xref: /freebsd/sys/netinet6/ip6_output.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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 
70 #include <sys/param.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/priv.h>
76 #include <sys/proc.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/syslog.h>
81 #include <sys/ucred.h>
82 
83 #include <net/if.h>
84 #include <net/netisr.h>
85 #include <net/route.h>
86 #include <net/pfil.h>
87 #include <net/vnet.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/tcp_var.h>
97 #include <netinet6/nd6.h>
98 
99 #ifdef IPSEC
100 #include <netipsec/ipsec.h>
101 #include <netipsec/ipsec6.h>
102 #include <netipsec/key.h>
103 #include <netinet6/ip6_ipsec.h>
104 #endif /* IPSEC */
105 
106 #include <netinet6/ip6protosw.h>
107 #include <netinet6/scope6_var.h>
108 
109 extern int in6_mcast_loop;
110 
111 struct ip6_exthdrs {
112 	struct mbuf *ip6e_ip6;
113 	struct mbuf *ip6e_hbh;
114 	struct mbuf *ip6e_dest1;
115 	struct mbuf *ip6e_rthdr;
116 	struct mbuf *ip6e_dest2;
117 };
118 
119 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
120 			   struct ucred *, int));
121 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
122 	struct socket *, struct sockopt *));
123 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
124 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
125 	struct ucred *, int, int, int));
126 
127 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
128 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
129 	struct ip6_frag **));
130 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
131 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
132 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
133 	struct ifnet *, struct in6_addr *, u_long *, int *));
134 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
135 
136 
137 /*
138  * Make an extension header from option data.  hp is the source, and
139  * mp is the destination.
140  */
141 #define MAKE_EXTHDR(hp, mp)						\
142     do {								\
143 	if (hp) {							\
144 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
145 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
146 		    ((eh)->ip6e_len + 1) << 3);				\
147 		if (error)						\
148 			goto freehdrs;					\
149 	}								\
150     } while (/*CONSTCOND*/ 0)
151 
152 /*
153  * Form a chain of extension headers.
154  * m is the extension header mbuf
155  * mp is the previous mbuf in the chain
156  * p is the next header
157  * i is the type of option.
158  */
159 #define MAKE_CHAIN(m, mp, p, i)\
160     do {\
161 	if (m) {\
162 		if (!hdrsplit) \
163 			panic("assumption failed: hdr not split"); \
164 		*mtod((m), u_char *) = *(p);\
165 		*(p) = (i);\
166 		p = mtod((m), u_char *);\
167 		(m)->m_next = (mp)->m_next;\
168 		(mp)->m_next = (m);\
169 		(mp) = (m);\
170 	}\
171     } while (/*CONSTCOND*/ 0)
172 
173 /*
174  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
175  * header (with pri, len, nxt, hlim, src, dst).
176  * This function may modify ver and hlim only.
177  * The mbuf chain containing the packet will be freed.
178  * The mbuf opt, if present, will not be freed.
179  *
180  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
181  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
182  * which is rt_rmx.rmx_mtu.
183  *
184  * ifpp - XXX: just for statistics
185  */
186 int
187 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
188     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
189     struct ifnet **ifpp, struct inpcb *inp)
190 {
191 	struct ip6_hdr *ip6, *mhip6;
192 	struct ifnet *ifp, *origifp;
193 	struct mbuf *m = m0;
194 	struct mbuf *mprev = NULL;
195 	int hlen, tlen, len, off;
196 	struct route_in6 ip6route;
197 	struct rtentry *rt = NULL;
198 	struct sockaddr_in6 *dst, src_sa, dst_sa;
199 	struct in6_addr odst;
200 	int error = 0;
201 	struct in6_ifaddr *ia = NULL;
202 	u_long mtu;
203 	int alwaysfrag, dontfrag;
204 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
205 	struct ip6_exthdrs exthdrs;
206 	struct in6_addr finaldst, src0, dst0;
207 	u_int32_t zone;
208 	struct route_in6 *ro_pmtu = NULL;
209 	int hdrsplit = 0;
210 	int needipsec = 0;
211 #ifdef IPSEC
212 	struct ipsec_output_state state;
213 	struct ip6_rthdr *rh = NULL;
214 	int needipsectun = 0;
215 	int segleft_org = 0;
216 	struct secpolicy *sp = NULL;
217 #endif /* IPSEC */
218 
219 	ip6 = mtod(m, struct ip6_hdr *);
220 	if (ip6 == NULL) {
221 		printf ("ip6 is NULL");
222 		goto bad;
223 	}
224 
225 	finaldst = ip6->ip6_dst;
226 
227 	bzero(&exthdrs, sizeof(exthdrs));
228 
229 	if (opt) {
230 		/* Hop-by-Hop options header */
231 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
232 		/* Destination options header(1st part) */
233 		if (opt->ip6po_rthdr) {
234 			/*
235 			 * Destination options header(1st part)
236 			 * This only makes sense with a routing header.
237 			 * See Section 9.2 of RFC 3542.
238 			 * Disabling this part just for MIP6 convenience is
239 			 * a bad idea.  We need to think carefully about a
240 			 * way to make the advanced API coexist with MIP6
241 			 * options, which might automatically be inserted in
242 			 * the kernel.
243 			 */
244 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
245 		}
246 		/* Routing header */
247 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
248 		/* Destination options header(2nd part) */
249 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
250 	}
251 
252 	/*
253 	 * IPSec checking which handles several cases.
254 	 * FAST IPSEC: We re-injected the packet.
255 	 */
256 #ifdef IPSEC
257 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
258 	{
259 	case 1:                 /* Bad packet */
260 		goto freehdrs;
261 	case -1:                /* Do IPSec */
262 		needipsec = 1;
263 	case 0:                 /* No IPSec */
264 	default:
265 		break;
266 	}
267 #endif /* IPSEC */
268 
269 	/*
270 	 * Calculate the total length of the extension header chain.
271 	 * Keep the length of the unfragmentable part for fragmentation.
272 	 */
273 	optlen = 0;
274 	if (exthdrs.ip6e_hbh)
275 		optlen += exthdrs.ip6e_hbh->m_len;
276 	if (exthdrs.ip6e_dest1)
277 		optlen += exthdrs.ip6e_dest1->m_len;
278 	if (exthdrs.ip6e_rthdr)
279 		optlen += exthdrs.ip6e_rthdr->m_len;
280 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
281 
282 	/* NOTE: we don't add AH/ESP length here. do that later. */
283 	if (exthdrs.ip6e_dest2)
284 		optlen += exthdrs.ip6e_dest2->m_len;
285 
286 	/*
287 	 * If we need IPsec, or there is at least one extension header,
288 	 * separate IP6 header from the payload.
289 	 */
290 	if ((needipsec || optlen) && !hdrsplit) {
291 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
292 			m = NULL;
293 			goto freehdrs;
294 		}
295 		m = exthdrs.ip6e_ip6;
296 		hdrsplit++;
297 	}
298 
299 	/* adjust pointer */
300 	ip6 = mtod(m, struct ip6_hdr *);
301 
302 	/* adjust mbuf packet header length */
303 	m->m_pkthdr.len += optlen;
304 	plen = m->m_pkthdr.len - sizeof(*ip6);
305 
306 	/* If this is a jumbo payload, insert a jumbo payload option. */
307 	if (plen > IPV6_MAXPACKET) {
308 		if (!hdrsplit) {
309 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
310 				m = NULL;
311 				goto freehdrs;
312 			}
313 			m = exthdrs.ip6e_ip6;
314 			hdrsplit++;
315 		}
316 		/* adjust pointer */
317 		ip6 = mtod(m, struct ip6_hdr *);
318 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
319 			goto freehdrs;
320 		ip6->ip6_plen = 0;
321 	} else
322 		ip6->ip6_plen = htons(plen);
323 
324 	/*
325 	 * Concatenate headers and fill in next header fields.
326 	 * Here we have, on "m"
327 	 *	IPv6 payload
328 	 * and we insert headers accordingly.  Finally, we should be getting:
329 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
330 	 *
331 	 * during the header composing process, "m" points to IPv6 header.
332 	 * "mprev" points to an extension header prior to esp.
333 	 */
334 	u_char *nexthdrp = &ip6->ip6_nxt;
335 	mprev = m;
336 
337 	/*
338 	 * we treat dest2 specially.  this makes IPsec processing
339 	 * much easier.  the goal here is to make mprev point the
340 	 * mbuf prior to dest2.
341 	 *
342 	 * result: IPv6 dest2 payload
343 	 * m and mprev will point to IPv6 header.
344 	 */
345 	if (exthdrs.ip6e_dest2) {
346 		if (!hdrsplit)
347 			panic("assumption failed: hdr not split");
348 		exthdrs.ip6e_dest2->m_next = m->m_next;
349 		m->m_next = exthdrs.ip6e_dest2;
350 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
351 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
352 	}
353 
354 	/*
355 	 * result: IPv6 hbh dest1 rthdr dest2 payload
356 	 * m will point to IPv6 header.  mprev will point to the
357 	 * extension header prior to dest2 (rthdr in the above case).
358 	 */
359 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
360 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
361 		   IPPROTO_DSTOPTS);
362 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
363 		   IPPROTO_ROUTING);
364 
365 #ifdef IPSEC
366 	if (!needipsec)
367 		goto skip_ipsec2;
368 
369 	/*
370 	 * pointers after IPsec headers are not valid any more.
371 	 * other pointers need a great care too.
372 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
373 	 */
374 	exthdrs.ip6e_dest2 = NULL;
375 
376 	if (exthdrs.ip6e_rthdr) {
377 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
378 		segleft_org = rh->ip6r_segleft;
379 		rh->ip6r_segleft = 0;
380 	}
381 
382 	bzero(&state, sizeof(state));
383 	state.m = m;
384 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
385 				    &needipsectun);
386 	m = state.m;
387 	if (error == EJUSTRETURN) {
388 		/*
389 		 * We had a SP with a level of 'use' and no SA. We
390 		 * will just continue to process the packet without
391 		 * IPsec processing.
392 		 */
393 		;
394 	} else if (error) {
395 		/* mbuf is already reclaimed in ipsec6_output_trans. */
396 		m = NULL;
397 		switch (error) {
398 		case EHOSTUNREACH:
399 		case ENETUNREACH:
400 		case EMSGSIZE:
401 		case ENOBUFS:
402 		case ENOMEM:
403 			break;
404 		default:
405 			printf("[%s:%d] (ipsec): error code %d\n",
406 			    __func__, __LINE__, error);
407 			/* FALLTHROUGH */
408 		case ENOENT:
409 			/* don't show these error codes to the user */
410 			error = 0;
411 			break;
412 		}
413 		goto bad;
414 	} else if (!needipsectun) {
415 		/*
416 		 * In the FAST IPSec case we have already
417 		 * re-injected the packet and it has been freed
418 		 * by the ipsec_done() function.  So, just clean
419 		 * up after ourselves.
420 		 */
421 		m = NULL;
422 		goto done;
423 	}
424 	if (exthdrs.ip6e_rthdr) {
425 		/* ah6_output doesn't modify mbuf chain */
426 		rh->ip6r_segleft = segleft_org;
427 	}
428 skip_ipsec2:;
429 #endif /* IPSEC */
430 
431 	/*
432 	 * If there is a routing header, discard the packet.
433 	 */
434 	if (exthdrs.ip6e_rthdr) {
435 		 error = EINVAL;
436 		 goto bad;
437 	}
438 
439 	/* Source address validation */
440 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
441 	    (flags & IPV6_UNSPECSRC) == 0) {
442 		error = EOPNOTSUPP;
443 		V_ip6stat.ip6s_badscope++;
444 		goto bad;
445 	}
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
447 		error = EOPNOTSUPP;
448 		V_ip6stat.ip6s_badscope++;
449 		goto bad;
450 	}
451 
452 	V_ip6stat.ip6s_localout++;
453 
454 	/*
455 	 * Route packet.
456 	 */
457 	if (ro == 0) {
458 		ro = &ip6route;
459 		bzero((caddr_t)ro, sizeof(*ro));
460 	}
461 	ro_pmtu = ro;
462 	if (opt && opt->ip6po_rthdr)
463 		ro = &opt->ip6po_route;
464 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
465 
466 again:
467 	/*
468 	 * if specified, try to fill in the traffic class field.
469 	 * do not override if a non-zero value is already set.
470 	 * we check the diffserv field and the ecn field separately.
471 	 */
472 	if (opt && opt->ip6po_tclass >= 0) {
473 		int mask = 0;
474 
475 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
476 			mask |= 0xfc;
477 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
478 			mask |= 0x03;
479 		if (mask != 0)
480 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
481 	}
482 
483 	/* fill in or override the hop limit field, if necessary. */
484 	if (opt && opt->ip6po_hlim != -1)
485 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
486 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
487 		if (im6o != NULL)
488 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
489 		else
490 			ip6->ip6_hlim = V_ip6_defmcasthlim;
491 	}
492 
493 #ifdef IPSEC
494 	/*
495 	 * We may re-inject packets into the stack here.
496 	 */
497 	if (needipsec && needipsectun) {
498 		struct ipsec_output_state state;
499 
500 		/*
501 		 * All the extension headers will become inaccessible
502 		 * (since they can be encrypted).
503 		 * Don't panic, we need no more updates to extension headers
504 		 * on inner IPv6 packet (since they are now encapsulated).
505 		 *
506 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
507 		 */
508 		bzero(&exthdrs, sizeof(exthdrs));
509 		exthdrs.ip6e_ip6 = m;
510 
511 		bzero(&state, sizeof(state));
512 		state.m = m;
513 		state.ro = (struct route *)ro;
514 		state.dst = (struct sockaddr *)dst;
515 
516 		error = ipsec6_output_tunnel(&state, sp, flags);
517 
518 		m = state.m;
519 		ro = (struct route_in6 *)state.ro;
520 		dst = (struct sockaddr_in6 *)state.dst;
521 		if (error == EJUSTRETURN) {
522 			/*
523 			 * We had a SP with a level of 'use' and no SA. We
524 			 * will just continue to process the packet without
525 			 * IPsec processing.
526 			 */
527 			;
528 		} else if (error) {
529 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
530 			m0 = m = NULL;
531 			m = NULL;
532 			switch (error) {
533 			case EHOSTUNREACH:
534 			case ENETUNREACH:
535 			case EMSGSIZE:
536 			case ENOBUFS:
537 			case ENOMEM:
538 				break;
539 			default:
540 				printf("[%s:%d] (ipsec): error code %d\n",
541 				    __func__, __LINE__, error);
542 				/* FALLTHROUGH */
543 			case ENOENT:
544 				/* don't show these error codes to the user */
545 				error = 0;
546 				break;
547 			}
548 			goto bad;
549 		} else {
550 			/*
551 			 * In the FAST IPSec case we have already
552 			 * re-injected the packet and it has been freed
553 			 * by the ipsec_done() function.  So, just clean
554 			 * up after ourselves.
555 			 */
556 			m = NULL;
557 			goto done;
558 		}
559 
560 		exthdrs.ip6e_ip6 = m;
561 	}
562 #endif /* IPSEC */
563 
564 	/* adjust pointer */
565 	ip6 = mtod(m, struct ip6_hdr *);
566 
567 	bzero(&dst_sa, sizeof(dst_sa));
568 	dst_sa.sin6_family = AF_INET6;
569 	dst_sa.sin6_len = sizeof(dst_sa);
570 	dst_sa.sin6_addr = ip6->ip6_dst;
571 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
572 	    &ifp, &rt)) != 0) {
573 		switch (error) {
574 		case EHOSTUNREACH:
575 			V_ip6stat.ip6s_noroute++;
576 			break;
577 		case EADDRNOTAVAIL:
578 		default:
579 			break; /* XXX statistics? */
580 		}
581 		if (ifp != NULL)
582 			in6_ifstat_inc(ifp, ifs6_out_discard);
583 		goto bad;
584 	}
585 	if (rt == NULL) {
586 		/*
587 		 * If in6_selectroute() does not return a route entry,
588 		 * dst may not have been updated.
589 		 */
590 		*dst = dst_sa;	/* XXX */
591 	}
592 
593 	/*
594 	 * then rt (for unicast) and ifp must be non-NULL valid values.
595 	 */
596 	if ((flags & IPV6_FORWARDING) == 0) {
597 		/* XXX: the FORWARDING flag can be set for mrouting. */
598 		in6_ifstat_inc(ifp, ifs6_out_request);
599 	}
600 	if (rt != NULL) {
601 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
602 		rt->rt_use++;
603 	}
604 
605 	/*
606 	 * The outgoing interface must be in the zone of source and
607 	 * destination addresses.  We should use ia_ifp to support the
608 	 * case of sending packets to an address of our own.
609 	 */
610 	if (ia != NULL && ia->ia_ifp)
611 		origifp = ia->ia_ifp;
612 	else
613 		origifp = ifp;
614 
615 	src0 = ip6->ip6_src;
616 	if (in6_setscope(&src0, origifp, &zone))
617 		goto badscope;
618 	bzero(&src_sa, sizeof(src_sa));
619 	src_sa.sin6_family = AF_INET6;
620 	src_sa.sin6_len = sizeof(src_sa);
621 	src_sa.sin6_addr = ip6->ip6_src;
622 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
623 		goto badscope;
624 
625 	dst0 = ip6->ip6_dst;
626 	if (in6_setscope(&dst0, origifp, &zone))
627 		goto badscope;
628 	/* re-initialize to be sure */
629 	bzero(&dst_sa, sizeof(dst_sa));
630 	dst_sa.sin6_family = AF_INET6;
631 	dst_sa.sin6_len = sizeof(dst_sa);
632 	dst_sa.sin6_addr = ip6->ip6_dst;
633 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
634 		goto badscope;
635 	}
636 
637 	/* scope check is done. */
638 	goto routefound;
639 
640   badscope:
641 	V_ip6stat.ip6s_badscope++;
642 	in6_ifstat_inc(origifp, ifs6_out_discard);
643 	if (error == 0)
644 		error = EHOSTUNREACH; /* XXX */
645 	goto bad;
646 
647   routefound:
648 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
649 		if (opt && opt->ip6po_nextroute.ro_rt) {
650 			/*
651 			 * The nexthop is explicitly specified by the
652 			 * application.  We assume the next hop is an IPv6
653 			 * address.
654 			 */
655 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
656 		}
657 		else if ((rt->rt_flags & RTF_GATEWAY))
658 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
659 	}
660 
661 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
662 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
663 	} else {
664 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
665 		in6_ifstat_inc(ifp, ifs6_out_mcast);
666 		/*
667 		 * Confirm that the outgoing interface supports multicast.
668 		 */
669 		if (!(ifp->if_flags & IFF_MULTICAST)) {
670 			V_ip6stat.ip6s_noroute++;
671 			in6_ifstat_inc(ifp, ifs6_out_discard);
672 			error = ENETUNREACH;
673 			goto bad;
674 		}
675 		if ((im6o == NULL && in6_mcast_loop) ||
676 		    (im6o && im6o->im6o_multicast_loop)) {
677 			/*
678 			 * Loop back multicast datagram if not expressly
679 			 * forbidden to do so, even if we have not joined
680 			 * the address; protocols will filter it later,
681 			 * thus deferring a hash lookup and lock acquisition
682 			 * at the expense of an m_copym().
683 			 */
684 			ip6_mloopback(ifp, m, dst);
685 		} else {
686 			/*
687 			 * If we are acting as a multicast router, perform
688 			 * multicast forwarding as if the packet had just
689 			 * arrived on the interface to which we are about
690 			 * to send.  The multicast forwarding function
691 			 * recursively calls this function, using the
692 			 * IPV6_FORWARDING flag to prevent infinite recursion.
693 			 *
694 			 * Multicasts that are looped back by ip6_mloopback(),
695 			 * above, will be forwarded by the ip6_input() routine,
696 			 * if necessary.
697 			 */
698 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
699 				/*
700 				 * XXX: ip6_mforward expects that rcvif is NULL
701 				 * when it is called from the originating path.
702 				 * However, it is not always the case, since
703 				 * some versions of MGETHDR() does not
704 				 * initialize the field.
705 				 */
706 				m->m_pkthdr.rcvif = NULL;
707 				if (ip6_mforward(ip6, ifp, m) != 0) {
708 					m_freem(m);
709 					goto done;
710 				}
711 			}
712 		}
713 		/*
714 		 * Multicasts with a hoplimit of zero may be looped back,
715 		 * above, but must not be transmitted on a network.
716 		 * Also, multicasts addressed to the loopback interface
717 		 * are not sent -- the above call to ip6_mloopback() will
718 		 * loop back a copy if this host actually belongs to the
719 		 * destination group on the loopback interface.
720 		 */
721 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
722 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
723 			m_freem(m);
724 			goto done;
725 		}
726 	}
727 
728 	/*
729 	 * Fill the outgoing inteface to tell the upper layer
730 	 * to increment per-interface statistics.
731 	 */
732 	if (ifpp)
733 		*ifpp = ifp;
734 
735 	/* Determine path MTU. */
736 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
737 	    &alwaysfrag)) != 0)
738 		goto bad;
739 
740 	/*
741 	 * The caller of this function may specify to use the minimum MTU
742 	 * in some cases.
743 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
744 	 * setting.  The logic is a bit complicated; by default, unicast
745 	 * packets will follow path MTU while multicast packets will be sent at
746 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
747 	 * including unicast ones will be sent at the minimum MTU.  Multicast
748 	 * packets will always be sent at the minimum MTU unless
749 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
750 	 * See RFC 3542 for more details.
751 	 */
752 	if (mtu > IPV6_MMTU) {
753 		if ((flags & IPV6_MINMTU))
754 			mtu = IPV6_MMTU;
755 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
756 			mtu = IPV6_MMTU;
757 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
758 			 (opt == NULL ||
759 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
760 			mtu = IPV6_MMTU;
761 		}
762 	}
763 
764 	/*
765 	 * clear embedded scope identifiers if necessary.
766 	 * in6_clearscope will touch the addresses only when necessary.
767 	 */
768 	in6_clearscope(&ip6->ip6_src);
769 	in6_clearscope(&ip6->ip6_dst);
770 
771 	/*
772 	 * If the outgoing packet contains a hop-by-hop options header,
773 	 * it must be examined and processed even by the source node.
774 	 * (RFC 2460, section 4.)
775 	 */
776 	if (exthdrs.ip6e_hbh) {
777 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
778 		u_int32_t dummy; /* XXX unused */
779 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
780 
781 #ifdef DIAGNOSTIC
782 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
783 			panic("ip6e_hbh is not continuous");
784 #endif
785 		/*
786 		 *  XXX: if we have to send an ICMPv6 error to the sender,
787 		 *       we need the M_LOOP flag since icmp6_error() expects
788 		 *       the IPv6 and the hop-by-hop options header are
789 		 *       continuous unless the flag is set.
790 		 */
791 		m->m_flags |= M_LOOP;
792 		m->m_pkthdr.rcvif = ifp;
793 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
794 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
795 		    &dummy, &plen) < 0) {
796 			/* m was already freed at this point */
797 			error = EINVAL;/* better error? */
798 			goto done;
799 		}
800 		m->m_flags &= ~M_LOOP; /* XXX */
801 		m->m_pkthdr.rcvif = NULL;
802 	}
803 
804 	/* Jump over all PFIL processing if hooks are not active. */
805 	if (!PFIL_HOOKED(&inet6_pfil_hook))
806 		goto passout;
807 
808 	odst = ip6->ip6_dst;
809 	/* Run through list of hooks for output packets. */
810 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
811 	if (error != 0 || m == NULL)
812 		goto done;
813 	ip6 = mtod(m, struct ip6_hdr *);
814 
815 	/* See if destination IP address was changed by packet filter. */
816 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
817 		m->m_flags |= M_SKIP_FIREWALL;
818 		/* If destination is now ourself drop to ip6_input(). */
819 		if (in6_localaddr(&ip6->ip6_dst)) {
820 			if (m->m_pkthdr.rcvif == NULL)
821 				m->m_pkthdr.rcvif = V_loif;
822 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
823 				m->m_pkthdr.csum_flags |=
824 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
825 				m->m_pkthdr.csum_data = 0xffff;
826 			}
827 			m->m_pkthdr.csum_flags |=
828 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
829 			error = netisr_queue(NETISR_IPV6, m);
830 			goto done;
831 		} else
832 			goto again;	/* Redo the routing table lookup. */
833 	}
834 
835 	/* XXX: IPFIREWALL_FORWARD */
836 
837 passout:
838 	/*
839 	 * Send the packet to the outgoing interface.
840 	 * If necessary, do IPv6 fragmentation before sending.
841 	 *
842 	 * the logic here is rather complex:
843 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
844 	 * 1-a:	send as is if tlen <= path mtu
845 	 * 1-b:	fragment if tlen > path mtu
846 	 *
847 	 * 2: if user asks us not to fragment (dontfrag == 1)
848 	 * 2-a:	send as is if tlen <= interface mtu
849 	 * 2-b:	error if tlen > interface mtu
850 	 *
851 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
852 	 *	always fragment
853 	 *
854 	 * 4: if dontfrag == 1 && alwaysfrag == 1
855 	 *	error, as we cannot handle this conflicting request
856 	 */
857 	tlen = m->m_pkthdr.len;
858 
859 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
860 		dontfrag = 1;
861 	else
862 		dontfrag = 0;
863 	if (dontfrag && alwaysfrag) {	/* case 4 */
864 		/* conflicting request - can't transmit */
865 		error = EMSGSIZE;
866 		goto bad;
867 	}
868 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
869 		/*
870 		 * Even if the DONTFRAG option is specified, we cannot send the
871 		 * packet when the data length is larger than the MTU of the
872 		 * outgoing interface.
873 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
874 		 * well as returning an error code (the latter is not described
875 		 * in the API spec.)
876 		 */
877 		u_int32_t mtu32;
878 		struct ip6ctlparam ip6cp;
879 
880 		mtu32 = (u_int32_t)mtu;
881 		bzero(&ip6cp, sizeof(ip6cp));
882 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
883 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
884 		    (void *)&ip6cp);
885 
886 		error = EMSGSIZE;
887 		goto bad;
888 	}
889 
890 	/*
891 	 * transmit packet without fragmentation
892 	 */
893 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
894 		struct in6_ifaddr *ia6;
895 
896 		ip6 = mtod(m, struct ip6_hdr *);
897 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
898 		if (ia6) {
899 			/* Record statistics for this interface address. */
900 			ia6->ia_ifa.if_opackets++;
901 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
902 			ifa_free(&ia6->ia_ifa);
903 		}
904 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
905 		goto done;
906 	}
907 
908 	/*
909 	 * try to fragment the packet.  case 1-b and 3
910 	 */
911 	if (mtu < IPV6_MMTU) {
912 		/* path MTU cannot be less than IPV6_MMTU */
913 		error = EMSGSIZE;
914 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
915 		goto bad;
916 	} else if (ip6->ip6_plen == 0) {
917 		/* jumbo payload cannot be fragmented */
918 		error = EMSGSIZE;
919 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
920 		goto bad;
921 	} else {
922 		struct mbuf **mnext, *m_frgpart;
923 		struct ip6_frag *ip6f;
924 		u_int32_t id = htonl(ip6_randomid());
925 		u_char nextproto;
926 
927 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
928 
929 		/*
930 		 * Too large for the destination or interface;
931 		 * fragment if possible.
932 		 * Must be able to put at least 8 bytes per fragment.
933 		 */
934 		hlen = unfragpartlen;
935 		if (mtu > IPV6_MAXPACKET)
936 			mtu = IPV6_MAXPACKET;
937 
938 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
939 		if (len < 8) {
940 			error = EMSGSIZE;
941 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
942 			goto bad;
943 		}
944 
945 		/*
946 		 * Verify that we have any chance at all of being able to queue
947 		 *      the packet or packet fragments
948 		 */
949 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
950 		    < tlen  /* - hlen */)) {
951 			error = ENOBUFS;
952 			V_ip6stat.ip6s_odropped++;
953 			goto bad;
954 		}
955 
956 		mnext = &m->m_nextpkt;
957 
958 		/*
959 		 * Change the next header field of the last header in the
960 		 * unfragmentable part.
961 		 */
962 		if (exthdrs.ip6e_rthdr) {
963 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
964 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
965 		} else if (exthdrs.ip6e_dest1) {
966 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
967 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
968 		} else if (exthdrs.ip6e_hbh) {
969 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
970 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
971 		} else {
972 			nextproto = ip6->ip6_nxt;
973 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
974 		}
975 
976 		/*
977 		 * Loop through length of segment after first fragment,
978 		 * make new header and copy data of each part and link onto
979 		 * chain.
980 		 */
981 		m0 = m;
982 		for (off = hlen; off < tlen; off += len) {
983 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
984 			if (!m) {
985 				error = ENOBUFS;
986 				V_ip6stat.ip6s_odropped++;
987 				goto sendorfree;
988 			}
989 			m->m_pkthdr.rcvif = NULL;
990 			m->m_flags = m0->m_flags & M_COPYFLAGS;
991 			*mnext = m;
992 			mnext = &m->m_nextpkt;
993 			m->m_data += max_linkhdr;
994 			mhip6 = mtod(m, struct ip6_hdr *);
995 			*mhip6 = *ip6;
996 			m->m_len = sizeof(*mhip6);
997 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
998 			if (error) {
999 				V_ip6stat.ip6s_odropped++;
1000 				goto sendorfree;
1001 			}
1002 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1003 			if (off + len >= tlen)
1004 				len = tlen - off;
1005 			else
1006 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1007 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1008 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1009 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1010 				error = ENOBUFS;
1011 				V_ip6stat.ip6s_odropped++;
1012 				goto sendorfree;
1013 			}
1014 			m_cat(m, m_frgpart);
1015 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1016 			m->m_pkthdr.rcvif = NULL;
1017 			ip6f->ip6f_reserved = 0;
1018 			ip6f->ip6f_ident = id;
1019 			ip6f->ip6f_nxt = nextproto;
1020 			V_ip6stat.ip6s_ofragments++;
1021 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1022 		}
1023 
1024 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1025 	}
1026 
1027 	/*
1028 	 * Remove leading garbages.
1029 	 */
1030 sendorfree:
1031 	m = m0->m_nextpkt;
1032 	m0->m_nextpkt = 0;
1033 	m_freem(m0);
1034 	for (m0 = m; m; m = m0) {
1035 		m0 = m->m_nextpkt;
1036 		m->m_nextpkt = 0;
1037 		if (error == 0) {
1038 			/* Record statistics for this interface address. */
1039 			if (ia) {
1040 				ia->ia_ifa.if_opackets++;
1041 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1042 			}
1043 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1044 		} else
1045 			m_freem(m);
1046 	}
1047 
1048 	if (error == 0)
1049 		V_ip6stat.ip6s_fragmented++;
1050 
1051 done:
1052 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1053 		RTFREE(ro->ro_rt);
1054 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1055 		RTFREE(ro_pmtu->ro_rt);
1056 	}
1057 #ifdef IPSEC
1058 	if (sp != NULL)
1059 		KEY_FREESP(&sp);
1060 #endif
1061 
1062 	return (error);
1063 
1064 freehdrs:
1065 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1066 	m_freem(exthdrs.ip6e_dest1);
1067 	m_freem(exthdrs.ip6e_rthdr);
1068 	m_freem(exthdrs.ip6e_dest2);
1069 	/* FALLTHROUGH */
1070 bad:
1071 	if (m)
1072 		m_freem(m);
1073 	goto done;
1074 }
1075 
1076 static int
1077 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1078 {
1079 	struct mbuf *m;
1080 
1081 	if (hlen > MCLBYTES)
1082 		return (ENOBUFS); /* XXX */
1083 
1084 	MGET(m, M_DONTWAIT, MT_DATA);
1085 	if (!m)
1086 		return (ENOBUFS);
1087 
1088 	if (hlen > MLEN) {
1089 		MCLGET(m, M_DONTWAIT);
1090 		if ((m->m_flags & M_EXT) == 0) {
1091 			m_free(m);
1092 			return (ENOBUFS);
1093 		}
1094 	}
1095 	m->m_len = hlen;
1096 	if (hdr)
1097 		bcopy(hdr, mtod(m, caddr_t), hlen);
1098 
1099 	*mp = m;
1100 	return (0);
1101 }
1102 
1103 /*
1104  * Insert jumbo payload option.
1105  */
1106 static int
1107 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1108 {
1109 	struct mbuf *mopt;
1110 	u_char *optbuf;
1111 	u_int32_t v;
1112 
1113 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1114 
1115 	/*
1116 	 * If there is no hop-by-hop options header, allocate new one.
1117 	 * If there is one but it doesn't have enough space to store the
1118 	 * jumbo payload option, allocate a cluster to store the whole options.
1119 	 * Otherwise, use it to store the options.
1120 	 */
1121 	if (exthdrs->ip6e_hbh == 0) {
1122 		MGET(mopt, M_DONTWAIT, MT_DATA);
1123 		if (mopt == 0)
1124 			return (ENOBUFS);
1125 		mopt->m_len = JUMBOOPTLEN;
1126 		optbuf = mtod(mopt, u_char *);
1127 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1128 		exthdrs->ip6e_hbh = mopt;
1129 	} else {
1130 		struct ip6_hbh *hbh;
1131 
1132 		mopt = exthdrs->ip6e_hbh;
1133 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1134 			/*
1135 			 * XXX assumption:
1136 			 * - exthdrs->ip6e_hbh is not referenced from places
1137 			 *   other than exthdrs.
1138 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1139 			 */
1140 			int oldoptlen = mopt->m_len;
1141 			struct mbuf *n;
1142 
1143 			/*
1144 			 * XXX: give up if the whole (new) hbh header does
1145 			 * not fit even in an mbuf cluster.
1146 			 */
1147 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1148 				return (ENOBUFS);
1149 
1150 			/*
1151 			 * As a consequence, we must always prepare a cluster
1152 			 * at this point.
1153 			 */
1154 			MGET(n, M_DONTWAIT, MT_DATA);
1155 			if (n) {
1156 				MCLGET(n, M_DONTWAIT);
1157 				if ((n->m_flags & M_EXT) == 0) {
1158 					m_freem(n);
1159 					n = NULL;
1160 				}
1161 			}
1162 			if (!n)
1163 				return (ENOBUFS);
1164 			n->m_len = oldoptlen + JUMBOOPTLEN;
1165 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1166 			    oldoptlen);
1167 			optbuf = mtod(n, caddr_t) + oldoptlen;
1168 			m_freem(mopt);
1169 			mopt = exthdrs->ip6e_hbh = n;
1170 		} else {
1171 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1172 			mopt->m_len += JUMBOOPTLEN;
1173 		}
1174 		optbuf[0] = IP6OPT_PADN;
1175 		optbuf[1] = 1;
1176 
1177 		/*
1178 		 * Adjust the header length according to the pad and
1179 		 * the jumbo payload option.
1180 		 */
1181 		hbh = mtod(mopt, struct ip6_hbh *);
1182 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1183 	}
1184 
1185 	/* fill in the option. */
1186 	optbuf[2] = IP6OPT_JUMBO;
1187 	optbuf[3] = 4;
1188 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1189 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1190 
1191 	/* finally, adjust the packet header length */
1192 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1193 
1194 	return (0);
1195 #undef JUMBOOPTLEN
1196 }
1197 
1198 /*
1199  * Insert fragment header and copy unfragmentable header portions.
1200  */
1201 static int
1202 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1203     struct ip6_frag **frghdrp)
1204 {
1205 	struct mbuf *n, *mlast;
1206 
1207 	if (hlen > sizeof(struct ip6_hdr)) {
1208 		n = m_copym(m0, sizeof(struct ip6_hdr),
1209 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1210 		if (n == 0)
1211 			return (ENOBUFS);
1212 		m->m_next = n;
1213 	} else
1214 		n = m;
1215 
1216 	/* Search for the last mbuf of unfragmentable part. */
1217 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1218 		;
1219 
1220 	if ((mlast->m_flags & M_EXT) == 0 &&
1221 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1222 		/* use the trailing space of the last mbuf for the fragment hdr */
1223 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1224 		    mlast->m_len);
1225 		mlast->m_len += sizeof(struct ip6_frag);
1226 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1227 	} else {
1228 		/* allocate a new mbuf for the fragment header */
1229 		struct mbuf *mfrg;
1230 
1231 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1232 		if (mfrg == 0)
1233 			return (ENOBUFS);
1234 		mfrg->m_len = sizeof(struct ip6_frag);
1235 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1236 		mlast->m_next = mfrg;
1237 	}
1238 
1239 	return (0);
1240 }
1241 
1242 static int
1243 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1244     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1245     int *alwaysfragp)
1246 {
1247 	u_int32_t mtu = 0;
1248 	int alwaysfrag = 0;
1249 	int error = 0;
1250 
1251 	if (ro_pmtu != ro) {
1252 		/* The first hop and the final destination may differ. */
1253 		struct sockaddr_in6 *sa6_dst =
1254 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1255 		if (ro_pmtu->ro_rt &&
1256 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1257 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1258 			RTFREE(ro_pmtu->ro_rt);
1259 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1260 		}
1261 		if (ro_pmtu->ro_rt == NULL) {
1262 			bzero(sa6_dst, sizeof(*sa6_dst));
1263 			sa6_dst->sin6_family = AF_INET6;
1264 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1265 			sa6_dst->sin6_addr = *dst;
1266 
1267 			rtalloc((struct route *)ro_pmtu);
1268 		}
1269 	}
1270 	if (ro_pmtu->ro_rt) {
1271 		u_int32_t ifmtu;
1272 		struct in_conninfo inc;
1273 
1274 		bzero(&inc, sizeof(inc));
1275 		inc.inc_flags |= INC_ISIPV6;
1276 		inc.inc6_faddr = *dst;
1277 
1278 		if (ifp == NULL)
1279 			ifp = ro_pmtu->ro_rt->rt_ifp;
1280 		ifmtu = IN6_LINKMTU(ifp);
1281 		mtu = tcp_hc_getmtu(&inc);
1282 		if (mtu)
1283 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1284 		else
1285 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1286 		if (mtu == 0)
1287 			mtu = ifmtu;
1288 		else if (mtu < IPV6_MMTU) {
1289 			/*
1290 			 * RFC2460 section 5, last paragraph:
1291 			 * if we record ICMPv6 too big message with
1292 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1293 			 * or smaller, with framgent header attached.
1294 			 * (fragment header is needed regardless from the
1295 			 * packet size, for translators to identify packets)
1296 			 */
1297 			alwaysfrag = 1;
1298 			mtu = IPV6_MMTU;
1299 		} else if (mtu > ifmtu) {
1300 			/*
1301 			 * The MTU on the route is larger than the MTU on
1302 			 * the interface!  This shouldn't happen, unless the
1303 			 * MTU of the interface has been changed after the
1304 			 * interface was brought up.  Change the MTU in the
1305 			 * route to match the interface MTU (as long as the
1306 			 * field isn't locked).
1307 			 */
1308 			mtu = ifmtu;
1309 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1310 		}
1311 	} else if (ifp) {
1312 		mtu = IN6_LINKMTU(ifp);
1313 	} else
1314 		error = EHOSTUNREACH; /* XXX */
1315 
1316 	*mtup = mtu;
1317 	if (alwaysfragp)
1318 		*alwaysfragp = alwaysfrag;
1319 	return (error);
1320 }
1321 
1322 /*
1323  * IP6 socket option processing.
1324  */
1325 int
1326 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1327 {
1328 	int optdatalen, uproto;
1329 	void *optdata;
1330 	struct inpcb *in6p = sotoinpcb(so);
1331 	int error, optval;
1332 	int level, op, optname;
1333 	int optlen;
1334 	struct thread *td;
1335 
1336 	level = sopt->sopt_level;
1337 	op = sopt->sopt_dir;
1338 	optname = sopt->sopt_name;
1339 	optlen = sopt->sopt_valsize;
1340 	td = sopt->sopt_td;
1341 	error = 0;
1342 	optval = 0;
1343 	uproto = (int)so->so_proto->pr_protocol;
1344 
1345 	if (level == IPPROTO_IPV6) {
1346 		switch (op) {
1347 
1348 		case SOPT_SET:
1349 			switch (optname) {
1350 			case IPV6_2292PKTOPTIONS:
1351 #ifdef IPV6_PKTOPTIONS
1352 			case IPV6_PKTOPTIONS:
1353 #endif
1354 			{
1355 				struct mbuf *m;
1356 
1357 				error = soopt_getm(sopt, &m); /* XXX */
1358 				if (error != 0)
1359 					break;
1360 				error = soopt_mcopyin(sopt, m); /* XXX */
1361 				if (error != 0)
1362 					break;
1363 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1364 						    m, so, sopt);
1365 				m_freem(m); /* XXX */
1366 				break;
1367 			}
1368 
1369 			/*
1370 			 * Use of some Hop-by-Hop options or some
1371 			 * Destination options, might require special
1372 			 * privilege.  That is, normal applications
1373 			 * (without special privilege) might be forbidden
1374 			 * from setting certain options in outgoing packets,
1375 			 * and might never see certain options in received
1376 			 * packets. [RFC 2292 Section 6]
1377 			 * KAME specific note:
1378 			 *  KAME prevents non-privileged users from sending or
1379 			 *  receiving ANY hbh/dst options in order to avoid
1380 			 *  overhead of parsing options in the kernel.
1381 			 */
1382 			case IPV6_RECVHOPOPTS:
1383 			case IPV6_RECVDSTOPTS:
1384 			case IPV6_RECVRTHDRDSTOPTS:
1385 				if (td != NULL) {
1386 					error = priv_check(td,
1387 					    PRIV_NETINET_SETHDROPTS);
1388 					if (error)
1389 						break;
1390 				}
1391 				/* FALLTHROUGH */
1392 			case IPV6_UNICAST_HOPS:
1393 			case IPV6_HOPLIMIT:
1394 			case IPV6_FAITH:
1395 
1396 			case IPV6_RECVPKTINFO:
1397 			case IPV6_RECVHOPLIMIT:
1398 			case IPV6_RECVRTHDR:
1399 			case IPV6_RECVPATHMTU:
1400 			case IPV6_RECVTCLASS:
1401 			case IPV6_V6ONLY:
1402 			case IPV6_AUTOFLOWLABEL:
1403 			case IPV6_BINDANY:
1404 				if (optname == IPV6_BINDANY && td != NULL) {
1405 					error = priv_check(td,
1406 					    PRIV_NETINET_BINDANY);
1407 					if (error)
1408 						break;
1409 				}
1410 
1411 				if (optlen != sizeof(int)) {
1412 					error = EINVAL;
1413 					break;
1414 				}
1415 				error = sooptcopyin(sopt, &optval,
1416 					sizeof optval, sizeof optval);
1417 				if (error)
1418 					break;
1419 				switch (optname) {
1420 
1421 				case IPV6_UNICAST_HOPS:
1422 					if (optval < -1 || optval >= 256)
1423 						error = EINVAL;
1424 					else {
1425 						/* -1 = kernel default */
1426 						in6p->in6p_hops = optval;
1427 						if ((in6p->inp_vflag &
1428 						     INP_IPV4) != 0)
1429 							in6p->inp_ip_ttl = optval;
1430 					}
1431 					break;
1432 #define OPTSET(bit) \
1433 do { \
1434 	if (optval) \
1435 		in6p->inp_flags |= (bit); \
1436 	else \
1437 		in6p->inp_flags &= ~(bit); \
1438 } while (/*CONSTCOND*/ 0)
1439 #define OPTSET2292(bit) \
1440 do { \
1441 	in6p->inp_flags |= IN6P_RFC2292; \
1442 	if (optval) \
1443 		in6p->inp_flags |= (bit); \
1444 	else \
1445 		in6p->inp_flags &= ~(bit); \
1446 } while (/*CONSTCOND*/ 0)
1447 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1448 
1449 				case IPV6_RECVPKTINFO:
1450 					/* cannot mix with RFC2292 */
1451 					if (OPTBIT(IN6P_RFC2292)) {
1452 						error = EINVAL;
1453 						break;
1454 					}
1455 					OPTSET(IN6P_PKTINFO);
1456 					break;
1457 
1458 				case IPV6_HOPLIMIT:
1459 				{
1460 					struct ip6_pktopts **optp;
1461 
1462 					/* cannot mix with RFC2292 */
1463 					if (OPTBIT(IN6P_RFC2292)) {
1464 						error = EINVAL;
1465 						break;
1466 					}
1467 					optp = &in6p->in6p_outputopts;
1468 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1469 					    (u_char *)&optval, sizeof(optval),
1470 					    optp, (td != NULL) ? td->td_ucred :
1471 					    NULL, uproto);
1472 					break;
1473 				}
1474 
1475 				case IPV6_RECVHOPLIMIT:
1476 					/* cannot mix with RFC2292 */
1477 					if (OPTBIT(IN6P_RFC2292)) {
1478 						error = EINVAL;
1479 						break;
1480 					}
1481 					OPTSET(IN6P_HOPLIMIT);
1482 					break;
1483 
1484 				case IPV6_RECVHOPOPTS:
1485 					/* cannot mix with RFC2292 */
1486 					if (OPTBIT(IN6P_RFC2292)) {
1487 						error = EINVAL;
1488 						break;
1489 					}
1490 					OPTSET(IN6P_HOPOPTS);
1491 					break;
1492 
1493 				case IPV6_RECVDSTOPTS:
1494 					/* cannot mix with RFC2292 */
1495 					if (OPTBIT(IN6P_RFC2292)) {
1496 						error = EINVAL;
1497 						break;
1498 					}
1499 					OPTSET(IN6P_DSTOPTS);
1500 					break;
1501 
1502 				case IPV6_RECVRTHDRDSTOPTS:
1503 					/* cannot mix with RFC2292 */
1504 					if (OPTBIT(IN6P_RFC2292)) {
1505 						error = EINVAL;
1506 						break;
1507 					}
1508 					OPTSET(IN6P_RTHDRDSTOPTS);
1509 					break;
1510 
1511 				case IPV6_RECVRTHDR:
1512 					/* cannot mix with RFC2292 */
1513 					if (OPTBIT(IN6P_RFC2292)) {
1514 						error = EINVAL;
1515 						break;
1516 					}
1517 					OPTSET(IN6P_RTHDR);
1518 					break;
1519 
1520 				case IPV6_FAITH:
1521 					OPTSET(INP_FAITH);
1522 					break;
1523 
1524 				case IPV6_RECVPATHMTU:
1525 					/*
1526 					 * We ignore this option for TCP
1527 					 * sockets.
1528 					 * (RFC3542 leaves this case
1529 					 * unspecified.)
1530 					 */
1531 					if (uproto != IPPROTO_TCP)
1532 						OPTSET(IN6P_MTU);
1533 					break;
1534 
1535 				case IPV6_V6ONLY:
1536 					/*
1537 					 * make setsockopt(IPV6_V6ONLY)
1538 					 * available only prior to bind(2).
1539 					 * see ipng mailing list, Jun 22 2001.
1540 					 */
1541 					if (in6p->inp_lport ||
1542 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1543 						error = EINVAL;
1544 						break;
1545 					}
1546 					OPTSET(IN6P_IPV6_V6ONLY);
1547 					if (optval)
1548 						in6p->inp_vflag &= ~INP_IPV4;
1549 					else
1550 						in6p->inp_vflag |= INP_IPV4;
1551 					break;
1552 				case IPV6_RECVTCLASS:
1553 					/* cannot mix with RFC2292 XXX */
1554 					if (OPTBIT(IN6P_RFC2292)) {
1555 						error = EINVAL;
1556 						break;
1557 					}
1558 					OPTSET(IN6P_TCLASS);
1559 					break;
1560 				case IPV6_AUTOFLOWLABEL:
1561 					OPTSET(IN6P_AUTOFLOWLABEL);
1562 					break;
1563 
1564 				case IPV6_BINDANY:
1565 					OPTSET(INP_BINDANY);
1566 					break;
1567 				}
1568 				break;
1569 
1570 			case IPV6_TCLASS:
1571 			case IPV6_DONTFRAG:
1572 			case IPV6_USE_MIN_MTU:
1573 			case IPV6_PREFER_TEMPADDR:
1574 				if (optlen != sizeof(optval)) {
1575 					error = EINVAL;
1576 					break;
1577 				}
1578 				error = sooptcopyin(sopt, &optval,
1579 					sizeof optval, sizeof optval);
1580 				if (error)
1581 					break;
1582 				{
1583 					struct ip6_pktopts **optp;
1584 					optp = &in6p->in6p_outputopts;
1585 					error = ip6_pcbopt(optname,
1586 					    (u_char *)&optval, sizeof(optval),
1587 					    optp, (td != NULL) ? td->td_ucred :
1588 					    NULL, uproto);
1589 					break;
1590 				}
1591 
1592 			case IPV6_2292PKTINFO:
1593 			case IPV6_2292HOPLIMIT:
1594 			case IPV6_2292HOPOPTS:
1595 			case IPV6_2292DSTOPTS:
1596 			case IPV6_2292RTHDR:
1597 				/* RFC 2292 */
1598 				if (optlen != sizeof(int)) {
1599 					error = EINVAL;
1600 					break;
1601 				}
1602 				error = sooptcopyin(sopt, &optval,
1603 					sizeof optval, sizeof optval);
1604 				if (error)
1605 					break;
1606 				switch (optname) {
1607 				case IPV6_2292PKTINFO:
1608 					OPTSET2292(IN6P_PKTINFO);
1609 					break;
1610 				case IPV6_2292HOPLIMIT:
1611 					OPTSET2292(IN6P_HOPLIMIT);
1612 					break;
1613 				case IPV6_2292HOPOPTS:
1614 					/*
1615 					 * Check super-user privilege.
1616 					 * See comments for IPV6_RECVHOPOPTS.
1617 					 */
1618 					if (td != NULL) {
1619 						error = priv_check(td,
1620 						    PRIV_NETINET_SETHDROPTS);
1621 						if (error)
1622 							return (error);
1623 					}
1624 					OPTSET2292(IN6P_HOPOPTS);
1625 					break;
1626 				case IPV6_2292DSTOPTS:
1627 					if (td != NULL) {
1628 						error = priv_check(td,
1629 						    PRIV_NETINET_SETHDROPTS);
1630 						if (error)
1631 							return (error);
1632 					}
1633 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1634 					break;
1635 				case IPV6_2292RTHDR:
1636 					OPTSET2292(IN6P_RTHDR);
1637 					break;
1638 				}
1639 				break;
1640 			case IPV6_PKTINFO:
1641 			case IPV6_HOPOPTS:
1642 			case IPV6_RTHDR:
1643 			case IPV6_DSTOPTS:
1644 			case IPV6_RTHDRDSTOPTS:
1645 			case IPV6_NEXTHOP:
1646 			{
1647 				/* new advanced API (RFC3542) */
1648 				u_char *optbuf;
1649 				u_char optbuf_storage[MCLBYTES];
1650 				int optlen;
1651 				struct ip6_pktopts **optp;
1652 
1653 				/* cannot mix with RFC2292 */
1654 				if (OPTBIT(IN6P_RFC2292)) {
1655 					error = EINVAL;
1656 					break;
1657 				}
1658 
1659 				/*
1660 				 * We only ensure valsize is not too large
1661 				 * here.  Further validation will be done
1662 				 * later.
1663 				 */
1664 				error = sooptcopyin(sopt, optbuf_storage,
1665 				    sizeof(optbuf_storage), 0);
1666 				if (error)
1667 					break;
1668 				optlen = sopt->sopt_valsize;
1669 				optbuf = optbuf_storage;
1670 				optp = &in6p->in6p_outputopts;
1671 				error = ip6_pcbopt(optname, optbuf, optlen,
1672 				    optp, (td != NULL) ? td->td_ucred : NULL,
1673 				    uproto);
1674 				break;
1675 			}
1676 #undef OPTSET
1677 
1678 			case IPV6_MULTICAST_IF:
1679 			case IPV6_MULTICAST_HOPS:
1680 			case IPV6_MULTICAST_LOOP:
1681 			case IPV6_JOIN_GROUP:
1682 			case IPV6_LEAVE_GROUP:
1683 			case IPV6_MSFILTER:
1684 			case MCAST_BLOCK_SOURCE:
1685 			case MCAST_UNBLOCK_SOURCE:
1686 			case MCAST_JOIN_GROUP:
1687 			case MCAST_LEAVE_GROUP:
1688 			case MCAST_JOIN_SOURCE_GROUP:
1689 			case MCAST_LEAVE_SOURCE_GROUP:
1690 				error = ip6_setmoptions(in6p, sopt);
1691 				break;
1692 
1693 			case IPV6_PORTRANGE:
1694 				error = sooptcopyin(sopt, &optval,
1695 				    sizeof optval, sizeof optval);
1696 				if (error)
1697 					break;
1698 
1699 				switch (optval) {
1700 				case IPV6_PORTRANGE_DEFAULT:
1701 					in6p->inp_flags &= ~(INP_LOWPORT);
1702 					in6p->inp_flags &= ~(INP_HIGHPORT);
1703 					break;
1704 
1705 				case IPV6_PORTRANGE_HIGH:
1706 					in6p->inp_flags &= ~(INP_LOWPORT);
1707 					in6p->inp_flags |= INP_HIGHPORT;
1708 					break;
1709 
1710 				case IPV6_PORTRANGE_LOW:
1711 					in6p->inp_flags &= ~(INP_HIGHPORT);
1712 					in6p->inp_flags |= INP_LOWPORT;
1713 					break;
1714 
1715 				default:
1716 					error = EINVAL;
1717 					break;
1718 				}
1719 				break;
1720 
1721 #ifdef IPSEC
1722 			case IPV6_IPSEC_POLICY:
1723 			{
1724 				caddr_t req;
1725 				struct mbuf *m;
1726 
1727 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1728 					break;
1729 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1730 					break;
1731 				req = mtod(m, caddr_t);
1732 				error = ipsec_set_policy(in6p, optname, req,
1733 				    m->m_len, (sopt->sopt_td != NULL) ?
1734 				    sopt->sopt_td->td_ucred : NULL);
1735 				m_freem(m);
1736 				break;
1737 			}
1738 #endif /* IPSEC */
1739 
1740 			default:
1741 				error = ENOPROTOOPT;
1742 				break;
1743 			}
1744 			break;
1745 
1746 		case SOPT_GET:
1747 			switch (optname) {
1748 
1749 			case IPV6_2292PKTOPTIONS:
1750 #ifdef IPV6_PKTOPTIONS
1751 			case IPV6_PKTOPTIONS:
1752 #endif
1753 				/*
1754 				 * RFC3542 (effectively) deprecated the
1755 				 * semantics of the 2292-style pktoptions.
1756 				 * Since it was not reliable in nature (i.e.,
1757 				 * applications had to expect the lack of some
1758 				 * information after all), it would make sense
1759 				 * to simplify this part by always returning
1760 				 * empty data.
1761 				 */
1762 				sopt->sopt_valsize = 0;
1763 				break;
1764 
1765 			case IPV6_RECVHOPOPTS:
1766 			case IPV6_RECVDSTOPTS:
1767 			case IPV6_RECVRTHDRDSTOPTS:
1768 			case IPV6_UNICAST_HOPS:
1769 			case IPV6_RECVPKTINFO:
1770 			case IPV6_RECVHOPLIMIT:
1771 			case IPV6_RECVRTHDR:
1772 			case IPV6_RECVPATHMTU:
1773 
1774 			case IPV6_FAITH:
1775 			case IPV6_V6ONLY:
1776 			case IPV6_PORTRANGE:
1777 			case IPV6_RECVTCLASS:
1778 			case IPV6_AUTOFLOWLABEL:
1779 				switch (optname) {
1780 
1781 				case IPV6_RECVHOPOPTS:
1782 					optval = OPTBIT(IN6P_HOPOPTS);
1783 					break;
1784 
1785 				case IPV6_RECVDSTOPTS:
1786 					optval = OPTBIT(IN6P_DSTOPTS);
1787 					break;
1788 
1789 				case IPV6_RECVRTHDRDSTOPTS:
1790 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1791 					break;
1792 
1793 				case IPV6_UNICAST_HOPS:
1794 					optval = in6p->in6p_hops;
1795 					break;
1796 
1797 				case IPV6_RECVPKTINFO:
1798 					optval = OPTBIT(IN6P_PKTINFO);
1799 					break;
1800 
1801 				case IPV6_RECVHOPLIMIT:
1802 					optval = OPTBIT(IN6P_HOPLIMIT);
1803 					break;
1804 
1805 				case IPV6_RECVRTHDR:
1806 					optval = OPTBIT(IN6P_RTHDR);
1807 					break;
1808 
1809 				case IPV6_RECVPATHMTU:
1810 					optval = OPTBIT(IN6P_MTU);
1811 					break;
1812 
1813 				case IPV6_FAITH:
1814 					optval = OPTBIT(INP_FAITH);
1815 					break;
1816 
1817 				case IPV6_V6ONLY:
1818 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1819 					break;
1820 
1821 				case IPV6_PORTRANGE:
1822 				    {
1823 					int flags;
1824 					flags = in6p->inp_flags;
1825 					if (flags & INP_HIGHPORT)
1826 						optval = IPV6_PORTRANGE_HIGH;
1827 					else if (flags & INP_LOWPORT)
1828 						optval = IPV6_PORTRANGE_LOW;
1829 					else
1830 						optval = 0;
1831 					break;
1832 				    }
1833 				case IPV6_RECVTCLASS:
1834 					optval = OPTBIT(IN6P_TCLASS);
1835 					break;
1836 
1837 				case IPV6_AUTOFLOWLABEL:
1838 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1839 					break;
1840 
1841 				case IPV6_BINDANY:
1842 					optval = OPTBIT(INP_BINDANY);
1843 					break;
1844 				}
1845 				if (error)
1846 					break;
1847 				error = sooptcopyout(sopt, &optval,
1848 					sizeof optval);
1849 				break;
1850 
1851 			case IPV6_PATHMTU:
1852 			{
1853 				u_long pmtu = 0;
1854 				struct ip6_mtuinfo mtuinfo;
1855 				struct route_in6 sro;
1856 
1857 				bzero(&sro, sizeof(sro));
1858 
1859 				if (!(so->so_state & SS_ISCONNECTED))
1860 					return (ENOTCONN);
1861 				/*
1862 				 * XXX: we dot not consider the case of source
1863 				 * routing, or optional information to specify
1864 				 * the outgoing interface.
1865 				 */
1866 				error = ip6_getpmtu(&sro, NULL, NULL,
1867 				    &in6p->in6p_faddr, &pmtu, NULL);
1868 				if (sro.ro_rt)
1869 					RTFREE(sro.ro_rt);
1870 				if (error)
1871 					break;
1872 				if (pmtu > IPV6_MAXPACKET)
1873 					pmtu = IPV6_MAXPACKET;
1874 
1875 				bzero(&mtuinfo, sizeof(mtuinfo));
1876 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1877 				optdata = (void *)&mtuinfo;
1878 				optdatalen = sizeof(mtuinfo);
1879 				error = sooptcopyout(sopt, optdata,
1880 				    optdatalen);
1881 				break;
1882 			}
1883 
1884 			case IPV6_2292PKTINFO:
1885 			case IPV6_2292HOPLIMIT:
1886 			case IPV6_2292HOPOPTS:
1887 			case IPV6_2292RTHDR:
1888 			case IPV6_2292DSTOPTS:
1889 				switch (optname) {
1890 				case IPV6_2292PKTINFO:
1891 					optval = OPTBIT(IN6P_PKTINFO);
1892 					break;
1893 				case IPV6_2292HOPLIMIT:
1894 					optval = OPTBIT(IN6P_HOPLIMIT);
1895 					break;
1896 				case IPV6_2292HOPOPTS:
1897 					optval = OPTBIT(IN6P_HOPOPTS);
1898 					break;
1899 				case IPV6_2292RTHDR:
1900 					optval = OPTBIT(IN6P_RTHDR);
1901 					break;
1902 				case IPV6_2292DSTOPTS:
1903 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1904 					break;
1905 				}
1906 				error = sooptcopyout(sopt, &optval,
1907 				    sizeof optval);
1908 				break;
1909 			case IPV6_PKTINFO:
1910 			case IPV6_HOPOPTS:
1911 			case IPV6_RTHDR:
1912 			case IPV6_DSTOPTS:
1913 			case IPV6_RTHDRDSTOPTS:
1914 			case IPV6_NEXTHOP:
1915 			case IPV6_TCLASS:
1916 			case IPV6_DONTFRAG:
1917 			case IPV6_USE_MIN_MTU:
1918 			case IPV6_PREFER_TEMPADDR:
1919 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1920 				    optname, sopt);
1921 				break;
1922 
1923 			case IPV6_MULTICAST_IF:
1924 			case IPV6_MULTICAST_HOPS:
1925 			case IPV6_MULTICAST_LOOP:
1926 			case IPV6_MSFILTER:
1927 				error = ip6_getmoptions(in6p, sopt);
1928 				break;
1929 
1930 #ifdef IPSEC
1931 			case IPV6_IPSEC_POLICY:
1932 			  {
1933 				caddr_t req = NULL;
1934 				size_t len = 0;
1935 				struct mbuf *m = NULL;
1936 				struct mbuf **mp = &m;
1937 				size_t ovalsize = sopt->sopt_valsize;
1938 				caddr_t oval = (caddr_t)sopt->sopt_val;
1939 
1940 				error = soopt_getm(sopt, &m); /* XXX */
1941 				if (error != 0)
1942 					break;
1943 				error = soopt_mcopyin(sopt, m); /* XXX */
1944 				if (error != 0)
1945 					break;
1946 				sopt->sopt_valsize = ovalsize;
1947 				sopt->sopt_val = oval;
1948 				if (m) {
1949 					req = mtod(m, caddr_t);
1950 					len = m->m_len;
1951 				}
1952 				error = ipsec_get_policy(in6p, req, len, mp);
1953 				if (error == 0)
1954 					error = soopt_mcopyout(sopt, m); /* XXX */
1955 				if (error == 0 && m)
1956 					m_freem(m);
1957 				break;
1958 			  }
1959 #endif /* IPSEC */
1960 
1961 			default:
1962 				error = ENOPROTOOPT;
1963 				break;
1964 			}
1965 			break;
1966 		}
1967 	} else {		/* level != IPPROTO_IPV6 */
1968 		error = EINVAL;
1969 	}
1970 	return (error);
1971 }
1972 
1973 int
1974 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
1975 {
1976 	int error = 0, optval, optlen;
1977 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1978 	struct inpcb *in6p = sotoinpcb(so);
1979 	int level, op, optname;
1980 
1981 	level = sopt->sopt_level;
1982 	op = sopt->sopt_dir;
1983 	optname = sopt->sopt_name;
1984 	optlen = sopt->sopt_valsize;
1985 
1986 	if (level != IPPROTO_IPV6) {
1987 		return (EINVAL);
1988 	}
1989 
1990 	switch (optname) {
1991 	case IPV6_CHECKSUM:
1992 		/*
1993 		 * For ICMPv6 sockets, no modification allowed for checksum
1994 		 * offset, permit "no change" values to help existing apps.
1995 		 *
1996 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1997 		 * for an ICMPv6 socket will fail."
1998 		 * The current behavior does not meet RFC3542.
1999 		 */
2000 		switch (op) {
2001 		case SOPT_SET:
2002 			if (optlen != sizeof(int)) {
2003 				error = EINVAL;
2004 				break;
2005 			}
2006 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2007 					    sizeof(optval));
2008 			if (error)
2009 				break;
2010 			if ((optval % 2) != 0) {
2011 				/* the API assumes even offset values */
2012 				error = EINVAL;
2013 			} else if (so->so_proto->pr_protocol ==
2014 			    IPPROTO_ICMPV6) {
2015 				if (optval != icmp6off)
2016 					error = EINVAL;
2017 			} else
2018 				in6p->in6p_cksum = optval;
2019 			break;
2020 
2021 		case SOPT_GET:
2022 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2023 				optval = icmp6off;
2024 			else
2025 				optval = in6p->in6p_cksum;
2026 
2027 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2028 			break;
2029 
2030 		default:
2031 			error = EINVAL;
2032 			break;
2033 		}
2034 		break;
2035 
2036 	default:
2037 		error = ENOPROTOOPT;
2038 		break;
2039 	}
2040 
2041 	return (error);
2042 }
2043 
2044 /*
2045  * Set up IP6 options in pcb for insertion in output packets or
2046  * specifying behavior of outgoing packets.
2047  */
2048 static int
2049 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2050     struct socket *so, struct sockopt *sopt)
2051 {
2052 	struct ip6_pktopts *opt = *pktopt;
2053 	int error = 0;
2054 	struct thread *td = sopt->sopt_td;
2055 
2056 	/* turn off any old options. */
2057 	if (opt) {
2058 #ifdef DIAGNOSTIC
2059 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2060 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2061 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2062 			printf("ip6_pcbopts: all specified options are cleared.\n");
2063 #endif
2064 		ip6_clearpktopts(opt, -1);
2065 	} else
2066 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2067 	*pktopt = NULL;
2068 
2069 	if (!m || m->m_len == 0) {
2070 		/*
2071 		 * Only turning off any previous options, regardless of
2072 		 * whether the opt is just created or given.
2073 		 */
2074 		free(opt, M_IP6OPT);
2075 		return (0);
2076 	}
2077 
2078 	/*  set options specified by user. */
2079 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2080 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2081 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2082 		free(opt, M_IP6OPT);
2083 		return (error);
2084 	}
2085 	*pktopt = opt;
2086 	return (0);
2087 }
2088 
2089 /*
2090  * initialize ip6_pktopts.  beware that there are non-zero default values in
2091  * the struct.
2092  */
2093 void
2094 ip6_initpktopts(struct ip6_pktopts *opt)
2095 {
2096 
2097 	bzero(opt, sizeof(*opt));
2098 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2099 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2100 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2101 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2102 }
2103 
2104 static int
2105 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2106     struct ucred *cred, int uproto)
2107 {
2108 	struct ip6_pktopts *opt;
2109 
2110 	if (*pktopt == NULL) {
2111 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2112 		    M_WAITOK);
2113 		ip6_initpktopts(*pktopt);
2114 	}
2115 	opt = *pktopt;
2116 
2117 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2118 }
2119 
2120 static int
2121 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2122 {
2123 	void *optdata = NULL;
2124 	int optdatalen = 0;
2125 	struct ip6_ext *ip6e;
2126 	int error = 0;
2127 	struct in6_pktinfo null_pktinfo;
2128 	int deftclass = 0, on;
2129 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2130 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2131 
2132 	switch (optname) {
2133 	case IPV6_PKTINFO:
2134 		if (pktopt && pktopt->ip6po_pktinfo)
2135 			optdata = (void *)pktopt->ip6po_pktinfo;
2136 		else {
2137 			/* XXX: we don't have to do this every time... */
2138 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2139 			optdata = (void *)&null_pktinfo;
2140 		}
2141 		optdatalen = sizeof(struct in6_pktinfo);
2142 		break;
2143 	case IPV6_TCLASS:
2144 		if (pktopt && pktopt->ip6po_tclass >= 0)
2145 			optdata = (void *)&pktopt->ip6po_tclass;
2146 		else
2147 			optdata = (void *)&deftclass;
2148 		optdatalen = sizeof(int);
2149 		break;
2150 	case IPV6_HOPOPTS:
2151 		if (pktopt && pktopt->ip6po_hbh) {
2152 			optdata = (void *)pktopt->ip6po_hbh;
2153 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2154 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2155 		}
2156 		break;
2157 	case IPV6_RTHDR:
2158 		if (pktopt && pktopt->ip6po_rthdr) {
2159 			optdata = (void *)pktopt->ip6po_rthdr;
2160 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2161 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2162 		}
2163 		break;
2164 	case IPV6_RTHDRDSTOPTS:
2165 		if (pktopt && pktopt->ip6po_dest1) {
2166 			optdata = (void *)pktopt->ip6po_dest1;
2167 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2168 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2169 		}
2170 		break;
2171 	case IPV6_DSTOPTS:
2172 		if (pktopt && pktopt->ip6po_dest2) {
2173 			optdata = (void *)pktopt->ip6po_dest2;
2174 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2175 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2176 		}
2177 		break;
2178 	case IPV6_NEXTHOP:
2179 		if (pktopt && pktopt->ip6po_nexthop) {
2180 			optdata = (void *)pktopt->ip6po_nexthop;
2181 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2182 		}
2183 		break;
2184 	case IPV6_USE_MIN_MTU:
2185 		if (pktopt)
2186 			optdata = (void *)&pktopt->ip6po_minmtu;
2187 		else
2188 			optdata = (void *)&defminmtu;
2189 		optdatalen = sizeof(int);
2190 		break;
2191 	case IPV6_DONTFRAG:
2192 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2193 			on = 1;
2194 		else
2195 			on = 0;
2196 		optdata = (void *)&on;
2197 		optdatalen = sizeof(on);
2198 		break;
2199 	case IPV6_PREFER_TEMPADDR:
2200 		if (pktopt)
2201 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2202 		else
2203 			optdata = (void *)&defpreftemp;
2204 		optdatalen = sizeof(int);
2205 		break;
2206 	default:		/* should not happen */
2207 #ifdef DIAGNOSTIC
2208 		panic("ip6_getpcbopt: unexpected option\n");
2209 #endif
2210 		return (ENOPROTOOPT);
2211 	}
2212 
2213 	error = sooptcopyout(sopt, optdata, optdatalen);
2214 
2215 	return (error);
2216 }
2217 
2218 void
2219 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2220 {
2221 	if (pktopt == NULL)
2222 		return;
2223 
2224 	if (optname == -1 || optname == IPV6_PKTINFO) {
2225 		if (pktopt->ip6po_pktinfo)
2226 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2227 		pktopt->ip6po_pktinfo = NULL;
2228 	}
2229 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2230 		pktopt->ip6po_hlim = -1;
2231 	if (optname == -1 || optname == IPV6_TCLASS)
2232 		pktopt->ip6po_tclass = -1;
2233 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2234 		if (pktopt->ip6po_nextroute.ro_rt) {
2235 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2236 			pktopt->ip6po_nextroute.ro_rt = NULL;
2237 		}
2238 		if (pktopt->ip6po_nexthop)
2239 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2240 		pktopt->ip6po_nexthop = NULL;
2241 	}
2242 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2243 		if (pktopt->ip6po_hbh)
2244 			free(pktopt->ip6po_hbh, M_IP6OPT);
2245 		pktopt->ip6po_hbh = NULL;
2246 	}
2247 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2248 		if (pktopt->ip6po_dest1)
2249 			free(pktopt->ip6po_dest1, M_IP6OPT);
2250 		pktopt->ip6po_dest1 = NULL;
2251 	}
2252 	if (optname == -1 || optname == IPV6_RTHDR) {
2253 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2254 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2255 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2256 		if (pktopt->ip6po_route.ro_rt) {
2257 			RTFREE(pktopt->ip6po_route.ro_rt);
2258 			pktopt->ip6po_route.ro_rt = NULL;
2259 		}
2260 	}
2261 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2262 		if (pktopt->ip6po_dest2)
2263 			free(pktopt->ip6po_dest2, M_IP6OPT);
2264 		pktopt->ip6po_dest2 = NULL;
2265 	}
2266 }
2267 
2268 #define PKTOPT_EXTHDRCPY(type) \
2269 do {\
2270 	if (src->type) {\
2271 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2272 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2273 		if (dst->type == NULL && canwait == M_NOWAIT)\
2274 			goto bad;\
2275 		bcopy(src->type, dst->type, hlen);\
2276 	}\
2277 } while (/*CONSTCOND*/ 0)
2278 
2279 static int
2280 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2281 {
2282 	if (dst == NULL || src == NULL)  {
2283 		printf("ip6_clearpktopts: invalid argument\n");
2284 		return (EINVAL);
2285 	}
2286 
2287 	dst->ip6po_hlim = src->ip6po_hlim;
2288 	dst->ip6po_tclass = src->ip6po_tclass;
2289 	dst->ip6po_flags = src->ip6po_flags;
2290 	if (src->ip6po_pktinfo) {
2291 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2292 		    M_IP6OPT, canwait);
2293 		if (dst->ip6po_pktinfo == NULL)
2294 			goto bad;
2295 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2296 	}
2297 	if (src->ip6po_nexthop) {
2298 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2299 		    M_IP6OPT, canwait);
2300 		if (dst->ip6po_nexthop == NULL)
2301 			goto bad;
2302 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2303 		    src->ip6po_nexthop->sa_len);
2304 	}
2305 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2306 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2307 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2308 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2309 	return (0);
2310 
2311   bad:
2312 	ip6_clearpktopts(dst, -1);
2313 	return (ENOBUFS);
2314 }
2315 #undef PKTOPT_EXTHDRCPY
2316 
2317 struct ip6_pktopts *
2318 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2319 {
2320 	int error;
2321 	struct ip6_pktopts *dst;
2322 
2323 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2324 	if (dst == NULL)
2325 		return (NULL);
2326 	ip6_initpktopts(dst);
2327 
2328 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2329 		free(dst, M_IP6OPT);
2330 		return (NULL);
2331 	}
2332 
2333 	return (dst);
2334 }
2335 
2336 void
2337 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2338 {
2339 	if (pktopt == NULL)
2340 		return;
2341 
2342 	ip6_clearpktopts(pktopt, -1);
2343 
2344 	free(pktopt, M_IP6OPT);
2345 }
2346 
2347 /*
2348  * Set IPv6 outgoing packet options based on advanced API.
2349  */
2350 int
2351 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2352     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2353 {
2354 	struct cmsghdr *cm = 0;
2355 
2356 	if (control == NULL || opt == NULL)
2357 		return (EINVAL);
2358 
2359 	ip6_initpktopts(opt);
2360 	if (stickyopt) {
2361 		int error;
2362 
2363 		/*
2364 		 * If stickyopt is provided, make a local copy of the options
2365 		 * for this particular packet, then override them by ancillary
2366 		 * objects.
2367 		 * XXX: copypktopts() does not copy the cached route to a next
2368 		 * hop (if any).  This is not very good in terms of efficiency,
2369 		 * but we can allow this since this option should be rarely
2370 		 * used.
2371 		 */
2372 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2373 			return (error);
2374 	}
2375 
2376 	/*
2377 	 * XXX: Currently, we assume all the optional information is stored
2378 	 * in a single mbuf.
2379 	 */
2380 	if (control->m_next)
2381 		return (EINVAL);
2382 
2383 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2384 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2385 		int error;
2386 
2387 		if (control->m_len < CMSG_LEN(0))
2388 			return (EINVAL);
2389 
2390 		cm = mtod(control, struct cmsghdr *);
2391 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2392 			return (EINVAL);
2393 		if (cm->cmsg_level != IPPROTO_IPV6)
2394 			continue;
2395 
2396 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2397 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2398 		if (error)
2399 			return (error);
2400 	}
2401 
2402 	return (0);
2403 }
2404 
2405 /*
2406  * Set a particular packet option, as a sticky option or an ancillary data
2407  * item.  "len" can be 0 only when it's a sticky option.
2408  * We have 4 cases of combination of "sticky" and "cmsg":
2409  * "sticky=0, cmsg=0": impossible
2410  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2411  * "sticky=1, cmsg=0": RFC3542 socket option
2412  * "sticky=1, cmsg=1": RFC2292 socket option
2413  */
2414 static int
2415 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2416     struct ucred *cred, int sticky, int cmsg, int uproto)
2417 {
2418 	int minmtupolicy, preftemp;
2419 	int error;
2420 
2421 	if (!sticky && !cmsg) {
2422 #ifdef DIAGNOSTIC
2423 		printf("ip6_setpktopt: impossible case\n");
2424 #endif
2425 		return (EINVAL);
2426 	}
2427 
2428 	/*
2429 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2430 	 * not be specified in the context of RFC3542.  Conversely,
2431 	 * RFC3542 types should not be specified in the context of RFC2292.
2432 	 */
2433 	if (!cmsg) {
2434 		switch (optname) {
2435 		case IPV6_2292PKTINFO:
2436 		case IPV6_2292HOPLIMIT:
2437 		case IPV6_2292NEXTHOP:
2438 		case IPV6_2292HOPOPTS:
2439 		case IPV6_2292DSTOPTS:
2440 		case IPV6_2292RTHDR:
2441 		case IPV6_2292PKTOPTIONS:
2442 			return (ENOPROTOOPT);
2443 		}
2444 	}
2445 	if (sticky && cmsg) {
2446 		switch (optname) {
2447 		case IPV6_PKTINFO:
2448 		case IPV6_HOPLIMIT:
2449 		case IPV6_NEXTHOP:
2450 		case IPV6_HOPOPTS:
2451 		case IPV6_DSTOPTS:
2452 		case IPV6_RTHDRDSTOPTS:
2453 		case IPV6_RTHDR:
2454 		case IPV6_USE_MIN_MTU:
2455 		case IPV6_DONTFRAG:
2456 		case IPV6_TCLASS:
2457 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2458 			return (ENOPROTOOPT);
2459 		}
2460 	}
2461 
2462 	switch (optname) {
2463 	case IPV6_2292PKTINFO:
2464 	case IPV6_PKTINFO:
2465 	{
2466 		struct ifnet *ifp = NULL;
2467 		struct in6_pktinfo *pktinfo;
2468 
2469 		if (len != sizeof(struct in6_pktinfo))
2470 			return (EINVAL);
2471 
2472 		pktinfo = (struct in6_pktinfo *)buf;
2473 
2474 		/*
2475 		 * An application can clear any sticky IPV6_PKTINFO option by
2476 		 * doing a "regular" setsockopt with ipi6_addr being
2477 		 * in6addr_any and ipi6_ifindex being zero.
2478 		 * [RFC 3542, Section 6]
2479 		 */
2480 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2481 		    pktinfo->ipi6_ifindex == 0 &&
2482 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2483 			ip6_clearpktopts(opt, optname);
2484 			break;
2485 		}
2486 
2487 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2488 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2489 			return (EINVAL);
2490 		}
2491 
2492 		/* validate the interface index if specified. */
2493 		if (pktinfo->ipi6_ifindex > V_if_index ||
2494 		    pktinfo->ipi6_ifindex < 0) {
2495 			 return (ENXIO);
2496 		}
2497 		if (pktinfo->ipi6_ifindex) {
2498 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2499 			if (ifp == NULL)
2500 				return (ENXIO);
2501 		}
2502 
2503 		/*
2504 		 * We store the address anyway, and let in6_selectsrc()
2505 		 * validate the specified address.  This is because ipi6_addr
2506 		 * may not have enough information about its scope zone, and
2507 		 * we may need additional information (such as outgoing
2508 		 * interface or the scope zone of a destination address) to
2509 		 * disambiguate the scope.
2510 		 * XXX: the delay of the validation may confuse the
2511 		 * application when it is used as a sticky option.
2512 		 */
2513 		if (opt->ip6po_pktinfo == NULL) {
2514 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2515 			    M_IP6OPT, M_NOWAIT);
2516 			if (opt->ip6po_pktinfo == NULL)
2517 				return (ENOBUFS);
2518 		}
2519 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2520 		break;
2521 	}
2522 
2523 	case IPV6_2292HOPLIMIT:
2524 	case IPV6_HOPLIMIT:
2525 	{
2526 		int *hlimp;
2527 
2528 		/*
2529 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2530 		 * to simplify the ordering among hoplimit options.
2531 		 */
2532 		if (optname == IPV6_HOPLIMIT && sticky)
2533 			return (ENOPROTOOPT);
2534 
2535 		if (len != sizeof(int))
2536 			return (EINVAL);
2537 		hlimp = (int *)buf;
2538 		if (*hlimp < -1 || *hlimp > 255)
2539 			return (EINVAL);
2540 
2541 		opt->ip6po_hlim = *hlimp;
2542 		break;
2543 	}
2544 
2545 	case IPV6_TCLASS:
2546 	{
2547 		int tclass;
2548 
2549 		if (len != sizeof(int))
2550 			return (EINVAL);
2551 		tclass = *(int *)buf;
2552 		if (tclass < -1 || tclass > 255)
2553 			return (EINVAL);
2554 
2555 		opt->ip6po_tclass = tclass;
2556 		break;
2557 	}
2558 
2559 	case IPV6_2292NEXTHOP:
2560 	case IPV6_NEXTHOP:
2561 		if (cred != NULL) {
2562 			error = priv_check_cred(cred,
2563 			    PRIV_NETINET_SETHDROPTS, 0);
2564 			if (error)
2565 				return (error);
2566 		}
2567 
2568 		if (len == 0) {	/* just remove the option */
2569 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2570 			break;
2571 		}
2572 
2573 		/* check if cmsg_len is large enough for sa_len */
2574 		if (len < sizeof(struct sockaddr) || len < *buf)
2575 			return (EINVAL);
2576 
2577 		switch (((struct sockaddr *)buf)->sa_family) {
2578 		case AF_INET6:
2579 		{
2580 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2581 			int error;
2582 
2583 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2584 				return (EINVAL);
2585 
2586 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2587 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2588 				return (EINVAL);
2589 			}
2590 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2591 			    != 0) {
2592 				return (error);
2593 			}
2594 			break;
2595 		}
2596 		case AF_LINK:	/* should eventually be supported */
2597 		default:
2598 			return (EAFNOSUPPORT);
2599 		}
2600 
2601 		/* turn off the previous option, then set the new option. */
2602 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2603 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2604 		if (opt->ip6po_nexthop == NULL)
2605 			return (ENOBUFS);
2606 		bcopy(buf, opt->ip6po_nexthop, *buf);
2607 		break;
2608 
2609 	case IPV6_2292HOPOPTS:
2610 	case IPV6_HOPOPTS:
2611 	{
2612 		struct ip6_hbh *hbh;
2613 		int hbhlen;
2614 
2615 		/*
2616 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2617 		 * options, since per-option restriction has too much
2618 		 * overhead.
2619 		 */
2620 		if (cred != NULL) {
2621 			error = priv_check_cred(cred,
2622 			    PRIV_NETINET_SETHDROPTS, 0);
2623 			if (error)
2624 				return (error);
2625 		}
2626 
2627 		if (len == 0) {
2628 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2629 			break;	/* just remove the option */
2630 		}
2631 
2632 		/* message length validation */
2633 		if (len < sizeof(struct ip6_hbh))
2634 			return (EINVAL);
2635 		hbh = (struct ip6_hbh *)buf;
2636 		hbhlen = (hbh->ip6h_len + 1) << 3;
2637 		if (len != hbhlen)
2638 			return (EINVAL);
2639 
2640 		/* turn off the previous option, then set the new option. */
2641 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2642 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2643 		if (opt->ip6po_hbh == NULL)
2644 			return (ENOBUFS);
2645 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2646 
2647 		break;
2648 	}
2649 
2650 	case IPV6_2292DSTOPTS:
2651 	case IPV6_DSTOPTS:
2652 	case IPV6_RTHDRDSTOPTS:
2653 	{
2654 		struct ip6_dest *dest, **newdest = NULL;
2655 		int destlen;
2656 
2657 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2658 			error = priv_check_cred(cred,
2659 			    PRIV_NETINET_SETHDROPTS, 0);
2660 			if (error)
2661 				return (error);
2662 		}
2663 
2664 		if (len == 0) {
2665 			ip6_clearpktopts(opt, optname);
2666 			break;	/* just remove the option */
2667 		}
2668 
2669 		/* message length validation */
2670 		if (len < sizeof(struct ip6_dest))
2671 			return (EINVAL);
2672 		dest = (struct ip6_dest *)buf;
2673 		destlen = (dest->ip6d_len + 1) << 3;
2674 		if (len != destlen)
2675 			return (EINVAL);
2676 
2677 		/*
2678 		 * Determine the position that the destination options header
2679 		 * should be inserted; before or after the routing header.
2680 		 */
2681 		switch (optname) {
2682 		case IPV6_2292DSTOPTS:
2683 			/*
2684 			 * The old advacned API is ambiguous on this point.
2685 			 * Our approach is to determine the position based
2686 			 * according to the existence of a routing header.
2687 			 * Note, however, that this depends on the order of the
2688 			 * extension headers in the ancillary data; the 1st
2689 			 * part of the destination options header must appear
2690 			 * before the routing header in the ancillary data,
2691 			 * too.
2692 			 * RFC3542 solved the ambiguity by introducing
2693 			 * separate ancillary data or option types.
2694 			 */
2695 			if (opt->ip6po_rthdr == NULL)
2696 				newdest = &opt->ip6po_dest1;
2697 			else
2698 				newdest = &opt->ip6po_dest2;
2699 			break;
2700 		case IPV6_RTHDRDSTOPTS:
2701 			newdest = &opt->ip6po_dest1;
2702 			break;
2703 		case IPV6_DSTOPTS:
2704 			newdest = &opt->ip6po_dest2;
2705 			break;
2706 		}
2707 
2708 		/* turn off the previous option, then set the new option. */
2709 		ip6_clearpktopts(opt, optname);
2710 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2711 		if (*newdest == NULL)
2712 			return (ENOBUFS);
2713 		bcopy(dest, *newdest, destlen);
2714 
2715 		break;
2716 	}
2717 
2718 	case IPV6_2292RTHDR:
2719 	case IPV6_RTHDR:
2720 	{
2721 		struct ip6_rthdr *rth;
2722 		int rthlen;
2723 
2724 		if (len == 0) {
2725 			ip6_clearpktopts(opt, IPV6_RTHDR);
2726 			break;	/* just remove the option */
2727 		}
2728 
2729 		/* message length validation */
2730 		if (len < sizeof(struct ip6_rthdr))
2731 			return (EINVAL);
2732 		rth = (struct ip6_rthdr *)buf;
2733 		rthlen = (rth->ip6r_len + 1) << 3;
2734 		if (len != rthlen)
2735 			return (EINVAL);
2736 
2737 		switch (rth->ip6r_type) {
2738 		case IPV6_RTHDR_TYPE_0:
2739 			if (rth->ip6r_len == 0)	/* must contain one addr */
2740 				return (EINVAL);
2741 			if (rth->ip6r_len % 2) /* length must be even */
2742 				return (EINVAL);
2743 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2744 				return (EINVAL);
2745 			break;
2746 		default:
2747 			return (EINVAL);	/* not supported */
2748 		}
2749 
2750 		/* turn off the previous option */
2751 		ip6_clearpktopts(opt, IPV6_RTHDR);
2752 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2753 		if (opt->ip6po_rthdr == NULL)
2754 			return (ENOBUFS);
2755 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2756 
2757 		break;
2758 	}
2759 
2760 	case IPV6_USE_MIN_MTU:
2761 		if (len != sizeof(int))
2762 			return (EINVAL);
2763 		minmtupolicy = *(int *)buf;
2764 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2765 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2766 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2767 			return (EINVAL);
2768 		}
2769 		opt->ip6po_minmtu = minmtupolicy;
2770 		break;
2771 
2772 	case IPV6_DONTFRAG:
2773 		if (len != sizeof(int))
2774 			return (EINVAL);
2775 
2776 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2777 			/*
2778 			 * we ignore this option for TCP sockets.
2779 			 * (RFC3542 leaves this case unspecified.)
2780 			 */
2781 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2782 		} else
2783 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2784 		break;
2785 
2786 	case IPV6_PREFER_TEMPADDR:
2787 		if (len != sizeof(int))
2788 			return (EINVAL);
2789 		preftemp = *(int *)buf;
2790 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2791 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2792 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2793 			return (EINVAL);
2794 		}
2795 		opt->ip6po_prefer_tempaddr = preftemp;
2796 		break;
2797 
2798 	default:
2799 		return (ENOPROTOOPT);
2800 	} /* end of switch */
2801 
2802 	return (0);
2803 }
2804 
2805 /*
2806  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2807  * packet to the input queue of a specified interface.  Note that this
2808  * calls the output routine of the loopback "driver", but with an interface
2809  * pointer that might NOT be &loif -- easier than replicating that code here.
2810  */
2811 void
2812 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2813 {
2814 	struct mbuf *copym;
2815 	struct ip6_hdr *ip6;
2816 
2817 	copym = m_copy(m, 0, M_COPYALL);
2818 	if (copym == NULL)
2819 		return;
2820 
2821 	/*
2822 	 * Make sure to deep-copy IPv6 header portion in case the data
2823 	 * is in an mbuf cluster, so that we can safely override the IPv6
2824 	 * header portion later.
2825 	 */
2826 	if ((copym->m_flags & M_EXT) != 0 ||
2827 	    copym->m_len < sizeof(struct ip6_hdr)) {
2828 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2829 		if (copym == NULL)
2830 			return;
2831 	}
2832 
2833 #ifdef DIAGNOSTIC
2834 	if (copym->m_len < sizeof(*ip6)) {
2835 		m_freem(copym);
2836 		return;
2837 	}
2838 #endif
2839 
2840 	ip6 = mtod(copym, struct ip6_hdr *);
2841 	/*
2842 	 * clear embedded scope identifiers if necessary.
2843 	 * in6_clearscope will touch the addresses only when necessary.
2844 	 */
2845 	in6_clearscope(&ip6->ip6_src);
2846 	in6_clearscope(&ip6->ip6_dst);
2847 
2848 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2849 }
2850 
2851 /*
2852  * Chop IPv6 header off from the payload.
2853  */
2854 static int
2855 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2856 {
2857 	struct mbuf *mh;
2858 	struct ip6_hdr *ip6;
2859 
2860 	ip6 = mtod(m, struct ip6_hdr *);
2861 	if (m->m_len > sizeof(*ip6)) {
2862 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2863 		if (mh == 0) {
2864 			m_freem(m);
2865 			return ENOBUFS;
2866 		}
2867 		M_MOVE_PKTHDR(mh, m);
2868 		MH_ALIGN(mh, sizeof(*ip6));
2869 		m->m_len -= sizeof(*ip6);
2870 		m->m_data += sizeof(*ip6);
2871 		mh->m_next = m;
2872 		m = mh;
2873 		m->m_len = sizeof(*ip6);
2874 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2875 	}
2876 	exthdrs->ip6e_ip6 = m;
2877 	return 0;
2878 }
2879 
2880 /*
2881  * Compute IPv6 extension header length.
2882  */
2883 int
2884 ip6_optlen(struct inpcb *in6p)
2885 {
2886 	int len;
2887 
2888 	if (!in6p->in6p_outputopts)
2889 		return 0;
2890 
2891 	len = 0;
2892 #define elen(x) \
2893     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2894 
2895 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2896 	if (in6p->in6p_outputopts->ip6po_rthdr)
2897 		/* dest1 is valid with rthdr only */
2898 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2899 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2900 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2901 	return len;
2902 #undef elen
2903 }
2904