xref: /freebsd/sys/netinet6/ip6_output.c (revision 26d8ca3bc9cafaf8fae8764a6ad2cb3fbb84c8ee)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 #include "opt_sctp.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_llatbl.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/rss_config.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #include <netinet6/in6_fib.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_rss.h>
109 
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/ipsec6.h>
113 #include <netipsec/key.h>
114 #include <netinet6/ip6_ipsec.h>
115 #endif /* IPSEC */
116 #ifdef SCTP
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/ip6protosw.h>
122 #include <netinet6/scope6_var.h>
123 
124 #ifdef FLOWTABLE
125 #include <net/flowtable.h>
126 #endif
127 
128 extern int in6_mcast_loop;
129 
130 struct ip6_exthdrs {
131 	struct mbuf *ip6e_ip6;
132 	struct mbuf *ip6e_hbh;
133 	struct mbuf *ip6e_dest1;
134 	struct mbuf *ip6e_rthdr;
135 	struct mbuf *ip6e_dest2;
136 };
137 
138 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
139 
140 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
141 			   struct ucred *, int);
142 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
143 	struct socket *, struct sockopt *);
144 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
145 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
146 	struct ucred *, int, int, int);
147 
148 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
149 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
150 	struct ip6_frag **);
151 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
152 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
153 static int ip6_getpmtu(struct route_in6 *, int,
154 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
155 	u_int);
156 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
157 	u_long *, int *, u_int);
158 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
159 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
160 
161 
162 /*
163  * Make an extension header from option data.  hp is the source, and
164  * mp is the destination.
165  */
166 #define MAKE_EXTHDR(hp, mp)						\
167     do {								\
168 	if (hp) {							\
169 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
170 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
171 		    ((eh)->ip6e_len + 1) << 3);				\
172 		if (error)						\
173 			goto freehdrs;					\
174 	}								\
175     } while (/*CONSTCOND*/ 0)
176 
177 /*
178  * Form a chain of extension headers.
179  * m is the extension header mbuf
180  * mp is the previous mbuf in the chain
181  * p is the next header
182  * i is the type of option.
183  */
184 #define MAKE_CHAIN(m, mp, p, i)\
185     do {\
186 	if (m) {\
187 		if (!hdrsplit) \
188 			panic("assumption failed: hdr not split"); \
189 		*mtod((m), u_char *) = *(p);\
190 		*(p) = (i);\
191 		p = mtod((m), u_char *);\
192 		(m)->m_next = (mp)->m_next;\
193 		(mp)->m_next = (m);\
194 		(mp) = (m);\
195 	}\
196     } while (/*CONSTCOND*/ 0)
197 
198 void
199 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
200 {
201 	u_short csum;
202 
203 	csum = in_cksum_skip(m, offset + plen, offset);
204 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
205 		csum = 0xffff;
206 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
207 
208 	if (offset + sizeof(u_short) > m->m_len) {
209 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
210 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
211 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
212 		/*
213 		 * XXX this should not happen, but if it does, the correct
214 		 * behavior may be to insert the checksum in the appropriate
215 		 * next mbuf in the chain.
216 		 */
217 		return;
218 	}
219 	*(u_short *)(m->m_data + offset) = csum;
220 }
221 
222 int
223 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
224     int mtu, uint32_t id)
225 {
226 	struct mbuf *m, **mnext, *m_frgpart;
227 	struct ip6_hdr *ip6, *mhip6;
228 	struct ip6_frag *ip6f;
229 	int off;
230 	int error;
231 	int tlen = m0->m_pkthdr.len;
232 
233 	m = m0;
234 	ip6 = mtod(m, struct ip6_hdr *);
235 	mnext = &m->m_nextpkt;
236 
237 	for (off = hlen; off < tlen; off += mtu) {
238 		m = m_gethdr(M_NOWAIT, MT_DATA);
239 		if (!m) {
240 			IP6STAT_INC(ip6s_odropped);
241 			return (ENOBUFS);
242 		}
243 		m->m_flags = m0->m_flags & M_COPYFLAGS;
244 		*mnext = m;
245 		mnext = &m->m_nextpkt;
246 		m->m_data += max_linkhdr;
247 		mhip6 = mtod(m, struct ip6_hdr *);
248 		*mhip6 = *ip6;
249 		m->m_len = sizeof(*mhip6);
250 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
251 		if (error) {
252 			IP6STAT_INC(ip6s_odropped);
253 			return (error);
254 		}
255 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
256 		if (off + mtu >= tlen)
257 			mtu = tlen - off;
258 		else
259 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
260 		mhip6->ip6_plen = htons((u_short)(mtu + hlen +
261 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
262 		if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) {
263 			IP6STAT_INC(ip6s_odropped);
264 			return (ENOBUFS);
265 		}
266 		m_cat(m, m_frgpart);
267 		m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
268 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
269 		m->m_pkthdr.rcvif = NULL;
270 		ip6f->ip6f_reserved = 0;
271 		ip6f->ip6f_ident = id;
272 		ip6f->ip6f_nxt = nextproto;
273 		IP6STAT_INC(ip6s_ofragments);
274 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
275 	}
276 
277 	return (0);
278 }
279 
280 /*
281  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
282  * header (with pri, len, nxt, hlim, src, dst).
283  * This function may modify ver and hlim only.
284  * The mbuf chain containing the packet will be freed.
285  * The mbuf opt, if present, will not be freed.
286  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
287  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
288  * then result of route lookup is stored in ro->ro_rt.
289  *
290  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
291  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
292  * which is rt_mtu.
293  *
294  * ifpp - XXX: just for statistics
295  */
296 /*
297  * XXX TODO: no flowid is assigned for outbound flows?
298  */
299 int
300 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
301     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
302     struct ifnet **ifpp, struct inpcb *inp)
303 {
304 	struct ip6_hdr *ip6;
305 	struct ifnet *ifp, *origifp;
306 	struct mbuf *m = m0;
307 	struct mbuf *mprev = NULL;
308 	int hlen, tlen, len;
309 	struct route_in6 ip6route;
310 	struct rtentry *rt = NULL;
311 	struct sockaddr_in6 *dst, src_sa, dst_sa;
312 	struct in6_addr odst;
313 	int error = 0;
314 	struct in6_ifaddr *ia = NULL;
315 	u_long mtu;
316 	int alwaysfrag, dontfrag;
317 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
318 	struct ip6_exthdrs exthdrs;
319 	struct in6_addr src0, dst0;
320 	u_int32_t zone;
321 	struct route_in6 *ro_pmtu = NULL;
322 	int hdrsplit = 0;
323 	int sw_csum, tso;
324 	int needfiblookup;
325 	uint32_t fibnum;
326 	struct m_tag *fwd_tag = NULL;
327 	uint32_t id;
328 
329 	if (inp != NULL) {
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 	}
337 
338 	bzero(&exthdrs, sizeof(exthdrs));
339 	if (opt) {
340 		/* Hop-by-Hop options header */
341 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
342 		/* Destination options header(1st part) */
343 		if (opt->ip6po_rthdr) {
344 			/*
345 			 * Destination options header(1st part)
346 			 * This only makes sense with a routing header.
347 			 * See Section 9.2 of RFC 3542.
348 			 * Disabling this part just for MIP6 convenience is
349 			 * a bad idea.  We need to think carefully about a
350 			 * way to make the advanced API coexist with MIP6
351 			 * options, which might automatically be inserted in
352 			 * the kernel.
353 			 */
354 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
355 		}
356 		/* Routing header */
357 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
358 		/* Destination options header(2nd part) */
359 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
360 	}
361 
362 #ifdef IPSEC
363 	/*
364 	 * IPSec checking which handles several cases.
365 	 * FAST IPSEC: We re-injected the packet.
366 	 * XXX: need scope argument.
367 	 */
368 	switch(ip6_ipsec_output(&m, inp, &error))
369 	{
370 	case 1:                 /* Bad packet */
371 		goto freehdrs;
372 	case -1:                /* IPSec done */
373 		goto done;
374 	case 0:                 /* No IPSec */
375 	default:
376 		break;
377 	}
378 #endif /* IPSEC */
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 	} else
505 		ro->ro_flags |= RT_LLE_CACHE;
506 	ro_pmtu = ro;
507 	if (opt && opt->ip6po_rthdr)
508 		ro = &opt->ip6po_route;
509 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
510 #ifdef FLOWTABLE
511 	if (ro->ro_rt == NULL)
512 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
513 #endif
514 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
515 again:
516 	/*
517 	 * if specified, try to fill in the traffic class field.
518 	 * do not override if a non-zero value is already set.
519 	 * we check the diffserv field and the ecn field separately.
520 	 */
521 	if (opt && opt->ip6po_tclass >= 0) {
522 		int mask = 0;
523 
524 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
525 			mask |= 0xfc;
526 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
527 			mask |= 0x03;
528 		if (mask != 0)
529 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
530 	}
531 
532 	/* fill in or override the hop limit field, if necessary. */
533 	if (opt && opt->ip6po_hlim != -1)
534 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
535 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
536 		if (im6o != NULL)
537 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
538 		else
539 			ip6->ip6_hlim = V_ip6_defmcasthlim;
540 	}
541 	/*
542 	 * Validate route against routing table additions;
543 	 * a better/more specific route might have been added.
544 	 * Make sure address family is set in route.
545 	 */
546 	if (inp) {
547 		ro->ro_dst.sin6_family = AF_INET6;
548 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
549 	}
550 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
551 	    ro->ro_dst.sin6_family == AF_INET6 &&
552 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
553 		rt = ro->ro_rt;
554 		ifp = ro->ro_rt->rt_ifp;
555 	} else {
556 		if (ro->ro_lle)
557 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
558 		ro->ro_lle = NULL;
559 		if (fwd_tag == NULL) {
560 			bzero(&dst_sa, sizeof(dst_sa));
561 			dst_sa.sin6_family = AF_INET6;
562 			dst_sa.sin6_len = sizeof(dst_sa);
563 			dst_sa.sin6_addr = ip6->ip6_dst;
564 		}
565 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
566 		    &rt, fibnum);
567 		if (error != 0) {
568 			if (ifp != NULL)
569 				in6_ifstat_inc(ifp, ifs6_out_discard);
570 			goto bad;
571 		}
572 	}
573 	if (rt == NULL) {
574 		/*
575 		 * If in6_selectroute() does not return a route entry,
576 		 * dst may not have been updated.
577 		 */
578 		*dst = dst_sa;	/* XXX */
579 	}
580 
581 	/*
582 	 * then rt (for unicast) and ifp must be non-NULL valid values.
583 	 */
584 	if ((flags & IPV6_FORWARDING) == 0) {
585 		/* XXX: the FORWARDING flag can be set for mrouting. */
586 		in6_ifstat_inc(ifp, ifs6_out_request);
587 	}
588 	if (rt != NULL) {
589 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
590 		counter_u64_add(rt->rt_pksent, 1);
591 	}
592 
593 
594 	/*
595 	 * The outgoing interface must be in the zone of source and
596 	 * destination addresses.
597 	 */
598 	origifp = ifp;
599 
600 	src0 = ip6->ip6_src;
601 	if (in6_setscope(&src0, origifp, &zone))
602 		goto badscope;
603 	bzero(&src_sa, sizeof(src_sa));
604 	src_sa.sin6_family = AF_INET6;
605 	src_sa.sin6_len = sizeof(src_sa);
606 	src_sa.sin6_addr = ip6->ip6_src;
607 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
608 		goto badscope;
609 
610 	dst0 = ip6->ip6_dst;
611 	if (in6_setscope(&dst0, origifp, &zone))
612 		goto badscope;
613 	/* re-initialize to be sure */
614 	bzero(&dst_sa, sizeof(dst_sa));
615 	dst_sa.sin6_family = AF_INET6;
616 	dst_sa.sin6_len = sizeof(dst_sa);
617 	dst_sa.sin6_addr = ip6->ip6_dst;
618 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
619 		goto badscope;
620 	}
621 
622 	/* We should use ia_ifp to support the case of
623 	 * sending packets to an address of our own.
624 	 */
625 	if (ia != NULL && ia->ia_ifp)
626 		ifp = ia->ia_ifp;
627 
628 	/* scope check is done. */
629 	goto routefound;
630 
631   badscope:
632 	IP6STAT_INC(ip6s_badscope);
633 	in6_ifstat_inc(origifp, ifs6_out_discard);
634 	if (error == 0)
635 		error = EHOSTUNREACH; /* XXX */
636 	goto bad;
637 
638   routefound:
639 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
640 		if (opt && opt->ip6po_nextroute.ro_rt) {
641 			/*
642 			 * The nexthop is explicitly specified by the
643 			 * application.  We assume the next hop is an IPv6
644 			 * address.
645 			 */
646 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
647 		}
648 		else if ((rt->rt_flags & RTF_GATEWAY))
649 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
650 	}
651 
652 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
653 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
654 	} else {
655 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
656 		in6_ifstat_inc(ifp, ifs6_out_mcast);
657 		/*
658 		 * Confirm that the outgoing interface supports multicast.
659 		 */
660 		if (!(ifp->if_flags & IFF_MULTICAST)) {
661 			IP6STAT_INC(ip6s_noroute);
662 			in6_ifstat_inc(ifp, ifs6_out_discard);
663 			error = ENETUNREACH;
664 			goto bad;
665 		}
666 		if ((im6o == NULL && in6_mcast_loop) ||
667 		    (im6o && im6o->im6o_multicast_loop)) {
668 			/*
669 			 * Loop back multicast datagram if not expressly
670 			 * forbidden to do so, even if we have not joined
671 			 * the address; protocols will filter it later,
672 			 * thus deferring a hash lookup and lock acquisition
673 			 * at the expense of an m_copym().
674 			 */
675 			ip6_mloopback(ifp, m);
676 		} else {
677 			/*
678 			 * If we are acting as a multicast router, perform
679 			 * multicast forwarding as if the packet had just
680 			 * arrived on the interface to which we are about
681 			 * to send.  The multicast forwarding function
682 			 * recursively calls this function, using the
683 			 * IPV6_FORWARDING flag to prevent infinite recursion.
684 			 *
685 			 * Multicasts that are looped back by ip6_mloopback(),
686 			 * above, will be forwarded by the ip6_input() routine,
687 			 * if necessary.
688 			 */
689 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
690 				/*
691 				 * XXX: ip6_mforward expects that rcvif is NULL
692 				 * when it is called from the originating path.
693 				 * However, it may not always be the case.
694 				 */
695 				m->m_pkthdr.rcvif = NULL;
696 				if (ip6_mforward(ip6, ifp, m) != 0) {
697 					m_freem(m);
698 					goto done;
699 				}
700 			}
701 		}
702 		/*
703 		 * Multicasts with a hoplimit of zero may be looped back,
704 		 * above, but must not be transmitted on a network.
705 		 * Also, multicasts addressed to the loopback interface
706 		 * are not sent -- the above call to ip6_mloopback() will
707 		 * loop back a copy if this host actually belongs to the
708 		 * destination group on the loopback interface.
709 		 */
710 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
711 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
712 			m_freem(m);
713 			goto done;
714 		}
715 	}
716 
717 	/*
718 	 * Fill the outgoing inteface to tell the upper layer
719 	 * to increment per-interface statistics.
720 	 */
721 	if (ifpp)
722 		*ifpp = ifp;
723 
724 	/* Determine path MTU. */
725 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
726 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
727 		goto bad;
728 
729 	/*
730 	 * The caller of this function may specify to use the minimum MTU
731 	 * in some cases.
732 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
733 	 * setting.  The logic is a bit complicated; by default, unicast
734 	 * packets will follow path MTU while multicast packets will be sent at
735 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
736 	 * including unicast ones will be sent at the minimum MTU.  Multicast
737 	 * packets will always be sent at the minimum MTU unless
738 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
739 	 * See RFC 3542 for more details.
740 	 */
741 	if (mtu > IPV6_MMTU) {
742 		if ((flags & IPV6_MINMTU))
743 			mtu = IPV6_MMTU;
744 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
745 			mtu = IPV6_MMTU;
746 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
747 			 (opt == NULL ||
748 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
749 			mtu = IPV6_MMTU;
750 		}
751 	}
752 
753 	/*
754 	 * clear embedded scope identifiers if necessary.
755 	 * in6_clearscope will touch the addresses only when necessary.
756 	 */
757 	in6_clearscope(&ip6->ip6_src);
758 	in6_clearscope(&ip6->ip6_dst);
759 
760 	/*
761 	 * If the outgoing packet contains a hop-by-hop options header,
762 	 * it must be examined and processed even by the source node.
763 	 * (RFC 2460, section 4.)
764 	 */
765 	if (exthdrs.ip6e_hbh) {
766 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
767 		u_int32_t dummy; /* XXX unused */
768 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
769 
770 #ifdef DIAGNOSTIC
771 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
772 			panic("ip6e_hbh is not contiguous");
773 #endif
774 		/*
775 		 *  XXX: if we have to send an ICMPv6 error to the sender,
776 		 *       we need the M_LOOP flag since icmp6_error() expects
777 		 *       the IPv6 and the hop-by-hop options header are
778 		 *       contiguous unless the flag is set.
779 		 */
780 		m->m_flags |= M_LOOP;
781 		m->m_pkthdr.rcvif = ifp;
782 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
783 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
784 		    &dummy, &plen) < 0) {
785 			/* m was already freed at this point */
786 			error = EINVAL;/* better error? */
787 			goto done;
788 		}
789 		m->m_flags &= ~M_LOOP; /* XXX */
790 		m->m_pkthdr.rcvif = NULL;
791 	}
792 
793 	/* Jump over all PFIL processing if hooks are not active. */
794 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
795 		goto passout;
796 
797 	odst = ip6->ip6_dst;
798 	/* Run through list of hooks for output packets. */
799 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
800 	if (error != 0 || m == NULL)
801 		goto done;
802 	/* adjust pointer */
803 	ip6 = mtod(m, struct ip6_hdr *);
804 
805 	needfiblookup = 0;
806 	/* See if destination IP address was changed by packet filter. */
807 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
808 		m->m_flags |= M_SKIP_FIREWALL;
809 		/* If destination is now ourself drop to ip6_input(). */
810 		if (in6_localip(&ip6->ip6_dst)) {
811 			m->m_flags |= M_FASTFWD_OURS;
812 			if (m->m_pkthdr.rcvif == NULL)
813 				m->m_pkthdr.rcvif = V_loif;
814 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
815 				m->m_pkthdr.csum_flags |=
816 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
817 				m->m_pkthdr.csum_data = 0xffff;
818 			}
819 #ifdef SCTP
820 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
821 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
822 #endif
823 			error = netisr_queue(NETISR_IPV6, m);
824 			goto done;
825 		} else {
826 			RO_RTFREE(ro);
827 			needfiblookup = 1; /* Redo the routing table lookup. */
828 			if (ro->ro_lle)
829 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
830 			ro->ro_lle = NULL;
831 		}
832 	}
833 	/* See if fib was changed by packet filter. */
834 	if (fibnum != M_GETFIB(m)) {
835 		m->m_flags |= M_SKIP_FIREWALL;
836 		fibnum = M_GETFIB(m);
837 		RO_RTFREE(ro);
838 		needfiblookup = 1;
839 		if (ro->ro_lle)
840 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
841 		ro->ro_lle = NULL;
842 	}
843 	if (needfiblookup)
844 		goto again;
845 
846 	/* See if local, if yes, send it to netisr. */
847 	if (m->m_flags & M_FASTFWD_OURS) {
848 		if (m->m_pkthdr.rcvif == NULL)
849 			m->m_pkthdr.rcvif = V_loif;
850 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
851 			m->m_pkthdr.csum_flags |=
852 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
853 			m->m_pkthdr.csum_data = 0xffff;
854 		}
855 #ifdef SCTP
856 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
857 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
858 #endif
859 		error = netisr_queue(NETISR_IPV6, m);
860 		goto done;
861 	}
862 	/* Or forward to some other address? */
863 	if ((m->m_flags & M_IP6_NEXTHOP) &&
864 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
865 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
866 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
867 		m->m_flags |= M_SKIP_FIREWALL;
868 		m->m_flags &= ~M_IP6_NEXTHOP;
869 		m_tag_delete(m, fwd_tag);
870 		goto again;
871 	}
872 
873 passout:
874 	/*
875 	 * Send the packet to the outgoing interface.
876 	 * If necessary, do IPv6 fragmentation before sending.
877 	 *
878 	 * the logic here is rather complex:
879 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
880 	 * 1-a:	send as is if tlen <= path mtu
881 	 * 1-b:	fragment if tlen > path mtu
882 	 *
883 	 * 2: if user asks us not to fragment (dontfrag == 1)
884 	 * 2-a:	send as is if tlen <= interface mtu
885 	 * 2-b:	error if tlen > interface mtu
886 	 *
887 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
888 	 *	always fragment
889 	 *
890 	 * 4: if dontfrag == 1 && alwaysfrag == 1
891 	 *	error, as we cannot handle this conflicting request
892 	 */
893 	sw_csum = m->m_pkthdr.csum_flags;
894 	if (!hdrsplit) {
895 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
896 		sw_csum &= ~ifp->if_hwassist;
897 	} else
898 		tso = 0;
899 	/*
900 	 * If we added extension headers, we will not do TSO and calculate the
901 	 * checksums ourselves for now.
902 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
903 	 * with ext. hdrs.
904 	 */
905 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
906 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
907 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
908 	}
909 #ifdef SCTP
910 	if (sw_csum & CSUM_SCTP_IPV6) {
911 		sw_csum &= ~CSUM_SCTP_IPV6;
912 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
913 	}
914 #endif
915 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
916 	tlen = m->m_pkthdr.len;
917 
918 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
919 		dontfrag = 1;
920 	else
921 		dontfrag = 0;
922 	if (dontfrag && alwaysfrag) {	/* case 4 */
923 		/* conflicting request - can't transmit */
924 		error = EMSGSIZE;
925 		goto bad;
926 	}
927 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
928 		/*
929 		 * Even if the DONTFRAG option is specified, we cannot send the
930 		 * packet when the data length is larger than the MTU of the
931 		 * outgoing interface.
932 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
933 		 * application wanted to know the MTU value. Also return an
934 		 * error code (this is not described in the API spec).
935 		 */
936 		if (inp != NULL)
937 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
938 		error = EMSGSIZE;
939 		goto bad;
940 	}
941 
942 	/*
943 	 * transmit packet without fragmentation
944 	 */
945 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
946 		struct in6_ifaddr *ia6;
947 
948 		ip6 = mtod(m, struct ip6_hdr *);
949 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
950 		if (ia6) {
951 			/* Record statistics for this interface address. */
952 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
953 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
954 			    m->m_pkthdr.len);
955 			ifa_free(&ia6->ia_ifa);
956 		}
957 		error = nd6_output_ifp(ifp, origifp, m, dst,
958 		    (struct route *)ro);
959 		goto done;
960 	}
961 
962 	/*
963 	 * try to fragment the packet.  case 1-b and 3
964 	 */
965 	if (mtu < IPV6_MMTU) {
966 		/* path MTU cannot be less than IPV6_MMTU */
967 		error = EMSGSIZE;
968 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
969 		goto bad;
970 	} else if (ip6->ip6_plen == 0) {
971 		/* jumbo payload cannot be fragmented */
972 		error = EMSGSIZE;
973 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
974 		goto bad;
975 	} else {
976 		u_char nextproto;
977 
978 		/*
979 		 * Too large for the destination or interface;
980 		 * fragment if possible.
981 		 * Must be able to put at least 8 bytes per fragment.
982 		 */
983 		hlen = unfragpartlen;
984 		if (mtu > IPV6_MAXPACKET)
985 			mtu = IPV6_MAXPACKET;
986 
987 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
988 		if (len < 8) {
989 			error = EMSGSIZE;
990 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
991 			goto bad;
992 		}
993 
994 		/*
995 		 * If the interface will not calculate checksums on
996 		 * fragmented packets, then do it here.
997 		 * XXX-BZ handle the hw offloading case.  Need flags.
998 		 */
999 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1000 			in6_delayed_cksum(m, plen, hlen);
1001 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1002 		}
1003 #ifdef SCTP
1004 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1005 			sctp_delayed_cksum(m, hlen);
1006 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1007 		}
1008 #endif
1009 		/*
1010 		 * Change the next header field of the last header in the
1011 		 * unfragmentable part.
1012 		 */
1013 		if (exthdrs.ip6e_rthdr) {
1014 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1015 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1016 		} else if (exthdrs.ip6e_dest1) {
1017 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1018 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1019 		} else if (exthdrs.ip6e_hbh) {
1020 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1021 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1022 		} else {
1023 			nextproto = ip6->ip6_nxt;
1024 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1025 		}
1026 
1027 		/*
1028 		 * Loop through length of segment after first fragment,
1029 		 * make new header and copy data of each part and link onto
1030 		 * chain.
1031 		 */
1032 		m0 = m;
1033 		id = htonl(ip6_randomid());
1034 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1035 			goto sendorfree;
1036 
1037 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1038 	}
1039 
1040 	/*
1041 	 * Remove leading garbages.
1042 	 */
1043 sendorfree:
1044 	m = m0->m_nextpkt;
1045 	m0->m_nextpkt = 0;
1046 	m_freem(m0);
1047 	for (m0 = m; m; m = m0) {
1048 		m0 = m->m_nextpkt;
1049 		m->m_nextpkt = 0;
1050 		if (error == 0) {
1051 			/* Record statistics for this interface address. */
1052 			if (ia) {
1053 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1054 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1055 				    m->m_pkthdr.len);
1056 			}
1057 			error = nd6_output_ifp(ifp, origifp, m, dst,
1058 			    (struct route *)ro);
1059 		} else
1060 			m_freem(m);
1061 	}
1062 
1063 	if (error == 0)
1064 		IP6STAT_INC(ip6s_fragmented);
1065 
1066 done:
1067 	if (ro == &ip6route)
1068 		RO_RTFREE(ro);
1069 	return (error);
1070 
1071 freehdrs:
1072 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1073 	m_freem(exthdrs.ip6e_dest1);
1074 	m_freem(exthdrs.ip6e_rthdr);
1075 	m_freem(exthdrs.ip6e_dest2);
1076 	/* FALLTHROUGH */
1077 bad:
1078 	if (m)
1079 		m_freem(m);
1080 	goto done;
1081 }
1082 
1083 static int
1084 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1085 {
1086 	struct mbuf *m;
1087 
1088 	if (hlen > MCLBYTES)
1089 		return (ENOBUFS); /* XXX */
1090 
1091 	if (hlen > MLEN)
1092 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1093 	else
1094 		m = m_get(M_NOWAIT, MT_DATA);
1095 	if (m == NULL)
1096 		return (ENOBUFS);
1097 	m->m_len = hlen;
1098 	if (hdr)
1099 		bcopy(hdr, mtod(m, caddr_t), hlen);
1100 
1101 	*mp = m;
1102 	return (0);
1103 }
1104 
1105 /*
1106  * Insert jumbo payload option.
1107  */
1108 static int
1109 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1110 {
1111 	struct mbuf *mopt;
1112 	u_char *optbuf;
1113 	u_int32_t v;
1114 
1115 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1116 
1117 	/*
1118 	 * If there is no hop-by-hop options header, allocate new one.
1119 	 * If there is one but it doesn't have enough space to store the
1120 	 * jumbo payload option, allocate a cluster to store the whole options.
1121 	 * Otherwise, use it to store the options.
1122 	 */
1123 	if (exthdrs->ip6e_hbh == NULL) {
1124 		mopt = m_get(M_NOWAIT, MT_DATA);
1125 		if (mopt == NULL)
1126 			return (ENOBUFS);
1127 		mopt->m_len = JUMBOOPTLEN;
1128 		optbuf = mtod(mopt, u_char *);
1129 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1130 		exthdrs->ip6e_hbh = mopt;
1131 	} else {
1132 		struct ip6_hbh *hbh;
1133 
1134 		mopt = exthdrs->ip6e_hbh;
1135 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1136 			/*
1137 			 * XXX assumption:
1138 			 * - exthdrs->ip6e_hbh is not referenced from places
1139 			 *   other than exthdrs.
1140 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1141 			 */
1142 			int oldoptlen = mopt->m_len;
1143 			struct mbuf *n;
1144 
1145 			/*
1146 			 * XXX: give up if the whole (new) hbh header does
1147 			 * not fit even in an mbuf cluster.
1148 			 */
1149 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1150 				return (ENOBUFS);
1151 
1152 			/*
1153 			 * As a consequence, we must always prepare a cluster
1154 			 * at this point.
1155 			 */
1156 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1157 			if (n == NULL)
1158 				return (ENOBUFS);
1159 			n->m_len = oldoptlen + JUMBOOPTLEN;
1160 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1161 			    oldoptlen);
1162 			optbuf = mtod(n, caddr_t) + oldoptlen;
1163 			m_freem(mopt);
1164 			mopt = exthdrs->ip6e_hbh = n;
1165 		} else {
1166 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1167 			mopt->m_len += JUMBOOPTLEN;
1168 		}
1169 		optbuf[0] = IP6OPT_PADN;
1170 		optbuf[1] = 1;
1171 
1172 		/*
1173 		 * Adjust the header length according to the pad and
1174 		 * the jumbo payload option.
1175 		 */
1176 		hbh = mtod(mopt, struct ip6_hbh *);
1177 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1178 	}
1179 
1180 	/* fill in the option. */
1181 	optbuf[2] = IP6OPT_JUMBO;
1182 	optbuf[3] = 4;
1183 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1184 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1185 
1186 	/* finally, adjust the packet header length */
1187 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1188 
1189 	return (0);
1190 #undef JUMBOOPTLEN
1191 }
1192 
1193 /*
1194  * Insert fragment header and copy unfragmentable header portions.
1195  */
1196 static int
1197 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1198     struct ip6_frag **frghdrp)
1199 {
1200 	struct mbuf *n, *mlast;
1201 
1202 	if (hlen > sizeof(struct ip6_hdr)) {
1203 		n = m_copym(m0, sizeof(struct ip6_hdr),
1204 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1205 		if (n == NULL)
1206 			return (ENOBUFS);
1207 		m->m_next = n;
1208 	} else
1209 		n = m;
1210 
1211 	/* Search for the last mbuf of unfragmentable part. */
1212 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1213 		;
1214 
1215 	if (M_WRITABLE(mlast) &&
1216 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1217 		/* use the trailing space of the last mbuf for the fragment hdr */
1218 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1219 		    mlast->m_len);
1220 		mlast->m_len += sizeof(struct ip6_frag);
1221 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1222 	} else {
1223 		/* allocate a new mbuf for the fragment header */
1224 		struct mbuf *mfrg;
1225 
1226 		mfrg = m_get(M_NOWAIT, MT_DATA);
1227 		if (mfrg == NULL)
1228 			return (ENOBUFS);
1229 		mfrg->m_len = sizeof(struct ip6_frag);
1230 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1231 		mlast->m_next = mfrg;
1232 	}
1233 
1234 	return (0);
1235 }
1236 
1237 /*
1238  * Calculates IPv6 path mtu for destination @dst.
1239  * Resulting MTU is stored in @mtup.
1240  *
1241  * Returns 0 on success.
1242  */
1243 static int
1244 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1245 {
1246 	struct nhop6_extended nh6;
1247 	struct in6_addr kdst;
1248 	uint32_t scopeid;
1249 	struct ifnet *ifp;
1250 	u_long mtu;
1251 	int error;
1252 
1253 	in6_splitscope(dst, &kdst, &scopeid);
1254 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1255 		return (EHOSTUNREACH);
1256 
1257 	ifp = nh6.nh_ifp;
1258 	mtu = nh6.nh_mtu;
1259 
1260 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1261 	fib6_free_nh_ext(fibnum, &nh6);
1262 
1263 	return (error);
1264 }
1265 
1266 /*
1267  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1268  * and cached data in @ro_pmtu.
1269  * MTU from (successful) route lookup is saved (along with dst)
1270  * inside @ro_pmtu to avoid subsequent route lookups after packet
1271  * filter processing.
1272  *
1273  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1274  * Returns 0 on success.
1275  */
1276 static int
1277 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1278     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1279     int *alwaysfragp, u_int fibnum, u_int proto)
1280 {
1281 	struct nhop6_basic nh6;
1282 	struct in6_addr kdst;
1283 	uint32_t scopeid;
1284 	struct sockaddr_in6 *sa6_dst;
1285 	u_long mtu;
1286 
1287 	mtu = 0;
1288 	if (do_lookup) {
1289 
1290 		/*
1291 		 * Here ro_pmtu has final destination address, while
1292 		 * ro might represent immediate destination.
1293 		 * Use ro_pmtu destination since mtu might differ.
1294 		 */
1295 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1296 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1297 			ro_pmtu->ro_mtu = 0;
1298 
1299 		if (ro_pmtu->ro_mtu == 0) {
1300 			bzero(sa6_dst, sizeof(*sa6_dst));
1301 			sa6_dst->sin6_family = AF_INET6;
1302 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1303 			sa6_dst->sin6_addr = *dst;
1304 
1305 			in6_splitscope(dst, &kdst, &scopeid);
1306 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1307 			    &nh6) == 0)
1308 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1309 		}
1310 
1311 		mtu = ro_pmtu->ro_mtu;
1312 	}
1313 
1314 	if (ro_pmtu->ro_rt)
1315 		mtu = ro_pmtu->ro_rt->rt_mtu;
1316 
1317 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1318 }
1319 
1320 /*
1321  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1322  * hostcache data for @dst.
1323  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1324  *
1325  * Returns 0 on success.
1326  */
1327 static int
1328 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1329     u_long *mtup, int *alwaysfragp, u_int proto)
1330 {
1331 	u_long mtu = 0;
1332 	int alwaysfrag = 0;
1333 	int error = 0;
1334 
1335 	if (rt_mtu > 0) {
1336 		u_int32_t ifmtu;
1337 		struct in_conninfo inc;
1338 
1339 		bzero(&inc, sizeof(inc));
1340 		inc.inc_flags |= INC_ISIPV6;
1341 		inc.inc6_faddr = *dst;
1342 
1343 		ifmtu = IN6_LINKMTU(ifp);
1344 
1345 		/* TCP is known to react to pmtu changes so skip hc */
1346 		if (proto != IPPROTO_TCP)
1347 			mtu = tcp_hc_getmtu(&inc);
1348 
1349 		if (mtu)
1350 			mtu = min(mtu, rt_mtu);
1351 		else
1352 			mtu = rt_mtu;
1353 		if (mtu == 0)
1354 			mtu = ifmtu;
1355 		else if (mtu < IPV6_MMTU) {
1356 			/*
1357 			 * RFC2460 section 5, last paragraph:
1358 			 * if we record ICMPv6 too big message with
1359 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1360 			 * or smaller, with framgent header attached.
1361 			 * (fragment header is needed regardless from the
1362 			 * packet size, for translators to identify packets)
1363 			 */
1364 			alwaysfrag = 1;
1365 			mtu = IPV6_MMTU;
1366 		}
1367 	} else if (ifp) {
1368 		mtu = IN6_LINKMTU(ifp);
1369 	} else
1370 		error = EHOSTUNREACH; /* XXX */
1371 
1372 	*mtup = mtu;
1373 	if (alwaysfragp)
1374 		*alwaysfragp = alwaysfrag;
1375 	return (error);
1376 }
1377 
1378 /*
1379  * IP6 socket option processing.
1380  */
1381 int
1382 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1383 {
1384 	int optdatalen, uproto;
1385 	void *optdata;
1386 	struct inpcb *in6p = sotoinpcb(so);
1387 	int error, optval;
1388 	int level, op, optname;
1389 	int optlen;
1390 	struct thread *td;
1391 #ifdef	RSS
1392 	uint32_t rss_bucket;
1393 	int retval;
1394 #endif
1395 
1396 	level = sopt->sopt_level;
1397 	op = sopt->sopt_dir;
1398 	optname = sopt->sopt_name;
1399 	optlen = sopt->sopt_valsize;
1400 	td = sopt->sopt_td;
1401 	error = 0;
1402 	optval = 0;
1403 	uproto = (int)so->so_proto->pr_protocol;
1404 
1405 	if (level != IPPROTO_IPV6) {
1406 		error = EINVAL;
1407 
1408 		if (sopt->sopt_level == SOL_SOCKET &&
1409 		    sopt->sopt_dir == SOPT_SET) {
1410 			switch (sopt->sopt_name) {
1411 			case SO_REUSEADDR:
1412 				INP_WLOCK(in6p);
1413 				if ((so->so_options & SO_REUSEADDR) != 0)
1414 					in6p->inp_flags2 |= INP_REUSEADDR;
1415 				else
1416 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1417 				INP_WUNLOCK(in6p);
1418 				error = 0;
1419 				break;
1420 			case SO_REUSEPORT:
1421 				INP_WLOCK(in6p);
1422 				if ((so->so_options & SO_REUSEPORT) != 0)
1423 					in6p->inp_flags2 |= INP_REUSEPORT;
1424 				else
1425 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1426 				INP_WUNLOCK(in6p);
1427 				error = 0;
1428 				break;
1429 			case SO_SETFIB:
1430 				INP_WLOCK(in6p);
1431 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1432 				INP_WUNLOCK(in6p);
1433 				error = 0;
1434 				break;
1435 			default:
1436 				break;
1437 			}
1438 		}
1439 	} else {		/* level == IPPROTO_IPV6 */
1440 		switch (op) {
1441 
1442 		case SOPT_SET:
1443 			switch (optname) {
1444 			case IPV6_2292PKTOPTIONS:
1445 #ifdef IPV6_PKTOPTIONS
1446 			case IPV6_PKTOPTIONS:
1447 #endif
1448 			{
1449 				struct mbuf *m;
1450 
1451 				error = soopt_getm(sopt, &m); /* XXX */
1452 				if (error != 0)
1453 					break;
1454 				error = soopt_mcopyin(sopt, m); /* XXX */
1455 				if (error != 0)
1456 					break;
1457 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1458 						    m, so, sopt);
1459 				m_freem(m); /* XXX */
1460 				break;
1461 			}
1462 
1463 			/*
1464 			 * Use of some Hop-by-Hop options or some
1465 			 * Destination options, might require special
1466 			 * privilege.  That is, normal applications
1467 			 * (without special privilege) might be forbidden
1468 			 * from setting certain options in outgoing packets,
1469 			 * and might never see certain options in received
1470 			 * packets. [RFC 2292 Section 6]
1471 			 * KAME specific note:
1472 			 *  KAME prevents non-privileged users from sending or
1473 			 *  receiving ANY hbh/dst options in order to avoid
1474 			 *  overhead of parsing options in the kernel.
1475 			 */
1476 			case IPV6_RECVHOPOPTS:
1477 			case IPV6_RECVDSTOPTS:
1478 			case IPV6_RECVRTHDRDSTOPTS:
1479 				if (td != NULL) {
1480 					error = priv_check(td,
1481 					    PRIV_NETINET_SETHDROPTS);
1482 					if (error)
1483 						break;
1484 				}
1485 				/* FALLTHROUGH */
1486 			case IPV6_UNICAST_HOPS:
1487 			case IPV6_HOPLIMIT:
1488 
1489 			case IPV6_RECVPKTINFO:
1490 			case IPV6_RECVHOPLIMIT:
1491 			case IPV6_RECVRTHDR:
1492 			case IPV6_RECVPATHMTU:
1493 			case IPV6_RECVTCLASS:
1494 			case IPV6_RECVFLOWID:
1495 #ifdef	RSS
1496 			case IPV6_RECVRSSBUCKETID:
1497 #endif
1498 			case IPV6_V6ONLY:
1499 			case IPV6_AUTOFLOWLABEL:
1500 			case IPV6_BINDANY:
1501 			case IPV6_BINDMULTI:
1502 #ifdef	RSS
1503 			case IPV6_RSS_LISTEN_BUCKET:
1504 #endif
1505 				if (optname == IPV6_BINDANY && td != NULL) {
1506 					error = priv_check(td,
1507 					    PRIV_NETINET_BINDANY);
1508 					if (error)
1509 						break;
1510 				}
1511 
1512 				if (optlen != sizeof(int)) {
1513 					error = EINVAL;
1514 					break;
1515 				}
1516 				error = sooptcopyin(sopt, &optval,
1517 					sizeof optval, sizeof optval);
1518 				if (error)
1519 					break;
1520 				switch (optname) {
1521 
1522 				case IPV6_UNICAST_HOPS:
1523 					if (optval < -1 || optval >= 256)
1524 						error = EINVAL;
1525 					else {
1526 						/* -1 = kernel default */
1527 						in6p->in6p_hops = optval;
1528 						if ((in6p->inp_vflag &
1529 						     INP_IPV4) != 0)
1530 							in6p->inp_ip_ttl = optval;
1531 					}
1532 					break;
1533 #define OPTSET(bit) \
1534 do { \
1535 	INP_WLOCK(in6p); \
1536 	if (optval) \
1537 		in6p->inp_flags |= (bit); \
1538 	else \
1539 		in6p->inp_flags &= ~(bit); \
1540 	INP_WUNLOCK(in6p); \
1541 } while (/*CONSTCOND*/ 0)
1542 #define OPTSET2292(bit) \
1543 do { \
1544 	INP_WLOCK(in6p); \
1545 	in6p->inp_flags |= IN6P_RFC2292; \
1546 	if (optval) \
1547 		in6p->inp_flags |= (bit); \
1548 	else \
1549 		in6p->inp_flags &= ~(bit); \
1550 	INP_WUNLOCK(in6p); \
1551 } while (/*CONSTCOND*/ 0)
1552 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1553 
1554 #define OPTSET2(bit, val) do {						\
1555 	INP_WLOCK(in6p);						\
1556 	if (val)							\
1557 		in6p->inp_flags2 |= bit;				\
1558 	else								\
1559 		in6p->inp_flags2 &= ~bit;				\
1560 	INP_WUNLOCK(in6p);						\
1561 } while (0)
1562 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1563 
1564 				case IPV6_RECVPKTINFO:
1565 					/* cannot mix with RFC2292 */
1566 					if (OPTBIT(IN6P_RFC2292)) {
1567 						error = EINVAL;
1568 						break;
1569 					}
1570 					OPTSET(IN6P_PKTINFO);
1571 					break;
1572 
1573 				case IPV6_HOPLIMIT:
1574 				{
1575 					struct ip6_pktopts **optp;
1576 
1577 					/* cannot mix with RFC2292 */
1578 					if (OPTBIT(IN6P_RFC2292)) {
1579 						error = EINVAL;
1580 						break;
1581 					}
1582 					optp = &in6p->in6p_outputopts;
1583 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1584 					    (u_char *)&optval, sizeof(optval),
1585 					    optp, (td != NULL) ? td->td_ucred :
1586 					    NULL, uproto);
1587 					break;
1588 				}
1589 
1590 				case IPV6_RECVHOPLIMIT:
1591 					/* cannot mix with RFC2292 */
1592 					if (OPTBIT(IN6P_RFC2292)) {
1593 						error = EINVAL;
1594 						break;
1595 					}
1596 					OPTSET(IN6P_HOPLIMIT);
1597 					break;
1598 
1599 				case IPV6_RECVHOPOPTS:
1600 					/* cannot mix with RFC2292 */
1601 					if (OPTBIT(IN6P_RFC2292)) {
1602 						error = EINVAL;
1603 						break;
1604 					}
1605 					OPTSET(IN6P_HOPOPTS);
1606 					break;
1607 
1608 				case IPV6_RECVDSTOPTS:
1609 					/* cannot mix with RFC2292 */
1610 					if (OPTBIT(IN6P_RFC2292)) {
1611 						error = EINVAL;
1612 						break;
1613 					}
1614 					OPTSET(IN6P_DSTOPTS);
1615 					break;
1616 
1617 				case IPV6_RECVRTHDRDSTOPTS:
1618 					/* cannot mix with RFC2292 */
1619 					if (OPTBIT(IN6P_RFC2292)) {
1620 						error = EINVAL;
1621 						break;
1622 					}
1623 					OPTSET(IN6P_RTHDRDSTOPTS);
1624 					break;
1625 
1626 				case IPV6_RECVRTHDR:
1627 					/* cannot mix with RFC2292 */
1628 					if (OPTBIT(IN6P_RFC2292)) {
1629 						error = EINVAL;
1630 						break;
1631 					}
1632 					OPTSET(IN6P_RTHDR);
1633 					break;
1634 
1635 				case IPV6_RECVPATHMTU:
1636 					/*
1637 					 * We ignore this option for TCP
1638 					 * sockets.
1639 					 * (RFC3542 leaves this case
1640 					 * unspecified.)
1641 					 */
1642 					if (uproto != IPPROTO_TCP)
1643 						OPTSET(IN6P_MTU);
1644 					break;
1645 
1646 				case IPV6_RECVFLOWID:
1647 					OPTSET2(INP_RECVFLOWID, optval);
1648 					break;
1649 
1650 #ifdef	RSS
1651 				case IPV6_RECVRSSBUCKETID:
1652 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1653 					break;
1654 #endif
1655 
1656 				case IPV6_V6ONLY:
1657 					/*
1658 					 * make setsockopt(IPV6_V6ONLY)
1659 					 * available only prior to bind(2).
1660 					 * see ipng mailing list, Jun 22 2001.
1661 					 */
1662 					if (in6p->inp_lport ||
1663 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1664 						error = EINVAL;
1665 						break;
1666 					}
1667 					OPTSET(IN6P_IPV6_V6ONLY);
1668 					if (optval)
1669 						in6p->inp_vflag &= ~INP_IPV4;
1670 					else
1671 						in6p->inp_vflag |= INP_IPV4;
1672 					break;
1673 				case IPV6_RECVTCLASS:
1674 					/* cannot mix with RFC2292 XXX */
1675 					if (OPTBIT(IN6P_RFC2292)) {
1676 						error = EINVAL;
1677 						break;
1678 					}
1679 					OPTSET(IN6P_TCLASS);
1680 					break;
1681 				case IPV6_AUTOFLOWLABEL:
1682 					OPTSET(IN6P_AUTOFLOWLABEL);
1683 					break;
1684 
1685 				case IPV6_BINDANY:
1686 					OPTSET(INP_BINDANY);
1687 					break;
1688 
1689 				case IPV6_BINDMULTI:
1690 					OPTSET2(INP_BINDMULTI, optval);
1691 					break;
1692 #ifdef	RSS
1693 				case IPV6_RSS_LISTEN_BUCKET:
1694 					if ((optval >= 0) &&
1695 					    (optval < rss_getnumbuckets())) {
1696 						in6p->inp_rss_listen_bucket = optval;
1697 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1698 					} else {
1699 						error = EINVAL;
1700 					}
1701 					break;
1702 #endif
1703 				}
1704 				break;
1705 
1706 			case IPV6_TCLASS:
1707 			case IPV6_DONTFRAG:
1708 			case IPV6_USE_MIN_MTU:
1709 			case IPV6_PREFER_TEMPADDR:
1710 				if (optlen != sizeof(optval)) {
1711 					error = EINVAL;
1712 					break;
1713 				}
1714 				error = sooptcopyin(sopt, &optval,
1715 					sizeof optval, sizeof optval);
1716 				if (error)
1717 					break;
1718 				{
1719 					struct ip6_pktopts **optp;
1720 					optp = &in6p->in6p_outputopts;
1721 					error = ip6_pcbopt(optname,
1722 					    (u_char *)&optval, sizeof(optval),
1723 					    optp, (td != NULL) ? td->td_ucred :
1724 					    NULL, uproto);
1725 					break;
1726 				}
1727 
1728 			case IPV6_2292PKTINFO:
1729 			case IPV6_2292HOPLIMIT:
1730 			case IPV6_2292HOPOPTS:
1731 			case IPV6_2292DSTOPTS:
1732 			case IPV6_2292RTHDR:
1733 				/* RFC 2292 */
1734 				if (optlen != sizeof(int)) {
1735 					error = EINVAL;
1736 					break;
1737 				}
1738 				error = sooptcopyin(sopt, &optval,
1739 					sizeof optval, sizeof optval);
1740 				if (error)
1741 					break;
1742 				switch (optname) {
1743 				case IPV6_2292PKTINFO:
1744 					OPTSET2292(IN6P_PKTINFO);
1745 					break;
1746 				case IPV6_2292HOPLIMIT:
1747 					OPTSET2292(IN6P_HOPLIMIT);
1748 					break;
1749 				case IPV6_2292HOPOPTS:
1750 					/*
1751 					 * Check super-user privilege.
1752 					 * See comments for IPV6_RECVHOPOPTS.
1753 					 */
1754 					if (td != NULL) {
1755 						error = priv_check(td,
1756 						    PRIV_NETINET_SETHDROPTS);
1757 						if (error)
1758 							return (error);
1759 					}
1760 					OPTSET2292(IN6P_HOPOPTS);
1761 					break;
1762 				case IPV6_2292DSTOPTS:
1763 					if (td != NULL) {
1764 						error = priv_check(td,
1765 						    PRIV_NETINET_SETHDROPTS);
1766 						if (error)
1767 							return (error);
1768 					}
1769 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1770 					break;
1771 				case IPV6_2292RTHDR:
1772 					OPTSET2292(IN6P_RTHDR);
1773 					break;
1774 				}
1775 				break;
1776 			case IPV6_PKTINFO:
1777 			case IPV6_HOPOPTS:
1778 			case IPV6_RTHDR:
1779 			case IPV6_DSTOPTS:
1780 			case IPV6_RTHDRDSTOPTS:
1781 			case IPV6_NEXTHOP:
1782 			{
1783 				/* new advanced API (RFC3542) */
1784 				u_char *optbuf;
1785 				u_char optbuf_storage[MCLBYTES];
1786 				int optlen;
1787 				struct ip6_pktopts **optp;
1788 
1789 				/* cannot mix with RFC2292 */
1790 				if (OPTBIT(IN6P_RFC2292)) {
1791 					error = EINVAL;
1792 					break;
1793 				}
1794 
1795 				/*
1796 				 * We only ensure valsize is not too large
1797 				 * here.  Further validation will be done
1798 				 * later.
1799 				 */
1800 				error = sooptcopyin(sopt, optbuf_storage,
1801 				    sizeof(optbuf_storage), 0);
1802 				if (error)
1803 					break;
1804 				optlen = sopt->sopt_valsize;
1805 				optbuf = optbuf_storage;
1806 				optp = &in6p->in6p_outputopts;
1807 				error = ip6_pcbopt(optname, optbuf, optlen,
1808 				    optp, (td != NULL) ? td->td_ucred : NULL,
1809 				    uproto);
1810 				break;
1811 			}
1812 #undef OPTSET
1813 
1814 			case IPV6_MULTICAST_IF:
1815 			case IPV6_MULTICAST_HOPS:
1816 			case IPV6_MULTICAST_LOOP:
1817 			case IPV6_JOIN_GROUP:
1818 			case IPV6_LEAVE_GROUP:
1819 			case IPV6_MSFILTER:
1820 			case MCAST_BLOCK_SOURCE:
1821 			case MCAST_UNBLOCK_SOURCE:
1822 			case MCAST_JOIN_GROUP:
1823 			case MCAST_LEAVE_GROUP:
1824 			case MCAST_JOIN_SOURCE_GROUP:
1825 			case MCAST_LEAVE_SOURCE_GROUP:
1826 				error = ip6_setmoptions(in6p, sopt);
1827 				break;
1828 
1829 			case IPV6_PORTRANGE:
1830 				error = sooptcopyin(sopt, &optval,
1831 				    sizeof optval, sizeof optval);
1832 				if (error)
1833 					break;
1834 
1835 				INP_WLOCK(in6p);
1836 				switch (optval) {
1837 				case IPV6_PORTRANGE_DEFAULT:
1838 					in6p->inp_flags &= ~(INP_LOWPORT);
1839 					in6p->inp_flags &= ~(INP_HIGHPORT);
1840 					break;
1841 
1842 				case IPV6_PORTRANGE_HIGH:
1843 					in6p->inp_flags &= ~(INP_LOWPORT);
1844 					in6p->inp_flags |= INP_HIGHPORT;
1845 					break;
1846 
1847 				case IPV6_PORTRANGE_LOW:
1848 					in6p->inp_flags &= ~(INP_HIGHPORT);
1849 					in6p->inp_flags |= INP_LOWPORT;
1850 					break;
1851 
1852 				default:
1853 					error = EINVAL;
1854 					break;
1855 				}
1856 				INP_WUNLOCK(in6p);
1857 				break;
1858 
1859 #ifdef IPSEC
1860 			case IPV6_IPSEC_POLICY:
1861 			{
1862 				caddr_t req;
1863 				struct mbuf *m;
1864 
1865 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1866 					break;
1867 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1868 					break;
1869 				req = mtod(m, caddr_t);
1870 				error = ipsec_set_policy(in6p, optname, req,
1871 				    m->m_len, (sopt->sopt_td != NULL) ?
1872 				    sopt->sopt_td->td_ucred : NULL);
1873 				m_freem(m);
1874 				break;
1875 			}
1876 #endif /* IPSEC */
1877 
1878 			default:
1879 				error = ENOPROTOOPT;
1880 				break;
1881 			}
1882 			break;
1883 
1884 		case SOPT_GET:
1885 			switch (optname) {
1886 
1887 			case IPV6_2292PKTOPTIONS:
1888 #ifdef IPV6_PKTOPTIONS
1889 			case IPV6_PKTOPTIONS:
1890 #endif
1891 				/*
1892 				 * RFC3542 (effectively) deprecated the
1893 				 * semantics of the 2292-style pktoptions.
1894 				 * Since it was not reliable in nature (i.e.,
1895 				 * applications had to expect the lack of some
1896 				 * information after all), it would make sense
1897 				 * to simplify this part by always returning
1898 				 * empty data.
1899 				 */
1900 				sopt->sopt_valsize = 0;
1901 				break;
1902 
1903 			case IPV6_RECVHOPOPTS:
1904 			case IPV6_RECVDSTOPTS:
1905 			case IPV6_RECVRTHDRDSTOPTS:
1906 			case IPV6_UNICAST_HOPS:
1907 			case IPV6_RECVPKTINFO:
1908 			case IPV6_RECVHOPLIMIT:
1909 			case IPV6_RECVRTHDR:
1910 			case IPV6_RECVPATHMTU:
1911 
1912 			case IPV6_V6ONLY:
1913 			case IPV6_PORTRANGE:
1914 			case IPV6_RECVTCLASS:
1915 			case IPV6_AUTOFLOWLABEL:
1916 			case IPV6_BINDANY:
1917 			case IPV6_FLOWID:
1918 			case IPV6_FLOWTYPE:
1919 			case IPV6_RECVFLOWID:
1920 #ifdef	RSS
1921 			case IPV6_RSSBUCKETID:
1922 			case IPV6_RECVRSSBUCKETID:
1923 #endif
1924 			case IPV6_BINDMULTI:
1925 				switch (optname) {
1926 
1927 				case IPV6_RECVHOPOPTS:
1928 					optval = OPTBIT(IN6P_HOPOPTS);
1929 					break;
1930 
1931 				case IPV6_RECVDSTOPTS:
1932 					optval = OPTBIT(IN6P_DSTOPTS);
1933 					break;
1934 
1935 				case IPV6_RECVRTHDRDSTOPTS:
1936 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1937 					break;
1938 
1939 				case IPV6_UNICAST_HOPS:
1940 					optval = in6p->in6p_hops;
1941 					break;
1942 
1943 				case IPV6_RECVPKTINFO:
1944 					optval = OPTBIT(IN6P_PKTINFO);
1945 					break;
1946 
1947 				case IPV6_RECVHOPLIMIT:
1948 					optval = OPTBIT(IN6P_HOPLIMIT);
1949 					break;
1950 
1951 				case IPV6_RECVRTHDR:
1952 					optval = OPTBIT(IN6P_RTHDR);
1953 					break;
1954 
1955 				case IPV6_RECVPATHMTU:
1956 					optval = OPTBIT(IN6P_MTU);
1957 					break;
1958 
1959 				case IPV6_V6ONLY:
1960 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1961 					break;
1962 
1963 				case IPV6_PORTRANGE:
1964 				    {
1965 					int flags;
1966 					flags = in6p->inp_flags;
1967 					if (flags & INP_HIGHPORT)
1968 						optval = IPV6_PORTRANGE_HIGH;
1969 					else if (flags & INP_LOWPORT)
1970 						optval = IPV6_PORTRANGE_LOW;
1971 					else
1972 						optval = 0;
1973 					break;
1974 				    }
1975 				case IPV6_RECVTCLASS:
1976 					optval = OPTBIT(IN6P_TCLASS);
1977 					break;
1978 
1979 				case IPV6_AUTOFLOWLABEL:
1980 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1981 					break;
1982 
1983 				case IPV6_BINDANY:
1984 					optval = OPTBIT(INP_BINDANY);
1985 					break;
1986 
1987 				case IPV6_FLOWID:
1988 					optval = in6p->inp_flowid;
1989 					break;
1990 
1991 				case IPV6_FLOWTYPE:
1992 					optval = in6p->inp_flowtype;
1993 					break;
1994 
1995 				case IPV6_RECVFLOWID:
1996 					optval = OPTBIT2(INP_RECVFLOWID);
1997 					break;
1998 #ifdef	RSS
1999 				case IPV6_RSSBUCKETID:
2000 					retval =
2001 					    rss_hash2bucket(in6p->inp_flowid,
2002 					    in6p->inp_flowtype,
2003 					    &rss_bucket);
2004 					if (retval == 0)
2005 						optval = rss_bucket;
2006 					else
2007 						error = EINVAL;
2008 					break;
2009 
2010 				case IPV6_RECVRSSBUCKETID:
2011 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2012 					break;
2013 #endif
2014 
2015 				case IPV6_BINDMULTI:
2016 					optval = OPTBIT2(INP_BINDMULTI);
2017 					break;
2018 
2019 				}
2020 				if (error)
2021 					break;
2022 				error = sooptcopyout(sopt, &optval,
2023 					sizeof optval);
2024 				break;
2025 
2026 			case IPV6_PATHMTU:
2027 			{
2028 				u_long pmtu = 0;
2029 				struct ip6_mtuinfo mtuinfo;
2030 
2031 				if (!(so->so_state & SS_ISCONNECTED))
2032 					return (ENOTCONN);
2033 				/*
2034 				 * XXX: we dot not consider the case of source
2035 				 * routing, or optional information to specify
2036 				 * the outgoing interface.
2037 				 */
2038 				error = ip6_getpmtu_ctl(so->so_fibnum,
2039 				    &in6p->in6p_faddr, &pmtu);
2040 				if (error)
2041 					break;
2042 				if (pmtu > IPV6_MAXPACKET)
2043 					pmtu = IPV6_MAXPACKET;
2044 
2045 				bzero(&mtuinfo, sizeof(mtuinfo));
2046 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2047 				optdata = (void *)&mtuinfo;
2048 				optdatalen = sizeof(mtuinfo);
2049 				error = sooptcopyout(sopt, optdata,
2050 				    optdatalen);
2051 				break;
2052 			}
2053 
2054 			case IPV6_2292PKTINFO:
2055 			case IPV6_2292HOPLIMIT:
2056 			case IPV6_2292HOPOPTS:
2057 			case IPV6_2292RTHDR:
2058 			case IPV6_2292DSTOPTS:
2059 				switch (optname) {
2060 				case IPV6_2292PKTINFO:
2061 					optval = OPTBIT(IN6P_PKTINFO);
2062 					break;
2063 				case IPV6_2292HOPLIMIT:
2064 					optval = OPTBIT(IN6P_HOPLIMIT);
2065 					break;
2066 				case IPV6_2292HOPOPTS:
2067 					optval = OPTBIT(IN6P_HOPOPTS);
2068 					break;
2069 				case IPV6_2292RTHDR:
2070 					optval = OPTBIT(IN6P_RTHDR);
2071 					break;
2072 				case IPV6_2292DSTOPTS:
2073 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2074 					break;
2075 				}
2076 				error = sooptcopyout(sopt, &optval,
2077 				    sizeof optval);
2078 				break;
2079 			case IPV6_PKTINFO:
2080 			case IPV6_HOPOPTS:
2081 			case IPV6_RTHDR:
2082 			case IPV6_DSTOPTS:
2083 			case IPV6_RTHDRDSTOPTS:
2084 			case IPV6_NEXTHOP:
2085 			case IPV6_TCLASS:
2086 			case IPV6_DONTFRAG:
2087 			case IPV6_USE_MIN_MTU:
2088 			case IPV6_PREFER_TEMPADDR:
2089 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2090 				    optname, sopt);
2091 				break;
2092 
2093 			case IPV6_MULTICAST_IF:
2094 			case IPV6_MULTICAST_HOPS:
2095 			case IPV6_MULTICAST_LOOP:
2096 			case IPV6_MSFILTER:
2097 				error = ip6_getmoptions(in6p, sopt);
2098 				break;
2099 
2100 #ifdef IPSEC
2101 			case IPV6_IPSEC_POLICY:
2102 			  {
2103 				caddr_t req = NULL;
2104 				size_t len = 0;
2105 				struct mbuf *m = NULL;
2106 				struct mbuf **mp = &m;
2107 				size_t ovalsize = sopt->sopt_valsize;
2108 				caddr_t oval = (caddr_t)sopt->sopt_val;
2109 
2110 				error = soopt_getm(sopt, &m); /* XXX */
2111 				if (error != 0)
2112 					break;
2113 				error = soopt_mcopyin(sopt, m); /* XXX */
2114 				if (error != 0)
2115 					break;
2116 				sopt->sopt_valsize = ovalsize;
2117 				sopt->sopt_val = oval;
2118 				if (m) {
2119 					req = mtod(m, caddr_t);
2120 					len = m->m_len;
2121 				}
2122 				error = ipsec_get_policy(in6p, req, len, mp);
2123 				if (error == 0)
2124 					error = soopt_mcopyout(sopt, m); /* XXX */
2125 				if (error == 0 && m)
2126 					m_freem(m);
2127 				break;
2128 			  }
2129 #endif /* IPSEC */
2130 
2131 			default:
2132 				error = ENOPROTOOPT;
2133 				break;
2134 			}
2135 			break;
2136 		}
2137 	}
2138 	return (error);
2139 }
2140 
2141 int
2142 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2143 {
2144 	int error = 0, optval, optlen;
2145 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2146 	struct inpcb *in6p = sotoinpcb(so);
2147 	int level, op, optname;
2148 
2149 	level = sopt->sopt_level;
2150 	op = sopt->sopt_dir;
2151 	optname = sopt->sopt_name;
2152 	optlen = sopt->sopt_valsize;
2153 
2154 	if (level != IPPROTO_IPV6) {
2155 		return (EINVAL);
2156 	}
2157 
2158 	switch (optname) {
2159 	case IPV6_CHECKSUM:
2160 		/*
2161 		 * For ICMPv6 sockets, no modification allowed for checksum
2162 		 * offset, permit "no change" values to help existing apps.
2163 		 *
2164 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2165 		 * for an ICMPv6 socket will fail."
2166 		 * The current behavior does not meet RFC3542.
2167 		 */
2168 		switch (op) {
2169 		case SOPT_SET:
2170 			if (optlen != sizeof(int)) {
2171 				error = EINVAL;
2172 				break;
2173 			}
2174 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2175 					    sizeof(optval));
2176 			if (error)
2177 				break;
2178 			if ((optval % 2) != 0) {
2179 				/* the API assumes even offset values */
2180 				error = EINVAL;
2181 			} else if (so->so_proto->pr_protocol ==
2182 			    IPPROTO_ICMPV6) {
2183 				if (optval != icmp6off)
2184 					error = EINVAL;
2185 			} else
2186 				in6p->in6p_cksum = optval;
2187 			break;
2188 
2189 		case SOPT_GET:
2190 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2191 				optval = icmp6off;
2192 			else
2193 				optval = in6p->in6p_cksum;
2194 
2195 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2196 			break;
2197 
2198 		default:
2199 			error = EINVAL;
2200 			break;
2201 		}
2202 		break;
2203 
2204 	default:
2205 		error = ENOPROTOOPT;
2206 		break;
2207 	}
2208 
2209 	return (error);
2210 }
2211 
2212 /*
2213  * Set up IP6 options in pcb for insertion in output packets or
2214  * specifying behavior of outgoing packets.
2215  */
2216 static int
2217 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2218     struct socket *so, struct sockopt *sopt)
2219 {
2220 	struct ip6_pktopts *opt = *pktopt;
2221 	int error = 0;
2222 	struct thread *td = sopt->sopt_td;
2223 
2224 	/* turn off any old options. */
2225 	if (opt) {
2226 #ifdef DIAGNOSTIC
2227 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2228 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2229 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2230 			printf("ip6_pcbopts: all specified options are cleared.\n");
2231 #endif
2232 		ip6_clearpktopts(opt, -1);
2233 	} else
2234 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2235 	*pktopt = NULL;
2236 
2237 	if (!m || m->m_len == 0) {
2238 		/*
2239 		 * Only turning off any previous options, regardless of
2240 		 * whether the opt is just created or given.
2241 		 */
2242 		free(opt, M_IP6OPT);
2243 		return (0);
2244 	}
2245 
2246 	/*  set options specified by user. */
2247 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2248 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2249 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2250 		free(opt, M_IP6OPT);
2251 		return (error);
2252 	}
2253 	*pktopt = opt;
2254 	return (0);
2255 }
2256 
2257 /*
2258  * initialize ip6_pktopts.  beware that there are non-zero default values in
2259  * the struct.
2260  */
2261 void
2262 ip6_initpktopts(struct ip6_pktopts *opt)
2263 {
2264 
2265 	bzero(opt, sizeof(*opt));
2266 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2267 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2268 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2269 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2270 }
2271 
2272 static int
2273 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2274     struct ucred *cred, int uproto)
2275 {
2276 	struct ip6_pktopts *opt;
2277 
2278 	if (*pktopt == NULL) {
2279 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2280 		    M_WAITOK);
2281 		ip6_initpktopts(*pktopt);
2282 	}
2283 	opt = *pktopt;
2284 
2285 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2286 }
2287 
2288 static int
2289 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2290 {
2291 	void *optdata = NULL;
2292 	int optdatalen = 0;
2293 	struct ip6_ext *ip6e;
2294 	int error = 0;
2295 	struct in6_pktinfo null_pktinfo;
2296 	int deftclass = 0, on;
2297 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2298 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2299 
2300 	switch (optname) {
2301 	case IPV6_PKTINFO:
2302 		optdata = (void *)&null_pktinfo;
2303 		if (pktopt && pktopt->ip6po_pktinfo) {
2304 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2305 			    sizeof(null_pktinfo));
2306 			in6_clearscope(&null_pktinfo.ipi6_addr);
2307 		} else {
2308 			/* XXX: we don't have to do this every time... */
2309 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2310 		}
2311 		optdatalen = sizeof(struct in6_pktinfo);
2312 		break;
2313 	case IPV6_TCLASS:
2314 		if (pktopt && pktopt->ip6po_tclass >= 0)
2315 			optdata = (void *)&pktopt->ip6po_tclass;
2316 		else
2317 			optdata = (void *)&deftclass;
2318 		optdatalen = sizeof(int);
2319 		break;
2320 	case IPV6_HOPOPTS:
2321 		if (pktopt && pktopt->ip6po_hbh) {
2322 			optdata = (void *)pktopt->ip6po_hbh;
2323 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2324 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2325 		}
2326 		break;
2327 	case IPV6_RTHDR:
2328 		if (pktopt && pktopt->ip6po_rthdr) {
2329 			optdata = (void *)pktopt->ip6po_rthdr;
2330 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2331 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2332 		}
2333 		break;
2334 	case IPV6_RTHDRDSTOPTS:
2335 		if (pktopt && pktopt->ip6po_dest1) {
2336 			optdata = (void *)pktopt->ip6po_dest1;
2337 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2338 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2339 		}
2340 		break;
2341 	case IPV6_DSTOPTS:
2342 		if (pktopt && pktopt->ip6po_dest2) {
2343 			optdata = (void *)pktopt->ip6po_dest2;
2344 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2345 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2346 		}
2347 		break;
2348 	case IPV6_NEXTHOP:
2349 		if (pktopt && pktopt->ip6po_nexthop) {
2350 			optdata = (void *)pktopt->ip6po_nexthop;
2351 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2352 		}
2353 		break;
2354 	case IPV6_USE_MIN_MTU:
2355 		if (pktopt)
2356 			optdata = (void *)&pktopt->ip6po_minmtu;
2357 		else
2358 			optdata = (void *)&defminmtu;
2359 		optdatalen = sizeof(int);
2360 		break;
2361 	case IPV6_DONTFRAG:
2362 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2363 			on = 1;
2364 		else
2365 			on = 0;
2366 		optdata = (void *)&on;
2367 		optdatalen = sizeof(on);
2368 		break;
2369 	case IPV6_PREFER_TEMPADDR:
2370 		if (pktopt)
2371 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2372 		else
2373 			optdata = (void *)&defpreftemp;
2374 		optdatalen = sizeof(int);
2375 		break;
2376 	default:		/* should not happen */
2377 #ifdef DIAGNOSTIC
2378 		panic("ip6_getpcbopt: unexpected option\n");
2379 #endif
2380 		return (ENOPROTOOPT);
2381 	}
2382 
2383 	error = sooptcopyout(sopt, optdata, optdatalen);
2384 
2385 	return (error);
2386 }
2387 
2388 void
2389 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2390 {
2391 	if (pktopt == NULL)
2392 		return;
2393 
2394 	if (optname == -1 || optname == IPV6_PKTINFO) {
2395 		if (pktopt->ip6po_pktinfo)
2396 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2397 		pktopt->ip6po_pktinfo = NULL;
2398 	}
2399 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2400 		pktopt->ip6po_hlim = -1;
2401 	if (optname == -1 || optname == IPV6_TCLASS)
2402 		pktopt->ip6po_tclass = -1;
2403 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2404 		if (pktopt->ip6po_nextroute.ro_rt) {
2405 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2406 			pktopt->ip6po_nextroute.ro_rt = NULL;
2407 		}
2408 		if (pktopt->ip6po_nexthop)
2409 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2410 		pktopt->ip6po_nexthop = NULL;
2411 	}
2412 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2413 		if (pktopt->ip6po_hbh)
2414 			free(pktopt->ip6po_hbh, M_IP6OPT);
2415 		pktopt->ip6po_hbh = NULL;
2416 	}
2417 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2418 		if (pktopt->ip6po_dest1)
2419 			free(pktopt->ip6po_dest1, M_IP6OPT);
2420 		pktopt->ip6po_dest1 = NULL;
2421 	}
2422 	if (optname == -1 || optname == IPV6_RTHDR) {
2423 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2424 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2425 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2426 		if (pktopt->ip6po_route.ro_rt) {
2427 			RTFREE(pktopt->ip6po_route.ro_rt);
2428 			pktopt->ip6po_route.ro_rt = NULL;
2429 		}
2430 	}
2431 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2432 		if (pktopt->ip6po_dest2)
2433 			free(pktopt->ip6po_dest2, M_IP6OPT);
2434 		pktopt->ip6po_dest2 = NULL;
2435 	}
2436 }
2437 
2438 #define PKTOPT_EXTHDRCPY(type) \
2439 do {\
2440 	if (src->type) {\
2441 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2442 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2443 		if (dst->type == NULL && canwait == M_NOWAIT)\
2444 			goto bad;\
2445 		bcopy(src->type, dst->type, hlen);\
2446 	}\
2447 } while (/*CONSTCOND*/ 0)
2448 
2449 static int
2450 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2451 {
2452 	if (dst == NULL || src == NULL)  {
2453 		printf("ip6_clearpktopts: invalid argument\n");
2454 		return (EINVAL);
2455 	}
2456 
2457 	dst->ip6po_hlim = src->ip6po_hlim;
2458 	dst->ip6po_tclass = src->ip6po_tclass;
2459 	dst->ip6po_flags = src->ip6po_flags;
2460 	dst->ip6po_minmtu = src->ip6po_minmtu;
2461 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2462 	if (src->ip6po_pktinfo) {
2463 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2464 		    M_IP6OPT, canwait);
2465 		if (dst->ip6po_pktinfo == NULL)
2466 			goto bad;
2467 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2468 	}
2469 	if (src->ip6po_nexthop) {
2470 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2471 		    M_IP6OPT, canwait);
2472 		if (dst->ip6po_nexthop == NULL)
2473 			goto bad;
2474 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2475 		    src->ip6po_nexthop->sa_len);
2476 	}
2477 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2478 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2479 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2480 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2481 	return (0);
2482 
2483   bad:
2484 	ip6_clearpktopts(dst, -1);
2485 	return (ENOBUFS);
2486 }
2487 #undef PKTOPT_EXTHDRCPY
2488 
2489 struct ip6_pktopts *
2490 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2491 {
2492 	int error;
2493 	struct ip6_pktopts *dst;
2494 
2495 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2496 	if (dst == NULL)
2497 		return (NULL);
2498 	ip6_initpktopts(dst);
2499 
2500 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2501 		free(dst, M_IP6OPT);
2502 		return (NULL);
2503 	}
2504 
2505 	return (dst);
2506 }
2507 
2508 void
2509 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2510 {
2511 	if (pktopt == NULL)
2512 		return;
2513 
2514 	ip6_clearpktopts(pktopt, -1);
2515 
2516 	free(pktopt, M_IP6OPT);
2517 }
2518 
2519 /*
2520  * Set IPv6 outgoing packet options based on advanced API.
2521  */
2522 int
2523 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2524     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2525 {
2526 	struct cmsghdr *cm = NULL;
2527 
2528 	if (control == NULL || opt == NULL)
2529 		return (EINVAL);
2530 
2531 	ip6_initpktopts(opt);
2532 	if (stickyopt) {
2533 		int error;
2534 
2535 		/*
2536 		 * If stickyopt is provided, make a local copy of the options
2537 		 * for this particular packet, then override them by ancillary
2538 		 * objects.
2539 		 * XXX: copypktopts() does not copy the cached route to a next
2540 		 * hop (if any).  This is not very good in terms of efficiency,
2541 		 * but we can allow this since this option should be rarely
2542 		 * used.
2543 		 */
2544 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2545 			return (error);
2546 	}
2547 
2548 	/*
2549 	 * XXX: Currently, we assume all the optional information is stored
2550 	 * in a single mbuf.
2551 	 */
2552 	if (control->m_next)
2553 		return (EINVAL);
2554 
2555 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2556 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2557 		int error;
2558 
2559 		if (control->m_len < CMSG_LEN(0))
2560 			return (EINVAL);
2561 
2562 		cm = mtod(control, struct cmsghdr *);
2563 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2564 			return (EINVAL);
2565 		if (cm->cmsg_level != IPPROTO_IPV6)
2566 			continue;
2567 
2568 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2569 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2570 		if (error)
2571 			return (error);
2572 	}
2573 
2574 	return (0);
2575 }
2576 
2577 /*
2578  * Set a particular packet option, as a sticky option or an ancillary data
2579  * item.  "len" can be 0 only when it's a sticky option.
2580  * We have 4 cases of combination of "sticky" and "cmsg":
2581  * "sticky=0, cmsg=0": impossible
2582  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2583  * "sticky=1, cmsg=0": RFC3542 socket option
2584  * "sticky=1, cmsg=1": RFC2292 socket option
2585  */
2586 static int
2587 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2588     struct ucred *cred, int sticky, int cmsg, int uproto)
2589 {
2590 	int minmtupolicy, preftemp;
2591 	int error;
2592 
2593 	if (!sticky && !cmsg) {
2594 #ifdef DIAGNOSTIC
2595 		printf("ip6_setpktopt: impossible case\n");
2596 #endif
2597 		return (EINVAL);
2598 	}
2599 
2600 	/*
2601 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2602 	 * not be specified in the context of RFC3542.  Conversely,
2603 	 * RFC3542 types should not be specified in the context of RFC2292.
2604 	 */
2605 	if (!cmsg) {
2606 		switch (optname) {
2607 		case IPV6_2292PKTINFO:
2608 		case IPV6_2292HOPLIMIT:
2609 		case IPV6_2292NEXTHOP:
2610 		case IPV6_2292HOPOPTS:
2611 		case IPV6_2292DSTOPTS:
2612 		case IPV6_2292RTHDR:
2613 		case IPV6_2292PKTOPTIONS:
2614 			return (ENOPROTOOPT);
2615 		}
2616 	}
2617 	if (sticky && cmsg) {
2618 		switch (optname) {
2619 		case IPV6_PKTINFO:
2620 		case IPV6_HOPLIMIT:
2621 		case IPV6_NEXTHOP:
2622 		case IPV6_HOPOPTS:
2623 		case IPV6_DSTOPTS:
2624 		case IPV6_RTHDRDSTOPTS:
2625 		case IPV6_RTHDR:
2626 		case IPV6_USE_MIN_MTU:
2627 		case IPV6_DONTFRAG:
2628 		case IPV6_TCLASS:
2629 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2630 			return (ENOPROTOOPT);
2631 		}
2632 	}
2633 
2634 	switch (optname) {
2635 	case IPV6_2292PKTINFO:
2636 	case IPV6_PKTINFO:
2637 	{
2638 		struct ifnet *ifp = NULL;
2639 		struct in6_pktinfo *pktinfo;
2640 
2641 		if (len != sizeof(struct in6_pktinfo))
2642 			return (EINVAL);
2643 
2644 		pktinfo = (struct in6_pktinfo *)buf;
2645 
2646 		/*
2647 		 * An application can clear any sticky IPV6_PKTINFO option by
2648 		 * doing a "regular" setsockopt with ipi6_addr being
2649 		 * in6addr_any and ipi6_ifindex being zero.
2650 		 * [RFC 3542, Section 6]
2651 		 */
2652 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2653 		    pktinfo->ipi6_ifindex == 0 &&
2654 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2655 			ip6_clearpktopts(opt, optname);
2656 			break;
2657 		}
2658 
2659 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2660 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2661 			return (EINVAL);
2662 		}
2663 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2664 			return (EINVAL);
2665 		/* validate the interface index if specified. */
2666 		if (pktinfo->ipi6_ifindex > V_if_index)
2667 			 return (ENXIO);
2668 		if (pktinfo->ipi6_ifindex) {
2669 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2670 			if (ifp == NULL)
2671 				return (ENXIO);
2672 		}
2673 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2674 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2675 			return (ENETDOWN);
2676 
2677 		if (ifp != NULL &&
2678 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2679 			struct in6_ifaddr *ia;
2680 
2681 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2682 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2683 			if (ia == NULL)
2684 				return (EADDRNOTAVAIL);
2685 			ifa_free(&ia->ia_ifa);
2686 		}
2687 		/*
2688 		 * We store the address anyway, and let in6_selectsrc()
2689 		 * validate the specified address.  This is because ipi6_addr
2690 		 * may not have enough information about its scope zone, and
2691 		 * we may need additional information (such as outgoing
2692 		 * interface or the scope zone of a destination address) to
2693 		 * disambiguate the scope.
2694 		 * XXX: the delay of the validation may confuse the
2695 		 * application when it is used as a sticky option.
2696 		 */
2697 		if (opt->ip6po_pktinfo == NULL) {
2698 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2699 			    M_IP6OPT, M_NOWAIT);
2700 			if (opt->ip6po_pktinfo == NULL)
2701 				return (ENOBUFS);
2702 		}
2703 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2704 		break;
2705 	}
2706 
2707 	case IPV6_2292HOPLIMIT:
2708 	case IPV6_HOPLIMIT:
2709 	{
2710 		int *hlimp;
2711 
2712 		/*
2713 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2714 		 * to simplify the ordering among hoplimit options.
2715 		 */
2716 		if (optname == IPV6_HOPLIMIT && sticky)
2717 			return (ENOPROTOOPT);
2718 
2719 		if (len != sizeof(int))
2720 			return (EINVAL);
2721 		hlimp = (int *)buf;
2722 		if (*hlimp < -1 || *hlimp > 255)
2723 			return (EINVAL);
2724 
2725 		opt->ip6po_hlim = *hlimp;
2726 		break;
2727 	}
2728 
2729 	case IPV6_TCLASS:
2730 	{
2731 		int tclass;
2732 
2733 		if (len != sizeof(int))
2734 			return (EINVAL);
2735 		tclass = *(int *)buf;
2736 		if (tclass < -1 || tclass > 255)
2737 			return (EINVAL);
2738 
2739 		opt->ip6po_tclass = tclass;
2740 		break;
2741 	}
2742 
2743 	case IPV6_2292NEXTHOP:
2744 	case IPV6_NEXTHOP:
2745 		if (cred != NULL) {
2746 			error = priv_check_cred(cred,
2747 			    PRIV_NETINET_SETHDROPTS, 0);
2748 			if (error)
2749 				return (error);
2750 		}
2751 
2752 		if (len == 0) {	/* just remove the option */
2753 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2754 			break;
2755 		}
2756 
2757 		/* check if cmsg_len is large enough for sa_len */
2758 		if (len < sizeof(struct sockaddr) || len < *buf)
2759 			return (EINVAL);
2760 
2761 		switch (((struct sockaddr *)buf)->sa_family) {
2762 		case AF_INET6:
2763 		{
2764 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2765 			int error;
2766 
2767 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2768 				return (EINVAL);
2769 
2770 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2771 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2772 				return (EINVAL);
2773 			}
2774 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2775 			    != 0) {
2776 				return (error);
2777 			}
2778 			break;
2779 		}
2780 		case AF_LINK:	/* should eventually be supported */
2781 		default:
2782 			return (EAFNOSUPPORT);
2783 		}
2784 
2785 		/* turn off the previous option, then set the new option. */
2786 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2787 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2788 		if (opt->ip6po_nexthop == NULL)
2789 			return (ENOBUFS);
2790 		bcopy(buf, opt->ip6po_nexthop, *buf);
2791 		break;
2792 
2793 	case IPV6_2292HOPOPTS:
2794 	case IPV6_HOPOPTS:
2795 	{
2796 		struct ip6_hbh *hbh;
2797 		int hbhlen;
2798 
2799 		/*
2800 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2801 		 * options, since per-option restriction has too much
2802 		 * overhead.
2803 		 */
2804 		if (cred != NULL) {
2805 			error = priv_check_cred(cred,
2806 			    PRIV_NETINET_SETHDROPTS, 0);
2807 			if (error)
2808 				return (error);
2809 		}
2810 
2811 		if (len == 0) {
2812 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2813 			break;	/* just remove the option */
2814 		}
2815 
2816 		/* message length validation */
2817 		if (len < sizeof(struct ip6_hbh))
2818 			return (EINVAL);
2819 		hbh = (struct ip6_hbh *)buf;
2820 		hbhlen = (hbh->ip6h_len + 1) << 3;
2821 		if (len != hbhlen)
2822 			return (EINVAL);
2823 
2824 		/* turn off the previous option, then set the new option. */
2825 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2826 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2827 		if (opt->ip6po_hbh == NULL)
2828 			return (ENOBUFS);
2829 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2830 
2831 		break;
2832 	}
2833 
2834 	case IPV6_2292DSTOPTS:
2835 	case IPV6_DSTOPTS:
2836 	case IPV6_RTHDRDSTOPTS:
2837 	{
2838 		struct ip6_dest *dest, **newdest = NULL;
2839 		int destlen;
2840 
2841 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2842 			error = priv_check_cred(cred,
2843 			    PRIV_NETINET_SETHDROPTS, 0);
2844 			if (error)
2845 				return (error);
2846 		}
2847 
2848 		if (len == 0) {
2849 			ip6_clearpktopts(opt, optname);
2850 			break;	/* just remove the option */
2851 		}
2852 
2853 		/* message length validation */
2854 		if (len < sizeof(struct ip6_dest))
2855 			return (EINVAL);
2856 		dest = (struct ip6_dest *)buf;
2857 		destlen = (dest->ip6d_len + 1) << 3;
2858 		if (len != destlen)
2859 			return (EINVAL);
2860 
2861 		/*
2862 		 * Determine the position that the destination options header
2863 		 * should be inserted; before or after the routing header.
2864 		 */
2865 		switch (optname) {
2866 		case IPV6_2292DSTOPTS:
2867 			/*
2868 			 * The old advacned API is ambiguous on this point.
2869 			 * Our approach is to determine the position based
2870 			 * according to the existence of a routing header.
2871 			 * Note, however, that this depends on the order of the
2872 			 * extension headers in the ancillary data; the 1st
2873 			 * part of the destination options header must appear
2874 			 * before the routing header in the ancillary data,
2875 			 * too.
2876 			 * RFC3542 solved the ambiguity by introducing
2877 			 * separate ancillary data or option types.
2878 			 */
2879 			if (opt->ip6po_rthdr == NULL)
2880 				newdest = &opt->ip6po_dest1;
2881 			else
2882 				newdest = &opt->ip6po_dest2;
2883 			break;
2884 		case IPV6_RTHDRDSTOPTS:
2885 			newdest = &opt->ip6po_dest1;
2886 			break;
2887 		case IPV6_DSTOPTS:
2888 			newdest = &opt->ip6po_dest2;
2889 			break;
2890 		}
2891 
2892 		/* turn off the previous option, then set the new option. */
2893 		ip6_clearpktopts(opt, optname);
2894 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2895 		if (*newdest == NULL)
2896 			return (ENOBUFS);
2897 		bcopy(dest, *newdest, destlen);
2898 
2899 		break;
2900 	}
2901 
2902 	case IPV6_2292RTHDR:
2903 	case IPV6_RTHDR:
2904 	{
2905 		struct ip6_rthdr *rth;
2906 		int rthlen;
2907 
2908 		if (len == 0) {
2909 			ip6_clearpktopts(opt, IPV6_RTHDR);
2910 			break;	/* just remove the option */
2911 		}
2912 
2913 		/* message length validation */
2914 		if (len < sizeof(struct ip6_rthdr))
2915 			return (EINVAL);
2916 		rth = (struct ip6_rthdr *)buf;
2917 		rthlen = (rth->ip6r_len + 1) << 3;
2918 		if (len != rthlen)
2919 			return (EINVAL);
2920 
2921 		switch (rth->ip6r_type) {
2922 		case IPV6_RTHDR_TYPE_0:
2923 			if (rth->ip6r_len == 0)	/* must contain one addr */
2924 				return (EINVAL);
2925 			if (rth->ip6r_len % 2) /* length must be even */
2926 				return (EINVAL);
2927 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2928 				return (EINVAL);
2929 			break;
2930 		default:
2931 			return (EINVAL);	/* not supported */
2932 		}
2933 
2934 		/* turn off the previous option */
2935 		ip6_clearpktopts(opt, IPV6_RTHDR);
2936 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2937 		if (opt->ip6po_rthdr == NULL)
2938 			return (ENOBUFS);
2939 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2940 
2941 		break;
2942 	}
2943 
2944 	case IPV6_USE_MIN_MTU:
2945 		if (len != sizeof(int))
2946 			return (EINVAL);
2947 		minmtupolicy = *(int *)buf;
2948 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2949 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2950 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2951 			return (EINVAL);
2952 		}
2953 		opt->ip6po_minmtu = minmtupolicy;
2954 		break;
2955 
2956 	case IPV6_DONTFRAG:
2957 		if (len != sizeof(int))
2958 			return (EINVAL);
2959 
2960 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2961 			/*
2962 			 * we ignore this option for TCP sockets.
2963 			 * (RFC3542 leaves this case unspecified.)
2964 			 */
2965 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2966 		} else
2967 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2968 		break;
2969 
2970 	case IPV6_PREFER_TEMPADDR:
2971 		if (len != sizeof(int))
2972 			return (EINVAL);
2973 		preftemp = *(int *)buf;
2974 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2975 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2976 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2977 			return (EINVAL);
2978 		}
2979 		opt->ip6po_prefer_tempaddr = preftemp;
2980 		break;
2981 
2982 	default:
2983 		return (ENOPROTOOPT);
2984 	} /* end of switch */
2985 
2986 	return (0);
2987 }
2988 
2989 /*
2990  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2991  * packet to the input queue of a specified interface.  Note that this
2992  * calls the output routine of the loopback "driver", but with an interface
2993  * pointer that might NOT be &loif -- easier than replicating that code here.
2994  */
2995 void
2996 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
2997 {
2998 	struct mbuf *copym;
2999 	struct ip6_hdr *ip6;
3000 
3001 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3002 	if (copym == NULL)
3003 		return;
3004 
3005 	/*
3006 	 * Make sure to deep-copy IPv6 header portion in case the data
3007 	 * is in an mbuf cluster, so that we can safely override the IPv6
3008 	 * header portion later.
3009 	 */
3010 	if (!M_WRITABLE(copym) ||
3011 	    copym->m_len < sizeof(struct ip6_hdr)) {
3012 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3013 		if (copym == NULL)
3014 			return;
3015 	}
3016 	ip6 = mtod(copym, struct ip6_hdr *);
3017 	/*
3018 	 * clear embedded scope identifiers if necessary.
3019 	 * in6_clearscope will touch the addresses only when necessary.
3020 	 */
3021 	in6_clearscope(&ip6->ip6_src);
3022 	in6_clearscope(&ip6->ip6_dst);
3023 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3024 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3025 		    CSUM_PSEUDO_HDR;
3026 		copym->m_pkthdr.csum_data = 0xffff;
3027 	}
3028 	if_simloop(ifp, copym, AF_INET6, 0);
3029 }
3030 
3031 /*
3032  * Chop IPv6 header off from the payload.
3033  */
3034 static int
3035 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3036 {
3037 	struct mbuf *mh;
3038 	struct ip6_hdr *ip6;
3039 
3040 	ip6 = mtod(m, struct ip6_hdr *);
3041 	if (m->m_len > sizeof(*ip6)) {
3042 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3043 		if (mh == NULL) {
3044 			m_freem(m);
3045 			return ENOBUFS;
3046 		}
3047 		m_move_pkthdr(mh, m);
3048 		M_ALIGN(mh, sizeof(*ip6));
3049 		m->m_len -= sizeof(*ip6);
3050 		m->m_data += sizeof(*ip6);
3051 		mh->m_next = m;
3052 		m = mh;
3053 		m->m_len = sizeof(*ip6);
3054 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3055 	}
3056 	exthdrs->ip6e_ip6 = m;
3057 	return 0;
3058 }
3059 
3060 /*
3061  * Compute IPv6 extension header length.
3062  */
3063 int
3064 ip6_optlen(struct inpcb *in6p)
3065 {
3066 	int len;
3067 
3068 	if (!in6p->in6p_outputopts)
3069 		return 0;
3070 
3071 	len = 0;
3072 #define elen(x) \
3073     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3074 
3075 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3076 	if (in6p->in6p_outputopts->ip6po_rthdr)
3077 		/* dest1 is valid with rthdr only */
3078 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3079 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3080 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3081 	return len;
3082 #undef elen
3083 }
3084