xref: /freebsd/sys/netinet6/ip6_output.c (revision f18976136625a7d016e97bfd9eabddf640b3e06d)
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, and
160  * mp is the destination.
161  */
162 #define MAKE_EXTHDR(hp, mp)						\
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 	}								\
171     } while (/*CONSTCOND*/ 0)
172 
173 /*
174  * Form a chain of extension headers.
175  * m is the extension header mbuf
176  * mp is the previous mbuf in the chain
177  * p is the next header
178  * i is the type of option.
179  */
180 #define MAKE_CHAIN(m, mp, p, i)\
181     do {\
182 	if (m) {\
183 		if (!hdrsplit) \
184 			panic("assumption failed: hdr not split"); \
185 		*mtod((m), u_char *) = *(p);\
186 		*(p) = (i);\
187 		p = mtod((m), u_char *);\
188 		(m)->m_next = (mp)->m_next;\
189 		(mp)->m_next = (m);\
190 		(mp) = (m);\
191 	}\
192     } while (/*CONSTCOND*/ 0)
193 
194 void
195 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
196 {
197 	u_short csum;
198 
199 	csum = in_cksum_skip(m, offset + plen, offset);
200 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
201 		csum = 0xffff;
202 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
203 
204 	if (offset + sizeof(csum) > m->m_len)
205 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
206 	else
207 		*(u_short *)mtodo(m, offset) = csum;
208 }
209 
210 int
211 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
212     int fraglen , uint32_t id)
213 {
214 	struct mbuf *m, **mnext, *m_frgpart;
215 	struct ip6_hdr *ip6, *mhip6;
216 	struct ip6_frag *ip6f;
217 	int off;
218 	int error;
219 	int tlen = m0->m_pkthdr.len;
220 
221 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
222 
223 	m = m0;
224 	ip6 = mtod(m, struct ip6_hdr *);
225 	mnext = &m->m_nextpkt;
226 
227 	for (off = hlen; off < tlen; off += fraglen) {
228 		m = m_gethdr(M_NOWAIT, MT_DATA);
229 		if (!m) {
230 			IP6STAT_INC(ip6s_odropped);
231 			return (ENOBUFS);
232 		}
233 
234 		/*
235 		 * Make sure the complete packet header gets copied
236 		 * from the originating mbuf to the newly created
237 		 * mbuf. This also ensures that existing firewall
238 		 * classification(s), VLAN tags and so on get copied
239 		 * to the resulting fragmented packet(s):
240 		 */
241 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
242 			m_free(m);
243 			IP6STAT_INC(ip6s_odropped);
244 			return (ENOBUFS);
245 		}
246 
247 		*mnext = m;
248 		mnext = &m->m_nextpkt;
249 		m->m_data += max_linkhdr;
250 		mhip6 = mtod(m, struct ip6_hdr *);
251 		*mhip6 = *ip6;
252 		m->m_len = sizeof(*mhip6);
253 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
254 		if (error) {
255 			IP6STAT_INC(ip6s_odropped);
256 			return (error);
257 		}
258 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
259 		if (off + fraglen >= tlen)
260 			fraglen = tlen - off;
261 		else
262 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
263 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
264 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
265 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
266 			IP6STAT_INC(ip6s_odropped);
267 			return (ENOBUFS);
268 		}
269 		m_cat(m, m_frgpart);
270 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
271 		ip6f->ip6f_reserved = 0;
272 		ip6f->ip6f_ident = id;
273 		ip6f->ip6f_nxt = nextproto;
274 		IP6STAT_INC(ip6s_ofragments);
275 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
276 	}
277 
278 	return (0);
279 }
280 
281 static int
282 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
283     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro)
284 {
285 #ifdef KERN_TLS
286 	struct ktls_session *tls = NULL;
287 #endif
288 	struct m_snd_tag *mst;
289 	int error;
290 
291 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
292 	mst = NULL;
293 
294 #ifdef KERN_TLS
295 	/*
296 	 * If this is an unencrypted TLS record, save a reference to
297 	 * the record.  This local reference is used to call
298 	 * ktls_output_eagain after the mbuf has been freed (thus
299 	 * dropping the mbuf's reference) in if_output.
300 	 */
301 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
302 		tls = ktls_hold(m->m_next->m_ext.ext_pgs->tls);
303 		mst = tls->snd_tag;
304 
305 		/*
306 		 * If a TLS session doesn't have a valid tag, it must
307 		 * have had an earlier ifp mismatch, so drop this
308 		 * packet.
309 		 */
310 		if (mst == NULL) {
311 			error = EAGAIN;
312 			goto done;
313 		}
314 	}
315 #endif
316 #ifdef RATELIMIT
317 	if (inp != NULL && mst == NULL) {
318 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
319 		    (inp->inp_snd_tag != NULL &&
320 		    inp->inp_snd_tag->ifp != ifp))
321 			in_pcboutput_txrtlmt(inp, ifp, m);
322 
323 		if (inp->inp_snd_tag != NULL)
324 			mst = inp->inp_snd_tag;
325 	}
326 #endif
327 	if (mst != NULL) {
328 		KASSERT(m->m_pkthdr.rcvif == NULL,
329 		    ("trying to add a send tag to a forwarded packet"));
330 		if (mst->ifp != ifp) {
331 			error = EAGAIN;
332 			goto done;
333 		}
334 
335 		/* stamp send tag on mbuf */
336 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
337 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
338 	}
339 
340 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
341 
342 done:
343 	/* Check for route change invalidating send tags. */
344 #ifdef KERN_TLS
345 	if (tls != NULL) {
346 		if (error == EAGAIN)
347 			error = ktls_output_eagain(inp, tls);
348 		ktls_free(tls);
349 	}
350 #endif
351 #ifdef RATELIMIT
352 	if (error == EAGAIN)
353 		in_pcboutput_eagain(inp);
354 #endif
355 	return (error);
356 }
357 
358 /*
359  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
360  * header (with pri, len, nxt, hlim, src, dst).
361  * This function may modify ver and hlim only.
362  * The mbuf chain containing the packet will be freed.
363  * The mbuf opt, if present, will not be freed.
364  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
365  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
366  * then result of route lookup is stored in ro->ro_rt.
367  *
368  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
369  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
370  * which is rt_mtu.
371  *
372  * ifpp - XXX: just for statistics
373  */
374 /*
375  * XXX TODO: no flowid is assigned for outbound flows?
376  */
377 int
378 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
379     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
380     struct ifnet **ifpp, struct inpcb *inp)
381 {
382 	struct ip6_hdr *ip6;
383 	struct ifnet *ifp, *origifp;
384 	struct mbuf *m = m0;
385 	struct mbuf *mprev = NULL;
386 	int hlen, tlen, len;
387 	struct epoch_tracker et;
388 	struct route_in6 ip6route;
389 	struct rtentry *rt = NULL;
390 	struct sockaddr_in6 *dst, src_sa, dst_sa;
391 	struct in6_addr odst;
392 	int error = 0;
393 	struct in6_ifaddr *ia = NULL;
394 	u_long mtu;
395 	int alwaysfrag, dontfrag;
396 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
397 	struct ip6_exthdrs exthdrs;
398 	struct in6_addr src0, dst0;
399 	u_int32_t zone;
400 	struct route_in6 *ro_pmtu = NULL;
401 	int hdrsplit = 0;
402 	int sw_csum, tso;
403 	int needfiblookup;
404 	uint32_t fibnum;
405 	struct m_tag *fwd_tag = NULL;
406 	uint32_t id;
407 
408 	if (inp != NULL) {
409 		INP_LOCK_ASSERT(inp);
410 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
411 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
412 			/* unconditionally set flowid */
413 			m->m_pkthdr.flowid = inp->inp_flowid;
414 			M_HASHTYPE_SET(m, inp->inp_flowtype);
415 		}
416 #ifdef NUMA
417 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
418 #endif
419 	}
420 
421 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
422 	/*
423 	 * IPSec checking which handles several cases.
424 	 * FAST IPSEC: We re-injected the packet.
425 	 * XXX: need scope argument.
426 	 */
427 	if (IPSEC_ENABLED(ipv6)) {
428 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
429 			if (error == EINPROGRESS)
430 				error = 0;
431 			goto done;
432 		}
433 	}
434 #endif /* IPSEC */
435 
436 	bzero(&exthdrs, sizeof(exthdrs));
437 	if (opt) {
438 		/* Hop-by-Hop options header */
439 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
440 		/* Destination options header(1st part) */
441 		if (opt->ip6po_rthdr) {
442 			/*
443 			 * Destination options header(1st part)
444 			 * This only makes sense with a routing header.
445 			 * See Section 9.2 of RFC 3542.
446 			 * Disabling this part just for MIP6 convenience is
447 			 * a bad idea.  We need to think carefully about a
448 			 * way to make the advanced API coexist with MIP6
449 			 * options, which might automatically be inserted in
450 			 * the kernel.
451 			 */
452 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
453 		}
454 		/* Routing header */
455 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
456 		/* Destination options header(2nd part) */
457 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
458 	}
459 
460 	/*
461 	 * Calculate the total length of the extension header chain.
462 	 * Keep the length of the unfragmentable part for fragmentation.
463 	 */
464 	optlen = 0;
465 	if (exthdrs.ip6e_hbh)
466 		optlen += exthdrs.ip6e_hbh->m_len;
467 	if (exthdrs.ip6e_dest1)
468 		optlen += exthdrs.ip6e_dest1->m_len;
469 	if (exthdrs.ip6e_rthdr)
470 		optlen += exthdrs.ip6e_rthdr->m_len;
471 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
472 
473 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
474 	if (exthdrs.ip6e_dest2)
475 		optlen += exthdrs.ip6e_dest2->m_len;
476 
477 	/*
478 	 * If there is at least one extension header,
479 	 * separate IP6 header from the payload.
480 	 */
481 	if (optlen && !hdrsplit) {
482 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
483 			m = NULL;
484 			goto freehdrs;
485 		}
486 		m = exthdrs.ip6e_ip6;
487 		hdrsplit++;
488 	}
489 
490 	ip6 = mtod(m, struct ip6_hdr *);
491 
492 	/* adjust mbuf packet header length */
493 	m->m_pkthdr.len += optlen;
494 	plen = m->m_pkthdr.len - sizeof(*ip6);
495 
496 	/* If this is a jumbo payload, insert a jumbo payload option. */
497 	if (plen > IPV6_MAXPACKET) {
498 		if (!hdrsplit) {
499 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
500 				m = NULL;
501 				goto freehdrs;
502 			}
503 			m = exthdrs.ip6e_ip6;
504 			hdrsplit++;
505 		}
506 		/* adjust pointer */
507 		ip6 = mtod(m, struct ip6_hdr *);
508 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
509 			goto freehdrs;
510 		ip6->ip6_plen = 0;
511 	} else
512 		ip6->ip6_plen = htons(plen);
513 
514 	/*
515 	 * Concatenate headers and fill in next header fields.
516 	 * Here we have, on "m"
517 	 *	IPv6 payload
518 	 * and we insert headers accordingly.  Finally, we should be getting:
519 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
520 	 *
521 	 * during the header composing process, "m" points to IPv6 header.
522 	 * "mprev" points to an extension header prior to esp.
523 	 */
524 	u_char *nexthdrp = &ip6->ip6_nxt;
525 	mprev = m;
526 
527 	/*
528 	 * we treat dest2 specially.  this makes IPsec processing
529 	 * much easier.  the goal here is to make mprev point the
530 	 * mbuf prior to dest2.
531 	 *
532 	 * result: IPv6 dest2 payload
533 	 * m and mprev will point to IPv6 header.
534 	 */
535 	if (exthdrs.ip6e_dest2) {
536 		if (!hdrsplit)
537 			panic("assumption failed: hdr not split");
538 		exthdrs.ip6e_dest2->m_next = m->m_next;
539 		m->m_next = exthdrs.ip6e_dest2;
540 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
541 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
542 	}
543 
544 	/*
545 	 * result: IPv6 hbh dest1 rthdr dest2 payload
546 	 * m will point to IPv6 header.  mprev will point to the
547 	 * extension header prior to dest2 (rthdr in the above case).
548 	 */
549 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
550 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
551 		   IPPROTO_DSTOPTS);
552 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
553 		   IPPROTO_ROUTING);
554 
555 	/*
556 	 * If there is a routing header, discard the packet.
557 	 */
558 	if (exthdrs.ip6e_rthdr) {
559 		 error = EINVAL;
560 		 goto bad;
561 	}
562 
563 	/* Source address validation */
564 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
565 	    (flags & IPV6_UNSPECSRC) == 0) {
566 		error = EOPNOTSUPP;
567 		IP6STAT_INC(ip6s_badscope);
568 		goto bad;
569 	}
570 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
571 		error = EOPNOTSUPP;
572 		IP6STAT_INC(ip6s_badscope);
573 		goto bad;
574 	}
575 
576 	IP6STAT_INC(ip6s_localout);
577 
578 	/*
579 	 * Route packet.
580 	 */
581 	if (ro == NULL) {
582 		ro = &ip6route;
583 		bzero((caddr_t)ro, sizeof(*ro));
584 	}
585 	ro_pmtu = ro;
586 	if (opt && opt->ip6po_rthdr)
587 		ro = &opt->ip6po_route;
588 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
589 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
590 	NET_EPOCH_ENTER(et);
591 again:
592 	/*
593 	 * if specified, try to fill in the traffic class field.
594 	 * do not override if a non-zero value is already set.
595 	 * we check the diffserv field and the ecn field separately.
596 	 */
597 	if (opt && opt->ip6po_tclass >= 0) {
598 		int mask = 0;
599 
600 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
601 			mask |= 0xfc;
602 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
603 			mask |= 0x03;
604 		if (mask != 0)
605 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
606 	}
607 
608 	/* fill in or override the hop limit field, if necessary. */
609 	if (opt && opt->ip6po_hlim != -1)
610 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
611 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
612 		if (im6o != NULL)
613 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
614 		else
615 			ip6->ip6_hlim = V_ip6_defmcasthlim;
616 	}
617 	/*
618 	 * Validate route against routing table additions;
619 	 * a better/more specific route might have been added.
620 	 * Make sure address family is set in route.
621 	 */
622 	if (inp) {
623 		ro->ro_dst.sin6_family = AF_INET6;
624 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
625 	}
626 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
627 	    ro->ro_dst.sin6_family == AF_INET6 &&
628 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
629 		rt = ro->ro_rt;
630 		ifp = ro->ro_rt->rt_ifp;
631 	} else {
632 		if (ro->ro_lle)
633 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
634 		ro->ro_lle = NULL;
635 		if (fwd_tag == NULL) {
636 			bzero(&dst_sa, sizeof(dst_sa));
637 			dst_sa.sin6_family = AF_INET6;
638 			dst_sa.sin6_len = sizeof(dst_sa);
639 			dst_sa.sin6_addr = ip6->ip6_dst;
640 		}
641 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
642 		    &rt, fibnum);
643 		if (error != 0) {
644 			if (ifp != NULL)
645 				in6_ifstat_inc(ifp, ifs6_out_discard);
646 			goto bad;
647 		}
648 	}
649 	if (rt == NULL) {
650 		/*
651 		 * If in6_selectroute() does not return a route entry,
652 		 * dst may not have been updated.
653 		 */
654 		*dst = dst_sa;	/* XXX */
655 	}
656 
657 	/*
658 	 * then rt (for unicast) and ifp must be non-NULL valid values.
659 	 */
660 	if ((flags & IPV6_FORWARDING) == 0) {
661 		/* XXX: the FORWARDING flag can be set for mrouting. */
662 		in6_ifstat_inc(ifp, ifs6_out_request);
663 	}
664 	if (rt != NULL) {
665 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
666 		counter_u64_add(rt->rt_pksent, 1);
667 	}
668 
669 	/* Setup data structures for scope ID checks. */
670 	src0 = ip6->ip6_src;
671 	bzero(&src_sa, sizeof(src_sa));
672 	src_sa.sin6_family = AF_INET6;
673 	src_sa.sin6_len = sizeof(src_sa);
674 	src_sa.sin6_addr = ip6->ip6_src;
675 
676 	dst0 = ip6->ip6_dst;
677 	/* re-initialize to be sure */
678 	bzero(&dst_sa, sizeof(dst_sa));
679 	dst_sa.sin6_family = AF_INET6;
680 	dst_sa.sin6_len = sizeof(dst_sa);
681 	dst_sa.sin6_addr = ip6->ip6_dst;
682 
683 	/* Check for valid scope ID. */
684 	if (in6_setscope(&src0, ifp, &zone) == 0 &&
685 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
686 	    in6_setscope(&dst0, ifp, &zone) == 0 &&
687 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
688 		/*
689 		 * The outgoing interface is in the zone of the source
690 		 * and destination addresses.
691 		 *
692 		 * Because the loopback interface cannot receive
693 		 * packets with a different scope ID than its own,
694 		 * there is a trick is to pretend the outgoing packet
695 		 * was received by the real network interface, by
696 		 * setting "origifp" different from "ifp". This is
697 		 * only allowed when "ifp" is a loopback network
698 		 * interface. Refer to code in nd6_output_ifp() for
699 		 * more details.
700 		 */
701 		origifp = ifp;
702 
703 		/*
704 		 * We should use ia_ifp to support the case of sending
705 		 * packets to an address of our own.
706 		 */
707 		if (ia != NULL && ia->ia_ifp)
708 			ifp = ia->ia_ifp;
709 
710 	} else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
711 	    sa6_recoverscope(&src_sa) != 0 ||
712 	    sa6_recoverscope(&dst_sa) != 0 ||
713 	    dst_sa.sin6_scope_id == 0 ||
714 	    (src_sa.sin6_scope_id != 0 &&
715 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
716 	    (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
717 		/*
718 		 * If the destination network interface is not a
719 		 * loopback interface, or the destination network
720 		 * address has no scope ID, or the source address has
721 		 * a scope ID set which is different from the
722 		 * destination address one, or there is no network
723 		 * interface representing this scope ID, the address
724 		 * pair is considered invalid.
725 		 */
726 		IP6STAT_INC(ip6s_badscope);
727 		in6_ifstat_inc(ifp, ifs6_out_discard);
728 		if (error == 0)
729 			error = EHOSTUNREACH; /* XXX */
730 		goto bad;
731 	}
732 
733 	/* All scope ID checks are successful. */
734 
735 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
736 		if (opt && opt->ip6po_nextroute.ro_rt) {
737 			/*
738 			 * The nexthop is explicitly specified by the
739 			 * application.  We assume the next hop is an IPv6
740 			 * address.
741 			 */
742 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
743 		}
744 		else if ((rt->rt_flags & RTF_GATEWAY))
745 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
746 	}
747 
748 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
749 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
750 	} else {
751 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
752 		in6_ifstat_inc(ifp, ifs6_out_mcast);
753 		/*
754 		 * Confirm that the outgoing interface supports multicast.
755 		 */
756 		if (!(ifp->if_flags & IFF_MULTICAST)) {
757 			IP6STAT_INC(ip6s_noroute);
758 			in6_ifstat_inc(ifp, ifs6_out_discard);
759 			error = ENETUNREACH;
760 			goto bad;
761 		}
762 		if ((im6o == NULL && in6_mcast_loop) ||
763 		    (im6o && im6o->im6o_multicast_loop)) {
764 			/*
765 			 * Loop back multicast datagram if not expressly
766 			 * forbidden to do so, even if we have not joined
767 			 * the address; protocols will filter it later,
768 			 * thus deferring a hash lookup and lock acquisition
769 			 * at the expense of an m_copym().
770 			 */
771 			ip6_mloopback(ifp, m);
772 		} else {
773 			/*
774 			 * If we are acting as a multicast router, perform
775 			 * multicast forwarding as if the packet had just
776 			 * arrived on the interface to which we are about
777 			 * to send.  The multicast forwarding function
778 			 * recursively calls this function, using the
779 			 * IPV6_FORWARDING flag to prevent infinite recursion.
780 			 *
781 			 * Multicasts that are looped back by ip6_mloopback(),
782 			 * above, will be forwarded by the ip6_input() routine,
783 			 * if necessary.
784 			 */
785 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
786 				/*
787 				 * XXX: ip6_mforward expects that rcvif is NULL
788 				 * when it is called from the originating path.
789 				 * However, it may not always be the case.
790 				 */
791 				m->m_pkthdr.rcvif = NULL;
792 				if (ip6_mforward(ip6, ifp, m) != 0) {
793 					m_freem(m);
794 					goto done;
795 				}
796 			}
797 		}
798 		/*
799 		 * Multicasts with a hoplimit of zero may be looped back,
800 		 * above, but must not be transmitted on a network.
801 		 * Also, multicasts addressed to the loopback interface
802 		 * are not sent -- the above call to ip6_mloopback() will
803 		 * loop back a copy if this host actually belongs to the
804 		 * destination group on the loopback interface.
805 		 */
806 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
807 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
808 			m_freem(m);
809 			goto done;
810 		}
811 	}
812 
813 	/*
814 	 * Fill the outgoing inteface to tell the upper layer
815 	 * to increment per-interface statistics.
816 	 */
817 	if (ifpp)
818 		*ifpp = ifp;
819 
820 	/* Determine path MTU. */
821 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
822 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
823 		goto bad;
824 
825 	/*
826 	 * The caller of this function may specify to use the minimum MTU
827 	 * in some cases.
828 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
829 	 * setting.  The logic is a bit complicated; by default, unicast
830 	 * packets will follow path MTU while multicast packets will be sent at
831 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
832 	 * including unicast ones will be sent at the minimum MTU.  Multicast
833 	 * packets will always be sent at the minimum MTU unless
834 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
835 	 * See RFC 3542 for more details.
836 	 */
837 	if (mtu > IPV6_MMTU) {
838 		if ((flags & IPV6_MINMTU))
839 			mtu = IPV6_MMTU;
840 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
841 			mtu = IPV6_MMTU;
842 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
843 			 (opt == NULL ||
844 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
845 			mtu = IPV6_MMTU;
846 		}
847 	}
848 
849 	/*
850 	 * clear embedded scope identifiers if necessary.
851 	 * in6_clearscope will touch the addresses only when necessary.
852 	 */
853 	in6_clearscope(&ip6->ip6_src);
854 	in6_clearscope(&ip6->ip6_dst);
855 
856 	/*
857 	 * If the outgoing packet contains a hop-by-hop options header,
858 	 * it must be examined and processed even by the source node.
859 	 * (RFC 2460, section 4.)
860 	 */
861 	if (exthdrs.ip6e_hbh) {
862 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
863 		u_int32_t dummy; /* XXX unused */
864 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
865 
866 #ifdef DIAGNOSTIC
867 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
868 			panic("ip6e_hbh is not contiguous");
869 #endif
870 		/*
871 		 *  XXX: if we have to send an ICMPv6 error to the sender,
872 		 *       we need the M_LOOP flag since icmp6_error() expects
873 		 *       the IPv6 and the hop-by-hop options header are
874 		 *       contiguous unless the flag is set.
875 		 */
876 		m->m_flags |= M_LOOP;
877 		m->m_pkthdr.rcvif = ifp;
878 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
879 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
880 		    &dummy, &plen) < 0) {
881 			/* m was already freed at this point */
882 			error = EINVAL;/* better error? */
883 			goto done;
884 		}
885 		m->m_flags &= ~M_LOOP; /* XXX */
886 		m->m_pkthdr.rcvif = NULL;
887 	}
888 
889 	/* Jump over all PFIL processing if hooks are not active. */
890 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
891 		goto passout;
892 
893 	odst = ip6->ip6_dst;
894 	/* Run through list of hooks for output packets. */
895 	switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
896 	case PFIL_PASS:
897 		ip6 = mtod(m, struct ip6_hdr *);
898 		break;
899 	case PFIL_DROPPED:
900 		error = EPERM;
901 		/* FALLTHROUGH */
902 	case PFIL_CONSUMED:
903 		goto done;
904 	}
905 
906 	needfiblookup = 0;
907 	/* See if destination IP address was changed by packet filter. */
908 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
909 		m->m_flags |= M_SKIP_FIREWALL;
910 		/* If destination is now ourself drop to ip6_input(). */
911 		if (in6_localip(&ip6->ip6_dst)) {
912 			m->m_flags |= M_FASTFWD_OURS;
913 			if (m->m_pkthdr.rcvif == NULL)
914 				m->m_pkthdr.rcvif = V_loif;
915 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
916 				m->m_pkthdr.csum_flags |=
917 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
918 				m->m_pkthdr.csum_data = 0xffff;
919 			}
920 #ifdef SCTP
921 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
922 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
923 #endif
924 			error = netisr_queue(NETISR_IPV6, m);
925 			goto done;
926 		} else {
927 			RO_INVALIDATE_CACHE(ro);
928 			needfiblookup = 1; /* Redo the routing table lookup. */
929 		}
930 	}
931 	/* See if fib was changed by packet filter. */
932 	if (fibnum != M_GETFIB(m)) {
933 		m->m_flags |= M_SKIP_FIREWALL;
934 		fibnum = M_GETFIB(m);
935 		RO_INVALIDATE_CACHE(ro);
936 		needfiblookup = 1;
937 	}
938 	if (needfiblookup)
939 		goto again;
940 
941 	/* See if local, if yes, send it to netisr. */
942 	if (m->m_flags & M_FASTFWD_OURS) {
943 		if (m->m_pkthdr.rcvif == NULL)
944 			m->m_pkthdr.rcvif = V_loif;
945 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
946 			m->m_pkthdr.csum_flags |=
947 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
948 			m->m_pkthdr.csum_data = 0xffff;
949 		}
950 #ifdef SCTP
951 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
952 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
953 #endif
954 		error = netisr_queue(NETISR_IPV6, m);
955 		goto done;
956 	}
957 	/* Or forward to some other address? */
958 	if ((m->m_flags & M_IP6_NEXTHOP) &&
959 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
960 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
961 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
962 		m->m_flags |= M_SKIP_FIREWALL;
963 		m->m_flags &= ~M_IP6_NEXTHOP;
964 		m_tag_delete(m, fwd_tag);
965 		goto again;
966 	}
967 
968 passout:
969 	/*
970 	 * Send the packet to the outgoing interface.
971 	 * If necessary, do IPv6 fragmentation before sending.
972 	 *
973 	 * the logic here is rather complex:
974 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
975 	 * 1-a:	send as is if tlen <= path mtu
976 	 * 1-b:	fragment if tlen > path mtu
977 	 *
978 	 * 2: if user asks us not to fragment (dontfrag == 1)
979 	 * 2-a:	send as is if tlen <= interface mtu
980 	 * 2-b:	error if tlen > interface mtu
981 	 *
982 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
983 	 *	always fragment
984 	 *
985 	 * 4: if dontfrag == 1 && alwaysfrag == 1
986 	 *	error, as we cannot handle this conflicting request
987 	 */
988 	sw_csum = m->m_pkthdr.csum_flags;
989 	if (!hdrsplit) {
990 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
991 		sw_csum &= ~ifp->if_hwassist;
992 	} else
993 		tso = 0;
994 	/*
995 	 * If we added extension headers, we will not do TSO and calculate the
996 	 * checksums ourselves for now.
997 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
998 	 * with ext. hdrs.
999 	 */
1000 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
1001 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
1002 		m = mb_unmapped_to_ext(m);
1003 		if (m == NULL) {
1004 			error = ENOBUFS;
1005 			IP6STAT_INC(ip6s_odropped);
1006 			goto bad;
1007 		}
1008 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
1009 	} else if ((ifp->if_capenable & IFCAP_NOMAP) == 0) {
1010 		m = mb_unmapped_to_ext(m);
1011 		if (m == NULL) {
1012 			error = ENOBUFS;
1013 			IP6STAT_INC(ip6s_odropped);
1014 			goto bad;
1015 		}
1016 	}
1017 #ifdef SCTP
1018 	if (sw_csum & CSUM_SCTP_IPV6) {
1019 		sw_csum &= ~CSUM_SCTP_IPV6;
1020 		m = mb_unmapped_to_ext(m);
1021 		if (m == NULL) {
1022 			error = ENOBUFS;
1023 			IP6STAT_INC(ip6s_odropped);
1024 			goto bad;
1025 		}
1026 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
1027 	}
1028 #endif
1029 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
1030 	tlen = m->m_pkthdr.len;
1031 
1032 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1033 		dontfrag = 1;
1034 	else
1035 		dontfrag = 0;
1036 	if (dontfrag && alwaysfrag) {	/* case 4 */
1037 		/* conflicting request - can't transmit */
1038 		error = EMSGSIZE;
1039 		goto bad;
1040 	}
1041 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
1042 		/*
1043 		 * Even if the DONTFRAG option is specified, we cannot send the
1044 		 * packet when the data length is larger than the MTU of the
1045 		 * outgoing interface.
1046 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1047 		 * application wanted to know the MTU value. Also return an
1048 		 * error code (this is not described in the API spec).
1049 		 */
1050 		if (inp != NULL)
1051 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1052 		error = EMSGSIZE;
1053 		goto bad;
1054 	}
1055 
1056 	/*
1057 	 * transmit packet without fragmentation
1058 	 */
1059 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
1060 		struct in6_ifaddr *ia6;
1061 
1062 		ip6 = mtod(m, struct ip6_hdr *);
1063 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1064 		if (ia6) {
1065 			/* Record statistics for this interface address. */
1066 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1067 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1068 			    m->m_pkthdr.len);
1069 			ifa_free(&ia6->ia_ifa);
1070 		}
1071 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1072 		goto done;
1073 	}
1074 
1075 	/*
1076 	 * try to fragment the packet.  case 1-b and 3
1077 	 */
1078 	if (mtu < IPV6_MMTU) {
1079 		/* path MTU cannot be less than IPV6_MMTU */
1080 		error = EMSGSIZE;
1081 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1082 		goto bad;
1083 	} else if (ip6->ip6_plen == 0) {
1084 		/* jumbo payload cannot be fragmented */
1085 		error = EMSGSIZE;
1086 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1087 		goto bad;
1088 	} else {
1089 		u_char nextproto;
1090 
1091 		/*
1092 		 * Too large for the destination or interface;
1093 		 * fragment if possible.
1094 		 * Must be able to put at least 8 bytes per fragment.
1095 		 */
1096 		hlen = unfragpartlen;
1097 		if (mtu > IPV6_MAXPACKET)
1098 			mtu = IPV6_MAXPACKET;
1099 
1100 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1101 		if (len < 8) {
1102 			error = EMSGSIZE;
1103 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1104 			goto bad;
1105 		}
1106 
1107 		/*
1108 		 * If the interface will not calculate checksums on
1109 		 * fragmented packets, then do it here.
1110 		 * XXX-BZ handle the hw offloading case.  Need flags.
1111 		 */
1112 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1113 			m = mb_unmapped_to_ext(m);
1114 			if (m == NULL) {
1115 				in6_ifstat_inc(ifp, ifs6_out_fragfail);
1116 				error = ENOBUFS;
1117 				goto bad;
1118 			}
1119 			in6_delayed_cksum(m, plen, hlen);
1120 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1121 		}
1122 #ifdef SCTP
1123 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1124 			m = mb_unmapped_to_ext(m);
1125 			if (m == NULL) {
1126 				in6_ifstat_inc(ifp, ifs6_out_fragfail);
1127 				error = ENOBUFS;
1128 				goto bad;
1129 			}
1130 			sctp_delayed_cksum(m, hlen);
1131 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1132 		}
1133 #endif
1134 		/*
1135 		 * Change the next header field of the last header in the
1136 		 * unfragmentable part.
1137 		 */
1138 		if (exthdrs.ip6e_rthdr) {
1139 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1140 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1141 		} else if (exthdrs.ip6e_dest1) {
1142 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1143 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1144 		} else if (exthdrs.ip6e_hbh) {
1145 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1146 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1147 		} else {
1148 			nextproto = ip6->ip6_nxt;
1149 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1150 		}
1151 
1152 		/*
1153 		 * Loop through length of segment after first fragment,
1154 		 * make new header and copy data of each part and link onto
1155 		 * chain.
1156 		 */
1157 		m0 = m;
1158 		id = htonl(ip6_randomid());
1159 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1160 			goto sendorfree;
1161 
1162 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1163 	}
1164 
1165 	/*
1166 	 * Remove leading garbages.
1167 	 */
1168 sendorfree:
1169 	m = m0->m_nextpkt;
1170 	m0->m_nextpkt = 0;
1171 	m_freem(m0);
1172 	for (; m; m = m0) {
1173 		m0 = m->m_nextpkt;
1174 		m->m_nextpkt = 0;
1175 		if (error == 0) {
1176 			/* Record statistics for this interface address. */
1177 			if (ia) {
1178 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1179 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1180 				    m->m_pkthdr.len);
1181 			}
1182 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro);
1183 		} else
1184 			m_freem(m);
1185 	}
1186 
1187 	if (error == 0)
1188 		IP6STAT_INC(ip6s_fragmented);
1189 
1190 done:
1191 	NET_EPOCH_EXIT(et);
1192 	if (ro == &ip6route)
1193 		RO_RTFREE(ro);
1194 	return (error);
1195 
1196 freehdrs:
1197 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1198 	m_freem(exthdrs.ip6e_dest1);
1199 	m_freem(exthdrs.ip6e_rthdr);
1200 	m_freem(exthdrs.ip6e_dest2);
1201 	/* FALLTHROUGH */
1202 bad:
1203 	if (m)
1204 		m_freem(m);
1205 	goto done;
1206 }
1207 
1208 static int
1209 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1210 {
1211 	struct mbuf *m;
1212 
1213 	if (hlen > MCLBYTES)
1214 		return (ENOBUFS); /* XXX */
1215 
1216 	if (hlen > MLEN)
1217 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1218 	else
1219 		m = m_get(M_NOWAIT, MT_DATA);
1220 	if (m == NULL)
1221 		return (ENOBUFS);
1222 	m->m_len = hlen;
1223 	if (hdr)
1224 		bcopy(hdr, mtod(m, caddr_t), hlen);
1225 
1226 	*mp = m;
1227 	return (0);
1228 }
1229 
1230 /*
1231  * Insert jumbo payload option.
1232  */
1233 static int
1234 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1235 {
1236 	struct mbuf *mopt;
1237 	u_char *optbuf;
1238 	u_int32_t v;
1239 
1240 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1241 
1242 	/*
1243 	 * If there is no hop-by-hop options header, allocate new one.
1244 	 * If there is one but it doesn't have enough space to store the
1245 	 * jumbo payload option, allocate a cluster to store the whole options.
1246 	 * Otherwise, use it to store the options.
1247 	 */
1248 	if (exthdrs->ip6e_hbh == NULL) {
1249 		mopt = m_get(M_NOWAIT, MT_DATA);
1250 		if (mopt == NULL)
1251 			return (ENOBUFS);
1252 		mopt->m_len = JUMBOOPTLEN;
1253 		optbuf = mtod(mopt, u_char *);
1254 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1255 		exthdrs->ip6e_hbh = mopt;
1256 	} else {
1257 		struct ip6_hbh *hbh;
1258 
1259 		mopt = exthdrs->ip6e_hbh;
1260 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1261 			/*
1262 			 * XXX assumption:
1263 			 * - exthdrs->ip6e_hbh is not referenced from places
1264 			 *   other than exthdrs.
1265 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1266 			 */
1267 			int oldoptlen = mopt->m_len;
1268 			struct mbuf *n;
1269 
1270 			/*
1271 			 * XXX: give up if the whole (new) hbh header does
1272 			 * not fit even in an mbuf cluster.
1273 			 */
1274 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1275 				return (ENOBUFS);
1276 
1277 			/*
1278 			 * As a consequence, we must always prepare a cluster
1279 			 * at this point.
1280 			 */
1281 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1282 			if (n == NULL)
1283 				return (ENOBUFS);
1284 			n->m_len = oldoptlen + JUMBOOPTLEN;
1285 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1286 			    oldoptlen);
1287 			optbuf = mtod(n, caddr_t) + oldoptlen;
1288 			m_freem(mopt);
1289 			mopt = exthdrs->ip6e_hbh = n;
1290 		} else {
1291 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1292 			mopt->m_len += JUMBOOPTLEN;
1293 		}
1294 		optbuf[0] = IP6OPT_PADN;
1295 		optbuf[1] = 1;
1296 
1297 		/*
1298 		 * Adjust the header length according to the pad and
1299 		 * the jumbo payload option.
1300 		 */
1301 		hbh = mtod(mopt, struct ip6_hbh *);
1302 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1303 	}
1304 
1305 	/* fill in the option. */
1306 	optbuf[2] = IP6OPT_JUMBO;
1307 	optbuf[3] = 4;
1308 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1309 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1310 
1311 	/* finally, adjust the packet header length */
1312 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1313 
1314 	return (0);
1315 #undef JUMBOOPTLEN
1316 }
1317 
1318 /*
1319  * Insert fragment header and copy unfragmentable header portions.
1320  */
1321 static int
1322 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1323     struct ip6_frag **frghdrp)
1324 {
1325 	struct mbuf *n, *mlast;
1326 
1327 	if (hlen > sizeof(struct ip6_hdr)) {
1328 		n = m_copym(m0, sizeof(struct ip6_hdr),
1329 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1330 		if (n == NULL)
1331 			return (ENOBUFS);
1332 		m->m_next = n;
1333 	} else
1334 		n = m;
1335 
1336 	/* Search for the last mbuf of unfragmentable part. */
1337 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1338 		;
1339 
1340 	if (M_WRITABLE(mlast) &&
1341 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1342 		/* use the trailing space of the last mbuf for the fragment hdr */
1343 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1344 		    mlast->m_len);
1345 		mlast->m_len += sizeof(struct ip6_frag);
1346 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1347 	} else {
1348 		/* allocate a new mbuf for the fragment header */
1349 		struct mbuf *mfrg;
1350 
1351 		mfrg = m_get(M_NOWAIT, MT_DATA);
1352 		if (mfrg == NULL)
1353 			return (ENOBUFS);
1354 		mfrg->m_len = sizeof(struct ip6_frag);
1355 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1356 		mlast->m_next = mfrg;
1357 	}
1358 
1359 	return (0);
1360 }
1361 
1362 /*
1363  * Calculates IPv6 path mtu for destination @dst.
1364  * Resulting MTU is stored in @mtup.
1365  *
1366  * Returns 0 on success.
1367  */
1368 static int
1369 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1370 {
1371 	struct nhop6_extended nh6;
1372 	struct in6_addr kdst;
1373 	uint32_t scopeid;
1374 	struct ifnet *ifp;
1375 	u_long mtu;
1376 	int error;
1377 
1378 	in6_splitscope(dst, &kdst, &scopeid);
1379 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1380 		return (EHOSTUNREACH);
1381 
1382 	ifp = nh6.nh_ifp;
1383 	mtu = nh6.nh_mtu;
1384 
1385 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1386 	fib6_free_nh_ext(fibnum, &nh6);
1387 
1388 	return (error);
1389 }
1390 
1391 /*
1392  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1393  * and cached data in @ro_pmtu.
1394  * MTU from (successful) route lookup is saved (along with dst)
1395  * inside @ro_pmtu to avoid subsequent route lookups after packet
1396  * filter processing.
1397  *
1398  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1399  * Returns 0 on success.
1400  */
1401 static int
1402 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1403     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1404     int *alwaysfragp, u_int fibnum, u_int proto)
1405 {
1406 	struct nhop6_basic nh6;
1407 	struct in6_addr kdst;
1408 	uint32_t scopeid;
1409 	struct sockaddr_in6 *sa6_dst;
1410 	u_long mtu;
1411 
1412 	mtu = 0;
1413 	if (do_lookup) {
1414 
1415 		/*
1416 		 * Here ro_pmtu has final destination address, while
1417 		 * ro might represent immediate destination.
1418 		 * Use ro_pmtu destination since mtu might differ.
1419 		 */
1420 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1421 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1422 			ro_pmtu->ro_mtu = 0;
1423 
1424 		if (ro_pmtu->ro_mtu == 0) {
1425 			bzero(sa6_dst, sizeof(*sa6_dst));
1426 			sa6_dst->sin6_family = AF_INET6;
1427 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1428 			sa6_dst->sin6_addr = *dst;
1429 
1430 			in6_splitscope(dst, &kdst, &scopeid);
1431 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1432 			    &nh6) == 0)
1433 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1434 		}
1435 
1436 		mtu = ro_pmtu->ro_mtu;
1437 	}
1438 
1439 	if (ro_pmtu->ro_rt)
1440 		mtu = ro_pmtu->ro_rt->rt_mtu;
1441 
1442 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1443 }
1444 
1445 /*
1446  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1447  * hostcache data for @dst.
1448  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1449  *
1450  * Returns 0 on success.
1451  */
1452 static int
1453 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1454     u_long *mtup, int *alwaysfragp, u_int proto)
1455 {
1456 	u_long mtu = 0;
1457 	int alwaysfrag = 0;
1458 	int error = 0;
1459 
1460 	if (rt_mtu > 0) {
1461 		u_int32_t ifmtu;
1462 		struct in_conninfo inc;
1463 
1464 		bzero(&inc, sizeof(inc));
1465 		inc.inc_flags |= INC_ISIPV6;
1466 		inc.inc6_faddr = *dst;
1467 
1468 		ifmtu = IN6_LINKMTU(ifp);
1469 
1470 		/* TCP is known to react to pmtu changes so skip hc */
1471 		if (proto != IPPROTO_TCP)
1472 			mtu = tcp_hc_getmtu(&inc);
1473 
1474 		if (mtu)
1475 			mtu = min(mtu, rt_mtu);
1476 		else
1477 			mtu = rt_mtu;
1478 		if (mtu == 0)
1479 			mtu = ifmtu;
1480 		else if (mtu < IPV6_MMTU) {
1481 			/*
1482 			 * RFC2460 section 5, last paragraph:
1483 			 * if we record ICMPv6 too big message with
1484 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1485 			 * or smaller, with framgent header attached.
1486 			 * (fragment header is needed regardless from the
1487 			 * packet size, for translators to identify packets)
1488 			 */
1489 			alwaysfrag = 1;
1490 			mtu = IPV6_MMTU;
1491 		}
1492 	} else if (ifp) {
1493 		mtu = IN6_LINKMTU(ifp);
1494 	} else
1495 		error = EHOSTUNREACH; /* XXX */
1496 
1497 	*mtup = mtu;
1498 	if (alwaysfragp)
1499 		*alwaysfragp = alwaysfrag;
1500 	return (error);
1501 }
1502 
1503 /*
1504  * IP6 socket option processing.
1505  */
1506 int
1507 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1508 {
1509 	int optdatalen, uproto;
1510 	void *optdata;
1511 	struct inpcb *inp = sotoinpcb(so);
1512 	int error, optval;
1513 	int level, op, optname;
1514 	int optlen;
1515 	struct thread *td;
1516 #ifdef	RSS
1517 	uint32_t rss_bucket;
1518 	int retval;
1519 #endif
1520 
1521 /*
1522  * Don't use more than a quarter of mbuf clusters.  N.B.:
1523  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1524  * on LP64 architectures, so cast to u_long to avoid undefined
1525  * behavior.  ILP32 architectures cannot have nmbclusters
1526  * large enough to overflow for other reasons.
1527  */
1528 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1529 
1530 	level = sopt->sopt_level;
1531 	op = sopt->sopt_dir;
1532 	optname = sopt->sopt_name;
1533 	optlen = sopt->sopt_valsize;
1534 	td = sopt->sopt_td;
1535 	error = 0;
1536 	optval = 0;
1537 	uproto = (int)so->so_proto->pr_protocol;
1538 
1539 	if (level != IPPROTO_IPV6) {
1540 		error = EINVAL;
1541 
1542 		if (sopt->sopt_level == SOL_SOCKET &&
1543 		    sopt->sopt_dir == SOPT_SET) {
1544 			switch (sopt->sopt_name) {
1545 			case SO_REUSEADDR:
1546 				INP_WLOCK(inp);
1547 				if ((so->so_options & SO_REUSEADDR) != 0)
1548 					inp->inp_flags2 |= INP_REUSEADDR;
1549 				else
1550 					inp->inp_flags2 &= ~INP_REUSEADDR;
1551 				INP_WUNLOCK(inp);
1552 				error = 0;
1553 				break;
1554 			case SO_REUSEPORT:
1555 				INP_WLOCK(inp);
1556 				if ((so->so_options & SO_REUSEPORT) != 0)
1557 					inp->inp_flags2 |= INP_REUSEPORT;
1558 				else
1559 					inp->inp_flags2 &= ~INP_REUSEPORT;
1560 				INP_WUNLOCK(inp);
1561 				error = 0;
1562 				break;
1563 			case SO_REUSEPORT_LB:
1564 				INP_WLOCK(inp);
1565 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1566 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1567 				else
1568 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1569 				INP_WUNLOCK(inp);
1570 				error = 0;
1571 				break;
1572 			case SO_SETFIB:
1573 				INP_WLOCK(inp);
1574 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1575 				INP_WUNLOCK(inp);
1576 				error = 0;
1577 				break;
1578 			case SO_MAX_PACING_RATE:
1579 #ifdef RATELIMIT
1580 				INP_WLOCK(inp);
1581 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1582 				INP_WUNLOCK(inp);
1583 				error = 0;
1584 #else
1585 				error = EOPNOTSUPP;
1586 #endif
1587 				break;
1588 			default:
1589 				break;
1590 			}
1591 		}
1592 	} else {		/* level == IPPROTO_IPV6 */
1593 		switch (op) {
1594 
1595 		case SOPT_SET:
1596 			switch (optname) {
1597 			case IPV6_2292PKTOPTIONS:
1598 #ifdef IPV6_PKTOPTIONS
1599 			case IPV6_PKTOPTIONS:
1600 #endif
1601 			{
1602 				struct mbuf *m;
1603 
1604 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1605 					printf("ip6_ctloutput: mbuf limit hit\n");
1606 					error = ENOBUFS;
1607 					break;
1608 				}
1609 
1610 				error = soopt_getm(sopt, &m); /* XXX */
1611 				if (error != 0)
1612 					break;
1613 				error = soopt_mcopyin(sopt, m); /* XXX */
1614 				if (error != 0)
1615 					break;
1616 				error = ip6_pcbopts(&inp->in6p_outputopts,
1617 						    m, so, sopt);
1618 				m_freem(m); /* XXX */
1619 				break;
1620 			}
1621 
1622 			/*
1623 			 * Use of some Hop-by-Hop options or some
1624 			 * Destination options, might require special
1625 			 * privilege.  That is, normal applications
1626 			 * (without special privilege) might be forbidden
1627 			 * from setting certain options in outgoing packets,
1628 			 * and might never see certain options in received
1629 			 * packets. [RFC 2292 Section 6]
1630 			 * KAME specific note:
1631 			 *  KAME prevents non-privileged users from sending or
1632 			 *  receiving ANY hbh/dst options in order to avoid
1633 			 *  overhead of parsing options in the kernel.
1634 			 */
1635 			case IPV6_RECVHOPOPTS:
1636 			case IPV6_RECVDSTOPTS:
1637 			case IPV6_RECVRTHDRDSTOPTS:
1638 				if (td != NULL) {
1639 					error = priv_check(td,
1640 					    PRIV_NETINET_SETHDROPTS);
1641 					if (error)
1642 						break;
1643 				}
1644 				/* FALLTHROUGH */
1645 			case IPV6_UNICAST_HOPS:
1646 			case IPV6_HOPLIMIT:
1647 
1648 			case IPV6_RECVPKTINFO:
1649 			case IPV6_RECVHOPLIMIT:
1650 			case IPV6_RECVRTHDR:
1651 			case IPV6_RECVPATHMTU:
1652 			case IPV6_RECVTCLASS:
1653 			case IPV6_RECVFLOWID:
1654 #ifdef	RSS
1655 			case IPV6_RECVRSSBUCKETID:
1656 #endif
1657 			case IPV6_V6ONLY:
1658 			case IPV6_AUTOFLOWLABEL:
1659 			case IPV6_ORIGDSTADDR:
1660 			case IPV6_BINDANY:
1661 			case IPV6_BINDMULTI:
1662 #ifdef	RSS
1663 			case IPV6_RSS_LISTEN_BUCKET:
1664 #endif
1665 				if (optname == IPV6_BINDANY && td != NULL) {
1666 					error = priv_check(td,
1667 					    PRIV_NETINET_BINDANY);
1668 					if (error)
1669 						break;
1670 				}
1671 
1672 				if (optlen != sizeof(int)) {
1673 					error = EINVAL;
1674 					break;
1675 				}
1676 				error = sooptcopyin(sopt, &optval,
1677 					sizeof optval, sizeof optval);
1678 				if (error)
1679 					break;
1680 				switch (optname) {
1681 
1682 				case IPV6_UNICAST_HOPS:
1683 					if (optval < -1 || optval >= 256)
1684 						error = EINVAL;
1685 					else {
1686 						/* -1 = kernel default */
1687 						inp->in6p_hops = optval;
1688 						if ((inp->inp_vflag &
1689 						     INP_IPV4) != 0)
1690 							inp->inp_ip_ttl = optval;
1691 					}
1692 					break;
1693 #define OPTSET(bit) \
1694 do { \
1695 	INP_WLOCK(inp); \
1696 	if (optval) \
1697 		inp->inp_flags |= (bit); \
1698 	else \
1699 		inp->inp_flags &= ~(bit); \
1700 	INP_WUNLOCK(inp); \
1701 } while (/*CONSTCOND*/ 0)
1702 #define OPTSET2292(bit) \
1703 do { \
1704 	INP_WLOCK(inp); \
1705 	inp->inp_flags |= IN6P_RFC2292; \
1706 	if (optval) \
1707 		inp->inp_flags |= (bit); \
1708 	else \
1709 		inp->inp_flags &= ~(bit); \
1710 	INP_WUNLOCK(inp); \
1711 } while (/*CONSTCOND*/ 0)
1712 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1713 
1714 #define OPTSET2_N(bit, val) do {					\
1715 	if (val)							\
1716 		inp->inp_flags2 |= bit;					\
1717 	else								\
1718 		inp->inp_flags2 &= ~bit;				\
1719 } while (0)
1720 #define OPTSET2(bit, val) do {						\
1721 	INP_WLOCK(inp);							\
1722 	OPTSET2_N(bit, val);						\
1723 	INP_WUNLOCK(inp);						\
1724 } while (0)
1725 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1726 #define OPTSET2292_EXCLUSIVE(bit)					\
1727 do {									\
1728 	INP_WLOCK(inp);							\
1729 	if (OPTBIT(IN6P_RFC2292)) {					\
1730 		error = EINVAL;						\
1731 	} else {							\
1732 		if (optval)						\
1733 			inp->inp_flags |= (bit);			\
1734 		else							\
1735 			inp->inp_flags &= ~(bit);			\
1736 	}								\
1737 	INP_WUNLOCK(inp);						\
1738 } while (/*CONSTCOND*/ 0)
1739 
1740 				case IPV6_RECVPKTINFO:
1741 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1742 					break;
1743 
1744 				case IPV6_HOPLIMIT:
1745 				{
1746 					struct ip6_pktopts **optp;
1747 
1748 					/* cannot mix with RFC2292 */
1749 					if (OPTBIT(IN6P_RFC2292)) {
1750 						error = EINVAL;
1751 						break;
1752 					}
1753 					INP_WLOCK(inp);
1754 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1755 						INP_WUNLOCK(inp);
1756 						return (ECONNRESET);
1757 					}
1758 					optp = &inp->in6p_outputopts;
1759 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1760 					    (u_char *)&optval, sizeof(optval),
1761 					    optp, (td != NULL) ? td->td_ucred :
1762 					    NULL, uproto);
1763 					INP_WUNLOCK(inp);
1764 					break;
1765 				}
1766 
1767 				case IPV6_RECVHOPLIMIT:
1768 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1769 					break;
1770 
1771 				case IPV6_RECVHOPOPTS:
1772 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1773 					break;
1774 
1775 				case IPV6_RECVDSTOPTS:
1776 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1777 					break;
1778 
1779 				case IPV6_RECVRTHDRDSTOPTS:
1780 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1781 					break;
1782 
1783 				case IPV6_RECVRTHDR:
1784 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1785 					break;
1786 
1787 				case IPV6_RECVPATHMTU:
1788 					/*
1789 					 * We ignore this option for TCP
1790 					 * sockets.
1791 					 * (RFC3542 leaves this case
1792 					 * unspecified.)
1793 					 */
1794 					if (uproto != IPPROTO_TCP)
1795 						OPTSET(IN6P_MTU);
1796 					break;
1797 
1798 				case IPV6_RECVFLOWID:
1799 					OPTSET2(INP_RECVFLOWID, optval);
1800 					break;
1801 
1802 #ifdef	RSS
1803 				case IPV6_RECVRSSBUCKETID:
1804 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1805 					break;
1806 #endif
1807 
1808 				case IPV6_V6ONLY:
1809 					INP_WLOCK(inp);
1810 					if (inp->inp_lport ||
1811 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1812 						/*
1813 						 * The socket is already bound.
1814 						 */
1815 						INP_WUNLOCK(inp);
1816 						error = EINVAL;
1817 						break;
1818 					}
1819 					if (optval) {
1820 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1821 						inp->inp_vflag &= ~INP_IPV4;
1822 					} else {
1823 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1824 						inp->inp_vflag |= INP_IPV4;
1825 					}
1826 					INP_WUNLOCK(inp);
1827 					break;
1828 				case IPV6_RECVTCLASS:
1829 					/* cannot mix with RFC2292 XXX */
1830 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1831 					break;
1832 				case IPV6_AUTOFLOWLABEL:
1833 					OPTSET(IN6P_AUTOFLOWLABEL);
1834 					break;
1835 
1836 				case IPV6_ORIGDSTADDR:
1837 					OPTSET2(INP_ORIGDSTADDR, optval);
1838 					break;
1839 				case IPV6_BINDANY:
1840 					OPTSET(INP_BINDANY);
1841 					break;
1842 
1843 				case IPV6_BINDMULTI:
1844 					OPTSET2(INP_BINDMULTI, optval);
1845 					break;
1846 #ifdef	RSS
1847 				case IPV6_RSS_LISTEN_BUCKET:
1848 					if ((optval >= 0) &&
1849 					    (optval < rss_getnumbuckets())) {
1850 						INP_WLOCK(inp);
1851 						inp->inp_rss_listen_bucket = optval;
1852 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1853 						INP_WUNLOCK(inp);
1854 					} else {
1855 						error = EINVAL;
1856 					}
1857 					break;
1858 #endif
1859 				}
1860 				break;
1861 
1862 			case IPV6_TCLASS:
1863 			case IPV6_DONTFRAG:
1864 			case IPV6_USE_MIN_MTU:
1865 			case IPV6_PREFER_TEMPADDR:
1866 				if (optlen != sizeof(optval)) {
1867 					error = EINVAL;
1868 					break;
1869 				}
1870 				error = sooptcopyin(sopt, &optval,
1871 					sizeof optval, sizeof optval);
1872 				if (error)
1873 					break;
1874 				{
1875 					struct ip6_pktopts **optp;
1876 					INP_WLOCK(inp);
1877 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1878 						INP_WUNLOCK(inp);
1879 						return (ECONNRESET);
1880 					}
1881 					optp = &inp->in6p_outputopts;
1882 					error = ip6_pcbopt(optname,
1883 					    (u_char *)&optval, sizeof(optval),
1884 					    optp, (td != NULL) ? td->td_ucred :
1885 					    NULL, uproto);
1886 					INP_WUNLOCK(inp);
1887 					break;
1888 				}
1889 
1890 			case IPV6_2292PKTINFO:
1891 			case IPV6_2292HOPLIMIT:
1892 			case IPV6_2292HOPOPTS:
1893 			case IPV6_2292DSTOPTS:
1894 			case IPV6_2292RTHDR:
1895 				/* RFC 2292 */
1896 				if (optlen != sizeof(int)) {
1897 					error = EINVAL;
1898 					break;
1899 				}
1900 				error = sooptcopyin(sopt, &optval,
1901 					sizeof optval, sizeof optval);
1902 				if (error)
1903 					break;
1904 				switch (optname) {
1905 				case IPV6_2292PKTINFO:
1906 					OPTSET2292(IN6P_PKTINFO);
1907 					break;
1908 				case IPV6_2292HOPLIMIT:
1909 					OPTSET2292(IN6P_HOPLIMIT);
1910 					break;
1911 				case IPV6_2292HOPOPTS:
1912 					/*
1913 					 * Check super-user privilege.
1914 					 * See comments for IPV6_RECVHOPOPTS.
1915 					 */
1916 					if (td != NULL) {
1917 						error = priv_check(td,
1918 						    PRIV_NETINET_SETHDROPTS);
1919 						if (error)
1920 							return (error);
1921 					}
1922 					OPTSET2292(IN6P_HOPOPTS);
1923 					break;
1924 				case IPV6_2292DSTOPTS:
1925 					if (td != NULL) {
1926 						error = priv_check(td,
1927 						    PRIV_NETINET_SETHDROPTS);
1928 						if (error)
1929 							return (error);
1930 					}
1931 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1932 					break;
1933 				case IPV6_2292RTHDR:
1934 					OPTSET2292(IN6P_RTHDR);
1935 					break;
1936 				}
1937 				break;
1938 			case IPV6_PKTINFO:
1939 			case IPV6_HOPOPTS:
1940 			case IPV6_RTHDR:
1941 			case IPV6_DSTOPTS:
1942 			case IPV6_RTHDRDSTOPTS:
1943 			case IPV6_NEXTHOP:
1944 			{
1945 				/* new advanced API (RFC3542) */
1946 				u_char *optbuf;
1947 				u_char optbuf_storage[MCLBYTES];
1948 				int optlen;
1949 				struct ip6_pktopts **optp;
1950 
1951 				/* cannot mix with RFC2292 */
1952 				if (OPTBIT(IN6P_RFC2292)) {
1953 					error = EINVAL;
1954 					break;
1955 				}
1956 
1957 				/*
1958 				 * We only ensure valsize is not too large
1959 				 * here.  Further validation will be done
1960 				 * later.
1961 				 */
1962 				error = sooptcopyin(sopt, optbuf_storage,
1963 				    sizeof(optbuf_storage), 0);
1964 				if (error)
1965 					break;
1966 				optlen = sopt->sopt_valsize;
1967 				optbuf = optbuf_storage;
1968 				INP_WLOCK(inp);
1969 				if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1970 					INP_WUNLOCK(inp);
1971 					return (ECONNRESET);
1972 				}
1973 				optp = &inp->in6p_outputopts;
1974 				error = ip6_pcbopt(optname, optbuf, optlen,
1975 				    optp, (td != NULL) ? td->td_ucred : NULL,
1976 				    uproto);
1977 				INP_WUNLOCK(inp);
1978 				break;
1979 			}
1980 #undef OPTSET
1981 
1982 			case IPV6_MULTICAST_IF:
1983 			case IPV6_MULTICAST_HOPS:
1984 			case IPV6_MULTICAST_LOOP:
1985 			case IPV6_JOIN_GROUP:
1986 			case IPV6_LEAVE_GROUP:
1987 			case IPV6_MSFILTER:
1988 			case MCAST_BLOCK_SOURCE:
1989 			case MCAST_UNBLOCK_SOURCE:
1990 			case MCAST_JOIN_GROUP:
1991 			case MCAST_LEAVE_GROUP:
1992 			case MCAST_JOIN_SOURCE_GROUP:
1993 			case MCAST_LEAVE_SOURCE_GROUP:
1994 				error = ip6_setmoptions(inp, sopt);
1995 				break;
1996 
1997 			case IPV6_PORTRANGE:
1998 				error = sooptcopyin(sopt, &optval,
1999 				    sizeof optval, sizeof optval);
2000 				if (error)
2001 					break;
2002 
2003 				INP_WLOCK(inp);
2004 				switch (optval) {
2005 				case IPV6_PORTRANGE_DEFAULT:
2006 					inp->inp_flags &= ~(INP_LOWPORT);
2007 					inp->inp_flags &= ~(INP_HIGHPORT);
2008 					break;
2009 
2010 				case IPV6_PORTRANGE_HIGH:
2011 					inp->inp_flags &= ~(INP_LOWPORT);
2012 					inp->inp_flags |= INP_HIGHPORT;
2013 					break;
2014 
2015 				case IPV6_PORTRANGE_LOW:
2016 					inp->inp_flags &= ~(INP_HIGHPORT);
2017 					inp->inp_flags |= INP_LOWPORT;
2018 					break;
2019 
2020 				default:
2021 					error = EINVAL;
2022 					break;
2023 				}
2024 				INP_WUNLOCK(inp);
2025 				break;
2026 
2027 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2028 			case IPV6_IPSEC_POLICY:
2029 				if (IPSEC_ENABLED(ipv6)) {
2030 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2031 					break;
2032 				}
2033 				/* FALLTHROUGH */
2034 #endif /* IPSEC */
2035 
2036 			default:
2037 				error = ENOPROTOOPT;
2038 				break;
2039 			}
2040 			break;
2041 
2042 		case SOPT_GET:
2043 			switch (optname) {
2044 
2045 			case IPV6_2292PKTOPTIONS:
2046 #ifdef IPV6_PKTOPTIONS
2047 			case IPV6_PKTOPTIONS:
2048 #endif
2049 				/*
2050 				 * RFC3542 (effectively) deprecated the
2051 				 * semantics of the 2292-style pktoptions.
2052 				 * Since it was not reliable in nature (i.e.,
2053 				 * applications had to expect the lack of some
2054 				 * information after all), it would make sense
2055 				 * to simplify this part by always returning
2056 				 * empty data.
2057 				 */
2058 				sopt->sopt_valsize = 0;
2059 				break;
2060 
2061 			case IPV6_RECVHOPOPTS:
2062 			case IPV6_RECVDSTOPTS:
2063 			case IPV6_RECVRTHDRDSTOPTS:
2064 			case IPV6_UNICAST_HOPS:
2065 			case IPV6_RECVPKTINFO:
2066 			case IPV6_RECVHOPLIMIT:
2067 			case IPV6_RECVRTHDR:
2068 			case IPV6_RECVPATHMTU:
2069 
2070 			case IPV6_V6ONLY:
2071 			case IPV6_PORTRANGE:
2072 			case IPV6_RECVTCLASS:
2073 			case IPV6_AUTOFLOWLABEL:
2074 			case IPV6_BINDANY:
2075 			case IPV6_FLOWID:
2076 			case IPV6_FLOWTYPE:
2077 			case IPV6_RECVFLOWID:
2078 #ifdef	RSS
2079 			case IPV6_RSSBUCKETID:
2080 			case IPV6_RECVRSSBUCKETID:
2081 #endif
2082 			case IPV6_BINDMULTI:
2083 				switch (optname) {
2084 
2085 				case IPV6_RECVHOPOPTS:
2086 					optval = OPTBIT(IN6P_HOPOPTS);
2087 					break;
2088 
2089 				case IPV6_RECVDSTOPTS:
2090 					optval = OPTBIT(IN6P_DSTOPTS);
2091 					break;
2092 
2093 				case IPV6_RECVRTHDRDSTOPTS:
2094 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2095 					break;
2096 
2097 				case IPV6_UNICAST_HOPS:
2098 					optval = inp->in6p_hops;
2099 					break;
2100 
2101 				case IPV6_RECVPKTINFO:
2102 					optval = OPTBIT(IN6P_PKTINFO);
2103 					break;
2104 
2105 				case IPV6_RECVHOPLIMIT:
2106 					optval = OPTBIT(IN6P_HOPLIMIT);
2107 					break;
2108 
2109 				case IPV6_RECVRTHDR:
2110 					optval = OPTBIT(IN6P_RTHDR);
2111 					break;
2112 
2113 				case IPV6_RECVPATHMTU:
2114 					optval = OPTBIT(IN6P_MTU);
2115 					break;
2116 
2117 				case IPV6_V6ONLY:
2118 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2119 					break;
2120 
2121 				case IPV6_PORTRANGE:
2122 				    {
2123 					int flags;
2124 					flags = inp->inp_flags;
2125 					if (flags & INP_HIGHPORT)
2126 						optval = IPV6_PORTRANGE_HIGH;
2127 					else if (flags & INP_LOWPORT)
2128 						optval = IPV6_PORTRANGE_LOW;
2129 					else
2130 						optval = 0;
2131 					break;
2132 				    }
2133 				case IPV6_RECVTCLASS:
2134 					optval = OPTBIT(IN6P_TCLASS);
2135 					break;
2136 
2137 				case IPV6_AUTOFLOWLABEL:
2138 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2139 					break;
2140 
2141 				case IPV6_ORIGDSTADDR:
2142 					optval = OPTBIT2(INP_ORIGDSTADDR);
2143 					break;
2144 
2145 				case IPV6_BINDANY:
2146 					optval = OPTBIT(INP_BINDANY);
2147 					break;
2148 
2149 				case IPV6_FLOWID:
2150 					optval = inp->inp_flowid;
2151 					break;
2152 
2153 				case IPV6_FLOWTYPE:
2154 					optval = inp->inp_flowtype;
2155 					break;
2156 
2157 				case IPV6_RECVFLOWID:
2158 					optval = OPTBIT2(INP_RECVFLOWID);
2159 					break;
2160 #ifdef	RSS
2161 				case IPV6_RSSBUCKETID:
2162 					retval =
2163 					    rss_hash2bucket(inp->inp_flowid,
2164 					    inp->inp_flowtype,
2165 					    &rss_bucket);
2166 					if (retval == 0)
2167 						optval = rss_bucket;
2168 					else
2169 						error = EINVAL;
2170 					break;
2171 
2172 				case IPV6_RECVRSSBUCKETID:
2173 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2174 					break;
2175 #endif
2176 
2177 				case IPV6_BINDMULTI:
2178 					optval = OPTBIT2(INP_BINDMULTI);
2179 					break;
2180 
2181 				}
2182 				if (error)
2183 					break;
2184 				error = sooptcopyout(sopt, &optval,
2185 					sizeof optval);
2186 				break;
2187 
2188 			case IPV6_PATHMTU:
2189 			{
2190 				u_long pmtu = 0;
2191 				struct ip6_mtuinfo mtuinfo;
2192 				struct in6_addr addr;
2193 
2194 				if (!(so->so_state & SS_ISCONNECTED))
2195 					return (ENOTCONN);
2196 				/*
2197 				 * XXX: we dot not consider the case of source
2198 				 * routing, or optional information to specify
2199 				 * the outgoing interface.
2200 				 * Copy faddr out of inp to avoid holding lock
2201 				 * on inp during route lookup.
2202 				 */
2203 				INP_RLOCK(inp);
2204 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2205 				INP_RUNLOCK(inp);
2206 				error = ip6_getpmtu_ctl(so->so_fibnum,
2207 				    &addr, &pmtu);
2208 				if (error)
2209 					break;
2210 				if (pmtu > IPV6_MAXPACKET)
2211 					pmtu = IPV6_MAXPACKET;
2212 
2213 				bzero(&mtuinfo, sizeof(mtuinfo));
2214 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2215 				optdata = (void *)&mtuinfo;
2216 				optdatalen = sizeof(mtuinfo);
2217 				error = sooptcopyout(sopt, optdata,
2218 				    optdatalen);
2219 				break;
2220 			}
2221 
2222 			case IPV6_2292PKTINFO:
2223 			case IPV6_2292HOPLIMIT:
2224 			case IPV6_2292HOPOPTS:
2225 			case IPV6_2292RTHDR:
2226 			case IPV6_2292DSTOPTS:
2227 				switch (optname) {
2228 				case IPV6_2292PKTINFO:
2229 					optval = OPTBIT(IN6P_PKTINFO);
2230 					break;
2231 				case IPV6_2292HOPLIMIT:
2232 					optval = OPTBIT(IN6P_HOPLIMIT);
2233 					break;
2234 				case IPV6_2292HOPOPTS:
2235 					optval = OPTBIT(IN6P_HOPOPTS);
2236 					break;
2237 				case IPV6_2292RTHDR:
2238 					optval = OPTBIT(IN6P_RTHDR);
2239 					break;
2240 				case IPV6_2292DSTOPTS:
2241 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2242 					break;
2243 				}
2244 				error = sooptcopyout(sopt, &optval,
2245 				    sizeof optval);
2246 				break;
2247 			case IPV6_PKTINFO:
2248 			case IPV6_HOPOPTS:
2249 			case IPV6_RTHDR:
2250 			case IPV6_DSTOPTS:
2251 			case IPV6_RTHDRDSTOPTS:
2252 			case IPV6_NEXTHOP:
2253 			case IPV6_TCLASS:
2254 			case IPV6_DONTFRAG:
2255 			case IPV6_USE_MIN_MTU:
2256 			case IPV6_PREFER_TEMPADDR:
2257 				error = ip6_getpcbopt(inp, optname, sopt);
2258 				break;
2259 
2260 			case IPV6_MULTICAST_IF:
2261 			case IPV6_MULTICAST_HOPS:
2262 			case IPV6_MULTICAST_LOOP:
2263 			case IPV6_MSFILTER:
2264 				error = ip6_getmoptions(inp, sopt);
2265 				break;
2266 
2267 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2268 			case IPV6_IPSEC_POLICY:
2269 				if (IPSEC_ENABLED(ipv6)) {
2270 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2271 					break;
2272 				}
2273 				/* FALLTHROUGH */
2274 #endif /* IPSEC */
2275 			default:
2276 				error = ENOPROTOOPT;
2277 				break;
2278 			}
2279 			break;
2280 		}
2281 	}
2282 	return (error);
2283 }
2284 
2285 int
2286 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2287 {
2288 	int error = 0, optval, optlen;
2289 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2290 	struct inpcb *inp = sotoinpcb(so);
2291 	int level, op, optname;
2292 
2293 	level = sopt->sopt_level;
2294 	op = sopt->sopt_dir;
2295 	optname = sopt->sopt_name;
2296 	optlen = sopt->sopt_valsize;
2297 
2298 	if (level != IPPROTO_IPV6) {
2299 		return (EINVAL);
2300 	}
2301 
2302 	switch (optname) {
2303 	case IPV6_CHECKSUM:
2304 		/*
2305 		 * For ICMPv6 sockets, no modification allowed for checksum
2306 		 * offset, permit "no change" values to help existing apps.
2307 		 *
2308 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2309 		 * for an ICMPv6 socket will fail."
2310 		 * The current behavior does not meet RFC3542.
2311 		 */
2312 		switch (op) {
2313 		case SOPT_SET:
2314 			if (optlen != sizeof(int)) {
2315 				error = EINVAL;
2316 				break;
2317 			}
2318 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2319 					    sizeof(optval));
2320 			if (error)
2321 				break;
2322 			if (optval < -1 || (optval % 2) != 0) {
2323 				/*
2324 				 * The API assumes non-negative even offset
2325 				 * values or -1 as a special value.
2326 				 */
2327 				error = EINVAL;
2328 			} else if (so->so_proto->pr_protocol ==
2329 			    IPPROTO_ICMPV6) {
2330 				if (optval != icmp6off)
2331 					error = EINVAL;
2332 			} else
2333 				inp->in6p_cksum = optval;
2334 			break;
2335 
2336 		case SOPT_GET:
2337 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2338 				optval = icmp6off;
2339 			else
2340 				optval = inp->in6p_cksum;
2341 
2342 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2343 			break;
2344 
2345 		default:
2346 			error = EINVAL;
2347 			break;
2348 		}
2349 		break;
2350 
2351 	default:
2352 		error = ENOPROTOOPT;
2353 		break;
2354 	}
2355 
2356 	return (error);
2357 }
2358 
2359 /*
2360  * Set up IP6 options in pcb for insertion in output packets or
2361  * specifying behavior of outgoing packets.
2362  */
2363 static int
2364 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2365     struct socket *so, struct sockopt *sopt)
2366 {
2367 	struct ip6_pktopts *opt = *pktopt;
2368 	int error = 0;
2369 	struct thread *td = sopt->sopt_td;
2370 
2371 	/* turn off any old options. */
2372 	if (opt) {
2373 #ifdef DIAGNOSTIC
2374 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2375 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2376 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2377 			printf("ip6_pcbopts: all specified options are cleared.\n");
2378 #endif
2379 		ip6_clearpktopts(opt, -1);
2380 	} else
2381 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2382 	*pktopt = NULL;
2383 
2384 	if (!m || m->m_len == 0) {
2385 		/*
2386 		 * Only turning off any previous options, regardless of
2387 		 * whether the opt is just created or given.
2388 		 */
2389 		free(opt, M_IP6OPT);
2390 		return (0);
2391 	}
2392 
2393 	/*  set options specified by user. */
2394 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2395 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2396 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2397 		free(opt, M_IP6OPT);
2398 		return (error);
2399 	}
2400 	*pktopt = opt;
2401 	return (0);
2402 }
2403 
2404 /*
2405  * initialize ip6_pktopts.  beware that there are non-zero default values in
2406  * the struct.
2407  */
2408 void
2409 ip6_initpktopts(struct ip6_pktopts *opt)
2410 {
2411 
2412 	bzero(opt, sizeof(*opt));
2413 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2414 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2415 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2416 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2417 }
2418 
2419 static int
2420 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2421     struct ucred *cred, int uproto)
2422 {
2423 	struct ip6_pktopts *opt;
2424 
2425 	if (*pktopt == NULL) {
2426 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2427 		    M_NOWAIT);
2428 		if (*pktopt == NULL)
2429 			return (ENOBUFS);
2430 		ip6_initpktopts(*pktopt);
2431 	}
2432 	opt = *pktopt;
2433 
2434 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2435 }
2436 
2437 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2438 	if (pktopt && pktopt->field) {						\
2439 		INP_RUNLOCK(inp);						\
2440 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2441 		malloc_optdata = true;						\
2442 		INP_RLOCK(inp);							\
2443 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2444 			INP_RUNLOCK(inp);					\
2445 			free(optdata, M_TEMP);					\
2446 			return (ECONNRESET);					\
2447 		}								\
2448 		pktopt = inp->in6p_outputopts;					\
2449 		if (pktopt && pktopt->field) {					\
2450 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2451 			bcopy(&pktopt->field, optdata, optdatalen);		\
2452 		} else {							\
2453 			free(optdata, M_TEMP);					\
2454 			optdata = NULL;						\
2455 			malloc_optdata = false;					\
2456 		}								\
2457 	}									\
2458 } while(0)
2459 
2460 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2461 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2462 
2463 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2464 	pktopt->field->sa_len)
2465 
2466 static int
2467 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2468 {
2469 	void *optdata = NULL;
2470 	bool malloc_optdata = false;
2471 	int optdatalen = 0;
2472 	int error = 0;
2473 	struct in6_pktinfo null_pktinfo;
2474 	int deftclass = 0, on;
2475 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2476 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2477 	struct ip6_pktopts *pktopt;
2478 
2479 	INP_RLOCK(inp);
2480 	pktopt = inp->in6p_outputopts;
2481 
2482 	switch (optname) {
2483 	case IPV6_PKTINFO:
2484 		optdata = (void *)&null_pktinfo;
2485 		if (pktopt && pktopt->ip6po_pktinfo) {
2486 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2487 			    sizeof(null_pktinfo));
2488 			in6_clearscope(&null_pktinfo.ipi6_addr);
2489 		} else {
2490 			/* XXX: we don't have to do this every time... */
2491 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2492 		}
2493 		optdatalen = sizeof(struct in6_pktinfo);
2494 		break;
2495 	case IPV6_TCLASS:
2496 		if (pktopt && pktopt->ip6po_tclass >= 0)
2497 			deftclass = pktopt->ip6po_tclass;
2498 		optdata = (void *)&deftclass;
2499 		optdatalen = sizeof(int);
2500 		break;
2501 	case IPV6_HOPOPTS:
2502 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2503 		break;
2504 	case IPV6_RTHDR:
2505 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2506 		break;
2507 	case IPV6_RTHDRDSTOPTS:
2508 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2509 		break;
2510 	case IPV6_DSTOPTS:
2511 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2512 		break;
2513 	case IPV6_NEXTHOP:
2514 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2515 		break;
2516 	case IPV6_USE_MIN_MTU:
2517 		if (pktopt)
2518 			defminmtu = pktopt->ip6po_minmtu;
2519 		optdata = (void *)&defminmtu;
2520 		optdatalen = sizeof(int);
2521 		break;
2522 	case IPV6_DONTFRAG:
2523 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2524 			on = 1;
2525 		else
2526 			on = 0;
2527 		optdata = (void *)&on;
2528 		optdatalen = sizeof(on);
2529 		break;
2530 	case IPV6_PREFER_TEMPADDR:
2531 		if (pktopt)
2532 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2533 		optdata = (void *)&defpreftemp;
2534 		optdatalen = sizeof(int);
2535 		break;
2536 	default:		/* should not happen */
2537 #ifdef DIAGNOSTIC
2538 		panic("ip6_getpcbopt: unexpected option\n");
2539 #endif
2540 		INP_RUNLOCK(inp);
2541 		return (ENOPROTOOPT);
2542 	}
2543 	INP_RUNLOCK(inp);
2544 
2545 	error = sooptcopyout(sopt, optdata, optdatalen);
2546 	if (malloc_optdata)
2547 		free(optdata, M_TEMP);
2548 
2549 	return (error);
2550 }
2551 
2552 void
2553 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2554 {
2555 	if (pktopt == NULL)
2556 		return;
2557 
2558 	if (optname == -1 || optname == IPV6_PKTINFO) {
2559 		if (pktopt->ip6po_pktinfo)
2560 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2561 		pktopt->ip6po_pktinfo = NULL;
2562 	}
2563 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2564 		pktopt->ip6po_hlim = -1;
2565 	if (optname == -1 || optname == IPV6_TCLASS)
2566 		pktopt->ip6po_tclass = -1;
2567 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2568 		if (pktopt->ip6po_nextroute.ro_rt) {
2569 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2570 			pktopt->ip6po_nextroute.ro_rt = NULL;
2571 		}
2572 		if (pktopt->ip6po_nexthop)
2573 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2574 		pktopt->ip6po_nexthop = NULL;
2575 	}
2576 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2577 		if (pktopt->ip6po_hbh)
2578 			free(pktopt->ip6po_hbh, M_IP6OPT);
2579 		pktopt->ip6po_hbh = NULL;
2580 	}
2581 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2582 		if (pktopt->ip6po_dest1)
2583 			free(pktopt->ip6po_dest1, M_IP6OPT);
2584 		pktopt->ip6po_dest1 = NULL;
2585 	}
2586 	if (optname == -1 || optname == IPV6_RTHDR) {
2587 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2588 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2589 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2590 		if (pktopt->ip6po_route.ro_rt) {
2591 			RTFREE(pktopt->ip6po_route.ro_rt);
2592 			pktopt->ip6po_route.ro_rt = NULL;
2593 		}
2594 	}
2595 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2596 		if (pktopt->ip6po_dest2)
2597 			free(pktopt->ip6po_dest2, M_IP6OPT);
2598 		pktopt->ip6po_dest2 = NULL;
2599 	}
2600 }
2601 
2602 #define PKTOPT_EXTHDRCPY(type) \
2603 do {\
2604 	if (src->type) {\
2605 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2606 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2607 		if (dst->type == NULL)\
2608 			goto bad;\
2609 		bcopy(src->type, dst->type, hlen);\
2610 	}\
2611 } while (/*CONSTCOND*/ 0)
2612 
2613 static int
2614 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2615 {
2616 	if (dst == NULL || src == NULL)  {
2617 		printf("ip6_clearpktopts: invalid argument\n");
2618 		return (EINVAL);
2619 	}
2620 
2621 	dst->ip6po_hlim = src->ip6po_hlim;
2622 	dst->ip6po_tclass = src->ip6po_tclass;
2623 	dst->ip6po_flags = src->ip6po_flags;
2624 	dst->ip6po_minmtu = src->ip6po_minmtu;
2625 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2626 	if (src->ip6po_pktinfo) {
2627 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2628 		    M_IP6OPT, canwait);
2629 		if (dst->ip6po_pktinfo == NULL)
2630 			goto bad;
2631 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2632 	}
2633 	if (src->ip6po_nexthop) {
2634 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2635 		    M_IP6OPT, canwait);
2636 		if (dst->ip6po_nexthop == NULL)
2637 			goto bad;
2638 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2639 		    src->ip6po_nexthop->sa_len);
2640 	}
2641 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2642 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2643 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2644 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2645 	return (0);
2646 
2647   bad:
2648 	ip6_clearpktopts(dst, -1);
2649 	return (ENOBUFS);
2650 }
2651 #undef PKTOPT_EXTHDRCPY
2652 
2653 struct ip6_pktopts *
2654 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2655 {
2656 	int error;
2657 	struct ip6_pktopts *dst;
2658 
2659 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2660 	if (dst == NULL)
2661 		return (NULL);
2662 	ip6_initpktopts(dst);
2663 
2664 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2665 		free(dst, M_IP6OPT);
2666 		return (NULL);
2667 	}
2668 
2669 	return (dst);
2670 }
2671 
2672 void
2673 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2674 {
2675 	if (pktopt == NULL)
2676 		return;
2677 
2678 	ip6_clearpktopts(pktopt, -1);
2679 
2680 	free(pktopt, M_IP6OPT);
2681 }
2682 
2683 /*
2684  * Set IPv6 outgoing packet options based on advanced API.
2685  */
2686 int
2687 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2688     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2689 {
2690 	struct cmsghdr *cm = NULL;
2691 
2692 	if (control == NULL || opt == NULL)
2693 		return (EINVAL);
2694 
2695 	ip6_initpktopts(opt);
2696 	if (stickyopt) {
2697 		int error;
2698 
2699 		/*
2700 		 * If stickyopt is provided, make a local copy of the options
2701 		 * for this particular packet, then override them by ancillary
2702 		 * objects.
2703 		 * XXX: copypktopts() does not copy the cached route to a next
2704 		 * hop (if any).  This is not very good in terms of efficiency,
2705 		 * but we can allow this since this option should be rarely
2706 		 * used.
2707 		 */
2708 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2709 			return (error);
2710 	}
2711 
2712 	/*
2713 	 * XXX: Currently, we assume all the optional information is stored
2714 	 * in a single mbuf.
2715 	 */
2716 	if (control->m_next)
2717 		return (EINVAL);
2718 
2719 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2720 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2721 		int error;
2722 
2723 		if (control->m_len < CMSG_LEN(0))
2724 			return (EINVAL);
2725 
2726 		cm = mtod(control, struct cmsghdr *);
2727 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2728 			return (EINVAL);
2729 		if (cm->cmsg_level != IPPROTO_IPV6)
2730 			continue;
2731 
2732 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2733 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2734 		if (error)
2735 			return (error);
2736 	}
2737 
2738 	return (0);
2739 }
2740 
2741 /*
2742  * Set a particular packet option, as a sticky option or an ancillary data
2743  * item.  "len" can be 0 only when it's a sticky option.
2744  * We have 4 cases of combination of "sticky" and "cmsg":
2745  * "sticky=0, cmsg=0": impossible
2746  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2747  * "sticky=1, cmsg=0": RFC3542 socket option
2748  * "sticky=1, cmsg=1": RFC2292 socket option
2749  */
2750 static int
2751 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2752     struct ucred *cred, int sticky, int cmsg, int uproto)
2753 {
2754 	int minmtupolicy, preftemp;
2755 	int error;
2756 
2757 	if (!sticky && !cmsg) {
2758 #ifdef DIAGNOSTIC
2759 		printf("ip6_setpktopt: impossible case\n");
2760 #endif
2761 		return (EINVAL);
2762 	}
2763 
2764 	/*
2765 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2766 	 * not be specified in the context of RFC3542.  Conversely,
2767 	 * RFC3542 types should not be specified in the context of RFC2292.
2768 	 */
2769 	if (!cmsg) {
2770 		switch (optname) {
2771 		case IPV6_2292PKTINFO:
2772 		case IPV6_2292HOPLIMIT:
2773 		case IPV6_2292NEXTHOP:
2774 		case IPV6_2292HOPOPTS:
2775 		case IPV6_2292DSTOPTS:
2776 		case IPV6_2292RTHDR:
2777 		case IPV6_2292PKTOPTIONS:
2778 			return (ENOPROTOOPT);
2779 		}
2780 	}
2781 	if (sticky && cmsg) {
2782 		switch (optname) {
2783 		case IPV6_PKTINFO:
2784 		case IPV6_HOPLIMIT:
2785 		case IPV6_NEXTHOP:
2786 		case IPV6_HOPOPTS:
2787 		case IPV6_DSTOPTS:
2788 		case IPV6_RTHDRDSTOPTS:
2789 		case IPV6_RTHDR:
2790 		case IPV6_USE_MIN_MTU:
2791 		case IPV6_DONTFRAG:
2792 		case IPV6_TCLASS:
2793 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2794 			return (ENOPROTOOPT);
2795 		}
2796 	}
2797 
2798 	switch (optname) {
2799 	case IPV6_2292PKTINFO:
2800 	case IPV6_PKTINFO:
2801 	{
2802 		struct ifnet *ifp = NULL;
2803 		struct in6_pktinfo *pktinfo;
2804 
2805 		if (len != sizeof(struct in6_pktinfo))
2806 			return (EINVAL);
2807 
2808 		pktinfo = (struct in6_pktinfo *)buf;
2809 
2810 		/*
2811 		 * An application can clear any sticky IPV6_PKTINFO option by
2812 		 * doing a "regular" setsockopt with ipi6_addr being
2813 		 * in6addr_any and ipi6_ifindex being zero.
2814 		 * [RFC 3542, Section 6]
2815 		 */
2816 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2817 		    pktinfo->ipi6_ifindex == 0 &&
2818 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2819 			ip6_clearpktopts(opt, optname);
2820 			break;
2821 		}
2822 
2823 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2824 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2825 			return (EINVAL);
2826 		}
2827 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2828 			return (EINVAL);
2829 		/* validate the interface index if specified. */
2830 		if (pktinfo->ipi6_ifindex > V_if_index)
2831 			 return (ENXIO);
2832 		if (pktinfo->ipi6_ifindex) {
2833 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2834 			if (ifp == NULL)
2835 				return (ENXIO);
2836 		}
2837 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2838 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2839 			return (ENETDOWN);
2840 
2841 		if (ifp != NULL &&
2842 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2843 			struct in6_ifaddr *ia;
2844 
2845 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2846 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2847 			if (ia == NULL)
2848 				return (EADDRNOTAVAIL);
2849 			ifa_free(&ia->ia_ifa);
2850 		}
2851 		/*
2852 		 * We store the address anyway, and let in6_selectsrc()
2853 		 * validate the specified address.  This is because ipi6_addr
2854 		 * may not have enough information about its scope zone, and
2855 		 * we may need additional information (such as outgoing
2856 		 * interface or the scope zone of a destination address) to
2857 		 * disambiguate the scope.
2858 		 * XXX: the delay of the validation may confuse the
2859 		 * application when it is used as a sticky option.
2860 		 */
2861 		if (opt->ip6po_pktinfo == NULL) {
2862 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2863 			    M_IP6OPT, M_NOWAIT);
2864 			if (opt->ip6po_pktinfo == NULL)
2865 				return (ENOBUFS);
2866 		}
2867 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2868 		break;
2869 	}
2870 
2871 	case IPV6_2292HOPLIMIT:
2872 	case IPV6_HOPLIMIT:
2873 	{
2874 		int *hlimp;
2875 
2876 		/*
2877 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2878 		 * to simplify the ordering among hoplimit options.
2879 		 */
2880 		if (optname == IPV6_HOPLIMIT && sticky)
2881 			return (ENOPROTOOPT);
2882 
2883 		if (len != sizeof(int))
2884 			return (EINVAL);
2885 		hlimp = (int *)buf;
2886 		if (*hlimp < -1 || *hlimp > 255)
2887 			return (EINVAL);
2888 
2889 		opt->ip6po_hlim = *hlimp;
2890 		break;
2891 	}
2892 
2893 	case IPV6_TCLASS:
2894 	{
2895 		int tclass;
2896 
2897 		if (len != sizeof(int))
2898 			return (EINVAL);
2899 		tclass = *(int *)buf;
2900 		if (tclass < -1 || tclass > 255)
2901 			return (EINVAL);
2902 
2903 		opt->ip6po_tclass = tclass;
2904 		break;
2905 	}
2906 
2907 	case IPV6_2292NEXTHOP:
2908 	case IPV6_NEXTHOP:
2909 		if (cred != NULL) {
2910 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2911 			if (error)
2912 				return (error);
2913 		}
2914 
2915 		if (len == 0) {	/* just remove the option */
2916 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2917 			break;
2918 		}
2919 
2920 		/* check if cmsg_len is large enough for sa_len */
2921 		if (len < sizeof(struct sockaddr) || len < *buf)
2922 			return (EINVAL);
2923 
2924 		switch (((struct sockaddr *)buf)->sa_family) {
2925 		case AF_INET6:
2926 		{
2927 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2928 			int error;
2929 
2930 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2931 				return (EINVAL);
2932 
2933 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2934 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2935 				return (EINVAL);
2936 			}
2937 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2938 			    != 0) {
2939 				return (error);
2940 			}
2941 			break;
2942 		}
2943 		case AF_LINK:	/* should eventually be supported */
2944 		default:
2945 			return (EAFNOSUPPORT);
2946 		}
2947 
2948 		/* turn off the previous option, then set the new option. */
2949 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2950 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2951 		if (opt->ip6po_nexthop == NULL)
2952 			return (ENOBUFS);
2953 		bcopy(buf, opt->ip6po_nexthop, *buf);
2954 		break;
2955 
2956 	case IPV6_2292HOPOPTS:
2957 	case IPV6_HOPOPTS:
2958 	{
2959 		struct ip6_hbh *hbh;
2960 		int hbhlen;
2961 
2962 		/*
2963 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2964 		 * options, since per-option restriction has too much
2965 		 * overhead.
2966 		 */
2967 		if (cred != NULL) {
2968 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2969 			if (error)
2970 				return (error);
2971 		}
2972 
2973 		if (len == 0) {
2974 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2975 			break;	/* just remove the option */
2976 		}
2977 
2978 		/* message length validation */
2979 		if (len < sizeof(struct ip6_hbh))
2980 			return (EINVAL);
2981 		hbh = (struct ip6_hbh *)buf;
2982 		hbhlen = (hbh->ip6h_len + 1) << 3;
2983 		if (len != hbhlen)
2984 			return (EINVAL);
2985 
2986 		/* turn off the previous option, then set the new option. */
2987 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2988 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2989 		if (opt->ip6po_hbh == NULL)
2990 			return (ENOBUFS);
2991 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2992 
2993 		break;
2994 	}
2995 
2996 	case IPV6_2292DSTOPTS:
2997 	case IPV6_DSTOPTS:
2998 	case IPV6_RTHDRDSTOPTS:
2999 	{
3000 		struct ip6_dest *dest, **newdest = NULL;
3001 		int destlen;
3002 
3003 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3004 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3005 			if (error)
3006 				return (error);
3007 		}
3008 
3009 		if (len == 0) {
3010 			ip6_clearpktopts(opt, optname);
3011 			break;	/* just remove the option */
3012 		}
3013 
3014 		/* message length validation */
3015 		if (len < sizeof(struct ip6_dest))
3016 			return (EINVAL);
3017 		dest = (struct ip6_dest *)buf;
3018 		destlen = (dest->ip6d_len + 1) << 3;
3019 		if (len != destlen)
3020 			return (EINVAL);
3021 
3022 		/*
3023 		 * Determine the position that the destination options header
3024 		 * should be inserted; before or after the routing header.
3025 		 */
3026 		switch (optname) {
3027 		case IPV6_2292DSTOPTS:
3028 			/*
3029 			 * The old advacned API is ambiguous on this point.
3030 			 * Our approach is to determine the position based
3031 			 * according to the existence of a routing header.
3032 			 * Note, however, that this depends on the order of the
3033 			 * extension headers in the ancillary data; the 1st
3034 			 * part of the destination options header must appear
3035 			 * before the routing header in the ancillary data,
3036 			 * too.
3037 			 * RFC3542 solved the ambiguity by introducing
3038 			 * separate ancillary data or option types.
3039 			 */
3040 			if (opt->ip6po_rthdr == NULL)
3041 				newdest = &opt->ip6po_dest1;
3042 			else
3043 				newdest = &opt->ip6po_dest2;
3044 			break;
3045 		case IPV6_RTHDRDSTOPTS:
3046 			newdest = &opt->ip6po_dest1;
3047 			break;
3048 		case IPV6_DSTOPTS:
3049 			newdest = &opt->ip6po_dest2;
3050 			break;
3051 		}
3052 
3053 		/* turn off the previous option, then set the new option. */
3054 		ip6_clearpktopts(opt, optname);
3055 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3056 		if (*newdest == NULL)
3057 			return (ENOBUFS);
3058 		bcopy(dest, *newdest, destlen);
3059 
3060 		break;
3061 	}
3062 
3063 	case IPV6_2292RTHDR:
3064 	case IPV6_RTHDR:
3065 	{
3066 		struct ip6_rthdr *rth;
3067 		int rthlen;
3068 
3069 		if (len == 0) {
3070 			ip6_clearpktopts(opt, IPV6_RTHDR);
3071 			break;	/* just remove the option */
3072 		}
3073 
3074 		/* message length validation */
3075 		if (len < sizeof(struct ip6_rthdr))
3076 			return (EINVAL);
3077 		rth = (struct ip6_rthdr *)buf;
3078 		rthlen = (rth->ip6r_len + 1) << 3;
3079 		if (len != rthlen)
3080 			return (EINVAL);
3081 
3082 		switch (rth->ip6r_type) {
3083 		case IPV6_RTHDR_TYPE_0:
3084 			if (rth->ip6r_len == 0)	/* must contain one addr */
3085 				return (EINVAL);
3086 			if (rth->ip6r_len % 2) /* length must be even */
3087 				return (EINVAL);
3088 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3089 				return (EINVAL);
3090 			break;
3091 		default:
3092 			return (EINVAL);	/* not supported */
3093 		}
3094 
3095 		/* turn off the previous option */
3096 		ip6_clearpktopts(opt, IPV6_RTHDR);
3097 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3098 		if (opt->ip6po_rthdr == NULL)
3099 			return (ENOBUFS);
3100 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3101 
3102 		break;
3103 	}
3104 
3105 	case IPV6_USE_MIN_MTU:
3106 		if (len != sizeof(int))
3107 			return (EINVAL);
3108 		minmtupolicy = *(int *)buf;
3109 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3110 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3111 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3112 			return (EINVAL);
3113 		}
3114 		opt->ip6po_minmtu = minmtupolicy;
3115 		break;
3116 
3117 	case IPV6_DONTFRAG:
3118 		if (len != sizeof(int))
3119 			return (EINVAL);
3120 
3121 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3122 			/*
3123 			 * we ignore this option for TCP sockets.
3124 			 * (RFC3542 leaves this case unspecified.)
3125 			 */
3126 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3127 		} else
3128 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3129 		break;
3130 
3131 	case IPV6_PREFER_TEMPADDR:
3132 		if (len != sizeof(int))
3133 			return (EINVAL);
3134 		preftemp = *(int *)buf;
3135 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3136 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3137 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3138 			return (EINVAL);
3139 		}
3140 		opt->ip6po_prefer_tempaddr = preftemp;
3141 		break;
3142 
3143 	default:
3144 		return (ENOPROTOOPT);
3145 	} /* end of switch */
3146 
3147 	return (0);
3148 }
3149 
3150 /*
3151  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3152  * packet to the input queue of a specified interface.  Note that this
3153  * calls the output routine of the loopback "driver", but with an interface
3154  * pointer that might NOT be &loif -- easier than replicating that code here.
3155  */
3156 void
3157 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3158 {
3159 	struct mbuf *copym;
3160 	struct ip6_hdr *ip6;
3161 
3162 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3163 	if (copym == NULL)
3164 		return;
3165 
3166 	/*
3167 	 * Make sure to deep-copy IPv6 header portion in case the data
3168 	 * is in an mbuf cluster, so that we can safely override the IPv6
3169 	 * header portion later.
3170 	 */
3171 	if (!M_WRITABLE(copym) ||
3172 	    copym->m_len < sizeof(struct ip6_hdr)) {
3173 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3174 		if (copym == NULL)
3175 			return;
3176 	}
3177 	ip6 = mtod(copym, struct ip6_hdr *);
3178 	/*
3179 	 * clear embedded scope identifiers if necessary.
3180 	 * in6_clearscope will touch the addresses only when necessary.
3181 	 */
3182 	in6_clearscope(&ip6->ip6_src);
3183 	in6_clearscope(&ip6->ip6_dst);
3184 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3185 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3186 		    CSUM_PSEUDO_HDR;
3187 		copym->m_pkthdr.csum_data = 0xffff;
3188 	}
3189 	if_simloop(ifp, copym, AF_INET6, 0);
3190 }
3191 
3192 /*
3193  * Chop IPv6 header off from the payload.
3194  */
3195 static int
3196 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3197 {
3198 	struct mbuf *mh;
3199 	struct ip6_hdr *ip6;
3200 
3201 	ip6 = mtod(m, struct ip6_hdr *);
3202 	if (m->m_len > sizeof(*ip6)) {
3203 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3204 		if (mh == NULL) {
3205 			m_freem(m);
3206 			return ENOBUFS;
3207 		}
3208 		m_move_pkthdr(mh, m);
3209 		M_ALIGN(mh, sizeof(*ip6));
3210 		m->m_len -= sizeof(*ip6);
3211 		m->m_data += sizeof(*ip6);
3212 		mh->m_next = m;
3213 		m = mh;
3214 		m->m_len = sizeof(*ip6);
3215 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3216 	}
3217 	exthdrs->ip6e_ip6 = m;
3218 	return 0;
3219 }
3220 
3221 /*
3222  * Compute IPv6 extension header length.
3223  */
3224 int
3225 ip6_optlen(struct inpcb *inp)
3226 {
3227 	int len;
3228 
3229 	if (!inp->in6p_outputopts)
3230 		return 0;
3231 
3232 	len = 0;
3233 #define elen(x) \
3234     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3235 
3236 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3237 	if (inp->in6p_outputopts->ip6po_rthdr)
3238 		/* dest1 is valid with rthdr only */
3239 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3240 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3241 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3242 	return len;
3243 #undef elen
3244 }
3245