xref: /freebsd/sys/netinet6/ip6_output.c (revision 2ef9ff7dd34a78a7890ba4d6de64da34d9c10942)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ratelimit.h"
71 #include "opt_ipsec.h"
72 #include "opt_sctp.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/errno.h>
81 #include <sys/priv.h>
82 #include <sys/proc.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/ucred.h>
88 
89 #include <machine/in_cksum.h>
90 
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_llatbl.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/pfil.h>
97 #include <net/rss_config.h>
98 #include <net/vnet.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_var.h>
102 #include <netinet/ip_var.h>
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/tcp_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/in6_rss.h>
112 
113 #include <netipsec/ipsec_support.h>
114 #ifdef SCTP
115 #include <netinet/sctp.h>
116 #include <netinet/sctp_crc32.h>
117 #endif
118 
119 #include <netinet6/ip6protosw.h>
120 #include <netinet6/scope6_var.h>
121 
122 extern int in6_mcast_loop;
123 
124 struct ip6_exthdrs {
125 	struct mbuf *ip6e_ip6;
126 	struct mbuf *ip6e_hbh;
127 	struct mbuf *ip6e_dest1;
128 	struct mbuf *ip6e_rthdr;
129 	struct mbuf *ip6e_dest2;
130 };
131 
132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
133 
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 			   struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 	struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 	struct ucred *, int, int, int);
141 
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 	struct ip6_frag **);
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static int ip6_getpmtu(struct route_in6 *, int,
148 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
149 	u_int);
150 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
151 	u_long *, int *, u_int);
152 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
153 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
154 
155 
156 /*
157  * Make an extension header from option data.  hp is the source, and
158  * mp is the destination.
159  */
160 #define MAKE_EXTHDR(hp, mp)						\
161     do {								\
162 	if (hp) {							\
163 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
164 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
165 		    ((eh)->ip6e_len + 1) << 3);				\
166 		if (error)						\
167 			goto freehdrs;					\
168 	}								\
169     } while (/*CONSTCOND*/ 0)
170 
171 /*
172  * Form a chain of extension headers.
173  * m is the extension header mbuf
174  * mp is the previous mbuf in the chain
175  * p is the next header
176  * i is the type of option.
177  */
178 #define MAKE_CHAIN(m, mp, p, i)\
179     do {\
180 	if (m) {\
181 		if (!hdrsplit) \
182 			panic("assumption failed: hdr not split"); \
183 		*mtod((m), u_char *) = *(p);\
184 		*(p) = (i);\
185 		p = mtod((m), u_char *);\
186 		(m)->m_next = (mp)->m_next;\
187 		(mp)->m_next = (m);\
188 		(mp) = (m);\
189 	}\
190     } while (/*CONSTCOND*/ 0)
191 
192 void
193 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
194 {
195 	u_short csum;
196 
197 	csum = in_cksum_skip(m, offset + plen, offset);
198 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
199 		csum = 0xffff;
200 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
201 
202 	if (offset + sizeof(csum) > m->m_len)
203 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
204 	else
205 		*(u_short *)mtodo(m, offset) = csum;
206 }
207 
208 int
209 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
210     int fraglen , uint32_t id)
211 {
212 	struct mbuf *m, **mnext, *m_frgpart;
213 	struct ip6_hdr *ip6, *mhip6;
214 	struct ip6_frag *ip6f;
215 	int off;
216 	int error;
217 	int tlen = m0->m_pkthdr.len;
218 
219 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
220 
221 	m = m0;
222 	ip6 = mtod(m, struct ip6_hdr *);
223 	mnext = &m->m_nextpkt;
224 
225 	for (off = hlen; off < tlen; off += fraglen) {
226 		m = m_gethdr(M_NOWAIT, MT_DATA);
227 		if (!m) {
228 			IP6STAT_INC(ip6s_odropped);
229 			return (ENOBUFS);
230 		}
231 
232 		/*
233 		 * Make sure the complete packet header gets copied
234 		 * from the originating mbuf to the newly created
235 		 * mbuf. This also ensures that existing firewall
236 		 * classification(s), VLAN tags and so on get copied
237 		 * to the resulting fragmented packet(s):
238 		 */
239 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
240 			m_free(m);
241 			IP6STAT_INC(ip6s_odropped);
242 			return (ENOBUFS);
243 		}
244 
245 		*mnext = m;
246 		mnext = &m->m_nextpkt;
247 		m->m_data += max_linkhdr;
248 		mhip6 = mtod(m, struct ip6_hdr *);
249 		*mhip6 = *ip6;
250 		m->m_len = sizeof(*mhip6);
251 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
252 		if (error) {
253 			IP6STAT_INC(ip6s_odropped);
254 			return (error);
255 		}
256 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
257 		if (off + fraglen >= tlen)
258 			fraglen = tlen - off;
259 		else
260 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
261 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
262 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
263 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
264 			IP6STAT_INC(ip6s_odropped);
265 			return (ENOBUFS);
266 		}
267 		m_cat(m, m_frgpart);
268 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
269 		ip6f->ip6f_reserved = 0;
270 		ip6f->ip6f_ident = id;
271 		ip6f->ip6f_nxt = nextproto;
272 		IP6STAT_INC(ip6s_ofragments);
273 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
274 	}
275 
276 	return (0);
277 }
278 
279 /*
280  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
281  * header (with pri, len, nxt, hlim, src, dst).
282  * This function may modify ver and hlim only.
283  * The mbuf chain containing the packet will be freed.
284  * The mbuf opt, if present, will not be freed.
285  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
286  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
287  * then result of route lookup is stored in ro->ro_rt.
288  *
289  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
290  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
291  * which is rt_mtu.
292  *
293  * ifpp - XXX: just for statistics
294  */
295 /*
296  * XXX TODO: no flowid is assigned for outbound flows?
297  */
298 int
299 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
300     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
301     struct ifnet **ifpp, struct inpcb *inp)
302 {
303 	struct ip6_hdr *ip6;
304 	struct ifnet *ifp, *origifp;
305 	struct mbuf *m = m0;
306 	struct mbuf *mprev = NULL;
307 	int hlen, tlen, len;
308 	struct route_in6 ip6route;
309 	struct rtentry *rt = NULL;
310 	struct sockaddr_in6 *dst, src_sa, dst_sa;
311 	struct in6_addr odst;
312 	int error = 0;
313 	struct in6_ifaddr *ia = NULL;
314 	u_long mtu;
315 	int alwaysfrag, dontfrag;
316 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
317 	struct ip6_exthdrs exthdrs;
318 	struct in6_addr src0, dst0;
319 	u_int32_t zone;
320 	struct route_in6 *ro_pmtu = NULL;
321 	int hdrsplit = 0;
322 	int sw_csum, tso;
323 	int needfiblookup;
324 	uint32_t fibnum;
325 	struct m_tag *fwd_tag = NULL;
326 	uint32_t id;
327 
328 	if (inp != NULL) {
329 		INP_LOCK_ASSERT(inp);
330 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
331 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
332 			/* unconditionally set flowid */
333 			m->m_pkthdr.flowid = inp->inp_flowid;
334 			M_HASHTYPE_SET(m, inp->inp_flowtype);
335 		}
336 #ifdef NUMA
337 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
338 #endif
339 	}
340 
341 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
342 	/*
343 	 * IPSec checking which handles several cases.
344 	 * FAST IPSEC: We re-injected the packet.
345 	 * XXX: need scope argument.
346 	 */
347 	if (IPSEC_ENABLED(ipv6)) {
348 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
349 			if (error == EINPROGRESS)
350 				error = 0;
351 			goto done;
352 		}
353 	}
354 #endif /* IPSEC */
355 
356 	bzero(&exthdrs, sizeof(exthdrs));
357 	if (opt) {
358 		/* Hop-by-Hop options header */
359 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
360 		/* Destination options header(1st part) */
361 		if (opt->ip6po_rthdr) {
362 			/*
363 			 * Destination options header(1st part)
364 			 * This only makes sense with a routing header.
365 			 * See Section 9.2 of RFC 3542.
366 			 * Disabling this part just for MIP6 convenience is
367 			 * a bad idea.  We need to think carefully about a
368 			 * way to make the advanced API coexist with MIP6
369 			 * options, which might automatically be inserted in
370 			 * the kernel.
371 			 */
372 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
373 		}
374 		/* Routing header */
375 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
376 		/* Destination options header(2nd part) */
377 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
378 	}
379 
380 	/*
381 	 * Calculate the total length of the extension header chain.
382 	 * Keep the length of the unfragmentable part for fragmentation.
383 	 */
384 	optlen = 0;
385 	if (exthdrs.ip6e_hbh)
386 		optlen += exthdrs.ip6e_hbh->m_len;
387 	if (exthdrs.ip6e_dest1)
388 		optlen += exthdrs.ip6e_dest1->m_len;
389 	if (exthdrs.ip6e_rthdr)
390 		optlen += exthdrs.ip6e_rthdr->m_len;
391 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
392 
393 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
394 	if (exthdrs.ip6e_dest2)
395 		optlen += exthdrs.ip6e_dest2->m_len;
396 
397 	/*
398 	 * If there is at least one extension header,
399 	 * separate IP6 header from the payload.
400 	 */
401 	if (optlen && !hdrsplit) {
402 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
403 			m = NULL;
404 			goto freehdrs;
405 		}
406 		m = exthdrs.ip6e_ip6;
407 		hdrsplit++;
408 	}
409 
410 	ip6 = mtod(m, struct ip6_hdr *);
411 
412 	/* adjust mbuf packet header length */
413 	m->m_pkthdr.len += optlen;
414 	plen = m->m_pkthdr.len - sizeof(*ip6);
415 
416 	/* If this is a jumbo payload, insert a jumbo payload option. */
417 	if (plen > IPV6_MAXPACKET) {
418 		if (!hdrsplit) {
419 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
420 				m = NULL;
421 				goto freehdrs;
422 			}
423 			m = exthdrs.ip6e_ip6;
424 			hdrsplit++;
425 		}
426 		/* adjust pointer */
427 		ip6 = mtod(m, struct ip6_hdr *);
428 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
429 			goto freehdrs;
430 		ip6->ip6_plen = 0;
431 	} else
432 		ip6->ip6_plen = htons(plen);
433 
434 	/*
435 	 * Concatenate headers and fill in next header fields.
436 	 * Here we have, on "m"
437 	 *	IPv6 payload
438 	 * and we insert headers accordingly.  Finally, we should be getting:
439 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
440 	 *
441 	 * during the header composing process, "m" points to IPv6 header.
442 	 * "mprev" points to an extension header prior to esp.
443 	 */
444 	u_char *nexthdrp = &ip6->ip6_nxt;
445 	mprev = m;
446 
447 	/*
448 	 * we treat dest2 specially.  this makes IPsec processing
449 	 * much easier.  the goal here is to make mprev point the
450 	 * mbuf prior to dest2.
451 	 *
452 	 * result: IPv6 dest2 payload
453 	 * m and mprev will point to IPv6 header.
454 	 */
455 	if (exthdrs.ip6e_dest2) {
456 		if (!hdrsplit)
457 			panic("assumption failed: hdr not split");
458 		exthdrs.ip6e_dest2->m_next = m->m_next;
459 		m->m_next = exthdrs.ip6e_dest2;
460 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
461 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
462 	}
463 
464 	/*
465 	 * result: IPv6 hbh dest1 rthdr dest2 payload
466 	 * m will point to IPv6 header.  mprev will point to the
467 	 * extension header prior to dest2 (rthdr in the above case).
468 	 */
469 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
470 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
471 		   IPPROTO_DSTOPTS);
472 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
473 		   IPPROTO_ROUTING);
474 
475 	/*
476 	 * If there is a routing header, discard the packet.
477 	 */
478 	if (exthdrs.ip6e_rthdr) {
479 		 error = EINVAL;
480 		 goto bad;
481 	}
482 
483 	/* Source address validation */
484 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
485 	    (flags & IPV6_UNSPECSRC) == 0) {
486 		error = EOPNOTSUPP;
487 		IP6STAT_INC(ip6s_badscope);
488 		goto bad;
489 	}
490 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
491 		error = EOPNOTSUPP;
492 		IP6STAT_INC(ip6s_badscope);
493 		goto bad;
494 	}
495 
496 	IP6STAT_INC(ip6s_localout);
497 
498 	/*
499 	 * Route packet.
500 	 */
501 	if (ro == NULL) {
502 		ro = &ip6route;
503 		bzero((caddr_t)ro, sizeof(*ro));
504 	}
505 	ro_pmtu = ro;
506 	if (opt && opt->ip6po_rthdr)
507 		ro = &opt->ip6po_route;
508 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
509 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
510 again:
511 	/*
512 	 * if specified, try to fill in the traffic class field.
513 	 * do not override if a non-zero value is already set.
514 	 * we check the diffserv field and the ecn field separately.
515 	 */
516 	if (opt && opt->ip6po_tclass >= 0) {
517 		int mask = 0;
518 
519 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
520 			mask |= 0xfc;
521 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
522 			mask |= 0x03;
523 		if (mask != 0)
524 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
525 	}
526 
527 	/* fill in or override the hop limit field, if necessary. */
528 	if (opt && opt->ip6po_hlim != -1)
529 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
530 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
531 		if (im6o != NULL)
532 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
533 		else
534 			ip6->ip6_hlim = V_ip6_defmcasthlim;
535 	}
536 	/*
537 	 * Validate route against routing table additions;
538 	 * a better/more specific route might have been added.
539 	 * Make sure address family is set in route.
540 	 */
541 	if (inp) {
542 		ro->ro_dst.sin6_family = AF_INET6;
543 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
544 	}
545 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
546 	    ro->ro_dst.sin6_family == AF_INET6 &&
547 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
548 		rt = ro->ro_rt;
549 		ifp = ro->ro_rt->rt_ifp;
550 	} else {
551 		if (ro->ro_lle)
552 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
553 		ro->ro_lle = NULL;
554 		if (fwd_tag == NULL) {
555 			bzero(&dst_sa, sizeof(dst_sa));
556 			dst_sa.sin6_family = AF_INET6;
557 			dst_sa.sin6_len = sizeof(dst_sa);
558 			dst_sa.sin6_addr = ip6->ip6_dst;
559 		}
560 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
561 		    &rt, fibnum);
562 		if (error != 0) {
563 			if (ifp != NULL)
564 				in6_ifstat_inc(ifp, ifs6_out_discard);
565 			goto bad;
566 		}
567 	}
568 	if (rt == NULL) {
569 		/*
570 		 * If in6_selectroute() does not return a route entry,
571 		 * dst may not have been updated.
572 		 */
573 		*dst = dst_sa;	/* XXX */
574 	}
575 
576 	/*
577 	 * then rt (for unicast) and ifp must be non-NULL valid values.
578 	 */
579 	if ((flags & IPV6_FORWARDING) == 0) {
580 		/* XXX: the FORWARDING flag can be set for mrouting. */
581 		in6_ifstat_inc(ifp, ifs6_out_request);
582 	}
583 	if (rt != NULL) {
584 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
585 		counter_u64_add(rt->rt_pksent, 1);
586 	}
587 
588 	/* Setup data structures for scope ID checks. */
589 	src0 = ip6->ip6_src;
590 	bzero(&src_sa, sizeof(src_sa));
591 	src_sa.sin6_family = AF_INET6;
592 	src_sa.sin6_len = sizeof(src_sa);
593 	src_sa.sin6_addr = ip6->ip6_src;
594 
595 	dst0 = ip6->ip6_dst;
596 	/* re-initialize to be sure */
597 	bzero(&dst_sa, sizeof(dst_sa));
598 	dst_sa.sin6_family = AF_INET6;
599 	dst_sa.sin6_len = sizeof(dst_sa);
600 	dst_sa.sin6_addr = ip6->ip6_dst;
601 
602 	/* Check for valid scope ID. */
603 	if (in6_setscope(&src0, ifp, &zone) == 0 &&
604 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
605 	    in6_setscope(&dst0, ifp, &zone) == 0 &&
606 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
607 		/*
608 		 * The outgoing interface is in the zone of the source
609 		 * and destination addresses.
610 		 *
611 		 * Because the loopback interface cannot receive
612 		 * packets with a different scope ID than its own,
613 		 * there is a trick is to pretend the outgoing packet
614 		 * was received by the real network interface, by
615 		 * setting "origifp" different from "ifp". This is
616 		 * only allowed when "ifp" is a loopback network
617 		 * interface. Refer to code in nd6_output_ifp() for
618 		 * more details.
619 		 */
620 		origifp = ifp;
621 
622 		/*
623 		 * We should use ia_ifp to support the case of sending
624 		 * packets to an address of our own.
625 		 */
626 		if (ia != NULL && ia->ia_ifp)
627 			ifp = ia->ia_ifp;
628 
629 	} else if ((ifp->if_flags & IFF_LOOPBACK) == 0 ||
630 	    sa6_recoverscope(&src_sa) != 0 ||
631 	    sa6_recoverscope(&dst_sa) != 0 ||
632 	    dst_sa.sin6_scope_id == 0 ||
633 	    (src_sa.sin6_scope_id != 0 &&
634 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
635 	    (origifp = ifnet_byindex(dst_sa.sin6_scope_id)) == NULL) {
636 		/*
637 		 * If the destination network interface is not a
638 		 * loopback interface, or the destination network
639 		 * address has no scope ID, or the source address has
640 		 * a scope ID set which is different from the
641 		 * destination address one, or there is no network
642 		 * interface representing this scope ID, the address
643 		 * pair is considered invalid.
644 		 */
645 		IP6STAT_INC(ip6s_badscope);
646 		in6_ifstat_inc(ifp, ifs6_out_discard);
647 		if (error == 0)
648 			error = EHOSTUNREACH; /* XXX */
649 		goto bad;
650 	}
651 
652 	/* All scope ID checks are successful. */
653 
654 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
655 		if (opt && opt->ip6po_nextroute.ro_rt) {
656 			/*
657 			 * The nexthop is explicitly specified by the
658 			 * application.  We assume the next hop is an IPv6
659 			 * address.
660 			 */
661 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
662 		}
663 		else if ((rt->rt_flags & RTF_GATEWAY))
664 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
665 	}
666 
667 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
668 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
669 	} else {
670 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
671 		in6_ifstat_inc(ifp, ifs6_out_mcast);
672 		/*
673 		 * Confirm that the outgoing interface supports multicast.
674 		 */
675 		if (!(ifp->if_flags & IFF_MULTICAST)) {
676 			IP6STAT_INC(ip6s_noroute);
677 			in6_ifstat_inc(ifp, ifs6_out_discard);
678 			error = ENETUNREACH;
679 			goto bad;
680 		}
681 		if ((im6o == NULL && in6_mcast_loop) ||
682 		    (im6o && im6o->im6o_multicast_loop)) {
683 			/*
684 			 * Loop back multicast datagram if not expressly
685 			 * forbidden to do so, even if we have not joined
686 			 * the address; protocols will filter it later,
687 			 * thus deferring a hash lookup and lock acquisition
688 			 * at the expense of an m_copym().
689 			 */
690 			ip6_mloopback(ifp, m);
691 		} else {
692 			/*
693 			 * If we are acting as a multicast router, perform
694 			 * multicast forwarding as if the packet had just
695 			 * arrived on the interface to which we are about
696 			 * to send.  The multicast forwarding function
697 			 * recursively calls this function, using the
698 			 * IPV6_FORWARDING flag to prevent infinite recursion.
699 			 *
700 			 * Multicasts that are looped back by ip6_mloopback(),
701 			 * above, will be forwarded by the ip6_input() routine,
702 			 * if necessary.
703 			 */
704 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
705 				/*
706 				 * XXX: ip6_mforward expects that rcvif is NULL
707 				 * when it is called from the originating path.
708 				 * However, it may not always be the case.
709 				 */
710 				m->m_pkthdr.rcvif = NULL;
711 				if (ip6_mforward(ip6, ifp, m) != 0) {
712 					m_freem(m);
713 					goto done;
714 				}
715 			}
716 		}
717 		/*
718 		 * Multicasts with a hoplimit of zero may be looped back,
719 		 * above, but must not be transmitted on a network.
720 		 * Also, multicasts addressed to the loopback interface
721 		 * are not sent -- the above call to ip6_mloopback() will
722 		 * loop back a copy if this host actually belongs to the
723 		 * destination group on the loopback interface.
724 		 */
725 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
726 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
727 			m_freem(m);
728 			goto done;
729 		}
730 	}
731 
732 	/*
733 	 * Fill the outgoing inteface to tell the upper layer
734 	 * to increment per-interface statistics.
735 	 */
736 	if (ifpp)
737 		*ifpp = ifp;
738 
739 	/* Determine path MTU. */
740 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
741 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
742 		goto bad;
743 
744 	/*
745 	 * The caller of this function may specify to use the minimum MTU
746 	 * in some cases.
747 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
748 	 * setting.  The logic is a bit complicated; by default, unicast
749 	 * packets will follow path MTU while multicast packets will be sent at
750 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
751 	 * including unicast ones will be sent at the minimum MTU.  Multicast
752 	 * packets will always be sent at the minimum MTU unless
753 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
754 	 * See RFC 3542 for more details.
755 	 */
756 	if (mtu > IPV6_MMTU) {
757 		if ((flags & IPV6_MINMTU))
758 			mtu = IPV6_MMTU;
759 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
760 			mtu = IPV6_MMTU;
761 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
762 			 (opt == NULL ||
763 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
764 			mtu = IPV6_MMTU;
765 		}
766 	}
767 
768 	/*
769 	 * clear embedded scope identifiers if necessary.
770 	 * in6_clearscope will touch the addresses only when necessary.
771 	 */
772 	in6_clearscope(&ip6->ip6_src);
773 	in6_clearscope(&ip6->ip6_dst);
774 
775 	/*
776 	 * If the outgoing packet contains a hop-by-hop options header,
777 	 * it must be examined and processed even by the source node.
778 	 * (RFC 2460, section 4.)
779 	 */
780 	if (exthdrs.ip6e_hbh) {
781 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
782 		u_int32_t dummy; /* XXX unused */
783 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
784 
785 #ifdef DIAGNOSTIC
786 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
787 			panic("ip6e_hbh is not contiguous");
788 #endif
789 		/*
790 		 *  XXX: if we have to send an ICMPv6 error to the sender,
791 		 *       we need the M_LOOP flag since icmp6_error() expects
792 		 *       the IPv6 and the hop-by-hop options header are
793 		 *       contiguous unless the flag is set.
794 		 */
795 		m->m_flags |= M_LOOP;
796 		m->m_pkthdr.rcvif = ifp;
797 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
798 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
799 		    &dummy, &plen) < 0) {
800 			/* m was already freed at this point */
801 			error = EINVAL;/* better error? */
802 			goto done;
803 		}
804 		m->m_flags &= ~M_LOOP; /* XXX */
805 		m->m_pkthdr.rcvif = NULL;
806 	}
807 
808 	/* Jump over all PFIL processing if hooks are not active. */
809 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
810 		goto passout;
811 
812 	odst = ip6->ip6_dst;
813 	/* Run through list of hooks for output packets. */
814 	switch (pfil_run_hooks(V_inet6_pfil_head, &m, ifp, PFIL_OUT, inp)) {
815 	case PFIL_PASS:
816 		ip6 = mtod(m, struct ip6_hdr *);
817 		break;
818 	case PFIL_DROPPED:
819 		error = EPERM;
820 		/* FALLTHROUGH */
821 	case PFIL_CONSUMED:
822 		goto done;
823 	}
824 
825 	needfiblookup = 0;
826 	/* See if destination IP address was changed by packet filter. */
827 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
828 		m->m_flags |= M_SKIP_FIREWALL;
829 		/* If destination is now ourself drop to ip6_input(). */
830 		if (in6_localip(&ip6->ip6_dst)) {
831 			m->m_flags |= M_FASTFWD_OURS;
832 			if (m->m_pkthdr.rcvif == NULL)
833 				m->m_pkthdr.rcvif = V_loif;
834 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
835 				m->m_pkthdr.csum_flags |=
836 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
837 				m->m_pkthdr.csum_data = 0xffff;
838 			}
839 #ifdef SCTP
840 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
841 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
842 #endif
843 			error = netisr_queue(NETISR_IPV6, m);
844 			goto done;
845 		} else {
846 			RO_INVALIDATE_CACHE(ro);
847 			needfiblookup = 1; /* Redo the routing table lookup. */
848 		}
849 	}
850 	/* See if fib was changed by packet filter. */
851 	if (fibnum != M_GETFIB(m)) {
852 		m->m_flags |= M_SKIP_FIREWALL;
853 		fibnum = M_GETFIB(m);
854 		RO_INVALIDATE_CACHE(ro);
855 		needfiblookup = 1;
856 	}
857 	if (needfiblookup)
858 		goto again;
859 
860 	/* See if local, if yes, send it to netisr. */
861 	if (m->m_flags & M_FASTFWD_OURS) {
862 		if (m->m_pkthdr.rcvif == NULL)
863 			m->m_pkthdr.rcvif = V_loif;
864 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
865 			m->m_pkthdr.csum_flags |=
866 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
867 			m->m_pkthdr.csum_data = 0xffff;
868 		}
869 #ifdef SCTP
870 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
871 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
872 #endif
873 		error = netisr_queue(NETISR_IPV6, m);
874 		goto done;
875 	}
876 	/* Or forward to some other address? */
877 	if ((m->m_flags & M_IP6_NEXTHOP) &&
878 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
879 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
880 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
881 		m->m_flags |= M_SKIP_FIREWALL;
882 		m->m_flags &= ~M_IP6_NEXTHOP;
883 		m_tag_delete(m, fwd_tag);
884 		goto again;
885 	}
886 
887 passout:
888 	/*
889 	 * Send the packet to the outgoing interface.
890 	 * If necessary, do IPv6 fragmentation before sending.
891 	 *
892 	 * the logic here is rather complex:
893 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
894 	 * 1-a:	send as is if tlen <= path mtu
895 	 * 1-b:	fragment if tlen > path mtu
896 	 *
897 	 * 2: if user asks us not to fragment (dontfrag == 1)
898 	 * 2-a:	send as is if tlen <= interface mtu
899 	 * 2-b:	error if tlen > interface mtu
900 	 *
901 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
902 	 *	always fragment
903 	 *
904 	 * 4: if dontfrag == 1 && alwaysfrag == 1
905 	 *	error, as we cannot handle this conflicting request
906 	 */
907 	sw_csum = m->m_pkthdr.csum_flags;
908 	if (!hdrsplit) {
909 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
910 		sw_csum &= ~ifp->if_hwassist;
911 	} else
912 		tso = 0;
913 	/*
914 	 * If we added extension headers, we will not do TSO and calculate the
915 	 * checksums ourselves for now.
916 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
917 	 * with ext. hdrs.
918 	 */
919 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
920 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
921 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
922 	}
923 #ifdef SCTP
924 	if (sw_csum & CSUM_SCTP_IPV6) {
925 		sw_csum &= ~CSUM_SCTP_IPV6;
926 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
927 	}
928 #endif
929 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
930 	tlen = m->m_pkthdr.len;
931 
932 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
933 		dontfrag = 1;
934 	else
935 		dontfrag = 0;
936 	if (dontfrag && alwaysfrag) {	/* case 4 */
937 		/* conflicting request - can't transmit */
938 		error = EMSGSIZE;
939 		goto bad;
940 	}
941 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
942 		/*
943 		 * Even if the DONTFRAG option is specified, we cannot send the
944 		 * packet when the data length is larger than the MTU of the
945 		 * outgoing interface.
946 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
947 		 * application wanted to know the MTU value. Also return an
948 		 * error code (this is not described in the API spec).
949 		 */
950 		if (inp != NULL)
951 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
952 		error = EMSGSIZE;
953 		goto bad;
954 	}
955 
956 	/*
957 	 * transmit packet without fragmentation
958 	 */
959 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
960 		struct in6_ifaddr *ia6;
961 
962 		ip6 = mtod(m, struct ip6_hdr *);
963 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
964 		if (ia6) {
965 			/* Record statistics for this interface address. */
966 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
967 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
968 			    m->m_pkthdr.len);
969 			ifa_free(&ia6->ia_ifa);
970 		}
971 #ifdef RATELIMIT
972 		if (inp != NULL) {
973 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
974 				in_pcboutput_txrtlmt(inp, ifp, m);
975 			/* stamp send tag on mbuf */
976 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
977 		} else {
978 			m->m_pkthdr.snd_tag = NULL;
979 		}
980 #endif
981 		error = nd6_output_ifp(ifp, origifp, m, dst,
982 		    (struct route *)ro);
983 #ifdef RATELIMIT
984 		/* check for route change */
985 		if (error == EAGAIN)
986 			in_pcboutput_eagain(inp);
987 #endif
988 		goto done;
989 	}
990 
991 	/*
992 	 * try to fragment the packet.  case 1-b and 3
993 	 */
994 	if (mtu < IPV6_MMTU) {
995 		/* path MTU cannot be less than IPV6_MMTU */
996 		error = EMSGSIZE;
997 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
998 		goto bad;
999 	} else if (ip6->ip6_plen == 0) {
1000 		/* jumbo payload cannot be fragmented */
1001 		error = EMSGSIZE;
1002 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1003 		goto bad;
1004 	} else {
1005 		u_char nextproto;
1006 
1007 		/*
1008 		 * Too large for the destination or interface;
1009 		 * fragment if possible.
1010 		 * Must be able to put at least 8 bytes per fragment.
1011 		 */
1012 		hlen = unfragpartlen;
1013 		if (mtu > IPV6_MAXPACKET)
1014 			mtu = IPV6_MAXPACKET;
1015 
1016 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1017 		if (len < 8) {
1018 			error = EMSGSIZE;
1019 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1020 			goto bad;
1021 		}
1022 
1023 		/*
1024 		 * If the interface will not calculate checksums on
1025 		 * fragmented packets, then do it here.
1026 		 * XXX-BZ handle the hw offloading case.  Need flags.
1027 		 */
1028 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1029 			in6_delayed_cksum(m, plen, hlen);
1030 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1031 		}
1032 #ifdef SCTP
1033 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1034 			sctp_delayed_cksum(m, hlen);
1035 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1036 		}
1037 #endif
1038 		/*
1039 		 * Change the next header field of the last header in the
1040 		 * unfragmentable part.
1041 		 */
1042 		if (exthdrs.ip6e_rthdr) {
1043 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1044 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1045 		} else if (exthdrs.ip6e_dest1) {
1046 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1047 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1048 		} else if (exthdrs.ip6e_hbh) {
1049 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1050 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1051 		} else {
1052 			nextproto = ip6->ip6_nxt;
1053 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1054 		}
1055 
1056 		/*
1057 		 * Loop through length of segment after first fragment,
1058 		 * make new header and copy data of each part and link onto
1059 		 * chain.
1060 		 */
1061 		m0 = m;
1062 		id = htonl(ip6_randomid());
1063 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1064 			goto sendorfree;
1065 
1066 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1067 	}
1068 
1069 	/*
1070 	 * Remove leading garbages.
1071 	 */
1072 sendorfree:
1073 	m = m0->m_nextpkt;
1074 	m0->m_nextpkt = 0;
1075 	m_freem(m0);
1076 	for (; m; m = m0) {
1077 		m0 = m->m_nextpkt;
1078 		m->m_nextpkt = 0;
1079 		if (error == 0) {
1080 			/* Record statistics for this interface address. */
1081 			if (ia) {
1082 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1083 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1084 				    m->m_pkthdr.len);
1085 			}
1086 #ifdef RATELIMIT
1087 			if (inp != NULL) {
1088 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1089 					in_pcboutput_txrtlmt(inp, ifp, m);
1090 				/* stamp send tag on mbuf */
1091 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1092 			} else {
1093 				m->m_pkthdr.snd_tag = NULL;
1094 			}
1095 #endif
1096 			error = nd6_output_ifp(ifp, origifp, m, dst,
1097 			    (struct route *)ro);
1098 #ifdef RATELIMIT
1099 			/* check for route change */
1100 			if (error == EAGAIN)
1101 				in_pcboutput_eagain(inp);
1102 #endif
1103 		} else
1104 			m_freem(m);
1105 	}
1106 
1107 	if (error == 0)
1108 		IP6STAT_INC(ip6s_fragmented);
1109 
1110 done:
1111 	if (ro == &ip6route)
1112 		RO_RTFREE(ro);
1113 	return (error);
1114 
1115 freehdrs:
1116 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1117 	m_freem(exthdrs.ip6e_dest1);
1118 	m_freem(exthdrs.ip6e_rthdr);
1119 	m_freem(exthdrs.ip6e_dest2);
1120 	/* FALLTHROUGH */
1121 bad:
1122 	if (m)
1123 		m_freem(m);
1124 	goto done;
1125 }
1126 
1127 static int
1128 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1129 {
1130 	struct mbuf *m;
1131 
1132 	if (hlen > MCLBYTES)
1133 		return (ENOBUFS); /* XXX */
1134 
1135 	if (hlen > MLEN)
1136 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1137 	else
1138 		m = m_get(M_NOWAIT, MT_DATA);
1139 	if (m == NULL)
1140 		return (ENOBUFS);
1141 	m->m_len = hlen;
1142 	if (hdr)
1143 		bcopy(hdr, mtod(m, caddr_t), hlen);
1144 
1145 	*mp = m;
1146 	return (0);
1147 }
1148 
1149 /*
1150  * Insert jumbo payload option.
1151  */
1152 static int
1153 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1154 {
1155 	struct mbuf *mopt;
1156 	u_char *optbuf;
1157 	u_int32_t v;
1158 
1159 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1160 
1161 	/*
1162 	 * If there is no hop-by-hop options header, allocate new one.
1163 	 * If there is one but it doesn't have enough space to store the
1164 	 * jumbo payload option, allocate a cluster to store the whole options.
1165 	 * Otherwise, use it to store the options.
1166 	 */
1167 	if (exthdrs->ip6e_hbh == NULL) {
1168 		mopt = m_get(M_NOWAIT, MT_DATA);
1169 		if (mopt == NULL)
1170 			return (ENOBUFS);
1171 		mopt->m_len = JUMBOOPTLEN;
1172 		optbuf = mtod(mopt, u_char *);
1173 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1174 		exthdrs->ip6e_hbh = mopt;
1175 	} else {
1176 		struct ip6_hbh *hbh;
1177 
1178 		mopt = exthdrs->ip6e_hbh;
1179 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1180 			/*
1181 			 * XXX assumption:
1182 			 * - exthdrs->ip6e_hbh is not referenced from places
1183 			 *   other than exthdrs.
1184 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1185 			 */
1186 			int oldoptlen = mopt->m_len;
1187 			struct mbuf *n;
1188 
1189 			/*
1190 			 * XXX: give up if the whole (new) hbh header does
1191 			 * not fit even in an mbuf cluster.
1192 			 */
1193 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1194 				return (ENOBUFS);
1195 
1196 			/*
1197 			 * As a consequence, we must always prepare a cluster
1198 			 * at this point.
1199 			 */
1200 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1201 			if (n == NULL)
1202 				return (ENOBUFS);
1203 			n->m_len = oldoptlen + JUMBOOPTLEN;
1204 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1205 			    oldoptlen);
1206 			optbuf = mtod(n, caddr_t) + oldoptlen;
1207 			m_freem(mopt);
1208 			mopt = exthdrs->ip6e_hbh = n;
1209 		} else {
1210 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1211 			mopt->m_len += JUMBOOPTLEN;
1212 		}
1213 		optbuf[0] = IP6OPT_PADN;
1214 		optbuf[1] = 1;
1215 
1216 		/*
1217 		 * Adjust the header length according to the pad and
1218 		 * the jumbo payload option.
1219 		 */
1220 		hbh = mtod(mopt, struct ip6_hbh *);
1221 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1222 	}
1223 
1224 	/* fill in the option. */
1225 	optbuf[2] = IP6OPT_JUMBO;
1226 	optbuf[3] = 4;
1227 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1228 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1229 
1230 	/* finally, adjust the packet header length */
1231 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1232 
1233 	return (0);
1234 #undef JUMBOOPTLEN
1235 }
1236 
1237 /*
1238  * Insert fragment header and copy unfragmentable header portions.
1239  */
1240 static int
1241 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1242     struct ip6_frag **frghdrp)
1243 {
1244 	struct mbuf *n, *mlast;
1245 
1246 	if (hlen > sizeof(struct ip6_hdr)) {
1247 		n = m_copym(m0, sizeof(struct ip6_hdr),
1248 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1249 		if (n == NULL)
1250 			return (ENOBUFS);
1251 		m->m_next = n;
1252 	} else
1253 		n = m;
1254 
1255 	/* Search for the last mbuf of unfragmentable part. */
1256 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1257 		;
1258 
1259 	if (M_WRITABLE(mlast) &&
1260 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1261 		/* use the trailing space of the last mbuf for the fragment hdr */
1262 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1263 		    mlast->m_len);
1264 		mlast->m_len += sizeof(struct ip6_frag);
1265 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1266 	} else {
1267 		/* allocate a new mbuf for the fragment header */
1268 		struct mbuf *mfrg;
1269 
1270 		mfrg = m_get(M_NOWAIT, MT_DATA);
1271 		if (mfrg == NULL)
1272 			return (ENOBUFS);
1273 		mfrg->m_len = sizeof(struct ip6_frag);
1274 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1275 		mlast->m_next = mfrg;
1276 	}
1277 
1278 	return (0);
1279 }
1280 
1281 /*
1282  * Calculates IPv6 path mtu for destination @dst.
1283  * Resulting MTU is stored in @mtup.
1284  *
1285  * Returns 0 on success.
1286  */
1287 static int
1288 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1289 {
1290 	struct nhop6_extended nh6;
1291 	struct in6_addr kdst;
1292 	uint32_t scopeid;
1293 	struct ifnet *ifp;
1294 	u_long mtu;
1295 	int error;
1296 
1297 	in6_splitscope(dst, &kdst, &scopeid);
1298 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1299 		return (EHOSTUNREACH);
1300 
1301 	ifp = nh6.nh_ifp;
1302 	mtu = nh6.nh_mtu;
1303 
1304 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1305 	fib6_free_nh_ext(fibnum, &nh6);
1306 
1307 	return (error);
1308 }
1309 
1310 /*
1311  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1312  * and cached data in @ro_pmtu.
1313  * MTU from (successful) route lookup is saved (along with dst)
1314  * inside @ro_pmtu to avoid subsequent route lookups after packet
1315  * filter processing.
1316  *
1317  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1318  * Returns 0 on success.
1319  */
1320 static int
1321 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1322     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1323     int *alwaysfragp, u_int fibnum, u_int proto)
1324 {
1325 	struct nhop6_basic nh6;
1326 	struct in6_addr kdst;
1327 	uint32_t scopeid;
1328 	struct sockaddr_in6 *sa6_dst;
1329 	u_long mtu;
1330 
1331 	mtu = 0;
1332 	if (do_lookup) {
1333 
1334 		/*
1335 		 * Here ro_pmtu has final destination address, while
1336 		 * ro might represent immediate destination.
1337 		 * Use ro_pmtu destination since mtu might differ.
1338 		 */
1339 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1340 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1341 			ro_pmtu->ro_mtu = 0;
1342 
1343 		if (ro_pmtu->ro_mtu == 0) {
1344 			bzero(sa6_dst, sizeof(*sa6_dst));
1345 			sa6_dst->sin6_family = AF_INET6;
1346 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1347 			sa6_dst->sin6_addr = *dst;
1348 
1349 			in6_splitscope(dst, &kdst, &scopeid);
1350 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1351 			    &nh6) == 0)
1352 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1353 		}
1354 
1355 		mtu = ro_pmtu->ro_mtu;
1356 	}
1357 
1358 	if (ro_pmtu->ro_rt)
1359 		mtu = ro_pmtu->ro_rt->rt_mtu;
1360 
1361 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1362 }
1363 
1364 /*
1365  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1366  * hostcache data for @dst.
1367  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1368  *
1369  * Returns 0 on success.
1370  */
1371 static int
1372 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1373     u_long *mtup, int *alwaysfragp, u_int proto)
1374 {
1375 	u_long mtu = 0;
1376 	int alwaysfrag = 0;
1377 	int error = 0;
1378 
1379 	if (rt_mtu > 0) {
1380 		u_int32_t ifmtu;
1381 		struct in_conninfo inc;
1382 
1383 		bzero(&inc, sizeof(inc));
1384 		inc.inc_flags |= INC_ISIPV6;
1385 		inc.inc6_faddr = *dst;
1386 
1387 		ifmtu = IN6_LINKMTU(ifp);
1388 
1389 		/* TCP is known to react to pmtu changes so skip hc */
1390 		if (proto != IPPROTO_TCP)
1391 			mtu = tcp_hc_getmtu(&inc);
1392 
1393 		if (mtu)
1394 			mtu = min(mtu, rt_mtu);
1395 		else
1396 			mtu = rt_mtu;
1397 		if (mtu == 0)
1398 			mtu = ifmtu;
1399 		else if (mtu < IPV6_MMTU) {
1400 			/*
1401 			 * RFC2460 section 5, last paragraph:
1402 			 * if we record ICMPv6 too big message with
1403 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1404 			 * or smaller, with framgent header attached.
1405 			 * (fragment header is needed regardless from the
1406 			 * packet size, for translators to identify packets)
1407 			 */
1408 			alwaysfrag = 1;
1409 			mtu = IPV6_MMTU;
1410 		}
1411 	} else if (ifp) {
1412 		mtu = IN6_LINKMTU(ifp);
1413 	} else
1414 		error = EHOSTUNREACH; /* XXX */
1415 
1416 	*mtup = mtu;
1417 	if (alwaysfragp)
1418 		*alwaysfragp = alwaysfrag;
1419 	return (error);
1420 }
1421 
1422 /*
1423  * IP6 socket option processing.
1424  */
1425 int
1426 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1427 {
1428 	int optdatalen, uproto;
1429 	void *optdata;
1430 	struct inpcb *in6p = sotoinpcb(so);
1431 	int error, optval;
1432 	int level, op, optname;
1433 	int optlen;
1434 	struct thread *td;
1435 #ifdef	RSS
1436 	uint32_t rss_bucket;
1437 	int retval;
1438 #endif
1439 
1440 /*
1441  * Don't use more than a quarter of mbuf clusters.  N.B.:
1442  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1443  * on LP64 architectures, so cast to u_long to avoid undefined
1444  * behavior.  ILP32 architectures cannot have nmbclusters
1445  * large enough to overflow for other reasons.
1446  */
1447 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1448 
1449 	level = sopt->sopt_level;
1450 	op = sopt->sopt_dir;
1451 	optname = sopt->sopt_name;
1452 	optlen = sopt->sopt_valsize;
1453 	td = sopt->sopt_td;
1454 	error = 0;
1455 	optval = 0;
1456 	uproto = (int)so->so_proto->pr_protocol;
1457 
1458 	if (level != IPPROTO_IPV6) {
1459 		error = EINVAL;
1460 
1461 		if (sopt->sopt_level == SOL_SOCKET &&
1462 		    sopt->sopt_dir == SOPT_SET) {
1463 			switch (sopt->sopt_name) {
1464 			case SO_REUSEADDR:
1465 				INP_WLOCK(in6p);
1466 				if ((so->so_options & SO_REUSEADDR) != 0)
1467 					in6p->inp_flags2 |= INP_REUSEADDR;
1468 				else
1469 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1470 				INP_WUNLOCK(in6p);
1471 				error = 0;
1472 				break;
1473 			case SO_REUSEPORT:
1474 				INP_WLOCK(in6p);
1475 				if ((so->so_options & SO_REUSEPORT) != 0)
1476 					in6p->inp_flags2 |= INP_REUSEPORT;
1477 				else
1478 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1479 				INP_WUNLOCK(in6p);
1480 				error = 0;
1481 				break;
1482 			case SO_REUSEPORT_LB:
1483 				INP_WLOCK(in6p);
1484 				if ((so->so_options & SO_REUSEPORT_LB) != 0)
1485 					in6p->inp_flags2 |= INP_REUSEPORT_LB;
1486 				else
1487 					in6p->inp_flags2 &= ~INP_REUSEPORT_LB;
1488 				INP_WUNLOCK(in6p);
1489 				error = 0;
1490 				break;
1491 			case SO_SETFIB:
1492 				INP_WLOCK(in6p);
1493 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1494 				INP_WUNLOCK(in6p);
1495 				error = 0;
1496 				break;
1497 			case SO_MAX_PACING_RATE:
1498 #ifdef RATELIMIT
1499 				INP_WLOCK(in6p);
1500 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1501 				INP_WUNLOCK(in6p);
1502 				error = 0;
1503 #else
1504 				error = EOPNOTSUPP;
1505 #endif
1506 				break;
1507 			default:
1508 				break;
1509 			}
1510 		}
1511 	} else {		/* level == IPPROTO_IPV6 */
1512 		switch (op) {
1513 
1514 		case SOPT_SET:
1515 			switch (optname) {
1516 			case IPV6_2292PKTOPTIONS:
1517 #ifdef IPV6_PKTOPTIONS
1518 			case IPV6_PKTOPTIONS:
1519 #endif
1520 			{
1521 				struct mbuf *m;
1522 
1523 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1524 					printf("ip6_ctloutput: mbuf limit hit\n");
1525 					error = ENOBUFS;
1526 					break;
1527 				}
1528 
1529 				error = soopt_getm(sopt, &m); /* XXX */
1530 				if (error != 0)
1531 					break;
1532 				error = soopt_mcopyin(sopt, m); /* XXX */
1533 				if (error != 0)
1534 					break;
1535 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1536 						    m, so, sopt);
1537 				m_freem(m); /* XXX */
1538 				break;
1539 			}
1540 
1541 			/*
1542 			 * Use of some Hop-by-Hop options or some
1543 			 * Destination options, might require special
1544 			 * privilege.  That is, normal applications
1545 			 * (without special privilege) might be forbidden
1546 			 * from setting certain options in outgoing packets,
1547 			 * and might never see certain options in received
1548 			 * packets. [RFC 2292 Section 6]
1549 			 * KAME specific note:
1550 			 *  KAME prevents non-privileged users from sending or
1551 			 *  receiving ANY hbh/dst options in order to avoid
1552 			 *  overhead of parsing options in the kernel.
1553 			 */
1554 			case IPV6_RECVHOPOPTS:
1555 			case IPV6_RECVDSTOPTS:
1556 			case IPV6_RECVRTHDRDSTOPTS:
1557 				if (td != NULL) {
1558 					error = priv_check(td,
1559 					    PRIV_NETINET_SETHDROPTS);
1560 					if (error)
1561 						break;
1562 				}
1563 				/* FALLTHROUGH */
1564 			case IPV6_UNICAST_HOPS:
1565 			case IPV6_HOPLIMIT:
1566 
1567 			case IPV6_RECVPKTINFO:
1568 			case IPV6_RECVHOPLIMIT:
1569 			case IPV6_RECVRTHDR:
1570 			case IPV6_RECVPATHMTU:
1571 			case IPV6_RECVTCLASS:
1572 			case IPV6_RECVFLOWID:
1573 #ifdef	RSS
1574 			case IPV6_RECVRSSBUCKETID:
1575 #endif
1576 			case IPV6_V6ONLY:
1577 			case IPV6_AUTOFLOWLABEL:
1578 			case IPV6_ORIGDSTADDR:
1579 			case IPV6_BINDANY:
1580 			case IPV6_BINDMULTI:
1581 #ifdef	RSS
1582 			case IPV6_RSS_LISTEN_BUCKET:
1583 #endif
1584 				if (optname == IPV6_BINDANY && td != NULL) {
1585 					error = priv_check(td,
1586 					    PRIV_NETINET_BINDANY);
1587 					if (error)
1588 						break;
1589 				}
1590 
1591 				if (optlen != sizeof(int)) {
1592 					error = EINVAL;
1593 					break;
1594 				}
1595 				error = sooptcopyin(sopt, &optval,
1596 					sizeof optval, sizeof optval);
1597 				if (error)
1598 					break;
1599 				switch (optname) {
1600 
1601 				case IPV6_UNICAST_HOPS:
1602 					if (optval < -1 || optval >= 256)
1603 						error = EINVAL;
1604 					else {
1605 						/* -1 = kernel default */
1606 						in6p->in6p_hops = optval;
1607 						if ((in6p->inp_vflag &
1608 						     INP_IPV4) != 0)
1609 							in6p->inp_ip_ttl = optval;
1610 					}
1611 					break;
1612 #define OPTSET(bit) \
1613 do { \
1614 	INP_WLOCK(in6p); \
1615 	if (optval) \
1616 		in6p->inp_flags |= (bit); \
1617 	else \
1618 		in6p->inp_flags &= ~(bit); \
1619 	INP_WUNLOCK(in6p); \
1620 } while (/*CONSTCOND*/ 0)
1621 #define OPTSET2292(bit) \
1622 do { \
1623 	INP_WLOCK(in6p); \
1624 	in6p->inp_flags |= IN6P_RFC2292; \
1625 	if (optval) \
1626 		in6p->inp_flags |= (bit); \
1627 	else \
1628 		in6p->inp_flags &= ~(bit); \
1629 	INP_WUNLOCK(in6p); \
1630 } while (/*CONSTCOND*/ 0)
1631 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1632 
1633 #define OPTSET2_N(bit, val) do {					\
1634 	if (val)							\
1635 		in6p->inp_flags2 |= bit;				\
1636 	else								\
1637 		in6p->inp_flags2 &= ~bit;				\
1638 } while (0)
1639 #define OPTSET2(bit, val) do {						\
1640 	INP_WLOCK(in6p);						\
1641 	OPTSET2_N(bit, val);						\
1642 	INP_WUNLOCK(in6p);						\
1643 } while (0)
1644 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1645 #define OPTSET2292_EXCLUSIVE(bit)					\
1646 do {									\
1647 	INP_WLOCK(in6p);						\
1648 	if (OPTBIT(IN6P_RFC2292)) {					\
1649 		error = EINVAL;						\
1650 	} else {							\
1651 		if (optval)						\
1652 			in6p->inp_flags |= (bit);			\
1653 		else							\
1654 			in6p->inp_flags &= ~(bit);			\
1655 	}								\
1656 	INP_WUNLOCK(in6p);						\
1657 } while (/*CONSTCOND*/ 0)
1658 
1659 				case IPV6_RECVPKTINFO:
1660 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1661 					break;
1662 
1663 				case IPV6_HOPLIMIT:
1664 				{
1665 					struct ip6_pktopts **optp;
1666 
1667 					/* cannot mix with RFC2292 */
1668 					if (OPTBIT(IN6P_RFC2292)) {
1669 						error = EINVAL;
1670 						break;
1671 					}
1672 					INP_WLOCK(in6p);
1673 					if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1674 						INP_WUNLOCK(in6p);
1675 						return (ECONNRESET);
1676 					}
1677 					optp = &in6p->in6p_outputopts;
1678 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1679 					    (u_char *)&optval, sizeof(optval),
1680 					    optp, (td != NULL) ? td->td_ucred :
1681 					    NULL, uproto);
1682 					INP_WUNLOCK(in6p);
1683 					break;
1684 				}
1685 
1686 				case IPV6_RECVHOPLIMIT:
1687 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1688 					break;
1689 
1690 				case IPV6_RECVHOPOPTS:
1691 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1692 					break;
1693 
1694 				case IPV6_RECVDSTOPTS:
1695 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1696 					break;
1697 
1698 				case IPV6_RECVRTHDRDSTOPTS:
1699 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1700 					break;
1701 
1702 				case IPV6_RECVRTHDR:
1703 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1704 					break;
1705 
1706 				case IPV6_RECVPATHMTU:
1707 					/*
1708 					 * We ignore this option for TCP
1709 					 * sockets.
1710 					 * (RFC3542 leaves this case
1711 					 * unspecified.)
1712 					 */
1713 					if (uproto != IPPROTO_TCP)
1714 						OPTSET(IN6P_MTU);
1715 					break;
1716 
1717 				case IPV6_RECVFLOWID:
1718 					OPTSET2(INP_RECVFLOWID, optval);
1719 					break;
1720 
1721 #ifdef	RSS
1722 				case IPV6_RECVRSSBUCKETID:
1723 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1724 					break;
1725 #endif
1726 
1727 				case IPV6_V6ONLY:
1728 					/*
1729 					 * make setsockopt(IPV6_V6ONLY)
1730 					 * available only prior to bind(2).
1731 					 * see ipng mailing list, Jun 22 2001.
1732 					 */
1733 					if (in6p->inp_lport ||
1734 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1735 						error = EINVAL;
1736 						break;
1737 					}
1738 					OPTSET(IN6P_IPV6_V6ONLY);
1739 					if (optval)
1740 						in6p->inp_vflag &= ~INP_IPV4;
1741 					else
1742 						in6p->inp_vflag |= INP_IPV4;
1743 					break;
1744 				case IPV6_RECVTCLASS:
1745 					/* cannot mix with RFC2292 XXX */
1746 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1747 					break;
1748 				case IPV6_AUTOFLOWLABEL:
1749 					OPTSET(IN6P_AUTOFLOWLABEL);
1750 					break;
1751 
1752 				case IPV6_ORIGDSTADDR:
1753 					OPTSET2(INP_ORIGDSTADDR, optval);
1754 					break;
1755 				case IPV6_BINDANY:
1756 					OPTSET(INP_BINDANY);
1757 					break;
1758 
1759 				case IPV6_BINDMULTI:
1760 					OPTSET2(INP_BINDMULTI, optval);
1761 					break;
1762 #ifdef	RSS
1763 				case IPV6_RSS_LISTEN_BUCKET:
1764 					if ((optval >= 0) &&
1765 					    (optval < rss_getnumbuckets())) {
1766 						INP_WLOCK(in6p);
1767 						in6p->inp_rss_listen_bucket = optval;
1768 						OPTSET2_N(INP_RSS_BUCKET_SET, 1);
1769 						INP_WUNLOCK(in6p);
1770 					} else {
1771 						error = EINVAL;
1772 					}
1773 					break;
1774 #endif
1775 				}
1776 				break;
1777 
1778 			case IPV6_TCLASS:
1779 			case IPV6_DONTFRAG:
1780 			case IPV6_USE_MIN_MTU:
1781 			case IPV6_PREFER_TEMPADDR:
1782 				if (optlen != sizeof(optval)) {
1783 					error = EINVAL;
1784 					break;
1785 				}
1786 				error = sooptcopyin(sopt, &optval,
1787 					sizeof optval, sizeof optval);
1788 				if (error)
1789 					break;
1790 				{
1791 					struct ip6_pktopts **optp;
1792 					INP_WLOCK(in6p);
1793 					if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1794 						INP_WUNLOCK(in6p);
1795 						return (ECONNRESET);
1796 					}
1797 					optp = &in6p->in6p_outputopts;
1798 					error = ip6_pcbopt(optname,
1799 					    (u_char *)&optval, sizeof(optval),
1800 					    optp, (td != NULL) ? td->td_ucred :
1801 					    NULL, uproto);
1802 					INP_WUNLOCK(in6p);
1803 					break;
1804 				}
1805 
1806 			case IPV6_2292PKTINFO:
1807 			case IPV6_2292HOPLIMIT:
1808 			case IPV6_2292HOPOPTS:
1809 			case IPV6_2292DSTOPTS:
1810 			case IPV6_2292RTHDR:
1811 				/* RFC 2292 */
1812 				if (optlen != sizeof(int)) {
1813 					error = EINVAL;
1814 					break;
1815 				}
1816 				error = sooptcopyin(sopt, &optval,
1817 					sizeof optval, sizeof optval);
1818 				if (error)
1819 					break;
1820 				switch (optname) {
1821 				case IPV6_2292PKTINFO:
1822 					OPTSET2292(IN6P_PKTINFO);
1823 					break;
1824 				case IPV6_2292HOPLIMIT:
1825 					OPTSET2292(IN6P_HOPLIMIT);
1826 					break;
1827 				case IPV6_2292HOPOPTS:
1828 					/*
1829 					 * Check super-user privilege.
1830 					 * See comments for IPV6_RECVHOPOPTS.
1831 					 */
1832 					if (td != NULL) {
1833 						error = priv_check(td,
1834 						    PRIV_NETINET_SETHDROPTS);
1835 						if (error)
1836 							return (error);
1837 					}
1838 					OPTSET2292(IN6P_HOPOPTS);
1839 					break;
1840 				case IPV6_2292DSTOPTS:
1841 					if (td != NULL) {
1842 						error = priv_check(td,
1843 						    PRIV_NETINET_SETHDROPTS);
1844 						if (error)
1845 							return (error);
1846 					}
1847 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1848 					break;
1849 				case IPV6_2292RTHDR:
1850 					OPTSET2292(IN6P_RTHDR);
1851 					break;
1852 				}
1853 				break;
1854 			case IPV6_PKTINFO:
1855 			case IPV6_HOPOPTS:
1856 			case IPV6_RTHDR:
1857 			case IPV6_DSTOPTS:
1858 			case IPV6_RTHDRDSTOPTS:
1859 			case IPV6_NEXTHOP:
1860 			{
1861 				/* new advanced API (RFC3542) */
1862 				u_char *optbuf;
1863 				u_char optbuf_storage[MCLBYTES];
1864 				int optlen;
1865 				struct ip6_pktopts **optp;
1866 
1867 				/* cannot mix with RFC2292 */
1868 				if (OPTBIT(IN6P_RFC2292)) {
1869 					error = EINVAL;
1870 					break;
1871 				}
1872 
1873 				/*
1874 				 * We only ensure valsize is not too large
1875 				 * here.  Further validation will be done
1876 				 * later.
1877 				 */
1878 				error = sooptcopyin(sopt, optbuf_storage,
1879 				    sizeof(optbuf_storage), 0);
1880 				if (error)
1881 					break;
1882 				optlen = sopt->sopt_valsize;
1883 				optbuf = optbuf_storage;
1884 				INP_WLOCK(in6p);
1885 				if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1886 					INP_WUNLOCK(in6p);
1887 					return (ECONNRESET);
1888 				}
1889 				optp = &in6p->in6p_outputopts;
1890 				error = ip6_pcbopt(optname, optbuf, optlen,
1891 				    optp, (td != NULL) ? td->td_ucred : NULL,
1892 				    uproto);
1893 				INP_WUNLOCK(in6p);
1894 				break;
1895 			}
1896 #undef OPTSET
1897 
1898 			case IPV6_MULTICAST_IF:
1899 			case IPV6_MULTICAST_HOPS:
1900 			case IPV6_MULTICAST_LOOP:
1901 			case IPV6_JOIN_GROUP:
1902 			case IPV6_LEAVE_GROUP:
1903 			case IPV6_MSFILTER:
1904 			case MCAST_BLOCK_SOURCE:
1905 			case MCAST_UNBLOCK_SOURCE:
1906 			case MCAST_JOIN_GROUP:
1907 			case MCAST_LEAVE_GROUP:
1908 			case MCAST_JOIN_SOURCE_GROUP:
1909 			case MCAST_LEAVE_SOURCE_GROUP:
1910 				error = ip6_setmoptions(in6p, sopt);
1911 				break;
1912 
1913 			case IPV6_PORTRANGE:
1914 				error = sooptcopyin(sopt, &optval,
1915 				    sizeof optval, sizeof optval);
1916 				if (error)
1917 					break;
1918 
1919 				INP_WLOCK(in6p);
1920 				switch (optval) {
1921 				case IPV6_PORTRANGE_DEFAULT:
1922 					in6p->inp_flags &= ~(INP_LOWPORT);
1923 					in6p->inp_flags &= ~(INP_HIGHPORT);
1924 					break;
1925 
1926 				case IPV6_PORTRANGE_HIGH:
1927 					in6p->inp_flags &= ~(INP_LOWPORT);
1928 					in6p->inp_flags |= INP_HIGHPORT;
1929 					break;
1930 
1931 				case IPV6_PORTRANGE_LOW:
1932 					in6p->inp_flags &= ~(INP_HIGHPORT);
1933 					in6p->inp_flags |= INP_LOWPORT;
1934 					break;
1935 
1936 				default:
1937 					error = EINVAL;
1938 					break;
1939 				}
1940 				INP_WUNLOCK(in6p);
1941 				break;
1942 
1943 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1944 			case IPV6_IPSEC_POLICY:
1945 				if (IPSEC_ENABLED(ipv6)) {
1946 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1947 					break;
1948 				}
1949 				/* FALLTHROUGH */
1950 #endif /* IPSEC */
1951 
1952 			default:
1953 				error = ENOPROTOOPT;
1954 				break;
1955 			}
1956 			break;
1957 
1958 		case SOPT_GET:
1959 			switch (optname) {
1960 
1961 			case IPV6_2292PKTOPTIONS:
1962 #ifdef IPV6_PKTOPTIONS
1963 			case IPV6_PKTOPTIONS:
1964 #endif
1965 				/*
1966 				 * RFC3542 (effectively) deprecated the
1967 				 * semantics of the 2292-style pktoptions.
1968 				 * Since it was not reliable in nature (i.e.,
1969 				 * applications had to expect the lack of some
1970 				 * information after all), it would make sense
1971 				 * to simplify this part by always returning
1972 				 * empty data.
1973 				 */
1974 				sopt->sopt_valsize = 0;
1975 				break;
1976 
1977 			case IPV6_RECVHOPOPTS:
1978 			case IPV6_RECVDSTOPTS:
1979 			case IPV6_RECVRTHDRDSTOPTS:
1980 			case IPV6_UNICAST_HOPS:
1981 			case IPV6_RECVPKTINFO:
1982 			case IPV6_RECVHOPLIMIT:
1983 			case IPV6_RECVRTHDR:
1984 			case IPV6_RECVPATHMTU:
1985 
1986 			case IPV6_V6ONLY:
1987 			case IPV6_PORTRANGE:
1988 			case IPV6_RECVTCLASS:
1989 			case IPV6_AUTOFLOWLABEL:
1990 			case IPV6_BINDANY:
1991 			case IPV6_FLOWID:
1992 			case IPV6_FLOWTYPE:
1993 			case IPV6_RECVFLOWID:
1994 #ifdef	RSS
1995 			case IPV6_RSSBUCKETID:
1996 			case IPV6_RECVRSSBUCKETID:
1997 #endif
1998 			case IPV6_BINDMULTI:
1999 				switch (optname) {
2000 
2001 				case IPV6_RECVHOPOPTS:
2002 					optval = OPTBIT(IN6P_HOPOPTS);
2003 					break;
2004 
2005 				case IPV6_RECVDSTOPTS:
2006 					optval = OPTBIT(IN6P_DSTOPTS);
2007 					break;
2008 
2009 				case IPV6_RECVRTHDRDSTOPTS:
2010 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2011 					break;
2012 
2013 				case IPV6_UNICAST_HOPS:
2014 					optval = in6p->in6p_hops;
2015 					break;
2016 
2017 				case IPV6_RECVPKTINFO:
2018 					optval = OPTBIT(IN6P_PKTINFO);
2019 					break;
2020 
2021 				case IPV6_RECVHOPLIMIT:
2022 					optval = OPTBIT(IN6P_HOPLIMIT);
2023 					break;
2024 
2025 				case IPV6_RECVRTHDR:
2026 					optval = OPTBIT(IN6P_RTHDR);
2027 					break;
2028 
2029 				case IPV6_RECVPATHMTU:
2030 					optval = OPTBIT(IN6P_MTU);
2031 					break;
2032 
2033 				case IPV6_V6ONLY:
2034 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2035 					break;
2036 
2037 				case IPV6_PORTRANGE:
2038 				    {
2039 					int flags;
2040 					flags = in6p->inp_flags;
2041 					if (flags & INP_HIGHPORT)
2042 						optval = IPV6_PORTRANGE_HIGH;
2043 					else if (flags & INP_LOWPORT)
2044 						optval = IPV6_PORTRANGE_LOW;
2045 					else
2046 						optval = 0;
2047 					break;
2048 				    }
2049 				case IPV6_RECVTCLASS:
2050 					optval = OPTBIT(IN6P_TCLASS);
2051 					break;
2052 
2053 				case IPV6_AUTOFLOWLABEL:
2054 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2055 					break;
2056 
2057 				case IPV6_ORIGDSTADDR:
2058 					optval = OPTBIT2(INP_ORIGDSTADDR);
2059 					break;
2060 
2061 				case IPV6_BINDANY:
2062 					optval = OPTBIT(INP_BINDANY);
2063 					break;
2064 
2065 				case IPV6_FLOWID:
2066 					optval = in6p->inp_flowid;
2067 					break;
2068 
2069 				case IPV6_FLOWTYPE:
2070 					optval = in6p->inp_flowtype;
2071 					break;
2072 
2073 				case IPV6_RECVFLOWID:
2074 					optval = OPTBIT2(INP_RECVFLOWID);
2075 					break;
2076 #ifdef	RSS
2077 				case IPV6_RSSBUCKETID:
2078 					retval =
2079 					    rss_hash2bucket(in6p->inp_flowid,
2080 					    in6p->inp_flowtype,
2081 					    &rss_bucket);
2082 					if (retval == 0)
2083 						optval = rss_bucket;
2084 					else
2085 						error = EINVAL;
2086 					break;
2087 
2088 				case IPV6_RECVRSSBUCKETID:
2089 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2090 					break;
2091 #endif
2092 
2093 				case IPV6_BINDMULTI:
2094 					optval = OPTBIT2(INP_BINDMULTI);
2095 					break;
2096 
2097 				}
2098 				if (error)
2099 					break;
2100 				error = sooptcopyout(sopt, &optval,
2101 					sizeof optval);
2102 				break;
2103 
2104 			case IPV6_PATHMTU:
2105 			{
2106 				u_long pmtu = 0;
2107 				struct ip6_mtuinfo mtuinfo;
2108 				struct in6_addr addr;
2109 
2110 				if (!(so->so_state & SS_ISCONNECTED))
2111 					return (ENOTCONN);
2112 				/*
2113 				 * XXX: we dot not consider the case of source
2114 				 * routing, or optional information to specify
2115 				 * the outgoing interface.
2116 				 * Copy faddr out of in6p to avoid holding lock
2117 				 * on inp during route lookup.
2118 				 */
2119 				INP_RLOCK(in6p);
2120 				bcopy(&in6p->in6p_faddr, &addr, sizeof(addr));
2121 				INP_RUNLOCK(in6p);
2122 				error = ip6_getpmtu_ctl(so->so_fibnum,
2123 				    &addr, &pmtu);
2124 				if (error)
2125 					break;
2126 				if (pmtu > IPV6_MAXPACKET)
2127 					pmtu = IPV6_MAXPACKET;
2128 
2129 				bzero(&mtuinfo, sizeof(mtuinfo));
2130 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2131 				optdata = (void *)&mtuinfo;
2132 				optdatalen = sizeof(mtuinfo);
2133 				error = sooptcopyout(sopt, optdata,
2134 				    optdatalen);
2135 				break;
2136 			}
2137 
2138 			case IPV6_2292PKTINFO:
2139 			case IPV6_2292HOPLIMIT:
2140 			case IPV6_2292HOPOPTS:
2141 			case IPV6_2292RTHDR:
2142 			case IPV6_2292DSTOPTS:
2143 				switch (optname) {
2144 				case IPV6_2292PKTINFO:
2145 					optval = OPTBIT(IN6P_PKTINFO);
2146 					break;
2147 				case IPV6_2292HOPLIMIT:
2148 					optval = OPTBIT(IN6P_HOPLIMIT);
2149 					break;
2150 				case IPV6_2292HOPOPTS:
2151 					optval = OPTBIT(IN6P_HOPOPTS);
2152 					break;
2153 				case IPV6_2292RTHDR:
2154 					optval = OPTBIT(IN6P_RTHDR);
2155 					break;
2156 				case IPV6_2292DSTOPTS:
2157 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2158 					break;
2159 				}
2160 				error = sooptcopyout(sopt, &optval,
2161 				    sizeof optval);
2162 				break;
2163 			case IPV6_PKTINFO:
2164 			case IPV6_HOPOPTS:
2165 			case IPV6_RTHDR:
2166 			case IPV6_DSTOPTS:
2167 			case IPV6_RTHDRDSTOPTS:
2168 			case IPV6_NEXTHOP:
2169 			case IPV6_TCLASS:
2170 			case IPV6_DONTFRAG:
2171 			case IPV6_USE_MIN_MTU:
2172 			case IPV6_PREFER_TEMPADDR:
2173 				error = ip6_getpcbopt(in6p, optname, sopt);
2174 				break;
2175 
2176 			case IPV6_MULTICAST_IF:
2177 			case IPV6_MULTICAST_HOPS:
2178 			case IPV6_MULTICAST_LOOP:
2179 			case IPV6_MSFILTER:
2180 				error = ip6_getmoptions(in6p, sopt);
2181 				break;
2182 
2183 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2184 			case IPV6_IPSEC_POLICY:
2185 				if (IPSEC_ENABLED(ipv6)) {
2186 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2187 					break;
2188 				}
2189 				/* FALLTHROUGH */
2190 #endif /* IPSEC */
2191 			default:
2192 				error = ENOPROTOOPT;
2193 				break;
2194 			}
2195 			break;
2196 		}
2197 	}
2198 	return (error);
2199 }
2200 
2201 int
2202 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2203 {
2204 	int error = 0, optval, optlen;
2205 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2206 	struct inpcb *in6p = sotoinpcb(so);
2207 	int level, op, optname;
2208 
2209 	level = sopt->sopt_level;
2210 	op = sopt->sopt_dir;
2211 	optname = sopt->sopt_name;
2212 	optlen = sopt->sopt_valsize;
2213 
2214 	if (level != IPPROTO_IPV6) {
2215 		return (EINVAL);
2216 	}
2217 
2218 	switch (optname) {
2219 	case IPV6_CHECKSUM:
2220 		/*
2221 		 * For ICMPv6 sockets, no modification allowed for checksum
2222 		 * offset, permit "no change" values to help existing apps.
2223 		 *
2224 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2225 		 * for an ICMPv6 socket will fail."
2226 		 * The current behavior does not meet RFC3542.
2227 		 */
2228 		switch (op) {
2229 		case SOPT_SET:
2230 			if (optlen != sizeof(int)) {
2231 				error = EINVAL;
2232 				break;
2233 			}
2234 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2235 					    sizeof(optval));
2236 			if (error)
2237 				break;
2238 			if (optval < -1 || (optval % 2) != 0) {
2239 				/*
2240 				 * The API assumes non-negative even offset
2241 				 * values or -1 as a special value.
2242 				 */
2243 				error = EINVAL;
2244 			} else if (so->so_proto->pr_protocol ==
2245 			    IPPROTO_ICMPV6) {
2246 				if (optval != icmp6off)
2247 					error = EINVAL;
2248 			} else
2249 				in6p->in6p_cksum = optval;
2250 			break;
2251 
2252 		case SOPT_GET:
2253 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2254 				optval = icmp6off;
2255 			else
2256 				optval = in6p->in6p_cksum;
2257 
2258 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2259 			break;
2260 
2261 		default:
2262 			error = EINVAL;
2263 			break;
2264 		}
2265 		break;
2266 
2267 	default:
2268 		error = ENOPROTOOPT;
2269 		break;
2270 	}
2271 
2272 	return (error);
2273 }
2274 
2275 /*
2276  * Set up IP6 options in pcb for insertion in output packets or
2277  * specifying behavior of outgoing packets.
2278  */
2279 static int
2280 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2281     struct socket *so, struct sockopt *sopt)
2282 {
2283 	struct ip6_pktopts *opt = *pktopt;
2284 	int error = 0;
2285 	struct thread *td = sopt->sopt_td;
2286 
2287 	/* turn off any old options. */
2288 	if (opt) {
2289 #ifdef DIAGNOSTIC
2290 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2291 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2292 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2293 			printf("ip6_pcbopts: all specified options are cleared.\n");
2294 #endif
2295 		ip6_clearpktopts(opt, -1);
2296 	} else
2297 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2298 	*pktopt = NULL;
2299 
2300 	if (!m || m->m_len == 0) {
2301 		/*
2302 		 * Only turning off any previous options, regardless of
2303 		 * whether the opt is just created or given.
2304 		 */
2305 		free(opt, M_IP6OPT);
2306 		return (0);
2307 	}
2308 
2309 	/*  set options specified by user. */
2310 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2311 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2312 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2313 		free(opt, M_IP6OPT);
2314 		return (error);
2315 	}
2316 	*pktopt = opt;
2317 	return (0);
2318 }
2319 
2320 /*
2321  * initialize ip6_pktopts.  beware that there are non-zero default values in
2322  * the struct.
2323  */
2324 void
2325 ip6_initpktopts(struct ip6_pktopts *opt)
2326 {
2327 
2328 	bzero(opt, sizeof(*opt));
2329 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2330 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2331 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2332 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2333 }
2334 
2335 static int
2336 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2337     struct ucred *cred, int uproto)
2338 {
2339 	struct ip6_pktopts *opt;
2340 
2341 	if (*pktopt == NULL) {
2342 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2343 		    M_NOWAIT);
2344 		if (*pktopt == NULL)
2345 			return (ENOBUFS);
2346 		ip6_initpktopts(*pktopt);
2347 	}
2348 	opt = *pktopt;
2349 
2350 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2351 }
2352 
2353 #define GET_PKTOPT_VAR(field, lenexpr) do {					\
2354 	if (pktopt && pktopt->field) {						\
2355 		INP_RUNLOCK(in6p);						\
2356 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);		\
2357 		malloc_optdata = true;						\
2358 		INP_RLOCK(in6p);						\
2359 		if (in6p->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
2360 			INP_RUNLOCK(in6p);					\
2361 			free(optdata, M_TEMP);					\
2362 			return (ECONNRESET);					\
2363 		}								\
2364 		pktopt = in6p->in6p_outputopts;					\
2365 		if (pktopt && pktopt->field) {					\
2366 			optdatalen = min(lenexpr, sopt->sopt_valsize);		\
2367 			bcopy(&pktopt->field, optdata, optdatalen);		\
2368 		} else {							\
2369 			free(optdata, M_TEMP);					\
2370 			optdata = NULL;						\
2371 			malloc_optdata = false;					\
2372 		}								\
2373 	}									\
2374 } while(0)
2375 
2376 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,				\
2377 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2378 
2379 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,			\
2380 	pktopt->field->sa_len)
2381 
2382 static int
2383 ip6_getpcbopt(struct inpcb *in6p, int optname, struct sockopt *sopt)
2384 {
2385 	void *optdata = NULL;
2386 	bool malloc_optdata = false;
2387 	int optdatalen = 0;
2388 	int error = 0;
2389 	struct in6_pktinfo null_pktinfo;
2390 	int deftclass = 0, on;
2391 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2392 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2393 	struct ip6_pktopts *pktopt;
2394 
2395 	INP_RLOCK(in6p);
2396 	pktopt = in6p->in6p_outputopts;
2397 
2398 	switch (optname) {
2399 	case IPV6_PKTINFO:
2400 		optdata = (void *)&null_pktinfo;
2401 		if (pktopt && pktopt->ip6po_pktinfo) {
2402 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2403 			    sizeof(null_pktinfo));
2404 			in6_clearscope(&null_pktinfo.ipi6_addr);
2405 		} else {
2406 			/* XXX: we don't have to do this every time... */
2407 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2408 		}
2409 		optdatalen = sizeof(struct in6_pktinfo);
2410 		break;
2411 	case IPV6_TCLASS:
2412 		if (pktopt && pktopt->ip6po_tclass >= 0)
2413 			deftclass = pktopt->ip6po_tclass;
2414 		optdata = (void *)&deftclass;
2415 		optdatalen = sizeof(int);
2416 		break;
2417 	case IPV6_HOPOPTS:
2418 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2419 		break;
2420 	case IPV6_RTHDR:
2421 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2422 		break;
2423 	case IPV6_RTHDRDSTOPTS:
2424 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2425 		break;
2426 	case IPV6_DSTOPTS:
2427 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2428 		break;
2429 	case IPV6_NEXTHOP:
2430 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2431 		break;
2432 	case IPV6_USE_MIN_MTU:
2433 		if (pktopt)
2434 			defminmtu = pktopt->ip6po_minmtu;
2435 		optdata = (void *)&defminmtu;
2436 		optdatalen = sizeof(int);
2437 		break;
2438 	case IPV6_DONTFRAG:
2439 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2440 			on = 1;
2441 		else
2442 			on = 0;
2443 		optdata = (void *)&on;
2444 		optdatalen = sizeof(on);
2445 		break;
2446 	case IPV6_PREFER_TEMPADDR:
2447 		if (pktopt)
2448 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2449 		optdata = (void *)&defpreftemp;
2450 		optdatalen = sizeof(int);
2451 		break;
2452 	default:		/* should not happen */
2453 #ifdef DIAGNOSTIC
2454 		panic("ip6_getpcbopt: unexpected option\n");
2455 #endif
2456 		INP_RUNLOCK(in6p);
2457 		return (ENOPROTOOPT);
2458 	}
2459 	INP_RUNLOCK(in6p);
2460 
2461 	error = sooptcopyout(sopt, optdata, optdatalen);
2462 	if (malloc_optdata)
2463 		free(optdata, M_TEMP);
2464 
2465 	return (error);
2466 }
2467 
2468 void
2469 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2470 {
2471 	if (pktopt == NULL)
2472 		return;
2473 
2474 	if (optname == -1 || optname == IPV6_PKTINFO) {
2475 		if (pktopt->ip6po_pktinfo)
2476 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2477 		pktopt->ip6po_pktinfo = NULL;
2478 	}
2479 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2480 		pktopt->ip6po_hlim = -1;
2481 	if (optname == -1 || optname == IPV6_TCLASS)
2482 		pktopt->ip6po_tclass = -1;
2483 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2484 		if (pktopt->ip6po_nextroute.ro_rt) {
2485 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2486 			pktopt->ip6po_nextroute.ro_rt = NULL;
2487 		}
2488 		if (pktopt->ip6po_nexthop)
2489 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2490 		pktopt->ip6po_nexthop = NULL;
2491 	}
2492 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2493 		if (pktopt->ip6po_hbh)
2494 			free(pktopt->ip6po_hbh, M_IP6OPT);
2495 		pktopt->ip6po_hbh = NULL;
2496 	}
2497 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2498 		if (pktopt->ip6po_dest1)
2499 			free(pktopt->ip6po_dest1, M_IP6OPT);
2500 		pktopt->ip6po_dest1 = NULL;
2501 	}
2502 	if (optname == -1 || optname == IPV6_RTHDR) {
2503 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2504 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2505 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2506 		if (pktopt->ip6po_route.ro_rt) {
2507 			RTFREE(pktopt->ip6po_route.ro_rt);
2508 			pktopt->ip6po_route.ro_rt = NULL;
2509 		}
2510 	}
2511 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2512 		if (pktopt->ip6po_dest2)
2513 			free(pktopt->ip6po_dest2, M_IP6OPT);
2514 		pktopt->ip6po_dest2 = NULL;
2515 	}
2516 }
2517 
2518 #define PKTOPT_EXTHDRCPY(type) \
2519 do {\
2520 	if (src->type) {\
2521 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2522 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2523 		if (dst->type == NULL)\
2524 			goto bad;\
2525 		bcopy(src->type, dst->type, hlen);\
2526 	}\
2527 } while (/*CONSTCOND*/ 0)
2528 
2529 static int
2530 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2531 {
2532 	if (dst == NULL || src == NULL)  {
2533 		printf("ip6_clearpktopts: invalid argument\n");
2534 		return (EINVAL);
2535 	}
2536 
2537 	dst->ip6po_hlim = src->ip6po_hlim;
2538 	dst->ip6po_tclass = src->ip6po_tclass;
2539 	dst->ip6po_flags = src->ip6po_flags;
2540 	dst->ip6po_minmtu = src->ip6po_minmtu;
2541 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2542 	if (src->ip6po_pktinfo) {
2543 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2544 		    M_IP6OPT, canwait);
2545 		if (dst->ip6po_pktinfo == NULL)
2546 			goto bad;
2547 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2548 	}
2549 	if (src->ip6po_nexthop) {
2550 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2551 		    M_IP6OPT, canwait);
2552 		if (dst->ip6po_nexthop == NULL)
2553 			goto bad;
2554 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2555 		    src->ip6po_nexthop->sa_len);
2556 	}
2557 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2558 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2559 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2560 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2561 	return (0);
2562 
2563   bad:
2564 	ip6_clearpktopts(dst, -1);
2565 	return (ENOBUFS);
2566 }
2567 #undef PKTOPT_EXTHDRCPY
2568 
2569 struct ip6_pktopts *
2570 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2571 {
2572 	int error;
2573 	struct ip6_pktopts *dst;
2574 
2575 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2576 	if (dst == NULL)
2577 		return (NULL);
2578 	ip6_initpktopts(dst);
2579 
2580 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2581 		free(dst, M_IP6OPT);
2582 		return (NULL);
2583 	}
2584 
2585 	return (dst);
2586 }
2587 
2588 void
2589 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2590 {
2591 	if (pktopt == NULL)
2592 		return;
2593 
2594 	ip6_clearpktopts(pktopt, -1);
2595 
2596 	free(pktopt, M_IP6OPT);
2597 }
2598 
2599 /*
2600  * Set IPv6 outgoing packet options based on advanced API.
2601  */
2602 int
2603 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2604     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2605 {
2606 	struct cmsghdr *cm = NULL;
2607 
2608 	if (control == NULL || opt == NULL)
2609 		return (EINVAL);
2610 
2611 	ip6_initpktopts(opt);
2612 	if (stickyopt) {
2613 		int error;
2614 
2615 		/*
2616 		 * If stickyopt is provided, make a local copy of the options
2617 		 * for this particular packet, then override them by ancillary
2618 		 * objects.
2619 		 * XXX: copypktopts() does not copy the cached route to a next
2620 		 * hop (if any).  This is not very good in terms of efficiency,
2621 		 * but we can allow this since this option should be rarely
2622 		 * used.
2623 		 */
2624 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2625 			return (error);
2626 	}
2627 
2628 	/*
2629 	 * XXX: Currently, we assume all the optional information is stored
2630 	 * in a single mbuf.
2631 	 */
2632 	if (control->m_next)
2633 		return (EINVAL);
2634 
2635 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2636 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2637 		int error;
2638 
2639 		if (control->m_len < CMSG_LEN(0))
2640 			return (EINVAL);
2641 
2642 		cm = mtod(control, struct cmsghdr *);
2643 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2644 			return (EINVAL);
2645 		if (cm->cmsg_level != IPPROTO_IPV6)
2646 			continue;
2647 
2648 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2649 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2650 		if (error)
2651 			return (error);
2652 	}
2653 
2654 	return (0);
2655 }
2656 
2657 /*
2658  * Set a particular packet option, as a sticky option or an ancillary data
2659  * item.  "len" can be 0 only when it's a sticky option.
2660  * We have 4 cases of combination of "sticky" and "cmsg":
2661  * "sticky=0, cmsg=0": impossible
2662  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2663  * "sticky=1, cmsg=0": RFC3542 socket option
2664  * "sticky=1, cmsg=1": RFC2292 socket option
2665  */
2666 static int
2667 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2668     struct ucred *cred, int sticky, int cmsg, int uproto)
2669 {
2670 	int minmtupolicy, preftemp;
2671 	int error;
2672 
2673 	if (!sticky && !cmsg) {
2674 #ifdef DIAGNOSTIC
2675 		printf("ip6_setpktopt: impossible case\n");
2676 #endif
2677 		return (EINVAL);
2678 	}
2679 
2680 	/*
2681 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2682 	 * not be specified in the context of RFC3542.  Conversely,
2683 	 * RFC3542 types should not be specified in the context of RFC2292.
2684 	 */
2685 	if (!cmsg) {
2686 		switch (optname) {
2687 		case IPV6_2292PKTINFO:
2688 		case IPV6_2292HOPLIMIT:
2689 		case IPV6_2292NEXTHOP:
2690 		case IPV6_2292HOPOPTS:
2691 		case IPV6_2292DSTOPTS:
2692 		case IPV6_2292RTHDR:
2693 		case IPV6_2292PKTOPTIONS:
2694 			return (ENOPROTOOPT);
2695 		}
2696 	}
2697 	if (sticky && cmsg) {
2698 		switch (optname) {
2699 		case IPV6_PKTINFO:
2700 		case IPV6_HOPLIMIT:
2701 		case IPV6_NEXTHOP:
2702 		case IPV6_HOPOPTS:
2703 		case IPV6_DSTOPTS:
2704 		case IPV6_RTHDRDSTOPTS:
2705 		case IPV6_RTHDR:
2706 		case IPV6_USE_MIN_MTU:
2707 		case IPV6_DONTFRAG:
2708 		case IPV6_TCLASS:
2709 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2710 			return (ENOPROTOOPT);
2711 		}
2712 	}
2713 
2714 	switch (optname) {
2715 	case IPV6_2292PKTINFO:
2716 	case IPV6_PKTINFO:
2717 	{
2718 		struct ifnet *ifp = NULL;
2719 		struct in6_pktinfo *pktinfo;
2720 
2721 		if (len != sizeof(struct in6_pktinfo))
2722 			return (EINVAL);
2723 
2724 		pktinfo = (struct in6_pktinfo *)buf;
2725 
2726 		/*
2727 		 * An application can clear any sticky IPV6_PKTINFO option by
2728 		 * doing a "regular" setsockopt with ipi6_addr being
2729 		 * in6addr_any and ipi6_ifindex being zero.
2730 		 * [RFC 3542, Section 6]
2731 		 */
2732 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2733 		    pktinfo->ipi6_ifindex == 0 &&
2734 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2735 			ip6_clearpktopts(opt, optname);
2736 			break;
2737 		}
2738 
2739 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2740 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2741 			return (EINVAL);
2742 		}
2743 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2744 			return (EINVAL);
2745 		/* validate the interface index if specified. */
2746 		if (pktinfo->ipi6_ifindex > V_if_index)
2747 			 return (ENXIO);
2748 		if (pktinfo->ipi6_ifindex) {
2749 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2750 			if (ifp == NULL)
2751 				return (ENXIO);
2752 		}
2753 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2754 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2755 			return (ENETDOWN);
2756 
2757 		if (ifp != NULL &&
2758 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2759 			struct in6_ifaddr *ia;
2760 
2761 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2762 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2763 			if (ia == NULL)
2764 				return (EADDRNOTAVAIL);
2765 			ifa_free(&ia->ia_ifa);
2766 		}
2767 		/*
2768 		 * We store the address anyway, and let in6_selectsrc()
2769 		 * validate the specified address.  This is because ipi6_addr
2770 		 * may not have enough information about its scope zone, and
2771 		 * we may need additional information (such as outgoing
2772 		 * interface or the scope zone of a destination address) to
2773 		 * disambiguate the scope.
2774 		 * XXX: the delay of the validation may confuse the
2775 		 * application when it is used as a sticky option.
2776 		 */
2777 		if (opt->ip6po_pktinfo == NULL) {
2778 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2779 			    M_IP6OPT, M_NOWAIT);
2780 			if (opt->ip6po_pktinfo == NULL)
2781 				return (ENOBUFS);
2782 		}
2783 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2784 		break;
2785 	}
2786 
2787 	case IPV6_2292HOPLIMIT:
2788 	case IPV6_HOPLIMIT:
2789 	{
2790 		int *hlimp;
2791 
2792 		/*
2793 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2794 		 * to simplify the ordering among hoplimit options.
2795 		 */
2796 		if (optname == IPV6_HOPLIMIT && sticky)
2797 			return (ENOPROTOOPT);
2798 
2799 		if (len != sizeof(int))
2800 			return (EINVAL);
2801 		hlimp = (int *)buf;
2802 		if (*hlimp < -1 || *hlimp > 255)
2803 			return (EINVAL);
2804 
2805 		opt->ip6po_hlim = *hlimp;
2806 		break;
2807 	}
2808 
2809 	case IPV6_TCLASS:
2810 	{
2811 		int tclass;
2812 
2813 		if (len != sizeof(int))
2814 			return (EINVAL);
2815 		tclass = *(int *)buf;
2816 		if (tclass < -1 || tclass > 255)
2817 			return (EINVAL);
2818 
2819 		opt->ip6po_tclass = tclass;
2820 		break;
2821 	}
2822 
2823 	case IPV6_2292NEXTHOP:
2824 	case IPV6_NEXTHOP:
2825 		if (cred != NULL) {
2826 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2827 			if (error)
2828 				return (error);
2829 		}
2830 
2831 		if (len == 0) {	/* just remove the option */
2832 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2833 			break;
2834 		}
2835 
2836 		/* check if cmsg_len is large enough for sa_len */
2837 		if (len < sizeof(struct sockaddr) || len < *buf)
2838 			return (EINVAL);
2839 
2840 		switch (((struct sockaddr *)buf)->sa_family) {
2841 		case AF_INET6:
2842 		{
2843 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2844 			int error;
2845 
2846 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2847 				return (EINVAL);
2848 
2849 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2850 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2851 				return (EINVAL);
2852 			}
2853 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2854 			    != 0) {
2855 				return (error);
2856 			}
2857 			break;
2858 		}
2859 		case AF_LINK:	/* should eventually be supported */
2860 		default:
2861 			return (EAFNOSUPPORT);
2862 		}
2863 
2864 		/* turn off the previous option, then set the new option. */
2865 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2866 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2867 		if (opt->ip6po_nexthop == NULL)
2868 			return (ENOBUFS);
2869 		bcopy(buf, opt->ip6po_nexthop, *buf);
2870 		break;
2871 
2872 	case IPV6_2292HOPOPTS:
2873 	case IPV6_HOPOPTS:
2874 	{
2875 		struct ip6_hbh *hbh;
2876 		int hbhlen;
2877 
2878 		/*
2879 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2880 		 * options, since per-option restriction has too much
2881 		 * overhead.
2882 		 */
2883 		if (cred != NULL) {
2884 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2885 			if (error)
2886 				return (error);
2887 		}
2888 
2889 		if (len == 0) {
2890 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2891 			break;	/* just remove the option */
2892 		}
2893 
2894 		/* message length validation */
2895 		if (len < sizeof(struct ip6_hbh))
2896 			return (EINVAL);
2897 		hbh = (struct ip6_hbh *)buf;
2898 		hbhlen = (hbh->ip6h_len + 1) << 3;
2899 		if (len != hbhlen)
2900 			return (EINVAL);
2901 
2902 		/* turn off the previous option, then set the new option. */
2903 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2904 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2905 		if (opt->ip6po_hbh == NULL)
2906 			return (ENOBUFS);
2907 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2908 
2909 		break;
2910 	}
2911 
2912 	case IPV6_2292DSTOPTS:
2913 	case IPV6_DSTOPTS:
2914 	case IPV6_RTHDRDSTOPTS:
2915 	{
2916 		struct ip6_dest *dest, **newdest = NULL;
2917 		int destlen;
2918 
2919 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2920 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
2921 			if (error)
2922 				return (error);
2923 		}
2924 
2925 		if (len == 0) {
2926 			ip6_clearpktopts(opt, optname);
2927 			break;	/* just remove the option */
2928 		}
2929 
2930 		/* message length validation */
2931 		if (len < sizeof(struct ip6_dest))
2932 			return (EINVAL);
2933 		dest = (struct ip6_dest *)buf;
2934 		destlen = (dest->ip6d_len + 1) << 3;
2935 		if (len != destlen)
2936 			return (EINVAL);
2937 
2938 		/*
2939 		 * Determine the position that the destination options header
2940 		 * should be inserted; before or after the routing header.
2941 		 */
2942 		switch (optname) {
2943 		case IPV6_2292DSTOPTS:
2944 			/*
2945 			 * The old advacned API is ambiguous on this point.
2946 			 * Our approach is to determine the position based
2947 			 * according to the existence of a routing header.
2948 			 * Note, however, that this depends on the order of the
2949 			 * extension headers in the ancillary data; the 1st
2950 			 * part of the destination options header must appear
2951 			 * before the routing header in the ancillary data,
2952 			 * too.
2953 			 * RFC3542 solved the ambiguity by introducing
2954 			 * separate ancillary data or option types.
2955 			 */
2956 			if (opt->ip6po_rthdr == NULL)
2957 				newdest = &opt->ip6po_dest1;
2958 			else
2959 				newdest = &opt->ip6po_dest2;
2960 			break;
2961 		case IPV6_RTHDRDSTOPTS:
2962 			newdest = &opt->ip6po_dest1;
2963 			break;
2964 		case IPV6_DSTOPTS:
2965 			newdest = &opt->ip6po_dest2;
2966 			break;
2967 		}
2968 
2969 		/* turn off the previous option, then set the new option. */
2970 		ip6_clearpktopts(opt, optname);
2971 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2972 		if (*newdest == NULL)
2973 			return (ENOBUFS);
2974 		bcopy(dest, *newdest, destlen);
2975 
2976 		break;
2977 	}
2978 
2979 	case IPV6_2292RTHDR:
2980 	case IPV6_RTHDR:
2981 	{
2982 		struct ip6_rthdr *rth;
2983 		int rthlen;
2984 
2985 		if (len == 0) {
2986 			ip6_clearpktopts(opt, IPV6_RTHDR);
2987 			break;	/* just remove the option */
2988 		}
2989 
2990 		/* message length validation */
2991 		if (len < sizeof(struct ip6_rthdr))
2992 			return (EINVAL);
2993 		rth = (struct ip6_rthdr *)buf;
2994 		rthlen = (rth->ip6r_len + 1) << 3;
2995 		if (len != rthlen)
2996 			return (EINVAL);
2997 
2998 		switch (rth->ip6r_type) {
2999 		case IPV6_RTHDR_TYPE_0:
3000 			if (rth->ip6r_len == 0)	/* must contain one addr */
3001 				return (EINVAL);
3002 			if (rth->ip6r_len % 2) /* length must be even */
3003 				return (EINVAL);
3004 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3005 				return (EINVAL);
3006 			break;
3007 		default:
3008 			return (EINVAL);	/* not supported */
3009 		}
3010 
3011 		/* turn off the previous option */
3012 		ip6_clearpktopts(opt, IPV6_RTHDR);
3013 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3014 		if (opt->ip6po_rthdr == NULL)
3015 			return (ENOBUFS);
3016 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3017 
3018 		break;
3019 	}
3020 
3021 	case IPV6_USE_MIN_MTU:
3022 		if (len != sizeof(int))
3023 			return (EINVAL);
3024 		minmtupolicy = *(int *)buf;
3025 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3026 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3027 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3028 			return (EINVAL);
3029 		}
3030 		opt->ip6po_minmtu = minmtupolicy;
3031 		break;
3032 
3033 	case IPV6_DONTFRAG:
3034 		if (len != sizeof(int))
3035 			return (EINVAL);
3036 
3037 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3038 			/*
3039 			 * we ignore this option for TCP sockets.
3040 			 * (RFC3542 leaves this case unspecified.)
3041 			 */
3042 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3043 		} else
3044 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3045 		break;
3046 
3047 	case IPV6_PREFER_TEMPADDR:
3048 		if (len != sizeof(int))
3049 			return (EINVAL);
3050 		preftemp = *(int *)buf;
3051 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3052 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3053 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3054 			return (EINVAL);
3055 		}
3056 		opt->ip6po_prefer_tempaddr = preftemp;
3057 		break;
3058 
3059 	default:
3060 		return (ENOPROTOOPT);
3061 	} /* end of switch */
3062 
3063 	return (0);
3064 }
3065 
3066 /*
3067  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3068  * packet to the input queue of a specified interface.  Note that this
3069  * calls the output routine of the loopback "driver", but with an interface
3070  * pointer that might NOT be &loif -- easier than replicating that code here.
3071  */
3072 void
3073 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3074 {
3075 	struct mbuf *copym;
3076 	struct ip6_hdr *ip6;
3077 
3078 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3079 	if (copym == NULL)
3080 		return;
3081 
3082 	/*
3083 	 * Make sure to deep-copy IPv6 header portion in case the data
3084 	 * is in an mbuf cluster, so that we can safely override the IPv6
3085 	 * header portion later.
3086 	 */
3087 	if (!M_WRITABLE(copym) ||
3088 	    copym->m_len < sizeof(struct ip6_hdr)) {
3089 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3090 		if (copym == NULL)
3091 			return;
3092 	}
3093 	ip6 = mtod(copym, struct ip6_hdr *);
3094 	/*
3095 	 * clear embedded scope identifiers if necessary.
3096 	 * in6_clearscope will touch the addresses only when necessary.
3097 	 */
3098 	in6_clearscope(&ip6->ip6_src);
3099 	in6_clearscope(&ip6->ip6_dst);
3100 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3101 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3102 		    CSUM_PSEUDO_HDR;
3103 		copym->m_pkthdr.csum_data = 0xffff;
3104 	}
3105 	if_simloop(ifp, copym, AF_INET6, 0);
3106 }
3107 
3108 /*
3109  * Chop IPv6 header off from the payload.
3110  */
3111 static int
3112 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3113 {
3114 	struct mbuf *mh;
3115 	struct ip6_hdr *ip6;
3116 
3117 	ip6 = mtod(m, struct ip6_hdr *);
3118 	if (m->m_len > sizeof(*ip6)) {
3119 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3120 		if (mh == NULL) {
3121 			m_freem(m);
3122 			return ENOBUFS;
3123 		}
3124 		m_move_pkthdr(mh, m);
3125 		M_ALIGN(mh, sizeof(*ip6));
3126 		m->m_len -= sizeof(*ip6);
3127 		m->m_data += sizeof(*ip6);
3128 		mh->m_next = m;
3129 		m = mh;
3130 		m->m_len = sizeof(*ip6);
3131 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3132 	}
3133 	exthdrs->ip6e_ip6 = m;
3134 	return 0;
3135 }
3136 
3137 /*
3138  * Compute IPv6 extension header length.
3139  */
3140 int
3141 ip6_optlen(struct inpcb *in6p)
3142 {
3143 	int len;
3144 
3145 	if (!in6p->in6p_outputopts)
3146 		return 0;
3147 
3148 	len = 0;
3149 #define elen(x) \
3150     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3151 
3152 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3153 	if (in6p->in6p_outputopts->ip6po_rthdr)
3154 		/* dest1 is valid with rthdr only */
3155 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3156 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3157 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3158 	return len;
3159 #undef elen
3160 }
3161