xref: /freebsd/sys/netinet6/ip6_output.c (revision b4a19ee2422c66243410081c83cbd08f1505fc84)
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_llatbl.h>
96 #include <net/netisr.h>
97 #include <net/route.h>
98 #include <net/pfil.h>
99 #include <net/rss_config.h>
100 #include <net/vnet.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip_var.h>
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet/tcp_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/in6_rss.h>
114 
115 #include <netipsec/ipsec_support.h>
116 #ifdef SCTP
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/ip6protosw.h>
122 #include <netinet6/scope6_var.h>
123 
124 extern int in6_mcast_loop;
125 
126 struct ip6_exthdrs {
127 	struct mbuf *ip6e_ip6;
128 	struct mbuf *ip6e_hbh;
129 	struct mbuf *ip6e_dest1;
130 	struct mbuf *ip6e_rthdr;
131 	struct mbuf *ip6e_dest2;
132 };
133 
134 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
135 
136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
137 			   struct ucred *, int);
138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
139 	struct socket *, struct sockopt *);
140 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
142 	struct ucred *, int, int, int);
143 
144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
146 	struct ip6_frag **);
147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
149 static int ip6_getpmtu(struct route_in6 *, int,
150 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
151 	u_int);
152 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
153 	u_long *, int *, u_int);
154 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
155 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
156 
157 
158 /*
159  * Make an extension header from option data.  hp is the source,
160  * mp is the destination, and _ol is the optlen.
161  */
162 #define	MAKE_EXTHDR(hp, mp, _ol)					\
163     do {								\
164 	if (hp) {							\
165 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
166 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
167 		    ((eh)->ip6e_len + 1) << 3);				\
168 		if (error)						\
169 			goto freehdrs;					\
170 		(_ol) += (*(mp))->m_len;				\
171 	}								\
172     } while (/*CONSTCOND*/ 0)
173 
174 /*
175  * Form a chain of extension headers.
176  * m is the extension header mbuf
177  * mp is the previous mbuf in the chain
178  * p is the next header
179  * i is the type of option.
180  */
181 #define MAKE_CHAIN(m, mp, p, i)\
182     do {\
183 	if (m) {\
184 		if (!hdrsplit) \
185 			panic("%s:%d: assumption failed: "\
186 			    "hdr not split: hdrsplit %d exthdrs %p",\
187 			    __func__, __LINE__, hdrsplit, &exthdrs);\
188 		*mtod((m), u_char *) = *(p);\
189 		*(p) = (i);\
190 		p = mtod((m), u_char *);\
191 		(m)->m_next = (mp)->m_next;\
192 		(mp)->m_next = (m);\
193 		(mp) = (m);\
194 	}\
195     } while (/*CONSTCOND*/ 0)
196 
197 void
198 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
199 {
200 	u_short csum;
201 
202 	csum = in_cksum_skip(m, offset + plen, offset);
203 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
204 		csum = 0xffff;
205 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
206 
207 	if (offset + sizeof(csum) > m->m_len)
208 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
209 	else
210 		*(u_short *)mtodo(m, offset) = csum;
211 }
212 
213 static int
214 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
215     int plen, int optlen, bool frag)
216 {
217 
218 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
219 	    "csum_flags %#x frag %d\n",
220 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags, frag));
221 
222 	if ((csum_flags & CSUM_DELAY_DATA_IPV6) ||
223 #ifdef SCTP
224 	    (csum_flags & CSUM_SCTP_IPV6) ||
225 #endif
226 	    (!frag && (ifp->if_capenable & IFCAP_NOMAP) == 0)) {
227 		m = mb_unmapped_to_ext(m);
228 		if (m == NULL) {
229 			if (frag)
230 				in6_ifstat_inc(ifp, ifs6_out_fragfail);
231 			else
232 				IP6STAT_INC(ip6s_odropped);
233 			return (ENOBUFS);
234 		}
235 		if (csum_flags & CSUM_DELAY_DATA_IPV6) {
236 			in6_delayed_cksum(m, plen - optlen,
237 			    sizeof(struct ip6_hdr) + optlen);
238 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
239 		}
240 #ifdef SCTP
241 		if (csum_flags & CSUM_SCTP_IPV6) {
242 			sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
243 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
244 		}
245 #endif
246 	}
247 
248 	return (0);
249 }
250 
251 int
252 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
253     int fraglen , uint32_t id)
254 {
255 	struct mbuf *m, **mnext, *m_frgpart;
256 	struct ip6_hdr *ip6, *mhip6;
257 	struct ip6_frag *ip6f;
258 	int off;
259 	int error;
260 	int tlen = m0->m_pkthdr.len;
261 
262 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
263 
264 	m = m0;
265 	ip6 = mtod(m, struct ip6_hdr *);
266 	mnext = &m->m_nextpkt;
267 
268 	for (off = hlen; off < tlen; off += fraglen) {
269 		m = m_gethdr(M_NOWAIT, MT_DATA);
270 		if (!m) {
271 			IP6STAT_INC(ip6s_odropped);
272 			return (ENOBUFS);
273 		}
274 
275 		/*
276 		 * Make sure the complete packet header gets copied
277 		 * from the originating mbuf to the newly created
278 		 * mbuf. This also ensures that existing firewall
279 		 * classification(s), VLAN tags and so on get copied
280 		 * to the resulting fragmented packet(s):
281 		 */
282 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
283 			m_free(m);
284 			IP6STAT_INC(ip6s_odropped);
285 			return (ENOBUFS);
286 		}
287 
288 		*mnext = m;
289 		mnext = &m->m_nextpkt;
290 		m->m_data += max_linkhdr;
291 		mhip6 = mtod(m, struct ip6_hdr *);
292 		*mhip6 = *ip6;
293 		m->m_len = sizeof(*mhip6);
294 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
295 		if (error) {
296 			IP6STAT_INC(ip6s_odropped);
297 			return (error);
298 		}
299 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
300 		if (off + fraglen >= tlen)
301 			fraglen = tlen - off;
302 		else
303 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
304 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
305 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
306 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
307 			IP6STAT_INC(ip6s_odropped);
308 			return (ENOBUFS);
309 		}
310 		m_cat(m, m_frgpart);
311 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
312 		ip6f->ip6f_reserved = 0;
313 		ip6f->ip6f_ident = id;
314 		ip6f->ip6f_nxt = nextproto;
315 		IP6STAT_INC(ip6s_ofragments);
316 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
317 	}
318 
319 	return (0);
320 }
321 
322 static int
323 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
324     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro)
325 {
326 #ifdef KERN_TLS
327 	struct ktls_session *tls = NULL;
328 #endif
329 	struct m_snd_tag *mst;
330 	int error;
331 
332 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
333 	mst = NULL;
334 
335 #ifdef KERN_TLS
336 	/*
337 	 * If this is an unencrypted TLS record, save a reference to
338 	 * the record.  This local reference is used to call
339 	 * ktls_output_eagain after the mbuf has been freed (thus
340 	 * dropping the mbuf's reference) in if_output.
341 	 */
342 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
343 		tls = ktls_hold(m->m_next->m_ext.ext_pgs->tls);
344 		mst = tls->snd_tag;
345 
346 		/*
347 		 * If a TLS session doesn't have a valid tag, it must
348 		 * have had an earlier ifp mismatch, so drop this
349 		 * packet.
350 		 */
351 		if (mst == NULL) {
352 			error = EAGAIN;
353 			goto done;
354 		}
355 	}
356 #endif
357 #ifdef RATELIMIT
358 	if (inp != NULL && mst == NULL) {
359 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
360 		    (inp->inp_snd_tag != NULL &&
361 		    inp->inp_snd_tag->ifp != ifp))
362 			in_pcboutput_txrtlmt(inp, ifp, m);
363 
364 		if (inp->inp_snd_tag != NULL)
365 			mst = inp->inp_snd_tag;
366 	}
367 #endif
368 	if (mst != NULL) {
369 		KASSERT(m->m_pkthdr.rcvif == NULL,
370 		    ("trying to add a send tag to a forwarded packet"));
371 		if (mst->ifp != ifp) {
372 			error = EAGAIN;
373 			goto done;
374 		}
375 
376 		/* stamp send tag on mbuf */
377 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
378 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
379 	}
380 
381 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
382 
383 done:
384 	/* Check for route change invalidating send tags. */
385 #ifdef KERN_TLS
386 	if (tls != NULL) {
387 		if (error == EAGAIN)
388 			error = ktls_output_eagain(inp, tls);
389 		ktls_free(tls);
390 	}
391 #endif
392 #ifdef RATELIMIT
393 	if (error == EAGAIN)
394 		in_pcboutput_eagain(inp);
395 #endif
396 	return (error);
397 }
398 
399 /*
400  * IP6 output.
401  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
402  * nxt, hlim, src, dst).
403  * This function may modify ver and hlim only.
404  * The mbuf chain containing the packet will be freed.
405  * The mbuf opt, if present, will not be freed.
406  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
407  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
408  * then result of route lookup is stored in ro->ro_rt.
409  *
410  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
411  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
412  *
413  * ifpp - XXX: just for statistics
414  */
415 /*
416  * XXX TODO: no flowid is assigned for outbound flows?
417  */
418 int
419 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
420     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
421     struct ifnet **ifpp, struct inpcb *inp)
422 {
423 	struct ip6_hdr *ip6;
424 	struct ifnet *ifp, *origifp;
425 	struct mbuf *m = m0;
426 	struct mbuf *mprev;
427 	struct route_in6 *ro_pmtu;
428 	struct rtentry *rt;
429 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
430 	struct in6_addr odst;
431 	u_char *nexthdrp;
432 	int tlen, len;
433 	int error = 0;
434 	struct in6_ifaddr *ia = NULL;
435 	u_long mtu;
436 	int alwaysfrag, dontfrag;
437 	u_int32_t optlen, plen = 0, unfragpartlen;
438 	struct ip6_exthdrs exthdrs;
439 	struct in6_addr src0, dst0;
440 	u_int32_t zone;
441 	bool hdrsplit;
442 	int sw_csum, tso;
443 	int needfiblookup;
444 	uint32_t fibnum;
445 	struct m_tag *fwd_tag = NULL;
446 	uint32_t id;
447 
448 	NET_EPOCH_ASSERT();
449 
450 	if (inp != NULL) {
451 		INP_LOCK_ASSERT(inp);
452 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
453 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
454 			/* Unconditionally set flowid. */
455 			m->m_pkthdr.flowid = inp->inp_flowid;
456 			M_HASHTYPE_SET(m, inp->inp_flowtype);
457 		}
458 #ifdef NUMA
459 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
460 #endif
461 	}
462 
463 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
464 	/*
465 	 * IPSec checking which handles several cases.
466 	 * FAST IPSEC: We re-injected the packet.
467 	 * XXX: need scope argument.
468 	 */
469 	if (IPSEC_ENABLED(ipv6)) {
470 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
471 			if (error == EINPROGRESS)
472 				error = 0;
473 			goto done;
474 		}
475 	}
476 #endif /* IPSEC */
477 
478 	/* Source address validation. */
479 	ip6 = mtod(m, struct ip6_hdr *);
480 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
481 	    (flags & IPV6_UNSPECSRC) == 0) {
482 		error = EOPNOTSUPP;
483 		IP6STAT_INC(ip6s_badscope);
484 		goto bad;
485 	}
486 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
487 		error = EOPNOTSUPP;
488 		IP6STAT_INC(ip6s_badscope);
489 		goto bad;
490 	}
491 
492 	/*
493 	 * If we are given packet options to add extension headers prepare them.
494 	 * Calculate the total length of the extension header chain.
495 	 * Keep the length of the unfragmentable part for fragmentation.
496 	 */
497 	bzero(&exthdrs, sizeof(exthdrs));
498 	optlen = 0;
499 	unfragpartlen = sizeof(struct ip6_hdr);
500 	if (opt) {
501 		/* Hop-by-Hop options header. */
502 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
503 
504 		/* Destination options header (1st part). */
505 		if (opt->ip6po_rthdr) {
506 #ifndef RTHDR_SUPPORT_IMPLEMENTED
507 			/*
508 			 * If there is a routing header, discard the packet
509 			 * right away here. RH0/1 are obsolete and we do not
510 			 * currently support RH2/3/4.
511 			 * People trying to use RH253/254 may want to disable
512 			 * this check.
513 			 * The moment we do support any routing header (again)
514 			 * this block should check the routing type more
515 			 * selectively.
516 			 */
517 			error = EINVAL;
518 			goto bad;
519 #endif
520 
521 			/*
522 			 * Destination options header (1st part).
523 			 * This only makes sense with a routing header.
524 			 * See Section 9.2 of RFC 3542.
525 			 * Disabling this part just for MIP6 convenience is
526 			 * a bad idea.  We need to think carefully about a
527 			 * way to make the advanced API coexist with MIP6
528 			 * options, which might automatically be inserted in
529 			 * the kernel.
530 			 */
531 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
532 			    optlen);
533 		}
534 		/* Routing header. */
535 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
536 
537 		unfragpartlen += optlen;
538 
539 		/*
540 		 * NOTE: we don't add AH/ESP length here (done in
541 		 * ip6_ipsec_output()).
542 		 */
543 
544 		/* Destination options header (2nd part). */
545 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
546 	}
547 
548 	/*
549 	 * If there is at least one extension header,
550 	 * separate IP6 header from the payload.
551 	 */
552 	hdrsplit = false;
553 	if (optlen) {
554 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
555 			m = NULL;
556 			goto freehdrs;
557 		}
558 		m = exthdrs.ip6e_ip6;
559 		ip6 = mtod(m, struct ip6_hdr *);
560 		hdrsplit = true;
561 	}
562 
563 	/* Adjust mbuf packet header length. */
564 	m->m_pkthdr.len += optlen;
565 	plen = m->m_pkthdr.len - sizeof(*ip6);
566 
567 	/* If this is a jumbo payload, insert a jumbo payload option. */
568 	if (plen > IPV6_MAXPACKET) {
569 		if (!hdrsplit) {
570 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
571 				m = NULL;
572 				goto freehdrs;
573 			}
574 			m = exthdrs.ip6e_ip6;
575 			ip6 = mtod(m, struct ip6_hdr *);
576 			hdrsplit = true;
577 		}
578 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
579 			goto freehdrs;
580 		ip6->ip6_plen = 0;
581 	} else
582 		ip6->ip6_plen = htons(plen);
583 	nexthdrp = &ip6->ip6_nxt;
584 
585 	if (optlen) {
586 		/*
587 		 * Concatenate headers and fill in next header fields.
588 		 * Here we have, on "m"
589 		 *	IPv6 payload
590 		 * and we insert headers accordingly.
591 		 * Finally, we should be getting:
592 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
593 		 *
594 		 * During the header composing process "m" points to IPv6
595 		 * header.  "mprev" points to an extension header prior to esp.
596 		 */
597 		mprev = m;
598 
599 		/*
600 		 * We treat dest2 specially.  This makes IPsec processing
601 		 * much easier.  The goal here is to make mprev point the
602 		 * mbuf prior to dest2.
603 		 *
604 		 * Result: IPv6 dest2 payload.
605 		 * m and mprev will point to IPv6 header.
606 		 */
607 		if (exthdrs.ip6e_dest2) {
608 			if (!hdrsplit)
609 				panic("%s:%d: assumption failed: "
610 				    "hdr not split: hdrsplit %d exthdrs %p",
611 				    __func__, __LINE__, hdrsplit, &exthdrs);
612 			exthdrs.ip6e_dest2->m_next = m->m_next;
613 			m->m_next = exthdrs.ip6e_dest2;
614 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
615 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
616 		}
617 
618 		/*
619 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
620 		 * m will point to IPv6 header.  mprev will point to the
621 		 * extension header prior to dest2 (rthdr in the above case).
622 		 */
623 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
624 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
625 			   IPPROTO_DSTOPTS);
626 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
627 			   IPPROTO_ROUTING);
628 	}
629 
630 	IP6STAT_INC(ip6s_localout);
631 
632 	/* Route packet. */
633 	ro_pmtu = ro;
634 	if (opt && opt->ip6po_rthdr)
635 		ro = &opt->ip6po_route;
636 	if (ro != NULL)
637 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
638 	else
639 		dst = &sin6;
640 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
641 
642 again:
643 	/*
644 	 * If specified, try to fill in the traffic class field.
645 	 * Do not override if a non-zero value is already set.
646 	 * We check the diffserv field and the ECN field separately.
647 	 */
648 	if (opt && opt->ip6po_tclass >= 0) {
649 		int mask = 0;
650 
651 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
652 			mask |= 0xfc;
653 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
654 			mask |= 0x03;
655 		if (mask != 0)
656 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
657 	}
658 
659 	/* Fill in or override the hop limit field, if necessary. */
660 	if (opt && opt->ip6po_hlim != -1)
661 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
662 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
663 		if (im6o != NULL)
664 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
665 		else
666 			ip6->ip6_hlim = V_ip6_defmcasthlim;
667 	}
668 
669 	if (ro == NULL || ro->ro_rt == NULL) {
670 		bzero(dst, sizeof(*dst));
671 		dst->sin6_family = AF_INET6;
672 		dst->sin6_len = sizeof(*dst);
673 		dst->sin6_addr = ip6->ip6_dst;
674 	}
675 	/*
676 	 * Validate route against routing table changes.
677 	 * Make sure that the address family is set in route.
678 	 */
679 	rt = NULL;
680 	ifp = NULL;
681 	mtu = 0;
682 	if (ro != NULL) {
683 		if (ro->ro_rt != NULL && inp != NULL) {
684 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
685 			RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
686 			    fibnum);
687 		}
688 		if (ro->ro_rt != NULL && fwd_tag == NULL &&
689 		    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
690 		    ro->ro_rt->rt_ifp == NULL ||
691 		    !RT_LINK_IS_UP(ro->ro_rt->rt_ifp) ||
692 		    ro->ro_dst.sin6_family != AF_INET6 ||
693 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
694 			RO_INVALIDATE_CACHE(ro);
695 
696 		if (ro->ro_rt != NULL && fwd_tag == NULL &&
697 		    (ro->ro_rt->rt_flags & RTF_UP) &&
698 		    ro->ro_dst.sin6_family == AF_INET6 &&
699 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
700 			rt = ro->ro_rt;
701 			ifp = ro->ro_rt->rt_ifp;
702 		} else {
703 			if (ro->ro_lle)
704 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
705 			ro->ro_lle = NULL;
706 			if (fwd_tag == NULL) {
707 				bzero(&dst_sa, sizeof(dst_sa));
708 				dst_sa.sin6_family = AF_INET6;
709 				dst_sa.sin6_len = sizeof(dst_sa);
710 				dst_sa.sin6_addr = ip6->ip6_dst;
711 			}
712 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
713 			    &rt, fibnum);
714 			if (error != 0) {
715 				IP6STAT_INC(ip6s_noroute);
716 				if (ifp != NULL)
717 					in6_ifstat_inc(ifp, ifs6_out_discard);
718 				goto bad;
719 			}
720 			if (ifp != NULL)
721 			    mtu = ifp->if_mtu;
722 		}
723 		if (rt == NULL) {
724 			/*
725 			 * If in6_selectroute() does not return a route entry
726 			 * dst may not have been updated.
727 			 */
728 			*dst = dst_sa;	/* XXX */
729 		} else {
730 			if (rt->rt_flags & RTF_HOST)
731 			    mtu = rt->rt_mtu;
732 			ia = (struct in6_ifaddr *)(rt->rt_ifa);
733 			counter_u64_add(rt->rt_pksent, 1);
734 		}
735 	} else {
736 		struct nhop6_extended nh6;
737 		struct in6_addr kdst;
738 		uint32_t scopeid;
739 
740 		if (fwd_tag == NULL) {
741 			bzero(&dst_sa, sizeof(dst_sa));
742 			dst_sa.sin6_family = AF_INET6;
743 			dst_sa.sin6_len = sizeof(dst_sa);
744 			dst_sa.sin6_addr = ip6->ip6_dst;
745 		}
746 
747 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
748 		    im6o != NULL &&
749 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
750 			/* We do not need a route lookup. */
751 			*dst = dst_sa;	/* XXX */
752 			goto nonh6lookup;
753 		}
754 
755 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
756 
757 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
758 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
759 			if (scopeid > 0) {
760 				ifp = in6_getlinkifnet(scopeid);
761 				*dst = dst_sa;	/* XXX */
762 				goto nonh6lookup;
763 			}
764 		}
765 
766 		error = fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0,
767 		    &nh6);
768 		if (error != 0) {
769 			IP6STAT_INC(ip6s_noroute);
770 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
771 			error = EHOSTUNREACH;;
772 			goto bad;
773 		}
774 
775 		ifp = nh6.nh_ifp;
776 		mtu = nh6.nh_mtu;
777 		dst->sin6_addr = nh6.nh_addr;
778 		ia = nh6.nh_ia;
779 		fib6_free_nh_ext(fibnum, &nh6);
780 nonh6lookup:
781 		;
782 	}
783 
784 	/* Then rt (for unicast) and ifp must be non-NULL valid values. */
785 	if ((flags & IPV6_FORWARDING) == 0) {
786 		/* XXX: the FORWARDING flag can be set for mrouting. */
787 		in6_ifstat_inc(ifp, ifs6_out_request);
788 	}
789 
790 	/* Setup data structures for scope ID checks. */
791 	src0 = ip6->ip6_src;
792 	bzero(&src_sa, sizeof(src_sa));
793 	src_sa.sin6_family = AF_INET6;
794 	src_sa.sin6_len = sizeof(src_sa);
795 	src_sa.sin6_addr = ip6->ip6_src;
796 
797 	dst0 = ip6->ip6_dst;
798 	/* Re-initialize to be sure. */
799 	bzero(&dst_sa, sizeof(dst_sa));
800 	dst_sa.sin6_family = AF_INET6;
801 	dst_sa.sin6_len = sizeof(dst_sa);
802 	dst_sa.sin6_addr = ip6->ip6_dst;
803 
804 	/* Check for valid scope ID. */
805 	if (in6_setscope(&src0, ifp, &zone) == 0 &&
806 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
807 	    in6_setscope(&dst0, ifp, &zone) == 0 &&
808 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
809 		/*
810 		 * The outgoing interface is in the zone of the source
811 		 * and destination addresses.
812 		 *
813 		 * Because the loopback interface cannot receive
814 		 * packets with a different scope ID than its own,
815 		 * there is a trick to pretend the outgoing packet
816 		 * was received by the real network interface, by
817 		 * setting "origifp" different from "ifp". This is
818 		 * only allowed when "ifp" is a loopback network
819 		 * interface. Refer to code in nd6_output_ifp() for
820 		 * more details.
821 		 */
822 		origifp = ifp;
823 
824 		/*
825 		 * We should use ia_ifp to support the case of sending
826 		 * packets to an address of our own.
827 		 */
828 		if (ia != NULL && ia->ia_ifp)
829 			ifp = ia->ia_ifp;
830 
831 	} else if ((ifp->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 	    (origifp = 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(ifp, 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 (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
856 		if (opt && opt->ip6po_nextroute.ro_rt) {
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 ((rt->rt_flags & RTF_GATEWAY))
865 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
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 #ifdef SCTP
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 #ifdef SCTP
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 	/*
1097 	 * Send the packet to the outgoing interface.
1098 	 * If necessary, do IPv6 fragmentation before sending.
1099 	 *
1100 	 * The logic here is rather complex:
1101 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
1102 	 * 1-a:	send as is if tlen <= path mtu
1103 	 * 1-b:	fragment if tlen > path mtu
1104 	 *
1105 	 * 2: if user asks us not to fragment (dontfrag == 1)
1106 	 * 2-a:	send as is if tlen <= interface mtu
1107 	 * 2-b:	error if tlen > interface mtu
1108 	 *
1109 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
1110 	 *	always fragment
1111 	 *
1112 	 * 4: if dontfrag == 1 && alwaysfrag == 1
1113 	 *	error, as we cannot handle this conflicting request.
1114 	 */
1115 	sw_csum = m->m_pkthdr.csum_flags;
1116 	if (!hdrsplit) {
1117 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
1118 		sw_csum &= ~ifp->if_hwassist;
1119 	} else
1120 		tso = 0;
1121 	/*
1122 	 * If we added extension headers, we will not do TSO and calculate the
1123 	 * checksums ourselves for now.
1124 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1125 	 * with ext. hdrs.
1126 	 */
1127 	error = ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen, false);
1128 	if (error != 0)
1129 		goto bad;
1130 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1131 	tlen = m->m_pkthdr.len;
1132 
1133 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1134 		dontfrag = 1;
1135 	else
1136 		dontfrag = 0;
1137 	if (dontfrag && alwaysfrag) {	/* Case 4. */
1138 		/* Conflicting request - can't transmit. */
1139 		error = EMSGSIZE;
1140 		goto bad;
1141 	}
1142 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1143 		/*
1144 		 * Even if the DONTFRAG option is specified, we cannot send the
1145 		 * packet when the data length is larger than the MTU of the
1146 		 * outgoing interface.
1147 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1148 		 * application wanted to know the MTU value. Also return an
1149 		 * error code (this is not described in the API spec).
1150 		 */
1151 		if (inp != NULL)
1152 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1153 		error = EMSGSIZE;
1154 		goto bad;
1155 	}
1156 
1157 	/* Transmit packet without fragmentation. */
1158 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* Cases 1-a and 2-a. */
1159 		struct in6_ifaddr *ia6;
1160 
1161 		ip6 = mtod(m, struct ip6_hdr *);
1162 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1163 		if (ia6) {
1164 			/* Record statistics for this interface address. */
1165 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1166 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1167 			    m->m_pkthdr.len);
1168 			ifa_free(&ia6->ia_ifa);
1169 		}
1170 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1171 		goto done;
1172 	}
1173 
1174 	/* Try to fragment the packet.  Cases 1-b and 3. */
1175 	if (mtu < IPV6_MMTU) {
1176 		/* Path MTU cannot be less than IPV6_MMTU. */
1177 		error = EMSGSIZE;
1178 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1179 		goto bad;
1180 	} else if (ip6->ip6_plen == 0) {
1181 		/* Jumbo payload cannot be fragmented. */
1182 		error = EMSGSIZE;
1183 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1184 		goto bad;
1185 	} else {
1186 		u_char nextproto;
1187 
1188 		/*
1189 		 * Too large for the destination or interface;
1190 		 * fragment if possible.
1191 		 * Must be able to put at least 8 bytes per fragment.
1192 		 */
1193 		if (mtu > IPV6_MAXPACKET)
1194 			mtu = IPV6_MAXPACKET;
1195 
1196 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1197 		if (len < 8) {
1198 			error = EMSGSIZE;
1199 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1200 			goto bad;
1201 		}
1202 
1203 		/*
1204 		 * If the interface will not calculate checksums on
1205 		 * fragmented packets, then do it here.
1206 		 * XXX-BZ handle the hw offloading case.  Need flags.
1207 		 */
1208 		error = ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags,
1209 		    plen, optlen, true);
1210 		if (error != 0)
1211 			goto bad;
1212 
1213 		/*
1214 		 * Change the next header field of the last header in the
1215 		 * unfragmentable part.
1216 		 */
1217 		if (exthdrs.ip6e_rthdr) {
1218 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1219 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1220 		} else if (exthdrs.ip6e_dest1) {
1221 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1222 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1223 		} else if (exthdrs.ip6e_hbh) {
1224 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1225 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1226 		} else {
1227 			ip6 = mtod(m, struct ip6_hdr *);
1228 			nextproto = ip6->ip6_nxt;
1229 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1230 		}
1231 
1232 		/*
1233 		 * Loop through length of segment after first fragment,
1234 		 * make new header and copy data of each part and link onto
1235 		 * chain.
1236 		 */
1237 		m0 = m;
1238 		id = htonl(ip6_randomid());
1239 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1240 		if (error != 0)
1241 			goto sendorfree;
1242 
1243 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1244 	}
1245 
1246 	/* Remove leading garbage. */
1247 sendorfree:
1248 	m = m0->m_nextpkt;
1249 	m0->m_nextpkt = 0;
1250 	m_freem(m0);
1251 	for (; m; m = m0) {
1252 		m0 = m->m_nextpkt;
1253 		m->m_nextpkt = 0;
1254 		if (error == 0) {
1255 			/* Record statistics for this interface address. */
1256 			if (ia) {
1257 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1258 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1259 				    m->m_pkthdr.len);
1260 			}
1261 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1262 		} else
1263 			m_freem(m);
1264 	}
1265 
1266 	if (error == 0)
1267 		IP6STAT_INC(ip6s_fragmented);
1268 
1269 done:
1270 	return (error);
1271 
1272 freehdrs:
1273 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1274 	m_freem(exthdrs.ip6e_dest1);
1275 	m_freem(exthdrs.ip6e_rthdr);
1276 	m_freem(exthdrs.ip6e_dest2);
1277 	/* FALLTHROUGH */
1278 bad:
1279 	if (m)
1280 		m_freem(m);
1281 	goto done;
1282 }
1283 
1284 static int
1285 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1286 {
1287 	struct mbuf *m;
1288 
1289 	if (hlen > MCLBYTES)
1290 		return (ENOBUFS); /* XXX */
1291 
1292 	if (hlen > MLEN)
1293 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1294 	else
1295 		m = m_get(M_NOWAIT, MT_DATA);
1296 	if (m == NULL)
1297 		return (ENOBUFS);
1298 	m->m_len = hlen;
1299 	if (hdr)
1300 		bcopy(hdr, mtod(m, caddr_t), hlen);
1301 
1302 	*mp = m;
1303 	return (0);
1304 }
1305 
1306 /*
1307  * Insert jumbo payload option.
1308  */
1309 static int
1310 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1311 {
1312 	struct mbuf *mopt;
1313 	u_char *optbuf;
1314 	u_int32_t v;
1315 
1316 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1317 
1318 	/*
1319 	 * If there is no hop-by-hop options header, allocate new one.
1320 	 * If there is one but it doesn't have enough space to store the
1321 	 * jumbo payload option, allocate a cluster to store the whole options.
1322 	 * Otherwise, use it to store the options.
1323 	 */
1324 	if (exthdrs->ip6e_hbh == NULL) {
1325 		mopt = m_get(M_NOWAIT, MT_DATA);
1326 		if (mopt == NULL)
1327 			return (ENOBUFS);
1328 		mopt->m_len = JUMBOOPTLEN;
1329 		optbuf = mtod(mopt, u_char *);
1330 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1331 		exthdrs->ip6e_hbh = mopt;
1332 	} else {
1333 		struct ip6_hbh *hbh;
1334 
1335 		mopt = exthdrs->ip6e_hbh;
1336 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1337 			/*
1338 			 * XXX assumption:
1339 			 * - exthdrs->ip6e_hbh is not referenced from places
1340 			 *   other than exthdrs.
1341 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1342 			 */
1343 			int oldoptlen = mopt->m_len;
1344 			struct mbuf *n;
1345 
1346 			/*
1347 			 * XXX: give up if the whole (new) hbh header does
1348 			 * not fit even in an mbuf cluster.
1349 			 */
1350 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1351 				return (ENOBUFS);
1352 
1353 			/*
1354 			 * As a consequence, we must always prepare a cluster
1355 			 * at this point.
1356 			 */
1357 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1358 			if (n == NULL)
1359 				return (ENOBUFS);
1360 			n->m_len = oldoptlen + JUMBOOPTLEN;
1361 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1362 			    oldoptlen);
1363 			optbuf = mtod(n, caddr_t) + oldoptlen;
1364 			m_freem(mopt);
1365 			mopt = exthdrs->ip6e_hbh = n;
1366 		} else {
1367 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1368 			mopt->m_len += JUMBOOPTLEN;
1369 		}
1370 		optbuf[0] = IP6OPT_PADN;
1371 		optbuf[1] = 1;
1372 
1373 		/*
1374 		 * Adjust the header length according to the pad and
1375 		 * the jumbo payload option.
1376 		 */
1377 		hbh = mtod(mopt, struct ip6_hbh *);
1378 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1379 	}
1380 
1381 	/* fill in the option. */
1382 	optbuf[2] = IP6OPT_JUMBO;
1383 	optbuf[3] = 4;
1384 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1385 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1386 
1387 	/* finally, adjust the packet header length */
1388 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1389 
1390 	return (0);
1391 #undef JUMBOOPTLEN
1392 }
1393 
1394 /*
1395  * Insert fragment header and copy unfragmentable header portions.
1396  */
1397 static int
1398 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1399     struct ip6_frag **frghdrp)
1400 {
1401 	struct mbuf *n, *mlast;
1402 
1403 	if (hlen > sizeof(struct ip6_hdr)) {
1404 		n = m_copym(m0, sizeof(struct ip6_hdr),
1405 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1406 		if (n == NULL)
1407 			return (ENOBUFS);
1408 		m->m_next = n;
1409 	} else
1410 		n = m;
1411 
1412 	/* Search for the last mbuf of unfragmentable part. */
1413 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1414 		;
1415 
1416 	if (M_WRITABLE(mlast) &&
1417 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1418 		/* use the trailing space of the last mbuf for the fragment hdr */
1419 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1420 		    mlast->m_len);
1421 		mlast->m_len += sizeof(struct ip6_frag);
1422 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1423 	} else {
1424 		/* allocate a new mbuf for the fragment header */
1425 		struct mbuf *mfrg;
1426 
1427 		mfrg = m_get(M_NOWAIT, MT_DATA);
1428 		if (mfrg == NULL)
1429 			return (ENOBUFS);
1430 		mfrg->m_len = sizeof(struct ip6_frag);
1431 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1432 		mlast->m_next = mfrg;
1433 	}
1434 
1435 	return (0);
1436 }
1437 
1438 /*
1439  * Calculates IPv6 path mtu for destination @dst.
1440  * Resulting MTU is stored in @mtup.
1441  *
1442  * Returns 0 on success.
1443  */
1444 static int
1445 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1446 {
1447 	struct nhop6_extended nh6;
1448 	struct in6_addr kdst;
1449 	uint32_t scopeid;
1450 	struct ifnet *ifp;
1451 	u_long mtu;
1452 	int error;
1453 
1454 	in6_splitscope(dst, &kdst, &scopeid);
1455 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1456 		return (EHOSTUNREACH);
1457 
1458 	ifp = nh6.nh_ifp;
1459 	mtu = nh6.nh_mtu;
1460 
1461 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1462 	fib6_free_nh_ext(fibnum, &nh6);
1463 
1464 	return (error);
1465 }
1466 
1467 /*
1468  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1469  * and cached data in @ro_pmtu.
1470  * MTU from (successful) route lookup is saved (along with dst)
1471  * inside @ro_pmtu to avoid subsequent route lookups after packet
1472  * filter processing.
1473  *
1474  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1475  * Returns 0 on success.
1476  */
1477 static int
1478 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1479     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1480     int *alwaysfragp, u_int fibnum, u_int proto)
1481 {
1482 	struct nhop6_basic nh6;
1483 	struct in6_addr kdst;
1484 	uint32_t scopeid;
1485 	struct sockaddr_in6 *sa6_dst, sin6;
1486 	u_long mtu;
1487 
1488 	mtu = 0;
1489 	if (ro_pmtu == NULL || do_lookup) {
1490 
1491 		/*
1492 		 * Here ro_pmtu has final destination address, while
1493 		 * ro might represent immediate destination.
1494 		 * Use ro_pmtu destination since mtu might differ.
1495 		 */
1496 		if (ro_pmtu != NULL) {
1497 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1498 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1499 				ro_pmtu->ro_mtu = 0;
1500 		} else
1501 			sa6_dst = &sin6;
1502 
1503 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1504 			bzero(sa6_dst, sizeof(*sa6_dst));
1505 			sa6_dst->sin6_family = AF_INET6;
1506 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1507 			sa6_dst->sin6_addr = *dst;
1508 
1509 			in6_splitscope(dst, &kdst, &scopeid);
1510 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1511 			    &nh6) == 0) {
1512 				mtu = nh6.nh_mtu;
1513 				if (ro_pmtu != NULL)
1514 					ro_pmtu->ro_mtu = mtu;
1515 			}
1516 		} else
1517 			mtu = ro_pmtu->ro_mtu;
1518 	}
1519 
1520 	if (ro_pmtu != NULL && ro_pmtu->ro_rt != NULL)
1521 		mtu = ro_pmtu->ro_rt->rt_mtu;
1522 
1523 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1524 }
1525 
1526 /*
1527  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1528  * hostcache data for @dst.
1529  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1530  *
1531  * Returns 0 on success.
1532  */
1533 static int
1534 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1535     u_long *mtup, int *alwaysfragp, u_int proto)
1536 {
1537 	u_long mtu = 0;
1538 	int alwaysfrag = 0;
1539 	int error = 0;
1540 
1541 	if (rt_mtu > 0) {
1542 		u_int32_t ifmtu;
1543 		struct in_conninfo inc;
1544 
1545 		bzero(&inc, sizeof(inc));
1546 		inc.inc_flags |= INC_ISIPV6;
1547 		inc.inc6_faddr = *dst;
1548 
1549 		ifmtu = IN6_LINKMTU(ifp);
1550 
1551 		/* TCP is known to react to pmtu changes so skip hc */
1552 		if (proto != IPPROTO_TCP)
1553 			mtu = tcp_hc_getmtu(&inc);
1554 
1555 		if (mtu)
1556 			mtu = min(mtu, rt_mtu);
1557 		else
1558 			mtu = rt_mtu;
1559 		if (mtu == 0)
1560 			mtu = ifmtu;
1561 		else if (mtu < IPV6_MMTU) {
1562 			/*
1563 			 * RFC2460 section 5, last paragraph:
1564 			 * if we record ICMPv6 too big message with
1565 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1566 			 * or smaller, with framgent header attached.
1567 			 * (fragment header is needed regardless from the
1568 			 * packet size, for translators to identify packets)
1569 			 */
1570 			alwaysfrag = 1;
1571 			mtu = IPV6_MMTU;
1572 		}
1573 	} else if (ifp) {
1574 		mtu = IN6_LINKMTU(ifp);
1575 	} else
1576 		error = EHOSTUNREACH; /* XXX */
1577 
1578 	*mtup = mtu;
1579 	if (alwaysfragp)
1580 		*alwaysfragp = alwaysfrag;
1581 	return (error);
1582 }
1583 
1584 /*
1585  * IP6 socket option processing.
1586  */
1587 int
1588 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1589 {
1590 	int optdatalen, uproto;
1591 	void *optdata;
1592 	struct inpcb *inp = sotoinpcb(so);
1593 	int error, optval;
1594 	int level, op, optname;
1595 	int optlen;
1596 	struct thread *td;
1597 #ifdef	RSS
1598 	uint32_t rss_bucket;
1599 	int retval;
1600 #endif
1601 
1602 /*
1603  * Don't use more than a quarter of mbuf clusters.  N.B.:
1604  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1605  * on LP64 architectures, so cast to u_long to avoid undefined
1606  * behavior.  ILP32 architectures cannot have nmbclusters
1607  * large enough to overflow for other reasons.
1608  */
1609 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1610 
1611 	level = sopt->sopt_level;
1612 	op = sopt->sopt_dir;
1613 	optname = sopt->sopt_name;
1614 	optlen = sopt->sopt_valsize;
1615 	td = sopt->sopt_td;
1616 	error = 0;
1617 	optval = 0;
1618 	uproto = (int)so->so_proto->pr_protocol;
1619 
1620 	if (level != IPPROTO_IPV6) {
1621 		error = EINVAL;
1622 
1623 		if (sopt->sopt_level == SOL_SOCKET &&
1624 		    sopt->sopt_dir == SOPT_SET) {
1625 			switch (sopt->sopt_name) {
1626 			case SO_REUSEADDR:
1627 				INP_WLOCK(inp);
1628 				if ((so->so_options & SO_REUSEADDR) != 0)
1629 					inp->inp_flags2 |= INP_REUSEADDR;
1630 				else
1631 					inp->inp_flags2 &= ~INP_REUSEADDR;
1632 				INP_WUNLOCK(inp);
1633 				error = 0;
1634 				break;
1635 			case SO_REUSEPORT:
1636 				INP_WLOCK(inp);
1637 				if ((so->so_options & SO_REUSEPORT) != 0)
1638 					inp->inp_flags2 |= INP_REUSEPORT;
1639 				else
1640 					inp->inp_flags2 &= ~INP_REUSEPORT;
1641 				INP_WUNLOCK(inp);
1642 				error = 0;
1643 				break;
1644 			case SO_REUSEPORT_LB:
1645 				INP_WLOCK(inp);
1646 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1647 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1648 				else
1649 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1650 				INP_WUNLOCK(inp);
1651 				error = 0;
1652 				break;
1653 			case SO_SETFIB:
1654 				INP_WLOCK(inp);
1655 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1656 				INP_WUNLOCK(inp);
1657 				error = 0;
1658 				break;
1659 			case SO_MAX_PACING_RATE:
1660 #ifdef RATELIMIT
1661 				INP_WLOCK(inp);
1662 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1663 				INP_WUNLOCK(inp);
1664 				error = 0;
1665 #else
1666 				error = EOPNOTSUPP;
1667 #endif
1668 				break;
1669 			default:
1670 				break;
1671 			}
1672 		}
1673 	} else {		/* level == IPPROTO_IPV6 */
1674 		switch (op) {
1675 
1676 		case SOPT_SET:
1677 			switch (optname) {
1678 			case IPV6_2292PKTOPTIONS:
1679 #ifdef IPV6_PKTOPTIONS
1680 			case IPV6_PKTOPTIONS:
1681 #endif
1682 			{
1683 				struct mbuf *m;
1684 
1685 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1686 					printf("ip6_ctloutput: mbuf limit hit\n");
1687 					error = ENOBUFS;
1688 					break;
1689 				}
1690 
1691 				error = soopt_getm(sopt, &m); /* XXX */
1692 				if (error != 0)
1693 					break;
1694 				error = soopt_mcopyin(sopt, m); /* XXX */
1695 				if (error != 0)
1696 					break;
1697 				INP_WLOCK(inp);
1698 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1699 				    so, sopt);
1700 				INP_WUNLOCK(inp);
1701 				m_freem(m); /* XXX */
1702 				break;
1703 			}
1704 
1705 			/*
1706 			 * Use of some Hop-by-Hop options or some
1707 			 * Destination options, might require special
1708 			 * privilege.  That is, normal applications
1709 			 * (without special privilege) might be forbidden
1710 			 * from setting certain options in outgoing packets,
1711 			 * and might never see certain options in received
1712 			 * packets. [RFC 2292 Section 6]
1713 			 * KAME specific note:
1714 			 *  KAME prevents non-privileged users from sending or
1715 			 *  receiving ANY hbh/dst options in order to avoid
1716 			 *  overhead of parsing options in the kernel.
1717 			 */
1718 			case IPV6_RECVHOPOPTS:
1719 			case IPV6_RECVDSTOPTS:
1720 			case IPV6_RECVRTHDRDSTOPTS:
1721 				if (td != NULL) {
1722 					error = priv_check(td,
1723 					    PRIV_NETINET_SETHDROPTS);
1724 					if (error)
1725 						break;
1726 				}
1727 				/* FALLTHROUGH */
1728 			case IPV6_UNICAST_HOPS:
1729 			case IPV6_HOPLIMIT:
1730 
1731 			case IPV6_RECVPKTINFO:
1732 			case IPV6_RECVHOPLIMIT:
1733 			case IPV6_RECVRTHDR:
1734 			case IPV6_RECVPATHMTU:
1735 			case IPV6_RECVTCLASS:
1736 			case IPV6_RECVFLOWID:
1737 #ifdef	RSS
1738 			case IPV6_RECVRSSBUCKETID:
1739 #endif
1740 			case IPV6_V6ONLY:
1741 			case IPV6_AUTOFLOWLABEL:
1742 			case IPV6_ORIGDSTADDR:
1743 			case IPV6_BINDANY:
1744 			case IPV6_BINDMULTI:
1745 #ifdef	RSS
1746 			case IPV6_RSS_LISTEN_BUCKET:
1747 #endif
1748 				if (optname == IPV6_BINDANY && td != NULL) {
1749 					error = priv_check(td,
1750 					    PRIV_NETINET_BINDANY);
1751 					if (error)
1752 						break;
1753 				}
1754 
1755 				if (optlen != sizeof(int)) {
1756 					error = EINVAL;
1757 					break;
1758 				}
1759 				error = sooptcopyin(sopt, &optval,
1760 					sizeof optval, sizeof optval);
1761 				if (error)
1762 					break;
1763 				switch (optname) {
1764 
1765 				case IPV6_UNICAST_HOPS:
1766 					if (optval < -1 || optval >= 256)
1767 						error = EINVAL;
1768 					else {
1769 						/* -1 = kernel default */
1770 						inp->in6p_hops = optval;
1771 						if ((inp->inp_vflag &
1772 						     INP_IPV4) != 0)
1773 							inp->inp_ip_ttl = optval;
1774 					}
1775 					break;
1776 #define OPTSET(bit) \
1777 do { \
1778 	INP_WLOCK(inp); \
1779 	if (optval) \
1780 		inp->inp_flags |= (bit); \
1781 	else \
1782 		inp->inp_flags &= ~(bit); \
1783 	INP_WUNLOCK(inp); \
1784 } while (/*CONSTCOND*/ 0)
1785 #define OPTSET2292(bit) \
1786 do { \
1787 	INP_WLOCK(inp); \
1788 	inp->inp_flags |= IN6P_RFC2292; \
1789 	if (optval) \
1790 		inp->inp_flags |= (bit); \
1791 	else \
1792 		inp->inp_flags &= ~(bit); \
1793 	INP_WUNLOCK(inp); \
1794 } while (/*CONSTCOND*/ 0)
1795 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1796 
1797 #define OPTSET2_N(bit, val) do {					\
1798 	if (val)							\
1799 		inp->inp_flags2 |= bit;					\
1800 	else								\
1801 		inp->inp_flags2 &= ~bit;				\
1802 } while (0)
1803 #define OPTSET2(bit, val) do {						\
1804 	INP_WLOCK(inp);							\
1805 	OPTSET2_N(bit, val);						\
1806 	INP_WUNLOCK(inp);						\
1807 } while (0)
1808 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1809 #define OPTSET2292_EXCLUSIVE(bit)					\
1810 do {									\
1811 	INP_WLOCK(inp);							\
1812 	if (OPTBIT(IN6P_RFC2292)) {					\
1813 		error = EINVAL;						\
1814 	} else {							\
1815 		if (optval)						\
1816 			inp->inp_flags |= (bit);			\
1817 		else							\
1818 			inp->inp_flags &= ~(bit);			\
1819 	}								\
1820 	INP_WUNLOCK(inp);						\
1821 } while (/*CONSTCOND*/ 0)
1822 
1823 				case IPV6_RECVPKTINFO:
1824 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1825 					break;
1826 
1827 				case IPV6_HOPLIMIT:
1828 				{
1829 					struct ip6_pktopts **optp;
1830 
1831 					/* cannot mix with RFC2292 */
1832 					if (OPTBIT(IN6P_RFC2292)) {
1833 						error = EINVAL;
1834 						break;
1835 					}
1836 					INP_WLOCK(inp);
1837 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1838 						INP_WUNLOCK(inp);
1839 						return (ECONNRESET);
1840 					}
1841 					optp = &inp->in6p_outputopts;
1842 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1843 					    (u_char *)&optval, sizeof(optval),
1844 					    optp, (td != NULL) ? td->td_ucred :
1845 					    NULL, uproto);
1846 					INP_WUNLOCK(inp);
1847 					break;
1848 				}
1849 
1850 				case IPV6_RECVHOPLIMIT:
1851 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1852 					break;
1853 
1854 				case IPV6_RECVHOPOPTS:
1855 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1856 					break;
1857 
1858 				case IPV6_RECVDSTOPTS:
1859 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1860 					break;
1861 
1862 				case IPV6_RECVRTHDRDSTOPTS:
1863 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1864 					break;
1865 
1866 				case IPV6_RECVRTHDR:
1867 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1868 					break;
1869 
1870 				case IPV6_RECVPATHMTU:
1871 					/*
1872 					 * We ignore this option for TCP
1873 					 * sockets.
1874 					 * (RFC3542 leaves this case
1875 					 * unspecified.)
1876 					 */
1877 					if (uproto != IPPROTO_TCP)
1878 						OPTSET(IN6P_MTU);
1879 					break;
1880 
1881 				case IPV6_RECVFLOWID:
1882 					OPTSET2(INP_RECVFLOWID, optval);
1883 					break;
1884 
1885 #ifdef	RSS
1886 				case IPV6_RECVRSSBUCKETID:
1887 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1888 					break;
1889 #endif
1890 
1891 				case IPV6_V6ONLY:
1892 					INP_WLOCK(inp);
1893 					if (inp->inp_lport ||
1894 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1895 						/*
1896 						 * The socket is already bound.
1897 						 */
1898 						INP_WUNLOCK(inp);
1899 						error = EINVAL;
1900 						break;
1901 					}
1902 					if (optval) {
1903 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1904 						inp->inp_vflag &= ~INP_IPV4;
1905 					} else {
1906 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1907 						inp->inp_vflag |= INP_IPV4;
1908 					}
1909 					INP_WUNLOCK(inp);
1910 					break;
1911 				case IPV6_RECVTCLASS:
1912 					/* cannot mix with RFC2292 XXX */
1913 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1914 					break;
1915 				case IPV6_AUTOFLOWLABEL:
1916 					OPTSET(IN6P_AUTOFLOWLABEL);
1917 					break;
1918 
1919 				case IPV6_ORIGDSTADDR:
1920 					OPTSET2(INP_ORIGDSTADDR, optval);
1921 					break;
1922 				case IPV6_BINDANY:
1923 					OPTSET(INP_BINDANY);
1924 					break;
1925 
1926 				case IPV6_BINDMULTI:
1927 					OPTSET2(INP_BINDMULTI, optval);
1928 					break;
1929 #ifdef	RSS
1930 				case IPV6_RSS_LISTEN_BUCKET:
1931 					if ((optval >= 0) &&
1932 					    (optval < rss_getnumbuckets())) {
1933 						INP_WLOCK(inp);
1934 						inp->inp_rss_listen_bucket = optval;
1935 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1936 						INP_WUNLOCK(inp);
1937 					} else {
1938 						error = EINVAL;
1939 					}
1940 					break;
1941 #endif
1942 				}
1943 				break;
1944 
1945 			case IPV6_TCLASS:
1946 			case IPV6_DONTFRAG:
1947 			case IPV6_USE_MIN_MTU:
1948 			case IPV6_PREFER_TEMPADDR:
1949 				if (optlen != sizeof(optval)) {
1950 					error = EINVAL;
1951 					break;
1952 				}
1953 				error = sooptcopyin(sopt, &optval,
1954 					sizeof optval, sizeof optval);
1955 				if (error)
1956 					break;
1957 				{
1958 					struct ip6_pktopts **optp;
1959 					INP_WLOCK(inp);
1960 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1961 						INP_WUNLOCK(inp);
1962 						return (ECONNRESET);
1963 					}
1964 					optp = &inp->in6p_outputopts;
1965 					error = ip6_pcbopt(optname,
1966 					    (u_char *)&optval, sizeof(optval),
1967 					    optp, (td != NULL) ? td->td_ucred :
1968 					    NULL, uproto);
1969 					INP_WUNLOCK(inp);
1970 					break;
1971 				}
1972 
1973 			case IPV6_2292PKTINFO:
1974 			case IPV6_2292HOPLIMIT:
1975 			case IPV6_2292HOPOPTS:
1976 			case IPV6_2292DSTOPTS:
1977 			case IPV6_2292RTHDR:
1978 				/* RFC 2292 */
1979 				if (optlen != sizeof(int)) {
1980 					error = EINVAL;
1981 					break;
1982 				}
1983 				error = sooptcopyin(sopt, &optval,
1984 					sizeof optval, sizeof optval);
1985 				if (error)
1986 					break;
1987 				switch (optname) {
1988 				case IPV6_2292PKTINFO:
1989 					OPTSET2292(IN6P_PKTINFO);
1990 					break;
1991 				case IPV6_2292HOPLIMIT:
1992 					OPTSET2292(IN6P_HOPLIMIT);
1993 					break;
1994 				case IPV6_2292HOPOPTS:
1995 					/*
1996 					 * Check super-user privilege.
1997 					 * See comments for IPV6_RECVHOPOPTS.
1998 					 */
1999 					if (td != NULL) {
2000 						error = priv_check(td,
2001 						    PRIV_NETINET_SETHDROPTS);
2002 						if (error)
2003 							return (error);
2004 					}
2005 					OPTSET2292(IN6P_HOPOPTS);
2006 					break;
2007 				case IPV6_2292DSTOPTS:
2008 					if (td != NULL) {
2009 						error = priv_check(td,
2010 						    PRIV_NETINET_SETHDROPTS);
2011 						if (error)
2012 							return (error);
2013 					}
2014 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2015 					break;
2016 				case IPV6_2292RTHDR:
2017 					OPTSET2292(IN6P_RTHDR);
2018 					break;
2019 				}
2020 				break;
2021 			case IPV6_PKTINFO:
2022 			case IPV6_HOPOPTS:
2023 			case IPV6_RTHDR:
2024 			case IPV6_DSTOPTS:
2025 			case IPV6_RTHDRDSTOPTS:
2026 			case IPV6_NEXTHOP:
2027 			{
2028 				/* new advanced API (RFC3542) */
2029 				u_char *optbuf;
2030 				u_char optbuf_storage[MCLBYTES];
2031 				int optlen;
2032 				struct ip6_pktopts **optp;
2033 
2034 				/* cannot mix with RFC2292 */
2035 				if (OPTBIT(IN6P_RFC2292)) {
2036 					error = EINVAL;
2037 					break;
2038 				}
2039 
2040 				/*
2041 				 * We only ensure valsize is not too large
2042 				 * here.  Further validation will be done
2043 				 * later.
2044 				 */
2045 				error = sooptcopyin(sopt, optbuf_storage,
2046 				    sizeof(optbuf_storage), 0);
2047 				if (error)
2048 					break;
2049 				optlen = sopt->sopt_valsize;
2050 				optbuf = optbuf_storage;
2051 				INP_WLOCK(inp);
2052 				if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2053 					INP_WUNLOCK(inp);
2054 					return (ECONNRESET);
2055 				}
2056 				optp = &inp->in6p_outputopts;
2057 				error = ip6_pcbopt(optname, optbuf, optlen,
2058 				    optp, (td != NULL) ? td->td_ucred : NULL,
2059 				    uproto);
2060 				INP_WUNLOCK(inp);
2061 				break;
2062 			}
2063 #undef OPTSET
2064 
2065 			case IPV6_MULTICAST_IF:
2066 			case IPV6_MULTICAST_HOPS:
2067 			case IPV6_MULTICAST_LOOP:
2068 			case IPV6_JOIN_GROUP:
2069 			case IPV6_LEAVE_GROUP:
2070 			case IPV6_MSFILTER:
2071 			case MCAST_BLOCK_SOURCE:
2072 			case MCAST_UNBLOCK_SOURCE:
2073 			case MCAST_JOIN_GROUP:
2074 			case MCAST_LEAVE_GROUP:
2075 			case MCAST_JOIN_SOURCE_GROUP:
2076 			case MCAST_LEAVE_SOURCE_GROUP:
2077 				error = ip6_setmoptions(inp, sopt);
2078 				break;
2079 
2080 			case IPV6_PORTRANGE:
2081 				error = sooptcopyin(sopt, &optval,
2082 				    sizeof optval, sizeof optval);
2083 				if (error)
2084 					break;
2085 
2086 				INP_WLOCK(inp);
2087 				switch (optval) {
2088 				case IPV6_PORTRANGE_DEFAULT:
2089 					inp->inp_flags &= ~(INP_LOWPORT);
2090 					inp->inp_flags &= ~(INP_HIGHPORT);
2091 					break;
2092 
2093 				case IPV6_PORTRANGE_HIGH:
2094 					inp->inp_flags &= ~(INP_LOWPORT);
2095 					inp->inp_flags |= INP_HIGHPORT;
2096 					break;
2097 
2098 				case IPV6_PORTRANGE_LOW:
2099 					inp->inp_flags &= ~(INP_HIGHPORT);
2100 					inp->inp_flags |= INP_LOWPORT;
2101 					break;
2102 
2103 				default:
2104 					error = EINVAL;
2105 					break;
2106 				}
2107 				INP_WUNLOCK(inp);
2108 				break;
2109 
2110 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2111 			case IPV6_IPSEC_POLICY:
2112 				if (IPSEC_ENABLED(ipv6)) {
2113 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2114 					break;
2115 				}
2116 				/* FALLTHROUGH */
2117 #endif /* IPSEC */
2118 
2119 			default:
2120 				error = ENOPROTOOPT;
2121 				break;
2122 			}
2123 			break;
2124 
2125 		case SOPT_GET:
2126 			switch (optname) {
2127 
2128 			case IPV6_2292PKTOPTIONS:
2129 #ifdef IPV6_PKTOPTIONS
2130 			case IPV6_PKTOPTIONS:
2131 #endif
2132 				/*
2133 				 * RFC3542 (effectively) deprecated the
2134 				 * semantics of the 2292-style pktoptions.
2135 				 * Since it was not reliable in nature (i.e.,
2136 				 * applications had to expect the lack of some
2137 				 * information after all), it would make sense
2138 				 * to simplify this part by always returning
2139 				 * empty data.
2140 				 */
2141 				sopt->sopt_valsize = 0;
2142 				break;
2143 
2144 			case IPV6_RECVHOPOPTS:
2145 			case IPV6_RECVDSTOPTS:
2146 			case IPV6_RECVRTHDRDSTOPTS:
2147 			case IPV6_UNICAST_HOPS:
2148 			case IPV6_RECVPKTINFO:
2149 			case IPV6_RECVHOPLIMIT:
2150 			case IPV6_RECVRTHDR:
2151 			case IPV6_RECVPATHMTU:
2152 
2153 			case IPV6_V6ONLY:
2154 			case IPV6_PORTRANGE:
2155 			case IPV6_RECVTCLASS:
2156 			case IPV6_AUTOFLOWLABEL:
2157 			case IPV6_BINDANY:
2158 			case IPV6_FLOWID:
2159 			case IPV6_FLOWTYPE:
2160 			case IPV6_RECVFLOWID:
2161 #ifdef	RSS
2162 			case IPV6_RSSBUCKETID:
2163 			case IPV6_RECVRSSBUCKETID:
2164 #endif
2165 			case IPV6_BINDMULTI:
2166 				switch (optname) {
2167 
2168 				case IPV6_RECVHOPOPTS:
2169 					optval = OPTBIT(IN6P_HOPOPTS);
2170 					break;
2171 
2172 				case IPV6_RECVDSTOPTS:
2173 					optval = OPTBIT(IN6P_DSTOPTS);
2174 					break;
2175 
2176 				case IPV6_RECVRTHDRDSTOPTS:
2177 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2178 					break;
2179 
2180 				case IPV6_UNICAST_HOPS:
2181 					optval = inp->in6p_hops;
2182 					break;
2183 
2184 				case IPV6_RECVPKTINFO:
2185 					optval = OPTBIT(IN6P_PKTINFO);
2186 					break;
2187 
2188 				case IPV6_RECVHOPLIMIT:
2189 					optval = OPTBIT(IN6P_HOPLIMIT);
2190 					break;
2191 
2192 				case IPV6_RECVRTHDR:
2193 					optval = OPTBIT(IN6P_RTHDR);
2194 					break;
2195 
2196 				case IPV6_RECVPATHMTU:
2197 					optval = OPTBIT(IN6P_MTU);
2198 					break;
2199 
2200 				case IPV6_V6ONLY:
2201 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2202 					break;
2203 
2204 				case IPV6_PORTRANGE:
2205 				    {
2206 					int flags;
2207 					flags = inp->inp_flags;
2208 					if (flags & INP_HIGHPORT)
2209 						optval = IPV6_PORTRANGE_HIGH;
2210 					else if (flags & INP_LOWPORT)
2211 						optval = IPV6_PORTRANGE_LOW;
2212 					else
2213 						optval = 0;
2214 					break;
2215 				    }
2216 				case IPV6_RECVTCLASS:
2217 					optval = OPTBIT(IN6P_TCLASS);
2218 					break;
2219 
2220 				case IPV6_AUTOFLOWLABEL:
2221 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2222 					break;
2223 
2224 				case IPV6_ORIGDSTADDR:
2225 					optval = OPTBIT2(INP_ORIGDSTADDR);
2226 					break;
2227 
2228 				case IPV6_BINDANY:
2229 					optval = OPTBIT(INP_BINDANY);
2230 					break;
2231 
2232 				case IPV6_FLOWID:
2233 					optval = inp->inp_flowid;
2234 					break;
2235 
2236 				case IPV6_FLOWTYPE:
2237 					optval = inp->inp_flowtype;
2238 					break;
2239 
2240 				case IPV6_RECVFLOWID:
2241 					optval = OPTBIT2(INP_RECVFLOWID);
2242 					break;
2243 #ifdef	RSS
2244 				case IPV6_RSSBUCKETID:
2245 					retval =
2246 					    rss_hash2bucket(inp->inp_flowid,
2247 					    inp->inp_flowtype,
2248 					    &rss_bucket);
2249 					if (retval == 0)
2250 						optval = rss_bucket;
2251 					else
2252 						error = EINVAL;
2253 					break;
2254 
2255 				case IPV6_RECVRSSBUCKETID:
2256 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2257 					break;
2258 #endif
2259 
2260 				case IPV6_BINDMULTI:
2261 					optval = OPTBIT2(INP_BINDMULTI);
2262 					break;
2263 
2264 				}
2265 				if (error)
2266 					break;
2267 				error = sooptcopyout(sopt, &optval,
2268 					sizeof optval);
2269 				break;
2270 
2271 			case IPV6_PATHMTU:
2272 			{
2273 				u_long pmtu = 0;
2274 				struct ip6_mtuinfo mtuinfo;
2275 				struct in6_addr addr;
2276 
2277 				if (!(so->so_state & SS_ISCONNECTED))
2278 					return (ENOTCONN);
2279 				/*
2280 				 * XXX: we dot not consider the case of source
2281 				 * routing, or optional information to specify
2282 				 * the outgoing interface.
2283 				 * Copy faddr out of inp to avoid holding lock
2284 				 * on inp during route lookup.
2285 				 */
2286 				INP_RLOCK(inp);
2287 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2288 				INP_RUNLOCK(inp);
2289 				error = ip6_getpmtu_ctl(so->so_fibnum,
2290 				    &addr, &pmtu);
2291 				if (error)
2292 					break;
2293 				if (pmtu > IPV6_MAXPACKET)
2294 					pmtu = IPV6_MAXPACKET;
2295 
2296 				bzero(&mtuinfo, sizeof(mtuinfo));
2297 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2298 				optdata = (void *)&mtuinfo;
2299 				optdatalen = sizeof(mtuinfo);
2300 				error = sooptcopyout(sopt, optdata,
2301 				    optdatalen);
2302 				break;
2303 			}
2304 
2305 			case IPV6_2292PKTINFO:
2306 			case IPV6_2292HOPLIMIT:
2307 			case IPV6_2292HOPOPTS:
2308 			case IPV6_2292RTHDR:
2309 			case IPV6_2292DSTOPTS:
2310 				switch (optname) {
2311 				case IPV6_2292PKTINFO:
2312 					optval = OPTBIT(IN6P_PKTINFO);
2313 					break;
2314 				case IPV6_2292HOPLIMIT:
2315 					optval = OPTBIT(IN6P_HOPLIMIT);
2316 					break;
2317 				case IPV6_2292HOPOPTS:
2318 					optval = OPTBIT(IN6P_HOPOPTS);
2319 					break;
2320 				case IPV6_2292RTHDR:
2321 					optval = OPTBIT(IN6P_RTHDR);
2322 					break;
2323 				case IPV6_2292DSTOPTS:
2324 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2325 					break;
2326 				}
2327 				error = sooptcopyout(sopt, &optval,
2328 				    sizeof optval);
2329 				break;
2330 			case IPV6_PKTINFO:
2331 			case IPV6_HOPOPTS:
2332 			case IPV6_RTHDR:
2333 			case IPV6_DSTOPTS:
2334 			case IPV6_RTHDRDSTOPTS:
2335 			case IPV6_NEXTHOP:
2336 			case IPV6_TCLASS:
2337 			case IPV6_DONTFRAG:
2338 			case IPV6_USE_MIN_MTU:
2339 			case IPV6_PREFER_TEMPADDR:
2340 				error = ip6_getpcbopt(inp, optname, sopt);
2341 				break;
2342 
2343 			case IPV6_MULTICAST_IF:
2344 			case IPV6_MULTICAST_HOPS:
2345 			case IPV6_MULTICAST_LOOP:
2346 			case IPV6_MSFILTER:
2347 				error = ip6_getmoptions(inp, sopt);
2348 				break;
2349 
2350 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2351 			case IPV6_IPSEC_POLICY:
2352 				if (IPSEC_ENABLED(ipv6)) {
2353 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2354 					break;
2355 				}
2356 				/* FALLTHROUGH */
2357 #endif /* IPSEC */
2358 			default:
2359 				error = ENOPROTOOPT;
2360 				break;
2361 			}
2362 			break;
2363 		}
2364 	}
2365 	return (error);
2366 }
2367 
2368 int
2369 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2370 {
2371 	int error = 0, optval, optlen;
2372 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2373 	struct inpcb *inp = sotoinpcb(so);
2374 	int level, op, optname;
2375 
2376 	level = sopt->sopt_level;
2377 	op = sopt->sopt_dir;
2378 	optname = sopt->sopt_name;
2379 	optlen = sopt->sopt_valsize;
2380 
2381 	if (level != IPPROTO_IPV6) {
2382 		return (EINVAL);
2383 	}
2384 
2385 	switch (optname) {
2386 	case IPV6_CHECKSUM:
2387 		/*
2388 		 * For ICMPv6 sockets, no modification allowed for checksum
2389 		 * offset, permit "no change" values to help existing apps.
2390 		 *
2391 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2392 		 * for an ICMPv6 socket will fail."
2393 		 * The current behavior does not meet RFC3542.
2394 		 */
2395 		switch (op) {
2396 		case SOPT_SET:
2397 			if (optlen != sizeof(int)) {
2398 				error = EINVAL;
2399 				break;
2400 			}
2401 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2402 					    sizeof(optval));
2403 			if (error)
2404 				break;
2405 			if (optval < -1 || (optval % 2) != 0) {
2406 				/*
2407 				 * The API assumes non-negative even offset
2408 				 * values or -1 as a special value.
2409 				 */
2410 				error = EINVAL;
2411 			} else if (so->so_proto->pr_protocol ==
2412 			    IPPROTO_ICMPV6) {
2413 				if (optval != icmp6off)
2414 					error = EINVAL;
2415 			} else
2416 				inp->in6p_cksum = optval;
2417 			break;
2418 
2419 		case SOPT_GET:
2420 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2421 				optval = icmp6off;
2422 			else
2423 				optval = inp->in6p_cksum;
2424 
2425 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2426 			break;
2427 
2428 		default:
2429 			error = EINVAL;
2430 			break;
2431 		}
2432 		break;
2433 
2434 	default:
2435 		error = ENOPROTOOPT;
2436 		break;
2437 	}
2438 
2439 	return (error);
2440 }
2441 
2442 /*
2443  * Set up IP6 options in pcb for insertion in output packets or
2444  * specifying behavior of outgoing packets.
2445  */
2446 static int
2447 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2448     struct socket *so, struct sockopt *sopt)
2449 {
2450 	struct ip6_pktopts *opt = *pktopt;
2451 	int error = 0;
2452 	struct thread *td = sopt->sopt_td;
2453 
2454 	/* turn off any old options. */
2455 	if (opt) {
2456 #ifdef DIAGNOSTIC
2457 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2458 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2459 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2460 			printf("ip6_pcbopts: all specified options are cleared.\n");
2461 #endif
2462 		ip6_clearpktopts(opt, -1);
2463 	} else {
2464 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2465 		if (opt == NULL)
2466 			return (ENOMEM);
2467 	}
2468 	*pktopt = NULL;
2469 
2470 	if (!m || m->m_len == 0) {
2471 		/*
2472 		 * Only turning off any previous options, regardless of
2473 		 * whether the opt is just created or given.
2474 		 */
2475 		free(opt, M_IP6OPT);
2476 		return (0);
2477 	}
2478 
2479 	/*  set options specified by user. */
2480 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2481 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2482 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2483 		free(opt, M_IP6OPT);
2484 		return (error);
2485 	}
2486 	*pktopt = opt;
2487 	return (0);
2488 }
2489 
2490 /*
2491  * initialize ip6_pktopts.  beware that there are non-zero default values in
2492  * the struct.
2493  */
2494 void
2495 ip6_initpktopts(struct ip6_pktopts *opt)
2496 {
2497 
2498 	bzero(opt, sizeof(*opt));
2499 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2500 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2501 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2502 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2503 }
2504 
2505 static int
2506 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2507     struct ucred *cred, int uproto)
2508 {
2509 	struct ip6_pktopts *opt;
2510 
2511 	if (*pktopt == NULL) {
2512 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2513 		    M_NOWAIT);
2514 		if (*pktopt == NULL)
2515 			return (ENOBUFS);
2516 		ip6_initpktopts(*pktopt);
2517 	}
2518 	opt = *pktopt;
2519 
2520 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2521 }
2522 
2523 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2524 	if (pktopt && pktopt->field) {						\
2525 		INP_RUNLOCK(inp);						\
2526 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2527 		malloc_optdata = true;						\
2528 		INP_RLOCK(inp);							\
2529 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2530 			INP_RUNLOCK(inp);					\
2531 			free(optdata, M_TEMP);					\
2532 			return (ECONNRESET);					\
2533 		}								\
2534 		pktopt = inp->in6p_outputopts;					\
2535 		if (pktopt && pktopt->field) {					\
2536 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2537 			bcopy(&pktopt->field, optdata, optdatalen);		\
2538 		} else {							\
2539 			free(optdata, M_TEMP);					\
2540 			optdata = NULL;						\
2541 			malloc_optdata = false;					\
2542 		}								\
2543 	}									\
2544 } while(0)
2545 
2546 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2547 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2548 
2549 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2550 	pktopt->field->sa_len)
2551 
2552 static int
2553 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2554 {
2555 	void *optdata = NULL;
2556 	bool malloc_optdata = false;
2557 	int optdatalen = 0;
2558 	int error = 0;
2559 	struct in6_pktinfo null_pktinfo;
2560 	int deftclass = 0, on;
2561 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2562 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2563 	struct ip6_pktopts *pktopt;
2564 
2565 	INP_RLOCK(inp);
2566 	pktopt = inp->in6p_outputopts;
2567 
2568 	switch (optname) {
2569 	case IPV6_PKTINFO:
2570 		optdata = (void *)&null_pktinfo;
2571 		if (pktopt && pktopt->ip6po_pktinfo) {
2572 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2573 			    sizeof(null_pktinfo));
2574 			in6_clearscope(&null_pktinfo.ipi6_addr);
2575 		} else {
2576 			/* XXX: we don't have to do this every time... */
2577 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2578 		}
2579 		optdatalen = sizeof(struct in6_pktinfo);
2580 		break;
2581 	case IPV6_TCLASS:
2582 		if (pktopt && pktopt->ip6po_tclass >= 0)
2583 			deftclass = pktopt->ip6po_tclass;
2584 		optdata = (void *)&deftclass;
2585 		optdatalen = sizeof(int);
2586 		break;
2587 	case IPV6_HOPOPTS:
2588 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2589 		break;
2590 	case IPV6_RTHDR:
2591 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2592 		break;
2593 	case IPV6_RTHDRDSTOPTS:
2594 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2595 		break;
2596 	case IPV6_DSTOPTS:
2597 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2598 		break;
2599 	case IPV6_NEXTHOP:
2600 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2601 		break;
2602 	case IPV6_USE_MIN_MTU:
2603 		if (pktopt)
2604 			defminmtu = pktopt->ip6po_minmtu;
2605 		optdata = (void *)&defminmtu;
2606 		optdatalen = sizeof(int);
2607 		break;
2608 	case IPV6_DONTFRAG:
2609 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2610 			on = 1;
2611 		else
2612 			on = 0;
2613 		optdata = (void *)&on;
2614 		optdatalen = sizeof(on);
2615 		break;
2616 	case IPV6_PREFER_TEMPADDR:
2617 		if (pktopt)
2618 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2619 		optdata = (void *)&defpreftemp;
2620 		optdatalen = sizeof(int);
2621 		break;
2622 	default:		/* should not happen */
2623 #ifdef DIAGNOSTIC
2624 		panic("ip6_getpcbopt: unexpected option\n");
2625 #endif
2626 		INP_RUNLOCK(inp);
2627 		return (ENOPROTOOPT);
2628 	}
2629 	INP_RUNLOCK(inp);
2630 
2631 	error = sooptcopyout(sopt, optdata, optdatalen);
2632 	if (malloc_optdata)
2633 		free(optdata, M_TEMP);
2634 
2635 	return (error);
2636 }
2637 
2638 void
2639 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2640 {
2641 	if (pktopt == NULL)
2642 		return;
2643 
2644 	if (optname == -1 || optname == IPV6_PKTINFO) {
2645 		if (pktopt->ip6po_pktinfo)
2646 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2647 		pktopt->ip6po_pktinfo = NULL;
2648 	}
2649 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2650 		pktopt->ip6po_hlim = -1;
2651 	if (optname == -1 || optname == IPV6_TCLASS)
2652 		pktopt->ip6po_tclass = -1;
2653 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2654 		if (pktopt->ip6po_nextroute.ro_rt) {
2655 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2656 			pktopt->ip6po_nextroute.ro_rt = NULL;
2657 		}
2658 		if (pktopt->ip6po_nexthop)
2659 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2660 		pktopt->ip6po_nexthop = NULL;
2661 	}
2662 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2663 		if (pktopt->ip6po_hbh)
2664 			free(pktopt->ip6po_hbh, M_IP6OPT);
2665 		pktopt->ip6po_hbh = NULL;
2666 	}
2667 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2668 		if (pktopt->ip6po_dest1)
2669 			free(pktopt->ip6po_dest1, M_IP6OPT);
2670 		pktopt->ip6po_dest1 = NULL;
2671 	}
2672 	if (optname == -1 || optname == IPV6_RTHDR) {
2673 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2674 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2675 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2676 		if (pktopt->ip6po_route.ro_rt) {
2677 			RTFREE(pktopt->ip6po_route.ro_rt);
2678 			pktopt->ip6po_route.ro_rt = NULL;
2679 		}
2680 	}
2681 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2682 		if (pktopt->ip6po_dest2)
2683 			free(pktopt->ip6po_dest2, M_IP6OPT);
2684 		pktopt->ip6po_dest2 = NULL;
2685 	}
2686 }
2687 
2688 #define PKTOPT_EXTHDRCPY(type) \
2689 do {\
2690 	if (src->type) {\
2691 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2692 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2693 		if (dst->type == NULL)\
2694 			goto bad;\
2695 		bcopy(src->type, dst->type, hlen);\
2696 	}\
2697 } while (/*CONSTCOND*/ 0)
2698 
2699 static int
2700 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2701 {
2702 	if (dst == NULL || src == NULL)  {
2703 		printf("ip6_clearpktopts: invalid argument\n");
2704 		return (EINVAL);
2705 	}
2706 
2707 	dst->ip6po_hlim = src->ip6po_hlim;
2708 	dst->ip6po_tclass = src->ip6po_tclass;
2709 	dst->ip6po_flags = src->ip6po_flags;
2710 	dst->ip6po_minmtu = src->ip6po_minmtu;
2711 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2712 	if (src->ip6po_pktinfo) {
2713 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2714 		    M_IP6OPT, canwait);
2715 		if (dst->ip6po_pktinfo == NULL)
2716 			goto bad;
2717 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2718 	}
2719 	if (src->ip6po_nexthop) {
2720 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2721 		    M_IP6OPT, canwait);
2722 		if (dst->ip6po_nexthop == NULL)
2723 			goto bad;
2724 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2725 		    src->ip6po_nexthop->sa_len);
2726 	}
2727 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2728 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2729 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2730 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2731 	return (0);
2732 
2733   bad:
2734 	ip6_clearpktopts(dst, -1);
2735 	return (ENOBUFS);
2736 }
2737 #undef PKTOPT_EXTHDRCPY
2738 
2739 struct ip6_pktopts *
2740 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2741 {
2742 	int error;
2743 	struct ip6_pktopts *dst;
2744 
2745 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2746 	if (dst == NULL)
2747 		return (NULL);
2748 	ip6_initpktopts(dst);
2749 
2750 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2751 		free(dst, M_IP6OPT);
2752 		return (NULL);
2753 	}
2754 
2755 	return (dst);
2756 }
2757 
2758 void
2759 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2760 {
2761 	if (pktopt == NULL)
2762 		return;
2763 
2764 	ip6_clearpktopts(pktopt, -1);
2765 
2766 	free(pktopt, M_IP6OPT);
2767 }
2768 
2769 /*
2770  * Set IPv6 outgoing packet options based on advanced API.
2771  */
2772 int
2773 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2774     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2775 {
2776 	struct cmsghdr *cm = NULL;
2777 
2778 	if (control == NULL || opt == NULL)
2779 		return (EINVAL);
2780 
2781 	ip6_initpktopts(opt);
2782 	if (stickyopt) {
2783 		int error;
2784 
2785 		/*
2786 		 * If stickyopt is provided, make a local copy of the options
2787 		 * for this particular packet, then override them by ancillary
2788 		 * objects.
2789 		 * XXX: copypktopts() does not copy the cached route to a next
2790 		 * hop (if any).  This is not very good in terms of efficiency,
2791 		 * but we can allow this since this option should be rarely
2792 		 * used.
2793 		 */
2794 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2795 			return (error);
2796 	}
2797 
2798 	/*
2799 	 * XXX: Currently, we assume all the optional information is stored
2800 	 * in a single mbuf.
2801 	 */
2802 	if (control->m_next)
2803 		return (EINVAL);
2804 
2805 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2806 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2807 		int error;
2808 
2809 		if (control->m_len < CMSG_LEN(0))
2810 			return (EINVAL);
2811 
2812 		cm = mtod(control, struct cmsghdr *);
2813 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2814 			return (EINVAL);
2815 		if (cm->cmsg_level != IPPROTO_IPV6)
2816 			continue;
2817 
2818 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2819 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2820 		if (error)
2821 			return (error);
2822 	}
2823 
2824 	return (0);
2825 }
2826 
2827 /*
2828  * Set a particular packet option, as a sticky option or an ancillary data
2829  * item.  "len" can be 0 only when it's a sticky option.
2830  * We have 4 cases of combination of "sticky" and "cmsg":
2831  * "sticky=0, cmsg=0": impossible
2832  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2833  * "sticky=1, cmsg=0": RFC3542 socket option
2834  * "sticky=1, cmsg=1": RFC2292 socket option
2835  */
2836 static int
2837 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2838     struct ucred *cred, int sticky, int cmsg, int uproto)
2839 {
2840 	int minmtupolicy, preftemp;
2841 	int error;
2842 
2843 	if (!sticky && !cmsg) {
2844 #ifdef DIAGNOSTIC
2845 		printf("ip6_setpktopt: impossible case\n");
2846 #endif
2847 		return (EINVAL);
2848 	}
2849 
2850 	/*
2851 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2852 	 * not be specified in the context of RFC3542.  Conversely,
2853 	 * RFC3542 types should not be specified in the context of RFC2292.
2854 	 */
2855 	if (!cmsg) {
2856 		switch (optname) {
2857 		case IPV6_2292PKTINFO:
2858 		case IPV6_2292HOPLIMIT:
2859 		case IPV6_2292NEXTHOP:
2860 		case IPV6_2292HOPOPTS:
2861 		case IPV6_2292DSTOPTS:
2862 		case IPV6_2292RTHDR:
2863 		case IPV6_2292PKTOPTIONS:
2864 			return (ENOPROTOOPT);
2865 		}
2866 	}
2867 	if (sticky && cmsg) {
2868 		switch (optname) {
2869 		case IPV6_PKTINFO:
2870 		case IPV6_HOPLIMIT:
2871 		case IPV6_NEXTHOP:
2872 		case IPV6_HOPOPTS:
2873 		case IPV6_DSTOPTS:
2874 		case IPV6_RTHDRDSTOPTS:
2875 		case IPV6_RTHDR:
2876 		case IPV6_USE_MIN_MTU:
2877 		case IPV6_DONTFRAG:
2878 		case IPV6_TCLASS:
2879 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2880 			return (ENOPROTOOPT);
2881 		}
2882 	}
2883 
2884 	switch (optname) {
2885 	case IPV6_2292PKTINFO:
2886 	case IPV6_PKTINFO:
2887 	{
2888 		struct ifnet *ifp = NULL;
2889 		struct in6_pktinfo *pktinfo;
2890 
2891 		if (len != sizeof(struct in6_pktinfo))
2892 			return (EINVAL);
2893 
2894 		pktinfo = (struct in6_pktinfo *)buf;
2895 
2896 		/*
2897 		 * An application can clear any sticky IPV6_PKTINFO option by
2898 		 * doing a "regular" setsockopt with ipi6_addr being
2899 		 * in6addr_any and ipi6_ifindex being zero.
2900 		 * [RFC 3542, Section 6]
2901 		 */
2902 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2903 		    pktinfo->ipi6_ifindex == 0 &&
2904 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2905 			ip6_clearpktopts(opt, optname);
2906 			break;
2907 		}
2908 
2909 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2910 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2911 			return (EINVAL);
2912 		}
2913 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2914 			return (EINVAL);
2915 		/* validate the interface index if specified. */
2916 		if (pktinfo->ipi6_ifindex > V_if_index)
2917 			 return (ENXIO);
2918 		if (pktinfo->ipi6_ifindex) {
2919 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2920 			if (ifp == NULL)
2921 				return (ENXIO);
2922 		}
2923 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2924 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2925 			return (ENETDOWN);
2926 
2927 		if (ifp != NULL &&
2928 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2929 			struct in6_ifaddr *ia;
2930 
2931 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2932 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2933 			if (ia == NULL)
2934 				return (EADDRNOTAVAIL);
2935 			ifa_free(&ia->ia_ifa);
2936 		}
2937 		/*
2938 		 * We store the address anyway, and let in6_selectsrc()
2939 		 * validate the specified address.  This is because ipi6_addr
2940 		 * may not have enough information about its scope zone, and
2941 		 * we may need additional information (such as outgoing
2942 		 * interface or the scope zone of a destination address) to
2943 		 * disambiguate the scope.
2944 		 * XXX: the delay of the validation may confuse the
2945 		 * application when it is used as a sticky option.
2946 		 */
2947 		if (opt->ip6po_pktinfo == NULL) {
2948 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2949 			    M_IP6OPT, M_NOWAIT);
2950 			if (opt->ip6po_pktinfo == NULL)
2951 				return (ENOBUFS);
2952 		}
2953 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2954 		break;
2955 	}
2956 
2957 	case IPV6_2292HOPLIMIT:
2958 	case IPV6_HOPLIMIT:
2959 	{
2960 		int *hlimp;
2961 
2962 		/*
2963 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2964 		 * to simplify the ordering among hoplimit options.
2965 		 */
2966 		if (optname == IPV6_HOPLIMIT && sticky)
2967 			return (ENOPROTOOPT);
2968 
2969 		if (len != sizeof(int))
2970 			return (EINVAL);
2971 		hlimp = (int *)buf;
2972 		if (*hlimp < -1 || *hlimp > 255)
2973 			return (EINVAL);
2974 
2975 		opt->ip6po_hlim = *hlimp;
2976 		break;
2977 	}
2978 
2979 	case IPV6_TCLASS:
2980 	{
2981 		int tclass;
2982 
2983 		if (len != sizeof(int))
2984 			return (EINVAL);
2985 		tclass = *(int *)buf;
2986 		if (tclass < -1 || tclass > 255)
2987 			return (EINVAL);
2988 
2989 		opt->ip6po_tclass = tclass;
2990 		break;
2991 	}
2992 
2993 	case IPV6_2292NEXTHOP:
2994 	case IPV6_NEXTHOP:
2995 		if (cred != NULL) {
2996 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2997 			if (error)
2998 				return (error);
2999 		}
3000 
3001 		if (len == 0) {	/* just remove the option */
3002 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3003 			break;
3004 		}
3005 
3006 		/* check if cmsg_len is large enough for sa_len */
3007 		if (len < sizeof(struct sockaddr) || len < *buf)
3008 			return (EINVAL);
3009 
3010 		switch (((struct sockaddr *)buf)->sa_family) {
3011 		case AF_INET6:
3012 		{
3013 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3014 			int error;
3015 
3016 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3017 				return (EINVAL);
3018 
3019 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3020 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3021 				return (EINVAL);
3022 			}
3023 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3024 			    != 0) {
3025 				return (error);
3026 			}
3027 			break;
3028 		}
3029 		case AF_LINK:	/* should eventually be supported */
3030 		default:
3031 			return (EAFNOSUPPORT);
3032 		}
3033 
3034 		/* turn off the previous option, then set the new option. */
3035 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3036 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3037 		if (opt->ip6po_nexthop == NULL)
3038 			return (ENOBUFS);
3039 		bcopy(buf, opt->ip6po_nexthop, *buf);
3040 		break;
3041 
3042 	case IPV6_2292HOPOPTS:
3043 	case IPV6_HOPOPTS:
3044 	{
3045 		struct ip6_hbh *hbh;
3046 		int hbhlen;
3047 
3048 		/*
3049 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3050 		 * options, since per-option restriction has too much
3051 		 * overhead.
3052 		 */
3053 		if (cred != NULL) {
3054 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3055 			if (error)
3056 				return (error);
3057 		}
3058 
3059 		if (len == 0) {
3060 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3061 			break;	/* just remove the option */
3062 		}
3063 
3064 		/* message length validation */
3065 		if (len < sizeof(struct ip6_hbh))
3066 			return (EINVAL);
3067 		hbh = (struct ip6_hbh *)buf;
3068 		hbhlen = (hbh->ip6h_len + 1) << 3;
3069 		if (len != hbhlen)
3070 			return (EINVAL);
3071 
3072 		/* turn off the previous option, then set the new option. */
3073 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3074 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3075 		if (opt->ip6po_hbh == NULL)
3076 			return (ENOBUFS);
3077 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3078 
3079 		break;
3080 	}
3081 
3082 	case IPV6_2292DSTOPTS:
3083 	case IPV6_DSTOPTS:
3084 	case IPV6_RTHDRDSTOPTS:
3085 	{
3086 		struct ip6_dest *dest, **newdest = NULL;
3087 		int destlen;
3088 
3089 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3090 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3091 			if (error)
3092 				return (error);
3093 		}
3094 
3095 		if (len == 0) {
3096 			ip6_clearpktopts(opt, optname);
3097 			break;	/* just remove the option */
3098 		}
3099 
3100 		/* message length validation */
3101 		if (len < sizeof(struct ip6_dest))
3102 			return (EINVAL);
3103 		dest = (struct ip6_dest *)buf;
3104 		destlen = (dest->ip6d_len + 1) << 3;
3105 		if (len != destlen)
3106 			return (EINVAL);
3107 
3108 		/*
3109 		 * Determine the position that the destination options header
3110 		 * should be inserted; before or after the routing header.
3111 		 */
3112 		switch (optname) {
3113 		case IPV6_2292DSTOPTS:
3114 			/*
3115 			 * The old advacned API is ambiguous on this point.
3116 			 * Our approach is to determine the position based
3117 			 * according to the existence of a routing header.
3118 			 * Note, however, that this depends on the order of the
3119 			 * extension headers in the ancillary data; the 1st
3120 			 * part of the destination options header must appear
3121 			 * before the routing header in the ancillary data,
3122 			 * too.
3123 			 * RFC3542 solved the ambiguity by introducing
3124 			 * separate ancillary data or option types.
3125 			 */
3126 			if (opt->ip6po_rthdr == NULL)
3127 				newdest = &opt->ip6po_dest1;
3128 			else
3129 				newdest = &opt->ip6po_dest2;
3130 			break;
3131 		case IPV6_RTHDRDSTOPTS:
3132 			newdest = &opt->ip6po_dest1;
3133 			break;
3134 		case IPV6_DSTOPTS:
3135 			newdest = &opt->ip6po_dest2;
3136 			break;
3137 		}
3138 
3139 		/* turn off the previous option, then set the new option. */
3140 		ip6_clearpktopts(opt, optname);
3141 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3142 		if (*newdest == NULL)
3143 			return (ENOBUFS);
3144 		bcopy(dest, *newdest, destlen);
3145 
3146 		break;
3147 	}
3148 
3149 	case IPV6_2292RTHDR:
3150 	case IPV6_RTHDR:
3151 	{
3152 		struct ip6_rthdr *rth;
3153 		int rthlen;
3154 
3155 		if (len == 0) {
3156 			ip6_clearpktopts(opt, IPV6_RTHDR);
3157 			break;	/* just remove the option */
3158 		}
3159 
3160 		/* message length validation */
3161 		if (len < sizeof(struct ip6_rthdr))
3162 			return (EINVAL);
3163 		rth = (struct ip6_rthdr *)buf;
3164 		rthlen = (rth->ip6r_len + 1) << 3;
3165 		if (len != rthlen)
3166 			return (EINVAL);
3167 
3168 		switch (rth->ip6r_type) {
3169 		case IPV6_RTHDR_TYPE_0:
3170 			if (rth->ip6r_len == 0)	/* must contain one addr */
3171 				return (EINVAL);
3172 			if (rth->ip6r_len % 2) /* length must be even */
3173 				return (EINVAL);
3174 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3175 				return (EINVAL);
3176 			break;
3177 		default:
3178 			return (EINVAL);	/* not supported */
3179 		}
3180 
3181 		/* turn off the previous option */
3182 		ip6_clearpktopts(opt, IPV6_RTHDR);
3183 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3184 		if (opt->ip6po_rthdr == NULL)
3185 			return (ENOBUFS);
3186 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3187 
3188 		break;
3189 	}
3190 
3191 	case IPV6_USE_MIN_MTU:
3192 		if (len != sizeof(int))
3193 			return (EINVAL);
3194 		minmtupolicy = *(int *)buf;
3195 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3196 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3197 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3198 			return (EINVAL);
3199 		}
3200 		opt->ip6po_minmtu = minmtupolicy;
3201 		break;
3202 
3203 	case IPV6_DONTFRAG:
3204 		if (len != sizeof(int))
3205 			return (EINVAL);
3206 
3207 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3208 			/*
3209 			 * we ignore this option for TCP sockets.
3210 			 * (RFC3542 leaves this case unspecified.)
3211 			 */
3212 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3213 		} else
3214 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3215 		break;
3216 
3217 	case IPV6_PREFER_TEMPADDR:
3218 		if (len != sizeof(int))
3219 			return (EINVAL);
3220 		preftemp = *(int *)buf;
3221 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3222 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3223 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3224 			return (EINVAL);
3225 		}
3226 		opt->ip6po_prefer_tempaddr = preftemp;
3227 		break;
3228 
3229 	default:
3230 		return (ENOPROTOOPT);
3231 	} /* end of switch */
3232 
3233 	return (0);
3234 }
3235 
3236 /*
3237  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3238  * packet to the input queue of a specified interface.  Note that this
3239  * calls the output routine of the loopback "driver", but with an interface
3240  * pointer that might NOT be &loif -- easier than replicating that code here.
3241  */
3242 void
3243 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3244 {
3245 	struct mbuf *copym;
3246 	struct ip6_hdr *ip6;
3247 
3248 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3249 	if (copym == NULL)
3250 		return;
3251 
3252 	/*
3253 	 * Make sure to deep-copy IPv6 header portion in case the data
3254 	 * is in an mbuf cluster, so that we can safely override the IPv6
3255 	 * header portion later.
3256 	 */
3257 	if (!M_WRITABLE(copym) ||
3258 	    copym->m_len < sizeof(struct ip6_hdr)) {
3259 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3260 		if (copym == NULL)
3261 			return;
3262 	}
3263 	ip6 = mtod(copym, struct ip6_hdr *);
3264 	/*
3265 	 * clear embedded scope identifiers if necessary.
3266 	 * in6_clearscope will touch the addresses only when necessary.
3267 	 */
3268 	in6_clearscope(&ip6->ip6_src);
3269 	in6_clearscope(&ip6->ip6_dst);
3270 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3271 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3272 		    CSUM_PSEUDO_HDR;
3273 		copym->m_pkthdr.csum_data = 0xffff;
3274 	}
3275 	if_simloop(ifp, copym, AF_INET6, 0);
3276 }
3277 
3278 /*
3279  * Chop IPv6 header off from the payload.
3280  */
3281 static int
3282 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3283 {
3284 	struct mbuf *mh;
3285 	struct ip6_hdr *ip6;
3286 
3287 	ip6 = mtod(m, struct ip6_hdr *);
3288 	if (m->m_len > sizeof(*ip6)) {
3289 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3290 		if (mh == NULL) {
3291 			m_freem(m);
3292 			return ENOBUFS;
3293 		}
3294 		m_move_pkthdr(mh, m);
3295 		M_ALIGN(mh, sizeof(*ip6));
3296 		m->m_len -= sizeof(*ip6);
3297 		m->m_data += sizeof(*ip6);
3298 		mh->m_next = m;
3299 		m = mh;
3300 		m->m_len = sizeof(*ip6);
3301 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3302 	}
3303 	exthdrs->ip6e_ip6 = m;
3304 	return 0;
3305 }
3306 
3307 /*
3308  * Compute IPv6 extension header length.
3309  */
3310 int
3311 ip6_optlen(struct inpcb *inp)
3312 {
3313 	int len;
3314 
3315 	if (!inp->in6p_outputopts)
3316 		return 0;
3317 
3318 	len = 0;
3319 #define elen(x) \
3320     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3321 
3322 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3323 	if (inp->in6p_outputopts->ip6po_rthdr)
3324 		/* dest1 is valid with rthdr only */
3325 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3326 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3327 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3328 	return len;
3329 #undef elen
3330 }
3331