xref: /freebsd/sys/netinet6/ip6_output.c (revision cacdd70cc751fb68dec4b86c5e5b8c969b6e26ef)
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 	level = sopt->sopt_level;
1379 	op = sopt->sopt_dir;
1380 	optname = sopt->sopt_name;
1381 	optlen = sopt->sopt_valsize;
1382 	td = sopt->sopt_td;
1383 	error = 0;
1384 	optval = 0;
1385 	uproto = (int)so->so_proto->pr_protocol;
1386 
1387 	if (level == IPPROTO_IPV6) {
1388 		switch (op) {
1389 
1390 		case SOPT_SET:
1391 			switch (optname) {
1392 			case IPV6_2292PKTOPTIONS:
1393 #ifdef IPV6_PKTOPTIONS
1394 			case IPV6_PKTOPTIONS:
1395 #endif
1396 			{
1397 				struct mbuf *m;
1398 
1399 				error = soopt_getm(sopt, &m); /* XXX */
1400 				if (error != 0)
1401 					break;
1402 				error = soopt_mcopyin(sopt, m); /* XXX */
1403 				if (error != 0)
1404 					break;
1405 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1406 						    m, so, sopt);
1407 				m_freem(m); /* XXX */
1408 				break;
1409 			}
1410 
1411 			/*
1412 			 * Use of some Hop-by-Hop options or some
1413 			 * Destination options, might require special
1414 			 * privilege.  That is, normal applications
1415 			 * (without special privilege) might be forbidden
1416 			 * from setting certain options in outgoing packets,
1417 			 * and might never see certain options in received
1418 			 * packets. [RFC 2292 Section 6]
1419 			 * KAME specific note:
1420 			 *  KAME prevents non-privileged users from sending or
1421 			 *  receiving ANY hbh/dst options in order to avoid
1422 			 *  overhead of parsing options in the kernel.
1423 			 */
1424 			case IPV6_RECVHOPOPTS:
1425 			case IPV6_RECVDSTOPTS:
1426 			case IPV6_RECVRTHDRDSTOPTS:
1427 				if (td != NULL) {
1428 					error = priv_check(td,
1429 					    PRIV_NETINET_SETHDROPTS);
1430 					if (error)
1431 						break;
1432 				}
1433 				/* FALLTHROUGH */
1434 			case IPV6_UNICAST_HOPS:
1435 			case IPV6_HOPLIMIT:
1436 			case IPV6_FAITH:
1437 
1438 			case IPV6_RECVPKTINFO:
1439 			case IPV6_RECVHOPLIMIT:
1440 			case IPV6_RECVRTHDR:
1441 			case IPV6_RECVPATHMTU:
1442 			case IPV6_RECVTCLASS:
1443 			case IPV6_V6ONLY:
1444 			case IPV6_AUTOFLOWLABEL:
1445 				if (optlen != sizeof(int)) {
1446 					error = EINVAL;
1447 					break;
1448 				}
1449 				error = sooptcopyin(sopt, &optval,
1450 					sizeof optval, sizeof optval);
1451 				if (error)
1452 					break;
1453 				switch (optname) {
1454 
1455 				case IPV6_UNICAST_HOPS:
1456 					if (optval < -1 || optval >= 256)
1457 						error = EINVAL;
1458 					else {
1459 						/* -1 = kernel default */
1460 						in6p->in6p_hops = optval;
1461 						if ((in6p->in6p_vflag &
1462 						     INP_IPV4) != 0)
1463 							in6p->inp_ip_ttl = optval;
1464 					}
1465 					break;
1466 #define OPTSET(bit) \
1467 do { \
1468 	if (optval) \
1469 		in6p->in6p_flags |= (bit); \
1470 	else \
1471 		in6p->in6p_flags &= ~(bit); \
1472 } while (/*CONSTCOND*/ 0)
1473 #define OPTSET2292(bit) \
1474 do { \
1475 	in6p->in6p_flags |= IN6P_RFC2292; \
1476 	if (optval) \
1477 		in6p->in6p_flags |= (bit); \
1478 	else \
1479 		in6p->in6p_flags &= ~(bit); \
1480 } while (/*CONSTCOND*/ 0)
1481 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1482 
1483 				case IPV6_RECVPKTINFO:
1484 					/* cannot mix with RFC2292 */
1485 					if (OPTBIT(IN6P_RFC2292)) {
1486 						error = EINVAL;
1487 						break;
1488 					}
1489 					OPTSET(IN6P_PKTINFO);
1490 					break;
1491 
1492 				case IPV6_HOPLIMIT:
1493 				{
1494 					struct ip6_pktopts **optp;
1495 
1496 					/* cannot mix with RFC2292 */
1497 					if (OPTBIT(IN6P_RFC2292)) {
1498 						error = EINVAL;
1499 						break;
1500 					}
1501 					optp = &in6p->in6p_outputopts;
1502 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1503 					    (u_char *)&optval, sizeof(optval),
1504 					    optp, (td != NULL) ? td->td_ucred :
1505 					    NULL, uproto);
1506 					break;
1507 				}
1508 
1509 				case IPV6_RECVHOPLIMIT:
1510 					/* cannot mix with RFC2292 */
1511 					if (OPTBIT(IN6P_RFC2292)) {
1512 						error = EINVAL;
1513 						break;
1514 					}
1515 					OPTSET(IN6P_HOPLIMIT);
1516 					break;
1517 
1518 				case IPV6_RECVHOPOPTS:
1519 					/* cannot mix with RFC2292 */
1520 					if (OPTBIT(IN6P_RFC2292)) {
1521 						error = EINVAL;
1522 						break;
1523 					}
1524 					OPTSET(IN6P_HOPOPTS);
1525 					break;
1526 
1527 				case IPV6_RECVDSTOPTS:
1528 					/* cannot mix with RFC2292 */
1529 					if (OPTBIT(IN6P_RFC2292)) {
1530 						error = EINVAL;
1531 						break;
1532 					}
1533 					OPTSET(IN6P_DSTOPTS);
1534 					break;
1535 
1536 				case IPV6_RECVRTHDRDSTOPTS:
1537 					/* cannot mix with RFC2292 */
1538 					if (OPTBIT(IN6P_RFC2292)) {
1539 						error = EINVAL;
1540 						break;
1541 					}
1542 					OPTSET(IN6P_RTHDRDSTOPTS);
1543 					break;
1544 
1545 				case IPV6_RECVRTHDR:
1546 					/* cannot mix with RFC2292 */
1547 					if (OPTBIT(IN6P_RFC2292)) {
1548 						error = EINVAL;
1549 						break;
1550 					}
1551 					OPTSET(IN6P_RTHDR);
1552 					break;
1553 
1554 				case IPV6_FAITH:
1555 					OPTSET(IN6P_FAITH);
1556 					break;
1557 
1558 				case IPV6_RECVPATHMTU:
1559 					/*
1560 					 * We ignore this option for TCP
1561 					 * sockets.
1562 					 * (RFC3542 leaves this case
1563 					 * unspecified.)
1564 					 */
1565 					if (uproto != IPPROTO_TCP)
1566 						OPTSET(IN6P_MTU);
1567 					break;
1568 
1569 				case IPV6_V6ONLY:
1570 					/*
1571 					 * make setsockopt(IPV6_V6ONLY)
1572 					 * available only prior to bind(2).
1573 					 * see ipng mailing list, Jun 22 2001.
1574 					 */
1575 					if (in6p->in6p_lport ||
1576 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1577 						error = EINVAL;
1578 						break;
1579 					}
1580 					OPTSET(IN6P_IPV6_V6ONLY);
1581 					if (optval)
1582 						in6p->in6p_vflag &= ~INP_IPV4;
1583 					else
1584 						in6p->in6p_vflag |= INP_IPV4;
1585 					break;
1586 				case IPV6_RECVTCLASS:
1587 					/* cannot mix with RFC2292 XXX */
1588 					if (OPTBIT(IN6P_RFC2292)) {
1589 						error = EINVAL;
1590 						break;
1591 					}
1592 					OPTSET(IN6P_TCLASS);
1593 					break;
1594 				case IPV6_AUTOFLOWLABEL:
1595 					OPTSET(IN6P_AUTOFLOWLABEL);
1596 					break;
1597 
1598 				}
1599 				break;
1600 
1601 			case IPV6_TCLASS:
1602 			case IPV6_DONTFRAG:
1603 			case IPV6_USE_MIN_MTU:
1604 			case IPV6_PREFER_TEMPADDR:
1605 				if (optlen != sizeof(optval)) {
1606 					error = EINVAL;
1607 					break;
1608 				}
1609 				error = sooptcopyin(sopt, &optval,
1610 					sizeof optval, sizeof optval);
1611 				if (error)
1612 					break;
1613 				{
1614 					struct ip6_pktopts **optp;
1615 					optp = &in6p->in6p_outputopts;
1616 					error = ip6_pcbopt(optname,
1617 					    (u_char *)&optval, sizeof(optval),
1618 					    optp, (td != NULL) ? td->td_ucred :
1619 					    NULL, uproto);
1620 					break;
1621 				}
1622 
1623 			case IPV6_2292PKTINFO:
1624 			case IPV6_2292HOPLIMIT:
1625 			case IPV6_2292HOPOPTS:
1626 			case IPV6_2292DSTOPTS:
1627 			case IPV6_2292RTHDR:
1628 				/* RFC 2292 */
1629 				if (optlen != sizeof(int)) {
1630 					error = EINVAL;
1631 					break;
1632 				}
1633 				error = sooptcopyin(sopt, &optval,
1634 					sizeof optval, sizeof optval);
1635 				if (error)
1636 					break;
1637 				switch (optname) {
1638 				case IPV6_2292PKTINFO:
1639 					OPTSET2292(IN6P_PKTINFO);
1640 					break;
1641 				case IPV6_2292HOPLIMIT:
1642 					OPTSET2292(IN6P_HOPLIMIT);
1643 					break;
1644 				case IPV6_2292HOPOPTS:
1645 					/*
1646 					 * Check super-user privilege.
1647 					 * See comments for IPV6_RECVHOPOPTS.
1648 					 */
1649 					if (td != NULL) {
1650 						error = priv_check(td,
1651 						    PRIV_NETINET_SETHDROPTS);
1652 						if (error)
1653 							return (error);
1654 					}
1655 					OPTSET2292(IN6P_HOPOPTS);
1656 					break;
1657 				case IPV6_2292DSTOPTS:
1658 					if (td != NULL) {
1659 						error = priv_check(td,
1660 						    PRIV_NETINET_SETHDROPTS);
1661 						if (error)
1662 							return (error);
1663 					}
1664 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1665 					break;
1666 				case IPV6_2292RTHDR:
1667 					OPTSET2292(IN6P_RTHDR);
1668 					break;
1669 				}
1670 				break;
1671 			case IPV6_PKTINFO:
1672 			case IPV6_HOPOPTS:
1673 			case IPV6_RTHDR:
1674 			case IPV6_DSTOPTS:
1675 			case IPV6_RTHDRDSTOPTS:
1676 			case IPV6_NEXTHOP:
1677 			{
1678 				/* new advanced API (RFC3542) */
1679 				u_char *optbuf;
1680 				u_char optbuf_storage[MCLBYTES];
1681 				int optlen;
1682 				struct ip6_pktopts **optp;
1683 
1684 				/* cannot mix with RFC2292 */
1685 				if (OPTBIT(IN6P_RFC2292)) {
1686 					error = EINVAL;
1687 					break;
1688 				}
1689 
1690 				/*
1691 				 * We only ensure valsize is not too large
1692 				 * here.  Further validation will be done
1693 				 * later.
1694 				 */
1695 				error = sooptcopyin(sopt, optbuf_storage,
1696 				    sizeof(optbuf_storage), 0);
1697 				if (error)
1698 					break;
1699 				optlen = sopt->sopt_valsize;
1700 				optbuf = optbuf_storage;
1701 				optp = &in6p->in6p_outputopts;
1702 				error = ip6_pcbopt(optname, optbuf, optlen,
1703 				    optp, (td != NULL) ? td->td_ucred : NULL,
1704 				    uproto);
1705 				break;
1706 			}
1707 #undef OPTSET
1708 
1709 			case IPV6_MULTICAST_IF:
1710 			case IPV6_MULTICAST_HOPS:
1711 			case IPV6_MULTICAST_LOOP:
1712 			case IPV6_JOIN_GROUP:
1713 			case IPV6_LEAVE_GROUP:
1714 			    {
1715 				if (sopt->sopt_valsize > MLEN) {
1716 					error = EMSGSIZE;
1717 					break;
1718 				}
1719 				/* XXX */
1720 			    }
1721 			    /* FALLTHROUGH */
1722 			    {
1723 				struct mbuf *m;
1724 
1725 				if (sopt->sopt_valsize > MCLBYTES) {
1726 					error = EMSGSIZE;
1727 					break;
1728 				}
1729 				/* XXX */
1730 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1731 				if (m == 0) {
1732 					error = ENOBUFS;
1733 					break;
1734 				}
1735 				if (sopt->sopt_valsize > MLEN) {
1736 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1737 					if ((m->m_flags & M_EXT) == 0) {
1738 						m_free(m);
1739 						error = ENOBUFS;
1740 						break;
1741 					}
1742 				}
1743 				m->m_len = sopt->sopt_valsize;
1744 				error = sooptcopyin(sopt, mtod(m, char *),
1745 						    m->m_len, m->m_len);
1746 				if (error) {
1747 					(void)m_free(m);
1748 					break;
1749 				}
1750 				error =	ip6_setmoptions(sopt->sopt_name,
1751 							&in6p->in6p_moptions,
1752 							m);
1753 				(void)m_free(m);
1754 			    }
1755 				break;
1756 
1757 			case IPV6_PORTRANGE:
1758 				error = sooptcopyin(sopt, &optval,
1759 				    sizeof optval, sizeof optval);
1760 				if (error)
1761 					break;
1762 
1763 				switch (optval) {
1764 				case IPV6_PORTRANGE_DEFAULT:
1765 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1766 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1767 					break;
1768 
1769 				case IPV6_PORTRANGE_HIGH:
1770 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1771 					in6p->in6p_flags |= IN6P_HIGHPORT;
1772 					break;
1773 
1774 				case IPV6_PORTRANGE_LOW:
1775 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1776 					in6p->in6p_flags |= IN6P_LOWPORT;
1777 					break;
1778 
1779 				default:
1780 					error = EINVAL;
1781 					break;
1782 				}
1783 				break;
1784 
1785 #ifdef IPSEC
1786 			case IPV6_IPSEC_POLICY:
1787 			{
1788 				caddr_t req;
1789 				struct mbuf *m;
1790 
1791 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1792 					break;
1793 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1794 					break;
1795 				req = mtod(m, caddr_t);
1796 				error = ipsec6_set_policy(in6p, optname, req,
1797 				    m->m_len, (sopt->sopt_td != NULL) ?
1798 				    sopt->sopt_td->td_ucred : NULL);
1799 				m_freem(m);
1800 				break;
1801 			}
1802 #endif /* IPSEC */
1803 
1804 			default:
1805 				error = ENOPROTOOPT;
1806 				break;
1807 			}
1808 			break;
1809 
1810 		case SOPT_GET:
1811 			switch (optname) {
1812 
1813 			case IPV6_2292PKTOPTIONS:
1814 #ifdef IPV6_PKTOPTIONS
1815 			case IPV6_PKTOPTIONS:
1816 #endif
1817 				/*
1818 				 * RFC3542 (effectively) deprecated the
1819 				 * semantics of the 2292-style pktoptions.
1820 				 * Since it was not reliable in nature (i.e.,
1821 				 * applications had to expect the lack of some
1822 				 * information after all), it would make sense
1823 				 * to simplify this part by always returning
1824 				 * empty data.
1825 				 */
1826 				sopt->sopt_valsize = 0;
1827 				break;
1828 
1829 			case IPV6_RECVHOPOPTS:
1830 			case IPV6_RECVDSTOPTS:
1831 			case IPV6_RECVRTHDRDSTOPTS:
1832 			case IPV6_UNICAST_HOPS:
1833 			case IPV6_RECVPKTINFO:
1834 			case IPV6_RECVHOPLIMIT:
1835 			case IPV6_RECVRTHDR:
1836 			case IPV6_RECVPATHMTU:
1837 
1838 			case IPV6_FAITH:
1839 			case IPV6_V6ONLY:
1840 			case IPV6_PORTRANGE:
1841 			case IPV6_RECVTCLASS:
1842 			case IPV6_AUTOFLOWLABEL:
1843 				switch (optname) {
1844 
1845 				case IPV6_RECVHOPOPTS:
1846 					optval = OPTBIT(IN6P_HOPOPTS);
1847 					break;
1848 
1849 				case IPV6_RECVDSTOPTS:
1850 					optval = OPTBIT(IN6P_DSTOPTS);
1851 					break;
1852 
1853 				case IPV6_RECVRTHDRDSTOPTS:
1854 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1855 					break;
1856 
1857 				case IPV6_UNICAST_HOPS:
1858 					optval = in6p->in6p_hops;
1859 					break;
1860 
1861 				case IPV6_RECVPKTINFO:
1862 					optval = OPTBIT(IN6P_PKTINFO);
1863 					break;
1864 
1865 				case IPV6_RECVHOPLIMIT:
1866 					optval = OPTBIT(IN6P_HOPLIMIT);
1867 					break;
1868 
1869 				case IPV6_RECVRTHDR:
1870 					optval = OPTBIT(IN6P_RTHDR);
1871 					break;
1872 
1873 				case IPV6_RECVPATHMTU:
1874 					optval = OPTBIT(IN6P_MTU);
1875 					break;
1876 
1877 				case IPV6_FAITH:
1878 					optval = OPTBIT(IN6P_FAITH);
1879 					break;
1880 
1881 				case IPV6_V6ONLY:
1882 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1883 					break;
1884 
1885 				case IPV6_PORTRANGE:
1886 				    {
1887 					int flags;
1888 					flags = in6p->in6p_flags;
1889 					if (flags & IN6P_HIGHPORT)
1890 						optval = IPV6_PORTRANGE_HIGH;
1891 					else if (flags & IN6P_LOWPORT)
1892 						optval = IPV6_PORTRANGE_LOW;
1893 					else
1894 						optval = 0;
1895 					break;
1896 				    }
1897 				case IPV6_RECVTCLASS:
1898 					optval = OPTBIT(IN6P_TCLASS);
1899 					break;
1900 
1901 				case IPV6_AUTOFLOWLABEL:
1902 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1903 					break;
1904 				}
1905 				if (error)
1906 					break;
1907 				error = sooptcopyout(sopt, &optval,
1908 					sizeof optval);
1909 				break;
1910 
1911 			case IPV6_PATHMTU:
1912 			{
1913 				u_long pmtu = 0;
1914 				struct ip6_mtuinfo mtuinfo;
1915 				struct route_in6 sro;
1916 
1917 				bzero(&sro, sizeof(sro));
1918 
1919 				if (!(so->so_state & SS_ISCONNECTED))
1920 					return (ENOTCONN);
1921 				/*
1922 				 * XXX: we dot not consider the case of source
1923 				 * routing, or optional information to specify
1924 				 * the outgoing interface.
1925 				 */
1926 				error = ip6_getpmtu(&sro, NULL, NULL,
1927 				    &in6p->in6p_faddr, &pmtu, NULL);
1928 				if (sro.ro_rt)
1929 					RTFREE(sro.ro_rt);
1930 				if (error)
1931 					break;
1932 				if (pmtu > IPV6_MAXPACKET)
1933 					pmtu = IPV6_MAXPACKET;
1934 
1935 				bzero(&mtuinfo, sizeof(mtuinfo));
1936 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1937 				optdata = (void *)&mtuinfo;
1938 				optdatalen = sizeof(mtuinfo);
1939 				error = sooptcopyout(sopt, optdata,
1940 				    optdatalen);
1941 				break;
1942 			}
1943 
1944 			case IPV6_2292PKTINFO:
1945 			case IPV6_2292HOPLIMIT:
1946 			case IPV6_2292HOPOPTS:
1947 			case IPV6_2292RTHDR:
1948 			case IPV6_2292DSTOPTS:
1949 				switch (optname) {
1950 				case IPV6_2292PKTINFO:
1951 					optval = OPTBIT(IN6P_PKTINFO);
1952 					break;
1953 				case IPV6_2292HOPLIMIT:
1954 					optval = OPTBIT(IN6P_HOPLIMIT);
1955 					break;
1956 				case IPV6_2292HOPOPTS:
1957 					optval = OPTBIT(IN6P_HOPOPTS);
1958 					break;
1959 				case IPV6_2292RTHDR:
1960 					optval = OPTBIT(IN6P_RTHDR);
1961 					break;
1962 				case IPV6_2292DSTOPTS:
1963 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1964 					break;
1965 				}
1966 				error = sooptcopyout(sopt, &optval,
1967 				    sizeof optval);
1968 				break;
1969 			case IPV6_PKTINFO:
1970 			case IPV6_HOPOPTS:
1971 			case IPV6_RTHDR:
1972 			case IPV6_DSTOPTS:
1973 			case IPV6_RTHDRDSTOPTS:
1974 			case IPV6_NEXTHOP:
1975 			case IPV6_TCLASS:
1976 			case IPV6_DONTFRAG:
1977 			case IPV6_USE_MIN_MTU:
1978 			case IPV6_PREFER_TEMPADDR:
1979 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1980 				    optname, sopt);
1981 				break;
1982 
1983 			case IPV6_MULTICAST_IF:
1984 			case IPV6_MULTICAST_HOPS:
1985 			case IPV6_MULTICAST_LOOP:
1986 			case IPV6_JOIN_GROUP:
1987 			case IPV6_LEAVE_GROUP:
1988 			    {
1989 				struct mbuf *m;
1990 				error = ip6_getmoptions(sopt->sopt_name,
1991 				    in6p->in6p_moptions, &m);
1992 				if (error == 0)
1993 					error = sooptcopyout(sopt,
1994 					    mtod(m, char *), m->m_len);
1995 				m_freem(m);
1996 			    }
1997 				break;
1998 
1999 #ifdef IPSEC
2000 			case IPV6_IPSEC_POLICY:
2001 			  {
2002 				caddr_t req = NULL;
2003 				size_t len = 0;
2004 				struct mbuf *m = NULL;
2005 				struct mbuf **mp = &m;
2006 				size_t ovalsize = sopt->sopt_valsize;
2007 				caddr_t oval = (caddr_t)sopt->sopt_val;
2008 
2009 				error = soopt_getm(sopt, &m); /* XXX */
2010 				if (error != 0)
2011 					break;
2012 				error = soopt_mcopyin(sopt, m); /* XXX */
2013 				if (error != 0)
2014 					break;
2015 				sopt->sopt_valsize = ovalsize;
2016 				sopt->sopt_val = oval;
2017 				if (m) {
2018 					req = mtod(m, caddr_t);
2019 					len = m->m_len;
2020 				}
2021 				error = ipsec6_get_policy(in6p, req, len, mp);
2022 				if (error == 0)
2023 					error = soopt_mcopyout(sopt, m); /* XXX */
2024 				if (error == 0 && m)
2025 					m_freem(m);
2026 				break;
2027 			  }
2028 #endif /* IPSEC */
2029 
2030 			default:
2031 				error = ENOPROTOOPT;
2032 				break;
2033 			}
2034 			break;
2035 		}
2036 	} else {		/* level != IPPROTO_IPV6 */
2037 		error = EINVAL;
2038 	}
2039 	return (error);
2040 }
2041 
2042 int
2043 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2044 {
2045 	int error = 0, optval, optlen;
2046 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2047 	struct in6pcb *in6p = sotoin6pcb(so);
2048 	int level, op, optname;
2049 
2050 	level = sopt->sopt_level;
2051 	op = sopt->sopt_dir;
2052 	optname = sopt->sopt_name;
2053 	optlen = sopt->sopt_valsize;
2054 
2055 	if (level != IPPROTO_IPV6) {
2056 		return (EINVAL);
2057 	}
2058 
2059 	switch (optname) {
2060 	case IPV6_CHECKSUM:
2061 		/*
2062 		 * For ICMPv6 sockets, no modification allowed for checksum
2063 		 * offset, permit "no change" values to help existing apps.
2064 		 *
2065 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2066 		 * for an ICMPv6 socket will fail."
2067 		 * The current behavior does not meet RFC3542.
2068 		 */
2069 		switch (op) {
2070 		case SOPT_SET:
2071 			if (optlen != sizeof(int)) {
2072 				error = EINVAL;
2073 				break;
2074 			}
2075 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2076 					    sizeof(optval));
2077 			if (error)
2078 				break;
2079 			if ((optval % 2) != 0) {
2080 				/* the API assumes even offset values */
2081 				error = EINVAL;
2082 			} else if (so->so_proto->pr_protocol ==
2083 			    IPPROTO_ICMPV6) {
2084 				if (optval != icmp6off)
2085 					error = EINVAL;
2086 			} else
2087 				in6p->in6p_cksum = optval;
2088 			break;
2089 
2090 		case SOPT_GET:
2091 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2092 				optval = icmp6off;
2093 			else
2094 				optval = in6p->in6p_cksum;
2095 
2096 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2097 			break;
2098 
2099 		default:
2100 			error = EINVAL;
2101 			break;
2102 		}
2103 		break;
2104 
2105 	default:
2106 		error = ENOPROTOOPT;
2107 		break;
2108 	}
2109 
2110 	return (error);
2111 }
2112 
2113 /*
2114  * Set up IP6 options in pcb for insertion in output packets or
2115  * specifying behavior of outgoing packets.
2116  */
2117 static int
2118 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2119     struct socket *so, struct sockopt *sopt)
2120 {
2121 	struct ip6_pktopts *opt = *pktopt;
2122 	int error = 0;
2123 	struct thread *td = sopt->sopt_td;
2124 
2125 	/* turn off any old options. */
2126 	if (opt) {
2127 #ifdef DIAGNOSTIC
2128 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2129 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2130 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2131 			printf("ip6_pcbopts: all specified options are cleared.\n");
2132 #endif
2133 		ip6_clearpktopts(opt, -1);
2134 	} else
2135 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2136 	*pktopt = NULL;
2137 
2138 	if (!m || m->m_len == 0) {
2139 		/*
2140 		 * Only turning off any previous options, regardless of
2141 		 * whether the opt is just created or given.
2142 		 */
2143 		free(opt, M_IP6OPT);
2144 		return (0);
2145 	}
2146 
2147 	/*  set options specified by user. */
2148 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2149 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2150 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2151 		free(opt, M_IP6OPT);
2152 		return (error);
2153 	}
2154 	*pktopt = opt;
2155 	return (0);
2156 }
2157 
2158 /*
2159  * initialize ip6_pktopts.  beware that there are non-zero default values in
2160  * the struct.
2161  */
2162 void
2163 ip6_initpktopts(struct ip6_pktopts *opt)
2164 {
2165 
2166 	bzero(opt, sizeof(*opt));
2167 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2168 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2169 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2170 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2171 }
2172 
2173 static int
2174 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2175     struct ucred *cred, int uproto)
2176 {
2177 	struct ip6_pktopts *opt;
2178 
2179 	if (*pktopt == NULL) {
2180 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2181 		    M_WAITOK);
2182 		ip6_initpktopts(*pktopt);
2183 	}
2184 	opt = *pktopt;
2185 
2186 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2187 }
2188 
2189 static int
2190 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2191 {
2192 	void *optdata = NULL;
2193 	int optdatalen = 0;
2194 	struct ip6_ext *ip6e;
2195 	int error = 0;
2196 	struct in6_pktinfo null_pktinfo;
2197 	int deftclass = 0, on;
2198 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2199 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2200 
2201 	switch (optname) {
2202 	case IPV6_PKTINFO:
2203 		if (pktopt && pktopt->ip6po_pktinfo)
2204 			optdata = (void *)pktopt->ip6po_pktinfo;
2205 		else {
2206 			/* XXX: we don't have to do this every time... */
2207 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2208 			optdata = (void *)&null_pktinfo;
2209 		}
2210 		optdatalen = sizeof(struct in6_pktinfo);
2211 		break;
2212 	case IPV6_TCLASS:
2213 		if (pktopt && pktopt->ip6po_tclass >= 0)
2214 			optdata = (void *)&pktopt->ip6po_tclass;
2215 		else
2216 			optdata = (void *)&deftclass;
2217 		optdatalen = sizeof(int);
2218 		break;
2219 	case IPV6_HOPOPTS:
2220 		if (pktopt && pktopt->ip6po_hbh) {
2221 			optdata = (void *)pktopt->ip6po_hbh;
2222 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2223 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2224 		}
2225 		break;
2226 	case IPV6_RTHDR:
2227 		if (pktopt && pktopt->ip6po_rthdr) {
2228 			optdata = (void *)pktopt->ip6po_rthdr;
2229 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2230 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2231 		}
2232 		break;
2233 	case IPV6_RTHDRDSTOPTS:
2234 		if (pktopt && pktopt->ip6po_dest1) {
2235 			optdata = (void *)pktopt->ip6po_dest1;
2236 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2237 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2238 		}
2239 		break;
2240 	case IPV6_DSTOPTS:
2241 		if (pktopt && pktopt->ip6po_dest2) {
2242 			optdata = (void *)pktopt->ip6po_dest2;
2243 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2244 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2245 		}
2246 		break;
2247 	case IPV6_NEXTHOP:
2248 		if (pktopt && pktopt->ip6po_nexthop) {
2249 			optdata = (void *)pktopt->ip6po_nexthop;
2250 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2251 		}
2252 		break;
2253 	case IPV6_USE_MIN_MTU:
2254 		if (pktopt)
2255 			optdata = (void *)&pktopt->ip6po_minmtu;
2256 		else
2257 			optdata = (void *)&defminmtu;
2258 		optdatalen = sizeof(int);
2259 		break;
2260 	case IPV6_DONTFRAG:
2261 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2262 			on = 1;
2263 		else
2264 			on = 0;
2265 		optdata = (void *)&on;
2266 		optdatalen = sizeof(on);
2267 		break;
2268 	case IPV6_PREFER_TEMPADDR:
2269 		if (pktopt)
2270 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2271 		else
2272 			optdata = (void *)&defpreftemp;
2273 		optdatalen = sizeof(int);
2274 		break;
2275 	default:		/* should not happen */
2276 #ifdef DIAGNOSTIC
2277 		panic("ip6_getpcbopt: unexpected option\n");
2278 #endif
2279 		return (ENOPROTOOPT);
2280 	}
2281 
2282 	error = sooptcopyout(sopt, optdata, optdatalen);
2283 
2284 	return (error);
2285 }
2286 
2287 void
2288 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2289 {
2290 	if (pktopt == NULL)
2291 		return;
2292 
2293 	if (optname == -1 || optname == IPV6_PKTINFO) {
2294 		if (pktopt->ip6po_pktinfo)
2295 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2296 		pktopt->ip6po_pktinfo = NULL;
2297 	}
2298 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2299 		pktopt->ip6po_hlim = -1;
2300 	if (optname == -1 || optname == IPV6_TCLASS)
2301 		pktopt->ip6po_tclass = -1;
2302 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2303 		if (pktopt->ip6po_nextroute.ro_rt) {
2304 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2305 			pktopt->ip6po_nextroute.ro_rt = NULL;
2306 		}
2307 		if (pktopt->ip6po_nexthop)
2308 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2309 		pktopt->ip6po_nexthop = NULL;
2310 	}
2311 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2312 		if (pktopt->ip6po_hbh)
2313 			free(pktopt->ip6po_hbh, M_IP6OPT);
2314 		pktopt->ip6po_hbh = NULL;
2315 	}
2316 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2317 		if (pktopt->ip6po_dest1)
2318 			free(pktopt->ip6po_dest1, M_IP6OPT);
2319 		pktopt->ip6po_dest1 = NULL;
2320 	}
2321 	if (optname == -1 || optname == IPV6_RTHDR) {
2322 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2323 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2324 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2325 		if (pktopt->ip6po_route.ro_rt) {
2326 			RTFREE(pktopt->ip6po_route.ro_rt);
2327 			pktopt->ip6po_route.ro_rt = NULL;
2328 		}
2329 	}
2330 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2331 		if (pktopt->ip6po_dest2)
2332 			free(pktopt->ip6po_dest2, M_IP6OPT);
2333 		pktopt->ip6po_dest2 = NULL;
2334 	}
2335 }
2336 
2337 #define PKTOPT_EXTHDRCPY(type) \
2338 do {\
2339 	if (src->type) {\
2340 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2341 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2342 		if (dst->type == NULL && canwait == M_NOWAIT)\
2343 			goto bad;\
2344 		bcopy(src->type, dst->type, hlen);\
2345 	}\
2346 } while (/*CONSTCOND*/ 0)
2347 
2348 static int
2349 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2350 {
2351 	if (dst == NULL || src == NULL)  {
2352 		printf("ip6_clearpktopts: invalid argument\n");
2353 		return (EINVAL);
2354 	}
2355 
2356 	dst->ip6po_hlim = src->ip6po_hlim;
2357 	dst->ip6po_tclass = src->ip6po_tclass;
2358 	dst->ip6po_flags = src->ip6po_flags;
2359 	if (src->ip6po_pktinfo) {
2360 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2361 		    M_IP6OPT, canwait);
2362 		if (dst->ip6po_pktinfo == NULL)
2363 			goto bad;
2364 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2365 	}
2366 	if (src->ip6po_nexthop) {
2367 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2368 		    M_IP6OPT, canwait);
2369 		if (dst->ip6po_nexthop == NULL)
2370 			goto bad;
2371 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2372 		    src->ip6po_nexthop->sa_len);
2373 	}
2374 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2375 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2376 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2377 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2378 	return (0);
2379 
2380   bad:
2381 	ip6_clearpktopts(dst, -1);
2382 	return (ENOBUFS);
2383 }
2384 #undef PKTOPT_EXTHDRCPY
2385 
2386 struct ip6_pktopts *
2387 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2388 {
2389 	int error;
2390 	struct ip6_pktopts *dst;
2391 
2392 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2393 	if (dst == NULL)
2394 		return (NULL);
2395 	ip6_initpktopts(dst);
2396 
2397 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2398 		free(dst, M_IP6OPT);
2399 		return (NULL);
2400 	}
2401 
2402 	return (dst);
2403 }
2404 
2405 void
2406 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2407 {
2408 	if (pktopt == NULL)
2409 		return;
2410 
2411 	ip6_clearpktopts(pktopt, -1);
2412 
2413 	free(pktopt, M_IP6OPT);
2414 }
2415 
2416 /*
2417  * Set the IP6 multicast options in response to user setsockopt().
2418  */
2419 static int
2420 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2421 {
2422 	int error = 0;
2423 	u_int loop, ifindex;
2424 	struct ipv6_mreq *mreq;
2425 	struct ifnet *ifp;
2426 	struct ip6_moptions *im6o = *im6op;
2427 	struct route_in6 ro;
2428 	struct in6_multi_mship *imm;
2429 
2430 	if (im6o == NULL) {
2431 		/*
2432 		 * No multicast option buffer attached to the pcb;
2433 		 * allocate one and initialize to default values.
2434 		 */
2435 		im6o = (struct ip6_moptions *)
2436 			malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2437 
2438 		if (im6o == NULL)
2439 			return (ENOBUFS);
2440 		*im6op = im6o;
2441 		im6o->im6o_multicast_ifp = NULL;
2442 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2443 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2444 		LIST_INIT(&im6o->im6o_memberships);
2445 	}
2446 
2447 	switch (optname) {
2448 
2449 	case IPV6_MULTICAST_IF:
2450 		/*
2451 		 * Select the interface for outgoing multicast packets.
2452 		 */
2453 		if (m == NULL || m->m_len != sizeof(u_int)) {
2454 			error = EINVAL;
2455 			break;
2456 		}
2457 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2458 		if (ifindex < 0 || if_index < ifindex) {
2459 			error = ENXIO;	/* XXX EINVAL? */
2460 			break;
2461 		}
2462 		ifp = ifnet_byindex(ifindex);
2463 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2464 			error = EADDRNOTAVAIL;
2465 			break;
2466 		}
2467 		im6o->im6o_multicast_ifp = ifp;
2468 		break;
2469 
2470 	case IPV6_MULTICAST_HOPS:
2471 	    {
2472 		/*
2473 		 * Set the IP6 hoplimit for outgoing multicast packets.
2474 		 */
2475 		int optval;
2476 		if (m == NULL || m->m_len != sizeof(int)) {
2477 			error = EINVAL;
2478 			break;
2479 		}
2480 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2481 		if (optval < -1 || optval >= 256)
2482 			error = EINVAL;
2483 		else if (optval == -1)
2484 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2485 		else
2486 			im6o->im6o_multicast_hlim = optval;
2487 		break;
2488 	    }
2489 
2490 	case IPV6_MULTICAST_LOOP:
2491 		/*
2492 		 * Set the loopback flag for outgoing multicast packets.
2493 		 * Must be zero or one.
2494 		 */
2495 		if (m == NULL || m->m_len != sizeof(u_int)) {
2496 			error = EINVAL;
2497 			break;
2498 		}
2499 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2500 		if (loop > 1) {
2501 			error = EINVAL;
2502 			break;
2503 		}
2504 		im6o->im6o_multicast_loop = loop;
2505 		break;
2506 
2507 	case IPV6_JOIN_GROUP:
2508 		/*
2509 		 * Add a multicast group membership.
2510 		 * Group must be a valid IP6 multicast address.
2511 		 */
2512 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2513 			error = EINVAL;
2514 			break;
2515 		}
2516 		mreq = mtod(m, struct ipv6_mreq *);
2517 
2518 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2519 			/*
2520 			 * We use the unspecified address to specify to accept
2521 			 * all multicast addresses. Only super user is allowed
2522 			 * to do this.
2523 			 */
2524 			/* XXX-BZ might need a better PRIV_NETINET_x for this */
2525 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
2526 			if (error)
2527 				break;
2528 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2529 			error = EINVAL;
2530 			break;
2531 		}
2532 
2533 		/*
2534 		 * If no interface was explicitly specified, choose an
2535 		 * appropriate one according to the given multicast address.
2536 		 */
2537 		if (mreq->ipv6mr_interface == 0) {
2538 			struct sockaddr_in6 *dst;
2539 
2540 			/*
2541 			 * Look up the routing table for the
2542 			 * address, and choose the outgoing interface.
2543 			 *   XXX: is it a good approach?
2544 			 */
2545 			ro.ro_rt = NULL;
2546 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2547 			bzero(dst, sizeof(*dst));
2548 			dst->sin6_family = AF_INET6;
2549 			dst->sin6_len = sizeof(*dst);
2550 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2551 			rtalloc((struct route *)&ro);
2552 			if (ro.ro_rt == NULL) {
2553 				error = EADDRNOTAVAIL;
2554 				break;
2555 			}
2556 			ifp = ro.ro_rt->rt_ifp;
2557 			RTFREE(ro.ro_rt);
2558 		} else {
2559 			/*
2560 			 * If the interface is specified, validate it.
2561 			 */
2562 			if (mreq->ipv6mr_interface < 0 ||
2563 			    if_index < mreq->ipv6mr_interface) {
2564 				error = ENXIO;	/* XXX EINVAL? */
2565 				break;
2566 			}
2567 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2568 			if (!ifp) {
2569 				error = ENXIO;	/* XXX EINVAL? */
2570 				break;
2571 			}
2572 		}
2573 
2574 		/*
2575 		 * See if we found an interface, and confirm that it
2576 		 * supports multicast
2577 		 */
2578 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2579 			error = EADDRNOTAVAIL;
2580 			break;
2581 		}
2582 
2583 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2584 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2585 			break;
2586 		}
2587 
2588 		/*
2589 		 * See if the membership already exists.
2590 		 */
2591 		for (imm = im6o->im6o_memberships.lh_first;
2592 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2593 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2594 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2595 					       &mreq->ipv6mr_multiaddr))
2596 				break;
2597 		if (imm != NULL) {
2598 			error = EADDRINUSE;
2599 			break;
2600 		}
2601 		/*
2602 		 * Everything looks good; add a new record to the multicast
2603 		 * address list for the given interface.
2604 		 */
2605 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2606 		if (imm == NULL)
2607 			break;
2608 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2609 		break;
2610 
2611 	case IPV6_LEAVE_GROUP:
2612 		/*
2613 		 * Drop a multicast group membership.
2614 		 * Group must be a valid IP6 multicast address.
2615 		 */
2616 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2617 			error = EINVAL;
2618 			break;
2619 		}
2620 		mreq = mtod(m, struct ipv6_mreq *);
2621 
2622 		/*
2623 		 * If an interface address was specified, get a pointer
2624 		 * to its ifnet structure.
2625 		 */
2626 		if (mreq->ipv6mr_interface < 0 ||
2627 		    if_index < mreq->ipv6mr_interface) {
2628 			error = ENXIO;	/* XXX EINVAL? */
2629 			break;
2630 		}
2631 		if (mreq->ipv6mr_interface == 0)
2632 			ifp = NULL;
2633 		else
2634 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2635 
2636 		/* Fill in the scope zone ID */
2637 		if (ifp) {
2638 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2639 				/* XXX: should not happen */
2640 				error = EADDRNOTAVAIL;
2641 				break;
2642 			}
2643 		} else if (mreq->ipv6mr_interface != 0) {
2644 			/*
2645 			 * This case happens when the (positive) index is in
2646 			 * the valid range, but the corresponding interface has
2647 			 * been detached dynamically (XXX).
2648 			 */
2649 			error = EADDRNOTAVAIL;
2650 			break;
2651 		} else {	/* ipv6mr_interface == 0 */
2652 			struct sockaddr_in6 sa6_mc;
2653 
2654 			/*
2655 			 * The API spec says as follows:
2656 			 *  If the interface index is specified as 0, the
2657 			 *  system may choose a multicast group membership to
2658 			 *  drop by matching the multicast address only.
2659 			 * On the other hand, we cannot disambiguate the scope
2660 			 * zone unless an interface is provided.  Thus, we
2661 			 * check if there's ambiguity with the default scope
2662 			 * zone as the last resort.
2663 			 */
2664 			bzero(&sa6_mc, sizeof(sa6_mc));
2665 			sa6_mc.sin6_family = AF_INET6;
2666 			sa6_mc.sin6_len = sizeof(sa6_mc);
2667 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2668 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2669 			if (error != 0)
2670 				break;
2671 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2672 		}
2673 
2674 		/*
2675 		 * Find the membership in the membership list.
2676 		 */
2677 		for (imm = im6o->im6o_memberships.lh_first;
2678 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2679 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2680 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2681 			    &mreq->ipv6mr_multiaddr))
2682 				break;
2683 		}
2684 		if (imm == NULL) {
2685 			/* Unable to resolve interface */
2686 			error = EADDRNOTAVAIL;
2687 			break;
2688 		}
2689 		/*
2690 		 * Give up the multicast address record to which the
2691 		 * membership points.
2692 		 */
2693 		LIST_REMOVE(imm, i6mm_chain);
2694 		in6_delmulti(imm->i6mm_maddr);
2695 		free(imm, M_IP6MADDR);
2696 		break;
2697 
2698 	default:
2699 		error = EOPNOTSUPP;
2700 		break;
2701 	}
2702 
2703 	/*
2704 	 * If all options have default values, no need to keep the mbuf.
2705 	 */
2706 	if (im6o->im6o_multicast_ifp == NULL &&
2707 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2708 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2709 	    im6o->im6o_memberships.lh_first == NULL) {
2710 		free(*im6op, M_IP6MOPTS);
2711 		*im6op = NULL;
2712 	}
2713 
2714 	return (error);
2715 }
2716 
2717 /*
2718  * Return the IP6 multicast options in response to user getsockopt().
2719  */
2720 static int
2721 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2722 {
2723 	u_int *hlim, *loop, *ifindex;
2724 
2725 	*mp = m_get(M_WAIT, MT_HEADER);		/* XXX */
2726 
2727 	switch (optname) {
2728 
2729 	case IPV6_MULTICAST_IF:
2730 		ifindex = mtod(*mp, u_int *);
2731 		(*mp)->m_len = sizeof(u_int);
2732 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2733 			*ifindex = 0;
2734 		else
2735 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2736 		return (0);
2737 
2738 	case IPV6_MULTICAST_HOPS:
2739 		hlim = mtod(*mp, u_int *);
2740 		(*mp)->m_len = sizeof(u_int);
2741 		if (im6o == NULL)
2742 			*hlim = ip6_defmcasthlim;
2743 		else
2744 			*hlim = im6o->im6o_multicast_hlim;
2745 		return (0);
2746 
2747 	case IPV6_MULTICAST_LOOP:
2748 		loop = mtod(*mp, u_int *);
2749 		(*mp)->m_len = sizeof(u_int);
2750 		if (im6o == NULL)
2751 			*loop = ip6_defmcasthlim;
2752 		else
2753 			*loop = im6o->im6o_multicast_loop;
2754 		return (0);
2755 
2756 	default:
2757 		return (EOPNOTSUPP);
2758 	}
2759 }
2760 
2761 /*
2762  * Discard the IP6 multicast options.
2763  */
2764 void
2765 ip6_freemoptions(struct ip6_moptions *im6o)
2766 {
2767 	struct in6_multi_mship *imm;
2768 
2769 	if (im6o == NULL)
2770 		return;
2771 
2772 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2773 		LIST_REMOVE(imm, i6mm_chain);
2774 		if (imm->i6mm_maddr)
2775 			in6_delmulti(imm->i6mm_maddr);
2776 		free(imm, M_IP6MADDR);
2777 	}
2778 	free(im6o, M_IP6MOPTS);
2779 }
2780 
2781 /*
2782  * Set IPv6 outgoing packet options based on advanced API.
2783  */
2784 int
2785 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2786     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2787 {
2788 	struct cmsghdr *cm = 0;
2789 
2790 	if (control == NULL || opt == NULL)
2791 		return (EINVAL);
2792 
2793 	ip6_initpktopts(opt);
2794 	if (stickyopt) {
2795 		int error;
2796 
2797 		/*
2798 		 * If stickyopt is provided, make a local copy of the options
2799 		 * for this particular packet, then override them by ancillary
2800 		 * objects.
2801 		 * XXX: copypktopts() does not copy the cached route to a next
2802 		 * hop (if any).  This is not very good in terms of efficiency,
2803 		 * but we can allow this since this option should be rarely
2804 		 * used.
2805 		 */
2806 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2807 			return (error);
2808 	}
2809 
2810 	/*
2811 	 * XXX: Currently, we assume all the optional information is stored
2812 	 * in a single mbuf.
2813 	 */
2814 	if (control->m_next)
2815 		return (EINVAL);
2816 
2817 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2818 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2819 		int error;
2820 
2821 		if (control->m_len < CMSG_LEN(0))
2822 			return (EINVAL);
2823 
2824 		cm = mtod(control, struct cmsghdr *);
2825 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2826 			return (EINVAL);
2827 		if (cm->cmsg_level != IPPROTO_IPV6)
2828 			continue;
2829 
2830 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2831 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2832 		if (error)
2833 			return (error);
2834 	}
2835 
2836 	return (0);
2837 }
2838 
2839 /*
2840  * Set a particular packet option, as a sticky option or an ancillary data
2841  * item.  "len" can be 0 only when it's a sticky option.
2842  * We have 4 cases of combination of "sticky" and "cmsg":
2843  * "sticky=0, cmsg=0": impossible
2844  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2845  * "sticky=1, cmsg=0": RFC3542 socket option
2846  * "sticky=1, cmsg=1": RFC2292 socket option
2847  */
2848 static int
2849 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2850     struct ucred *cred, int sticky, int cmsg, int uproto)
2851 {
2852 	int minmtupolicy, preftemp;
2853 	int error;
2854 
2855 	if (!sticky && !cmsg) {
2856 #ifdef DIAGNOSTIC
2857 		printf("ip6_setpktopt: impossible case\n");
2858 #endif
2859 		return (EINVAL);
2860 	}
2861 
2862 	/*
2863 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2864 	 * not be specified in the context of RFC3542.  Conversely,
2865 	 * RFC3542 types should not be specified in the context of RFC2292.
2866 	 */
2867 	if (!cmsg) {
2868 		switch (optname) {
2869 		case IPV6_2292PKTINFO:
2870 		case IPV6_2292HOPLIMIT:
2871 		case IPV6_2292NEXTHOP:
2872 		case IPV6_2292HOPOPTS:
2873 		case IPV6_2292DSTOPTS:
2874 		case IPV6_2292RTHDR:
2875 		case IPV6_2292PKTOPTIONS:
2876 			return (ENOPROTOOPT);
2877 		}
2878 	}
2879 	if (sticky && cmsg) {
2880 		switch (optname) {
2881 		case IPV6_PKTINFO:
2882 		case IPV6_HOPLIMIT:
2883 		case IPV6_NEXTHOP:
2884 		case IPV6_HOPOPTS:
2885 		case IPV6_DSTOPTS:
2886 		case IPV6_RTHDRDSTOPTS:
2887 		case IPV6_RTHDR:
2888 		case IPV6_USE_MIN_MTU:
2889 		case IPV6_DONTFRAG:
2890 		case IPV6_TCLASS:
2891 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2892 			return (ENOPROTOOPT);
2893 		}
2894 	}
2895 
2896 	switch (optname) {
2897 	case IPV6_2292PKTINFO:
2898 	case IPV6_PKTINFO:
2899 	{
2900 		struct ifnet *ifp = NULL;
2901 		struct in6_pktinfo *pktinfo;
2902 
2903 		if (len != sizeof(struct in6_pktinfo))
2904 			return (EINVAL);
2905 
2906 		pktinfo = (struct in6_pktinfo *)buf;
2907 
2908 		/*
2909 		 * An application can clear any sticky IPV6_PKTINFO option by
2910 		 * doing a "regular" setsockopt with ipi6_addr being
2911 		 * in6addr_any and ipi6_ifindex being zero.
2912 		 * [RFC 3542, Section 6]
2913 		 */
2914 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2915 		    pktinfo->ipi6_ifindex == 0 &&
2916 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2917 			ip6_clearpktopts(opt, optname);
2918 			break;
2919 		}
2920 
2921 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2922 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2923 			return (EINVAL);
2924 		}
2925 
2926 		/* validate the interface index if specified. */
2927 		if (pktinfo->ipi6_ifindex > if_index ||
2928 		    pktinfo->ipi6_ifindex < 0) {
2929 			 return (ENXIO);
2930 		}
2931 		if (pktinfo->ipi6_ifindex) {
2932 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2933 			if (ifp == NULL)
2934 				return (ENXIO);
2935 		}
2936 
2937 		/*
2938 		 * We store the address anyway, and let in6_selectsrc()
2939 		 * validate the specified address.  This is because ipi6_addr
2940 		 * may not have enough information about its scope zone, and
2941 		 * we may need additional information (such as outgoing
2942 		 * interface or the scope zone of a destination address) to
2943 		 * disambiguate the scope.
2944 		 * XXX: the delay of the validation may confuse the
2945 		 * application when it is used as a sticky option.
2946 		 */
2947 		if (opt->ip6po_pktinfo == NULL) {
2948 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2949 			    M_IP6OPT, M_NOWAIT);
2950 			if (opt->ip6po_pktinfo == NULL)
2951 				return (ENOBUFS);
2952 		}
2953 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2954 		break;
2955 	}
2956 
2957 	case IPV6_2292HOPLIMIT:
2958 	case IPV6_HOPLIMIT:
2959 	{
2960 		int *hlimp;
2961 
2962 		/*
2963 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2964 		 * to simplify the ordering among hoplimit options.
2965 		 */
2966 		if (optname == IPV6_HOPLIMIT && sticky)
2967 			return (ENOPROTOOPT);
2968 
2969 		if (len != sizeof(int))
2970 			return (EINVAL);
2971 		hlimp = (int *)buf;
2972 		if (*hlimp < -1 || *hlimp > 255)
2973 			return (EINVAL);
2974 
2975 		opt->ip6po_hlim = *hlimp;
2976 		break;
2977 	}
2978 
2979 	case IPV6_TCLASS:
2980 	{
2981 		int tclass;
2982 
2983 		if (len != sizeof(int))
2984 			return (EINVAL);
2985 		tclass = *(int *)buf;
2986 		if (tclass < -1 || tclass > 255)
2987 			return (EINVAL);
2988 
2989 		opt->ip6po_tclass = tclass;
2990 		break;
2991 	}
2992 
2993 	case IPV6_2292NEXTHOP:
2994 	case IPV6_NEXTHOP:
2995 		if (cred != NULL) {
2996 			error = priv_check_cred(cred,
2997 			    PRIV_NETINET_SETHDROPTS, 0);
2998 			if (error)
2999 				return (error);
3000 		}
3001 
3002 		if (len == 0) {	/* just remove the option */
3003 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3004 			break;
3005 		}
3006 
3007 		/* check if cmsg_len is large enough for sa_len */
3008 		if (len < sizeof(struct sockaddr) || len < *buf)
3009 			return (EINVAL);
3010 
3011 		switch (((struct sockaddr *)buf)->sa_family) {
3012 		case AF_INET6:
3013 		{
3014 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3015 			int error;
3016 
3017 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3018 				return (EINVAL);
3019 
3020 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3021 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3022 				return (EINVAL);
3023 			}
3024 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3025 			    != 0) {
3026 				return (error);
3027 			}
3028 			break;
3029 		}
3030 		case AF_LINK:	/* should eventually be supported */
3031 		default:
3032 			return (EAFNOSUPPORT);
3033 		}
3034 
3035 		/* turn off the previous option, then set the new option. */
3036 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3037 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3038 		if (opt->ip6po_nexthop == NULL)
3039 			return (ENOBUFS);
3040 		bcopy(buf, opt->ip6po_nexthop, *buf);
3041 		break;
3042 
3043 	case IPV6_2292HOPOPTS:
3044 	case IPV6_HOPOPTS:
3045 	{
3046 		struct ip6_hbh *hbh;
3047 		int hbhlen;
3048 
3049 		/*
3050 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3051 		 * options, since per-option restriction has too much
3052 		 * overhead.
3053 		 */
3054 		if (cred != NULL) {
3055 			error = priv_check_cred(cred,
3056 			    PRIV_NETINET_SETHDROPTS, 0);
3057 			if (error)
3058 				return (error);
3059 		}
3060 
3061 		if (len == 0) {
3062 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3063 			break;	/* just remove the option */
3064 		}
3065 
3066 		/* message length validation */
3067 		if (len < sizeof(struct ip6_hbh))
3068 			return (EINVAL);
3069 		hbh = (struct ip6_hbh *)buf;
3070 		hbhlen = (hbh->ip6h_len + 1) << 3;
3071 		if (len != hbhlen)
3072 			return (EINVAL);
3073 
3074 		/* turn off the previous option, then set the new option. */
3075 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3076 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3077 		if (opt->ip6po_hbh == NULL)
3078 			return (ENOBUFS);
3079 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3080 
3081 		break;
3082 	}
3083 
3084 	case IPV6_2292DSTOPTS:
3085 	case IPV6_DSTOPTS:
3086 	case IPV6_RTHDRDSTOPTS:
3087 	{
3088 		struct ip6_dest *dest, **newdest = NULL;
3089 		int destlen;
3090 
3091 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3092 			error = priv_check_cred(cred,
3093 			    PRIV_NETINET_SETHDROPTS, 0);
3094 			if (error)
3095 				return (error);
3096 		}
3097 
3098 		if (len == 0) {
3099 			ip6_clearpktopts(opt, optname);
3100 			break;	/* just remove the option */
3101 		}
3102 
3103 		/* message length validation */
3104 		if (len < sizeof(struct ip6_dest))
3105 			return (EINVAL);
3106 		dest = (struct ip6_dest *)buf;
3107 		destlen = (dest->ip6d_len + 1) << 3;
3108 		if (len != destlen)
3109 			return (EINVAL);
3110 
3111 		/*
3112 		 * Determine the position that the destination options header
3113 		 * should be inserted; before or after the routing header.
3114 		 */
3115 		switch (optname) {
3116 		case IPV6_2292DSTOPTS:
3117 			/*
3118 			 * The old advacned API is ambiguous on this point.
3119 			 * Our approach is to determine the position based
3120 			 * according to the existence of a routing header.
3121 			 * Note, however, that this depends on the order of the
3122 			 * extension headers in the ancillary data; the 1st
3123 			 * part of the destination options header must appear
3124 			 * before the routing header in the ancillary data,
3125 			 * too.
3126 			 * RFC3542 solved the ambiguity by introducing
3127 			 * separate ancillary data or option types.
3128 			 */
3129 			if (opt->ip6po_rthdr == NULL)
3130 				newdest = &opt->ip6po_dest1;
3131 			else
3132 				newdest = &opt->ip6po_dest2;
3133 			break;
3134 		case IPV6_RTHDRDSTOPTS:
3135 			newdest = &opt->ip6po_dest1;
3136 			break;
3137 		case IPV6_DSTOPTS:
3138 			newdest = &opt->ip6po_dest2;
3139 			break;
3140 		}
3141 
3142 		/* turn off the previous option, then set the new option. */
3143 		ip6_clearpktopts(opt, optname);
3144 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3145 		if (*newdest == NULL)
3146 			return (ENOBUFS);
3147 		bcopy(dest, *newdest, destlen);
3148 
3149 		break;
3150 	}
3151 
3152 	case IPV6_2292RTHDR:
3153 	case IPV6_RTHDR:
3154 	{
3155 		struct ip6_rthdr *rth;
3156 		int rthlen;
3157 
3158 		if (len == 0) {
3159 			ip6_clearpktopts(opt, IPV6_RTHDR);
3160 			break;	/* just remove the option */
3161 		}
3162 
3163 		/* message length validation */
3164 		if (len < sizeof(struct ip6_rthdr))
3165 			return (EINVAL);
3166 		rth = (struct ip6_rthdr *)buf;
3167 		rthlen = (rth->ip6r_len + 1) << 3;
3168 		if (len != rthlen)
3169 			return (EINVAL);
3170 
3171 		switch (rth->ip6r_type) {
3172 		case IPV6_RTHDR_TYPE_0:
3173 			if (rth->ip6r_len == 0)	/* must contain one addr */
3174 				return (EINVAL);
3175 			if (rth->ip6r_len % 2) /* length must be even */
3176 				return (EINVAL);
3177 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3178 				return (EINVAL);
3179 			break;
3180 		default:
3181 			return (EINVAL);	/* not supported */
3182 		}
3183 
3184 		/* turn off the previous option */
3185 		ip6_clearpktopts(opt, IPV6_RTHDR);
3186 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3187 		if (opt->ip6po_rthdr == NULL)
3188 			return (ENOBUFS);
3189 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3190 
3191 		break;
3192 	}
3193 
3194 	case IPV6_USE_MIN_MTU:
3195 		if (len != sizeof(int))
3196 			return (EINVAL);
3197 		minmtupolicy = *(int *)buf;
3198 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3199 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3200 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3201 			return (EINVAL);
3202 		}
3203 		opt->ip6po_minmtu = minmtupolicy;
3204 		break;
3205 
3206 	case IPV6_DONTFRAG:
3207 		if (len != sizeof(int))
3208 			return (EINVAL);
3209 
3210 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3211 			/*
3212 			 * we ignore this option for TCP sockets.
3213 			 * (RFC3542 leaves this case unspecified.)
3214 			 */
3215 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3216 		} else
3217 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3218 		break;
3219 
3220 	case IPV6_PREFER_TEMPADDR:
3221 		if (len != sizeof(int))
3222 			return (EINVAL);
3223 		preftemp = *(int *)buf;
3224 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3225 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3226 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3227 			return (EINVAL);
3228 		}
3229 		opt->ip6po_prefer_tempaddr = preftemp;
3230 		break;
3231 
3232 	default:
3233 		return (ENOPROTOOPT);
3234 	} /* end of switch */
3235 
3236 	return (0);
3237 }
3238 
3239 /*
3240  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3241  * packet to the input queue of a specified interface.  Note that this
3242  * calls the output routine of the loopback "driver", but with an interface
3243  * pointer that might NOT be &loif -- easier than replicating that code here.
3244  */
3245 void
3246 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3247 {
3248 	struct mbuf *copym;
3249 	struct ip6_hdr *ip6;
3250 
3251 	copym = m_copy(m, 0, M_COPYALL);
3252 	if (copym == NULL)
3253 		return;
3254 
3255 	/*
3256 	 * Make sure to deep-copy IPv6 header portion in case the data
3257 	 * is in an mbuf cluster, so that we can safely override the IPv6
3258 	 * header portion later.
3259 	 */
3260 	if ((copym->m_flags & M_EXT) != 0 ||
3261 	    copym->m_len < sizeof(struct ip6_hdr)) {
3262 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3263 		if (copym == NULL)
3264 			return;
3265 	}
3266 
3267 #ifdef DIAGNOSTIC
3268 	if (copym->m_len < sizeof(*ip6)) {
3269 		m_freem(copym);
3270 		return;
3271 	}
3272 #endif
3273 
3274 	ip6 = mtod(copym, struct ip6_hdr *);
3275 	/*
3276 	 * clear embedded scope identifiers if necessary.
3277 	 * in6_clearscope will touch the addresses only when necessary.
3278 	 */
3279 	in6_clearscope(&ip6->ip6_src);
3280 	in6_clearscope(&ip6->ip6_dst);
3281 
3282 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3283 }
3284 
3285 /*
3286  * Chop IPv6 header off from the payload.
3287  */
3288 static int
3289 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3290 {
3291 	struct mbuf *mh;
3292 	struct ip6_hdr *ip6;
3293 
3294 	ip6 = mtod(m, struct ip6_hdr *);
3295 	if (m->m_len > sizeof(*ip6)) {
3296 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3297 		if (mh == 0) {
3298 			m_freem(m);
3299 			return ENOBUFS;
3300 		}
3301 		M_MOVE_PKTHDR(mh, m);
3302 		MH_ALIGN(mh, sizeof(*ip6));
3303 		m->m_len -= sizeof(*ip6);
3304 		m->m_data += sizeof(*ip6);
3305 		mh->m_next = m;
3306 		m = mh;
3307 		m->m_len = sizeof(*ip6);
3308 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3309 	}
3310 	exthdrs->ip6e_ip6 = m;
3311 	return 0;
3312 }
3313 
3314 /*
3315  * Compute IPv6 extension header length.
3316  */
3317 int
3318 ip6_optlen(struct in6pcb *in6p)
3319 {
3320 	int len;
3321 
3322 	if (!in6p->in6p_outputopts)
3323 		return 0;
3324 
3325 	len = 0;
3326 #define elen(x) \
3327     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3328 
3329 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3330 	if (in6p->in6p_outputopts->ip6po_rthdr)
3331 		/* dest1 is valid with rthdr only */
3332 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3333 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3334 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3335 	return len;
3336 #undef elen
3337 }
3338