xref: /freebsd/sys/netinet6/ip6_output.c (revision d4eeb02986980bf33dd56c41ceb9fc5f180c0d47)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_kern_tls.h"
72 #include "opt_ratelimit.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 #include "opt_sctp.h"
76 
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/ktls.h>
80 #include <sys/malloc.h>
81 #include <sys/mbuf.h>
82 #include <sys/errno.h>
83 #include <sys/priv.h>
84 #include <sys/proc.h>
85 #include <sys/protosw.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/syslog.h>
89 #include <sys/ucred.h>
90 
91 #include <machine/in_cksum.h>
92 
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_vlan_var.h>
96 #include <net/if_llatbl.h>
97 #include <net/ethernet.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/route/nhop.h>
101 #include <net/pfil.h>
102 #include <net/rss_config.h>
103 #include <net/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip_var.h>
108 #include <netinet6/in6_fib.h>
109 #include <netinet6/in6_var.h>
110 #include <netinet/ip6.h>
111 #include <netinet/icmp6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/tcp_var.h>
115 #include <netinet6/nd6.h>
116 #include <netinet6/in6_rss.h>
117 
118 #include <netipsec/ipsec_support.h>
119 #if defined(SCTP) || defined(SCTP_SUPPORT)
120 #include <netinet/sctp.h>
121 #include <netinet/sctp_crc32.h>
122 #endif
123 
124 #include <netinet6/ip6protosw.h>
125 #include <netinet6/scope6_var.h>
126 
127 extern int in6_mcast_loop;
128 
129 struct ip6_exthdrs {
130 	struct mbuf *ip6e_ip6;
131 	struct mbuf *ip6e_hbh;
132 	struct mbuf *ip6e_dest1;
133 	struct mbuf *ip6e_rthdr;
134 	struct mbuf *ip6e_dest2;
135 };
136 
137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
138 
139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
140 			   struct ucred *, int);
141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
142 	struct socket *, struct sockopt *);
143 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
145 	struct ucred *, int, int, int);
146 
147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
149 	struct ip6_frag **);
150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
152 static int ip6_getpmtu(struct route_in6 *, int,
153 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
154 	u_int);
155 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
156 	u_long *, int *, u_int);
157 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
158 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
159 
160 /*
161  * Make an extension header from option data.  hp is the source,
162  * mp is the destination, and _ol is the optlen.
163  */
164 #define	MAKE_EXTHDR(hp, mp, _ol)					\
165     do {								\
166 	if (hp) {							\
167 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
168 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
169 		    ((eh)->ip6e_len + 1) << 3);				\
170 		if (error)						\
171 			goto freehdrs;					\
172 		(_ol) += (*(mp))->m_len;				\
173 	}								\
174     } while (/*CONSTCOND*/ 0)
175 
176 /*
177  * Form a chain of extension headers.
178  * m is the extension header mbuf
179  * mp is the previous mbuf in the chain
180  * p is the next header
181  * i is the type of option.
182  */
183 #define MAKE_CHAIN(m, mp, p, i)\
184     do {\
185 	if (m) {\
186 		if (!hdrsplit) \
187 			panic("%s:%d: assumption failed: "\
188 			    "hdr not split: hdrsplit %d exthdrs %p",\
189 			    __func__, __LINE__, hdrsplit, &exthdrs);\
190 		*mtod((m), u_char *) = *(p);\
191 		*(p) = (i);\
192 		p = mtod((m), u_char *);\
193 		(m)->m_next = (mp)->m_next;\
194 		(mp)->m_next = (m);\
195 		(mp) = (m);\
196 	}\
197     } while (/*CONSTCOND*/ 0)
198 
199 void
200 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
201 {
202 	u_short csum;
203 
204 	csum = in_cksum_skip(m, offset + plen, offset);
205 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
206 		csum = 0xffff;
207 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
208 
209 	if (offset + sizeof(csum) > m->m_len)
210 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
211 	else
212 		*(u_short *)mtodo(m, offset) = csum;
213 }
214 
215 static void
216 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
217     int plen, int optlen)
218 {
219 
220 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
221 	    "csum_flags %#x",
222 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
223 
224 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
225 		in6_delayed_cksum(m, plen - optlen,
226 		    sizeof(struct ip6_hdr) + optlen);
227 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
228 	}
229 #if defined(SCTP) || defined(SCTP_SUPPORT)
230 	if (csum_flags & CSUM_SCTP_IPV6) {
231 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
232 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
233 	}
234 #endif
235 }
236 
237 int
238 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
239     int fraglen , uint32_t id)
240 {
241 	struct mbuf *m, **mnext, *m_frgpart;
242 	struct ip6_hdr *ip6, *mhip6;
243 	struct ip6_frag *ip6f;
244 	int off;
245 	int error;
246 	int tlen = m0->m_pkthdr.len;
247 
248 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
249 
250 	m = m0;
251 	ip6 = mtod(m, struct ip6_hdr *);
252 	mnext = &m->m_nextpkt;
253 
254 	for (off = hlen; off < tlen; off += fraglen) {
255 		m = m_gethdr(M_NOWAIT, MT_DATA);
256 		if (!m) {
257 			IP6STAT_INC(ip6s_odropped);
258 			return (ENOBUFS);
259 		}
260 
261 		/*
262 		 * Make sure the complete packet header gets copied
263 		 * from the originating mbuf to the newly created
264 		 * mbuf. This also ensures that existing firewall
265 		 * classification(s), VLAN tags and so on get copied
266 		 * to the resulting fragmented packet(s):
267 		 */
268 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
269 			m_free(m);
270 			IP6STAT_INC(ip6s_odropped);
271 			return (ENOBUFS);
272 		}
273 
274 		*mnext = m;
275 		mnext = &m->m_nextpkt;
276 		m->m_data += max_linkhdr;
277 		mhip6 = mtod(m, struct ip6_hdr *);
278 		*mhip6 = *ip6;
279 		m->m_len = sizeof(*mhip6);
280 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
281 		if (error) {
282 			IP6STAT_INC(ip6s_odropped);
283 			return (error);
284 		}
285 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
286 		if (off + fraglen >= tlen)
287 			fraglen = tlen - off;
288 		else
289 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
290 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
291 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
292 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
293 			IP6STAT_INC(ip6s_odropped);
294 			return (ENOBUFS);
295 		}
296 		m_cat(m, m_frgpart);
297 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
298 		ip6f->ip6f_reserved = 0;
299 		ip6f->ip6f_ident = id;
300 		ip6f->ip6f_nxt = nextproto;
301 		IP6STAT_INC(ip6s_ofragments);
302 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
303 	}
304 
305 	return (0);
306 }
307 
308 static int
309 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
310     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
311     bool stamp_tag)
312 {
313 #ifdef KERN_TLS
314 	struct ktls_session *tls = NULL;
315 #endif
316 	struct m_snd_tag *mst;
317 	int error;
318 
319 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
320 	mst = NULL;
321 
322 #ifdef KERN_TLS
323 	/*
324 	 * If this is an unencrypted TLS record, save a reference to
325 	 * the record.  This local reference is used to call
326 	 * ktls_output_eagain after the mbuf has been freed (thus
327 	 * dropping the mbuf's reference) in if_output.
328 	 */
329 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
330 		tls = ktls_hold(m->m_next->m_epg_tls);
331 		mst = tls->snd_tag;
332 
333 		/*
334 		 * If a TLS session doesn't have a valid tag, it must
335 		 * have had an earlier ifp mismatch, so drop this
336 		 * packet.
337 		 */
338 		if (mst == NULL) {
339 			m_freem(m);
340 			error = EAGAIN;
341 			goto done;
342 		}
343 		/*
344 		 * Always stamp tags that include NIC ktls.
345 		 */
346 		stamp_tag = true;
347 	}
348 #endif
349 #ifdef RATELIMIT
350 	if (inp != NULL && mst == NULL) {
351 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
352 		    (inp->inp_snd_tag != NULL &&
353 		    inp->inp_snd_tag->ifp != ifp))
354 			in_pcboutput_txrtlmt(inp, ifp, m);
355 
356 		if (inp->inp_snd_tag != NULL)
357 			mst = inp->inp_snd_tag;
358 	}
359 #endif
360 	if (stamp_tag && mst != NULL) {
361 		KASSERT(m->m_pkthdr.rcvif == NULL,
362 		    ("trying to add a send tag to a forwarded packet"));
363 		if (mst->ifp != ifp) {
364 			m_freem(m);
365 			error = EAGAIN;
366 			goto done;
367 		}
368 
369 		/* stamp send tag on mbuf */
370 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
371 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
372 	}
373 
374 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
375 
376 done:
377 	/* Check for route change invalidating send tags. */
378 #ifdef KERN_TLS
379 	if (tls != NULL) {
380 		if (error == EAGAIN)
381 			error = ktls_output_eagain(inp, tls);
382 		ktls_free(tls);
383 	}
384 #endif
385 #ifdef RATELIMIT
386 	if (error == EAGAIN)
387 		in_pcboutput_eagain(inp);
388 #endif
389 	return (error);
390 }
391 
392 /*
393  * IP6 output.
394  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
395  * nxt, hlim, src, dst).
396  * This function may modify ver and hlim only.
397  * The mbuf chain containing the packet will be freed.
398  * The mbuf opt, if present, will not be freed.
399  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
400  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
401  * then result of route lookup is stored in ro->ro_nh.
402  *
403  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
404  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
405  *
406  * ifpp - XXX: just for statistics
407  */
408 int
409 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
410     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
411     struct ifnet **ifpp, struct inpcb *inp)
412 {
413 	struct ip6_hdr *ip6;
414 	struct ifnet *ifp, *origifp;
415 	struct mbuf *m = m0;
416 	struct mbuf *mprev;
417 	struct route_in6 *ro_pmtu;
418 	struct nhop_object *nh;
419 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
420 	struct in6_addr odst;
421 	u_char *nexthdrp;
422 	int tlen, len;
423 	int error = 0;
424 	int vlan_pcp = -1;
425 	struct in6_ifaddr *ia = NULL;
426 	u_long mtu;
427 	int alwaysfrag, dontfrag;
428 	u_int32_t optlen, plen = 0, unfragpartlen;
429 	struct ip6_exthdrs exthdrs;
430 	struct in6_addr src0, dst0;
431 	u_int32_t zone;
432 	bool hdrsplit;
433 	int sw_csum, tso;
434 	int needfiblookup;
435 	uint32_t fibnum;
436 	struct m_tag *fwd_tag = NULL;
437 	uint32_t id;
438 
439 	NET_EPOCH_ASSERT();
440 
441 	if (inp != NULL) {
442 		INP_LOCK_ASSERT(inp);
443 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
444 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
445 			/* Unconditionally set flowid. */
446 			m->m_pkthdr.flowid = inp->inp_flowid;
447 			M_HASHTYPE_SET(m, inp->inp_flowtype);
448 		}
449 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
450 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
451 			    INP_2PCP_SHIFT;
452 #ifdef NUMA
453 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
454 #endif
455 	}
456 
457 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
458 	/*
459 	 * IPSec checking which handles several cases.
460 	 * FAST IPSEC: We re-injected the packet.
461 	 * XXX: need scope argument.
462 	 */
463 	if (IPSEC_ENABLED(ipv6)) {
464 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
465 			if (error == EINPROGRESS)
466 				error = 0;
467 			goto done;
468 		}
469 	}
470 #endif /* IPSEC */
471 
472 	/* Source address validation. */
473 	ip6 = mtod(m, struct ip6_hdr *);
474 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
475 	    (flags & IPV6_UNSPECSRC) == 0) {
476 		error = EOPNOTSUPP;
477 		IP6STAT_INC(ip6s_badscope);
478 		goto bad;
479 	}
480 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
481 		error = EOPNOTSUPP;
482 		IP6STAT_INC(ip6s_badscope);
483 		goto bad;
484 	}
485 
486 	/*
487 	 * If we are given packet options to add extension headers prepare them.
488 	 * Calculate the total length of the extension header chain.
489 	 * Keep the length of the unfragmentable part for fragmentation.
490 	 */
491 	bzero(&exthdrs, sizeof(exthdrs));
492 	optlen = 0;
493 	unfragpartlen = sizeof(struct ip6_hdr);
494 	if (opt) {
495 		/* Hop-by-Hop options header. */
496 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
497 
498 		/* Destination options header (1st part). */
499 		if (opt->ip6po_rthdr) {
500 #ifndef RTHDR_SUPPORT_IMPLEMENTED
501 			/*
502 			 * If there is a routing header, discard the packet
503 			 * right away here. RH0/1 are obsolete and we do not
504 			 * currently support RH2/3/4.
505 			 * People trying to use RH253/254 may want to disable
506 			 * this check.
507 			 * The moment we do support any routing header (again)
508 			 * this block should check the routing type more
509 			 * selectively.
510 			 */
511 			error = EINVAL;
512 			goto bad;
513 #endif
514 
515 			/*
516 			 * Destination options header (1st part).
517 			 * This only makes sense with a routing header.
518 			 * See Section 9.2 of RFC 3542.
519 			 * Disabling this part just for MIP6 convenience is
520 			 * a bad idea.  We need to think carefully about a
521 			 * way to make the advanced API coexist with MIP6
522 			 * options, which might automatically be inserted in
523 			 * the kernel.
524 			 */
525 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
526 			    optlen);
527 		}
528 		/* Routing header. */
529 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
530 
531 		unfragpartlen += optlen;
532 
533 		/*
534 		 * NOTE: we don't add AH/ESP length here (done in
535 		 * ip6_ipsec_output()).
536 		 */
537 
538 		/* Destination options header (2nd part). */
539 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
540 	}
541 
542 	/*
543 	 * If there is at least one extension header,
544 	 * separate IP6 header from the payload.
545 	 */
546 	hdrsplit = false;
547 	if (optlen) {
548 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
549 			m = NULL;
550 			goto freehdrs;
551 		}
552 		m = exthdrs.ip6e_ip6;
553 		ip6 = mtod(m, struct ip6_hdr *);
554 		hdrsplit = true;
555 	}
556 
557 	/* Adjust mbuf packet header length. */
558 	m->m_pkthdr.len += optlen;
559 	plen = m->m_pkthdr.len - sizeof(*ip6);
560 
561 	/* If this is a jumbo payload, insert a jumbo payload option. */
562 	if (plen > IPV6_MAXPACKET) {
563 		if (!hdrsplit) {
564 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
565 				m = NULL;
566 				goto freehdrs;
567 			}
568 			m = exthdrs.ip6e_ip6;
569 			ip6 = mtod(m, struct ip6_hdr *);
570 			hdrsplit = true;
571 		}
572 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
573 			goto freehdrs;
574 		ip6->ip6_plen = 0;
575 	} else
576 		ip6->ip6_plen = htons(plen);
577 	nexthdrp = &ip6->ip6_nxt;
578 
579 	if (optlen) {
580 		/*
581 		 * Concatenate headers and fill in next header fields.
582 		 * Here we have, on "m"
583 		 *	IPv6 payload
584 		 * and we insert headers accordingly.
585 		 * Finally, we should be getting:
586 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
587 		 *
588 		 * During the header composing process "m" points to IPv6
589 		 * header.  "mprev" points to an extension header prior to esp.
590 		 */
591 		mprev = m;
592 
593 		/*
594 		 * We treat dest2 specially.  This makes IPsec processing
595 		 * much easier.  The goal here is to make mprev point the
596 		 * mbuf prior to dest2.
597 		 *
598 		 * Result: IPv6 dest2 payload.
599 		 * m and mprev will point to IPv6 header.
600 		 */
601 		if (exthdrs.ip6e_dest2) {
602 			if (!hdrsplit)
603 				panic("%s:%d: assumption failed: "
604 				    "hdr not split: hdrsplit %d exthdrs %p",
605 				    __func__, __LINE__, hdrsplit, &exthdrs);
606 			exthdrs.ip6e_dest2->m_next = m->m_next;
607 			m->m_next = exthdrs.ip6e_dest2;
608 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
609 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
610 		}
611 
612 		/*
613 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
614 		 * m will point to IPv6 header.  mprev will point to the
615 		 * extension header prior to dest2 (rthdr in the above case).
616 		 */
617 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
618 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
619 			   IPPROTO_DSTOPTS);
620 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
621 			   IPPROTO_ROUTING);
622 	}
623 
624 	IP6STAT_INC(ip6s_localout);
625 
626 	/* Route packet. */
627 	ro_pmtu = ro;
628 	if (opt && opt->ip6po_rthdr)
629 		ro = &opt->ip6po_route;
630 	if (ro != NULL)
631 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
632 	else
633 		dst = &sin6;
634 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
635 
636 again:
637 	/*
638 	 * If specified, try to fill in the traffic class field.
639 	 * Do not override if a non-zero value is already set.
640 	 * We check the diffserv field and the ECN field separately.
641 	 */
642 	if (opt && opt->ip6po_tclass >= 0) {
643 		int mask = 0;
644 
645 		if (IPV6_DSCP(ip6) == 0)
646 			mask |= 0xfc;
647 		if (IPV6_ECN(ip6) == 0)
648 			mask |= 0x03;
649 		if (mask != 0)
650 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
651 	}
652 
653 	/* Fill in or override the hop limit field, if necessary. */
654 	if (opt && opt->ip6po_hlim != -1)
655 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
656 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
657 		if (im6o != NULL)
658 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
659 		else
660 			ip6->ip6_hlim = V_ip6_defmcasthlim;
661 	}
662 
663 	if (ro == NULL || ro->ro_nh == NULL) {
664 		bzero(dst, sizeof(*dst));
665 		dst->sin6_family = AF_INET6;
666 		dst->sin6_len = sizeof(*dst);
667 		dst->sin6_addr = ip6->ip6_dst;
668 	}
669 	/*
670 	 * Validate route against routing table changes.
671 	 * Make sure that the address family is set in route.
672 	 */
673 	nh = NULL;
674 	ifp = NULL;
675 	mtu = 0;
676 	if (ro != NULL) {
677 		if (ro->ro_nh != NULL && inp != NULL) {
678 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
679 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
680 			    fibnum);
681 		}
682 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
683 		    (!NH_IS_VALID(ro->ro_nh) ||
684 		    ro->ro_dst.sin6_family != AF_INET6 ||
685 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
686 			RO_INVALIDATE_CACHE(ro);
687 
688 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
689 		    ro->ro_dst.sin6_family == AF_INET6 &&
690 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
691 			/* Nexthop is valid and contains valid ifp */
692 			nh = ro->ro_nh;
693 		} else {
694 			if (ro->ro_lle)
695 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
696 			ro->ro_lle = NULL;
697 			if (fwd_tag == NULL) {
698 				bzero(&dst_sa, sizeof(dst_sa));
699 				dst_sa.sin6_family = AF_INET6;
700 				dst_sa.sin6_len = sizeof(dst_sa);
701 				dst_sa.sin6_addr = ip6->ip6_dst;
702 			}
703 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
704 			    &nh, fibnum, m->m_pkthdr.flowid);
705 			if (error != 0) {
706 				IP6STAT_INC(ip6s_noroute);
707 				if (ifp != NULL)
708 					in6_ifstat_inc(ifp, ifs6_out_discard);
709 				goto bad;
710 			}
711 			/*
712 			 * At this point at least @ifp is not NULL
713 			 * Can be the case when dst is multicast, link-local or
714 			 * interface is explicitly specificed by the caller.
715 			 */
716 		}
717 		if (nh == NULL) {
718 			/*
719 			 * If in6_selectroute() does not return a nexthop
720 			 * dst may not have been updated.
721 			 */
722 			*dst = dst_sa;	/* XXX */
723 			origifp = ifp;
724 			mtu = ifp->if_mtu;
725 		} else {
726 			ifp = nh->nh_ifp;
727 			origifp = nh->nh_aifp;
728 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
729 			counter_u64_add(nh->nh_pksent, 1);
730 		}
731 	} else {
732 		struct nhop_object *nh;
733 		struct in6_addr kdst;
734 		uint32_t scopeid;
735 
736 		if (fwd_tag == NULL) {
737 			bzero(&dst_sa, sizeof(dst_sa));
738 			dst_sa.sin6_family = AF_INET6;
739 			dst_sa.sin6_len = sizeof(dst_sa);
740 			dst_sa.sin6_addr = ip6->ip6_dst;
741 		}
742 
743 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
744 		    im6o != NULL &&
745 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
746 			/* We do not need a route lookup. */
747 			*dst = dst_sa;	/* XXX */
748 			origifp = ifp;
749 			goto nonh6lookup;
750 		}
751 
752 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
753 
754 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
755 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
756 			if (scopeid > 0) {
757 				ifp = in6_getlinkifnet(scopeid);
758 				if (ifp == NULL) {
759 					error = EHOSTUNREACH;
760 					goto bad;
761 				}
762 				*dst = dst_sa;	/* XXX */
763 				origifp = ifp;
764 				goto nonh6lookup;
765 			}
766 		}
767 
768 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
769 		    m->m_pkthdr.flowid);
770 		if (nh == NULL) {
771 			IP6STAT_INC(ip6s_noroute);
772 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
773 			error = EHOSTUNREACH;;
774 			goto bad;
775 		}
776 
777 		ifp = nh->nh_ifp;
778 		origifp = nh->nh_aifp;
779 		ia = ifatoia6(nh->nh_ifa);
780 		if (nh->nh_flags & NHF_GATEWAY)
781 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
782 		else if (fwd_tag != NULL)
783 			dst->sin6_addr = dst_sa.sin6_addr;
784 nonh6lookup:
785 		;
786 	}
787 	/*
788 	 * At this point ifp MUST be pointing to the valid transmit ifp.
789 	 * origifp MUST be valid and pointing to either the same ifp or,
790 	 * in case of loopback output, to the interface which ip6_src
791 	 * belongs to.
792 	 * Examples:
793 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
794 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
795 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
796 	 *
797 	 * mtu can be 0 and will be refined later.
798 	 */
799 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
800 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
801 
802 	if ((flags & IPV6_FORWARDING) == 0) {
803 		/* XXX: the FORWARDING flag can be set for mrouting. */
804 		in6_ifstat_inc(ifp, ifs6_out_request);
805 	}
806 
807 	/* Setup data structures for scope ID checks. */
808 	src0 = ip6->ip6_src;
809 	bzero(&src_sa, sizeof(src_sa));
810 	src_sa.sin6_family = AF_INET6;
811 	src_sa.sin6_len = sizeof(src_sa);
812 	src_sa.sin6_addr = ip6->ip6_src;
813 
814 	dst0 = ip6->ip6_dst;
815 	/* Re-initialize to be sure. */
816 	bzero(&dst_sa, sizeof(dst_sa));
817 	dst_sa.sin6_family = AF_INET6;
818 	dst_sa.sin6_len = sizeof(dst_sa);
819 	dst_sa.sin6_addr = ip6->ip6_dst;
820 
821 	/* Check for valid scope ID. */
822 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
823 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
824 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
825 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
826 		/*
827 		 * The outgoing interface is in the zone of the source
828 		 * and destination addresses.
829 		 *
830 		 */
831 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
832 	    sa6_recoverscope(&src_sa) != 0 ||
833 	    sa6_recoverscope(&dst_sa) != 0 ||
834 	    dst_sa.sin6_scope_id == 0 ||
835 	    (src_sa.sin6_scope_id != 0 &&
836 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
837 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
838 		/*
839 		 * If the destination network interface is not a
840 		 * loopback interface, or the destination network
841 		 * address has no scope ID, or the source address has
842 		 * a scope ID set which is different from the
843 		 * destination address one, or there is no network
844 		 * interface representing this scope ID, the address
845 		 * pair is considered invalid.
846 		 */
847 		IP6STAT_INC(ip6s_badscope);
848 		in6_ifstat_inc(origifp, ifs6_out_discard);
849 		if (error == 0)
850 			error = EHOSTUNREACH; /* XXX */
851 		goto bad;
852 	}
853 	/* All scope ID checks are successful. */
854 
855 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
856 		if (opt && opt->ip6po_nextroute.ro_nh) {
857 			/*
858 			 * The nexthop is explicitly specified by the
859 			 * application.  We assume the next hop is an IPv6
860 			 * address.
861 			 */
862 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
863 		}
864 		else if ((nh->nh_flags & NHF_GATEWAY))
865 			dst = &nh->gw6_sa;
866 	}
867 
868 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
869 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
870 	} else {
871 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
872 		in6_ifstat_inc(ifp, ifs6_out_mcast);
873 
874 		/* Confirm that the outgoing interface supports multicast. */
875 		if (!(ifp->if_flags & IFF_MULTICAST)) {
876 			IP6STAT_INC(ip6s_noroute);
877 			in6_ifstat_inc(ifp, ifs6_out_discard);
878 			error = ENETUNREACH;
879 			goto bad;
880 		}
881 		if ((im6o == NULL && in6_mcast_loop) ||
882 		    (im6o && im6o->im6o_multicast_loop)) {
883 			/*
884 			 * Loop back multicast datagram if not expressly
885 			 * forbidden to do so, even if we have not joined
886 			 * the address; protocols will filter it later,
887 			 * thus deferring a hash lookup and lock acquisition
888 			 * at the expense of an m_copym().
889 			 */
890 			ip6_mloopback(ifp, m);
891 		} else {
892 			/*
893 			 * If we are acting as a multicast router, perform
894 			 * multicast forwarding as if the packet had just
895 			 * arrived on the interface to which we are about
896 			 * to send.  The multicast forwarding function
897 			 * recursively calls this function, using the
898 			 * IPV6_FORWARDING flag to prevent infinite recursion.
899 			 *
900 			 * Multicasts that are looped back by ip6_mloopback(),
901 			 * above, will be forwarded by the ip6_input() routine,
902 			 * if necessary.
903 			 */
904 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
905 				/*
906 				 * XXX: ip6_mforward expects that rcvif is NULL
907 				 * when it is called from the originating path.
908 				 * However, it may not always be the case.
909 				 */
910 				m->m_pkthdr.rcvif = NULL;
911 				if (ip6_mforward(ip6, ifp, m) != 0) {
912 					m_freem(m);
913 					goto done;
914 				}
915 			}
916 		}
917 		/*
918 		 * Multicasts with a hoplimit of zero may be looped back,
919 		 * above, but must not be transmitted on a network.
920 		 * Also, multicasts addressed to the loopback interface
921 		 * are not sent -- the above call to ip6_mloopback() will
922 		 * loop back a copy if this host actually belongs to the
923 		 * destination group on the loopback interface.
924 		 */
925 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
926 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
927 			m_freem(m);
928 			goto done;
929 		}
930 	}
931 
932 	/*
933 	 * Fill the outgoing inteface to tell the upper layer
934 	 * to increment per-interface statistics.
935 	 */
936 	if (ifpp)
937 		*ifpp = ifp;
938 
939 	/* Determine path MTU. */
940 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
941 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
942 		goto bad;
943 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p "
944 	    "alwaysfrag %d fibnum %u\n", __func__, __LINE__, mtu, ro_pmtu, ro,
945 	    ifp, alwaysfrag, fibnum));
946 
947 	/*
948 	 * The caller of this function may specify to use the minimum MTU
949 	 * in some cases.
950 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
951 	 * setting.  The logic is a bit complicated; by default, unicast
952 	 * packets will follow path MTU while multicast packets will be sent at
953 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
954 	 * including unicast ones will be sent at the minimum MTU.  Multicast
955 	 * packets will always be sent at the minimum MTU unless
956 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
957 	 * See RFC 3542 for more details.
958 	 */
959 	if (mtu > IPV6_MMTU) {
960 		if ((flags & IPV6_MINMTU))
961 			mtu = IPV6_MMTU;
962 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
963 			mtu = IPV6_MMTU;
964 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
965 			 (opt == NULL ||
966 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
967 			mtu = IPV6_MMTU;
968 		}
969 	}
970 
971 	/*
972 	 * Clear embedded scope identifiers if necessary.
973 	 * in6_clearscope() will touch the addresses only when necessary.
974 	 */
975 	in6_clearscope(&ip6->ip6_src);
976 	in6_clearscope(&ip6->ip6_dst);
977 
978 	/*
979 	 * If the outgoing packet contains a hop-by-hop options header,
980 	 * it must be examined and processed even by the source node.
981 	 * (RFC 2460, section 4.)
982 	 */
983 	if (exthdrs.ip6e_hbh) {
984 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
985 		u_int32_t dummy; /* XXX unused */
986 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
987 
988 #ifdef DIAGNOSTIC
989 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
990 			panic("ip6e_hbh is not contiguous");
991 #endif
992 		/*
993 		 *  XXX: if we have to send an ICMPv6 error to the sender,
994 		 *       we need the M_LOOP flag since icmp6_error() expects
995 		 *       the IPv6 and the hop-by-hop options header are
996 		 *       contiguous unless the flag is set.
997 		 */
998 		m->m_flags |= M_LOOP;
999 		m->m_pkthdr.rcvif = ifp;
1000 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
1001 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1002 		    &dummy, &plen) < 0) {
1003 			/* m was already freed at this point. */
1004 			error = EINVAL;/* better error? */
1005 			goto done;
1006 		}
1007 		m->m_flags &= ~M_LOOP; /* XXX */
1008 		m->m_pkthdr.rcvif = NULL;
1009 	}
1010 
1011 	/* Jump over all PFIL processing if hooks are not active. */
1012 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1013 		goto passout;
1014 
1015 	odst = ip6->ip6_dst;
1016 	/* Run through list of hooks for output packets. */
1017 	switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
1018 	case PFIL_PASS:
1019 		ip6 = mtod(m, struct ip6_hdr *);
1020 		break;
1021 	case PFIL_DROPPED:
1022 		error = EACCES;
1023 		/* FALLTHROUGH */
1024 	case PFIL_CONSUMED:
1025 		goto done;
1026 	}
1027 
1028 	needfiblookup = 0;
1029 	/* See if destination IP address was changed by packet filter. */
1030 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1031 		m->m_flags |= M_SKIP_FIREWALL;
1032 		/* If destination is now ourself drop to ip6_input(). */
1033 		if (in6_localip(&ip6->ip6_dst)) {
1034 			m->m_flags |= M_FASTFWD_OURS;
1035 			if (m->m_pkthdr.rcvif == NULL)
1036 				m->m_pkthdr.rcvif = V_loif;
1037 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1038 				m->m_pkthdr.csum_flags |=
1039 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1040 				m->m_pkthdr.csum_data = 0xffff;
1041 			}
1042 #if defined(SCTP) || defined(SCTP_SUPPORT)
1043 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1044 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1045 #endif
1046 			error = netisr_queue(NETISR_IPV6, m);
1047 			goto done;
1048 		} else {
1049 			if (ro != NULL)
1050 				RO_INVALIDATE_CACHE(ro);
1051 			needfiblookup = 1; /* Redo the routing table lookup. */
1052 		}
1053 	}
1054 	/* See if fib was changed by packet filter. */
1055 	if (fibnum != M_GETFIB(m)) {
1056 		m->m_flags |= M_SKIP_FIREWALL;
1057 		fibnum = M_GETFIB(m);
1058 		if (ro != NULL)
1059 			RO_INVALIDATE_CACHE(ro);
1060 		needfiblookup = 1;
1061 	}
1062 	if (needfiblookup)
1063 		goto again;
1064 
1065 	/* See if local, if yes, send it to netisr. */
1066 	if (m->m_flags & M_FASTFWD_OURS) {
1067 		if (m->m_pkthdr.rcvif == NULL)
1068 			m->m_pkthdr.rcvif = V_loif;
1069 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1070 			m->m_pkthdr.csum_flags |=
1071 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1072 			m->m_pkthdr.csum_data = 0xffff;
1073 		}
1074 #if defined(SCTP) || defined(SCTP_SUPPORT)
1075 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1076 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1077 #endif
1078 		error = netisr_queue(NETISR_IPV6, m);
1079 		goto done;
1080 	}
1081 	/* Or forward to some other address? */
1082 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1083 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1084 		if (ro != NULL)
1085 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1086 		else
1087 			dst = &sin6;
1088 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1089 		m->m_flags |= M_SKIP_FIREWALL;
1090 		m->m_flags &= ~M_IP6_NEXTHOP;
1091 		m_tag_delete(m, fwd_tag);
1092 		goto again;
1093 	}
1094 
1095 passout:
1096 	if (vlan_pcp > -1)
1097 		EVL_APPLY_PRI(m, vlan_pcp);
1098 
1099 	/* Ensure the packet data is mapped if the interface requires it. */
1100 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1101 		m = mb_unmapped_to_ext(m);
1102 		if (m == NULL) {
1103 			IP6STAT_INC(ip6s_odropped);
1104 			return (ENOBUFS);
1105 		}
1106 	}
1107 
1108 	/*
1109 	 * Send the packet to the outgoing interface.
1110 	 * If necessary, do IPv6 fragmentation before sending.
1111 	 *
1112 	 * The logic here is rather complex:
1113 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1114 	 * 1-a:	send as is if tlen <= path mtu
1115 	 * 1-b:	fragment if tlen > path mtu
1116 	 *
1117 	 * 2: if user asks us not to fragment (dontfrag == 1)
1118 	 * 2-a:	send as is if tlen <= interface mtu
1119 	 * 2-b:	error if tlen > interface mtu
1120 	 *
1121 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1122 	 *	always fragment
1123 	 *
1124 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1125 	 *	error, as we cannot handle this conflicting request.
1126 	 */
1127 	sw_csum = m->m_pkthdr.csum_flags;
1128 	if (!hdrsplit) {
1129 		tso = ((sw_csum & ifp->if_hwassist &
1130 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1131 		sw_csum &= ~ifp->if_hwassist;
1132 	} else
1133 		tso = 0;
1134 	/*
1135 	 * If we added extension headers, we will not do TSO and calculate the
1136 	 * checksums ourselves for now.
1137 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1138 	 * with ext. hdrs.
1139 	 */
1140 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1141 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1142 	tlen = m->m_pkthdr.len;
1143 
1144 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1145 		dontfrag = 1;
1146 	else
1147 		dontfrag = 0;
1148 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1149 		/* Conflicting request - can't transmit. */
1150 		error = EMSGSIZE;
1151 		goto bad;
1152 	}
1153 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1154 		/*
1155 		 * Even if the DONTFRAG option is specified, we cannot send the
1156 		 * packet when the data length is larger than the MTU of the
1157 		 * outgoing interface.
1158 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1159 		 * application wanted to know the MTU value. Also return an
1160 		 * error code (this is not described in the API spec).
1161 		 */
1162 		if (inp != NULL)
1163 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1164 		error = EMSGSIZE;
1165 		goto bad;
1166 	}
1167 
1168 	/* Transmit packet without fragmentation. */
1169 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1170 		struct in6_ifaddr *ia6;
1171 
1172 		ip6 = mtod(m, struct ip6_hdr *);
1173 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1174 		if (ia6) {
1175 			/* Record statistics for this interface address. */
1176 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1177 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1178 			    m->m_pkthdr.len);
1179 		}
1180 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1181 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1182 		goto done;
1183 	}
1184 
1185 	/* Try to fragment the packet.  Cases 1-b and 3. */
1186 	if (mtu < IPV6_MMTU) {
1187 		/* Path MTU cannot be less than IPV6_MMTU. */
1188 		error = EMSGSIZE;
1189 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1190 		goto bad;
1191 	} else if (ip6->ip6_plen == 0) {
1192 		/* Jumbo payload cannot be fragmented. */
1193 		error = EMSGSIZE;
1194 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1195 		goto bad;
1196 	} else {
1197 		u_char nextproto;
1198 
1199 		/*
1200 		 * Too large for the destination or interface;
1201 		 * fragment if possible.
1202 		 * Must be able to put at least 8 bytes per fragment.
1203 		 */
1204 		if (mtu > IPV6_MAXPACKET)
1205 			mtu = IPV6_MAXPACKET;
1206 
1207 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1208 		if (len < 8) {
1209 			error = EMSGSIZE;
1210 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1211 			goto bad;
1212 		}
1213 
1214 		/*
1215 		 * If the interface will not calculate checksums on
1216 		 * fragmented packets, then do it here.
1217 		 * XXX-BZ handle the hw offloading case.  Need flags.
1218 		 */
1219 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1220 		    optlen);
1221 
1222 		/*
1223 		 * Change the next header field of the last header in the
1224 		 * unfragmentable part.
1225 		 */
1226 		if (exthdrs.ip6e_rthdr) {
1227 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1228 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1229 		} else if (exthdrs.ip6e_dest1) {
1230 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1231 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1232 		} else if (exthdrs.ip6e_hbh) {
1233 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1234 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1235 		} else {
1236 			ip6 = mtod(m, struct ip6_hdr *);
1237 			nextproto = ip6->ip6_nxt;
1238 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1239 		}
1240 
1241 		/*
1242 		 * Loop through length of segment after first fragment,
1243 		 * make new header and copy data of each part and link onto
1244 		 * chain.
1245 		 */
1246 		m0 = m;
1247 		id = htonl(ip6_randomid());
1248 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1249 		if (error != 0)
1250 			goto sendorfree;
1251 
1252 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1253 	}
1254 
1255 	/* Remove leading garbage. */
1256 sendorfree:
1257 	m = m0->m_nextpkt;
1258 	m0->m_nextpkt = 0;
1259 	m_freem(m0);
1260 	for (; m; m = m0) {
1261 		m0 = m->m_nextpkt;
1262 		m->m_nextpkt = 0;
1263 		if (error == 0) {
1264 			/* Record statistics for this interface address. */
1265 			if (ia) {
1266 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1267 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1268 				    m->m_pkthdr.len);
1269 			}
1270 			if (vlan_pcp > -1)
1271 				EVL_APPLY_PRI(m, vlan_pcp);
1272 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1273 			    true);
1274 		} else
1275 			m_freem(m);
1276 	}
1277 
1278 	if (error == 0)
1279 		IP6STAT_INC(ip6s_fragmented);
1280 
1281 done:
1282 	return (error);
1283 
1284 freehdrs:
1285 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1286 	m_freem(exthdrs.ip6e_dest1);
1287 	m_freem(exthdrs.ip6e_rthdr);
1288 	m_freem(exthdrs.ip6e_dest2);
1289 	/* FALLTHROUGH */
1290 bad:
1291 	if (m)
1292 		m_freem(m);
1293 	goto done;
1294 }
1295 
1296 static int
1297 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1298 {
1299 	struct mbuf *m;
1300 
1301 	if (hlen > MCLBYTES)
1302 		return (ENOBUFS); /* XXX */
1303 
1304 	if (hlen > MLEN)
1305 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1306 	else
1307 		m = m_get(M_NOWAIT, MT_DATA);
1308 	if (m == NULL)
1309 		return (ENOBUFS);
1310 	m->m_len = hlen;
1311 	if (hdr)
1312 		bcopy(hdr, mtod(m, caddr_t), hlen);
1313 
1314 	*mp = m;
1315 	return (0);
1316 }
1317 
1318 /*
1319  * Insert jumbo payload option.
1320  */
1321 static int
1322 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1323 {
1324 	struct mbuf *mopt;
1325 	u_char *optbuf;
1326 	u_int32_t v;
1327 
1328 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1329 
1330 	/*
1331 	 * If there is no hop-by-hop options header, allocate new one.
1332 	 * If there is one but it doesn't have enough space to store the
1333 	 * jumbo payload option, allocate a cluster to store the whole options.
1334 	 * Otherwise, use it to store the options.
1335 	 */
1336 	if (exthdrs->ip6e_hbh == NULL) {
1337 		mopt = m_get(M_NOWAIT, MT_DATA);
1338 		if (mopt == NULL)
1339 			return (ENOBUFS);
1340 		mopt->m_len = JUMBOOPTLEN;
1341 		optbuf = mtod(mopt, u_char *);
1342 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1343 		exthdrs->ip6e_hbh = mopt;
1344 	} else {
1345 		struct ip6_hbh *hbh;
1346 
1347 		mopt = exthdrs->ip6e_hbh;
1348 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1349 			/*
1350 			 * XXX assumption:
1351 			 * - exthdrs->ip6e_hbh is not referenced from places
1352 			 *   other than exthdrs.
1353 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1354 			 */
1355 			int oldoptlen = mopt->m_len;
1356 			struct mbuf *n;
1357 
1358 			/*
1359 			 * XXX: give up if the whole (new) hbh header does
1360 			 * not fit even in an mbuf cluster.
1361 			 */
1362 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1363 				return (ENOBUFS);
1364 
1365 			/*
1366 			 * As a consequence, we must always prepare a cluster
1367 			 * at this point.
1368 			 */
1369 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1370 			if (n == NULL)
1371 				return (ENOBUFS);
1372 			n->m_len = oldoptlen + JUMBOOPTLEN;
1373 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1374 			    oldoptlen);
1375 			optbuf = mtod(n, caddr_t) + oldoptlen;
1376 			m_freem(mopt);
1377 			mopt = exthdrs->ip6e_hbh = n;
1378 		} else {
1379 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1380 			mopt->m_len += JUMBOOPTLEN;
1381 		}
1382 		optbuf[0] = IP6OPT_PADN;
1383 		optbuf[1] = 1;
1384 
1385 		/*
1386 		 * Adjust the header length according to the pad and
1387 		 * the jumbo payload option.
1388 		 */
1389 		hbh = mtod(mopt, struct ip6_hbh *);
1390 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1391 	}
1392 
1393 	/* fill in the option. */
1394 	optbuf[2] = IP6OPT_JUMBO;
1395 	optbuf[3] = 4;
1396 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1397 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1398 
1399 	/* finally, adjust the packet header length */
1400 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1401 
1402 	return (0);
1403 #undef JUMBOOPTLEN
1404 }
1405 
1406 /*
1407  * Insert fragment header and copy unfragmentable header portions.
1408  */
1409 static int
1410 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1411     struct ip6_frag **frghdrp)
1412 {
1413 	struct mbuf *n, *mlast;
1414 
1415 	if (hlen > sizeof(struct ip6_hdr)) {
1416 		n = m_copym(m0, sizeof(struct ip6_hdr),
1417 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1418 		if (n == NULL)
1419 			return (ENOBUFS);
1420 		m->m_next = n;
1421 	} else
1422 		n = m;
1423 
1424 	/* Search for the last mbuf of unfragmentable part. */
1425 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1426 		;
1427 
1428 	if (M_WRITABLE(mlast) &&
1429 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1430 		/* use the trailing space of the last mbuf for the fragment hdr */
1431 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1432 		    mlast->m_len);
1433 		mlast->m_len += sizeof(struct ip6_frag);
1434 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1435 	} else {
1436 		/* allocate a new mbuf for the fragment header */
1437 		struct mbuf *mfrg;
1438 
1439 		mfrg = m_get(M_NOWAIT, MT_DATA);
1440 		if (mfrg == NULL)
1441 			return (ENOBUFS);
1442 		mfrg->m_len = sizeof(struct ip6_frag);
1443 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1444 		mlast->m_next = mfrg;
1445 	}
1446 
1447 	return (0);
1448 }
1449 
1450 /*
1451  * Calculates IPv6 path mtu for destination @dst.
1452  * Resulting MTU is stored in @mtup.
1453  *
1454  * Returns 0 on success.
1455  */
1456 static int
1457 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1458 {
1459 	struct epoch_tracker et;
1460 	struct nhop_object *nh;
1461 	struct in6_addr kdst;
1462 	uint32_t scopeid;
1463 	int error;
1464 
1465 	in6_splitscope(dst, &kdst, &scopeid);
1466 
1467 	NET_EPOCH_ENTER(et);
1468 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1469 	if (nh != NULL)
1470 		error = ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, NULL, 0);
1471 	else
1472 		error = EHOSTUNREACH;
1473 	NET_EPOCH_EXIT(et);
1474 
1475 	return (error);
1476 }
1477 
1478 /*
1479  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1480  * and cached data in @ro_pmtu.
1481  * MTU from (successful) route lookup is saved (along with dst)
1482  * inside @ro_pmtu to avoid subsequent route lookups after packet
1483  * filter processing.
1484  *
1485  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1486  * Returns 0 on success.
1487  */
1488 static int
1489 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1490     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1491     int *alwaysfragp, u_int fibnum, u_int proto)
1492 {
1493 	struct nhop_object *nh;
1494 	struct in6_addr kdst;
1495 	uint32_t scopeid;
1496 	struct sockaddr_in6 *sa6_dst, sin6;
1497 	u_long mtu;
1498 
1499 	NET_EPOCH_ASSERT();
1500 
1501 	mtu = 0;
1502 	if (ro_pmtu == NULL || do_lookup) {
1503 		/*
1504 		 * Here ro_pmtu has final destination address, while
1505 		 * ro might represent immediate destination.
1506 		 * Use ro_pmtu destination since mtu might differ.
1507 		 */
1508 		if (ro_pmtu != NULL) {
1509 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1510 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1511 				ro_pmtu->ro_mtu = 0;
1512 		} else
1513 			sa6_dst = &sin6;
1514 
1515 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1516 			bzero(sa6_dst, sizeof(*sa6_dst));
1517 			sa6_dst->sin6_family = AF_INET6;
1518 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1519 			sa6_dst->sin6_addr = *dst;
1520 
1521 			in6_splitscope(dst, &kdst, &scopeid);
1522 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1523 			if (nh != NULL) {
1524 				mtu = nh->nh_mtu;
1525 				if (ro_pmtu != NULL)
1526 					ro_pmtu->ro_mtu = mtu;
1527 			}
1528 		} else
1529 			mtu = ro_pmtu->ro_mtu;
1530 	}
1531 
1532 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1533 		mtu = ro_pmtu->ro_nh->nh_mtu;
1534 
1535 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1536 }
1537 
1538 /*
1539  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1540  * hostcache data for @dst.
1541  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1542  *
1543  * Returns 0 on success.
1544  */
1545 static int
1546 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1547     u_long *mtup, int *alwaysfragp, u_int proto)
1548 {
1549 	u_long mtu = 0;
1550 	int alwaysfrag = 0;
1551 	int error = 0;
1552 
1553 	if (rt_mtu > 0) {
1554 		u_int32_t ifmtu;
1555 		struct in_conninfo inc;
1556 
1557 		bzero(&inc, sizeof(inc));
1558 		inc.inc_flags |= INC_ISIPV6;
1559 		inc.inc6_faddr = *dst;
1560 
1561 		ifmtu = IN6_LINKMTU(ifp);
1562 
1563 		/* TCP is known to react to pmtu changes so skip hc */
1564 		if (proto != IPPROTO_TCP)
1565 			mtu = tcp_hc_getmtu(&inc);
1566 
1567 		if (mtu)
1568 			mtu = min(mtu, rt_mtu);
1569 		else
1570 			mtu = rt_mtu;
1571 		if (mtu == 0)
1572 			mtu = ifmtu;
1573 		else if (mtu < IPV6_MMTU) {
1574 			/*
1575 			 * RFC2460 section 5, last paragraph:
1576 			 * if we record ICMPv6 too big message with
1577 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1578 			 * or smaller, with framgent header attached.
1579 			 * (fragment header is needed regardless from the
1580 			 * packet size, for translators to identify packets)
1581 			 */
1582 			alwaysfrag = 1;
1583 			mtu = IPV6_MMTU;
1584 		}
1585 	} else if (ifp) {
1586 		mtu = IN6_LINKMTU(ifp);
1587 	} else
1588 		error = EHOSTUNREACH; /* XXX */
1589 
1590 	*mtup = mtu;
1591 	if (alwaysfragp)
1592 		*alwaysfragp = alwaysfrag;
1593 	return (error);
1594 }
1595 
1596 /*
1597  * IP6 socket option processing.
1598  */
1599 int
1600 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1601 {
1602 	int optdatalen, uproto;
1603 	void *optdata;
1604 	struct inpcb *inp = sotoinpcb(so);
1605 	int error, optval;
1606 	int level, op, optname;
1607 	int optlen;
1608 	struct thread *td;
1609 #ifdef	RSS
1610 	uint32_t rss_bucket;
1611 	int retval;
1612 #endif
1613 
1614 /*
1615  * Don't use more than a quarter of mbuf clusters.  N.B.:
1616  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1617  * on LP64 architectures, so cast to u_long to avoid undefined
1618  * behavior.  ILP32 architectures cannot have nmbclusters
1619  * large enough to overflow for other reasons.
1620  */
1621 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1622 
1623 	level = sopt->sopt_level;
1624 	op = sopt->sopt_dir;
1625 	optname = sopt->sopt_name;
1626 	optlen = sopt->sopt_valsize;
1627 	td = sopt->sopt_td;
1628 	error = 0;
1629 	optval = 0;
1630 	uproto = (int)so->so_proto->pr_protocol;
1631 
1632 	if (level != IPPROTO_IPV6) {
1633 		error = EINVAL;
1634 
1635 		if (sopt->sopt_level == SOL_SOCKET &&
1636 		    sopt->sopt_dir == SOPT_SET) {
1637 			switch (sopt->sopt_name) {
1638 			case SO_REUSEADDR:
1639 				INP_WLOCK(inp);
1640 				if ((so->so_options & SO_REUSEADDR) != 0)
1641 					inp->inp_flags2 |= INP_REUSEADDR;
1642 				else
1643 					inp->inp_flags2 &= ~INP_REUSEADDR;
1644 				INP_WUNLOCK(inp);
1645 				error = 0;
1646 				break;
1647 			case SO_REUSEPORT:
1648 				INP_WLOCK(inp);
1649 				if ((so->so_options & SO_REUSEPORT) != 0)
1650 					inp->inp_flags2 |= INP_REUSEPORT;
1651 				else
1652 					inp->inp_flags2 &= ~INP_REUSEPORT;
1653 				INP_WUNLOCK(inp);
1654 				error = 0;
1655 				break;
1656 			case SO_REUSEPORT_LB:
1657 				INP_WLOCK(inp);
1658 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1659 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1660 				else
1661 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1662 				INP_WUNLOCK(inp);
1663 				error = 0;
1664 				break;
1665 			case SO_SETFIB:
1666 				INP_WLOCK(inp);
1667 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1668 				INP_WUNLOCK(inp);
1669 				error = 0;
1670 				break;
1671 			case SO_MAX_PACING_RATE:
1672 #ifdef RATELIMIT
1673 				INP_WLOCK(inp);
1674 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1675 				INP_WUNLOCK(inp);
1676 				error = 0;
1677 #else
1678 				error = EOPNOTSUPP;
1679 #endif
1680 				break;
1681 			default:
1682 				break;
1683 			}
1684 		}
1685 	} else {		/* level == IPPROTO_IPV6 */
1686 		switch (op) {
1687 		case SOPT_SET:
1688 			switch (optname) {
1689 			case IPV6_2292PKTOPTIONS:
1690 #ifdef IPV6_PKTOPTIONS
1691 			case IPV6_PKTOPTIONS:
1692 #endif
1693 			{
1694 				struct mbuf *m;
1695 
1696 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1697 					printf("ip6_ctloutput: mbuf limit hit\n");
1698 					error = ENOBUFS;
1699 					break;
1700 				}
1701 
1702 				error = soopt_getm(sopt, &m); /* XXX */
1703 				if (error != 0)
1704 					break;
1705 				error = soopt_mcopyin(sopt, m); /* XXX */
1706 				if (error != 0)
1707 					break;
1708 				INP_WLOCK(inp);
1709 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1710 				    so, sopt);
1711 				INP_WUNLOCK(inp);
1712 				m_freem(m); /* XXX */
1713 				break;
1714 			}
1715 
1716 			/*
1717 			 * Use of some Hop-by-Hop options or some
1718 			 * Destination options, might require special
1719 			 * privilege.  That is, normal applications
1720 			 * (without special privilege) might be forbidden
1721 			 * from setting certain options in outgoing packets,
1722 			 * and might never see certain options in received
1723 			 * packets. [RFC 2292 Section 6]
1724 			 * KAME specific note:
1725 			 *  KAME prevents non-privileged users from sending or
1726 			 *  receiving ANY hbh/dst options in order to avoid
1727 			 *  overhead of parsing options in the kernel.
1728 			 */
1729 			case IPV6_RECVHOPOPTS:
1730 			case IPV6_RECVDSTOPTS:
1731 			case IPV6_RECVRTHDRDSTOPTS:
1732 				if (td != NULL) {
1733 					error = priv_check(td,
1734 					    PRIV_NETINET_SETHDROPTS);
1735 					if (error)
1736 						break;
1737 				}
1738 				/* FALLTHROUGH */
1739 			case IPV6_UNICAST_HOPS:
1740 			case IPV6_HOPLIMIT:
1741 
1742 			case IPV6_RECVPKTINFO:
1743 			case IPV6_RECVHOPLIMIT:
1744 			case IPV6_RECVRTHDR:
1745 			case IPV6_RECVPATHMTU:
1746 			case IPV6_RECVTCLASS:
1747 			case IPV6_RECVFLOWID:
1748 #ifdef	RSS
1749 			case IPV6_RECVRSSBUCKETID:
1750 #endif
1751 			case IPV6_V6ONLY:
1752 			case IPV6_AUTOFLOWLABEL:
1753 			case IPV6_ORIGDSTADDR:
1754 			case IPV6_BINDANY:
1755 			case IPV6_BINDMULTI:
1756 #ifdef	RSS
1757 			case IPV6_RSS_LISTEN_BUCKET:
1758 #endif
1759 			case IPV6_VLAN_PCP:
1760 				if (optname == IPV6_BINDANY && td != NULL) {
1761 					error = priv_check(td,
1762 					    PRIV_NETINET_BINDANY);
1763 					if (error)
1764 						break;
1765 				}
1766 
1767 				if (optlen != sizeof(int)) {
1768 					error = EINVAL;
1769 					break;
1770 				}
1771 				error = sooptcopyin(sopt, &optval,
1772 					sizeof optval, sizeof optval);
1773 				if (error)
1774 					break;
1775 				switch (optname) {
1776 				case IPV6_UNICAST_HOPS:
1777 					if (optval < -1 || optval >= 256)
1778 						error = EINVAL;
1779 					else {
1780 						/* -1 = kernel default */
1781 						inp->in6p_hops = optval;
1782 						if ((inp->inp_vflag &
1783 						     INP_IPV4) != 0)
1784 							inp->inp_ip_ttl = optval;
1785 					}
1786 					break;
1787 #define OPTSET(bit) \
1788 do { \
1789 	INP_WLOCK(inp); \
1790 	if (optval) \
1791 		inp->inp_flags |= (bit); \
1792 	else \
1793 		inp->inp_flags &= ~(bit); \
1794 	INP_WUNLOCK(inp); \
1795 } while (/*CONSTCOND*/ 0)
1796 #define OPTSET2292(bit) \
1797 do { \
1798 	INP_WLOCK(inp); \
1799 	inp->inp_flags |= IN6P_RFC2292; \
1800 	if (optval) \
1801 		inp->inp_flags |= (bit); \
1802 	else \
1803 		inp->inp_flags &= ~(bit); \
1804 	INP_WUNLOCK(inp); \
1805 } while (/*CONSTCOND*/ 0)
1806 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1807 
1808 #define OPTSET2_N(bit, val) do {					\
1809 	if (val)							\
1810 		inp->inp_flags2 |= bit;					\
1811 	else								\
1812 		inp->inp_flags2 &= ~bit;				\
1813 } while (0)
1814 #define OPTSET2(bit, val) do {						\
1815 	INP_WLOCK(inp);							\
1816 	OPTSET2_N(bit, val);						\
1817 	INP_WUNLOCK(inp);						\
1818 } while (0)
1819 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1820 #define OPTSET2292_EXCLUSIVE(bit)					\
1821 do {									\
1822 	INP_WLOCK(inp);							\
1823 	if (OPTBIT(IN6P_RFC2292)) {					\
1824 		error = EINVAL;						\
1825 	} else {							\
1826 		if (optval)						\
1827 			inp->inp_flags |= (bit);			\
1828 		else							\
1829 			inp->inp_flags &= ~(bit);			\
1830 	}								\
1831 	INP_WUNLOCK(inp);						\
1832 } while (/*CONSTCOND*/ 0)
1833 
1834 				case IPV6_RECVPKTINFO:
1835 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1836 					break;
1837 
1838 				case IPV6_HOPLIMIT:
1839 				{
1840 					struct ip6_pktopts **optp;
1841 
1842 					/* cannot mix with RFC2292 */
1843 					if (OPTBIT(IN6P_RFC2292)) {
1844 						error = EINVAL;
1845 						break;
1846 					}
1847 					INP_WLOCK(inp);
1848 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1849 						INP_WUNLOCK(inp);
1850 						return (ECONNRESET);
1851 					}
1852 					optp = &inp->in6p_outputopts;
1853 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1854 					    (u_char *)&optval, sizeof(optval),
1855 					    optp, (td != NULL) ? td->td_ucred :
1856 					    NULL, uproto);
1857 					INP_WUNLOCK(inp);
1858 					break;
1859 				}
1860 
1861 				case IPV6_RECVHOPLIMIT:
1862 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1863 					break;
1864 
1865 				case IPV6_RECVHOPOPTS:
1866 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1867 					break;
1868 
1869 				case IPV6_RECVDSTOPTS:
1870 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1871 					break;
1872 
1873 				case IPV6_RECVRTHDRDSTOPTS:
1874 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1875 					break;
1876 
1877 				case IPV6_RECVRTHDR:
1878 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1879 					break;
1880 
1881 				case IPV6_RECVPATHMTU:
1882 					/*
1883 					 * We ignore this option for TCP
1884 					 * sockets.
1885 					 * (RFC3542 leaves this case
1886 					 * unspecified.)
1887 					 */
1888 					if (uproto != IPPROTO_TCP)
1889 						OPTSET(IN6P_MTU);
1890 					break;
1891 
1892 				case IPV6_RECVFLOWID:
1893 					OPTSET2(INP_RECVFLOWID, optval);
1894 					break;
1895 
1896 #ifdef	RSS
1897 				case IPV6_RECVRSSBUCKETID:
1898 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1899 					break;
1900 #endif
1901 
1902 				case IPV6_V6ONLY:
1903 					INP_WLOCK(inp);
1904 					if (inp->inp_lport ||
1905 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1906 						/*
1907 						 * The socket is already bound.
1908 						 */
1909 						INP_WUNLOCK(inp);
1910 						error = EINVAL;
1911 						break;
1912 					}
1913 					if (optval) {
1914 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1915 						inp->inp_vflag &= ~INP_IPV4;
1916 					} else {
1917 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1918 						inp->inp_vflag |= INP_IPV4;
1919 					}
1920 					INP_WUNLOCK(inp);
1921 					break;
1922 				case IPV6_RECVTCLASS:
1923 					/* cannot mix with RFC2292 XXX */
1924 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1925 					break;
1926 				case IPV6_AUTOFLOWLABEL:
1927 					OPTSET(IN6P_AUTOFLOWLABEL);
1928 					break;
1929 
1930 				case IPV6_ORIGDSTADDR:
1931 					OPTSET2(INP_ORIGDSTADDR, optval);
1932 					break;
1933 				case IPV6_BINDANY:
1934 					OPTSET(INP_BINDANY);
1935 					break;
1936 
1937 				case IPV6_BINDMULTI:
1938 					OPTSET2(INP_BINDMULTI, optval);
1939 					break;
1940 #ifdef	RSS
1941 				case IPV6_RSS_LISTEN_BUCKET:
1942 					if ((optval >= 0) &&
1943 					    (optval < rss_getnumbuckets())) {
1944 						INP_WLOCK(inp);
1945 						inp->inp_rss_listen_bucket = optval;
1946 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1947 						INP_WUNLOCK(inp);
1948 					} else {
1949 						error = EINVAL;
1950 					}
1951 					break;
1952 #endif
1953 				case IPV6_VLAN_PCP:
1954 					if ((optval >= -1) && (optval <=
1955 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1956 						if (optval == -1) {
1957 							INP_WLOCK(inp);
1958 							inp->inp_flags2 &=
1959 							    ~(INP_2PCP_SET |
1960 							    INP_2PCP_MASK);
1961 							INP_WUNLOCK(inp);
1962 						} else {
1963 							INP_WLOCK(inp);
1964 							inp->inp_flags2 |=
1965 							    INP_2PCP_SET;
1966 							inp->inp_flags2 &=
1967 							    ~INP_2PCP_MASK;
1968 							inp->inp_flags2 |=
1969 							    optval <<
1970 							    INP_2PCP_SHIFT;
1971 							INP_WUNLOCK(inp);
1972 						}
1973 					} else
1974 						error = EINVAL;
1975 					break;
1976 				}
1977 				break;
1978 
1979 			case IPV6_TCLASS:
1980 			case IPV6_DONTFRAG:
1981 			case IPV6_USE_MIN_MTU:
1982 			case IPV6_PREFER_TEMPADDR:
1983 				if (optlen != sizeof(optval)) {
1984 					error = EINVAL;
1985 					break;
1986 				}
1987 				error = sooptcopyin(sopt, &optval,
1988 					sizeof optval, sizeof optval);
1989 				if (error)
1990 					break;
1991 				{
1992 					struct ip6_pktopts **optp;
1993 					INP_WLOCK(inp);
1994 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1995 						INP_WUNLOCK(inp);
1996 						return (ECONNRESET);
1997 					}
1998 					optp = &inp->in6p_outputopts;
1999 					error = ip6_pcbopt(optname,
2000 					    (u_char *)&optval, sizeof(optval),
2001 					    optp, (td != NULL) ? td->td_ucred :
2002 					    NULL, uproto);
2003 					INP_WUNLOCK(inp);
2004 					break;
2005 				}
2006 
2007 			case IPV6_2292PKTINFO:
2008 			case IPV6_2292HOPLIMIT:
2009 			case IPV6_2292HOPOPTS:
2010 			case IPV6_2292DSTOPTS:
2011 			case IPV6_2292RTHDR:
2012 				/* RFC 2292 */
2013 				if (optlen != sizeof(int)) {
2014 					error = EINVAL;
2015 					break;
2016 				}
2017 				error = sooptcopyin(sopt, &optval,
2018 					sizeof optval, sizeof optval);
2019 				if (error)
2020 					break;
2021 				switch (optname) {
2022 				case IPV6_2292PKTINFO:
2023 					OPTSET2292(IN6P_PKTINFO);
2024 					break;
2025 				case IPV6_2292HOPLIMIT:
2026 					OPTSET2292(IN6P_HOPLIMIT);
2027 					break;
2028 				case IPV6_2292HOPOPTS:
2029 					/*
2030 					 * Check super-user privilege.
2031 					 * See comments for IPV6_RECVHOPOPTS.
2032 					 */
2033 					if (td != NULL) {
2034 						error = priv_check(td,
2035 						    PRIV_NETINET_SETHDROPTS);
2036 						if (error)
2037 							return (error);
2038 					}
2039 					OPTSET2292(IN6P_HOPOPTS);
2040 					break;
2041 				case IPV6_2292DSTOPTS:
2042 					if (td != NULL) {
2043 						error = priv_check(td,
2044 						    PRIV_NETINET_SETHDROPTS);
2045 						if (error)
2046 							return (error);
2047 					}
2048 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2049 					break;
2050 				case IPV6_2292RTHDR:
2051 					OPTSET2292(IN6P_RTHDR);
2052 					break;
2053 				}
2054 				break;
2055 			case IPV6_PKTINFO:
2056 			case IPV6_HOPOPTS:
2057 			case IPV6_RTHDR:
2058 			case IPV6_DSTOPTS:
2059 			case IPV6_RTHDRDSTOPTS:
2060 			case IPV6_NEXTHOP:
2061 			{
2062 				/* new advanced API (RFC3542) */
2063 				u_char *optbuf;
2064 				u_char optbuf_storage[MCLBYTES];
2065 				int optlen;
2066 				struct ip6_pktopts **optp;
2067 
2068 				/* cannot mix with RFC2292 */
2069 				if (OPTBIT(IN6P_RFC2292)) {
2070 					error = EINVAL;
2071 					break;
2072 				}
2073 
2074 				/*
2075 				 * We only ensure valsize is not too large
2076 				 * here.  Further validation will be done
2077 				 * later.
2078 				 */
2079 				error = sooptcopyin(sopt, optbuf_storage,
2080 				    sizeof(optbuf_storage), 0);
2081 				if (error)
2082 					break;
2083 				optlen = sopt->sopt_valsize;
2084 				optbuf = optbuf_storage;
2085 				INP_WLOCK(inp);
2086 				if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2087 					INP_WUNLOCK(inp);
2088 					return (ECONNRESET);
2089 				}
2090 				optp = &inp->in6p_outputopts;
2091 				error = ip6_pcbopt(optname, optbuf, optlen,
2092 				    optp, (td != NULL) ? td->td_ucred : NULL,
2093 				    uproto);
2094 				INP_WUNLOCK(inp);
2095 				break;
2096 			}
2097 #undef OPTSET
2098 
2099 			case IPV6_MULTICAST_IF:
2100 			case IPV6_MULTICAST_HOPS:
2101 			case IPV6_MULTICAST_LOOP:
2102 			case IPV6_JOIN_GROUP:
2103 			case IPV6_LEAVE_GROUP:
2104 			case IPV6_MSFILTER:
2105 			case MCAST_BLOCK_SOURCE:
2106 			case MCAST_UNBLOCK_SOURCE:
2107 			case MCAST_JOIN_GROUP:
2108 			case MCAST_LEAVE_GROUP:
2109 			case MCAST_JOIN_SOURCE_GROUP:
2110 			case MCAST_LEAVE_SOURCE_GROUP:
2111 				error = ip6_setmoptions(inp, sopt);
2112 				break;
2113 
2114 			case IPV6_PORTRANGE:
2115 				error = sooptcopyin(sopt, &optval,
2116 				    sizeof optval, sizeof optval);
2117 				if (error)
2118 					break;
2119 
2120 				INP_WLOCK(inp);
2121 				switch (optval) {
2122 				case IPV6_PORTRANGE_DEFAULT:
2123 					inp->inp_flags &= ~(INP_LOWPORT);
2124 					inp->inp_flags &= ~(INP_HIGHPORT);
2125 					break;
2126 
2127 				case IPV6_PORTRANGE_HIGH:
2128 					inp->inp_flags &= ~(INP_LOWPORT);
2129 					inp->inp_flags |= INP_HIGHPORT;
2130 					break;
2131 
2132 				case IPV6_PORTRANGE_LOW:
2133 					inp->inp_flags &= ~(INP_HIGHPORT);
2134 					inp->inp_flags |= INP_LOWPORT;
2135 					break;
2136 
2137 				default:
2138 					error = EINVAL;
2139 					break;
2140 				}
2141 				INP_WUNLOCK(inp);
2142 				break;
2143 
2144 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2145 			case IPV6_IPSEC_POLICY:
2146 				if (IPSEC_ENABLED(ipv6)) {
2147 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2148 					break;
2149 				}
2150 				/* FALLTHROUGH */
2151 #endif /* IPSEC */
2152 
2153 			default:
2154 				error = ENOPROTOOPT;
2155 				break;
2156 			}
2157 			break;
2158 
2159 		case SOPT_GET:
2160 			switch (optname) {
2161 			case IPV6_2292PKTOPTIONS:
2162 #ifdef IPV6_PKTOPTIONS
2163 			case IPV6_PKTOPTIONS:
2164 #endif
2165 				/*
2166 				 * RFC3542 (effectively) deprecated the
2167 				 * semantics of the 2292-style pktoptions.
2168 				 * Since it was not reliable in nature (i.e.,
2169 				 * applications had to expect the lack of some
2170 				 * information after all), it would make sense
2171 				 * to simplify this part by always returning
2172 				 * empty data.
2173 				 */
2174 				sopt->sopt_valsize = 0;
2175 				break;
2176 
2177 			case IPV6_RECVHOPOPTS:
2178 			case IPV6_RECVDSTOPTS:
2179 			case IPV6_RECVRTHDRDSTOPTS:
2180 			case IPV6_UNICAST_HOPS:
2181 			case IPV6_RECVPKTINFO:
2182 			case IPV6_RECVHOPLIMIT:
2183 			case IPV6_RECVRTHDR:
2184 			case IPV6_RECVPATHMTU:
2185 
2186 			case IPV6_V6ONLY:
2187 			case IPV6_PORTRANGE:
2188 			case IPV6_RECVTCLASS:
2189 			case IPV6_AUTOFLOWLABEL:
2190 			case IPV6_BINDANY:
2191 			case IPV6_FLOWID:
2192 			case IPV6_FLOWTYPE:
2193 			case IPV6_RECVFLOWID:
2194 #ifdef	RSS
2195 			case IPV6_RSSBUCKETID:
2196 			case IPV6_RECVRSSBUCKETID:
2197 #endif
2198 			case IPV6_BINDMULTI:
2199 			case IPV6_VLAN_PCP:
2200 				switch (optname) {
2201 				case IPV6_RECVHOPOPTS:
2202 					optval = OPTBIT(IN6P_HOPOPTS);
2203 					break;
2204 
2205 				case IPV6_RECVDSTOPTS:
2206 					optval = OPTBIT(IN6P_DSTOPTS);
2207 					break;
2208 
2209 				case IPV6_RECVRTHDRDSTOPTS:
2210 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2211 					break;
2212 
2213 				case IPV6_UNICAST_HOPS:
2214 					optval = inp->in6p_hops;
2215 					break;
2216 
2217 				case IPV6_RECVPKTINFO:
2218 					optval = OPTBIT(IN6P_PKTINFO);
2219 					break;
2220 
2221 				case IPV6_RECVHOPLIMIT:
2222 					optval = OPTBIT(IN6P_HOPLIMIT);
2223 					break;
2224 
2225 				case IPV6_RECVRTHDR:
2226 					optval = OPTBIT(IN6P_RTHDR);
2227 					break;
2228 
2229 				case IPV6_RECVPATHMTU:
2230 					optval = OPTBIT(IN6P_MTU);
2231 					break;
2232 
2233 				case IPV6_V6ONLY:
2234 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2235 					break;
2236 
2237 				case IPV6_PORTRANGE:
2238 				    {
2239 					int flags;
2240 					flags = inp->inp_flags;
2241 					if (flags & INP_HIGHPORT)
2242 						optval = IPV6_PORTRANGE_HIGH;
2243 					else if (flags & INP_LOWPORT)
2244 						optval = IPV6_PORTRANGE_LOW;
2245 					else
2246 						optval = 0;
2247 					break;
2248 				    }
2249 				case IPV6_RECVTCLASS:
2250 					optval = OPTBIT(IN6P_TCLASS);
2251 					break;
2252 
2253 				case IPV6_AUTOFLOWLABEL:
2254 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2255 					break;
2256 
2257 				case IPV6_ORIGDSTADDR:
2258 					optval = OPTBIT2(INP_ORIGDSTADDR);
2259 					break;
2260 
2261 				case IPV6_BINDANY:
2262 					optval = OPTBIT(INP_BINDANY);
2263 					break;
2264 
2265 				case IPV6_FLOWID:
2266 					optval = inp->inp_flowid;
2267 					break;
2268 
2269 				case IPV6_FLOWTYPE:
2270 					optval = inp->inp_flowtype;
2271 					break;
2272 
2273 				case IPV6_RECVFLOWID:
2274 					optval = OPTBIT2(INP_RECVFLOWID);
2275 					break;
2276 #ifdef	RSS
2277 				case IPV6_RSSBUCKETID:
2278 					retval =
2279 					    rss_hash2bucket(inp->inp_flowid,
2280 					    inp->inp_flowtype,
2281 					    &rss_bucket);
2282 					if (retval == 0)
2283 						optval = rss_bucket;
2284 					else
2285 						error = EINVAL;
2286 					break;
2287 
2288 				case IPV6_RECVRSSBUCKETID:
2289 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2290 					break;
2291 #endif
2292 
2293 				case IPV6_BINDMULTI:
2294 					optval = OPTBIT2(INP_BINDMULTI);
2295 					break;
2296 
2297 				case IPV6_VLAN_PCP:
2298 					if (OPTBIT2(INP_2PCP_SET)) {
2299 						optval = (inp->inp_flags2 &
2300 							    INP_2PCP_MASK) >>
2301 							    INP_2PCP_SHIFT;
2302 					} else {
2303 						optval = -1;
2304 					}
2305 					break;
2306 				}
2307 
2308 				if (error)
2309 					break;
2310 				error = sooptcopyout(sopt, &optval,
2311 					sizeof optval);
2312 				break;
2313 
2314 			case IPV6_PATHMTU:
2315 			{
2316 				u_long pmtu = 0;
2317 				struct ip6_mtuinfo mtuinfo;
2318 				struct in6_addr addr;
2319 
2320 				if (!(so->so_state & SS_ISCONNECTED))
2321 					return (ENOTCONN);
2322 				/*
2323 				 * XXX: we dot not consider the case of source
2324 				 * routing, or optional information to specify
2325 				 * the outgoing interface.
2326 				 * Copy faddr out of inp to avoid holding lock
2327 				 * on inp during route lookup.
2328 				 */
2329 				INP_RLOCK(inp);
2330 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2331 				INP_RUNLOCK(inp);
2332 				error = ip6_getpmtu_ctl(so->so_fibnum,
2333 				    &addr, &pmtu);
2334 				if (error)
2335 					break;
2336 				if (pmtu > IPV6_MAXPACKET)
2337 					pmtu = IPV6_MAXPACKET;
2338 
2339 				bzero(&mtuinfo, sizeof(mtuinfo));
2340 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2341 				optdata = (void *)&mtuinfo;
2342 				optdatalen = sizeof(mtuinfo);
2343 				error = sooptcopyout(sopt, optdata,
2344 				    optdatalen);
2345 				break;
2346 			}
2347 
2348 			case IPV6_2292PKTINFO:
2349 			case IPV6_2292HOPLIMIT:
2350 			case IPV6_2292HOPOPTS:
2351 			case IPV6_2292RTHDR:
2352 			case IPV6_2292DSTOPTS:
2353 				switch (optname) {
2354 				case IPV6_2292PKTINFO:
2355 					optval = OPTBIT(IN6P_PKTINFO);
2356 					break;
2357 				case IPV6_2292HOPLIMIT:
2358 					optval = OPTBIT(IN6P_HOPLIMIT);
2359 					break;
2360 				case IPV6_2292HOPOPTS:
2361 					optval = OPTBIT(IN6P_HOPOPTS);
2362 					break;
2363 				case IPV6_2292RTHDR:
2364 					optval = OPTBIT(IN6P_RTHDR);
2365 					break;
2366 				case IPV6_2292DSTOPTS:
2367 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2368 					break;
2369 				}
2370 				error = sooptcopyout(sopt, &optval,
2371 				    sizeof optval);
2372 				break;
2373 			case IPV6_PKTINFO:
2374 			case IPV6_HOPOPTS:
2375 			case IPV6_RTHDR:
2376 			case IPV6_DSTOPTS:
2377 			case IPV6_RTHDRDSTOPTS:
2378 			case IPV6_NEXTHOP:
2379 			case IPV6_TCLASS:
2380 			case IPV6_DONTFRAG:
2381 			case IPV6_USE_MIN_MTU:
2382 			case IPV6_PREFER_TEMPADDR:
2383 				error = ip6_getpcbopt(inp, optname, sopt);
2384 				break;
2385 
2386 			case IPV6_MULTICAST_IF:
2387 			case IPV6_MULTICAST_HOPS:
2388 			case IPV6_MULTICAST_LOOP:
2389 			case IPV6_MSFILTER:
2390 				error = ip6_getmoptions(inp, sopt);
2391 				break;
2392 
2393 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2394 			case IPV6_IPSEC_POLICY:
2395 				if (IPSEC_ENABLED(ipv6)) {
2396 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2397 					break;
2398 				}
2399 				/* FALLTHROUGH */
2400 #endif /* IPSEC */
2401 			default:
2402 				error = ENOPROTOOPT;
2403 				break;
2404 			}
2405 			break;
2406 		}
2407 	}
2408 	return (error);
2409 }
2410 
2411 int
2412 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2413 {
2414 	int error = 0, optval, optlen;
2415 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2416 	struct inpcb *inp = sotoinpcb(so);
2417 	int level, op, optname;
2418 
2419 	level = sopt->sopt_level;
2420 	op = sopt->sopt_dir;
2421 	optname = sopt->sopt_name;
2422 	optlen = sopt->sopt_valsize;
2423 
2424 	if (level != IPPROTO_IPV6) {
2425 		return (EINVAL);
2426 	}
2427 
2428 	switch (optname) {
2429 	case IPV6_CHECKSUM:
2430 		/*
2431 		 * For ICMPv6 sockets, no modification allowed for checksum
2432 		 * offset, permit "no change" values to help existing apps.
2433 		 *
2434 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2435 		 * for an ICMPv6 socket will fail."
2436 		 * The current behavior does not meet RFC3542.
2437 		 */
2438 		switch (op) {
2439 		case SOPT_SET:
2440 			if (optlen != sizeof(int)) {
2441 				error = EINVAL;
2442 				break;
2443 			}
2444 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2445 					    sizeof(optval));
2446 			if (error)
2447 				break;
2448 			if (optval < -1 || (optval % 2) != 0) {
2449 				/*
2450 				 * The API assumes non-negative even offset
2451 				 * values or -1 as a special value.
2452 				 */
2453 				error = EINVAL;
2454 			} else if (so->so_proto->pr_protocol ==
2455 			    IPPROTO_ICMPV6) {
2456 				if (optval != icmp6off)
2457 					error = EINVAL;
2458 			} else
2459 				inp->in6p_cksum = optval;
2460 			break;
2461 
2462 		case SOPT_GET:
2463 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2464 				optval = icmp6off;
2465 			else
2466 				optval = inp->in6p_cksum;
2467 
2468 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2469 			break;
2470 
2471 		default:
2472 			error = EINVAL;
2473 			break;
2474 		}
2475 		break;
2476 
2477 	default:
2478 		error = ENOPROTOOPT;
2479 		break;
2480 	}
2481 
2482 	return (error);
2483 }
2484 
2485 /*
2486  * Set up IP6 options in pcb for insertion in output packets or
2487  * specifying behavior of outgoing packets.
2488  */
2489 static int
2490 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2491     struct socket *so, struct sockopt *sopt)
2492 {
2493 	struct ip6_pktopts *opt = *pktopt;
2494 	int error = 0;
2495 	struct thread *td = sopt->sopt_td;
2496 	struct epoch_tracker et;
2497 
2498 	/* turn off any old options. */
2499 	if (opt) {
2500 #ifdef DIAGNOSTIC
2501 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2502 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2503 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2504 			printf("ip6_pcbopts: all specified options are cleared.\n");
2505 #endif
2506 		ip6_clearpktopts(opt, -1);
2507 	} else {
2508 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2509 		if (opt == NULL)
2510 			return (ENOMEM);
2511 	}
2512 	*pktopt = NULL;
2513 
2514 	if (!m || m->m_len == 0) {
2515 		/*
2516 		 * Only turning off any previous options, regardless of
2517 		 * whether the opt is just created or given.
2518 		 */
2519 		free(opt, M_IP6OPT);
2520 		return (0);
2521 	}
2522 
2523 	/*  set options specified by user. */
2524 	NET_EPOCH_ENTER(et);
2525 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2526 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2527 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2528 		free(opt, M_IP6OPT);
2529 		NET_EPOCH_EXIT(et);
2530 		return (error);
2531 	}
2532 	NET_EPOCH_EXIT(et);
2533 	*pktopt = opt;
2534 	return (0);
2535 }
2536 
2537 /*
2538  * initialize ip6_pktopts.  beware that there are non-zero default values in
2539  * the struct.
2540  */
2541 void
2542 ip6_initpktopts(struct ip6_pktopts *opt)
2543 {
2544 
2545 	bzero(opt, sizeof(*opt));
2546 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2547 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2548 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2549 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2550 }
2551 
2552 static int
2553 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2554     struct ucred *cred, int uproto)
2555 {
2556 	struct epoch_tracker et;
2557 	struct ip6_pktopts *opt;
2558 	int ret;
2559 
2560 	if (*pktopt == NULL) {
2561 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2562 		    M_NOWAIT);
2563 		if (*pktopt == NULL)
2564 			return (ENOBUFS);
2565 		ip6_initpktopts(*pktopt);
2566 	}
2567 	opt = *pktopt;
2568 
2569 	NET_EPOCH_ENTER(et);
2570 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2571 	NET_EPOCH_EXIT(et);
2572 
2573 	return (ret);
2574 }
2575 
2576 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2577 	if (pktopt && pktopt->field) {						\
2578 		INP_RUNLOCK(inp);						\
2579 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2580 		malloc_optdata = true;						\
2581 		INP_RLOCK(inp);							\
2582 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2583 			INP_RUNLOCK(inp);					\
2584 			free(optdata, M_TEMP);					\
2585 			return (ECONNRESET);					\
2586 		}								\
2587 		pktopt = inp->in6p_outputopts;					\
2588 		if (pktopt && pktopt->field) {					\
2589 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2590 			bcopy(pktopt->field, optdata, optdatalen);		\
2591 		} else {							\
2592 			free(optdata, M_TEMP);					\
2593 			optdata = NULL;						\
2594 			malloc_optdata = false;					\
2595 		}								\
2596 	}									\
2597 } while(0)
2598 
2599 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2600 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2601 
2602 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2603 	pktopt->field->sa_len)
2604 
2605 static int
2606 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2607 {
2608 	void *optdata = NULL;
2609 	bool malloc_optdata = false;
2610 	int optdatalen = 0;
2611 	int error = 0;
2612 	struct in6_pktinfo null_pktinfo;
2613 	int deftclass = 0, on;
2614 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2615 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2616 	struct ip6_pktopts *pktopt;
2617 
2618 	INP_RLOCK(inp);
2619 	pktopt = inp->in6p_outputopts;
2620 
2621 	switch (optname) {
2622 	case IPV6_PKTINFO:
2623 		optdata = (void *)&null_pktinfo;
2624 		if (pktopt && pktopt->ip6po_pktinfo) {
2625 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2626 			    sizeof(null_pktinfo));
2627 			in6_clearscope(&null_pktinfo.ipi6_addr);
2628 		} else {
2629 			/* XXX: we don't have to do this every time... */
2630 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2631 		}
2632 		optdatalen = sizeof(struct in6_pktinfo);
2633 		break;
2634 	case IPV6_TCLASS:
2635 		if (pktopt && pktopt->ip6po_tclass >= 0)
2636 			deftclass = pktopt->ip6po_tclass;
2637 		optdata = (void *)&deftclass;
2638 		optdatalen = sizeof(int);
2639 		break;
2640 	case IPV6_HOPOPTS:
2641 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2642 		break;
2643 	case IPV6_RTHDR:
2644 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2645 		break;
2646 	case IPV6_RTHDRDSTOPTS:
2647 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2648 		break;
2649 	case IPV6_DSTOPTS:
2650 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2651 		break;
2652 	case IPV6_NEXTHOP:
2653 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2654 		break;
2655 	case IPV6_USE_MIN_MTU:
2656 		if (pktopt)
2657 			defminmtu = pktopt->ip6po_minmtu;
2658 		optdata = (void *)&defminmtu;
2659 		optdatalen = sizeof(int);
2660 		break;
2661 	case IPV6_DONTFRAG:
2662 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2663 			on = 1;
2664 		else
2665 			on = 0;
2666 		optdata = (void *)&on;
2667 		optdatalen = sizeof(on);
2668 		break;
2669 	case IPV6_PREFER_TEMPADDR:
2670 		if (pktopt)
2671 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2672 		optdata = (void *)&defpreftemp;
2673 		optdatalen = sizeof(int);
2674 		break;
2675 	default:		/* should not happen */
2676 #ifdef DIAGNOSTIC
2677 		panic("ip6_getpcbopt: unexpected option\n");
2678 #endif
2679 		INP_RUNLOCK(inp);
2680 		return (ENOPROTOOPT);
2681 	}
2682 	INP_RUNLOCK(inp);
2683 
2684 	error = sooptcopyout(sopt, optdata, optdatalen);
2685 	if (malloc_optdata)
2686 		free(optdata, M_TEMP);
2687 
2688 	return (error);
2689 }
2690 
2691 void
2692 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2693 {
2694 	if (pktopt == NULL)
2695 		return;
2696 
2697 	if (optname == -1 || optname == IPV6_PKTINFO) {
2698 		if (pktopt->ip6po_pktinfo)
2699 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2700 		pktopt->ip6po_pktinfo = NULL;
2701 	}
2702 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2703 		pktopt->ip6po_hlim = -1;
2704 	if (optname == -1 || optname == IPV6_TCLASS)
2705 		pktopt->ip6po_tclass = -1;
2706 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2707 		if (pktopt->ip6po_nextroute.ro_nh) {
2708 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2709 			pktopt->ip6po_nextroute.ro_nh = NULL;
2710 		}
2711 		if (pktopt->ip6po_nexthop)
2712 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2713 		pktopt->ip6po_nexthop = NULL;
2714 	}
2715 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2716 		if (pktopt->ip6po_hbh)
2717 			free(pktopt->ip6po_hbh, M_IP6OPT);
2718 		pktopt->ip6po_hbh = NULL;
2719 	}
2720 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2721 		if (pktopt->ip6po_dest1)
2722 			free(pktopt->ip6po_dest1, M_IP6OPT);
2723 		pktopt->ip6po_dest1 = NULL;
2724 	}
2725 	if (optname == -1 || optname == IPV6_RTHDR) {
2726 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2727 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2728 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2729 		if (pktopt->ip6po_route.ro_nh) {
2730 			NH_FREE(pktopt->ip6po_route.ro_nh);
2731 			pktopt->ip6po_route.ro_nh = NULL;
2732 		}
2733 	}
2734 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2735 		if (pktopt->ip6po_dest2)
2736 			free(pktopt->ip6po_dest2, M_IP6OPT);
2737 		pktopt->ip6po_dest2 = NULL;
2738 	}
2739 }
2740 
2741 #define PKTOPT_EXTHDRCPY(type) \
2742 do {\
2743 	if (src->type) {\
2744 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2745 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2746 		if (dst->type == NULL)\
2747 			goto bad;\
2748 		bcopy(src->type, dst->type, hlen);\
2749 	}\
2750 } while (/*CONSTCOND*/ 0)
2751 
2752 static int
2753 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2754 {
2755 	if (dst == NULL || src == NULL)  {
2756 		printf("ip6_clearpktopts: invalid argument\n");
2757 		return (EINVAL);
2758 	}
2759 
2760 	dst->ip6po_hlim = src->ip6po_hlim;
2761 	dst->ip6po_tclass = src->ip6po_tclass;
2762 	dst->ip6po_flags = src->ip6po_flags;
2763 	dst->ip6po_minmtu = src->ip6po_minmtu;
2764 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2765 	if (src->ip6po_pktinfo) {
2766 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2767 		    M_IP6OPT, canwait);
2768 		if (dst->ip6po_pktinfo == NULL)
2769 			goto bad;
2770 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2771 	}
2772 	if (src->ip6po_nexthop) {
2773 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2774 		    M_IP6OPT, canwait);
2775 		if (dst->ip6po_nexthop == NULL)
2776 			goto bad;
2777 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2778 		    src->ip6po_nexthop->sa_len);
2779 	}
2780 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2781 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2782 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2783 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2784 	return (0);
2785 
2786   bad:
2787 	ip6_clearpktopts(dst, -1);
2788 	return (ENOBUFS);
2789 }
2790 #undef PKTOPT_EXTHDRCPY
2791 
2792 struct ip6_pktopts *
2793 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2794 {
2795 	int error;
2796 	struct ip6_pktopts *dst;
2797 
2798 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2799 	if (dst == NULL)
2800 		return (NULL);
2801 	ip6_initpktopts(dst);
2802 
2803 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2804 		free(dst, M_IP6OPT);
2805 		return (NULL);
2806 	}
2807 
2808 	return (dst);
2809 }
2810 
2811 void
2812 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2813 {
2814 	if (pktopt == NULL)
2815 		return;
2816 
2817 	ip6_clearpktopts(pktopt, -1);
2818 
2819 	free(pktopt, M_IP6OPT);
2820 }
2821 
2822 /*
2823  * Set IPv6 outgoing packet options based on advanced API.
2824  */
2825 int
2826 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2827     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2828 {
2829 	struct cmsghdr *cm = NULL;
2830 
2831 	if (control == NULL || opt == NULL)
2832 		return (EINVAL);
2833 
2834 	/*
2835 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2836 	 * are in the network epoch here.
2837 	 */
2838 	NET_EPOCH_ASSERT();
2839 
2840 	ip6_initpktopts(opt);
2841 	if (stickyopt) {
2842 		int error;
2843 
2844 		/*
2845 		 * If stickyopt is provided, make a local copy of the options
2846 		 * for this particular packet, then override them by ancillary
2847 		 * objects.
2848 		 * XXX: copypktopts() does not copy the cached route to a next
2849 		 * hop (if any).  This is not very good in terms of efficiency,
2850 		 * but we can allow this since this option should be rarely
2851 		 * used.
2852 		 */
2853 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2854 			return (error);
2855 	}
2856 
2857 	/*
2858 	 * XXX: Currently, we assume all the optional information is stored
2859 	 * in a single mbuf.
2860 	 */
2861 	if (control->m_next)
2862 		return (EINVAL);
2863 
2864 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2865 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2866 		int error;
2867 
2868 		if (control->m_len < CMSG_LEN(0))
2869 			return (EINVAL);
2870 
2871 		cm = mtod(control, struct cmsghdr *);
2872 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2873 			return (EINVAL);
2874 		if (cm->cmsg_level != IPPROTO_IPV6)
2875 			continue;
2876 
2877 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2878 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2879 		if (error)
2880 			return (error);
2881 	}
2882 
2883 	return (0);
2884 }
2885 
2886 /*
2887  * Set a particular packet option, as a sticky option or an ancillary data
2888  * item.  "len" can be 0 only when it's a sticky option.
2889  * We have 4 cases of combination of "sticky" and "cmsg":
2890  * "sticky=0, cmsg=0": impossible
2891  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2892  * "sticky=1, cmsg=0": RFC3542 socket option
2893  * "sticky=1, cmsg=1": RFC2292 socket option
2894  */
2895 static int
2896 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2897     struct ucred *cred, int sticky, int cmsg, int uproto)
2898 {
2899 	int minmtupolicy, preftemp;
2900 	int error;
2901 
2902 	NET_EPOCH_ASSERT();
2903 
2904 	if (!sticky && !cmsg) {
2905 #ifdef DIAGNOSTIC
2906 		printf("ip6_setpktopt: impossible case\n");
2907 #endif
2908 		return (EINVAL);
2909 	}
2910 
2911 	/*
2912 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2913 	 * not be specified in the context of RFC3542.  Conversely,
2914 	 * RFC3542 types should not be specified in the context of RFC2292.
2915 	 */
2916 	if (!cmsg) {
2917 		switch (optname) {
2918 		case IPV6_2292PKTINFO:
2919 		case IPV6_2292HOPLIMIT:
2920 		case IPV6_2292NEXTHOP:
2921 		case IPV6_2292HOPOPTS:
2922 		case IPV6_2292DSTOPTS:
2923 		case IPV6_2292RTHDR:
2924 		case IPV6_2292PKTOPTIONS:
2925 			return (ENOPROTOOPT);
2926 		}
2927 	}
2928 	if (sticky && cmsg) {
2929 		switch (optname) {
2930 		case IPV6_PKTINFO:
2931 		case IPV6_HOPLIMIT:
2932 		case IPV6_NEXTHOP:
2933 		case IPV6_HOPOPTS:
2934 		case IPV6_DSTOPTS:
2935 		case IPV6_RTHDRDSTOPTS:
2936 		case IPV6_RTHDR:
2937 		case IPV6_USE_MIN_MTU:
2938 		case IPV6_DONTFRAG:
2939 		case IPV6_TCLASS:
2940 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2941 			return (ENOPROTOOPT);
2942 		}
2943 	}
2944 
2945 	switch (optname) {
2946 	case IPV6_2292PKTINFO:
2947 	case IPV6_PKTINFO:
2948 	{
2949 		struct ifnet *ifp = NULL;
2950 		struct in6_pktinfo *pktinfo;
2951 
2952 		if (len != sizeof(struct in6_pktinfo))
2953 			return (EINVAL);
2954 
2955 		pktinfo = (struct in6_pktinfo *)buf;
2956 
2957 		/*
2958 		 * An application can clear any sticky IPV6_PKTINFO option by
2959 		 * doing a "regular" setsockopt with ipi6_addr being
2960 		 * in6addr_any and ipi6_ifindex being zero.
2961 		 * [RFC 3542, Section 6]
2962 		 */
2963 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2964 		    pktinfo->ipi6_ifindex == 0 &&
2965 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2966 			ip6_clearpktopts(opt, optname);
2967 			break;
2968 		}
2969 
2970 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2971 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2972 			return (EINVAL);
2973 		}
2974 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2975 			return (EINVAL);
2976 		/* validate the interface index if specified. */
2977 		if (pktinfo->ipi6_ifindex) {
2978 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2979 			if (ifp == NULL)
2980 				return (ENXIO);
2981 		}
2982 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2983 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2984 			return (ENETDOWN);
2985 
2986 		if (ifp != NULL &&
2987 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2988 			struct in6_ifaddr *ia;
2989 
2990 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2991 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2992 			if (ia == NULL)
2993 				return (EADDRNOTAVAIL);
2994 			ifa_free(&ia->ia_ifa);
2995 		}
2996 		/*
2997 		 * We store the address anyway, and let in6_selectsrc()
2998 		 * validate the specified address.  This is because ipi6_addr
2999 		 * may not have enough information about its scope zone, and
3000 		 * we may need additional information (such as outgoing
3001 		 * interface or the scope zone of a destination address) to
3002 		 * disambiguate the scope.
3003 		 * XXX: the delay of the validation may confuse the
3004 		 * application when it is used as a sticky option.
3005 		 */
3006 		if (opt->ip6po_pktinfo == NULL) {
3007 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3008 			    M_IP6OPT, M_NOWAIT);
3009 			if (opt->ip6po_pktinfo == NULL)
3010 				return (ENOBUFS);
3011 		}
3012 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3013 		break;
3014 	}
3015 
3016 	case IPV6_2292HOPLIMIT:
3017 	case IPV6_HOPLIMIT:
3018 	{
3019 		int *hlimp;
3020 
3021 		/*
3022 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3023 		 * to simplify the ordering among hoplimit options.
3024 		 */
3025 		if (optname == IPV6_HOPLIMIT && sticky)
3026 			return (ENOPROTOOPT);
3027 
3028 		if (len != sizeof(int))
3029 			return (EINVAL);
3030 		hlimp = (int *)buf;
3031 		if (*hlimp < -1 || *hlimp > 255)
3032 			return (EINVAL);
3033 
3034 		opt->ip6po_hlim = *hlimp;
3035 		break;
3036 	}
3037 
3038 	case IPV6_TCLASS:
3039 	{
3040 		int tclass;
3041 
3042 		if (len != sizeof(int))
3043 			return (EINVAL);
3044 		tclass = *(int *)buf;
3045 		if (tclass < -1 || tclass > 255)
3046 			return (EINVAL);
3047 
3048 		opt->ip6po_tclass = tclass;
3049 		break;
3050 	}
3051 
3052 	case IPV6_2292NEXTHOP:
3053 	case IPV6_NEXTHOP:
3054 		if (cred != NULL) {
3055 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3056 			if (error)
3057 				return (error);
3058 		}
3059 
3060 		if (len == 0) {	/* just remove the option */
3061 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3062 			break;
3063 		}
3064 
3065 		/* check if cmsg_len is large enough for sa_len */
3066 		if (len < sizeof(struct sockaddr) || len < *buf)
3067 			return (EINVAL);
3068 
3069 		switch (((struct sockaddr *)buf)->sa_family) {
3070 		case AF_INET6:
3071 		{
3072 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3073 			int error;
3074 
3075 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3076 				return (EINVAL);
3077 
3078 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3079 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3080 				return (EINVAL);
3081 			}
3082 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3083 			    != 0) {
3084 				return (error);
3085 			}
3086 			break;
3087 		}
3088 		case AF_LINK:	/* should eventually be supported */
3089 		default:
3090 			return (EAFNOSUPPORT);
3091 		}
3092 
3093 		/* turn off the previous option, then set the new option. */
3094 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3095 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3096 		if (opt->ip6po_nexthop == NULL)
3097 			return (ENOBUFS);
3098 		bcopy(buf, opt->ip6po_nexthop, *buf);
3099 		break;
3100 
3101 	case IPV6_2292HOPOPTS:
3102 	case IPV6_HOPOPTS:
3103 	{
3104 		struct ip6_hbh *hbh;
3105 		int hbhlen;
3106 
3107 		/*
3108 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3109 		 * options, since per-option restriction has too much
3110 		 * overhead.
3111 		 */
3112 		if (cred != NULL) {
3113 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3114 			if (error)
3115 				return (error);
3116 		}
3117 
3118 		if (len == 0) {
3119 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3120 			break;	/* just remove the option */
3121 		}
3122 
3123 		/* message length validation */
3124 		if (len < sizeof(struct ip6_hbh))
3125 			return (EINVAL);
3126 		hbh = (struct ip6_hbh *)buf;
3127 		hbhlen = (hbh->ip6h_len + 1) << 3;
3128 		if (len != hbhlen)
3129 			return (EINVAL);
3130 
3131 		/* turn off the previous option, then set the new option. */
3132 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3133 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3134 		if (opt->ip6po_hbh == NULL)
3135 			return (ENOBUFS);
3136 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3137 
3138 		break;
3139 	}
3140 
3141 	case IPV6_2292DSTOPTS:
3142 	case IPV6_DSTOPTS:
3143 	case IPV6_RTHDRDSTOPTS:
3144 	{
3145 		struct ip6_dest *dest, **newdest = NULL;
3146 		int destlen;
3147 
3148 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3149 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3150 			if (error)
3151 				return (error);
3152 		}
3153 
3154 		if (len == 0) {
3155 			ip6_clearpktopts(opt, optname);
3156 			break;	/* just remove the option */
3157 		}
3158 
3159 		/* message length validation */
3160 		if (len < sizeof(struct ip6_dest))
3161 			return (EINVAL);
3162 		dest = (struct ip6_dest *)buf;
3163 		destlen = (dest->ip6d_len + 1) << 3;
3164 		if (len != destlen)
3165 			return (EINVAL);
3166 
3167 		/*
3168 		 * Determine the position that the destination options header
3169 		 * should be inserted; before or after the routing header.
3170 		 */
3171 		switch (optname) {
3172 		case IPV6_2292DSTOPTS:
3173 			/*
3174 			 * The old advacned API is ambiguous on this point.
3175 			 * Our approach is to determine the position based
3176 			 * according to the existence of a routing header.
3177 			 * Note, however, that this depends on the order of the
3178 			 * extension headers in the ancillary data; the 1st
3179 			 * part of the destination options header must appear
3180 			 * before the routing header in the ancillary data,
3181 			 * too.
3182 			 * RFC3542 solved the ambiguity by introducing
3183 			 * separate ancillary data or option types.
3184 			 */
3185 			if (opt->ip6po_rthdr == NULL)
3186 				newdest = &opt->ip6po_dest1;
3187 			else
3188 				newdest = &opt->ip6po_dest2;
3189 			break;
3190 		case IPV6_RTHDRDSTOPTS:
3191 			newdest = &opt->ip6po_dest1;
3192 			break;
3193 		case IPV6_DSTOPTS:
3194 			newdest = &opt->ip6po_dest2;
3195 			break;
3196 		}
3197 
3198 		/* turn off the previous option, then set the new option. */
3199 		ip6_clearpktopts(opt, optname);
3200 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3201 		if (*newdest == NULL)
3202 			return (ENOBUFS);
3203 		bcopy(dest, *newdest, destlen);
3204 
3205 		break;
3206 	}
3207 
3208 	case IPV6_2292RTHDR:
3209 	case IPV6_RTHDR:
3210 	{
3211 		struct ip6_rthdr *rth;
3212 		int rthlen;
3213 
3214 		if (len == 0) {
3215 			ip6_clearpktopts(opt, IPV6_RTHDR);
3216 			break;	/* just remove the option */
3217 		}
3218 
3219 		/* message length validation */
3220 		if (len < sizeof(struct ip6_rthdr))
3221 			return (EINVAL);
3222 		rth = (struct ip6_rthdr *)buf;
3223 		rthlen = (rth->ip6r_len + 1) << 3;
3224 		if (len != rthlen)
3225 			return (EINVAL);
3226 
3227 		switch (rth->ip6r_type) {
3228 		case IPV6_RTHDR_TYPE_0:
3229 			if (rth->ip6r_len == 0)	/* must contain one addr */
3230 				return (EINVAL);
3231 			if (rth->ip6r_len % 2) /* length must be even */
3232 				return (EINVAL);
3233 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3234 				return (EINVAL);
3235 			break;
3236 		default:
3237 			return (EINVAL);	/* not supported */
3238 		}
3239 
3240 		/* turn off the previous option */
3241 		ip6_clearpktopts(opt, IPV6_RTHDR);
3242 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3243 		if (opt->ip6po_rthdr == NULL)
3244 			return (ENOBUFS);
3245 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3246 
3247 		break;
3248 	}
3249 
3250 	case IPV6_USE_MIN_MTU:
3251 		if (len != sizeof(int))
3252 			return (EINVAL);
3253 		minmtupolicy = *(int *)buf;
3254 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3255 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3256 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3257 			return (EINVAL);
3258 		}
3259 		opt->ip6po_minmtu = minmtupolicy;
3260 		break;
3261 
3262 	case IPV6_DONTFRAG:
3263 		if (len != sizeof(int))
3264 			return (EINVAL);
3265 
3266 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3267 			/*
3268 			 * we ignore this option for TCP sockets.
3269 			 * (RFC3542 leaves this case unspecified.)
3270 			 */
3271 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3272 		} else
3273 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3274 		break;
3275 
3276 	case IPV6_PREFER_TEMPADDR:
3277 		if (len != sizeof(int))
3278 			return (EINVAL);
3279 		preftemp = *(int *)buf;
3280 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3281 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3282 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3283 			return (EINVAL);
3284 		}
3285 		opt->ip6po_prefer_tempaddr = preftemp;
3286 		break;
3287 
3288 	default:
3289 		return (ENOPROTOOPT);
3290 	} /* end of switch */
3291 
3292 	return (0);
3293 }
3294 
3295 /*
3296  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3297  * packet to the input queue of a specified interface.  Note that this
3298  * calls the output routine of the loopback "driver", but with an interface
3299  * pointer that might NOT be &loif -- easier than replicating that code here.
3300  */
3301 void
3302 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3303 {
3304 	struct mbuf *copym;
3305 	struct ip6_hdr *ip6;
3306 
3307 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3308 	if (copym == NULL)
3309 		return;
3310 
3311 	/*
3312 	 * Make sure to deep-copy IPv6 header portion in case the data
3313 	 * is in an mbuf cluster, so that we can safely override the IPv6
3314 	 * header portion later.
3315 	 */
3316 	if (!M_WRITABLE(copym) ||
3317 	    copym->m_len < sizeof(struct ip6_hdr)) {
3318 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3319 		if (copym == NULL)
3320 			return;
3321 	}
3322 	ip6 = mtod(copym, struct ip6_hdr *);
3323 	/*
3324 	 * clear embedded scope identifiers if necessary.
3325 	 * in6_clearscope will touch the addresses only when necessary.
3326 	 */
3327 	in6_clearscope(&ip6->ip6_src);
3328 	in6_clearscope(&ip6->ip6_dst);
3329 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3330 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3331 		    CSUM_PSEUDO_HDR;
3332 		copym->m_pkthdr.csum_data = 0xffff;
3333 	}
3334 	if_simloop(ifp, copym, AF_INET6, 0);
3335 }
3336 
3337 /*
3338  * Chop IPv6 header off from the payload.
3339  */
3340 static int
3341 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3342 {
3343 	struct mbuf *mh;
3344 	struct ip6_hdr *ip6;
3345 
3346 	ip6 = mtod(m, struct ip6_hdr *);
3347 	if (m->m_len > sizeof(*ip6)) {
3348 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3349 		if (mh == NULL) {
3350 			m_freem(m);
3351 			return ENOBUFS;
3352 		}
3353 		m_move_pkthdr(mh, m);
3354 		M_ALIGN(mh, sizeof(*ip6));
3355 		m->m_len -= sizeof(*ip6);
3356 		m->m_data += sizeof(*ip6);
3357 		mh->m_next = m;
3358 		m = mh;
3359 		m->m_len = sizeof(*ip6);
3360 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3361 	}
3362 	exthdrs->ip6e_ip6 = m;
3363 	return 0;
3364 }
3365 
3366 /*
3367  * Compute IPv6 extension header length.
3368  */
3369 int
3370 ip6_optlen(struct inpcb *inp)
3371 {
3372 	int len;
3373 
3374 	if (!inp->in6p_outputopts)
3375 		return 0;
3376 
3377 	len = 0;
3378 #define elen(x) \
3379     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3380 
3381 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3382 	if (inp->in6p_outputopts->ip6po_rthdr)
3383 		/* dest1 is valid with rthdr only */
3384 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3385 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3386 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3387 	return len;
3388 #undef elen
3389 }
3390