xref: /freebsd/sys/netinet6/ip6_output.c (revision 74ca7bf1d4c7173d5575ba168bc4b5f6d181ff5a)
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_ratelimit.h"
71 #include "opt_ipsec.h"
72 #include "opt_sctp.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/errno.h>
81 #include <sys/priv.h>
82 #include <sys/proc.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/ucred.h>
88 
89 #include <machine/in_cksum.h>
90 
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/pfil.h>
97 #include <net/rss_config.h>
98 #include <net/vnet.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_var.h>
102 #include <netinet/ip_var.h>
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/tcp_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/in6_rss.h>
112 
113 #include <netipsec/ipsec_support.h>
114 #ifdef SCTP
115 #include <netinet/sctp.h>
116 #include <netinet/sctp_crc32.h>
117 #endif
118 
119 #include <netinet6/ip6protosw.h>
120 #include <netinet6/scope6_var.h>
121 
122 extern int in6_mcast_loop;
123 
124 struct ip6_exthdrs {
125 	struct mbuf *ip6e_ip6;
126 	struct mbuf *ip6e_hbh;
127 	struct mbuf *ip6e_dest1;
128 	struct mbuf *ip6e_rthdr;
129 	struct mbuf *ip6e_dest2;
130 };
131 
132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
133 
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 			   struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 	struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 	struct ucred *, int, int, int);
141 
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 	struct ip6_frag **);
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static int ip6_getpmtu(struct route_in6 *, int,
148 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
149 	u_int);
150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
151 	u_long *, int *, u_int);
152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
154 
155 
156 /*
157  * Make an extension header from option data.  hp is the source, and
158  * mp is the destination.
159  */
160 #define MAKE_EXTHDR(hp, mp)						\
161     do {								\
162 	if (hp) {							\
163 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
164 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
165 		    ((eh)->ip6e_len + 1) << 3);				\
166 		if (error)						\
167 			goto freehdrs;					\
168 	}								\
169     } while (/*CONSTCOND*/ 0)
170 
171 /*
172  * Form a chain of extension headers.
173  * m is the extension header mbuf
174  * mp is the previous mbuf in the chain
175  * p is the next header
176  * i is the type of option.
177  */
178 #define MAKE_CHAIN(m, mp, p, i)\
179     do {\
180 	if (m) {\
181 		if (!hdrsplit) \
182 			panic("assumption failed: hdr not split"); \
183 		*mtod((m), u_char *) = *(p);\
184 		*(p) = (i);\
185 		p = mtod((m), u_char *);\
186 		(m)->m_next = (mp)->m_next;\
187 		(mp)->m_next = (m);\
188 		(mp) = (m);\
189 	}\
190     } while (/*CONSTCOND*/ 0)
191 
192 void
193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
194 {
195 	u_short csum;
196 
197 	csum = in_cksum_skip(m, offset + plen, offset);
198 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
199 		csum = 0xffff;
200 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
201 
202 	if (offset + sizeof(u_short) > m->m_len) {
203 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
204 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
205 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
206 		/*
207 		 * XXX this should not happen, but if it does, the correct
208 		 * behavior may be to insert the checksum in the appropriate
209 		 * next mbuf in the chain.
210 		 */
211 		return;
212 	}
213 	*(u_short *)(m->m_data + offset) = csum;
214 }
215 
216 int
217 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
218     int fraglen , uint32_t id)
219 {
220 	struct mbuf *m, **mnext, *m_frgpart;
221 	struct ip6_hdr *ip6, *mhip6;
222 	struct ip6_frag *ip6f;
223 	int off;
224 	int error;
225 	int tlen = m0->m_pkthdr.len;
226 
227 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
228 
229 	m = m0;
230 	ip6 = mtod(m, struct ip6_hdr *);
231 	mnext = &m->m_nextpkt;
232 
233 	for (off = hlen; off < tlen; off += fraglen) {
234 		m = m_gethdr(M_NOWAIT, MT_DATA);
235 		if (!m) {
236 			IP6STAT_INC(ip6s_odropped);
237 			return (ENOBUFS);
238 		}
239 		m->m_flags = m0->m_flags & M_COPYFLAGS;
240 		*mnext = m;
241 		mnext = &m->m_nextpkt;
242 		m->m_data += max_linkhdr;
243 		mhip6 = mtod(m, struct ip6_hdr *);
244 		*mhip6 = *ip6;
245 		m->m_len = sizeof(*mhip6);
246 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
247 		if (error) {
248 			IP6STAT_INC(ip6s_odropped);
249 			return (error);
250 		}
251 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
252 		if (off + fraglen >= tlen)
253 			fraglen = tlen - off;
254 		else
255 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
256 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
257 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
258 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
259 			IP6STAT_INC(ip6s_odropped);
260 			return (ENOBUFS);
261 		}
262 		m_cat(m, m_frgpart);
263 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
264 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
265 		m->m_pkthdr.rcvif = NULL;
266 		ip6f->ip6f_reserved = 0;
267 		ip6f->ip6f_ident = id;
268 		ip6f->ip6f_nxt = nextproto;
269 		IP6STAT_INC(ip6s_ofragments);
270 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
271 	}
272 
273 	return (0);
274 }
275 
276 /*
277  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
278  * header (with pri, len, nxt, hlim, src, dst).
279  * This function may modify ver and hlim only.
280  * The mbuf chain containing the packet will be freed.
281  * The mbuf opt, if present, will not be freed.
282  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
283  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
284  * then result of route lookup is stored in ro->ro_rt.
285  *
286  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
287  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
288  * which is rt_mtu.
289  *
290  * ifpp - XXX: just for statistics
291  */
292 /*
293  * XXX TODO: no flowid is assigned for outbound flows?
294  */
295 int
296 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
297     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
298     struct ifnet **ifpp, struct inpcb *inp)
299 {
300 	struct ip6_hdr *ip6;
301 	struct ifnet *ifp, *origifp;
302 	struct mbuf *m = m0;
303 	struct mbuf *mprev = NULL;
304 	int hlen, tlen, len;
305 	struct route_in6 ip6route;
306 	struct rtentry *rt = NULL;
307 	struct sockaddr_in6 *dst, src_sa, dst_sa;
308 	struct in6_addr odst;
309 	int error = 0;
310 	struct in6_ifaddr *ia = NULL;
311 	u_long mtu;
312 	int alwaysfrag, dontfrag;
313 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
314 	struct ip6_exthdrs exthdrs;
315 	struct in6_addr src0, dst0;
316 	u_int32_t zone;
317 	struct route_in6 *ro_pmtu = NULL;
318 	int hdrsplit = 0;
319 	int sw_csum, tso;
320 	int needfiblookup;
321 	uint32_t fibnum;
322 	struct m_tag *fwd_tag = NULL;
323 	uint32_t id;
324 
325 	if (inp != NULL) {
326 		INP_LOCK_ASSERT(inp);
327 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
328 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
329 			/* unconditionally set flowid */
330 			m->m_pkthdr.flowid = inp->inp_flowid;
331 			M_HASHTYPE_SET(m, inp->inp_flowtype);
332 		}
333 	}
334 
335 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
336 	/*
337 	 * IPSec checking which handles several cases.
338 	 * FAST IPSEC: We re-injected the packet.
339 	 * XXX: need scope argument.
340 	 */
341 	if (IPSEC_ENABLED(ipv6)) {
342 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
343 			if (error == EINPROGRESS)
344 				error = 0;
345 			goto done;
346 		}
347 	}
348 #endif /* IPSEC */
349 
350 	bzero(&exthdrs, sizeof(exthdrs));
351 	if (opt) {
352 		/* Hop-by-Hop options header */
353 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
354 		/* Destination options header(1st part) */
355 		if (opt->ip6po_rthdr) {
356 			/*
357 			 * Destination options header(1st part)
358 			 * This only makes sense with a routing header.
359 			 * See Section 9.2 of RFC 3542.
360 			 * Disabling this part just for MIP6 convenience is
361 			 * a bad idea.  We need to think carefully about a
362 			 * way to make the advanced API coexist with MIP6
363 			 * options, which might automatically be inserted in
364 			 * the kernel.
365 			 */
366 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
367 		}
368 		/* Routing header */
369 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
370 		/* Destination options header(2nd part) */
371 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
372 	}
373 
374 	/*
375 	 * Calculate the total length of the extension header chain.
376 	 * Keep the length of the unfragmentable part for fragmentation.
377 	 */
378 	optlen = 0;
379 	if (exthdrs.ip6e_hbh)
380 		optlen += exthdrs.ip6e_hbh->m_len;
381 	if (exthdrs.ip6e_dest1)
382 		optlen += exthdrs.ip6e_dest1->m_len;
383 	if (exthdrs.ip6e_rthdr)
384 		optlen += exthdrs.ip6e_rthdr->m_len;
385 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
386 
387 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
388 	if (exthdrs.ip6e_dest2)
389 		optlen += exthdrs.ip6e_dest2->m_len;
390 
391 	/*
392 	 * If there is at least one extension header,
393 	 * separate IP6 header from the payload.
394 	 */
395 	if (optlen && !hdrsplit) {
396 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
397 			m = NULL;
398 			goto freehdrs;
399 		}
400 		m = exthdrs.ip6e_ip6;
401 		hdrsplit++;
402 	}
403 
404 	ip6 = mtod(m, struct ip6_hdr *);
405 
406 	/* adjust mbuf packet header length */
407 	m->m_pkthdr.len += optlen;
408 	plen = m->m_pkthdr.len - sizeof(*ip6);
409 
410 	/* If this is a jumbo payload, insert a jumbo payload option. */
411 	if (plen > IPV6_MAXPACKET) {
412 		if (!hdrsplit) {
413 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
414 				m = NULL;
415 				goto freehdrs;
416 			}
417 			m = exthdrs.ip6e_ip6;
418 			hdrsplit++;
419 		}
420 		/* adjust pointer */
421 		ip6 = mtod(m, struct ip6_hdr *);
422 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
423 			goto freehdrs;
424 		ip6->ip6_plen = 0;
425 	} else
426 		ip6->ip6_plen = htons(plen);
427 
428 	/*
429 	 * Concatenate headers and fill in next header fields.
430 	 * Here we have, on "m"
431 	 *	IPv6 payload
432 	 * and we insert headers accordingly.  Finally, we should be getting:
433 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
434 	 *
435 	 * during the header composing process, "m" points to IPv6 header.
436 	 * "mprev" points to an extension header prior to esp.
437 	 */
438 	u_char *nexthdrp = &ip6->ip6_nxt;
439 	mprev = m;
440 
441 	/*
442 	 * we treat dest2 specially.  this makes IPsec processing
443 	 * much easier.  the goal here is to make mprev point the
444 	 * mbuf prior to dest2.
445 	 *
446 	 * result: IPv6 dest2 payload
447 	 * m and mprev will point to IPv6 header.
448 	 */
449 	if (exthdrs.ip6e_dest2) {
450 		if (!hdrsplit)
451 			panic("assumption failed: hdr not split");
452 		exthdrs.ip6e_dest2->m_next = m->m_next;
453 		m->m_next = exthdrs.ip6e_dest2;
454 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
455 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
456 	}
457 
458 	/*
459 	 * result: IPv6 hbh dest1 rthdr dest2 payload
460 	 * m will point to IPv6 header.  mprev will point to the
461 	 * extension header prior to dest2 (rthdr in the above case).
462 	 */
463 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
464 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
465 		   IPPROTO_DSTOPTS);
466 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
467 		   IPPROTO_ROUTING);
468 
469 	/*
470 	 * If there is a routing header, discard the packet.
471 	 */
472 	if (exthdrs.ip6e_rthdr) {
473 		 error = EINVAL;
474 		 goto bad;
475 	}
476 
477 	/* Source address validation */
478 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
479 	    (flags & IPV6_UNSPECSRC) == 0) {
480 		error = EOPNOTSUPP;
481 		IP6STAT_INC(ip6s_badscope);
482 		goto bad;
483 	}
484 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
485 		error = EOPNOTSUPP;
486 		IP6STAT_INC(ip6s_badscope);
487 		goto bad;
488 	}
489 
490 	IP6STAT_INC(ip6s_localout);
491 
492 	/*
493 	 * Route packet.
494 	 */
495 	if (ro == NULL) {
496 		ro = &ip6route;
497 		bzero((caddr_t)ro, sizeof(*ro));
498 	}
499 	ro_pmtu = ro;
500 	if (opt && opt->ip6po_rthdr)
501 		ro = &opt->ip6po_route;
502 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
503 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
504 again:
505 	/*
506 	 * if specified, try to fill in the traffic class field.
507 	 * do not override if a non-zero value is already set.
508 	 * we check the diffserv field and the ecn field separately.
509 	 */
510 	if (opt && opt->ip6po_tclass >= 0) {
511 		int mask = 0;
512 
513 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
514 			mask |= 0xfc;
515 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
516 			mask |= 0x03;
517 		if (mask != 0)
518 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
519 	}
520 
521 	/* fill in or override the hop limit field, if necessary. */
522 	if (opt && opt->ip6po_hlim != -1)
523 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
524 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
525 		if (im6o != NULL)
526 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
527 		else
528 			ip6->ip6_hlim = V_ip6_defmcasthlim;
529 	}
530 	/*
531 	 * Validate route against routing table additions;
532 	 * a better/more specific route might have been added.
533 	 * Make sure address family is set in route.
534 	 */
535 	if (inp) {
536 		ro->ro_dst.sin6_family = AF_INET6;
537 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
538 	}
539 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
540 	    ro->ro_dst.sin6_family == AF_INET6 &&
541 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
542 		rt = ro->ro_rt;
543 		ifp = ro->ro_rt->rt_ifp;
544 	} else {
545 		if (ro->ro_lle)
546 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
547 		ro->ro_lle = NULL;
548 		if (fwd_tag == NULL) {
549 			bzero(&dst_sa, sizeof(dst_sa));
550 			dst_sa.sin6_family = AF_INET6;
551 			dst_sa.sin6_len = sizeof(dst_sa);
552 			dst_sa.sin6_addr = ip6->ip6_dst;
553 		}
554 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
555 		    &rt, fibnum);
556 		if (error != 0) {
557 			if (ifp != NULL)
558 				in6_ifstat_inc(ifp, ifs6_out_discard);
559 			goto bad;
560 		}
561 	}
562 	if (rt == NULL) {
563 		/*
564 		 * If in6_selectroute() does not return a route entry,
565 		 * dst may not have been updated.
566 		 */
567 		*dst = dst_sa;	/* XXX */
568 	}
569 
570 	/*
571 	 * then rt (for unicast) and ifp must be non-NULL valid values.
572 	 */
573 	if ((flags & IPV6_FORWARDING) == 0) {
574 		/* XXX: the FORWARDING flag can be set for mrouting. */
575 		in6_ifstat_inc(ifp, ifs6_out_request);
576 	}
577 	if (rt != NULL) {
578 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
579 		counter_u64_add(rt->rt_pksent, 1);
580 	}
581 
582 
583 	/*
584 	 * The outgoing interface must be in the zone of source and
585 	 * destination addresses.
586 	 */
587 	origifp = ifp;
588 
589 	src0 = ip6->ip6_src;
590 	if (in6_setscope(&src0, origifp, &zone))
591 		goto badscope;
592 	bzero(&src_sa, sizeof(src_sa));
593 	src_sa.sin6_family = AF_INET6;
594 	src_sa.sin6_len = sizeof(src_sa);
595 	src_sa.sin6_addr = ip6->ip6_src;
596 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
597 		goto badscope;
598 
599 	dst0 = ip6->ip6_dst;
600 	if (in6_setscope(&dst0, origifp, &zone))
601 		goto badscope;
602 	/* re-initialize to be sure */
603 	bzero(&dst_sa, sizeof(dst_sa));
604 	dst_sa.sin6_family = AF_INET6;
605 	dst_sa.sin6_len = sizeof(dst_sa);
606 	dst_sa.sin6_addr = ip6->ip6_dst;
607 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
608 		goto badscope;
609 	}
610 
611 	/* We should use ia_ifp to support the case of
612 	 * sending packets to an address of our own.
613 	 */
614 	if (ia != NULL && ia->ia_ifp)
615 		ifp = ia->ia_ifp;
616 
617 	/* scope check is done. */
618 	goto routefound;
619 
620   badscope:
621 	IP6STAT_INC(ip6s_badscope);
622 	in6_ifstat_inc(origifp, ifs6_out_discard);
623 	if (error == 0)
624 		error = EHOSTUNREACH; /* XXX */
625 	goto bad;
626 
627   routefound:
628 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
629 		if (opt && opt->ip6po_nextroute.ro_rt) {
630 			/*
631 			 * The nexthop is explicitly specified by the
632 			 * application.  We assume the next hop is an IPv6
633 			 * address.
634 			 */
635 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
636 		}
637 		else if ((rt->rt_flags & RTF_GATEWAY))
638 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
639 	}
640 
641 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
642 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
643 	} else {
644 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
645 		in6_ifstat_inc(ifp, ifs6_out_mcast);
646 		/*
647 		 * Confirm that the outgoing interface supports multicast.
648 		 */
649 		if (!(ifp->if_flags & IFF_MULTICAST)) {
650 			IP6STAT_INC(ip6s_noroute);
651 			in6_ifstat_inc(ifp, ifs6_out_discard);
652 			error = ENETUNREACH;
653 			goto bad;
654 		}
655 		if ((im6o == NULL && in6_mcast_loop) ||
656 		    (im6o && im6o->im6o_multicast_loop)) {
657 			/*
658 			 * Loop back multicast datagram if not expressly
659 			 * forbidden to do so, even if we have not joined
660 			 * the address; protocols will filter it later,
661 			 * thus deferring a hash lookup and lock acquisition
662 			 * at the expense of an m_copym().
663 			 */
664 			ip6_mloopback(ifp, m);
665 		} else {
666 			/*
667 			 * If we are acting as a multicast router, perform
668 			 * multicast forwarding as if the packet had just
669 			 * arrived on the interface to which we are about
670 			 * to send.  The multicast forwarding function
671 			 * recursively calls this function, using the
672 			 * IPV6_FORWARDING flag to prevent infinite recursion.
673 			 *
674 			 * Multicasts that are looped back by ip6_mloopback(),
675 			 * above, will be forwarded by the ip6_input() routine,
676 			 * if necessary.
677 			 */
678 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
679 				/*
680 				 * XXX: ip6_mforward expects that rcvif is NULL
681 				 * when it is called from the originating path.
682 				 * However, it may not always be the case.
683 				 */
684 				m->m_pkthdr.rcvif = NULL;
685 				if (ip6_mforward(ip6, ifp, m) != 0) {
686 					m_freem(m);
687 					goto done;
688 				}
689 			}
690 		}
691 		/*
692 		 * Multicasts with a hoplimit of zero may be looped back,
693 		 * above, but must not be transmitted on a network.
694 		 * Also, multicasts addressed to the loopback interface
695 		 * are not sent -- the above call to ip6_mloopback() will
696 		 * loop back a copy if this host actually belongs to the
697 		 * destination group on the loopback interface.
698 		 */
699 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
700 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
701 			m_freem(m);
702 			goto done;
703 		}
704 	}
705 
706 	/*
707 	 * Fill the outgoing inteface to tell the upper layer
708 	 * to increment per-interface statistics.
709 	 */
710 	if (ifpp)
711 		*ifpp = ifp;
712 
713 	/* Determine path MTU. */
714 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
715 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
716 		goto bad;
717 
718 	/*
719 	 * The caller of this function may specify to use the minimum MTU
720 	 * in some cases.
721 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
722 	 * setting.  The logic is a bit complicated; by default, unicast
723 	 * packets will follow path MTU while multicast packets will be sent at
724 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
725 	 * including unicast ones will be sent at the minimum MTU.  Multicast
726 	 * packets will always be sent at the minimum MTU unless
727 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
728 	 * See RFC 3542 for more details.
729 	 */
730 	if (mtu > IPV6_MMTU) {
731 		if ((flags & IPV6_MINMTU))
732 			mtu = IPV6_MMTU;
733 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
734 			mtu = IPV6_MMTU;
735 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
736 			 (opt == NULL ||
737 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
738 			mtu = IPV6_MMTU;
739 		}
740 	}
741 
742 	/*
743 	 * clear embedded scope identifiers if necessary.
744 	 * in6_clearscope will touch the addresses only when necessary.
745 	 */
746 	in6_clearscope(&ip6->ip6_src);
747 	in6_clearscope(&ip6->ip6_dst);
748 
749 	/*
750 	 * If the outgoing packet contains a hop-by-hop options header,
751 	 * it must be examined and processed even by the source node.
752 	 * (RFC 2460, section 4.)
753 	 */
754 	if (exthdrs.ip6e_hbh) {
755 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
756 		u_int32_t dummy; /* XXX unused */
757 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
758 
759 #ifdef DIAGNOSTIC
760 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
761 			panic("ip6e_hbh is not contiguous");
762 #endif
763 		/*
764 		 *  XXX: if we have to send an ICMPv6 error to the sender,
765 		 *       we need the M_LOOP flag since icmp6_error() expects
766 		 *       the IPv6 and the hop-by-hop options header are
767 		 *       contiguous unless the flag is set.
768 		 */
769 		m->m_flags |= M_LOOP;
770 		m->m_pkthdr.rcvif = ifp;
771 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
772 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
773 		    &dummy, &plen) < 0) {
774 			/* m was already freed at this point */
775 			error = EINVAL;/* better error? */
776 			goto done;
777 		}
778 		m->m_flags &= ~M_LOOP; /* XXX */
779 		m->m_pkthdr.rcvif = NULL;
780 	}
781 
782 	/* Jump over all PFIL processing if hooks are not active. */
783 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
784 		goto passout;
785 
786 	odst = ip6->ip6_dst;
787 	/* Run through list of hooks for output packets. */
788 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, 0, inp);
789 	if (error != 0 || m == NULL)
790 		goto done;
791 	/* adjust pointer */
792 	ip6 = mtod(m, struct ip6_hdr *);
793 
794 	needfiblookup = 0;
795 	/* See if destination IP address was changed by packet filter. */
796 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
797 		m->m_flags |= M_SKIP_FIREWALL;
798 		/* If destination is now ourself drop to ip6_input(). */
799 		if (in6_localip(&ip6->ip6_dst)) {
800 			m->m_flags |= M_FASTFWD_OURS;
801 			if (m->m_pkthdr.rcvif == NULL)
802 				m->m_pkthdr.rcvif = V_loif;
803 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
804 				m->m_pkthdr.csum_flags |=
805 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
806 				m->m_pkthdr.csum_data = 0xffff;
807 			}
808 #ifdef SCTP
809 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
810 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
811 #endif
812 			error = netisr_queue(NETISR_IPV6, m);
813 			goto done;
814 		} else {
815 			RO_RTFREE(ro);
816 			needfiblookup = 1; /* Redo the routing table lookup. */
817 			if (ro->ro_lle)
818 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
819 			ro->ro_lle = NULL;
820 		}
821 	}
822 	/* See if fib was changed by packet filter. */
823 	if (fibnum != M_GETFIB(m)) {
824 		m->m_flags |= M_SKIP_FIREWALL;
825 		fibnum = M_GETFIB(m);
826 		RO_RTFREE(ro);
827 		needfiblookup = 1;
828 		if (ro->ro_lle)
829 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
830 		ro->ro_lle = NULL;
831 	}
832 	if (needfiblookup)
833 		goto again;
834 
835 	/* See if local, if yes, send it to netisr. */
836 	if (m->m_flags & M_FASTFWD_OURS) {
837 		if (m->m_pkthdr.rcvif == NULL)
838 			m->m_pkthdr.rcvif = V_loif;
839 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
840 			m->m_pkthdr.csum_flags |=
841 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
842 			m->m_pkthdr.csum_data = 0xffff;
843 		}
844 #ifdef SCTP
845 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
846 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
847 #endif
848 		error = netisr_queue(NETISR_IPV6, m);
849 		goto done;
850 	}
851 	/* Or forward to some other address? */
852 	if ((m->m_flags & M_IP6_NEXTHOP) &&
853 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
854 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
855 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
856 		m->m_flags |= M_SKIP_FIREWALL;
857 		m->m_flags &= ~M_IP6_NEXTHOP;
858 		m_tag_delete(m, fwd_tag);
859 		goto again;
860 	}
861 
862 passout:
863 	/*
864 	 * Send the packet to the outgoing interface.
865 	 * If necessary, do IPv6 fragmentation before sending.
866 	 *
867 	 * the logic here is rather complex:
868 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
869 	 * 1-a:	send as is if tlen <= path mtu
870 	 * 1-b:	fragment if tlen > path mtu
871 	 *
872 	 * 2: if user asks us not to fragment (dontfrag == 1)
873 	 * 2-a:	send as is if tlen <= interface mtu
874 	 * 2-b:	error if tlen > interface mtu
875 	 *
876 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
877 	 *	always fragment
878 	 *
879 	 * 4: if dontfrag == 1 && alwaysfrag == 1
880 	 *	error, as we cannot handle this conflicting request
881 	 */
882 	sw_csum = m->m_pkthdr.csum_flags;
883 	if (!hdrsplit) {
884 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
885 		sw_csum &= ~ifp->if_hwassist;
886 	} else
887 		tso = 0;
888 	/*
889 	 * If we added extension headers, we will not do TSO and calculate the
890 	 * checksums ourselves for now.
891 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
892 	 * with ext. hdrs.
893 	 */
894 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
895 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
896 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
897 	}
898 #ifdef SCTP
899 	if (sw_csum & CSUM_SCTP_IPV6) {
900 		sw_csum &= ~CSUM_SCTP_IPV6;
901 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
902 	}
903 #endif
904 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
905 	tlen = m->m_pkthdr.len;
906 
907 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
908 		dontfrag = 1;
909 	else
910 		dontfrag = 0;
911 	if (dontfrag && alwaysfrag) {	/* case 4 */
912 		/* conflicting request - can't transmit */
913 		error = EMSGSIZE;
914 		goto bad;
915 	}
916 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
917 		/*
918 		 * Even if the DONTFRAG option is specified, we cannot send the
919 		 * packet when the data length is larger than the MTU of the
920 		 * outgoing interface.
921 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
922 		 * application wanted to know the MTU value. Also return an
923 		 * error code (this is not described in the API spec).
924 		 */
925 		if (inp != NULL)
926 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
927 		error = EMSGSIZE;
928 		goto bad;
929 	}
930 
931 	/*
932 	 * transmit packet without fragmentation
933 	 */
934 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
935 		struct in6_ifaddr *ia6;
936 
937 		ip6 = mtod(m, struct ip6_hdr *);
938 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
939 		if (ia6) {
940 			/* Record statistics for this interface address. */
941 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
942 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
943 			    m->m_pkthdr.len);
944 			ifa_free(&ia6->ia_ifa);
945 		}
946 #ifdef RATELIMIT
947 		if (inp != NULL) {
948 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
949 				in_pcboutput_txrtlmt(inp, ifp, m);
950 			/* stamp send tag on mbuf */
951 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
952 		} else {
953 			m->m_pkthdr.snd_tag = NULL;
954 		}
955 #endif
956 		error = nd6_output_ifp(ifp, origifp, m, dst,
957 		    (struct route *)ro);
958 #ifdef RATELIMIT
959 		/* check for route change */
960 		if (error == EAGAIN)
961 			in_pcboutput_eagain(inp);
962 #endif
963 		goto done;
964 	}
965 
966 	/*
967 	 * try to fragment the packet.  case 1-b and 3
968 	 */
969 	if (mtu < IPV6_MMTU) {
970 		/* path MTU cannot be less than IPV6_MMTU */
971 		error = EMSGSIZE;
972 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
973 		goto bad;
974 	} else if (ip6->ip6_plen == 0) {
975 		/* jumbo payload cannot be fragmented */
976 		error = EMSGSIZE;
977 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
978 		goto bad;
979 	} else {
980 		u_char nextproto;
981 
982 		/*
983 		 * Too large for the destination or interface;
984 		 * fragment if possible.
985 		 * Must be able to put at least 8 bytes per fragment.
986 		 */
987 		hlen = unfragpartlen;
988 		if (mtu > IPV6_MAXPACKET)
989 			mtu = IPV6_MAXPACKET;
990 
991 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
992 		if (len < 8) {
993 			error = EMSGSIZE;
994 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
995 			goto bad;
996 		}
997 
998 		/*
999 		 * If the interface will not calculate checksums on
1000 		 * fragmented packets, then do it here.
1001 		 * XXX-BZ handle the hw offloading case.  Need flags.
1002 		 */
1003 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1004 			in6_delayed_cksum(m, plen, hlen);
1005 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1006 		}
1007 #ifdef SCTP
1008 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1009 			sctp_delayed_cksum(m, hlen);
1010 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1011 		}
1012 #endif
1013 		/*
1014 		 * Change the next header field of the last header in the
1015 		 * unfragmentable part.
1016 		 */
1017 		if (exthdrs.ip6e_rthdr) {
1018 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1019 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1020 		} else if (exthdrs.ip6e_dest1) {
1021 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1022 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1023 		} else if (exthdrs.ip6e_hbh) {
1024 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1025 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1026 		} else {
1027 			nextproto = ip6->ip6_nxt;
1028 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1029 		}
1030 
1031 		/*
1032 		 * Loop through length of segment after first fragment,
1033 		 * make new header and copy data of each part and link onto
1034 		 * chain.
1035 		 */
1036 		m0 = m;
1037 		id = htonl(ip6_randomid());
1038 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1039 			goto sendorfree;
1040 
1041 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1042 	}
1043 
1044 	/*
1045 	 * Remove leading garbages.
1046 	 */
1047 sendorfree:
1048 	m = m0->m_nextpkt;
1049 	m0->m_nextpkt = 0;
1050 	m_freem(m0);
1051 	for (; m; m = m0) {
1052 		m0 = m->m_nextpkt;
1053 		m->m_nextpkt = 0;
1054 		if (error == 0) {
1055 			/* Record statistics for this interface address. */
1056 			if (ia) {
1057 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1058 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1059 				    m->m_pkthdr.len);
1060 			}
1061 #ifdef RATELIMIT
1062 			if (inp != NULL) {
1063 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1064 					in_pcboutput_txrtlmt(inp, ifp, m);
1065 				/* stamp send tag on mbuf */
1066 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1067 			} else {
1068 				m->m_pkthdr.snd_tag = NULL;
1069 			}
1070 #endif
1071 			error = nd6_output_ifp(ifp, origifp, m, dst,
1072 			    (struct route *)ro);
1073 #ifdef RATELIMIT
1074 			/* check for route change */
1075 			if (error == EAGAIN)
1076 				in_pcboutput_eagain(inp);
1077 #endif
1078 		} else
1079 			m_freem(m);
1080 	}
1081 
1082 	if (error == 0)
1083 		IP6STAT_INC(ip6s_fragmented);
1084 
1085 done:
1086 	if (ro == &ip6route)
1087 		RO_RTFREE(ro);
1088 	return (error);
1089 
1090 freehdrs:
1091 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1092 	m_freem(exthdrs.ip6e_dest1);
1093 	m_freem(exthdrs.ip6e_rthdr);
1094 	m_freem(exthdrs.ip6e_dest2);
1095 	/* FALLTHROUGH */
1096 bad:
1097 	if (m)
1098 		m_freem(m);
1099 	goto done;
1100 }
1101 
1102 static int
1103 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1104 {
1105 	struct mbuf *m;
1106 
1107 	if (hlen > MCLBYTES)
1108 		return (ENOBUFS); /* XXX */
1109 
1110 	if (hlen > MLEN)
1111 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1112 	else
1113 		m = m_get(M_NOWAIT, MT_DATA);
1114 	if (m == NULL)
1115 		return (ENOBUFS);
1116 	m->m_len = hlen;
1117 	if (hdr)
1118 		bcopy(hdr, mtod(m, caddr_t), hlen);
1119 
1120 	*mp = m;
1121 	return (0);
1122 }
1123 
1124 /*
1125  * Insert jumbo payload option.
1126  */
1127 static int
1128 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1129 {
1130 	struct mbuf *mopt;
1131 	u_char *optbuf;
1132 	u_int32_t v;
1133 
1134 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1135 
1136 	/*
1137 	 * If there is no hop-by-hop options header, allocate new one.
1138 	 * If there is one but it doesn't have enough space to store the
1139 	 * jumbo payload option, allocate a cluster to store the whole options.
1140 	 * Otherwise, use it to store the options.
1141 	 */
1142 	if (exthdrs->ip6e_hbh == NULL) {
1143 		mopt = m_get(M_NOWAIT, MT_DATA);
1144 		if (mopt == NULL)
1145 			return (ENOBUFS);
1146 		mopt->m_len = JUMBOOPTLEN;
1147 		optbuf = mtod(mopt, u_char *);
1148 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1149 		exthdrs->ip6e_hbh = mopt;
1150 	} else {
1151 		struct ip6_hbh *hbh;
1152 
1153 		mopt = exthdrs->ip6e_hbh;
1154 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1155 			/*
1156 			 * XXX assumption:
1157 			 * - exthdrs->ip6e_hbh is not referenced from places
1158 			 *   other than exthdrs.
1159 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1160 			 */
1161 			int oldoptlen = mopt->m_len;
1162 			struct mbuf *n;
1163 
1164 			/*
1165 			 * XXX: give up if the whole (new) hbh header does
1166 			 * not fit even in an mbuf cluster.
1167 			 */
1168 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1169 				return (ENOBUFS);
1170 
1171 			/*
1172 			 * As a consequence, we must always prepare a cluster
1173 			 * at this point.
1174 			 */
1175 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1176 			if (n == NULL)
1177 				return (ENOBUFS);
1178 			n->m_len = oldoptlen + JUMBOOPTLEN;
1179 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1180 			    oldoptlen);
1181 			optbuf = mtod(n, caddr_t) + oldoptlen;
1182 			m_freem(mopt);
1183 			mopt = exthdrs->ip6e_hbh = n;
1184 		} else {
1185 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1186 			mopt->m_len += JUMBOOPTLEN;
1187 		}
1188 		optbuf[0] = IP6OPT_PADN;
1189 		optbuf[1] = 1;
1190 
1191 		/*
1192 		 * Adjust the header length according to the pad and
1193 		 * the jumbo payload option.
1194 		 */
1195 		hbh = mtod(mopt, struct ip6_hbh *);
1196 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1197 	}
1198 
1199 	/* fill in the option. */
1200 	optbuf[2] = IP6OPT_JUMBO;
1201 	optbuf[3] = 4;
1202 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1203 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1204 
1205 	/* finally, adjust the packet header length */
1206 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1207 
1208 	return (0);
1209 #undef JUMBOOPTLEN
1210 }
1211 
1212 /*
1213  * Insert fragment header and copy unfragmentable header portions.
1214  */
1215 static int
1216 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1217     struct ip6_frag **frghdrp)
1218 {
1219 	struct mbuf *n, *mlast;
1220 
1221 	if (hlen > sizeof(struct ip6_hdr)) {
1222 		n = m_copym(m0, sizeof(struct ip6_hdr),
1223 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1224 		if (n == NULL)
1225 			return (ENOBUFS);
1226 		m->m_next = n;
1227 	} else
1228 		n = m;
1229 
1230 	/* Search for the last mbuf of unfragmentable part. */
1231 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1232 		;
1233 
1234 	if (M_WRITABLE(mlast) &&
1235 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1236 		/* use the trailing space of the last mbuf for the fragment hdr */
1237 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1238 		    mlast->m_len);
1239 		mlast->m_len += sizeof(struct ip6_frag);
1240 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1241 	} else {
1242 		/* allocate a new mbuf for the fragment header */
1243 		struct mbuf *mfrg;
1244 
1245 		mfrg = m_get(M_NOWAIT, MT_DATA);
1246 		if (mfrg == NULL)
1247 			return (ENOBUFS);
1248 		mfrg->m_len = sizeof(struct ip6_frag);
1249 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1250 		mlast->m_next = mfrg;
1251 	}
1252 
1253 	return (0);
1254 }
1255 
1256 /*
1257  * Calculates IPv6 path mtu for destination @dst.
1258  * Resulting MTU is stored in @mtup.
1259  *
1260  * Returns 0 on success.
1261  */
1262 static int
1263 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1264 {
1265 	struct nhop6_extended nh6;
1266 	struct in6_addr kdst;
1267 	uint32_t scopeid;
1268 	struct ifnet *ifp;
1269 	u_long mtu;
1270 	int error;
1271 
1272 	in6_splitscope(dst, &kdst, &scopeid);
1273 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1274 		return (EHOSTUNREACH);
1275 
1276 	ifp = nh6.nh_ifp;
1277 	mtu = nh6.nh_mtu;
1278 
1279 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1280 	fib6_free_nh_ext(fibnum, &nh6);
1281 
1282 	return (error);
1283 }
1284 
1285 /*
1286  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1287  * and cached data in @ro_pmtu.
1288  * MTU from (successful) route lookup is saved (along with dst)
1289  * inside @ro_pmtu to avoid subsequent route lookups after packet
1290  * filter processing.
1291  *
1292  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1293  * Returns 0 on success.
1294  */
1295 static int
1296 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1297     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1298     int *alwaysfragp, u_int fibnum, u_int proto)
1299 {
1300 	struct nhop6_basic nh6;
1301 	struct in6_addr kdst;
1302 	uint32_t scopeid;
1303 	struct sockaddr_in6 *sa6_dst;
1304 	u_long mtu;
1305 
1306 	mtu = 0;
1307 	if (do_lookup) {
1308 
1309 		/*
1310 		 * Here ro_pmtu has final destination address, while
1311 		 * ro might represent immediate destination.
1312 		 * Use ro_pmtu destination since mtu might differ.
1313 		 */
1314 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1315 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1316 			ro_pmtu->ro_mtu = 0;
1317 
1318 		if (ro_pmtu->ro_mtu == 0) {
1319 			bzero(sa6_dst, sizeof(*sa6_dst));
1320 			sa6_dst->sin6_family = AF_INET6;
1321 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1322 			sa6_dst->sin6_addr = *dst;
1323 
1324 			in6_splitscope(dst, &kdst, &scopeid);
1325 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1326 			    &nh6) == 0)
1327 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1328 		}
1329 
1330 		mtu = ro_pmtu->ro_mtu;
1331 	}
1332 
1333 	if (ro_pmtu->ro_rt)
1334 		mtu = ro_pmtu->ro_rt->rt_mtu;
1335 
1336 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1337 }
1338 
1339 /*
1340  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1341  * hostcache data for @dst.
1342  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1343  *
1344  * Returns 0 on success.
1345  */
1346 static int
1347 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1348     u_long *mtup, int *alwaysfragp, u_int proto)
1349 {
1350 	u_long mtu = 0;
1351 	int alwaysfrag = 0;
1352 	int error = 0;
1353 
1354 	if (rt_mtu > 0) {
1355 		u_int32_t ifmtu;
1356 		struct in_conninfo inc;
1357 
1358 		bzero(&inc, sizeof(inc));
1359 		inc.inc_flags |= INC_ISIPV6;
1360 		inc.inc6_faddr = *dst;
1361 
1362 		ifmtu = IN6_LINKMTU(ifp);
1363 
1364 		/* TCP is known to react to pmtu changes so skip hc */
1365 		if (proto != IPPROTO_TCP)
1366 			mtu = tcp_hc_getmtu(&inc);
1367 
1368 		if (mtu)
1369 			mtu = min(mtu, rt_mtu);
1370 		else
1371 			mtu = rt_mtu;
1372 		if (mtu == 0)
1373 			mtu = ifmtu;
1374 		else if (mtu < IPV6_MMTU) {
1375 			/*
1376 			 * RFC2460 section 5, last paragraph:
1377 			 * if we record ICMPv6 too big message with
1378 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1379 			 * or smaller, with framgent header attached.
1380 			 * (fragment header is needed regardless from the
1381 			 * packet size, for translators to identify packets)
1382 			 */
1383 			alwaysfrag = 1;
1384 			mtu = IPV6_MMTU;
1385 		}
1386 	} else if (ifp) {
1387 		mtu = IN6_LINKMTU(ifp);
1388 	} else
1389 		error = EHOSTUNREACH; /* XXX */
1390 
1391 	*mtup = mtu;
1392 	if (alwaysfragp)
1393 		*alwaysfragp = alwaysfrag;
1394 	return (error);
1395 }
1396 
1397 /*
1398  * IP6 socket option processing.
1399  */
1400 int
1401 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1402 {
1403 	int optdatalen, uproto;
1404 	void *optdata;
1405 	struct inpcb *in6p = sotoinpcb(so);
1406 	int error, optval;
1407 	int level, op, optname;
1408 	int optlen;
1409 	struct thread *td;
1410 #ifdef	RSS
1411 	uint32_t rss_bucket;
1412 	int retval;
1413 #endif
1414 
1415 /*
1416  * Don't use more than a quarter of mbuf clusters.  N.B.:
1417  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1418  * on LP64 architectures, so cast to u_long to avoid undefined
1419  * behavior.  ILP32 architectures cannot have nmbclusters
1420  * large enough to overflow for other reasons.
1421  */
1422 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1423 
1424 	level = sopt->sopt_level;
1425 	op = sopt->sopt_dir;
1426 	optname = sopt->sopt_name;
1427 	optlen = sopt->sopt_valsize;
1428 	td = sopt->sopt_td;
1429 	error = 0;
1430 	optval = 0;
1431 	uproto = (int)so->so_proto->pr_protocol;
1432 
1433 	if (level != IPPROTO_IPV6) {
1434 		error = EINVAL;
1435 
1436 		if (sopt->sopt_level == SOL_SOCKET &&
1437 		    sopt->sopt_dir == SOPT_SET) {
1438 			switch (sopt->sopt_name) {
1439 			case SO_REUSEADDR:
1440 				INP_WLOCK(in6p);
1441 				if ((so->so_options & SO_REUSEADDR) != 0)
1442 					in6p->inp_flags2 |= INP_REUSEADDR;
1443 				else
1444 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1445 				INP_WUNLOCK(in6p);
1446 				error = 0;
1447 				break;
1448 			case SO_REUSEPORT:
1449 				INP_WLOCK(in6p);
1450 				if ((so->so_options & SO_REUSEPORT) != 0)
1451 					in6p->inp_flags2 |= INP_REUSEPORT;
1452 				else
1453 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1454 				INP_WUNLOCK(in6p);
1455 				error = 0;
1456 				break;
1457 			case SO_SETFIB:
1458 				INP_WLOCK(in6p);
1459 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1460 				INP_WUNLOCK(in6p);
1461 				error = 0;
1462 				break;
1463 			case SO_MAX_PACING_RATE:
1464 #ifdef RATELIMIT
1465 				INP_WLOCK(in6p);
1466 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1467 				INP_WUNLOCK(in6p);
1468 				error = 0;
1469 #else
1470 				error = EOPNOTSUPP;
1471 #endif
1472 				break;
1473 			default:
1474 				break;
1475 			}
1476 		}
1477 	} else {		/* level == IPPROTO_IPV6 */
1478 		switch (op) {
1479 
1480 		case SOPT_SET:
1481 			switch (optname) {
1482 			case IPV6_2292PKTOPTIONS:
1483 #ifdef IPV6_PKTOPTIONS
1484 			case IPV6_PKTOPTIONS:
1485 #endif
1486 			{
1487 				struct mbuf *m;
1488 
1489 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1490 					printf("ip6_ctloutput: mbuf limit hit\n");
1491 					error = ENOBUFS;
1492 					break;
1493 				}
1494 
1495 				error = soopt_getm(sopt, &m); /* XXX */
1496 				if (error != 0)
1497 					break;
1498 				error = soopt_mcopyin(sopt, m); /* XXX */
1499 				if (error != 0)
1500 					break;
1501 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1502 						    m, so, sopt);
1503 				m_freem(m); /* XXX */
1504 				break;
1505 			}
1506 
1507 			/*
1508 			 * Use of some Hop-by-Hop options or some
1509 			 * Destination options, might require special
1510 			 * privilege.  That is, normal applications
1511 			 * (without special privilege) might be forbidden
1512 			 * from setting certain options in outgoing packets,
1513 			 * and might never see certain options in received
1514 			 * packets. [RFC 2292 Section 6]
1515 			 * KAME specific note:
1516 			 *  KAME prevents non-privileged users from sending or
1517 			 *  receiving ANY hbh/dst options in order to avoid
1518 			 *  overhead of parsing options in the kernel.
1519 			 */
1520 			case IPV6_RECVHOPOPTS:
1521 			case IPV6_RECVDSTOPTS:
1522 			case IPV6_RECVRTHDRDSTOPTS:
1523 				if (td != NULL) {
1524 					error = priv_check(td,
1525 					    PRIV_NETINET_SETHDROPTS);
1526 					if (error)
1527 						break;
1528 				}
1529 				/* FALLTHROUGH */
1530 			case IPV6_UNICAST_HOPS:
1531 			case IPV6_HOPLIMIT:
1532 
1533 			case IPV6_RECVPKTINFO:
1534 			case IPV6_RECVHOPLIMIT:
1535 			case IPV6_RECVRTHDR:
1536 			case IPV6_RECVPATHMTU:
1537 			case IPV6_RECVTCLASS:
1538 			case IPV6_RECVFLOWID:
1539 #ifdef	RSS
1540 			case IPV6_RECVRSSBUCKETID:
1541 #endif
1542 			case IPV6_V6ONLY:
1543 			case IPV6_AUTOFLOWLABEL:
1544 			case IPV6_ORIGDSTADDR:
1545 			case IPV6_BINDANY:
1546 			case IPV6_BINDMULTI:
1547 #ifdef	RSS
1548 			case IPV6_RSS_LISTEN_BUCKET:
1549 #endif
1550 				if (optname == IPV6_BINDANY && td != NULL) {
1551 					error = priv_check(td,
1552 					    PRIV_NETINET_BINDANY);
1553 					if (error)
1554 						break;
1555 				}
1556 
1557 				if (optlen != sizeof(int)) {
1558 					error = EINVAL;
1559 					break;
1560 				}
1561 				error = sooptcopyin(sopt, &optval,
1562 					sizeof optval, sizeof optval);
1563 				if (error)
1564 					break;
1565 				switch (optname) {
1566 
1567 				case IPV6_UNICAST_HOPS:
1568 					if (optval < -1 || optval >= 256)
1569 						error = EINVAL;
1570 					else {
1571 						/* -1 = kernel default */
1572 						in6p->in6p_hops = optval;
1573 						if ((in6p->inp_vflag &
1574 						     INP_IPV4) != 0)
1575 							in6p->inp_ip_ttl = optval;
1576 					}
1577 					break;
1578 #define OPTSET(bit) \
1579 do { \
1580 	INP_WLOCK(in6p); \
1581 	if (optval) \
1582 		in6p->inp_flags |= (bit); \
1583 	else \
1584 		in6p->inp_flags &= ~(bit); \
1585 	INP_WUNLOCK(in6p); \
1586 } while (/*CONSTCOND*/ 0)
1587 #define OPTSET2292(bit) \
1588 do { \
1589 	INP_WLOCK(in6p); \
1590 	in6p->inp_flags |= IN6P_RFC2292; \
1591 	if (optval) \
1592 		in6p->inp_flags |= (bit); \
1593 	else \
1594 		in6p->inp_flags &= ~(bit); \
1595 	INP_WUNLOCK(in6p); \
1596 } while (/*CONSTCOND*/ 0)
1597 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1598 
1599 #define OPTSET2_N(bit, val) do {					\
1600 	if (val)							\
1601 		in6p->inp_flags2 |= bit;				\
1602 	else								\
1603 		in6p->inp_flags2 &= ~bit;				\
1604 } while (0)
1605 #define OPTSET2(bit, val) do {						\
1606 	INP_WLOCK(in6p);						\
1607 	OPTSET2_N(bit, val);						\
1608 	INP_WUNLOCK(in6p);						\
1609 } while (0)
1610 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1611 #define OPTSET2292_EXCLUSIVE(bit)					\
1612 do {									\
1613 	INP_WLOCK(in6p);						\
1614 	if (OPTBIT(IN6P_RFC2292)) {					\
1615 		error = EINVAL;						\
1616 	} else {							\
1617 		if (optval)						\
1618 			in6p->inp_flags |= (bit);			\
1619 		else							\
1620 			in6p->inp_flags &= ~(bit);			\
1621 	}								\
1622 	INP_WUNLOCK(in6p);						\
1623 } while (/*CONSTCOND*/ 0)
1624 
1625 				case IPV6_RECVPKTINFO:
1626 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1627 					break;
1628 
1629 				case IPV6_HOPLIMIT:
1630 				{
1631 					struct ip6_pktopts **optp;
1632 
1633 					/* cannot mix with RFC2292 */
1634 					if (OPTBIT(IN6P_RFC2292)) {
1635 						error = EINVAL;
1636 						break;
1637 					}
1638 					optp = &in6p->in6p_outputopts;
1639 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1640 					    (u_char *)&optval, sizeof(optval),
1641 					    optp, (td != NULL) ? td->td_ucred :
1642 					    NULL, uproto);
1643 					break;
1644 				}
1645 
1646 				case IPV6_RECVHOPLIMIT:
1647 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1648 					break;
1649 
1650 				case IPV6_RECVHOPOPTS:
1651 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1652 					break;
1653 
1654 				case IPV6_RECVDSTOPTS:
1655 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1656 					break;
1657 
1658 				case IPV6_RECVRTHDRDSTOPTS:
1659 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1660 					break;
1661 
1662 				case IPV6_RECVRTHDR:
1663 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1664 					break;
1665 
1666 				case IPV6_RECVPATHMTU:
1667 					/*
1668 					 * We ignore this option for TCP
1669 					 * sockets.
1670 					 * (RFC3542 leaves this case
1671 					 * unspecified.)
1672 					 */
1673 					if (uproto != IPPROTO_TCP)
1674 						OPTSET(IN6P_MTU);
1675 					break;
1676 
1677 				case IPV6_RECVFLOWID:
1678 					OPTSET2(INP_RECVFLOWID, optval);
1679 					break;
1680 
1681 #ifdef	RSS
1682 				case IPV6_RECVRSSBUCKETID:
1683 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1684 					break;
1685 #endif
1686 
1687 				case IPV6_V6ONLY:
1688 					/*
1689 					 * make setsockopt(IPV6_V6ONLY)
1690 					 * available only prior to bind(2).
1691 					 * see ipng mailing list, Jun 22 2001.
1692 					 */
1693 					if (in6p->inp_lport ||
1694 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1695 						error = EINVAL;
1696 						break;
1697 					}
1698 					OPTSET(IN6P_IPV6_V6ONLY);
1699 					if (optval)
1700 						in6p->inp_vflag &= ~INP_IPV4;
1701 					else
1702 						in6p->inp_vflag |= INP_IPV4;
1703 					break;
1704 				case IPV6_RECVTCLASS:
1705 					/* cannot mix with RFC2292 XXX */
1706 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1707 					break;
1708 				case IPV6_AUTOFLOWLABEL:
1709 					OPTSET(IN6P_AUTOFLOWLABEL);
1710 					break;
1711 
1712 				case IPV6_ORIGDSTADDR:
1713 					OPTSET2(INP_ORIGDSTADDR, optval);
1714 					break;
1715 				case IPV6_BINDANY:
1716 					OPTSET(INP_BINDANY);
1717 					break;
1718 
1719 				case IPV6_BINDMULTI:
1720 					OPTSET2(INP_BINDMULTI, optval);
1721 					break;
1722 #ifdef	RSS
1723 				case IPV6_RSS_LISTEN_BUCKET:
1724 					if ((optval >= 0) &&
1725 					    (optval < rss_getnumbuckets())) {
1726 						INP_WLOCK(in6p);
1727 						in6p->inp_rss_listen_bucket = optval;
1728 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1729 						INP_WUNLOCK(in6p);
1730 					} else {
1731 						error = EINVAL;
1732 					}
1733 					break;
1734 #endif
1735 				}
1736 				break;
1737 
1738 			case IPV6_TCLASS:
1739 			case IPV6_DONTFRAG:
1740 			case IPV6_USE_MIN_MTU:
1741 			case IPV6_PREFER_TEMPADDR:
1742 				if (optlen != sizeof(optval)) {
1743 					error = EINVAL;
1744 					break;
1745 				}
1746 				error = sooptcopyin(sopt, &optval,
1747 					sizeof optval, sizeof optval);
1748 				if (error)
1749 					break;
1750 				{
1751 					struct ip6_pktopts **optp;
1752 					optp = &in6p->in6p_outputopts;
1753 					error = ip6_pcbopt(optname,
1754 					    (u_char *)&optval, sizeof(optval),
1755 					    optp, (td != NULL) ? td->td_ucred :
1756 					    NULL, uproto);
1757 					break;
1758 				}
1759 
1760 			case IPV6_2292PKTINFO:
1761 			case IPV6_2292HOPLIMIT:
1762 			case IPV6_2292HOPOPTS:
1763 			case IPV6_2292DSTOPTS:
1764 			case IPV6_2292RTHDR:
1765 				/* RFC 2292 */
1766 				if (optlen != sizeof(int)) {
1767 					error = EINVAL;
1768 					break;
1769 				}
1770 				error = sooptcopyin(sopt, &optval,
1771 					sizeof optval, sizeof optval);
1772 				if (error)
1773 					break;
1774 				switch (optname) {
1775 				case IPV6_2292PKTINFO:
1776 					OPTSET2292(IN6P_PKTINFO);
1777 					break;
1778 				case IPV6_2292HOPLIMIT:
1779 					OPTSET2292(IN6P_HOPLIMIT);
1780 					break;
1781 				case IPV6_2292HOPOPTS:
1782 					/*
1783 					 * Check super-user privilege.
1784 					 * See comments for IPV6_RECVHOPOPTS.
1785 					 */
1786 					if (td != NULL) {
1787 						error = priv_check(td,
1788 						    PRIV_NETINET_SETHDROPTS);
1789 						if (error)
1790 							return (error);
1791 					}
1792 					OPTSET2292(IN6P_HOPOPTS);
1793 					break;
1794 				case IPV6_2292DSTOPTS:
1795 					if (td != NULL) {
1796 						error = priv_check(td,
1797 						    PRIV_NETINET_SETHDROPTS);
1798 						if (error)
1799 							return (error);
1800 					}
1801 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1802 					break;
1803 				case IPV6_2292RTHDR:
1804 					OPTSET2292(IN6P_RTHDR);
1805 					break;
1806 				}
1807 				break;
1808 			case IPV6_PKTINFO:
1809 			case IPV6_HOPOPTS:
1810 			case IPV6_RTHDR:
1811 			case IPV6_DSTOPTS:
1812 			case IPV6_RTHDRDSTOPTS:
1813 			case IPV6_NEXTHOP:
1814 			{
1815 				/* new advanced API (RFC3542) */
1816 				u_char *optbuf;
1817 				u_char optbuf_storage[MCLBYTES];
1818 				int optlen;
1819 				struct ip6_pktopts **optp;
1820 
1821 				/* cannot mix with RFC2292 */
1822 				if (OPTBIT(IN6P_RFC2292)) {
1823 					error = EINVAL;
1824 					break;
1825 				}
1826 
1827 				/*
1828 				 * We only ensure valsize is not too large
1829 				 * here.  Further validation will be done
1830 				 * later.
1831 				 */
1832 				error = sooptcopyin(sopt, optbuf_storage,
1833 				    sizeof(optbuf_storage), 0);
1834 				if (error)
1835 					break;
1836 				optlen = sopt->sopt_valsize;
1837 				optbuf = optbuf_storage;
1838 				optp = &in6p->in6p_outputopts;
1839 				error = ip6_pcbopt(optname, optbuf, optlen,
1840 				    optp, (td != NULL) ? td->td_ucred : NULL,
1841 				    uproto);
1842 				break;
1843 			}
1844 #undef OPTSET
1845 
1846 			case IPV6_MULTICAST_IF:
1847 			case IPV6_MULTICAST_HOPS:
1848 			case IPV6_MULTICAST_LOOP:
1849 			case IPV6_JOIN_GROUP:
1850 			case IPV6_LEAVE_GROUP:
1851 			case IPV6_MSFILTER:
1852 			case MCAST_BLOCK_SOURCE:
1853 			case MCAST_UNBLOCK_SOURCE:
1854 			case MCAST_JOIN_GROUP:
1855 			case MCAST_LEAVE_GROUP:
1856 			case MCAST_JOIN_SOURCE_GROUP:
1857 			case MCAST_LEAVE_SOURCE_GROUP:
1858 				error = ip6_setmoptions(in6p, sopt);
1859 				break;
1860 
1861 			case IPV6_PORTRANGE:
1862 				error = sooptcopyin(sopt, &optval,
1863 				    sizeof optval, sizeof optval);
1864 				if (error)
1865 					break;
1866 
1867 				INP_WLOCK(in6p);
1868 				switch (optval) {
1869 				case IPV6_PORTRANGE_DEFAULT:
1870 					in6p->inp_flags &= ~(INP_LOWPORT);
1871 					in6p->inp_flags &= ~(INP_HIGHPORT);
1872 					break;
1873 
1874 				case IPV6_PORTRANGE_HIGH:
1875 					in6p->inp_flags &= ~(INP_LOWPORT);
1876 					in6p->inp_flags |= INP_HIGHPORT;
1877 					break;
1878 
1879 				case IPV6_PORTRANGE_LOW:
1880 					in6p->inp_flags &= ~(INP_HIGHPORT);
1881 					in6p->inp_flags |= INP_LOWPORT;
1882 					break;
1883 
1884 				default:
1885 					error = EINVAL;
1886 					break;
1887 				}
1888 				INP_WUNLOCK(in6p);
1889 				break;
1890 
1891 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1892 			case IPV6_IPSEC_POLICY:
1893 				if (IPSEC_ENABLED(ipv6)) {
1894 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1895 					break;
1896 				}
1897 				/* FALLTHROUGH */
1898 #endif /* IPSEC */
1899 
1900 			default:
1901 				error = ENOPROTOOPT;
1902 				break;
1903 			}
1904 			break;
1905 
1906 		case SOPT_GET:
1907 			switch (optname) {
1908 
1909 			case IPV6_2292PKTOPTIONS:
1910 #ifdef IPV6_PKTOPTIONS
1911 			case IPV6_PKTOPTIONS:
1912 #endif
1913 				/*
1914 				 * RFC3542 (effectively) deprecated the
1915 				 * semantics of the 2292-style pktoptions.
1916 				 * Since it was not reliable in nature (i.e.,
1917 				 * applications had to expect the lack of some
1918 				 * information after all), it would make sense
1919 				 * to simplify this part by always returning
1920 				 * empty data.
1921 				 */
1922 				sopt->sopt_valsize = 0;
1923 				break;
1924 
1925 			case IPV6_RECVHOPOPTS:
1926 			case IPV6_RECVDSTOPTS:
1927 			case IPV6_RECVRTHDRDSTOPTS:
1928 			case IPV6_UNICAST_HOPS:
1929 			case IPV6_RECVPKTINFO:
1930 			case IPV6_RECVHOPLIMIT:
1931 			case IPV6_RECVRTHDR:
1932 			case IPV6_RECVPATHMTU:
1933 
1934 			case IPV6_V6ONLY:
1935 			case IPV6_PORTRANGE:
1936 			case IPV6_RECVTCLASS:
1937 			case IPV6_AUTOFLOWLABEL:
1938 			case IPV6_BINDANY:
1939 			case IPV6_FLOWID:
1940 			case IPV6_FLOWTYPE:
1941 			case IPV6_RECVFLOWID:
1942 #ifdef	RSS
1943 			case IPV6_RSSBUCKETID:
1944 			case IPV6_RECVRSSBUCKETID:
1945 #endif
1946 			case IPV6_BINDMULTI:
1947 				switch (optname) {
1948 
1949 				case IPV6_RECVHOPOPTS:
1950 					optval = OPTBIT(IN6P_HOPOPTS);
1951 					break;
1952 
1953 				case IPV6_RECVDSTOPTS:
1954 					optval = OPTBIT(IN6P_DSTOPTS);
1955 					break;
1956 
1957 				case IPV6_RECVRTHDRDSTOPTS:
1958 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1959 					break;
1960 
1961 				case IPV6_UNICAST_HOPS:
1962 					optval = in6p->in6p_hops;
1963 					break;
1964 
1965 				case IPV6_RECVPKTINFO:
1966 					optval = OPTBIT(IN6P_PKTINFO);
1967 					break;
1968 
1969 				case IPV6_RECVHOPLIMIT:
1970 					optval = OPTBIT(IN6P_HOPLIMIT);
1971 					break;
1972 
1973 				case IPV6_RECVRTHDR:
1974 					optval = OPTBIT(IN6P_RTHDR);
1975 					break;
1976 
1977 				case IPV6_RECVPATHMTU:
1978 					optval = OPTBIT(IN6P_MTU);
1979 					break;
1980 
1981 				case IPV6_V6ONLY:
1982 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1983 					break;
1984 
1985 				case IPV6_PORTRANGE:
1986 				    {
1987 					int flags;
1988 					flags = in6p->inp_flags;
1989 					if (flags & INP_HIGHPORT)
1990 						optval = IPV6_PORTRANGE_HIGH;
1991 					else if (flags & INP_LOWPORT)
1992 						optval = IPV6_PORTRANGE_LOW;
1993 					else
1994 						optval = 0;
1995 					break;
1996 				    }
1997 				case IPV6_RECVTCLASS:
1998 					optval = OPTBIT(IN6P_TCLASS);
1999 					break;
2000 
2001 				case IPV6_AUTOFLOWLABEL:
2002 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2003 					break;
2004 
2005 				case IPV6_ORIGDSTADDR:
2006 					optval = OPTBIT2(INP_ORIGDSTADDR);
2007 					break;
2008 
2009 				case IPV6_BINDANY:
2010 					optval = OPTBIT(INP_BINDANY);
2011 					break;
2012 
2013 				case IPV6_FLOWID:
2014 					optval = in6p->inp_flowid;
2015 					break;
2016 
2017 				case IPV6_FLOWTYPE:
2018 					optval = in6p->inp_flowtype;
2019 					break;
2020 
2021 				case IPV6_RECVFLOWID:
2022 					optval = OPTBIT2(INP_RECVFLOWID);
2023 					break;
2024 #ifdef	RSS
2025 				case IPV6_RSSBUCKETID:
2026 					retval =
2027 					    rss_hash2bucket(in6p->inp_flowid,
2028 					    in6p->inp_flowtype,
2029 					    &rss_bucket);
2030 					if (retval == 0)
2031 						optval = rss_bucket;
2032 					else
2033 						error = EINVAL;
2034 					break;
2035 
2036 				case IPV6_RECVRSSBUCKETID:
2037 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2038 					break;
2039 #endif
2040 
2041 				case IPV6_BINDMULTI:
2042 					optval = OPTBIT2(INP_BINDMULTI);
2043 					break;
2044 
2045 				}
2046 				if (error)
2047 					break;
2048 				error = sooptcopyout(sopt, &optval,
2049 					sizeof optval);
2050 				break;
2051 
2052 			case IPV6_PATHMTU:
2053 			{
2054 				u_long pmtu = 0;
2055 				struct ip6_mtuinfo mtuinfo;
2056 				struct in6_addr addr;
2057 
2058 				if (!(so->so_state & SS_ISCONNECTED))
2059 					return (ENOTCONN);
2060 				/*
2061 				 * XXX: we dot not consider the case of source
2062 				 * routing, or optional information to specify
2063 				 * the outgoing interface.
2064 				 * Copy faddr out of in6p to avoid holding lock
2065 				 * on inp during route lookup.
2066 				 */
2067 				INP_RLOCK(in6p);
2068 				bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2069 				INP_RUNLOCK(in6p);
2070 				error = ip6_getpmtu_ctl(so->so_fibnum,
2071 				    &addr, &pmtu);
2072 				if (error)
2073 					break;
2074 				if (pmtu > IPV6_MAXPACKET)
2075 					pmtu = IPV6_MAXPACKET;
2076 
2077 				bzero(&mtuinfo, sizeof(mtuinfo));
2078 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2079 				optdata = (void *)&mtuinfo;
2080 				optdatalen = sizeof(mtuinfo);
2081 				error = sooptcopyout(sopt, optdata,
2082 				    optdatalen);
2083 				break;
2084 			}
2085 
2086 			case IPV6_2292PKTINFO:
2087 			case IPV6_2292HOPLIMIT:
2088 			case IPV6_2292HOPOPTS:
2089 			case IPV6_2292RTHDR:
2090 			case IPV6_2292DSTOPTS:
2091 				switch (optname) {
2092 				case IPV6_2292PKTINFO:
2093 					optval = OPTBIT(IN6P_PKTINFO);
2094 					break;
2095 				case IPV6_2292HOPLIMIT:
2096 					optval = OPTBIT(IN6P_HOPLIMIT);
2097 					break;
2098 				case IPV6_2292HOPOPTS:
2099 					optval = OPTBIT(IN6P_HOPOPTS);
2100 					break;
2101 				case IPV6_2292RTHDR:
2102 					optval = OPTBIT(IN6P_RTHDR);
2103 					break;
2104 				case IPV6_2292DSTOPTS:
2105 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2106 					break;
2107 				}
2108 				error = sooptcopyout(sopt, &optval,
2109 				    sizeof optval);
2110 				break;
2111 			case IPV6_PKTINFO:
2112 			case IPV6_HOPOPTS:
2113 			case IPV6_RTHDR:
2114 			case IPV6_DSTOPTS:
2115 			case IPV6_RTHDRDSTOPTS:
2116 			case IPV6_NEXTHOP:
2117 			case IPV6_TCLASS:
2118 			case IPV6_DONTFRAG:
2119 			case IPV6_USE_MIN_MTU:
2120 			case IPV6_PREFER_TEMPADDR:
2121 				error = ip6_getpcbopt(in6p, optname, sopt);
2122 				break;
2123 
2124 			case IPV6_MULTICAST_IF:
2125 			case IPV6_MULTICAST_HOPS:
2126 			case IPV6_MULTICAST_LOOP:
2127 			case IPV6_MSFILTER:
2128 				error = ip6_getmoptions(in6p, sopt);
2129 				break;
2130 
2131 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2132 			case IPV6_IPSEC_POLICY:
2133 				if (IPSEC_ENABLED(ipv6)) {
2134 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2135 					break;
2136 				}
2137 				/* FALLTHROUGH */
2138 #endif /* IPSEC */
2139 			default:
2140 				error = ENOPROTOOPT;
2141 				break;
2142 			}
2143 			break;
2144 		}
2145 	}
2146 	return (error);
2147 }
2148 
2149 int
2150 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2151 {
2152 	int error = 0, optval, optlen;
2153 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2154 	struct inpcb *in6p = sotoinpcb(so);
2155 	int level, op, optname;
2156 
2157 	level = sopt->sopt_level;
2158 	op = sopt->sopt_dir;
2159 	optname = sopt->sopt_name;
2160 	optlen = sopt->sopt_valsize;
2161 
2162 	if (level != IPPROTO_IPV6) {
2163 		return (EINVAL);
2164 	}
2165 
2166 	switch (optname) {
2167 	case IPV6_CHECKSUM:
2168 		/*
2169 		 * For ICMPv6 sockets, no modification allowed for checksum
2170 		 * offset, permit "no change" values to help existing apps.
2171 		 *
2172 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2173 		 * for an ICMPv6 socket will fail."
2174 		 * The current behavior does not meet RFC3542.
2175 		 */
2176 		switch (op) {
2177 		case SOPT_SET:
2178 			if (optlen != sizeof(int)) {
2179 				error = EINVAL;
2180 				break;
2181 			}
2182 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2183 					    sizeof(optval));
2184 			if (error)
2185 				break;
2186 			if ((optval % 2) != 0) {
2187 				/* the API assumes even offset values */
2188 				error = EINVAL;
2189 			} else if (so->so_proto->pr_protocol ==
2190 			    IPPROTO_ICMPV6) {
2191 				if (optval != icmp6off)
2192 					error = EINVAL;
2193 			} else
2194 				in6p->in6p_cksum = optval;
2195 			break;
2196 
2197 		case SOPT_GET:
2198 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2199 				optval = icmp6off;
2200 			else
2201 				optval = in6p->in6p_cksum;
2202 
2203 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2204 			break;
2205 
2206 		default:
2207 			error = EINVAL;
2208 			break;
2209 		}
2210 		break;
2211 
2212 	default:
2213 		error = ENOPROTOOPT;
2214 		break;
2215 	}
2216 
2217 	return (error);
2218 }
2219 
2220 /*
2221  * Set up IP6 options in pcb for insertion in output packets or
2222  * specifying behavior of outgoing packets.
2223  */
2224 static int
2225 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2226     struct socket *so, struct sockopt *sopt)
2227 {
2228 	struct ip6_pktopts *opt = *pktopt;
2229 	int error = 0;
2230 	struct thread *td = sopt->sopt_td;
2231 
2232 	/* turn off any old options. */
2233 	if (opt) {
2234 #ifdef DIAGNOSTIC
2235 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2236 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2237 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2238 			printf("ip6_pcbopts: all specified options are cleared.\n");
2239 #endif
2240 		ip6_clearpktopts(opt, -1);
2241 	} else
2242 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2243 	*pktopt = NULL;
2244 
2245 	if (!m || m->m_len == 0) {
2246 		/*
2247 		 * Only turning off any previous options, regardless of
2248 		 * whether the opt is just created or given.
2249 		 */
2250 		free(opt, M_IP6OPT);
2251 		return (0);
2252 	}
2253 
2254 	/*  set options specified by user. */
2255 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2256 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2257 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2258 		free(opt, M_IP6OPT);
2259 		return (error);
2260 	}
2261 	*pktopt = opt;
2262 	return (0);
2263 }
2264 
2265 /*
2266  * initialize ip6_pktopts.  beware that there are non-zero default values in
2267  * the struct.
2268  */
2269 void
2270 ip6_initpktopts(struct ip6_pktopts *opt)
2271 {
2272 
2273 	bzero(opt, sizeof(*opt));
2274 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2275 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2276 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2277 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2278 }
2279 
2280 static int
2281 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2282     struct ucred *cred, int uproto)
2283 {
2284 	struct ip6_pktopts *opt;
2285 
2286 	if (*pktopt == NULL) {
2287 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2288 		    M_WAITOK);
2289 		ip6_initpktopts(*pktopt);
2290 	}
2291 	opt = *pktopt;
2292 
2293 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2294 }
2295 
2296 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2297 	if (pktopt && pktopt->field) {						\
2298 		INP_RUNLOCK(in6p);						\
2299 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2300 		malloc_optdata = true;						\
2301 		INP_RLOCK(in6p);						\
2302 		if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2303 			INP_RUNLOCK(in6p);					\
2304 			free(optdata, M_TEMP);					\
2305 			return (ECONNRESET);					\
2306 		}								\
2307 		pktopt = in6p->in6p_outputopts;					\
2308 		if (pktopt && pktopt->field) {					\
2309 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2310 			bcopy(&pktopt->field, optdata, optdatalen);		\
2311 		} else {							\
2312 			free(optdata, M_TEMP);					\
2313 			optdata = NULL;						\
2314 			malloc_optdata = false;					\
2315 		}								\
2316 	}									\
2317 } while(0)
2318 
2319 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2320 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2321 
2322 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2323 	pktopt->field->sa_len)
2324 
2325 static int
2326 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2327 {
2328 	void *optdata = NULL;
2329 	bool malloc_optdata = false;
2330 	int optdatalen = 0;
2331 	int error = 0;
2332 	struct in6_pktinfo null_pktinfo;
2333 	int deftclass = 0, on;
2334 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2335 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2336 	struct ip6_pktopts *pktopt;
2337 
2338 	INP_RLOCK(in6p);
2339 	pktopt = in6p->in6p_outputopts;
2340 
2341 	switch (optname) {
2342 	case IPV6_PKTINFO:
2343 		optdata = (void *)&null_pktinfo;
2344 		if (pktopt && pktopt->ip6po_pktinfo) {
2345 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2346 			    sizeof(null_pktinfo));
2347 			in6_clearscope(&null_pktinfo.ipi6_addr);
2348 		} else {
2349 			/* XXX: we don't have to do this every time... */
2350 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2351 		}
2352 		optdatalen = sizeof(struct in6_pktinfo);
2353 		break;
2354 	case IPV6_TCLASS:
2355 		if (pktopt && pktopt->ip6po_tclass >= 0)
2356 			deftclass = pktopt->ip6po_tclass;
2357 		optdata = (void *)&deftclass;
2358 		optdatalen = sizeof(int);
2359 		break;
2360 	case IPV6_HOPOPTS:
2361 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2362 		break;
2363 	case IPV6_RTHDR:
2364 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2365 		break;
2366 	case IPV6_RTHDRDSTOPTS:
2367 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2368 		break;
2369 	case IPV6_DSTOPTS:
2370 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2371 		break;
2372 	case IPV6_NEXTHOP:
2373 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2374 		break;
2375 	case IPV6_USE_MIN_MTU:
2376 		if (pktopt)
2377 			defminmtu = pktopt->ip6po_minmtu;
2378 		optdata = (void *)&defminmtu;
2379 		optdatalen = sizeof(int);
2380 		break;
2381 	case IPV6_DONTFRAG:
2382 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2383 			on = 1;
2384 		else
2385 			on = 0;
2386 		optdata = (void *)&on;
2387 		optdatalen = sizeof(on);
2388 		break;
2389 	case IPV6_PREFER_TEMPADDR:
2390 		if (pktopt)
2391 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2392 		optdata = (void *)&defpreftemp;
2393 		optdatalen = sizeof(int);
2394 		break;
2395 	default:		/* should not happen */
2396 #ifdef DIAGNOSTIC
2397 		panic("ip6_getpcbopt: unexpected option\n");
2398 #endif
2399 		INP_RUNLOCK(in6p);
2400 		return (ENOPROTOOPT);
2401 	}
2402 	INP_RUNLOCK(in6p);
2403 
2404 	error = sooptcopyout(sopt, optdata, optdatalen);
2405 	if (malloc_optdata)
2406 		free(optdata, M_TEMP);
2407 
2408 	return (error);
2409 }
2410 
2411 void
2412 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2413 {
2414 	if (pktopt == NULL)
2415 		return;
2416 
2417 	if (optname == -1 || optname == IPV6_PKTINFO) {
2418 		if (pktopt->ip6po_pktinfo)
2419 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2420 		pktopt->ip6po_pktinfo = NULL;
2421 	}
2422 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2423 		pktopt->ip6po_hlim = -1;
2424 	if (optname == -1 || optname == IPV6_TCLASS)
2425 		pktopt->ip6po_tclass = -1;
2426 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2427 		if (pktopt->ip6po_nextroute.ro_rt) {
2428 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2429 			pktopt->ip6po_nextroute.ro_rt = NULL;
2430 		}
2431 		if (pktopt->ip6po_nexthop)
2432 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2433 		pktopt->ip6po_nexthop = NULL;
2434 	}
2435 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2436 		if (pktopt->ip6po_hbh)
2437 			free(pktopt->ip6po_hbh, M_IP6OPT);
2438 		pktopt->ip6po_hbh = NULL;
2439 	}
2440 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2441 		if (pktopt->ip6po_dest1)
2442 			free(pktopt->ip6po_dest1, M_IP6OPT);
2443 		pktopt->ip6po_dest1 = NULL;
2444 	}
2445 	if (optname == -1 || optname == IPV6_RTHDR) {
2446 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2447 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2448 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2449 		if (pktopt->ip6po_route.ro_rt) {
2450 			RTFREE(pktopt->ip6po_route.ro_rt);
2451 			pktopt->ip6po_route.ro_rt = NULL;
2452 		}
2453 	}
2454 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2455 		if (pktopt->ip6po_dest2)
2456 			free(pktopt->ip6po_dest2, M_IP6OPT);
2457 		pktopt->ip6po_dest2 = NULL;
2458 	}
2459 }
2460 
2461 #define PKTOPT_EXTHDRCPY(type) \
2462 do {\
2463 	if (src->type) {\
2464 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2465 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2466 		if (dst->type == NULL)\
2467 			goto bad;\
2468 		bcopy(src->type, dst->type, hlen);\
2469 	}\
2470 } while (/*CONSTCOND*/ 0)
2471 
2472 static int
2473 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2474 {
2475 	if (dst == NULL || src == NULL)  {
2476 		printf("ip6_clearpktopts: invalid argument\n");
2477 		return (EINVAL);
2478 	}
2479 
2480 	dst->ip6po_hlim = src->ip6po_hlim;
2481 	dst->ip6po_tclass = src->ip6po_tclass;
2482 	dst->ip6po_flags = src->ip6po_flags;
2483 	dst->ip6po_minmtu = src->ip6po_minmtu;
2484 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2485 	if (src->ip6po_pktinfo) {
2486 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2487 		    M_IP6OPT, canwait);
2488 		if (dst->ip6po_pktinfo == NULL)
2489 			goto bad;
2490 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2491 	}
2492 	if (src->ip6po_nexthop) {
2493 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2494 		    M_IP6OPT, canwait);
2495 		if (dst->ip6po_nexthop == NULL)
2496 			goto bad;
2497 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2498 		    src->ip6po_nexthop->sa_len);
2499 	}
2500 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2501 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2502 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2503 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2504 	return (0);
2505 
2506   bad:
2507 	ip6_clearpktopts(dst, -1);
2508 	return (ENOBUFS);
2509 }
2510 #undef PKTOPT_EXTHDRCPY
2511 
2512 struct ip6_pktopts *
2513 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2514 {
2515 	int error;
2516 	struct ip6_pktopts *dst;
2517 
2518 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2519 	if (dst == NULL)
2520 		return (NULL);
2521 	ip6_initpktopts(dst);
2522 
2523 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2524 		free(dst, M_IP6OPT);
2525 		return (NULL);
2526 	}
2527 
2528 	return (dst);
2529 }
2530 
2531 void
2532 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2533 {
2534 	if (pktopt == NULL)
2535 		return;
2536 
2537 	ip6_clearpktopts(pktopt, -1);
2538 
2539 	free(pktopt, M_IP6OPT);
2540 }
2541 
2542 /*
2543  * Set IPv6 outgoing packet options based on advanced API.
2544  */
2545 int
2546 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2547     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2548 {
2549 	struct cmsghdr *cm = NULL;
2550 
2551 	if (control == NULL || opt == NULL)
2552 		return (EINVAL);
2553 
2554 	ip6_initpktopts(opt);
2555 	if (stickyopt) {
2556 		int error;
2557 
2558 		/*
2559 		 * If stickyopt is provided, make a local copy of the options
2560 		 * for this particular packet, then override them by ancillary
2561 		 * objects.
2562 		 * XXX: copypktopts() does not copy the cached route to a next
2563 		 * hop (if any).  This is not very good in terms of efficiency,
2564 		 * but we can allow this since this option should be rarely
2565 		 * used.
2566 		 */
2567 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2568 			return (error);
2569 	}
2570 
2571 	/*
2572 	 * XXX: Currently, we assume all the optional information is stored
2573 	 * in a single mbuf.
2574 	 */
2575 	if (control->m_next)
2576 		return (EINVAL);
2577 
2578 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2579 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2580 		int error;
2581 
2582 		if (control->m_len < CMSG_LEN(0))
2583 			return (EINVAL);
2584 
2585 		cm = mtod(control, struct cmsghdr *);
2586 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2587 			return (EINVAL);
2588 		if (cm->cmsg_level != IPPROTO_IPV6)
2589 			continue;
2590 
2591 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2592 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2593 		if (error)
2594 			return (error);
2595 	}
2596 
2597 	return (0);
2598 }
2599 
2600 /*
2601  * Set a particular packet option, as a sticky option or an ancillary data
2602  * item.  "len" can be 0 only when it's a sticky option.
2603  * We have 4 cases of combination of "sticky" and "cmsg":
2604  * "sticky=0, cmsg=0": impossible
2605  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2606  * "sticky=1, cmsg=0": RFC3542 socket option
2607  * "sticky=1, cmsg=1": RFC2292 socket option
2608  */
2609 static int
2610 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2611     struct ucred *cred, int sticky, int cmsg, int uproto)
2612 {
2613 	int minmtupolicy, preftemp;
2614 	int error;
2615 
2616 	if (!sticky && !cmsg) {
2617 #ifdef DIAGNOSTIC
2618 		printf("ip6_setpktopt: impossible case\n");
2619 #endif
2620 		return (EINVAL);
2621 	}
2622 
2623 	/*
2624 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2625 	 * not be specified in the context of RFC3542.  Conversely,
2626 	 * RFC3542 types should not be specified in the context of RFC2292.
2627 	 */
2628 	if (!cmsg) {
2629 		switch (optname) {
2630 		case IPV6_2292PKTINFO:
2631 		case IPV6_2292HOPLIMIT:
2632 		case IPV6_2292NEXTHOP:
2633 		case IPV6_2292HOPOPTS:
2634 		case IPV6_2292DSTOPTS:
2635 		case IPV6_2292RTHDR:
2636 		case IPV6_2292PKTOPTIONS:
2637 			return (ENOPROTOOPT);
2638 		}
2639 	}
2640 	if (sticky && cmsg) {
2641 		switch (optname) {
2642 		case IPV6_PKTINFO:
2643 		case IPV6_HOPLIMIT:
2644 		case IPV6_NEXTHOP:
2645 		case IPV6_HOPOPTS:
2646 		case IPV6_DSTOPTS:
2647 		case IPV6_RTHDRDSTOPTS:
2648 		case IPV6_RTHDR:
2649 		case IPV6_USE_MIN_MTU:
2650 		case IPV6_DONTFRAG:
2651 		case IPV6_TCLASS:
2652 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2653 			return (ENOPROTOOPT);
2654 		}
2655 	}
2656 
2657 	switch (optname) {
2658 	case IPV6_2292PKTINFO:
2659 	case IPV6_PKTINFO:
2660 	{
2661 		struct ifnet *ifp = NULL;
2662 		struct in6_pktinfo *pktinfo;
2663 
2664 		if (len != sizeof(struct in6_pktinfo))
2665 			return (EINVAL);
2666 
2667 		pktinfo = (struct in6_pktinfo *)buf;
2668 
2669 		/*
2670 		 * An application can clear any sticky IPV6_PKTINFO option by
2671 		 * doing a "regular" setsockopt with ipi6_addr being
2672 		 * in6addr_any and ipi6_ifindex being zero.
2673 		 * [RFC 3542, Section 6]
2674 		 */
2675 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2676 		    pktinfo->ipi6_ifindex == 0 &&
2677 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678 			ip6_clearpktopts(opt, optname);
2679 			break;
2680 		}
2681 
2682 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2683 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2684 			return (EINVAL);
2685 		}
2686 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2687 			return (EINVAL);
2688 		/* validate the interface index if specified. */
2689 		if (pktinfo->ipi6_ifindex > V_if_index)
2690 			 return (ENXIO);
2691 		if (pktinfo->ipi6_ifindex) {
2692 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2693 			if (ifp == NULL)
2694 				return (ENXIO);
2695 		}
2696 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2697 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2698 			return (ENETDOWN);
2699 
2700 		if (ifp != NULL &&
2701 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2702 			struct in6_ifaddr *ia;
2703 
2704 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2705 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2706 			if (ia == NULL)
2707 				return (EADDRNOTAVAIL);
2708 			ifa_free(&ia->ia_ifa);
2709 		}
2710 		/*
2711 		 * We store the address anyway, and let in6_selectsrc()
2712 		 * validate the specified address.  This is because ipi6_addr
2713 		 * may not have enough information about its scope zone, and
2714 		 * we may need additional information (such as outgoing
2715 		 * interface or the scope zone of a destination address) to
2716 		 * disambiguate the scope.
2717 		 * XXX: the delay of the validation may confuse the
2718 		 * application when it is used as a sticky option.
2719 		 */
2720 		if (opt->ip6po_pktinfo == NULL) {
2721 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2722 			    M_IP6OPT, M_NOWAIT);
2723 			if (opt->ip6po_pktinfo == NULL)
2724 				return (ENOBUFS);
2725 		}
2726 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2727 		break;
2728 	}
2729 
2730 	case IPV6_2292HOPLIMIT:
2731 	case IPV6_HOPLIMIT:
2732 	{
2733 		int *hlimp;
2734 
2735 		/*
2736 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2737 		 * to simplify the ordering among hoplimit options.
2738 		 */
2739 		if (optname == IPV6_HOPLIMIT && sticky)
2740 			return (ENOPROTOOPT);
2741 
2742 		if (len != sizeof(int))
2743 			return (EINVAL);
2744 		hlimp = (int *)buf;
2745 		if (*hlimp < -1 || *hlimp > 255)
2746 			return (EINVAL);
2747 
2748 		opt->ip6po_hlim = *hlimp;
2749 		break;
2750 	}
2751 
2752 	case IPV6_TCLASS:
2753 	{
2754 		int tclass;
2755 
2756 		if (len != sizeof(int))
2757 			return (EINVAL);
2758 		tclass = *(int *)buf;
2759 		if (tclass < -1 || tclass > 255)
2760 			return (EINVAL);
2761 
2762 		opt->ip6po_tclass = tclass;
2763 		break;
2764 	}
2765 
2766 	case IPV6_2292NEXTHOP:
2767 	case IPV6_NEXTHOP:
2768 		if (cred != NULL) {
2769 			error = priv_check_cred(cred,
2770 			    PRIV_NETINET_SETHDROPTS, 0);
2771 			if (error)
2772 				return (error);
2773 		}
2774 
2775 		if (len == 0) {	/* just remove the option */
2776 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2777 			break;
2778 		}
2779 
2780 		/* check if cmsg_len is large enough for sa_len */
2781 		if (len < sizeof(struct sockaddr) || len < *buf)
2782 			return (EINVAL);
2783 
2784 		switch (((struct sockaddr *)buf)->sa_family) {
2785 		case AF_INET6:
2786 		{
2787 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2788 			int error;
2789 
2790 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2791 				return (EINVAL);
2792 
2793 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2794 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2795 				return (EINVAL);
2796 			}
2797 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2798 			    != 0) {
2799 				return (error);
2800 			}
2801 			break;
2802 		}
2803 		case AF_LINK:	/* should eventually be supported */
2804 		default:
2805 			return (EAFNOSUPPORT);
2806 		}
2807 
2808 		/* turn off the previous option, then set the new option. */
2809 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2810 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2811 		if (opt->ip6po_nexthop == NULL)
2812 			return (ENOBUFS);
2813 		bcopy(buf, opt->ip6po_nexthop, *buf);
2814 		break;
2815 
2816 	case IPV6_2292HOPOPTS:
2817 	case IPV6_HOPOPTS:
2818 	{
2819 		struct ip6_hbh *hbh;
2820 		int hbhlen;
2821 
2822 		/*
2823 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2824 		 * options, since per-option restriction has too much
2825 		 * overhead.
2826 		 */
2827 		if (cred != NULL) {
2828 			error = priv_check_cred(cred,
2829 			    PRIV_NETINET_SETHDROPTS, 0);
2830 			if (error)
2831 				return (error);
2832 		}
2833 
2834 		if (len == 0) {
2835 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2836 			break;	/* just remove the option */
2837 		}
2838 
2839 		/* message length validation */
2840 		if (len < sizeof(struct ip6_hbh))
2841 			return (EINVAL);
2842 		hbh = (struct ip6_hbh *)buf;
2843 		hbhlen = (hbh->ip6h_len + 1) << 3;
2844 		if (len != hbhlen)
2845 			return (EINVAL);
2846 
2847 		/* turn off the previous option, then set the new option. */
2848 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2849 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2850 		if (opt->ip6po_hbh == NULL)
2851 			return (ENOBUFS);
2852 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2853 
2854 		break;
2855 	}
2856 
2857 	case IPV6_2292DSTOPTS:
2858 	case IPV6_DSTOPTS:
2859 	case IPV6_RTHDRDSTOPTS:
2860 	{
2861 		struct ip6_dest *dest, **newdest = NULL;
2862 		int destlen;
2863 
2864 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2865 			error = priv_check_cred(cred,
2866 			    PRIV_NETINET_SETHDROPTS, 0);
2867 			if (error)
2868 				return (error);
2869 		}
2870 
2871 		if (len == 0) {
2872 			ip6_clearpktopts(opt, optname);
2873 			break;	/* just remove the option */
2874 		}
2875 
2876 		/* message length validation */
2877 		if (len < sizeof(struct ip6_dest))
2878 			return (EINVAL);
2879 		dest = (struct ip6_dest *)buf;
2880 		destlen = (dest->ip6d_len + 1) << 3;
2881 		if (len != destlen)
2882 			return (EINVAL);
2883 
2884 		/*
2885 		 * Determine the position that the destination options header
2886 		 * should be inserted; before or after the routing header.
2887 		 */
2888 		switch (optname) {
2889 		case IPV6_2292DSTOPTS:
2890 			/*
2891 			 * The old advacned API is ambiguous on this point.
2892 			 * Our approach is to determine the position based
2893 			 * according to the existence of a routing header.
2894 			 * Note, however, that this depends on the order of the
2895 			 * extension headers in the ancillary data; the 1st
2896 			 * part of the destination options header must appear
2897 			 * before the routing header in the ancillary data,
2898 			 * too.
2899 			 * RFC3542 solved the ambiguity by introducing
2900 			 * separate ancillary data or option types.
2901 			 */
2902 			if (opt->ip6po_rthdr == NULL)
2903 				newdest = &opt->ip6po_dest1;
2904 			else
2905 				newdest = &opt->ip6po_dest2;
2906 			break;
2907 		case IPV6_RTHDRDSTOPTS:
2908 			newdest = &opt->ip6po_dest1;
2909 			break;
2910 		case IPV6_DSTOPTS:
2911 			newdest = &opt->ip6po_dest2;
2912 			break;
2913 		}
2914 
2915 		/* turn off the previous option, then set the new option. */
2916 		ip6_clearpktopts(opt, optname);
2917 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2918 		if (*newdest == NULL)
2919 			return (ENOBUFS);
2920 		bcopy(dest, *newdest, destlen);
2921 
2922 		break;
2923 	}
2924 
2925 	case IPV6_2292RTHDR:
2926 	case IPV6_RTHDR:
2927 	{
2928 		struct ip6_rthdr *rth;
2929 		int rthlen;
2930 
2931 		if (len == 0) {
2932 			ip6_clearpktopts(opt, IPV6_RTHDR);
2933 			break;	/* just remove the option */
2934 		}
2935 
2936 		/* message length validation */
2937 		if (len < sizeof(struct ip6_rthdr))
2938 			return (EINVAL);
2939 		rth = (struct ip6_rthdr *)buf;
2940 		rthlen = (rth->ip6r_len + 1) << 3;
2941 		if (len != rthlen)
2942 			return (EINVAL);
2943 
2944 		switch (rth->ip6r_type) {
2945 		case IPV6_RTHDR_TYPE_0:
2946 			if (rth->ip6r_len == 0)	/* must contain one addr */
2947 				return (EINVAL);
2948 			if (rth->ip6r_len % 2) /* length must be even */
2949 				return (EINVAL);
2950 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2951 				return (EINVAL);
2952 			break;
2953 		default:
2954 			return (EINVAL);	/* not supported */
2955 		}
2956 
2957 		/* turn off the previous option */
2958 		ip6_clearpktopts(opt, IPV6_RTHDR);
2959 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2960 		if (opt->ip6po_rthdr == NULL)
2961 			return (ENOBUFS);
2962 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2963 
2964 		break;
2965 	}
2966 
2967 	case IPV6_USE_MIN_MTU:
2968 		if (len != sizeof(int))
2969 			return (EINVAL);
2970 		minmtupolicy = *(int *)buf;
2971 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2972 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2973 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2974 			return (EINVAL);
2975 		}
2976 		opt->ip6po_minmtu = minmtupolicy;
2977 		break;
2978 
2979 	case IPV6_DONTFRAG:
2980 		if (len != sizeof(int))
2981 			return (EINVAL);
2982 
2983 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2984 			/*
2985 			 * we ignore this option for TCP sockets.
2986 			 * (RFC3542 leaves this case unspecified.)
2987 			 */
2988 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2989 		} else
2990 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2991 		break;
2992 
2993 	case IPV6_PREFER_TEMPADDR:
2994 		if (len != sizeof(int))
2995 			return (EINVAL);
2996 		preftemp = *(int *)buf;
2997 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2998 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2999 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3000 			return (EINVAL);
3001 		}
3002 		opt->ip6po_prefer_tempaddr = preftemp;
3003 		break;
3004 
3005 	default:
3006 		return (ENOPROTOOPT);
3007 	} /* end of switch */
3008 
3009 	return (0);
3010 }
3011 
3012 /*
3013  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3014  * packet to the input queue of a specified interface.  Note that this
3015  * calls the output routine of the loopback "driver", but with an interface
3016  * pointer that might NOT be &loif -- easier than replicating that code here.
3017  */
3018 void
3019 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3020 {
3021 	struct mbuf *copym;
3022 	struct ip6_hdr *ip6;
3023 
3024 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3025 	if (copym == NULL)
3026 		return;
3027 
3028 	/*
3029 	 * Make sure to deep-copy IPv6 header portion in case the data
3030 	 * is in an mbuf cluster, so that we can safely override the IPv6
3031 	 * header portion later.
3032 	 */
3033 	if (!M_WRITABLE(copym) ||
3034 	    copym->m_len < sizeof(struct ip6_hdr)) {
3035 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3036 		if (copym == NULL)
3037 			return;
3038 	}
3039 	ip6 = mtod(copym, struct ip6_hdr *);
3040 	/*
3041 	 * clear embedded scope identifiers if necessary.
3042 	 * in6_clearscope will touch the addresses only when necessary.
3043 	 */
3044 	in6_clearscope(&ip6->ip6_src);
3045 	in6_clearscope(&ip6->ip6_dst);
3046 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3047 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3048 		    CSUM_PSEUDO_HDR;
3049 		copym->m_pkthdr.csum_data = 0xffff;
3050 	}
3051 	if_simloop(ifp, copym, AF_INET6, 0);
3052 }
3053 
3054 /*
3055  * Chop IPv6 header off from the payload.
3056  */
3057 static int
3058 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3059 {
3060 	struct mbuf *mh;
3061 	struct ip6_hdr *ip6;
3062 
3063 	ip6 = mtod(m, struct ip6_hdr *);
3064 	if (m->m_len > sizeof(*ip6)) {
3065 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3066 		if (mh == NULL) {
3067 			m_freem(m);
3068 			return ENOBUFS;
3069 		}
3070 		m_move_pkthdr(mh, m);
3071 		M_ALIGN(mh, sizeof(*ip6));
3072 		m->m_len -= sizeof(*ip6);
3073 		m->m_data += sizeof(*ip6);
3074 		mh->m_next = m;
3075 		m = mh;
3076 		m->m_len = sizeof(*ip6);
3077 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3078 	}
3079 	exthdrs->ip6e_ip6 = m;
3080 	return 0;
3081 }
3082 
3083 /*
3084  * Compute IPv6 extension header length.
3085  */
3086 int
3087 ip6_optlen(struct inpcb *in6p)
3088 {
3089 	int len;
3090 
3091 	if (!in6p->in6p_outputopts)
3092 		return 0;
3093 
3094 	len = 0;
3095 #define elen(x) \
3096     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3097 
3098 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3099 	if (in6p->in6p_outputopts->ip6po_rthdr)
3100 		/* dest1 is valid with rthdr only */
3101 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3102 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3103 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3104 	return len;
3105 #undef elen
3106 }
3107