xref: /freebsd/sys/netinet6/ip6_output.c (revision cfc39718e9cc18943a6f8428c560b02c6f590b16)
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/scope6_var.h>
125 
126 extern int in6_mcast_loop;
127 
128 struct ip6_exthdrs {
129 	struct mbuf *ip6e_ip6;
130 	struct mbuf *ip6e_hbh;
131 	struct mbuf *ip6e_dest1;
132 	struct mbuf *ip6e_rthdr;
133 	struct mbuf *ip6e_dest2;
134 };
135 
136 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
137 
138 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
139 			   struct ucred *, int);
140 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
141 	struct socket *, struct sockopt *);
142 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
143 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
144 	struct ucred *, int, int, int);
145 
146 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
147 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
148 	struct ip6_frag **);
149 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
150 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
151 static int ip6_getpmtu(struct route_in6 *, int,
152 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
153 	u_int);
154 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
155 	u_long *, int *, u_int);
156 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
157 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
158 
159 /*
160  * Make an extension header from option data.  hp is the source,
161  * mp is the destination, and _ol is the optlen.
162  */
163 #define	MAKE_EXTHDR(hp, mp, _ol)					\
164     do {								\
165 	if (hp) {							\
166 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
167 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
168 		    ((eh)->ip6e_len + 1) << 3);				\
169 		if (error)						\
170 			goto freehdrs;					\
171 		(_ol) += (*(mp))->m_len;				\
172 	}								\
173     } while (/*CONSTCOND*/ 0)
174 
175 /*
176  * Form a chain of extension headers.
177  * m is the extension header mbuf
178  * mp is the previous mbuf in the chain
179  * p is the next header
180  * i is the type of option.
181  */
182 #define MAKE_CHAIN(m, mp, p, i)\
183     do {\
184 	if (m) {\
185 		if (!hdrsplit) \
186 			panic("%s:%d: assumption failed: "\
187 			    "hdr not split: hdrsplit %d exthdrs %p",\
188 			    __func__, __LINE__, hdrsplit, &exthdrs);\
189 		*mtod((m), u_char *) = *(p);\
190 		*(p) = (i);\
191 		p = mtod((m), u_char *);\
192 		(m)->m_next = (mp)->m_next;\
193 		(mp)->m_next = (m);\
194 		(mp) = (m);\
195 	}\
196     } while (/*CONSTCOND*/ 0)
197 
198 void
199 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
200 {
201 	u_short csum;
202 
203 	csum = in_cksum_skip(m, offset + plen, offset);
204 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
205 		csum = 0xffff;
206 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
207 
208 	if (offset + sizeof(csum) > m->m_len)
209 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
210 	else
211 		*(u_short *)mtodo(m, offset) = csum;
212 }
213 
214 static void
215 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
216     int plen, int optlen)
217 {
218 
219 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
220 	    "csum_flags %#x",
221 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
222 
223 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
224 		in6_delayed_cksum(m, plen - optlen,
225 		    sizeof(struct ip6_hdr) + optlen);
226 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
227 	}
228 #if defined(SCTP) || defined(SCTP_SUPPORT)
229 	if (csum_flags & CSUM_SCTP_IPV6) {
230 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
231 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
232 	}
233 #endif
234 }
235 
236 int
237 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
238     int fraglen , uint32_t id)
239 {
240 	struct mbuf *m, **mnext, *m_frgpart;
241 	struct ip6_hdr *ip6, *mhip6;
242 	struct ip6_frag *ip6f;
243 	int off;
244 	int error;
245 	int tlen = m0->m_pkthdr.len;
246 
247 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
248 
249 	m = m0;
250 	ip6 = mtod(m, struct ip6_hdr *);
251 	mnext = &m->m_nextpkt;
252 
253 	for (off = hlen; off < tlen; off += fraglen) {
254 		m = m_gethdr(M_NOWAIT, MT_DATA);
255 		if (!m) {
256 			IP6STAT_INC(ip6s_odropped);
257 			return (ENOBUFS);
258 		}
259 
260 		/*
261 		 * Make sure the complete packet header gets copied
262 		 * from the originating mbuf to the newly created
263 		 * mbuf. This also ensures that existing firewall
264 		 * classification(s), VLAN tags and so on get copied
265 		 * to the resulting fragmented packet(s):
266 		 */
267 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
268 			m_free(m);
269 			IP6STAT_INC(ip6s_odropped);
270 			return (ENOBUFS);
271 		}
272 
273 		*mnext = m;
274 		mnext = &m->m_nextpkt;
275 		m->m_data += max_linkhdr;
276 		mhip6 = mtod(m, struct ip6_hdr *);
277 		*mhip6 = *ip6;
278 		m->m_len = sizeof(*mhip6);
279 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
280 		if (error) {
281 			IP6STAT_INC(ip6s_odropped);
282 			return (error);
283 		}
284 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
285 		if (off + fraglen >= tlen)
286 			fraglen = tlen - off;
287 		else
288 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
289 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
290 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
291 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
292 			IP6STAT_INC(ip6s_odropped);
293 			return (ENOBUFS);
294 		}
295 		m_cat(m, m_frgpart);
296 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
297 		ip6f->ip6f_reserved = 0;
298 		ip6f->ip6f_ident = id;
299 		ip6f->ip6f_nxt = nextproto;
300 		IP6STAT_INC(ip6s_ofragments);
301 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
302 	}
303 
304 	return (0);
305 }
306 
307 static int
308 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
309     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
310     bool stamp_tag)
311 {
312 #ifdef KERN_TLS
313 	struct ktls_session *tls = NULL;
314 #endif
315 	struct m_snd_tag *mst;
316 	int error;
317 
318 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
319 	mst = NULL;
320 
321 #ifdef KERN_TLS
322 	/*
323 	 * If this is an unencrypted TLS record, save a reference to
324 	 * the record.  This local reference is used to call
325 	 * ktls_output_eagain after the mbuf has been freed (thus
326 	 * dropping the mbuf's reference) in if_output.
327 	 */
328 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
329 		tls = ktls_hold(m->m_next->m_epg_tls);
330 		mst = tls->snd_tag;
331 
332 		/*
333 		 * If a TLS session doesn't have a valid tag, it must
334 		 * have had an earlier ifp mismatch, so drop this
335 		 * packet.
336 		 */
337 		if (mst == NULL) {
338 			m_freem(m);
339 			error = EAGAIN;
340 			goto done;
341 		}
342 		/*
343 		 * Always stamp tags that include NIC ktls.
344 		 */
345 		stamp_tag = true;
346 	}
347 #endif
348 #ifdef RATELIMIT
349 	if (inp != NULL && mst == NULL) {
350 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
351 		    (inp->inp_snd_tag != NULL &&
352 		    inp->inp_snd_tag->ifp != ifp))
353 			in_pcboutput_txrtlmt(inp, ifp, m);
354 
355 		if (inp->inp_snd_tag != NULL)
356 			mst = inp->inp_snd_tag;
357 	}
358 #endif
359 	if (stamp_tag && mst != NULL) {
360 		KASSERT(m->m_pkthdr.rcvif == NULL,
361 		    ("trying to add a send tag to a forwarded packet"));
362 		if (mst->ifp != ifp) {
363 			m_freem(m);
364 			error = EAGAIN;
365 			goto done;
366 		}
367 
368 		/* stamp send tag on mbuf */
369 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
370 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
371 	}
372 
373 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
374 
375 done:
376 	/* Check for route change invalidating send tags. */
377 #ifdef KERN_TLS
378 	if (tls != NULL) {
379 		if (error == EAGAIN)
380 			error = ktls_output_eagain(inp, tls);
381 		ktls_free(tls);
382 	}
383 #endif
384 #ifdef RATELIMIT
385 	if (error == EAGAIN)
386 		in_pcboutput_eagain(inp);
387 #endif
388 	return (error);
389 }
390 
391 /*
392  * IP6 output.
393  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
394  * nxt, hlim, src, dst).
395  * This function may modify ver and hlim only.
396  * The mbuf chain containing the packet will be freed.
397  * The mbuf opt, if present, will not be freed.
398  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
399  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
400  * then result of route lookup is stored in ro->ro_nh.
401  *
402  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
403  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
404  *
405  * ifpp - XXX: just for statistics
406  */
407 int
408 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
409     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
410     struct ifnet **ifpp, struct inpcb *inp)
411 {
412 	struct ip6_hdr *ip6;
413 	struct ifnet *ifp, *origifp;
414 	struct mbuf *m = m0;
415 	struct mbuf *mprev;
416 	struct route_in6 *ro_pmtu;
417 	struct nhop_object *nh;
418 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
419 	struct in6_addr odst;
420 	u_char *nexthdrp;
421 	int tlen, len;
422 	int error = 0;
423 	int vlan_pcp = -1;
424 	struct in6_ifaddr *ia = NULL;
425 	u_long mtu;
426 	int alwaysfrag, dontfrag;
427 	u_int32_t optlen, plen = 0, unfragpartlen;
428 	struct ip6_exthdrs exthdrs;
429 	struct in6_addr src0, dst0;
430 	u_int32_t zone;
431 	bool hdrsplit;
432 	int sw_csum, tso;
433 	int needfiblookup;
434 	uint32_t fibnum;
435 	struct m_tag *fwd_tag = NULL;
436 	uint32_t id;
437 
438 	NET_EPOCH_ASSERT();
439 
440 	if (inp != NULL) {
441 		INP_LOCK_ASSERT(inp);
442 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
443 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
444 			/* Unconditionally set flowid. */
445 			m->m_pkthdr.flowid = inp->inp_flowid;
446 			M_HASHTYPE_SET(m, inp->inp_flowtype);
447 		}
448 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
449 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
450 			    INP_2PCP_SHIFT;
451 #ifdef NUMA
452 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
453 #endif
454 	}
455 
456 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
457 	/*
458 	 * IPSec checking which handles several cases.
459 	 * FAST IPSEC: We re-injected the packet.
460 	 * XXX: need scope argument.
461 	 */
462 	if (IPSEC_ENABLED(ipv6)) {
463 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
464 			if (error == EINPROGRESS)
465 				error = 0;
466 			goto done;
467 		}
468 	}
469 #endif /* IPSEC */
470 
471 	/* Source address validation. */
472 	ip6 = mtod(m, struct ip6_hdr *);
473 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
474 	    (flags & IPV6_UNSPECSRC) == 0) {
475 		error = EOPNOTSUPP;
476 		IP6STAT_INC(ip6s_badscope);
477 		goto bad;
478 	}
479 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
480 		error = EOPNOTSUPP;
481 		IP6STAT_INC(ip6s_badscope);
482 		goto bad;
483 	}
484 
485 	/*
486 	 * If we are given packet options to add extension headers prepare them.
487 	 * Calculate the total length of the extension header chain.
488 	 * Keep the length of the unfragmentable part for fragmentation.
489 	 */
490 	bzero(&exthdrs, sizeof(exthdrs));
491 	optlen = 0;
492 	unfragpartlen = sizeof(struct ip6_hdr);
493 	if (opt) {
494 		/* Hop-by-Hop options header. */
495 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
496 
497 		/* Destination options header (1st part). */
498 		if (opt->ip6po_rthdr) {
499 #ifndef RTHDR_SUPPORT_IMPLEMENTED
500 			/*
501 			 * If there is a routing header, discard the packet
502 			 * right away here. RH0/1 are obsolete and we do not
503 			 * currently support RH2/3/4.
504 			 * People trying to use RH253/254 may want to disable
505 			 * this check.
506 			 * The moment we do support any routing header (again)
507 			 * this block should check the routing type more
508 			 * selectively.
509 			 */
510 			error = EINVAL;
511 			goto bad;
512 #endif
513 
514 			/*
515 			 * Destination options header (1st part).
516 			 * This only makes sense with a routing header.
517 			 * See Section 9.2 of RFC 3542.
518 			 * Disabling this part just for MIP6 convenience is
519 			 * a bad idea.  We need to think carefully about a
520 			 * way to make the advanced API coexist with MIP6
521 			 * options, which might automatically be inserted in
522 			 * the kernel.
523 			 */
524 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
525 			    optlen);
526 		}
527 		/* Routing header. */
528 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
529 
530 		unfragpartlen += optlen;
531 
532 		/*
533 		 * NOTE: we don't add AH/ESP length here (done in
534 		 * ip6_ipsec_output()).
535 		 */
536 
537 		/* Destination options header (2nd part). */
538 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
539 	}
540 
541 	/*
542 	 * If there is at least one extension header,
543 	 * separate IP6 header from the payload.
544 	 */
545 	hdrsplit = false;
546 	if (optlen) {
547 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
548 			m = NULL;
549 			goto freehdrs;
550 		}
551 		m = exthdrs.ip6e_ip6;
552 		ip6 = mtod(m, struct ip6_hdr *);
553 		hdrsplit = true;
554 	}
555 
556 	/* Adjust mbuf packet header length. */
557 	m->m_pkthdr.len += optlen;
558 	plen = m->m_pkthdr.len - sizeof(*ip6);
559 
560 	/* If this is a jumbo payload, insert a jumbo payload option. */
561 	if (plen > IPV6_MAXPACKET) {
562 		if (!hdrsplit) {
563 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
564 				m = NULL;
565 				goto freehdrs;
566 			}
567 			m = exthdrs.ip6e_ip6;
568 			ip6 = mtod(m, struct ip6_hdr *);
569 			hdrsplit = true;
570 		}
571 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
572 			goto freehdrs;
573 		ip6->ip6_plen = 0;
574 	} else
575 		ip6->ip6_plen = htons(plen);
576 	nexthdrp = &ip6->ip6_nxt;
577 
578 	if (optlen) {
579 		/*
580 		 * Concatenate headers and fill in next header fields.
581 		 * Here we have, on "m"
582 		 *	IPv6 payload
583 		 * and we insert headers accordingly.
584 		 * Finally, we should be getting:
585 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
586 		 *
587 		 * During the header composing process "m" points to IPv6
588 		 * header.  "mprev" points to an extension header prior to esp.
589 		 */
590 		mprev = m;
591 
592 		/*
593 		 * We treat dest2 specially.  This makes IPsec processing
594 		 * much easier.  The goal here is to make mprev point the
595 		 * mbuf prior to dest2.
596 		 *
597 		 * Result: IPv6 dest2 payload.
598 		 * m and mprev will point to IPv6 header.
599 		 */
600 		if (exthdrs.ip6e_dest2) {
601 			if (!hdrsplit)
602 				panic("%s:%d: assumption failed: "
603 				    "hdr not split: hdrsplit %d exthdrs %p",
604 				    __func__, __LINE__, hdrsplit, &exthdrs);
605 			exthdrs.ip6e_dest2->m_next = m->m_next;
606 			m->m_next = exthdrs.ip6e_dest2;
607 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
608 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
609 		}
610 
611 		/*
612 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
613 		 * m will point to IPv6 header.  mprev will point to the
614 		 * extension header prior to dest2 (rthdr in the above case).
615 		 */
616 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
617 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
618 			   IPPROTO_DSTOPTS);
619 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
620 			   IPPROTO_ROUTING);
621 	}
622 
623 	IP6STAT_INC(ip6s_localout);
624 
625 	/* Route packet. */
626 	ro_pmtu = ro;
627 	if (opt && opt->ip6po_rthdr)
628 		ro = &opt->ip6po_route;
629 	if (ro != NULL)
630 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
631 	else
632 		dst = &sin6;
633 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
634 
635 again:
636 	/*
637 	 * If specified, try to fill in the traffic class field.
638 	 * Do not override if a non-zero value is already set.
639 	 * We check the diffserv field and the ECN field separately.
640 	 */
641 	if (opt && opt->ip6po_tclass >= 0) {
642 		int mask = 0;
643 
644 		if (IPV6_DSCP(ip6) == 0)
645 			mask |= 0xfc;
646 		if (IPV6_ECN(ip6) == 0)
647 			mask |= 0x03;
648 		if (mask != 0)
649 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
650 	}
651 
652 	/* Fill in or override the hop limit field, if necessary. */
653 	if (opt && opt->ip6po_hlim != -1)
654 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
655 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
656 		if (im6o != NULL)
657 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
658 		else
659 			ip6->ip6_hlim = V_ip6_defmcasthlim;
660 	}
661 
662 	if (ro == NULL || ro->ro_nh == NULL) {
663 		bzero(dst, sizeof(*dst));
664 		dst->sin6_family = AF_INET6;
665 		dst->sin6_len = sizeof(*dst);
666 		dst->sin6_addr = ip6->ip6_dst;
667 	}
668 	/*
669 	 * Validate route against routing table changes.
670 	 * Make sure that the address family is set in route.
671 	 */
672 	nh = NULL;
673 	ifp = NULL;
674 	mtu = 0;
675 	if (ro != NULL) {
676 		if (ro->ro_nh != NULL && inp != NULL) {
677 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
678 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
679 			    fibnum);
680 		}
681 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
682 		    (!NH_IS_VALID(ro->ro_nh) ||
683 		    ro->ro_dst.sin6_family != AF_INET6 ||
684 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
685 			RO_INVALIDATE_CACHE(ro);
686 
687 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
688 		    ro->ro_dst.sin6_family == AF_INET6 &&
689 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
690 			/* Nexthop is valid and contains valid ifp */
691 			nh = ro->ro_nh;
692 		} else {
693 			if (ro->ro_lle)
694 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
695 			ro->ro_lle = NULL;
696 			if (fwd_tag == NULL) {
697 				bzero(&dst_sa, sizeof(dst_sa));
698 				dst_sa.sin6_family = AF_INET6;
699 				dst_sa.sin6_len = sizeof(dst_sa);
700 				dst_sa.sin6_addr = ip6->ip6_dst;
701 			}
702 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
703 			    &nh, fibnum, m->m_pkthdr.flowid);
704 			if (error != 0) {
705 				IP6STAT_INC(ip6s_noroute);
706 				if (ifp != NULL)
707 					in6_ifstat_inc(ifp, ifs6_out_discard);
708 				goto bad;
709 			}
710 			/*
711 			 * At this point at least @ifp is not NULL
712 			 * Can be the case when dst is multicast, link-local or
713 			 * interface is explicitly specificed by the caller.
714 			 */
715 		}
716 		if (nh == NULL) {
717 			/*
718 			 * If in6_selectroute() does not return a nexthop
719 			 * dst may not have been updated.
720 			 */
721 			*dst = dst_sa;	/* XXX */
722 			origifp = ifp;
723 			mtu = ifp->if_mtu;
724 		} else {
725 			ifp = nh->nh_ifp;
726 			origifp = nh->nh_aifp;
727 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
728 			counter_u64_add(nh->nh_pksent, 1);
729 		}
730 	} else {
731 		struct nhop_object *nh;
732 		struct in6_addr kdst;
733 		uint32_t scopeid;
734 
735 		if (fwd_tag == NULL) {
736 			bzero(&dst_sa, sizeof(dst_sa));
737 			dst_sa.sin6_family = AF_INET6;
738 			dst_sa.sin6_len = sizeof(dst_sa);
739 			dst_sa.sin6_addr = ip6->ip6_dst;
740 		}
741 
742 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
743 		    im6o != NULL &&
744 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
745 			/* We do not need a route lookup. */
746 			*dst = dst_sa;	/* XXX */
747 			origifp = ifp;
748 			goto nonh6lookup;
749 		}
750 
751 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
752 
753 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
754 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
755 			if (scopeid > 0) {
756 				ifp = in6_getlinkifnet(scopeid);
757 				if (ifp == NULL) {
758 					error = EHOSTUNREACH;
759 					goto bad;
760 				}
761 				*dst = dst_sa;	/* XXX */
762 				origifp = ifp;
763 				goto nonh6lookup;
764 			}
765 		}
766 
767 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
768 		    m->m_pkthdr.flowid);
769 		if (nh == NULL) {
770 			IP6STAT_INC(ip6s_noroute);
771 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
772 			error = EHOSTUNREACH;;
773 			goto bad;
774 		}
775 
776 		ifp = nh->nh_ifp;
777 		origifp = nh->nh_aifp;
778 		ia = ifatoia6(nh->nh_ifa);
779 		if (nh->nh_flags & NHF_GATEWAY)
780 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
781 		else if (fwd_tag != NULL)
782 			dst->sin6_addr = dst_sa.sin6_addr;
783 nonh6lookup:
784 		;
785 	}
786 	/*
787 	 * At this point ifp MUST be pointing to the valid transmit ifp.
788 	 * origifp MUST be valid and pointing to either the same ifp or,
789 	 * in case of loopback output, to the interface which ip6_src
790 	 * belongs to.
791 	 * Examples:
792 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
793 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
794 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
795 	 *
796 	 * mtu can be 0 and will be refined later.
797 	 */
798 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
799 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
800 
801 	if ((flags & IPV6_FORWARDING) == 0) {
802 		/* XXX: the FORWARDING flag can be set for mrouting. */
803 		in6_ifstat_inc(ifp, ifs6_out_request);
804 	}
805 
806 	/* Setup data structures for scope ID checks. */
807 	src0 = ip6->ip6_src;
808 	bzero(&src_sa, sizeof(src_sa));
809 	src_sa.sin6_family = AF_INET6;
810 	src_sa.sin6_len = sizeof(src_sa);
811 	src_sa.sin6_addr = ip6->ip6_src;
812 
813 	dst0 = ip6->ip6_dst;
814 	/* Re-initialize to be sure. */
815 	bzero(&dst_sa, sizeof(dst_sa));
816 	dst_sa.sin6_family = AF_INET6;
817 	dst_sa.sin6_len = sizeof(dst_sa);
818 	dst_sa.sin6_addr = ip6->ip6_dst;
819 
820 	/* Check for valid scope ID. */
821 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
822 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
823 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
824 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
825 		/*
826 		 * The outgoing interface is in the zone of the source
827 		 * and destination addresses.
828 		 *
829 		 */
830 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
831 	    sa6_recoverscope(&src_sa) != 0 ||
832 	    sa6_recoverscope(&dst_sa) != 0 ||
833 	    dst_sa.sin6_scope_id == 0 ||
834 	    (src_sa.sin6_scope_id != 0 &&
835 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
836 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
837 		/*
838 		 * If the destination network interface is not a
839 		 * loopback interface, or the destination network
840 		 * address has no scope ID, or the source address has
841 		 * a scope ID set which is different from the
842 		 * destination address one, or there is no network
843 		 * interface representing this scope ID, the address
844 		 * pair is considered invalid.
845 		 */
846 		IP6STAT_INC(ip6s_badscope);
847 		in6_ifstat_inc(origifp, ifs6_out_discard);
848 		if (error == 0)
849 			error = EHOSTUNREACH; /* XXX */
850 		goto bad;
851 	}
852 	/* All scope ID checks are successful. */
853 
854 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
855 		if (opt && opt->ip6po_nextroute.ro_nh) {
856 			/*
857 			 * The nexthop is explicitly specified by the
858 			 * application.  We assume the next hop is an IPv6
859 			 * address.
860 			 */
861 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
862 		}
863 		else if ((nh->nh_flags & NHF_GATEWAY))
864 			dst = &nh->gw6_sa;
865 	}
866 
867 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
868 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
869 	} else {
870 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
871 		in6_ifstat_inc(ifp, ifs6_out_mcast);
872 
873 		/* Confirm that the outgoing interface supports multicast. */
874 		if (!(ifp->if_flags & IFF_MULTICAST)) {
875 			IP6STAT_INC(ip6s_noroute);
876 			in6_ifstat_inc(ifp, ifs6_out_discard);
877 			error = ENETUNREACH;
878 			goto bad;
879 		}
880 		if ((im6o == NULL && in6_mcast_loop) ||
881 		    (im6o && im6o->im6o_multicast_loop)) {
882 			/*
883 			 * Loop back multicast datagram if not expressly
884 			 * forbidden to do so, even if we have not joined
885 			 * the address; protocols will filter it later,
886 			 * thus deferring a hash lookup and lock acquisition
887 			 * at the expense of an m_copym().
888 			 */
889 			ip6_mloopback(ifp, m);
890 		} else {
891 			/*
892 			 * If we are acting as a multicast router, perform
893 			 * multicast forwarding as if the packet had just
894 			 * arrived on the interface to which we are about
895 			 * to send.  The multicast forwarding function
896 			 * recursively calls this function, using the
897 			 * IPV6_FORWARDING flag to prevent infinite recursion.
898 			 *
899 			 * Multicasts that are looped back by ip6_mloopback(),
900 			 * above, will be forwarded by the ip6_input() routine,
901 			 * if necessary.
902 			 */
903 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
904 				/*
905 				 * XXX: ip6_mforward expects that rcvif is NULL
906 				 * when it is called from the originating path.
907 				 * However, it may not always be the case.
908 				 */
909 				m->m_pkthdr.rcvif = NULL;
910 				if (ip6_mforward(ip6, ifp, m) != 0) {
911 					m_freem(m);
912 					goto done;
913 				}
914 			}
915 		}
916 		/*
917 		 * Multicasts with a hoplimit of zero may be looped back,
918 		 * above, but must not be transmitted on a network.
919 		 * Also, multicasts addressed to the loopback interface
920 		 * are not sent -- the above call to ip6_mloopback() will
921 		 * loop back a copy if this host actually belongs to the
922 		 * destination group on the loopback interface.
923 		 */
924 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
925 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
926 			m_freem(m);
927 			goto done;
928 		}
929 	}
930 
931 	/*
932 	 * Fill the outgoing inteface to tell the upper layer
933 	 * to increment per-interface statistics.
934 	 */
935 	if (ifpp)
936 		*ifpp = ifp;
937 
938 	/* Determine path MTU. */
939 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
940 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
941 		goto bad;
942 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
943 	    "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
944 	    ifp, alwaysfrag, fibnum));
945 
946 	/*
947 	 * The caller of this function may specify to use the minimum MTU
948 	 * in some cases.
949 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
950 	 * setting.  The logic is a bit complicated; by default, unicast
951 	 * packets will follow path MTU while multicast packets will be sent at
952 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
953 	 * including unicast ones will be sent at the minimum MTU.  Multicast
954 	 * packets will always be sent at the minimum MTU unless
955 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
956 	 * See RFC 3542 for more details.
957 	 */
958 	if (mtu > IPV6_MMTU) {
959 		if ((flags & IPV6_MINMTU))
960 			mtu = IPV6_MMTU;
961 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
962 			mtu = IPV6_MMTU;
963 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
964 			 (opt == NULL ||
965 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
966 			mtu = IPV6_MMTU;
967 		}
968 	}
969 
970 	/*
971 	 * Clear embedded scope identifiers if necessary.
972 	 * in6_clearscope() will touch the addresses only when necessary.
973 	 */
974 	in6_clearscope(&ip6->ip6_src);
975 	in6_clearscope(&ip6->ip6_dst);
976 
977 	/*
978 	 * If the outgoing packet contains a hop-by-hop options header,
979 	 * it must be examined and processed even by the source node.
980 	 * (RFC 2460, section 4.)
981 	 */
982 	if (exthdrs.ip6e_hbh) {
983 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
984 		u_int32_t dummy; /* XXX unused */
985 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
986 
987 #ifdef DIAGNOSTIC
988 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
989 			panic("ip6e_hbh is not contiguous");
990 #endif
991 		/*
992 		 *  XXX: if we have to send an ICMPv6 error to the sender,
993 		 *       we need the M_LOOP flag since icmp6_error() expects
994 		 *       the IPv6 and the hop-by-hop options header are
995 		 *       contiguous unless the flag is set.
996 		 */
997 		m->m_flags |= M_LOOP;
998 		m->m_pkthdr.rcvif = ifp;
999 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1000 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1001 		    &dummy, &plen) < 0) {
1002 			/* m was already freed at this point. */
1003 			error = EINVAL;/* better error? */
1004 			goto done;
1005 		}
1006 		m->m_flags &= ~M_LOOP; /* XXX */
1007 		m->m_pkthdr.rcvif = NULL;
1008 	}
1009 
1010 	/* Jump over all PFIL processing if hooks are not active. */
1011 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1012 		goto passout;
1013 
1014 	odst = ip6->ip6_dst;
1015 	/* Run through list of hooks for output packets. */
1016 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1017 	case PFIL_PASS:
1018 		ip6 = mtod(m, struct ip6_hdr *);
1019 		break;
1020 	case PFIL_DROPPED:
1021 		error = EACCES;
1022 		/* FALLTHROUGH */
1023 	case PFIL_CONSUMED:
1024 		goto done;
1025 	}
1026 
1027 	needfiblookup = 0;
1028 	/* See if destination IP address was changed by packet filter. */
1029 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1030 		m->m_flags |= M_SKIP_FIREWALL;
1031 		/* If destination is now ourself drop to ip6_input(). */
1032 		if (in6_localip(&ip6->ip6_dst)) {
1033 			m->m_flags |= M_FASTFWD_OURS;
1034 			if (m->m_pkthdr.rcvif == NULL)
1035 				m->m_pkthdr.rcvif = V_loif;
1036 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1037 				m->m_pkthdr.csum_flags |=
1038 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1039 				m->m_pkthdr.csum_data = 0xffff;
1040 			}
1041 #if defined(SCTP) || defined(SCTP_SUPPORT)
1042 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1043 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1044 #endif
1045 			error = netisr_queue(NETISR_IPV6, m);
1046 			goto done;
1047 		} else {
1048 			if (ro != NULL)
1049 				RO_INVALIDATE_CACHE(ro);
1050 			needfiblookup = 1; /* Redo the routing table lookup. */
1051 		}
1052 	}
1053 	/* See if fib was changed by packet filter. */
1054 	if (fibnum != M_GETFIB(m)) {
1055 		m->m_flags |= M_SKIP_FIREWALL;
1056 		fibnum = M_GETFIB(m);
1057 		if (ro != NULL)
1058 			RO_INVALIDATE_CACHE(ro);
1059 		needfiblookup = 1;
1060 	}
1061 	if (needfiblookup)
1062 		goto again;
1063 
1064 	/* See if local, if yes, send it to netisr. */
1065 	if (m->m_flags & M_FASTFWD_OURS) {
1066 		if (m->m_pkthdr.rcvif == NULL)
1067 			m->m_pkthdr.rcvif = V_loif;
1068 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1069 			m->m_pkthdr.csum_flags |=
1070 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1071 			m->m_pkthdr.csum_data = 0xffff;
1072 		}
1073 #if defined(SCTP) || defined(SCTP_SUPPORT)
1074 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1075 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1076 #endif
1077 		error = netisr_queue(NETISR_IPV6, m);
1078 		goto done;
1079 	}
1080 	/* Or forward to some other address? */
1081 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1082 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1083 		if (ro != NULL)
1084 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1085 		else
1086 			dst = &sin6;
1087 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1088 		m->m_flags |= M_SKIP_FIREWALL;
1089 		m->m_flags &= ~M_IP6_NEXTHOP;
1090 		m_tag_delete(m, fwd_tag);
1091 		goto again;
1092 	}
1093 
1094 passout:
1095 	if (vlan_pcp > -1)
1096 		EVL_APPLY_PRI(m, vlan_pcp);
1097 
1098 	/* Ensure the packet data is mapped if the interface requires it. */
1099 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1100 		m = mb_unmapped_to_ext(m);
1101 		if (m == NULL) {
1102 			IP6STAT_INC(ip6s_odropped);
1103 			return (ENOBUFS);
1104 		}
1105 	}
1106 
1107 	/*
1108 	 * Send the packet to the outgoing interface.
1109 	 * If necessary, do IPv6 fragmentation before sending.
1110 	 *
1111 	 * The logic here is rather complex:
1112 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1113 	 * 1-a:	send as is if tlen <= path mtu
1114 	 * 1-b:	fragment if tlen > path mtu
1115 	 *
1116 	 * 2: if user asks us not to fragment (dontfrag == 1)
1117 	 * 2-a:	send as is if tlen <= interface mtu
1118 	 * 2-b:	error if tlen > interface mtu
1119 	 *
1120 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1121 	 *	always fragment
1122 	 *
1123 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1124 	 *	error, as we cannot handle this conflicting request.
1125 	 */
1126 	sw_csum = m->m_pkthdr.csum_flags;
1127 	if (!hdrsplit) {
1128 		tso = ((sw_csum & ifp->if_hwassist &
1129 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1130 		sw_csum &= ~ifp->if_hwassist;
1131 	} else
1132 		tso = 0;
1133 	/*
1134 	 * If we added extension headers, we will not do TSO and calculate the
1135 	 * checksums ourselves for now.
1136 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1137 	 * with ext. hdrs.
1138 	 */
1139 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1140 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1141 	tlen = m->m_pkthdr.len;
1142 
1143 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1144 		dontfrag = 1;
1145 	else
1146 		dontfrag = 0;
1147 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1148 		/* Conflicting request - can't transmit. */
1149 		error = EMSGSIZE;
1150 		goto bad;
1151 	}
1152 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1153 		/*
1154 		 * Even if the DONTFRAG option is specified, we cannot send the
1155 		 * packet when the data length is larger than the MTU of the
1156 		 * outgoing interface.
1157 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1158 		 * application wanted to know the MTU value. Also return an
1159 		 * error code (this is not described in the API spec).
1160 		 */
1161 		if (inp != NULL)
1162 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1163 		error = EMSGSIZE;
1164 		goto bad;
1165 	}
1166 
1167 	/* Transmit packet without fragmentation. */
1168 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1169 		struct in6_ifaddr *ia6;
1170 
1171 		ip6 = mtod(m, struct ip6_hdr *);
1172 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1173 		if (ia6) {
1174 			/* Record statistics for this interface address. */
1175 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1176 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1177 			    m->m_pkthdr.len);
1178 		}
1179 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1180 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1181 		goto done;
1182 	}
1183 
1184 	/* Try to fragment the packet.  Cases 1-b and 3. */
1185 	if (mtu < IPV6_MMTU) {
1186 		/* Path MTU cannot be less than IPV6_MMTU. */
1187 		error = EMSGSIZE;
1188 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1189 		goto bad;
1190 	} else if (ip6->ip6_plen == 0) {
1191 		/* Jumbo payload cannot be fragmented. */
1192 		error = EMSGSIZE;
1193 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1194 		goto bad;
1195 	} else {
1196 		u_char nextproto;
1197 
1198 		/*
1199 		 * Too large for the destination or interface;
1200 		 * fragment if possible.
1201 		 * Must be able to put at least 8 bytes per fragment.
1202 		 */
1203 		if (mtu > IPV6_MAXPACKET)
1204 			mtu = IPV6_MAXPACKET;
1205 
1206 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1207 		if (len < 8) {
1208 			error = EMSGSIZE;
1209 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1210 			goto bad;
1211 		}
1212 
1213 		/*
1214 		 * If the interface will not calculate checksums on
1215 		 * fragmented packets, then do it here.
1216 		 * XXX-BZ handle the hw offloading case.  Need flags.
1217 		 */
1218 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1219 		    optlen);
1220 
1221 		/*
1222 		 * Change the next header field of the last header in the
1223 		 * unfragmentable part.
1224 		 */
1225 		if (exthdrs.ip6e_rthdr) {
1226 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1227 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1228 		} else if (exthdrs.ip6e_dest1) {
1229 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1230 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1231 		} else if (exthdrs.ip6e_hbh) {
1232 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1233 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1234 		} else {
1235 			ip6 = mtod(m, struct ip6_hdr *);
1236 			nextproto = ip6->ip6_nxt;
1237 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1238 		}
1239 
1240 		/*
1241 		 * Loop through length of segment after first fragment,
1242 		 * make new header and copy data of each part and link onto
1243 		 * chain.
1244 		 */
1245 		m0 = m;
1246 		id = htonl(ip6_randomid());
1247 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1248 		if (error != 0)
1249 			goto sendorfree;
1250 
1251 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1252 	}
1253 
1254 	/* Remove leading garbage. */
1255 sendorfree:
1256 	m = m0->m_nextpkt;
1257 	m0->m_nextpkt = 0;
1258 	m_freem(m0);
1259 	for (; m; m = m0) {
1260 		m0 = m->m_nextpkt;
1261 		m->m_nextpkt = 0;
1262 		if (error == 0) {
1263 			/* Record statistics for this interface address. */
1264 			if (ia) {
1265 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1266 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1267 				    m->m_pkthdr.len);
1268 			}
1269 			if (vlan_pcp > -1)
1270 				EVL_APPLY_PRI(m, vlan_pcp);
1271 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1272 			    true);
1273 		} else
1274 			m_freem(m);
1275 	}
1276 
1277 	if (error == 0)
1278 		IP6STAT_INC(ip6s_fragmented);
1279 
1280 done:
1281 	return (error);
1282 
1283 freehdrs:
1284 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1285 	m_freem(exthdrs.ip6e_dest1);
1286 	m_freem(exthdrs.ip6e_rthdr);
1287 	m_freem(exthdrs.ip6e_dest2);
1288 	/* FALLTHROUGH */
1289 bad:
1290 	if (m)
1291 		m_freem(m);
1292 	goto done;
1293 }
1294 
1295 static int
1296 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1297 {
1298 	struct mbuf *m;
1299 
1300 	if (hlen > MCLBYTES)
1301 		return (ENOBUFS); /* XXX */
1302 
1303 	if (hlen > MLEN)
1304 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1305 	else
1306 		m = m_get(M_NOWAIT, MT_DATA);
1307 	if (m == NULL)
1308 		return (ENOBUFS);
1309 	m->m_len = hlen;
1310 	if (hdr)
1311 		bcopy(hdr, mtod(m, caddr_t), hlen);
1312 
1313 	*mp = m;
1314 	return (0);
1315 }
1316 
1317 /*
1318  * Insert jumbo payload option.
1319  */
1320 static int
1321 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1322 {
1323 	struct mbuf *mopt;
1324 	u_char *optbuf;
1325 	u_int32_t v;
1326 
1327 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1328 
1329 	/*
1330 	 * If there is no hop-by-hop options header, allocate new one.
1331 	 * If there is one but it doesn't have enough space to store the
1332 	 * jumbo payload option, allocate a cluster to store the whole options.
1333 	 * Otherwise, use it to store the options.
1334 	 */
1335 	if (exthdrs->ip6e_hbh == NULL) {
1336 		mopt = m_get(M_NOWAIT, MT_DATA);
1337 		if (mopt == NULL)
1338 			return (ENOBUFS);
1339 		mopt->m_len = JUMBOOPTLEN;
1340 		optbuf = mtod(mopt, u_char *);
1341 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1342 		exthdrs->ip6e_hbh = mopt;
1343 	} else {
1344 		struct ip6_hbh *hbh;
1345 
1346 		mopt = exthdrs->ip6e_hbh;
1347 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1348 			/*
1349 			 * XXX assumption:
1350 			 * - exthdrs->ip6e_hbh is not referenced from places
1351 			 *   other than exthdrs.
1352 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1353 			 */
1354 			int oldoptlen = mopt->m_len;
1355 			struct mbuf *n;
1356 
1357 			/*
1358 			 * XXX: give up if the whole (new) hbh header does
1359 			 * not fit even in an mbuf cluster.
1360 			 */
1361 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1362 				return (ENOBUFS);
1363 
1364 			/*
1365 			 * As a consequence, we must always prepare a cluster
1366 			 * at this point.
1367 			 */
1368 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1369 			if (n == NULL)
1370 				return (ENOBUFS);
1371 			n->m_len = oldoptlen + JUMBOOPTLEN;
1372 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1373 			    oldoptlen);
1374 			optbuf = mtod(n, caddr_t) + oldoptlen;
1375 			m_freem(mopt);
1376 			mopt = exthdrs->ip6e_hbh = n;
1377 		} else {
1378 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1379 			mopt->m_len += JUMBOOPTLEN;
1380 		}
1381 		optbuf[0] = IP6OPT_PADN;
1382 		optbuf[1] = 1;
1383 
1384 		/*
1385 		 * Adjust the header length according to the pad and
1386 		 * the jumbo payload option.
1387 		 */
1388 		hbh = mtod(mopt, struct ip6_hbh *);
1389 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1390 	}
1391 
1392 	/* fill in the option. */
1393 	optbuf[2] = IP6OPT_JUMBO;
1394 	optbuf[3] = 4;
1395 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1396 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1397 
1398 	/* finally, adjust the packet header length */
1399 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1400 
1401 	return (0);
1402 #undef JUMBOOPTLEN
1403 }
1404 
1405 /*
1406  * Insert fragment header and copy unfragmentable header portions.
1407  */
1408 static int
1409 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1410     struct ip6_frag **frghdrp)
1411 {
1412 	struct mbuf *n, *mlast;
1413 
1414 	if (hlen > sizeof(struct ip6_hdr)) {
1415 		n = m_copym(m0, sizeof(struct ip6_hdr),
1416 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1417 		if (n == NULL)
1418 			return (ENOBUFS);
1419 		m->m_next = n;
1420 	} else
1421 		n = m;
1422 
1423 	/* Search for the last mbuf of unfragmentable part. */
1424 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1425 		;
1426 
1427 	if (M_WRITABLE(mlast) &&
1428 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1429 		/* use the trailing space of the last mbuf for the fragment hdr */
1430 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1431 		    mlast->m_len);
1432 		mlast->m_len += sizeof(struct ip6_frag);
1433 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1434 	} else {
1435 		/* allocate a new mbuf for the fragment header */
1436 		struct mbuf *mfrg;
1437 
1438 		mfrg = m_get(M_NOWAIT, MT_DATA);
1439 		if (mfrg == NULL)
1440 			return (ENOBUFS);
1441 		mfrg->m_len = sizeof(struct ip6_frag);
1442 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1443 		mlast->m_next = mfrg;
1444 	}
1445 
1446 	return (0);
1447 }
1448 
1449 /*
1450  * Calculates IPv6 path mtu for destination @dst.
1451  * Resulting MTU is stored in @mtup.
1452  *
1453  * Returns 0 on success.
1454  */
1455 static int
1456 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1457 {
1458 	struct epoch_tracker et;
1459 	struct nhop_object *nh;
1460 	struct in6_addr kdst;
1461 	uint32_t scopeid;
1462 	int error;
1463 
1464 	in6_splitscope(dst, &kdst, &scopeid);
1465 
1466 	NET_EPOCH_ENTER(et);
1467 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1468 	if (nh != NULL)
1469 		error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1470 	else
1471 		error = EHOSTUNREACH;
1472 	NET_EPOCH_EXIT(et);
1473 
1474 	return (error);
1475 }
1476 
1477 /*
1478  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1479  * and cached data in @ro_pmtu.
1480  * MTU from (successful) route lookup is saved (along with dst)
1481  * inside @ro_pmtu to avoid subsequent route lookups after packet
1482  * filter processing.
1483  *
1484  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1485  * Returns 0 on success.
1486  */
1487 static int
1488 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1489     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1490     int *alwaysfragp, u_int fibnum, u_int proto)
1491 {
1492 	struct nhop_object *nh;
1493 	struct in6_addr kdst;
1494 	uint32_t scopeid;
1495 	struct sockaddr_in6 *sa6_dst, sin6;
1496 	u_long mtu;
1497 
1498 	NET_EPOCH_ASSERT();
1499 
1500 	mtu = 0;
1501 	if (ro_pmtu == NULL || do_lookup) {
1502 		/*
1503 		 * Here ro_pmtu has final destination address, while
1504 		 * ro might represent immediate destination.
1505 		 * Use ro_pmtu destination since mtu might differ.
1506 		 */
1507 		if (ro_pmtu != NULL) {
1508 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1509 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1510 				ro_pmtu->ro_mtu = 0;
1511 		} else
1512 			sa6_dst = &sin6;
1513 
1514 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1515 			bzero(sa6_dst, sizeof(*sa6_dst));
1516 			sa6_dst->sin6_family = AF_INET6;
1517 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1518 			sa6_dst->sin6_addr = *dst;
1519 
1520 			in6_splitscope(dst, &kdst, &scopeid);
1521 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1522 			if (nh != NULL) {
1523 				mtu = nh->nh_mtu;
1524 				if (ro_pmtu != NULL)
1525 					ro_pmtu->ro_mtu = mtu;
1526 			}
1527 		} else
1528 			mtu = ro_pmtu->ro_mtu;
1529 	}
1530 
1531 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1532 		mtu = ro_pmtu->ro_nh->nh_mtu;
1533 
1534 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1535 }
1536 
1537 /*
1538  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1539  * hostcache data for @dst.
1540  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1541  *
1542  * Returns 0 on success.
1543  */
1544 static int
1545 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1546     u_long *mtup, int *alwaysfragp, u_int proto)
1547 {
1548 	u_long mtu = 0;
1549 	int alwaysfrag = 0;
1550 	int error = 0;
1551 
1552 	if (rt_mtu > 0) {
1553 		u_int32_t ifmtu;
1554 		struct in_conninfo inc;
1555 
1556 		bzero(&inc, sizeof(inc));
1557 		inc.inc_flags |= INC_ISIPV6;
1558 		inc.inc6_faddr = *dst;
1559 
1560 		ifmtu = IN6_LINKMTU(ifp);
1561 
1562 		/* TCP is known to react to pmtu changes so skip hc */
1563 		if (proto != IPPROTO_TCP)
1564 			mtu = tcp_hc_getmtu(&inc);
1565 
1566 		if (mtu)
1567 			mtu = min(mtu, rt_mtu);
1568 		else
1569 			mtu = rt_mtu;
1570 		if (mtu == 0)
1571 			mtu = ifmtu;
1572 		else if (mtu < IPV6_MMTU) {
1573 			/*
1574 			 * RFC2460 section 5, last paragraph:
1575 			 * if we record ICMPv6 too big message with
1576 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1577 			 * or smaller, with framgent header attached.
1578 			 * (fragment header is needed regardless from the
1579 			 * packet size, for translators to identify packets)
1580 			 */
1581 			alwaysfrag = 1;
1582 			mtu = IPV6_MMTU;
1583 		}
1584 	} else if (ifp) {
1585 		mtu = IN6_LINKMTU(ifp);
1586 	} else
1587 		error = EHOSTUNREACH; /* XXX */
1588 
1589 	*mtup = mtu;
1590 	if (alwaysfragp)
1591 		*alwaysfragp = alwaysfrag;
1592 	return (error);
1593 }
1594 
1595 /*
1596  * IP6 socket option processing.
1597  */
1598 int
1599 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1600 {
1601 	int optdatalen, uproto;
1602 	void *optdata;
1603 	struct inpcb *inp = sotoinpcb(so);
1604 	int error, optval;
1605 	int level, op, optname;
1606 	int optlen;
1607 	struct thread *td;
1608 #ifdef	RSS
1609 	uint32_t rss_bucket;
1610 	int retval;
1611 #endif
1612 
1613 /*
1614  * Don't use more than a quarter of mbuf clusters.  N.B.:
1615  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1616  * on LP64 architectures, so cast to u_long to avoid undefined
1617  * behavior.  ILP32 architectures cannot have nmbclusters
1618  * large enough to overflow for other reasons.
1619  */
1620 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1621 
1622 	level = sopt->sopt_level;
1623 	op = sopt->sopt_dir;
1624 	optname = sopt->sopt_name;
1625 	optlen = sopt->sopt_valsize;
1626 	td = sopt->sopt_td;
1627 	error = 0;
1628 	optval = 0;
1629 	uproto = (int)so->so_proto->pr_protocol;
1630 
1631 	if (level != IPPROTO_IPV6) {
1632 		error = EINVAL;
1633 
1634 		if (sopt->sopt_level == SOL_SOCKET &&
1635 		    sopt->sopt_dir == SOPT_SET) {
1636 			switch (sopt->sopt_name) {
1637 			case SO_REUSEADDR:
1638 				INP_WLOCK(inp);
1639 				if ((so->so_options & SO_REUSEADDR) != 0)
1640 					inp->inp_flags2 |= INP_REUSEADDR;
1641 				else
1642 					inp->inp_flags2 &= ~INP_REUSEADDR;
1643 				INP_WUNLOCK(inp);
1644 				error = 0;
1645 				break;
1646 			case SO_REUSEPORT:
1647 				INP_WLOCK(inp);
1648 				if ((so->so_options & SO_REUSEPORT) != 0)
1649 					inp->inp_flags2 |= INP_REUSEPORT;
1650 				else
1651 					inp->inp_flags2 &= ~INP_REUSEPORT;
1652 				INP_WUNLOCK(inp);
1653 				error = 0;
1654 				break;
1655 			case SO_REUSEPORT_LB:
1656 				INP_WLOCK(inp);
1657 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1658 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1659 				else
1660 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1661 				INP_WUNLOCK(inp);
1662 				error = 0;
1663 				break;
1664 			case SO_SETFIB:
1665 				INP_WLOCK(inp);
1666 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1667 				INP_WUNLOCK(inp);
1668 				error = 0;
1669 				break;
1670 			case SO_MAX_PACING_RATE:
1671 #ifdef RATELIMIT
1672 				INP_WLOCK(inp);
1673 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1674 				INP_WUNLOCK(inp);
1675 				error = 0;
1676 #else
1677 				error = EOPNOTSUPP;
1678 #endif
1679 				break;
1680 			default:
1681 				break;
1682 			}
1683 		}
1684 	} else {		/* level == IPPROTO_IPV6 */
1685 		switch (op) {
1686 		case SOPT_SET:
1687 			switch (optname) {
1688 			case IPV6_2292PKTOPTIONS:
1689 #ifdef IPV6_PKTOPTIONS
1690 			case IPV6_PKTOPTIONS:
1691 #endif
1692 			{
1693 				struct mbuf *m;
1694 
1695 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1696 					printf("ip6_ctloutput: mbuf limit hit\n");
1697 					error = ENOBUFS;
1698 					break;
1699 				}
1700 
1701 				error = soopt_getm(sopt, &m); /* XXX */
1702 				if (error != 0)
1703 					break;
1704 				error = soopt_mcopyin(sopt, m); /* XXX */
1705 				if (error != 0)
1706 					break;
1707 				INP_WLOCK(inp);
1708 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1709 				    so, sopt);
1710 				INP_WUNLOCK(inp);
1711 				m_freem(m); /* XXX */
1712 				break;
1713 			}
1714 
1715 			/*
1716 			 * Use of some Hop-by-Hop options or some
1717 			 * Destination options, might require special
1718 			 * privilege.  That is, normal applications
1719 			 * (without special privilege) might be forbidden
1720 			 * from setting certain options in outgoing packets,
1721 			 * and might never see certain options in received
1722 			 * packets. [RFC 2292 Section 6]
1723 			 * KAME specific note:
1724 			 *  KAME prevents non-privileged users from sending or
1725 			 *  receiving ANY hbh/dst options in order to avoid
1726 			 *  overhead of parsing options in the kernel.
1727 			 */
1728 			case IPV6_RECVHOPOPTS:
1729 			case IPV6_RECVDSTOPTS:
1730 			case IPV6_RECVRTHDRDSTOPTS:
1731 				if (td != NULL) {
1732 					error = priv_check(td,
1733 					    PRIV_NETINET_SETHDROPTS);
1734 					if (error)
1735 						break;
1736 				}
1737 				/* FALLTHROUGH */
1738 			case IPV6_UNICAST_HOPS:
1739 			case IPV6_HOPLIMIT:
1740 
1741 			case IPV6_RECVPKTINFO:
1742 			case IPV6_RECVHOPLIMIT:
1743 			case IPV6_RECVRTHDR:
1744 			case IPV6_RECVPATHMTU:
1745 			case IPV6_RECVTCLASS:
1746 			case IPV6_RECVFLOWID:
1747 #ifdef	RSS
1748 			case IPV6_RECVRSSBUCKETID:
1749 #endif
1750 			case IPV6_V6ONLY:
1751 			case IPV6_AUTOFLOWLABEL:
1752 			case IPV6_ORIGDSTADDR:
1753 			case IPV6_BINDANY:
1754 			case IPV6_BINDMULTI:
1755 #ifdef	RSS
1756 			case IPV6_RSS_LISTEN_BUCKET:
1757 #endif
1758 			case IPV6_VLAN_PCP:
1759 				if (optname == IPV6_BINDANY && td != NULL) {
1760 					error = priv_check(td,
1761 					    PRIV_NETINET_BINDANY);
1762 					if (error)
1763 						break;
1764 				}
1765 
1766 				if (optlen != sizeof(int)) {
1767 					error = EINVAL;
1768 					break;
1769 				}
1770 				error = sooptcopyin(sopt, &optval,
1771 					sizeof optval, sizeof optval);
1772 				if (error)
1773 					break;
1774 				switch (optname) {
1775 				case IPV6_UNICAST_HOPS:
1776 					if (optval < -1 || optval >= 256)
1777 						error = EINVAL;
1778 					else {
1779 						/* -1 = kernel default */
1780 						inp->in6p_hops = optval;
1781 						if ((inp->inp_vflag &
1782 						     INP_IPV4) != 0)
1783 							inp->inp_ip_ttl = optval;
1784 					}
1785 					break;
1786 #define OPTSET(bit) \
1787 do { \
1788 	INP_WLOCK(inp); \
1789 	if (optval) \
1790 		inp->inp_flags |= (bit); \
1791 	else \
1792 		inp->inp_flags &= ~(bit); \
1793 	INP_WUNLOCK(inp); \
1794 } while (/*CONSTCOND*/ 0)
1795 #define OPTSET2292(bit) \
1796 do { \
1797 	INP_WLOCK(inp); \
1798 	inp->inp_flags |= IN6P_RFC2292; \
1799 	if (optval) \
1800 		inp->inp_flags |= (bit); \
1801 	else \
1802 		inp->inp_flags &= ~(bit); \
1803 	INP_WUNLOCK(inp); \
1804 } while (/*CONSTCOND*/ 0)
1805 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1806 
1807 #define OPTSET2_N(bit, val) do {					\
1808 	if (val)							\
1809 		inp->inp_flags2 |= bit;					\
1810 	else								\
1811 		inp->inp_flags2 &= ~bit;				\
1812 } while (0)
1813 #define OPTSET2(bit, val) do {						\
1814 	INP_WLOCK(inp);							\
1815 	OPTSET2_N(bit, val);						\
1816 	INP_WUNLOCK(inp);						\
1817 } while (0)
1818 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1819 #define OPTSET2292_EXCLUSIVE(bit)					\
1820 do {									\
1821 	INP_WLOCK(inp);							\
1822 	if (OPTBIT(IN6P_RFC2292)) {					\
1823 		error = EINVAL;						\
1824 	} else {							\
1825 		if (optval)						\
1826 			inp->inp_flags |= (bit);			\
1827 		else							\
1828 			inp->inp_flags &= ~(bit);			\
1829 	}								\
1830 	INP_WUNLOCK(inp);						\
1831 } while (/*CONSTCOND*/ 0)
1832 
1833 				case IPV6_RECVPKTINFO:
1834 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1835 					break;
1836 
1837 				case IPV6_HOPLIMIT:
1838 				{
1839 					struct ip6_pktopts **optp;
1840 
1841 					/* cannot mix with RFC2292 */
1842 					if (OPTBIT(IN6P_RFC2292)) {
1843 						error = EINVAL;
1844 						break;
1845 					}
1846 					INP_WLOCK(inp);
1847 					if (inp->inp_flags & INP_DROPPED) {
1848 						INP_WUNLOCK(inp);
1849 						return (ECONNRESET);
1850 					}
1851 					optp = &inp->in6p_outputopts;
1852 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1853 					    (u_char *)&optval, sizeof(optval),
1854 					    optp, (td != NULL) ? td->td_ucred :
1855 					    NULL, uproto);
1856 					INP_WUNLOCK(inp);
1857 					break;
1858 				}
1859 
1860 				case IPV6_RECVHOPLIMIT:
1861 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1862 					break;
1863 
1864 				case IPV6_RECVHOPOPTS:
1865 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1866 					break;
1867 
1868 				case IPV6_RECVDSTOPTS:
1869 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1870 					break;
1871 
1872 				case IPV6_RECVRTHDRDSTOPTS:
1873 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1874 					break;
1875 
1876 				case IPV6_RECVRTHDR:
1877 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1878 					break;
1879 
1880 				case IPV6_RECVPATHMTU:
1881 					/*
1882 					 * We ignore this option for TCP
1883 					 * sockets.
1884 					 * (RFC3542 leaves this case
1885 					 * unspecified.)
1886 					 */
1887 					if (uproto != IPPROTO_TCP)
1888 						OPTSET(IN6P_MTU);
1889 					break;
1890 
1891 				case IPV6_RECVFLOWID:
1892 					OPTSET2(INP_RECVFLOWID, optval);
1893 					break;
1894 
1895 #ifdef	RSS
1896 				case IPV6_RECVRSSBUCKETID:
1897 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1898 					break;
1899 #endif
1900 
1901 				case IPV6_V6ONLY:
1902 					INP_WLOCK(inp);
1903 					if (inp->inp_lport ||
1904 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1905 						/*
1906 						 * The socket is already bound.
1907 						 */
1908 						INP_WUNLOCK(inp);
1909 						error = EINVAL;
1910 						break;
1911 					}
1912 					if (optval) {
1913 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1914 						inp->inp_vflag &= ~INP_IPV4;
1915 					} else {
1916 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1917 						inp->inp_vflag |= INP_IPV4;
1918 					}
1919 					INP_WUNLOCK(inp);
1920 					break;
1921 				case IPV6_RECVTCLASS:
1922 					/* cannot mix with RFC2292 XXX */
1923 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1924 					break;
1925 				case IPV6_AUTOFLOWLABEL:
1926 					OPTSET(IN6P_AUTOFLOWLABEL);
1927 					break;
1928 
1929 				case IPV6_ORIGDSTADDR:
1930 					OPTSET2(INP_ORIGDSTADDR, optval);
1931 					break;
1932 				case IPV6_BINDANY:
1933 					OPTSET(INP_BINDANY);
1934 					break;
1935 
1936 				case IPV6_BINDMULTI:
1937 					OPTSET2(INP_BINDMULTI, optval);
1938 					break;
1939 #ifdef	RSS
1940 				case IPV6_RSS_LISTEN_BUCKET:
1941 					if ((optval >= 0) &&
1942 					    (optval < rss_getnumbuckets())) {
1943 						INP_WLOCK(inp);
1944 						inp->inp_rss_listen_bucket = optval;
1945 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1946 						INP_WUNLOCK(inp);
1947 					} else {
1948 						error = EINVAL;
1949 					}
1950 					break;
1951 #endif
1952 				case IPV6_VLAN_PCP:
1953 					if ((optval >= -1) && (optval <=
1954 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1955 						if (optval == -1) {
1956 							INP_WLOCK(inp);
1957 							inp->inp_flags2 &=
1958 							    ~(INP_2PCP_SET |
1959 							    INP_2PCP_MASK);
1960 							INP_WUNLOCK(inp);
1961 						} else {
1962 							INP_WLOCK(inp);
1963 							inp->inp_flags2 |=
1964 							    INP_2PCP_SET;
1965 							inp->inp_flags2 &=
1966 							    ~INP_2PCP_MASK;
1967 							inp->inp_flags2 |=
1968 							    optval <<
1969 							    INP_2PCP_SHIFT;
1970 							INP_WUNLOCK(inp);
1971 						}
1972 					} else
1973 						error = EINVAL;
1974 					break;
1975 				}
1976 				break;
1977 
1978 			case IPV6_TCLASS:
1979 			case IPV6_DONTFRAG:
1980 			case IPV6_USE_MIN_MTU:
1981 			case IPV6_PREFER_TEMPADDR:
1982 				if (optlen != sizeof(optval)) {
1983 					error = EINVAL;
1984 					break;
1985 				}
1986 				error = sooptcopyin(sopt, &optval,
1987 					sizeof optval, sizeof optval);
1988 				if (error)
1989 					break;
1990 				{
1991 					struct ip6_pktopts **optp;
1992 					INP_WLOCK(inp);
1993 					if (inp->inp_flags & INP_DROPPED) {
1994 						INP_WUNLOCK(inp);
1995 						return (ECONNRESET);
1996 					}
1997 					optp = &inp->in6p_outputopts;
1998 					error = ip6_pcbopt(optname,
1999 					    (u_char *)&optval, sizeof(optval),
2000 					    optp, (td != NULL) ? td->td_ucred :
2001 					    NULL, uproto);
2002 					INP_WUNLOCK(inp);
2003 					break;
2004 				}
2005 
2006 			case IPV6_2292PKTINFO:
2007 			case IPV6_2292HOPLIMIT:
2008 			case IPV6_2292HOPOPTS:
2009 			case IPV6_2292DSTOPTS:
2010 			case IPV6_2292RTHDR:
2011 				/* RFC 2292 */
2012 				if (optlen != sizeof(int)) {
2013 					error = EINVAL;
2014 					break;
2015 				}
2016 				error = sooptcopyin(sopt, &optval,
2017 					sizeof optval, sizeof optval);
2018 				if (error)
2019 					break;
2020 				switch (optname) {
2021 				case IPV6_2292PKTINFO:
2022 					OPTSET2292(IN6P_PKTINFO);
2023 					break;
2024 				case IPV6_2292HOPLIMIT:
2025 					OPTSET2292(IN6P_HOPLIMIT);
2026 					break;
2027 				case IPV6_2292HOPOPTS:
2028 					/*
2029 					 * Check super-user privilege.
2030 					 * See comments for IPV6_RECVHOPOPTS.
2031 					 */
2032 					if (td != NULL) {
2033 						error = priv_check(td,
2034 						    PRIV_NETINET_SETHDROPTS);
2035 						if (error)
2036 							return (error);
2037 					}
2038 					OPTSET2292(IN6P_HOPOPTS);
2039 					break;
2040 				case IPV6_2292DSTOPTS:
2041 					if (td != NULL) {
2042 						error = priv_check(td,
2043 						    PRIV_NETINET_SETHDROPTS);
2044 						if (error)
2045 							return (error);
2046 					}
2047 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2048 					break;
2049 				case IPV6_2292RTHDR:
2050 					OPTSET2292(IN6P_RTHDR);
2051 					break;
2052 				}
2053 				break;
2054 			case IPV6_PKTINFO:
2055 			case IPV6_HOPOPTS:
2056 			case IPV6_RTHDR:
2057 			case IPV6_DSTOPTS:
2058 			case IPV6_RTHDRDSTOPTS:
2059 			case IPV6_NEXTHOP:
2060 			{
2061 				/* new advanced API (RFC3542) */
2062 				u_char *optbuf;
2063 				u_char optbuf_storage[MCLBYTES];
2064 				int optlen;
2065 				struct ip6_pktopts **optp;
2066 
2067 				/* cannot mix with RFC2292 */
2068 				if (OPTBIT(IN6P_RFC2292)) {
2069 					error = EINVAL;
2070 					break;
2071 				}
2072 
2073 				/*
2074 				 * We only ensure valsize is not too large
2075 				 * here.  Further validation will be done
2076 				 * later.
2077 				 */
2078 				error = sooptcopyin(sopt, optbuf_storage,
2079 				    sizeof(optbuf_storage), 0);
2080 				if (error)
2081 					break;
2082 				optlen = sopt->sopt_valsize;
2083 				optbuf = optbuf_storage;
2084 				INP_WLOCK(inp);
2085 				if (inp->inp_flags & INP_DROPPED) {
2086 					INP_WUNLOCK(inp);
2087 					return (ECONNRESET);
2088 				}
2089 				optp = &inp->in6p_outputopts;
2090 				error = ip6_pcbopt(optname, optbuf, optlen,
2091 				    optp, (td != NULL) ? td->td_ucred : NULL,
2092 				    uproto);
2093 				INP_WUNLOCK(inp);
2094 				break;
2095 			}
2096 #undef OPTSET
2097 
2098 			case IPV6_MULTICAST_IF:
2099 			case IPV6_MULTICAST_HOPS:
2100 			case IPV6_MULTICAST_LOOP:
2101 			case IPV6_JOIN_GROUP:
2102 			case IPV6_LEAVE_GROUP:
2103 			case IPV6_MSFILTER:
2104 			case MCAST_BLOCK_SOURCE:
2105 			case MCAST_UNBLOCK_SOURCE:
2106 			case MCAST_JOIN_GROUP:
2107 			case MCAST_LEAVE_GROUP:
2108 			case MCAST_JOIN_SOURCE_GROUP:
2109 			case MCAST_LEAVE_SOURCE_GROUP:
2110 				error = ip6_setmoptions(inp, sopt);
2111 				break;
2112 
2113 			case IPV6_PORTRANGE:
2114 				error = sooptcopyin(sopt, &optval,
2115 				    sizeof optval, sizeof optval);
2116 				if (error)
2117 					break;
2118 
2119 				INP_WLOCK(inp);
2120 				switch (optval) {
2121 				case IPV6_PORTRANGE_DEFAULT:
2122 					inp->inp_flags &= ~(INP_LOWPORT);
2123 					inp->inp_flags &= ~(INP_HIGHPORT);
2124 					break;
2125 
2126 				case IPV6_PORTRANGE_HIGH:
2127 					inp->inp_flags &= ~(INP_LOWPORT);
2128 					inp->inp_flags |= INP_HIGHPORT;
2129 					break;
2130 
2131 				case IPV6_PORTRANGE_LOW:
2132 					inp->inp_flags &= ~(INP_HIGHPORT);
2133 					inp->inp_flags |= INP_LOWPORT;
2134 					break;
2135 
2136 				default:
2137 					error = EINVAL;
2138 					break;
2139 				}
2140 				INP_WUNLOCK(inp);
2141 				break;
2142 
2143 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2144 			case IPV6_IPSEC_POLICY:
2145 				if (IPSEC_ENABLED(ipv6)) {
2146 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2147 					break;
2148 				}
2149 				/* FALLTHROUGH */
2150 #endif /* IPSEC */
2151 
2152 			default:
2153 				error = ENOPROTOOPT;
2154 				break;
2155 			}
2156 			break;
2157 
2158 		case SOPT_GET:
2159 			switch (optname) {
2160 			case IPV6_2292PKTOPTIONS:
2161 #ifdef IPV6_PKTOPTIONS
2162 			case IPV6_PKTOPTIONS:
2163 #endif
2164 				/*
2165 				 * RFC3542 (effectively) deprecated the
2166 				 * semantics of the 2292-style pktoptions.
2167 				 * Since it was not reliable in nature (i.e.,
2168 				 * applications had to expect the lack of some
2169 				 * information after all), it would make sense
2170 				 * to simplify this part by always returning
2171 				 * empty data.
2172 				 */
2173 				sopt->sopt_valsize = 0;
2174 				break;
2175 
2176 			case IPV6_RECVHOPOPTS:
2177 			case IPV6_RECVDSTOPTS:
2178 			case IPV6_RECVRTHDRDSTOPTS:
2179 			case IPV6_UNICAST_HOPS:
2180 			case IPV6_RECVPKTINFO:
2181 			case IPV6_RECVHOPLIMIT:
2182 			case IPV6_RECVRTHDR:
2183 			case IPV6_RECVPATHMTU:
2184 
2185 			case IPV6_V6ONLY:
2186 			case IPV6_PORTRANGE:
2187 			case IPV6_RECVTCLASS:
2188 			case IPV6_AUTOFLOWLABEL:
2189 			case IPV6_BINDANY:
2190 			case IPV6_FLOWID:
2191 			case IPV6_FLOWTYPE:
2192 			case IPV6_RECVFLOWID:
2193 #ifdef	RSS
2194 			case IPV6_RSSBUCKETID:
2195 			case IPV6_RECVRSSBUCKETID:
2196 #endif
2197 			case IPV6_BINDMULTI:
2198 			case IPV6_VLAN_PCP:
2199 				switch (optname) {
2200 				case IPV6_RECVHOPOPTS:
2201 					optval = OPTBIT(IN6P_HOPOPTS);
2202 					break;
2203 
2204 				case IPV6_RECVDSTOPTS:
2205 					optval = OPTBIT(IN6P_DSTOPTS);
2206 					break;
2207 
2208 				case IPV6_RECVRTHDRDSTOPTS:
2209 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2210 					break;
2211 
2212 				case IPV6_UNICAST_HOPS:
2213 					optval = inp->in6p_hops;
2214 					break;
2215 
2216 				case IPV6_RECVPKTINFO:
2217 					optval = OPTBIT(IN6P_PKTINFO);
2218 					break;
2219 
2220 				case IPV6_RECVHOPLIMIT:
2221 					optval = OPTBIT(IN6P_HOPLIMIT);
2222 					break;
2223 
2224 				case IPV6_RECVRTHDR:
2225 					optval = OPTBIT(IN6P_RTHDR);
2226 					break;
2227 
2228 				case IPV6_RECVPATHMTU:
2229 					optval = OPTBIT(IN6P_MTU);
2230 					break;
2231 
2232 				case IPV6_V6ONLY:
2233 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2234 					break;
2235 
2236 				case IPV6_PORTRANGE:
2237 				    {
2238 					int flags;
2239 					flags = inp->inp_flags;
2240 					if (flags & INP_HIGHPORT)
2241 						optval = IPV6_PORTRANGE_HIGH;
2242 					else if (flags & INP_LOWPORT)
2243 						optval = IPV6_PORTRANGE_LOW;
2244 					else
2245 						optval = 0;
2246 					break;
2247 				    }
2248 				case IPV6_RECVTCLASS:
2249 					optval = OPTBIT(IN6P_TCLASS);
2250 					break;
2251 
2252 				case IPV6_AUTOFLOWLABEL:
2253 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2254 					break;
2255 
2256 				case IPV6_ORIGDSTADDR:
2257 					optval = OPTBIT2(INP_ORIGDSTADDR);
2258 					break;
2259 
2260 				case IPV6_BINDANY:
2261 					optval = OPTBIT(INP_BINDANY);
2262 					break;
2263 
2264 				case IPV6_FLOWID:
2265 					optval = inp->inp_flowid;
2266 					break;
2267 
2268 				case IPV6_FLOWTYPE:
2269 					optval = inp->inp_flowtype;
2270 					break;
2271 
2272 				case IPV6_RECVFLOWID:
2273 					optval = OPTBIT2(INP_RECVFLOWID);
2274 					break;
2275 #ifdef	RSS
2276 				case IPV6_RSSBUCKETID:
2277 					retval =
2278 					    rss_hash2bucket(inp->inp_flowid,
2279 					    inp->inp_flowtype,
2280 					    &rss_bucket);
2281 					if (retval == 0)
2282 						optval = rss_bucket;
2283 					else
2284 						error = EINVAL;
2285 					break;
2286 
2287 				case IPV6_RECVRSSBUCKETID:
2288 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2289 					break;
2290 #endif
2291 
2292 				case IPV6_BINDMULTI:
2293 					optval = OPTBIT2(INP_BINDMULTI);
2294 					break;
2295 
2296 				case IPV6_VLAN_PCP:
2297 					if (OPTBIT2(INP_2PCP_SET)) {
2298 						optval = (inp->inp_flags2 &
2299 							    INP_2PCP_MASK) >>
2300 							    INP_2PCP_SHIFT;
2301 					} else {
2302 						optval = -1;
2303 					}
2304 					break;
2305 				}
2306 
2307 				if (error)
2308 					break;
2309 				error = sooptcopyout(sopt, &optval,
2310 					sizeof optval);
2311 				break;
2312 
2313 			case IPV6_PATHMTU:
2314 			{
2315 				u_long pmtu = 0;
2316 				struct ip6_mtuinfo mtuinfo;
2317 				struct in6_addr addr;
2318 
2319 				if (!(so->so_state & SS_ISCONNECTED))
2320 					return (ENOTCONN);
2321 				/*
2322 				 * XXX: we dot not consider the case of source
2323 				 * routing, or optional information to specify
2324 				 * the outgoing interface.
2325 				 * Copy faddr out of inp to avoid holding lock
2326 				 * on inp during route lookup.
2327 				 */
2328 				INP_RLOCK(inp);
2329 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2330 				INP_RUNLOCK(inp);
2331 				error = ip6_getpmtu_ctl(so->so_fibnum,
2332 				    &addr, &pmtu);
2333 				if (error)
2334 					break;
2335 				if (pmtu > IPV6_MAXPACKET)
2336 					pmtu = IPV6_MAXPACKET;
2337 
2338 				bzero(&mtuinfo, sizeof(mtuinfo));
2339 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2340 				optdata = (void *)&mtuinfo;
2341 				optdatalen = sizeof(mtuinfo);
2342 				error = sooptcopyout(sopt, optdata,
2343 				    optdatalen);
2344 				break;
2345 			}
2346 
2347 			case IPV6_2292PKTINFO:
2348 			case IPV6_2292HOPLIMIT:
2349 			case IPV6_2292HOPOPTS:
2350 			case IPV6_2292RTHDR:
2351 			case IPV6_2292DSTOPTS:
2352 				switch (optname) {
2353 				case IPV6_2292PKTINFO:
2354 					optval = OPTBIT(IN6P_PKTINFO);
2355 					break;
2356 				case IPV6_2292HOPLIMIT:
2357 					optval = OPTBIT(IN6P_HOPLIMIT);
2358 					break;
2359 				case IPV6_2292HOPOPTS:
2360 					optval = OPTBIT(IN6P_HOPOPTS);
2361 					break;
2362 				case IPV6_2292RTHDR:
2363 					optval = OPTBIT(IN6P_RTHDR);
2364 					break;
2365 				case IPV6_2292DSTOPTS:
2366 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2367 					break;
2368 				}
2369 				error = sooptcopyout(sopt, &optval,
2370 				    sizeof optval);
2371 				break;
2372 			case IPV6_PKTINFO:
2373 			case IPV6_HOPOPTS:
2374 			case IPV6_RTHDR:
2375 			case IPV6_DSTOPTS:
2376 			case IPV6_RTHDRDSTOPTS:
2377 			case IPV6_NEXTHOP:
2378 			case IPV6_TCLASS:
2379 			case IPV6_DONTFRAG:
2380 			case IPV6_USE_MIN_MTU:
2381 			case IPV6_PREFER_TEMPADDR:
2382 				error = ip6_getpcbopt(inp, optname, sopt);
2383 				break;
2384 
2385 			case IPV6_MULTICAST_IF:
2386 			case IPV6_MULTICAST_HOPS:
2387 			case IPV6_MULTICAST_LOOP:
2388 			case IPV6_MSFILTER:
2389 				error = ip6_getmoptions(inp, sopt);
2390 				break;
2391 
2392 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2393 			case IPV6_IPSEC_POLICY:
2394 				if (IPSEC_ENABLED(ipv6)) {
2395 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2396 					break;
2397 				}
2398 				/* FALLTHROUGH */
2399 #endif /* IPSEC */
2400 			default:
2401 				error = ENOPROTOOPT;
2402 				break;
2403 			}
2404 			break;
2405 		}
2406 	}
2407 	return (error);
2408 }
2409 
2410 int
2411 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2412 {
2413 	int error = 0, optval, optlen;
2414 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2415 	struct inpcb *inp = sotoinpcb(so);
2416 	int level, op, optname;
2417 
2418 	level = sopt->sopt_level;
2419 	op = sopt->sopt_dir;
2420 	optname = sopt->sopt_name;
2421 	optlen = sopt->sopt_valsize;
2422 
2423 	if (level != IPPROTO_IPV6) {
2424 		return (EINVAL);
2425 	}
2426 
2427 	switch (optname) {
2428 	case IPV6_CHECKSUM:
2429 		/*
2430 		 * For ICMPv6 sockets, no modification allowed for checksum
2431 		 * offset, permit "no change" values to help existing apps.
2432 		 *
2433 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2434 		 * for an ICMPv6 socket will fail."
2435 		 * The current behavior does not meet RFC3542.
2436 		 */
2437 		switch (op) {
2438 		case SOPT_SET:
2439 			if (optlen != sizeof(int)) {
2440 				error = EINVAL;
2441 				break;
2442 			}
2443 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2444 					    sizeof(optval));
2445 			if (error)
2446 				break;
2447 			if (optval < -1 || (optval % 2) != 0) {
2448 				/*
2449 				 * The API assumes non-negative even offset
2450 				 * values or -1 as a special value.
2451 				 */
2452 				error = EINVAL;
2453 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2454 				if (optval != icmp6off)
2455 					error = EINVAL;
2456 			} else
2457 				inp->in6p_cksum = optval;
2458 			break;
2459 
2460 		case SOPT_GET:
2461 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2462 				optval = icmp6off;
2463 			else
2464 				optval = inp->in6p_cksum;
2465 
2466 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2467 			break;
2468 
2469 		default:
2470 			error = EINVAL;
2471 			break;
2472 		}
2473 		break;
2474 
2475 	default:
2476 		error = ENOPROTOOPT;
2477 		break;
2478 	}
2479 
2480 	return (error);
2481 }
2482 
2483 /*
2484  * Set up IP6 options in pcb for insertion in output packets or
2485  * specifying behavior of outgoing packets.
2486  */
2487 static int
2488 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2489     struct socket *so, struct sockopt *sopt)
2490 {
2491 	struct ip6_pktopts *opt = *pktopt;
2492 	int error = 0;
2493 	struct thread *td = sopt->sopt_td;
2494 	struct epoch_tracker et;
2495 
2496 	/* turn off any old options. */
2497 	if (opt) {
2498 #ifdef DIAGNOSTIC
2499 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2500 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2501 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2502 			printf("ip6_pcbopts: all specified options are cleared.\n");
2503 #endif
2504 		ip6_clearpktopts(opt, -1);
2505 	} else {
2506 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2507 		if (opt == NULL)
2508 			return (ENOMEM);
2509 	}
2510 	*pktopt = NULL;
2511 
2512 	if (!m || m->m_len == 0) {
2513 		/*
2514 		 * Only turning off any previous options, regardless of
2515 		 * whether the opt is just created or given.
2516 		 */
2517 		free(opt, M_IP6OPT);
2518 		return (0);
2519 	}
2520 
2521 	/*  set options specified by user. */
2522 	NET_EPOCH_ENTER(et);
2523 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2524 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2525 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2526 		free(opt, M_IP6OPT);
2527 		NET_EPOCH_EXIT(et);
2528 		return (error);
2529 	}
2530 	NET_EPOCH_EXIT(et);
2531 	*pktopt = opt;
2532 	return (0);
2533 }
2534 
2535 /*
2536  * initialize ip6_pktopts.  beware that there are non-zero default values in
2537  * the struct.
2538  */
2539 void
2540 ip6_initpktopts(struct ip6_pktopts *opt)
2541 {
2542 
2543 	bzero(opt, sizeof(*opt));
2544 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2545 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2546 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2547 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2548 }
2549 
2550 static int
2551 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2552     struct ucred *cred, int uproto)
2553 {
2554 	struct epoch_tracker et;
2555 	struct ip6_pktopts *opt;
2556 	int ret;
2557 
2558 	if (*pktopt == NULL) {
2559 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2560 		    M_NOWAIT);
2561 		if (*pktopt == NULL)
2562 			return (ENOBUFS);
2563 		ip6_initpktopts(*pktopt);
2564 	}
2565 	opt = *pktopt;
2566 
2567 	NET_EPOCH_ENTER(et);
2568 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2569 	NET_EPOCH_EXIT(et);
2570 
2571 	return (ret);
2572 }
2573 
2574 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2575 	if (pktopt && pktopt->field) {					\
2576 		INP_RUNLOCK(inp);					\
2577 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2578 		malloc_optdata = true;					\
2579 		INP_RLOCK(inp);						\
2580 		if (inp->inp_flags & INP_DROPPED) {			\
2581 			INP_RUNLOCK(inp);				\
2582 			free(optdata, M_TEMP);				\
2583 			return (ECONNRESET);				\
2584 		}							\
2585 		pktopt = inp->in6p_outputopts;				\
2586 		if (pktopt && pktopt->field) {				\
2587 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2588 			bcopy(pktopt->field, optdata, optdatalen);	\
2589 		} else {						\
2590 			free(optdata, M_TEMP);				\
2591 			optdata = NULL;					\
2592 			malloc_optdata = false;				\
2593 		}							\
2594 	}								\
2595 } while(0)
2596 
2597 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2598 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2599 
2600 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2601 	pktopt->field->sa_len)
2602 
2603 static int
2604 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2605 {
2606 	void *optdata = NULL;
2607 	bool malloc_optdata = false;
2608 	int optdatalen = 0;
2609 	int error = 0;
2610 	struct in6_pktinfo null_pktinfo;
2611 	int deftclass = 0, on;
2612 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2613 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2614 	struct ip6_pktopts *pktopt;
2615 
2616 	INP_RLOCK(inp);
2617 	pktopt = inp->in6p_outputopts;
2618 
2619 	switch (optname) {
2620 	case IPV6_PKTINFO:
2621 		optdata = (void *)&null_pktinfo;
2622 		if (pktopt && pktopt->ip6po_pktinfo) {
2623 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2624 			    sizeof(null_pktinfo));
2625 			in6_clearscope(&null_pktinfo.ipi6_addr);
2626 		} else {
2627 			/* XXX: we don't have to do this every time... */
2628 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2629 		}
2630 		optdatalen = sizeof(struct in6_pktinfo);
2631 		break;
2632 	case IPV6_TCLASS:
2633 		if (pktopt && pktopt->ip6po_tclass >= 0)
2634 			deftclass = pktopt->ip6po_tclass;
2635 		optdata = (void *)&deftclass;
2636 		optdatalen = sizeof(int);
2637 		break;
2638 	case IPV6_HOPOPTS:
2639 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2640 		break;
2641 	case IPV6_RTHDR:
2642 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2643 		break;
2644 	case IPV6_RTHDRDSTOPTS:
2645 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2646 		break;
2647 	case IPV6_DSTOPTS:
2648 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2649 		break;
2650 	case IPV6_NEXTHOP:
2651 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2652 		break;
2653 	case IPV6_USE_MIN_MTU:
2654 		if (pktopt)
2655 			defminmtu = pktopt->ip6po_minmtu;
2656 		optdata = (void *)&defminmtu;
2657 		optdatalen = sizeof(int);
2658 		break;
2659 	case IPV6_DONTFRAG:
2660 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2661 			on = 1;
2662 		else
2663 			on = 0;
2664 		optdata = (void *)&on;
2665 		optdatalen = sizeof(on);
2666 		break;
2667 	case IPV6_PREFER_TEMPADDR:
2668 		if (pktopt)
2669 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2670 		optdata = (void *)&defpreftemp;
2671 		optdatalen = sizeof(int);
2672 		break;
2673 	default:		/* should not happen */
2674 #ifdef DIAGNOSTIC
2675 		panic("ip6_getpcbopt: unexpected option\n");
2676 #endif
2677 		INP_RUNLOCK(inp);
2678 		return (ENOPROTOOPT);
2679 	}
2680 	INP_RUNLOCK(inp);
2681 
2682 	error = sooptcopyout(sopt, optdata, optdatalen);
2683 	if (malloc_optdata)
2684 		free(optdata, M_TEMP);
2685 
2686 	return (error);
2687 }
2688 
2689 void
2690 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2691 {
2692 	if (pktopt == NULL)
2693 		return;
2694 
2695 	if (optname == -1 || optname == IPV6_PKTINFO) {
2696 		if (pktopt->ip6po_pktinfo)
2697 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2698 		pktopt->ip6po_pktinfo = NULL;
2699 	}
2700 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2701 		pktopt->ip6po_hlim = -1;
2702 	if (optname == -1 || optname == IPV6_TCLASS)
2703 		pktopt->ip6po_tclass = -1;
2704 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2705 		if (pktopt->ip6po_nextroute.ro_nh) {
2706 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2707 			pktopt->ip6po_nextroute.ro_nh = NULL;
2708 		}
2709 		if (pktopt->ip6po_nexthop)
2710 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2711 		pktopt->ip6po_nexthop = NULL;
2712 	}
2713 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2714 		if (pktopt->ip6po_hbh)
2715 			free(pktopt->ip6po_hbh, M_IP6OPT);
2716 		pktopt->ip6po_hbh = NULL;
2717 	}
2718 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2719 		if (pktopt->ip6po_dest1)
2720 			free(pktopt->ip6po_dest1, M_IP6OPT);
2721 		pktopt->ip6po_dest1 = NULL;
2722 	}
2723 	if (optname == -1 || optname == IPV6_RTHDR) {
2724 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2725 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2726 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2727 		if (pktopt->ip6po_route.ro_nh) {
2728 			NH_FREE(pktopt->ip6po_route.ro_nh);
2729 			pktopt->ip6po_route.ro_nh = NULL;
2730 		}
2731 	}
2732 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2733 		if (pktopt->ip6po_dest2)
2734 			free(pktopt->ip6po_dest2, M_IP6OPT);
2735 		pktopt->ip6po_dest2 = NULL;
2736 	}
2737 }
2738 
2739 #define PKTOPT_EXTHDRCPY(type) \
2740 do {\
2741 	if (src->type) {\
2742 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2743 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2744 		if (dst->type == NULL)\
2745 			goto bad;\
2746 		bcopy(src->type, dst->type, hlen);\
2747 	}\
2748 } while (/*CONSTCOND*/ 0)
2749 
2750 static int
2751 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2752 {
2753 	if (dst == NULL || src == NULL)  {
2754 		printf("ip6_clearpktopts: invalid argument\n");
2755 		return (EINVAL);
2756 	}
2757 
2758 	dst->ip6po_hlim = src->ip6po_hlim;
2759 	dst->ip6po_tclass = src->ip6po_tclass;
2760 	dst->ip6po_flags = src->ip6po_flags;
2761 	dst->ip6po_minmtu = src->ip6po_minmtu;
2762 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2763 	if (src->ip6po_pktinfo) {
2764 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2765 		    M_IP6OPT, canwait);
2766 		if (dst->ip6po_pktinfo == NULL)
2767 			goto bad;
2768 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2769 	}
2770 	if (src->ip6po_nexthop) {
2771 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2772 		    M_IP6OPT, canwait);
2773 		if (dst->ip6po_nexthop == NULL)
2774 			goto bad;
2775 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2776 		    src->ip6po_nexthop->sa_len);
2777 	}
2778 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2779 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2780 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2781 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2782 	return (0);
2783 
2784   bad:
2785 	ip6_clearpktopts(dst, -1);
2786 	return (ENOBUFS);
2787 }
2788 #undef PKTOPT_EXTHDRCPY
2789 
2790 struct ip6_pktopts *
2791 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2792 {
2793 	int error;
2794 	struct ip6_pktopts *dst;
2795 
2796 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2797 	if (dst == NULL)
2798 		return (NULL);
2799 	ip6_initpktopts(dst);
2800 
2801 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2802 		free(dst, M_IP6OPT);
2803 		return (NULL);
2804 	}
2805 
2806 	return (dst);
2807 }
2808 
2809 void
2810 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2811 {
2812 	if (pktopt == NULL)
2813 		return;
2814 
2815 	ip6_clearpktopts(pktopt, -1);
2816 
2817 	free(pktopt, M_IP6OPT);
2818 }
2819 
2820 /*
2821  * Set IPv6 outgoing packet options based on advanced API.
2822  */
2823 int
2824 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2825     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2826 {
2827 	struct cmsghdr *cm = NULL;
2828 
2829 	if (control == NULL || opt == NULL)
2830 		return (EINVAL);
2831 
2832 	/*
2833 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2834 	 * are in the network epoch here.
2835 	 */
2836 	NET_EPOCH_ASSERT();
2837 
2838 	ip6_initpktopts(opt);
2839 	if (stickyopt) {
2840 		int error;
2841 
2842 		/*
2843 		 * If stickyopt is provided, make a local copy of the options
2844 		 * for this particular packet, then override them by ancillary
2845 		 * objects.
2846 		 * XXX: copypktopts() does not copy the cached route to a next
2847 		 * hop (if any).  This is not very good in terms of efficiency,
2848 		 * but we can allow this since this option should be rarely
2849 		 * used.
2850 		 */
2851 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2852 			return (error);
2853 	}
2854 
2855 	/*
2856 	 * XXX: Currently, we assume all the optional information is stored
2857 	 * in a single mbuf.
2858 	 */
2859 	if (control->m_next)
2860 		return (EINVAL);
2861 
2862 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2863 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2864 		int error;
2865 
2866 		if (control->m_len < CMSG_LEN(0))
2867 			return (EINVAL);
2868 
2869 		cm = mtod(control, struct cmsghdr *);
2870 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2871 			return (EINVAL);
2872 		if (cm->cmsg_level != IPPROTO_IPV6)
2873 			continue;
2874 
2875 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2876 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2877 		if (error)
2878 			return (error);
2879 	}
2880 
2881 	return (0);
2882 }
2883 
2884 /*
2885  * Set a particular packet option, as a sticky option or an ancillary data
2886  * item.  "len" can be 0 only when it's a sticky option.
2887  * We have 4 cases of combination of "sticky" and "cmsg":
2888  * "sticky=0, cmsg=0": impossible
2889  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2890  * "sticky=1, cmsg=0": RFC3542 socket option
2891  * "sticky=1, cmsg=1": RFC2292 socket option
2892  */
2893 static int
2894 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2895     struct ucred *cred, int sticky, int cmsg, int uproto)
2896 {
2897 	int minmtupolicy, preftemp;
2898 	int error;
2899 
2900 	NET_EPOCH_ASSERT();
2901 
2902 	if (!sticky && !cmsg) {
2903 #ifdef DIAGNOSTIC
2904 		printf("ip6_setpktopt: impossible case\n");
2905 #endif
2906 		return (EINVAL);
2907 	}
2908 
2909 	/*
2910 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2911 	 * not be specified in the context of RFC3542.  Conversely,
2912 	 * RFC3542 types should not be specified in the context of RFC2292.
2913 	 */
2914 	if (!cmsg) {
2915 		switch (optname) {
2916 		case IPV6_2292PKTINFO:
2917 		case IPV6_2292HOPLIMIT:
2918 		case IPV6_2292NEXTHOP:
2919 		case IPV6_2292HOPOPTS:
2920 		case IPV6_2292DSTOPTS:
2921 		case IPV6_2292RTHDR:
2922 		case IPV6_2292PKTOPTIONS:
2923 			return (ENOPROTOOPT);
2924 		}
2925 	}
2926 	if (sticky && cmsg) {
2927 		switch (optname) {
2928 		case IPV6_PKTINFO:
2929 		case IPV6_HOPLIMIT:
2930 		case IPV6_NEXTHOP:
2931 		case IPV6_HOPOPTS:
2932 		case IPV6_DSTOPTS:
2933 		case IPV6_RTHDRDSTOPTS:
2934 		case IPV6_RTHDR:
2935 		case IPV6_USE_MIN_MTU:
2936 		case IPV6_DONTFRAG:
2937 		case IPV6_TCLASS:
2938 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2939 			return (ENOPROTOOPT);
2940 		}
2941 	}
2942 
2943 	switch (optname) {
2944 	case IPV6_2292PKTINFO:
2945 	case IPV6_PKTINFO:
2946 	{
2947 		struct ifnet *ifp = NULL;
2948 		struct in6_pktinfo *pktinfo;
2949 
2950 		if (len != sizeof(struct in6_pktinfo))
2951 			return (EINVAL);
2952 
2953 		pktinfo = (struct in6_pktinfo *)buf;
2954 
2955 		/*
2956 		 * An application can clear any sticky IPV6_PKTINFO option by
2957 		 * doing a "regular" setsockopt with ipi6_addr being
2958 		 * in6addr_any and ipi6_ifindex being zero.
2959 		 * [RFC 3542, Section 6]
2960 		 */
2961 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2962 		    pktinfo->ipi6_ifindex == 0 &&
2963 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2964 			ip6_clearpktopts(opt, optname);
2965 			break;
2966 		}
2967 
2968 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2969 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2970 			return (EINVAL);
2971 		}
2972 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2973 			return (EINVAL);
2974 		/* validate the interface index if specified. */
2975 		if (pktinfo->ipi6_ifindex) {
2976 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2977 			if (ifp == NULL)
2978 				return (ENXIO);
2979 		}
2980 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2981 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2982 			return (ENETDOWN);
2983 
2984 		if (ifp != NULL &&
2985 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2986 			struct in6_ifaddr *ia;
2987 
2988 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2989 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2990 			if (ia == NULL)
2991 				return (EADDRNOTAVAIL);
2992 			ifa_free(&ia->ia_ifa);
2993 		}
2994 		/*
2995 		 * We store the address anyway, and let in6_selectsrc()
2996 		 * validate the specified address.  This is because ipi6_addr
2997 		 * may not have enough information about its scope zone, and
2998 		 * we may need additional information (such as outgoing
2999 		 * interface or the scope zone of a destination address) to
3000 		 * disambiguate the scope.
3001 		 * XXX: the delay of the validation may confuse the
3002 		 * application when it is used as a sticky option.
3003 		 */
3004 		if (opt->ip6po_pktinfo == NULL) {
3005 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3006 			    M_IP6OPT, M_NOWAIT);
3007 			if (opt->ip6po_pktinfo == NULL)
3008 				return (ENOBUFS);
3009 		}
3010 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3011 		break;
3012 	}
3013 
3014 	case IPV6_2292HOPLIMIT:
3015 	case IPV6_HOPLIMIT:
3016 	{
3017 		int *hlimp;
3018 
3019 		/*
3020 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3021 		 * to simplify the ordering among hoplimit options.
3022 		 */
3023 		if (optname == IPV6_HOPLIMIT && sticky)
3024 			return (ENOPROTOOPT);
3025 
3026 		if (len != sizeof(int))
3027 			return (EINVAL);
3028 		hlimp = (int *)buf;
3029 		if (*hlimp < -1 || *hlimp > 255)
3030 			return (EINVAL);
3031 
3032 		opt->ip6po_hlim = *hlimp;
3033 		break;
3034 	}
3035 
3036 	case IPV6_TCLASS:
3037 	{
3038 		int tclass;
3039 
3040 		if (len != sizeof(int))
3041 			return (EINVAL);
3042 		tclass = *(int *)buf;
3043 		if (tclass < -1 || tclass > 255)
3044 			return (EINVAL);
3045 
3046 		opt->ip6po_tclass = tclass;
3047 		break;
3048 	}
3049 
3050 	case IPV6_2292NEXTHOP:
3051 	case IPV6_NEXTHOP:
3052 		if (cred != NULL) {
3053 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3054 			if (error)
3055 				return (error);
3056 		}
3057 
3058 		if (len == 0) {	/* just remove the option */
3059 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3060 			break;
3061 		}
3062 
3063 		/* check if cmsg_len is large enough for sa_len */
3064 		if (len < sizeof(struct sockaddr) || len < *buf)
3065 			return (EINVAL);
3066 
3067 		switch (((struct sockaddr *)buf)->sa_family) {
3068 		case AF_INET6:
3069 		{
3070 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3071 			int error;
3072 
3073 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3074 				return (EINVAL);
3075 
3076 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3077 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3078 				return (EINVAL);
3079 			}
3080 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3081 			    != 0) {
3082 				return (error);
3083 			}
3084 			break;
3085 		}
3086 		case AF_LINK:	/* should eventually be supported */
3087 		default:
3088 			return (EAFNOSUPPORT);
3089 		}
3090 
3091 		/* turn off the previous option, then set the new option. */
3092 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3093 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3094 		if (opt->ip6po_nexthop == NULL)
3095 			return (ENOBUFS);
3096 		bcopy(buf, opt->ip6po_nexthop, *buf);
3097 		break;
3098 
3099 	case IPV6_2292HOPOPTS:
3100 	case IPV6_HOPOPTS:
3101 	{
3102 		struct ip6_hbh *hbh;
3103 		int hbhlen;
3104 
3105 		/*
3106 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3107 		 * options, since per-option restriction has too much
3108 		 * overhead.
3109 		 */
3110 		if (cred != NULL) {
3111 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3112 			if (error)
3113 				return (error);
3114 		}
3115 
3116 		if (len == 0) {
3117 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3118 			break;	/* just remove the option */
3119 		}
3120 
3121 		/* message length validation */
3122 		if (len < sizeof(struct ip6_hbh))
3123 			return (EINVAL);
3124 		hbh = (struct ip6_hbh *)buf;
3125 		hbhlen = (hbh->ip6h_len + 1) << 3;
3126 		if (len != hbhlen)
3127 			return (EINVAL);
3128 
3129 		/* turn off the previous option, then set the new option. */
3130 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3131 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3132 		if (opt->ip6po_hbh == NULL)
3133 			return (ENOBUFS);
3134 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3135 
3136 		break;
3137 	}
3138 
3139 	case IPV6_2292DSTOPTS:
3140 	case IPV6_DSTOPTS:
3141 	case IPV6_RTHDRDSTOPTS:
3142 	{
3143 		struct ip6_dest *dest, **newdest = NULL;
3144 		int destlen;
3145 
3146 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3147 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3148 			if (error)
3149 				return (error);
3150 		}
3151 
3152 		if (len == 0) {
3153 			ip6_clearpktopts(opt, optname);
3154 			break;	/* just remove the option */
3155 		}
3156 
3157 		/* message length validation */
3158 		if (len < sizeof(struct ip6_dest))
3159 			return (EINVAL);
3160 		dest = (struct ip6_dest *)buf;
3161 		destlen = (dest->ip6d_len + 1) << 3;
3162 		if (len != destlen)
3163 			return (EINVAL);
3164 
3165 		/*
3166 		 * Determine the position that the destination options header
3167 		 * should be inserted; before or after the routing header.
3168 		 */
3169 		switch (optname) {
3170 		case IPV6_2292DSTOPTS:
3171 			/*
3172 			 * The old advacned API is ambiguous on this point.
3173 			 * Our approach is to determine the position based
3174 			 * according to the existence of a routing header.
3175 			 * Note, however, that this depends on the order of the
3176 			 * extension headers in the ancillary data; the 1st
3177 			 * part of the destination options header must appear
3178 			 * before the routing header in the ancillary data,
3179 			 * too.
3180 			 * RFC3542 solved the ambiguity by introducing
3181 			 * separate ancillary data or option types.
3182 			 */
3183 			if (opt->ip6po_rthdr == NULL)
3184 				newdest = &opt->ip6po_dest1;
3185 			else
3186 				newdest = &opt->ip6po_dest2;
3187 			break;
3188 		case IPV6_RTHDRDSTOPTS:
3189 			newdest = &opt->ip6po_dest1;
3190 			break;
3191 		case IPV6_DSTOPTS:
3192 			newdest = &opt->ip6po_dest2;
3193 			break;
3194 		}
3195 
3196 		/* turn off the previous option, then set the new option. */
3197 		ip6_clearpktopts(opt, optname);
3198 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3199 		if (*newdest == NULL)
3200 			return (ENOBUFS);
3201 		bcopy(dest, *newdest, destlen);
3202 
3203 		break;
3204 	}
3205 
3206 	case IPV6_2292RTHDR:
3207 	case IPV6_RTHDR:
3208 	{
3209 		struct ip6_rthdr *rth;
3210 		int rthlen;
3211 
3212 		if (len == 0) {
3213 			ip6_clearpktopts(opt, IPV6_RTHDR);
3214 			break;	/* just remove the option */
3215 		}
3216 
3217 		/* message length validation */
3218 		if (len < sizeof(struct ip6_rthdr))
3219 			return (EINVAL);
3220 		rth = (struct ip6_rthdr *)buf;
3221 		rthlen = (rth->ip6r_len + 1) << 3;
3222 		if (len != rthlen)
3223 			return (EINVAL);
3224 
3225 		switch (rth->ip6r_type) {
3226 		case IPV6_RTHDR_TYPE_0:
3227 			if (rth->ip6r_len == 0)	/* must contain one addr */
3228 				return (EINVAL);
3229 			if (rth->ip6r_len % 2) /* length must be even */
3230 				return (EINVAL);
3231 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3232 				return (EINVAL);
3233 			break;
3234 		default:
3235 			return (EINVAL);	/* not supported */
3236 		}
3237 
3238 		/* turn off the previous option */
3239 		ip6_clearpktopts(opt, IPV6_RTHDR);
3240 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3241 		if (opt->ip6po_rthdr == NULL)
3242 			return (ENOBUFS);
3243 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3244 
3245 		break;
3246 	}
3247 
3248 	case IPV6_USE_MIN_MTU:
3249 		if (len != sizeof(int))
3250 			return (EINVAL);
3251 		minmtupolicy = *(int *)buf;
3252 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3253 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3254 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3255 			return (EINVAL);
3256 		}
3257 		opt->ip6po_minmtu = minmtupolicy;
3258 		break;
3259 
3260 	case IPV6_DONTFRAG:
3261 		if (len != sizeof(int))
3262 			return (EINVAL);
3263 
3264 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3265 			/*
3266 			 * we ignore this option for TCP sockets.
3267 			 * (RFC3542 leaves this case unspecified.)
3268 			 */
3269 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3270 		} else
3271 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3272 		break;
3273 
3274 	case IPV6_PREFER_TEMPADDR:
3275 		if (len != sizeof(int))
3276 			return (EINVAL);
3277 		preftemp = *(int *)buf;
3278 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3279 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3280 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3281 			return (EINVAL);
3282 		}
3283 		opt->ip6po_prefer_tempaddr = preftemp;
3284 		break;
3285 
3286 	default:
3287 		return (ENOPROTOOPT);
3288 	} /* end of switch */
3289 
3290 	return (0);
3291 }
3292 
3293 /*
3294  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3295  * packet to the input queue of a specified interface.  Note that this
3296  * calls the output routine of the loopback "driver", but with an interface
3297  * pointer that might NOT be &loif -- easier than replicating that code here.
3298  */
3299 void
3300 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3301 {
3302 	struct mbuf *copym;
3303 	struct ip6_hdr *ip6;
3304 
3305 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3306 	if (copym == NULL)
3307 		return;
3308 
3309 	/*
3310 	 * Make sure to deep-copy IPv6 header portion in case the data
3311 	 * is in an mbuf cluster, so that we can safely override the IPv6
3312 	 * header portion later.
3313 	 */
3314 	if (!M_WRITABLE(copym) ||
3315 	    copym->m_len < sizeof(struct ip6_hdr)) {
3316 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3317 		if (copym == NULL)
3318 			return;
3319 	}
3320 	ip6 = mtod(copym, struct ip6_hdr *);
3321 	/*
3322 	 * clear embedded scope identifiers if necessary.
3323 	 * in6_clearscope will touch the addresses only when necessary.
3324 	 */
3325 	in6_clearscope(&ip6->ip6_src);
3326 	in6_clearscope(&ip6->ip6_dst);
3327 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3328 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3329 		    CSUM_PSEUDO_HDR;
3330 		copym->m_pkthdr.csum_data = 0xffff;
3331 	}
3332 	if_simloop(ifp, copym, AF_INET6, 0);
3333 }
3334 
3335 /*
3336  * Chop IPv6 header off from the payload.
3337  */
3338 static int
3339 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3340 {
3341 	struct mbuf *mh;
3342 	struct ip6_hdr *ip6;
3343 
3344 	ip6 = mtod(m, struct ip6_hdr *);
3345 	if (m->m_len > sizeof(*ip6)) {
3346 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3347 		if (mh == NULL) {
3348 			m_freem(m);
3349 			return ENOBUFS;
3350 		}
3351 		m_move_pkthdr(mh, m);
3352 		M_ALIGN(mh, sizeof(*ip6));
3353 		m->m_len -= sizeof(*ip6);
3354 		m->m_data += sizeof(*ip6);
3355 		mh->m_next = m;
3356 		m = mh;
3357 		m->m_len = sizeof(*ip6);
3358 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3359 	}
3360 	exthdrs->ip6e_ip6 = m;
3361 	return 0;
3362 }
3363 
3364 /*
3365  * Compute IPv6 extension header length.
3366  */
3367 int
3368 ip6_optlen(struct inpcb *inp)
3369 {
3370 	int len;
3371 
3372 	if (!inp->in6p_outputopts)
3373 		return 0;
3374 
3375 	len = 0;
3376 #define elen(x) \
3377     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3378 
3379 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3380 	if (inp->in6p_outputopts->ip6po_rthdr)
3381 		/* dest1 is valid with rthdr only */
3382 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3383 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3384 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3385 	return len;
3386 #undef elen
3387 }
3388