xref: /freebsd/sys/netinet6/ip6_output.c (revision 4ca50eab86aec8001978a612a42f9ae8388eceab)
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 #ifdef FLOWTABLE
121 #include <net/flowtable.h>
122 #endif
123 
124 extern int in6_mcast_loop;
125 
126 struct ip6_exthdrs {
127 	struct mbuf *ip6e_ip6;
128 	struct mbuf *ip6e_hbh;
129 	struct mbuf *ip6e_dest1;
130 	struct mbuf *ip6e_rthdr;
131 	struct mbuf *ip6e_dest2;
132 };
133 
134 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
135 
136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
137 			   struct ucred *, int);
138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
139 	struct socket *, struct sockopt *);
140 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
142 	struct ucred *, int, int, int);
143 
144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
146 	struct ip6_frag **);
147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
149 static int ip6_getpmtu(struct route_in6 *, int,
150 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
151 	u_int);
152 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
153 	u_long *, int *, u_int);
154 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
155 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
156 
157 
158 /*
159  * Make an extension header from option data.  hp is the source, and
160  * mp is the destination.
161  */
162 #define MAKE_EXTHDR(hp, mp)						\
163     do {								\
164 	if (hp) {							\
165 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
166 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
167 		    ((eh)->ip6e_len + 1) << 3);				\
168 		if (error)						\
169 			goto freehdrs;					\
170 	}								\
171     } while (/*CONSTCOND*/ 0)
172 
173 /*
174  * Form a chain of extension headers.
175  * m is the extension header mbuf
176  * mp is the previous mbuf in the chain
177  * p is the next header
178  * i is the type of option.
179  */
180 #define MAKE_CHAIN(m, mp, p, i)\
181     do {\
182 	if (m) {\
183 		if (!hdrsplit) \
184 			panic("assumption failed: hdr not split"); \
185 		*mtod((m), u_char *) = *(p);\
186 		*(p) = (i);\
187 		p = mtod((m), u_char *);\
188 		(m)->m_next = (mp)->m_next;\
189 		(mp)->m_next = (m);\
190 		(mp) = (m);\
191 	}\
192     } while (/*CONSTCOND*/ 0)
193 
194 void
195 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
196 {
197 	u_short csum;
198 
199 	csum = in_cksum_skip(m, offset + plen, offset);
200 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
201 		csum = 0xffff;
202 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
203 
204 	if (offset + sizeof(u_short) > m->m_len) {
205 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
206 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
207 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
208 		/*
209 		 * XXX this should not happen, but if it does, the correct
210 		 * behavior may be to insert the checksum in the appropriate
211 		 * next mbuf in the chain.
212 		 */
213 		return;
214 	}
215 	*(u_short *)(m->m_data + offset) = csum;
216 }
217 
218 int
219 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
220     int mtu, uint32_t id)
221 {
222 	struct mbuf *m, **mnext, *m_frgpart;
223 	struct ip6_hdr *ip6, *mhip6;
224 	struct ip6_frag *ip6f;
225 	int off;
226 	int error;
227 	int tlen = m0->m_pkthdr.len;
228 
229 	m = m0;
230 	ip6 = mtod(m, struct ip6_hdr *);
231 	mnext = &m->m_nextpkt;
232 
233 	for (off = hlen; off < tlen; off += mtu) {
234 		m = m_gethdr(M_NOWAIT, MT_DATA);
235 		if (!m) {
236 			IP6STAT_INC(ip6s_odropped);
237 			return (ENOBUFS);
238 		}
239 		m->m_flags = m0->m_flags & M_COPYFLAGS;
240 		*mnext = m;
241 		mnext = &m->m_nextpkt;
242 		m->m_data += max_linkhdr;
243 		mhip6 = mtod(m, struct ip6_hdr *);
244 		*mhip6 = *ip6;
245 		m->m_len = sizeof(*mhip6);
246 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
247 		if (error) {
248 			IP6STAT_INC(ip6s_odropped);
249 			return (error);
250 		}
251 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
252 		if (off + mtu >= tlen)
253 			mtu = tlen - off;
254 		else
255 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
256 		mhip6->ip6_plen = htons((u_short)(mtu + hlen +
257 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
258 		if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) {
259 			IP6STAT_INC(ip6s_odropped);
260 			return (ENOBUFS);
261 		}
262 		m_cat(m, m_frgpart);
263 		m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
264 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
265 		m->m_pkthdr.rcvif = NULL;
266 		ip6f->ip6f_reserved = 0;
267 		ip6f->ip6f_ident = id;
268 		ip6f->ip6f_nxt = nextproto;
269 		IP6STAT_INC(ip6s_ofragments);
270 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
271 	}
272 
273 	return (0);
274 }
275 
276 /*
277  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
278  * header (with pri, len, nxt, hlim, src, dst).
279  * This function may modify ver and hlim only.
280  * The mbuf chain containing the packet will be freed.
281  * The mbuf opt, if present, will not be freed.
282  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
283  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
284  * then result of route lookup is stored in ro->ro_rt.
285  *
286  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
287  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
288  * which is rt_mtu.
289  *
290  * ifpp - XXX: just for statistics
291  */
292 /*
293  * XXX TODO: no flowid is assigned for outbound flows?
294  */
295 int
296 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
297     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
298     struct ifnet **ifpp, struct inpcb *inp)
299 {
300 	struct ip6_hdr *ip6;
301 	struct ifnet *ifp, *origifp;
302 	struct mbuf *m = m0;
303 	struct mbuf *mprev = NULL;
304 	int hlen, tlen, len;
305 	struct route_in6 ip6route;
306 	struct rtentry *rt = NULL;
307 	struct sockaddr_in6 *dst, src_sa, dst_sa;
308 	struct in6_addr odst;
309 	int error = 0;
310 	struct in6_ifaddr *ia = NULL;
311 	u_long mtu;
312 	int alwaysfrag, dontfrag;
313 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
314 	struct ip6_exthdrs exthdrs;
315 	struct in6_addr src0, dst0;
316 	u_int32_t zone;
317 	struct route_in6 *ro_pmtu = NULL;
318 	int hdrsplit = 0;
319 	int sw_csum, tso;
320 	int needfiblookup;
321 	uint32_t fibnum;
322 	struct m_tag *fwd_tag = NULL;
323 	uint32_t id;
324 
325 	if (inp != NULL) {
326 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
327 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
328 			/* unconditionally set flowid */
329 			m->m_pkthdr.flowid = inp->inp_flowid;
330 			M_HASHTYPE_SET(m, inp->inp_flowtype);
331 		}
332 	}
333 
334 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
335 	/*
336 	 * IPSec checking which handles several cases.
337 	 * FAST IPSEC: We re-injected the packet.
338 	 * XXX: need scope argument.
339 	 */
340 	if (IPSEC_ENABLED(ipv6)) {
341 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
342 			if (error == EINPROGRESS)
343 				error = 0;
344 			goto done;
345 		}
346 	}
347 #endif /* IPSEC */
348 
349 	bzero(&exthdrs, sizeof(exthdrs));
350 	if (opt) {
351 		/* Hop-by-Hop options header */
352 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
353 		/* Destination options header(1st part) */
354 		if (opt->ip6po_rthdr) {
355 			/*
356 			 * Destination options header(1st part)
357 			 * This only makes sense with a routing header.
358 			 * See Section 9.2 of RFC 3542.
359 			 * Disabling this part just for MIP6 convenience is
360 			 * a bad idea.  We need to think carefully about a
361 			 * way to make the advanced API coexist with MIP6
362 			 * options, which might automatically be inserted in
363 			 * the kernel.
364 			 */
365 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
366 		}
367 		/* Routing header */
368 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
369 		/* Destination options header(2nd part) */
370 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
371 	}
372 
373 	/*
374 	 * Calculate the total length of the extension header chain.
375 	 * Keep the length of the unfragmentable part for fragmentation.
376 	 */
377 	optlen = 0;
378 	if (exthdrs.ip6e_hbh)
379 		optlen += exthdrs.ip6e_hbh->m_len;
380 	if (exthdrs.ip6e_dest1)
381 		optlen += exthdrs.ip6e_dest1->m_len;
382 	if (exthdrs.ip6e_rthdr)
383 		optlen += exthdrs.ip6e_rthdr->m_len;
384 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
385 
386 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
387 	if (exthdrs.ip6e_dest2)
388 		optlen += exthdrs.ip6e_dest2->m_len;
389 
390 	/*
391 	 * If there is at least one extension header,
392 	 * separate IP6 header from the payload.
393 	 */
394 	if (optlen && !hdrsplit) {
395 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
396 			m = NULL;
397 			goto freehdrs;
398 		}
399 		m = exthdrs.ip6e_ip6;
400 		hdrsplit++;
401 	}
402 
403 	ip6 = mtod(m, struct ip6_hdr *);
404 
405 	/* adjust mbuf packet header length */
406 	m->m_pkthdr.len += optlen;
407 	plen = m->m_pkthdr.len - sizeof(*ip6);
408 
409 	/* If this is a jumbo payload, insert a jumbo payload option. */
410 	if (plen > IPV6_MAXPACKET) {
411 		if (!hdrsplit) {
412 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
413 				m = NULL;
414 				goto freehdrs;
415 			}
416 			m = exthdrs.ip6e_ip6;
417 			hdrsplit++;
418 		}
419 		/* adjust pointer */
420 		ip6 = mtod(m, struct ip6_hdr *);
421 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
422 			goto freehdrs;
423 		ip6->ip6_plen = 0;
424 	} else
425 		ip6->ip6_plen = htons(plen);
426 
427 	/*
428 	 * Concatenate headers and fill in next header fields.
429 	 * Here we have, on "m"
430 	 *	IPv6 payload
431 	 * and we insert headers accordingly.  Finally, we should be getting:
432 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
433 	 *
434 	 * during the header composing process, "m" points to IPv6 header.
435 	 * "mprev" points to an extension header prior to esp.
436 	 */
437 	u_char *nexthdrp = &ip6->ip6_nxt;
438 	mprev = m;
439 
440 	/*
441 	 * we treat dest2 specially.  this makes IPsec processing
442 	 * much easier.  the goal here is to make mprev point the
443 	 * mbuf prior to dest2.
444 	 *
445 	 * result: IPv6 dest2 payload
446 	 * m and mprev will point to IPv6 header.
447 	 */
448 	if (exthdrs.ip6e_dest2) {
449 		if (!hdrsplit)
450 			panic("assumption failed: hdr not split");
451 		exthdrs.ip6e_dest2->m_next = m->m_next;
452 		m->m_next = exthdrs.ip6e_dest2;
453 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
454 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
455 	}
456 
457 	/*
458 	 * result: IPv6 hbh dest1 rthdr dest2 payload
459 	 * m will point to IPv6 header.  mprev will point to the
460 	 * extension header prior to dest2 (rthdr in the above case).
461 	 */
462 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
463 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
464 		   IPPROTO_DSTOPTS);
465 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
466 		   IPPROTO_ROUTING);
467 
468 	/*
469 	 * If there is a routing header, discard the packet.
470 	 */
471 	if (exthdrs.ip6e_rthdr) {
472 		 error = EINVAL;
473 		 goto bad;
474 	}
475 
476 	/* Source address validation */
477 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
478 	    (flags & IPV6_UNSPECSRC) == 0) {
479 		error = EOPNOTSUPP;
480 		IP6STAT_INC(ip6s_badscope);
481 		goto bad;
482 	}
483 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
484 		error = EOPNOTSUPP;
485 		IP6STAT_INC(ip6s_badscope);
486 		goto bad;
487 	}
488 
489 	IP6STAT_INC(ip6s_localout);
490 
491 	/*
492 	 * Route packet.
493 	 */
494 	if (ro == NULL) {
495 		ro = &ip6route;
496 		bzero((caddr_t)ro, sizeof(*ro));
497 	}
498 	ro_pmtu = ro;
499 	if (opt && opt->ip6po_rthdr)
500 		ro = &opt->ip6po_route;
501 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
502 #ifdef FLOWTABLE
503 	if (ro->ro_rt == NULL)
504 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
505 #endif
506 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
507 again:
508 	/*
509 	 * if specified, try to fill in the traffic class field.
510 	 * do not override if a non-zero value is already set.
511 	 * we check the diffserv field and the ecn field separately.
512 	 */
513 	if (opt && opt->ip6po_tclass >= 0) {
514 		int mask = 0;
515 
516 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
517 			mask |= 0xfc;
518 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
519 			mask |= 0x03;
520 		if (mask != 0)
521 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
522 	}
523 
524 	/* fill in or override the hop limit field, if necessary. */
525 	if (opt && opt->ip6po_hlim != -1)
526 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
527 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
528 		if (im6o != NULL)
529 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
530 		else
531 			ip6->ip6_hlim = V_ip6_defmcasthlim;
532 	}
533 	/*
534 	 * Validate route against routing table additions;
535 	 * a better/more specific route might have been added.
536 	 * Make sure address family is set in route.
537 	 */
538 	if (inp) {
539 		ro->ro_dst.sin6_family = AF_INET6;
540 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
541 	}
542 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
543 	    ro->ro_dst.sin6_family == AF_INET6 &&
544 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
545 		rt = ro->ro_rt;
546 		ifp = ro->ro_rt->rt_ifp;
547 	} else {
548 		if (ro->ro_lle)
549 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
550 		ro->ro_lle = NULL;
551 		if (fwd_tag == NULL) {
552 			bzero(&dst_sa, sizeof(dst_sa));
553 			dst_sa.sin6_family = AF_INET6;
554 			dst_sa.sin6_len = sizeof(dst_sa);
555 			dst_sa.sin6_addr = ip6->ip6_dst;
556 		}
557 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
558 		    &rt, fibnum);
559 		if (error != 0) {
560 			if (ifp != NULL)
561 				in6_ifstat_inc(ifp, ifs6_out_discard);
562 			goto bad;
563 		}
564 	}
565 	if (rt == NULL) {
566 		/*
567 		 * If in6_selectroute() does not return a route entry,
568 		 * dst may not have been updated.
569 		 */
570 		*dst = dst_sa;	/* XXX */
571 	}
572 
573 	/*
574 	 * then rt (for unicast) and ifp must be non-NULL valid values.
575 	 */
576 	if ((flags & IPV6_FORWARDING) == 0) {
577 		/* XXX: the FORWARDING flag can be set for mrouting. */
578 		in6_ifstat_inc(ifp, ifs6_out_request);
579 	}
580 	if (rt != NULL) {
581 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
582 		counter_u64_add(rt->rt_pksent, 1);
583 	}
584 
585 
586 	/*
587 	 * The outgoing interface must be in the zone of source and
588 	 * destination addresses.
589 	 */
590 	origifp = ifp;
591 
592 	src0 = ip6->ip6_src;
593 	if (in6_setscope(&src0, origifp, &zone))
594 		goto badscope;
595 	bzero(&src_sa, sizeof(src_sa));
596 	src_sa.sin6_family = AF_INET6;
597 	src_sa.sin6_len = sizeof(src_sa);
598 	src_sa.sin6_addr = ip6->ip6_src;
599 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
600 		goto badscope;
601 
602 	dst0 = ip6->ip6_dst;
603 	if (in6_setscope(&dst0, origifp, &zone))
604 		goto badscope;
605 	/* re-initialize to be sure */
606 	bzero(&dst_sa, sizeof(dst_sa));
607 	dst_sa.sin6_family = AF_INET6;
608 	dst_sa.sin6_len = sizeof(dst_sa);
609 	dst_sa.sin6_addr = ip6->ip6_dst;
610 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
611 		goto badscope;
612 	}
613 
614 	/* We should use ia_ifp to support the case of
615 	 * sending packets to an address of our own.
616 	 */
617 	if (ia != NULL && ia->ia_ifp)
618 		ifp = ia->ia_ifp;
619 
620 	/* scope check is done. */
621 	goto routefound;
622 
623   badscope:
624 	IP6STAT_INC(ip6s_badscope);
625 	in6_ifstat_inc(origifp, ifs6_out_discard);
626 	if (error == 0)
627 		error = EHOSTUNREACH; /* XXX */
628 	goto bad;
629 
630   routefound:
631 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
632 		if (opt && opt->ip6po_nextroute.ro_rt) {
633 			/*
634 			 * The nexthop is explicitly specified by the
635 			 * application.  We assume the next hop is an IPv6
636 			 * address.
637 			 */
638 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
639 		}
640 		else if ((rt->rt_flags & RTF_GATEWAY))
641 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
642 	}
643 
644 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
645 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
646 	} else {
647 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
648 		in6_ifstat_inc(ifp, ifs6_out_mcast);
649 		/*
650 		 * Confirm that the outgoing interface supports multicast.
651 		 */
652 		if (!(ifp->if_flags & IFF_MULTICAST)) {
653 			IP6STAT_INC(ip6s_noroute);
654 			in6_ifstat_inc(ifp, ifs6_out_discard);
655 			error = ENETUNREACH;
656 			goto bad;
657 		}
658 		if ((im6o == NULL && in6_mcast_loop) ||
659 		    (im6o && im6o->im6o_multicast_loop)) {
660 			/*
661 			 * Loop back multicast datagram if not expressly
662 			 * forbidden to do so, even if we have not joined
663 			 * the address; protocols will filter it later,
664 			 * thus deferring a hash lookup and lock acquisition
665 			 * at the expense of an m_copym().
666 			 */
667 			ip6_mloopback(ifp, m);
668 		} else {
669 			/*
670 			 * If we are acting as a multicast router, perform
671 			 * multicast forwarding as if the packet had just
672 			 * arrived on the interface to which we are about
673 			 * to send.  The multicast forwarding function
674 			 * recursively calls this function, using the
675 			 * IPV6_FORWARDING flag to prevent infinite recursion.
676 			 *
677 			 * Multicasts that are looped back by ip6_mloopback(),
678 			 * above, will be forwarded by the ip6_input() routine,
679 			 * if necessary.
680 			 */
681 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
682 				/*
683 				 * XXX: ip6_mforward expects that rcvif is NULL
684 				 * when it is called from the originating path.
685 				 * However, it may not always be the case.
686 				 */
687 				m->m_pkthdr.rcvif = NULL;
688 				if (ip6_mforward(ip6, ifp, m) != 0) {
689 					m_freem(m);
690 					goto done;
691 				}
692 			}
693 		}
694 		/*
695 		 * Multicasts with a hoplimit of zero may be looped back,
696 		 * above, but must not be transmitted on a network.
697 		 * Also, multicasts addressed to the loopback interface
698 		 * are not sent -- the above call to ip6_mloopback() will
699 		 * loop back a copy if this host actually belongs to the
700 		 * destination group on the loopback interface.
701 		 */
702 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
703 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
704 			m_freem(m);
705 			goto done;
706 		}
707 	}
708 
709 	/*
710 	 * Fill the outgoing inteface to tell the upper layer
711 	 * to increment per-interface statistics.
712 	 */
713 	if (ifpp)
714 		*ifpp = ifp;
715 
716 	/* Determine path MTU. */
717 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
718 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
719 		goto bad;
720 
721 	/*
722 	 * The caller of this function may specify to use the minimum MTU
723 	 * in some cases.
724 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
725 	 * setting.  The logic is a bit complicated; by default, unicast
726 	 * packets will follow path MTU while multicast packets will be sent at
727 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
728 	 * including unicast ones will be sent at the minimum MTU.  Multicast
729 	 * packets will always be sent at the minimum MTU unless
730 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
731 	 * See RFC 3542 for more details.
732 	 */
733 	if (mtu > IPV6_MMTU) {
734 		if ((flags & IPV6_MINMTU))
735 			mtu = IPV6_MMTU;
736 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
737 			mtu = IPV6_MMTU;
738 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
739 			 (opt == NULL ||
740 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
741 			mtu = IPV6_MMTU;
742 		}
743 	}
744 
745 	/*
746 	 * clear embedded scope identifiers if necessary.
747 	 * in6_clearscope will touch the addresses only when necessary.
748 	 */
749 	in6_clearscope(&ip6->ip6_src);
750 	in6_clearscope(&ip6->ip6_dst);
751 
752 	/*
753 	 * If the outgoing packet contains a hop-by-hop options header,
754 	 * it must be examined and processed even by the source node.
755 	 * (RFC 2460, section 4.)
756 	 */
757 	if (exthdrs.ip6e_hbh) {
758 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
759 		u_int32_t dummy; /* XXX unused */
760 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
761 
762 #ifdef DIAGNOSTIC
763 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
764 			panic("ip6e_hbh is not contiguous");
765 #endif
766 		/*
767 		 *  XXX: if we have to send an ICMPv6 error to the sender,
768 		 *       we need the M_LOOP flag since icmp6_error() expects
769 		 *       the IPv6 and the hop-by-hop options header are
770 		 *       contiguous unless the flag is set.
771 		 */
772 		m->m_flags |= M_LOOP;
773 		m->m_pkthdr.rcvif = ifp;
774 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
775 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
776 		    &dummy, &plen) < 0) {
777 			/* m was already freed at this point */
778 			error = EINVAL;/* better error? */
779 			goto done;
780 		}
781 		m->m_flags &= ~M_LOOP; /* XXX */
782 		m->m_pkthdr.rcvif = NULL;
783 	}
784 
785 	/* Jump over all PFIL processing if hooks are not active. */
786 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
787 		goto passout;
788 
789 	odst = ip6->ip6_dst;
790 	/* Run through list of hooks for output packets. */
791 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
792 	if (error != 0 || m == NULL)
793 		goto done;
794 	/* adjust pointer */
795 	ip6 = mtod(m, struct ip6_hdr *);
796 
797 	needfiblookup = 0;
798 	/* See if destination IP address was changed by packet filter. */
799 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
800 		m->m_flags |= M_SKIP_FIREWALL;
801 		/* If destination is now ourself drop to ip6_input(). */
802 		if (in6_localip(&ip6->ip6_dst)) {
803 			m->m_flags |= M_FASTFWD_OURS;
804 			if (m->m_pkthdr.rcvif == NULL)
805 				m->m_pkthdr.rcvif = V_loif;
806 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
807 				m->m_pkthdr.csum_flags |=
808 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
809 				m->m_pkthdr.csum_data = 0xffff;
810 			}
811 #ifdef SCTP
812 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
813 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
814 #endif
815 			error = netisr_queue(NETISR_IPV6, m);
816 			goto done;
817 		} else {
818 			RO_RTFREE(ro);
819 			needfiblookup = 1; /* Redo the routing table lookup. */
820 			if (ro->ro_lle)
821 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
822 			ro->ro_lle = NULL;
823 		}
824 	}
825 	/* See if fib was changed by packet filter. */
826 	if (fibnum != M_GETFIB(m)) {
827 		m->m_flags |= M_SKIP_FIREWALL;
828 		fibnum = M_GETFIB(m);
829 		RO_RTFREE(ro);
830 		needfiblookup = 1;
831 		if (ro->ro_lle)
832 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
833 		ro->ro_lle = NULL;
834 	}
835 	if (needfiblookup)
836 		goto again;
837 
838 	/* See if local, if yes, send it to netisr. */
839 	if (m->m_flags & M_FASTFWD_OURS) {
840 		if (m->m_pkthdr.rcvif == NULL)
841 			m->m_pkthdr.rcvif = V_loif;
842 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
843 			m->m_pkthdr.csum_flags |=
844 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
845 			m->m_pkthdr.csum_data = 0xffff;
846 		}
847 #ifdef SCTP
848 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
849 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
850 #endif
851 		error = netisr_queue(NETISR_IPV6, m);
852 		goto done;
853 	}
854 	/* Or forward to some other address? */
855 	if ((m->m_flags & M_IP6_NEXTHOP) &&
856 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
857 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
858 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
859 		m->m_flags |= M_SKIP_FIREWALL;
860 		m->m_flags &= ~M_IP6_NEXTHOP;
861 		m_tag_delete(m, fwd_tag);
862 		goto again;
863 	}
864 
865 passout:
866 	/*
867 	 * Send the packet to the outgoing interface.
868 	 * If necessary, do IPv6 fragmentation before sending.
869 	 *
870 	 * the logic here is rather complex:
871 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
872 	 * 1-a:	send as is if tlen <= path mtu
873 	 * 1-b:	fragment if tlen > path mtu
874 	 *
875 	 * 2: if user asks us not to fragment (dontfrag == 1)
876 	 * 2-a:	send as is if tlen <= interface mtu
877 	 * 2-b:	error if tlen > interface mtu
878 	 *
879 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
880 	 *	always fragment
881 	 *
882 	 * 4: if dontfrag == 1 && alwaysfrag == 1
883 	 *	error, as we cannot handle this conflicting request
884 	 */
885 	sw_csum = m->m_pkthdr.csum_flags;
886 	if (!hdrsplit) {
887 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
888 		sw_csum &= ~ifp->if_hwassist;
889 	} else
890 		tso = 0;
891 	/*
892 	 * If we added extension headers, we will not do TSO and calculate the
893 	 * checksums ourselves for now.
894 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
895 	 * with ext. hdrs.
896 	 */
897 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
898 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
899 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
900 	}
901 #ifdef SCTP
902 	if (sw_csum & CSUM_SCTP_IPV6) {
903 		sw_csum &= ~CSUM_SCTP_IPV6;
904 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
905 	}
906 #endif
907 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
908 	tlen = m->m_pkthdr.len;
909 
910 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
911 		dontfrag = 1;
912 	else
913 		dontfrag = 0;
914 	if (dontfrag && alwaysfrag) {	/* case 4 */
915 		/* conflicting request - can't transmit */
916 		error = EMSGSIZE;
917 		goto bad;
918 	}
919 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
920 		/*
921 		 * Even if the DONTFRAG option is specified, we cannot send the
922 		 * packet when the data length is larger than the MTU of the
923 		 * outgoing interface.
924 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
925 		 * application wanted to know the MTU value. Also return an
926 		 * error code (this is not described in the API spec).
927 		 */
928 		if (inp != NULL)
929 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
930 		error = EMSGSIZE;
931 		goto bad;
932 	}
933 
934 	/*
935 	 * transmit packet without fragmentation
936 	 */
937 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
938 		struct in6_ifaddr *ia6;
939 
940 		ip6 = mtod(m, struct ip6_hdr *);
941 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
942 		if (ia6) {
943 			/* Record statistics for this interface address. */
944 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
945 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
946 			    m->m_pkthdr.len);
947 			ifa_free(&ia6->ia_ifa);
948 		}
949 #ifdef RATELIMIT
950 		if (inp != NULL) {
951 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
952 				in_pcboutput_txrtlmt(inp, ifp, m);
953 			/* stamp send tag on mbuf */
954 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
955 		} else {
956 			m->m_pkthdr.snd_tag = NULL;
957 		}
958 #endif
959 		error = nd6_output_ifp(ifp, origifp, m, dst,
960 		    (struct route *)ro);
961 #ifdef RATELIMIT
962 		/* check for route change */
963 		if (error == EAGAIN)
964 			in_pcboutput_eagain(inp);
965 #endif
966 		goto done;
967 	}
968 
969 	/*
970 	 * try to fragment the packet.  case 1-b and 3
971 	 */
972 	if (mtu < IPV6_MMTU) {
973 		/* path MTU cannot be less than IPV6_MMTU */
974 		error = EMSGSIZE;
975 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
976 		goto bad;
977 	} else if (ip6->ip6_plen == 0) {
978 		/* jumbo payload cannot be fragmented */
979 		error = EMSGSIZE;
980 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
981 		goto bad;
982 	} else {
983 		u_char nextproto;
984 
985 		/*
986 		 * Too large for the destination or interface;
987 		 * fragment if possible.
988 		 * Must be able to put at least 8 bytes per fragment.
989 		 */
990 		hlen = unfragpartlen;
991 		if (mtu > IPV6_MAXPACKET)
992 			mtu = IPV6_MAXPACKET;
993 
994 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
995 		if (len < 8) {
996 			error = EMSGSIZE;
997 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
998 			goto bad;
999 		}
1000 
1001 		/*
1002 		 * If the interface will not calculate checksums on
1003 		 * fragmented packets, then do it here.
1004 		 * XXX-BZ handle the hw offloading case.  Need flags.
1005 		 */
1006 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1007 			in6_delayed_cksum(m, plen, hlen);
1008 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1009 		}
1010 #ifdef SCTP
1011 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1012 			sctp_delayed_cksum(m, hlen);
1013 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1014 		}
1015 #endif
1016 		/*
1017 		 * Change the next header field of the last header in the
1018 		 * unfragmentable part.
1019 		 */
1020 		if (exthdrs.ip6e_rthdr) {
1021 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1022 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1023 		} else if (exthdrs.ip6e_dest1) {
1024 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1025 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1026 		} else if (exthdrs.ip6e_hbh) {
1027 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1028 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1029 		} else {
1030 			nextproto = ip6->ip6_nxt;
1031 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1032 		}
1033 
1034 		/*
1035 		 * Loop through length of segment after first fragment,
1036 		 * make new header and copy data of each part and link onto
1037 		 * chain.
1038 		 */
1039 		m0 = m;
1040 		id = htonl(ip6_randomid());
1041 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1042 			goto sendorfree;
1043 
1044 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1045 	}
1046 
1047 	/*
1048 	 * Remove leading garbages.
1049 	 */
1050 sendorfree:
1051 	m = m0->m_nextpkt;
1052 	m0->m_nextpkt = 0;
1053 	m_freem(m0);
1054 	for (m0 = m; m; m = m0) {
1055 		m0 = m->m_nextpkt;
1056 		m->m_nextpkt = 0;
1057 		if (error == 0) {
1058 			/* Record statistics for this interface address. */
1059 			if (ia) {
1060 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1061 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1062 				    m->m_pkthdr.len);
1063 			}
1064 #ifdef RATELIMIT
1065 			if (inp != NULL) {
1066 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1067 					in_pcboutput_txrtlmt(inp, ifp, m);
1068 				/* stamp send tag on mbuf */
1069 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1070 			} else {
1071 				m->m_pkthdr.snd_tag = NULL;
1072 			}
1073 #endif
1074 			error = nd6_output_ifp(ifp, origifp, m, dst,
1075 			    (struct route *)ro);
1076 #ifdef RATELIMIT
1077 			/* check for route change */
1078 			if (error == EAGAIN)
1079 				in_pcboutput_eagain(inp);
1080 #endif
1081 		} else
1082 			m_freem(m);
1083 	}
1084 
1085 	if (error == 0)
1086 		IP6STAT_INC(ip6s_fragmented);
1087 
1088 done:
1089 	if (ro == &ip6route)
1090 		RO_RTFREE(ro);
1091 	return (error);
1092 
1093 freehdrs:
1094 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1095 	m_freem(exthdrs.ip6e_dest1);
1096 	m_freem(exthdrs.ip6e_rthdr);
1097 	m_freem(exthdrs.ip6e_dest2);
1098 	/* FALLTHROUGH */
1099 bad:
1100 	if (m)
1101 		m_freem(m);
1102 	goto done;
1103 }
1104 
1105 static int
1106 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1107 {
1108 	struct mbuf *m;
1109 
1110 	if (hlen > MCLBYTES)
1111 		return (ENOBUFS); /* XXX */
1112 
1113 	if (hlen > MLEN)
1114 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1115 	else
1116 		m = m_get(M_NOWAIT, MT_DATA);
1117 	if (m == NULL)
1118 		return (ENOBUFS);
1119 	m->m_len = hlen;
1120 	if (hdr)
1121 		bcopy(hdr, mtod(m, caddr_t), hlen);
1122 
1123 	*mp = m;
1124 	return (0);
1125 }
1126 
1127 /*
1128  * Insert jumbo payload option.
1129  */
1130 static int
1131 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1132 {
1133 	struct mbuf *mopt;
1134 	u_char *optbuf;
1135 	u_int32_t v;
1136 
1137 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1138 
1139 	/*
1140 	 * If there is no hop-by-hop options header, allocate new one.
1141 	 * If there is one but it doesn't have enough space to store the
1142 	 * jumbo payload option, allocate a cluster to store the whole options.
1143 	 * Otherwise, use it to store the options.
1144 	 */
1145 	if (exthdrs->ip6e_hbh == NULL) {
1146 		mopt = m_get(M_NOWAIT, MT_DATA);
1147 		if (mopt == NULL)
1148 			return (ENOBUFS);
1149 		mopt->m_len = JUMBOOPTLEN;
1150 		optbuf = mtod(mopt, u_char *);
1151 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1152 		exthdrs->ip6e_hbh = mopt;
1153 	} else {
1154 		struct ip6_hbh *hbh;
1155 
1156 		mopt = exthdrs->ip6e_hbh;
1157 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1158 			/*
1159 			 * XXX assumption:
1160 			 * - exthdrs->ip6e_hbh is not referenced from places
1161 			 *   other than exthdrs.
1162 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1163 			 */
1164 			int oldoptlen = mopt->m_len;
1165 			struct mbuf *n;
1166 
1167 			/*
1168 			 * XXX: give up if the whole (new) hbh header does
1169 			 * not fit even in an mbuf cluster.
1170 			 */
1171 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1172 				return (ENOBUFS);
1173 
1174 			/*
1175 			 * As a consequence, we must always prepare a cluster
1176 			 * at this point.
1177 			 */
1178 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1179 			if (n == NULL)
1180 				return (ENOBUFS);
1181 			n->m_len = oldoptlen + JUMBOOPTLEN;
1182 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1183 			    oldoptlen);
1184 			optbuf = mtod(n, caddr_t) + oldoptlen;
1185 			m_freem(mopt);
1186 			mopt = exthdrs->ip6e_hbh = n;
1187 		} else {
1188 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1189 			mopt->m_len += JUMBOOPTLEN;
1190 		}
1191 		optbuf[0] = IP6OPT_PADN;
1192 		optbuf[1] = 1;
1193 
1194 		/*
1195 		 * Adjust the header length according to the pad and
1196 		 * the jumbo payload option.
1197 		 */
1198 		hbh = mtod(mopt, struct ip6_hbh *);
1199 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1200 	}
1201 
1202 	/* fill in the option. */
1203 	optbuf[2] = IP6OPT_JUMBO;
1204 	optbuf[3] = 4;
1205 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1206 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1207 
1208 	/* finally, adjust the packet header length */
1209 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1210 
1211 	return (0);
1212 #undef JUMBOOPTLEN
1213 }
1214 
1215 /*
1216  * Insert fragment header and copy unfragmentable header portions.
1217  */
1218 static int
1219 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1220     struct ip6_frag **frghdrp)
1221 {
1222 	struct mbuf *n, *mlast;
1223 
1224 	if (hlen > sizeof(struct ip6_hdr)) {
1225 		n = m_copym(m0, sizeof(struct ip6_hdr),
1226 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1227 		if (n == NULL)
1228 			return (ENOBUFS);
1229 		m->m_next = n;
1230 	} else
1231 		n = m;
1232 
1233 	/* Search for the last mbuf of unfragmentable part. */
1234 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1235 		;
1236 
1237 	if (M_WRITABLE(mlast) &&
1238 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1239 		/* use the trailing space of the last mbuf for the fragment hdr */
1240 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1241 		    mlast->m_len);
1242 		mlast->m_len += sizeof(struct ip6_frag);
1243 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1244 	} else {
1245 		/* allocate a new mbuf for the fragment header */
1246 		struct mbuf *mfrg;
1247 
1248 		mfrg = m_get(M_NOWAIT, MT_DATA);
1249 		if (mfrg == NULL)
1250 			return (ENOBUFS);
1251 		mfrg->m_len = sizeof(struct ip6_frag);
1252 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1253 		mlast->m_next = mfrg;
1254 	}
1255 
1256 	return (0);
1257 }
1258 
1259 /*
1260  * Calculates IPv6 path mtu for destination @dst.
1261  * Resulting MTU is stored in @mtup.
1262  *
1263  * Returns 0 on success.
1264  */
1265 static int
1266 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1267 {
1268 	struct nhop6_extended nh6;
1269 	struct in6_addr kdst;
1270 	uint32_t scopeid;
1271 	struct ifnet *ifp;
1272 	u_long mtu;
1273 	int error;
1274 
1275 	in6_splitscope(dst, &kdst, &scopeid);
1276 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1277 		return (EHOSTUNREACH);
1278 
1279 	ifp = nh6.nh_ifp;
1280 	mtu = nh6.nh_mtu;
1281 
1282 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1283 	fib6_free_nh_ext(fibnum, &nh6);
1284 
1285 	return (error);
1286 }
1287 
1288 /*
1289  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1290  * and cached data in @ro_pmtu.
1291  * MTU from (successful) route lookup is saved (along with dst)
1292  * inside @ro_pmtu to avoid subsequent route lookups after packet
1293  * filter processing.
1294  *
1295  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1296  * Returns 0 on success.
1297  */
1298 static int
1299 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1300     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1301     int *alwaysfragp, u_int fibnum, u_int proto)
1302 {
1303 	struct nhop6_basic nh6;
1304 	struct in6_addr kdst;
1305 	uint32_t scopeid;
1306 	struct sockaddr_in6 *sa6_dst;
1307 	u_long mtu;
1308 
1309 	mtu = 0;
1310 	if (do_lookup) {
1311 
1312 		/*
1313 		 * Here ro_pmtu has final destination address, while
1314 		 * ro might represent immediate destination.
1315 		 * Use ro_pmtu destination since mtu might differ.
1316 		 */
1317 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1318 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1319 			ro_pmtu->ro_mtu = 0;
1320 
1321 		if (ro_pmtu->ro_mtu == 0) {
1322 			bzero(sa6_dst, sizeof(*sa6_dst));
1323 			sa6_dst->sin6_family = AF_INET6;
1324 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1325 			sa6_dst->sin6_addr = *dst;
1326 
1327 			in6_splitscope(dst, &kdst, &scopeid);
1328 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1329 			    &nh6) == 0)
1330 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1331 		}
1332 
1333 		mtu = ro_pmtu->ro_mtu;
1334 	}
1335 
1336 	if (ro_pmtu->ro_rt)
1337 		mtu = ro_pmtu->ro_rt->rt_mtu;
1338 
1339 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1340 }
1341 
1342 /*
1343  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1344  * hostcache data for @dst.
1345  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1346  *
1347  * Returns 0 on success.
1348  */
1349 static int
1350 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1351     u_long *mtup, int *alwaysfragp, u_int proto)
1352 {
1353 	u_long mtu = 0;
1354 	int alwaysfrag = 0;
1355 	int error = 0;
1356 
1357 	if (rt_mtu > 0) {
1358 		u_int32_t ifmtu;
1359 		struct in_conninfo inc;
1360 
1361 		bzero(&inc, sizeof(inc));
1362 		inc.inc_flags |= INC_ISIPV6;
1363 		inc.inc6_faddr = *dst;
1364 
1365 		ifmtu = IN6_LINKMTU(ifp);
1366 
1367 		/* TCP is known to react to pmtu changes so skip hc */
1368 		if (proto != IPPROTO_TCP)
1369 			mtu = tcp_hc_getmtu(&inc);
1370 
1371 		if (mtu)
1372 			mtu = min(mtu, rt_mtu);
1373 		else
1374 			mtu = rt_mtu;
1375 		if (mtu == 0)
1376 			mtu = ifmtu;
1377 		else if (mtu < IPV6_MMTU) {
1378 			/*
1379 			 * RFC2460 section 5, last paragraph:
1380 			 * if we record ICMPv6 too big message with
1381 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1382 			 * or smaller, with framgent header attached.
1383 			 * (fragment header is needed regardless from the
1384 			 * packet size, for translators to identify packets)
1385 			 */
1386 			alwaysfrag = 1;
1387 			mtu = IPV6_MMTU;
1388 		}
1389 	} else if (ifp) {
1390 		mtu = IN6_LINKMTU(ifp);
1391 	} else
1392 		error = EHOSTUNREACH; /* XXX */
1393 
1394 	*mtup = mtu;
1395 	if (alwaysfragp)
1396 		*alwaysfragp = alwaysfrag;
1397 	return (error);
1398 }
1399 
1400 /*
1401  * IP6 socket option processing.
1402  */
1403 int
1404 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1405 {
1406 	int optdatalen, uproto;
1407 	void *optdata;
1408 	struct inpcb *in6p = sotoinpcb(so);
1409 	int error, optval;
1410 	int level, op, optname;
1411 	int optlen;
1412 	struct thread *td;
1413 #ifdef	RSS
1414 	uint32_t rss_bucket;
1415 	int retval;
1416 #endif
1417 
1418 /*
1419  * Don't use more than a quarter of mbuf clusters.  N.B.:
1420  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1421  * on LP64 architectures, so cast to u_long to avoid undefined
1422  * behavior.  ILP32 architectures cannot have nmbclusters
1423  * large enough to overflow for other reasons.
1424  */
1425 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1426 
1427 	level = sopt->sopt_level;
1428 	op = sopt->sopt_dir;
1429 	optname = sopt->sopt_name;
1430 	optlen = sopt->sopt_valsize;
1431 	td = sopt->sopt_td;
1432 	error = 0;
1433 	optval = 0;
1434 	uproto = (int)so->so_proto->pr_protocol;
1435 
1436 	if (level != IPPROTO_IPV6) {
1437 		error = EINVAL;
1438 
1439 		if (sopt->sopt_level == SOL_SOCKET &&
1440 		    sopt->sopt_dir == SOPT_SET) {
1441 			switch (sopt->sopt_name) {
1442 			case SO_REUSEADDR:
1443 				INP_WLOCK(in6p);
1444 				if ((so->so_options & SO_REUSEADDR) != 0)
1445 					in6p->inp_flags2 |= INP_REUSEADDR;
1446 				else
1447 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1448 				INP_WUNLOCK(in6p);
1449 				error = 0;
1450 				break;
1451 			case SO_REUSEPORT:
1452 				INP_WLOCK(in6p);
1453 				if ((so->so_options & SO_REUSEPORT) != 0)
1454 					in6p->inp_flags2 |= INP_REUSEPORT;
1455 				else
1456 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1457 				INP_WUNLOCK(in6p);
1458 				error = 0;
1459 				break;
1460 			case SO_SETFIB:
1461 				INP_WLOCK(in6p);
1462 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1463 				INP_WUNLOCK(in6p);
1464 				error = 0;
1465 				break;
1466 			case SO_MAX_PACING_RATE:
1467 #ifdef RATELIMIT
1468 				INP_WLOCK(in6p);
1469 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1470 				INP_WUNLOCK(in6p);
1471 				error = 0;
1472 #else
1473 				error = EOPNOTSUPP;
1474 #endif
1475 				break;
1476 			default:
1477 				break;
1478 			}
1479 		}
1480 	} else {		/* level == IPPROTO_IPV6 */
1481 		switch (op) {
1482 
1483 		case SOPT_SET:
1484 			switch (optname) {
1485 			case IPV6_2292PKTOPTIONS:
1486 #ifdef IPV6_PKTOPTIONS
1487 			case IPV6_PKTOPTIONS:
1488 #endif
1489 			{
1490 				struct mbuf *m;
1491 
1492 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1493 					printf("ip6_ctloutput: mbuf limit hit\n");
1494 					error = ENOBUFS;
1495 					break;
1496 				}
1497 
1498 				error = soopt_getm(sopt, &m); /* XXX */
1499 				if (error != 0)
1500 					break;
1501 				error = soopt_mcopyin(sopt, m); /* XXX */
1502 				if (error != 0)
1503 					break;
1504 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1505 						    m, so, sopt);
1506 				m_freem(m); /* XXX */
1507 				break;
1508 			}
1509 
1510 			/*
1511 			 * Use of some Hop-by-Hop options or some
1512 			 * Destination options, might require special
1513 			 * privilege.  That is, normal applications
1514 			 * (without special privilege) might be forbidden
1515 			 * from setting certain options in outgoing packets,
1516 			 * and might never see certain options in received
1517 			 * packets. [RFC 2292 Section 6]
1518 			 * KAME specific note:
1519 			 *  KAME prevents non-privileged users from sending or
1520 			 *  receiving ANY hbh/dst options in order to avoid
1521 			 *  overhead of parsing options in the kernel.
1522 			 */
1523 			case IPV6_RECVHOPOPTS:
1524 			case IPV6_RECVDSTOPTS:
1525 			case IPV6_RECVRTHDRDSTOPTS:
1526 				if (td != NULL) {
1527 					error = priv_check(td,
1528 					    PRIV_NETINET_SETHDROPTS);
1529 					if (error)
1530 						break;
1531 				}
1532 				/* FALLTHROUGH */
1533 			case IPV6_UNICAST_HOPS:
1534 			case IPV6_HOPLIMIT:
1535 
1536 			case IPV6_RECVPKTINFO:
1537 			case IPV6_RECVHOPLIMIT:
1538 			case IPV6_RECVRTHDR:
1539 			case IPV6_RECVPATHMTU:
1540 			case IPV6_RECVTCLASS:
1541 			case IPV6_RECVFLOWID:
1542 #ifdef	RSS
1543 			case IPV6_RECVRSSBUCKETID:
1544 #endif
1545 			case IPV6_V6ONLY:
1546 			case IPV6_AUTOFLOWLABEL:
1547 			case IPV6_ORIGDSTADDR:
1548 			case IPV6_BINDANY:
1549 			case IPV6_BINDMULTI:
1550 #ifdef	RSS
1551 			case IPV6_RSS_LISTEN_BUCKET:
1552 #endif
1553 				if (optname == IPV6_BINDANY && td != NULL) {
1554 					error = priv_check(td,
1555 					    PRIV_NETINET_BINDANY);
1556 					if (error)
1557 						break;
1558 				}
1559 
1560 				if (optlen != sizeof(int)) {
1561 					error = EINVAL;
1562 					break;
1563 				}
1564 				error = sooptcopyin(sopt, &optval,
1565 					sizeof optval, sizeof optval);
1566 				if (error)
1567 					break;
1568 				switch (optname) {
1569 
1570 				case IPV6_UNICAST_HOPS:
1571 					if (optval < -1 || optval >= 256)
1572 						error = EINVAL;
1573 					else {
1574 						/* -1 = kernel default */
1575 						in6p->in6p_hops = optval;
1576 						if ((in6p->inp_vflag &
1577 						     INP_IPV4) != 0)
1578 							in6p->inp_ip_ttl = optval;
1579 					}
1580 					break;
1581 #define OPTSET(bit) \
1582 do { \
1583 	INP_WLOCK(in6p); \
1584 	if (optval) \
1585 		in6p->inp_flags |= (bit); \
1586 	else \
1587 		in6p->inp_flags &= ~(bit); \
1588 	INP_WUNLOCK(in6p); \
1589 } while (/*CONSTCOND*/ 0)
1590 #define OPTSET2292(bit) \
1591 do { \
1592 	INP_WLOCK(in6p); \
1593 	in6p->inp_flags |= IN6P_RFC2292; \
1594 	if (optval) \
1595 		in6p->inp_flags |= (bit); \
1596 	else \
1597 		in6p->inp_flags &= ~(bit); \
1598 	INP_WUNLOCK(in6p); \
1599 } while (/*CONSTCOND*/ 0)
1600 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1601 
1602 #define OPTSET2(bit, val) do {						\
1603 	INP_WLOCK(in6p);						\
1604 	if (val)							\
1605 		in6p->inp_flags2 |= bit;				\
1606 	else								\
1607 		in6p->inp_flags2 &= ~bit;				\
1608 	INP_WUNLOCK(in6p);						\
1609 } while (0)
1610 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1611 
1612 				case IPV6_RECVPKTINFO:
1613 					/* cannot mix with RFC2292 */
1614 					if (OPTBIT(IN6P_RFC2292)) {
1615 						error = EINVAL;
1616 						break;
1617 					}
1618 					OPTSET(IN6P_PKTINFO);
1619 					break;
1620 
1621 				case IPV6_HOPLIMIT:
1622 				{
1623 					struct ip6_pktopts **optp;
1624 
1625 					/* cannot mix with RFC2292 */
1626 					if (OPTBIT(IN6P_RFC2292)) {
1627 						error = EINVAL;
1628 						break;
1629 					}
1630 					optp = &in6p->in6p_outputopts;
1631 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1632 					    (u_char *)&optval, sizeof(optval),
1633 					    optp, (td != NULL) ? td->td_ucred :
1634 					    NULL, uproto);
1635 					break;
1636 				}
1637 
1638 				case IPV6_RECVHOPLIMIT:
1639 					/* cannot mix with RFC2292 */
1640 					if (OPTBIT(IN6P_RFC2292)) {
1641 						error = EINVAL;
1642 						break;
1643 					}
1644 					OPTSET(IN6P_HOPLIMIT);
1645 					break;
1646 
1647 				case IPV6_RECVHOPOPTS:
1648 					/* cannot mix with RFC2292 */
1649 					if (OPTBIT(IN6P_RFC2292)) {
1650 						error = EINVAL;
1651 						break;
1652 					}
1653 					OPTSET(IN6P_HOPOPTS);
1654 					break;
1655 
1656 				case IPV6_RECVDSTOPTS:
1657 					/* cannot mix with RFC2292 */
1658 					if (OPTBIT(IN6P_RFC2292)) {
1659 						error = EINVAL;
1660 						break;
1661 					}
1662 					OPTSET(IN6P_DSTOPTS);
1663 					break;
1664 
1665 				case IPV6_RECVRTHDRDSTOPTS:
1666 					/* cannot mix with RFC2292 */
1667 					if (OPTBIT(IN6P_RFC2292)) {
1668 						error = EINVAL;
1669 						break;
1670 					}
1671 					OPTSET(IN6P_RTHDRDSTOPTS);
1672 					break;
1673 
1674 				case IPV6_RECVRTHDR:
1675 					/* cannot mix with RFC2292 */
1676 					if (OPTBIT(IN6P_RFC2292)) {
1677 						error = EINVAL;
1678 						break;
1679 					}
1680 					OPTSET(IN6P_RTHDR);
1681 					break;
1682 
1683 				case IPV6_RECVPATHMTU:
1684 					/*
1685 					 * We ignore this option for TCP
1686 					 * sockets.
1687 					 * (RFC3542 leaves this case
1688 					 * unspecified.)
1689 					 */
1690 					if (uproto != IPPROTO_TCP)
1691 						OPTSET(IN6P_MTU);
1692 					break;
1693 
1694 				case IPV6_RECVFLOWID:
1695 					OPTSET2(INP_RECVFLOWID, optval);
1696 					break;
1697 
1698 #ifdef	RSS
1699 				case IPV6_RECVRSSBUCKETID:
1700 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1701 					break;
1702 #endif
1703 
1704 				case IPV6_V6ONLY:
1705 					/*
1706 					 * make setsockopt(IPV6_V6ONLY)
1707 					 * available only prior to bind(2).
1708 					 * see ipng mailing list, Jun 22 2001.
1709 					 */
1710 					if (in6p->inp_lport ||
1711 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1712 						error = EINVAL;
1713 						break;
1714 					}
1715 					OPTSET(IN6P_IPV6_V6ONLY);
1716 					if (optval)
1717 						in6p->inp_vflag &= ~INP_IPV4;
1718 					else
1719 						in6p->inp_vflag |= INP_IPV4;
1720 					break;
1721 				case IPV6_RECVTCLASS:
1722 					/* cannot mix with RFC2292 XXX */
1723 					if (OPTBIT(IN6P_RFC2292)) {
1724 						error = EINVAL;
1725 						break;
1726 					}
1727 					OPTSET(IN6P_TCLASS);
1728 					break;
1729 				case IPV6_AUTOFLOWLABEL:
1730 					OPTSET(IN6P_AUTOFLOWLABEL);
1731 					break;
1732 
1733 				case IPV6_ORIGDSTADDR:
1734 					OPTSET2(INP_ORIGDSTADDR, optval);
1735 					break;
1736 				case IPV6_BINDANY:
1737 					OPTSET(INP_BINDANY);
1738 					break;
1739 
1740 				case IPV6_BINDMULTI:
1741 					OPTSET2(INP_BINDMULTI, optval);
1742 					break;
1743 #ifdef	RSS
1744 				case IPV6_RSS_LISTEN_BUCKET:
1745 					if ((optval >= 0) &&
1746 					    (optval < rss_getnumbuckets())) {
1747 						in6p->inp_rss_listen_bucket = optval;
1748 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1749 					} else {
1750 						error = EINVAL;
1751 					}
1752 					break;
1753 #endif
1754 				}
1755 				break;
1756 
1757 			case IPV6_TCLASS:
1758 			case IPV6_DONTFRAG:
1759 			case IPV6_USE_MIN_MTU:
1760 			case IPV6_PREFER_TEMPADDR:
1761 				if (optlen != sizeof(optval)) {
1762 					error = EINVAL;
1763 					break;
1764 				}
1765 				error = sooptcopyin(sopt, &optval,
1766 					sizeof optval, sizeof optval);
1767 				if (error)
1768 					break;
1769 				{
1770 					struct ip6_pktopts **optp;
1771 					optp = &in6p->in6p_outputopts;
1772 					error = ip6_pcbopt(optname,
1773 					    (u_char *)&optval, sizeof(optval),
1774 					    optp, (td != NULL) ? td->td_ucred :
1775 					    NULL, uproto);
1776 					break;
1777 				}
1778 
1779 			case IPV6_2292PKTINFO:
1780 			case IPV6_2292HOPLIMIT:
1781 			case IPV6_2292HOPOPTS:
1782 			case IPV6_2292DSTOPTS:
1783 			case IPV6_2292RTHDR:
1784 				/* RFC 2292 */
1785 				if (optlen != sizeof(int)) {
1786 					error = EINVAL;
1787 					break;
1788 				}
1789 				error = sooptcopyin(sopt, &optval,
1790 					sizeof optval, sizeof optval);
1791 				if (error)
1792 					break;
1793 				switch (optname) {
1794 				case IPV6_2292PKTINFO:
1795 					OPTSET2292(IN6P_PKTINFO);
1796 					break;
1797 				case IPV6_2292HOPLIMIT:
1798 					OPTSET2292(IN6P_HOPLIMIT);
1799 					break;
1800 				case IPV6_2292HOPOPTS:
1801 					/*
1802 					 * Check super-user privilege.
1803 					 * See comments for IPV6_RECVHOPOPTS.
1804 					 */
1805 					if (td != NULL) {
1806 						error = priv_check(td,
1807 						    PRIV_NETINET_SETHDROPTS);
1808 						if (error)
1809 							return (error);
1810 					}
1811 					OPTSET2292(IN6P_HOPOPTS);
1812 					break;
1813 				case IPV6_2292DSTOPTS:
1814 					if (td != NULL) {
1815 						error = priv_check(td,
1816 						    PRIV_NETINET_SETHDROPTS);
1817 						if (error)
1818 							return (error);
1819 					}
1820 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1821 					break;
1822 				case IPV6_2292RTHDR:
1823 					OPTSET2292(IN6P_RTHDR);
1824 					break;
1825 				}
1826 				break;
1827 			case IPV6_PKTINFO:
1828 			case IPV6_HOPOPTS:
1829 			case IPV6_RTHDR:
1830 			case IPV6_DSTOPTS:
1831 			case IPV6_RTHDRDSTOPTS:
1832 			case IPV6_NEXTHOP:
1833 			{
1834 				/* new advanced API (RFC3542) */
1835 				u_char *optbuf;
1836 				u_char optbuf_storage[MCLBYTES];
1837 				int optlen;
1838 				struct ip6_pktopts **optp;
1839 
1840 				/* cannot mix with RFC2292 */
1841 				if (OPTBIT(IN6P_RFC2292)) {
1842 					error = EINVAL;
1843 					break;
1844 				}
1845 
1846 				/*
1847 				 * We only ensure valsize is not too large
1848 				 * here.  Further validation will be done
1849 				 * later.
1850 				 */
1851 				error = sooptcopyin(sopt, optbuf_storage,
1852 				    sizeof(optbuf_storage), 0);
1853 				if (error)
1854 					break;
1855 				optlen = sopt->sopt_valsize;
1856 				optbuf = optbuf_storage;
1857 				optp = &in6p->in6p_outputopts;
1858 				error = ip6_pcbopt(optname, optbuf, optlen,
1859 				    optp, (td != NULL) ? td->td_ucred : NULL,
1860 				    uproto);
1861 				break;
1862 			}
1863 #undef OPTSET
1864 
1865 			case IPV6_MULTICAST_IF:
1866 			case IPV6_MULTICAST_HOPS:
1867 			case IPV6_MULTICAST_LOOP:
1868 			case IPV6_JOIN_GROUP:
1869 			case IPV6_LEAVE_GROUP:
1870 			case IPV6_MSFILTER:
1871 			case MCAST_BLOCK_SOURCE:
1872 			case MCAST_UNBLOCK_SOURCE:
1873 			case MCAST_JOIN_GROUP:
1874 			case MCAST_LEAVE_GROUP:
1875 			case MCAST_JOIN_SOURCE_GROUP:
1876 			case MCAST_LEAVE_SOURCE_GROUP:
1877 				error = ip6_setmoptions(in6p, sopt);
1878 				break;
1879 
1880 			case IPV6_PORTRANGE:
1881 				error = sooptcopyin(sopt, &optval,
1882 				    sizeof optval, sizeof optval);
1883 				if (error)
1884 					break;
1885 
1886 				INP_WLOCK(in6p);
1887 				switch (optval) {
1888 				case IPV6_PORTRANGE_DEFAULT:
1889 					in6p->inp_flags &= ~(INP_LOWPORT);
1890 					in6p->inp_flags &= ~(INP_HIGHPORT);
1891 					break;
1892 
1893 				case IPV6_PORTRANGE_HIGH:
1894 					in6p->inp_flags &= ~(INP_LOWPORT);
1895 					in6p->inp_flags |= INP_HIGHPORT;
1896 					break;
1897 
1898 				case IPV6_PORTRANGE_LOW:
1899 					in6p->inp_flags &= ~(INP_HIGHPORT);
1900 					in6p->inp_flags |= INP_LOWPORT;
1901 					break;
1902 
1903 				default:
1904 					error = EINVAL;
1905 					break;
1906 				}
1907 				INP_WUNLOCK(in6p);
1908 				break;
1909 
1910 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1911 			case IPV6_IPSEC_POLICY:
1912 				if (IPSEC_ENABLED(ipv6)) {
1913 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1914 					break;
1915 				}
1916 				/* FALLTHROUGH */
1917 #endif /* IPSEC */
1918 
1919 			default:
1920 				error = ENOPROTOOPT;
1921 				break;
1922 			}
1923 			break;
1924 
1925 		case SOPT_GET:
1926 			switch (optname) {
1927 
1928 			case IPV6_2292PKTOPTIONS:
1929 #ifdef IPV6_PKTOPTIONS
1930 			case IPV6_PKTOPTIONS:
1931 #endif
1932 				/*
1933 				 * RFC3542 (effectively) deprecated the
1934 				 * semantics of the 2292-style pktoptions.
1935 				 * Since it was not reliable in nature (i.e.,
1936 				 * applications had to expect the lack of some
1937 				 * information after all), it would make sense
1938 				 * to simplify this part by always returning
1939 				 * empty data.
1940 				 */
1941 				sopt->sopt_valsize = 0;
1942 				break;
1943 
1944 			case IPV6_RECVHOPOPTS:
1945 			case IPV6_RECVDSTOPTS:
1946 			case IPV6_RECVRTHDRDSTOPTS:
1947 			case IPV6_UNICAST_HOPS:
1948 			case IPV6_RECVPKTINFO:
1949 			case IPV6_RECVHOPLIMIT:
1950 			case IPV6_RECVRTHDR:
1951 			case IPV6_RECVPATHMTU:
1952 
1953 			case IPV6_V6ONLY:
1954 			case IPV6_PORTRANGE:
1955 			case IPV6_RECVTCLASS:
1956 			case IPV6_AUTOFLOWLABEL:
1957 			case IPV6_BINDANY:
1958 			case IPV6_FLOWID:
1959 			case IPV6_FLOWTYPE:
1960 			case IPV6_RECVFLOWID:
1961 #ifdef	RSS
1962 			case IPV6_RSSBUCKETID:
1963 			case IPV6_RECVRSSBUCKETID:
1964 #endif
1965 			case IPV6_BINDMULTI:
1966 				switch (optname) {
1967 
1968 				case IPV6_RECVHOPOPTS:
1969 					optval = OPTBIT(IN6P_HOPOPTS);
1970 					break;
1971 
1972 				case IPV6_RECVDSTOPTS:
1973 					optval = OPTBIT(IN6P_DSTOPTS);
1974 					break;
1975 
1976 				case IPV6_RECVRTHDRDSTOPTS:
1977 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1978 					break;
1979 
1980 				case IPV6_UNICAST_HOPS:
1981 					optval = in6p->in6p_hops;
1982 					break;
1983 
1984 				case IPV6_RECVPKTINFO:
1985 					optval = OPTBIT(IN6P_PKTINFO);
1986 					break;
1987 
1988 				case IPV6_RECVHOPLIMIT:
1989 					optval = OPTBIT(IN6P_HOPLIMIT);
1990 					break;
1991 
1992 				case IPV6_RECVRTHDR:
1993 					optval = OPTBIT(IN6P_RTHDR);
1994 					break;
1995 
1996 				case IPV6_RECVPATHMTU:
1997 					optval = OPTBIT(IN6P_MTU);
1998 					break;
1999 
2000 				case IPV6_V6ONLY:
2001 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2002 					break;
2003 
2004 				case IPV6_PORTRANGE:
2005 				    {
2006 					int flags;
2007 					flags = in6p->inp_flags;
2008 					if (flags & INP_HIGHPORT)
2009 						optval = IPV6_PORTRANGE_HIGH;
2010 					else if (flags & INP_LOWPORT)
2011 						optval = IPV6_PORTRANGE_LOW;
2012 					else
2013 						optval = 0;
2014 					break;
2015 				    }
2016 				case IPV6_RECVTCLASS:
2017 					optval = OPTBIT(IN6P_TCLASS);
2018 					break;
2019 
2020 				case IPV6_AUTOFLOWLABEL:
2021 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2022 					break;
2023 
2024 				case IPV6_ORIGDSTADDR:
2025 					optval = OPTBIT2(INP_ORIGDSTADDR);
2026 					break;
2027 
2028 				case IPV6_BINDANY:
2029 					optval = OPTBIT(INP_BINDANY);
2030 					break;
2031 
2032 				case IPV6_FLOWID:
2033 					optval = in6p->inp_flowid;
2034 					break;
2035 
2036 				case IPV6_FLOWTYPE:
2037 					optval = in6p->inp_flowtype;
2038 					break;
2039 
2040 				case IPV6_RECVFLOWID:
2041 					optval = OPTBIT2(INP_RECVFLOWID);
2042 					break;
2043 #ifdef	RSS
2044 				case IPV6_RSSBUCKETID:
2045 					retval =
2046 					    rss_hash2bucket(in6p->inp_flowid,
2047 					    in6p->inp_flowtype,
2048 					    &rss_bucket);
2049 					if (retval == 0)
2050 						optval = rss_bucket;
2051 					else
2052 						error = EINVAL;
2053 					break;
2054 
2055 				case IPV6_RECVRSSBUCKETID:
2056 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2057 					break;
2058 #endif
2059 
2060 				case IPV6_BINDMULTI:
2061 					optval = OPTBIT2(INP_BINDMULTI);
2062 					break;
2063 
2064 				}
2065 				if (error)
2066 					break;
2067 				error = sooptcopyout(sopt, &optval,
2068 					sizeof optval);
2069 				break;
2070 
2071 			case IPV6_PATHMTU:
2072 			{
2073 				u_long pmtu = 0;
2074 				struct ip6_mtuinfo mtuinfo;
2075 
2076 				if (!(so->so_state & SS_ISCONNECTED))
2077 					return (ENOTCONN);
2078 				/*
2079 				 * XXX: we dot not consider the case of source
2080 				 * routing, or optional information to specify
2081 				 * the outgoing interface.
2082 				 */
2083 				error = ip6_getpmtu_ctl(so->so_fibnum,
2084 				    &in6p->in6p_faddr, &pmtu);
2085 				if (error)
2086 					break;
2087 				if (pmtu > IPV6_MAXPACKET)
2088 					pmtu = IPV6_MAXPACKET;
2089 
2090 				bzero(&mtuinfo, sizeof(mtuinfo));
2091 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2092 				optdata = (void *)&mtuinfo;
2093 				optdatalen = sizeof(mtuinfo);
2094 				error = sooptcopyout(sopt, optdata,
2095 				    optdatalen);
2096 				break;
2097 			}
2098 
2099 			case IPV6_2292PKTINFO:
2100 			case IPV6_2292HOPLIMIT:
2101 			case IPV6_2292HOPOPTS:
2102 			case IPV6_2292RTHDR:
2103 			case IPV6_2292DSTOPTS:
2104 				switch (optname) {
2105 				case IPV6_2292PKTINFO:
2106 					optval = OPTBIT(IN6P_PKTINFO);
2107 					break;
2108 				case IPV6_2292HOPLIMIT:
2109 					optval = OPTBIT(IN6P_HOPLIMIT);
2110 					break;
2111 				case IPV6_2292HOPOPTS:
2112 					optval = OPTBIT(IN6P_HOPOPTS);
2113 					break;
2114 				case IPV6_2292RTHDR:
2115 					optval = OPTBIT(IN6P_RTHDR);
2116 					break;
2117 				case IPV6_2292DSTOPTS:
2118 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2119 					break;
2120 				}
2121 				error = sooptcopyout(sopt, &optval,
2122 				    sizeof optval);
2123 				break;
2124 			case IPV6_PKTINFO:
2125 			case IPV6_HOPOPTS:
2126 			case IPV6_RTHDR:
2127 			case IPV6_DSTOPTS:
2128 			case IPV6_RTHDRDSTOPTS:
2129 			case IPV6_NEXTHOP:
2130 			case IPV6_TCLASS:
2131 			case IPV6_DONTFRAG:
2132 			case IPV6_USE_MIN_MTU:
2133 			case IPV6_PREFER_TEMPADDR:
2134 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2135 				    optname, sopt);
2136 				break;
2137 
2138 			case IPV6_MULTICAST_IF:
2139 			case IPV6_MULTICAST_HOPS:
2140 			case IPV6_MULTICAST_LOOP:
2141 			case IPV6_MSFILTER:
2142 				error = ip6_getmoptions(in6p, sopt);
2143 				break;
2144 
2145 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2146 			case IPV6_IPSEC_POLICY:
2147 				if (IPSEC_ENABLED(ipv6)) {
2148 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2149 					break;
2150 				}
2151 				/* FALLTHROUGH */
2152 #endif /* IPSEC */
2153 			default:
2154 				error = ENOPROTOOPT;
2155 				break;
2156 			}
2157 			break;
2158 		}
2159 	}
2160 	return (error);
2161 }
2162 
2163 int
2164 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2165 {
2166 	int error = 0, optval, optlen;
2167 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2168 	struct inpcb *in6p = sotoinpcb(so);
2169 	int level, op, optname;
2170 
2171 	level = sopt->sopt_level;
2172 	op = sopt->sopt_dir;
2173 	optname = sopt->sopt_name;
2174 	optlen = sopt->sopt_valsize;
2175 
2176 	if (level != IPPROTO_IPV6) {
2177 		return (EINVAL);
2178 	}
2179 
2180 	switch (optname) {
2181 	case IPV6_CHECKSUM:
2182 		/*
2183 		 * For ICMPv6 sockets, no modification allowed for checksum
2184 		 * offset, permit "no change" values to help existing apps.
2185 		 *
2186 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2187 		 * for an ICMPv6 socket will fail."
2188 		 * The current behavior does not meet RFC3542.
2189 		 */
2190 		switch (op) {
2191 		case SOPT_SET:
2192 			if (optlen != sizeof(int)) {
2193 				error = EINVAL;
2194 				break;
2195 			}
2196 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2197 					    sizeof(optval));
2198 			if (error)
2199 				break;
2200 			if ((optval % 2) != 0) {
2201 				/* the API assumes even offset values */
2202 				error = EINVAL;
2203 			} else if (so->so_proto->pr_protocol ==
2204 			    IPPROTO_ICMPV6) {
2205 				if (optval != icmp6off)
2206 					error = EINVAL;
2207 			} else
2208 				in6p->in6p_cksum = optval;
2209 			break;
2210 
2211 		case SOPT_GET:
2212 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2213 				optval = icmp6off;
2214 			else
2215 				optval = in6p->in6p_cksum;
2216 
2217 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2218 			break;
2219 
2220 		default:
2221 			error = EINVAL;
2222 			break;
2223 		}
2224 		break;
2225 
2226 	default:
2227 		error = ENOPROTOOPT;
2228 		break;
2229 	}
2230 
2231 	return (error);
2232 }
2233 
2234 /*
2235  * Set up IP6 options in pcb for insertion in output packets or
2236  * specifying behavior of outgoing packets.
2237  */
2238 static int
2239 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2240     struct socket *so, struct sockopt *sopt)
2241 {
2242 	struct ip6_pktopts *opt = *pktopt;
2243 	int error = 0;
2244 	struct thread *td = sopt->sopt_td;
2245 
2246 	/* turn off any old options. */
2247 	if (opt) {
2248 #ifdef DIAGNOSTIC
2249 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2250 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2251 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2252 			printf("ip6_pcbopts: all specified options are cleared.\n");
2253 #endif
2254 		ip6_clearpktopts(opt, -1);
2255 	} else
2256 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2257 	*pktopt = NULL;
2258 
2259 	if (!m || m->m_len == 0) {
2260 		/*
2261 		 * Only turning off any previous options, regardless of
2262 		 * whether the opt is just created or given.
2263 		 */
2264 		free(opt, M_IP6OPT);
2265 		return (0);
2266 	}
2267 
2268 	/*  set options specified by user. */
2269 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2270 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2271 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2272 		free(opt, M_IP6OPT);
2273 		return (error);
2274 	}
2275 	*pktopt = opt;
2276 	return (0);
2277 }
2278 
2279 /*
2280  * initialize ip6_pktopts.  beware that there are non-zero default values in
2281  * the struct.
2282  */
2283 void
2284 ip6_initpktopts(struct ip6_pktopts *opt)
2285 {
2286 
2287 	bzero(opt, sizeof(*opt));
2288 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2289 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2290 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2291 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2292 }
2293 
2294 static int
2295 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2296     struct ucred *cred, int uproto)
2297 {
2298 	struct ip6_pktopts *opt;
2299 
2300 	if (*pktopt == NULL) {
2301 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2302 		    M_WAITOK);
2303 		ip6_initpktopts(*pktopt);
2304 	}
2305 	opt = *pktopt;
2306 
2307 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2308 }
2309 
2310 static int
2311 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2312 {
2313 	void *optdata = NULL;
2314 	int optdatalen = 0;
2315 	struct ip6_ext *ip6e;
2316 	int error = 0;
2317 	struct in6_pktinfo null_pktinfo;
2318 	int deftclass = 0, on;
2319 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2320 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2321 
2322 	switch (optname) {
2323 	case IPV6_PKTINFO:
2324 		optdata = (void *)&null_pktinfo;
2325 		if (pktopt && pktopt->ip6po_pktinfo) {
2326 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2327 			    sizeof(null_pktinfo));
2328 			in6_clearscope(&null_pktinfo.ipi6_addr);
2329 		} else {
2330 			/* XXX: we don't have to do this every time... */
2331 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2332 		}
2333 		optdatalen = sizeof(struct in6_pktinfo);
2334 		break;
2335 	case IPV6_TCLASS:
2336 		if (pktopt && pktopt->ip6po_tclass >= 0)
2337 			optdata = (void *)&pktopt->ip6po_tclass;
2338 		else
2339 			optdata = (void *)&deftclass;
2340 		optdatalen = sizeof(int);
2341 		break;
2342 	case IPV6_HOPOPTS:
2343 		if (pktopt && pktopt->ip6po_hbh) {
2344 			optdata = (void *)pktopt->ip6po_hbh;
2345 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2346 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2347 		}
2348 		break;
2349 	case IPV6_RTHDR:
2350 		if (pktopt && pktopt->ip6po_rthdr) {
2351 			optdata = (void *)pktopt->ip6po_rthdr;
2352 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2353 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2354 		}
2355 		break;
2356 	case IPV6_RTHDRDSTOPTS:
2357 		if (pktopt && pktopt->ip6po_dest1) {
2358 			optdata = (void *)pktopt->ip6po_dest1;
2359 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2360 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2361 		}
2362 		break;
2363 	case IPV6_DSTOPTS:
2364 		if (pktopt && pktopt->ip6po_dest2) {
2365 			optdata = (void *)pktopt->ip6po_dest2;
2366 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2367 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2368 		}
2369 		break;
2370 	case IPV6_NEXTHOP:
2371 		if (pktopt && pktopt->ip6po_nexthop) {
2372 			optdata = (void *)pktopt->ip6po_nexthop;
2373 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2374 		}
2375 		break;
2376 	case IPV6_USE_MIN_MTU:
2377 		if (pktopt)
2378 			optdata = (void *)&pktopt->ip6po_minmtu;
2379 		else
2380 			optdata = (void *)&defminmtu;
2381 		optdatalen = sizeof(int);
2382 		break;
2383 	case IPV6_DONTFRAG:
2384 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2385 			on = 1;
2386 		else
2387 			on = 0;
2388 		optdata = (void *)&on;
2389 		optdatalen = sizeof(on);
2390 		break;
2391 	case IPV6_PREFER_TEMPADDR:
2392 		if (pktopt)
2393 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2394 		else
2395 			optdata = (void *)&defpreftemp;
2396 		optdatalen = sizeof(int);
2397 		break;
2398 	default:		/* should not happen */
2399 #ifdef DIAGNOSTIC
2400 		panic("ip6_getpcbopt: unexpected option\n");
2401 #endif
2402 		return (ENOPROTOOPT);
2403 	}
2404 
2405 	error = sooptcopyout(sopt, optdata, optdatalen);
2406 
2407 	return (error);
2408 }
2409 
2410 void
2411 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2412 {
2413 	if (pktopt == NULL)
2414 		return;
2415 
2416 	if (optname == -1 || optname == IPV6_PKTINFO) {
2417 		if (pktopt->ip6po_pktinfo)
2418 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2419 		pktopt->ip6po_pktinfo = NULL;
2420 	}
2421 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2422 		pktopt->ip6po_hlim = -1;
2423 	if (optname == -1 || optname == IPV6_TCLASS)
2424 		pktopt->ip6po_tclass = -1;
2425 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2426 		if (pktopt->ip6po_nextroute.ro_rt) {
2427 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2428 			pktopt->ip6po_nextroute.ro_rt = NULL;
2429 		}
2430 		if (pktopt->ip6po_nexthop)
2431 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2432 		pktopt->ip6po_nexthop = NULL;
2433 	}
2434 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2435 		if (pktopt->ip6po_hbh)
2436 			free(pktopt->ip6po_hbh, M_IP6OPT);
2437 		pktopt->ip6po_hbh = NULL;
2438 	}
2439 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2440 		if (pktopt->ip6po_dest1)
2441 			free(pktopt->ip6po_dest1, M_IP6OPT);
2442 		pktopt->ip6po_dest1 = NULL;
2443 	}
2444 	if (optname == -1 || optname == IPV6_RTHDR) {
2445 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2446 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2447 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2448 		if (pktopt->ip6po_route.ro_rt) {
2449 			RTFREE(pktopt->ip6po_route.ro_rt);
2450 			pktopt->ip6po_route.ro_rt = NULL;
2451 		}
2452 	}
2453 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2454 		if (pktopt->ip6po_dest2)
2455 			free(pktopt->ip6po_dest2, M_IP6OPT);
2456 		pktopt->ip6po_dest2 = NULL;
2457 	}
2458 }
2459 
2460 #define PKTOPT_EXTHDRCPY(type) \
2461 do {\
2462 	if (src->type) {\
2463 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2464 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2465 		if (dst->type == NULL && canwait == M_NOWAIT)\
2466 			goto bad;\
2467 		bcopy(src->type, dst->type, hlen);\
2468 	}\
2469 } while (/*CONSTCOND*/ 0)
2470 
2471 static int
2472 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2473 {
2474 	if (dst == NULL || src == NULL)  {
2475 		printf("ip6_clearpktopts: invalid argument\n");
2476 		return (EINVAL);
2477 	}
2478 
2479 	dst->ip6po_hlim = src->ip6po_hlim;
2480 	dst->ip6po_tclass = src->ip6po_tclass;
2481 	dst->ip6po_flags = src->ip6po_flags;
2482 	dst->ip6po_minmtu = src->ip6po_minmtu;
2483 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2484 	if (src->ip6po_pktinfo) {
2485 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2486 		    M_IP6OPT, canwait);
2487 		if (dst->ip6po_pktinfo == NULL)
2488 			goto bad;
2489 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2490 	}
2491 	if (src->ip6po_nexthop) {
2492 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2493 		    M_IP6OPT, canwait);
2494 		if (dst->ip6po_nexthop == NULL)
2495 			goto bad;
2496 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2497 		    src->ip6po_nexthop->sa_len);
2498 	}
2499 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2500 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2501 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2502 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2503 	return (0);
2504 
2505   bad:
2506 	ip6_clearpktopts(dst, -1);
2507 	return (ENOBUFS);
2508 }
2509 #undef PKTOPT_EXTHDRCPY
2510 
2511 struct ip6_pktopts *
2512 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2513 {
2514 	int error;
2515 	struct ip6_pktopts *dst;
2516 
2517 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2518 	if (dst == NULL)
2519 		return (NULL);
2520 	ip6_initpktopts(dst);
2521 
2522 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2523 		free(dst, M_IP6OPT);
2524 		return (NULL);
2525 	}
2526 
2527 	return (dst);
2528 }
2529 
2530 void
2531 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2532 {
2533 	if (pktopt == NULL)
2534 		return;
2535 
2536 	ip6_clearpktopts(pktopt, -1);
2537 
2538 	free(pktopt, M_IP6OPT);
2539 }
2540 
2541 /*
2542  * Set IPv6 outgoing packet options based on advanced API.
2543  */
2544 int
2545 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2546     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2547 {
2548 	struct cmsghdr *cm = NULL;
2549 
2550 	if (control == NULL || opt == NULL)
2551 		return (EINVAL);
2552 
2553 	ip6_initpktopts(opt);
2554 	if (stickyopt) {
2555 		int error;
2556 
2557 		/*
2558 		 * If stickyopt is provided, make a local copy of the options
2559 		 * for this particular packet, then override them by ancillary
2560 		 * objects.
2561 		 * XXX: copypktopts() does not copy the cached route to a next
2562 		 * hop (if any).  This is not very good in terms of efficiency,
2563 		 * but we can allow this since this option should be rarely
2564 		 * used.
2565 		 */
2566 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2567 			return (error);
2568 	}
2569 
2570 	/*
2571 	 * XXX: Currently, we assume all the optional information is stored
2572 	 * in a single mbuf.
2573 	 */
2574 	if (control->m_next)
2575 		return (EINVAL);
2576 
2577 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2578 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2579 		int error;
2580 
2581 		if (control->m_len < CMSG_LEN(0))
2582 			return (EINVAL);
2583 
2584 		cm = mtod(control, struct cmsghdr *);
2585 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2586 			return (EINVAL);
2587 		if (cm->cmsg_level != IPPROTO_IPV6)
2588 			continue;
2589 
2590 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2591 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2592 		if (error)
2593 			return (error);
2594 	}
2595 
2596 	return (0);
2597 }
2598 
2599 /*
2600  * Set a particular packet option, as a sticky option or an ancillary data
2601  * item.  "len" can be 0 only when it's a sticky option.
2602  * We have 4 cases of combination of "sticky" and "cmsg":
2603  * "sticky=0, cmsg=0": impossible
2604  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2605  * "sticky=1, cmsg=0": RFC3542 socket option
2606  * "sticky=1, cmsg=1": RFC2292 socket option
2607  */
2608 static int
2609 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2610     struct ucred *cred, int sticky, int cmsg, int uproto)
2611 {
2612 	int minmtupolicy, preftemp;
2613 	int error;
2614 
2615 	if (!sticky && !cmsg) {
2616 #ifdef DIAGNOSTIC
2617 		printf("ip6_setpktopt: impossible case\n");
2618 #endif
2619 		return (EINVAL);
2620 	}
2621 
2622 	/*
2623 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2624 	 * not be specified in the context of RFC3542.  Conversely,
2625 	 * RFC3542 types should not be specified in the context of RFC2292.
2626 	 */
2627 	if (!cmsg) {
2628 		switch (optname) {
2629 		case IPV6_2292PKTINFO:
2630 		case IPV6_2292HOPLIMIT:
2631 		case IPV6_2292NEXTHOP:
2632 		case IPV6_2292HOPOPTS:
2633 		case IPV6_2292DSTOPTS:
2634 		case IPV6_2292RTHDR:
2635 		case IPV6_2292PKTOPTIONS:
2636 			return (ENOPROTOOPT);
2637 		}
2638 	}
2639 	if (sticky && cmsg) {
2640 		switch (optname) {
2641 		case IPV6_PKTINFO:
2642 		case IPV6_HOPLIMIT:
2643 		case IPV6_NEXTHOP:
2644 		case IPV6_HOPOPTS:
2645 		case IPV6_DSTOPTS:
2646 		case IPV6_RTHDRDSTOPTS:
2647 		case IPV6_RTHDR:
2648 		case IPV6_USE_MIN_MTU:
2649 		case IPV6_DONTFRAG:
2650 		case IPV6_TCLASS:
2651 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2652 			return (ENOPROTOOPT);
2653 		}
2654 	}
2655 
2656 	switch (optname) {
2657 	case IPV6_2292PKTINFO:
2658 	case IPV6_PKTINFO:
2659 	{
2660 		struct ifnet *ifp = NULL;
2661 		struct in6_pktinfo *pktinfo;
2662 
2663 		if (len != sizeof(struct in6_pktinfo))
2664 			return (EINVAL);
2665 
2666 		pktinfo = (struct in6_pktinfo *)buf;
2667 
2668 		/*
2669 		 * An application can clear any sticky IPV6_PKTINFO option by
2670 		 * doing a "regular" setsockopt with ipi6_addr being
2671 		 * in6addr_any and ipi6_ifindex being zero.
2672 		 * [RFC 3542, Section 6]
2673 		 */
2674 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2675 		    pktinfo->ipi6_ifindex == 0 &&
2676 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2677 			ip6_clearpktopts(opt, optname);
2678 			break;
2679 		}
2680 
2681 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2682 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2683 			return (EINVAL);
2684 		}
2685 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2686 			return (EINVAL);
2687 		/* validate the interface index if specified. */
2688 		if (pktinfo->ipi6_ifindex > V_if_index)
2689 			 return (ENXIO);
2690 		if (pktinfo->ipi6_ifindex) {
2691 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2692 			if (ifp == NULL)
2693 				return (ENXIO);
2694 		}
2695 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2696 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2697 			return (ENETDOWN);
2698 
2699 		if (ifp != NULL &&
2700 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2701 			struct in6_ifaddr *ia;
2702 
2703 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2704 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2705 			if (ia == NULL)
2706 				return (EADDRNOTAVAIL);
2707 			ifa_free(&ia->ia_ifa);
2708 		}
2709 		/*
2710 		 * We store the address anyway, and let in6_selectsrc()
2711 		 * validate the specified address.  This is because ipi6_addr
2712 		 * may not have enough information about its scope zone, and
2713 		 * we may need additional information (such as outgoing
2714 		 * interface or the scope zone of a destination address) to
2715 		 * disambiguate the scope.
2716 		 * XXX: the delay of the validation may confuse the
2717 		 * application when it is used as a sticky option.
2718 		 */
2719 		if (opt->ip6po_pktinfo == NULL) {
2720 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2721 			    M_IP6OPT, M_NOWAIT);
2722 			if (opt->ip6po_pktinfo == NULL)
2723 				return (ENOBUFS);
2724 		}
2725 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2726 		break;
2727 	}
2728 
2729 	case IPV6_2292HOPLIMIT:
2730 	case IPV6_HOPLIMIT:
2731 	{
2732 		int *hlimp;
2733 
2734 		/*
2735 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2736 		 * to simplify the ordering among hoplimit options.
2737 		 */
2738 		if (optname == IPV6_HOPLIMIT && sticky)
2739 			return (ENOPROTOOPT);
2740 
2741 		if (len != sizeof(int))
2742 			return (EINVAL);
2743 		hlimp = (int *)buf;
2744 		if (*hlimp < -1 || *hlimp > 255)
2745 			return (EINVAL);
2746 
2747 		opt->ip6po_hlim = *hlimp;
2748 		break;
2749 	}
2750 
2751 	case IPV6_TCLASS:
2752 	{
2753 		int tclass;
2754 
2755 		if (len != sizeof(int))
2756 			return (EINVAL);
2757 		tclass = *(int *)buf;
2758 		if (tclass < -1 || tclass > 255)
2759 			return (EINVAL);
2760 
2761 		opt->ip6po_tclass = tclass;
2762 		break;
2763 	}
2764 
2765 	case IPV6_2292NEXTHOP:
2766 	case IPV6_NEXTHOP:
2767 		if (cred != NULL) {
2768 			error = priv_check_cred(cred,
2769 			    PRIV_NETINET_SETHDROPTS, 0);
2770 			if (error)
2771 				return (error);
2772 		}
2773 
2774 		if (len == 0) {	/* just remove the option */
2775 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2776 			break;
2777 		}
2778 
2779 		/* check if cmsg_len is large enough for sa_len */
2780 		if (len < sizeof(struct sockaddr) || len < *buf)
2781 			return (EINVAL);
2782 
2783 		switch (((struct sockaddr *)buf)->sa_family) {
2784 		case AF_INET6:
2785 		{
2786 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2787 			int error;
2788 
2789 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2790 				return (EINVAL);
2791 
2792 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2793 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2794 				return (EINVAL);
2795 			}
2796 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2797 			    != 0) {
2798 				return (error);
2799 			}
2800 			break;
2801 		}
2802 		case AF_LINK:	/* should eventually be supported */
2803 		default:
2804 			return (EAFNOSUPPORT);
2805 		}
2806 
2807 		/* turn off the previous option, then set the new option. */
2808 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2809 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2810 		if (opt->ip6po_nexthop == NULL)
2811 			return (ENOBUFS);
2812 		bcopy(buf, opt->ip6po_nexthop, *buf);
2813 		break;
2814 
2815 	case IPV6_2292HOPOPTS:
2816 	case IPV6_HOPOPTS:
2817 	{
2818 		struct ip6_hbh *hbh;
2819 		int hbhlen;
2820 
2821 		/*
2822 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2823 		 * options, since per-option restriction has too much
2824 		 * overhead.
2825 		 */
2826 		if (cred != NULL) {
2827 			error = priv_check_cred(cred,
2828 			    PRIV_NETINET_SETHDROPTS, 0);
2829 			if (error)
2830 				return (error);
2831 		}
2832 
2833 		if (len == 0) {
2834 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2835 			break;	/* just remove the option */
2836 		}
2837 
2838 		/* message length validation */
2839 		if (len < sizeof(struct ip6_hbh))
2840 			return (EINVAL);
2841 		hbh = (struct ip6_hbh *)buf;
2842 		hbhlen = (hbh->ip6h_len + 1) << 3;
2843 		if (len != hbhlen)
2844 			return (EINVAL);
2845 
2846 		/* turn off the previous option, then set the new option. */
2847 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2848 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2849 		if (opt->ip6po_hbh == NULL)
2850 			return (ENOBUFS);
2851 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2852 
2853 		break;
2854 	}
2855 
2856 	case IPV6_2292DSTOPTS:
2857 	case IPV6_DSTOPTS:
2858 	case IPV6_RTHDRDSTOPTS:
2859 	{
2860 		struct ip6_dest *dest, **newdest = NULL;
2861 		int destlen;
2862 
2863 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2864 			error = priv_check_cred(cred,
2865 			    PRIV_NETINET_SETHDROPTS, 0);
2866 			if (error)
2867 				return (error);
2868 		}
2869 
2870 		if (len == 0) {
2871 			ip6_clearpktopts(opt, optname);
2872 			break;	/* just remove the option */
2873 		}
2874 
2875 		/* message length validation */
2876 		if (len < sizeof(struct ip6_dest))
2877 			return (EINVAL);
2878 		dest = (struct ip6_dest *)buf;
2879 		destlen = (dest->ip6d_len + 1) << 3;
2880 		if (len != destlen)
2881 			return (EINVAL);
2882 
2883 		/*
2884 		 * Determine the position that the destination options header
2885 		 * should be inserted; before or after the routing header.
2886 		 */
2887 		switch (optname) {
2888 		case IPV6_2292DSTOPTS:
2889 			/*
2890 			 * The old advacned API is ambiguous on this point.
2891 			 * Our approach is to determine the position based
2892 			 * according to the existence of a routing header.
2893 			 * Note, however, that this depends on the order of the
2894 			 * extension headers in the ancillary data; the 1st
2895 			 * part of the destination options header must appear
2896 			 * before the routing header in the ancillary data,
2897 			 * too.
2898 			 * RFC3542 solved the ambiguity by introducing
2899 			 * separate ancillary data or option types.
2900 			 */
2901 			if (opt->ip6po_rthdr == NULL)
2902 				newdest = &opt->ip6po_dest1;
2903 			else
2904 				newdest = &opt->ip6po_dest2;
2905 			break;
2906 		case IPV6_RTHDRDSTOPTS:
2907 			newdest = &opt->ip6po_dest1;
2908 			break;
2909 		case IPV6_DSTOPTS:
2910 			newdest = &opt->ip6po_dest2;
2911 			break;
2912 		}
2913 
2914 		/* turn off the previous option, then set the new option. */
2915 		ip6_clearpktopts(opt, optname);
2916 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2917 		if (*newdest == NULL)
2918 			return (ENOBUFS);
2919 		bcopy(dest, *newdest, destlen);
2920 
2921 		break;
2922 	}
2923 
2924 	case IPV6_2292RTHDR:
2925 	case IPV6_RTHDR:
2926 	{
2927 		struct ip6_rthdr *rth;
2928 		int rthlen;
2929 
2930 		if (len == 0) {
2931 			ip6_clearpktopts(opt, IPV6_RTHDR);
2932 			break;	/* just remove the option */
2933 		}
2934 
2935 		/* message length validation */
2936 		if (len < sizeof(struct ip6_rthdr))
2937 			return (EINVAL);
2938 		rth = (struct ip6_rthdr *)buf;
2939 		rthlen = (rth->ip6r_len + 1) << 3;
2940 		if (len != rthlen)
2941 			return (EINVAL);
2942 
2943 		switch (rth->ip6r_type) {
2944 		case IPV6_RTHDR_TYPE_0:
2945 			if (rth->ip6r_len == 0)	/* must contain one addr */
2946 				return (EINVAL);
2947 			if (rth->ip6r_len % 2) /* length must be even */
2948 				return (EINVAL);
2949 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2950 				return (EINVAL);
2951 			break;
2952 		default:
2953 			return (EINVAL);	/* not supported */
2954 		}
2955 
2956 		/* turn off the previous option */
2957 		ip6_clearpktopts(opt, IPV6_RTHDR);
2958 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2959 		if (opt->ip6po_rthdr == NULL)
2960 			return (ENOBUFS);
2961 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2962 
2963 		break;
2964 	}
2965 
2966 	case IPV6_USE_MIN_MTU:
2967 		if (len != sizeof(int))
2968 			return (EINVAL);
2969 		minmtupolicy = *(int *)buf;
2970 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2971 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2972 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2973 			return (EINVAL);
2974 		}
2975 		opt->ip6po_minmtu = minmtupolicy;
2976 		break;
2977 
2978 	case IPV6_DONTFRAG:
2979 		if (len != sizeof(int))
2980 			return (EINVAL);
2981 
2982 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2983 			/*
2984 			 * we ignore this option for TCP sockets.
2985 			 * (RFC3542 leaves this case unspecified.)
2986 			 */
2987 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2988 		} else
2989 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2990 		break;
2991 
2992 	case IPV6_PREFER_TEMPADDR:
2993 		if (len != sizeof(int))
2994 			return (EINVAL);
2995 		preftemp = *(int *)buf;
2996 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2997 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2998 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2999 			return (EINVAL);
3000 		}
3001 		opt->ip6po_prefer_tempaddr = preftemp;
3002 		break;
3003 
3004 	default:
3005 		return (ENOPROTOOPT);
3006 	} /* end of switch */
3007 
3008 	return (0);
3009 }
3010 
3011 /*
3012  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3013  * packet to the input queue of a specified interface.  Note that this
3014  * calls the output routine of the loopback "driver", but with an interface
3015  * pointer that might NOT be &loif -- easier than replicating that code here.
3016  */
3017 void
3018 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3019 {
3020 	struct mbuf *copym;
3021 	struct ip6_hdr *ip6;
3022 
3023 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3024 	if (copym == NULL)
3025 		return;
3026 
3027 	/*
3028 	 * Make sure to deep-copy IPv6 header portion in case the data
3029 	 * is in an mbuf cluster, so that we can safely override the IPv6
3030 	 * header portion later.
3031 	 */
3032 	if (!M_WRITABLE(copym) ||
3033 	    copym->m_len < sizeof(struct ip6_hdr)) {
3034 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3035 		if (copym == NULL)
3036 			return;
3037 	}
3038 	ip6 = mtod(copym, struct ip6_hdr *);
3039 	/*
3040 	 * clear embedded scope identifiers if necessary.
3041 	 * in6_clearscope will touch the addresses only when necessary.
3042 	 */
3043 	in6_clearscope(&ip6->ip6_src);
3044 	in6_clearscope(&ip6->ip6_dst);
3045 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3046 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3047 		    CSUM_PSEUDO_HDR;
3048 		copym->m_pkthdr.csum_data = 0xffff;
3049 	}
3050 	if_simloop(ifp, copym, AF_INET6, 0);
3051 }
3052 
3053 /*
3054  * Chop IPv6 header off from the payload.
3055  */
3056 static int
3057 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3058 {
3059 	struct mbuf *mh;
3060 	struct ip6_hdr *ip6;
3061 
3062 	ip6 = mtod(m, struct ip6_hdr *);
3063 	if (m->m_len > sizeof(*ip6)) {
3064 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3065 		if (mh == NULL) {
3066 			m_freem(m);
3067 			return ENOBUFS;
3068 		}
3069 		m_move_pkthdr(mh, m);
3070 		M_ALIGN(mh, sizeof(*ip6));
3071 		m->m_len -= sizeof(*ip6);
3072 		m->m_data += sizeof(*ip6);
3073 		mh->m_next = m;
3074 		m = mh;
3075 		m->m_len = sizeof(*ip6);
3076 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3077 	}
3078 	exthdrs->ip6e_ip6 = m;
3079 	return 0;
3080 }
3081 
3082 /*
3083  * Compute IPv6 extension header length.
3084  */
3085 int
3086 ip6_optlen(struct inpcb *in6p)
3087 {
3088 	int len;
3089 
3090 	if (!in6p->in6p_outputopts)
3091 		return 0;
3092 
3093 	len = 0;
3094 #define elen(x) \
3095     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3096 
3097 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3098 	if (in6p->in6p_outputopts->ip6po_rthdr)
3099 		/* dest1 is valid with rthdr only */
3100 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3101 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3102 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3103 	return len;
3104 #undef elen
3105 }
3106