xref: /freebsd/sys/netinet6/ip6_output.c (revision f173c2b77e0287807852bb7ba982643f33bdf8a0)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ratelimit.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_llatbl.h>
92 #include <net/netisr.h>
93 #include <net/route.h>
94 #include <net/pfil.h>
95 #include <net/rss_config.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip_var.h>
101 #include <netinet6/in6_fib.h>
102 #include <netinet6/in6_var.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp6.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/tcp_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/in6_rss.h>
110 
111 #include <netipsec/ipsec_support.h>
112 #ifdef SCTP
113 #include <netinet/sctp.h>
114 #include <netinet/sctp_crc32.h>
115 #endif
116 
117 #include <netinet6/ip6protosw.h>
118 #include <netinet6/scope6_var.h>
119 
120 extern int in6_mcast_loop;
121 
122 struct ip6_exthdrs {
123 	struct mbuf *ip6e_ip6;
124 	struct mbuf *ip6e_hbh;
125 	struct mbuf *ip6e_dest1;
126 	struct mbuf *ip6e_rthdr;
127 	struct mbuf *ip6e_dest2;
128 };
129 
130 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
131 
132 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
133 			   struct ucred *, int);
134 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
135 	struct socket *, struct sockopt *);
136 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
137 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
138 	struct ucred *, int, int, int);
139 
140 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
141 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
142 	struct ip6_frag **);
143 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
144 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
145 static int ip6_getpmtu(struct route_in6 *, int,
146 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
147 	u_int);
148 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
149 	u_long *, int *, u_int);
150 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
151 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
152 
153 
154 /*
155  * Make an extension header from option data.  hp is the source, and
156  * mp is the destination.
157  */
158 #define MAKE_EXTHDR(hp, mp)						\
159     do {								\
160 	if (hp) {							\
161 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
162 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
163 		    ((eh)->ip6e_len + 1) << 3);				\
164 		if (error)						\
165 			goto freehdrs;					\
166 	}								\
167     } while (/*CONSTCOND*/ 0)
168 
169 /*
170  * Form a chain of extension headers.
171  * m is the extension header mbuf
172  * mp is the previous mbuf in the chain
173  * p is the next header
174  * i is the type of option.
175  */
176 #define MAKE_CHAIN(m, mp, p, i)\
177     do {\
178 	if (m) {\
179 		if (!hdrsplit) \
180 			panic("assumption failed: hdr not split"); \
181 		*mtod((m), u_char *) = *(p);\
182 		*(p) = (i);\
183 		p = mtod((m), u_char *);\
184 		(m)->m_next = (mp)->m_next;\
185 		(mp)->m_next = (m);\
186 		(mp) = (m);\
187 	}\
188     } while (/*CONSTCOND*/ 0)
189 
190 void
191 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
192 {
193 	u_short csum;
194 
195 	csum = in_cksum_skip(m, offset + plen, offset);
196 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
197 		csum = 0xffff;
198 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
199 
200 	if (offset + sizeof(u_short) > m->m_len) {
201 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
202 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
203 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
204 		/*
205 		 * XXX this should not happen, but if it does, the correct
206 		 * behavior may be to insert the checksum in the appropriate
207 		 * next mbuf in the chain.
208 		 */
209 		return;
210 	}
211 	*(u_short *)(m->m_data + offset) = csum;
212 }
213 
214 int
215 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
216     int fraglen , uint32_t id)
217 {
218 	struct mbuf *m, **mnext, *m_frgpart;
219 	struct ip6_hdr *ip6, *mhip6;
220 	struct ip6_frag *ip6f;
221 	int off;
222 	int error;
223 	int tlen = m0->m_pkthdr.len;
224 
225 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
226 
227 	m = m0;
228 	ip6 = mtod(m, struct ip6_hdr *);
229 	mnext = &m->m_nextpkt;
230 
231 	for (off = hlen; off < tlen; off += fraglen) {
232 		m = m_gethdr(M_NOWAIT, MT_DATA);
233 		if (!m) {
234 			IP6STAT_INC(ip6s_odropped);
235 			return (ENOBUFS);
236 		}
237 		m->m_flags = m0->m_flags & M_COPYFLAGS;
238 		*mnext = m;
239 		mnext = &m->m_nextpkt;
240 		m->m_data += max_linkhdr;
241 		mhip6 = mtod(m, struct ip6_hdr *);
242 		*mhip6 = *ip6;
243 		m->m_len = sizeof(*mhip6);
244 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
245 		if (error) {
246 			IP6STAT_INC(ip6s_odropped);
247 			return (error);
248 		}
249 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
250 		if (off + fraglen >= tlen)
251 			fraglen = tlen - off;
252 		else
253 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
254 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
255 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
256 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
257 			IP6STAT_INC(ip6s_odropped);
258 			return (ENOBUFS);
259 		}
260 		m_cat(m, m_frgpart);
261 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
262 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
263 		m->m_pkthdr.rcvif = NULL;
264 		ip6f->ip6f_reserved = 0;
265 		ip6f->ip6f_ident = id;
266 		ip6f->ip6f_nxt = nextproto;
267 		IP6STAT_INC(ip6s_ofragments);
268 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
269 	}
270 
271 	return (0);
272 }
273 
274 /*
275  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
276  * header (with pri, len, nxt, hlim, src, dst).
277  * This function may modify ver and hlim only.
278  * The mbuf chain containing the packet will be freed.
279  * The mbuf opt, if present, will not be freed.
280  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
281  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
282  * then result of route lookup is stored in ro->ro_rt.
283  *
284  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
285  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
286  * which is rt_mtu.
287  *
288  * ifpp - XXX: just for statistics
289  */
290 /*
291  * XXX TODO: no flowid is assigned for outbound flows?
292  */
293 int
294 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
295     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
296     struct ifnet **ifpp, struct inpcb *inp)
297 {
298 	struct ip6_hdr *ip6;
299 	struct ifnet *ifp, *origifp;
300 	struct mbuf *m = m0;
301 	struct mbuf *mprev = NULL;
302 	int hlen, tlen, len;
303 	struct route_in6 ip6route;
304 	struct rtentry *rt = NULL;
305 	struct sockaddr_in6 *dst, src_sa, dst_sa;
306 	struct in6_addr odst;
307 	int error = 0;
308 	struct in6_ifaddr *ia = NULL;
309 	u_long mtu;
310 	int alwaysfrag, dontfrag;
311 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
312 	struct ip6_exthdrs exthdrs;
313 	struct in6_addr src0, dst0;
314 	u_int32_t zone;
315 	struct route_in6 *ro_pmtu = NULL;
316 	int hdrsplit = 0;
317 	int sw_csum, tso;
318 	int needfiblookup;
319 	uint32_t fibnum;
320 	struct m_tag *fwd_tag = NULL;
321 	uint32_t id;
322 
323 	if (inp != NULL) {
324 		INP_LOCK_ASSERT(inp);
325 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
326 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
327 			/* unconditionally set flowid */
328 			m->m_pkthdr.flowid = inp->inp_flowid;
329 			M_HASHTYPE_SET(m, inp->inp_flowtype);
330 		}
331 	}
332 
333 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
334 	/*
335 	 * IPSec checking which handles several cases.
336 	 * FAST IPSEC: We re-injected the packet.
337 	 * XXX: need scope argument.
338 	 */
339 	if (IPSEC_ENABLED(ipv6)) {
340 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
341 			if (error == EINPROGRESS)
342 				error = 0;
343 			goto done;
344 		}
345 	}
346 #endif /* IPSEC */
347 
348 	bzero(&exthdrs, sizeof(exthdrs));
349 	if (opt) {
350 		/* Hop-by-Hop options header */
351 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
352 		/* Destination options header(1st part) */
353 		if (opt->ip6po_rthdr) {
354 			/*
355 			 * Destination options header(1st part)
356 			 * This only makes sense with a routing header.
357 			 * See Section 9.2 of RFC 3542.
358 			 * Disabling this part just for MIP6 convenience is
359 			 * a bad idea.  We need to think carefully about a
360 			 * way to make the advanced API coexist with MIP6
361 			 * options, which might automatically be inserted in
362 			 * the kernel.
363 			 */
364 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
365 		}
366 		/* Routing header */
367 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
368 		/* Destination options header(2nd part) */
369 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
370 	}
371 
372 	/*
373 	 * Calculate the total length of the extension header chain.
374 	 * Keep the length of the unfragmentable part for fragmentation.
375 	 */
376 	optlen = 0;
377 	if (exthdrs.ip6e_hbh)
378 		optlen += exthdrs.ip6e_hbh->m_len;
379 	if (exthdrs.ip6e_dest1)
380 		optlen += exthdrs.ip6e_dest1->m_len;
381 	if (exthdrs.ip6e_rthdr)
382 		optlen += exthdrs.ip6e_rthdr->m_len;
383 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
384 
385 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
386 	if (exthdrs.ip6e_dest2)
387 		optlen += exthdrs.ip6e_dest2->m_len;
388 
389 	/*
390 	 * If there is at least one extension header,
391 	 * separate IP6 header from the payload.
392 	 */
393 	if (optlen && !hdrsplit) {
394 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
395 			m = NULL;
396 			goto freehdrs;
397 		}
398 		m = exthdrs.ip6e_ip6;
399 		hdrsplit++;
400 	}
401 
402 	ip6 = mtod(m, struct ip6_hdr *);
403 
404 	/* adjust mbuf packet header length */
405 	m->m_pkthdr.len += optlen;
406 	plen = m->m_pkthdr.len - sizeof(*ip6);
407 
408 	/* If this is a jumbo payload, insert a jumbo payload option. */
409 	if (plen > IPV6_MAXPACKET) {
410 		if (!hdrsplit) {
411 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
412 				m = NULL;
413 				goto freehdrs;
414 			}
415 			m = exthdrs.ip6e_ip6;
416 			hdrsplit++;
417 		}
418 		/* adjust pointer */
419 		ip6 = mtod(m, struct ip6_hdr *);
420 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
421 			goto freehdrs;
422 		ip6->ip6_plen = 0;
423 	} else
424 		ip6->ip6_plen = htons(plen);
425 
426 	/*
427 	 * Concatenate headers and fill in next header fields.
428 	 * Here we have, on "m"
429 	 *	IPv6 payload
430 	 * and we insert headers accordingly.  Finally, we should be getting:
431 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
432 	 *
433 	 * during the header composing process, "m" points to IPv6 header.
434 	 * "mprev" points to an extension header prior to esp.
435 	 */
436 	u_char *nexthdrp = &ip6->ip6_nxt;
437 	mprev = m;
438 
439 	/*
440 	 * we treat dest2 specially.  this makes IPsec processing
441 	 * much easier.  the goal here is to make mprev point the
442 	 * mbuf prior to dest2.
443 	 *
444 	 * result: IPv6 dest2 payload
445 	 * m and mprev will point to IPv6 header.
446 	 */
447 	if (exthdrs.ip6e_dest2) {
448 		if (!hdrsplit)
449 			panic("assumption failed: hdr not split");
450 		exthdrs.ip6e_dest2->m_next = m->m_next;
451 		m->m_next = exthdrs.ip6e_dest2;
452 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
453 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
454 	}
455 
456 	/*
457 	 * result: IPv6 hbh dest1 rthdr dest2 payload
458 	 * m will point to IPv6 header.  mprev will point to the
459 	 * extension header prior to dest2 (rthdr in the above case).
460 	 */
461 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
462 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
463 		   IPPROTO_DSTOPTS);
464 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
465 		   IPPROTO_ROUTING);
466 
467 	/*
468 	 * If there is a routing header, discard the packet.
469 	 */
470 	if (exthdrs.ip6e_rthdr) {
471 		 error = EINVAL;
472 		 goto bad;
473 	}
474 
475 	/* Source address validation */
476 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
477 	    (flags & IPV6_UNSPECSRC) == 0) {
478 		error = EOPNOTSUPP;
479 		IP6STAT_INC(ip6s_badscope);
480 		goto bad;
481 	}
482 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
483 		error = EOPNOTSUPP;
484 		IP6STAT_INC(ip6s_badscope);
485 		goto bad;
486 	}
487 
488 	IP6STAT_INC(ip6s_localout);
489 
490 	/*
491 	 * Route packet.
492 	 */
493 	if (ro == NULL) {
494 		ro = &ip6route;
495 		bzero((caddr_t)ro, sizeof(*ro));
496 	}
497 	ro_pmtu = ro;
498 	if (opt && opt->ip6po_rthdr)
499 		ro = &opt->ip6po_route;
500 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
501 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
502 again:
503 	/*
504 	 * if specified, try to fill in the traffic class field.
505 	 * do not override if a non-zero value is already set.
506 	 * we check the diffserv field and the ecn field separately.
507 	 */
508 	if (opt && opt->ip6po_tclass >= 0) {
509 		int mask = 0;
510 
511 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
512 			mask |= 0xfc;
513 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
514 			mask |= 0x03;
515 		if (mask != 0)
516 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
517 	}
518 
519 	/* fill in or override the hop limit field, if necessary. */
520 	if (opt && opt->ip6po_hlim != -1)
521 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
522 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
523 		if (im6o != NULL)
524 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
525 		else
526 			ip6->ip6_hlim = V_ip6_defmcasthlim;
527 	}
528 	/*
529 	 * Validate route against routing table additions;
530 	 * a better/more specific route might have been added.
531 	 * Make sure address family is set in route.
532 	 */
533 	if (inp) {
534 		ro->ro_dst.sin6_family = AF_INET6;
535 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
536 	}
537 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
538 	    ro->ro_dst.sin6_family == AF_INET6 &&
539 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
540 		rt = ro->ro_rt;
541 		ifp = ro->ro_rt->rt_ifp;
542 	} else {
543 		if (ro->ro_lle)
544 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
545 		ro->ro_lle = NULL;
546 		if (fwd_tag == NULL) {
547 			bzero(&dst_sa, sizeof(dst_sa));
548 			dst_sa.sin6_family = AF_INET6;
549 			dst_sa.sin6_len = sizeof(dst_sa);
550 			dst_sa.sin6_addr = ip6->ip6_dst;
551 		}
552 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
553 		    &rt, fibnum);
554 		if (error != 0) {
555 			if (ifp != NULL)
556 				in6_ifstat_inc(ifp, ifs6_out_discard);
557 			goto bad;
558 		}
559 	}
560 	if (rt == NULL) {
561 		/*
562 		 * If in6_selectroute() does not return a route entry,
563 		 * dst may not have been updated.
564 		 */
565 		*dst = dst_sa;	/* XXX */
566 	}
567 
568 	/*
569 	 * then rt (for unicast) and ifp must be non-NULL valid values.
570 	 */
571 	if ((flags & IPV6_FORWARDING) == 0) {
572 		/* XXX: the FORWARDING flag can be set for mrouting. */
573 		in6_ifstat_inc(ifp, ifs6_out_request);
574 	}
575 	if (rt != NULL) {
576 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
577 		counter_u64_add(rt->rt_pksent, 1);
578 	}
579 
580 
581 	/*
582 	 * The outgoing interface must be in the zone of source and
583 	 * destination addresses.
584 	 */
585 	origifp = ifp;
586 
587 	src0 = ip6->ip6_src;
588 	if (in6_setscope(&src0, origifp, &zone))
589 		goto badscope;
590 	bzero(&src_sa, sizeof(src_sa));
591 	src_sa.sin6_family = AF_INET6;
592 	src_sa.sin6_len = sizeof(src_sa);
593 	src_sa.sin6_addr = ip6->ip6_src;
594 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
595 		goto badscope;
596 
597 	dst0 = ip6->ip6_dst;
598 	if (in6_setscope(&dst0, origifp, &zone))
599 		goto badscope;
600 	/* re-initialize to be sure */
601 	bzero(&dst_sa, sizeof(dst_sa));
602 	dst_sa.sin6_family = AF_INET6;
603 	dst_sa.sin6_len = sizeof(dst_sa);
604 	dst_sa.sin6_addr = ip6->ip6_dst;
605 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
606 		goto badscope;
607 	}
608 
609 	/* We should use ia_ifp to support the case of
610 	 * sending packets to an address of our own.
611 	 */
612 	if (ia != NULL && ia->ia_ifp)
613 		ifp = ia->ia_ifp;
614 
615 	/* scope check is done. */
616 	goto routefound;
617 
618   badscope:
619 	IP6STAT_INC(ip6s_badscope);
620 	in6_ifstat_inc(origifp, ifs6_out_discard);
621 	if (error == 0)
622 		error = EHOSTUNREACH; /* XXX */
623 	goto bad;
624 
625   routefound:
626 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
627 		if (opt && opt->ip6po_nextroute.ro_rt) {
628 			/*
629 			 * The nexthop is explicitly specified by the
630 			 * application.  We assume the next hop is an IPv6
631 			 * address.
632 			 */
633 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
634 		}
635 		else if ((rt->rt_flags & RTF_GATEWAY))
636 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
637 	}
638 
639 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
640 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
641 	} else {
642 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
643 		in6_ifstat_inc(ifp, ifs6_out_mcast);
644 		/*
645 		 * Confirm that the outgoing interface supports multicast.
646 		 */
647 		if (!(ifp->if_flags & IFF_MULTICAST)) {
648 			IP6STAT_INC(ip6s_noroute);
649 			in6_ifstat_inc(ifp, ifs6_out_discard);
650 			error = ENETUNREACH;
651 			goto bad;
652 		}
653 		if ((im6o == NULL && in6_mcast_loop) ||
654 		    (im6o && im6o->im6o_multicast_loop)) {
655 			/*
656 			 * Loop back multicast datagram if not expressly
657 			 * forbidden to do so, even if we have not joined
658 			 * the address; protocols will filter it later,
659 			 * thus deferring a hash lookup and lock acquisition
660 			 * at the expense of an m_copym().
661 			 */
662 			ip6_mloopback(ifp, m);
663 		} else {
664 			/*
665 			 * If we are acting as a multicast router, perform
666 			 * multicast forwarding as if the packet had just
667 			 * arrived on the interface to which we are about
668 			 * to send.  The multicast forwarding function
669 			 * recursively calls this function, using the
670 			 * IPV6_FORWARDING flag to prevent infinite recursion.
671 			 *
672 			 * Multicasts that are looped back by ip6_mloopback(),
673 			 * above, will be forwarded by the ip6_input() routine,
674 			 * if necessary.
675 			 */
676 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
677 				/*
678 				 * XXX: ip6_mforward expects that rcvif is NULL
679 				 * when it is called from the originating path.
680 				 * However, it may not always be the case.
681 				 */
682 				m->m_pkthdr.rcvif = NULL;
683 				if (ip6_mforward(ip6, ifp, m) != 0) {
684 					m_freem(m);
685 					goto done;
686 				}
687 			}
688 		}
689 		/*
690 		 * Multicasts with a hoplimit of zero may be looped back,
691 		 * above, but must not be transmitted on a network.
692 		 * Also, multicasts addressed to the loopback interface
693 		 * are not sent -- the above call to ip6_mloopback() will
694 		 * loop back a copy if this host actually belongs to the
695 		 * destination group on the loopback interface.
696 		 */
697 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
698 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
699 			m_freem(m);
700 			goto done;
701 		}
702 	}
703 
704 	/*
705 	 * Fill the outgoing inteface to tell the upper layer
706 	 * to increment per-interface statistics.
707 	 */
708 	if (ifpp)
709 		*ifpp = ifp;
710 
711 	/* Determine path MTU. */
712 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
713 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
714 		goto bad;
715 
716 	/*
717 	 * The caller of this function may specify to use the minimum MTU
718 	 * in some cases.
719 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
720 	 * setting.  The logic is a bit complicated; by default, unicast
721 	 * packets will follow path MTU while multicast packets will be sent at
722 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
723 	 * including unicast ones will be sent at the minimum MTU.  Multicast
724 	 * packets will always be sent at the minimum MTU unless
725 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
726 	 * See RFC 3542 for more details.
727 	 */
728 	if (mtu > IPV6_MMTU) {
729 		if ((flags & IPV6_MINMTU))
730 			mtu = IPV6_MMTU;
731 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
732 			mtu = IPV6_MMTU;
733 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
734 			 (opt == NULL ||
735 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
736 			mtu = IPV6_MMTU;
737 		}
738 	}
739 
740 	/*
741 	 * clear embedded scope identifiers if necessary.
742 	 * in6_clearscope will touch the addresses only when necessary.
743 	 */
744 	in6_clearscope(&ip6->ip6_src);
745 	in6_clearscope(&ip6->ip6_dst);
746 
747 	/*
748 	 * If the outgoing packet contains a hop-by-hop options header,
749 	 * it must be examined and processed even by the source node.
750 	 * (RFC 2460, section 4.)
751 	 */
752 	if (exthdrs.ip6e_hbh) {
753 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
754 		u_int32_t dummy; /* XXX unused */
755 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
756 
757 #ifdef DIAGNOSTIC
758 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
759 			panic("ip6e_hbh is not contiguous");
760 #endif
761 		/*
762 		 *  XXX: if we have to send an ICMPv6 error to the sender,
763 		 *       we need the M_LOOP flag since icmp6_error() expects
764 		 *       the IPv6 and the hop-by-hop options header are
765 		 *       contiguous unless the flag is set.
766 		 */
767 		m->m_flags |= M_LOOP;
768 		m->m_pkthdr.rcvif = ifp;
769 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
770 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
771 		    &dummy, &plen) < 0) {
772 			/* m was already freed at this point */
773 			error = EINVAL;/* better error? */
774 			goto done;
775 		}
776 		m->m_flags &= ~M_LOOP; /* XXX */
777 		m->m_pkthdr.rcvif = NULL;
778 	}
779 
780 	/* Jump over all PFIL processing if hooks are not active. */
781 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
782 		goto passout;
783 
784 	odst = ip6->ip6_dst;
785 	/* Run through list of hooks for output packets. */
786 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
787 	if (error != 0 || m == NULL)
788 		goto done;
789 	/* adjust pointer */
790 	ip6 = mtod(m, struct ip6_hdr *);
791 
792 	needfiblookup = 0;
793 	/* See if destination IP address was changed by packet filter. */
794 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
795 		m->m_flags |= M_SKIP_FIREWALL;
796 		/* If destination is now ourself drop to ip6_input(). */
797 		if (in6_localip(&ip6->ip6_dst)) {
798 			m->m_flags |= M_FASTFWD_OURS;
799 			if (m->m_pkthdr.rcvif == NULL)
800 				m->m_pkthdr.rcvif = V_loif;
801 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
802 				m->m_pkthdr.csum_flags |=
803 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
804 				m->m_pkthdr.csum_data = 0xffff;
805 			}
806 #ifdef SCTP
807 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
808 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
809 #endif
810 			error = netisr_queue(NETISR_IPV6, m);
811 			goto done;
812 		} else {
813 			RO_RTFREE(ro);
814 			needfiblookup = 1; /* Redo the routing table lookup. */
815 			if (ro->ro_lle)
816 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
817 			ro->ro_lle = NULL;
818 		}
819 	}
820 	/* See if fib was changed by packet filter. */
821 	if (fibnum != M_GETFIB(m)) {
822 		m->m_flags |= M_SKIP_FIREWALL;
823 		fibnum = M_GETFIB(m);
824 		RO_RTFREE(ro);
825 		needfiblookup = 1;
826 		if (ro->ro_lle)
827 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
828 		ro->ro_lle = NULL;
829 	}
830 	if (needfiblookup)
831 		goto again;
832 
833 	/* See if local, if yes, send it to netisr. */
834 	if (m->m_flags & M_FASTFWD_OURS) {
835 		if (m->m_pkthdr.rcvif == NULL)
836 			m->m_pkthdr.rcvif = V_loif;
837 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
838 			m->m_pkthdr.csum_flags |=
839 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
840 			m->m_pkthdr.csum_data = 0xffff;
841 		}
842 #ifdef SCTP
843 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
844 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
845 #endif
846 		error = netisr_queue(NETISR_IPV6, m);
847 		goto done;
848 	}
849 	/* Or forward to some other address? */
850 	if ((m->m_flags & M_IP6_NEXTHOP) &&
851 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
852 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
853 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
854 		m->m_flags |= M_SKIP_FIREWALL;
855 		m->m_flags &= ~M_IP6_NEXTHOP;
856 		m_tag_delete(m, fwd_tag);
857 		goto again;
858 	}
859 
860 passout:
861 	/*
862 	 * Send the packet to the outgoing interface.
863 	 * If necessary, do IPv6 fragmentation before sending.
864 	 *
865 	 * the logic here is rather complex:
866 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
867 	 * 1-a:	send as is if tlen <= path mtu
868 	 * 1-b:	fragment if tlen > path mtu
869 	 *
870 	 * 2: if user asks us not to fragment (dontfrag == 1)
871 	 * 2-a:	send as is if tlen <= interface mtu
872 	 * 2-b:	error if tlen > interface mtu
873 	 *
874 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
875 	 *	always fragment
876 	 *
877 	 * 4: if dontfrag == 1 && alwaysfrag == 1
878 	 *	error, as we cannot handle this conflicting request
879 	 */
880 	sw_csum = m->m_pkthdr.csum_flags;
881 	if (!hdrsplit) {
882 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
883 		sw_csum &= ~ifp->if_hwassist;
884 	} else
885 		tso = 0;
886 	/*
887 	 * If we added extension headers, we will not do TSO and calculate the
888 	 * checksums ourselves for now.
889 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
890 	 * with ext. hdrs.
891 	 */
892 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
893 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
894 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
895 	}
896 #ifdef SCTP
897 	if (sw_csum & CSUM_SCTP_IPV6) {
898 		sw_csum &= ~CSUM_SCTP_IPV6;
899 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
900 	}
901 #endif
902 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
903 	tlen = m->m_pkthdr.len;
904 
905 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
906 		dontfrag = 1;
907 	else
908 		dontfrag = 0;
909 	if (dontfrag && alwaysfrag) {	/* case 4 */
910 		/* conflicting request - can't transmit */
911 		error = EMSGSIZE;
912 		goto bad;
913 	}
914 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
915 		/*
916 		 * Even if the DONTFRAG option is specified, we cannot send the
917 		 * packet when the data length is larger than the MTU of the
918 		 * outgoing interface.
919 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
920 		 * application wanted to know the MTU value. Also return an
921 		 * error code (this is not described in the API spec).
922 		 */
923 		if (inp != NULL)
924 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
925 		error = EMSGSIZE;
926 		goto bad;
927 	}
928 
929 	/*
930 	 * transmit packet without fragmentation
931 	 */
932 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
933 		struct in6_ifaddr *ia6;
934 
935 		ip6 = mtod(m, struct ip6_hdr *);
936 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
937 		if (ia6) {
938 			/* Record statistics for this interface address. */
939 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
940 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
941 			    m->m_pkthdr.len);
942 			ifa_free(&ia6->ia_ifa);
943 		}
944 #ifdef RATELIMIT
945 		if (inp != NULL) {
946 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
947 				in_pcboutput_txrtlmt(inp, ifp, m);
948 			/* stamp send tag on mbuf */
949 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
950 		} else {
951 			m->m_pkthdr.snd_tag = NULL;
952 		}
953 #endif
954 		error = nd6_output_ifp(ifp, origifp, m, dst,
955 		    (struct route *)ro);
956 #ifdef RATELIMIT
957 		/* check for route change */
958 		if (error == EAGAIN)
959 			in_pcboutput_eagain(inp);
960 #endif
961 		goto done;
962 	}
963 
964 	/*
965 	 * try to fragment the packet.  case 1-b and 3
966 	 */
967 	if (mtu < IPV6_MMTU) {
968 		/* path MTU cannot be less than IPV6_MMTU */
969 		error = EMSGSIZE;
970 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
971 		goto bad;
972 	} else if (ip6->ip6_plen == 0) {
973 		/* jumbo payload cannot be fragmented */
974 		error = EMSGSIZE;
975 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
976 		goto bad;
977 	} else {
978 		u_char nextproto;
979 
980 		/*
981 		 * Too large for the destination or interface;
982 		 * fragment if possible.
983 		 * Must be able to put at least 8 bytes per fragment.
984 		 */
985 		hlen = unfragpartlen;
986 		if (mtu > IPV6_MAXPACKET)
987 			mtu = IPV6_MAXPACKET;
988 
989 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
990 		if (len < 8) {
991 			error = EMSGSIZE;
992 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
993 			goto bad;
994 		}
995 
996 		/*
997 		 * If the interface will not calculate checksums on
998 		 * fragmented packets, then do it here.
999 		 * XXX-BZ handle the hw offloading case.  Need flags.
1000 		 */
1001 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1002 			in6_delayed_cksum(m, plen, hlen);
1003 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1004 		}
1005 #ifdef SCTP
1006 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1007 			sctp_delayed_cksum(m, hlen);
1008 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1009 		}
1010 #endif
1011 		/*
1012 		 * Change the next header field of the last header in the
1013 		 * unfragmentable part.
1014 		 */
1015 		if (exthdrs.ip6e_rthdr) {
1016 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1017 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1018 		} else if (exthdrs.ip6e_dest1) {
1019 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1020 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1021 		} else if (exthdrs.ip6e_hbh) {
1022 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1023 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1024 		} else {
1025 			nextproto = ip6->ip6_nxt;
1026 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1027 		}
1028 
1029 		/*
1030 		 * Loop through length of segment after first fragment,
1031 		 * make new header and copy data of each part and link onto
1032 		 * chain.
1033 		 */
1034 		m0 = m;
1035 		id = htonl(ip6_randomid());
1036 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1037 			goto sendorfree;
1038 
1039 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1040 	}
1041 
1042 	/*
1043 	 * Remove leading garbages.
1044 	 */
1045 sendorfree:
1046 	m = m0->m_nextpkt;
1047 	m0->m_nextpkt = 0;
1048 	m_freem(m0);
1049 	for (m0 = m; m; m = m0) {
1050 		m0 = m->m_nextpkt;
1051 		m->m_nextpkt = 0;
1052 		if (error == 0) {
1053 			/* Record statistics for this interface address. */
1054 			if (ia) {
1055 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1056 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1057 				    m->m_pkthdr.len);
1058 			}
1059 #ifdef RATELIMIT
1060 			if (inp != NULL) {
1061 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1062 					in_pcboutput_txrtlmt(inp, ifp, m);
1063 				/* stamp send tag on mbuf */
1064 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1065 			} else {
1066 				m->m_pkthdr.snd_tag = NULL;
1067 			}
1068 #endif
1069 			error = nd6_output_ifp(ifp, origifp, m, dst,
1070 			    (struct route *)ro);
1071 #ifdef RATELIMIT
1072 			/* check for route change */
1073 			if (error == EAGAIN)
1074 				in_pcboutput_eagain(inp);
1075 #endif
1076 		} else
1077 			m_freem(m);
1078 	}
1079 
1080 	if (error == 0)
1081 		IP6STAT_INC(ip6s_fragmented);
1082 
1083 done:
1084 	if (ro == &ip6route)
1085 		RO_RTFREE(ro);
1086 	return (error);
1087 
1088 freehdrs:
1089 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1090 	m_freem(exthdrs.ip6e_dest1);
1091 	m_freem(exthdrs.ip6e_rthdr);
1092 	m_freem(exthdrs.ip6e_dest2);
1093 	/* FALLTHROUGH */
1094 bad:
1095 	if (m)
1096 		m_freem(m);
1097 	goto done;
1098 }
1099 
1100 static int
1101 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1102 {
1103 	struct mbuf *m;
1104 
1105 	if (hlen > MCLBYTES)
1106 		return (ENOBUFS); /* XXX */
1107 
1108 	if (hlen > MLEN)
1109 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1110 	else
1111 		m = m_get(M_NOWAIT, MT_DATA);
1112 	if (m == NULL)
1113 		return (ENOBUFS);
1114 	m->m_len = hlen;
1115 	if (hdr)
1116 		bcopy(hdr, mtod(m, caddr_t), hlen);
1117 
1118 	*mp = m;
1119 	return (0);
1120 }
1121 
1122 /*
1123  * Insert jumbo payload option.
1124  */
1125 static int
1126 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1127 {
1128 	struct mbuf *mopt;
1129 	u_char *optbuf;
1130 	u_int32_t v;
1131 
1132 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1133 
1134 	/*
1135 	 * If there is no hop-by-hop options header, allocate new one.
1136 	 * If there is one but it doesn't have enough space to store the
1137 	 * jumbo payload option, allocate a cluster to store the whole options.
1138 	 * Otherwise, use it to store the options.
1139 	 */
1140 	if (exthdrs->ip6e_hbh == NULL) {
1141 		mopt = m_get(M_NOWAIT, MT_DATA);
1142 		if (mopt == NULL)
1143 			return (ENOBUFS);
1144 		mopt->m_len = JUMBOOPTLEN;
1145 		optbuf = mtod(mopt, u_char *);
1146 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1147 		exthdrs->ip6e_hbh = mopt;
1148 	} else {
1149 		struct ip6_hbh *hbh;
1150 
1151 		mopt = exthdrs->ip6e_hbh;
1152 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1153 			/*
1154 			 * XXX assumption:
1155 			 * - exthdrs->ip6e_hbh is not referenced from places
1156 			 *   other than exthdrs.
1157 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1158 			 */
1159 			int oldoptlen = mopt->m_len;
1160 			struct mbuf *n;
1161 
1162 			/*
1163 			 * XXX: give up if the whole (new) hbh header does
1164 			 * not fit even in an mbuf cluster.
1165 			 */
1166 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1167 				return (ENOBUFS);
1168 
1169 			/*
1170 			 * As a consequence, we must always prepare a cluster
1171 			 * at this point.
1172 			 */
1173 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1174 			if (n == NULL)
1175 				return (ENOBUFS);
1176 			n->m_len = oldoptlen + JUMBOOPTLEN;
1177 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1178 			    oldoptlen);
1179 			optbuf = mtod(n, caddr_t) + oldoptlen;
1180 			m_freem(mopt);
1181 			mopt = exthdrs->ip6e_hbh = n;
1182 		} else {
1183 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1184 			mopt->m_len += JUMBOOPTLEN;
1185 		}
1186 		optbuf[0] = IP6OPT_PADN;
1187 		optbuf[1] = 1;
1188 
1189 		/*
1190 		 * Adjust the header length according to the pad and
1191 		 * the jumbo payload option.
1192 		 */
1193 		hbh = mtod(mopt, struct ip6_hbh *);
1194 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1195 	}
1196 
1197 	/* fill in the option. */
1198 	optbuf[2] = IP6OPT_JUMBO;
1199 	optbuf[3] = 4;
1200 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1201 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1202 
1203 	/* finally, adjust the packet header length */
1204 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1205 
1206 	return (0);
1207 #undef JUMBOOPTLEN
1208 }
1209 
1210 /*
1211  * Insert fragment header and copy unfragmentable header portions.
1212  */
1213 static int
1214 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1215     struct ip6_frag **frghdrp)
1216 {
1217 	struct mbuf *n, *mlast;
1218 
1219 	if (hlen > sizeof(struct ip6_hdr)) {
1220 		n = m_copym(m0, sizeof(struct ip6_hdr),
1221 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1222 		if (n == NULL)
1223 			return (ENOBUFS);
1224 		m->m_next = n;
1225 	} else
1226 		n = m;
1227 
1228 	/* Search for the last mbuf of unfragmentable part. */
1229 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1230 		;
1231 
1232 	if (M_WRITABLE(mlast) &&
1233 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1234 		/* use the trailing space of the last mbuf for the fragment hdr */
1235 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1236 		    mlast->m_len);
1237 		mlast->m_len += sizeof(struct ip6_frag);
1238 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1239 	} else {
1240 		/* allocate a new mbuf for the fragment header */
1241 		struct mbuf *mfrg;
1242 
1243 		mfrg = m_get(M_NOWAIT, MT_DATA);
1244 		if (mfrg == NULL)
1245 			return (ENOBUFS);
1246 		mfrg->m_len = sizeof(struct ip6_frag);
1247 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1248 		mlast->m_next = mfrg;
1249 	}
1250 
1251 	return (0);
1252 }
1253 
1254 /*
1255  * Calculates IPv6 path mtu for destination @dst.
1256  * Resulting MTU is stored in @mtup.
1257  *
1258  * Returns 0 on success.
1259  */
1260 static int
1261 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1262 {
1263 	struct nhop6_extended nh6;
1264 	struct in6_addr kdst;
1265 	uint32_t scopeid;
1266 	struct ifnet *ifp;
1267 	u_long mtu;
1268 	int error;
1269 
1270 	in6_splitscope(dst, &kdst, &scopeid);
1271 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1272 		return (EHOSTUNREACH);
1273 
1274 	ifp = nh6.nh_ifp;
1275 	mtu = nh6.nh_mtu;
1276 
1277 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1278 	fib6_free_nh_ext(fibnum, &nh6);
1279 
1280 	return (error);
1281 }
1282 
1283 /*
1284  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1285  * and cached data in @ro_pmtu.
1286  * MTU from (successful) route lookup is saved (along with dst)
1287  * inside @ro_pmtu to avoid subsequent route lookups after packet
1288  * filter processing.
1289  *
1290  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1291  * Returns 0 on success.
1292  */
1293 static int
1294 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1295     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1296     int *alwaysfragp, u_int fibnum, u_int proto)
1297 {
1298 	struct nhop6_basic nh6;
1299 	struct in6_addr kdst;
1300 	uint32_t scopeid;
1301 	struct sockaddr_in6 *sa6_dst;
1302 	u_long mtu;
1303 
1304 	mtu = 0;
1305 	if (do_lookup) {
1306 
1307 		/*
1308 		 * Here ro_pmtu has final destination address, while
1309 		 * ro might represent immediate destination.
1310 		 * Use ro_pmtu destination since mtu might differ.
1311 		 */
1312 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1313 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1314 			ro_pmtu->ro_mtu = 0;
1315 
1316 		if (ro_pmtu->ro_mtu == 0) {
1317 			bzero(sa6_dst, sizeof(*sa6_dst));
1318 			sa6_dst->sin6_family = AF_INET6;
1319 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1320 			sa6_dst->sin6_addr = *dst;
1321 
1322 			in6_splitscope(dst, &kdst, &scopeid);
1323 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1324 			    &nh6) == 0)
1325 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1326 		}
1327 
1328 		mtu = ro_pmtu->ro_mtu;
1329 	}
1330 
1331 	if (ro_pmtu->ro_rt)
1332 		mtu = ro_pmtu->ro_rt->rt_mtu;
1333 
1334 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1335 }
1336 
1337 /*
1338  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1339  * hostcache data for @dst.
1340  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1341  *
1342  * Returns 0 on success.
1343  */
1344 static int
1345 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1346     u_long *mtup, int *alwaysfragp, u_int proto)
1347 {
1348 	u_long mtu = 0;
1349 	int alwaysfrag = 0;
1350 	int error = 0;
1351 
1352 	if (rt_mtu > 0) {
1353 		u_int32_t ifmtu;
1354 		struct in_conninfo inc;
1355 
1356 		bzero(&inc, sizeof(inc));
1357 		inc.inc_flags |= INC_ISIPV6;
1358 		inc.inc6_faddr = *dst;
1359 
1360 		ifmtu = IN6_LINKMTU(ifp);
1361 
1362 		/* TCP is known to react to pmtu changes so skip hc */
1363 		if (proto != IPPROTO_TCP)
1364 			mtu = tcp_hc_getmtu(&inc);
1365 
1366 		if (mtu)
1367 			mtu = min(mtu, rt_mtu);
1368 		else
1369 			mtu = rt_mtu;
1370 		if (mtu == 0)
1371 			mtu = ifmtu;
1372 		else if (mtu < IPV6_MMTU) {
1373 			/*
1374 			 * RFC2460 section 5, last paragraph:
1375 			 * if we record ICMPv6 too big message with
1376 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1377 			 * or smaller, with framgent header attached.
1378 			 * (fragment header is needed regardless from the
1379 			 * packet size, for translators to identify packets)
1380 			 */
1381 			alwaysfrag = 1;
1382 			mtu = IPV6_MMTU;
1383 		}
1384 	} else if (ifp) {
1385 		mtu = IN6_LINKMTU(ifp);
1386 	} else
1387 		error = EHOSTUNREACH; /* XXX */
1388 
1389 	*mtup = mtu;
1390 	if (alwaysfragp)
1391 		*alwaysfragp = alwaysfrag;
1392 	return (error);
1393 }
1394 
1395 /*
1396  * IP6 socket option processing.
1397  */
1398 int
1399 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1400 {
1401 	int optdatalen, uproto;
1402 	void *optdata;
1403 	struct inpcb *in6p = sotoinpcb(so);
1404 	int error, optval;
1405 	int level, op, optname;
1406 	int optlen;
1407 	struct thread *td;
1408 #ifdef	RSS
1409 	uint32_t rss_bucket;
1410 	int retval;
1411 #endif
1412 
1413 /*
1414  * Don't use more than a quarter of mbuf clusters.  N.B.:
1415  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1416  * on LP64 architectures, so cast to u_long to avoid undefined
1417  * behavior.  ILP32 architectures cannot have nmbclusters
1418  * large enough to overflow for other reasons.
1419  */
1420 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1421 
1422 	level = sopt->sopt_level;
1423 	op = sopt->sopt_dir;
1424 	optname = sopt->sopt_name;
1425 	optlen = sopt->sopt_valsize;
1426 	td = sopt->sopt_td;
1427 	error = 0;
1428 	optval = 0;
1429 	uproto = (int)so->so_proto->pr_protocol;
1430 
1431 	if (level != IPPROTO_IPV6) {
1432 		error = EINVAL;
1433 
1434 		if (sopt->sopt_level == SOL_SOCKET &&
1435 		    sopt->sopt_dir == SOPT_SET) {
1436 			switch (sopt->sopt_name) {
1437 			case SO_REUSEADDR:
1438 				INP_WLOCK(in6p);
1439 				if ((so->so_options & SO_REUSEADDR) != 0)
1440 					in6p->inp_flags2 |= INP_REUSEADDR;
1441 				else
1442 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1443 				INP_WUNLOCK(in6p);
1444 				error = 0;
1445 				break;
1446 			case SO_REUSEPORT:
1447 				INP_WLOCK(in6p);
1448 				if ((so->so_options & SO_REUSEPORT) != 0)
1449 					in6p->inp_flags2 |= INP_REUSEPORT;
1450 				else
1451 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1452 				INP_WUNLOCK(in6p);
1453 				error = 0;
1454 				break;
1455 			case SO_SETFIB:
1456 				INP_WLOCK(in6p);
1457 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1458 				INP_WUNLOCK(in6p);
1459 				error = 0;
1460 				break;
1461 			case SO_MAX_PACING_RATE:
1462 #ifdef RATELIMIT
1463 				INP_WLOCK(in6p);
1464 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1465 				INP_WUNLOCK(in6p);
1466 				error = 0;
1467 #else
1468 				error = EOPNOTSUPP;
1469 #endif
1470 				break;
1471 			default:
1472 				break;
1473 			}
1474 		}
1475 	} else {		/* level == IPPROTO_IPV6 */
1476 		switch (op) {
1477 
1478 		case SOPT_SET:
1479 			switch (optname) {
1480 			case IPV6_2292PKTOPTIONS:
1481 #ifdef IPV6_PKTOPTIONS
1482 			case IPV6_PKTOPTIONS:
1483 #endif
1484 			{
1485 				struct mbuf *m;
1486 
1487 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1488 					printf("ip6_ctloutput: mbuf limit hit\n");
1489 					error = ENOBUFS;
1490 					break;
1491 				}
1492 
1493 				error = soopt_getm(sopt, &m); /* XXX */
1494 				if (error != 0)
1495 					break;
1496 				error = soopt_mcopyin(sopt, m); /* XXX */
1497 				if (error != 0)
1498 					break;
1499 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1500 						    m, so, sopt);
1501 				m_freem(m); /* XXX */
1502 				break;
1503 			}
1504 
1505 			/*
1506 			 * Use of some Hop-by-Hop options or some
1507 			 * Destination options, might require special
1508 			 * privilege.  That is, normal applications
1509 			 * (without special privilege) might be forbidden
1510 			 * from setting certain options in outgoing packets,
1511 			 * and might never see certain options in received
1512 			 * packets. [RFC 2292 Section 6]
1513 			 * KAME specific note:
1514 			 *  KAME prevents non-privileged users from sending or
1515 			 *  receiving ANY hbh/dst options in order to avoid
1516 			 *  overhead of parsing options in the kernel.
1517 			 */
1518 			case IPV6_RECVHOPOPTS:
1519 			case IPV6_RECVDSTOPTS:
1520 			case IPV6_RECVRTHDRDSTOPTS:
1521 				if (td != NULL) {
1522 					error = priv_check(td,
1523 					    PRIV_NETINET_SETHDROPTS);
1524 					if (error)
1525 						break;
1526 				}
1527 				/* FALLTHROUGH */
1528 			case IPV6_UNICAST_HOPS:
1529 			case IPV6_HOPLIMIT:
1530 
1531 			case IPV6_RECVPKTINFO:
1532 			case IPV6_RECVHOPLIMIT:
1533 			case IPV6_RECVRTHDR:
1534 			case IPV6_RECVPATHMTU:
1535 			case IPV6_RECVTCLASS:
1536 			case IPV6_RECVFLOWID:
1537 #ifdef	RSS
1538 			case IPV6_RECVRSSBUCKETID:
1539 #endif
1540 			case IPV6_V6ONLY:
1541 			case IPV6_AUTOFLOWLABEL:
1542 			case IPV6_ORIGDSTADDR:
1543 			case IPV6_BINDANY:
1544 			case IPV6_BINDMULTI:
1545 #ifdef	RSS
1546 			case IPV6_RSS_LISTEN_BUCKET:
1547 #endif
1548 				if (optname == IPV6_BINDANY && td != NULL) {
1549 					error = priv_check(td,
1550 					    PRIV_NETINET_BINDANY);
1551 					if (error)
1552 						break;
1553 				}
1554 
1555 				if (optlen != sizeof(int)) {
1556 					error = EINVAL;
1557 					break;
1558 				}
1559 				error = sooptcopyin(sopt, &optval,
1560 					sizeof optval, sizeof optval);
1561 				if (error)
1562 					break;
1563 				switch (optname) {
1564 
1565 				case IPV6_UNICAST_HOPS:
1566 					if (optval < -1 || optval >= 256)
1567 						error = EINVAL;
1568 					else {
1569 						/* -1 = kernel default */
1570 						in6p->in6p_hops = optval;
1571 						if ((in6p->inp_vflag &
1572 						     INP_IPV4) != 0)
1573 							in6p->inp_ip_ttl = optval;
1574 					}
1575 					break;
1576 #define OPTSET(bit) \
1577 do { \
1578 	INP_WLOCK(in6p); \
1579 	if (optval) \
1580 		in6p->inp_flags |= (bit); \
1581 	else \
1582 		in6p->inp_flags &= ~(bit); \
1583 	INP_WUNLOCK(in6p); \
1584 } while (/*CONSTCOND*/ 0)
1585 #define OPTSET2292(bit) \
1586 do { \
1587 	INP_WLOCK(in6p); \
1588 	in6p->inp_flags |= IN6P_RFC2292; \
1589 	if (optval) \
1590 		in6p->inp_flags |= (bit); \
1591 	else \
1592 		in6p->inp_flags &= ~(bit); \
1593 	INP_WUNLOCK(in6p); \
1594 } while (/*CONSTCOND*/ 0)
1595 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1596 
1597 #define OPTSET2(bit, val) do {						\
1598 	INP_WLOCK(in6p);						\
1599 	if (val)							\
1600 		in6p->inp_flags2 |= bit;				\
1601 	else								\
1602 		in6p->inp_flags2 &= ~bit;				\
1603 	INP_WUNLOCK(in6p);						\
1604 } while (0)
1605 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1606 
1607 				case IPV6_RECVPKTINFO:
1608 					/* cannot mix with RFC2292 */
1609 					if (OPTBIT(IN6P_RFC2292)) {
1610 						error = EINVAL;
1611 						break;
1612 					}
1613 					OPTSET(IN6P_PKTINFO);
1614 					break;
1615 
1616 				case IPV6_HOPLIMIT:
1617 				{
1618 					struct ip6_pktopts **optp;
1619 
1620 					/* cannot mix with RFC2292 */
1621 					if (OPTBIT(IN6P_RFC2292)) {
1622 						error = EINVAL;
1623 						break;
1624 					}
1625 					optp = &in6p->in6p_outputopts;
1626 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1627 					    (u_char *)&optval, sizeof(optval),
1628 					    optp, (td != NULL) ? td->td_ucred :
1629 					    NULL, uproto);
1630 					break;
1631 				}
1632 
1633 				case IPV6_RECVHOPLIMIT:
1634 					/* cannot mix with RFC2292 */
1635 					if (OPTBIT(IN6P_RFC2292)) {
1636 						error = EINVAL;
1637 						break;
1638 					}
1639 					OPTSET(IN6P_HOPLIMIT);
1640 					break;
1641 
1642 				case IPV6_RECVHOPOPTS:
1643 					/* cannot mix with RFC2292 */
1644 					if (OPTBIT(IN6P_RFC2292)) {
1645 						error = EINVAL;
1646 						break;
1647 					}
1648 					OPTSET(IN6P_HOPOPTS);
1649 					break;
1650 
1651 				case IPV6_RECVDSTOPTS:
1652 					/* cannot mix with RFC2292 */
1653 					if (OPTBIT(IN6P_RFC2292)) {
1654 						error = EINVAL;
1655 						break;
1656 					}
1657 					OPTSET(IN6P_DSTOPTS);
1658 					break;
1659 
1660 				case IPV6_RECVRTHDRDSTOPTS:
1661 					/* cannot mix with RFC2292 */
1662 					if (OPTBIT(IN6P_RFC2292)) {
1663 						error = EINVAL;
1664 						break;
1665 					}
1666 					OPTSET(IN6P_RTHDRDSTOPTS);
1667 					break;
1668 
1669 				case IPV6_RECVRTHDR:
1670 					/* cannot mix with RFC2292 */
1671 					if (OPTBIT(IN6P_RFC2292)) {
1672 						error = EINVAL;
1673 						break;
1674 					}
1675 					OPTSET(IN6P_RTHDR);
1676 					break;
1677 
1678 				case IPV6_RECVPATHMTU:
1679 					/*
1680 					 * We ignore this option for TCP
1681 					 * sockets.
1682 					 * (RFC3542 leaves this case
1683 					 * unspecified.)
1684 					 */
1685 					if (uproto != IPPROTO_TCP)
1686 						OPTSET(IN6P_MTU);
1687 					break;
1688 
1689 				case IPV6_RECVFLOWID:
1690 					OPTSET2(INP_RECVFLOWID, optval);
1691 					break;
1692 
1693 #ifdef	RSS
1694 				case IPV6_RECVRSSBUCKETID:
1695 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1696 					break;
1697 #endif
1698 
1699 				case IPV6_V6ONLY:
1700 					/*
1701 					 * make setsockopt(IPV6_V6ONLY)
1702 					 * available only prior to bind(2).
1703 					 * see ipng mailing list, Jun 22 2001.
1704 					 */
1705 					if (in6p->inp_lport ||
1706 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1707 						error = EINVAL;
1708 						break;
1709 					}
1710 					OPTSET(IN6P_IPV6_V6ONLY);
1711 					if (optval)
1712 						in6p->inp_vflag &= ~INP_IPV4;
1713 					else
1714 						in6p->inp_vflag |= INP_IPV4;
1715 					break;
1716 				case IPV6_RECVTCLASS:
1717 					/* cannot mix with RFC2292 XXX */
1718 					if (OPTBIT(IN6P_RFC2292)) {
1719 						error = EINVAL;
1720 						break;
1721 					}
1722 					OPTSET(IN6P_TCLASS);
1723 					break;
1724 				case IPV6_AUTOFLOWLABEL:
1725 					OPTSET(IN6P_AUTOFLOWLABEL);
1726 					break;
1727 
1728 				case IPV6_ORIGDSTADDR:
1729 					OPTSET2(INP_ORIGDSTADDR, optval);
1730 					break;
1731 				case IPV6_BINDANY:
1732 					OPTSET(INP_BINDANY);
1733 					break;
1734 
1735 				case IPV6_BINDMULTI:
1736 					OPTSET2(INP_BINDMULTI, optval);
1737 					break;
1738 #ifdef	RSS
1739 				case IPV6_RSS_LISTEN_BUCKET:
1740 					if ((optval >= 0) &&
1741 					    (optval < rss_getnumbuckets())) {
1742 						in6p->inp_rss_listen_bucket = optval;
1743 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1744 					} else {
1745 						error = EINVAL;
1746 					}
1747 					break;
1748 #endif
1749 				}
1750 				break;
1751 
1752 			case IPV6_TCLASS:
1753 			case IPV6_DONTFRAG:
1754 			case IPV6_USE_MIN_MTU:
1755 			case IPV6_PREFER_TEMPADDR:
1756 				if (optlen != sizeof(optval)) {
1757 					error = EINVAL;
1758 					break;
1759 				}
1760 				error = sooptcopyin(sopt, &optval,
1761 					sizeof optval, sizeof optval);
1762 				if (error)
1763 					break;
1764 				{
1765 					struct ip6_pktopts **optp;
1766 					optp = &in6p->in6p_outputopts;
1767 					error = ip6_pcbopt(optname,
1768 					    (u_char *)&optval, sizeof(optval),
1769 					    optp, (td != NULL) ? td->td_ucred :
1770 					    NULL, uproto);
1771 					break;
1772 				}
1773 
1774 			case IPV6_2292PKTINFO:
1775 			case IPV6_2292HOPLIMIT:
1776 			case IPV6_2292HOPOPTS:
1777 			case IPV6_2292DSTOPTS:
1778 			case IPV6_2292RTHDR:
1779 				/* RFC 2292 */
1780 				if (optlen != sizeof(int)) {
1781 					error = EINVAL;
1782 					break;
1783 				}
1784 				error = sooptcopyin(sopt, &optval,
1785 					sizeof optval, sizeof optval);
1786 				if (error)
1787 					break;
1788 				switch (optname) {
1789 				case IPV6_2292PKTINFO:
1790 					OPTSET2292(IN6P_PKTINFO);
1791 					break;
1792 				case IPV6_2292HOPLIMIT:
1793 					OPTSET2292(IN6P_HOPLIMIT);
1794 					break;
1795 				case IPV6_2292HOPOPTS:
1796 					/*
1797 					 * Check super-user privilege.
1798 					 * See comments for IPV6_RECVHOPOPTS.
1799 					 */
1800 					if (td != NULL) {
1801 						error = priv_check(td,
1802 						    PRIV_NETINET_SETHDROPTS);
1803 						if (error)
1804 							return (error);
1805 					}
1806 					OPTSET2292(IN6P_HOPOPTS);
1807 					break;
1808 				case IPV6_2292DSTOPTS:
1809 					if (td != NULL) {
1810 						error = priv_check(td,
1811 						    PRIV_NETINET_SETHDROPTS);
1812 						if (error)
1813 							return (error);
1814 					}
1815 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1816 					break;
1817 				case IPV6_2292RTHDR:
1818 					OPTSET2292(IN6P_RTHDR);
1819 					break;
1820 				}
1821 				break;
1822 			case IPV6_PKTINFO:
1823 			case IPV6_HOPOPTS:
1824 			case IPV6_RTHDR:
1825 			case IPV6_DSTOPTS:
1826 			case IPV6_RTHDRDSTOPTS:
1827 			case IPV6_NEXTHOP:
1828 			{
1829 				/* new advanced API (RFC3542) */
1830 				u_char *optbuf;
1831 				u_char optbuf_storage[MCLBYTES];
1832 				int optlen;
1833 				struct ip6_pktopts **optp;
1834 
1835 				/* cannot mix with RFC2292 */
1836 				if (OPTBIT(IN6P_RFC2292)) {
1837 					error = EINVAL;
1838 					break;
1839 				}
1840 
1841 				/*
1842 				 * We only ensure valsize is not too large
1843 				 * here.  Further validation will be done
1844 				 * later.
1845 				 */
1846 				error = sooptcopyin(sopt, optbuf_storage,
1847 				    sizeof(optbuf_storage), 0);
1848 				if (error)
1849 					break;
1850 				optlen = sopt->sopt_valsize;
1851 				optbuf = optbuf_storage;
1852 				optp = &in6p->in6p_outputopts;
1853 				error = ip6_pcbopt(optname, optbuf, optlen,
1854 				    optp, (td != NULL) ? td->td_ucred : NULL,
1855 				    uproto);
1856 				break;
1857 			}
1858 #undef OPTSET
1859 
1860 			case IPV6_MULTICAST_IF:
1861 			case IPV6_MULTICAST_HOPS:
1862 			case IPV6_MULTICAST_LOOP:
1863 			case IPV6_JOIN_GROUP:
1864 			case IPV6_LEAVE_GROUP:
1865 			case IPV6_MSFILTER:
1866 			case MCAST_BLOCK_SOURCE:
1867 			case MCAST_UNBLOCK_SOURCE:
1868 			case MCAST_JOIN_GROUP:
1869 			case MCAST_LEAVE_GROUP:
1870 			case MCAST_JOIN_SOURCE_GROUP:
1871 			case MCAST_LEAVE_SOURCE_GROUP:
1872 				error = ip6_setmoptions(in6p, sopt);
1873 				break;
1874 
1875 			case IPV6_PORTRANGE:
1876 				error = sooptcopyin(sopt, &optval,
1877 				    sizeof optval, sizeof optval);
1878 				if (error)
1879 					break;
1880 
1881 				INP_WLOCK(in6p);
1882 				switch (optval) {
1883 				case IPV6_PORTRANGE_DEFAULT:
1884 					in6p->inp_flags &= ~(INP_LOWPORT);
1885 					in6p->inp_flags &= ~(INP_HIGHPORT);
1886 					break;
1887 
1888 				case IPV6_PORTRANGE_HIGH:
1889 					in6p->inp_flags &= ~(INP_LOWPORT);
1890 					in6p->inp_flags |= INP_HIGHPORT;
1891 					break;
1892 
1893 				case IPV6_PORTRANGE_LOW:
1894 					in6p->inp_flags &= ~(INP_HIGHPORT);
1895 					in6p->inp_flags |= INP_LOWPORT;
1896 					break;
1897 
1898 				default:
1899 					error = EINVAL;
1900 					break;
1901 				}
1902 				INP_WUNLOCK(in6p);
1903 				break;
1904 
1905 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1906 			case IPV6_IPSEC_POLICY:
1907 				if (IPSEC_ENABLED(ipv6)) {
1908 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1909 					break;
1910 				}
1911 				/* FALLTHROUGH */
1912 #endif /* IPSEC */
1913 
1914 			default:
1915 				error = ENOPROTOOPT;
1916 				break;
1917 			}
1918 			break;
1919 
1920 		case SOPT_GET:
1921 			switch (optname) {
1922 
1923 			case IPV6_2292PKTOPTIONS:
1924 #ifdef IPV6_PKTOPTIONS
1925 			case IPV6_PKTOPTIONS:
1926 #endif
1927 				/*
1928 				 * RFC3542 (effectively) deprecated the
1929 				 * semantics of the 2292-style pktoptions.
1930 				 * Since it was not reliable in nature (i.e.,
1931 				 * applications had to expect the lack of some
1932 				 * information after all), it would make sense
1933 				 * to simplify this part by always returning
1934 				 * empty data.
1935 				 */
1936 				sopt->sopt_valsize = 0;
1937 				break;
1938 
1939 			case IPV6_RECVHOPOPTS:
1940 			case IPV6_RECVDSTOPTS:
1941 			case IPV6_RECVRTHDRDSTOPTS:
1942 			case IPV6_UNICAST_HOPS:
1943 			case IPV6_RECVPKTINFO:
1944 			case IPV6_RECVHOPLIMIT:
1945 			case IPV6_RECVRTHDR:
1946 			case IPV6_RECVPATHMTU:
1947 
1948 			case IPV6_V6ONLY:
1949 			case IPV6_PORTRANGE:
1950 			case IPV6_RECVTCLASS:
1951 			case IPV6_AUTOFLOWLABEL:
1952 			case IPV6_BINDANY:
1953 			case IPV6_FLOWID:
1954 			case IPV6_FLOWTYPE:
1955 			case IPV6_RECVFLOWID:
1956 #ifdef	RSS
1957 			case IPV6_RSSBUCKETID:
1958 			case IPV6_RECVRSSBUCKETID:
1959 #endif
1960 			case IPV6_BINDMULTI:
1961 				switch (optname) {
1962 
1963 				case IPV6_RECVHOPOPTS:
1964 					optval = OPTBIT(IN6P_HOPOPTS);
1965 					break;
1966 
1967 				case IPV6_RECVDSTOPTS:
1968 					optval = OPTBIT(IN6P_DSTOPTS);
1969 					break;
1970 
1971 				case IPV6_RECVRTHDRDSTOPTS:
1972 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1973 					break;
1974 
1975 				case IPV6_UNICAST_HOPS:
1976 					optval = in6p->in6p_hops;
1977 					break;
1978 
1979 				case IPV6_RECVPKTINFO:
1980 					optval = OPTBIT(IN6P_PKTINFO);
1981 					break;
1982 
1983 				case IPV6_RECVHOPLIMIT:
1984 					optval = OPTBIT(IN6P_HOPLIMIT);
1985 					break;
1986 
1987 				case IPV6_RECVRTHDR:
1988 					optval = OPTBIT(IN6P_RTHDR);
1989 					break;
1990 
1991 				case IPV6_RECVPATHMTU:
1992 					optval = OPTBIT(IN6P_MTU);
1993 					break;
1994 
1995 				case IPV6_V6ONLY:
1996 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1997 					break;
1998 
1999 				case IPV6_PORTRANGE:
2000 				    {
2001 					int flags;
2002 					flags = in6p->inp_flags;
2003 					if (flags & INP_HIGHPORT)
2004 						optval = IPV6_PORTRANGE_HIGH;
2005 					else if (flags & INP_LOWPORT)
2006 						optval = IPV6_PORTRANGE_LOW;
2007 					else
2008 						optval = 0;
2009 					break;
2010 				    }
2011 				case IPV6_RECVTCLASS:
2012 					optval = OPTBIT(IN6P_TCLASS);
2013 					break;
2014 
2015 				case IPV6_AUTOFLOWLABEL:
2016 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2017 					break;
2018 
2019 				case IPV6_ORIGDSTADDR:
2020 					optval = OPTBIT2(INP_ORIGDSTADDR);
2021 					break;
2022 
2023 				case IPV6_BINDANY:
2024 					optval = OPTBIT(INP_BINDANY);
2025 					break;
2026 
2027 				case IPV6_FLOWID:
2028 					optval = in6p->inp_flowid;
2029 					break;
2030 
2031 				case IPV6_FLOWTYPE:
2032 					optval = in6p->inp_flowtype;
2033 					break;
2034 
2035 				case IPV6_RECVFLOWID:
2036 					optval = OPTBIT2(INP_RECVFLOWID);
2037 					break;
2038 #ifdef	RSS
2039 				case IPV6_RSSBUCKETID:
2040 					retval =
2041 					    rss_hash2bucket(in6p->inp_flowid,
2042 					    in6p->inp_flowtype,
2043 					    &rss_bucket);
2044 					if (retval == 0)
2045 						optval = rss_bucket;
2046 					else
2047 						error = EINVAL;
2048 					break;
2049 
2050 				case IPV6_RECVRSSBUCKETID:
2051 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2052 					break;
2053 #endif
2054 
2055 				case IPV6_BINDMULTI:
2056 					optval = OPTBIT2(INP_BINDMULTI);
2057 					break;
2058 
2059 				}
2060 				if (error)
2061 					break;
2062 				error = sooptcopyout(sopt, &optval,
2063 					sizeof optval);
2064 				break;
2065 
2066 			case IPV6_PATHMTU:
2067 			{
2068 				u_long pmtu = 0;
2069 				struct ip6_mtuinfo mtuinfo;
2070 
2071 				if (!(so->so_state & SS_ISCONNECTED))
2072 					return (ENOTCONN);
2073 				/*
2074 				 * XXX: we dot not consider the case of source
2075 				 * routing, or optional information to specify
2076 				 * the outgoing interface.
2077 				 */
2078 				error = ip6_getpmtu_ctl(so->so_fibnum,
2079 				    &in6p->in6p_faddr, &pmtu);
2080 				if (error)
2081 					break;
2082 				if (pmtu > IPV6_MAXPACKET)
2083 					pmtu = IPV6_MAXPACKET;
2084 
2085 				bzero(&mtuinfo, sizeof(mtuinfo));
2086 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2087 				optdata = (void *)&mtuinfo;
2088 				optdatalen = sizeof(mtuinfo);
2089 				error = sooptcopyout(sopt, optdata,
2090 				    optdatalen);
2091 				break;
2092 			}
2093 
2094 			case IPV6_2292PKTINFO:
2095 			case IPV6_2292HOPLIMIT:
2096 			case IPV6_2292HOPOPTS:
2097 			case IPV6_2292RTHDR:
2098 			case IPV6_2292DSTOPTS:
2099 				switch (optname) {
2100 				case IPV6_2292PKTINFO:
2101 					optval = OPTBIT(IN6P_PKTINFO);
2102 					break;
2103 				case IPV6_2292HOPLIMIT:
2104 					optval = OPTBIT(IN6P_HOPLIMIT);
2105 					break;
2106 				case IPV6_2292HOPOPTS:
2107 					optval = OPTBIT(IN6P_HOPOPTS);
2108 					break;
2109 				case IPV6_2292RTHDR:
2110 					optval = OPTBIT(IN6P_RTHDR);
2111 					break;
2112 				case IPV6_2292DSTOPTS:
2113 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2114 					break;
2115 				}
2116 				error = sooptcopyout(sopt, &optval,
2117 				    sizeof optval);
2118 				break;
2119 			case IPV6_PKTINFO:
2120 			case IPV6_HOPOPTS:
2121 			case IPV6_RTHDR:
2122 			case IPV6_DSTOPTS:
2123 			case IPV6_RTHDRDSTOPTS:
2124 			case IPV6_NEXTHOP:
2125 			case IPV6_TCLASS:
2126 			case IPV6_DONTFRAG:
2127 			case IPV6_USE_MIN_MTU:
2128 			case IPV6_PREFER_TEMPADDR:
2129 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2130 				    optname, sopt);
2131 				break;
2132 
2133 			case IPV6_MULTICAST_IF:
2134 			case IPV6_MULTICAST_HOPS:
2135 			case IPV6_MULTICAST_LOOP:
2136 			case IPV6_MSFILTER:
2137 				error = ip6_getmoptions(in6p, sopt);
2138 				break;
2139 
2140 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2141 			case IPV6_IPSEC_POLICY:
2142 				if (IPSEC_ENABLED(ipv6)) {
2143 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2144 					break;
2145 				}
2146 				/* FALLTHROUGH */
2147 #endif /* IPSEC */
2148 			default:
2149 				error = ENOPROTOOPT;
2150 				break;
2151 			}
2152 			break;
2153 		}
2154 	}
2155 	return (error);
2156 }
2157 
2158 int
2159 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2160 {
2161 	int error = 0, optval, optlen;
2162 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2163 	struct inpcb *in6p = sotoinpcb(so);
2164 	int level, op, optname;
2165 
2166 	level = sopt->sopt_level;
2167 	op = sopt->sopt_dir;
2168 	optname = sopt->sopt_name;
2169 	optlen = sopt->sopt_valsize;
2170 
2171 	if (level != IPPROTO_IPV6) {
2172 		return (EINVAL);
2173 	}
2174 
2175 	switch (optname) {
2176 	case IPV6_CHECKSUM:
2177 		/*
2178 		 * For ICMPv6 sockets, no modification allowed for checksum
2179 		 * offset, permit "no change" values to help existing apps.
2180 		 *
2181 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2182 		 * for an ICMPv6 socket will fail."
2183 		 * The current behavior does not meet RFC3542.
2184 		 */
2185 		switch (op) {
2186 		case SOPT_SET:
2187 			if (optlen != sizeof(int)) {
2188 				error = EINVAL;
2189 				break;
2190 			}
2191 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2192 					    sizeof(optval));
2193 			if (error)
2194 				break;
2195 			if ((optval % 2) != 0) {
2196 				/* the API assumes even offset values */
2197 				error = EINVAL;
2198 			} else if (so->so_proto->pr_protocol ==
2199 			    IPPROTO_ICMPV6) {
2200 				if (optval != icmp6off)
2201 					error = EINVAL;
2202 			} else
2203 				in6p->in6p_cksum = optval;
2204 			break;
2205 
2206 		case SOPT_GET:
2207 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2208 				optval = icmp6off;
2209 			else
2210 				optval = in6p->in6p_cksum;
2211 
2212 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2213 			break;
2214 
2215 		default:
2216 			error = EINVAL;
2217 			break;
2218 		}
2219 		break;
2220 
2221 	default:
2222 		error = ENOPROTOOPT;
2223 		break;
2224 	}
2225 
2226 	return (error);
2227 }
2228 
2229 /*
2230  * Set up IP6 options in pcb for insertion in output packets or
2231  * specifying behavior of outgoing packets.
2232  */
2233 static int
2234 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2235     struct socket *so, struct sockopt *sopt)
2236 {
2237 	struct ip6_pktopts *opt = *pktopt;
2238 	int error = 0;
2239 	struct thread *td = sopt->sopt_td;
2240 
2241 	/* turn off any old options. */
2242 	if (opt) {
2243 #ifdef DIAGNOSTIC
2244 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2245 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2246 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2247 			printf("ip6_pcbopts: all specified options are cleared.\n");
2248 #endif
2249 		ip6_clearpktopts(opt, -1);
2250 	} else
2251 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2252 	*pktopt = NULL;
2253 
2254 	if (!m || m->m_len == 0) {
2255 		/*
2256 		 * Only turning off any previous options, regardless of
2257 		 * whether the opt is just created or given.
2258 		 */
2259 		free(opt, M_IP6OPT);
2260 		return (0);
2261 	}
2262 
2263 	/*  set options specified by user. */
2264 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2265 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2266 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2267 		free(opt, M_IP6OPT);
2268 		return (error);
2269 	}
2270 	*pktopt = opt;
2271 	return (0);
2272 }
2273 
2274 /*
2275  * initialize ip6_pktopts.  beware that there are non-zero default values in
2276  * the struct.
2277  */
2278 void
2279 ip6_initpktopts(struct ip6_pktopts *opt)
2280 {
2281 
2282 	bzero(opt, sizeof(*opt));
2283 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2284 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2285 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2286 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2287 }
2288 
2289 static int
2290 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2291     struct ucred *cred, int uproto)
2292 {
2293 	struct ip6_pktopts *opt;
2294 
2295 	if (*pktopt == NULL) {
2296 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2297 		    M_WAITOK);
2298 		ip6_initpktopts(*pktopt);
2299 	}
2300 	opt = *pktopt;
2301 
2302 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2303 }
2304 
2305 static int
2306 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2307 {
2308 	void *optdata = NULL;
2309 	int optdatalen = 0;
2310 	struct ip6_ext *ip6e;
2311 	int error = 0;
2312 	struct in6_pktinfo null_pktinfo;
2313 	int deftclass = 0, on;
2314 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2315 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2316 
2317 	switch (optname) {
2318 	case IPV6_PKTINFO:
2319 		optdata = (void *)&null_pktinfo;
2320 		if (pktopt && pktopt->ip6po_pktinfo) {
2321 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2322 			    sizeof(null_pktinfo));
2323 			in6_clearscope(&null_pktinfo.ipi6_addr);
2324 		} else {
2325 			/* XXX: we don't have to do this every time... */
2326 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2327 		}
2328 		optdatalen = sizeof(struct in6_pktinfo);
2329 		break;
2330 	case IPV6_TCLASS:
2331 		if (pktopt && pktopt->ip6po_tclass >= 0)
2332 			optdata = (void *)&pktopt->ip6po_tclass;
2333 		else
2334 			optdata = (void *)&deftclass;
2335 		optdatalen = sizeof(int);
2336 		break;
2337 	case IPV6_HOPOPTS:
2338 		if (pktopt && pktopt->ip6po_hbh) {
2339 			optdata = (void *)pktopt->ip6po_hbh;
2340 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2341 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2342 		}
2343 		break;
2344 	case IPV6_RTHDR:
2345 		if (pktopt && pktopt->ip6po_rthdr) {
2346 			optdata = (void *)pktopt->ip6po_rthdr;
2347 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2348 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2349 		}
2350 		break;
2351 	case IPV6_RTHDRDSTOPTS:
2352 		if (pktopt && pktopt->ip6po_dest1) {
2353 			optdata = (void *)pktopt->ip6po_dest1;
2354 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2355 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2356 		}
2357 		break;
2358 	case IPV6_DSTOPTS:
2359 		if (pktopt && pktopt->ip6po_dest2) {
2360 			optdata = (void *)pktopt->ip6po_dest2;
2361 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2362 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2363 		}
2364 		break;
2365 	case IPV6_NEXTHOP:
2366 		if (pktopt && pktopt->ip6po_nexthop) {
2367 			optdata = (void *)pktopt->ip6po_nexthop;
2368 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2369 		}
2370 		break;
2371 	case IPV6_USE_MIN_MTU:
2372 		if (pktopt)
2373 			optdata = (void *)&pktopt->ip6po_minmtu;
2374 		else
2375 			optdata = (void *)&defminmtu;
2376 		optdatalen = sizeof(int);
2377 		break;
2378 	case IPV6_DONTFRAG:
2379 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2380 			on = 1;
2381 		else
2382 			on = 0;
2383 		optdata = (void *)&on;
2384 		optdatalen = sizeof(on);
2385 		break;
2386 	case IPV6_PREFER_TEMPADDR:
2387 		if (pktopt)
2388 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2389 		else
2390 			optdata = (void *)&defpreftemp;
2391 		optdatalen = sizeof(int);
2392 		break;
2393 	default:		/* should not happen */
2394 #ifdef DIAGNOSTIC
2395 		panic("ip6_getpcbopt: unexpected option\n");
2396 #endif
2397 		return (ENOPROTOOPT);
2398 	}
2399 
2400 	error = sooptcopyout(sopt, optdata, optdatalen);
2401 
2402 	return (error);
2403 }
2404 
2405 void
2406 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2407 {
2408 	if (pktopt == NULL)
2409 		return;
2410 
2411 	if (optname == -1 || optname == IPV6_PKTINFO) {
2412 		if (pktopt->ip6po_pktinfo)
2413 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2414 		pktopt->ip6po_pktinfo = NULL;
2415 	}
2416 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2417 		pktopt->ip6po_hlim = -1;
2418 	if (optname == -1 || optname == IPV6_TCLASS)
2419 		pktopt->ip6po_tclass = -1;
2420 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2421 		if (pktopt->ip6po_nextroute.ro_rt) {
2422 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2423 			pktopt->ip6po_nextroute.ro_rt = NULL;
2424 		}
2425 		if (pktopt->ip6po_nexthop)
2426 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2427 		pktopt->ip6po_nexthop = NULL;
2428 	}
2429 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2430 		if (pktopt->ip6po_hbh)
2431 			free(pktopt->ip6po_hbh, M_IP6OPT);
2432 		pktopt->ip6po_hbh = NULL;
2433 	}
2434 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2435 		if (pktopt->ip6po_dest1)
2436 			free(pktopt->ip6po_dest1, M_IP6OPT);
2437 		pktopt->ip6po_dest1 = NULL;
2438 	}
2439 	if (optname == -1 || optname == IPV6_RTHDR) {
2440 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2441 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2442 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2443 		if (pktopt->ip6po_route.ro_rt) {
2444 			RTFREE(pktopt->ip6po_route.ro_rt);
2445 			pktopt->ip6po_route.ro_rt = NULL;
2446 		}
2447 	}
2448 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2449 		if (pktopt->ip6po_dest2)
2450 			free(pktopt->ip6po_dest2, M_IP6OPT);
2451 		pktopt->ip6po_dest2 = NULL;
2452 	}
2453 }
2454 
2455 #define PKTOPT_EXTHDRCPY(type) \
2456 do {\
2457 	if (src->type) {\
2458 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2459 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2460 		if (dst->type == NULL)\
2461 			goto bad;\
2462 		bcopy(src->type, dst->type, hlen);\
2463 	}\
2464 } while (/*CONSTCOND*/ 0)
2465 
2466 static int
2467 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2468 {
2469 	if (dst == NULL || src == NULL)  {
2470 		printf("ip6_clearpktopts: invalid argument\n");
2471 		return (EINVAL);
2472 	}
2473 
2474 	dst->ip6po_hlim = src->ip6po_hlim;
2475 	dst->ip6po_tclass = src->ip6po_tclass;
2476 	dst->ip6po_flags = src->ip6po_flags;
2477 	dst->ip6po_minmtu = src->ip6po_minmtu;
2478 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2479 	if (src->ip6po_pktinfo) {
2480 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2481 		    M_IP6OPT, canwait);
2482 		if (dst->ip6po_pktinfo == NULL)
2483 			goto bad;
2484 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2485 	}
2486 	if (src->ip6po_nexthop) {
2487 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2488 		    M_IP6OPT, canwait);
2489 		if (dst->ip6po_nexthop == NULL)
2490 			goto bad;
2491 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2492 		    src->ip6po_nexthop->sa_len);
2493 	}
2494 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2495 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2496 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2497 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2498 	return (0);
2499 
2500   bad:
2501 	ip6_clearpktopts(dst, -1);
2502 	return (ENOBUFS);
2503 }
2504 #undef PKTOPT_EXTHDRCPY
2505 
2506 struct ip6_pktopts *
2507 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2508 {
2509 	int error;
2510 	struct ip6_pktopts *dst;
2511 
2512 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2513 	if (dst == NULL)
2514 		return (NULL);
2515 	ip6_initpktopts(dst);
2516 
2517 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2518 		free(dst, M_IP6OPT);
2519 		return (NULL);
2520 	}
2521 
2522 	return (dst);
2523 }
2524 
2525 void
2526 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2527 {
2528 	if (pktopt == NULL)
2529 		return;
2530 
2531 	ip6_clearpktopts(pktopt, -1);
2532 
2533 	free(pktopt, M_IP6OPT);
2534 }
2535 
2536 /*
2537  * Set IPv6 outgoing packet options based on advanced API.
2538  */
2539 int
2540 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2541     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2542 {
2543 	struct cmsghdr *cm = NULL;
2544 
2545 	if (control == NULL || opt == NULL)
2546 		return (EINVAL);
2547 
2548 	ip6_initpktopts(opt);
2549 	if (stickyopt) {
2550 		int error;
2551 
2552 		/*
2553 		 * If stickyopt is provided, make a local copy of the options
2554 		 * for this particular packet, then override them by ancillary
2555 		 * objects.
2556 		 * XXX: copypktopts() does not copy the cached route to a next
2557 		 * hop (if any).  This is not very good in terms of efficiency,
2558 		 * but we can allow this since this option should be rarely
2559 		 * used.
2560 		 */
2561 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2562 			return (error);
2563 	}
2564 
2565 	/*
2566 	 * XXX: Currently, we assume all the optional information is stored
2567 	 * in a single mbuf.
2568 	 */
2569 	if (control->m_next)
2570 		return (EINVAL);
2571 
2572 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2573 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2574 		int error;
2575 
2576 		if (control->m_len < CMSG_LEN(0))
2577 			return (EINVAL);
2578 
2579 		cm = mtod(control, struct cmsghdr *);
2580 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2581 			return (EINVAL);
2582 		if (cm->cmsg_level != IPPROTO_IPV6)
2583 			continue;
2584 
2585 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2586 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2587 		if (error)
2588 			return (error);
2589 	}
2590 
2591 	return (0);
2592 }
2593 
2594 /*
2595  * Set a particular packet option, as a sticky option or an ancillary data
2596  * item.  "len" can be 0 only when it's a sticky option.
2597  * We have 4 cases of combination of "sticky" and "cmsg":
2598  * "sticky=0, cmsg=0": impossible
2599  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2600  * "sticky=1, cmsg=0": RFC3542 socket option
2601  * "sticky=1, cmsg=1": RFC2292 socket option
2602  */
2603 static int
2604 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2605     struct ucred *cred, int sticky, int cmsg, int uproto)
2606 {
2607 	int minmtupolicy, preftemp;
2608 	int error;
2609 
2610 	if (!sticky && !cmsg) {
2611 #ifdef DIAGNOSTIC
2612 		printf("ip6_setpktopt: impossible case\n");
2613 #endif
2614 		return (EINVAL);
2615 	}
2616 
2617 	/*
2618 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2619 	 * not be specified in the context of RFC3542.  Conversely,
2620 	 * RFC3542 types should not be specified in the context of RFC2292.
2621 	 */
2622 	if (!cmsg) {
2623 		switch (optname) {
2624 		case IPV6_2292PKTINFO:
2625 		case IPV6_2292HOPLIMIT:
2626 		case IPV6_2292NEXTHOP:
2627 		case IPV6_2292HOPOPTS:
2628 		case IPV6_2292DSTOPTS:
2629 		case IPV6_2292RTHDR:
2630 		case IPV6_2292PKTOPTIONS:
2631 			return (ENOPROTOOPT);
2632 		}
2633 	}
2634 	if (sticky && cmsg) {
2635 		switch (optname) {
2636 		case IPV6_PKTINFO:
2637 		case IPV6_HOPLIMIT:
2638 		case IPV6_NEXTHOP:
2639 		case IPV6_HOPOPTS:
2640 		case IPV6_DSTOPTS:
2641 		case IPV6_RTHDRDSTOPTS:
2642 		case IPV6_RTHDR:
2643 		case IPV6_USE_MIN_MTU:
2644 		case IPV6_DONTFRAG:
2645 		case IPV6_TCLASS:
2646 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2647 			return (ENOPROTOOPT);
2648 		}
2649 	}
2650 
2651 	switch (optname) {
2652 	case IPV6_2292PKTINFO:
2653 	case IPV6_PKTINFO:
2654 	{
2655 		struct ifnet *ifp = NULL;
2656 		struct in6_pktinfo *pktinfo;
2657 
2658 		if (len != sizeof(struct in6_pktinfo))
2659 			return (EINVAL);
2660 
2661 		pktinfo = (struct in6_pktinfo *)buf;
2662 
2663 		/*
2664 		 * An application can clear any sticky IPV6_PKTINFO option by
2665 		 * doing a "regular" setsockopt with ipi6_addr being
2666 		 * in6addr_any and ipi6_ifindex being zero.
2667 		 * [RFC 3542, Section 6]
2668 		 */
2669 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2670 		    pktinfo->ipi6_ifindex == 0 &&
2671 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2672 			ip6_clearpktopts(opt, optname);
2673 			break;
2674 		}
2675 
2676 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2677 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678 			return (EINVAL);
2679 		}
2680 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2681 			return (EINVAL);
2682 		/* validate the interface index if specified. */
2683 		if (pktinfo->ipi6_ifindex > V_if_index)
2684 			 return (ENXIO);
2685 		if (pktinfo->ipi6_ifindex) {
2686 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2687 			if (ifp == NULL)
2688 				return (ENXIO);
2689 		}
2690 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2691 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2692 			return (ENETDOWN);
2693 
2694 		if (ifp != NULL &&
2695 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2696 			struct in6_ifaddr *ia;
2697 
2698 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2699 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2700 			if (ia == NULL)
2701 				return (EADDRNOTAVAIL);
2702 			ifa_free(&ia->ia_ifa);
2703 		}
2704 		/*
2705 		 * We store the address anyway, and let in6_selectsrc()
2706 		 * validate the specified address.  This is because ipi6_addr
2707 		 * may not have enough information about its scope zone, and
2708 		 * we may need additional information (such as outgoing
2709 		 * interface or the scope zone of a destination address) to
2710 		 * disambiguate the scope.
2711 		 * XXX: the delay of the validation may confuse the
2712 		 * application when it is used as a sticky option.
2713 		 */
2714 		if (opt->ip6po_pktinfo == NULL) {
2715 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2716 			    M_IP6OPT, M_NOWAIT);
2717 			if (opt->ip6po_pktinfo == NULL)
2718 				return (ENOBUFS);
2719 		}
2720 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2721 		break;
2722 	}
2723 
2724 	case IPV6_2292HOPLIMIT:
2725 	case IPV6_HOPLIMIT:
2726 	{
2727 		int *hlimp;
2728 
2729 		/*
2730 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2731 		 * to simplify the ordering among hoplimit options.
2732 		 */
2733 		if (optname == IPV6_HOPLIMIT && sticky)
2734 			return (ENOPROTOOPT);
2735 
2736 		if (len != sizeof(int))
2737 			return (EINVAL);
2738 		hlimp = (int *)buf;
2739 		if (*hlimp < -1 || *hlimp > 255)
2740 			return (EINVAL);
2741 
2742 		opt->ip6po_hlim = *hlimp;
2743 		break;
2744 	}
2745 
2746 	case IPV6_TCLASS:
2747 	{
2748 		int tclass;
2749 
2750 		if (len != sizeof(int))
2751 			return (EINVAL);
2752 		tclass = *(int *)buf;
2753 		if (tclass < -1 || tclass > 255)
2754 			return (EINVAL);
2755 
2756 		opt->ip6po_tclass = tclass;
2757 		break;
2758 	}
2759 
2760 	case IPV6_2292NEXTHOP:
2761 	case IPV6_NEXTHOP:
2762 		if (cred != NULL) {
2763 			error = priv_check_cred(cred,
2764 			    PRIV_NETINET_SETHDROPTS, 0);
2765 			if (error)
2766 				return (error);
2767 		}
2768 
2769 		if (len == 0) {	/* just remove the option */
2770 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2771 			break;
2772 		}
2773 
2774 		/* check if cmsg_len is large enough for sa_len */
2775 		if (len < sizeof(struct sockaddr) || len < *buf)
2776 			return (EINVAL);
2777 
2778 		switch (((struct sockaddr *)buf)->sa_family) {
2779 		case AF_INET6:
2780 		{
2781 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2782 			int error;
2783 
2784 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2785 				return (EINVAL);
2786 
2787 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2788 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2789 				return (EINVAL);
2790 			}
2791 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2792 			    != 0) {
2793 				return (error);
2794 			}
2795 			break;
2796 		}
2797 		case AF_LINK:	/* should eventually be supported */
2798 		default:
2799 			return (EAFNOSUPPORT);
2800 		}
2801 
2802 		/* turn off the previous option, then set the new option. */
2803 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2804 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2805 		if (opt->ip6po_nexthop == NULL)
2806 			return (ENOBUFS);
2807 		bcopy(buf, opt->ip6po_nexthop, *buf);
2808 		break;
2809 
2810 	case IPV6_2292HOPOPTS:
2811 	case IPV6_HOPOPTS:
2812 	{
2813 		struct ip6_hbh *hbh;
2814 		int hbhlen;
2815 
2816 		/*
2817 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2818 		 * options, since per-option restriction has too much
2819 		 * overhead.
2820 		 */
2821 		if (cred != NULL) {
2822 			error = priv_check_cred(cred,
2823 			    PRIV_NETINET_SETHDROPTS, 0);
2824 			if (error)
2825 				return (error);
2826 		}
2827 
2828 		if (len == 0) {
2829 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2830 			break;	/* just remove the option */
2831 		}
2832 
2833 		/* message length validation */
2834 		if (len < sizeof(struct ip6_hbh))
2835 			return (EINVAL);
2836 		hbh = (struct ip6_hbh *)buf;
2837 		hbhlen = (hbh->ip6h_len + 1) << 3;
2838 		if (len != hbhlen)
2839 			return (EINVAL);
2840 
2841 		/* turn off the previous option, then set the new option. */
2842 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2843 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2844 		if (opt->ip6po_hbh == NULL)
2845 			return (ENOBUFS);
2846 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2847 
2848 		break;
2849 	}
2850 
2851 	case IPV6_2292DSTOPTS:
2852 	case IPV6_DSTOPTS:
2853 	case IPV6_RTHDRDSTOPTS:
2854 	{
2855 		struct ip6_dest *dest, **newdest = NULL;
2856 		int destlen;
2857 
2858 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2859 			error = priv_check_cred(cred,
2860 			    PRIV_NETINET_SETHDROPTS, 0);
2861 			if (error)
2862 				return (error);
2863 		}
2864 
2865 		if (len == 0) {
2866 			ip6_clearpktopts(opt, optname);
2867 			break;	/* just remove the option */
2868 		}
2869 
2870 		/* message length validation */
2871 		if (len < sizeof(struct ip6_dest))
2872 			return (EINVAL);
2873 		dest = (struct ip6_dest *)buf;
2874 		destlen = (dest->ip6d_len + 1) << 3;
2875 		if (len != destlen)
2876 			return (EINVAL);
2877 
2878 		/*
2879 		 * Determine the position that the destination options header
2880 		 * should be inserted; before or after the routing header.
2881 		 */
2882 		switch (optname) {
2883 		case IPV6_2292DSTOPTS:
2884 			/*
2885 			 * The old advacned API is ambiguous on this point.
2886 			 * Our approach is to determine the position based
2887 			 * according to the existence of a routing header.
2888 			 * Note, however, that this depends on the order of the
2889 			 * extension headers in the ancillary data; the 1st
2890 			 * part of the destination options header must appear
2891 			 * before the routing header in the ancillary data,
2892 			 * too.
2893 			 * RFC3542 solved the ambiguity by introducing
2894 			 * separate ancillary data or option types.
2895 			 */
2896 			if (opt->ip6po_rthdr == NULL)
2897 				newdest = &opt->ip6po_dest1;
2898 			else
2899 				newdest = &opt->ip6po_dest2;
2900 			break;
2901 		case IPV6_RTHDRDSTOPTS:
2902 			newdest = &opt->ip6po_dest1;
2903 			break;
2904 		case IPV6_DSTOPTS:
2905 			newdest = &opt->ip6po_dest2;
2906 			break;
2907 		}
2908 
2909 		/* turn off the previous option, then set the new option. */
2910 		ip6_clearpktopts(opt, optname);
2911 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2912 		if (*newdest == NULL)
2913 			return (ENOBUFS);
2914 		bcopy(dest, *newdest, destlen);
2915 
2916 		break;
2917 	}
2918 
2919 	case IPV6_2292RTHDR:
2920 	case IPV6_RTHDR:
2921 	{
2922 		struct ip6_rthdr *rth;
2923 		int rthlen;
2924 
2925 		if (len == 0) {
2926 			ip6_clearpktopts(opt, IPV6_RTHDR);
2927 			break;	/* just remove the option */
2928 		}
2929 
2930 		/* message length validation */
2931 		if (len < sizeof(struct ip6_rthdr))
2932 			return (EINVAL);
2933 		rth = (struct ip6_rthdr *)buf;
2934 		rthlen = (rth->ip6r_len + 1) << 3;
2935 		if (len != rthlen)
2936 			return (EINVAL);
2937 
2938 		switch (rth->ip6r_type) {
2939 		case IPV6_RTHDR_TYPE_0:
2940 			if (rth->ip6r_len == 0)	/* must contain one addr */
2941 				return (EINVAL);
2942 			if (rth->ip6r_len % 2) /* length must be even */
2943 				return (EINVAL);
2944 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2945 				return (EINVAL);
2946 			break;
2947 		default:
2948 			return (EINVAL);	/* not supported */
2949 		}
2950 
2951 		/* turn off the previous option */
2952 		ip6_clearpktopts(opt, IPV6_RTHDR);
2953 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2954 		if (opt->ip6po_rthdr == NULL)
2955 			return (ENOBUFS);
2956 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2957 
2958 		break;
2959 	}
2960 
2961 	case IPV6_USE_MIN_MTU:
2962 		if (len != sizeof(int))
2963 			return (EINVAL);
2964 		minmtupolicy = *(int *)buf;
2965 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2966 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2967 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2968 			return (EINVAL);
2969 		}
2970 		opt->ip6po_minmtu = minmtupolicy;
2971 		break;
2972 
2973 	case IPV6_DONTFRAG:
2974 		if (len != sizeof(int))
2975 			return (EINVAL);
2976 
2977 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2978 			/*
2979 			 * we ignore this option for TCP sockets.
2980 			 * (RFC3542 leaves this case unspecified.)
2981 			 */
2982 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2983 		} else
2984 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2985 		break;
2986 
2987 	case IPV6_PREFER_TEMPADDR:
2988 		if (len != sizeof(int))
2989 			return (EINVAL);
2990 		preftemp = *(int *)buf;
2991 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2992 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2993 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2994 			return (EINVAL);
2995 		}
2996 		opt->ip6po_prefer_tempaddr = preftemp;
2997 		break;
2998 
2999 	default:
3000 		return (ENOPROTOOPT);
3001 	} /* end of switch */
3002 
3003 	return (0);
3004 }
3005 
3006 /*
3007  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3008  * packet to the input queue of a specified interface.  Note that this
3009  * calls the output routine of the loopback "driver", but with an interface
3010  * pointer that might NOT be &loif -- easier than replicating that code here.
3011  */
3012 void
3013 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3014 {
3015 	struct mbuf *copym;
3016 	struct ip6_hdr *ip6;
3017 
3018 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3019 	if (copym == NULL)
3020 		return;
3021 
3022 	/*
3023 	 * Make sure to deep-copy IPv6 header portion in case the data
3024 	 * is in an mbuf cluster, so that we can safely override the IPv6
3025 	 * header portion later.
3026 	 */
3027 	if (!M_WRITABLE(copym) ||
3028 	    copym->m_len < sizeof(struct ip6_hdr)) {
3029 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3030 		if (copym == NULL)
3031 			return;
3032 	}
3033 	ip6 = mtod(copym, struct ip6_hdr *);
3034 	/*
3035 	 * clear embedded scope identifiers if necessary.
3036 	 * in6_clearscope will touch the addresses only when necessary.
3037 	 */
3038 	in6_clearscope(&ip6->ip6_src);
3039 	in6_clearscope(&ip6->ip6_dst);
3040 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3041 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3042 		    CSUM_PSEUDO_HDR;
3043 		copym->m_pkthdr.csum_data = 0xffff;
3044 	}
3045 	if_simloop(ifp, copym, AF_INET6, 0);
3046 }
3047 
3048 /*
3049  * Chop IPv6 header off from the payload.
3050  */
3051 static int
3052 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3053 {
3054 	struct mbuf *mh;
3055 	struct ip6_hdr *ip6;
3056 
3057 	ip6 = mtod(m, struct ip6_hdr *);
3058 	if (m->m_len > sizeof(*ip6)) {
3059 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3060 		if (mh == NULL) {
3061 			m_freem(m);
3062 			return ENOBUFS;
3063 		}
3064 		m_move_pkthdr(mh, m);
3065 		M_ALIGN(mh, sizeof(*ip6));
3066 		m->m_len -= sizeof(*ip6);
3067 		m->m_data += sizeof(*ip6);
3068 		mh->m_next = m;
3069 		m = mh;
3070 		m->m_len = sizeof(*ip6);
3071 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3072 	}
3073 	exthdrs->ip6e_ip6 = m;
3074 	return 0;
3075 }
3076 
3077 /*
3078  * Compute IPv6 extension header length.
3079  */
3080 int
3081 ip6_optlen(struct inpcb *in6p)
3082 {
3083 	int len;
3084 
3085 	if (!in6p->in6p_outputopts)
3086 		return 0;
3087 
3088 	len = 0;
3089 #define elen(x) \
3090     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3091 
3092 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3093 	if (in6p->in6p_outputopts->ip6po_rthdr)
3094 		/* dest1 is valid with rthdr only */
3095 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3096 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3097 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3098 	return len;
3099 #undef elen
3100 }
3101