xref: /freebsd/sys/netinet6/ip6_output.c (revision 10f0bcab61ef441cb5af32fb706688d8cbd55dc0)
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/ucred.h>
81 
82 #include <net/if.h>
83 #include <net/netisr.h>
84 #include <net/route.h>
85 #include <net/pfil.h>
86 
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet6/in6_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/tcp_var.h>
95 #include <netinet6/nd6.h>
96 
97 #ifdef IPSEC
98 #include <netipsec/ipsec.h>
99 #include <netipsec/ipsec6.h>
100 #include <netipsec/key.h>
101 #include <netinet6/ip6_ipsec.h>
102 #endif /* IPSEC */
103 
104 #include <netinet6/ip6protosw.h>
105 #include <netinet6/scope6_var.h>
106 
107 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
108 
109 struct ip6_exthdrs {
110 	struct mbuf *ip6e_ip6;
111 	struct mbuf *ip6e_hbh;
112 	struct mbuf *ip6e_dest1;
113 	struct mbuf *ip6e_rthdr;
114 	struct mbuf *ip6e_dest2;
115 };
116 
117 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
118 			   struct ucred *, int));
119 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
120 	struct socket *, struct sockopt *));
121 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
122 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
123 	struct ucred *, int, int, int));
124 
125 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
126 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
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, replace the destination address field
433 	 * with the first hop of the routing header.
434 	 */
435 	if (exthdrs.ip6e_rthdr) {
436 		struct ip6_rthdr *rh =
437 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
438 						  struct ip6_rthdr *));
439 		struct ip6_rthdr0 *rh0;
440 		struct in6_addr *addr;
441 		struct sockaddr_in6 sa;
442 
443 		switch (rh->ip6r_type) {
444 		case IPV6_RTHDR_TYPE_0:
445 			 rh0 = (struct ip6_rthdr0 *)rh;
446 			 addr = (struct in6_addr *)(rh0 + 1);
447 
448 			 /*
449 			  * construct a sockaddr_in6 form of
450 			  * the first hop.
451 			  *
452 			  * XXX: we may not have enough
453 			  * information about its scope zone;
454 			  * there is no standard API to pass
455 			  * the information from the
456 			  * application.
457 			  */
458 			 bzero(&sa, sizeof(sa));
459 			 sa.sin6_family = AF_INET6;
460 			 sa.sin6_len = sizeof(sa);
461 			 sa.sin6_addr = addr[0];
462 			 if ((error = sa6_embedscope(&sa,
463 			     ip6_use_defzone)) != 0) {
464 				 goto bad;
465 			 }
466 			 ip6->ip6_dst = sa.sin6_addr;
467 			 bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
468 			     * (rh0->ip6r0_segleft - 1));
469 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
470 			 /* XXX */
471 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
472 			 break;
473 		default:	/* is it possible? */
474 			 error = EINVAL;
475 			 goto bad;
476 		}
477 	}
478 
479 	/* Source address validation */
480 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
481 	    (flags & IPV6_UNSPECSRC) == 0) {
482 		error = EOPNOTSUPP;
483 		ip6stat.ip6s_badscope++;
484 		goto bad;
485 	}
486 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
487 		error = EOPNOTSUPP;
488 		ip6stat.ip6s_badscope++;
489 		goto bad;
490 	}
491 
492 	ip6stat.ip6s_localout++;
493 
494 	/*
495 	 * Route packet.
496 	 */
497 	if (ro == 0) {
498 		ro = &ip6route;
499 		bzero((caddr_t)ro, sizeof(*ro));
500 	}
501 	ro_pmtu = ro;
502 	if (opt && opt->ip6po_rthdr)
503 		ro = &opt->ip6po_route;
504 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
505 
506 again:
507 	/*
508 	 * if specified, try to fill in the traffic class field.
509 	 * do not override if a non-zero value is already set.
510 	 * we check the diffserv field and the ecn field separately.
511 	 */
512 	if (opt && opt->ip6po_tclass >= 0) {
513 		int mask = 0;
514 
515 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
516 			mask |= 0xfc;
517 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
518 			mask |= 0x03;
519 		if (mask != 0)
520 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
521 	}
522 
523 	/* fill in or override the hop limit field, if necessary. */
524 	if (opt && opt->ip6po_hlim != -1)
525 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
526 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
527 		if (im6o != NULL)
528 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
529 		else
530 			ip6->ip6_hlim = ip6_defmcasthlim;
531 	}
532 
533 #ifdef IPSEC
534 	/*
535 	 * We may re-inject packets into the stack here.
536 	 */
537 	if (needipsec && needipsectun) {
538 		struct ipsec_output_state state;
539 
540 		/*
541 		 * All the extension headers will become inaccessible
542 		 * (since they can be encrypted).
543 		 * Don't panic, we need no more updates to extension headers
544 		 * on inner IPv6 packet (since they are now encapsulated).
545 		 *
546 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
547 		 */
548 		bzero(&exthdrs, sizeof(exthdrs));
549 		exthdrs.ip6e_ip6 = m;
550 
551 		bzero(&state, sizeof(state));
552 		state.m = m;
553 		state.ro = (struct route *)ro;
554 		state.dst = (struct sockaddr *)dst;
555 
556 		error = ipsec6_output_tunnel(&state, sp, flags);
557 
558 		m = state.m;
559 		ro = (struct route_in6 *)state.ro;
560 		dst = (struct sockaddr_in6 *)state.dst;
561 		if (error == EJUSTRETURN) {
562 			/*
563 			 * We had a SP with a level of 'use' and no SA. We
564 			 * will just continue to process the packet without
565 			 * IPsec processing.
566 			 */
567 			;
568 		} else if (error) {
569 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
570 			m0 = m = NULL;
571 			m = NULL;
572 			switch (error) {
573 			case EHOSTUNREACH:
574 			case ENETUNREACH:
575 			case EMSGSIZE:
576 			case ENOBUFS:
577 			case ENOMEM:
578 				break;
579 			default:
580 				printf("[%s:%d] (ipsec): error code %d\n",
581 				    __func__, __LINE__, error);
582 				/* FALLTHROUGH */
583 			case ENOENT:
584 				/* don't show these error codes to the user */
585 				error = 0;
586 				break;
587 			}
588 			goto bad;
589 		} else {
590 			/*
591 			 * In the FAST IPSec case we have already
592 			 * re-injected the packet and it has been freed
593 			 * by the ipsec_done() function.  So, just clean
594 			 * up after ourselves.
595 			 */
596 			m = NULL;
597 			goto done;
598 		}
599 
600 		exthdrs.ip6e_ip6 = m;
601 	}
602 #endif /* IPSEC */
603 
604 	/* adjust pointer */
605 	ip6 = mtod(m, struct ip6_hdr *);
606 
607 	bzero(&dst_sa, sizeof(dst_sa));
608 	dst_sa.sin6_family = AF_INET6;
609 	dst_sa.sin6_len = sizeof(dst_sa);
610 	dst_sa.sin6_addr = ip6->ip6_dst;
611 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
612 	    &ifp, &rt, 0)) != 0) {
613 		switch (error) {
614 		case EHOSTUNREACH:
615 			ip6stat.ip6s_noroute++;
616 			break;
617 		case EADDRNOTAVAIL:
618 		default:
619 			break; /* XXX statistics? */
620 		}
621 		if (ifp != NULL)
622 			in6_ifstat_inc(ifp, ifs6_out_discard);
623 		goto bad;
624 	}
625 	if (rt == NULL) {
626 		/*
627 		 * If in6_selectroute() does not return a route entry,
628 		 * dst may not have been updated.
629 		 */
630 		*dst = dst_sa;	/* XXX */
631 	}
632 
633 	/*
634 	 * then rt (for unicast) and ifp must be non-NULL valid values.
635 	 */
636 	if ((flags & IPV6_FORWARDING) == 0) {
637 		/* XXX: the FORWARDING flag can be set for mrouting. */
638 		in6_ifstat_inc(ifp, ifs6_out_request);
639 	}
640 	if (rt != NULL) {
641 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
642 		rt->rt_use++;
643 	}
644 
645 	/*
646 	 * The outgoing interface must be in the zone of source and
647 	 * destination addresses.  We should use ia_ifp to support the
648 	 * case of sending packets to an address of our own.
649 	 */
650 	if (ia != NULL && ia->ia_ifp)
651 		origifp = ia->ia_ifp;
652 	else
653 		origifp = ifp;
654 
655 	src0 = ip6->ip6_src;
656 	if (in6_setscope(&src0, origifp, &zone))
657 		goto badscope;
658 	bzero(&src_sa, sizeof(src_sa));
659 	src_sa.sin6_family = AF_INET6;
660 	src_sa.sin6_len = sizeof(src_sa);
661 	src_sa.sin6_addr = ip6->ip6_src;
662 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
663 		goto badscope;
664 
665 	dst0 = ip6->ip6_dst;
666 	if (in6_setscope(&dst0, origifp, &zone))
667 		goto badscope;
668 	/* re-initialize to be sure */
669 	bzero(&dst_sa, sizeof(dst_sa));
670 	dst_sa.sin6_family = AF_INET6;
671 	dst_sa.sin6_len = sizeof(dst_sa);
672 	dst_sa.sin6_addr = ip6->ip6_dst;
673 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
674 		goto badscope;
675 	}
676 
677 	/* scope check is done. */
678 	goto routefound;
679 
680   badscope:
681 	ip6stat.ip6s_badscope++;
682 	in6_ifstat_inc(origifp, ifs6_out_discard);
683 	if (error == 0)
684 		error = EHOSTUNREACH; /* XXX */
685 	goto bad;
686 
687   routefound:
688 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
689 		if (opt && opt->ip6po_nextroute.ro_rt) {
690 			/*
691 			 * The nexthop is explicitly specified by the
692 			 * application.  We assume the next hop is an IPv6
693 			 * address.
694 			 */
695 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
696 		}
697 		else if ((rt->rt_flags & RTF_GATEWAY))
698 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
699 	}
700 
701 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
702 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
703 	} else {
704 		struct	in6_multi *in6m;
705 
706 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
707 
708 		in6_ifstat_inc(ifp, ifs6_out_mcast);
709 
710 		/*
711 		 * Confirm that the outgoing interface supports multicast.
712 		 */
713 		if (!(ifp->if_flags & IFF_MULTICAST)) {
714 			ip6stat.ip6s_noroute++;
715 			in6_ifstat_inc(ifp, ifs6_out_discard);
716 			error = ENETUNREACH;
717 			goto bad;
718 		}
719 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
720 		if (in6m != NULL &&
721 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
722 			/*
723 			 * If we belong to the destination multicast group
724 			 * on the outgoing interface, and the caller did not
725 			 * forbid loopback, loop back a copy.
726 			 */
727 			ip6_mloopback(ifp, m, dst);
728 		} else {
729 			/*
730 			 * If we are acting as a multicast router, perform
731 			 * multicast forwarding as if the packet had just
732 			 * arrived on the interface to which we are about
733 			 * to send.  The multicast forwarding function
734 			 * recursively calls this function, using the
735 			 * IPV6_FORWARDING flag to prevent infinite recursion.
736 			 *
737 			 * Multicasts that are looped back by ip6_mloopback(),
738 			 * above, will be forwarded by the ip6_input() routine,
739 			 * if necessary.
740 			 */
741 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
742 				/*
743 				 * XXX: ip6_mforward expects that rcvif is NULL
744 				 * when it is called from the originating path.
745 				 * However, it is not always the case, since
746 				 * some versions of MGETHDR() does not
747 				 * initialize the field.
748 				 */
749 				m->m_pkthdr.rcvif = NULL;
750 				if (ip6_mforward(ip6, ifp, m) != 0) {
751 					m_freem(m);
752 					goto done;
753 				}
754 			}
755 		}
756 		/*
757 		 * Multicasts with a hoplimit of zero may be looped back,
758 		 * above, but must not be transmitted on a network.
759 		 * Also, multicasts addressed to the loopback interface
760 		 * are not sent -- the above call to ip6_mloopback() will
761 		 * loop back a copy if this host actually belongs to the
762 		 * destination group on the loopback interface.
763 		 */
764 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
765 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
766 			m_freem(m);
767 			goto done;
768 		}
769 	}
770 
771 	/*
772 	 * Fill the outgoing inteface to tell the upper layer
773 	 * to increment per-interface statistics.
774 	 */
775 	if (ifpp)
776 		*ifpp = ifp;
777 
778 	/* Determine path MTU. */
779 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
780 	    &alwaysfrag)) != 0)
781 		goto bad;
782 
783 	/*
784 	 * The caller of this function may specify to use the minimum MTU
785 	 * in some cases.
786 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
787 	 * setting.  The logic is a bit complicated; by default, unicast
788 	 * packets will follow path MTU while multicast packets will be sent at
789 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
790 	 * including unicast ones will be sent at the minimum MTU.  Multicast
791 	 * packets will always be sent at the minimum MTU unless
792 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
793 	 * See RFC 3542 for more details.
794 	 */
795 	if (mtu > IPV6_MMTU) {
796 		if ((flags & IPV6_MINMTU))
797 			mtu = IPV6_MMTU;
798 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
799 			mtu = IPV6_MMTU;
800 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
801 			 (opt == NULL ||
802 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
803 			mtu = IPV6_MMTU;
804 		}
805 	}
806 
807 	/*
808 	 * clear embedded scope identifiers if necessary.
809 	 * in6_clearscope will touch the addresses only when necessary.
810 	 */
811 	in6_clearscope(&ip6->ip6_src);
812 	in6_clearscope(&ip6->ip6_dst);
813 
814 	/*
815 	 * If the outgoing packet contains a hop-by-hop options header,
816 	 * it must be examined and processed even by the source node.
817 	 * (RFC 2460, section 4.)
818 	 */
819 	if (exthdrs.ip6e_hbh) {
820 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
821 		u_int32_t dummy; /* XXX unused */
822 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
823 
824 #ifdef DIAGNOSTIC
825 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
826 			panic("ip6e_hbh is not continuous");
827 #endif
828 		/*
829 		 *  XXX: if we have to send an ICMPv6 error to the sender,
830 		 *       we need the M_LOOP flag since icmp6_error() expects
831 		 *       the IPv6 and the hop-by-hop options header are
832 		 *       continuous unless the flag is set.
833 		 */
834 		m->m_flags |= M_LOOP;
835 		m->m_pkthdr.rcvif = ifp;
836 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
837 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
838 		    &dummy, &plen) < 0) {
839 			/* m was already freed at this point */
840 			error = EINVAL;/* better error? */
841 			goto done;
842 		}
843 		m->m_flags &= ~M_LOOP; /* XXX */
844 		m->m_pkthdr.rcvif = NULL;
845 	}
846 
847 	/* Jump over all PFIL processing if hooks are not active. */
848 	if (!PFIL_HOOKED(&inet6_pfil_hook))
849 		goto passout;
850 
851 	odst = ip6->ip6_dst;
852 	/* Run through list of hooks for output packets. */
853 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
854 	if (error != 0 || m == NULL)
855 		goto done;
856 	ip6 = mtod(m, struct ip6_hdr *);
857 
858 	/* See if destination IP address was changed by packet filter. */
859 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
860 		m->m_flags |= M_SKIP_FIREWALL;
861 		/* If destination is now ourself drop to ip6_input(). */
862 		if (in6_localaddr(&ip6->ip6_dst)) {
863 			if (m->m_pkthdr.rcvif == NULL)
864 				m->m_pkthdr.rcvif = loif;
865 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
866 				m->m_pkthdr.csum_flags |=
867 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
868 				m->m_pkthdr.csum_data = 0xffff;
869 			}
870 			m->m_pkthdr.csum_flags |=
871 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
872 			error = netisr_queue(NETISR_IPV6, m);
873 			goto done;
874 		} else
875 			goto again;	/* Redo the routing table lookup. */
876 	}
877 
878 	/* XXX: IPFIREWALL_FORWARD */
879 
880 passout:
881 	/*
882 	 * Send the packet to the outgoing interface.
883 	 * If necessary, do IPv6 fragmentation before sending.
884 	 *
885 	 * the logic here is rather complex:
886 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
887 	 * 1-a:	send as is if tlen <= path mtu
888 	 * 1-b:	fragment if tlen > path mtu
889 	 *
890 	 * 2: if user asks us not to fragment (dontfrag == 1)
891 	 * 2-a:	send as is if tlen <= interface mtu
892 	 * 2-b:	error if tlen > interface mtu
893 	 *
894 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
895 	 *	always fragment
896 	 *
897 	 * 4: if dontfrag == 1 && alwaysfrag == 1
898 	 *	error, as we cannot handle this conflicting request
899 	 */
900 	tlen = m->m_pkthdr.len;
901 
902 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
903 		dontfrag = 1;
904 	else
905 		dontfrag = 0;
906 	if (dontfrag && alwaysfrag) {	/* case 4 */
907 		/* conflicting request - can't transmit */
908 		error = EMSGSIZE;
909 		goto bad;
910 	}
911 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
912 		/*
913 		 * Even if the DONTFRAG option is specified, we cannot send the
914 		 * packet when the data length is larger than the MTU of the
915 		 * outgoing interface.
916 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
917 		 * well as returning an error code (the latter is not described
918 		 * in the API spec.)
919 		 */
920 		u_int32_t mtu32;
921 		struct ip6ctlparam ip6cp;
922 
923 		mtu32 = (u_int32_t)mtu;
924 		bzero(&ip6cp, sizeof(ip6cp));
925 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
926 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
927 		    (void *)&ip6cp);
928 
929 		error = EMSGSIZE;
930 		goto bad;
931 	}
932 
933 	/*
934 	 * transmit packet without fragmentation
935 	 */
936 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
937 		struct in6_ifaddr *ia6;
938 
939 		ip6 = mtod(m, struct ip6_hdr *);
940 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
941 		if (ia6) {
942 			/* Record statistics for this interface address. */
943 			ia6->ia_ifa.if_opackets++;
944 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
945 		}
946 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
947 		goto done;
948 	}
949 
950 	/*
951 	 * try to fragment the packet.  case 1-b and 3
952 	 */
953 	if (mtu < IPV6_MMTU) {
954 		/* path MTU cannot be less than IPV6_MMTU */
955 		error = EMSGSIZE;
956 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
957 		goto bad;
958 	} else if (ip6->ip6_plen == 0) {
959 		/* jumbo payload cannot be fragmented */
960 		error = EMSGSIZE;
961 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
962 		goto bad;
963 	} else {
964 		struct mbuf **mnext, *m_frgpart;
965 		struct ip6_frag *ip6f;
966 		u_int32_t id = htonl(ip6_randomid());
967 		u_char nextproto;
968 
969 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
970 
971 		/*
972 		 * Too large for the destination or interface;
973 		 * fragment if possible.
974 		 * Must be able to put at least 8 bytes per fragment.
975 		 */
976 		hlen = unfragpartlen;
977 		if (mtu > IPV6_MAXPACKET)
978 			mtu = IPV6_MAXPACKET;
979 
980 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
981 		if (len < 8) {
982 			error = EMSGSIZE;
983 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
984 			goto bad;
985 		}
986 
987 		/*
988 		 * Verify that we have any chance at all of being able to queue
989 		 *      the packet or packet fragments
990 		 */
991 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
992 		    < tlen  /* - hlen */)) {
993 			error = ENOBUFS;
994 			ip6stat.ip6s_odropped++;
995 			goto bad;
996 		}
997 
998 		mnext = &m->m_nextpkt;
999 
1000 		/*
1001 		 * Change the next header field of the last header in the
1002 		 * unfragmentable part.
1003 		 */
1004 		if (exthdrs.ip6e_rthdr) {
1005 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1006 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1007 		} else if (exthdrs.ip6e_dest1) {
1008 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1009 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1010 		} else if (exthdrs.ip6e_hbh) {
1011 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1012 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1013 		} else {
1014 			nextproto = ip6->ip6_nxt;
1015 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1016 		}
1017 
1018 		/*
1019 		 * Loop through length of segment after first fragment,
1020 		 * make new header and copy data of each part and link onto
1021 		 * chain.
1022 		 */
1023 		m0 = m;
1024 		for (off = hlen; off < tlen; off += len) {
1025 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1026 			if (!m) {
1027 				error = ENOBUFS;
1028 				ip6stat.ip6s_odropped++;
1029 				goto sendorfree;
1030 			}
1031 			m->m_pkthdr.rcvif = NULL;
1032 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1033 			*mnext = m;
1034 			mnext = &m->m_nextpkt;
1035 			m->m_data += max_linkhdr;
1036 			mhip6 = mtod(m, struct ip6_hdr *);
1037 			*mhip6 = *ip6;
1038 			m->m_len = sizeof(*mhip6);
1039 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1040 			if (error) {
1041 				ip6stat.ip6s_odropped++;
1042 				goto sendorfree;
1043 			}
1044 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1045 			if (off + len >= tlen)
1046 				len = tlen - off;
1047 			else
1048 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1049 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1050 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1051 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1052 				error = ENOBUFS;
1053 				ip6stat.ip6s_odropped++;
1054 				goto sendorfree;
1055 			}
1056 			m_cat(m, m_frgpart);
1057 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1058 			m->m_pkthdr.rcvif = NULL;
1059 			ip6f->ip6f_reserved = 0;
1060 			ip6f->ip6f_ident = id;
1061 			ip6f->ip6f_nxt = nextproto;
1062 			ip6stat.ip6s_ofragments++;
1063 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1064 		}
1065 
1066 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1067 	}
1068 
1069 	/*
1070 	 * Remove leading garbages.
1071 	 */
1072 sendorfree:
1073 	m = m0->m_nextpkt;
1074 	m0->m_nextpkt = 0;
1075 	m_freem(m0);
1076 	for (m0 = m; m; m = m0) {
1077 		m0 = m->m_nextpkt;
1078 		m->m_nextpkt = 0;
1079 		if (error == 0) {
1080 			/* Record statistics for this interface address. */
1081 			if (ia) {
1082 				ia->ia_ifa.if_opackets++;
1083 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1084 			}
1085 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1086 		} else
1087 			m_freem(m);
1088 	}
1089 
1090 	if (error == 0)
1091 		ip6stat.ip6s_fragmented++;
1092 
1093 done:
1094 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1095 		RTFREE(ro->ro_rt);
1096 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1097 		RTFREE(ro_pmtu->ro_rt);
1098 	}
1099 #ifdef IPSEC
1100 	if (sp != NULL)
1101 		KEY_FREESP(&sp);
1102 #endif
1103 
1104 	return (error);
1105 
1106 freehdrs:
1107 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1108 	m_freem(exthdrs.ip6e_dest1);
1109 	m_freem(exthdrs.ip6e_rthdr);
1110 	m_freem(exthdrs.ip6e_dest2);
1111 	/* FALLTHROUGH */
1112 bad:
1113 	if (m)
1114 		m_freem(m);
1115 	goto done;
1116 }
1117 
1118 static int
1119 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1120 {
1121 	struct mbuf *m;
1122 
1123 	if (hlen > MCLBYTES)
1124 		return (ENOBUFS); /* XXX */
1125 
1126 	MGET(m, M_DONTWAIT, MT_DATA);
1127 	if (!m)
1128 		return (ENOBUFS);
1129 
1130 	if (hlen > MLEN) {
1131 		MCLGET(m, M_DONTWAIT);
1132 		if ((m->m_flags & M_EXT) == 0) {
1133 			m_free(m);
1134 			return (ENOBUFS);
1135 		}
1136 	}
1137 	m->m_len = hlen;
1138 	if (hdr)
1139 		bcopy(hdr, mtod(m, caddr_t), hlen);
1140 
1141 	*mp = m;
1142 	return (0);
1143 }
1144 
1145 /*
1146  * Insert jumbo payload option.
1147  */
1148 static int
1149 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1150 {
1151 	struct mbuf *mopt;
1152 	u_char *optbuf;
1153 	u_int32_t v;
1154 
1155 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1156 
1157 	/*
1158 	 * If there is no hop-by-hop options header, allocate new one.
1159 	 * If there is one but it doesn't have enough space to store the
1160 	 * jumbo payload option, allocate a cluster to store the whole options.
1161 	 * Otherwise, use it to store the options.
1162 	 */
1163 	if (exthdrs->ip6e_hbh == 0) {
1164 		MGET(mopt, M_DONTWAIT, MT_DATA);
1165 		if (mopt == 0)
1166 			return (ENOBUFS);
1167 		mopt->m_len = JUMBOOPTLEN;
1168 		optbuf = mtod(mopt, u_char *);
1169 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1170 		exthdrs->ip6e_hbh = mopt;
1171 	} else {
1172 		struct ip6_hbh *hbh;
1173 
1174 		mopt = exthdrs->ip6e_hbh;
1175 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1176 			/*
1177 			 * XXX assumption:
1178 			 * - exthdrs->ip6e_hbh is not referenced from places
1179 			 *   other than exthdrs.
1180 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1181 			 */
1182 			int oldoptlen = mopt->m_len;
1183 			struct mbuf *n;
1184 
1185 			/*
1186 			 * XXX: give up if the whole (new) hbh header does
1187 			 * not fit even in an mbuf cluster.
1188 			 */
1189 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1190 				return (ENOBUFS);
1191 
1192 			/*
1193 			 * As a consequence, we must always prepare a cluster
1194 			 * at this point.
1195 			 */
1196 			MGET(n, M_DONTWAIT, MT_DATA);
1197 			if (n) {
1198 				MCLGET(n, M_DONTWAIT);
1199 				if ((n->m_flags & M_EXT) == 0) {
1200 					m_freem(n);
1201 					n = NULL;
1202 				}
1203 			}
1204 			if (!n)
1205 				return (ENOBUFS);
1206 			n->m_len = oldoptlen + JUMBOOPTLEN;
1207 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1208 			    oldoptlen);
1209 			optbuf = mtod(n, caddr_t) + oldoptlen;
1210 			m_freem(mopt);
1211 			mopt = exthdrs->ip6e_hbh = n;
1212 		} else {
1213 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1214 			mopt->m_len += JUMBOOPTLEN;
1215 		}
1216 		optbuf[0] = IP6OPT_PADN;
1217 		optbuf[1] = 1;
1218 
1219 		/*
1220 		 * Adjust the header length according to the pad and
1221 		 * the jumbo payload option.
1222 		 */
1223 		hbh = mtod(mopt, struct ip6_hbh *);
1224 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1225 	}
1226 
1227 	/* fill in the option. */
1228 	optbuf[2] = IP6OPT_JUMBO;
1229 	optbuf[3] = 4;
1230 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1231 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1232 
1233 	/* finally, adjust the packet header length */
1234 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1235 
1236 	return (0);
1237 #undef JUMBOOPTLEN
1238 }
1239 
1240 /*
1241  * Insert fragment header and copy unfragmentable header portions.
1242  */
1243 static int
1244 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1245     struct ip6_frag **frghdrp)
1246 {
1247 	struct mbuf *n, *mlast;
1248 
1249 	if (hlen > sizeof(struct ip6_hdr)) {
1250 		n = m_copym(m0, sizeof(struct ip6_hdr),
1251 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1252 		if (n == 0)
1253 			return (ENOBUFS);
1254 		m->m_next = n;
1255 	} else
1256 		n = m;
1257 
1258 	/* Search for the last mbuf of unfragmentable part. */
1259 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1260 		;
1261 
1262 	if ((mlast->m_flags & M_EXT) == 0 &&
1263 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1264 		/* use the trailing space of the last mbuf for the fragment hdr */
1265 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1266 		    mlast->m_len);
1267 		mlast->m_len += sizeof(struct ip6_frag);
1268 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1269 	} else {
1270 		/* allocate a new mbuf for the fragment header */
1271 		struct mbuf *mfrg;
1272 
1273 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1274 		if (mfrg == 0)
1275 			return (ENOBUFS);
1276 		mfrg->m_len = sizeof(struct ip6_frag);
1277 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1278 		mlast->m_next = mfrg;
1279 	}
1280 
1281 	return (0);
1282 }
1283 
1284 static int
1285 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1286     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1287     int *alwaysfragp)
1288 {
1289 	u_int32_t mtu = 0;
1290 	int alwaysfrag = 0;
1291 	int error = 0;
1292 
1293 	if (ro_pmtu != ro) {
1294 		/* The first hop and the final destination may differ. */
1295 		struct sockaddr_in6 *sa6_dst =
1296 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1297 		if (ro_pmtu->ro_rt &&
1298 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1299 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1300 			RTFREE(ro_pmtu->ro_rt);
1301 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1302 		}
1303 		if (ro_pmtu->ro_rt == NULL) {
1304 			bzero(sa6_dst, sizeof(*sa6_dst));
1305 			sa6_dst->sin6_family = AF_INET6;
1306 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1307 			sa6_dst->sin6_addr = *dst;
1308 
1309 			rtalloc((struct route *)ro_pmtu);
1310 		}
1311 	}
1312 	if (ro_pmtu->ro_rt) {
1313 		u_int32_t ifmtu;
1314 		struct in_conninfo inc;
1315 
1316 		bzero(&inc, sizeof(inc));
1317 		inc.inc_flags = 1; /* IPv6 */
1318 		inc.inc6_faddr = *dst;
1319 
1320 		if (ifp == NULL)
1321 			ifp = ro_pmtu->ro_rt->rt_ifp;
1322 		ifmtu = IN6_LINKMTU(ifp);
1323 		mtu = tcp_hc_getmtu(&inc);
1324 		if (mtu)
1325 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1326 		else
1327 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1328 		if (mtu == 0)
1329 			mtu = ifmtu;
1330 		else if (mtu < IPV6_MMTU) {
1331 			/*
1332 			 * RFC2460 section 5, last paragraph:
1333 			 * if we record ICMPv6 too big message with
1334 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1335 			 * or smaller, with framgent header attached.
1336 			 * (fragment header is needed regardless from the
1337 			 * packet size, for translators to identify packets)
1338 			 */
1339 			alwaysfrag = 1;
1340 			mtu = IPV6_MMTU;
1341 		} else if (mtu > ifmtu) {
1342 			/*
1343 			 * The MTU on the route is larger than the MTU on
1344 			 * the interface!  This shouldn't happen, unless the
1345 			 * MTU of the interface has been changed after the
1346 			 * interface was brought up.  Change the MTU in the
1347 			 * route to match the interface MTU (as long as the
1348 			 * field isn't locked).
1349 			 */
1350 			mtu = ifmtu;
1351 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1352 		}
1353 	} else if (ifp) {
1354 		mtu = IN6_LINKMTU(ifp);
1355 	} else
1356 		error = EHOSTUNREACH; /* XXX */
1357 
1358 	*mtup = mtu;
1359 	if (alwaysfragp)
1360 		*alwaysfragp = alwaysfrag;
1361 	return (error);
1362 }
1363 
1364 /*
1365  * IP6 socket option processing.
1366  */
1367 int
1368 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1369 {
1370 	int optdatalen, uproto;
1371 	void *optdata;
1372 	struct inpcb *in6p = sotoinpcb(so);
1373 	int error, optval;
1374 	int level, op, optname;
1375 	int optlen;
1376 	struct thread *td;
1377 
1378 	if (sopt) {
1379 		level = sopt->sopt_level;
1380 		op = sopt->sopt_dir;
1381 		optname = sopt->sopt_name;
1382 		optlen = sopt->sopt_valsize;
1383 		td = sopt->sopt_td;
1384 	} else {
1385 		panic("ip6_ctloutput: arg soopt is NULL");
1386 	}
1387 	error = optval = 0;
1388 
1389 	uproto = (int)so->so_proto->pr_protocol;
1390 
1391 	if (level == IPPROTO_IPV6) {
1392 		switch (op) {
1393 
1394 		case SOPT_SET:
1395 			switch (optname) {
1396 			case IPV6_2292PKTOPTIONS:
1397 #ifdef IPV6_PKTOPTIONS
1398 			case IPV6_PKTOPTIONS:
1399 #endif
1400 			{
1401 				struct mbuf *m;
1402 
1403 				error = soopt_getm(sopt, &m); /* XXX */
1404 				if (error != 0)
1405 					break;
1406 				error = soopt_mcopyin(sopt, m); /* XXX */
1407 				if (error != 0)
1408 					break;
1409 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1410 						    m, so, sopt);
1411 				m_freem(m); /* XXX */
1412 				break;
1413 			}
1414 
1415 			/*
1416 			 * Use of some Hop-by-Hop options or some
1417 			 * Destination options, might require special
1418 			 * privilege.  That is, normal applications
1419 			 * (without special privilege) might be forbidden
1420 			 * from setting certain options in outgoing packets,
1421 			 * and might never see certain options in received
1422 			 * packets. [RFC 2292 Section 6]
1423 			 * KAME specific note:
1424 			 *  KAME prevents non-privileged users from sending or
1425 			 *  receiving ANY hbh/dst options in order to avoid
1426 			 *  overhead of parsing options in the kernel.
1427 			 */
1428 			case IPV6_RECVHOPOPTS:
1429 			case IPV6_RECVDSTOPTS:
1430 			case IPV6_RECVRTHDRDSTOPTS:
1431 				if (td != NULL) {
1432 					error = priv_check(td,
1433 					    PRIV_NETINET_SETHDROPTS);
1434 					if (error)
1435 						break;
1436 				}
1437 				/* FALLTHROUGH */
1438 			case IPV6_UNICAST_HOPS:
1439 			case IPV6_HOPLIMIT:
1440 			case IPV6_FAITH:
1441 
1442 			case IPV6_RECVPKTINFO:
1443 			case IPV6_RECVHOPLIMIT:
1444 			case IPV6_RECVRTHDR:
1445 			case IPV6_RECVPATHMTU:
1446 			case IPV6_RECVTCLASS:
1447 			case IPV6_V6ONLY:
1448 			case IPV6_AUTOFLOWLABEL:
1449 				if (optlen != sizeof(int)) {
1450 					error = EINVAL;
1451 					break;
1452 				}
1453 				error = sooptcopyin(sopt, &optval,
1454 					sizeof optval, sizeof optval);
1455 				if (error)
1456 					break;
1457 				switch (optname) {
1458 
1459 				case IPV6_UNICAST_HOPS:
1460 					if (optval < -1 || optval >= 256)
1461 						error = EINVAL;
1462 					else {
1463 						/* -1 = kernel default */
1464 						in6p->in6p_hops = optval;
1465 						if ((in6p->in6p_vflag &
1466 						     INP_IPV4) != 0)
1467 							in6p->inp_ip_ttl = optval;
1468 					}
1469 					break;
1470 #define OPTSET(bit) \
1471 do { \
1472 	if (optval) \
1473 		in6p->in6p_flags |= (bit); \
1474 	else \
1475 		in6p->in6p_flags &= ~(bit); \
1476 } while (/*CONSTCOND*/ 0)
1477 #define OPTSET2292(bit) \
1478 do { \
1479 	in6p->in6p_flags |= IN6P_RFC2292; \
1480 	if (optval) \
1481 		in6p->in6p_flags |= (bit); \
1482 	else \
1483 		in6p->in6p_flags &= ~(bit); \
1484 } while (/*CONSTCOND*/ 0)
1485 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1486 
1487 				case IPV6_RECVPKTINFO:
1488 					/* cannot mix with RFC2292 */
1489 					if (OPTBIT(IN6P_RFC2292)) {
1490 						error = EINVAL;
1491 						break;
1492 					}
1493 					OPTSET(IN6P_PKTINFO);
1494 					break;
1495 
1496 				case IPV6_HOPLIMIT:
1497 				{
1498 					struct ip6_pktopts **optp;
1499 
1500 					/* cannot mix with RFC2292 */
1501 					if (OPTBIT(IN6P_RFC2292)) {
1502 						error = EINVAL;
1503 						break;
1504 					}
1505 					optp = &in6p->in6p_outputopts;
1506 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1507 					    (u_char *)&optval, sizeof(optval),
1508 					    optp, (td != NULL) ? td->td_ucred :
1509 					    NULL, uproto);
1510 					break;
1511 				}
1512 
1513 				case IPV6_RECVHOPLIMIT:
1514 					/* cannot mix with RFC2292 */
1515 					if (OPTBIT(IN6P_RFC2292)) {
1516 						error = EINVAL;
1517 						break;
1518 					}
1519 					OPTSET(IN6P_HOPLIMIT);
1520 					break;
1521 
1522 				case IPV6_RECVHOPOPTS:
1523 					/* cannot mix with RFC2292 */
1524 					if (OPTBIT(IN6P_RFC2292)) {
1525 						error = EINVAL;
1526 						break;
1527 					}
1528 					OPTSET(IN6P_HOPOPTS);
1529 					break;
1530 
1531 				case IPV6_RECVDSTOPTS:
1532 					/* cannot mix with RFC2292 */
1533 					if (OPTBIT(IN6P_RFC2292)) {
1534 						error = EINVAL;
1535 						break;
1536 					}
1537 					OPTSET(IN6P_DSTOPTS);
1538 					break;
1539 
1540 				case IPV6_RECVRTHDRDSTOPTS:
1541 					/* cannot mix with RFC2292 */
1542 					if (OPTBIT(IN6P_RFC2292)) {
1543 						error = EINVAL;
1544 						break;
1545 					}
1546 					OPTSET(IN6P_RTHDRDSTOPTS);
1547 					break;
1548 
1549 				case IPV6_RECVRTHDR:
1550 					/* cannot mix with RFC2292 */
1551 					if (OPTBIT(IN6P_RFC2292)) {
1552 						error = EINVAL;
1553 						break;
1554 					}
1555 					OPTSET(IN6P_RTHDR);
1556 					break;
1557 
1558 				case IPV6_FAITH:
1559 					OPTSET(IN6P_FAITH);
1560 					break;
1561 
1562 				case IPV6_RECVPATHMTU:
1563 					/*
1564 					 * We ignore this option for TCP
1565 					 * sockets.
1566 					 * (RFC3542 leaves this case
1567 					 * unspecified.)
1568 					 */
1569 					if (uproto != IPPROTO_TCP)
1570 						OPTSET(IN6P_MTU);
1571 					break;
1572 
1573 				case IPV6_V6ONLY:
1574 					/*
1575 					 * make setsockopt(IPV6_V6ONLY)
1576 					 * available only prior to bind(2).
1577 					 * see ipng mailing list, Jun 22 2001.
1578 					 */
1579 					if (in6p->in6p_lport ||
1580 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1581 						error = EINVAL;
1582 						break;
1583 					}
1584 					OPTSET(IN6P_IPV6_V6ONLY);
1585 					if (optval)
1586 						in6p->in6p_vflag &= ~INP_IPV4;
1587 					else
1588 						in6p->in6p_vflag |= INP_IPV4;
1589 					break;
1590 				case IPV6_RECVTCLASS:
1591 					/* cannot mix with RFC2292 XXX */
1592 					if (OPTBIT(IN6P_RFC2292)) {
1593 						error = EINVAL;
1594 						break;
1595 					}
1596 					OPTSET(IN6P_TCLASS);
1597 					break;
1598 				case IPV6_AUTOFLOWLABEL:
1599 					OPTSET(IN6P_AUTOFLOWLABEL);
1600 					break;
1601 
1602 				}
1603 				break;
1604 
1605 			case IPV6_TCLASS:
1606 			case IPV6_DONTFRAG:
1607 			case IPV6_USE_MIN_MTU:
1608 			case IPV6_PREFER_TEMPADDR:
1609 				if (optlen != sizeof(optval)) {
1610 					error = EINVAL;
1611 					break;
1612 				}
1613 				error = sooptcopyin(sopt, &optval,
1614 					sizeof optval, sizeof optval);
1615 				if (error)
1616 					break;
1617 				{
1618 					struct ip6_pktopts **optp;
1619 					optp = &in6p->in6p_outputopts;
1620 					error = ip6_pcbopt(optname,
1621 					    (u_char *)&optval, sizeof(optval),
1622 					    optp, (td != NULL) ? td->td_ucred :
1623 					    NULL, uproto);
1624 					break;
1625 				}
1626 
1627 			case IPV6_2292PKTINFO:
1628 			case IPV6_2292HOPLIMIT:
1629 			case IPV6_2292HOPOPTS:
1630 			case IPV6_2292DSTOPTS:
1631 			case IPV6_2292RTHDR:
1632 				/* RFC 2292 */
1633 				if (optlen != sizeof(int)) {
1634 					error = EINVAL;
1635 					break;
1636 				}
1637 				error = sooptcopyin(sopt, &optval,
1638 					sizeof optval, sizeof optval);
1639 				if (error)
1640 					break;
1641 				switch (optname) {
1642 				case IPV6_2292PKTINFO:
1643 					OPTSET2292(IN6P_PKTINFO);
1644 					break;
1645 				case IPV6_2292HOPLIMIT:
1646 					OPTSET2292(IN6P_HOPLIMIT);
1647 					break;
1648 				case IPV6_2292HOPOPTS:
1649 					/*
1650 					 * Check super-user privilege.
1651 					 * See comments for IPV6_RECVHOPOPTS.
1652 					 */
1653 					if (td != NULL) {
1654 						error = priv_check(td,
1655 						    PRIV_NETINET_SETHDROPTS);
1656 						if (error)
1657 							return (error);
1658 					}
1659 					OPTSET2292(IN6P_HOPOPTS);
1660 					break;
1661 				case IPV6_2292DSTOPTS:
1662 					if (td != NULL) {
1663 						error = priv_check(td,
1664 						    PRIV_NETINET_SETHDROPTS);
1665 						if (error)
1666 							return (error);
1667 					}
1668 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1669 					break;
1670 				case IPV6_2292RTHDR:
1671 					OPTSET2292(IN6P_RTHDR);
1672 					break;
1673 				}
1674 				break;
1675 			case IPV6_PKTINFO:
1676 			case IPV6_HOPOPTS:
1677 			case IPV6_RTHDR:
1678 			case IPV6_DSTOPTS:
1679 			case IPV6_RTHDRDSTOPTS:
1680 			case IPV6_NEXTHOP:
1681 			{
1682 				/* new advanced API (RFC3542) */
1683 				u_char *optbuf;
1684 				u_char optbuf_storage[MCLBYTES];
1685 				int optlen;
1686 				struct ip6_pktopts **optp;
1687 
1688 				/* cannot mix with RFC2292 */
1689 				if (OPTBIT(IN6P_RFC2292)) {
1690 					error = EINVAL;
1691 					break;
1692 				}
1693 
1694 				/*
1695 				 * We only ensure valsize is not too large
1696 				 * here.  Further validation will be done
1697 				 * later.
1698 				 */
1699 				error = sooptcopyin(sopt, optbuf_storage,
1700 				    sizeof(optbuf_storage), 0);
1701 				if (error)
1702 					break;
1703 				optlen = sopt->sopt_valsize;
1704 				optbuf = optbuf_storage;
1705 				optp = &in6p->in6p_outputopts;
1706 				error = ip6_pcbopt(optname, optbuf, optlen,
1707 				    optp, (td != NULL) ? td->td_ucred : NULL,
1708 				    uproto);
1709 				break;
1710 			}
1711 #undef OPTSET
1712 
1713 			case IPV6_MULTICAST_IF:
1714 			case IPV6_MULTICAST_HOPS:
1715 			case IPV6_MULTICAST_LOOP:
1716 			case IPV6_JOIN_GROUP:
1717 			case IPV6_LEAVE_GROUP:
1718 			    {
1719 				if (sopt->sopt_valsize > MLEN) {
1720 					error = EMSGSIZE;
1721 					break;
1722 				}
1723 				/* XXX */
1724 			    }
1725 			    /* FALLTHROUGH */
1726 			    {
1727 				struct mbuf *m;
1728 
1729 				if (sopt->sopt_valsize > MCLBYTES) {
1730 					error = EMSGSIZE;
1731 					break;
1732 				}
1733 				/* XXX */
1734 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1735 				if (m == 0) {
1736 					error = ENOBUFS;
1737 					break;
1738 				}
1739 				if (sopt->sopt_valsize > MLEN) {
1740 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1741 					if ((m->m_flags & M_EXT) == 0) {
1742 						m_free(m);
1743 						error = ENOBUFS;
1744 						break;
1745 					}
1746 				}
1747 				m->m_len = sopt->sopt_valsize;
1748 				error = sooptcopyin(sopt, mtod(m, char *),
1749 						    m->m_len, m->m_len);
1750 				if (error) {
1751 					(void)m_free(m);
1752 					break;
1753 				}
1754 				error =	ip6_setmoptions(sopt->sopt_name,
1755 							&in6p->in6p_moptions,
1756 							m);
1757 				(void)m_free(m);
1758 			    }
1759 				break;
1760 
1761 			case IPV6_PORTRANGE:
1762 				error = sooptcopyin(sopt, &optval,
1763 				    sizeof optval, sizeof optval);
1764 				if (error)
1765 					break;
1766 
1767 				switch (optval) {
1768 				case IPV6_PORTRANGE_DEFAULT:
1769 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1770 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1771 					break;
1772 
1773 				case IPV6_PORTRANGE_HIGH:
1774 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1775 					in6p->in6p_flags |= IN6P_HIGHPORT;
1776 					break;
1777 
1778 				case IPV6_PORTRANGE_LOW:
1779 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1780 					in6p->in6p_flags |= IN6P_LOWPORT;
1781 					break;
1782 
1783 				default:
1784 					error = EINVAL;
1785 					break;
1786 				}
1787 				break;
1788 
1789 #ifdef IPSEC
1790 			case IPV6_IPSEC_POLICY:
1791 			{
1792 				caddr_t req;
1793 				struct mbuf *m;
1794 
1795 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1796 					break;
1797 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1798 					break;
1799 				req = mtod(m, caddr_t);
1800 				error = ipsec6_set_policy(in6p, optname, req,
1801 				    m->m_len, (sopt->sopt_td != NULL) ?
1802 				    sopt->sopt_td->td_ucred : NULL);
1803 				m_freem(m);
1804 				break;
1805 			}
1806 #endif /* IPSEC */
1807 
1808 			default:
1809 				error = ENOPROTOOPT;
1810 				break;
1811 			}
1812 			break;
1813 
1814 		case SOPT_GET:
1815 			switch (optname) {
1816 
1817 			case IPV6_2292PKTOPTIONS:
1818 #ifdef IPV6_PKTOPTIONS
1819 			case IPV6_PKTOPTIONS:
1820 #endif
1821 				/*
1822 				 * RFC3542 (effectively) deprecated the
1823 				 * semantics of the 2292-style pktoptions.
1824 				 * Since it was not reliable in nature (i.e.,
1825 				 * applications had to expect the lack of some
1826 				 * information after all), it would make sense
1827 				 * to simplify this part by always returning
1828 				 * empty data.
1829 				 */
1830 				sopt->sopt_valsize = 0;
1831 				break;
1832 
1833 			case IPV6_RECVHOPOPTS:
1834 			case IPV6_RECVDSTOPTS:
1835 			case IPV6_RECVRTHDRDSTOPTS:
1836 			case IPV6_UNICAST_HOPS:
1837 			case IPV6_RECVPKTINFO:
1838 			case IPV6_RECVHOPLIMIT:
1839 			case IPV6_RECVRTHDR:
1840 			case IPV6_RECVPATHMTU:
1841 
1842 			case IPV6_FAITH:
1843 			case IPV6_V6ONLY:
1844 			case IPV6_PORTRANGE:
1845 			case IPV6_RECVTCLASS:
1846 			case IPV6_AUTOFLOWLABEL:
1847 				switch (optname) {
1848 
1849 				case IPV6_RECVHOPOPTS:
1850 					optval = OPTBIT(IN6P_HOPOPTS);
1851 					break;
1852 
1853 				case IPV6_RECVDSTOPTS:
1854 					optval = OPTBIT(IN6P_DSTOPTS);
1855 					break;
1856 
1857 				case IPV6_RECVRTHDRDSTOPTS:
1858 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1859 					break;
1860 
1861 				case IPV6_UNICAST_HOPS:
1862 					optval = in6p->in6p_hops;
1863 					break;
1864 
1865 				case IPV6_RECVPKTINFO:
1866 					optval = OPTBIT(IN6P_PKTINFO);
1867 					break;
1868 
1869 				case IPV6_RECVHOPLIMIT:
1870 					optval = OPTBIT(IN6P_HOPLIMIT);
1871 					break;
1872 
1873 				case IPV6_RECVRTHDR:
1874 					optval = OPTBIT(IN6P_RTHDR);
1875 					break;
1876 
1877 				case IPV6_RECVPATHMTU:
1878 					optval = OPTBIT(IN6P_MTU);
1879 					break;
1880 
1881 				case IPV6_FAITH:
1882 					optval = OPTBIT(IN6P_FAITH);
1883 					break;
1884 
1885 				case IPV6_V6ONLY:
1886 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1887 					break;
1888 
1889 				case IPV6_PORTRANGE:
1890 				    {
1891 					int flags;
1892 					flags = in6p->in6p_flags;
1893 					if (flags & IN6P_HIGHPORT)
1894 						optval = IPV6_PORTRANGE_HIGH;
1895 					else if (flags & IN6P_LOWPORT)
1896 						optval = IPV6_PORTRANGE_LOW;
1897 					else
1898 						optval = 0;
1899 					break;
1900 				    }
1901 				case IPV6_RECVTCLASS:
1902 					optval = OPTBIT(IN6P_TCLASS);
1903 					break;
1904 
1905 				case IPV6_AUTOFLOWLABEL:
1906 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1907 					break;
1908 				}
1909 				if (error)
1910 					break;
1911 				error = sooptcopyout(sopt, &optval,
1912 					sizeof optval);
1913 				break;
1914 
1915 			case IPV6_PATHMTU:
1916 			{
1917 				u_long pmtu = 0;
1918 				struct ip6_mtuinfo mtuinfo;
1919 				struct route_in6 sro;
1920 
1921 				bzero(&sro, sizeof(sro));
1922 
1923 				if (!(so->so_state & SS_ISCONNECTED))
1924 					return (ENOTCONN);
1925 				/*
1926 				 * XXX: we dot not consider the case of source
1927 				 * routing, or optional information to specify
1928 				 * the outgoing interface.
1929 				 */
1930 				error = ip6_getpmtu(&sro, NULL, NULL,
1931 				    &in6p->in6p_faddr, &pmtu, NULL);
1932 				if (sro.ro_rt)
1933 					RTFREE(sro.ro_rt);
1934 				if (error)
1935 					break;
1936 				if (pmtu > IPV6_MAXPACKET)
1937 					pmtu = IPV6_MAXPACKET;
1938 
1939 				bzero(&mtuinfo, sizeof(mtuinfo));
1940 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1941 				optdata = (void *)&mtuinfo;
1942 				optdatalen = sizeof(mtuinfo);
1943 				error = sooptcopyout(sopt, optdata,
1944 				    optdatalen);
1945 				break;
1946 			}
1947 
1948 			case IPV6_2292PKTINFO:
1949 			case IPV6_2292HOPLIMIT:
1950 			case IPV6_2292HOPOPTS:
1951 			case IPV6_2292RTHDR:
1952 			case IPV6_2292DSTOPTS:
1953 				switch (optname) {
1954 				case IPV6_2292PKTINFO:
1955 					optval = OPTBIT(IN6P_PKTINFO);
1956 					break;
1957 				case IPV6_2292HOPLIMIT:
1958 					optval = OPTBIT(IN6P_HOPLIMIT);
1959 					break;
1960 				case IPV6_2292HOPOPTS:
1961 					optval = OPTBIT(IN6P_HOPOPTS);
1962 					break;
1963 				case IPV6_2292RTHDR:
1964 					optval = OPTBIT(IN6P_RTHDR);
1965 					break;
1966 				case IPV6_2292DSTOPTS:
1967 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1968 					break;
1969 				}
1970 				error = sooptcopyout(sopt, &optval,
1971 				    sizeof optval);
1972 				break;
1973 			case IPV6_PKTINFO:
1974 			case IPV6_HOPOPTS:
1975 			case IPV6_RTHDR:
1976 			case IPV6_DSTOPTS:
1977 			case IPV6_RTHDRDSTOPTS:
1978 			case IPV6_NEXTHOP:
1979 			case IPV6_TCLASS:
1980 			case IPV6_DONTFRAG:
1981 			case IPV6_USE_MIN_MTU:
1982 			case IPV6_PREFER_TEMPADDR:
1983 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1984 				    optname, sopt);
1985 				break;
1986 
1987 			case IPV6_MULTICAST_IF:
1988 			case IPV6_MULTICAST_HOPS:
1989 			case IPV6_MULTICAST_LOOP:
1990 			case IPV6_JOIN_GROUP:
1991 			case IPV6_LEAVE_GROUP:
1992 			    {
1993 				struct mbuf *m;
1994 				error = ip6_getmoptions(sopt->sopt_name,
1995 				    in6p->in6p_moptions, &m);
1996 				if (error == 0)
1997 					error = sooptcopyout(sopt,
1998 					    mtod(m, char *), m->m_len);
1999 				m_freem(m);
2000 			    }
2001 				break;
2002 
2003 #ifdef IPSEC
2004 			case IPV6_IPSEC_POLICY:
2005 			  {
2006 				caddr_t req = NULL;
2007 				size_t len = 0;
2008 				struct mbuf *m = NULL;
2009 				struct mbuf **mp = &m;
2010 				size_t ovalsize = sopt->sopt_valsize;
2011 				caddr_t oval = (caddr_t)sopt->sopt_val;
2012 
2013 				error = soopt_getm(sopt, &m); /* XXX */
2014 				if (error != 0)
2015 					break;
2016 				error = soopt_mcopyin(sopt, m); /* XXX */
2017 				if (error != 0)
2018 					break;
2019 				sopt->sopt_valsize = ovalsize;
2020 				sopt->sopt_val = oval;
2021 				if (m) {
2022 					req = mtod(m, caddr_t);
2023 					len = m->m_len;
2024 				}
2025 				error = ipsec6_get_policy(in6p, req, len, mp);
2026 				if (error == 0)
2027 					error = soopt_mcopyout(sopt, m); /* XXX */
2028 				if (error == 0 && m)
2029 					m_freem(m);
2030 				break;
2031 			  }
2032 #endif /* IPSEC */
2033 
2034 			default:
2035 				error = ENOPROTOOPT;
2036 				break;
2037 			}
2038 			break;
2039 		}
2040 	} else {		/* level != IPPROTO_IPV6 */
2041 		error = EINVAL;
2042 	}
2043 	return (error);
2044 }
2045 
2046 int
2047 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2048 {
2049 	int error = 0, optval, optlen;
2050 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2051 	struct in6pcb *in6p = sotoin6pcb(so);
2052 	int level, op, optname;
2053 
2054 	if (sopt) {
2055 		level = sopt->sopt_level;
2056 		op = sopt->sopt_dir;
2057 		optname = sopt->sopt_name;
2058 		optlen = sopt->sopt_valsize;
2059 	} else
2060 		panic("ip6_raw_ctloutput: arg soopt is NULL");
2061 
2062 	if (level != IPPROTO_IPV6) {
2063 		return (EINVAL);
2064 	}
2065 
2066 	switch (optname) {
2067 	case IPV6_CHECKSUM:
2068 		/*
2069 		 * For ICMPv6 sockets, no modification allowed for checksum
2070 		 * offset, permit "no change" values to help existing apps.
2071 		 *
2072 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2073 		 * for an ICMPv6 socket will fail."
2074 		 * The current behavior does not meet RFC3542.
2075 		 */
2076 		switch (op) {
2077 		case SOPT_SET:
2078 			if (optlen != sizeof(int)) {
2079 				error = EINVAL;
2080 				break;
2081 			}
2082 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2083 					    sizeof(optval));
2084 			if (error)
2085 				break;
2086 			if ((optval % 2) != 0) {
2087 				/* the API assumes even offset values */
2088 				error = EINVAL;
2089 			} else if (so->so_proto->pr_protocol ==
2090 			    IPPROTO_ICMPV6) {
2091 				if (optval != icmp6off)
2092 					error = EINVAL;
2093 			} else
2094 				in6p->in6p_cksum = optval;
2095 			break;
2096 
2097 		case SOPT_GET:
2098 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2099 				optval = icmp6off;
2100 			else
2101 				optval = in6p->in6p_cksum;
2102 
2103 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2104 			break;
2105 
2106 		default:
2107 			error = EINVAL;
2108 			break;
2109 		}
2110 		break;
2111 
2112 	default:
2113 		error = ENOPROTOOPT;
2114 		break;
2115 	}
2116 
2117 	return (error);
2118 }
2119 
2120 /*
2121  * Set up IP6 options in pcb for insertion in output packets or
2122  * specifying behavior of outgoing packets.
2123  */
2124 static int
2125 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2126     struct socket *so, struct sockopt *sopt)
2127 {
2128 	struct ip6_pktopts *opt = *pktopt;
2129 	int error = 0;
2130 	struct thread *td = sopt->sopt_td;
2131 
2132 	/* turn off any old options. */
2133 	if (opt) {
2134 #ifdef DIAGNOSTIC
2135 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2136 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2137 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2138 			printf("ip6_pcbopts: all specified options are cleared.\n");
2139 #endif
2140 		ip6_clearpktopts(opt, -1);
2141 	} else
2142 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2143 	*pktopt = NULL;
2144 
2145 	if (!m || m->m_len == 0) {
2146 		/*
2147 		 * Only turning off any previous options, regardless of
2148 		 * whether the opt is just created or given.
2149 		 */
2150 		free(opt, M_IP6OPT);
2151 		return (0);
2152 	}
2153 
2154 	/*  set options specified by user. */
2155 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2156 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2157 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2158 		free(opt, M_IP6OPT);
2159 		return (error);
2160 	}
2161 	*pktopt = opt;
2162 	return (0);
2163 }
2164 
2165 /*
2166  * initialize ip6_pktopts.  beware that there are non-zero default values in
2167  * the struct.
2168  */
2169 void
2170 ip6_initpktopts(struct ip6_pktopts *opt)
2171 {
2172 
2173 	bzero(opt, sizeof(*opt));
2174 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2175 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2176 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2177 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2178 }
2179 
2180 static int
2181 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2182     struct ucred *cred, int uproto)
2183 {
2184 	struct ip6_pktopts *opt;
2185 
2186 	if (*pktopt == NULL) {
2187 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2188 		    M_WAITOK);
2189 		ip6_initpktopts(*pktopt);
2190 	}
2191 	opt = *pktopt;
2192 
2193 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2194 }
2195 
2196 static int
2197 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2198 {
2199 	void *optdata = NULL;
2200 	int optdatalen = 0;
2201 	struct ip6_ext *ip6e;
2202 	int error = 0;
2203 	struct in6_pktinfo null_pktinfo;
2204 	int deftclass = 0, on;
2205 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2206 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2207 
2208 	switch (optname) {
2209 	case IPV6_PKTINFO:
2210 		if (pktopt && pktopt->ip6po_pktinfo)
2211 			optdata = (void *)pktopt->ip6po_pktinfo;
2212 		else {
2213 			/* XXX: we don't have to do this every time... */
2214 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2215 			optdata = (void *)&null_pktinfo;
2216 		}
2217 		optdatalen = sizeof(struct in6_pktinfo);
2218 		break;
2219 	case IPV6_TCLASS:
2220 		if (pktopt && pktopt->ip6po_tclass >= 0)
2221 			optdata = (void *)&pktopt->ip6po_tclass;
2222 		else
2223 			optdata = (void *)&deftclass;
2224 		optdatalen = sizeof(int);
2225 		break;
2226 	case IPV6_HOPOPTS:
2227 		if (pktopt && pktopt->ip6po_hbh) {
2228 			optdata = (void *)pktopt->ip6po_hbh;
2229 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2230 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2231 		}
2232 		break;
2233 	case IPV6_RTHDR:
2234 		if (pktopt && pktopt->ip6po_rthdr) {
2235 			optdata = (void *)pktopt->ip6po_rthdr;
2236 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2237 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2238 		}
2239 		break;
2240 	case IPV6_RTHDRDSTOPTS:
2241 		if (pktopt && pktopt->ip6po_dest1) {
2242 			optdata = (void *)pktopt->ip6po_dest1;
2243 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2244 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2245 		}
2246 		break;
2247 	case IPV6_DSTOPTS:
2248 		if (pktopt && pktopt->ip6po_dest2) {
2249 			optdata = (void *)pktopt->ip6po_dest2;
2250 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2251 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2252 		}
2253 		break;
2254 	case IPV6_NEXTHOP:
2255 		if (pktopt && pktopt->ip6po_nexthop) {
2256 			optdata = (void *)pktopt->ip6po_nexthop;
2257 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2258 		}
2259 		break;
2260 	case IPV6_USE_MIN_MTU:
2261 		if (pktopt)
2262 			optdata = (void *)&pktopt->ip6po_minmtu;
2263 		else
2264 			optdata = (void *)&defminmtu;
2265 		optdatalen = sizeof(int);
2266 		break;
2267 	case IPV6_DONTFRAG:
2268 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2269 			on = 1;
2270 		else
2271 			on = 0;
2272 		optdata = (void *)&on;
2273 		optdatalen = sizeof(on);
2274 		break;
2275 	case IPV6_PREFER_TEMPADDR:
2276 		if (pktopt)
2277 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2278 		else
2279 			optdata = (void *)&defpreftemp;
2280 		optdatalen = sizeof(int);
2281 		break;
2282 	default:		/* should not happen */
2283 #ifdef DIAGNOSTIC
2284 		panic("ip6_getpcbopt: unexpected option\n");
2285 #endif
2286 		return (ENOPROTOOPT);
2287 	}
2288 
2289 	error = sooptcopyout(sopt, optdata, optdatalen);
2290 
2291 	return (error);
2292 }
2293 
2294 void
2295 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2296 {
2297 	if (pktopt == NULL)
2298 		return;
2299 
2300 	if (optname == -1 || optname == IPV6_PKTINFO) {
2301 		if (pktopt->ip6po_pktinfo)
2302 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2303 		pktopt->ip6po_pktinfo = NULL;
2304 	}
2305 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2306 		pktopt->ip6po_hlim = -1;
2307 	if (optname == -1 || optname == IPV6_TCLASS)
2308 		pktopt->ip6po_tclass = -1;
2309 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2310 		if (pktopt->ip6po_nextroute.ro_rt) {
2311 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2312 			pktopt->ip6po_nextroute.ro_rt = NULL;
2313 		}
2314 		if (pktopt->ip6po_nexthop)
2315 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2316 		pktopt->ip6po_nexthop = NULL;
2317 	}
2318 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2319 		if (pktopt->ip6po_hbh)
2320 			free(pktopt->ip6po_hbh, M_IP6OPT);
2321 		pktopt->ip6po_hbh = NULL;
2322 	}
2323 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2324 		if (pktopt->ip6po_dest1)
2325 			free(pktopt->ip6po_dest1, M_IP6OPT);
2326 		pktopt->ip6po_dest1 = NULL;
2327 	}
2328 	if (optname == -1 || optname == IPV6_RTHDR) {
2329 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2330 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2331 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2332 		if (pktopt->ip6po_route.ro_rt) {
2333 			RTFREE(pktopt->ip6po_route.ro_rt);
2334 			pktopt->ip6po_route.ro_rt = NULL;
2335 		}
2336 	}
2337 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2338 		if (pktopt->ip6po_dest2)
2339 			free(pktopt->ip6po_dest2, M_IP6OPT);
2340 		pktopt->ip6po_dest2 = NULL;
2341 	}
2342 }
2343 
2344 #define PKTOPT_EXTHDRCPY(type) \
2345 do {\
2346 	if (src->type) {\
2347 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2348 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2349 		if (dst->type == NULL && canwait == M_NOWAIT)\
2350 			goto bad;\
2351 		bcopy(src->type, dst->type, hlen);\
2352 	}\
2353 } while (/*CONSTCOND*/ 0)
2354 
2355 static int
2356 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2357 {
2358 	if (dst == NULL || src == NULL)  {
2359 		printf("ip6_clearpktopts: invalid argument\n");
2360 		return (EINVAL);
2361 	}
2362 
2363 	dst->ip6po_hlim = src->ip6po_hlim;
2364 	dst->ip6po_tclass = src->ip6po_tclass;
2365 	dst->ip6po_flags = src->ip6po_flags;
2366 	if (src->ip6po_pktinfo) {
2367 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2368 		    M_IP6OPT, canwait);
2369 		if (dst->ip6po_pktinfo == NULL)
2370 			goto bad;
2371 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2372 	}
2373 	if (src->ip6po_nexthop) {
2374 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2375 		    M_IP6OPT, canwait);
2376 		if (dst->ip6po_nexthop == NULL)
2377 			goto bad;
2378 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2379 		    src->ip6po_nexthop->sa_len);
2380 	}
2381 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2382 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2383 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2384 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2385 	return (0);
2386 
2387   bad:
2388 	ip6_clearpktopts(dst, -1);
2389 	return (ENOBUFS);
2390 }
2391 #undef PKTOPT_EXTHDRCPY
2392 
2393 struct ip6_pktopts *
2394 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2395 {
2396 	int error;
2397 	struct ip6_pktopts *dst;
2398 
2399 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2400 	if (dst == NULL)
2401 		return (NULL);
2402 	ip6_initpktopts(dst);
2403 
2404 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2405 		free(dst, M_IP6OPT);
2406 		return (NULL);
2407 	}
2408 
2409 	return (dst);
2410 }
2411 
2412 void
2413 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2414 {
2415 	if (pktopt == NULL)
2416 		return;
2417 
2418 	ip6_clearpktopts(pktopt, -1);
2419 
2420 	free(pktopt, M_IP6OPT);
2421 }
2422 
2423 /*
2424  * Set the IP6 multicast options in response to user setsockopt().
2425  */
2426 static int
2427 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2428 {
2429 	int error = 0;
2430 	u_int loop, ifindex;
2431 	struct ipv6_mreq *mreq;
2432 	struct ifnet *ifp;
2433 	struct ip6_moptions *im6o = *im6op;
2434 	struct route_in6 ro;
2435 	struct in6_multi_mship *imm;
2436 
2437 	if (im6o == NULL) {
2438 		/*
2439 		 * No multicast option buffer attached to the pcb;
2440 		 * allocate one and initialize to default values.
2441 		 */
2442 		im6o = (struct ip6_moptions *)
2443 			malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2444 
2445 		if (im6o == NULL)
2446 			return (ENOBUFS);
2447 		*im6op = im6o;
2448 		im6o->im6o_multicast_ifp = NULL;
2449 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2450 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2451 		LIST_INIT(&im6o->im6o_memberships);
2452 	}
2453 
2454 	switch (optname) {
2455 
2456 	case IPV6_MULTICAST_IF:
2457 		/*
2458 		 * Select the interface for outgoing multicast packets.
2459 		 */
2460 		if (m == NULL || m->m_len != sizeof(u_int)) {
2461 			error = EINVAL;
2462 			break;
2463 		}
2464 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2465 		if (ifindex < 0 || if_index < ifindex) {
2466 			error = ENXIO;	/* XXX EINVAL? */
2467 			break;
2468 		}
2469 		ifp = ifnet_byindex(ifindex);
2470 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2471 			error = EADDRNOTAVAIL;
2472 			break;
2473 		}
2474 		im6o->im6o_multicast_ifp = ifp;
2475 		break;
2476 
2477 	case IPV6_MULTICAST_HOPS:
2478 	    {
2479 		/*
2480 		 * Set the IP6 hoplimit for outgoing multicast packets.
2481 		 */
2482 		int optval;
2483 		if (m == NULL || m->m_len != sizeof(int)) {
2484 			error = EINVAL;
2485 			break;
2486 		}
2487 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2488 		if (optval < -1 || optval >= 256)
2489 			error = EINVAL;
2490 		else if (optval == -1)
2491 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2492 		else
2493 			im6o->im6o_multicast_hlim = optval;
2494 		break;
2495 	    }
2496 
2497 	case IPV6_MULTICAST_LOOP:
2498 		/*
2499 		 * Set the loopback flag for outgoing multicast packets.
2500 		 * Must be zero or one.
2501 		 */
2502 		if (m == NULL || m->m_len != sizeof(u_int)) {
2503 			error = EINVAL;
2504 			break;
2505 		}
2506 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2507 		if (loop > 1) {
2508 			error = EINVAL;
2509 			break;
2510 		}
2511 		im6o->im6o_multicast_loop = loop;
2512 		break;
2513 
2514 	case IPV6_JOIN_GROUP:
2515 		/*
2516 		 * Add a multicast group membership.
2517 		 * Group must be a valid IP6 multicast address.
2518 		 */
2519 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2520 			error = EINVAL;
2521 			break;
2522 		}
2523 		mreq = mtod(m, struct ipv6_mreq *);
2524 
2525 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2526 			/*
2527 			 * We use the unspecified address to specify to accept
2528 			 * all multicast addresses. Only super user is allowed
2529 			 * to do this.
2530 			 */
2531 			/* XXX-BZ might need a better PRIV_NETINET_x for this */
2532 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
2533 			if (error)
2534 				break;
2535 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2536 			error = EINVAL;
2537 			break;
2538 		}
2539 
2540 		/*
2541 		 * If no interface was explicitly specified, choose an
2542 		 * appropriate one according to the given multicast address.
2543 		 */
2544 		if (mreq->ipv6mr_interface == 0) {
2545 			struct sockaddr_in6 *dst;
2546 
2547 			/*
2548 			 * Look up the routing table for the
2549 			 * address, and choose the outgoing interface.
2550 			 *   XXX: is it a good approach?
2551 			 */
2552 			ro.ro_rt = NULL;
2553 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2554 			bzero(dst, sizeof(*dst));
2555 			dst->sin6_family = AF_INET6;
2556 			dst->sin6_len = sizeof(*dst);
2557 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2558 			rtalloc((struct route *)&ro);
2559 			if (ro.ro_rt == NULL) {
2560 				error = EADDRNOTAVAIL;
2561 				break;
2562 			}
2563 			ifp = ro.ro_rt->rt_ifp;
2564 			RTFREE(ro.ro_rt);
2565 		} else {
2566 			/*
2567 			 * If the interface is specified, validate it.
2568 			 */
2569 			if (mreq->ipv6mr_interface < 0 ||
2570 			    if_index < mreq->ipv6mr_interface) {
2571 				error = ENXIO;	/* XXX EINVAL? */
2572 				break;
2573 			}
2574 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2575 			if (!ifp) {
2576 				error = ENXIO;	/* XXX EINVAL? */
2577 				break;
2578 			}
2579 		}
2580 
2581 		/*
2582 		 * See if we found an interface, and confirm that it
2583 		 * supports multicast
2584 		 */
2585 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2586 			error = EADDRNOTAVAIL;
2587 			break;
2588 		}
2589 
2590 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2591 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2592 			break;
2593 		}
2594 
2595 		/*
2596 		 * See if the membership already exists.
2597 		 */
2598 		for (imm = im6o->im6o_memberships.lh_first;
2599 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2600 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2601 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2602 					       &mreq->ipv6mr_multiaddr))
2603 				break;
2604 		if (imm != NULL) {
2605 			error = EADDRINUSE;
2606 			break;
2607 		}
2608 		/*
2609 		 * Everything looks good; add a new record to the multicast
2610 		 * address list for the given interface.
2611 		 */
2612 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2613 		if (imm == NULL)
2614 			break;
2615 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2616 		break;
2617 
2618 	case IPV6_LEAVE_GROUP:
2619 		/*
2620 		 * Drop a multicast group membership.
2621 		 * Group must be a valid IP6 multicast address.
2622 		 */
2623 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2624 			error = EINVAL;
2625 			break;
2626 		}
2627 		mreq = mtod(m, struct ipv6_mreq *);
2628 
2629 		/*
2630 		 * If an interface address was specified, get a pointer
2631 		 * to its ifnet structure.
2632 		 */
2633 		if (mreq->ipv6mr_interface < 0 ||
2634 		    if_index < mreq->ipv6mr_interface) {
2635 			error = ENXIO;	/* XXX EINVAL? */
2636 			break;
2637 		}
2638 		if (mreq->ipv6mr_interface == 0)
2639 			ifp = NULL;
2640 		else
2641 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2642 
2643 		/* Fill in the scope zone ID */
2644 		if (ifp) {
2645 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2646 				/* XXX: should not happen */
2647 				error = EADDRNOTAVAIL;
2648 				break;
2649 			}
2650 		} else if (mreq->ipv6mr_interface != 0) {
2651 			/*
2652 			 * This case happens when the (positive) index is in
2653 			 * the valid range, but the corresponding interface has
2654 			 * been detached dynamically (XXX).
2655 			 */
2656 			error = EADDRNOTAVAIL;
2657 			break;
2658 		} else {	/* ipv6mr_interface == 0 */
2659 			struct sockaddr_in6 sa6_mc;
2660 
2661 			/*
2662 			 * The API spec says as follows:
2663 			 *  If the interface index is specified as 0, the
2664 			 *  system may choose a multicast group membership to
2665 			 *  drop by matching the multicast address only.
2666 			 * On the other hand, we cannot disambiguate the scope
2667 			 * zone unless an interface is provided.  Thus, we
2668 			 * check if there's ambiguity with the default scope
2669 			 * zone as the last resort.
2670 			 */
2671 			bzero(&sa6_mc, sizeof(sa6_mc));
2672 			sa6_mc.sin6_family = AF_INET6;
2673 			sa6_mc.sin6_len = sizeof(sa6_mc);
2674 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2675 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2676 			if (error != 0)
2677 				break;
2678 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2679 		}
2680 
2681 		/*
2682 		 * Find the membership in the membership list.
2683 		 */
2684 		for (imm = im6o->im6o_memberships.lh_first;
2685 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2686 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2687 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2688 			    &mreq->ipv6mr_multiaddr))
2689 				break;
2690 		}
2691 		if (imm == NULL) {
2692 			/* Unable to resolve interface */
2693 			error = EADDRNOTAVAIL;
2694 			break;
2695 		}
2696 		/*
2697 		 * Give up the multicast address record to which the
2698 		 * membership points.
2699 		 */
2700 		LIST_REMOVE(imm, i6mm_chain);
2701 		in6_delmulti(imm->i6mm_maddr);
2702 		free(imm, M_IP6MADDR);
2703 		break;
2704 
2705 	default:
2706 		error = EOPNOTSUPP;
2707 		break;
2708 	}
2709 
2710 	/*
2711 	 * If all options have default values, no need to keep the mbuf.
2712 	 */
2713 	if (im6o->im6o_multicast_ifp == NULL &&
2714 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2715 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2716 	    im6o->im6o_memberships.lh_first == NULL) {
2717 		free(*im6op, M_IP6MOPTS);
2718 		*im6op = NULL;
2719 	}
2720 
2721 	return (error);
2722 }
2723 
2724 /*
2725  * Return the IP6 multicast options in response to user getsockopt().
2726  */
2727 static int
2728 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2729 {
2730 	u_int *hlim, *loop, *ifindex;
2731 
2732 	*mp = m_get(M_WAIT, MT_HEADER);		/* XXX */
2733 
2734 	switch (optname) {
2735 
2736 	case IPV6_MULTICAST_IF:
2737 		ifindex = mtod(*mp, u_int *);
2738 		(*mp)->m_len = sizeof(u_int);
2739 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2740 			*ifindex = 0;
2741 		else
2742 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2743 		return (0);
2744 
2745 	case IPV6_MULTICAST_HOPS:
2746 		hlim = mtod(*mp, u_int *);
2747 		(*mp)->m_len = sizeof(u_int);
2748 		if (im6o == NULL)
2749 			*hlim = ip6_defmcasthlim;
2750 		else
2751 			*hlim = im6o->im6o_multicast_hlim;
2752 		return (0);
2753 
2754 	case IPV6_MULTICAST_LOOP:
2755 		loop = mtod(*mp, u_int *);
2756 		(*mp)->m_len = sizeof(u_int);
2757 		if (im6o == NULL)
2758 			*loop = ip6_defmcasthlim;
2759 		else
2760 			*loop = im6o->im6o_multicast_loop;
2761 		return (0);
2762 
2763 	default:
2764 		return (EOPNOTSUPP);
2765 	}
2766 }
2767 
2768 /*
2769  * Discard the IP6 multicast options.
2770  */
2771 void
2772 ip6_freemoptions(struct ip6_moptions *im6o)
2773 {
2774 	struct in6_multi_mship *imm;
2775 
2776 	if (im6o == NULL)
2777 		return;
2778 
2779 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2780 		LIST_REMOVE(imm, i6mm_chain);
2781 		if (imm->i6mm_maddr)
2782 			in6_delmulti(imm->i6mm_maddr);
2783 		free(imm, M_IP6MADDR);
2784 	}
2785 	free(im6o, M_IP6MOPTS);
2786 }
2787 
2788 /*
2789  * Set IPv6 outgoing packet options based on advanced API.
2790  */
2791 int
2792 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2793     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2794 {
2795 	struct cmsghdr *cm = 0;
2796 
2797 	if (control == NULL || opt == NULL)
2798 		return (EINVAL);
2799 
2800 	ip6_initpktopts(opt);
2801 	if (stickyopt) {
2802 		int error;
2803 
2804 		/*
2805 		 * If stickyopt is provided, make a local copy of the options
2806 		 * for this particular packet, then override them by ancillary
2807 		 * objects.
2808 		 * XXX: copypktopts() does not copy the cached route to a next
2809 		 * hop (if any).  This is not very good in terms of efficiency,
2810 		 * but we can allow this since this option should be rarely
2811 		 * used.
2812 		 */
2813 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2814 			return (error);
2815 	}
2816 
2817 	/*
2818 	 * XXX: Currently, we assume all the optional information is stored
2819 	 * in a single mbuf.
2820 	 */
2821 	if (control->m_next)
2822 		return (EINVAL);
2823 
2824 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2825 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2826 		int error;
2827 
2828 		if (control->m_len < CMSG_LEN(0))
2829 			return (EINVAL);
2830 
2831 		cm = mtod(control, struct cmsghdr *);
2832 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2833 			return (EINVAL);
2834 		if (cm->cmsg_level != IPPROTO_IPV6)
2835 			continue;
2836 
2837 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2838 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2839 		if (error)
2840 			return (error);
2841 	}
2842 
2843 	return (0);
2844 }
2845 
2846 /*
2847  * Set a particular packet option, as a sticky option or an ancillary data
2848  * item.  "len" can be 0 only when it's a sticky option.
2849  * We have 4 cases of combination of "sticky" and "cmsg":
2850  * "sticky=0, cmsg=0": impossible
2851  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2852  * "sticky=1, cmsg=0": RFC3542 socket option
2853  * "sticky=1, cmsg=1": RFC2292 socket option
2854  */
2855 static int
2856 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2857     struct ucred *cred, int sticky, int cmsg, int uproto)
2858 {
2859 	int minmtupolicy, preftemp;
2860 	int error;
2861 
2862 	if (!sticky && !cmsg) {
2863 #ifdef DIAGNOSTIC
2864 		printf("ip6_setpktopt: impossible case\n");
2865 #endif
2866 		return (EINVAL);
2867 	}
2868 
2869 	/*
2870 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2871 	 * not be specified in the context of RFC3542.  Conversely,
2872 	 * RFC3542 types should not be specified in the context of RFC2292.
2873 	 */
2874 	if (!cmsg) {
2875 		switch (optname) {
2876 		case IPV6_2292PKTINFO:
2877 		case IPV6_2292HOPLIMIT:
2878 		case IPV6_2292NEXTHOP:
2879 		case IPV6_2292HOPOPTS:
2880 		case IPV6_2292DSTOPTS:
2881 		case IPV6_2292RTHDR:
2882 		case IPV6_2292PKTOPTIONS:
2883 			return (ENOPROTOOPT);
2884 		}
2885 	}
2886 	if (sticky && cmsg) {
2887 		switch (optname) {
2888 		case IPV6_PKTINFO:
2889 		case IPV6_HOPLIMIT:
2890 		case IPV6_NEXTHOP:
2891 		case IPV6_HOPOPTS:
2892 		case IPV6_DSTOPTS:
2893 		case IPV6_RTHDRDSTOPTS:
2894 		case IPV6_RTHDR:
2895 		case IPV6_USE_MIN_MTU:
2896 		case IPV6_DONTFRAG:
2897 		case IPV6_TCLASS:
2898 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2899 			return (ENOPROTOOPT);
2900 		}
2901 	}
2902 
2903 	switch (optname) {
2904 	case IPV6_2292PKTINFO:
2905 	case IPV6_PKTINFO:
2906 	{
2907 		struct ifnet *ifp = NULL;
2908 		struct in6_pktinfo *pktinfo;
2909 
2910 		if (len != sizeof(struct in6_pktinfo))
2911 			return (EINVAL);
2912 
2913 		pktinfo = (struct in6_pktinfo *)buf;
2914 
2915 		/*
2916 		 * An application can clear any sticky IPV6_PKTINFO option by
2917 		 * doing a "regular" setsockopt with ipi6_addr being
2918 		 * in6addr_any and ipi6_ifindex being zero.
2919 		 * [RFC 3542, Section 6]
2920 		 */
2921 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2922 		    pktinfo->ipi6_ifindex == 0 &&
2923 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2924 			ip6_clearpktopts(opt, optname);
2925 			break;
2926 		}
2927 
2928 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2929 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2930 			return (EINVAL);
2931 		}
2932 
2933 		/* validate the interface index if specified. */
2934 		if (pktinfo->ipi6_ifindex > if_index ||
2935 		    pktinfo->ipi6_ifindex < 0) {
2936 			 return (ENXIO);
2937 		}
2938 		if (pktinfo->ipi6_ifindex) {
2939 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2940 			if (ifp == NULL)
2941 				return (ENXIO);
2942 		}
2943 
2944 		/*
2945 		 * We store the address anyway, and let in6_selectsrc()
2946 		 * validate the specified address.  This is because ipi6_addr
2947 		 * may not have enough information about its scope zone, and
2948 		 * we may need additional information (such as outgoing
2949 		 * interface or the scope zone of a destination address) to
2950 		 * disambiguate the scope.
2951 		 * XXX: the delay of the validation may confuse the
2952 		 * application when it is used as a sticky option.
2953 		 */
2954 		if (opt->ip6po_pktinfo == NULL) {
2955 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2956 			    M_IP6OPT, M_NOWAIT);
2957 			if (opt->ip6po_pktinfo == NULL)
2958 				return (ENOBUFS);
2959 		}
2960 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2961 		break;
2962 	}
2963 
2964 	case IPV6_2292HOPLIMIT:
2965 	case IPV6_HOPLIMIT:
2966 	{
2967 		int *hlimp;
2968 
2969 		/*
2970 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2971 		 * to simplify the ordering among hoplimit options.
2972 		 */
2973 		if (optname == IPV6_HOPLIMIT && sticky)
2974 			return (ENOPROTOOPT);
2975 
2976 		if (len != sizeof(int))
2977 			return (EINVAL);
2978 		hlimp = (int *)buf;
2979 		if (*hlimp < -1 || *hlimp > 255)
2980 			return (EINVAL);
2981 
2982 		opt->ip6po_hlim = *hlimp;
2983 		break;
2984 	}
2985 
2986 	case IPV6_TCLASS:
2987 	{
2988 		int tclass;
2989 
2990 		if (len != sizeof(int))
2991 			return (EINVAL);
2992 		tclass = *(int *)buf;
2993 		if (tclass < -1 || tclass > 255)
2994 			return (EINVAL);
2995 
2996 		opt->ip6po_tclass = tclass;
2997 		break;
2998 	}
2999 
3000 	case IPV6_2292NEXTHOP:
3001 	case IPV6_NEXTHOP:
3002 		if (cred != NULL) {
3003 			error = priv_check_cred(cred,
3004 			    PRIV_NETINET_SETHDROPTS, 0);
3005 			if (error)
3006 				return (error);
3007 		}
3008 
3009 		if (len == 0) {	/* just remove the option */
3010 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3011 			break;
3012 		}
3013 
3014 		/* check if cmsg_len is large enough for sa_len */
3015 		if (len < sizeof(struct sockaddr) || len < *buf)
3016 			return (EINVAL);
3017 
3018 		switch (((struct sockaddr *)buf)->sa_family) {
3019 		case AF_INET6:
3020 		{
3021 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3022 			int error;
3023 
3024 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3025 				return (EINVAL);
3026 
3027 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3028 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3029 				return (EINVAL);
3030 			}
3031 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3032 			    != 0) {
3033 				return (error);
3034 			}
3035 			break;
3036 		}
3037 		case AF_LINK:	/* should eventually be supported */
3038 		default:
3039 			return (EAFNOSUPPORT);
3040 		}
3041 
3042 		/* turn off the previous option, then set the new option. */
3043 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3044 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3045 		if (opt->ip6po_nexthop == NULL)
3046 			return (ENOBUFS);
3047 		bcopy(buf, opt->ip6po_nexthop, *buf);
3048 		break;
3049 
3050 	case IPV6_2292HOPOPTS:
3051 	case IPV6_HOPOPTS:
3052 	{
3053 		struct ip6_hbh *hbh;
3054 		int hbhlen;
3055 
3056 		/*
3057 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3058 		 * options, since per-option restriction has too much
3059 		 * overhead.
3060 		 */
3061 		if (cred != NULL) {
3062 			error = priv_check_cred(cred,
3063 			    PRIV_NETINET_SETHDROPTS, 0);
3064 			if (error)
3065 				return (error);
3066 		}
3067 
3068 		if (len == 0) {
3069 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3070 			break;	/* just remove the option */
3071 		}
3072 
3073 		/* message length validation */
3074 		if (len < sizeof(struct ip6_hbh))
3075 			return (EINVAL);
3076 		hbh = (struct ip6_hbh *)buf;
3077 		hbhlen = (hbh->ip6h_len + 1) << 3;
3078 		if (len != hbhlen)
3079 			return (EINVAL);
3080 
3081 		/* turn off the previous option, then set the new option. */
3082 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3083 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3084 		if (opt->ip6po_hbh == NULL)
3085 			return (ENOBUFS);
3086 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3087 
3088 		break;
3089 	}
3090 
3091 	case IPV6_2292DSTOPTS:
3092 	case IPV6_DSTOPTS:
3093 	case IPV6_RTHDRDSTOPTS:
3094 	{
3095 		struct ip6_dest *dest, **newdest = NULL;
3096 		int destlen;
3097 
3098 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3099 			error = priv_check_cred(cred,
3100 			    PRIV_NETINET_SETHDROPTS, 0);
3101 			if (error)
3102 				return (error);
3103 		}
3104 
3105 		if (len == 0) {
3106 			ip6_clearpktopts(opt, optname);
3107 			break;	/* just remove the option */
3108 		}
3109 
3110 		/* message length validation */
3111 		if (len < sizeof(struct ip6_dest))
3112 			return (EINVAL);
3113 		dest = (struct ip6_dest *)buf;
3114 		destlen = (dest->ip6d_len + 1) << 3;
3115 		if (len != destlen)
3116 			return (EINVAL);
3117 
3118 		/*
3119 		 * Determine the position that the destination options header
3120 		 * should be inserted; before or after the routing header.
3121 		 */
3122 		switch (optname) {
3123 		case IPV6_2292DSTOPTS:
3124 			/*
3125 			 * The old advacned API is ambiguous on this point.
3126 			 * Our approach is to determine the position based
3127 			 * according to the existence of a routing header.
3128 			 * Note, however, that this depends on the order of the
3129 			 * extension headers in the ancillary data; the 1st
3130 			 * part of the destination options header must appear
3131 			 * before the routing header in the ancillary data,
3132 			 * too.
3133 			 * RFC3542 solved the ambiguity by introducing
3134 			 * separate ancillary data or option types.
3135 			 */
3136 			if (opt->ip6po_rthdr == NULL)
3137 				newdest = &opt->ip6po_dest1;
3138 			else
3139 				newdest = &opt->ip6po_dest2;
3140 			break;
3141 		case IPV6_RTHDRDSTOPTS:
3142 			newdest = &opt->ip6po_dest1;
3143 			break;
3144 		case IPV6_DSTOPTS:
3145 			newdest = &opt->ip6po_dest2;
3146 			break;
3147 		}
3148 
3149 		/* turn off the previous option, then set the new option. */
3150 		ip6_clearpktopts(opt, optname);
3151 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3152 		if (*newdest == NULL)
3153 			return (ENOBUFS);
3154 		bcopy(dest, *newdest, destlen);
3155 
3156 		break;
3157 	}
3158 
3159 	case IPV6_2292RTHDR:
3160 	case IPV6_RTHDR:
3161 	{
3162 		struct ip6_rthdr *rth;
3163 		int rthlen;
3164 
3165 		if (len == 0) {
3166 			ip6_clearpktopts(opt, IPV6_RTHDR);
3167 			break;	/* just remove the option */
3168 		}
3169 
3170 		/* message length validation */
3171 		if (len < sizeof(struct ip6_rthdr))
3172 			return (EINVAL);
3173 		rth = (struct ip6_rthdr *)buf;
3174 		rthlen = (rth->ip6r_len + 1) << 3;
3175 		if (len != rthlen)
3176 			return (EINVAL);
3177 
3178 		switch (rth->ip6r_type) {
3179 		case IPV6_RTHDR_TYPE_0:
3180 			if (rth->ip6r_len == 0)	/* must contain one addr */
3181 				return (EINVAL);
3182 			if (rth->ip6r_len % 2) /* length must be even */
3183 				return (EINVAL);
3184 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3185 				return (EINVAL);
3186 			break;
3187 		default:
3188 			return (EINVAL);	/* not supported */
3189 		}
3190 
3191 		/* turn off the previous option */
3192 		ip6_clearpktopts(opt, IPV6_RTHDR);
3193 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3194 		if (opt->ip6po_rthdr == NULL)
3195 			return (ENOBUFS);
3196 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3197 
3198 		break;
3199 	}
3200 
3201 	case IPV6_USE_MIN_MTU:
3202 		if (len != sizeof(int))
3203 			return (EINVAL);
3204 		minmtupolicy = *(int *)buf;
3205 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3206 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3207 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3208 			return (EINVAL);
3209 		}
3210 		opt->ip6po_minmtu = minmtupolicy;
3211 		break;
3212 
3213 	case IPV6_DONTFRAG:
3214 		if (len != sizeof(int))
3215 			return (EINVAL);
3216 
3217 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3218 			/*
3219 			 * we ignore this option for TCP sockets.
3220 			 * (RFC3542 leaves this case unspecified.)
3221 			 */
3222 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3223 		} else
3224 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3225 		break;
3226 
3227 	case IPV6_PREFER_TEMPADDR:
3228 		if (len != sizeof(int))
3229 			return (EINVAL);
3230 		preftemp = *(int *)buf;
3231 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3232 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3233 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3234 			return (EINVAL);
3235 		}
3236 		opt->ip6po_prefer_tempaddr = preftemp;
3237 		break;
3238 
3239 	default:
3240 		return (ENOPROTOOPT);
3241 	} /* end of switch */
3242 
3243 	return (0);
3244 }
3245 
3246 /*
3247  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3248  * packet to the input queue of a specified interface.  Note that this
3249  * calls the output routine of the loopback "driver", but with an interface
3250  * pointer that might NOT be &loif -- easier than replicating that code here.
3251  */
3252 void
3253 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3254 {
3255 	struct mbuf *copym;
3256 	struct ip6_hdr *ip6;
3257 
3258 	copym = m_copy(m, 0, M_COPYALL);
3259 	if (copym == NULL)
3260 		return;
3261 
3262 	/*
3263 	 * Make sure to deep-copy IPv6 header portion in case the data
3264 	 * is in an mbuf cluster, so that we can safely override the IPv6
3265 	 * header portion later.
3266 	 */
3267 	if ((copym->m_flags & M_EXT) != 0 ||
3268 	    copym->m_len < sizeof(struct ip6_hdr)) {
3269 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3270 		if (copym == NULL)
3271 			return;
3272 	}
3273 
3274 #ifdef DIAGNOSTIC
3275 	if (copym->m_len < sizeof(*ip6)) {
3276 		m_freem(copym);
3277 		return;
3278 	}
3279 #endif
3280 
3281 	ip6 = mtod(copym, struct ip6_hdr *);
3282 	/*
3283 	 * clear embedded scope identifiers if necessary.
3284 	 * in6_clearscope will touch the addresses only when necessary.
3285 	 */
3286 	in6_clearscope(&ip6->ip6_src);
3287 	in6_clearscope(&ip6->ip6_dst);
3288 
3289 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3290 }
3291 
3292 /*
3293  * Chop IPv6 header off from the payload.
3294  */
3295 static int
3296 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3297 {
3298 	struct mbuf *mh;
3299 	struct ip6_hdr *ip6;
3300 
3301 	ip6 = mtod(m, struct ip6_hdr *);
3302 	if (m->m_len > sizeof(*ip6)) {
3303 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3304 		if (mh == 0) {
3305 			m_freem(m);
3306 			return ENOBUFS;
3307 		}
3308 		M_MOVE_PKTHDR(mh, m);
3309 		MH_ALIGN(mh, sizeof(*ip6));
3310 		m->m_len -= sizeof(*ip6);
3311 		m->m_data += sizeof(*ip6);
3312 		mh->m_next = m;
3313 		m = mh;
3314 		m->m_len = sizeof(*ip6);
3315 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3316 	}
3317 	exthdrs->ip6e_ip6 = m;
3318 	return 0;
3319 }
3320 
3321 /*
3322  * Compute IPv6 extension header length.
3323  */
3324 int
3325 ip6_optlen(struct in6pcb *in6p)
3326 {
3327 	int len;
3328 
3329 	if (!in6p->in6p_outputopts)
3330 		return 0;
3331 
3332 	len = 0;
3333 #define elen(x) \
3334     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3335 
3336 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3337 	if (in6p->in6p_outputopts->ip6po_rthdr)
3338 		/* dest1 is valid with rthdr only */
3339 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3340 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3341 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3342 	return len;
3343 #undef elen
3344 }
3345