval =
2276 					    rss_hash2bucket(inp->inp_flowid,
2277 					    inp->inp_flowtype,
2278 					    &rss_bucket);
2279 					if (retval == 0)
2280 						optval = rss_bucket;
2281 					else
2282 						error = EINVAL;
2283 					break;
2284 
2285 				case IPV6_RECVRSSBUCKETID:
2286 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2287 					break;
2288 #endif
2289 
2290 				case IPV6_BINDMULTI:
2291 					optval = OPTBIT2(INP_BINDMULTI);
2292 					break;
2293 
2294 				case IPV6_VLAN_PCP:
2295 					if (OPTBIT2(INP_2PCP_SET)) {
2296 						optval = (inp->inp_flags2 &
2297 							    INP_2PCP_MASK) >>
2298 							    INP_2PCP_SHIFT;
2299 					} else {
2300 						optval = -1;
2301 					}
2302 					break;
2303 				}
2304 
2305 				if (error)
2306 					break;
2307 				error = sooptcopyout(sopt, &optval,
2308 					sizeof optval);
2309 				break;
2310 
2311 			case IPV6_PATHMTU:
2312 			{
2313 				u_long pmtu = 0;
2314 				struct ip6_mtuinfo mtuinfo;
2315 				struct in6_addr addr;
2316 
2317 				if (!(so->so_state & SS_ISCONNECTED))
2318 					return (ENOTCONN);
2319 				/*
2320 				 * XXX: we dot not consider the case of source
2321 				 * routing, or optional information to specify
2322 				 * the outgoing interface.
2323 				 * Copy faddr out of inp to avoid holding lock
2324 				 * on inp during route lookup.
2325 				 */
2326 				INP_RLOCK(inp);
2327 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2328 				INP_RUNLOCK(inp);
2329 				error = ip6_getpmtu_ctl(so->so_fibnum,
2330 				    &addr, &pmtu);
2331 				if (error)
2332 					break;
2333 				if (pmtu > IPV6_MAXPACKET)
2334 					pmtu = IPV6_MAXPACKET;
2335 
2336 				bzero(&mtuinfo, sizeof(mtuinfo));
2337 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2338 				optdata = (void *)&mtuinfo;
2339 				optdatalen = sizeof(mtuinfo);
2340 				error = sooptcopyout(sopt, optdata,
2341 				    optdatalen);
2342 				break;
2343 			}
2344 
2345 			case IPV6_2292PKTINFO:
2346 			case IPV6_2292HOPLIMIT:
2347 			case IPV6_2292HOPOPTS:
2348 			case IPV6_2292RTHDR:
2349 			case IPV6_2292DSTOPTS:
2350 				switch (optname) {
2351 				case IPV6_2292PKTINFO:
2352 					optval = OPTBIT(IN6P_PKTINFO);
2353 					break;
2354 				case IPV6_2292HOPLIMIT:
2355 					optval = OPTBIT(IN6P_HOPLIMIT);
2356 					break;
2357 				case IPV6_2292HOPOPTS:
2358 					optval = OPTBIT(IN6P_HOPOPTS);
2359 					break;
2360 				case IPV6_2292RTHDR:
2361 					optval = OPTBIT(IN6P_RTHDR);
2362 					break;
2363 				case IPV6_2292DSTOPTS:
2364 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2365 					break;
2366 				}
2367 				error = sooptcopyout(sopt, &optval,
2368 				    sizeof optval);
2369 				break;
2370 			case IPV6_PKTINFO:
2371 			case IPV6_HOPOPTS:
2372 			case IPV6_RTHDR:
2373 			case IPV6_DSTOPTS:
2374 			case IPV6_RTHDRDSTOPTS:
2375 			case IPV6_NEXTHOP:
2376 			case IPV6_TCLASS:
2377 			case IPV6_DONTFRAG:
2378 			case IPV6_USE_MIN_MTU:
2379 			case IPV6_PREFER_TEMPADDR:
2380 				error = ip6_getpcbopt(inp, optname, sopt);
2381 				break;
2382 
2383 			case IPV6_MULTICAST_IF:
2384 			case IPV6_MULTICAST_HOPS:
2385 			case IPV6_MULTICAST_LOOP:
2386 			case IPV6_MSFILTER:
2387 				error = ip6_getmoptions(inp, sopt);
2388 				break;
2389 
2390 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2391 			case IPV6_IPSEC_POLICY:
2392 				if (IPSEC_ENABLED(ipv6)) {
2393 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2394 					break;
2395 				}
2396 				/* FALLTHROUGH */
2397 #endif /* IPSEC */
2398 			default:
2399 				error = ENOPROTOOPT;
2400 				break;
2401 			}
2402 			break;
2403 		}
2404 	}
2405 	return (error);
2406 }
2407 
2408 int
2409 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2410 {
2411 	int error = 0, optval, optlen;
2412 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2413 	struct inpcb *inp = sotoinpcb(so);
2414 	int level, op, optname;
2415 
2416 	level = sopt->sopt_level;
2417 	op = sopt->sopt_dir;
2418 	optname = sopt->sopt_name;
2419 	optlen = sopt->sopt_valsize;
2420 
2421 	if (level != IPPROTO_IPV6) {
2422 		return (EINVAL);
2423 	}
2424 
2425 	switch (optname) {
2426 	case IPV6_CHECKSUM:
2427 		/*
2428 		 * For ICMPv6 sockets, no modification allowed for checksum
2429 		 * offset, permit "no change" values to help existing apps.
2430 		 *
2431 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2432 		 * for an ICMPv6 socket will fail."
2433 		 * The current behavior does not meet RFC3542.
2434 		 */
2435 		switch (op) {
2436 		case SOPT_SET:
2437 			if (optlen != sizeof(int)) {
2438 				error = EINVAL;
2439 				break;
2440 			}
2441 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2442 					    sizeof(optval));
2443 			if (error)
2444 				break;
2445 			if (optval < -1 || (optval % 2) != 0) {
2446 				/*
2447 				 * The API assumes non-negative even offset
2448 				 * values or -1 as a special value.
2449 				 */
2450 				error = EINVAL;
2451 			} else if (so->so_proto->pr_protocol ==
2452 			    IPPROTO_ICMPV6) {
2453 				if (optval != icmp6off)
2454 					error = EINVAL;
2455 			} else
2456 				inp->in6p_cksum = optval;
2457 			break;
2458 
2459 		case SOPT_GET:
2460 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2461 				optval = icmp6off;
2462 			else
2463 				optval = inp->in6p_cksum;
2464 
2465 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2466 			break;
2467 
2468 		default:
2469 			error = EINVAL;
2470 			break;
2471 		}
2472 		break;
2473 
2474 	default:
2475 		error = ENOPROTOOPT;
2476 		break;
2477 	}
2478 
2479 	return (error);
2480 }
2481 
2482 /*
2483  * Set up IP6 options in pcb for insertion in output packets or
2484  * specifying behavior of outgoing packets.
2485  */
2486 static int
2487 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2488     struct socket *so, struct sockopt *sopt)
2489 {
2490 	struct ip6_pktopts *opt = *pktopt;
2491 	int error = 0;
2492 	struct thread *td = sopt->sopt_td;
2493 	struct epoch_tracker et;
2494 
2495 	/* turn off any old options. */
2496 	if (opt) {
2497 #ifdef DIAGNOSTIC
2498 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2499 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2500 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2501 			printf("ip6_pcbopts: all specified options are cleared.\n");
2502 #endif
2503 		ip6_clearpktopts(opt, -1);
2504 	} else {
2505 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2506 		if (opt == NULL)
2507 			return (ENOMEM);
2508 	}
2509 	*pktopt = NULL;
2510 
2511 	if (!m || m->m_len == 0) {
2512 		/*
2513 		 * Only turning off any previous options, regardless of
2514 		 * whether the opt is just created or given.
2515 		 */
2516 		free(opt, M_IP6OPT);
2517 		return (0);
2518 	}
2519 
2520 	/*  set options specified by user. */
2521 	NET_EPOCH_ENTER(et);
2522 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2523 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2524 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2525 		free(opt, M_IP6OPT);
2526 		NET_EPOCH_EXIT(et);
2527 		return (error);
2528 	}
2529 	NET_EPOCH_EXIT(et);
2530 	*pktopt = opt;
2531 	return (0);
2532 }
2533 
2534 /*
2535  * initialize ip6_pktopts.  beware that there are non-zero default values in
2536  * the struct.
2537  */
2538 void
2539 ip6_initpktopts(struct ip6_pktopts *opt)
2540 {
2541 
2542 	bzero(opt, sizeof(*opt));
2543 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2544 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2545 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2546 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2547 }
2548 
2549 static int
2550 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2551     struct ucred *cred, int uproto)
2552 {
2553 	struct epoch_tracker et;
2554 	struct ip6_pktopts *opt;
2555 	int ret;
2556 
2557 	if (*pktopt == NULL) {
2558 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2559 		    M_NOWAIT);
2560 		if (*pktopt == NULL)
2561 			return (ENOBUFS);
2562 		ip6_initpktopts(*pktopt);
2563 	}
2564 	opt = *pktopt;
2565 
2566 	NET_EPOCH_ENTER(et);
2567 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2568 	NET_EPOCH_EXIT(et);
2569 
2570 	return (ret);
2571 }
2572 
2573 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2574 	if (pktopt && pktopt->field) {						\
2575 		INP_RUNLOCK(inp);						\
2576 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2577 		malloc_optdata = true;						\
2578 		INP_RLOCK(inp);							\
2579 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2580 			INP_RUNLOCK(inp);					\
2581 			free(optdata, M_TEMP);					\
2582 			return (ECONNRESET);					\
2583 		}								\
2584 		pktopt = inp->in6p_outputopts;					\
2585 		if (pktopt && pktopt->field) {					\
2586 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2587 			bcopy(pktopt->field, optdata, optdatalen);		\
2588 		} else {							\
2589 			free(optdata, M_TEMP);					\
2590 			optdata = NULL;						\
2591 			malloc_optdata = false;					\
2592 		}								\
2593 	}									\
2594 } while(0)
2595 
2596 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2597 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2598 
2599 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2600 	pktopt->field->sa_len)
2601 
2602 static int
2603 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2604 {
2605 	void *optdata = NULL;
2606 	bool malloc_optdata = false;
2607 	int optdatalen = 0;
2608 	int error = 0;
2609 	struct in6_pktinfo null_pktinfo;
2610 	int deftclass = 0, on;
2611 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2612 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2613 	struct ip6_pktopts *pktopt;
2614 
2615 	INP_RLOCK(inp);
2616 	pktopt = inp->in6p_outputopts;
2617 
2618 	switch (optname) {
2619 	case IPV6_PKTINFO:
2620 		optdata = (void *)&null_pktinfo;
2621 		if (pktopt && pktopt->ip6po_pktinfo) {
2622 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2623 			    sizeof(null_pktinfo));
2624 			in6_clearscope(&null_pktinfo.ipi6_addr);
2625 		} else {
2626 			/* XXX: we don't have to do this every time... */
2627 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2628 		}
2629 		optdatalen = sizeof(struct in6_pktinfo);
2630 		break;
2631 	case IPV6_TCLASS:
2632 		if (pktopt && pktopt->ip6po_tclass >= 0)
2633 			deftclass = pktopt->ip6po_tclass;
2634 		optdata = (void *)&deftclass;
2635 		optdatalen = sizeof(int);
2636 		break;
2637 	case IPV6_HOPOPTS:
2638 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2639 		break;
2640 	case IPV6_RTHDR:
2641 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2642 		break;
2643 	case IPV6_RTHDRDSTOPTS:
2644 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2645 		break;
2646 	case IPV6_DSTOPTS:
2647 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2648 		break;
2649 	case IPV6_NEXTHOP:
2650 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2651 		break;
2652 	case IPV6_USE_MIN_MTU:
2653 		if (pktopt)
2654 			defminmtu = pktopt->ip6po_minmtu;
2655 		optdata = (void *)&defminmtu;
2656 		optdatalen = sizeof(int);
2657 		break;
2658 	case IPV6_DONTFRAG:
2659 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2660 			on = 1;
2661 		else
2662 			on = 0;
2663 		optdata = (void *)&on;
2664 		optdatalen = sizeof(on);
2665 		break;
2666 	case IPV6_PREFER_TEMPADDR:
2667 		if (pktopt)
2668 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2669 		optdata = (void *)&defpreftemp;
2670 		optdatalen = sizeof(int);
2671 		break;
2672 	default:		/* should not happen */
2673 #ifdef DIAGNOSTIC
2674 		panic("ip6_getpcbopt: unexpected option\n");
2675 #endif
2676 		INP_RUNLOCK(inp);
2677 		return (ENOPROTOOPT);
2678 	}
2679 	INP_RUNLOCK(inp);
2680 
2681 	error = sooptcopyout(sopt, optdata, optdatalen);
2682 	if (malloc_optdata)
2683 		free(optdata, M_TEMP);
2684 
2685 	return (error);
2686 }
2687 
2688 void
2689 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2690 {
2691 	if (pktopt == NULL)
2692 		return;
2693 
2694 	if (optname == -1 || optname == IPV6_PKTINFO) {
2695 		if (pktopt->ip6po_pktinfo)
2696 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2697 		pktopt->ip6po_pktinfo = NULL;
2698 	}
2699 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2700 		pktopt->ip6po_hlim = -1;
2701 	if (optname == -1 || optname == IPV6_TCLASS)
2702 		pktopt->ip6po_tclass = -1;
2703 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2704 		if (pktopt->ip6po_nextroute.ro_nh) {
2705 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2706 			pktopt->ip6po_nextroute.ro_nh = NULL;
2707 		}
2708 		if (pktopt->ip6po_nexthop)
2709 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2710 		pktopt->ip6po_nexthop = NULL;
2711 	}
2712 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2713 		if (pktopt->ip6po_hbh)
2714 			free(pktopt->ip6po_hbh, M_IP6OPT);
2715 		pktopt->ip6po_hbh = NULL;
2716 	}
2717 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2718 		if (pktopt->ip6po_dest1)
2719 			free(pktopt->ip6po_dest1, M_IP6OPT);
2720 		pktopt->ip6po_dest1 = NULL;
2721 	}
2722 	if (optname == -1 || optname == IPV6_RTHDR) {
2723 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2724 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2725 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2726 		if (pktopt->ip6po_route.ro_nh) {
2727 			NH_FREE(pktopt->ip6po_route.ro_nh);
2728 			pktopt->ip6po_route.ro_nh = NULL;
2729 		}
2730 	}
2731 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2732 		if (pktopt->ip6po_dest2)
2733 			free(pktopt->ip6po_dest2, M_IP6OPT);
2734 		pktopt->ip6po_dest2 = NULL;
2735 	}
2736 }
2737 
2738 #define PKTOPT_EXTHDRCPY(type) \
2739 do {\
2740 	if (src->type) {\
2741 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2742 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2743 		if (dst->type == NULL)\
2744 			goto bad;\
2745 		bcopy(src->type, dst->type, hlen);\
2746 	}\
2747 } while (/*CONSTCOND*/ 0)
2748 
2749 static int
2750 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2751 {
2752 	if (dst == NULL || src == NULL)  {
2753 		printf("ip6_clearpktopts: invalid argument\n");
2754 		return (EINVAL);
2755 	}
2756 
2757 	dst->ip6po_hlim = src->ip6po_hlim;
2758 	dst->ip6po_tclass = src->ip6po_tclass;
2759 	dst->ip6po_flags = src->ip6po_flags;
2760 	dst->ip6po_minmtu = src->ip6po_minmtu;
2761 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2762 	if (src->ip6po_pktinfo) {
2763 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2764 		    M_IP6OPT, canwait);
2765 		if (dst->ip6po_pktinfo == NULL)
2766 			goto bad;
2767 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2768 	}
2769 	if (src->ip6po_nexthop) {
2770 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2771 		    M_IP6OPT, canwait);
2772 		if (dst->ip6po_nexthop == NULL)
2773 			goto bad;
2774 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2775 		    src->ip6po_nexthop->sa_len);
2776 	}
2777 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2778 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2779 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2780 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2781 	return (0);
2782 
2783   bad:
2784 	ip6_clearpktopts(dst, -1);
2785 	return (ENOBUFS);
2786 }
2787 #undef PKTOPT_EXTHDRCPY
2788 
2789 struct ip6_pktopts *
2790 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2791 {
2792 	int error;
2793 	struct ip6_pktopts *dst;
2794 
2795 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2796 	if (dst == NULL)
2797 		return (NULL);
2798 	ip6_initpktopts(dst);
2799 
2800 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2801 		free(dst, M_IP6OPT);
2802 		return (NULL);
2803 	}
2804 
2805 	return (dst);
2806 }
2807 
2808 void
2809 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2810 {
2811 	if (pktopt == NULL)
2812 		return;
2813 
2814 	ip6_clearpktopts(pktopt, -1);
2815 
2816 	free(pktopt, M_IP6OPT);
2817 }
2818 
2819 /*
2820  * Set IPv6 outgoing packet options based on advanced API.
2821  */
2822 int
2823 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2824     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2825 {
2826 	struct cmsghdr *cm = NULL;
2827 
2828 	if (control == NULL || opt == NULL)
2829 		return (EINVAL);
2830 
2831 	/*
2832 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2833 	 * are in the network epoch here.
2834 	 */
2835 	NET_EPOCH_ASSERT();
2836 
2837 	ip6_initpktopts(opt);
2838 	if (stickyopt) {
2839 		int error;
2840 
2841 		/*
2842 		 * If stickyopt is provided, make a local copy of the options
2843 		 * for this particular packet, then override them by ancillary
2844 		 * objects.
2845 		 * XXX: copypktopts() does not copy the cached route to a next
2846 		 * hop (if any).  This is not very good in terms of efficiency,
2847 		 * but we can allow this since this option should be rarely
2848 		 * used.
2849 		 */
2850 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2851 			return (error);
2852 	}
2853 
2854 	/*
2855 	 * XXX: Currently, we assume all the optional information is stored
2856 	 * in a single mbuf.
2857 	 */
2858 	if (control->m_next)
2859 		return (EINVAL);
2860 
2861 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2862 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2863 		int error;
2864 
2865 		if (control->m_len < CMSG_LEN(0))
2866 			return (EINVAL);
2867 
2868 		cm = mtod(control, struct cmsghdr *);
2869 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2870 			return (EINVAL);
2871 		if (cm->cmsg_level != IPPROTO_IPV6)
2872 			continue;
2873 
2874 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2875 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2876 		if (error)
2877 			return (error);
2878 	}
2879 
2880 	return (0);
2881 }
2882 
2883 /*
2884  * Set a particular packet option, as a sticky option or an ancillary data
2885  * item.  "len" can be 0 only when it's a sticky option.
2886  * We have 4 cases of combination of "sticky" and "cmsg":
2887  * "sticky=0, cmsg=0": impossible
2888  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2889  * "sticky=1, cmsg=0": RFC3542 socket option
2890  * "sticky=1, cmsg=1": RFC2292 socket option
2891  */
2892 static int
2893 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2894     struct ucred *cred, int sticky, int cmsg, int uproto)
2895 {
2896 	int minmtupolicy, preftemp;
2897 	int error;
2898 
2899 	NET_EPOCH_ASSERT();
2900 
2901 	if (!sticky && !cmsg) {
2902 #ifdef DIAGNOSTIC
2903 		printf("ip6_setpktopt: impossible case\n");
2904 #endif
2905 		return (EINVAL);
2906 	}
2907 
2908 	/*
2909 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2910 	 * not be specified in the context of RFC3542.  Conversely,
2911 	 * RFC3542 types should not be specified in the context of RFC2292.
2912 	 */
2913 	if (!cmsg) {
2914 		switch (optname) {
2915 		case IPV6_2292PKTINFO:
2916 		case IPV6_2292HOPLIMIT:
2917 		case IPV6_2292NEXTHOP:
2918 		case IPV6_2292HOPOPTS:
2919 		case IPV6_2292DSTOPTS:
2920 		case IPV6_2292RTHDR:
2921 		case IPV6_2292PKTOPTIONS:
2922 			return (ENOPROTOOPT);
2923 		}
2924 	}
2925 	if (sticky && cmsg) {
2926 		switch (optname) {
2927 		case IPV6_PKTINFO:
2928 		case IPV6_HOPLIMIT:
2929 		case IPV6_NEXTHOP:
2930 		case IPV6_HOPOPTS:
2931 		case IPV6_DSTOPTS:
2932 		case IPV6_RTHDRDSTOPTS:
2933 		case IPV6_RTHDR:
2934 		case IPV6_USE_MIN_MTU:
2935 		case IPV6_DONTFRAG:
2936 		case IPV6_TCLASS:
2937 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2938 			return (ENOPROTOOPT);
2939 		}
2940 	}
2941 
2942 	switch (optname) {
2943 	case IPV6_2292PKTINFO:
2944 	case IPV6_PKTINFO:
2945 	{
2946 		struct ifnet *ifp = NULL;
2947 		struct in6_pktinfo *pktinfo;
2948 
2949 		if (len != sizeof(struct in6_pktinfo))
2950 			return (EINVAL);
2951 
2952 		pktinfo = (struct in6_pktinfo *)buf;
2953 
2954 		/*
2955 		 * An application can clear any sticky IPV6_PKTINFO option by
2956 		 * doing a "regular" setsockopt with ipi6_addr being
2957 		 * in6addr_any and ipi6_ifindex being zero.
2958 		 * [RFC 3542, Section 6]
2959 		 */
2960 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2961 		    pktinfo->ipi6_ifindex == 0 &&
2962 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2963 			ip6_clearpktopts(opt, optname);
2964 			break;
2965 		}
2966 
2967 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2968 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2969 			return (EINVAL);
2970 		}
2971 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2972 			return (EINVAL);
2973 		/* validate the interface index if specified. */
2974 		if (pktinfo->ipi6_ifindex) {
2975 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2976 			if (ifp == NULL)
2977 				return (ENXIO);
2978 		}
2979 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2980 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2981 			return (ENETDOWN);
2982 
2983 		if (ifp != NULL &&
2984 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2985 			struct in6_ifaddr *ia;
2986 
2987 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2988 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2989 			if (ia == NULL)
2990 				return (EADDRNOTAVAIL);
2991 			ifa_free(&ia->ia_ifa);
2992 		}
2993 		/*
2994 		 * We store the address anyway, and let in6_selectsrc()
2995 		 * validate the specified address.  This is because ipi6_addr
2996 		 * may not have enough information about its scope zone, and
2997 		 * we may need additional information (such as outgoing
2998 		 * interface or the scope zone of a destination address) to
2999 		 * disambiguate the scope.
3000 		 * XXX: the delay of the validation may confuse the
3001 		 * application when it is used as a sticky option.
3002 		 */
3003 		if (opt->ip6po_pktinfo == NULL) {
3004 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3005 			    M_IP6OPT, M_NOWAIT);
3006 			if (opt->ip6po_pktinfo == NULL)
3007 				return (ENOBUFS);
3008 		}
3009 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3010 		break;
3011 	}
3012 
3013 	case IPV6_2292HOPLIMIT:
3014 	case IPV6_HOPLIMIT:
3015 	{
3016 		int *hlimp;
3017 
3018 		/*
3019 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3020 		 * to simplify the ordering among hoplimit options.
3021 		 */
3022 		if (optname == IPV6_HOPLIMIT && sticky)
3023 			return (ENOPROTOOPT);
3024 
3025 		if (len != sizeof(int))
3026 			return (EINVAL);
3027 		hlimp = (int *)buf;
3028 		if (*hlimp < -1 || *hlimp > 255)
3029 			return (EINVAL);
3030 
3031 		opt->ip6po_hlim = *hlimp;
3032 		break;
3033 	}
3034 
3035 	case IPV6_TCLASS:
3036 	{
3037 		int tclass;
3038 
3039 		if (len != sizeof(int))
3040 			return (EINVAL);
3041 		tclass = *(int *)buf;
3042 		if (tclass < -1 || tclass > 255)
3043 			return (EINVAL);
3044 
3045 		opt->ip6po_tclass = tclass;
3046 		break;
3047 	}
3048 
3049 	case IPV6_2292NEXTHOP:
3050 	case IPV6_NEXTHOP:
3051 		if (cred != NULL) {
3052 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3053 			if (error)
3054 				return (error);
3055 		}
3056 
3057 		if (len == 0) {	/* just remove the option */
3058 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3059 			break;
3060 		}
3061 
3062 		/* check if cmsg_len is large enough for sa_len */
3063 		if (len < sizeof(struct sockaddr) || len < *buf)
3064 			return (EINVAL);
3065 
3066 		switch (((struct sockaddr *)buf)->sa_family) {
3067 		case AF_INET6:
3068 		{
3069 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3070 			int error;
3071 
3072 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3073 				return (EINVAL);
3074 
3075 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3076 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3077 				return (EINVAL);
3078 			}
3079 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3080 			    != 0) {
3081 				return (error);
3082 			}
3083 			break;
3084 		}
3085 		case AF_LINK:	/* should eventually be supported */
3086 		default:
3087 			return (EAFNOSUPPORT);
3088 		}
3089 
3090 		/* turn off the previous option, then set the new option. */
3091 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3092 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3093 		if (opt->ip6po_nexthop == NULL)
3094 			return (ENOBUFS);
3095 		bcopy(buf, opt->ip6po_nexthop, *buf);
3096 		break;
3097 
3098 	case IPV6_2292HOPOPTS:
3099 	case IPV6_HOPOPTS:
3100 	{
3101 		struct ip6_hbh *hbh;
3102 		int hbhlen;
3103 
3104 		/*
3105 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3106 		 * options, since per-option restriction has too much
3107 		 * overhead.
3108 		 */
3109 		if (cred != NULL) {
3110 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3111 			if (error)
3112 				return (error);
3113 		}
3114 
3115 		if (len == 0) {
3116 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3117 			break;	/* just remove the option */
3118 		}
3119 
3120 		/* message length validation */
3121 		if (len < sizeof(struct ip6_hbh))
3122 			return (EINVAL);
3123 		hbh = (struct ip6_hbh *)buf;
3124 		hbhlen = (hbh->ip6h_len + 1) << 3;
3125 		if (len != hbhlen)
3126 			return (EINVAL);
3127 
3128 		/* turn off the previous option, then set the new option. */
3129 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3130 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3131 		if (opt->ip6po_hbh == NULL)
3132 			return (ENOBUFS);
3133 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3134 
3135 		break;
3136 	}
3137 
3138 	case IPV6_2292DSTOPTS:
3139 	case IPV6_DSTOPTS:
3140 	case IPV6_RTHDRDSTOPTS:
3141 	{
3142 		struct ip6_dest *dest, **newdest = NULL;
3143 		int destlen;
3144 
3145 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3146 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3147 			if (error)
3148 				return (error);
3149 		}
3150 
3151 		if (len == 0) {
3152 			ip6_clearpktopts(opt, optname);
3153 			break;	/* just remove the option */
3154 		}
3155 
3156 		/* message length validation */
3157 		if (len < sizeof(struct ip6_dest))
3158 			return (EINVAL);
3159 		dest = (struct ip6_dest *)buf;
3160 		destlen = (dest->ip6d_len + 1) << 3;
3161 		if (len != destlen)
3162 			return (EINVAL);
3163 
3164 		/*
3165 		 * Determine the position that the destination options header
3166 		 * should be inserted; before or after the routing header.
3167 		 */
3168 		switch (optname) {
3169 		case IPV6_2292DSTOPTS:
3170 			/*
3171 			 * The old advacned API is ambiguous on this point.
3172 			 * Our approach is to determine the position based
3173 			 * according to the existence of a routing header.
3174 			 * Note, however, that this depends on the order of the
3175 			 * extension headers in the ancillary data; the 1st
3176 			 * part of the destination options header must appear
3177 			 * before the routing header in the ancillary data,
3178 			 * too.
3179 			 * RFC3542 solved the ambiguity by introducing
3180 			 * separate ancillary data or option types.
3181 			 */
3182 			if (opt->ip6po_rthdr == NULL)
3183 				newdest = &opt->ip6po_dest1;
3184 			else
3185 				newdest = &opt->ip6po_dest2;
3186 			break;
3187 		case IPV6_RTHDRDSTOPTS:
3188 			newdest = &opt->ip6po_dest1;
3189 			break;
3190 		case IPV6_DSTOPTS:
3191 			newdest = &opt->ip6po_dest2;
3192 			break;
3193 		}
3194 
3195 		/* turn off the previous option, then set the new option. */
3196 		ip6_clearpktopts(opt, optname);
3197 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3198 		if (*newdest == NULL)
3199 			return (ENOBUFS);
3200 		bcopy(dest, *newdest, destlen);
3201 
3202 		break;
3203 	}
3204 
3205 	case IPV6_2292RTHDR:
3206 	case IPV6_RTHDR:
3207 	{
3208 		struct ip6_rthdr *rth;
3209 		int rthlen;
3210 
3211 		if (len == 0) {
3212 			ip6_clearpktopts(opt, IPV6_RTHDR);
3213 			break;	/* just remove the option */
3214 		}
3215 
3216 		/* message length validation */
3217 		if (len < sizeof(struct ip6_rthdr))
3218 			return (EINVAL);
3219 		rth = (struct ip6_rthdr *)buf;
3220 		rthlen = (rth->ip6r_len + 1) << 3;
3221 		if (len != rthlen)
3222 			return (EINVAL);
3223 
3224 		switch (rth->ip6r_type) {
3225 		case IPV6_RTHDR_TYPE_0:
3226 			if (rth->ip6r_len == 0)	/* must contain one addr */
3227 				return (EINVAL);
3228 			if (rth->ip6r_len % 2) /* length must be even */
3229 				return (EINVAL);
3230 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3231 				return (EINVAL);
3232 			break;
3233 		default:
3234 			return (EINVAL);	/* not supported */
3235 		}
3236 
3237 		/* turn off the previous option */
3238 		ip6_clearpktopts(opt, IPV6_RTHDR);
3239 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3240 		if (opt->ip6po_rthdr == NULL)
3241 			return (ENOBUFS);
3242 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3243 
3244 		break;
3245 	}
3246 
3247 	case IPV6_USE_MIN_MTU:
3248 		if (len != sizeof(int))
3249 			return (EINVAL);
3250 		minmtupolicy = *(int *)buf;
3251 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3252 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3253 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3254 			return (EINVAL);
3255 		}
3256 		opt->ip6po_minmtu = minmtupolicy;
3257 		break;
3258 
3259 	case IPV6_DONTFRAG:
3260 		if (len != sizeof(int))
3261 			return (EINVAL);
3262 
3263 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3264 			/*
3265 			 * we ignore this option for TCP sockets.
3266 			 * (RFC3542 leaves this case unspecified.)
3267 			 */
3268 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3269 		} else
3270 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3271 		break;
3272 
3273 	case IPV6_PREFER_TEMPADDR:
3274 		if (len != sizeof(int))
3275 			return (EINVAL);
3276 		preftemp = *(int *)buf;
3277 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3278 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3279 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3280 			return (EINVAL);
3281 		}
3282 		opt->ip6po_prefer_tempaddr = preftemp;
3283 		break;
3284 
3285 	default:
3286 		return (ENOPROTOOPT);
3287 	} /* end of switch */
3288 
3289 	return (0);
3290 }
3291 
3292 /*
3293  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3294  * packet to the input queue of a specified interface.  Note that this
3295  * calls the output routine of the loopback "driver", but with an interface
3296  * pointer that might NOT be &loif -- easier than replicating that code here.
3297  */
3298 void
3299 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3300 {
3301 	struct mbuf *copym;
3302 	struct ip6_hdr *ip6;
3303 
3304 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3305 	if (copym == NULL)
3306 		return;
3307 
3308 	/*
3309 	 * Make sure to deep-copy IPv6 header portion in case the data
3310 	 * is in an mbuf cluster, so that we can safely override the IPv6
3311 	 * header portion later.
3312 	 */
3313 	if (!M_WRITABLE(copym) ||
3314 	    copym->m_len < sizeof(struct ip6_hdr)) {
3315 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3316 		if (copym == NULL)
3317 			return;
3318 	}
3319 	ip6 = mtod(copym, struct ip6_hdr *);
3320 	/*
3321 	 * clear embedded scope identifiers if necessary.
3322 	 * in6_clearscope will touch the addresses only when necessary.
3323 	 */
3324 	in6_clearscope(&ip6->ip6_src);
3325 	in6_clearscope(&ip6->ip6_dst);
3326 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3327 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3328 		    CSUM_PSEUDO_HDR;
3329 		copym->m_pkthdr.csum_data = 0xffff;
3330 	}
3331 	if_simloop(ifp, copym, AF_INET6, 0);
3332 }
3333 
3334 /*
3335  * Chop IPv6 header off from the payload.
3336  */
3337 static int
3338 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3339 {
3340 	struct mbuf *mh;
3341 	struct ip6_hdr *ip6;
3342 
3343 	ip6 = mtod(m, struct ip6_hdr *);
3344 	if (m->m_len > sizeof(*ip6)) {
3345 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3346 		if (mh == NULL) {
3347 			m_freem(m);
3348 			return ENOBUFS;
3349 		}
3350 		m_move_pkthdr(mh, m);
3351 		M_ALIGN(mh, sizeof(*ip6));
3352 		m->m_len -= sizeof(*ip6);
3353 		m->m_data += sizeof(*ip6);
3354 		mh->m_next = m;
3355 		m = mh;
3356 		m->m_len = sizeof(*ip6);
3357 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3358 	}
3359 	exthdrs->ip6e_ip6 = m;
3360 	return 0;
3361 }
3362 
3363 /*
3364  * Compute IPv6 extension header length.
3365  */
3366 int
3367 ip6_optlen(struct inpcb *inp)
3368 {
3369 	int len;
3370 
3371 	if (!inp->in6p_outputopts)
3372 		return 0;
3373 
3374 	len = 0;
3375 #define elen(x) \
3376     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3377 
3378 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3379 	if (inp->in6p_outputopts->ip6po_rthdr)
3380 		/* dest1 is valid with rthdr only */
3381 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3382 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3383 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3384 	return len;
3385 #undef elen
3386 }
3387