xref: /freebsd/sys/netinet6/ip6_output.c (revision d2b2128a286a00ee53d79cb88b4e59bf42525cf9)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 #include "opt_route.h"
70 
71 #include <sys/param.h>
72 #include <sys/kernel.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/priv.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/syslog.h>
82 #include <sys/ucred.h>
83 #include <sys/vimage.h>
84 
85 #include <net/if.h>
86 #include <net/netisr.h>
87 #include <net/route.h>
88 #include <net/pfil.h>
89 #include <net/vnet.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet6/in6_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/tcp_var.h>
99 #include <netinet6/nd6.h>
100 #include <netinet/vinet.h>
101 
102 #ifdef IPSEC
103 #include <netipsec/ipsec.h>
104 #include <netipsec/ipsec6.h>
105 #include <netipsec/key.h>
106 #include <netinet6/ip6_ipsec.h>
107 #endif /* IPSEC */
108 
109 #include <netinet6/ip6protosw.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/vinet6.h>
112 
113 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
114 
115 struct ip6_exthdrs {
116 	struct mbuf *ip6e_ip6;
117 	struct mbuf *ip6e_hbh;
118 	struct mbuf *ip6e_dest1;
119 	struct mbuf *ip6e_rthdr;
120 	struct mbuf *ip6e_dest2;
121 };
122 
123 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
124 			   struct ucred *, int));
125 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
126 	struct socket *, struct sockopt *));
127 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
128 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
129 	struct ucred *, int, int, int));
130 
131 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
132 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
133 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
134 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
135 	struct ip6_frag **));
136 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
137 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
138 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
139 	struct ifnet *, struct in6_addr *, u_long *, int *));
140 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
141 
142 
143 /*
144  * Make an extension header from option data.  hp is the source, and
145  * mp is the destination.
146  */
147 #define MAKE_EXTHDR(hp, mp)						\
148     do {								\
149 	if (hp) {							\
150 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
151 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
152 		    ((eh)->ip6e_len + 1) << 3);				\
153 		if (error)						\
154 			goto freehdrs;					\
155 	}								\
156     } while (/*CONSTCOND*/ 0)
157 
158 /*
159  * Form a chain of extension headers.
160  * m is the extension header mbuf
161  * mp is the previous mbuf in the chain
162  * p is the next header
163  * i is the type of option.
164  */
165 #define MAKE_CHAIN(m, mp, p, i)\
166     do {\
167 	if (m) {\
168 		if (!hdrsplit) \
169 			panic("assumption failed: hdr not split"); \
170 		*mtod((m), u_char *) = *(p);\
171 		*(p) = (i);\
172 		p = mtod((m), u_char *);\
173 		(m)->m_next = (mp)->m_next;\
174 		(mp)->m_next = (m);\
175 		(mp) = (m);\
176 	}\
177     } while (/*CONSTCOND*/ 0)
178 
179 /*
180  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
181  * header (with pri, len, nxt, hlim, src, dst).
182  * This function may modify ver and hlim only.
183  * The mbuf chain containing the packet will be freed.
184  * The mbuf opt, if present, will not be freed.
185  *
186  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
187  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
188  * which is rt_rmx.rmx_mtu.
189  *
190  * ifpp - XXX: just for statistics
191  */
192 int
193 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
194     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
195     struct ifnet **ifpp, struct inpcb *inp)
196 {
197 	INIT_VNET_NET(curvnet);
198 	INIT_VNET_INET6(curvnet);
199 	struct ip6_hdr *ip6, *mhip6;
200 	struct ifnet *ifp, *origifp;
201 	struct mbuf *m = m0;
202 	struct mbuf *mprev = NULL;
203 	int hlen, tlen, len, off;
204 	struct route_in6 ip6route;
205 	struct rtentry *rt = NULL;
206 	struct sockaddr_in6 *dst, src_sa, dst_sa;
207 	struct in6_addr odst;
208 	int error = 0;
209 	struct in6_ifaddr *ia = NULL;
210 	u_long mtu;
211 	int alwaysfrag, dontfrag;
212 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
213 	struct ip6_exthdrs exthdrs;
214 	struct in6_addr finaldst, src0, dst0;
215 	u_int32_t zone;
216 	struct route_in6 *ro_pmtu = NULL;
217 	int hdrsplit = 0;
218 	int needipsec = 0;
219 #ifdef IPSEC
220 	struct ipsec_output_state state;
221 	struct ip6_rthdr *rh = NULL;
222 	int needipsectun = 0;
223 	int segleft_org = 0;
224 	struct secpolicy *sp = NULL;
225 #endif /* IPSEC */
226 
227 	ip6 = mtod(m, struct ip6_hdr *);
228 	if (ip6 == NULL) {
229 		printf ("ip6 is NULL");
230 		goto bad;
231 	}
232 
233 	finaldst = ip6->ip6_dst;
234 
235 	bzero(&exthdrs, sizeof(exthdrs));
236 
237 	if (opt) {
238 		/* Hop-by-Hop options header */
239 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
240 		/* Destination options header(1st part) */
241 		if (opt->ip6po_rthdr) {
242 			/*
243 			 * Destination options header(1st part)
244 			 * This only makes sense with a routing header.
245 			 * See Section 9.2 of RFC 3542.
246 			 * Disabling this part just for MIP6 convenience is
247 			 * a bad idea.  We need to think carefully about a
248 			 * way to make the advanced API coexist with MIP6
249 			 * options, which might automatically be inserted in
250 			 * the kernel.
251 			 */
252 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
253 		}
254 		/* Routing header */
255 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
256 		/* Destination options header(2nd part) */
257 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
258 	}
259 
260 	/*
261 	 * IPSec checking which handles several cases.
262 	 * FAST IPSEC: We re-injected the packet.
263 	 */
264 #ifdef IPSEC
265 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
266 	{
267 	case 1:                 /* Bad packet */
268 		goto freehdrs;
269 	case -1:                /* Do IPSec */
270 		needipsec = 1;
271 	case 0:                 /* No IPSec */
272 	default:
273 		break;
274 	}
275 #endif /* IPSEC */
276 
277 	/*
278 	 * Calculate the total length of the extension header chain.
279 	 * Keep the length of the unfragmentable part for fragmentation.
280 	 */
281 	optlen = 0;
282 	if (exthdrs.ip6e_hbh)
283 		optlen += exthdrs.ip6e_hbh->m_len;
284 	if (exthdrs.ip6e_dest1)
285 		optlen += exthdrs.ip6e_dest1->m_len;
286 	if (exthdrs.ip6e_rthdr)
287 		optlen += exthdrs.ip6e_rthdr->m_len;
288 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
289 
290 	/* NOTE: we don't add AH/ESP length here. do that later. */
291 	if (exthdrs.ip6e_dest2)
292 		optlen += exthdrs.ip6e_dest2->m_len;
293 
294 	/*
295 	 * If we need IPsec, or there is at least one extension header,
296 	 * separate IP6 header from the payload.
297 	 */
298 	if ((needipsec || optlen) && !hdrsplit) {
299 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
300 			m = NULL;
301 			goto freehdrs;
302 		}
303 		m = exthdrs.ip6e_ip6;
304 		hdrsplit++;
305 	}
306 
307 	/* adjust pointer */
308 	ip6 = mtod(m, struct ip6_hdr *);
309 
310 	/* adjust mbuf packet header length */
311 	m->m_pkthdr.len += optlen;
312 	plen = m->m_pkthdr.len - sizeof(*ip6);
313 
314 	/* If this is a jumbo payload, insert a jumbo payload option. */
315 	if (plen > IPV6_MAXPACKET) {
316 		if (!hdrsplit) {
317 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
318 				m = NULL;
319 				goto freehdrs;
320 			}
321 			m = exthdrs.ip6e_ip6;
322 			hdrsplit++;
323 		}
324 		/* adjust pointer */
325 		ip6 = mtod(m, struct ip6_hdr *);
326 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
327 			goto freehdrs;
328 		ip6->ip6_plen = 0;
329 	} else
330 		ip6->ip6_plen = htons(plen);
331 
332 	/*
333 	 * Concatenate headers and fill in next header fields.
334 	 * Here we have, on "m"
335 	 *	IPv6 payload
336 	 * and we insert headers accordingly.  Finally, we should be getting:
337 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
338 	 *
339 	 * during the header composing process, "m" points to IPv6 header.
340 	 * "mprev" points to an extension header prior to esp.
341 	 */
342 	u_char *nexthdrp = &ip6->ip6_nxt;
343 	mprev = m;
344 
345 	/*
346 	 * we treat dest2 specially.  this makes IPsec processing
347 	 * much easier.  the goal here is to make mprev point the
348 	 * mbuf prior to dest2.
349 	 *
350 	 * result: IPv6 dest2 payload
351 	 * m and mprev will point to IPv6 header.
352 	 */
353 	if (exthdrs.ip6e_dest2) {
354 		if (!hdrsplit)
355 			panic("assumption failed: hdr not split");
356 		exthdrs.ip6e_dest2->m_next = m->m_next;
357 		m->m_next = exthdrs.ip6e_dest2;
358 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
359 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
360 	}
361 
362 	/*
363 	 * result: IPv6 hbh dest1 rthdr dest2 payload
364 	 * m will point to IPv6 header.  mprev will point to the
365 	 * extension header prior to dest2 (rthdr in the above case).
366 	 */
367 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
368 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
369 		   IPPROTO_DSTOPTS);
370 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
371 		   IPPROTO_ROUTING);
372 
373 #ifdef IPSEC
374 	if (!needipsec)
375 		goto skip_ipsec2;
376 
377 	/*
378 	 * pointers after IPsec headers are not valid any more.
379 	 * other pointers need a great care too.
380 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
381 	 */
382 	exthdrs.ip6e_dest2 = NULL;
383 
384 	if (exthdrs.ip6e_rthdr) {
385 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
386 		segleft_org = rh->ip6r_segleft;
387 		rh->ip6r_segleft = 0;
388 	}
389 
390 	bzero(&state, sizeof(state));
391 	state.m = m;
392 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
393 				    &needipsectun);
394 	m = state.m;
395 	if (error == EJUSTRETURN) {
396 		/*
397 		 * We had a SP with a level of 'use' and no SA. We
398 		 * will just continue to process the packet without
399 		 * IPsec processing.
400 		 */
401 		;
402 	} else if (error) {
403 		/* mbuf is already reclaimed in ipsec6_output_trans. */
404 		m = NULL;
405 		switch (error) {
406 		case EHOSTUNREACH:
407 		case ENETUNREACH:
408 		case EMSGSIZE:
409 		case ENOBUFS:
410 		case ENOMEM:
411 			break;
412 		default:
413 			printf("[%s:%d] (ipsec): error code %d\n",
414 			    __func__, __LINE__, error);
415 			/* FALLTHROUGH */
416 		case ENOENT:
417 			/* don't show these error codes to the user */
418 			error = 0;
419 			break;
420 		}
421 		goto bad;
422 	} else if (!needipsectun) {
423 		/*
424 		 * In the FAST IPSec case we have already
425 		 * re-injected the packet and it has been freed
426 		 * by the ipsec_done() function.  So, just clean
427 		 * up after ourselves.
428 		 */
429 		m = NULL;
430 		goto done;
431 	}
432 	if (exthdrs.ip6e_rthdr) {
433 		/* ah6_output doesn't modify mbuf chain */
434 		rh->ip6r_segleft = segleft_org;
435 	}
436 skip_ipsec2:;
437 #endif /* IPSEC */
438 
439 	/*
440 	 * If there is a routing header, replace the destination address field
441 	 * with the first hop of the routing header.
442 	 */
443 	if (exthdrs.ip6e_rthdr) {
444 		struct ip6_rthdr *rh =
445 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
446 						  struct ip6_rthdr *));
447 
448 		/*
449 		 * While this switch may look gratuitous, leave it in
450 		 * in favour of RH2 implementations, etc.
451 		 */
452 		switch (rh->ip6r_type) {
453 #ifndef BURN_BRIDGES
454 		case IPV6_RTHDR_TYPE_0:
455 			/*
456 			 * According to RFC 5095 we should not implement
457 			 * it in any way but we may want to give the user
458 			 * a hint for now.
459 			 */
460 			log(LOG_INFO, "[%s:%d] IPv6 Type 0 Routing Headers are "
461 			    "deprecated.\n", __func__, __LINE__);
462 			/* FALLTHROUGH */
463 #endif
464 		default:	/* is it possible? */
465 			 error = EINVAL;
466 			 goto bad;
467 		}
468 	}
469 
470 	/* Source address validation */
471 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
472 	    (flags & IPV6_UNSPECSRC) == 0) {
473 		error = EOPNOTSUPP;
474 		V_ip6stat.ip6s_badscope++;
475 		goto bad;
476 	}
477 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
478 		error = EOPNOTSUPP;
479 		V_ip6stat.ip6s_badscope++;
480 		goto bad;
481 	}
482 
483 	V_ip6stat.ip6s_localout++;
484 
485 	/*
486 	 * Route packet.
487 	 */
488 	if (ro == 0) {
489 		ro = &ip6route;
490 		bzero((caddr_t)ro, sizeof(*ro));
491 	}
492 	ro_pmtu = ro;
493 	if (opt && opt->ip6po_rthdr)
494 		ro = &opt->ip6po_route;
495 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
496 
497 again:
498 	/*
499 	 * if specified, try to fill in the traffic class field.
500 	 * do not override if a non-zero value is already set.
501 	 * we check the diffserv field and the ecn field separately.
502 	 */
503 	if (opt && opt->ip6po_tclass >= 0) {
504 		int mask = 0;
505 
506 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
507 			mask |= 0xfc;
508 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
509 			mask |= 0x03;
510 		if (mask != 0)
511 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
512 	}
513 
514 	/* fill in or override the hop limit field, if necessary. */
515 	if (opt && opt->ip6po_hlim != -1)
516 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
517 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
518 		if (im6o != NULL)
519 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
520 		else
521 			ip6->ip6_hlim = V_ip6_defmcasthlim;
522 	}
523 
524 #ifdef IPSEC
525 	/*
526 	 * We may re-inject packets into the stack here.
527 	 */
528 	if (needipsec && needipsectun) {
529 		struct ipsec_output_state state;
530 
531 		/*
532 		 * All the extension headers will become inaccessible
533 		 * (since they can be encrypted).
534 		 * Don't panic, we need no more updates to extension headers
535 		 * on inner IPv6 packet (since they are now encapsulated).
536 		 *
537 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
538 		 */
539 		bzero(&exthdrs, sizeof(exthdrs));
540 		exthdrs.ip6e_ip6 = m;
541 
542 		bzero(&state, sizeof(state));
543 		state.m = m;
544 		state.ro = (struct route *)ro;
545 		state.dst = (struct sockaddr *)dst;
546 
547 		error = ipsec6_output_tunnel(&state, sp, flags);
548 
549 		m = state.m;
550 		ro = (struct route_in6 *)state.ro;
551 		dst = (struct sockaddr_in6 *)state.dst;
552 		if (error == EJUSTRETURN) {
553 			/*
554 			 * We had a SP with a level of 'use' and no SA. We
555 			 * will just continue to process the packet without
556 			 * IPsec processing.
557 			 */
558 			;
559 		} else if (error) {
560 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
561 			m0 = m = NULL;
562 			m = NULL;
563 			switch (error) {
564 			case EHOSTUNREACH:
565 			case ENETUNREACH:
566 			case EMSGSIZE:
567 			case ENOBUFS:
568 			case ENOMEM:
569 				break;
570 			default:
571 				printf("[%s:%d] (ipsec): error code %d\n",
572 				    __func__, __LINE__, error);
573 				/* FALLTHROUGH */
574 			case ENOENT:
575 				/* don't show these error codes to the user */
576 				error = 0;
577 				break;
578 			}
579 			goto bad;
580 		} else {
581 			/*
582 			 * In the FAST IPSec case we have already
583 			 * re-injected the packet and it has been freed
584 			 * by the ipsec_done() function.  So, just clean
585 			 * up after ourselves.
586 			 */
587 			m = NULL;
588 			goto done;
589 		}
590 
591 		exthdrs.ip6e_ip6 = m;
592 	}
593 #endif /* IPSEC */
594 
595 	/* adjust pointer */
596 	ip6 = mtod(m, struct ip6_hdr *);
597 
598 	bzero(&dst_sa, sizeof(dst_sa));
599 	dst_sa.sin6_family = AF_INET6;
600 	dst_sa.sin6_len = sizeof(dst_sa);
601 	dst_sa.sin6_addr = ip6->ip6_dst;
602 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
603 	    &ifp, &rt)) != 0) {
604 		switch (error) {
605 		case EHOSTUNREACH:
606 			V_ip6stat.ip6s_noroute++;
607 			break;
608 		case EADDRNOTAVAIL:
609 		default:
610 			break; /* XXX statistics? */
611 		}
612 		if (ifp != NULL)
613 			in6_ifstat_inc(ifp, ifs6_out_discard);
614 		goto bad;
615 	}
616 	if (rt == NULL) {
617 		/*
618 		 * If in6_selectroute() does not return a route entry,
619 		 * dst may not have been updated.
620 		 */
621 		*dst = dst_sa;	/* XXX */
622 	}
623 
624 	/*
625 	 * then rt (for unicast) and ifp must be non-NULL valid values.
626 	 */
627 	if ((flags & IPV6_FORWARDING) == 0) {
628 		/* XXX: the FORWARDING flag can be set for mrouting. */
629 		in6_ifstat_inc(ifp, ifs6_out_request);
630 	}
631 	if (rt != NULL) {
632 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
633 		rt->rt_use++;
634 	}
635 
636 	/*
637 	 * The outgoing interface must be in the zone of source and
638 	 * destination addresses.  We should use ia_ifp to support the
639 	 * case of sending packets to an address of our own.
640 	 */
641 	if (ia != NULL && ia->ia_ifp)
642 		origifp = ia->ia_ifp;
643 	else
644 		origifp = ifp;
645 
646 	src0 = ip6->ip6_src;
647 	if (in6_setscope(&src0, origifp, &zone))
648 		goto badscope;
649 	bzero(&src_sa, sizeof(src_sa));
650 	src_sa.sin6_family = AF_INET6;
651 	src_sa.sin6_len = sizeof(src_sa);
652 	src_sa.sin6_addr = ip6->ip6_src;
653 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
654 		goto badscope;
655 
656 	dst0 = ip6->ip6_dst;
657 	if (in6_setscope(&dst0, origifp, &zone))
658 		goto badscope;
659 	/* re-initialize to be sure */
660 	bzero(&dst_sa, sizeof(dst_sa));
661 	dst_sa.sin6_family = AF_INET6;
662 	dst_sa.sin6_len = sizeof(dst_sa);
663 	dst_sa.sin6_addr = ip6->ip6_dst;
664 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
665 		goto badscope;
666 	}
667 
668 	/* scope check is done. */
669 	goto routefound;
670 
671   badscope:
672 	V_ip6stat.ip6s_badscope++;
673 	in6_ifstat_inc(origifp, ifs6_out_discard);
674 	if (error == 0)
675 		error = EHOSTUNREACH; /* XXX */
676 	goto bad;
677 
678   routefound:
679 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
680 		if (opt && opt->ip6po_nextroute.ro_rt) {
681 			/*
682 			 * The nexthop is explicitly specified by the
683 			 * application.  We assume the next hop is an IPv6
684 			 * address.
685 			 */
686 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
687 		}
688 		else if ((rt->rt_flags & RTF_GATEWAY))
689 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
690 	}
691 
692 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
693 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
694 	} else {
695 		struct	in6_multi *in6m;
696 
697 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
698 
699 		in6_ifstat_inc(ifp, ifs6_out_mcast);
700 
701 		/*
702 		 * Confirm that the outgoing interface supports multicast.
703 		 */
704 		if (!(ifp->if_flags & IFF_MULTICAST)) {
705 			V_ip6stat.ip6s_noroute++;
706 			in6_ifstat_inc(ifp, ifs6_out_discard);
707 			error = ENETUNREACH;
708 			goto bad;
709 		}
710 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
711 		if (in6m != NULL &&
712 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
713 			/*
714 			 * If we belong to the destination multicast group
715 			 * on the outgoing interface, and the caller did not
716 			 * forbid loopback, loop back a copy.
717 			 */
718 			ip6_mloopback(ifp, m, dst);
719 		} else {
720 			/*
721 			 * If we are acting as a multicast router, perform
722 			 * multicast forwarding as if the packet had just
723 			 * arrived on the interface to which we are about
724 			 * to send.  The multicast forwarding function
725 			 * recursively calls this function, using the
726 			 * IPV6_FORWARDING flag to prevent infinite recursion.
727 			 *
728 			 * Multicasts that are looped back by ip6_mloopback(),
729 			 * above, will be forwarded by the ip6_input() routine,
730 			 * if necessary.
731 			 */
732 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
733 				/*
734 				 * XXX: ip6_mforward expects that rcvif is NULL
735 				 * when it is called from the originating path.
736 				 * However, it is not always the case, since
737 				 * some versions of MGETHDR() does not
738 				 * initialize the field.
739 				 */
740 				m->m_pkthdr.rcvif = NULL;
741 				if (ip6_mforward(ip6, ifp, m) != 0) {
742 					m_freem(m);
743 					goto done;
744 				}
745 			}
746 		}
747 		/*
748 		 * Multicasts with a hoplimit of zero may be looped back,
749 		 * above, but must not be transmitted on a network.
750 		 * Also, multicasts addressed to the loopback interface
751 		 * are not sent -- the above call to ip6_mloopback() will
752 		 * loop back a copy if this host actually belongs to the
753 		 * destination group on the loopback interface.
754 		 */
755 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
756 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
757 			m_freem(m);
758 			goto done;
759 		}
760 	}
761 
762 	/*
763 	 * Fill the outgoing inteface to tell the upper layer
764 	 * to increment per-interface statistics.
765 	 */
766 	if (ifpp)
767 		*ifpp = ifp;
768 
769 	/* Determine path MTU. */
770 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
771 	    &alwaysfrag)) != 0)
772 		goto bad;
773 
774 	/*
775 	 * The caller of this function may specify to use the minimum MTU
776 	 * in some cases.
777 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
778 	 * setting.  The logic is a bit complicated; by default, unicast
779 	 * packets will follow path MTU while multicast packets will be sent at
780 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
781 	 * including unicast ones will be sent at the minimum MTU.  Multicast
782 	 * packets will always be sent at the minimum MTU unless
783 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
784 	 * See RFC 3542 for more details.
785 	 */
786 	if (mtu > IPV6_MMTU) {
787 		if ((flags & IPV6_MINMTU))
788 			mtu = IPV6_MMTU;
789 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
790 			mtu = IPV6_MMTU;
791 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
792 			 (opt == NULL ||
793 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
794 			mtu = IPV6_MMTU;
795 		}
796 	}
797 
798 	/*
799 	 * clear embedded scope identifiers if necessary.
800 	 * in6_clearscope will touch the addresses only when necessary.
801 	 */
802 	in6_clearscope(&ip6->ip6_src);
803 	in6_clearscope(&ip6->ip6_dst);
804 
805 	/*
806 	 * If the outgoing packet contains a hop-by-hop options header,
807 	 * it must be examined and processed even by the source node.
808 	 * (RFC 2460, section 4.)
809 	 */
810 	if (exthdrs.ip6e_hbh) {
811 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
812 		u_int32_t dummy; /* XXX unused */
813 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
814 
815 #ifdef DIAGNOSTIC
816 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
817 			panic("ip6e_hbh is not continuous");
818 #endif
819 		/*
820 		 *  XXX: if we have to send an ICMPv6 error to the sender,
821 		 *       we need the M_LOOP flag since icmp6_error() expects
822 		 *       the IPv6 and the hop-by-hop options header are
823 		 *       continuous unless the flag is set.
824 		 */
825 		m->m_flags |= M_LOOP;
826 		m->m_pkthdr.rcvif = ifp;
827 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
828 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
829 		    &dummy, &plen) < 0) {
830 			/* m was already freed at this point */
831 			error = EINVAL;/* better error? */
832 			goto done;
833 		}
834 		m->m_flags &= ~M_LOOP; /* XXX */
835 		m->m_pkthdr.rcvif = NULL;
836 	}
837 
838 	/* Jump over all PFIL processing if hooks are not active. */
839 	if (!PFIL_HOOKED(&inet6_pfil_hook))
840 		goto passout;
841 
842 	odst = ip6->ip6_dst;
843 	/* Run through list of hooks for output packets. */
844 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
845 	if (error != 0 || m == NULL)
846 		goto done;
847 	ip6 = mtod(m, struct ip6_hdr *);
848 
849 	/* See if destination IP address was changed by packet filter. */
850 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
851 		m->m_flags |= M_SKIP_FIREWALL;
852 		/* If destination is now ourself drop to ip6_input(). */
853 		if (in6_localaddr(&ip6->ip6_dst)) {
854 			if (m->m_pkthdr.rcvif == NULL)
855 				m->m_pkthdr.rcvif = V_loif;
856 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
857 				m->m_pkthdr.csum_flags |=
858 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
859 				m->m_pkthdr.csum_data = 0xffff;
860 			}
861 			m->m_pkthdr.csum_flags |=
862 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
863 			error = netisr_queue(NETISR_IPV6, m);
864 			goto done;
865 		} else
866 			goto again;	/* Redo the routing table lookup. */
867 	}
868 
869 	/* XXX: IPFIREWALL_FORWARD */
870 
871 passout:
872 	/*
873 	 * Send the packet to the outgoing interface.
874 	 * If necessary, do IPv6 fragmentation before sending.
875 	 *
876 	 * the logic here is rather complex:
877 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
878 	 * 1-a:	send as is if tlen <= path mtu
879 	 * 1-b:	fragment if tlen > path mtu
880 	 *
881 	 * 2: if user asks us not to fragment (dontfrag == 1)
882 	 * 2-a:	send as is if tlen <= interface mtu
883 	 * 2-b:	error if tlen > interface mtu
884 	 *
885 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
886 	 *	always fragment
887 	 *
888 	 * 4: if dontfrag == 1 && alwaysfrag == 1
889 	 *	error, as we cannot handle this conflicting request
890 	 */
891 	tlen = m->m_pkthdr.len;
892 
893 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
894 		dontfrag = 1;
895 	else
896 		dontfrag = 0;
897 	if (dontfrag && alwaysfrag) {	/* case 4 */
898 		/* conflicting request - can't transmit */
899 		error = EMSGSIZE;
900 		goto bad;
901 	}
902 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
903 		/*
904 		 * Even if the DONTFRAG option is specified, we cannot send the
905 		 * packet when the data length is larger than the MTU of the
906 		 * outgoing interface.
907 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
908 		 * well as returning an error code (the latter is not described
909 		 * in the API spec.)
910 		 */
911 		u_int32_t mtu32;
912 		struct ip6ctlparam ip6cp;
913 
914 		mtu32 = (u_int32_t)mtu;
915 		bzero(&ip6cp, sizeof(ip6cp));
916 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
917 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
918 		    (void *)&ip6cp);
919 
920 		error = EMSGSIZE;
921 		goto bad;
922 	}
923 
924 	/*
925 	 * transmit packet without fragmentation
926 	 */
927 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
928 		struct in6_ifaddr *ia6;
929 
930 		ip6 = mtod(m, struct ip6_hdr *);
931 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
932 		if (ia6) {
933 			/* Record statistics for this interface address. */
934 			ia6->ia_ifa.if_opackets++;
935 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
936 		}
937 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
938 		goto done;
939 	}
940 
941 	/*
942 	 * try to fragment the packet.  case 1-b and 3
943 	 */
944 	if (mtu < IPV6_MMTU) {
945 		/* path MTU cannot be less than IPV6_MMTU */
946 		error = EMSGSIZE;
947 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
948 		goto bad;
949 	} else if (ip6->ip6_plen == 0) {
950 		/* jumbo payload cannot be fragmented */
951 		error = EMSGSIZE;
952 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
953 		goto bad;
954 	} else {
955 		struct mbuf **mnext, *m_frgpart;
956 		struct ip6_frag *ip6f;
957 		u_int32_t id = htonl(ip6_randomid());
958 		u_char nextproto;
959 
960 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
961 
962 		/*
963 		 * Too large for the destination or interface;
964 		 * fragment if possible.
965 		 * Must be able to put at least 8 bytes per fragment.
966 		 */
967 		hlen = unfragpartlen;
968 		if (mtu > IPV6_MAXPACKET)
969 			mtu = IPV6_MAXPACKET;
970 
971 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
972 		if (len < 8) {
973 			error = EMSGSIZE;
974 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
975 			goto bad;
976 		}
977 
978 		/*
979 		 * Verify that we have any chance at all of being able to queue
980 		 *      the packet or packet fragments
981 		 */
982 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
983 		    < tlen  /* - hlen */)) {
984 			error = ENOBUFS;
985 			V_ip6stat.ip6s_odropped++;
986 			goto bad;
987 		}
988 
989 		mnext = &m->m_nextpkt;
990 
991 		/*
992 		 * Change the next header field of the last header in the
993 		 * unfragmentable part.
994 		 */
995 		if (exthdrs.ip6e_rthdr) {
996 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
997 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
998 		} else if (exthdrs.ip6e_dest1) {
999 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1000 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1001 		} else if (exthdrs.ip6e_hbh) {
1002 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1003 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1004 		} else {
1005 			nextproto = ip6->ip6_nxt;
1006 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1007 		}
1008 
1009 		/*
1010 		 * Loop through length of segment after first fragment,
1011 		 * make new header and copy data of each part and link onto
1012 		 * chain.
1013 		 */
1014 		m0 = m;
1015 		for (off = hlen; off < tlen; off += len) {
1016 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1017 			if (!m) {
1018 				error = ENOBUFS;
1019 				V_ip6stat.ip6s_odropped++;
1020 				goto sendorfree;
1021 			}
1022 			m->m_pkthdr.rcvif = NULL;
1023 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1024 			*mnext = m;
1025 			mnext = &m->m_nextpkt;
1026 			m->m_data += max_linkhdr;
1027 			mhip6 = mtod(m, struct ip6_hdr *);
1028 			*mhip6 = *ip6;
1029 			m->m_len = sizeof(*mhip6);
1030 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1031 			if (error) {
1032 				V_ip6stat.ip6s_odropped++;
1033 				goto sendorfree;
1034 			}
1035 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1036 			if (off + len >= tlen)
1037 				len = tlen - off;
1038 			else
1039 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1040 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1041 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1042 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1043 				error = ENOBUFS;
1044 				V_ip6stat.ip6s_odropped++;
1045 				goto sendorfree;
1046 			}
1047 			m_cat(m, m_frgpart);
1048 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1049 			m->m_pkthdr.rcvif = NULL;
1050 			ip6f->ip6f_reserved = 0;
1051 			ip6f->ip6f_ident = id;
1052 			ip6f->ip6f_nxt = nextproto;
1053 			V_ip6stat.ip6s_ofragments++;
1054 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1055 		}
1056 
1057 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1058 	}
1059 
1060 	/*
1061 	 * Remove leading garbages.
1062 	 */
1063 sendorfree:
1064 	m = m0->m_nextpkt;
1065 	m0->m_nextpkt = 0;
1066 	m_freem(m0);
1067 	for (m0 = m; m; m = m0) {
1068 		m0 = m->m_nextpkt;
1069 		m->m_nextpkt = 0;
1070 		if (error == 0) {
1071 			/* Record statistics for this interface address. */
1072 			if (ia) {
1073 				ia->ia_ifa.if_opackets++;
1074 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1075 			}
1076 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1077 		} else
1078 			m_freem(m);
1079 	}
1080 
1081 	if (error == 0)
1082 		V_ip6stat.ip6s_fragmented++;
1083 
1084 done:
1085 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1086 		RTFREE(ro->ro_rt);
1087 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1088 		RTFREE(ro_pmtu->ro_rt);
1089 	}
1090 #ifdef IPSEC
1091 	if (sp != NULL)
1092 		KEY_FREESP(&sp);
1093 #endif
1094 
1095 	return (error);
1096 
1097 freehdrs:
1098 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1099 	m_freem(exthdrs.ip6e_dest1);
1100 	m_freem(exthdrs.ip6e_rthdr);
1101 	m_freem(exthdrs.ip6e_dest2);
1102 	/* FALLTHROUGH */
1103 bad:
1104 	if (m)
1105 		m_freem(m);
1106 	goto done;
1107 }
1108 
1109 static int
1110 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1111 {
1112 	struct mbuf *m;
1113 
1114 	if (hlen > MCLBYTES)
1115 		return (ENOBUFS); /* XXX */
1116 
1117 	MGET(m, M_DONTWAIT, MT_DATA);
1118 	if (!m)
1119 		return (ENOBUFS);
1120 
1121 	if (hlen > MLEN) {
1122 		MCLGET(m, M_DONTWAIT);
1123 		if ((m->m_flags & M_EXT) == 0) {
1124 			m_free(m);
1125 			return (ENOBUFS);
1126 		}
1127 	}
1128 	m->m_len = hlen;
1129 	if (hdr)
1130 		bcopy(hdr, mtod(m, caddr_t), hlen);
1131 
1132 	*mp = m;
1133 	return (0);
1134 }
1135 
1136 /*
1137  * Insert jumbo payload option.
1138  */
1139 static int
1140 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1141 {
1142 	struct mbuf *mopt;
1143 	u_char *optbuf;
1144 	u_int32_t v;
1145 
1146 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1147 
1148 	/*
1149 	 * If there is no hop-by-hop options header, allocate new one.
1150 	 * If there is one but it doesn't have enough space to store the
1151 	 * jumbo payload option, allocate a cluster to store the whole options.
1152 	 * Otherwise, use it to store the options.
1153 	 */
1154 	if (exthdrs->ip6e_hbh == 0) {
1155 		MGET(mopt, M_DONTWAIT, MT_DATA);
1156 		if (mopt == 0)
1157 			return (ENOBUFS);
1158 		mopt->m_len = JUMBOOPTLEN;
1159 		optbuf = mtod(mopt, u_char *);
1160 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1161 		exthdrs->ip6e_hbh = mopt;
1162 	} else {
1163 		struct ip6_hbh *hbh;
1164 
1165 		mopt = exthdrs->ip6e_hbh;
1166 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1167 			/*
1168 			 * XXX assumption:
1169 			 * - exthdrs->ip6e_hbh is not referenced from places
1170 			 *   other than exthdrs.
1171 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1172 			 */
1173 			int oldoptlen = mopt->m_len;
1174 			struct mbuf *n;
1175 
1176 			/*
1177 			 * XXX: give up if the whole (new) hbh header does
1178 			 * not fit even in an mbuf cluster.
1179 			 */
1180 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1181 				return (ENOBUFS);
1182 
1183 			/*
1184 			 * As a consequence, we must always prepare a cluster
1185 			 * at this point.
1186 			 */
1187 			MGET(n, M_DONTWAIT, MT_DATA);
1188 			if (n) {
1189 				MCLGET(n, M_DONTWAIT);
1190 				if ((n->m_flags & M_EXT) == 0) {
1191 					m_freem(n);
1192 					n = NULL;
1193 				}
1194 			}
1195 			if (!n)
1196 				return (ENOBUFS);
1197 			n->m_len = oldoptlen + JUMBOOPTLEN;
1198 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1199 			    oldoptlen);
1200 			optbuf = mtod(n, caddr_t) + oldoptlen;
1201 			m_freem(mopt);
1202 			mopt = exthdrs->ip6e_hbh = n;
1203 		} else {
1204 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1205 			mopt->m_len += JUMBOOPTLEN;
1206 		}
1207 		optbuf[0] = IP6OPT_PADN;
1208 		optbuf[1] = 1;
1209 
1210 		/*
1211 		 * Adjust the header length according to the pad and
1212 		 * the jumbo payload option.
1213 		 */
1214 		hbh = mtod(mopt, struct ip6_hbh *);
1215 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1216 	}
1217 
1218 	/* fill in the option. */
1219 	optbuf[2] = IP6OPT_JUMBO;
1220 	optbuf[3] = 4;
1221 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1222 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1223 
1224 	/* finally, adjust the packet header length */
1225 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1226 
1227 	return (0);
1228 #undef JUMBOOPTLEN
1229 }
1230 
1231 /*
1232  * Insert fragment header and copy unfragmentable header portions.
1233  */
1234 static int
1235 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1236     struct ip6_frag **frghdrp)
1237 {
1238 	struct mbuf *n, *mlast;
1239 
1240 	if (hlen > sizeof(struct ip6_hdr)) {
1241 		n = m_copym(m0, sizeof(struct ip6_hdr),
1242 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1243 		if (n == 0)
1244 			return (ENOBUFS);
1245 		m->m_next = n;
1246 	} else
1247 		n = m;
1248 
1249 	/* Search for the last mbuf of unfragmentable part. */
1250 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1251 		;
1252 
1253 	if ((mlast->m_flags & M_EXT) == 0 &&
1254 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1255 		/* use the trailing space of the last mbuf for the fragment hdr */
1256 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1257 		    mlast->m_len);
1258 		mlast->m_len += sizeof(struct ip6_frag);
1259 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1260 	} else {
1261 		/* allocate a new mbuf for the fragment header */
1262 		struct mbuf *mfrg;
1263 
1264 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1265 		if (mfrg == 0)
1266 			return (ENOBUFS);
1267 		mfrg->m_len = sizeof(struct ip6_frag);
1268 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1269 		mlast->m_next = mfrg;
1270 	}
1271 
1272 	return (0);
1273 }
1274 
1275 static int
1276 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1277     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1278     int *alwaysfragp)
1279 {
1280 	u_int32_t mtu = 0;
1281 	int alwaysfrag = 0;
1282 	int error = 0;
1283 
1284 	if (ro_pmtu != ro) {
1285 		/* The first hop and the final destination may differ. */
1286 		struct sockaddr_in6 *sa6_dst =
1287 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1288 		if (ro_pmtu->ro_rt &&
1289 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1290 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1291 			RTFREE(ro_pmtu->ro_rt);
1292 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1293 		}
1294 		if (ro_pmtu->ro_rt == NULL) {
1295 			bzero(sa6_dst, sizeof(*sa6_dst));
1296 			sa6_dst->sin6_family = AF_INET6;
1297 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1298 			sa6_dst->sin6_addr = *dst;
1299 
1300 			rtalloc((struct route *)ro_pmtu);
1301 		}
1302 	}
1303 	if (ro_pmtu->ro_rt) {
1304 		u_int32_t ifmtu;
1305 		struct in_conninfo inc;
1306 
1307 		bzero(&inc, sizeof(inc));
1308 		inc.inc_flags |= INC_ISIPV6;
1309 		inc.inc6_faddr = *dst;
1310 
1311 		if (ifp == NULL)
1312 			ifp = ro_pmtu->ro_rt->rt_ifp;
1313 		ifmtu = IN6_LINKMTU(ifp);
1314 		mtu = tcp_hc_getmtu(&inc);
1315 		if (mtu)
1316 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1317 		else
1318 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1319 		if (mtu == 0)
1320 			mtu = ifmtu;
1321 		else if (mtu < IPV6_MMTU) {
1322 			/*
1323 			 * RFC2460 section 5, last paragraph:
1324 			 * if we record ICMPv6 too big message with
1325 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1326 			 * or smaller, with framgent header attached.
1327 			 * (fragment header is needed regardless from the
1328 			 * packet size, for translators to identify packets)
1329 			 */
1330 			alwaysfrag = 1;
1331 			mtu = IPV6_MMTU;
1332 		} else if (mtu > ifmtu) {
1333 			/*
1334 			 * The MTU on the route is larger than the MTU on
1335 			 * the interface!  This shouldn't happen, unless the
1336 			 * MTU of the interface has been changed after the
1337 			 * interface was brought up.  Change the MTU in the
1338 			 * route to match the interface MTU (as long as the
1339 			 * field isn't locked).
1340 			 */
1341 			mtu = ifmtu;
1342 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1343 		}
1344 	} else if (ifp) {
1345 		mtu = IN6_LINKMTU(ifp);
1346 	} else
1347 		error = EHOSTUNREACH; /* XXX */
1348 
1349 	*mtup = mtu;
1350 	if (alwaysfragp)
1351 		*alwaysfragp = alwaysfrag;
1352 	return (error);
1353 }
1354 
1355 /*
1356  * IP6 socket option processing.
1357  */
1358 int
1359 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1360 {
1361 	int optdatalen, uproto;
1362 	void *optdata;
1363 	struct inpcb *in6p = sotoinpcb(so);
1364 	int error, optval;
1365 	int level, op, optname;
1366 	int optlen;
1367 	struct thread *td;
1368 
1369 	level = sopt->sopt_level;
1370 	op = sopt->sopt_dir;
1371 	optname = sopt->sopt_name;
1372 	optlen = sopt->sopt_valsize;
1373 	td = sopt->sopt_td;
1374 	error = 0;
1375 	optval = 0;
1376 	uproto = (int)so->so_proto->pr_protocol;
1377 
1378 	if (level == IPPROTO_IPV6) {
1379 		switch (op) {
1380 
1381 		case SOPT_SET:
1382 			switch (optname) {
1383 			case IPV6_2292PKTOPTIONS:
1384 #ifdef IPV6_PKTOPTIONS
1385 			case IPV6_PKTOPTIONS:
1386 #endif
1387 			{
1388 				struct mbuf *m;
1389 
1390 				error = soopt_getm(sopt, &m); /* XXX */
1391 				if (error != 0)
1392 					break;
1393 				error = soopt_mcopyin(sopt, m); /* XXX */
1394 				if (error != 0)
1395 					break;
1396 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1397 						    m, so, sopt);
1398 				m_freem(m); /* XXX */
1399 				break;
1400 			}
1401 
1402 			/*
1403 			 * Use of some Hop-by-Hop options or some
1404 			 * Destination options, might require special
1405 			 * privilege.  That is, normal applications
1406 			 * (without special privilege) might be forbidden
1407 			 * from setting certain options in outgoing packets,
1408 			 * and might never see certain options in received
1409 			 * packets. [RFC 2292 Section 6]
1410 			 * KAME specific note:
1411 			 *  KAME prevents non-privileged users from sending or
1412 			 *  receiving ANY hbh/dst options in order to avoid
1413 			 *  overhead of parsing options in the kernel.
1414 			 */
1415 			case IPV6_RECVHOPOPTS:
1416 			case IPV6_RECVDSTOPTS:
1417 			case IPV6_RECVRTHDRDSTOPTS:
1418 				if (td != NULL) {
1419 					error = priv_check(td,
1420 					    PRIV_NETINET_SETHDROPTS);
1421 					if (error)
1422 						break;
1423 				}
1424 				/* FALLTHROUGH */
1425 			case IPV6_UNICAST_HOPS:
1426 			case IPV6_HOPLIMIT:
1427 			case IPV6_FAITH:
1428 
1429 			case IPV6_RECVPKTINFO:
1430 			case IPV6_RECVHOPLIMIT:
1431 			case IPV6_RECVRTHDR:
1432 			case IPV6_RECVPATHMTU:
1433 			case IPV6_RECVTCLASS:
1434 			case IPV6_V6ONLY:
1435 			case IPV6_AUTOFLOWLABEL:
1436 				if (optlen != sizeof(int)) {
1437 					error = EINVAL;
1438 					break;
1439 				}
1440 				error = sooptcopyin(sopt, &optval,
1441 					sizeof optval, sizeof optval);
1442 				if (error)
1443 					break;
1444 				switch (optname) {
1445 
1446 				case IPV6_UNICAST_HOPS:
1447 					if (optval < -1 || optval >= 256)
1448 						error = EINVAL;
1449 					else {
1450 						/* -1 = kernel default */
1451 						in6p->in6p_hops = optval;
1452 						if ((in6p->inp_vflag &
1453 						     INP_IPV4) != 0)
1454 							in6p->inp_ip_ttl = optval;
1455 					}
1456 					break;
1457 #define OPTSET(bit) \
1458 do { \
1459 	if (optval) \
1460 		in6p->inp_flags |= (bit); \
1461 	else \
1462 		in6p->inp_flags &= ~(bit); \
1463 } while (/*CONSTCOND*/ 0)
1464 #define OPTSET2292(bit) \
1465 do { \
1466 	in6p->inp_flags |= IN6P_RFC2292; \
1467 	if (optval) \
1468 		in6p->inp_flags |= (bit); \
1469 	else \
1470 		in6p->inp_flags &= ~(bit); \
1471 } while (/*CONSTCOND*/ 0)
1472 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1473 
1474 				case IPV6_RECVPKTINFO:
1475 					/* cannot mix with RFC2292 */
1476 					if (OPTBIT(IN6P_RFC2292)) {
1477 						error = EINVAL;
1478 						break;
1479 					}
1480 					OPTSET(IN6P_PKTINFO);
1481 					break;
1482 
1483 				case IPV6_HOPLIMIT:
1484 				{
1485 					struct ip6_pktopts **optp;
1486 
1487 					/* cannot mix with RFC2292 */
1488 					if (OPTBIT(IN6P_RFC2292)) {
1489 						error = EINVAL;
1490 						break;
1491 					}
1492 					optp = &in6p->in6p_outputopts;
1493 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1494 					    (u_char *)&optval, sizeof(optval),
1495 					    optp, (td != NULL) ? td->td_ucred :
1496 					    NULL, uproto);
1497 					break;
1498 				}
1499 
1500 				case IPV6_RECVHOPLIMIT:
1501 					/* cannot mix with RFC2292 */
1502 					if (OPTBIT(IN6P_RFC2292)) {
1503 						error = EINVAL;
1504 						break;
1505 					}
1506 					OPTSET(IN6P_HOPLIMIT);
1507 					break;
1508 
1509 				case IPV6_RECVHOPOPTS:
1510 					/* cannot mix with RFC2292 */
1511 					if (OPTBIT(IN6P_RFC2292)) {
1512 						error = EINVAL;
1513 						break;
1514 					}
1515 					OPTSET(IN6P_HOPOPTS);
1516 					break;
1517 
1518 				case IPV6_RECVDSTOPTS:
1519 					/* cannot mix with RFC2292 */
1520 					if (OPTBIT(IN6P_RFC2292)) {
1521 						error = EINVAL;
1522 						break;
1523 					}
1524 					OPTSET(IN6P_DSTOPTS);
1525 					break;
1526 
1527 				case IPV6_RECVRTHDRDSTOPTS:
1528 					/* cannot mix with RFC2292 */
1529 					if (OPTBIT(IN6P_RFC2292)) {
1530 						error = EINVAL;
1531 						break;
1532 					}
1533 					OPTSET(IN6P_RTHDRDSTOPTS);
1534 					break;
1535 
1536 				case IPV6_RECVRTHDR:
1537 					/* cannot mix with RFC2292 */
1538 					if (OPTBIT(IN6P_RFC2292)) {
1539 						error = EINVAL;
1540 						break;
1541 					}
1542 					OPTSET(IN6P_RTHDR);
1543 					break;
1544 
1545 				case IPV6_FAITH:
1546 					OPTSET(INP_FAITH);
1547 					break;
1548 
1549 				case IPV6_RECVPATHMTU:
1550 					/*
1551 					 * We ignore this option for TCP
1552 					 * sockets.
1553 					 * (RFC3542 leaves this case
1554 					 * unspecified.)
1555 					 */
1556 					if (uproto != IPPROTO_TCP)
1557 						OPTSET(IN6P_MTU);
1558 					break;
1559 
1560 				case IPV6_V6ONLY:
1561 					/*
1562 					 * make setsockopt(IPV6_V6ONLY)
1563 					 * available only prior to bind(2).
1564 					 * see ipng mailing list, Jun 22 2001.
1565 					 */
1566 					if (in6p->inp_lport ||
1567 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1568 						error = EINVAL;
1569 						break;
1570 					}
1571 					OPTSET(IN6P_IPV6_V6ONLY);
1572 					if (optval)
1573 						in6p->inp_vflag &= ~INP_IPV4;
1574 					else
1575 						in6p->inp_vflag |= INP_IPV4;
1576 					break;
1577 				case IPV6_RECVTCLASS:
1578 					/* cannot mix with RFC2292 XXX */
1579 					if (OPTBIT(IN6P_RFC2292)) {
1580 						error = EINVAL;
1581 						break;
1582 					}
1583 					OPTSET(IN6P_TCLASS);
1584 					break;
1585 				case IPV6_AUTOFLOWLABEL:
1586 					OPTSET(IN6P_AUTOFLOWLABEL);
1587 					break;
1588 
1589 				}
1590 				break;
1591 
1592 			case IPV6_TCLASS:
1593 			case IPV6_DONTFRAG:
1594 			case IPV6_USE_MIN_MTU:
1595 			case IPV6_PREFER_TEMPADDR:
1596 				if (optlen != sizeof(optval)) {
1597 					error = EINVAL;
1598 					break;
1599 				}
1600 				error = sooptcopyin(sopt, &optval,
1601 					sizeof optval, sizeof optval);
1602 				if (error)
1603 					break;
1604 				{
1605 					struct ip6_pktopts **optp;
1606 					optp = &in6p->in6p_outputopts;
1607 					error = ip6_pcbopt(optname,
1608 					    (u_char *)&optval, sizeof(optval),
1609 					    optp, (td != NULL) ? td->td_ucred :
1610 					    NULL, uproto);
1611 					break;
1612 				}
1613 
1614 			case IPV6_2292PKTINFO:
1615 			case IPV6_2292HOPLIMIT:
1616 			case IPV6_2292HOPOPTS:
1617 			case IPV6_2292DSTOPTS:
1618 			case IPV6_2292RTHDR:
1619 				/* RFC 2292 */
1620 				if (optlen != sizeof(int)) {
1621 					error = EINVAL;
1622 					break;
1623 				}
1624 				error = sooptcopyin(sopt, &optval,
1625 					sizeof optval, sizeof optval);
1626 				if (error)
1627 					break;
1628 				switch (optname) {
1629 				case IPV6_2292PKTINFO:
1630 					OPTSET2292(IN6P_PKTINFO);
1631 					break;
1632 				case IPV6_2292HOPLIMIT:
1633 					OPTSET2292(IN6P_HOPLIMIT);
1634 					break;
1635 				case IPV6_2292HOPOPTS:
1636 					/*
1637 					 * Check super-user privilege.
1638 					 * See comments for IPV6_RECVHOPOPTS.
1639 					 */
1640 					if (td != NULL) {
1641 						error = priv_check(td,
1642 						    PRIV_NETINET_SETHDROPTS);
1643 						if (error)
1644 							return (error);
1645 					}
1646 					OPTSET2292(IN6P_HOPOPTS);
1647 					break;
1648 				case IPV6_2292DSTOPTS:
1649 					if (td != NULL) {
1650 						error = priv_check(td,
1651 						    PRIV_NETINET_SETHDROPTS);
1652 						if (error)
1653 							return (error);
1654 					}
1655 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1656 					break;
1657 				case IPV6_2292RTHDR:
1658 					OPTSET2292(IN6P_RTHDR);
1659 					break;
1660 				}
1661 				break;
1662 			case IPV6_PKTINFO:
1663 			case IPV6_HOPOPTS:
1664 			case IPV6_RTHDR:
1665 			case IPV6_DSTOPTS:
1666 			case IPV6_RTHDRDSTOPTS:
1667 			case IPV6_NEXTHOP:
1668 			{
1669 				/* new advanced API (RFC3542) */
1670 				u_char *optbuf;
1671 				u_char optbuf_storage[MCLBYTES];
1672 				int optlen;
1673 				struct ip6_pktopts **optp;
1674 
1675 				/* cannot mix with RFC2292 */
1676 				if (OPTBIT(IN6P_RFC2292)) {
1677 					error = EINVAL;
1678 					break;
1679 				}
1680 
1681 				/*
1682 				 * We only ensure valsize is not too large
1683 				 * here.  Further validation will be done
1684 				 * later.
1685 				 */
1686 				error = sooptcopyin(sopt, optbuf_storage,
1687 				    sizeof(optbuf_storage), 0);
1688 				if (error)
1689 					break;
1690 				optlen = sopt->sopt_valsize;
1691 				optbuf = optbuf_storage;
1692 				optp = &in6p->in6p_outputopts;
1693 				error = ip6_pcbopt(optname, optbuf, optlen,
1694 				    optp, (td != NULL) ? td->td_ucred : NULL,
1695 				    uproto);
1696 				break;
1697 			}
1698 #undef OPTSET
1699 
1700 			case IPV6_MULTICAST_IF:
1701 			case IPV6_MULTICAST_HOPS:
1702 			case IPV6_MULTICAST_LOOP:
1703 			case IPV6_JOIN_GROUP:
1704 			case IPV6_LEAVE_GROUP:
1705 			    {
1706 				if (sopt->sopt_valsize > MLEN) {
1707 					error = EMSGSIZE;
1708 					break;
1709 				}
1710 				/* XXX */
1711 			    }
1712 			    /* FALLTHROUGH */
1713 			    {
1714 				struct mbuf *m;
1715 
1716 				if (sopt->sopt_valsize > MCLBYTES) {
1717 					error = EMSGSIZE;
1718 					break;
1719 				}
1720 				/* XXX */
1721 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1722 				if (m == 0) {
1723 					error = ENOBUFS;
1724 					break;
1725 				}
1726 				if (sopt->sopt_valsize > MLEN) {
1727 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1728 					if ((m->m_flags & M_EXT) == 0) {
1729 						m_free(m);
1730 						error = ENOBUFS;
1731 						break;
1732 					}
1733 				}
1734 				m->m_len = sopt->sopt_valsize;
1735 				error = sooptcopyin(sopt, mtod(m, char *),
1736 						    m->m_len, m->m_len);
1737 				if (error) {
1738 					(void)m_free(m);
1739 					break;
1740 				}
1741 				error =	ip6_setmoptions(sopt->sopt_name,
1742 							&in6p->in6p_moptions,
1743 							m);
1744 				(void)m_free(m);
1745 			    }
1746 				break;
1747 
1748 			case IPV6_PORTRANGE:
1749 				error = sooptcopyin(sopt, &optval,
1750 				    sizeof optval, sizeof optval);
1751 				if (error)
1752 					break;
1753 
1754 				switch (optval) {
1755 				case IPV6_PORTRANGE_DEFAULT:
1756 					in6p->inp_flags &= ~(INP_LOWPORT);
1757 					in6p->inp_flags &= ~(INP_HIGHPORT);
1758 					break;
1759 
1760 				case IPV6_PORTRANGE_HIGH:
1761 					in6p->inp_flags &= ~(INP_LOWPORT);
1762 					in6p->inp_flags |= INP_HIGHPORT;
1763 					break;
1764 
1765 				case IPV6_PORTRANGE_LOW:
1766 					in6p->inp_flags &= ~(INP_HIGHPORT);
1767 					in6p->inp_flags |= INP_LOWPORT;
1768 					break;
1769 
1770 				default:
1771 					error = EINVAL;
1772 					break;
1773 				}
1774 				break;
1775 
1776 #ifdef IPSEC
1777 			case IPV6_IPSEC_POLICY:
1778 			{
1779 				caddr_t req;
1780 				struct mbuf *m;
1781 
1782 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1783 					break;
1784 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1785 					break;
1786 				req = mtod(m, caddr_t);
1787 				error = ipsec_set_policy(in6p, optname, req,
1788 				    m->m_len, (sopt->sopt_td != NULL) ?
1789 				    sopt->sopt_td->td_ucred : NULL);
1790 				m_freem(m);
1791 				break;
1792 			}
1793 #endif /* IPSEC */
1794 
1795 			default:
1796 				error = ENOPROTOOPT;
1797 				break;
1798 			}
1799 			break;
1800 
1801 		case SOPT_GET:
1802 			switch (optname) {
1803 
1804 			case IPV6_2292PKTOPTIONS:
1805 #ifdef IPV6_PKTOPTIONS
1806 			case IPV6_PKTOPTIONS:
1807 #endif
1808 				/*
1809 				 * RFC3542 (effectively) deprecated the
1810 				 * semantics of the 2292-style pktoptions.
1811 				 * Since it was not reliable in nature (i.e.,
1812 				 * applications had to expect the lack of some
1813 				 * information after all), it would make sense
1814 				 * to simplify this part by always returning
1815 				 * empty data.
1816 				 */
1817 				sopt->sopt_valsize = 0;
1818 				break;
1819 
1820 			case IPV6_RECVHOPOPTS:
1821 			case IPV6_RECVDSTOPTS:
1822 			case IPV6_RECVRTHDRDSTOPTS:
1823 			case IPV6_UNICAST_HOPS:
1824 			case IPV6_RECVPKTINFO:
1825 			case IPV6_RECVHOPLIMIT:
1826 			case IPV6_RECVRTHDR:
1827 			case IPV6_RECVPATHMTU:
1828 
1829 			case IPV6_FAITH:
1830 			case IPV6_V6ONLY:
1831 			case IPV6_PORTRANGE:
1832 			case IPV6_RECVTCLASS:
1833 			case IPV6_AUTOFLOWLABEL:
1834 				switch (optname) {
1835 
1836 				case IPV6_RECVHOPOPTS:
1837 					optval = OPTBIT(IN6P_HOPOPTS);
1838 					break;
1839 
1840 				case IPV6_RECVDSTOPTS:
1841 					optval = OPTBIT(IN6P_DSTOPTS);
1842 					break;
1843 
1844 				case IPV6_RECVRTHDRDSTOPTS:
1845 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1846 					break;
1847 
1848 				case IPV6_UNICAST_HOPS:
1849 					optval = in6p->in6p_hops;
1850 					break;
1851 
1852 				case IPV6_RECVPKTINFO:
1853 					optval = OPTBIT(IN6P_PKTINFO);
1854 					break;
1855 
1856 				case IPV6_RECVHOPLIMIT:
1857 					optval = OPTBIT(IN6P_HOPLIMIT);
1858 					break;
1859 
1860 				case IPV6_RECVRTHDR:
1861 					optval = OPTBIT(IN6P_RTHDR);
1862 					break;
1863 
1864 				case IPV6_RECVPATHMTU:
1865 					optval = OPTBIT(IN6P_MTU);
1866 					break;
1867 
1868 				case IPV6_FAITH:
1869 					optval = OPTBIT(INP_FAITH);
1870 					break;
1871 
1872 				case IPV6_V6ONLY:
1873 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1874 					break;
1875 
1876 				case IPV6_PORTRANGE:
1877 				    {
1878 					int flags;
1879 					flags = in6p->inp_flags;
1880 					if (flags & INP_HIGHPORT)
1881 						optval = IPV6_PORTRANGE_HIGH;
1882 					else if (flags & INP_LOWPORT)
1883 						optval = IPV6_PORTRANGE_LOW;
1884 					else
1885 						optval = 0;
1886 					break;
1887 				    }
1888 				case IPV6_RECVTCLASS:
1889 					optval = OPTBIT(IN6P_TCLASS);
1890 					break;
1891 
1892 				case IPV6_AUTOFLOWLABEL:
1893 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1894 					break;
1895 				}
1896 				if (error)
1897 					break;
1898 				error = sooptcopyout(sopt, &optval,
1899 					sizeof optval);
1900 				break;
1901 
1902 			case IPV6_PATHMTU:
1903 			{
1904 				u_long pmtu = 0;
1905 				struct ip6_mtuinfo mtuinfo;
1906 				struct route_in6 sro;
1907 
1908 				bzero(&sro, sizeof(sro));
1909 
1910 				if (!(so->so_state & SS_ISCONNECTED))
1911 					return (ENOTCONN);
1912 				/*
1913 				 * XXX: we dot not consider the case of source
1914 				 * routing, or optional information to specify
1915 				 * the outgoing interface.
1916 				 */
1917 				error = ip6_getpmtu(&sro, NULL, NULL,
1918 				    &in6p->in6p_faddr, &pmtu, NULL);
1919 				if (sro.ro_rt)
1920 					RTFREE(sro.ro_rt);
1921 				if (error)
1922 					break;
1923 				if (pmtu > IPV6_MAXPACKET)
1924 					pmtu = IPV6_MAXPACKET;
1925 
1926 				bzero(&mtuinfo, sizeof(mtuinfo));
1927 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1928 				optdata = (void *)&mtuinfo;
1929 				optdatalen = sizeof(mtuinfo);
1930 				error = sooptcopyout(sopt, optdata,
1931 				    optdatalen);
1932 				break;
1933 			}
1934 
1935 			case IPV6_2292PKTINFO:
1936 			case IPV6_2292HOPLIMIT:
1937 			case IPV6_2292HOPOPTS:
1938 			case IPV6_2292RTHDR:
1939 			case IPV6_2292DSTOPTS:
1940 				switch (optname) {
1941 				case IPV6_2292PKTINFO:
1942 					optval = OPTBIT(IN6P_PKTINFO);
1943 					break;
1944 				case IPV6_2292HOPLIMIT:
1945 					optval = OPTBIT(IN6P_HOPLIMIT);
1946 					break;
1947 				case IPV6_2292HOPOPTS:
1948 					optval = OPTBIT(IN6P_HOPOPTS);
1949 					break;
1950 				case IPV6_2292RTHDR:
1951 					optval = OPTBIT(IN6P_RTHDR);
1952 					break;
1953 				case IPV6_2292DSTOPTS:
1954 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1955 					break;
1956 				}
1957 				error = sooptcopyout(sopt, &optval,
1958 				    sizeof optval);
1959 				break;
1960 			case IPV6_PKTINFO:
1961 			case IPV6_HOPOPTS:
1962 			case IPV6_RTHDR:
1963 			case IPV6_DSTOPTS:
1964 			case IPV6_RTHDRDSTOPTS:
1965 			case IPV6_NEXTHOP:
1966 			case IPV6_TCLASS:
1967 			case IPV6_DONTFRAG:
1968 			case IPV6_USE_MIN_MTU:
1969 			case IPV6_PREFER_TEMPADDR:
1970 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1971 				    optname, sopt);
1972 				break;
1973 
1974 			case IPV6_MULTICAST_IF:
1975 			case IPV6_MULTICAST_HOPS:
1976 			case IPV6_MULTICAST_LOOP:
1977 			case IPV6_JOIN_GROUP:
1978 			case IPV6_LEAVE_GROUP:
1979 			    {
1980 				struct mbuf *m;
1981 				error = ip6_getmoptions(sopt->sopt_name,
1982 				    in6p->in6p_moptions, &m);
1983 				if (error == 0)
1984 					error = sooptcopyout(sopt,
1985 					    mtod(m, char *), m->m_len);
1986 				m_freem(m);
1987 			    }
1988 				break;
1989 
1990 #ifdef IPSEC
1991 			case IPV6_IPSEC_POLICY:
1992 			  {
1993 				caddr_t req = NULL;
1994 				size_t len = 0;
1995 				struct mbuf *m = NULL;
1996 				struct mbuf **mp = &m;
1997 				size_t ovalsize = sopt->sopt_valsize;
1998 				caddr_t oval = (caddr_t)sopt->sopt_val;
1999 
2000 				error = soopt_getm(sopt, &m); /* XXX */
2001 				if (error != 0)
2002 					break;
2003 				error = soopt_mcopyin(sopt, m); /* XXX */
2004 				if (error != 0)
2005 					break;
2006 				sopt->sopt_valsize = ovalsize;
2007 				sopt->sopt_val = oval;
2008 				if (m) {
2009 					req = mtod(m, caddr_t);
2010 					len = m->m_len;
2011 				}
2012 				error = ipsec_get_policy(in6p, req, len, mp);
2013 				if (error == 0)
2014 					error = soopt_mcopyout(sopt, m); /* XXX */
2015 				if (error == 0 && m)
2016 					m_freem(m);
2017 				break;
2018 			  }
2019 #endif /* IPSEC */
2020 
2021 			default:
2022 				error = ENOPROTOOPT;
2023 				break;
2024 			}
2025 			break;
2026 		}
2027 	} else {		/* level != IPPROTO_IPV6 */
2028 		error = EINVAL;
2029 	}
2030 	return (error);
2031 }
2032 
2033 int
2034 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2035 {
2036 	int error = 0, optval, optlen;
2037 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2038 	struct inpcb *in6p = sotoinpcb(so);
2039 	int level, op, optname;
2040 
2041 	level = sopt->sopt_level;
2042 	op = sopt->sopt_dir;
2043 	optname = sopt->sopt_name;
2044 	optlen = sopt->sopt_valsize;
2045 
2046 	if (level != IPPROTO_IPV6) {
2047 		return (EINVAL);
2048 	}
2049 
2050 	switch (optname) {
2051 	case IPV6_CHECKSUM:
2052 		/*
2053 		 * For ICMPv6 sockets, no modification allowed for checksum
2054 		 * offset, permit "no change" values to help existing apps.
2055 		 *
2056 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2057 		 * for an ICMPv6 socket will fail."
2058 		 * The current behavior does not meet RFC3542.
2059 		 */
2060 		switch (op) {
2061 		case SOPT_SET:
2062 			if (optlen != sizeof(int)) {
2063 				error = EINVAL;
2064 				break;
2065 			}
2066 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2067 					    sizeof(optval));
2068 			if (error)
2069 				break;
2070 			if ((optval % 2) != 0) {
2071 				/* the API assumes even offset values */
2072 				error = EINVAL;
2073 			} else if (so->so_proto->pr_protocol ==
2074 			    IPPROTO_ICMPV6) {
2075 				if (optval != icmp6off)
2076 					error = EINVAL;
2077 			} else
2078 				in6p->in6p_cksum = optval;
2079 			break;
2080 
2081 		case SOPT_GET:
2082 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2083 				optval = icmp6off;
2084 			else
2085 				optval = in6p->in6p_cksum;
2086 
2087 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2088 			break;
2089 
2090 		default:
2091 			error = EINVAL;
2092 			break;
2093 		}
2094 		break;
2095 
2096 	default:
2097 		error = ENOPROTOOPT;
2098 		break;
2099 	}
2100 
2101 	return (error);
2102 }
2103 
2104 /*
2105  * Set up IP6 options in pcb for insertion in output packets or
2106  * specifying behavior of outgoing packets.
2107  */
2108 static int
2109 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2110     struct socket *so, struct sockopt *sopt)
2111 {
2112 	struct ip6_pktopts *opt = *pktopt;
2113 	int error = 0;
2114 	struct thread *td = sopt->sopt_td;
2115 
2116 	/* turn off any old options. */
2117 	if (opt) {
2118 #ifdef DIAGNOSTIC
2119 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2120 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2121 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2122 			printf("ip6_pcbopts: all specified options are cleared.\n");
2123 #endif
2124 		ip6_clearpktopts(opt, -1);
2125 	} else
2126 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2127 	*pktopt = NULL;
2128 
2129 	if (!m || m->m_len == 0) {
2130 		/*
2131 		 * Only turning off any previous options, regardless of
2132 		 * whether the opt is just created or given.
2133 		 */
2134 		free(opt, M_IP6OPT);
2135 		return (0);
2136 	}
2137 
2138 	/*  set options specified by user. */
2139 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2140 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2141 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2142 		free(opt, M_IP6OPT);
2143 		return (error);
2144 	}
2145 	*pktopt = opt;
2146 	return (0);
2147 }
2148 
2149 /*
2150  * initialize ip6_pktopts.  beware that there are non-zero default values in
2151  * the struct.
2152  */
2153 void
2154 ip6_initpktopts(struct ip6_pktopts *opt)
2155 {
2156 
2157 	bzero(opt, sizeof(*opt));
2158 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2159 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2160 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2161 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2162 }
2163 
2164 static int
2165 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2166     struct ucred *cred, int uproto)
2167 {
2168 	struct ip6_pktopts *opt;
2169 
2170 	if (*pktopt == NULL) {
2171 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2172 		    M_WAITOK);
2173 		ip6_initpktopts(*pktopt);
2174 	}
2175 	opt = *pktopt;
2176 
2177 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2178 }
2179 
2180 static int
2181 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2182 {
2183 	void *optdata = NULL;
2184 	int optdatalen = 0;
2185 	struct ip6_ext *ip6e;
2186 	int error = 0;
2187 	struct in6_pktinfo null_pktinfo;
2188 	int deftclass = 0, on;
2189 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2190 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2191 
2192 	switch (optname) {
2193 	case IPV6_PKTINFO:
2194 		if (pktopt && pktopt->ip6po_pktinfo)
2195 			optdata = (void *)pktopt->ip6po_pktinfo;
2196 		else {
2197 			/* XXX: we don't have to do this every time... */
2198 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2199 			optdata = (void *)&null_pktinfo;
2200 		}
2201 		optdatalen = sizeof(struct in6_pktinfo);
2202 		break;
2203 	case IPV6_TCLASS:
2204 		if (pktopt && pktopt->ip6po_tclass >= 0)
2205 			optdata = (void *)&pktopt->ip6po_tclass;
2206 		else
2207 			optdata = (void *)&deftclass;
2208 		optdatalen = sizeof(int);
2209 		break;
2210 	case IPV6_HOPOPTS:
2211 		if (pktopt && pktopt->ip6po_hbh) {
2212 			optdata = (void *)pktopt->ip6po_hbh;
2213 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2214 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2215 		}
2216 		break;
2217 	case IPV6_RTHDR:
2218 		if (pktopt && pktopt->ip6po_rthdr) {
2219 			optdata = (void *)pktopt->ip6po_rthdr;
2220 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2221 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2222 		}
2223 		break;
2224 	case IPV6_RTHDRDSTOPTS:
2225 		if (pktopt && pktopt->ip6po_dest1) {
2226 			optdata = (void *)pktopt->ip6po_dest1;
2227 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2228 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2229 		}
2230 		break;
2231 	case IPV6_DSTOPTS:
2232 		if (pktopt && pktopt->ip6po_dest2) {
2233 			optdata = (void *)pktopt->ip6po_dest2;
2234 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2235 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2236 		}
2237 		break;
2238 	case IPV6_NEXTHOP:
2239 		if (pktopt && pktopt->ip6po_nexthop) {
2240 			optdata = (void *)pktopt->ip6po_nexthop;
2241 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2242 		}
2243 		break;
2244 	case IPV6_USE_MIN_MTU:
2245 		if (pktopt)
2246 			optdata = (void *)&pktopt->ip6po_minmtu;
2247 		else
2248 			optdata = (void *)&defminmtu;
2249 		optdatalen = sizeof(int);
2250 		break;
2251 	case IPV6_DONTFRAG:
2252 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2253 			on = 1;
2254 		else
2255 			on = 0;
2256 		optdata = (void *)&on;
2257 		optdatalen = sizeof(on);
2258 		break;
2259 	case IPV6_PREFER_TEMPADDR:
2260 		if (pktopt)
2261 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2262 		else
2263 			optdata = (void *)&defpreftemp;
2264 		optdatalen = sizeof(int);
2265 		break;
2266 	default:		/* should not happen */
2267 #ifdef DIAGNOSTIC
2268 		panic("ip6_getpcbopt: unexpected option\n");
2269 #endif
2270 		return (ENOPROTOOPT);
2271 	}
2272 
2273 	error = sooptcopyout(sopt, optdata, optdatalen);
2274 
2275 	return (error);
2276 }
2277 
2278 void
2279 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2280 {
2281 	if (pktopt == NULL)
2282 		return;
2283 
2284 	if (optname == -1 || optname == IPV6_PKTINFO) {
2285 		if (pktopt->ip6po_pktinfo)
2286 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2287 		pktopt->ip6po_pktinfo = NULL;
2288 	}
2289 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2290 		pktopt->ip6po_hlim = -1;
2291 	if (optname == -1 || optname == IPV6_TCLASS)
2292 		pktopt->ip6po_tclass = -1;
2293 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2294 		if (pktopt->ip6po_nextroute.ro_rt) {
2295 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2296 			pktopt->ip6po_nextroute.ro_rt = NULL;
2297 		}
2298 		if (pktopt->ip6po_nexthop)
2299 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2300 		pktopt->ip6po_nexthop = NULL;
2301 	}
2302 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2303 		if (pktopt->ip6po_hbh)
2304 			free(pktopt->ip6po_hbh, M_IP6OPT);
2305 		pktopt->ip6po_hbh = NULL;
2306 	}
2307 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2308 		if (pktopt->ip6po_dest1)
2309 			free(pktopt->ip6po_dest1, M_IP6OPT);
2310 		pktopt->ip6po_dest1 = NULL;
2311 	}
2312 	if (optname == -1 || optname == IPV6_RTHDR) {
2313 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2314 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2315 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2316 		if (pktopt->ip6po_route.ro_rt) {
2317 			RTFREE(pktopt->ip6po_route.ro_rt);
2318 			pktopt->ip6po_route.ro_rt = NULL;
2319 		}
2320 	}
2321 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2322 		if (pktopt->ip6po_dest2)
2323 			free(pktopt->ip6po_dest2, M_IP6OPT);
2324 		pktopt->ip6po_dest2 = NULL;
2325 	}
2326 }
2327 
2328 #define PKTOPT_EXTHDRCPY(type) \
2329 do {\
2330 	if (src->type) {\
2331 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2332 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2333 		if (dst->type == NULL && canwait == M_NOWAIT)\
2334 			goto bad;\
2335 		bcopy(src->type, dst->type, hlen);\
2336 	}\
2337 } while (/*CONSTCOND*/ 0)
2338 
2339 static int
2340 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2341 {
2342 	if (dst == NULL || src == NULL)  {
2343 		printf("ip6_clearpktopts: invalid argument\n");
2344 		return (EINVAL);
2345 	}
2346 
2347 	dst->ip6po_hlim = src->ip6po_hlim;
2348 	dst->ip6po_tclass = src->ip6po_tclass;
2349 	dst->ip6po_flags = src->ip6po_flags;
2350 	if (src->ip6po_pktinfo) {
2351 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2352 		    M_IP6OPT, canwait);
2353 		if (dst->ip6po_pktinfo == NULL)
2354 			goto bad;
2355 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2356 	}
2357 	if (src->ip6po_nexthop) {
2358 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2359 		    M_IP6OPT, canwait);
2360 		if (dst->ip6po_nexthop == NULL)
2361 			goto bad;
2362 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2363 		    src->ip6po_nexthop->sa_len);
2364 	}
2365 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2366 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2367 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2368 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2369 	return (0);
2370 
2371   bad:
2372 	ip6_clearpktopts(dst, -1);
2373 	return (ENOBUFS);
2374 }
2375 #undef PKTOPT_EXTHDRCPY
2376 
2377 struct ip6_pktopts *
2378 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2379 {
2380 	int error;
2381 	struct ip6_pktopts *dst;
2382 
2383 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2384 	if (dst == NULL)
2385 		return (NULL);
2386 	ip6_initpktopts(dst);
2387 
2388 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2389 		free(dst, M_IP6OPT);
2390 		return (NULL);
2391 	}
2392 
2393 	return (dst);
2394 }
2395 
2396 void
2397 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2398 {
2399 	if (pktopt == NULL)
2400 		return;
2401 
2402 	ip6_clearpktopts(pktopt, -1);
2403 
2404 	free(pktopt, M_IP6OPT);
2405 }
2406 
2407 /*
2408  * Set the IP6 multicast options in response to user setsockopt().
2409  */
2410 static int
2411 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2412 {
2413 	INIT_VNET_NET(curvnet);
2414 	INIT_VNET_INET6(curvnet);
2415 	int error = 0;
2416 	u_int loop, ifindex;
2417 	struct ipv6_mreq *mreq;
2418 	struct ifnet *ifp;
2419 	struct ip6_moptions *im6o = *im6op;
2420 	struct route_in6 ro;
2421 	struct in6_multi_mship *imm;
2422 
2423 	if (im6o == NULL) {
2424 		/*
2425 		 * No multicast option buffer attached to the pcb;
2426 		 * allocate one and initialize to default values.
2427 		 */
2428 		im6o = (struct ip6_moptions *)
2429 			malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2430 
2431 		if (im6o == NULL)
2432 			return (ENOBUFS);
2433 		*im6op = im6o;
2434 		im6o->im6o_multicast_ifp = NULL;
2435 		im6o->im6o_multicast_hlim = V_ip6_defmcasthlim;
2436 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2437 		LIST_INIT(&im6o->im6o_memberships);
2438 	}
2439 
2440 	switch (optname) {
2441 
2442 	case IPV6_MULTICAST_IF:
2443 		/*
2444 		 * Select the interface for outgoing multicast packets.
2445 		 */
2446 		if (m == NULL || m->m_len != sizeof(u_int)) {
2447 			error = EINVAL;
2448 			break;
2449 		}
2450 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2451 		if (ifindex < 0 || V_if_index < ifindex) {
2452 			error = ENXIO;	/* XXX EINVAL? */
2453 			break;
2454 		}
2455 		ifp = ifnet_byindex(ifindex);
2456 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2457 			error = EADDRNOTAVAIL;
2458 			break;
2459 		}
2460 		im6o->im6o_multicast_ifp = ifp;
2461 		break;
2462 
2463 	case IPV6_MULTICAST_HOPS:
2464 	    {
2465 		/*
2466 		 * Set the IP6 hoplimit for outgoing multicast packets.
2467 		 */
2468 		int optval;
2469 		if (m == NULL || m->m_len != sizeof(int)) {
2470 			error = EINVAL;
2471 			break;
2472 		}
2473 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2474 		if (optval < -1 || optval >= 256)
2475 			error = EINVAL;
2476 		else if (optval == -1)
2477 			im6o->im6o_multicast_hlim = V_ip6_defmcasthlim;
2478 		else
2479 			im6o->im6o_multicast_hlim = optval;
2480 		break;
2481 	    }
2482 
2483 	case IPV6_MULTICAST_LOOP:
2484 		/*
2485 		 * Set the loopback flag for outgoing multicast packets.
2486 		 * Must be zero or one.
2487 		 */
2488 		if (m == NULL || m->m_len != sizeof(u_int)) {
2489 			error = EINVAL;
2490 			break;
2491 		}
2492 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2493 		if (loop > 1) {
2494 			error = EINVAL;
2495 			break;
2496 		}
2497 		im6o->im6o_multicast_loop = loop;
2498 		break;
2499 
2500 	case IPV6_JOIN_GROUP:
2501 		/*
2502 		 * Add a multicast group membership.
2503 		 * Group must be a valid IP6 multicast address.
2504 		 */
2505 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2506 			error = EINVAL;
2507 			break;
2508 		}
2509 		mreq = mtod(m, struct ipv6_mreq *);
2510 
2511 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2512 			/*
2513 			 * We use the unspecified address to specify to accept
2514 			 * all multicast addresses. Only super user is allowed
2515 			 * to do this.
2516 			 */
2517 			/* XXX-BZ might need a better PRIV_NETINET_x for this */
2518 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
2519 			if (error)
2520 				break;
2521 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2522 			error = EINVAL;
2523 			break;
2524 		}
2525 
2526 		/*
2527 		 * If no interface was explicitly specified, choose an
2528 		 * appropriate one according to the given multicast address.
2529 		 */
2530 		if (mreq->ipv6mr_interface == 0) {
2531 			struct sockaddr_in6 *dst;
2532 
2533 			/*
2534 			 * Look up the routing table for the
2535 			 * address, and choose the outgoing interface.
2536 			 *   XXX: is it a good approach?
2537 			 */
2538 			ro.ro_rt = NULL;
2539 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2540 			bzero(dst, sizeof(*dst));
2541 			dst->sin6_family = AF_INET6;
2542 			dst->sin6_len = sizeof(*dst);
2543 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2544 			rtalloc((struct route *)&ro);
2545 			if (ro.ro_rt == NULL) {
2546 				error = EADDRNOTAVAIL;
2547 				break;
2548 			}
2549 			ifp = ro.ro_rt->rt_ifp;
2550 			RTFREE(ro.ro_rt);
2551 		} else {
2552 			/*
2553 			 * If the interface is specified, validate it.
2554 			 */
2555 			if (mreq->ipv6mr_interface < 0 ||
2556 			    V_if_index < mreq->ipv6mr_interface) {
2557 				error = ENXIO;	/* XXX EINVAL? */
2558 				break;
2559 			}
2560 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2561 			if (!ifp) {
2562 				error = ENXIO;	/* XXX EINVAL? */
2563 				break;
2564 			}
2565 		}
2566 
2567 		/*
2568 		 * See if we found an interface, and confirm that it
2569 		 * supports multicast
2570 		 */
2571 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2572 			error = EADDRNOTAVAIL;
2573 			break;
2574 		}
2575 
2576 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2577 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2578 			break;
2579 		}
2580 
2581 		/*
2582 		 * See if the membership already exists.
2583 		 */
2584 		for (imm = im6o->im6o_memberships.lh_first;
2585 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2586 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2587 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2588 					       &mreq->ipv6mr_multiaddr))
2589 				break;
2590 		if (imm != NULL) {
2591 			error = EADDRINUSE;
2592 			break;
2593 		}
2594 		/*
2595 		 * Everything looks good; add a new record to the multicast
2596 		 * address list for the given interface.
2597 		 */
2598 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2599 		if (imm == NULL)
2600 			break;
2601 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2602 		break;
2603 
2604 	case IPV6_LEAVE_GROUP:
2605 		/*
2606 		 * Drop a multicast group membership.
2607 		 * Group must be a valid IP6 multicast address.
2608 		 */
2609 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2610 			error = EINVAL;
2611 			break;
2612 		}
2613 		mreq = mtod(m, struct ipv6_mreq *);
2614 
2615 		/*
2616 		 * If an interface address was specified, get a pointer
2617 		 * to its ifnet structure.
2618 		 */
2619 		if (mreq->ipv6mr_interface < 0 ||
2620 		    V_if_index < mreq->ipv6mr_interface) {
2621 			error = ENXIO;	/* XXX EINVAL? */
2622 			break;
2623 		}
2624 		if (mreq->ipv6mr_interface == 0)
2625 			ifp = NULL;
2626 		else
2627 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2628 
2629 		/* Fill in the scope zone ID */
2630 		if (ifp) {
2631 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2632 				/* XXX: should not happen */
2633 				error = EADDRNOTAVAIL;
2634 				break;
2635 			}
2636 		} else if (mreq->ipv6mr_interface != 0) {
2637 			/*
2638 			 * This case happens when the (positive) index is in
2639 			 * the valid range, but the corresponding interface has
2640 			 * been detached dynamically (XXX).
2641 			 */
2642 			error = EADDRNOTAVAIL;
2643 			break;
2644 		} else {	/* ipv6mr_interface == 0 */
2645 			struct sockaddr_in6 sa6_mc;
2646 
2647 			/*
2648 			 * The API spec says as follows:
2649 			 *  If the interface index is specified as 0, the
2650 			 *  system may choose a multicast group membership to
2651 			 *  drop by matching the multicast address only.
2652 			 * On the other hand, we cannot disambiguate the scope
2653 			 * zone unless an interface is provided.  Thus, we
2654 			 * check if there's ambiguity with the default scope
2655 			 * zone as the last resort.
2656 			 */
2657 			bzero(&sa6_mc, sizeof(sa6_mc));
2658 			sa6_mc.sin6_family = AF_INET6;
2659 			sa6_mc.sin6_len = sizeof(sa6_mc);
2660 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2661 			error = sa6_embedscope(&sa6_mc, V_ip6_use_defzone);
2662 			if (error != 0)
2663 				break;
2664 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2665 		}
2666 
2667 		/*
2668 		 * Find the membership in the membership list.
2669 		 */
2670 		for (imm = im6o->im6o_memberships.lh_first;
2671 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2672 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2673 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2674 			    &mreq->ipv6mr_multiaddr))
2675 				break;
2676 		}
2677 		if (imm == NULL) {
2678 			/* Unable to resolve interface */
2679 			error = EADDRNOTAVAIL;
2680 			break;
2681 		}
2682 		/*
2683 		 * Give up the multicast address record to which the
2684 		 * membership points.
2685 		 */
2686 		LIST_REMOVE(imm, i6mm_chain);
2687 		in6_delmulti(imm->i6mm_maddr);
2688 		free(imm, M_IP6MADDR);
2689 		break;
2690 
2691 	default:
2692 		error = EOPNOTSUPP;
2693 		break;
2694 	}
2695 
2696 	/*
2697 	 * If all options have default values, no need to keep the mbuf.
2698 	 */
2699 	if (im6o->im6o_multicast_ifp == NULL &&
2700 	    im6o->im6o_multicast_hlim == V_ip6_defmcasthlim &&
2701 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2702 	    im6o->im6o_memberships.lh_first == NULL) {
2703 		free(*im6op, M_IP6MOPTS);
2704 		*im6op = NULL;
2705 	}
2706 
2707 	return (error);
2708 }
2709 
2710 /*
2711  * Return the IP6 multicast options in response to user getsockopt().
2712  */
2713 static int
2714 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2715 {
2716 	INIT_VNET_INET6(curvnet);
2717 	u_int *hlim, *loop, *ifindex;
2718 
2719 	*mp = m_get(M_WAIT, MT_HEADER);		/* XXX */
2720 
2721 	switch (optname) {
2722 
2723 	case IPV6_MULTICAST_IF:
2724 		ifindex = mtod(*mp, u_int *);
2725 		(*mp)->m_len = sizeof(u_int);
2726 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2727 			*ifindex = 0;
2728 		else
2729 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2730 		return (0);
2731 
2732 	case IPV6_MULTICAST_HOPS:
2733 		hlim = mtod(*mp, u_int *);
2734 		(*mp)->m_len = sizeof(u_int);
2735 		if (im6o == NULL)
2736 			*hlim = V_ip6_defmcasthlim;
2737 		else
2738 			*hlim = im6o->im6o_multicast_hlim;
2739 		return (0);
2740 
2741 	case IPV6_MULTICAST_LOOP:
2742 		loop = mtod(*mp, u_int *);
2743 		(*mp)->m_len = sizeof(u_int);
2744 		if (im6o == NULL)
2745 			*loop = V_ip6_defmcasthlim;
2746 		else
2747 			*loop = im6o->im6o_multicast_loop;
2748 		return (0);
2749 
2750 	default:
2751 		return (EOPNOTSUPP);
2752 	}
2753 }
2754 
2755 /*
2756  * Discard the IP6 multicast options.
2757  */
2758 void
2759 ip6_freemoptions(struct ip6_moptions *im6o)
2760 {
2761 	struct in6_multi_mship *imm;
2762 
2763 	if (im6o == NULL)
2764 		return;
2765 
2766 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2767 		LIST_REMOVE(imm, i6mm_chain);
2768 		if (imm->i6mm_maddr)
2769 			in6_delmulti(imm->i6mm_maddr);
2770 		free(imm, M_IP6MADDR);
2771 	}
2772 	free(im6o, M_IP6MOPTS);
2773 }
2774 
2775 /*
2776  * Set IPv6 outgoing packet options based on advanced API.
2777  */
2778 int
2779 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2780     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2781 {
2782 	struct cmsghdr *cm = 0;
2783 
2784 	if (control == NULL || opt == NULL)
2785 		return (EINVAL);
2786 
2787 	ip6_initpktopts(opt);
2788 	if (stickyopt) {
2789 		int error;
2790 
2791 		/*
2792 		 * If stickyopt is provided, make a local copy of the options
2793 		 * for this particular packet, then override them by ancillary
2794 		 * objects.
2795 		 * XXX: copypktopts() does not copy the cached route to a next
2796 		 * hop (if any).  This is not very good in terms of efficiency,
2797 		 * but we can allow this since this option should be rarely
2798 		 * used.
2799 		 */
2800 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2801 			return (error);
2802 	}
2803 
2804 	/*
2805 	 * XXX: Currently, we assume all the optional information is stored
2806 	 * in a single mbuf.
2807 	 */
2808 	if (control->m_next)
2809 		return (EINVAL);
2810 
2811 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2812 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2813 		int error;
2814 
2815 		if (control->m_len < CMSG_LEN(0))
2816 			return (EINVAL);
2817 
2818 		cm = mtod(control, struct cmsghdr *);
2819 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2820 			return (EINVAL);
2821 		if (cm->cmsg_level != IPPROTO_IPV6)
2822 			continue;
2823 
2824 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2825 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2826 		if (error)
2827 			return (error);
2828 	}
2829 
2830 	return (0);
2831 }
2832 
2833 /*
2834  * Set a particular packet option, as a sticky option or an ancillary data
2835  * item.  "len" can be 0 only when it's a sticky option.
2836  * We have 4 cases of combination of "sticky" and "cmsg":
2837  * "sticky=0, cmsg=0": impossible
2838  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2839  * "sticky=1, cmsg=0": RFC3542 socket option
2840  * "sticky=1, cmsg=1": RFC2292 socket option
2841  */
2842 static int
2843 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2844     struct ucred *cred, int sticky, int cmsg, int uproto)
2845 {
2846 	INIT_VNET_NET(curvnet);
2847 	INIT_VNET_INET6(curvnet);
2848 	int minmtupolicy, preftemp;
2849 	int error;
2850 
2851 	if (!sticky && !cmsg) {
2852 #ifdef DIAGNOSTIC
2853 		printf("ip6_setpktopt: impossible case\n");
2854 #endif
2855 		return (EINVAL);
2856 	}
2857 
2858 	/*
2859 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2860 	 * not be specified in the context of RFC3542.  Conversely,
2861 	 * RFC3542 types should not be specified in the context of RFC2292.
2862 	 */
2863 	if (!cmsg) {
2864 		switch (optname) {
2865 		case IPV6_2292PKTINFO:
2866 		case IPV6_2292HOPLIMIT:
2867 		case IPV6_2292NEXTHOP:
2868 		case IPV6_2292HOPOPTS:
2869 		case IPV6_2292DSTOPTS:
2870 		case IPV6_2292RTHDR:
2871 		case IPV6_2292PKTOPTIONS:
2872 			return (ENOPROTOOPT);
2873 		}
2874 	}
2875 	if (sticky && cmsg) {
2876 		switch (optname) {
2877 		case IPV6_PKTINFO:
2878 		case IPV6_HOPLIMIT:
2879 		case IPV6_NEXTHOP:
2880 		case IPV6_HOPOPTS:
2881 		case IPV6_DSTOPTS:
2882 		case IPV6_RTHDRDSTOPTS:
2883 		case IPV6_RTHDR:
2884 		case IPV6_USE_MIN_MTU:
2885 		case IPV6_DONTFRAG:
2886 		case IPV6_TCLASS:
2887 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2888 			return (ENOPROTOOPT);
2889 		}
2890 	}
2891 
2892 	switch (optname) {
2893 	case IPV6_2292PKTINFO:
2894 	case IPV6_PKTINFO:
2895 	{
2896 		struct ifnet *ifp = NULL;
2897 		struct in6_pktinfo *pktinfo;
2898 
2899 		if (len != sizeof(struct in6_pktinfo))
2900 			return (EINVAL);
2901 
2902 		pktinfo = (struct in6_pktinfo *)buf;
2903 
2904 		/*
2905 		 * An application can clear any sticky IPV6_PKTINFO option by
2906 		 * doing a "regular" setsockopt with ipi6_addr being
2907 		 * in6addr_any and ipi6_ifindex being zero.
2908 		 * [RFC 3542, Section 6]
2909 		 */
2910 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2911 		    pktinfo->ipi6_ifindex == 0 &&
2912 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2913 			ip6_clearpktopts(opt, optname);
2914 			break;
2915 		}
2916 
2917 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2918 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2919 			return (EINVAL);
2920 		}
2921 
2922 		/* validate the interface index if specified. */
2923 		if (pktinfo->ipi6_ifindex > V_if_index ||
2924 		    pktinfo->ipi6_ifindex < 0) {
2925 			 return (ENXIO);
2926 		}
2927 		if (pktinfo->ipi6_ifindex) {
2928 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2929 			if (ifp == NULL)
2930 				return (ENXIO);
2931 		}
2932 
2933 		/*
2934 		 * We store the address anyway, and let in6_selectsrc()
2935 		 * validate the specified address.  This is because ipi6_addr
2936 		 * may not have enough information about its scope zone, and
2937 		 * we may need additional information (such as outgoing
2938 		 * interface or the scope zone of a destination address) to
2939 		 * disambiguate the scope.
2940 		 * XXX: the delay of the validation may confuse the
2941 		 * application when it is used as a sticky option.
2942 		 */
2943 		if (opt->ip6po_pktinfo == NULL) {
2944 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2945 			    M_IP6OPT, M_NOWAIT);
2946 			if (opt->ip6po_pktinfo == NULL)
2947 				return (ENOBUFS);
2948 		}
2949 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2950 		break;
2951 	}
2952 
2953 	case IPV6_2292HOPLIMIT:
2954 	case IPV6_HOPLIMIT:
2955 	{
2956 		int *hlimp;
2957 
2958 		/*
2959 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2960 		 * to simplify the ordering among hoplimit options.
2961 		 */
2962 		if (optname == IPV6_HOPLIMIT && sticky)
2963 			return (ENOPROTOOPT);
2964 
2965 		if (len != sizeof(int))
2966 			return (EINVAL);
2967 		hlimp = (int *)buf;
2968 		if (*hlimp < -1 || *hlimp > 255)
2969 			return (EINVAL);
2970 
2971 		opt->ip6po_hlim = *hlimp;
2972 		break;
2973 	}
2974 
2975 	case IPV6_TCLASS:
2976 	{
2977 		int tclass;
2978 
2979 		if (len != sizeof(int))
2980 			return (EINVAL);
2981 		tclass = *(int *)buf;
2982 		if (tclass < -1 || tclass > 255)
2983 			return (EINVAL);
2984 
2985 		opt->ip6po_tclass = tclass;
2986 		break;
2987 	}
2988 
2989 	case IPV6_2292NEXTHOP:
2990 	case IPV6_NEXTHOP:
2991 		if (cred != NULL) {
2992 			error = priv_check_cred(cred,
2993 			    PRIV_NETINET_SETHDROPTS, 0);
2994 			if (error)
2995 				return (error);
2996 		}
2997 
2998 		if (len == 0) {	/* just remove the option */
2999 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3000 			break;
3001 		}
3002 
3003 		/* check if cmsg_len is large enough for sa_len */
3004 		if (len < sizeof(struct sockaddr) || len < *buf)
3005 			return (EINVAL);
3006 
3007 		switch (((struct sockaddr *)buf)->sa_family) {
3008 		case AF_INET6:
3009 		{
3010 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3011 			int error;
3012 
3013 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3014 				return (EINVAL);
3015 
3016 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3017 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3018 				return (EINVAL);
3019 			}
3020 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3021 			    != 0) {
3022 				return (error);
3023 			}
3024 			break;
3025 		}
3026 		case AF_LINK:	/* should eventually be supported */
3027 		default:
3028 			return (EAFNOSUPPORT);
3029 		}
3030 
3031 		/* turn off the previous option, then set the new option. */
3032 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3033 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3034 		if (opt->ip6po_nexthop == NULL)
3035 			return (ENOBUFS);
3036 		bcopy(buf, opt->ip6po_nexthop, *buf);
3037 		break;
3038 
3039 	case IPV6_2292HOPOPTS:
3040 	case IPV6_HOPOPTS:
3041 	{
3042 		struct ip6_hbh *hbh;
3043 		int hbhlen;
3044 
3045 		/*
3046 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3047 		 * options, since per-option restriction has too much
3048 		 * overhead.
3049 		 */
3050 		if (cred != NULL) {
3051 			error = priv_check_cred(cred,
3052 			    PRIV_NETINET_SETHDROPTS, 0);
3053 			if (error)
3054 				return (error);
3055 		}
3056 
3057 		if (len == 0) {
3058 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3059 			break;	/* just remove the option */
3060 		}
3061 
3062 		/* message length validation */
3063 		if (len < sizeof(struct ip6_hbh))
3064 			return (EINVAL);
3065 		hbh = (struct ip6_hbh *)buf;
3066 		hbhlen = (hbh->ip6h_len + 1) << 3;
3067 		if (len != hbhlen)
3068 			return (EINVAL);
3069 
3070 		/* turn off the previous option, then set the new option. */
3071 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3072 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3073 		if (opt->ip6po_hbh == NULL)
3074 			return (ENOBUFS);
3075 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3076 
3077 		break;
3078 	}
3079 
3080 	case IPV6_2292DSTOPTS:
3081 	case IPV6_DSTOPTS:
3082 	case IPV6_RTHDRDSTOPTS:
3083 	{
3084 		struct ip6_dest *dest, **newdest = NULL;
3085 		int destlen;
3086 
3087 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3088 			error = priv_check_cred(cred,
3089 			    PRIV_NETINET_SETHDROPTS, 0);
3090 			if (error)
3091 				return (error);
3092 		}
3093 
3094 		if (len == 0) {
3095 			ip6_clearpktopts(opt, optname);
3096 			break;	/* just remove the option */
3097 		}
3098 
3099 		/* message length validation */
3100 		if (len < sizeof(struct ip6_dest))
3101 			return (EINVAL);
3102 		dest = (struct ip6_dest *)buf;
3103 		destlen = (dest->ip6d_len + 1) << 3;
3104 		if (len != destlen)
3105 			return (EINVAL);
3106 
3107 		/*
3108 		 * Determine the position that the destination options header
3109 		 * should be inserted; before or after the routing header.
3110 		 */
3111 		switch (optname) {
3112 		case IPV6_2292DSTOPTS:
3113 			/*
3114 			 * The old advacned API is ambiguous on this point.
3115 			 * Our approach is to determine the position based
3116 			 * according to the existence of a routing header.
3117 			 * Note, however, that this depends on the order of the
3118 			 * extension headers in the ancillary data; the 1st
3119 			 * part of the destination options header must appear
3120 			 * before the routing header in the ancillary data,
3121 			 * too.
3122 			 * RFC3542 solved the ambiguity by introducing
3123 			 * separate ancillary data or option types.
3124 			 */
3125 			if (opt->ip6po_rthdr == NULL)
3126 				newdest = &opt->ip6po_dest1;
3127 			else
3128 				newdest = &opt->ip6po_dest2;
3129 			break;
3130 		case IPV6_RTHDRDSTOPTS:
3131 			newdest = &opt->ip6po_dest1;
3132 			break;
3133 		case IPV6_DSTOPTS:
3134 			newdest = &opt->ip6po_dest2;
3135 			break;
3136 		}
3137 
3138 		/* turn off the previous option, then set the new option. */
3139 		ip6_clearpktopts(opt, optname);
3140 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3141 		if (*newdest == NULL)
3142 			return (ENOBUFS);
3143 		bcopy(dest, *newdest, destlen);
3144 
3145 		break;
3146 	}
3147 
3148 	case IPV6_2292RTHDR:
3149 	case IPV6_RTHDR:
3150 	{
3151 		struct ip6_rthdr *rth;
3152 		int rthlen;
3153 
3154 		if (len == 0) {
3155 			ip6_clearpktopts(opt, IPV6_RTHDR);
3156 			break;	/* just remove the option */
3157 		}
3158 
3159 		/* message length validation */
3160 		if (len < sizeof(struct ip6_rthdr))
3161 			return (EINVAL);
3162 		rth = (struct ip6_rthdr *)buf;
3163 		rthlen = (rth->ip6r_len + 1) << 3;
3164 		if (len != rthlen)
3165 			return (EINVAL);
3166 
3167 		switch (rth->ip6r_type) {
3168 		case IPV6_RTHDR_TYPE_0:
3169 			if (rth->ip6r_len == 0)	/* must contain one addr */
3170 				return (EINVAL);
3171 			if (rth->ip6r_len % 2) /* length must be even */
3172 				return (EINVAL);
3173 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3174 				return (EINVAL);
3175 			break;
3176 		default:
3177 			return (EINVAL);	/* not supported */
3178 		}
3179 
3180 		/* turn off the previous option */
3181 		ip6_clearpktopts(opt, IPV6_RTHDR);
3182 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3183 		if (opt->ip6po_rthdr == NULL)
3184 			return (ENOBUFS);
3185 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3186 
3187 		break;
3188 	}
3189 
3190 	case IPV6_USE_MIN_MTU:
3191 		if (len != sizeof(int))
3192 			return (EINVAL);
3193 		minmtupolicy = *(int *)buf;
3194 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3195 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3196 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3197 			return (EINVAL);
3198 		}
3199 		opt->ip6po_minmtu = minmtupolicy;
3200 		break;
3201 
3202 	case IPV6_DONTFRAG:
3203 		if (len != sizeof(int))
3204 			return (EINVAL);
3205 
3206 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3207 			/*
3208 			 * we ignore this option for TCP sockets.
3209 			 * (RFC3542 leaves this case unspecified.)
3210 			 */
3211 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3212 		} else
3213 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3214 		break;
3215 
3216 	case IPV6_PREFER_TEMPADDR:
3217 		if (len != sizeof(int))
3218 			return (EINVAL);
3219 		preftemp = *(int *)buf;
3220 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3221 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3222 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3223 			return (EINVAL);
3224 		}
3225 		opt->ip6po_prefer_tempaddr = preftemp;
3226 		break;
3227 
3228 	default:
3229 		return (ENOPROTOOPT);
3230 	} /* end of switch */
3231 
3232 	return (0);
3233 }
3234 
3235 /*
3236  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3237  * packet to the input queue of a specified interface.  Note that this
3238  * calls the output routine of the loopback "driver", but with an interface
3239  * pointer that might NOT be &loif -- easier than replicating that code here.
3240  */
3241 void
3242 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3243 {
3244 	struct mbuf *copym;
3245 	struct ip6_hdr *ip6;
3246 
3247 	copym = m_copy(m, 0, M_COPYALL);
3248 	if (copym == NULL)
3249 		return;
3250 
3251 	/*
3252 	 * Make sure to deep-copy IPv6 header portion in case the data
3253 	 * is in an mbuf cluster, so that we can safely override the IPv6
3254 	 * header portion later.
3255 	 */
3256 	if ((copym->m_flags & M_EXT) != 0 ||
3257 	    copym->m_len < sizeof(struct ip6_hdr)) {
3258 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3259 		if (copym == NULL)
3260 			return;
3261 	}
3262 
3263 #ifdef DIAGNOSTIC
3264 	if (copym->m_len < sizeof(*ip6)) {
3265 		m_freem(copym);
3266 		return;
3267 	}
3268 #endif
3269 
3270 	ip6 = mtod(copym, struct ip6_hdr *);
3271 	/*
3272 	 * clear embedded scope identifiers if necessary.
3273 	 * in6_clearscope will touch the addresses only when necessary.
3274 	 */
3275 	in6_clearscope(&ip6->ip6_src);
3276 	in6_clearscope(&ip6->ip6_dst);
3277 
3278 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3279 }
3280 
3281 /*
3282  * Chop IPv6 header off from the payload.
3283  */
3284 static int
3285 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3286 {
3287 	struct mbuf *mh;
3288 	struct ip6_hdr *ip6;
3289 
3290 	ip6 = mtod(m, struct ip6_hdr *);
3291 	if (m->m_len > sizeof(*ip6)) {
3292 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3293 		if (mh == 0) {
3294 			m_freem(m);
3295 			return ENOBUFS;
3296 		}
3297 		M_MOVE_PKTHDR(mh, m);
3298 		MH_ALIGN(mh, sizeof(*ip6));
3299 		m->m_len -= sizeof(*ip6);
3300 		m->m_data += sizeof(*ip6);
3301 		mh->m_next = m;
3302 		m = mh;
3303 		m->m_len = sizeof(*ip6);
3304 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3305 	}
3306 	exthdrs->ip6e_ip6 = m;
3307 	return 0;
3308 }
3309 
3310 /*
3311  * Compute IPv6 extension header length.
3312  */
3313 int
3314 ip6_optlen(struct inpcb *in6p)
3315 {
3316 	int len;
3317 
3318 	if (!in6p->in6p_outputopts)
3319 		return 0;
3320 
3321 	len = 0;
3322 #define elen(x) \
3323     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3324 
3325 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3326 	if (in6p->in6p_outputopts->ip6po_rthdr)
3327 		/* dest1 is valid with rthdr only */
3328 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3329 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3330 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3331 	return len;
3332 #undef elen
3333 }
3334