xref: /freebsd/sys/netinet6/ip6_output.c (revision f2530c80db7b29b95368fce956b3a778f096b368)
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 route_in6 ip6route;
388 	struct rtentry *rt = NULL;
389 	struct sockaddr_in6 *dst, src_sa, dst_sa;
390 	struct in6_addr odst;
391 	int error = 0;
392 	struct in6_ifaddr *ia = NULL;
393 	u_long mtu;
394 	int alwaysfrag, dontfrag;
395 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
396 	struct ip6_exthdrs exthdrs;
397 	struct in6_addr src0, dst0;
398 	u_int32_t zone;
399 	struct route_in6 *ro_pmtu = NULL;
400 	int hdrsplit = 0;
401 	int sw_csum, tso;
402 	int needfiblookup;
403 	uint32_t fibnum;
404 	struct m_tag *fwd_tag = NULL;
405 	uint32_t id;
406 
407 	NET_EPOCH_ASSERT();
408 
409 	if (inp != NULL) {
410 		INP_LOCK_ASSERT(inp);
411 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
412 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
413 			/* unconditionally set flowid */
414 			m->m_pkthdr.flowid = inp->inp_flowid;
415 			M_HASHTYPE_SET(m, inp->inp_flowtype);
416 		}
417 #ifdef NUMA
418 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
419 #endif
420 	}
421 
422 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
423 	/*
424 	 * IPSec checking which handles several cases.
425 	 * FAST IPSEC: We re-injected the packet.
426 	 * XXX: need scope argument.
427 	 */
428 	if (IPSEC_ENABLED(ipv6)) {
429 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
430 			if (error == EINPROGRESS)
431 				error = 0;
432 			goto done;
433 		}
434 	}
435 #endif /* IPSEC */
436 
437 	bzero(&exthdrs, sizeof(exthdrs));
438 	if (opt) {
439 		/* Hop-by-Hop options header */
440 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
441 		/* Destination options header(1st part) */
442 		if (opt->ip6po_rthdr) {
443 			/*
444 			 * Destination options header(1st part)
445 			 * This only makes sense with a routing header.
446 			 * See Section 9.2 of RFC 3542.
447 			 * Disabling this part just for MIP6 convenience is
448 			 * a bad idea.  We need to think carefully about a
449 			 * way to make the advanced API coexist with MIP6
450 			 * options, which might automatically be inserted in
451 			 * the kernel.
452 			 */
453 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
454 		}
455 		/* Routing header */
456 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
457 		/* Destination options header(2nd part) */
458 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
459 	}
460 
461 	/*
462 	 * Calculate the total length of the extension header chain.
463 	 * Keep the length of the unfragmentable part for fragmentation.
464 	 */
465 	optlen = 0;
466 	if (exthdrs.ip6e_hbh)
467 		optlen += exthdrs.ip6e_hbh->m_len;
468 	if (exthdrs.ip6e_dest1)
469 		optlen += exthdrs.ip6e_dest1->m_len;
470 	if (exthdrs.ip6e_rthdr)
471 		optlen += exthdrs.ip6e_rthdr->m_len;
472 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
473 
474 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
475 	if (exthdrs.ip6e_dest2)
476 		optlen += exthdrs.ip6e_dest2->m_len;
477 
478 	/*
479 	 * If there is at least one extension header,
480 	 * separate IP6 header from the payload.
481 	 */
482 	if (optlen && !hdrsplit) {
483 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
484 			m = NULL;
485 			goto freehdrs;
486 		}
487 		m = exthdrs.ip6e_ip6;
488 		hdrsplit++;
489 	}
490 
491 	ip6 = mtod(m, struct ip6_hdr *);
492 
493 	/* adjust mbuf packet header length */
494 	m->m_pkthdr.len += optlen;
495 	plen = m->m_pkthdr.len - sizeof(*ip6);
496 
497 	/* If this is a jumbo payload, insert a jumbo payload option. */
498 	if (plen > IPV6_MAXPACKET) {
499 		if (!hdrsplit) {
500 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
501 				m = NULL;
502 				goto freehdrs;
503 			}
504 			m = exthdrs.ip6e_ip6;
505 			hdrsplit++;
506 		}
507 		/* adjust pointer */
508 		ip6 = mtod(m, struct ip6_hdr *);
509 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
510 			goto freehdrs;
511 		ip6->ip6_plen = 0;
512 	} else
513 		ip6->ip6_plen = htons(plen);
514 
515 	/*
516 	 * Concatenate headers and fill in next header fields.
517 	 * Here we have, on "m"
518 	 *	IPv6 payload
519 	 * and we insert headers accordingly.  Finally, we should be getting:
520 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
521 	 *
522 	 * during the header composing process, "m" points to IPv6 header.
523 	 * "mprev" points to an extension header prior to esp.
524 	 */
525 	u_char *nexthdrp = &ip6->ip6_nxt;
526 	mprev = m;
527 
528 	/*
529 	 * we treat dest2 specially.  this makes IPsec processing
530 	 * much easier.  the goal here is to make mprev point the
531 	 * mbuf prior to dest2.
532 	 *
533 	 * result: IPv6 dest2 payload
534 	 * m and mprev will point to IPv6 header.
535 	 */
536 	if (exthdrs.ip6e_dest2) {
537 		if (!hdrsplit)
538 			panic("assumption failed: hdr not split");
539 		exthdrs.ip6e_dest2->m_next = m->m_next;
540 		m->m_next = exthdrs.ip6e_dest2;
541 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
542 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
543 	}
544 
545 	/*
546 	 * result: IPv6 hbh dest1 rthdr dest2 payload
547 	 * m will point to IPv6 header.  mprev will point to the
548 	 * extension header prior to dest2 (rthdr in the above case).
549 	 */
550 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
551 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
552 		   IPPROTO_DSTOPTS);
553 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
554 		   IPPROTO_ROUTING);
555 
556 	/*
557 	 * If there is a routing header, discard the packet.
558 	 */
559 	if (exthdrs.ip6e_rthdr) {
560 		 error = EINVAL;
561 		 goto bad;
562 	}
563 
564 	/* Source address validation */
565 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
566 	    (flags & IPV6_UNSPECSRC) == 0) {
567 		error = EOPNOTSUPP;
568 		IP6STAT_INC(ip6s_badscope);
569 		goto bad;
570 	}
571 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
572 		error = EOPNOTSUPP;
573 		IP6STAT_INC(ip6s_badscope);
574 		goto bad;
575 	}
576 
577 	IP6STAT_INC(ip6s_localout);
578 
579 	/*
580 	 * Route packet.
581 	 */
582 	if (ro == NULL) {
583 		ro = &ip6route;
584 		bzero((caddr_t)ro, sizeof(*ro));
585 	}
586 	ro_pmtu = ro;
587 	if (opt && opt->ip6po_rthdr)
588 		ro = &opt->ip6po_route;
589 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
590 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
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 = EACCES;
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 	if (ro == &ip6route)
1192 		RO_RTFREE(ro);
1193 	return (error);
1194 
1195 freehdrs:
1196 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1197 	m_freem(exthdrs.ip6e_dest1);
1198 	m_freem(exthdrs.ip6e_rthdr);
1199 	m_freem(exthdrs.ip6e_dest2);
1200 	/* FALLTHROUGH */
1201 bad:
1202 	if (m)
1203 		m_freem(m);
1204 	goto done;
1205 }
1206 
1207 static int
1208 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1209 {
1210 	struct mbuf *m;
1211 
1212 	if (hlen > MCLBYTES)
1213 		return (ENOBUFS); /* XXX */
1214 
1215 	if (hlen > MLEN)
1216 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1217 	else
1218 		m = m_get(M_NOWAIT, MT_DATA);
1219 	if (m == NULL)
1220 		return (ENOBUFS);
1221 	m->m_len = hlen;
1222 	if (hdr)
1223 		bcopy(hdr, mtod(m, caddr_t), hlen);
1224 
1225 	*mp = m;
1226 	return (0);
1227 }
1228 
1229 /*
1230  * Insert jumbo payload option.
1231  */
1232 static int
1233 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1234 {
1235 	struct mbuf *mopt;
1236 	u_char *optbuf;
1237 	u_int32_t v;
1238 
1239 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1240 
1241 	/*
1242 	 * If there is no hop-by-hop options header, allocate new one.
1243 	 * If there is one but it doesn't have enough space to store the
1244 	 * jumbo payload option, allocate a cluster to store the whole options.
1245 	 * Otherwise, use it to store the options.
1246 	 */
1247 	if (exthdrs->ip6e_hbh == NULL) {
1248 		mopt = m_get(M_NOWAIT, MT_DATA);
1249 		if (mopt == NULL)
1250 			return (ENOBUFS);
1251 		mopt->m_len = JUMBOOPTLEN;
1252 		optbuf = mtod(mopt, u_char *);
1253 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1254 		exthdrs->ip6e_hbh = mopt;
1255 	} else {
1256 		struct ip6_hbh *hbh;
1257 
1258 		mopt = exthdrs->ip6e_hbh;
1259 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1260 			/*
1261 			 * XXX assumption:
1262 			 * - exthdrs->ip6e_hbh is not referenced from places
1263 			 *   other than exthdrs.
1264 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1265 			 */
1266 			int oldoptlen = mopt->m_len;
1267 			struct mbuf *n;
1268 
1269 			/*
1270 			 * XXX: give up if the whole (new) hbh header does
1271 			 * not fit even in an mbuf cluster.
1272 			 */
1273 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1274 				return (ENOBUFS);
1275 
1276 			/*
1277 			 * As a consequence, we must always prepare a cluster
1278 			 * at this point.
1279 			 */
1280 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1281 			if (n == NULL)
1282 				return (ENOBUFS);
1283 			n->m_len = oldoptlen + JUMBOOPTLEN;
1284 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1285 			    oldoptlen);
1286 			optbuf = mtod(n, caddr_t) + oldoptlen;
1287 			m_freem(mopt);
1288 			mopt = exthdrs->ip6e_hbh = n;
1289 		} else {
1290 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1291 			mopt->m_len += JUMBOOPTLEN;
1292 		}
1293 		optbuf[0] = IP6OPT_PADN;
1294 		optbuf[1] = 1;
1295 
1296 		/*
1297 		 * Adjust the header length according to the pad and
1298 		 * the jumbo payload option.
1299 		 */
1300 		hbh = mtod(mopt, struct ip6_hbh *);
1301 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1302 	}
1303 
1304 	/* fill in the option. */
1305 	optbuf[2] = IP6OPT_JUMBO;
1306 	optbuf[3] = 4;
1307 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1308 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1309 
1310 	/* finally, adjust the packet header length */
1311 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1312 
1313 	return (0);
1314 #undef JUMBOOPTLEN
1315 }
1316 
1317 /*
1318  * Insert fragment header and copy unfragmentable header portions.
1319  */
1320 static int
1321 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1322     struct ip6_frag **frghdrp)
1323 {
1324 	struct mbuf *n, *mlast;
1325 
1326 	if (hlen > sizeof(struct ip6_hdr)) {
1327 		n = m_copym(m0, sizeof(struct ip6_hdr),
1328 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1329 		if (n == NULL)
1330 			return (ENOBUFS);
1331 		m->m_next = n;
1332 	} else
1333 		n = m;
1334 
1335 	/* Search for the last mbuf of unfragmentable part. */
1336 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1337 		;
1338 
1339 	if (M_WRITABLE(mlast) &&
1340 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1341 		/* use the trailing space of the last mbuf for the fragment hdr */
1342 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1343 		    mlast->m_len);
1344 		mlast->m_len += sizeof(struct ip6_frag);
1345 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1346 	} else {
1347 		/* allocate a new mbuf for the fragment header */
1348 		struct mbuf *mfrg;
1349 
1350 		mfrg = m_get(M_NOWAIT, MT_DATA);
1351 		if (mfrg == NULL)
1352 			return (ENOBUFS);
1353 		mfrg->m_len = sizeof(struct ip6_frag);
1354 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1355 		mlast->m_next = mfrg;
1356 	}
1357 
1358 	return (0);
1359 }
1360 
1361 /*
1362  * Calculates IPv6 path mtu for destination @dst.
1363  * Resulting MTU is stored in @mtup.
1364  *
1365  * Returns 0 on success.
1366  */
1367 static int
1368 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1369 {
1370 	struct nhop6_extended nh6;
1371 	struct in6_addr kdst;
1372 	uint32_t scopeid;
1373 	struct ifnet *ifp;
1374 	u_long mtu;
1375 	int error;
1376 
1377 	in6_splitscope(dst, &kdst, &scopeid);
1378 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1379 		return (EHOSTUNREACH);
1380 
1381 	ifp = nh6.nh_ifp;
1382 	mtu = nh6.nh_mtu;
1383 
1384 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1385 	fib6_free_nh_ext(fibnum, &nh6);
1386 
1387 	return (error);
1388 }
1389 
1390 /*
1391  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1392  * and cached data in @ro_pmtu.
1393  * MTU from (successful) route lookup is saved (along with dst)
1394  * inside @ro_pmtu to avoid subsequent route lookups after packet
1395  * filter processing.
1396  *
1397  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1398  * Returns 0 on success.
1399  */
1400 static int
1401 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1402     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1403     int *alwaysfragp, u_int fibnum, u_int proto)
1404 {
1405 	struct nhop6_basic nh6;
1406 	struct in6_addr kdst;
1407 	uint32_t scopeid;
1408 	struct sockaddr_in6 *sa6_dst;
1409 	u_long mtu;
1410 
1411 	mtu = 0;
1412 	if (do_lookup) {
1413 
1414 		/*
1415 		 * Here ro_pmtu has final destination address, while
1416 		 * ro might represent immediate destination.
1417 		 * Use ro_pmtu destination since mtu might differ.
1418 		 */
1419 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1420 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1421 			ro_pmtu->ro_mtu = 0;
1422 
1423 		if (ro_pmtu->ro_mtu == 0) {
1424 			bzero(sa6_dst, sizeof(*sa6_dst));
1425 			sa6_dst->sin6_family = AF_INET6;
1426 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1427 			sa6_dst->sin6_addr = *dst;
1428 
1429 			in6_splitscope(dst, &kdst, &scopeid);
1430 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1431 			    &nh6) == 0)
1432 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1433 		}
1434 
1435 		mtu = ro_pmtu->ro_mtu;
1436 	}
1437 
1438 	if (ro_pmtu->ro_rt)
1439 		mtu = ro_pmtu->ro_rt->rt_mtu;
1440 
1441 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1442 }
1443 
1444 /*
1445  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1446  * hostcache data for @dst.
1447  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1448  *
1449  * Returns 0 on success.
1450  */
1451 static int
1452 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1453     u_long *mtup, int *alwaysfragp, u_int proto)
1454 {
1455 	u_long mtu = 0;
1456 	int alwaysfrag = 0;
1457 	int error = 0;
1458 
1459 	if (rt_mtu > 0) {
1460 		u_int32_t ifmtu;
1461 		struct in_conninfo inc;
1462 
1463 		bzero(&inc, sizeof(inc));
1464 		inc.inc_flags |= INC_ISIPV6;
1465 		inc.inc6_faddr = *dst;
1466 
1467 		ifmtu = IN6_LINKMTU(ifp);
1468 
1469 		/* TCP is known to react to pmtu changes so skip hc */
1470 		if (proto != IPPROTO_TCP)
1471 			mtu = tcp_hc_getmtu(&inc);
1472 
1473 		if (mtu)
1474 			mtu = min(mtu, rt_mtu);
1475 		else
1476 			mtu = rt_mtu;
1477 		if (mtu == 0)
1478 			mtu = ifmtu;
1479 		else if (mtu < IPV6_MMTU) {
1480 			/*
1481 			 * RFC2460 section 5, last paragraph:
1482 			 * if we record ICMPv6 too big message with
1483 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1484 			 * or smaller, with framgent header attached.
1485 			 * (fragment header is needed regardless from the
1486 			 * packet size, for translators to identify packets)
1487 			 */
1488 			alwaysfrag = 1;
1489 			mtu = IPV6_MMTU;
1490 		}
1491 	} else if (ifp) {
1492 		mtu = IN6_LINKMTU(ifp);
1493 	} else
1494 		error = EHOSTUNREACH; /* XXX */
1495 
1496 	*mtup = mtu;
1497 	if (alwaysfragp)
1498 		*alwaysfragp = alwaysfrag;
1499 	return (error);
1500 }
1501 
1502 /*
1503  * IP6 socket option processing.
1504  */
1505 int
1506 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1507 {
1508 	int optdatalen, uproto;
1509 	void *optdata;
1510 	struct inpcb *inp = sotoinpcb(so);
1511 	int error, optval;
1512 	int level, op, optname;
1513 	int optlen;
1514 	struct thread *td;
1515 #ifdef	RSS
1516 	uint32_t rss_bucket;
1517 	int retval;
1518 #endif
1519 
1520 /*
1521  * Don't use more than a quarter of mbuf clusters.  N.B.:
1522  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1523  * on LP64 architectures, so cast to u_long to avoid undefined
1524  * behavior.  ILP32 architectures cannot have nmbclusters
1525  * large enough to overflow for other reasons.
1526  */
1527 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1528 
1529 	level = sopt->sopt_level;
1530 	op = sopt->sopt_dir;
1531 	optname = sopt->sopt_name;
1532 	optlen = sopt->sopt_valsize;
1533 	td = sopt->sopt_td;
1534 	error = 0;
1535 	optval = 0;
1536 	uproto = (int)so->so_proto->pr_protocol;
1537 
1538 	if (level != IPPROTO_IPV6) {
1539 		error = EINVAL;
1540 
1541 		if (sopt->sopt_level == SOL_SOCKET &&
1542 		    sopt->sopt_dir == SOPT_SET) {
1543 			switch (sopt->sopt_name) {
1544 			case SO_REUSEADDR:
1545 				INP_WLOCK(inp);
1546 				if ((so->so_options & SO_REUSEADDR) != 0)
1547 					inp->inp_flags2 |= INP_REUSEADDR;
1548 				else
1549 					inp->inp_flags2 &= ~INP_REUSEADDR;
1550 				INP_WUNLOCK(inp);
1551 				error = 0;
1552 				break;
1553 			case SO_REUSEPORT:
1554 				INP_WLOCK(inp);
1555 				if ((so->so_options & SO_REUSEPORT) != 0)
1556 					inp->inp_flags2 |= INP_REUSEPORT;
1557 				else
1558 					inp->inp_flags2 &= ~INP_REUSEPORT;
1559 				INP_WUNLOCK(inp);
1560 				error = 0;
1561 				break;
1562 			case SO_REUSEPORT_LB:
1563 				INP_WLOCK(inp);
1564 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1565 					inp->inp_flags2 |= INP_REUSEPORT_LB;
1566 				else
1567 					inp->inp_flags2 &= ~INP_REUSEPORT_LB;
1568 				INP_WUNLOCK(inp);
1569 				error = 0;
1570 				break;
1571 			case SO_SETFIB:
1572 				INP_WLOCK(inp);
1573 				inp->inp_inc.inc_fibnum = so->so_fibnum;
1574 				INP_WUNLOCK(inp);
1575 				error = 0;
1576 				break;
1577 			case SO_MAX_PACING_RATE:
1578 #ifdef RATELIMIT
1579 				INP_WLOCK(inp);
1580 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1581 				INP_WUNLOCK(inp);
1582 				error = 0;
1583 #else
1584 				error = EOPNOTSUPP;
1585 #endif
1586 				break;
1587 			default:
1588 				break;
1589 			}
1590 		}
1591 	} else {		/* level == IPPROTO_IPV6 */
1592 		switch (op) {
1593 
1594 		case SOPT_SET:
1595 			switch (optname) {
1596 			case IPV6_2292PKTOPTIONS:
1597 #ifdef IPV6_PKTOPTIONS
1598 			case IPV6_PKTOPTIONS:
1599 #endif
1600 			{
1601 				struct mbuf *m;
1602 
1603 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1604 					printf("ip6_ctloutput: mbuf limit hit\n");
1605 					error = ENOBUFS;
1606 					break;
1607 				}
1608 
1609 				error = soopt_getm(sopt, &m); /* XXX */
1610 				if (error != 0)
1611 					break;
1612 				error = soopt_mcopyin(sopt, m); /* XXX */
1613 				if (error != 0)
1614 					break;
1615 				error = ip6_pcbopts(&inp->in6p_outputopts,
1616 						    m, so, sopt);
1617 				m_freem(m); /* XXX */
1618 				break;
1619 			}
1620 
1621 			/*
1622 			 * Use of some Hop-by-Hop options or some
1623 			 * Destination options, might require special
1624 			 * privilege.  That is, normal applications
1625 			 * (without special privilege) might be forbidden
1626 			 * from setting certain options in outgoing packets,
1627 			 * and might never see certain options in received
1628 			 * packets. [RFC 2292 Section 6]
1629 			 * KAME specific note:
1630 			 *  KAME prevents non-privileged users from sending or
1631 			 *  receiving ANY hbh/dst options in order to avoid
1632 			 *  overhead of parsing options in the kernel.
1633 			 */
1634 			case IPV6_RECVHOPOPTS:
1635 			case IPV6_RECVDSTOPTS:
1636 			case IPV6_RECVRTHDRDSTOPTS:
1637 				if (td != NULL) {
1638 					error = priv_check(td,
1639 					    PRIV_NETINET_SETHDROPTS);
1640 					if (error)
1641 						break;
1642 				}
1643 				/* FALLTHROUGH */
1644 			case IPV6_UNICAST_HOPS:
1645 			case IPV6_HOPLIMIT:
1646 
1647 			case IPV6_RECVPKTINFO:
1648 			case IPV6_RECVHOPLIMIT:
1649 			case IPV6_RECVRTHDR:
1650 			case IPV6_RECVPATHMTU:
1651 			case IPV6_RECVTCLASS:
1652 			case IPV6_RECVFLOWID:
1653 #ifdef	RSS
1654 			case IPV6_RECVRSSBUCKETID:
1655 #endif
1656 			case IPV6_V6ONLY:
1657 			case IPV6_AUTOFLOWLABEL:
1658 			case IPV6_ORIGDSTADDR:
1659 			case IPV6_BINDANY:
1660 			case IPV6_BINDMULTI:
1661 #ifdef	RSS
1662 			case IPV6_RSS_LISTEN_BUCKET:
1663 #endif
1664 				if (optname == IPV6_BINDANY && td != NULL) {
1665 					error = priv_check(td,
1666 					    PRIV_NETINET_BINDANY);
1667 					if (error)
1668 						break;
1669 				}
1670 
1671 				if (optlen != sizeof(int)) {
1672 					error = EINVAL;
1673 					break;
1674 				}
1675 				error = sooptcopyin(sopt, &optval,
1676 					sizeof optval, sizeof optval);
1677 				if (error)
1678 					break;
1679 				switch (optname) {
1680 
1681 				case IPV6_UNICAST_HOPS:
1682 					if (optval < -1 || optval >= 256)
1683 						error = EINVAL;
1684 					else {
1685 						/* -1 = kernel default */
1686 						inp->in6p_hops = optval;
1687 						if ((inp->inp_vflag &
1688 						     INP_IPV4) != 0)
1689 							inp->inp_ip_ttl = optval;
1690 					}
1691 					break;
1692 #define OPTSET(bit) \
1693 do { \
1694 	INP_WLOCK(inp); \
1695 	if (optval) \
1696 		inp->inp_flags |= (bit); \
1697 	else \
1698 		inp->inp_flags &= ~(bit); \
1699 	INP_WUNLOCK(inp); \
1700 } while (/*CONSTCOND*/ 0)
1701 #define OPTSET2292(bit) \
1702 do { \
1703 	INP_WLOCK(inp); \
1704 	inp->inp_flags |= IN6P_RFC2292; \
1705 	if (optval) \
1706 		inp->inp_flags |= (bit); \
1707 	else \
1708 		inp->inp_flags &= ~(bit); \
1709 	INP_WUNLOCK(inp); \
1710 } while (/*CONSTCOND*/ 0)
1711 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1712 
1713 #define OPTSET2_N(bit, val) do {					\
1714 	if (val)							\
1715 		inp->inp_flags2 |= bit;					\
1716 	else								\
1717 		inp->inp_flags2 &= ~bit;				\
1718 } while (0)
1719 #define OPTSET2(bit, val) do {						\
1720 	INP_WLOCK(inp);							\
1721 	OPTSET2_N(bit, val);						\
1722 	INP_WUNLOCK(inp);						\
1723 } while (0)
1724 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1725 #define OPTSET2292_EXCLUSIVE(bit)					\
1726 do {									\
1727 	INP_WLOCK(inp);							\
1728 	if (OPTBIT(IN6P_RFC2292)) {					\
1729 		error = EINVAL;						\
1730 	} else {							\
1731 		if (optval)						\
1732 			inp->inp_flags |= (bit);			\
1733 		else							\
1734 			inp->inp_flags &= ~(bit);			\
1735 	}								\
1736 	INP_WUNLOCK(inp);						\
1737 } while (/*CONSTCOND*/ 0)
1738 
1739 				case IPV6_RECVPKTINFO:
1740 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1741 					break;
1742 
1743 				case IPV6_HOPLIMIT:
1744 				{
1745 					struct ip6_pktopts **optp;
1746 
1747 					/* cannot mix with RFC2292 */
1748 					if (OPTBIT(IN6P_RFC2292)) {
1749 						error = EINVAL;
1750 						break;
1751 					}
1752 					INP_WLOCK(inp);
1753 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1754 						INP_WUNLOCK(inp);
1755 						return (ECONNRESET);
1756 					}
1757 					optp = &inp->in6p_outputopts;
1758 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1759 					    (u_char *)&optval, sizeof(optval),
1760 					    optp, (td != NULL) ? td->td_ucred :
1761 					    NULL, uproto);
1762 					INP_WUNLOCK(inp);
1763 					break;
1764 				}
1765 
1766 				case IPV6_RECVHOPLIMIT:
1767 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1768 					break;
1769 
1770 				case IPV6_RECVHOPOPTS:
1771 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1772 					break;
1773 
1774 				case IPV6_RECVDSTOPTS:
1775 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1776 					break;
1777 
1778 				case IPV6_RECVRTHDRDSTOPTS:
1779 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1780 					break;
1781 
1782 				case IPV6_RECVRTHDR:
1783 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1784 					break;
1785 
1786 				case IPV6_RECVPATHMTU:
1787 					/*
1788 					 * We ignore this option for TCP
1789 					 * sockets.
1790 					 * (RFC3542 leaves this case
1791 					 * unspecified.)
1792 					 */
1793 					if (uproto != IPPROTO_TCP)
1794 						OPTSET(IN6P_MTU);
1795 					break;
1796 
1797 				case IPV6_RECVFLOWID:
1798 					OPTSET2(INP_RECVFLOWID, optval);
1799 					break;
1800 
1801 #ifdef	RSS
1802 				case IPV6_RECVRSSBUCKETID:
1803 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1804 					break;
1805 #endif
1806 
1807 				case IPV6_V6ONLY:
1808 					INP_WLOCK(inp);
1809 					if (inp->inp_lport ||
1810 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1811 						/*
1812 						 * The socket is already bound.
1813 						 */
1814 						INP_WUNLOCK(inp);
1815 						error = EINVAL;
1816 						break;
1817 					}
1818 					if (optval) {
1819 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1820 						inp->inp_vflag &= ~INP_IPV4;
1821 					} else {
1822 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1823 						inp->inp_vflag |= INP_IPV4;
1824 					}
1825 					INP_WUNLOCK(inp);
1826 					break;
1827 				case IPV6_RECVTCLASS:
1828 					/* cannot mix with RFC2292 XXX */
1829 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1830 					break;
1831 				case IPV6_AUTOFLOWLABEL:
1832 					OPTSET(IN6P_AUTOFLOWLABEL);
1833 					break;
1834 
1835 				case IPV6_ORIGDSTADDR:
1836 					OPTSET2(INP_ORIGDSTADDR, optval);
1837 					break;
1838 				case IPV6_BINDANY:
1839 					OPTSET(INP_BINDANY);
1840 					break;
1841 
1842 				case IPV6_BINDMULTI:
1843 					OPTSET2(INP_BINDMULTI, optval);
1844 					break;
1845 #ifdef	RSS
1846 				case IPV6_RSS_LISTEN_BUCKET:
1847 					if ((optval >= 0) &&
1848 					    (optval < rss_getnumbuckets())) {
1849 						INP_WLOCK(inp);
1850 						inp->inp_rss_listen_bucket = optval;
1851 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1852 						INP_WUNLOCK(inp);
1853 					} else {
1854 						error = EINVAL;
1855 					}
1856 					break;
1857 #endif
1858 				}
1859 				break;
1860 
1861 			case IPV6_TCLASS:
1862 			case IPV6_DONTFRAG:
1863 			case IPV6_USE_MIN_MTU:
1864 			case IPV6_PREFER_TEMPADDR:
1865 				if (optlen != sizeof(optval)) {
1866 					error = EINVAL;
1867 					break;
1868 				}
1869 				error = sooptcopyin(sopt, &optval,
1870 					sizeof optval, sizeof optval);
1871 				if (error)
1872 					break;
1873 				{
1874 					struct ip6_pktopts **optp;
1875 					INP_WLOCK(inp);
1876 					if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1877 						INP_WUNLOCK(inp);
1878 						return (ECONNRESET);
1879 					}
1880 					optp = &inp->in6p_outputopts;
1881 					error = ip6_pcbopt(optname,
1882 					    (u_char *)&optval, sizeof(optval),
1883 					    optp, (td != NULL) ? td->td_ucred :
1884 					    NULL, uproto);
1885 					INP_WUNLOCK(inp);
1886 					break;
1887 				}
1888 
1889 			case IPV6_2292PKTINFO:
1890 			case IPV6_2292HOPLIMIT:
1891 			case IPV6_2292HOPOPTS:
1892 			case IPV6_2292DSTOPTS:
1893 			case IPV6_2292RTHDR:
1894 				/* RFC 2292 */
1895 				if (optlen != sizeof(int)) {
1896 					error = EINVAL;
1897 					break;
1898 				}
1899 				error = sooptcopyin(sopt, &optval,
1900 					sizeof optval, sizeof optval);
1901 				if (error)
1902 					break;
1903 				switch (optname) {
1904 				case IPV6_2292PKTINFO:
1905 					OPTSET2292(IN6P_PKTINFO);
1906 					break;
1907 				case IPV6_2292HOPLIMIT:
1908 					OPTSET2292(IN6P_HOPLIMIT);
1909 					break;
1910 				case IPV6_2292HOPOPTS:
1911 					/*
1912 					 * Check super-user privilege.
1913 					 * See comments for IPV6_RECVHOPOPTS.
1914 					 */
1915 					if (td != NULL) {
1916 						error = priv_check(td,
1917 						    PRIV_NETINET_SETHDROPTS);
1918 						if (error)
1919 							return (error);
1920 					}
1921 					OPTSET2292(IN6P_HOPOPTS);
1922 					break;
1923 				case IPV6_2292DSTOPTS:
1924 					if (td != NULL) {
1925 						error = priv_check(td,
1926 						    PRIV_NETINET_SETHDROPTS);
1927 						if (error)
1928 							return (error);
1929 					}
1930 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1931 					break;
1932 				case IPV6_2292RTHDR:
1933 					OPTSET2292(IN6P_RTHDR);
1934 					break;
1935 				}
1936 				break;
1937 			case IPV6_PKTINFO:
1938 			case IPV6_HOPOPTS:
1939 			case IPV6_RTHDR:
1940 			case IPV6_DSTOPTS:
1941 			case IPV6_RTHDRDSTOPTS:
1942 			case IPV6_NEXTHOP:
1943 			{
1944 				/* new advanced API (RFC3542) */
1945 				u_char *optbuf;
1946 				u_char optbuf_storage[MCLBYTES];
1947 				int optlen;
1948 				struct ip6_pktopts **optp;
1949 
1950 				/* cannot mix with RFC2292 */
1951 				if (OPTBIT(IN6P_RFC2292)) {
1952 					error = EINVAL;
1953 					break;
1954 				}
1955 
1956 				/*
1957 				 * We only ensure valsize is not too large
1958 				 * here.  Further validation will be done
1959 				 * later.
1960 				 */
1961 				error = sooptcopyin(sopt, optbuf_storage,
1962 				    sizeof(optbuf_storage), 0);
1963 				if (error)
1964 					break;
1965 				optlen = sopt->sopt_valsize;
1966 				optbuf = optbuf_storage;
1967 				INP_WLOCK(inp);
1968 				if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1969 					INP_WUNLOCK(inp);
1970 					return (ECONNRESET);
1971 				}
1972 				optp = &inp->in6p_outputopts;
1973 				error = ip6_pcbopt(optname, optbuf, optlen,
1974 				    optp, (td != NULL) ? td->td_ucred : NULL,
1975 				    uproto);
1976 				INP_WUNLOCK(inp);
1977 				break;
1978 			}
1979 #undef OPTSET
1980 
1981 			case IPV6_MULTICAST_IF:
1982 			case IPV6_MULTICAST_HOPS:
1983 			case IPV6_MULTICAST_LOOP:
1984 			case IPV6_JOIN_GROUP:
1985 			case IPV6_LEAVE_GROUP:
1986 			case IPV6_MSFILTER:
1987 			case MCAST_BLOCK_SOURCE:
1988 			case MCAST_UNBLOCK_SOURCE:
1989 			case MCAST_JOIN_GROUP:
1990 			case MCAST_LEAVE_GROUP:
1991 			case MCAST_JOIN_SOURCE_GROUP:
1992 			case MCAST_LEAVE_SOURCE_GROUP:
1993 				error = ip6_setmoptions(inp, sopt);
1994 				break;
1995 
1996 			case IPV6_PORTRANGE:
1997 				error = sooptcopyin(sopt, &optval,
1998 				    sizeof optval, sizeof optval);
1999 				if (error)
2000 					break;
2001 
2002 				INP_WLOCK(inp);
2003 				switch (optval) {
2004 				case IPV6_PORTRANGE_DEFAULT:
2005 					inp->inp_flags &= ~(INP_LOWPORT);
2006 					inp->inp_flags &= ~(INP_HIGHPORT);
2007 					break;
2008 
2009 				case IPV6_PORTRANGE_HIGH:
2010 					inp->inp_flags &= ~(INP_LOWPORT);
2011 					inp->inp_flags |= INP_HIGHPORT;
2012 					break;
2013 
2014 				case IPV6_PORTRANGE_LOW:
2015 					inp->inp_flags &= ~(INP_HIGHPORT);
2016 					inp->inp_flags |= INP_LOWPORT;
2017 					break;
2018 
2019 				default:
2020 					error = EINVAL;
2021 					break;
2022 				}
2023 				INP_WUNLOCK(inp);
2024 				break;
2025 
2026 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2027 			case IPV6_IPSEC_POLICY:
2028 				if (IPSEC_ENABLED(ipv6)) {
2029 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2030 					break;
2031 				}
2032 				/* FALLTHROUGH */
2033 #endif /* IPSEC */
2034 
2035 			default:
2036 				error = ENOPROTOOPT;
2037 				break;
2038 			}
2039 			break;
2040 
2041 		case SOPT_GET:
2042 			switch (optname) {
2043 
2044 			case IPV6_2292PKTOPTIONS:
2045 #ifdef IPV6_PKTOPTIONS
2046 			case IPV6_PKTOPTIONS:
2047 #endif
2048 				/*
2049 				 * RFC3542 (effectively) deprecated the
2050 				 * semantics of the 2292-style pktoptions.
2051 				 * Since it was not reliable in nature (i.e.,
2052 				 * applications had to expect the lack of some
2053 				 * information after all), it would make sense
2054 				 * to simplify this part by always returning
2055 				 * empty data.
2056 				 */
2057 				sopt->sopt_valsize = 0;
2058 				break;
2059 
2060 			case IPV6_RECVHOPOPTS:
2061 			case IPV6_RECVDSTOPTS:
2062 			case IPV6_RECVRTHDRDSTOPTS:
2063 			case IPV6_UNICAST_HOPS:
2064 			case IPV6_RECVPKTINFO:
2065 			case IPV6_RECVHOPLIMIT:
2066 			case IPV6_RECVRTHDR:
2067 			case IPV6_RECVPATHMTU:
2068 
2069 			case IPV6_V6ONLY:
2070 			case IPV6_PORTRANGE:
2071 			case IPV6_RECVTCLASS:
2072 			case IPV6_AUTOFLOWLABEL:
2073 			case IPV6_BINDANY:
2074 			case IPV6_FLOWID:
2075 			case IPV6_FLOWTYPE:
2076 			case IPV6_RECVFLOWID:
2077 #ifdef	RSS
2078 			case IPV6_RSSBUCKETID:
2079 			case IPV6_RECVRSSBUCKETID:
2080 #endif
2081 			case IPV6_BINDMULTI:
2082 				switch (optname) {
2083 
2084 				case IPV6_RECVHOPOPTS:
2085 					optval = OPTBIT(IN6P_HOPOPTS);
2086 					break;
2087 
2088 				case IPV6_RECVDSTOPTS:
2089 					optval = OPTBIT(IN6P_DSTOPTS);
2090 					break;
2091 
2092 				case IPV6_RECVRTHDRDSTOPTS:
2093 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2094 					break;
2095 
2096 				case IPV6_UNICAST_HOPS:
2097 					optval = inp->in6p_hops;
2098 					break;
2099 
2100 				case IPV6_RECVPKTINFO:
2101 					optval = OPTBIT(IN6P_PKTINFO);
2102 					break;
2103 
2104 				case IPV6_RECVHOPLIMIT:
2105 					optval = OPTBIT(IN6P_HOPLIMIT);
2106 					break;
2107 
2108 				case IPV6_RECVRTHDR:
2109 					optval = OPTBIT(IN6P_RTHDR);
2110 					break;
2111 
2112 				case IPV6_RECVPATHMTU:
2113 					optval = OPTBIT(IN6P_MTU);
2114 					break;
2115 
2116 				case IPV6_V6ONLY:
2117 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2118 					break;
2119 
2120 				case IPV6_PORTRANGE:
2121 				    {
2122 					int flags;
2123 					flags = inp->inp_flags;
2124 					if (flags & INP_HIGHPORT)
2125 						optval = IPV6_PORTRANGE_HIGH;
2126 					else if (flags & INP_LOWPORT)
2127 						optval = IPV6_PORTRANGE_LOW;
2128 					else
2129 						optval = 0;
2130 					break;
2131 				    }
2132 				case IPV6_RECVTCLASS:
2133 					optval = OPTBIT(IN6P_TCLASS);
2134 					break;
2135 
2136 				case IPV6_AUTOFLOWLABEL:
2137 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2138 					break;
2139 
2140 				case IPV6_ORIGDSTADDR:
2141 					optval = OPTBIT2(INP_ORIGDSTADDR);
2142 					break;
2143 
2144 				case IPV6_BINDANY:
2145 					optval = OPTBIT(INP_BINDANY);
2146 					break;
2147 
2148 				case IPV6_FLOWID:
2149 					optval = inp->inp_flowid;
2150 					break;
2151 
2152 				case IPV6_FLOWTYPE:
2153 					optval = inp->inp_flowtype;
2154 					break;
2155 
2156 				case IPV6_RECVFLOWID:
2157 					optval = OPTBIT2(INP_RECVFLOWID);
2158 					break;
2159 #ifdef	RSS
2160 				case IPV6_RSSBUCKETID:
2161 					retval =
2162 					    rss_hash2bucket(inp->inp_flowid,
2163 					    inp->inp_flowtype,
2164 					    &rss_bucket);
2165 					if (retval == 0)
2166 						optval = rss_bucket;
2167 					else
2168 						error = EINVAL;
2169 					break;
2170 
2171 				case IPV6_RECVRSSBUCKETID:
2172 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2173 					break;
2174 #endif
2175 
2176 				case IPV6_BINDMULTI:
2177 					optval = OPTBIT2(INP_BINDMULTI);
2178 					break;
2179 
2180 				}
2181 				if (error)
2182 					break;
2183 				error = sooptcopyout(sopt, &optval,
2184 					sizeof optval);
2185 				break;
2186 
2187 			case IPV6_PATHMTU:
2188 			{
2189 				u_long pmtu = 0;
2190 				struct ip6_mtuinfo mtuinfo;
2191 				struct in6_addr addr;
2192 
2193 				if (!(so->so_state & SS_ISCONNECTED))
2194 					return (ENOTCONN);
2195 				/*
2196 				 * XXX: we dot not consider the case of source
2197 				 * routing, or optional information to specify
2198 				 * the outgoing interface.
2199 				 * Copy faddr out of inp to avoid holding lock
2200 				 * on inp during route lookup.
2201 				 */
2202 				INP_RLOCK(inp);
2203 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2204 				INP_RUNLOCK(inp);
2205 				error = ip6_getpmtu_ctl(so->so_fibnum,
2206 				    &addr, &pmtu);
2207 				if (error)
2208 					break;
2209 				if (pmtu > IPV6_MAXPACKET)
2210 					pmtu = IPV6_MAXPACKET;
2211 
2212 				bzero(&mtuinfo, sizeof(mtuinfo));
2213 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2214 				optdata = (void *)&mtuinfo;
2215 				optdatalen = sizeof(mtuinfo);
2216 				error = sooptcopyout(sopt, optdata,
2217 				    optdatalen);
2218 				break;
2219 			}
2220 
2221 			case IPV6_2292PKTINFO:
2222 			case IPV6_2292HOPLIMIT:
2223 			case IPV6_2292HOPOPTS:
2224 			case IPV6_2292RTHDR:
2225 			case IPV6_2292DSTOPTS:
2226 				switch (optname) {
2227 				case IPV6_2292PKTINFO:
2228 					optval = OPTBIT(IN6P_PKTINFO);
2229 					break;
2230 				case IPV6_2292HOPLIMIT:
2231 					optval = OPTBIT(IN6P_HOPLIMIT);
2232 					break;
2233 				case IPV6_2292HOPOPTS:
2234 					optval = OPTBIT(IN6P_HOPOPTS);
2235 					break;
2236 				case IPV6_2292RTHDR:
2237 					optval = OPTBIT(IN6P_RTHDR);
2238 					break;
2239 				case IPV6_2292DSTOPTS:
2240 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2241 					break;
2242 				}
2243 				error = sooptcopyout(sopt, &optval,
2244 				    sizeof optval);
2245 				break;
2246 			case IPV6_PKTINFO:
2247 			case IPV6_HOPOPTS:
2248 			case IPV6_RTHDR:
2249 			case IPV6_DSTOPTS:
2250 			case IPV6_RTHDRDSTOPTS:
2251 			case IPV6_NEXTHOP:
2252 			case IPV6_TCLASS:
2253 			case IPV6_DONTFRAG:
2254 			case IPV6_USE_MIN_MTU:
2255 			case IPV6_PREFER_TEMPADDR:
2256 				error = ip6_getpcbopt(inp, optname, sopt);
2257 				break;
2258 
2259 			case IPV6_MULTICAST_IF:
2260 			case IPV6_MULTICAST_HOPS:
2261 			case IPV6_MULTICAST_LOOP:
2262 			case IPV6_MSFILTER:
2263 				error = ip6_getmoptions(inp, sopt);
2264 				break;
2265 
2266 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2267 			case IPV6_IPSEC_POLICY:
2268 				if (IPSEC_ENABLED(ipv6)) {
2269 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2270 					break;
2271 				}
2272 				/* FALLTHROUGH */
2273 #endif /* IPSEC */
2274 			default:
2275 				error = ENOPROTOOPT;
2276 				break;
2277 			}
2278 			break;
2279 		}
2280 	}
2281 	return (error);
2282 }
2283 
2284 int
2285 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2286 {
2287 	int error = 0, optval, optlen;
2288 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2289 	struct inpcb *inp = sotoinpcb(so);
2290 	int level, op, optname;
2291 
2292 	level = sopt->sopt_level;
2293 	op = sopt->sopt_dir;
2294 	optname = sopt->sopt_name;
2295 	optlen = sopt->sopt_valsize;
2296 
2297 	if (level != IPPROTO_IPV6) {
2298 		return (EINVAL);
2299 	}
2300 
2301 	switch (optname) {
2302 	case IPV6_CHECKSUM:
2303 		/*
2304 		 * For ICMPv6 sockets, no modification allowed for checksum
2305 		 * offset, permit "no change" values to help existing apps.
2306 		 *
2307 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2308 		 * for an ICMPv6 socket will fail."
2309 		 * The current behavior does not meet RFC3542.
2310 		 */
2311 		switch (op) {
2312 		case SOPT_SET:
2313 			if (optlen != sizeof(int)) {
2314 				error = EINVAL;
2315 				break;
2316 			}
2317 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2318 					    sizeof(optval));
2319 			if (error)
2320 				break;
2321 			if (optval < -1 || (optval % 2) != 0) {
2322 				/*
2323 				 * The API assumes non-negative even offset
2324 				 * values or -1 as a special value.
2325 				 */
2326 				error = EINVAL;
2327 			} else if (so->so_proto->pr_protocol ==
2328 			    IPPROTO_ICMPV6) {
2329 				if (optval != icmp6off)
2330 					error = EINVAL;
2331 			} else
2332 				inp->in6p_cksum = optval;
2333 			break;
2334 
2335 		case SOPT_GET:
2336 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2337 				optval = icmp6off;
2338 			else
2339 				optval = inp->in6p_cksum;
2340 
2341 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2342 			break;
2343 
2344 		default:
2345 			error = EINVAL;
2346 			break;
2347 		}
2348 		break;
2349 
2350 	default:
2351 		error = ENOPROTOOPT;
2352 		break;
2353 	}
2354 
2355 	return (error);
2356 }
2357 
2358 /*
2359  * Set up IP6 options in pcb for insertion in output packets or
2360  * specifying behavior of outgoing packets.
2361  */
2362 static int
2363 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2364     struct socket *so, struct sockopt *sopt)
2365 {
2366 	struct ip6_pktopts *opt = *pktopt;
2367 	int error = 0;
2368 	struct thread *td = sopt->sopt_td;
2369 
2370 	/* turn off any old options. */
2371 	if (opt) {
2372 #ifdef DIAGNOSTIC
2373 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2374 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2375 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2376 			printf("ip6_pcbopts: all specified options are cleared.\n");
2377 #endif
2378 		ip6_clearpktopts(opt, -1);
2379 	} else
2380 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2381 	*pktopt = NULL;
2382 
2383 	if (!m || m->m_len == 0) {
2384 		/*
2385 		 * Only turning off any previous options, regardless of
2386 		 * whether the opt is just created or given.
2387 		 */
2388 		free(opt, M_IP6OPT);
2389 		return (0);
2390 	}
2391 
2392 	/*  set options specified by user. */
2393 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2394 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2395 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2396 		free(opt, M_IP6OPT);
2397 		return (error);
2398 	}
2399 	*pktopt = opt;
2400 	return (0);
2401 }
2402 
2403 /*
2404  * initialize ip6_pktopts.  beware that there are non-zero default values in
2405  * the struct.
2406  */
2407 void
2408 ip6_initpktopts(struct ip6_pktopts *opt)
2409 {
2410 
2411 	bzero(opt, sizeof(*opt));
2412 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2413 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2414 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2415 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2416 }
2417 
2418 static int
2419 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2420     struct ucred *cred, int uproto)
2421 {
2422 	struct ip6_pktopts *opt;
2423 
2424 	if (*pktopt == NULL) {
2425 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2426 		    M_NOWAIT);
2427 		if (*pktopt == NULL)
2428 			return (ENOBUFS);
2429 		ip6_initpktopts(*pktopt);
2430 	}
2431 	opt = *pktopt;
2432 
2433 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2434 }
2435 
2436 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2437 	if (pktopt && pktopt->field) {						\
2438 		INP_RUNLOCK(inp);						\
2439 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2440 		malloc_optdata = true;						\
2441 		INP_RLOCK(inp);							\
2442 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2443 			INP_RUNLOCK(inp);					\
2444 			free(optdata, M_TEMP);					\
2445 			return (ECONNRESET);					\
2446 		}								\
2447 		pktopt = inp->in6p_outputopts;					\
2448 		if (pktopt && pktopt->field) {					\
2449 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2450 			bcopy(&pktopt->field, optdata, optdatalen);		\
2451 		} else {							\
2452 			free(optdata, M_TEMP);					\
2453 			optdata = NULL;						\
2454 			malloc_optdata = false;					\
2455 		}								\
2456 	}									\
2457 } while(0)
2458 
2459 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2460 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2461 
2462 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2463 	pktopt->field->sa_len)
2464 
2465 static int
2466 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2467 {
2468 	void *optdata = NULL;
2469 	bool malloc_optdata = false;
2470 	int optdatalen = 0;
2471 	int error = 0;
2472 	struct in6_pktinfo null_pktinfo;
2473 	int deftclass = 0, on;
2474 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2475 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2476 	struct ip6_pktopts *pktopt;
2477 
2478 	INP_RLOCK(inp);
2479 	pktopt = inp->in6p_outputopts;
2480 
2481 	switch (optname) {
2482 	case IPV6_PKTINFO:
2483 		optdata = (void *)&null_pktinfo;
2484 		if (pktopt && pktopt->ip6po_pktinfo) {
2485 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2486 			    sizeof(null_pktinfo));
2487 			in6_clearscope(&null_pktinfo.ipi6_addr);
2488 		} else {
2489 			/* XXX: we don't have to do this every time... */
2490 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2491 		}
2492 		optdatalen = sizeof(struct in6_pktinfo);
2493 		break;
2494 	case IPV6_TCLASS:
2495 		if (pktopt && pktopt->ip6po_tclass >= 0)
2496 			deftclass = pktopt->ip6po_tclass;
2497 		optdata = (void *)&deftclass;
2498 		optdatalen = sizeof(int);
2499 		break;
2500 	case IPV6_HOPOPTS:
2501 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2502 		break;
2503 	case IPV6_RTHDR:
2504 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2505 		break;
2506 	case IPV6_RTHDRDSTOPTS:
2507 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2508 		break;
2509 	case IPV6_DSTOPTS:
2510 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2511 		break;
2512 	case IPV6_NEXTHOP:
2513 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2514 		break;
2515 	case IPV6_USE_MIN_MTU:
2516 		if (pktopt)
2517 			defminmtu = pktopt->ip6po_minmtu;
2518 		optdata = (void *)&defminmtu;
2519 		optdatalen = sizeof(int);
2520 		break;
2521 	case IPV6_DONTFRAG:
2522 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2523 			on = 1;
2524 		else
2525 			on = 0;
2526 		optdata = (void *)&on;
2527 		optdatalen = sizeof(on);
2528 		break;
2529 	case IPV6_PREFER_TEMPADDR:
2530 		if (pktopt)
2531 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2532 		optdata = (void *)&defpreftemp;
2533 		optdatalen = sizeof(int);
2534 		break;
2535 	default:		/* should not happen */
2536 #ifdef DIAGNOSTIC
2537 		panic("ip6_getpcbopt: unexpected option\n");
2538 #endif
2539 		INP_RUNLOCK(inp);
2540 		return (ENOPROTOOPT);
2541 	}
2542 	INP_RUNLOCK(inp);
2543 
2544 	error = sooptcopyout(sopt, optdata, optdatalen);
2545 	if (malloc_optdata)
2546 		free(optdata, M_TEMP);
2547 
2548 	return (error);
2549 }
2550 
2551 void
2552 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2553 {
2554 	if (pktopt == NULL)
2555 		return;
2556 
2557 	if (optname == -1 || optname == IPV6_PKTINFO) {
2558 		if (pktopt->ip6po_pktinfo)
2559 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2560 		pktopt->ip6po_pktinfo = NULL;
2561 	}
2562 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2563 		pktopt->ip6po_hlim = -1;
2564 	if (optname == -1 || optname == IPV6_TCLASS)
2565 		pktopt->ip6po_tclass = -1;
2566 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2567 		if (pktopt->ip6po_nextroute.ro_rt) {
2568 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2569 			pktopt->ip6po_nextroute.ro_rt = NULL;
2570 		}
2571 		if (pktopt->ip6po_nexthop)
2572 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2573 		pktopt->ip6po_nexthop = NULL;
2574 	}
2575 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2576 		if (pktopt->ip6po_hbh)
2577 			free(pktopt->ip6po_hbh, M_IP6OPT);
2578 		pktopt->ip6po_hbh = NULL;
2579 	}
2580 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2581 		if (pktopt->ip6po_dest1)
2582 			free(pktopt->ip6po_dest1, M_IP6OPT);
2583 		pktopt->ip6po_dest1 = NULL;
2584 	}
2585 	if (optname == -1 || optname == IPV6_RTHDR) {
2586 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2587 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2588 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2589 		if (pktopt->ip6po_route.ro_rt) {
2590 			RTFREE(pktopt->ip6po_route.ro_rt);
2591 			pktopt->ip6po_route.ro_rt = NULL;
2592 		}
2593 	}
2594 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2595 		if (pktopt->ip6po_dest2)
2596 			free(pktopt->ip6po_dest2, M_IP6OPT);
2597 		pktopt->ip6po_dest2 = NULL;
2598 	}
2599 }
2600 
2601 #define PKTOPT_EXTHDRCPY(type) \
2602 do {\
2603 	if (src->type) {\
2604 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2605 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2606 		if (dst->type == NULL)\
2607 			goto bad;\
2608 		bcopy(src->type, dst->type, hlen);\
2609 	}\
2610 } while (/*CONSTCOND*/ 0)
2611 
2612 static int
2613 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2614 {
2615 	if (dst == NULL || src == NULL)  {
2616 		printf("ip6_clearpktopts: invalid argument\n");
2617 		return (EINVAL);
2618 	}
2619 
2620 	dst->ip6po_hlim = src->ip6po_hlim;
2621 	dst->ip6po_tclass = src->ip6po_tclass;
2622 	dst->ip6po_flags = src->ip6po_flags;
2623 	dst->ip6po_minmtu = src->ip6po_minmtu;
2624 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2625 	if (src->ip6po_pktinfo) {
2626 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2627 		    M_IP6OPT, canwait);
2628 		if (dst->ip6po_pktinfo == NULL)
2629 			goto bad;
2630 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2631 	}
2632 	if (src->ip6po_nexthop) {
2633 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2634 		    M_IP6OPT, canwait);
2635 		if (dst->ip6po_nexthop == NULL)
2636 			goto bad;
2637 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2638 		    src->ip6po_nexthop->sa_len);
2639 	}
2640 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2641 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2642 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2643 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2644 	return (0);
2645 
2646   bad:
2647 	ip6_clearpktopts(dst, -1);
2648 	return (ENOBUFS);
2649 }
2650 #undef PKTOPT_EXTHDRCPY
2651 
2652 struct ip6_pktopts *
2653 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2654 {
2655 	int error;
2656 	struct ip6_pktopts *dst;
2657 
2658 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2659 	if (dst == NULL)
2660 		return (NULL);
2661 	ip6_initpktopts(dst);
2662 
2663 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2664 		free(dst, M_IP6OPT);
2665 		return (NULL);
2666 	}
2667 
2668 	return (dst);
2669 }
2670 
2671 void
2672 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2673 {
2674 	if (pktopt == NULL)
2675 		return;
2676 
2677 	ip6_clearpktopts(pktopt, -1);
2678 
2679 	free(pktopt, M_IP6OPT);
2680 }
2681 
2682 /*
2683  * Set IPv6 outgoing packet options based on advanced API.
2684  */
2685 int
2686 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2687     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2688 {
2689 	struct cmsghdr *cm = NULL;
2690 
2691 	if (control == NULL || opt == NULL)
2692 		return (EINVAL);
2693 
2694 	ip6_initpktopts(opt);
2695 	if (stickyopt) {
2696 		int error;
2697 
2698 		/*
2699 		 * If stickyopt is provided, make a local copy of the options
2700 		 * for this particular packet, then override them by ancillary
2701 		 * objects.
2702 		 * XXX: copypktopts() does not copy the cached route to a next
2703 		 * hop (if any).  This is not very good in terms of efficiency,
2704 		 * but we can allow this since this option should be rarely
2705 		 * used.
2706 		 */
2707 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2708 			return (error);
2709 	}
2710 
2711 	/*
2712 	 * XXX: Currently, we assume all the optional information is stored
2713 	 * in a single mbuf.
2714 	 */
2715 	if (control->m_next)
2716 		return (EINVAL);
2717 
2718 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2719 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2720 		int error;
2721 
2722 		if (control->m_len < CMSG_LEN(0))
2723 			return (EINVAL);
2724 
2725 		cm = mtod(control, struct cmsghdr *);
2726 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2727 			return (EINVAL);
2728 		if (cm->cmsg_level != IPPROTO_IPV6)
2729 			continue;
2730 
2731 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2732 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2733 		if (error)
2734 			return (error);
2735 	}
2736 
2737 	return (0);
2738 }
2739 
2740 /*
2741  * Set a particular packet option, as a sticky option or an ancillary data
2742  * item.  "len" can be 0 only when it's a sticky option.
2743  * We have 4 cases of combination of "sticky" and "cmsg":
2744  * "sticky=0, cmsg=0": impossible
2745  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2746  * "sticky=1, cmsg=0": RFC3542 socket option
2747  * "sticky=1, cmsg=1": RFC2292 socket option
2748  */
2749 static int
2750 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2751     struct ucred *cred, int sticky, int cmsg, int uproto)
2752 {
2753 	int minmtupolicy, preftemp;
2754 	int error;
2755 
2756 	if (!sticky && !cmsg) {
2757 #ifdef DIAGNOSTIC
2758 		printf("ip6_setpktopt: impossible case\n");
2759 #endif
2760 		return (EINVAL);
2761 	}
2762 
2763 	/*
2764 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2765 	 * not be specified in the context of RFC3542.  Conversely,
2766 	 * RFC3542 types should not be specified in the context of RFC2292.
2767 	 */
2768 	if (!cmsg) {
2769 		switch (optname) {
2770 		case IPV6_2292PKTINFO:
2771 		case IPV6_2292HOPLIMIT:
2772 		case IPV6_2292NEXTHOP:
2773 		case IPV6_2292HOPOPTS:
2774 		case IPV6_2292DSTOPTS:
2775 		case IPV6_2292RTHDR:
2776 		case IPV6_2292PKTOPTIONS:
2777 			return (ENOPROTOOPT);
2778 		}
2779 	}
2780 	if (sticky && cmsg) {
2781 		switch (optname) {
2782 		case IPV6_PKTINFO:
2783 		case IPV6_HOPLIMIT:
2784 		case IPV6_NEXTHOP:
2785 		case IPV6_HOPOPTS:
2786 		case IPV6_DSTOPTS:
2787 		case IPV6_RTHDRDSTOPTS:
2788 		case IPV6_RTHDR:
2789 		case IPV6_USE_MIN_MTU:
2790 		case IPV6_DONTFRAG:
2791 		case IPV6_TCLASS:
2792 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2793 			return (ENOPROTOOPT);
2794 		}
2795 	}
2796 
2797 	switch (optname) {
2798 	case IPV6_2292PKTINFO:
2799 	case IPV6_PKTINFO:
2800 	{
2801 		struct ifnet *ifp = NULL;
2802 		struct in6_pktinfo *pktinfo;
2803 
2804 		if (len != sizeof(struct in6_pktinfo))
2805 			return (EINVAL);
2806 
2807 		pktinfo = (struct in6_pktinfo *)buf;
2808 
2809 		/*
2810 		 * An application can clear any sticky IPV6_PKTINFO option by
2811 		 * doing a "regular" setsockopt with ipi6_addr being
2812 		 * in6addr_any and ipi6_ifindex being zero.
2813 		 * [RFC 3542, Section 6]
2814 		 */
2815 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2816 		    pktinfo->ipi6_ifindex == 0 &&
2817 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2818 			ip6_clearpktopts(opt, optname);
2819 			break;
2820 		}
2821 
2822 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2823 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2824 			return (EINVAL);
2825 		}
2826 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2827 			return (EINVAL);
2828 		/* validate the interface index if specified. */
2829 		if (pktinfo->ipi6_ifindex > V_if_index)
2830 			 return (ENXIO);
2831 		if (pktinfo->ipi6_ifindex) {
2832 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2833 			if (ifp == NULL)
2834 				return (ENXIO);
2835 		}
2836 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2837 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2838 			return (ENETDOWN);
2839 
2840 		if (ifp != NULL &&
2841 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2842 			struct in6_ifaddr *ia;
2843 
2844 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2845 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2846 			if (ia == NULL)
2847 				return (EADDRNOTAVAIL);
2848 			ifa_free(&ia->ia_ifa);
2849 		}
2850 		/*
2851 		 * We store the address anyway, and let in6_selectsrc()
2852 		 * validate the specified address.  This is because ipi6_addr
2853 		 * may not have enough information about its scope zone, and
2854 		 * we may need additional information (such as outgoing
2855 		 * interface or the scope zone of a destination address) to
2856 		 * disambiguate the scope.
2857 		 * XXX: the delay of the validation may confuse the
2858 		 * application when it is used as a sticky option.
2859 		 */
2860 		if (opt->ip6po_pktinfo == NULL) {
2861 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2862 			    M_IP6OPT, M_NOWAIT);
2863 			if (opt->ip6po_pktinfo == NULL)
2864 				return (ENOBUFS);
2865 		}
2866 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2867 		break;
2868 	}
2869 
2870 	case IPV6_2292HOPLIMIT:
2871 	case IPV6_HOPLIMIT:
2872 	{
2873 		int *hlimp;
2874 
2875 		/*
2876 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2877 		 * to simplify the ordering among hoplimit options.
2878 		 */
2879 		if (optname == IPV6_HOPLIMIT && sticky)
2880 			return (ENOPROTOOPT);
2881 
2882 		if (len != sizeof(int))
2883 			return (EINVAL);
2884 		hlimp = (int *)buf;
2885 		if (*hlimp < -1 || *hlimp > 255)
2886 			return (EINVAL);
2887 
2888 		opt->ip6po_hlim = *hlimp;
2889 		break;
2890 	}
2891 
2892 	case IPV6_TCLASS:
2893 	{
2894 		int tclass;
2895 
2896 		if (len != sizeof(int))
2897 			return (EINVAL);
2898 		tclass = *(int *)buf;
2899 		if (tclass < -1 || tclass > 255)
2900 			return (EINVAL);
2901 
2902 		opt->ip6po_tclass = tclass;
2903 		break;
2904 	}
2905 
2906 	case IPV6_2292NEXTHOP:
2907 	case IPV6_NEXTHOP:
2908 		if (cred != NULL) {
2909 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2910 			if (error)
2911 				return (error);
2912 		}
2913 
2914 		if (len == 0) {	/* just remove the option */
2915 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2916 			break;
2917 		}
2918 
2919 		/* check if cmsg_len is large enough for sa_len */
2920 		if (len < sizeof(struct sockaddr) || len < *buf)
2921 			return (EINVAL);
2922 
2923 		switch (((struct sockaddr *)buf)->sa_family) {
2924 		case AF_INET6:
2925 		{
2926 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2927 			int error;
2928 
2929 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2930 				return (EINVAL);
2931 
2932 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2933 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2934 				return (EINVAL);
2935 			}
2936 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2937 			    != 0) {
2938 				return (error);
2939 			}
2940 			break;
2941 		}
2942 		case AF_LINK:	/* should eventually be supported */
2943 		default:
2944 			return (EAFNOSUPPORT);
2945 		}
2946 
2947 		/* turn off the previous option, then set the new option. */
2948 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2949 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2950 		if (opt->ip6po_nexthop == NULL)
2951 			return (ENOBUFS);
2952 		bcopy(buf, opt->ip6po_nexthop, *buf);
2953 		break;
2954 
2955 	case IPV6_2292HOPOPTS:
2956 	case IPV6_HOPOPTS:
2957 	{
2958 		struct ip6_hbh *hbh;
2959 		int hbhlen;
2960 
2961 		/*
2962 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2963 		 * options, since per-option restriction has too much
2964 		 * overhead.
2965 		 */
2966 		if (cred != NULL) {
2967 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2968 			if (error)
2969 				return (error);
2970 		}
2971 
2972 		if (len == 0) {
2973 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2974 			break;	/* just remove the option */
2975 		}
2976 
2977 		/* message length validation */
2978 		if (len < sizeof(struct ip6_hbh))
2979 			return (EINVAL);
2980 		hbh = (struct ip6_hbh *)buf;
2981 		hbhlen = (hbh->ip6h_len + 1) << 3;
2982 		if (len != hbhlen)
2983 			return (EINVAL);
2984 
2985 		/* turn off the previous option, then set the new option. */
2986 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2987 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2988 		if (opt->ip6po_hbh == NULL)
2989 			return (ENOBUFS);
2990 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2991 
2992 		break;
2993 	}
2994 
2995 	case IPV6_2292DSTOPTS:
2996 	case IPV6_DSTOPTS:
2997 	case IPV6_RTHDRDSTOPTS:
2998 	{
2999 		struct ip6_dest *dest, **newdest = NULL;
3000 		int destlen;
3001 
3002 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3003 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3004 			if (error)
3005 				return (error);
3006 		}
3007 
3008 		if (len == 0) {
3009 			ip6_clearpktopts(opt, optname);
3010 			break;	/* just remove the option */
3011 		}
3012 
3013 		/* message length validation */
3014 		if (len < sizeof(struct ip6_dest))
3015 			return (EINVAL);
3016 		dest = (struct ip6_dest *)buf;
3017 		destlen = (dest->ip6d_len + 1) << 3;
3018 		if (len != destlen)
3019 			return (EINVAL);
3020 
3021 		/*
3022 		 * Determine the position that the destination options header
3023 		 * should be inserted; before or after the routing header.
3024 		 */
3025 		switch (optname) {
3026 		case IPV6_2292DSTOPTS:
3027 			/*
3028 			 * The old advacned API is ambiguous on this point.
3029 			 * Our approach is to determine the position based
3030 			 * according to the existence of a routing header.
3031 			 * Note, however, that this depends on the order of the
3032 			 * extension headers in the ancillary data; the 1st
3033 			 * part of the destination options header must appear
3034 			 * before the routing header in the ancillary data,
3035 			 * too.
3036 			 * RFC3542 solved the ambiguity by introducing
3037 			 * separate ancillary data or option types.
3038 			 */
3039 			if (opt->ip6po_rthdr == NULL)
3040 				newdest = &opt->ip6po_dest1;
3041 			else
3042 				newdest = &opt->ip6po_dest2;
3043 			break;
3044 		case IPV6_RTHDRDSTOPTS:
3045 			newdest = &opt->ip6po_dest1;
3046 			break;
3047 		case IPV6_DSTOPTS:
3048 			newdest = &opt->ip6po_dest2;
3049 			break;
3050 		}
3051 
3052 		/* turn off the previous option, then set the new option. */
3053 		ip6_clearpktopts(opt, optname);
3054 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3055 		if (*newdest == NULL)
3056 			return (ENOBUFS);
3057 		bcopy(dest, *newdest, destlen);
3058 
3059 		break;
3060 	}
3061 
3062 	case IPV6_2292RTHDR:
3063 	case IPV6_RTHDR:
3064 	{
3065 		struct ip6_rthdr *rth;
3066 		int rthlen;
3067 
3068 		if (len == 0) {
3069 			ip6_clearpktopts(opt, IPV6_RTHDR);
3070 			break;	/* just remove the option */
3071 		}
3072 
3073 		/* message length validation */
3074 		if (len < sizeof(struct ip6_rthdr))
3075 			return (EINVAL);
3076 		rth = (struct ip6_rthdr *)buf;
3077 		rthlen = (rth->ip6r_len + 1) << 3;
3078 		if (len != rthlen)
3079 			return (EINVAL);
3080 
3081 		switch (rth->ip6r_type) {
3082 		case IPV6_RTHDR_TYPE_0:
3083 			if (rth->ip6r_len == 0)	/* must contain one addr */
3084 				return (EINVAL);
3085 			if (rth->ip6r_len % 2) /* length must be even */
3086 				return (EINVAL);
3087 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3088 				return (EINVAL);
3089 			break;
3090 		default:
3091 			return (EINVAL);	/* not supported */
3092 		}
3093 
3094 		/* turn off the previous option */
3095 		ip6_clearpktopts(opt, IPV6_RTHDR);
3096 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3097 		if (opt->ip6po_rthdr == NULL)
3098 			return (ENOBUFS);
3099 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3100 
3101 		break;
3102 	}
3103 
3104 	case IPV6_USE_MIN_MTU:
3105 		if (len != sizeof(int))
3106 			return (EINVAL);
3107 		minmtupolicy = *(int *)buf;
3108 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3109 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3110 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3111 			return (EINVAL);
3112 		}
3113 		opt->ip6po_minmtu = minmtupolicy;
3114 		break;
3115 
3116 	case IPV6_DONTFRAG:
3117 		if (len != sizeof(int))
3118 			return (EINVAL);
3119 
3120 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3121 			/*
3122 			 * we ignore this option for TCP sockets.
3123 			 * (RFC3542 leaves this case unspecified.)
3124 			 */
3125 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3126 		} else
3127 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3128 		break;
3129 
3130 	case IPV6_PREFER_TEMPADDR:
3131 		if (len != sizeof(int))
3132 			return (EINVAL);
3133 		preftemp = *(int *)buf;
3134 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3135 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3136 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3137 			return (EINVAL);
3138 		}
3139 		opt->ip6po_prefer_tempaddr = preftemp;
3140 		break;
3141 
3142 	default:
3143 		return (ENOPROTOOPT);
3144 	} /* end of switch */
3145 
3146 	return (0);
3147 }
3148 
3149 /*
3150  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3151  * packet to the input queue of a specified interface.  Note that this
3152  * calls the output routine of the loopback "driver", but with an interface
3153  * pointer that might NOT be &loif -- easier than replicating that code here.
3154  */
3155 void
3156 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3157 {
3158 	struct mbuf *copym;
3159 	struct ip6_hdr *ip6;
3160 
3161 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3162 	if (copym == NULL)
3163 		return;
3164 
3165 	/*
3166 	 * Make sure to deep-copy IPv6 header portion in case the data
3167 	 * is in an mbuf cluster, so that we can safely override the IPv6
3168 	 * header portion later.
3169 	 */
3170 	if (!M_WRITABLE(copym) ||
3171 	    copym->m_len < sizeof(struct ip6_hdr)) {
3172 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3173 		if (copym == NULL)
3174 			return;
3175 	}
3176 	ip6 = mtod(copym, struct ip6_hdr *);
3177 	/*
3178 	 * clear embedded scope identifiers if necessary.
3179 	 * in6_clearscope will touch the addresses only when necessary.
3180 	 */
3181 	in6_clearscope(&ip6->ip6_src);
3182 	in6_clearscope(&ip6->ip6_dst);
3183 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3184 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3185 		    CSUM_PSEUDO_HDR;
3186 		copym->m_pkthdr.csum_data = 0xffff;
3187 	}
3188 	if_simloop(ifp, copym, AF_INET6, 0);
3189 }
3190 
3191 /*
3192  * Chop IPv6 header off from the payload.
3193  */
3194 static int
3195 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3196 {
3197 	struct mbuf *mh;
3198 	struct ip6_hdr *ip6;
3199 
3200 	ip6 = mtod(m, struct ip6_hdr *);
3201 	if (m->m_len > sizeof(*ip6)) {
3202 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3203 		if (mh == NULL) {
3204 			m_freem(m);
3205 			return ENOBUFS;
3206 		}
3207 		m_move_pkthdr(mh, m);
3208 		M_ALIGN(mh, sizeof(*ip6));
3209 		m->m_len -= sizeof(*ip6);
3210 		m->m_data += sizeof(*ip6);
3211 		mh->m_next = m;
3212 		m = mh;
3213 		m->m_len = sizeof(*ip6);
3214 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3215 	}
3216 	exthdrs->ip6e_ip6 = m;
3217 	return 0;
3218 }
3219 
3220 /*
3221  * Compute IPv6 extension header length.
3222  */
3223 int
3224 ip6_optlen(struct inpcb *inp)
3225 {
3226 	int len;
3227 
3228 	if (!inp->in6p_outputopts)
3229 		return 0;
3230 
3231 	len = 0;
3232 #define elen(x) \
3233     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3234 
3235 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3236 	if (inp->in6p_outputopts->ip6po_rthdr)
3237 		/* dest1 is valid with rthdr only */
3238 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3239 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3240 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3241 	return len;
3242 #undef elen
3243 }
3244