xref: /freebsd/sys/netinet6/ip6_output.c (revision 22cf89c938886d14f5796fc49f9f020c23ea8eaf)
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 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 #include "opt_kern_tls.h"
70 #include "opt_ratelimit.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 #include "opt_sctp.h"
74 
75 #include <sys/param.h>
76 #include <sys/kernel.h>
77 #include <sys/ktls.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/errno.h>
81 #include <sys/priv.h>
82 #include <sys/proc.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/ucred.h>
88 
89 #include <machine/in_cksum.h>
90 
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_private.h>
94 #include <net/if_vlan_var.h>
95 #include <net/if_llatbl.h>
96 #include <net/ethernet.h>
97 #include <net/netisr.h>
98 #include <net/route.h>
99 #include <net/route/nhop.h>
100 #include <net/pfil.h>
101 #include <net/rss_config.h>
102 #include <net/vnet.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip_var.h>
107 #include <netinet6/in6_fib.h>
108 #include <netinet6/in6_var.h>
109 #include <netinet/ip6.h>
110 #include <netinet/icmp6.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet/in_pcb.h>
113 #include <netinet/tcp_var.h>
114 #include <netinet6/nd6.h>
115 #include <netinet6/in6_rss.h>
116 
117 #include <netipsec/ipsec_support.h>
118 #if defined(SCTP) || defined(SCTP_SUPPORT)
119 #include <netinet/sctp.h>
120 #include <netinet/sctp_crc32.h>
121 #endif
122 
123 #include <netinet6/scope6_var.h>
124 
125 extern int in6_mcast_loop;
126 
127 struct ip6_exthdrs {
128 	struct mbuf *ip6e_ip6;
129 	struct mbuf *ip6e_hbh;
130 	struct mbuf *ip6e_dest1;
131 	struct mbuf *ip6e_rthdr;
132 	struct mbuf *ip6e_dest2;
133 };
134 
135 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
136 
137 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
138 			   struct ucred *, int);
139 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
140 	struct socket *, struct sockopt *);
141 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
142 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
143 	struct ucred *, int, int, int);
144 
145 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
146 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
147 	struct ip6_frag **);
148 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
149 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
150 static int ip6_getpmtu(struct route_in6 *, int,
151 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
152 	u_int);
153 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
154 	u_long *, int *, u_int);
155 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
156 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
157 
158 /*
159  * Make an extension header from option data.  hp is the source,
160  * mp is the destination, and _ol is the optlen.
161  */
162 #define	MAKE_EXTHDR(hp, mp, _ol)					\
163     do {								\
164 	if (hp) {							\
165 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
166 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
167 		    ((eh)->ip6e_len + 1) << 3);				\
168 		if (error)						\
169 			goto freehdrs;					\
170 		(_ol) += (*(mp))->m_len;				\
171 	}								\
172     } while (/*CONSTCOND*/ 0)
173 
174 /*
175  * Form a chain of extension headers.
176  * m is the extension header mbuf
177  * mp is the previous mbuf in the chain
178  * p is the next header
179  * i is the type of option.
180  */
181 #define MAKE_CHAIN(m, mp, p, i)\
182     do {\
183 	if (m) {\
184 		if (!hdrsplit) \
185 			panic("%s:%d: assumption failed: "\
186 			    "hdr not split: hdrsplit %d exthdrs %p",\
187 			    __func__, __LINE__, hdrsplit, &exthdrs);\
188 		*mtod((m), u_char *) = *(p);\
189 		*(p) = (i);\
190 		p = mtod((m), u_char *);\
191 		(m)->m_next = (mp)->m_next;\
192 		(mp)->m_next = (m);\
193 		(mp) = (m);\
194 	}\
195     } while (/*CONSTCOND*/ 0)
196 
197 void
198 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
199 {
200 	u_short csum;
201 
202 	csum = in_cksum_skip(m, offset + plen, offset);
203 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
204 		csum = 0xffff;
205 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
206 
207 	if (offset + sizeof(csum) > m->m_len)
208 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
209 	else
210 		*(u_short *)mtodo(m, offset) = csum;
211 }
212 
213 static void
214 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
215     int plen, int optlen)
216 {
217 
218 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
219 	    "csum_flags %#x",
220 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
221 
222 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
223 		in6_delayed_cksum(m, plen - optlen,
224 		    sizeof(struct ip6_hdr) + optlen);
225 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
226 	}
227 #if defined(SCTP) || defined(SCTP_SUPPORT)
228 	if (csum_flags & CSUM_SCTP_IPV6) {
229 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
230 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
231 	}
232 #endif
233 }
234 
235 int
236 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
237     int fraglen , uint32_t id)
238 {
239 	struct mbuf *m, **mnext, *m_frgpart;
240 	struct ip6_hdr *ip6, *mhip6;
241 	struct ip6_frag *ip6f;
242 	int off;
243 	int error;
244 	int tlen = m0->m_pkthdr.len;
245 
246 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
247 
248 	m = m0;
249 	ip6 = mtod(m, struct ip6_hdr *);
250 	mnext = &m->m_nextpkt;
251 
252 	for (off = hlen; off < tlen; off += fraglen) {
253 		m = m_gethdr(M_NOWAIT, MT_DATA);
254 		if (!m) {
255 			IP6STAT_INC(ip6s_odropped);
256 			return (ENOBUFS);
257 		}
258 
259 		/*
260 		 * Make sure the complete packet header gets copied
261 		 * from the originating mbuf to the newly created
262 		 * mbuf. This also ensures that existing firewall
263 		 * classification(s), VLAN tags and so on get copied
264 		 * to the resulting fragmented packet(s):
265 		 */
266 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
267 			m_free(m);
268 			IP6STAT_INC(ip6s_odropped);
269 			return (ENOBUFS);
270 		}
271 
272 		*mnext = m;
273 		mnext = &m->m_nextpkt;
274 		m->m_data += max_linkhdr;
275 		mhip6 = mtod(m, struct ip6_hdr *);
276 		*mhip6 = *ip6;
277 		m->m_len = sizeof(*mhip6);
278 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
279 		if (error) {
280 			IP6STAT_INC(ip6s_odropped);
281 			return (error);
282 		}
283 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
284 		if (off + fraglen >= tlen)
285 			fraglen = tlen - off;
286 		else
287 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
288 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
289 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
290 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
291 			IP6STAT_INC(ip6s_odropped);
292 			return (ENOBUFS);
293 		}
294 		m_cat(m, m_frgpart);
295 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
296 		ip6f->ip6f_reserved = 0;
297 		ip6f->ip6f_ident = id;
298 		ip6f->ip6f_nxt = nextproto;
299 		IP6STAT_INC(ip6s_ofragments);
300 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
301 	}
302 
303 	return (0);
304 }
305 
306 static int
307 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
308     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
309     bool stamp_tag)
310 {
311 #ifdef KERN_TLS
312 	struct ktls_session *tls = NULL;
313 #endif
314 	struct m_snd_tag *mst;
315 	int error;
316 
317 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
318 	mst = NULL;
319 
320 #ifdef KERN_TLS
321 	/*
322 	 * If this is an unencrypted TLS record, save a reference to
323 	 * the record.  This local reference is used to call
324 	 * ktls_output_eagain after the mbuf has been freed (thus
325 	 * dropping the mbuf's reference) in if_output.
326 	 */
327 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
328 		tls = ktls_hold(m->m_next->m_epg_tls);
329 		mst = tls->snd_tag;
330 
331 		/*
332 		 * If a TLS session doesn't have a valid tag, it must
333 		 * have had an earlier ifp mismatch, so drop this
334 		 * packet.
335 		 */
336 		if (mst == NULL) {
337 			m_freem(m);
338 			error = EAGAIN;
339 			goto done;
340 		}
341 		/*
342 		 * Always stamp tags that include NIC ktls.
343 		 */
344 		stamp_tag = true;
345 	}
346 #endif
347 #ifdef RATELIMIT
348 	if (inp != NULL && mst == NULL) {
349 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
350 		    (inp->inp_snd_tag != NULL &&
351 		    inp->inp_snd_tag->ifp != ifp))
352 			in_pcboutput_txrtlmt(inp, ifp, m);
353 
354 		if (inp->inp_snd_tag != NULL)
355 			mst = inp->inp_snd_tag;
356 	}
357 #endif
358 	if (stamp_tag && mst != NULL) {
359 		KASSERT(m->m_pkthdr.rcvif == NULL,
360 		    ("trying to add a send tag to a forwarded packet"));
361 		if (mst->ifp != ifp) {
362 			m_freem(m);
363 			error = EAGAIN;
364 			goto done;
365 		}
366 
367 		/* stamp send tag on mbuf */
368 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
369 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
370 	}
371 
372 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
373 
374 done:
375 	/* Check for route change invalidating send tags. */
376 #ifdef KERN_TLS
377 	if (tls != NULL) {
378 		if (error == EAGAIN)
379 			error = ktls_output_eagain(inp, tls);
380 		ktls_free(tls);
381 	}
382 #endif
383 #ifdef RATELIMIT
384 	if (error == EAGAIN)
385 		in_pcboutput_eagain(inp);
386 #endif
387 	return (error);
388 }
389 
390 /*
391  * IP6 output.
392  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
393  * nxt, hlim, src, dst).
394  * This function may modify ver and hlim only.
395  * The mbuf chain containing the packet will be freed.
396  * The mbuf opt, if present, will not be freed.
397  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
398  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
399  * then result of route lookup is stored in ro->ro_nh.
400  *
401  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
402  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
403  *
404  * ifpp - XXX: just for statistics
405  */
406 int
407 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
408     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
409     struct ifnet **ifpp, struct inpcb *inp)
410 {
411 	struct ip6_hdr *ip6;
412 	struct ifnet *ifp, *origifp;
413 	struct mbuf *m = m0;
414 	struct mbuf *mprev;
415 	struct route_in6 *ro_pmtu;
416 	struct nhop_object *nh;
417 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
418 	struct in6_addr odst;
419 	u_char *nexthdrp;
420 	int tlen, len;
421 	int error = 0;
422 	int vlan_pcp = -1;
423 	struct in6_ifaddr *ia = NULL;
424 	u_long mtu;
425 	int alwaysfrag, dontfrag;
426 	u_int32_t optlen, plen = 0, unfragpartlen;
427 	struct ip6_exthdrs exthdrs;
428 	struct in6_addr src0, dst0;
429 	u_int32_t zone;
430 	bool hdrsplit;
431 	int sw_csum, tso;
432 	int needfiblookup;
433 	uint32_t fibnum;
434 	struct m_tag *fwd_tag = NULL;
435 	uint32_t id;
436 
437 	NET_EPOCH_ASSERT();
438 
439 	if (inp != NULL) {
440 		INP_LOCK_ASSERT(inp);
441 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
442 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
443 			/* Unconditionally set flowid. */
444 			m->m_pkthdr.flowid = inp->inp_flowid;
445 			M_HASHTYPE_SET(m, inp->inp_flowtype);
446 		}
447 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
448 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
449 			    INP_2PCP_SHIFT;
450 #ifdef NUMA
451 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
452 #endif
453 	}
454 
455 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
456 	/*
457 	 * IPSec checking which handles several cases.
458 	 * FAST IPSEC: We re-injected the packet.
459 	 * XXX: need scope argument.
460 	 */
461 	if (IPSEC_ENABLED(ipv6)) {
462 		m = mb_unmapped_to_ext(m);
463 		if (m == NULL) {
464 			IP6STAT_INC(ip6s_odropped);
465 			error = ENOBUFS;
466 			goto bad;
467 		}
468 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
469 			if (error == EINPROGRESS)
470 				error = 0;
471 			goto done;
472 		}
473 	}
474 #endif /* IPSEC */
475 
476 	/* Source address validation. */
477 	ip6 = mtod(m, struct ip6_hdr *);
478 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
479 	    (flags & IPV6_UNSPECSRC) == 0) {
480 		error = EOPNOTSUPP;
481 		IP6STAT_INC(ip6s_badscope);
482 		goto bad;
483 	}
484 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
485 		error = EOPNOTSUPP;
486 		IP6STAT_INC(ip6s_badscope);
487 		goto bad;
488 	}
489 
490 	/*
491 	 * If we are given packet options to add extension headers prepare them.
492 	 * Calculate the total length of the extension header chain.
493 	 * Keep the length of the unfragmentable part for fragmentation.
494 	 */
495 	bzero(&exthdrs, sizeof(exthdrs));
496 	optlen = 0;
497 	unfragpartlen = sizeof(struct ip6_hdr);
498 	if (opt) {
499 		/* Hop-by-Hop options header. */
500 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
501 
502 		/* Destination options header (1st part). */
503 		if (opt->ip6po_rthdr) {
504 #ifndef RTHDR_SUPPORT_IMPLEMENTED
505 			/*
506 			 * If there is a routing header, discard the packet
507 			 * right away here. RH0/1 are obsolete and we do not
508 			 * currently support RH2/3/4.
509 			 * People trying to use RH253/254 may want to disable
510 			 * this check.
511 			 * The moment we do support any routing header (again)
512 			 * this block should check the routing type more
513 			 * selectively.
514 			 */
515 			error = EINVAL;
516 			goto bad;
517 #endif
518 
519 			/*
520 			 * Destination options header (1st part).
521 			 * This only makes sense with a routing header.
522 			 * See Section 9.2 of RFC 3542.
523 			 * Disabling this part just for MIP6 convenience is
524 			 * a bad idea.  We need to think carefully about a
525 			 * way to make the advanced API coexist with MIP6
526 			 * options, which might automatically be inserted in
527 			 * the kernel.
528 			 */
529 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
530 			    optlen);
531 		}
532 		/* Routing header. */
533 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
534 
535 		unfragpartlen += optlen;
536 
537 		/*
538 		 * NOTE: we don't add AH/ESP length here (done in
539 		 * ip6_ipsec_output()).
540 		 */
541 
542 		/* Destination options header (2nd part). */
543 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
544 	}
545 
546 	/*
547 	 * If there is at least one extension header,
548 	 * separate IP6 header from the payload.
549 	 */
550 	hdrsplit = false;
551 	if (optlen) {
552 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
553 			m = NULL;
554 			goto freehdrs;
555 		}
556 		m = exthdrs.ip6e_ip6;
557 		ip6 = mtod(m, struct ip6_hdr *);
558 		hdrsplit = true;
559 	}
560 
561 	/* Adjust mbuf packet header length. */
562 	m->m_pkthdr.len += optlen;
563 	plen = m->m_pkthdr.len - sizeof(*ip6);
564 
565 	/* If this is a jumbo payload, insert a jumbo payload option. */
566 	if (plen > IPV6_MAXPACKET) {
567 		if (!hdrsplit) {
568 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
569 				m = NULL;
570 				goto freehdrs;
571 			}
572 			m = exthdrs.ip6e_ip6;
573 			ip6 = mtod(m, struct ip6_hdr *);
574 			hdrsplit = true;
575 		}
576 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
577 			goto freehdrs;
578 		ip6->ip6_plen = 0;
579 	} else
580 		ip6->ip6_plen = htons(plen);
581 	nexthdrp = &ip6->ip6_nxt;
582 
583 	if (optlen) {
584 		/*
585 		 * Concatenate headers and fill in next header fields.
586 		 * Here we have, on "m"
587 		 *	IPv6 payload
588 		 * and we insert headers accordingly.
589 		 * Finally, we should be getting:
590 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
591 		 *
592 		 * During the header composing process "m" points to IPv6
593 		 * header.  "mprev" points to an extension header prior to esp.
594 		 */
595 		mprev = m;
596 
597 		/*
598 		 * We treat dest2 specially.  This makes IPsec processing
599 		 * much easier.  The goal here is to make mprev point the
600 		 * mbuf prior to dest2.
601 		 *
602 		 * Result: IPv6 dest2 payload.
603 		 * m and mprev will point to IPv6 header.
604 		 */
605 		if (exthdrs.ip6e_dest2) {
606 			if (!hdrsplit)
607 				panic("%s:%d: assumption failed: "
608 				    "hdr not split: hdrsplit %d exthdrs %p",
609 				    __func__, __LINE__, hdrsplit, &exthdrs);
610 			exthdrs.ip6e_dest2->m_next = m->m_next;
611 			m->m_next = exthdrs.ip6e_dest2;
612 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
613 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
614 		}
615 
616 		/*
617 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
618 		 * m will point to IPv6 header.  mprev will point to the
619 		 * extension header prior to dest2 (rthdr in the above case).
620 		 */
621 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
622 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
623 			   IPPROTO_DSTOPTS);
624 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
625 			   IPPROTO_ROUTING);
626 	}
627 
628 	IP6STAT_INC(ip6s_localout);
629 
630 	/* Route packet. */
631 	ro_pmtu = ro;
632 	if (opt && opt->ip6po_rthdr)
633 		ro = &opt->ip6po_route;
634 	if (ro != NULL)
635 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
636 	else
637 		dst = &sin6;
638 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
639 
640 again:
641 	/*
642 	 * If specified, try to fill in the traffic class field.
643 	 * Do not override if a non-zero value is already set.
644 	 * We check the diffserv field and the ECN field separately.
645 	 */
646 	if (opt && opt->ip6po_tclass >= 0) {
647 		int mask = 0;
648 
649 		if (IPV6_DSCP(ip6) == 0)
650 			mask |= 0xfc;
651 		if (IPV6_ECN(ip6) == 0)
652 			mask |= 0x03;
653 		if (mask != 0)
654 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
655 	}
656 
657 	/* Fill in or override the hop limit field, if necessary. */
658 	if (opt && opt->ip6po_hlim != -1)
659 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
660 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
661 		if (im6o != NULL)
662 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
663 		else
664 			ip6->ip6_hlim = V_ip6_defmcasthlim;
665 	}
666 
667 	if (ro == NULL || ro->ro_nh == NULL) {
668 		bzero(dst, sizeof(*dst));
669 		dst->sin6_family = AF_INET6;
670 		dst->sin6_len = sizeof(*dst);
671 		dst->sin6_addr = ip6->ip6_dst;
672 	}
673 	/*
674 	 * Validate route against routing table changes.
675 	 * Make sure that the address family is set in route.
676 	 */
677 	nh = NULL;
678 	ifp = NULL;
679 	mtu = 0;
680 	if (ro != NULL) {
681 		if (ro->ro_nh != NULL && inp != NULL) {
682 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
683 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
684 			    fibnum);
685 		}
686 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
687 		    (!NH_IS_VALID(ro->ro_nh) ||
688 		    ro->ro_dst.sin6_family != AF_INET6 ||
689 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
690 			RO_INVALIDATE_CACHE(ro);
691 
692 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
693 		    ro->ro_dst.sin6_family == AF_INET6 &&
694 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
695 			/* Nexthop is valid and contains valid ifp */
696 			nh = ro->ro_nh;
697 		} else {
698 			if (ro->ro_lle)
699 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
700 			ro->ro_lle = NULL;
701 			if (fwd_tag == NULL) {
702 				bzero(&dst_sa, sizeof(dst_sa));
703 				dst_sa.sin6_family = AF_INET6;
704 				dst_sa.sin6_len = sizeof(dst_sa);
705 				dst_sa.sin6_addr = ip6->ip6_dst;
706 			}
707 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
708 			    &nh, fibnum, m->m_pkthdr.flowid);
709 			if (error != 0) {
710 				IP6STAT_INC(ip6s_noroute);
711 				if (ifp != NULL)
712 					in6_ifstat_inc(ifp, ifs6_out_discard);
713 				goto bad;
714 			}
715 			/*
716 			 * At this point at least @ifp is not NULL
717 			 * Can be the case when dst is multicast, link-local or
718 			 * interface is explicitly specificed by the caller.
719 			 */
720 		}
721 		if (nh == NULL) {
722 			/*
723 			 * If in6_selectroute() does not return a nexthop
724 			 * dst may not have been updated.
725 			 */
726 			*dst = dst_sa;	/* XXX */
727 			origifp = ifp;
728 			mtu = ifp->if_mtu;
729 		} else {
730 			ifp = nh->nh_ifp;
731 			origifp = nh->nh_aifp;
732 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
733 			counter_u64_add(nh->nh_pksent, 1);
734 		}
735 	} else {
736 		struct nhop_object *nh;
737 		struct in6_addr kdst;
738 		uint32_t scopeid;
739 
740 		if (fwd_tag == NULL) {
741 			bzero(&dst_sa, sizeof(dst_sa));
742 			dst_sa.sin6_family = AF_INET6;
743 			dst_sa.sin6_len = sizeof(dst_sa);
744 			dst_sa.sin6_addr = ip6->ip6_dst;
745 		}
746 
747 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
748 		    im6o != NULL &&
749 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
750 			/* We do not need a route lookup. */
751 			*dst = dst_sa;	/* XXX */
752 			origifp = ifp;
753 			goto nonh6lookup;
754 		}
755 
756 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
757 
758 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
759 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
760 			if (scopeid > 0) {
761 				ifp = in6_getlinkifnet(scopeid);
762 				if (ifp == NULL) {
763 					error = EHOSTUNREACH;
764 					goto bad;
765 				}
766 				*dst = dst_sa;	/* XXX */
767 				origifp = ifp;
768 				goto nonh6lookup;
769 			}
770 		}
771 
772 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
773 		    m->m_pkthdr.flowid);
774 		if (nh == NULL) {
775 			IP6STAT_INC(ip6s_noroute);
776 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
777 			error = EHOSTUNREACH;
778 			goto bad;
779 		}
780 
781 		ifp = nh->nh_ifp;
782 		origifp = nh->nh_aifp;
783 		ia = ifatoia6(nh->nh_ifa);
784 		if (nh->nh_flags & NHF_GATEWAY)
785 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
786 		else if (fwd_tag != NULL)
787 			dst->sin6_addr = dst_sa.sin6_addr;
788 nonh6lookup:
789 		;
790 	}
791 	/*
792 	 * At this point ifp MUST be pointing to the valid transmit ifp.
793 	 * origifp MUST be valid and pointing to either the same ifp or,
794 	 * in case of loopback output, to the interface which ip6_src
795 	 * belongs to.
796 	 * Examples:
797 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
798 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
799 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
800 	 *
801 	 * mtu can be 0 and will be refined later.
802 	 */
803 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
804 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
805 
806 	if ((flags & IPV6_FORWARDING) == 0) {
807 		/* XXX: the FORWARDING flag can be set for mrouting. */
808 		in6_ifstat_inc(ifp, ifs6_out_request);
809 	}
810 
811 	/* Setup data structures for scope ID checks. */
812 	src0 = ip6->ip6_src;
813 	bzero(&src_sa, sizeof(src_sa));
814 	src_sa.sin6_family = AF_INET6;
815 	src_sa.sin6_len = sizeof(src_sa);
816 	src_sa.sin6_addr = ip6->ip6_src;
817 
818 	dst0 = ip6->ip6_dst;
819 	/* Re-initialize to be sure. */
820 	bzero(&dst_sa, sizeof(dst_sa));
821 	dst_sa.sin6_family = AF_INET6;
822 	dst_sa.sin6_len = sizeof(dst_sa);
823 	dst_sa.sin6_addr = ip6->ip6_dst;
824 
825 	/* Check for valid scope ID. */
826 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
827 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
828 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
829 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
830 		/*
831 		 * The outgoing interface is in the zone of the source
832 		 * and destination addresses.
833 		 *
834 		 */
835 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
836 	    sa6_recoverscope(&src_sa) != 0 ||
837 	    sa6_recoverscope(&dst_sa) != 0 ||
838 	    dst_sa.sin6_scope_id == 0 ||
839 	    (src_sa.sin6_scope_id != 0 &&
840 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
841 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
842 		/*
843 		 * If the destination network interface is not a
844 		 * loopback interface, or the destination network
845 		 * address has no scope ID, or the source address has
846 		 * a scope ID set which is different from the
847 		 * destination address one, or there is no network
848 		 * interface representing this scope ID, the address
849 		 * pair is considered invalid.
850 		 */
851 		IP6STAT_INC(ip6s_badscope);
852 		in6_ifstat_inc(origifp, ifs6_out_discard);
853 		if (error == 0)
854 			error = EHOSTUNREACH; /* XXX */
855 		goto bad;
856 	}
857 	/* All scope ID checks are successful. */
858 
859 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
860 		if (opt && opt->ip6po_nextroute.ro_nh) {
861 			/*
862 			 * The nexthop is explicitly specified by the
863 			 * application.  We assume the next hop is an IPv6
864 			 * address.
865 			 */
866 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
867 		}
868 		else if ((nh->nh_flags & NHF_GATEWAY))
869 			dst = &nh->gw6_sa;
870 	}
871 
872 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
873 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
874 	} else {
875 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
876 		in6_ifstat_inc(ifp, ifs6_out_mcast);
877 
878 		/* Confirm that the outgoing interface supports multicast. */
879 		if (!(ifp->if_flags & IFF_MULTICAST)) {
880 			IP6STAT_INC(ip6s_noroute);
881 			in6_ifstat_inc(ifp, ifs6_out_discard);
882 			error = ENETUNREACH;
883 			goto bad;
884 		}
885 		if ((im6o == NULL && in6_mcast_loop) ||
886 		    (im6o && im6o->im6o_multicast_loop)) {
887 			/*
888 			 * Loop back multicast datagram if not expressly
889 			 * forbidden to do so, even if we have not joined
890 			 * the address; protocols will filter it later,
891 			 * thus deferring a hash lookup and lock acquisition
892 			 * at the expense of an m_copym().
893 			 */
894 			ip6_mloopback(ifp, m);
895 		} else {
896 			/*
897 			 * If we are acting as a multicast router, perform
898 			 * multicast forwarding as if the packet had just
899 			 * arrived on the interface to which we are about
900 			 * to send.  The multicast forwarding function
901 			 * recursively calls this function, using the
902 			 * IPV6_FORWARDING flag to prevent infinite recursion.
903 			 *
904 			 * Multicasts that are looped back by ip6_mloopback(),
905 			 * above, will be forwarded by the ip6_input() routine,
906 			 * if necessary.
907 			 */
908 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
909 				/*
910 				 * XXX: ip6_mforward expects that rcvif is NULL
911 				 * when it is called from the originating path.
912 				 * However, it may not always be the case.
913 				 */
914 				m->m_pkthdr.rcvif = NULL;
915 				if (ip6_mforward(ip6, ifp, m) != 0) {
916 					m_freem(m);
917 					goto done;
918 				}
919 			}
920 		}
921 		/*
922 		 * Multicasts with a hoplimit of zero may be looped back,
923 		 * above, but must not be transmitted on a network.
924 		 * Also, multicasts addressed to the loopback interface
925 		 * are not sent -- the above call to ip6_mloopback() will
926 		 * loop back a copy if this host actually belongs to the
927 		 * destination group on the loopback interface.
928 		 */
929 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
930 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
931 			m_freem(m);
932 			goto done;
933 		}
934 	}
935 
936 	/*
937 	 * Fill the outgoing inteface to tell the upper layer
938 	 * to increment per-interface statistics.
939 	 */
940 	if (ifpp)
941 		*ifpp = ifp;
942 
943 	/* Determine path MTU. */
944 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
945 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
946 		goto bad;
947 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
948 	    "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
949 	    ifp, alwaysfrag, fibnum));
950 
951 	/*
952 	 * The caller of this function may specify to use the minimum MTU
953 	 * in some cases.
954 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
955 	 * setting.  The logic is a bit complicated; by default, unicast
956 	 * packets will follow path MTU while multicast packets will be sent at
957 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
958 	 * including unicast ones will be sent at the minimum MTU.  Multicast
959 	 * packets will always be sent at the minimum MTU unless
960 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
961 	 * See RFC 3542 for more details.
962 	 */
963 	if (mtu > IPV6_MMTU) {
964 		if ((flags & IPV6_MINMTU))
965 			mtu = IPV6_MMTU;
966 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
967 			mtu = IPV6_MMTU;
968 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
969 			 (opt == NULL ||
970 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
971 			mtu = IPV6_MMTU;
972 		}
973 	}
974 
975 	/*
976 	 * Clear embedded scope identifiers if necessary.
977 	 * in6_clearscope() will touch the addresses only when necessary.
978 	 */
979 	in6_clearscope(&ip6->ip6_src);
980 	in6_clearscope(&ip6->ip6_dst);
981 
982 	/*
983 	 * If the outgoing packet contains a hop-by-hop options header,
984 	 * it must be examined and processed even by the source node.
985 	 * (RFC 2460, section 4.)
986 	 */
987 	if (exthdrs.ip6e_hbh) {
988 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
989 		u_int32_t dummy; /* XXX unused */
990 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
991 
992 #ifdef DIAGNOSTIC
993 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
994 			panic("ip6e_hbh is not contiguous");
995 #endif
996 		/*
997 		 *  XXX: if we have to send an ICMPv6 error to the sender,
998 		 *       we need the M_LOOP flag since icmp6_error() expects
999 		 *       the IPv6 and the hop-by-hop options header are
1000 		 *       contiguous unless the flag is set.
1001 		 */
1002 		m->m_flags |= M_LOOP;
1003 		m->m_pkthdr.rcvif = ifp;
1004 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1005 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1006 		    &dummy, &plen) < 0) {
1007 			/* m was already freed at this point. */
1008 			error = EINVAL;/* better error? */
1009 			goto done;
1010 		}
1011 		m->m_flags &= ~M_LOOP; /* XXX */
1012 		m->m_pkthdr.rcvif = NULL;
1013 	}
1014 
1015 	/* Jump over all PFIL processing if hooks are not active. */
1016 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1017 		goto passout;
1018 
1019 	odst = ip6->ip6_dst;
1020 	/* Run through list of hooks for output packets. */
1021 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1022 	case PFIL_PASS:
1023 		ip6 = mtod(m, struct ip6_hdr *);
1024 		break;
1025 	case PFIL_DROPPED:
1026 		error = EACCES;
1027 		/* FALLTHROUGH */
1028 	case PFIL_CONSUMED:
1029 		goto done;
1030 	}
1031 
1032 	needfiblookup = 0;
1033 	/* See if destination IP address was changed by packet filter. */
1034 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1035 		m->m_flags |= M_SKIP_FIREWALL;
1036 		/* If destination is now ourself drop to ip6_input(). */
1037 		if (in6_localip(&ip6->ip6_dst)) {
1038 			m->m_flags |= M_FASTFWD_OURS;
1039 			if (m->m_pkthdr.rcvif == NULL)
1040 				m->m_pkthdr.rcvif = V_loif;
1041 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1042 				m->m_pkthdr.csum_flags |=
1043 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1044 				m->m_pkthdr.csum_data = 0xffff;
1045 			}
1046 #if defined(SCTP) || defined(SCTP_SUPPORT)
1047 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1048 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1049 #endif
1050 			error = netisr_queue(NETISR_IPV6, m);
1051 			goto done;
1052 		} else {
1053 			if (ro != NULL)
1054 				RO_INVALIDATE_CACHE(ro);
1055 			needfiblookup = 1; /* Redo the routing table lookup. */
1056 		}
1057 	}
1058 	/* See if fib was changed by packet filter. */
1059 	if (fibnum != M_GETFIB(m)) {
1060 		m->m_flags |= M_SKIP_FIREWALL;
1061 		fibnum = M_GETFIB(m);
1062 		if (ro != NULL)
1063 			RO_INVALIDATE_CACHE(ro);
1064 		needfiblookup = 1;
1065 	}
1066 	if (needfiblookup)
1067 		goto again;
1068 
1069 	/* See if local, if yes, send it to netisr. */
1070 	if (m->m_flags & M_FASTFWD_OURS) {
1071 		if (m->m_pkthdr.rcvif == NULL)
1072 			m->m_pkthdr.rcvif = V_loif;
1073 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1074 			m->m_pkthdr.csum_flags |=
1075 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1076 			m->m_pkthdr.csum_data = 0xffff;
1077 		}
1078 #if defined(SCTP) || defined(SCTP_SUPPORT)
1079 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1080 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1081 #endif
1082 		error = netisr_queue(NETISR_IPV6, m);
1083 		goto done;
1084 	}
1085 	/* Or forward to some other address? */
1086 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1087 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1088 		if (ro != NULL)
1089 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1090 		else
1091 			dst = &sin6;
1092 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1093 		m->m_flags |= M_SKIP_FIREWALL;
1094 		m->m_flags &= ~M_IP6_NEXTHOP;
1095 		m_tag_delete(m, fwd_tag);
1096 		goto again;
1097 	}
1098 
1099 passout:
1100 	if (vlan_pcp > -1)
1101 		EVL_APPLY_PRI(m, vlan_pcp);
1102 
1103 	/* Ensure the packet data is mapped if the interface requires it. */
1104 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1105 		m = mb_unmapped_to_ext(m);
1106 		if (m == NULL) {
1107 			IP6STAT_INC(ip6s_odropped);
1108 			return (ENOBUFS);
1109 		}
1110 	}
1111 
1112 	/*
1113 	 * Send the packet to the outgoing interface.
1114 	 * If necessary, do IPv6 fragmentation before sending.
1115 	 *
1116 	 * The logic here is rather complex:
1117 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1118 	 * 1-a:	send as is if tlen <= path mtu
1119 	 * 1-b:	fragment if tlen > path mtu
1120 	 *
1121 	 * 2: if user asks us not to fragment (dontfrag == 1)
1122 	 * 2-a:	send as is if tlen <= interface mtu
1123 	 * 2-b:	error if tlen > interface mtu
1124 	 *
1125 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1126 	 *	always fragment
1127 	 *
1128 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1129 	 *	error, as we cannot handle this conflicting request.
1130 	 */
1131 	sw_csum = m->m_pkthdr.csum_flags;
1132 	if (!hdrsplit) {
1133 		tso = ((sw_csum & ifp->if_hwassist &
1134 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1135 		sw_csum &= ~ifp->if_hwassist;
1136 	} else
1137 		tso = 0;
1138 	/*
1139 	 * If we added extension headers, we will not do TSO and calculate the
1140 	 * checksums ourselves for now.
1141 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1142 	 * with ext. hdrs.
1143 	 */
1144 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1145 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1146 	tlen = m->m_pkthdr.len;
1147 
1148 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1149 		dontfrag = 1;
1150 	else
1151 		dontfrag = 0;
1152 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1153 		/* Conflicting request - can't transmit. */
1154 		error = EMSGSIZE;
1155 		goto bad;
1156 	}
1157 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1158 		/*
1159 		 * Even if the DONTFRAG option is specified, we cannot send the
1160 		 * packet when the data length is larger than the MTU of the
1161 		 * outgoing interface.
1162 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1163 		 * application wanted to know the MTU value. Also return an
1164 		 * error code (this is not described in the API spec).
1165 		 */
1166 		if (inp != NULL)
1167 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1168 		error = EMSGSIZE;
1169 		goto bad;
1170 	}
1171 
1172 	/* Transmit packet without fragmentation. */
1173 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1174 		struct in6_ifaddr *ia6;
1175 
1176 		ip6 = mtod(m, struct ip6_hdr *);
1177 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1178 		if (ia6) {
1179 			/* Record statistics for this interface address. */
1180 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1181 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1182 			    m->m_pkthdr.len);
1183 		}
1184 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1185 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1186 		goto done;
1187 	}
1188 
1189 	/* Try to fragment the packet.  Cases 1-b and 3. */
1190 	if (mtu < IPV6_MMTU) {
1191 		/* Path MTU cannot be less than IPV6_MMTU. */
1192 		error = EMSGSIZE;
1193 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1194 		goto bad;
1195 	} else if (ip6->ip6_plen == 0) {
1196 		/* Jumbo payload cannot be fragmented. */
1197 		error = EMSGSIZE;
1198 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1199 		goto bad;
1200 	} else {
1201 		u_char nextproto;
1202 
1203 		/*
1204 		 * Too large for the destination or interface;
1205 		 * fragment if possible.
1206 		 * Must be able to put at least 8 bytes per fragment.
1207 		 */
1208 		if (mtu > IPV6_MAXPACKET)
1209 			mtu = IPV6_MAXPACKET;
1210 
1211 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1212 		if (len < 8) {
1213 			error = EMSGSIZE;
1214 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1215 			goto bad;
1216 		}
1217 
1218 		/*
1219 		 * If the interface will not calculate checksums on
1220 		 * fragmented packets, then do it here.
1221 		 * XXX-BZ handle the hw offloading case.  Need flags.
1222 		 */
1223 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1224 		    optlen);
1225 
1226 		/*
1227 		 * Change the next header field of the last header in the
1228 		 * unfragmentable part.
1229 		 */
1230 		if (exthdrs.ip6e_rthdr) {
1231 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1232 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1233 		} else if (exthdrs.ip6e_dest1) {
1234 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1235 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1236 		} else if (exthdrs.ip6e_hbh) {
1237 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1238 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1239 		} else {
1240 			ip6 = mtod(m, struct ip6_hdr *);
1241 			nextproto = ip6->ip6_nxt;
1242 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1243 		}
1244 
1245 		/*
1246 		 * Loop through length of segment after first fragment,
1247 		 * make new header and copy data of each part and link onto
1248 		 * chain.
1249 		 */
1250 		m0 = m;
1251 		id = htonl(ip6_randomid());
1252 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1253 		if (error != 0)
1254 			goto sendorfree;
1255 
1256 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1257 	}
1258 
1259 	/* Remove leading garbage. */
1260 sendorfree:
1261 	m = m0->m_nextpkt;
1262 	m0->m_nextpkt = 0;
1263 	m_freem(m0);
1264 	for (; m; m = m0) {
1265 		m0 = m->m_nextpkt;
1266 		m->m_nextpkt = 0;
1267 		if (error == 0) {
1268 			/* Record statistics for this interface address. */
1269 			if (ia) {
1270 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1271 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1272 				    m->m_pkthdr.len);
1273 			}
1274 			if (vlan_pcp > -1)
1275 				EVL_APPLY_PRI(m, vlan_pcp);
1276 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1277 			    true);
1278 		} else
1279 			m_freem(m);
1280 	}
1281 
1282 	if (error == 0)
1283 		IP6STAT_INC(ip6s_fragmented);
1284 
1285 done:
1286 	return (error);
1287 
1288 freehdrs:
1289 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1290 	m_freem(exthdrs.ip6e_dest1);
1291 	m_freem(exthdrs.ip6e_rthdr);
1292 	m_freem(exthdrs.ip6e_dest2);
1293 	/* FALLTHROUGH */
1294 bad:
1295 	if (m)
1296 		m_freem(m);
1297 	goto done;
1298 }
1299 
1300 static int
1301 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1302 {
1303 	struct mbuf *m;
1304 
1305 	if (hlen > MCLBYTES)
1306 		return (ENOBUFS); /* XXX */
1307 
1308 	if (hlen > MLEN)
1309 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1310 	else
1311 		m = m_get(M_NOWAIT, MT_DATA);
1312 	if (m == NULL)
1313 		return (ENOBUFS);
1314 	m->m_len = hlen;
1315 	if (hdr)
1316 		bcopy(hdr, mtod(m, caddr_t), hlen);
1317 
1318 	*mp = m;
1319 	return (0);
1320 }
1321 
1322 /*
1323  * Insert jumbo payload option.
1324  */
1325 static int
1326 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1327 {
1328 	struct mbuf *mopt;
1329 	u_char *optbuf;
1330 	u_int32_t v;
1331 
1332 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1333 
1334 	/*
1335 	 * If there is no hop-by-hop options header, allocate new one.
1336 	 * If there is one but it doesn't have enough space to store the
1337 	 * jumbo payload option, allocate a cluster to store the whole options.
1338 	 * Otherwise, use it to store the options.
1339 	 */
1340 	if (exthdrs->ip6e_hbh == NULL) {
1341 		mopt = m_get(M_NOWAIT, MT_DATA);
1342 		if (mopt == NULL)
1343 			return (ENOBUFS);
1344 		mopt->m_len = JUMBOOPTLEN;
1345 		optbuf = mtod(mopt, u_char *);
1346 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1347 		exthdrs->ip6e_hbh = mopt;
1348 	} else {
1349 		struct ip6_hbh *hbh;
1350 
1351 		mopt = exthdrs->ip6e_hbh;
1352 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1353 			/*
1354 			 * XXX assumption:
1355 			 * - exthdrs->ip6e_hbh is not referenced from places
1356 			 *   other than exthdrs.
1357 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1358 			 */
1359 			int oldoptlen = mopt->m_len;
1360 			struct mbuf *n;
1361 
1362 			/*
1363 			 * XXX: give up if the whole (new) hbh header does
1364 			 * not fit even in an mbuf cluster.
1365 			 */
1366 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1367 				return (ENOBUFS);
1368 
1369 			/*
1370 			 * As a consequence, we must always prepare a cluster
1371 			 * at this point.
1372 			 */
1373 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1374 			if (n == NULL)
1375 				return (ENOBUFS);
1376 			n->m_len = oldoptlen + JUMBOOPTLEN;
1377 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1378 			    oldoptlen);
1379 			optbuf = mtod(n, caddr_t) + oldoptlen;
1380 			m_freem(mopt);
1381 			mopt = exthdrs->ip6e_hbh = n;
1382 		} else {
1383 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1384 			mopt->m_len += JUMBOOPTLEN;
1385 		}
1386 		optbuf[0] = IP6OPT_PADN;
1387 		optbuf[1] = 1;
1388 
1389 		/*
1390 		 * Adjust the header length according to the pad and
1391 		 * the jumbo payload option.
1392 		 */
1393 		hbh = mtod(mopt, struct ip6_hbh *);
1394 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1395 	}
1396 
1397 	/* fill in the option. */
1398 	optbuf[2] = IP6OPT_JUMBO;
1399 	optbuf[3] = 4;
1400 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1401 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1402 
1403 	/* finally, adjust the packet header length */
1404 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1405 
1406 	return (0);
1407 #undef JUMBOOPTLEN
1408 }
1409 
1410 /*
1411  * Insert fragment header and copy unfragmentable header portions.
1412  */
1413 static int
1414 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1415     struct ip6_frag **frghdrp)
1416 {
1417 	struct mbuf *n, *mlast;
1418 
1419 	if (hlen > sizeof(struct ip6_hdr)) {
1420 		n = m_copym(m0, sizeof(struct ip6_hdr),
1421 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1422 		if (n == NULL)
1423 			return (ENOBUFS);
1424 		m->m_next = n;
1425 	} else
1426 		n = m;
1427 
1428 	/* Search for the last mbuf of unfragmentable part. */
1429 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1430 		;
1431 
1432 	if (M_WRITABLE(mlast) &&
1433 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1434 		/* use the trailing space of the last mbuf for the fragment hdr */
1435 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1436 		    mlast->m_len);
1437 		mlast->m_len += sizeof(struct ip6_frag);
1438 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1439 	} else {
1440 		/* allocate a new mbuf for the fragment header */
1441 		struct mbuf *mfrg;
1442 
1443 		mfrg = m_get(M_NOWAIT, MT_DATA);
1444 		if (mfrg == NULL)
1445 			return (ENOBUFS);
1446 		mfrg->m_len = sizeof(struct ip6_frag);
1447 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1448 		mlast->m_next = mfrg;
1449 	}
1450 
1451 	return (0);
1452 }
1453 
1454 /*
1455  * Calculates IPv6 path mtu for destination @dst.
1456  * Resulting MTU is stored in @mtup.
1457  *
1458  * Returns 0 on success.
1459  */
1460 static int
1461 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1462 {
1463 	struct epoch_tracker et;
1464 	struct nhop_object *nh;
1465 	struct in6_addr kdst;
1466 	uint32_t scopeid;
1467 	int error;
1468 
1469 	in6_splitscope(dst, &kdst, &scopeid);
1470 
1471 	NET_EPOCH_ENTER(et);
1472 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1473 	if (nh != NULL)
1474 		error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1475 	else
1476 		error = EHOSTUNREACH;
1477 	NET_EPOCH_EXIT(et);
1478 
1479 	return (error);
1480 }
1481 
1482 /*
1483  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1484  * and cached data in @ro_pmtu.
1485  * MTU from (successful) route lookup is saved (along with dst)
1486  * inside @ro_pmtu to avoid subsequent route lookups after packet
1487  * filter processing.
1488  *
1489  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1490  * Returns 0 on success.
1491  */
1492 static int
1493 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1494     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1495     int *alwaysfragp, u_int fibnum, u_int proto)
1496 {
1497 	struct nhop_object *nh;
1498 	struct in6_addr kdst;
1499 	uint32_t scopeid;
1500 	struct sockaddr_in6 *sa6_dst, sin6;
1501 	u_long mtu;
1502 
1503 	NET_EPOCH_ASSERT();
1504 
1505 	mtu = 0;
1506 	if (ro_pmtu == NULL || do_lookup) {
1507 		/*
1508 		 * Here ro_pmtu has final destination address, while
1509 		 * ro might represent immediate destination.
1510 		 * Use ro_pmtu destination since mtu might differ.
1511 		 */
1512 		if (ro_pmtu != NULL) {
1513 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1514 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1515 				ro_pmtu->ro_mtu = 0;
1516 		} else
1517 			sa6_dst = &sin6;
1518 
1519 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1520 			bzero(sa6_dst, sizeof(*sa6_dst));
1521 			sa6_dst->sin6_family = AF_INET6;
1522 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1523 			sa6_dst->sin6_addr = *dst;
1524 
1525 			in6_splitscope(dst, &kdst, &scopeid);
1526 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1527 			if (nh != NULL) {
1528 				mtu = nh->nh_mtu;
1529 				if (ro_pmtu != NULL)
1530 					ro_pmtu->ro_mtu = mtu;
1531 			}
1532 		} else
1533 			mtu = ro_pmtu->ro_mtu;
1534 	}
1535 
1536 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1537 		mtu = ro_pmtu->ro_nh->nh_mtu;
1538 
1539 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1540 }
1541 
1542 /*
1543  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1544  * hostcache data for @dst.
1545  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1546  *
1547  * Returns 0 on success.
1548  */
1549 static int
1550 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1551     u_long *mtup, int *alwaysfragp, u_int proto)
1552 {
1553 	u_long mtu = 0;
1554 	int alwaysfrag = 0;
1555 	int error = 0;
1556 
1557 	if (rt_mtu > 0) {
1558 		u_int32_t ifmtu;
1559 		struct in_conninfo inc;
1560 
1561 		bzero(&inc, sizeof(inc));
1562 		inc.inc_flags |= INC_ISIPV6;
1563 		inc.inc6_faddr = *dst;
1564 
1565 		ifmtu = IN6_LINKMTU(ifp);
1566 
1567 		/* TCP is known to react to pmtu changes so skip hc */
1568 		if (proto != IPPROTO_TCP)
1569 			mtu = tcp_hc_getmtu(&inc);
1570 
1571 		if (mtu)
1572 			mtu = min(mtu, rt_mtu);
1573 		else
1574 			mtu = rt_mtu;
1575 		if (mtu == 0)
1576 			mtu = ifmtu;
1577 		else if (mtu < IPV6_MMTU) {
1578 			/*
1579 			 * RFC2460 section 5, last paragraph:
1580 			 * if we record ICMPv6 too big message with
1581 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1582 			 * or smaller, with framgent header attached.
1583 			 * (fragment header is needed regardless from the
1584 			 * packet size, for translators to identify packets)
1585 			 */
1586 			alwaysfrag = 1;
1587 			mtu = IPV6_MMTU;
1588 		}
1589 	} else if (ifp) {
1590 		mtu = IN6_LINKMTU(ifp);
1591 	} else
1592 		error = EHOSTUNREACH; /* XXX */
1593 
1594 	*mtup = mtu;
1595 	if (alwaysfragp)
1596 		*alwaysfragp = alwaysfrag;
1597 	return (error);
1598 }
1599 
1600 /*
1601  * IP6 socket option processing.
1602  */
1603 int
1604 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1605 {
1606 	int optdatalen, uproto;
1607 	void *optdata;
1608 	struct inpcb *inp = sotoinpcb(so);
1609 	int error, optval;
1610 	int level, op, optname;
1611 	int optlen;
1612 	struct thread *td;
1613 #ifdef	RSS
1614 	uint32_t rss_bucket;
1615 	int retval;
1616 #endif
1617 
1618 /*
1619  * Don't use more than a quarter of mbuf clusters.  N.B.:
1620  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1621  * on LP64 architectures, so cast to u_long to avoid undefined
1622  * behavior.  ILP32 architectures cannot have nmbclusters
1623  * large enough to overflow for other reasons.
1624  */
1625 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1626 
1627 	level = sopt->sopt_level;
1628 	op = sopt->sopt_dir;
1629 	optname = sopt->sopt_name;
1630 	optlen = sopt->sopt_valsize;
1631 	td = sopt->sopt_td;
1632 	error = 0;
1633 	optval = 0;
1634 	uproto = (int)so->so_proto->pr_protocol;
1635 
1636 	if (level != IPPROTO_IPV6) {
1637 		error = EINVAL;
1638 
1639 		if (sopt->sopt_level == SOL_SOCKET &&
1640 		    sopt->sopt_dir == SOPT_SET) {
1641 			switch (sopt->sopt_name) {
1642 			case SO_SETFIB:
1643 				INP_WLOCK(inp);
1644 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1645 				INP_WUNLOCK(inp);
1646 				error = 0;
1647 				break;
1648 			case SO_MAX_PACING_RATE:
1649 #ifdef RATELIMIT
1650 				INP_WLOCK(inp);
1651 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1652 				INP_WUNLOCK(inp);
1653 				error = 0;
1654 #else
1655 				error = EOPNOTSUPP;
1656 #endif
1657 				break;
1658 			default:
1659 				break;
1660 			}
1661 		}
1662 	} else {		/* level == IPPROTO_IPV6 */
1663 		switch (op) {
1664 		case SOPT_SET:
1665 			switch (optname) {
1666 			case IPV6_2292PKTOPTIONS:
1667 #ifdef IPV6_PKTOPTIONS
1668 			case IPV6_PKTOPTIONS:
1669 #endif
1670 			{
1671 				struct mbuf *m;
1672 
1673 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1674 					printf("ip6_ctloutput: mbuf limit hit\n");
1675 					error = ENOBUFS;
1676 					break;
1677 				}
1678 
1679 				error = soopt_getm(sopt, &m); /* XXX */
1680 				if (error != 0)
1681 					break;
1682 				error = soopt_mcopyin(sopt, m); /* XXX */
1683 				if (error != 0)
1684 					break;
1685 				INP_WLOCK(inp);
1686 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1687 				    so, sopt);
1688 				INP_WUNLOCK(inp);
1689 				m_freem(m); /* XXX */
1690 				break;
1691 			}
1692 
1693 			/*
1694 			 * Use of some Hop-by-Hop options or some
1695 			 * Destination options, might require special
1696 			 * privilege.  That is, normal applications
1697 			 * (without special privilege) might be forbidden
1698 			 * from setting certain options in outgoing packets,
1699 			 * and might never see certain options in received
1700 			 * packets. [RFC 2292 Section 6]
1701 			 * KAME specific note:
1702 			 *  KAME prevents non-privileged users from sending or
1703 			 *  receiving ANY hbh/dst options in order to avoid
1704 			 *  overhead of parsing options in the kernel.
1705 			 */
1706 			case IPV6_RECVHOPOPTS:
1707 			case IPV6_RECVDSTOPTS:
1708 			case IPV6_RECVRTHDRDSTOPTS:
1709 				if (td != NULL) {
1710 					error = priv_check(td,
1711 					    PRIV_NETINET_SETHDROPTS);
1712 					if (error)
1713 						break;
1714 				}
1715 				/* FALLTHROUGH */
1716 			case IPV6_UNICAST_HOPS:
1717 			case IPV6_HOPLIMIT:
1718 
1719 			case IPV6_RECVPKTINFO:
1720 			case IPV6_RECVHOPLIMIT:
1721 			case IPV6_RECVRTHDR:
1722 			case IPV6_RECVPATHMTU:
1723 			case IPV6_RECVTCLASS:
1724 			case IPV6_RECVFLOWID:
1725 #ifdef	RSS
1726 			case IPV6_RECVRSSBUCKETID:
1727 #endif
1728 			case IPV6_V6ONLY:
1729 			case IPV6_AUTOFLOWLABEL:
1730 			case IPV6_ORIGDSTADDR:
1731 			case IPV6_BINDANY:
1732 			case IPV6_VLAN_PCP:
1733 				if (optname == IPV6_BINDANY && td != NULL) {
1734 					error = priv_check(td,
1735 					    PRIV_NETINET_BINDANY);
1736 					if (error)
1737 						break;
1738 				}
1739 
1740 				if (optlen != sizeof(int)) {
1741 					error = EINVAL;
1742 					break;
1743 				}
1744 				error = sooptcopyin(sopt, &optval,
1745 					sizeof optval, sizeof optval);
1746 				if (error)
1747 					break;
1748 				switch (optname) {
1749 				case IPV6_UNICAST_HOPS:
1750 					if (optval < -1 || optval >= 256)
1751 						error = EINVAL;
1752 					else {
1753 						/* -1 = kernel default */
1754 						inp->in6p_hops = optval;
1755 						if ((inp->inp_vflag &
1756 						     INP_IPV4) != 0)
1757 							inp->inp_ip_ttl = optval;
1758 					}
1759 					break;
1760 #define OPTSET(bit) \
1761 do { \
1762 	INP_WLOCK(inp); \
1763 	if (optval) \
1764 		inp->inp_flags |= (bit); \
1765 	else \
1766 		inp->inp_flags &= ~(bit); \
1767 	INP_WUNLOCK(inp); \
1768 } while (/*CONSTCOND*/ 0)
1769 #define OPTSET2292(bit) \
1770 do { \
1771 	INP_WLOCK(inp); \
1772 	inp->inp_flags |= IN6P_RFC2292; \
1773 	if (optval) \
1774 		inp->inp_flags |= (bit); \
1775 	else \
1776 		inp->inp_flags &= ~(bit); \
1777 	INP_WUNLOCK(inp); \
1778 } while (/*CONSTCOND*/ 0)
1779 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1780 
1781 #define OPTSET2_N(bit, val) do {					\
1782 	if (val)							\
1783 		inp->inp_flags2 |= bit;					\
1784 	else								\
1785 		inp->inp_flags2 &= ~bit;				\
1786 } while (0)
1787 #define OPTSET2(bit, val) do {						\
1788 	INP_WLOCK(inp);							\
1789 	OPTSET2_N(bit, val);						\
1790 	INP_WUNLOCK(inp);						\
1791 } while (0)
1792 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1793 #define OPTSET2292_EXCLUSIVE(bit)					\
1794 do {									\
1795 	INP_WLOCK(inp);							\
1796 	if (OPTBIT(IN6P_RFC2292)) {					\
1797 		error = EINVAL;						\
1798 	} else {							\
1799 		if (optval)						\
1800 			inp->inp_flags |= (bit);			\
1801 		else							\
1802 			inp->inp_flags &= ~(bit);			\
1803 	}								\
1804 	INP_WUNLOCK(inp);						\
1805 } while (/*CONSTCOND*/ 0)
1806 
1807 				case IPV6_RECVPKTINFO:
1808 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1809 					break;
1810 
1811 				case IPV6_HOPLIMIT:
1812 				{
1813 					struct ip6_pktopts **optp;
1814 
1815 					/* cannot mix with RFC2292 */
1816 					if (OPTBIT(IN6P_RFC2292)) {
1817 						error = EINVAL;
1818 						break;
1819 					}
1820 					INP_WLOCK(inp);
1821 					if (inp->inp_flags & INP_DROPPED) {
1822 						INP_WUNLOCK(inp);
1823 						return (ECONNRESET);
1824 					}
1825 					optp = &inp->in6p_outputopts;
1826 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1827 					    (u_char *)&optval, sizeof(optval),
1828 					    optp, (td != NULL) ? td->td_ucred :
1829 					    NULL, uproto);
1830 					INP_WUNLOCK(inp);
1831 					break;
1832 				}
1833 
1834 				case IPV6_RECVHOPLIMIT:
1835 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1836 					break;
1837 
1838 				case IPV6_RECVHOPOPTS:
1839 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1840 					break;
1841 
1842 				case IPV6_RECVDSTOPTS:
1843 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1844 					break;
1845 
1846 				case IPV6_RECVRTHDRDSTOPTS:
1847 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1848 					break;
1849 
1850 				case IPV6_RECVRTHDR:
1851 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1852 					break;
1853 
1854 				case IPV6_RECVPATHMTU:
1855 					/*
1856 					 * We ignore this option for TCP
1857 					 * sockets.
1858 					 * (RFC3542 leaves this case
1859 					 * unspecified.)
1860 					 */
1861 					if (uproto != IPPROTO_TCP)
1862 						OPTSET(IN6P_MTU);
1863 					break;
1864 
1865 				case IPV6_RECVFLOWID:
1866 					OPTSET2(INP_RECVFLOWID, optval);
1867 					break;
1868 
1869 #ifdef	RSS
1870 				case IPV6_RECVRSSBUCKETID:
1871 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1872 					break;
1873 #endif
1874 
1875 				case IPV6_V6ONLY:
1876 					INP_WLOCK(inp);
1877 					if (inp->inp_lport ||
1878 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1879 						/*
1880 						 * The socket is already bound.
1881 						 */
1882 						INP_WUNLOCK(inp);
1883 						error = EINVAL;
1884 						break;
1885 					}
1886 					if (optval) {
1887 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1888 						inp->inp_vflag &= ~INP_IPV4;
1889 					} else {
1890 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1891 						inp->inp_vflag |= INP_IPV4;
1892 					}
1893 					INP_WUNLOCK(inp);
1894 					break;
1895 				case IPV6_RECVTCLASS:
1896 					/* cannot mix with RFC2292 XXX */
1897 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1898 					break;
1899 				case IPV6_AUTOFLOWLABEL:
1900 					OPTSET(IN6P_AUTOFLOWLABEL);
1901 					break;
1902 
1903 				case IPV6_ORIGDSTADDR:
1904 					OPTSET2(INP_ORIGDSTADDR, optval);
1905 					break;
1906 				case IPV6_BINDANY:
1907 					OPTSET(INP_BINDANY);
1908 					break;
1909 				case IPV6_VLAN_PCP:
1910 					if ((optval >= -1) && (optval <=
1911 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1912 						if (optval == -1) {
1913 							INP_WLOCK(inp);
1914 							inp->inp_flags2 &=
1915 							    ~(INP_2PCP_SET |
1916 							    INP_2PCP_MASK);
1917 							INP_WUNLOCK(inp);
1918 						} else {
1919 							INP_WLOCK(inp);
1920 							inp->inp_flags2 |=
1921 							    INP_2PCP_SET;
1922 							inp->inp_flags2 &=
1923 							    ~INP_2PCP_MASK;
1924 							inp->inp_flags2 |=
1925 							    optval <<
1926 							    INP_2PCP_SHIFT;
1927 							INP_WUNLOCK(inp);
1928 						}
1929 					} else
1930 						error = EINVAL;
1931 					break;
1932 				}
1933 				break;
1934 
1935 			case IPV6_TCLASS:
1936 			case IPV6_DONTFRAG:
1937 			case IPV6_USE_MIN_MTU:
1938 			case IPV6_PREFER_TEMPADDR:
1939 				if (optlen != sizeof(optval)) {
1940 					error = EINVAL;
1941 					break;
1942 				}
1943 				error = sooptcopyin(sopt, &optval,
1944 					sizeof optval, sizeof optval);
1945 				if (error)
1946 					break;
1947 				{
1948 					struct ip6_pktopts **optp;
1949 					INP_WLOCK(inp);
1950 					if (inp->inp_flags & INP_DROPPED) {
1951 						INP_WUNLOCK(inp);
1952 						return (ECONNRESET);
1953 					}
1954 					optp = &inp->in6p_outputopts;
1955 					error = ip6_pcbopt(optname,
1956 					    (u_char *)&optval, sizeof(optval),
1957 					    optp, (td != NULL) ? td->td_ucred :
1958 					    NULL, uproto);
1959 					INP_WUNLOCK(inp);
1960 					break;
1961 				}
1962 
1963 			case IPV6_2292PKTINFO:
1964 			case IPV6_2292HOPLIMIT:
1965 			case IPV6_2292HOPOPTS:
1966 			case IPV6_2292DSTOPTS:
1967 			case IPV6_2292RTHDR:
1968 				/* RFC 2292 */
1969 				if (optlen != sizeof(int)) {
1970 					error = EINVAL;
1971 					break;
1972 				}
1973 				error = sooptcopyin(sopt, &optval,
1974 					sizeof optval, sizeof optval);
1975 				if (error)
1976 					break;
1977 				switch (optname) {
1978 				case IPV6_2292PKTINFO:
1979 					OPTSET2292(IN6P_PKTINFO);
1980 					break;
1981 				case IPV6_2292HOPLIMIT:
1982 					OPTSET2292(IN6P_HOPLIMIT);
1983 					break;
1984 				case IPV6_2292HOPOPTS:
1985 					/*
1986 					 * Check super-user privilege.
1987 					 * See comments for IPV6_RECVHOPOPTS.
1988 					 */
1989 					if (td != NULL) {
1990 						error = priv_check(td,
1991 						    PRIV_NETINET_SETHDROPTS);
1992 						if (error)
1993 							return (error);
1994 					}
1995 					OPTSET2292(IN6P_HOPOPTS);
1996 					break;
1997 				case IPV6_2292DSTOPTS:
1998 					if (td != NULL) {
1999 						error = priv_check(td,
2000 						    PRIV_NETINET_SETHDROPTS);
2001 						if (error)
2002 							return (error);
2003 					}
2004 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2005 					break;
2006 				case IPV6_2292RTHDR:
2007 					OPTSET2292(IN6P_RTHDR);
2008 					break;
2009 				}
2010 				break;
2011 			case IPV6_PKTINFO:
2012 			case IPV6_HOPOPTS:
2013 			case IPV6_RTHDR:
2014 			case IPV6_DSTOPTS:
2015 			case IPV6_RTHDRDSTOPTS:
2016 			case IPV6_NEXTHOP:
2017 			{
2018 				/* new advanced API (RFC3542) */
2019 				u_char *optbuf;
2020 				u_char optbuf_storage[MCLBYTES];
2021 				int optlen;
2022 				struct ip6_pktopts **optp;
2023 
2024 				/* cannot mix with RFC2292 */
2025 				if (OPTBIT(IN6P_RFC2292)) {
2026 					error = EINVAL;
2027 					break;
2028 				}
2029 
2030 				/*
2031 				 * We only ensure valsize is not too large
2032 				 * here.  Further validation will be done
2033 				 * later.
2034 				 */
2035 				error = sooptcopyin(sopt, optbuf_storage,
2036 				    sizeof(optbuf_storage), 0);
2037 				if (error)
2038 					break;
2039 				optlen = sopt->sopt_valsize;
2040 				optbuf = optbuf_storage;
2041 				INP_WLOCK(inp);
2042 				if (inp->inp_flags & INP_DROPPED) {
2043 					INP_WUNLOCK(inp);
2044 					return (ECONNRESET);
2045 				}
2046 				optp = &inp->in6p_outputopts;
2047 				error = ip6_pcbopt(optname, optbuf, optlen,
2048 				    optp, (td != NULL) ? td->td_ucred : NULL,
2049 				    uproto);
2050 				INP_WUNLOCK(inp);
2051 				break;
2052 			}
2053 #undef OPTSET
2054 
2055 			case IPV6_MULTICAST_IF:
2056 			case IPV6_MULTICAST_HOPS:
2057 			case IPV6_MULTICAST_LOOP:
2058 			case IPV6_JOIN_GROUP:
2059 			case IPV6_LEAVE_GROUP:
2060 			case IPV6_MSFILTER:
2061 			case MCAST_BLOCK_SOURCE:
2062 			case MCAST_UNBLOCK_SOURCE:
2063 			case MCAST_JOIN_GROUP:
2064 			case MCAST_LEAVE_GROUP:
2065 			case MCAST_JOIN_SOURCE_GROUP:
2066 			case MCAST_LEAVE_SOURCE_GROUP:
2067 				error = ip6_setmoptions(inp, sopt);
2068 				break;
2069 
2070 			case IPV6_PORTRANGE:
2071 				error = sooptcopyin(sopt, &optval,
2072 				    sizeof optval, sizeof optval);
2073 				if (error)
2074 					break;
2075 
2076 				INP_WLOCK(inp);
2077 				switch (optval) {
2078 				case IPV6_PORTRANGE_DEFAULT:
2079 					inp->inp_flags &= ~(INP_LOWPORT);
2080 					inp->inp_flags &= ~(INP_HIGHPORT);
2081 					break;
2082 
2083 				case IPV6_PORTRANGE_HIGH:
2084 					inp->inp_flags &= ~(INP_LOWPORT);
2085 					inp->inp_flags |= INP_HIGHPORT;
2086 					break;
2087 
2088 				case IPV6_PORTRANGE_LOW:
2089 					inp->inp_flags &= ~(INP_HIGHPORT);
2090 					inp->inp_flags |= INP_LOWPORT;
2091 					break;
2092 
2093 				default:
2094 					error = EINVAL;
2095 					break;
2096 				}
2097 				INP_WUNLOCK(inp);
2098 				break;
2099 
2100 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2101 			case IPV6_IPSEC_POLICY:
2102 				if (IPSEC_ENABLED(ipv6)) {
2103 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2104 					break;
2105 				}
2106 				/* FALLTHROUGH */
2107 #endif /* IPSEC */
2108 
2109 			default:
2110 				error = ENOPROTOOPT;
2111 				break;
2112 			}
2113 			break;
2114 
2115 		case SOPT_GET:
2116 			switch (optname) {
2117 			case IPV6_2292PKTOPTIONS:
2118 #ifdef IPV6_PKTOPTIONS
2119 			case IPV6_PKTOPTIONS:
2120 #endif
2121 				/*
2122 				 * RFC3542 (effectively) deprecated the
2123 				 * semantics of the 2292-style pktoptions.
2124 				 * Since it was not reliable in nature (i.e.,
2125 				 * applications had to expect the lack of some
2126 				 * information after all), it would make sense
2127 				 * to simplify this part by always returning
2128 				 * empty data.
2129 				 */
2130 				sopt->sopt_valsize = 0;
2131 				break;
2132 
2133 			case IPV6_RECVHOPOPTS:
2134 			case IPV6_RECVDSTOPTS:
2135 			case IPV6_RECVRTHDRDSTOPTS:
2136 			case IPV6_UNICAST_HOPS:
2137 			case IPV6_RECVPKTINFO:
2138 			case IPV6_RECVHOPLIMIT:
2139 			case IPV6_RECVRTHDR:
2140 			case IPV6_RECVPATHMTU:
2141 
2142 			case IPV6_V6ONLY:
2143 			case IPV6_PORTRANGE:
2144 			case IPV6_RECVTCLASS:
2145 			case IPV6_AUTOFLOWLABEL:
2146 			case IPV6_BINDANY:
2147 			case IPV6_FLOWID:
2148 			case IPV6_FLOWTYPE:
2149 			case IPV6_RECVFLOWID:
2150 #ifdef	RSS
2151 			case IPV6_RSSBUCKETID:
2152 			case IPV6_RECVRSSBUCKETID:
2153 #endif
2154 			case IPV6_VLAN_PCP:
2155 				switch (optname) {
2156 				case IPV6_RECVHOPOPTS:
2157 					optval = OPTBIT(IN6P_HOPOPTS);
2158 					break;
2159 
2160 				case IPV6_RECVDSTOPTS:
2161 					optval = OPTBIT(IN6P_DSTOPTS);
2162 					break;
2163 
2164 				case IPV6_RECVRTHDRDSTOPTS:
2165 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2166 					break;
2167 
2168 				case IPV6_UNICAST_HOPS:
2169 					optval = inp->in6p_hops;
2170 					break;
2171 
2172 				case IPV6_RECVPKTINFO:
2173 					optval = OPTBIT(IN6P_PKTINFO);
2174 					break;
2175 
2176 				case IPV6_RECVHOPLIMIT:
2177 					optval = OPTBIT(IN6P_HOPLIMIT);
2178 					break;
2179 
2180 				case IPV6_RECVRTHDR:
2181 					optval = OPTBIT(IN6P_RTHDR);
2182 					break;
2183 
2184 				case IPV6_RECVPATHMTU:
2185 					optval = OPTBIT(IN6P_MTU);
2186 					break;
2187 
2188 				case IPV6_V6ONLY:
2189 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2190 					break;
2191 
2192 				case IPV6_PORTRANGE:
2193 				    {
2194 					int flags;
2195 					flags = inp->inp_flags;
2196 					if (flags & INP_HIGHPORT)
2197 						optval = IPV6_PORTRANGE_HIGH;
2198 					else if (flags & INP_LOWPORT)
2199 						optval = IPV6_PORTRANGE_LOW;
2200 					else
2201 						optval = 0;
2202 					break;
2203 				    }
2204 				case IPV6_RECVTCLASS:
2205 					optval = OPTBIT(IN6P_TCLASS);
2206 					break;
2207 
2208 				case IPV6_AUTOFLOWLABEL:
2209 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2210 					break;
2211 
2212 				case IPV6_ORIGDSTADDR:
2213 					optval = OPTBIT2(INP_ORIGDSTADDR);
2214 					break;
2215 
2216 				case IPV6_BINDANY:
2217 					optval = OPTBIT(INP_BINDANY);
2218 					break;
2219 
2220 				case IPV6_FLOWID:
2221 					optval = inp->inp_flowid;
2222 					break;
2223 
2224 				case IPV6_FLOWTYPE:
2225 					optval = inp->inp_flowtype;
2226 					break;
2227 
2228 				case IPV6_RECVFLOWID:
2229 					optval = OPTBIT2(INP_RECVFLOWID);
2230 					break;
2231 #ifdef	RSS
2232 				case IPV6_RSSBUCKETID:
2233 					retval =
2234 					    rss_hash2bucket(inp->inp_flowid,
2235 					    inp->inp_flowtype,
2236 					    &rss_bucket);
2237 					if (retval == 0)
2238 						optval = rss_bucket;
2239 					else
2240 						error = EINVAL;
2241 					break;
2242 
2243 				case IPV6_RECVRSSBUCKETID:
2244 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2245 					break;
2246 #endif
2247 
2248 
2249 				case IPV6_VLAN_PCP:
2250 					if (OPTBIT2(INP_2PCP_SET)) {
2251 						optval = (inp->inp_flags2 &
2252 							    INP_2PCP_MASK) >>
2253 							    INP_2PCP_SHIFT;
2254 					} else {
2255 						optval = -1;
2256 					}
2257 					break;
2258 				}
2259 
2260 				if (error)
2261 					break;
2262 				error = sooptcopyout(sopt, &optval,
2263 					sizeof optval);
2264 				break;
2265 
2266 			case IPV6_PATHMTU:
2267 			{
2268 				u_long pmtu = 0;
2269 				struct ip6_mtuinfo mtuinfo;
2270 				struct in6_addr addr;
2271 
2272 				if (!(so->so_state & SS_ISCONNECTED))
2273 					return (ENOTCONN);
2274 				/*
2275 				 * XXX: we dot not consider the case of source
2276 				 * routing, or optional information to specify
2277 				 * the outgoing interface.
2278 				 * Copy faddr out of inp to avoid holding lock
2279 				 * on inp during route lookup.
2280 				 */
2281 				INP_RLOCK(inp);
2282 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2283 				INP_RUNLOCK(inp);
2284 				error = ip6_getpmtu_ctl(so->so_fibnum,
2285 				    &addr, &pmtu);
2286 				if (error)
2287 					break;
2288 				if (pmtu > IPV6_MAXPACKET)
2289 					pmtu = IPV6_MAXPACKET;
2290 
2291 				bzero(&mtuinfo, sizeof(mtuinfo));
2292 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2293 				optdata = (void *)&mtuinfo;
2294 				optdatalen = sizeof(mtuinfo);
2295 				error = sooptcopyout(sopt, optdata,
2296 				    optdatalen);
2297 				break;
2298 			}
2299 
2300 			case IPV6_2292PKTINFO:
2301 			case IPV6_2292HOPLIMIT:
2302 			case IPV6_2292HOPOPTS:
2303 			case IPV6_2292RTHDR:
2304 			case IPV6_2292DSTOPTS:
2305 				switch (optname) {
2306 				case IPV6_2292PKTINFO:
2307 					optval = OPTBIT(IN6P_PKTINFO);
2308 					break;
2309 				case IPV6_2292HOPLIMIT:
2310 					optval = OPTBIT(IN6P_HOPLIMIT);
2311 					break;
2312 				case IPV6_2292HOPOPTS:
2313 					optval = OPTBIT(IN6P_HOPOPTS);
2314 					break;
2315 				case IPV6_2292RTHDR:
2316 					optval = OPTBIT(IN6P_RTHDR);
2317 					break;
2318 				case IPV6_2292DSTOPTS:
2319 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2320 					break;
2321 				}
2322 				error = sooptcopyout(sopt, &optval,
2323 				    sizeof optval);
2324 				break;
2325 			case IPV6_PKTINFO:
2326 			case IPV6_HOPOPTS:
2327 			case IPV6_RTHDR:
2328 			case IPV6_DSTOPTS:
2329 			case IPV6_RTHDRDSTOPTS:
2330 			case IPV6_NEXTHOP:
2331 			case IPV6_TCLASS:
2332 			case IPV6_DONTFRAG:
2333 			case IPV6_USE_MIN_MTU:
2334 			case IPV6_PREFER_TEMPADDR:
2335 				error = ip6_getpcbopt(inp, optname, sopt);
2336 				break;
2337 
2338 			case IPV6_MULTICAST_IF:
2339 			case IPV6_MULTICAST_HOPS:
2340 			case IPV6_MULTICAST_LOOP:
2341 			case IPV6_MSFILTER:
2342 				error = ip6_getmoptions(inp, sopt);
2343 				break;
2344 
2345 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2346 			case IPV6_IPSEC_POLICY:
2347 				if (IPSEC_ENABLED(ipv6)) {
2348 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2349 					break;
2350 				}
2351 				/* FALLTHROUGH */
2352 #endif /* IPSEC */
2353 			default:
2354 				error = ENOPROTOOPT;
2355 				break;
2356 			}
2357 			break;
2358 		}
2359 	}
2360 	return (error);
2361 }
2362 
2363 int
2364 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2365 {
2366 	int error = 0, optval, optlen;
2367 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2368 	struct inpcb *inp = sotoinpcb(so);
2369 	int level, op, optname;
2370 
2371 	level = sopt->sopt_level;
2372 	op = sopt->sopt_dir;
2373 	optname = sopt->sopt_name;
2374 	optlen = sopt->sopt_valsize;
2375 
2376 	if (level != IPPROTO_IPV6) {
2377 		return (EINVAL);
2378 	}
2379 
2380 	switch (optname) {
2381 	case IPV6_CHECKSUM:
2382 		/*
2383 		 * For ICMPv6 sockets, no modification allowed for checksum
2384 		 * offset, permit "no change" values to help existing apps.
2385 		 *
2386 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2387 		 * for an ICMPv6 socket will fail."
2388 		 * The current behavior does not meet RFC3542.
2389 		 */
2390 		switch (op) {
2391 		case SOPT_SET:
2392 			if (optlen != sizeof(int)) {
2393 				error = EINVAL;
2394 				break;
2395 			}
2396 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2397 					    sizeof(optval));
2398 			if (error)
2399 				break;
2400 			if (optval < -1 || (optval % 2) != 0) {
2401 				/*
2402 				 * The API assumes non-negative even offset
2403 				 * values or -1 as a special value.
2404 				 */
2405 				error = EINVAL;
2406 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2407 				if (optval != icmp6off)
2408 					error = EINVAL;
2409 			} else
2410 				inp->in6p_cksum = optval;
2411 			break;
2412 
2413 		case SOPT_GET:
2414 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2415 				optval = icmp6off;
2416 			else
2417 				optval = inp->in6p_cksum;
2418 
2419 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2420 			break;
2421 
2422 		default:
2423 			error = EINVAL;
2424 			break;
2425 		}
2426 		break;
2427 
2428 	default:
2429 		error = ENOPROTOOPT;
2430 		break;
2431 	}
2432 
2433 	return (error);
2434 }
2435 
2436 /*
2437  * Set up IP6 options in pcb for insertion in output packets or
2438  * specifying behavior of outgoing packets.
2439  */
2440 static int
2441 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2442     struct socket *so, struct sockopt *sopt)
2443 {
2444 	struct ip6_pktopts *opt = *pktopt;
2445 	int error = 0;
2446 	struct thread *td = sopt->sopt_td;
2447 	struct epoch_tracker et;
2448 
2449 	/* turn off any old options. */
2450 	if (opt) {
2451 #ifdef DIAGNOSTIC
2452 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2453 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2454 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2455 			printf("ip6_pcbopts: all specified options are cleared.\n");
2456 #endif
2457 		ip6_clearpktopts(opt, -1);
2458 	} else {
2459 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2460 		if (opt == NULL)
2461 			return (ENOMEM);
2462 	}
2463 	*pktopt = NULL;
2464 
2465 	if (!m || m->m_len == 0) {
2466 		/*
2467 		 * Only turning off any previous options, regardless of
2468 		 * whether the opt is just created or given.
2469 		 */
2470 		free(opt, M_IP6OPT);
2471 		return (0);
2472 	}
2473 
2474 	/*  set options specified by user. */
2475 	NET_EPOCH_ENTER(et);
2476 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2477 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2478 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2479 		free(opt, M_IP6OPT);
2480 		NET_EPOCH_EXIT(et);
2481 		return (error);
2482 	}
2483 	NET_EPOCH_EXIT(et);
2484 	*pktopt = opt;
2485 	return (0);
2486 }
2487 
2488 /*
2489  * initialize ip6_pktopts.  beware that there are non-zero default values in
2490  * the struct.
2491  */
2492 void
2493 ip6_initpktopts(struct ip6_pktopts *opt)
2494 {
2495 
2496 	bzero(opt, sizeof(*opt));
2497 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2498 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2499 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2500 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2501 }
2502 
2503 static int
2504 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2505     struct ucred *cred, int uproto)
2506 {
2507 	struct epoch_tracker et;
2508 	struct ip6_pktopts *opt;
2509 	int ret;
2510 
2511 	if (*pktopt == NULL) {
2512 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2513 		    M_NOWAIT);
2514 		if (*pktopt == NULL)
2515 			return (ENOBUFS);
2516 		ip6_initpktopts(*pktopt);
2517 	}
2518 	opt = *pktopt;
2519 
2520 	NET_EPOCH_ENTER(et);
2521 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2522 	NET_EPOCH_EXIT(et);
2523 
2524 	return (ret);
2525 }
2526 
2527 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2528 	if (pktopt && pktopt->field) {					\
2529 		INP_RUNLOCK(inp);					\
2530 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2531 		malloc_optdata = true;					\
2532 		INP_RLOCK(inp);						\
2533 		if (inp->inp_flags & INP_DROPPED) {			\
2534 			INP_RUNLOCK(inp);				\
2535 			free(optdata, M_TEMP);				\
2536 			return (ECONNRESET);				\
2537 		}							\
2538 		pktopt = inp->in6p_outputopts;				\
2539 		if (pktopt && pktopt->field) {				\
2540 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2541 			bcopy(pktopt->field, optdata, optdatalen);	\
2542 		} else {						\
2543 			free(optdata, M_TEMP);				\
2544 			optdata = NULL;					\
2545 			malloc_optdata = false;				\
2546 		}							\
2547 	}								\
2548 } while(0)
2549 
2550 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2551 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2552 
2553 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2554 	pktopt->field->sa_len)
2555 
2556 static int
2557 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2558 {
2559 	void *optdata = NULL;
2560 	bool malloc_optdata = false;
2561 	int optdatalen = 0;
2562 	int error = 0;
2563 	struct in6_pktinfo null_pktinfo;
2564 	int deftclass = 0, on;
2565 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2566 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2567 	struct ip6_pktopts *pktopt;
2568 
2569 	INP_RLOCK(inp);
2570 	pktopt = inp->in6p_outputopts;
2571 
2572 	switch (optname) {
2573 	case IPV6_PKTINFO:
2574 		optdata = (void *)&null_pktinfo;
2575 		if (pktopt && pktopt->ip6po_pktinfo) {
2576 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2577 			    sizeof(null_pktinfo));
2578 			in6_clearscope(&null_pktinfo.ipi6_addr);
2579 		} else {
2580 			/* XXX: we don't have to do this every time... */
2581 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2582 		}
2583 		optdatalen = sizeof(struct in6_pktinfo);
2584 		break;
2585 	case IPV6_TCLASS:
2586 		if (pktopt && pktopt->ip6po_tclass >= 0)
2587 			deftclass = pktopt->ip6po_tclass;
2588 		optdata = (void *)&deftclass;
2589 		optdatalen = sizeof(int);
2590 		break;
2591 	case IPV6_HOPOPTS:
2592 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2593 		break;
2594 	case IPV6_RTHDR:
2595 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2596 		break;
2597 	case IPV6_RTHDRDSTOPTS:
2598 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2599 		break;
2600 	case IPV6_DSTOPTS:
2601 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2602 		break;
2603 	case IPV6_NEXTHOP:
2604 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2605 		break;
2606 	case IPV6_USE_MIN_MTU:
2607 		if (pktopt)
2608 			defminmtu = pktopt->ip6po_minmtu;
2609 		optdata = (void *)&defminmtu;
2610 		optdatalen = sizeof(int);
2611 		break;
2612 	case IPV6_DONTFRAG:
2613 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2614 			on = 1;
2615 		else
2616 			on = 0;
2617 		optdata = (void *)&on;
2618 		optdatalen = sizeof(on);
2619 		break;
2620 	case IPV6_PREFER_TEMPADDR:
2621 		if (pktopt)
2622 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2623 		optdata = (void *)&defpreftemp;
2624 		optdatalen = sizeof(int);
2625 		break;
2626 	default:		/* should not happen */
2627 #ifdef DIAGNOSTIC
2628 		panic("ip6_getpcbopt: unexpected option\n");
2629 #endif
2630 		INP_RUNLOCK(inp);
2631 		return (ENOPROTOOPT);
2632 	}
2633 	INP_RUNLOCK(inp);
2634 
2635 	error = sooptcopyout(sopt, optdata, optdatalen);
2636 	if (malloc_optdata)
2637 		free(optdata, M_TEMP);
2638 
2639 	return (error);
2640 }
2641 
2642 void
2643 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2644 {
2645 	if (pktopt == NULL)
2646 		return;
2647 
2648 	if (optname == -1 || optname == IPV6_PKTINFO) {
2649 		if (pktopt->ip6po_pktinfo)
2650 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2651 		pktopt->ip6po_pktinfo = NULL;
2652 	}
2653 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2654 		pktopt->ip6po_hlim = -1;
2655 	if (optname == -1 || optname == IPV6_TCLASS)
2656 		pktopt->ip6po_tclass = -1;
2657 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2658 		if (pktopt->ip6po_nextroute.ro_nh) {
2659 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2660 			pktopt->ip6po_nextroute.ro_nh = NULL;
2661 		}
2662 		if (pktopt->ip6po_nexthop)
2663 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2664 		pktopt->ip6po_nexthop = NULL;
2665 	}
2666 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2667 		if (pktopt->ip6po_hbh)
2668 			free(pktopt->ip6po_hbh, M_IP6OPT);
2669 		pktopt->ip6po_hbh = NULL;
2670 	}
2671 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2672 		if (pktopt->ip6po_dest1)
2673 			free(pktopt->ip6po_dest1, M_IP6OPT);
2674 		pktopt->ip6po_dest1 = NULL;
2675 	}
2676 	if (optname == -1 || optname == IPV6_RTHDR) {
2677 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2678 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2679 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2680 		if (pktopt->ip6po_route.ro_nh) {
2681 			NH_FREE(pktopt->ip6po_route.ro_nh);
2682 			pktopt->ip6po_route.ro_nh = NULL;
2683 		}
2684 	}
2685 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2686 		if (pktopt->ip6po_dest2)
2687 			free(pktopt->ip6po_dest2, M_IP6OPT);
2688 		pktopt->ip6po_dest2 = NULL;
2689 	}
2690 }
2691 
2692 #define PKTOPT_EXTHDRCPY(type) \
2693 do {\
2694 	if (src->type) {\
2695 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2696 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2697 		if (dst->type == NULL)\
2698 			goto bad;\
2699 		bcopy(src->type, dst->type, hlen);\
2700 	}\
2701 } while (/*CONSTCOND*/ 0)
2702 
2703 static int
2704 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2705 {
2706 	if (dst == NULL || src == NULL)  {
2707 		printf("ip6_clearpktopts: invalid argument\n");
2708 		return (EINVAL);
2709 	}
2710 
2711 	dst->ip6po_hlim = src->ip6po_hlim;
2712 	dst->ip6po_tclass = src->ip6po_tclass;
2713 	dst->ip6po_flags = src->ip6po_flags;
2714 	dst->ip6po_minmtu = src->ip6po_minmtu;
2715 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2716 	if (src->ip6po_pktinfo) {
2717 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2718 		    M_IP6OPT, canwait);
2719 		if (dst->ip6po_pktinfo == NULL)
2720 			goto bad;
2721 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2722 	}
2723 	if (src->ip6po_nexthop) {
2724 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2725 		    M_IP6OPT, canwait);
2726 		if (dst->ip6po_nexthop == NULL)
2727 			goto bad;
2728 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2729 		    src->ip6po_nexthop->sa_len);
2730 	}
2731 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2732 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2733 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2734 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2735 	return (0);
2736 
2737   bad:
2738 	ip6_clearpktopts(dst, -1);
2739 	return (ENOBUFS);
2740 }
2741 #undef PKTOPT_EXTHDRCPY
2742 
2743 struct ip6_pktopts *
2744 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2745 {
2746 	int error;
2747 	struct ip6_pktopts *dst;
2748 
2749 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2750 	if (dst == NULL)
2751 		return (NULL);
2752 	ip6_initpktopts(dst);
2753 
2754 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2755 		free(dst, M_IP6OPT);
2756 		return (NULL);
2757 	}
2758 
2759 	return (dst);
2760 }
2761 
2762 void
2763 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2764 {
2765 	if (pktopt == NULL)
2766 		return;
2767 
2768 	ip6_clearpktopts(pktopt, -1);
2769 
2770 	free(pktopt, M_IP6OPT);
2771 }
2772 
2773 /*
2774  * Set IPv6 outgoing packet options based on advanced API.
2775  */
2776 int
2777 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2778     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2779 {
2780 	struct cmsghdr *cm = NULL;
2781 
2782 	if (control == NULL || opt == NULL)
2783 		return (EINVAL);
2784 
2785 	/*
2786 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2787 	 * are in the network epoch here.
2788 	 */
2789 	NET_EPOCH_ASSERT();
2790 
2791 	ip6_initpktopts(opt);
2792 	if (stickyopt) {
2793 		int error;
2794 
2795 		/*
2796 		 * If stickyopt is provided, make a local copy of the options
2797 		 * for this particular packet, then override them by ancillary
2798 		 * objects.
2799 		 * XXX: copypktopts() does not copy the cached route to a next
2800 		 * hop (if any).  This is not very good in terms of efficiency,
2801 		 * but we can allow this since this option should be rarely
2802 		 * used.
2803 		 */
2804 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2805 			return (error);
2806 	}
2807 
2808 	/*
2809 	 * XXX: Currently, we assume all the optional information is stored
2810 	 * in a single mbuf.
2811 	 */
2812 	if (control->m_next)
2813 		return (EINVAL);
2814 
2815 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2816 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2817 		int error;
2818 
2819 		if (control->m_len < CMSG_LEN(0))
2820 			return (EINVAL);
2821 
2822 		cm = mtod(control, struct cmsghdr *);
2823 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2824 			return (EINVAL);
2825 		if (cm->cmsg_level != IPPROTO_IPV6)
2826 			continue;
2827 
2828 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2829 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2830 		if (error)
2831 			return (error);
2832 	}
2833 
2834 	return (0);
2835 }
2836 
2837 /*
2838  * Set a particular packet option, as a sticky option or an ancillary data
2839  * item.  "len" can be 0 only when it's a sticky option.
2840  * We have 4 cases of combination of "sticky" and "cmsg":
2841  * "sticky=0, cmsg=0": impossible
2842  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2843  * "sticky=1, cmsg=0": RFC3542 socket option
2844  * "sticky=1, cmsg=1": RFC2292 socket option
2845  */
2846 static int
2847 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2848     struct ucred *cred, int sticky, int cmsg, int uproto)
2849 {
2850 	int minmtupolicy, preftemp;
2851 	int error;
2852 
2853 	NET_EPOCH_ASSERT();
2854 
2855 	if (!sticky && !cmsg) {
2856 #ifdef DIAGNOSTIC
2857 		printf("ip6_setpktopt: impossible case\n");
2858 #endif
2859 		return (EINVAL);
2860 	}
2861 
2862 	/*
2863 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2864 	 * not be specified in the context of RFC3542.  Conversely,
2865 	 * RFC3542 types should not be specified in the context of RFC2292.
2866 	 */
2867 	if (!cmsg) {
2868 		switch (optname) {
2869 		case IPV6_2292PKTINFO:
2870 		case IPV6_2292HOPLIMIT:
2871 		case IPV6_2292NEXTHOP:
2872 		case IPV6_2292HOPOPTS:
2873 		case IPV6_2292DSTOPTS:
2874 		case IPV6_2292RTHDR:
2875 		case IPV6_2292PKTOPTIONS:
2876 			return (ENOPROTOOPT);
2877 		}
2878 	}
2879 	if (sticky && cmsg) {
2880 		switch (optname) {
2881 		case IPV6_PKTINFO:
2882 		case IPV6_HOPLIMIT:
2883 		case IPV6_NEXTHOP:
2884 		case IPV6_HOPOPTS:
2885 		case IPV6_DSTOPTS:
2886 		case IPV6_RTHDRDSTOPTS:
2887 		case IPV6_RTHDR:
2888 		case IPV6_USE_MIN_MTU:
2889 		case IPV6_DONTFRAG:
2890 		case IPV6_TCLASS:
2891 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2892 			return (ENOPROTOOPT);
2893 		}
2894 	}
2895 
2896 	switch (optname) {
2897 	case IPV6_2292PKTINFO:
2898 	case IPV6_PKTINFO:
2899 	{
2900 		struct ifnet *ifp = NULL;
2901 		struct in6_pktinfo *pktinfo;
2902 
2903 		if (len != sizeof(struct in6_pktinfo))
2904 			return (EINVAL);
2905 
2906 		pktinfo = (struct in6_pktinfo *)buf;
2907 
2908 		/*
2909 		 * An application can clear any sticky IPV6_PKTINFO option by
2910 		 * doing a "regular" setsockopt with ipi6_addr being
2911 		 * in6addr_any and ipi6_ifindex being zero.
2912 		 * [RFC 3542, Section 6]
2913 		 */
2914 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2915 		    pktinfo->ipi6_ifindex == 0 &&
2916 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2917 			ip6_clearpktopts(opt, optname);
2918 			break;
2919 		}
2920 
2921 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2922 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2923 			return (EINVAL);
2924 		}
2925 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2926 			return (EINVAL);
2927 		/* validate the interface index if specified. */
2928 		if (pktinfo->ipi6_ifindex) {
2929 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2930 			if (ifp == NULL)
2931 				return (ENXIO);
2932 		}
2933 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2934 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2935 			return (ENETDOWN);
2936 
2937 		if (ifp != NULL &&
2938 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2939 			struct in6_ifaddr *ia;
2940 
2941 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2942 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2943 			if (ia == NULL)
2944 				return (EADDRNOTAVAIL);
2945 			ifa_free(&ia->ia_ifa);
2946 		}
2947 		/*
2948 		 * We store the address anyway, and let in6_selectsrc()
2949 		 * validate the specified address.  This is because ipi6_addr
2950 		 * may not have enough information about its scope zone, and
2951 		 * we may need additional information (such as outgoing
2952 		 * interface or the scope zone of a destination address) to
2953 		 * disambiguate the scope.
2954 		 * XXX: the delay of the validation may confuse the
2955 		 * application when it is used as a sticky option.
2956 		 */
2957 		if (opt->ip6po_pktinfo == NULL) {
2958 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2959 			    M_IP6OPT, M_NOWAIT);
2960 			if (opt->ip6po_pktinfo == NULL)
2961 				return (ENOBUFS);
2962 		}
2963 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2964 		break;
2965 	}
2966 
2967 	case IPV6_2292HOPLIMIT:
2968 	case IPV6_HOPLIMIT:
2969 	{
2970 		int *hlimp;
2971 
2972 		/*
2973 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2974 		 * to simplify the ordering among hoplimit options.
2975 		 */
2976 		if (optname == IPV6_HOPLIMIT && sticky)
2977 			return (ENOPROTOOPT);
2978 
2979 		if (len != sizeof(int))
2980 			return (EINVAL);
2981 		hlimp = (int *)buf;
2982 		if (*hlimp < -1 || *hlimp > 255)
2983 			return (EINVAL);
2984 
2985 		opt->ip6po_hlim = *hlimp;
2986 		break;
2987 	}
2988 
2989 	case IPV6_TCLASS:
2990 	{
2991 		int tclass;
2992 
2993 		if (len != sizeof(int))
2994 			return (EINVAL);
2995 		tclass = *(int *)buf;
2996 		if (tclass < -1 || tclass > 255)
2997 			return (EINVAL);
2998 
2999 		opt->ip6po_tclass = tclass;
3000 		break;
3001 	}
3002 
3003 	case IPV6_2292NEXTHOP:
3004 	case IPV6_NEXTHOP:
3005 		if (cred != NULL) {
3006 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3007 			if (error)
3008 				return (error);
3009 		}
3010 
3011 		if (len == 0) {	/* just remove the option */
3012 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3013 			break;
3014 		}
3015 
3016 		/* check if cmsg_len is large enough for sa_len */
3017 		if (len < sizeof(struct sockaddr) || len < *buf)
3018 			return (EINVAL);
3019 
3020 		switch (((struct sockaddr *)buf)->sa_family) {
3021 		case AF_INET6:
3022 		{
3023 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3024 			int error;
3025 
3026 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3027 				return (EINVAL);
3028 
3029 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3030 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3031 				return (EINVAL);
3032 			}
3033 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3034 			    != 0) {
3035 				return (error);
3036 			}
3037 			break;
3038 		}
3039 		case AF_LINK:	/* should eventually be supported */
3040 		default:
3041 			return (EAFNOSUPPORT);
3042 		}
3043 
3044 		/* turn off the previous option, then set the new option. */
3045 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3046 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3047 		if (opt->ip6po_nexthop == NULL)
3048 			return (ENOBUFS);
3049 		bcopy(buf, opt->ip6po_nexthop, *buf);
3050 		break;
3051 
3052 	case IPV6_2292HOPOPTS:
3053 	case IPV6_HOPOPTS:
3054 	{
3055 		struct ip6_hbh *hbh;
3056 		int hbhlen;
3057 
3058 		/*
3059 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3060 		 * options, since per-option restriction has too much
3061 		 * overhead.
3062 		 */
3063 		if (cred != NULL) {
3064 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3065 			if (error)
3066 				return (error);
3067 		}
3068 
3069 		if (len == 0) {
3070 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3071 			break;	/* just remove the option */
3072 		}
3073 
3074 		/* message length validation */
3075 		if (len < sizeof(struct ip6_hbh))
3076 			return (EINVAL);
3077 		hbh = (struct ip6_hbh *)buf;
3078 		hbhlen = (hbh->ip6h_len + 1) << 3;
3079 		if (len != hbhlen)
3080 			return (EINVAL);
3081 
3082 		/* turn off the previous option, then set the new option. */
3083 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3084 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3085 		if (opt->ip6po_hbh == NULL)
3086 			return (ENOBUFS);
3087 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3088 
3089 		break;
3090 	}
3091 
3092 	case IPV6_2292DSTOPTS:
3093 	case IPV6_DSTOPTS:
3094 	case IPV6_RTHDRDSTOPTS:
3095 	{
3096 		struct ip6_dest *dest, **newdest = NULL;
3097 		int destlen;
3098 
3099 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3100 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3101 			if (error)
3102 				return (error);
3103 		}
3104 
3105 		if (len == 0) {
3106 			ip6_clearpktopts(opt, optname);
3107 			break;	/* just remove the option */
3108 		}
3109 
3110 		/* message length validation */
3111 		if (len < sizeof(struct ip6_dest))
3112 			return (EINVAL);
3113 		dest = (struct ip6_dest *)buf;
3114 		destlen = (dest->ip6d_len + 1) << 3;
3115 		if (len != destlen)
3116 			return (EINVAL);
3117 
3118 		/*
3119 		 * Determine the position that the destination options header
3120 		 * should be inserted; before or after the routing header.
3121 		 */
3122 		switch (optname) {
3123 		case IPV6_2292DSTOPTS:
3124 			/*
3125 			 * The old advacned API is ambiguous on this point.
3126 			 * Our approach is to determine the position based
3127 			 * according to the existence of a routing header.
3128 			 * Note, however, that this depends on the order of the
3129 			 * extension headers in the ancillary data; the 1st
3130 			 * part of the destination options header must appear
3131 			 * before the routing header in the ancillary data,
3132 			 * too.
3133 			 * RFC3542 solved the ambiguity by introducing
3134 			 * separate ancillary data or option types.
3135 			 */
3136 			if (opt->ip6po_rthdr == NULL)
3137 				newdest = &opt->ip6po_dest1;
3138 			else
3139 				newdest = &opt->ip6po_dest2;
3140 			break;
3141 		case IPV6_RTHDRDSTOPTS:
3142 			newdest = &opt->ip6po_dest1;
3143 			break;
3144 		case IPV6_DSTOPTS:
3145 			newdest = &opt->ip6po_dest2;
3146 			break;
3147 		}
3148 
3149 		/* turn off the previous option, then set the new option. */
3150 		ip6_clearpktopts(opt, optname);
3151 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3152 		if (*newdest == NULL)
3153 			return (ENOBUFS);
3154 		bcopy(dest, *newdest, destlen);
3155 
3156 		break;
3157 	}
3158 
3159 	case IPV6_2292RTHDR:
3160 	case IPV6_RTHDR:
3161 	{
3162 		struct ip6_rthdr *rth;
3163 		int rthlen;
3164 
3165 		if (len == 0) {
3166 			ip6_clearpktopts(opt, IPV6_RTHDR);
3167 			break;	/* just remove the option */
3168 		}
3169 
3170 		/* message length validation */
3171 		if (len < sizeof(struct ip6_rthdr))
3172 			return (EINVAL);
3173 		rth = (struct ip6_rthdr *)buf;
3174 		rthlen = (rth->ip6r_len + 1) << 3;
3175 		if (len != rthlen)
3176 			return (EINVAL);
3177 
3178 		switch (rth->ip6r_type) {
3179 		case IPV6_RTHDR_TYPE_0:
3180 			if (rth->ip6r_len == 0)	/* must contain one addr */
3181 				return (EINVAL);
3182 			if (rth->ip6r_len % 2) /* length must be even */
3183 				return (EINVAL);
3184 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3185 				return (EINVAL);
3186 			break;
3187 		default:
3188 			return (EINVAL);	/* not supported */
3189 		}
3190 
3191 		/* turn off the previous option */
3192 		ip6_clearpktopts(opt, IPV6_RTHDR);
3193 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3194 		if (opt->ip6po_rthdr == NULL)
3195 			return (ENOBUFS);
3196 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3197 
3198 		break;
3199 	}
3200 
3201 	case IPV6_USE_MIN_MTU:
3202 		if (len != sizeof(int))
3203 			return (EINVAL);
3204 		minmtupolicy = *(int *)buf;
3205 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3206 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3207 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3208 			return (EINVAL);
3209 		}
3210 		opt->ip6po_minmtu = minmtupolicy;
3211 		break;
3212 
3213 	case IPV6_DONTFRAG:
3214 		if (len != sizeof(int))
3215 			return (EINVAL);
3216 
3217 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3218 			/*
3219 			 * we ignore this option for TCP sockets.
3220 			 * (RFC3542 leaves this case unspecified.)
3221 			 */
3222 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3223 		} else
3224 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3225 		break;
3226 
3227 	case IPV6_PREFER_TEMPADDR:
3228 		if (len != sizeof(int))
3229 			return (EINVAL);
3230 		preftemp = *(int *)buf;
3231 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3232 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3233 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3234 			return (EINVAL);
3235 		}
3236 		opt->ip6po_prefer_tempaddr = preftemp;
3237 		break;
3238 
3239 	default:
3240 		return (ENOPROTOOPT);
3241 	} /* end of switch */
3242 
3243 	return (0);
3244 }
3245 
3246 /*
3247  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3248  * packet to the input queue of a specified interface.  Note that this
3249  * calls the output routine of the loopback "driver", but with an interface
3250  * pointer that might NOT be &loif -- easier than replicating that code here.
3251  */
3252 void
3253 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3254 {
3255 	struct mbuf *copym;
3256 	struct ip6_hdr *ip6;
3257 
3258 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3259 	if (copym == NULL)
3260 		return;
3261 
3262 	/*
3263 	 * Make sure to deep-copy IPv6 header portion in case the data
3264 	 * is in an mbuf cluster, so that we can safely override the IPv6
3265 	 * header portion later.
3266 	 */
3267 	if (!M_WRITABLE(copym) ||
3268 	    copym->m_len < sizeof(struct ip6_hdr)) {
3269 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3270 		if (copym == NULL)
3271 			return;
3272 	}
3273 	ip6 = mtod(copym, struct ip6_hdr *);
3274 	/*
3275 	 * clear embedded scope identifiers if necessary.
3276 	 * in6_clearscope will touch the addresses only when necessary.
3277 	 */
3278 	in6_clearscope(&ip6->ip6_src);
3279 	in6_clearscope(&ip6->ip6_dst);
3280 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3281 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3282 		    CSUM_PSEUDO_HDR;
3283 		copym->m_pkthdr.csum_data = 0xffff;
3284 	}
3285 	if_simloop(ifp, copym, AF_INET6, 0);
3286 }
3287 
3288 /*
3289  * Chop IPv6 header off from the payload.
3290  */
3291 static int
3292 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3293 {
3294 	struct mbuf *mh;
3295 	struct ip6_hdr *ip6;
3296 
3297 	ip6 = mtod(m, struct ip6_hdr *);
3298 	if (m->m_len > sizeof(*ip6)) {
3299 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3300 		if (mh == NULL) {
3301 			m_freem(m);
3302 			return ENOBUFS;
3303 		}
3304 		m_move_pkthdr(mh, m);
3305 		M_ALIGN(mh, sizeof(*ip6));
3306 		m->m_len -= sizeof(*ip6);
3307 		m->m_data += sizeof(*ip6);
3308 		mh->m_next = m;
3309 		m = mh;
3310 		m->m_len = sizeof(*ip6);
3311 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3312 	}
3313 	exthdrs->ip6e_ip6 = m;
3314 	return 0;
3315 }
3316 
3317 /*
3318  * Compute IPv6 extension header length.
3319  */
3320 int
3321 ip6_optlen(struct inpcb *inp)
3322 {
3323 	int len;
3324 
3325 	if (!inp->in6p_outputopts)
3326 		return 0;
3327 
3328 	len = 0;
3329 #define elen(x) \
3330     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3331 
3332 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3333 	if (inp->in6p_outputopts->ip6po_rthdr)
3334 		/* dest1 is valid with rthdr only */
3335 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3336 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3337 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3338 	return len;
3339 #undef elen
3340 }
3341