xref: /freebsd/sys/netinet6/ip6_output.c (revision 9a14aa017b21c292740c00ee098195cd46642730)
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_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <net/if.h>
87 #include <net/netisr.h>
88 #include <net/route.h>
89 #include <net/pfil.h>
90 #include <net/vnet.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip_var.h>
95 #include <netinet6/in6_var.h>
96 #include <netinet/ip6.h>
97 #include <netinet/icmp6.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/tcp_var.h>
101 #include <netinet6/nd6.h>
102 
103 #ifdef IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #include <netinet6/ip6_ipsec.h>
108 #endif /* IPSEC */
109 #ifdef SCTP
110 #include <netinet/sctp.h>
111 #include <netinet/sctp_crc32.h>
112 #endif
113 
114 #include <netinet6/ip6protosw.h>
115 #include <netinet6/scope6_var.h>
116 
117 #ifdef FLOWTABLE
118 #include <net/flowtable.h>
119 #endif
120 
121 extern int in6_mcast_loop;
122 
123 struct ip6_exthdrs {
124 	struct mbuf *ip6e_ip6;
125 	struct mbuf *ip6e_hbh;
126 	struct mbuf *ip6e_dest1;
127 	struct mbuf *ip6e_rthdr;
128 	struct mbuf *ip6e_dest2;
129 };
130 
131 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
132 			   struct ucred *, int));
133 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
134 	struct socket *, struct sockopt *));
135 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
136 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
137 	struct ucred *, int, int, int));
138 
139 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
140 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
141 	struct ip6_frag **));
142 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
143 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
144 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
145 	struct ifnet *, struct in6_addr *, u_long *, int *));
146 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
147 
148 
149 /*
150  * Make an extension header from option data.  hp is the source, and
151  * mp is the destination.
152  */
153 #define MAKE_EXTHDR(hp, mp)						\
154     do {								\
155 	if (hp) {							\
156 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
157 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
158 		    ((eh)->ip6e_len + 1) << 3);				\
159 		if (error)						\
160 			goto freehdrs;					\
161 	}								\
162     } while (/*CONSTCOND*/ 0)
163 
164 /*
165  * Form a chain of extension headers.
166  * m is the extension header mbuf
167  * mp is the previous mbuf in the chain
168  * p is the next header
169  * i is the type of option.
170  */
171 #define MAKE_CHAIN(m, mp, p, i)\
172     do {\
173 	if (m) {\
174 		if (!hdrsplit) \
175 			panic("assumption failed: hdr not split"); \
176 		*mtod((m), u_char *) = *(p);\
177 		*(p) = (i);\
178 		p = mtod((m), u_char *);\
179 		(m)->m_next = (mp)->m_next;\
180 		(mp)->m_next = (m);\
181 		(mp) = (m);\
182 	}\
183     } while (/*CONSTCOND*/ 0)
184 
185 /*
186  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
187  * header (with pri, len, nxt, hlim, src, dst).
188  * This function may modify ver and hlim only.
189  * The mbuf chain containing the packet will be freed.
190  * The mbuf opt, if present, will not be freed.
191  *
192  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
193  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
194  * which is rt_rmx.rmx_mtu.
195  *
196  * ifpp - XXX: just for statistics
197  */
198 int
199 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
200     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
201     struct ifnet **ifpp, struct inpcb *inp)
202 {
203 	struct ip6_hdr *ip6, *mhip6;
204 	struct ifnet *ifp, *origifp;
205 	struct mbuf *m = m0;
206 	struct mbuf *mprev = NULL;
207 	int hlen, tlen, len, off;
208 	struct route_in6 ip6route;
209 	struct rtentry *rt = NULL;
210 	struct sockaddr_in6 *dst, src_sa, dst_sa;
211 	struct in6_addr odst;
212 	int error = 0;
213 	struct in6_ifaddr *ia = NULL;
214 	u_long mtu;
215 	int alwaysfrag, dontfrag;
216 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
217 	struct ip6_exthdrs exthdrs;
218 	struct in6_addr finaldst, src0, dst0;
219 	u_int32_t zone;
220 	struct route_in6 *ro_pmtu = NULL;
221 	int flevalid = 0;
222 	int hdrsplit = 0;
223 	int needipsec = 0;
224 #ifdef SCTP
225 	int sw_csum;
226 #endif
227 #ifdef IPSEC
228 	struct ipsec_output_state state;
229 	struct ip6_rthdr *rh = NULL;
230 	int needipsectun = 0;
231 	int segleft_org = 0;
232 	struct secpolicy *sp = NULL;
233 #endif /* IPSEC */
234 #ifdef IPFIREWALL_FORWARD
235 	struct m_tag *fwd_tag;
236 #endif
237 
238 	ip6 = mtod(m, struct ip6_hdr *);
239 	if (ip6 == NULL) {
240 		printf ("ip6 is NULL");
241 		goto bad;
242 	}
243 
244 	finaldst = ip6->ip6_dst;
245 	bzero(&exthdrs, sizeof(exthdrs));
246 	if (opt) {
247 		/* Hop-by-Hop options header */
248 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
249 		/* Destination options header(1st part) */
250 		if (opt->ip6po_rthdr) {
251 			/*
252 			 * Destination options header(1st part)
253 			 * This only makes sense with a routing header.
254 			 * See Section 9.2 of RFC 3542.
255 			 * Disabling this part just for MIP6 convenience is
256 			 * a bad idea.  We need to think carefully about a
257 			 * way to make the advanced API coexist with MIP6
258 			 * options, which might automatically be inserted in
259 			 * the kernel.
260 			 */
261 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
262 		}
263 		/* Routing header */
264 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
265 		/* Destination options header(2nd part) */
266 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
267 	}
268 
269 	/*
270 	 * IPSec checking which handles several cases.
271 	 * FAST IPSEC: We re-injected the packet.
272 	 */
273 #ifdef IPSEC
274 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
275 	{
276 	case 1:                 /* Bad packet */
277 		goto freehdrs;
278 	case -1:                /* Do IPSec */
279 		needipsec = 1;
280 	case 0:                 /* No IPSec */
281 	default:
282 		break;
283 	}
284 #endif /* IPSEC */
285 
286 	/*
287 	 * Calculate the total length of the extension header chain.
288 	 * Keep the length of the unfragmentable part for fragmentation.
289 	 */
290 	optlen = 0;
291 	if (exthdrs.ip6e_hbh)
292 		optlen += exthdrs.ip6e_hbh->m_len;
293 	if (exthdrs.ip6e_dest1)
294 		optlen += exthdrs.ip6e_dest1->m_len;
295 	if (exthdrs.ip6e_rthdr)
296 		optlen += exthdrs.ip6e_rthdr->m_len;
297 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
298 
299 	/* NOTE: we don't add AH/ESP length here. do that later. */
300 	if (exthdrs.ip6e_dest2)
301 		optlen += exthdrs.ip6e_dest2->m_len;
302 
303 	/*
304 	 * If we need IPsec, or there is at least one extension header,
305 	 * separate IP6 header from the payload.
306 	 */
307 	if ((needipsec || optlen) && !hdrsplit) {
308 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
309 			m = NULL;
310 			goto freehdrs;
311 		}
312 		m = exthdrs.ip6e_ip6;
313 		hdrsplit++;
314 	}
315 
316 	/* adjust pointer */
317 	ip6 = mtod(m, struct ip6_hdr *);
318 
319 	/* adjust mbuf packet header length */
320 	m->m_pkthdr.len += optlen;
321 	plen = m->m_pkthdr.len - sizeof(*ip6);
322 
323 	/* If this is a jumbo payload, insert a jumbo payload option. */
324 	if (plen > IPV6_MAXPACKET) {
325 		if (!hdrsplit) {
326 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
327 				m = NULL;
328 				goto freehdrs;
329 			}
330 			m = exthdrs.ip6e_ip6;
331 			hdrsplit++;
332 		}
333 		/* adjust pointer */
334 		ip6 = mtod(m, struct ip6_hdr *);
335 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
336 			goto freehdrs;
337 		ip6->ip6_plen = 0;
338 	} else
339 		ip6->ip6_plen = htons(plen);
340 
341 	/*
342 	 * Concatenate headers and fill in next header fields.
343 	 * Here we have, on "m"
344 	 *	IPv6 payload
345 	 * and we insert headers accordingly.  Finally, we should be getting:
346 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
347 	 *
348 	 * during the header composing process, "m" points to IPv6 header.
349 	 * "mprev" points to an extension header prior to esp.
350 	 */
351 	u_char *nexthdrp = &ip6->ip6_nxt;
352 	mprev = m;
353 
354 	/*
355 	 * we treat dest2 specially.  this makes IPsec processing
356 	 * much easier.  the goal here is to make mprev point the
357 	 * mbuf prior to dest2.
358 	 *
359 	 * result: IPv6 dest2 payload
360 	 * m and mprev will point to IPv6 header.
361 	 */
362 	if (exthdrs.ip6e_dest2) {
363 		if (!hdrsplit)
364 			panic("assumption failed: hdr not split");
365 		exthdrs.ip6e_dest2->m_next = m->m_next;
366 		m->m_next = exthdrs.ip6e_dest2;
367 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
368 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
369 	}
370 
371 	/*
372 	 * result: IPv6 hbh dest1 rthdr dest2 payload
373 	 * m will point to IPv6 header.  mprev will point to the
374 	 * extension header prior to dest2 (rthdr in the above case).
375 	 */
376 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
377 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
378 		   IPPROTO_DSTOPTS);
379 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
380 		   IPPROTO_ROUTING);
381 
382 #ifdef IPSEC
383 	if (!needipsec)
384 		goto skip_ipsec2;
385 
386 	/*
387 	 * pointers after IPsec headers are not valid any more.
388 	 * other pointers need a great care too.
389 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
390 	 */
391 	exthdrs.ip6e_dest2 = NULL;
392 
393 	if (exthdrs.ip6e_rthdr) {
394 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
395 		segleft_org = rh->ip6r_segleft;
396 		rh->ip6r_segleft = 0;
397 	}
398 
399 	bzero(&state, sizeof(state));
400 	state.m = m;
401 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
402 				    &needipsectun);
403 	m = state.m;
404 	if (error == EJUSTRETURN) {
405 		/*
406 		 * We had a SP with a level of 'use' and no SA. We
407 		 * will just continue to process the packet without
408 		 * IPsec processing.
409 		 */
410 		;
411 	} else if (error) {
412 		/* mbuf is already reclaimed in ipsec6_output_trans. */
413 		m = NULL;
414 		switch (error) {
415 		case EHOSTUNREACH:
416 		case ENETUNREACH:
417 		case EMSGSIZE:
418 		case ENOBUFS:
419 		case ENOMEM:
420 			break;
421 		default:
422 			printf("[%s:%d] (ipsec): error code %d\n",
423 			    __func__, __LINE__, error);
424 			/* FALLTHROUGH */
425 		case ENOENT:
426 			/* don't show these error codes to the user */
427 			error = 0;
428 			break;
429 		}
430 		goto bad;
431 	} else if (!needipsectun) {
432 		/*
433 		 * In the FAST IPSec case we have already
434 		 * re-injected the packet and it has been freed
435 		 * by the ipsec_done() function.  So, just clean
436 		 * up after ourselves.
437 		 */
438 		m = NULL;
439 		goto done;
440 	}
441 	if (exthdrs.ip6e_rthdr) {
442 		/* ah6_output doesn't modify mbuf chain */
443 		rh->ip6r_segleft = segleft_org;
444 	}
445 skip_ipsec2:;
446 #endif /* IPSEC */
447 
448 	/*
449 	 * If there is a routing header, discard the packet.
450 	 */
451 	if (exthdrs.ip6e_rthdr) {
452 		 error = EINVAL;
453 		 goto bad;
454 	}
455 
456 	/* Source address validation */
457 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
458 	    (flags & IPV6_UNSPECSRC) == 0) {
459 		error = EOPNOTSUPP;
460 		V_ip6stat.ip6s_badscope++;
461 		goto bad;
462 	}
463 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
464 		error = EOPNOTSUPP;
465 		V_ip6stat.ip6s_badscope++;
466 		goto bad;
467 	}
468 
469 	V_ip6stat.ip6s_localout++;
470 
471 	/*
472 	 * Route packet.
473 	 */
474 	if (ro == 0) {
475 		ro = &ip6route;
476 		bzero((caddr_t)ro, sizeof(*ro));
477 	}
478 	ro_pmtu = ro;
479 	if (opt && opt->ip6po_rthdr)
480 		ro = &opt->ip6po_route;
481 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
482 #ifdef FLOWTABLE
483 	if (ro == &ip6route) {
484 		struct flentry *fle;
485 
486 		/*
487 		 * The flow table returns route entries valid for up to 30
488 		 * seconds; we rely on the remainder of ip_output() taking no
489 		 * longer than that long for the stability of ro_rt.  The
490 		 * flow ID assignment must have happened before this point.
491 		 */
492 		if ((fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6)) != NULL) {
493 			flow_to_route_in6(fle, ro);
494 			if (ro->ro_rt != NULL && ro->ro_lle != NULL)
495 				flevalid = 1;
496 		}
497 	}
498 #endif
499 again:
500 	/*
501 	 * if specified, try to fill in the traffic class field.
502 	 * do not override if a non-zero value is already set.
503 	 * we check the diffserv field and the ecn field separately.
504 	 */
505 	if (opt && opt->ip6po_tclass >= 0) {
506 		int mask = 0;
507 
508 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
509 			mask |= 0xfc;
510 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
511 			mask |= 0x03;
512 		if (mask != 0)
513 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
514 	}
515 
516 	/* fill in or override the hop limit field, if necessary. */
517 	if (opt && opt->ip6po_hlim != -1)
518 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
519 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
520 		if (im6o != NULL)
521 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
522 		else
523 			ip6->ip6_hlim = V_ip6_defmcasthlim;
524 	}
525 
526 #ifdef IPSEC
527 	/*
528 	 * We may re-inject packets into the stack here.
529 	 */
530 	if (needipsec && needipsectun) {
531 		struct ipsec_output_state state;
532 
533 		/*
534 		 * All the extension headers will become inaccessible
535 		 * (since they can be encrypted).
536 		 * Don't panic, we need no more updates to extension headers
537 		 * on inner IPv6 packet (since they are now encapsulated).
538 		 *
539 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
540 		 */
541 		bzero(&exthdrs, sizeof(exthdrs));
542 		exthdrs.ip6e_ip6 = m;
543 
544 		bzero(&state, sizeof(state));
545 		state.m = m;
546 		state.ro = (struct route *)ro;
547 		state.dst = (struct sockaddr *)dst;
548 
549 		error = ipsec6_output_tunnel(&state, sp, flags);
550 
551 		m = state.m;
552 		ro = (struct route_in6 *)state.ro;
553 		dst = (struct sockaddr_in6 *)state.dst;
554 		if (error == EJUSTRETURN) {
555 			/*
556 			 * We had a SP with a level of 'use' and no SA. We
557 			 * will just continue to process the packet without
558 			 * IPsec processing.
559 			 */
560 			;
561 		} else if (error) {
562 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
563 			m0 = m = NULL;
564 			m = NULL;
565 			switch (error) {
566 			case EHOSTUNREACH:
567 			case ENETUNREACH:
568 			case EMSGSIZE:
569 			case ENOBUFS:
570 			case ENOMEM:
571 				break;
572 			default:
573 				printf("[%s:%d] (ipsec): error code %d\n",
574 				    __func__, __LINE__, error);
575 				/* FALLTHROUGH */
576 			case ENOENT:
577 				/* don't show these error codes to the user */
578 				error = 0;
579 				break;
580 			}
581 			goto bad;
582 		} else {
583 			/*
584 			 * In the FAST IPSec case we have already
585 			 * re-injected the packet and it has been freed
586 			 * by the ipsec_done() function.  So, just clean
587 			 * up after ourselves.
588 			 */
589 			m = NULL;
590 			goto done;
591 		}
592 
593 		exthdrs.ip6e_ip6 = m;
594 	}
595 #endif /* IPSEC */
596 
597 	/* adjust pointer */
598 	ip6 = mtod(m, struct ip6_hdr *);
599 
600 	bzero(&dst_sa, sizeof(dst_sa));
601 	dst_sa.sin6_family = AF_INET6;
602 	dst_sa.sin6_len = sizeof(dst_sa);
603 	dst_sa.sin6_addr = ip6->ip6_dst;
604 	if (flevalid) {
605 		rt = ro->ro_rt;
606 		ifp = ro->ro_rt->rt_ifp;
607 	} else if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
608 	    &ifp, &rt)) != 0) {
609 		switch (error) {
610 		case EHOSTUNREACH:
611 			V_ip6stat.ip6s_noroute++;
612 			break;
613 		case EADDRNOTAVAIL:
614 		default:
615 			break; /* XXX statistics? */
616 		}
617 		if (ifp != NULL)
618 			in6_ifstat_inc(ifp, ifs6_out_discard);
619 		goto bad;
620 	}
621 	if (rt == NULL) {
622 		/*
623 		 * If in6_selectroute() does not return a route entry,
624 		 * dst may not have been updated.
625 		 */
626 		*dst = dst_sa;	/* XXX */
627 	}
628 
629 	/*
630 	 * then rt (for unicast) and ifp must be non-NULL valid values.
631 	 */
632 	if ((flags & IPV6_FORWARDING) == 0) {
633 		/* XXX: the FORWARDING flag can be set for mrouting. */
634 		in6_ifstat_inc(ifp, ifs6_out_request);
635 	}
636 	if (rt != NULL) {
637 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
638 		rt->rt_use++;
639 	}
640 
641 
642 	/*
643 	 * The outgoing interface must be in the zone of source and
644 	 * destination addresses.
645 	 */
646 	origifp = ifp;
647 
648 	src0 = ip6->ip6_src;
649 	if (in6_setscope(&src0, origifp, &zone))
650 		goto badscope;
651 	bzero(&src_sa, sizeof(src_sa));
652 	src_sa.sin6_family = AF_INET6;
653 	src_sa.sin6_len = sizeof(src_sa);
654 	src_sa.sin6_addr = ip6->ip6_src;
655 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
656 		goto badscope;
657 
658 	dst0 = ip6->ip6_dst;
659 	if (in6_setscope(&dst0, origifp, &zone))
660 		goto badscope;
661 	/* re-initialize to be sure */
662 	bzero(&dst_sa, sizeof(dst_sa));
663 	dst_sa.sin6_family = AF_INET6;
664 	dst_sa.sin6_len = sizeof(dst_sa);
665 	dst_sa.sin6_addr = ip6->ip6_dst;
666 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
667 		goto badscope;
668 	}
669 
670 	/* We should use ia_ifp to support the case of
671 	 * sending packets to an address of our own.
672 	 */
673 	if (ia != NULL && ia->ia_ifp)
674 		ifp = ia->ia_ifp;
675 
676 	/* scope check is done. */
677 	goto routefound;
678 
679   badscope:
680 	V_ip6stat.ip6s_badscope++;
681 	in6_ifstat_inc(origifp, ifs6_out_discard);
682 	if (error == 0)
683 		error = EHOSTUNREACH; /* XXX */
684 	goto bad;
685 
686   routefound:
687 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
688 		if (opt && opt->ip6po_nextroute.ro_rt) {
689 			/*
690 			 * The nexthop is explicitly specified by the
691 			 * application.  We assume the next hop is an IPv6
692 			 * address.
693 			 */
694 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
695 		}
696 		else if ((rt->rt_flags & RTF_GATEWAY))
697 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
698 	}
699 
700 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
701 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
702 	} else {
703 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
704 		in6_ifstat_inc(ifp, ifs6_out_mcast);
705 		/*
706 		 * Confirm that the outgoing interface supports multicast.
707 		 */
708 		if (!(ifp->if_flags & IFF_MULTICAST)) {
709 			V_ip6stat.ip6s_noroute++;
710 			in6_ifstat_inc(ifp, ifs6_out_discard);
711 			error = ENETUNREACH;
712 			goto bad;
713 		}
714 		if ((im6o == NULL && in6_mcast_loop) ||
715 		    (im6o && im6o->im6o_multicast_loop)) {
716 			/*
717 			 * Loop back multicast datagram if not expressly
718 			 * forbidden to do so, even if we have not joined
719 			 * the address; protocols will filter it later,
720 			 * thus deferring a hash lookup and lock acquisition
721 			 * at the expense of an m_copym().
722 			 */
723 			ip6_mloopback(ifp, m, dst);
724 		} else {
725 			/*
726 			 * If we are acting as a multicast router, perform
727 			 * multicast forwarding as if the packet had just
728 			 * arrived on the interface to which we are about
729 			 * to send.  The multicast forwarding function
730 			 * recursively calls this function, using the
731 			 * IPV6_FORWARDING flag to prevent infinite recursion.
732 			 *
733 			 * Multicasts that are looped back by ip6_mloopback(),
734 			 * above, will be forwarded by the ip6_input() routine,
735 			 * if necessary.
736 			 */
737 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
738 				/*
739 				 * XXX: ip6_mforward expects that rcvif is NULL
740 				 * when it is called from the originating path.
741 				 * However, it is not always the case, since
742 				 * some versions of MGETHDR() does not
743 				 * initialize the field.
744 				 */
745 				m->m_pkthdr.rcvif = NULL;
746 				if (ip6_mforward(ip6, ifp, m) != 0) {
747 					m_freem(m);
748 					goto done;
749 				}
750 			}
751 		}
752 		/*
753 		 * Multicasts with a hoplimit of zero may be looped back,
754 		 * above, but must not be transmitted on a network.
755 		 * Also, multicasts addressed to the loopback interface
756 		 * are not sent -- the above call to ip6_mloopback() will
757 		 * loop back a copy if this host actually belongs to the
758 		 * destination group on the loopback interface.
759 		 */
760 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
761 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
762 			m_freem(m);
763 			goto done;
764 		}
765 	}
766 
767 	/*
768 	 * Fill the outgoing inteface to tell the upper layer
769 	 * to increment per-interface statistics.
770 	 */
771 	if (ifpp)
772 		*ifpp = ifp;
773 
774 	/* Determine path MTU. */
775 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
776 	    &alwaysfrag)) != 0)
777 		goto bad;
778 
779 	/*
780 	 * The caller of this function may specify to use the minimum MTU
781 	 * in some cases.
782 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
783 	 * setting.  The logic is a bit complicated; by default, unicast
784 	 * packets will follow path MTU while multicast packets will be sent at
785 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
786 	 * including unicast ones will be sent at the minimum MTU.  Multicast
787 	 * packets will always be sent at the minimum MTU unless
788 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
789 	 * See RFC 3542 for more details.
790 	 */
791 	if (mtu > IPV6_MMTU) {
792 		if ((flags & IPV6_MINMTU))
793 			mtu = IPV6_MMTU;
794 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
795 			mtu = IPV6_MMTU;
796 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
797 			 (opt == NULL ||
798 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
799 			mtu = IPV6_MMTU;
800 		}
801 	}
802 
803 	/*
804 	 * clear embedded scope identifiers if necessary.
805 	 * in6_clearscope will touch the addresses only when necessary.
806 	 */
807 	in6_clearscope(&ip6->ip6_src);
808 	in6_clearscope(&ip6->ip6_dst);
809 
810 	/*
811 	 * If the outgoing packet contains a hop-by-hop options header,
812 	 * it must be examined and processed even by the source node.
813 	 * (RFC 2460, section 4.)
814 	 */
815 	if (exthdrs.ip6e_hbh) {
816 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
817 		u_int32_t dummy; /* XXX unused */
818 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
819 
820 #ifdef DIAGNOSTIC
821 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
822 			panic("ip6e_hbh is not contiguous");
823 #endif
824 		/*
825 		 *  XXX: if we have to send an ICMPv6 error to the sender,
826 		 *       we need the M_LOOP flag since icmp6_error() expects
827 		 *       the IPv6 and the hop-by-hop options header are
828 		 *       contiguous unless the flag is set.
829 		 */
830 		m->m_flags |= M_LOOP;
831 		m->m_pkthdr.rcvif = ifp;
832 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
833 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
834 		    &dummy, &plen) < 0) {
835 			/* m was already freed at this point */
836 			error = EINVAL;/* better error? */
837 			goto done;
838 		}
839 		m->m_flags &= ~M_LOOP; /* XXX */
840 		m->m_pkthdr.rcvif = NULL;
841 	}
842 
843 	/* Jump over all PFIL processing if hooks are not active. */
844 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
845 		goto passout;
846 
847 	odst = ip6->ip6_dst;
848 	/* Run through list of hooks for output packets. */
849 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
850 	if (error != 0 || m == NULL)
851 		goto done;
852 	ip6 = mtod(m, struct ip6_hdr *);
853 
854 	/* See if destination IP address was changed by packet filter. */
855 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
856 		m->m_flags |= M_SKIP_FIREWALL;
857 		/* If destination is now ourself drop to ip6_input(). */
858 		if (in6_localip(&ip6->ip6_dst)) {
859 			m->m_flags |= M_FASTFWD_OURS;
860 			if (m->m_pkthdr.rcvif == NULL)
861 				m->m_pkthdr.rcvif = V_loif;
862 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
863 				m->m_pkthdr.csum_flags |=
864 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
865 				m->m_pkthdr.csum_data = 0xffff;
866 			}
867 			m->m_pkthdr.csum_flags |=
868 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
869 #ifdef SCTP
870 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
871 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
872 #endif
873 			error = netisr_queue(NETISR_IPV6, m);
874 			goto done;
875 		} else
876 			goto again;	/* Redo the routing table lookup. */
877 	}
878 
879 #ifdef IPFIREWALL_FORWARD
880 	/* See if local, if yes, send it to netisr. */
881 	if (m->m_flags & M_FASTFWD_OURS) {
882 		if (m->m_pkthdr.rcvif == NULL)
883 			m->m_pkthdr.rcvif = V_loif;
884 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
885 			m->m_pkthdr.csum_flags |=
886 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
887 			m->m_pkthdr.csum_data = 0xffff;
888 		}
889 #ifdef SCTP
890 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
891 		m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
892 #endif
893 		error = netisr_queue(NETISR_IPV6, m);
894 		goto done;
895 	}
896 	/* Or forward to some other address? */
897 	fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
898 	if (fwd_tag) {
899 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
900 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
901 		m->m_flags |= M_SKIP_FIREWALL;
902 		m_tag_delete(m, fwd_tag);
903 		goto again;
904 	}
905 #endif /* IPFIREWALL_FORWARD */
906 
907 passout:
908 	/*
909 	 * Send the packet to the outgoing interface.
910 	 * If necessary, do IPv6 fragmentation before sending.
911 	 *
912 	 * the logic here is rather complex:
913 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
914 	 * 1-a:	send as is if tlen <= path mtu
915 	 * 1-b:	fragment if tlen > path mtu
916 	 *
917 	 * 2: if user asks us not to fragment (dontfrag == 1)
918 	 * 2-a:	send as is if tlen <= interface mtu
919 	 * 2-b:	error if tlen > interface mtu
920 	 *
921 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
922 	 *	always fragment
923 	 *
924 	 * 4: if dontfrag == 1 && alwaysfrag == 1
925 	 *	error, as we cannot handle this conflicting request
926 	 */
927 #ifdef SCTP
928 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
929 	if (sw_csum & CSUM_SCTP) {
930 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
931 		sw_csum &= ~CSUM_SCTP;
932 	}
933 #endif
934 	tlen = m->m_pkthdr.len;
935 
936 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
937 		dontfrag = 1;
938 	else
939 		dontfrag = 0;
940 	if (dontfrag && alwaysfrag) {	/* case 4 */
941 		/* conflicting request - can't transmit */
942 		error = EMSGSIZE;
943 		goto bad;
944 	}
945 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
946 		/*
947 		 * Even if the DONTFRAG option is specified, we cannot send the
948 		 * packet when the data length is larger than the MTU of the
949 		 * outgoing interface.
950 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
951 		 * well as returning an error code (the latter is not described
952 		 * in the API spec.)
953 		 */
954 		u_int32_t mtu32;
955 		struct ip6ctlparam ip6cp;
956 
957 		mtu32 = (u_int32_t)mtu;
958 		bzero(&ip6cp, sizeof(ip6cp));
959 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
960 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
961 		    (void *)&ip6cp);
962 
963 		error = EMSGSIZE;
964 		goto bad;
965 	}
966 
967 	/*
968 	 * transmit packet without fragmentation
969 	 */
970 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
971 		struct in6_ifaddr *ia6;
972 
973 		ip6 = mtod(m, struct ip6_hdr *);
974 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
975 		if (ia6) {
976 			/* Record statistics for this interface address. */
977 			ia6->ia_ifa.if_opackets++;
978 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
979 			ifa_free(&ia6->ia_ifa);
980 		}
981 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
982 		goto done;
983 	}
984 
985 	/*
986 	 * try to fragment the packet.  case 1-b and 3
987 	 */
988 	if (mtu < IPV6_MMTU) {
989 		/* path MTU cannot be less than IPV6_MMTU */
990 		error = EMSGSIZE;
991 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
992 		goto bad;
993 	} else if (ip6->ip6_plen == 0) {
994 		/* jumbo payload cannot be fragmented */
995 		error = EMSGSIZE;
996 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
997 		goto bad;
998 	} else {
999 		struct mbuf **mnext, *m_frgpart;
1000 		struct ip6_frag *ip6f;
1001 		u_int32_t id = htonl(ip6_randomid());
1002 		u_char nextproto;
1003 
1004 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1005 
1006 		/*
1007 		 * Too large for the destination or interface;
1008 		 * fragment if possible.
1009 		 * Must be able to put at least 8 bytes per fragment.
1010 		 */
1011 		hlen = unfragpartlen;
1012 		if (mtu > IPV6_MAXPACKET)
1013 			mtu = IPV6_MAXPACKET;
1014 
1015 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1016 		if (len < 8) {
1017 			error = EMSGSIZE;
1018 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1019 			goto bad;
1020 		}
1021 
1022 		/*
1023 		 * Verify that we have any chance at all of being able to queue
1024 		 *      the packet or packet fragments
1025 		 */
1026 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1027 		    < tlen  /* - hlen */)) {
1028 			error = ENOBUFS;
1029 			V_ip6stat.ip6s_odropped++;
1030 			goto bad;
1031 		}
1032 
1033 		mnext = &m->m_nextpkt;
1034 
1035 		/*
1036 		 * Change the next header field of the last header in the
1037 		 * unfragmentable part.
1038 		 */
1039 		if (exthdrs.ip6e_rthdr) {
1040 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1041 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1042 		} else if (exthdrs.ip6e_dest1) {
1043 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1044 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1045 		} else if (exthdrs.ip6e_hbh) {
1046 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1047 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1048 		} else {
1049 			nextproto = ip6->ip6_nxt;
1050 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1051 		}
1052 
1053 		/*
1054 		 * Loop through length of segment after first fragment,
1055 		 * make new header and copy data of each part and link onto
1056 		 * chain.
1057 		 */
1058 		m0 = m;
1059 		for (off = hlen; off < tlen; off += len) {
1060 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1061 			if (!m) {
1062 				error = ENOBUFS;
1063 				V_ip6stat.ip6s_odropped++;
1064 				goto sendorfree;
1065 			}
1066 			m->m_pkthdr.rcvif = NULL;
1067 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1068 			*mnext = m;
1069 			mnext = &m->m_nextpkt;
1070 			m->m_data += max_linkhdr;
1071 			mhip6 = mtod(m, struct ip6_hdr *);
1072 			*mhip6 = *ip6;
1073 			m->m_len = sizeof(*mhip6);
1074 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1075 			if (error) {
1076 				V_ip6stat.ip6s_odropped++;
1077 				goto sendorfree;
1078 			}
1079 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1080 			if (off + len >= tlen)
1081 				len = tlen - off;
1082 			else
1083 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1084 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1085 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1086 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1087 				error = ENOBUFS;
1088 				V_ip6stat.ip6s_odropped++;
1089 				goto sendorfree;
1090 			}
1091 			m_cat(m, m_frgpart);
1092 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1093 			m->m_pkthdr.rcvif = NULL;
1094 			ip6f->ip6f_reserved = 0;
1095 			ip6f->ip6f_ident = id;
1096 			ip6f->ip6f_nxt = nextproto;
1097 			V_ip6stat.ip6s_ofragments++;
1098 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1099 		}
1100 
1101 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1102 	}
1103 
1104 	/*
1105 	 * Remove leading garbages.
1106 	 */
1107 sendorfree:
1108 	m = m0->m_nextpkt;
1109 	m0->m_nextpkt = 0;
1110 	m_freem(m0);
1111 	for (m0 = m; m; m = m0) {
1112 		m0 = m->m_nextpkt;
1113 		m->m_nextpkt = 0;
1114 		if (error == 0) {
1115 			/* Record statistics for this interface address. */
1116 			if (ia) {
1117 				ia->ia_ifa.if_opackets++;
1118 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1119 			}
1120 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1121 		} else
1122 			m_freem(m);
1123 	}
1124 
1125 	if (error == 0)
1126 		V_ip6stat.ip6s_fragmented++;
1127 
1128 done:
1129 	if (ro == &ip6route && ro->ro_rt && flevalid == 0) {
1130                 /* brace necessary for RTFREE */
1131 		RTFREE(ro->ro_rt);
1132 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt &&
1133 	    ((flevalid == 0) || (ro_pmtu != ro))) {
1134 		RTFREE(ro_pmtu->ro_rt);
1135 	}
1136 #ifdef IPSEC
1137 	if (sp != NULL)
1138 		KEY_FREESP(&sp);
1139 #endif
1140 
1141 	return (error);
1142 
1143 freehdrs:
1144 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1145 	m_freem(exthdrs.ip6e_dest1);
1146 	m_freem(exthdrs.ip6e_rthdr);
1147 	m_freem(exthdrs.ip6e_dest2);
1148 	/* FALLTHROUGH */
1149 bad:
1150 	if (m)
1151 		m_freem(m);
1152 	goto done;
1153 }
1154 
1155 static int
1156 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1157 {
1158 	struct mbuf *m;
1159 
1160 	if (hlen > MCLBYTES)
1161 		return (ENOBUFS); /* XXX */
1162 
1163 	MGET(m, M_DONTWAIT, MT_DATA);
1164 	if (!m)
1165 		return (ENOBUFS);
1166 
1167 	if (hlen > MLEN) {
1168 		MCLGET(m, M_DONTWAIT);
1169 		if ((m->m_flags & M_EXT) == 0) {
1170 			m_free(m);
1171 			return (ENOBUFS);
1172 		}
1173 	}
1174 	m->m_len = hlen;
1175 	if (hdr)
1176 		bcopy(hdr, mtod(m, caddr_t), hlen);
1177 
1178 	*mp = m;
1179 	return (0);
1180 }
1181 
1182 /*
1183  * Insert jumbo payload option.
1184  */
1185 static int
1186 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1187 {
1188 	struct mbuf *mopt;
1189 	u_char *optbuf;
1190 	u_int32_t v;
1191 
1192 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1193 
1194 	/*
1195 	 * If there is no hop-by-hop options header, allocate new one.
1196 	 * If there is one but it doesn't have enough space to store the
1197 	 * jumbo payload option, allocate a cluster to store the whole options.
1198 	 * Otherwise, use it to store the options.
1199 	 */
1200 	if (exthdrs->ip6e_hbh == 0) {
1201 		MGET(mopt, M_DONTWAIT, MT_DATA);
1202 		if (mopt == 0)
1203 			return (ENOBUFS);
1204 		mopt->m_len = JUMBOOPTLEN;
1205 		optbuf = mtod(mopt, u_char *);
1206 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1207 		exthdrs->ip6e_hbh = mopt;
1208 	} else {
1209 		struct ip6_hbh *hbh;
1210 
1211 		mopt = exthdrs->ip6e_hbh;
1212 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1213 			/*
1214 			 * XXX assumption:
1215 			 * - exthdrs->ip6e_hbh is not referenced from places
1216 			 *   other than exthdrs.
1217 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1218 			 */
1219 			int oldoptlen = mopt->m_len;
1220 			struct mbuf *n;
1221 
1222 			/*
1223 			 * XXX: give up if the whole (new) hbh header does
1224 			 * not fit even in an mbuf cluster.
1225 			 */
1226 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1227 				return (ENOBUFS);
1228 
1229 			/*
1230 			 * As a consequence, we must always prepare a cluster
1231 			 * at this point.
1232 			 */
1233 			MGET(n, M_DONTWAIT, MT_DATA);
1234 			if (n) {
1235 				MCLGET(n, M_DONTWAIT);
1236 				if ((n->m_flags & M_EXT) == 0) {
1237 					m_freem(n);
1238 					n = NULL;
1239 				}
1240 			}
1241 			if (!n)
1242 				return (ENOBUFS);
1243 			n->m_len = oldoptlen + JUMBOOPTLEN;
1244 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1245 			    oldoptlen);
1246 			optbuf = mtod(n, caddr_t) + oldoptlen;
1247 			m_freem(mopt);
1248 			mopt = exthdrs->ip6e_hbh = n;
1249 		} else {
1250 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1251 			mopt->m_len += JUMBOOPTLEN;
1252 		}
1253 		optbuf[0] = IP6OPT_PADN;
1254 		optbuf[1] = 1;
1255 
1256 		/*
1257 		 * Adjust the header length according to the pad and
1258 		 * the jumbo payload option.
1259 		 */
1260 		hbh = mtod(mopt, struct ip6_hbh *);
1261 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1262 	}
1263 
1264 	/* fill in the option. */
1265 	optbuf[2] = IP6OPT_JUMBO;
1266 	optbuf[3] = 4;
1267 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1268 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1269 
1270 	/* finally, adjust the packet header length */
1271 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1272 
1273 	return (0);
1274 #undef JUMBOOPTLEN
1275 }
1276 
1277 /*
1278  * Insert fragment header and copy unfragmentable header portions.
1279  */
1280 static int
1281 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1282     struct ip6_frag **frghdrp)
1283 {
1284 	struct mbuf *n, *mlast;
1285 
1286 	if (hlen > sizeof(struct ip6_hdr)) {
1287 		n = m_copym(m0, sizeof(struct ip6_hdr),
1288 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1289 		if (n == 0)
1290 			return (ENOBUFS);
1291 		m->m_next = n;
1292 	} else
1293 		n = m;
1294 
1295 	/* Search for the last mbuf of unfragmentable part. */
1296 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1297 		;
1298 
1299 	if ((mlast->m_flags & M_EXT) == 0 &&
1300 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1301 		/* use the trailing space of the last mbuf for the fragment hdr */
1302 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1303 		    mlast->m_len);
1304 		mlast->m_len += sizeof(struct ip6_frag);
1305 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1306 	} else {
1307 		/* allocate a new mbuf for the fragment header */
1308 		struct mbuf *mfrg;
1309 
1310 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1311 		if (mfrg == 0)
1312 			return (ENOBUFS);
1313 		mfrg->m_len = sizeof(struct ip6_frag);
1314 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1315 		mlast->m_next = mfrg;
1316 	}
1317 
1318 	return (0);
1319 }
1320 
1321 static int
1322 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1323     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1324     int *alwaysfragp)
1325 {
1326 	u_int32_t mtu = 0;
1327 	int alwaysfrag = 0;
1328 	int error = 0;
1329 
1330 	if (ro_pmtu != ro) {
1331 		/* The first hop and the final destination may differ. */
1332 		struct sockaddr_in6 *sa6_dst =
1333 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1334 		if (ro_pmtu->ro_rt &&
1335 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1336 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1337 			RTFREE(ro_pmtu->ro_rt);
1338 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1339 		}
1340 		if (ro_pmtu->ro_rt == NULL) {
1341 			bzero(sa6_dst, sizeof(*sa6_dst));
1342 			sa6_dst->sin6_family = AF_INET6;
1343 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1344 			sa6_dst->sin6_addr = *dst;
1345 
1346 			rtalloc((struct route *)ro_pmtu);
1347 		}
1348 	}
1349 	if (ro_pmtu->ro_rt) {
1350 		u_int32_t ifmtu;
1351 		struct in_conninfo inc;
1352 
1353 		bzero(&inc, sizeof(inc));
1354 		inc.inc_flags |= INC_ISIPV6;
1355 		inc.inc6_faddr = *dst;
1356 
1357 		if (ifp == NULL)
1358 			ifp = ro_pmtu->ro_rt->rt_ifp;
1359 		ifmtu = IN6_LINKMTU(ifp);
1360 		mtu = tcp_hc_getmtu(&inc);
1361 		if (mtu)
1362 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1363 		else
1364 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1365 		if (mtu == 0)
1366 			mtu = ifmtu;
1367 		else if (mtu < IPV6_MMTU) {
1368 			/*
1369 			 * RFC2460 section 5, last paragraph:
1370 			 * if we record ICMPv6 too big message with
1371 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1372 			 * or smaller, with framgent header attached.
1373 			 * (fragment header is needed regardless from the
1374 			 * packet size, for translators to identify packets)
1375 			 */
1376 			alwaysfrag = 1;
1377 			mtu = IPV6_MMTU;
1378 		} else if (mtu > ifmtu) {
1379 			/*
1380 			 * The MTU on the route is larger than the MTU on
1381 			 * the interface!  This shouldn't happen, unless the
1382 			 * MTU of the interface has been changed after the
1383 			 * interface was brought up.  Change the MTU in the
1384 			 * route to match the interface MTU (as long as the
1385 			 * field isn't locked).
1386 			 */
1387 			mtu = ifmtu;
1388 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1389 		}
1390 	} else if (ifp) {
1391 		mtu = IN6_LINKMTU(ifp);
1392 	} else
1393 		error = EHOSTUNREACH; /* XXX */
1394 
1395 	*mtup = mtu;
1396 	if (alwaysfragp)
1397 		*alwaysfragp = alwaysfrag;
1398 	return (error);
1399 }
1400 
1401 /*
1402  * IP6 socket option processing.
1403  */
1404 int
1405 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1406 {
1407 	int optdatalen, uproto;
1408 	void *optdata;
1409 	struct inpcb *in6p = sotoinpcb(so);
1410 	int error, optval;
1411 	int level, op, optname;
1412 	int optlen;
1413 	struct thread *td;
1414 
1415 	level = sopt->sopt_level;
1416 	op = sopt->sopt_dir;
1417 	optname = sopt->sopt_name;
1418 	optlen = sopt->sopt_valsize;
1419 	td = sopt->sopt_td;
1420 	error = 0;
1421 	optval = 0;
1422 	uproto = (int)so->so_proto->pr_protocol;
1423 
1424 	if (level != IPPROTO_IPV6) {
1425 		error = EINVAL;
1426 
1427 		if (sopt->sopt_level == SOL_SOCKET &&
1428 		    sopt->sopt_dir == SOPT_SET) {
1429 			switch (sopt->sopt_name) {
1430 			case SO_REUSEADDR:
1431 				INP_WLOCK(in6p);
1432 				if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1433 					if ((so->so_options &
1434 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1435 						in6p->inp_flags2 |= INP_REUSEPORT;
1436 					else
1437 						in6p->inp_flags2 &= ~INP_REUSEPORT;
1438 				}
1439 				INP_WUNLOCK(in6p);
1440 				error = 0;
1441 				break;
1442 			case SO_REUSEPORT:
1443 				INP_WLOCK(in6p);
1444 				if ((so->so_options & SO_REUSEPORT) != 0)
1445 					in6p->inp_flags2 |= INP_REUSEPORT;
1446 				else
1447 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1448 				INP_WUNLOCK(in6p);
1449 				error = 0;
1450 				break;
1451 			default:
1452 				break;
1453 			}
1454 		}
1455 	} else {		/* level == IPPROTO_IPV6 */
1456 		switch (op) {
1457 
1458 		case SOPT_SET:
1459 			switch (optname) {
1460 			case IPV6_2292PKTOPTIONS:
1461 #ifdef IPV6_PKTOPTIONS
1462 			case IPV6_PKTOPTIONS:
1463 #endif
1464 			{
1465 				struct mbuf *m;
1466 
1467 				error = soopt_getm(sopt, &m); /* XXX */
1468 				if (error != 0)
1469 					break;
1470 				error = soopt_mcopyin(sopt, m); /* XXX */
1471 				if (error != 0)
1472 					break;
1473 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1474 						    m, so, sopt);
1475 				m_freem(m); /* XXX */
1476 				break;
1477 			}
1478 
1479 			/*
1480 			 * Use of some Hop-by-Hop options or some
1481 			 * Destination options, might require special
1482 			 * privilege.  That is, normal applications
1483 			 * (without special privilege) might be forbidden
1484 			 * from setting certain options in outgoing packets,
1485 			 * and might never see certain options in received
1486 			 * packets. [RFC 2292 Section 6]
1487 			 * KAME specific note:
1488 			 *  KAME prevents non-privileged users from sending or
1489 			 *  receiving ANY hbh/dst options in order to avoid
1490 			 *  overhead of parsing options in the kernel.
1491 			 */
1492 			case IPV6_RECVHOPOPTS:
1493 			case IPV6_RECVDSTOPTS:
1494 			case IPV6_RECVRTHDRDSTOPTS:
1495 				if (td != NULL) {
1496 					error = priv_check(td,
1497 					    PRIV_NETINET_SETHDROPTS);
1498 					if (error)
1499 						break;
1500 				}
1501 				/* FALLTHROUGH */
1502 			case IPV6_UNICAST_HOPS:
1503 			case IPV6_HOPLIMIT:
1504 			case IPV6_FAITH:
1505 
1506 			case IPV6_RECVPKTINFO:
1507 			case IPV6_RECVHOPLIMIT:
1508 			case IPV6_RECVRTHDR:
1509 			case IPV6_RECVPATHMTU:
1510 			case IPV6_RECVTCLASS:
1511 			case IPV6_V6ONLY:
1512 			case IPV6_AUTOFLOWLABEL:
1513 			case IPV6_BINDANY:
1514 				if (optname == IPV6_BINDANY && td != NULL) {
1515 					error = priv_check(td,
1516 					    PRIV_NETINET_BINDANY);
1517 					if (error)
1518 						break;
1519 				}
1520 
1521 				if (optlen != sizeof(int)) {
1522 					error = EINVAL;
1523 					break;
1524 				}
1525 				error = sooptcopyin(sopt, &optval,
1526 					sizeof optval, sizeof optval);
1527 				if (error)
1528 					break;
1529 				switch (optname) {
1530 
1531 				case IPV6_UNICAST_HOPS:
1532 					if (optval < -1 || optval >= 256)
1533 						error = EINVAL;
1534 					else {
1535 						/* -1 = kernel default */
1536 						in6p->in6p_hops = optval;
1537 						if ((in6p->inp_vflag &
1538 						     INP_IPV4) != 0)
1539 							in6p->inp_ip_ttl = optval;
1540 					}
1541 					break;
1542 #define OPTSET(bit) \
1543 do { \
1544 	if (optval) \
1545 		in6p->inp_flags |= (bit); \
1546 	else \
1547 		in6p->inp_flags &= ~(bit); \
1548 } while (/*CONSTCOND*/ 0)
1549 #define OPTSET2292(bit) \
1550 do { \
1551 	in6p->inp_flags |= IN6P_RFC2292; \
1552 	if (optval) \
1553 		in6p->inp_flags |= (bit); \
1554 	else \
1555 		in6p->inp_flags &= ~(bit); \
1556 } while (/*CONSTCOND*/ 0)
1557 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1558 
1559 				case IPV6_RECVPKTINFO:
1560 					/* cannot mix with RFC2292 */
1561 					if (OPTBIT(IN6P_RFC2292)) {
1562 						error = EINVAL;
1563 						break;
1564 					}
1565 					OPTSET(IN6P_PKTINFO);
1566 					break;
1567 
1568 				case IPV6_HOPLIMIT:
1569 				{
1570 					struct ip6_pktopts **optp;
1571 
1572 					/* cannot mix with RFC2292 */
1573 					if (OPTBIT(IN6P_RFC2292)) {
1574 						error = EINVAL;
1575 						break;
1576 					}
1577 					optp = &in6p->in6p_outputopts;
1578 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1579 					    (u_char *)&optval, sizeof(optval),
1580 					    optp, (td != NULL) ? td->td_ucred :
1581 					    NULL, uproto);
1582 					break;
1583 				}
1584 
1585 				case IPV6_RECVHOPLIMIT:
1586 					/* cannot mix with RFC2292 */
1587 					if (OPTBIT(IN6P_RFC2292)) {
1588 						error = EINVAL;
1589 						break;
1590 					}
1591 					OPTSET(IN6P_HOPLIMIT);
1592 					break;
1593 
1594 				case IPV6_RECVHOPOPTS:
1595 					/* cannot mix with RFC2292 */
1596 					if (OPTBIT(IN6P_RFC2292)) {
1597 						error = EINVAL;
1598 						break;
1599 					}
1600 					OPTSET(IN6P_HOPOPTS);
1601 					break;
1602 
1603 				case IPV6_RECVDSTOPTS:
1604 					/* cannot mix with RFC2292 */
1605 					if (OPTBIT(IN6P_RFC2292)) {
1606 						error = EINVAL;
1607 						break;
1608 					}
1609 					OPTSET(IN6P_DSTOPTS);
1610 					break;
1611 
1612 				case IPV6_RECVRTHDRDSTOPTS:
1613 					/* cannot mix with RFC2292 */
1614 					if (OPTBIT(IN6P_RFC2292)) {
1615 						error = EINVAL;
1616 						break;
1617 					}
1618 					OPTSET(IN6P_RTHDRDSTOPTS);
1619 					break;
1620 
1621 				case IPV6_RECVRTHDR:
1622 					/* cannot mix with RFC2292 */
1623 					if (OPTBIT(IN6P_RFC2292)) {
1624 						error = EINVAL;
1625 						break;
1626 					}
1627 					OPTSET(IN6P_RTHDR);
1628 					break;
1629 
1630 				case IPV6_FAITH:
1631 					OPTSET(INP_FAITH);
1632 					break;
1633 
1634 				case IPV6_RECVPATHMTU:
1635 					/*
1636 					 * We ignore this option for TCP
1637 					 * sockets.
1638 					 * (RFC3542 leaves this case
1639 					 * unspecified.)
1640 					 */
1641 					if (uproto != IPPROTO_TCP)
1642 						OPTSET(IN6P_MTU);
1643 					break;
1644 
1645 				case IPV6_V6ONLY:
1646 					/*
1647 					 * make setsockopt(IPV6_V6ONLY)
1648 					 * available only prior to bind(2).
1649 					 * see ipng mailing list, Jun 22 2001.
1650 					 */
1651 					if (in6p->inp_lport ||
1652 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1653 						error = EINVAL;
1654 						break;
1655 					}
1656 					OPTSET(IN6P_IPV6_V6ONLY);
1657 					if (optval)
1658 						in6p->inp_vflag &= ~INP_IPV4;
1659 					else
1660 						in6p->inp_vflag |= INP_IPV4;
1661 					break;
1662 				case IPV6_RECVTCLASS:
1663 					/* cannot mix with RFC2292 XXX */
1664 					if (OPTBIT(IN6P_RFC2292)) {
1665 						error = EINVAL;
1666 						break;
1667 					}
1668 					OPTSET(IN6P_TCLASS);
1669 					break;
1670 				case IPV6_AUTOFLOWLABEL:
1671 					OPTSET(IN6P_AUTOFLOWLABEL);
1672 					break;
1673 
1674 				case IPV6_BINDANY:
1675 					OPTSET(INP_BINDANY);
1676 					break;
1677 				}
1678 				break;
1679 
1680 			case IPV6_TCLASS:
1681 			case IPV6_DONTFRAG:
1682 			case IPV6_USE_MIN_MTU:
1683 			case IPV6_PREFER_TEMPADDR:
1684 				if (optlen != sizeof(optval)) {
1685 					error = EINVAL;
1686 					break;
1687 				}
1688 				error = sooptcopyin(sopt, &optval,
1689 					sizeof optval, sizeof optval);
1690 				if (error)
1691 					break;
1692 				{
1693 					struct ip6_pktopts **optp;
1694 					optp = &in6p->in6p_outputopts;
1695 					error = ip6_pcbopt(optname,
1696 					    (u_char *)&optval, sizeof(optval),
1697 					    optp, (td != NULL) ? td->td_ucred :
1698 					    NULL, uproto);
1699 					break;
1700 				}
1701 
1702 			case IPV6_2292PKTINFO:
1703 			case IPV6_2292HOPLIMIT:
1704 			case IPV6_2292HOPOPTS:
1705 			case IPV6_2292DSTOPTS:
1706 			case IPV6_2292RTHDR:
1707 				/* RFC 2292 */
1708 				if (optlen != sizeof(int)) {
1709 					error = EINVAL;
1710 					break;
1711 				}
1712 				error = sooptcopyin(sopt, &optval,
1713 					sizeof optval, sizeof optval);
1714 				if (error)
1715 					break;
1716 				switch (optname) {
1717 				case IPV6_2292PKTINFO:
1718 					OPTSET2292(IN6P_PKTINFO);
1719 					break;
1720 				case IPV6_2292HOPLIMIT:
1721 					OPTSET2292(IN6P_HOPLIMIT);
1722 					break;
1723 				case IPV6_2292HOPOPTS:
1724 					/*
1725 					 * Check super-user privilege.
1726 					 * See comments for IPV6_RECVHOPOPTS.
1727 					 */
1728 					if (td != NULL) {
1729 						error = priv_check(td,
1730 						    PRIV_NETINET_SETHDROPTS);
1731 						if (error)
1732 							return (error);
1733 					}
1734 					OPTSET2292(IN6P_HOPOPTS);
1735 					break;
1736 				case IPV6_2292DSTOPTS:
1737 					if (td != NULL) {
1738 						error = priv_check(td,
1739 						    PRIV_NETINET_SETHDROPTS);
1740 						if (error)
1741 							return (error);
1742 					}
1743 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1744 					break;
1745 				case IPV6_2292RTHDR:
1746 					OPTSET2292(IN6P_RTHDR);
1747 					break;
1748 				}
1749 				break;
1750 			case IPV6_PKTINFO:
1751 			case IPV6_HOPOPTS:
1752 			case IPV6_RTHDR:
1753 			case IPV6_DSTOPTS:
1754 			case IPV6_RTHDRDSTOPTS:
1755 			case IPV6_NEXTHOP:
1756 			{
1757 				/* new advanced API (RFC3542) */
1758 				u_char *optbuf;
1759 				u_char optbuf_storage[MCLBYTES];
1760 				int optlen;
1761 				struct ip6_pktopts **optp;
1762 
1763 				/* cannot mix with RFC2292 */
1764 				if (OPTBIT(IN6P_RFC2292)) {
1765 					error = EINVAL;
1766 					break;
1767 				}
1768 
1769 				/*
1770 				 * We only ensure valsize is not too large
1771 				 * here.  Further validation will be done
1772 				 * later.
1773 				 */
1774 				error = sooptcopyin(sopt, optbuf_storage,
1775 				    sizeof(optbuf_storage), 0);
1776 				if (error)
1777 					break;
1778 				optlen = sopt->sopt_valsize;
1779 				optbuf = optbuf_storage;
1780 				optp = &in6p->in6p_outputopts;
1781 				error = ip6_pcbopt(optname, optbuf, optlen,
1782 				    optp, (td != NULL) ? td->td_ucred : NULL,
1783 				    uproto);
1784 				break;
1785 			}
1786 #undef OPTSET
1787 
1788 			case IPV6_MULTICAST_IF:
1789 			case IPV6_MULTICAST_HOPS:
1790 			case IPV6_MULTICAST_LOOP:
1791 			case IPV6_JOIN_GROUP:
1792 			case IPV6_LEAVE_GROUP:
1793 			case IPV6_MSFILTER:
1794 			case MCAST_BLOCK_SOURCE:
1795 			case MCAST_UNBLOCK_SOURCE:
1796 			case MCAST_JOIN_GROUP:
1797 			case MCAST_LEAVE_GROUP:
1798 			case MCAST_JOIN_SOURCE_GROUP:
1799 			case MCAST_LEAVE_SOURCE_GROUP:
1800 				error = ip6_setmoptions(in6p, sopt);
1801 				break;
1802 
1803 			case IPV6_PORTRANGE:
1804 				error = sooptcopyin(sopt, &optval,
1805 				    sizeof optval, sizeof optval);
1806 				if (error)
1807 					break;
1808 
1809 				switch (optval) {
1810 				case IPV6_PORTRANGE_DEFAULT:
1811 					in6p->inp_flags &= ~(INP_LOWPORT);
1812 					in6p->inp_flags &= ~(INP_HIGHPORT);
1813 					break;
1814 
1815 				case IPV6_PORTRANGE_HIGH:
1816 					in6p->inp_flags &= ~(INP_LOWPORT);
1817 					in6p->inp_flags |= INP_HIGHPORT;
1818 					break;
1819 
1820 				case IPV6_PORTRANGE_LOW:
1821 					in6p->inp_flags &= ~(INP_HIGHPORT);
1822 					in6p->inp_flags |= INP_LOWPORT;
1823 					break;
1824 
1825 				default:
1826 					error = EINVAL;
1827 					break;
1828 				}
1829 				break;
1830 
1831 #ifdef IPSEC
1832 			case IPV6_IPSEC_POLICY:
1833 			{
1834 				caddr_t req;
1835 				struct mbuf *m;
1836 
1837 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1838 					break;
1839 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1840 					break;
1841 				req = mtod(m, caddr_t);
1842 				error = ipsec_set_policy(in6p, optname, req,
1843 				    m->m_len, (sopt->sopt_td != NULL) ?
1844 				    sopt->sopt_td->td_ucred : NULL);
1845 				m_freem(m);
1846 				break;
1847 			}
1848 #endif /* IPSEC */
1849 
1850 			default:
1851 				error = ENOPROTOOPT;
1852 				break;
1853 			}
1854 			break;
1855 
1856 		case SOPT_GET:
1857 			switch (optname) {
1858 
1859 			case IPV6_2292PKTOPTIONS:
1860 #ifdef IPV6_PKTOPTIONS
1861 			case IPV6_PKTOPTIONS:
1862 #endif
1863 				/*
1864 				 * RFC3542 (effectively) deprecated the
1865 				 * semantics of the 2292-style pktoptions.
1866 				 * Since it was not reliable in nature (i.e.,
1867 				 * applications had to expect the lack of some
1868 				 * information after all), it would make sense
1869 				 * to simplify this part by always returning
1870 				 * empty data.
1871 				 */
1872 				sopt->sopt_valsize = 0;
1873 				break;
1874 
1875 			case IPV6_RECVHOPOPTS:
1876 			case IPV6_RECVDSTOPTS:
1877 			case IPV6_RECVRTHDRDSTOPTS:
1878 			case IPV6_UNICAST_HOPS:
1879 			case IPV6_RECVPKTINFO:
1880 			case IPV6_RECVHOPLIMIT:
1881 			case IPV6_RECVRTHDR:
1882 			case IPV6_RECVPATHMTU:
1883 
1884 			case IPV6_FAITH:
1885 			case IPV6_V6ONLY:
1886 			case IPV6_PORTRANGE:
1887 			case IPV6_RECVTCLASS:
1888 			case IPV6_AUTOFLOWLABEL:
1889 			case IPV6_BINDANY:
1890 				switch (optname) {
1891 
1892 				case IPV6_RECVHOPOPTS:
1893 					optval = OPTBIT(IN6P_HOPOPTS);
1894 					break;
1895 
1896 				case IPV6_RECVDSTOPTS:
1897 					optval = OPTBIT(IN6P_DSTOPTS);
1898 					break;
1899 
1900 				case IPV6_RECVRTHDRDSTOPTS:
1901 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1902 					break;
1903 
1904 				case IPV6_UNICAST_HOPS:
1905 					optval = in6p->in6p_hops;
1906 					break;
1907 
1908 				case IPV6_RECVPKTINFO:
1909 					optval = OPTBIT(IN6P_PKTINFO);
1910 					break;
1911 
1912 				case IPV6_RECVHOPLIMIT:
1913 					optval = OPTBIT(IN6P_HOPLIMIT);
1914 					break;
1915 
1916 				case IPV6_RECVRTHDR:
1917 					optval = OPTBIT(IN6P_RTHDR);
1918 					break;
1919 
1920 				case IPV6_RECVPATHMTU:
1921 					optval = OPTBIT(IN6P_MTU);
1922 					break;
1923 
1924 				case IPV6_FAITH:
1925 					optval = OPTBIT(INP_FAITH);
1926 					break;
1927 
1928 				case IPV6_V6ONLY:
1929 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1930 					break;
1931 
1932 				case IPV6_PORTRANGE:
1933 				    {
1934 					int flags;
1935 					flags = in6p->inp_flags;
1936 					if (flags & INP_HIGHPORT)
1937 						optval = IPV6_PORTRANGE_HIGH;
1938 					else if (flags & INP_LOWPORT)
1939 						optval = IPV6_PORTRANGE_LOW;
1940 					else
1941 						optval = 0;
1942 					break;
1943 				    }
1944 				case IPV6_RECVTCLASS:
1945 					optval = OPTBIT(IN6P_TCLASS);
1946 					break;
1947 
1948 				case IPV6_AUTOFLOWLABEL:
1949 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1950 					break;
1951 
1952 				case IPV6_BINDANY:
1953 					optval = OPTBIT(INP_BINDANY);
1954 					break;
1955 				}
1956 				if (error)
1957 					break;
1958 				error = sooptcopyout(sopt, &optval,
1959 					sizeof optval);
1960 				break;
1961 
1962 			case IPV6_PATHMTU:
1963 			{
1964 				u_long pmtu = 0;
1965 				struct ip6_mtuinfo mtuinfo;
1966 				struct route_in6 sro;
1967 
1968 				bzero(&sro, sizeof(sro));
1969 
1970 				if (!(so->so_state & SS_ISCONNECTED))
1971 					return (ENOTCONN);
1972 				/*
1973 				 * XXX: we dot not consider the case of source
1974 				 * routing, or optional information to specify
1975 				 * the outgoing interface.
1976 				 */
1977 				error = ip6_getpmtu(&sro, NULL, NULL,
1978 				    &in6p->in6p_faddr, &pmtu, NULL);
1979 				if (sro.ro_rt)
1980 					RTFREE(sro.ro_rt);
1981 				if (error)
1982 					break;
1983 				if (pmtu > IPV6_MAXPACKET)
1984 					pmtu = IPV6_MAXPACKET;
1985 
1986 				bzero(&mtuinfo, sizeof(mtuinfo));
1987 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1988 				optdata = (void *)&mtuinfo;
1989 				optdatalen = sizeof(mtuinfo);
1990 				error = sooptcopyout(sopt, optdata,
1991 				    optdatalen);
1992 				break;
1993 			}
1994 
1995 			case IPV6_2292PKTINFO:
1996 			case IPV6_2292HOPLIMIT:
1997 			case IPV6_2292HOPOPTS:
1998 			case IPV6_2292RTHDR:
1999 			case IPV6_2292DSTOPTS:
2000 				switch (optname) {
2001 				case IPV6_2292PKTINFO:
2002 					optval = OPTBIT(IN6P_PKTINFO);
2003 					break;
2004 				case IPV6_2292HOPLIMIT:
2005 					optval = OPTBIT(IN6P_HOPLIMIT);
2006 					break;
2007 				case IPV6_2292HOPOPTS:
2008 					optval = OPTBIT(IN6P_HOPOPTS);
2009 					break;
2010 				case IPV6_2292RTHDR:
2011 					optval = OPTBIT(IN6P_RTHDR);
2012 					break;
2013 				case IPV6_2292DSTOPTS:
2014 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2015 					break;
2016 				}
2017 				error = sooptcopyout(sopt, &optval,
2018 				    sizeof optval);
2019 				break;
2020 			case IPV6_PKTINFO:
2021 			case IPV6_HOPOPTS:
2022 			case IPV6_RTHDR:
2023 			case IPV6_DSTOPTS:
2024 			case IPV6_RTHDRDSTOPTS:
2025 			case IPV6_NEXTHOP:
2026 			case IPV6_TCLASS:
2027 			case IPV6_DONTFRAG:
2028 			case IPV6_USE_MIN_MTU:
2029 			case IPV6_PREFER_TEMPADDR:
2030 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2031 				    optname, sopt);
2032 				break;
2033 
2034 			case IPV6_MULTICAST_IF:
2035 			case IPV6_MULTICAST_HOPS:
2036 			case IPV6_MULTICAST_LOOP:
2037 			case IPV6_MSFILTER:
2038 				error = ip6_getmoptions(in6p, sopt);
2039 				break;
2040 
2041 #ifdef IPSEC
2042 			case IPV6_IPSEC_POLICY:
2043 			  {
2044 				caddr_t req = NULL;
2045 				size_t len = 0;
2046 				struct mbuf *m = NULL;
2047 				struct mbuf **mp = &m;
2048 				size_t ovalsize = sopt->sopt_valsize;
2049 				caddr_t oval = (caddr_t)sopt->sopt_val;
2050 
2051 				error = soopt_getm(sopt, &m); /* XXX */
2052 				if (error != 0)
2053 					break;
2054 				error = soopt_mcopyin(sopt, m); /* XXX */
2055 				if (error != 0)
2056 					break;
2057 				sopt->sopt_valsize = ovalsize;
2058 				sopt->sopt_val = oval;
2059 				if (m) {
2060 					req = mtod(m, caddr_t);
2061 					len = m->m_len;
2062 				}
2063 				error = ipsec_get_policy(in6p, req, len, mp);
2064 				if (error == 0)
2065 					error = soopt_mcopyout(sopt, m); /* XXX */
2066 				if (error == 0 && m)
2067 					m_freem(m);
2068 				break;
2069 			  }
2070 #endif /* IPSEC */
2071 
2072 			default:
2073 				error = ENOPROTOOPT;
2074 				break;
2075 			}
2076 			break;
2077 		}
2078 	}
2079 	return (error);
2080 }
2081 
2082 int
2083 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2084 {
2085 	int error = 0, optval, optlen;
2086 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2087 	struct inpcb *in6p = sotoinpcb(so);
2088 	int level, op, optname;
2089 
2090 	level = sopt->sopt_level;
2091 	op = sopt->sopt_dir;
2092 	optname = sopt->sopt_name;
2093 	optlen = sopt->sopt_valsize;
2094 
2095 	if (level != IPPROTO_IPV6) {
2096 		return (EINVAL);
2097 	}
2098 
2099 	switch (optname) {
2100 	case IPV6_CHECKSUM:
2101 		/*
2102 		 * For ICMPv6 sockets, no modification allowed for checksum
2103 		 * offset, permit "no change" values to help existing apps.
2104 		 *
2105 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2106 		 * for an ICMPv6 socket will fail."
2107 		 * The current behavior does not meet RFC3542.
2108 		 */
2109 		switch (op) {
2110 		case SOPT_SET:
2111 			if (optlen != sizeof(int)) {
2112 				error = EINVAL;
2113 				break;
2114 			}
2115 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2116 					    sizeof(optval));
2117 			if (error)
2118 				break;
2119 			if ((optval % 2) != 0) {
2120 				/* the API assumes even offset values */
2121 				error = EINVAL;
2122 			} else if (so->so_proto->pr_protocol ==
2123 			    IPPROTO_ICMPV6) {
2124 				if (optval != icmp6off)
2125 					error = EINVAL;
2126 			} else
2127 				in6p->in6p_cksum = optval;
2128 			break;
2129 
2130 		case SOPT_GET:
2131 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2132 				optval = icmp6off;
2133 			else
2134 				optval = in6p->in6p_cksum;
2135 
2136 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2137 			break;
2138 
2139 		default:
2140 			error = EINVAL;
2141 			break;
2142 		}
2143 		break;
2144 
2145 	default:
2146 		error = ENOPROTOOPT;
2147 		break;
2148 	}
2149 
2150 	return (error);
2151 }
2152 
2153 /*
2154  * Set up IP6 options in pcb for insertion in output packets or
2155  * specifying behavior of outgoing packets.
2156  */
2157 static int
2158 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2159     struct socket *so, struct sockopt *sopt)
2160 {
2161 	struct ip6_pktopts *opt = *pktopt;
2162 	int error = 0;
2163 	struct thread *td = sopt->sopt_td;
2164 
2165 	/* turn off any old options. */
2166 	if (opt) {
2167 #ifdef DIAGNOSTIC
2168 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2169 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2170 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2171 			printf("ip6_pcbopts: all specified options are cleared.\n");
2172 #endif
2173 		ip6_clearpktopts(opt, -1);
2174 	} else
2175 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2176 	*pktopt = NULL;
2177 
2178 	if (!m || m->m_len == 0) {
2179 		/*
2180 		 * Only turning off any previous options, regardless of
2181 		 * whether the opt is just created or given.
2182 		 */
2183 		free(opt, M_IP6OPT);
2184 		return (0);
2185 	}
2186 
2187 	/*  set options specified by user. */
2188 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2189 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2190 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2191 		free(opt, M_IP6OPT);
2192 		return (error);
2193 	}
2194 	*pktopt = opt;
2195 	return (0);
2196 }
2197 
2198 /*
2199  * initialize ip6_pktopts.  beware that there are non-zero default values in
2200  * the struct.
2201  */
2202 void
2203 ip6_initpktopts(struct ip6_pktopts *opt)
2204 {
2205 
2206 	bzero(opt, sizeof(*opt));
2207 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2208 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2209 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2210 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2211 }
2212 
2213 static int
2214 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2215     struct ucred *cred, int uproto)
2216 {
2217 	struct ip6_pktopts *opt;
2218 
2219 	if (*pktopt == NULL) {
2220 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2221 		    M_WAITOK);
2222 		ip6_initpktopts(*pktopt);
2223 	}
2224 	opt = *pktopt;
2225 
2226 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2227 }
2228 
2229 static int
2230 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2231 {
2232 	void *optdata = NULL;
2233 	int optdatalen = 0;
2234 	struct ip6_ext *ip6e;
2235 	int error = 0;
2236 	struct in6_pktinfo null_pktinfo;
2237 	int deftclass = 0, on;
2238 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2239 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2240 
2241 	switch (optname) {
2242 	case IPV6_PKTINFO:
2243 		if (pktopt && pktopt->ip6po_pktinfo)
2244 			optdata = (void *)pktopt->ip6po_pktinfo;
2245 		else {
2246 			/* XXX: we don't have to do this every time... */
2247 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2248 			optdata = (void *)&null_pktinfo;
2249 		}
2250 		optdatalen = sizeof(struct in6_pktinfo);
2251 		break;
2252 	case IPV6_TCLASS:
2253 		if (pktopt && pktopt->ip6po_tclass >= 0)
2254 			optdata = (void *)&pktopt->ip6po_tclass;
2255 		else
2256 			optdata = (void *)&deftclass;
2257 		optdatalen = sizeof(int);
2258 		break;
2259 	case IPV6_HOPOPTS:
2260 		if (pktopt && pktopt->ip6po_hbh) {
2261 			optdata = (void *)pktopt->ip6po_hbh;
2262 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2263 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2264 		}
2265 		break;
2266 	case IPV6_RTHDR:
2267 		if (pktopt && pktopt->ip6po_rthdr) {
2268 			optdata = (void *)pktopt->ip6po_rthdr;
2269 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2270 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2271 		}
2272 		break;
2273 	case IPV6_RTHDRDSTOPTS:
2274 		if (pktopt && pktopt->ip6po_dest1) {
2275 			optdata = (void *)pktopt->ip6po_dest1;
2276 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2277 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2278 		}
2279 		break;
2280 	case IPV6_DSTOPTS:
2281 		if (pktopt && pktopt->ip6po_dest2) {
2282 			optdata = (void *)pktopt->ip6po_dest2;
2283 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2284 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2285 		}
2286 		break;
2287 	case IPV6_NEXTHOP:
2288 		if (pktopt && pktopt->ip6po_nexthop) {
2289 			optdata = (void *)pktopt->ip6po_nexthop;
2290 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2291 		}
2292 		break;
2293 	case IPV6_USE_MIN_MTU:
2294 		if (pktopt)
2295 			optdata = (void *)&pktopt->ip6po_minmtu;
2296 		else
2297 			optdata = (void *)&defminmtu;
2298 		optdatalen = sizeof(int);
2299 		break;
2300 	case IPV6_DONTFRAG:
2301 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2302 			on = 1;
2303 		else
2304 			on = 0;
2305 		optdata = (void *)&on;
2306 		optdatalen = sizeof(on);
2307 		break;
2308 	case IPV6_PREFER_TEMPADDR:
2309 		if (pktopt)
2310 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2311 		else
2312 			optdata = (void *)&defpreftemp;
2313 		optdatalen = sizeof(int);
2314 		break;
2315 	default:		/* should not happen */
2316 #ifdef DIAGNOSTIC
2317 		panic("ip6_getpcbopt: unexpected option\n");
2318 #endif
2319 		return (ENOPROTOOPT);
2320 	}
2321 
2322 	error = sooptcopyout(sopt, optdata, optdatalen);
2323 
2324 	return (error);
2325 }
2326 
2327 void
2328 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2329 {
2330 	if (pktopt == NULL)
2331 		return;
2332 
2333 	if (optname == -1 || optname == IPV6_PKTINFO) {
2334 		if (pktopt->ip6po_pktinfo)
2335 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2336 		pktopt->ip6po_pktinfo = NULL;
2337 	}
2338 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2339 		pktopt->ip6po_hlim = -1;
2340 	if (optname == -1 || optname == IPV6_TCLASS)
2341 		pktopt->ip6po_tclass = -1;
2342 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2343 		if (pktopt->ip6po_nextroute.ro_rt) {
2344 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2345 			pktopt->ip6po_nextroute.ro_rt = NULL;
2346 		}
2347 		if (pktopt->ip6po_nexthop)
2348 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2349 		pktopt->ip6po_nexthop = NULL;
2350 	}
2351 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2352 		if (pktopt->ip6po_hbh)
2353 			free(pktopt->ip6po_hbh, M_IP6OPT);
2354 		pktopt->ip6po_hbh = NULL;
2355 	}
2356 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2357 		if (pktopt->ip6po_dest1)
2358 			free(pktopt->ip6po_dest1, M_IP6OPT);
2359 		pktopt->ip6po_dest1 = NULL;
2360 	}
2361 	if (optname == -1 || optname == IPV6_RTHDR) {
2362 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2363 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2364 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2365 		if (pktopt->ip6po_route.ro_rt) {
2366 			RTFREE(pktopt->ip6po_route.ro_rt);
2367 			pktopt->ip6po_route.ro_rt = NULL;
2368 		}
2369 	}
2370 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2371 		if (pktopt->ip6po_dest2)
2372 			free(pktopt->ip6po_dest2, M_IP6OPT);
2373 		pktopt->ip6po_dest2 = NULL;
2374 	}
2375 }
2376 
2377 #define PKTOPT_EXTHDRCPY(type) \
2378 do {\
2379 	if (src->type) {\
2380 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2381 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2382 		if (dst->type == NULL && canwait == M_NOWAIT)\
2383 			goto bad;\
2384 		bcopy(src->type, dst->type, hlen);\
2385 	}\
2386 } while (/*CONSTCOND*/ 0)
2387 
2388 static int
2389 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2390 {
2391 	if (dst == NULL || src == NULL)  {
2392 		printf("ip6_clearpktopts: invalid argument\n");
2393 		return (EINVAL);
2394 	}
2395 
2396 	dst->ip6po_hlim = src->ip6po_hlim;
2397 	dst->ip6po_tclass = src->ip6po_tclass;
2398 	dst->ip6po_flags = src->ip6po_flags;
2399 	dst->ip6po_minmtu = src->ip6po_minmtu;
2400 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2401 	if (src->ip6po_pktinfo) {
2402 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2403 		    M_IP6OPT, canwait);
2404 		if (dst->ip6po_pktinfo == NULL)
2405 			goto bad;
2406 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2407 	}
2408 	if (src->ip6po_nexthop) {
2409 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2410 		    M_IP6OPT, canwait);
2411 		if (dst->ip6po_nexthop == NULL)
2412 			goto bad;
2413 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2414 		    src->ip6po_nexthop->sa_len);
2415 	}
2416 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2417 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2418 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2419 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2420 	return (0);
2421 
2422   bad:
2423 	ip6_clearpktopts(dst, -1);
2424 	return (ENOBUFS);
2425 }
2426 #undef PKTOPT_EXTHDRCPY
2427 
2428 struct ip6_pktopts *
2429 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2430 {
2431 	int error;
2432 	struct ip6_pktopts *dst;
2433 
2434 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2435 	if (dst == NULL)
2436 		return (NULL);
2437 	ip6_initpktopts(dst);
2438 
2439 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2440 		free(dst, M_IP6OPT);
2441 		return (NULL);
2442 	}
2443 
2444 	return (dst);
2445 }
2446 
2447 void
2448 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2449 {
2450 	if (pktopt == NULL)
2451 		return;
2452 
2453 	ip6_clearpktopts(pktopt, -1);
2454 
2455 	free(pktopt, M_IP6OPT);
2456 }
2457 
2458 /*
2459  * Set IPv6 outgoing packet options based on advanced API.
2460  */
2461 int
2462 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2463     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2464 {
2465 	struct cmsghdr *cm = 0;
2466 
2467 	if (control == NULL || opt == NULL)
2468 		return (EINVAL);
2469 
2470 	ip6_initpktopts(opt);
2471 	if (stickyopt) {
2472 		int error;
2473 
2474 		/*
2475 		 * If stickyopt is provided, make a local copy of the options
2476 		 * for this particular packet, then override them by ancillary
2477 		 * objects.
2478 		 * XXX: copypktopts() does not copy the cached route to a next
2479 		 * hop (if any).  This is not very good in terms of efficiency,
2480 		 * but we can allow this since this option should be rarely
2481 		 * used.
2482 		 */
2483 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2484 			return (error);
2485 	}
2486 
2487 	/*
2488 	 * XXX: Currently, we assume all the optional information is stored
2489 	 * in a single mbuf.
2490 	 */
2491 	if (control->m_next)
2492 		return (EINVAL);
2493 
2494 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2495 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2496 		int error;
2497 
2498 		if (control->m_len < CMSG_LEN(0))
2499 			return (EINVAL);
2500 
2501 		cm = mtod(control, struct cmsghdr *);
2502 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2503 			return (EINVAL);
2504 		if (cm->cmsg_level != IPPROTO_IPV6)
2505 			continue;
2506 
2507 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2508 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2509 		if (error)
2510 			return (error);
2511 	}
2512 
2513 	return (0);
2514 }
2515 
2516 /*
2517  * Set a particular packet option, as a sticky option or an ancillary data
2518  * item.  "len" can be 0 only when it's a sticky option.
2519  * We have 4 cases of combination of "sticky" and "cmsg":
2520  * "sticky=0, cmsg=0": impossible
2521  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2522  * "sticky=1, cmsg=0": RFC3542 socket option
2523  * "sticky=1, cmsg=1": RFC2292 socket option
2524  */
2525 static int
2526 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2527     struct ucred *cred, int sticky, int cmsg, int uproto)
2528 {
2529 	int minmtupolicy, preftemp;
2530 	int error;
2531 
2532 	if (!sticky && !cmsg) {
2533 #ifdef DIAGNOSTIC
2534 		printf("ip6_setpktopt: impossible case\n");
2535 #endif
2536 		return (EINVAL);
2537 	}
2538 
2539 	/*
2540 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2541 	 * not be specified in the context of RFC3542.  Conversely,
2542 	 * RFC3542 types should not be specified in the context of RFC2292.
2543 	 */
2544 	if (!cmsg) {
2545 		switch (optname) {
2546 		case IPV6_2292PKTINFO:
2547 		case IPV6_2292HOPLIMIT:
2548 		case IPV6_2292NEXTHOP:
2549 		case IPV6_2292HOPOPTS:
2550 		case IPV6_2292DSTOPTS:
2551 		case IPV6_2292RTHDR:
2552 		case IPV6_2292PKTOPTIONS:
2553 			return (ENOPROTOOPT);
2554 		}
2555 	}
2556 	if (sticky && cmsg) {
2557 		switch (optname) {
2558 		case IPV6_PKTINFO:
2559 		case IPV6_HOPLIMIT:
2560 		case IPV6_NEXTHOP:
2561 		case IPV6_HOPOPTS:
2562 		case IPV6_DSTOPTS:
2563 		case IPV6_RTHDRDSTOPTS:
2564 		case IPV6_RTHDR:
2565 		case IPV6_USE_MIN_MTU:
2566 		case IPV6_DONTFRAG:
2567 		case IPV6_TCLASS:
2568 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2569 			return (ENOPROTOOPT);
2570 		}
2571 	}
2572 
2573 	switch (optname) {
2574 	case IPV6_2292PKTINFO:
2575 	case IPV6_PKTINFO:
2576 	{
2577 		struct ifnet *ifp = NULL;
2578 		struct in6_pktinfo *pktinfo;
2579 
2580 		if (len != sizeof(struct in6_pktinfo))
2581 			return (EINVAL);
2582 
2583 		pktinfo = (struct in6_pktinfo *)buf;
2584 
2585 		/*
2586 		 * An application can clear any sticky IPV6_PKTINFO option by
2587 		 * doing a "regular" setsockopt with ipi6_addr being
2588 		 * in6addr_any and ipi6_ifindex being zero.
2589 		 * [RFC 3542, Section 6]
2590 		 */
2591 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2592 		    pktinfo->ipi6_ifindex == 0 &&
2593 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2594 			ip6_clearpktopts(opt, optname);
2595 			break;
2596 		}
2597 
2598 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2599 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2600 			return (EINVAL);
2601 		}
2602 
2603 		/* validate the interface index if specified. */
2604 		if (pktinfo->ipi6_ifindex > V_if_index ||
2605 		    pktinfo->ipi6_ifindex < 0) {
2606 			 return (ENXIO);
2607 		}
2608 		if (pktinfo->ipi6_ifindex) {
2609 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2610 			if (ifp == NULL)
2611 				return (ENXIO);
2612 		}
2613 
2614 		/*
2615 		 * We store the address anyway, and let in6_selectsrc()
2616 		 * validate the specified address.  This is because ipi6_addr
2617 		 * may not have enough information about its scope zone, and
2618 		 * we may need additional information (such as outgoing
2619 		 * interface or the scope zone of a destination address) to
2620 		 * disambiguate the scope.
2621 		 * XXX: the delay of the validation may confuse the
2622 		 * application when it is used as a sticky option.
2623 		 */
2624 		if (opt->ip6po_pktinfo == NULL) {
2625 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2626 			    M_IP6OPT, M_NOWAIT);
2627 			if (opt->ip6po_pktinfo == NULL)
2628 				return (ENOBUFS);
2629 		}
2630 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2631 		break;
2632 	}
2633 
2634 	case IPV6_2292HOPLIMIT:
2635 	case IPV6_HOPLIMIT:
2636 	{
2637 		int *hlimp;
2638 
2639 		/*
2640 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2641 		 * to simplify the ordering among hoplimit options.
2642 		 */
2643 		if (optname == IPV6_HOPLIMIT && sticky)
2644 			return (ENOPROTOOPT);
2645 
2646 		if (len != sizeof(int))
2647 			return (EINVAL);
2648 		hlimp = (int *)buf;
2649 		if (*hlimp < -1 || *hlimp > 255)
2650 			return (EINVAL);
2651 
2652 		opt->ip6po_hlim = *hlimp;
2653 		break;
2654 	}
2655 
2656 	case IPV6_TCLASS:
2657 	{
2658 		int tclass;
2659 
2660 		if (len != sizeof(int))
2661 			return (EINVAL);
2662 		tclass = *(int *)buf;
2663 		if (tclass < -1 || tclass > 255)
2664 			return (EINVAL);
2665 
2666 		opt->ip6po_tclass = tclass;
2667 		break;
2668 	}
2669 
2670 	case IPV6_2292NEXTHOP:
2671 	case IPV6_NEXTHOP:
2672 		if (cred != NULL) {
2673 			error = priv_check_cred(cred,
2674 			    PRIV_NETINET_SETHDROPTS, 0);
2675 			if (error)
2676 				return (error);
2677 		}
2678 
2679 		if (len == 0) {	/* just remove the option */
2680 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2681 			break;
2682 		}
2683 
2684 		/* check if cmsg_len is large enough for sa_len */
2685 		if (len < sizeof(struct sockaddr) || len < *buf)
2686 			return (EINVAL);
2687 
2688 		switch (((struct sockaddr *)buf)->sa_family) {
2689 		case AF_INET6:
2690 		{
2691 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2692 			int error;
2693 
2694 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2695 				return (EINVAL);
2696 
2697 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2698 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2699 				return (EINVAL);
2700 			}
2701 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2702 			    != 0) {
2703 				return (error);
2704 			}
2705 			break;
2706 		}
2707 		case AF_LINK:	/* should eventually be supported */
2708 		default:
2709 			return (EAFNOSUPPORT);
2710 		}
2711 
2712 		/* turn off the previous option, then set the new option. */
2713 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2714 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2715 		if (opt->ip6po_nexthop == NULL)
2716 			return (ENOBUFS);
2717 		bcopy(buf, opt->ip6po_nexthop, *buf);
2718 		break;
2719 
2720 	case IPV6_2292HOPOPTS:
2721 	case IPV6_HOPOPTS:
2722 	{
2723 		struct ip6_hbh *hbh;
2724 		int hbhlen;
2725 
2726 		/*
2727 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2728 		 * options, since per-option restriction has too much
2729 		 * overhead.
2730 		 */
2731 		if (cred != NULL) {
2732 			error = priv_check_cred(cred,
2733 			    PRIV_NETINET_SETHDROPTS, 0);
2734 			if (error)
2735 				return (error);
2736 		}
2737 
2738 		if (len == 0) {
2739 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2740 			break;	/* just remove the option */
2741 		}
2742 
2743 		/* message length validation */
2744 		if (len < sizeof(struct ip6_hbh))
2745 			return (EINVAL);
2746 		hbh = (struct ip6_hbh *)buf;
2747 		hbhlen = (hbh->ip6h_len + 1) << 3;
2748 		if (len != hbhlen)
2749 			return (EINVAL);
2750 
2751 		/* turn off the previous option, then set the new option. */
2752 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2753 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2754 		if (opt->ip6po_hbh == NULL)
2755 			return (ENOBUFS);
2756 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2757 
2758 		break;
2759 	}
2760 
2761 	case IPV6_2292DSTOPTS:
2762 	case IPV6_DSTOPTS:
2763 	case IPV6_RTHDRDSTOPTS:
2764 	{
2765 		struct ip6_dest *dest, **newdest = NULL;
2766 		int destlen;
2767 
2768 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2769 			error = priv_check_cred(cred,
2770 			    PRIV_NETINET_SETHDROPTS, 0);
2771 			if (error)
2772 				return (error);
2773 		}
2774 
2775 		if (len == 0) {
2776 			ip6_clearpktopts(opt, optname);
2777 			break;	/* just remove the option */
2778 		}
2779 
2780 		/* message length validation */
2781 		if (len < sizeof(struct ip6_dest))
2782 			return (EINVAL);
2783 		dest = (struct ip6_dest *)buf;
2784 		destlen = (dest->ip6d_len + 1) << 3;
2785 		if (len != destlen)
2786 			return (EINVAL);
2787 
2788 		/*
2789 		 * Determine the position that the destination options header
2790 		 * should be inserted; before or after the routing header.
2791 		 */
2792 		switch (optname) {
2793 		case IPV6_2292DSTOPTS:
2794 			/*
2795 			 * The old advacned API is ambiguous on this point.
2796 			 * Our approach is to determine the position based
2797 			 * according to the existence of a routing header.
2798 			 * Note, however, that this depends on the order of the
2799 			 * extension headers in the ancillary data; the 1st
2800 			 * part of the destination options header must appear
2801 			 * before the routing header in the ancillary data,
2802 			 * too.
2803 			 * RFC3542 solved the ambiguity by introducing
2804 			 * separate ancillary data or option types.
2805 			 */
2806 			if (opt->ip6po_rthdr == NULL)
2807 				newdest = &opt->ip6po_dest1;
2808 			else
2809 				newdest = &opt->ip6po_dest2;
2810 			break;
2811 		case IPV6_RTHDRDSTOPTS:
2812 			newdest = &opt->ip6po_dest1;
2813 			break;
2814 		case IPV6_DSTOPTS:
2815 			newdest = &opt->ip6po_dest2;
2816 			break;
2817 		}
2818 
2819 		/* turn off the previous option, then set the new option. */
2820 		ip6_clearpktopts(opt, optname);
2821 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2822 		if (*newdest == NULL)
2823 			return (ENOBUFS);
2824 		bcopy(dest, *newdest, destlen);
2825 
2826 		break;
2827 	}
2828 
2829 	case IPV6_2292RTHDR:
2830 	case IPV6_RTHDR:
2831 	{
2832 		struct ip6_rthdr *rth;
2833 		int rthlen;
2834 
2835 		if (len == 0) {
2836 			ip6_clearpktopts(opt, IPV6_RTHDR);
2837 			break;	/* just remove the option */
2838 		}
2839 
2840 		/* message length validation */
2841 		if (len < sizeof(struct ip6_rthdr))
2842 			return (EINVAL);
2843 		rth = (struct ip6_rthdr *)buf;
2844 		rthlen = (rth->ip6r_len + 1) << 3;
2845 		if (len != rthlen)
2846 			return (EINVAL);
2847 
2848 		switch (rth->ip6r_type) {
2849 		case IPV6_RTHDR_TYPE_0:
2850 			if (rth->ip6r_len == 0)	/* must contain one addr */
2851 				return (EINVAL);
2852 			if (rth->ip6r_len % 2) /* length must be even */
2853 				return (EINVAL);
2854 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2855 				return (EINVAL);
2856 			break;
2857 		default:
2858 			return (EINVAL);	/* not supported */
2859 		}
2860 
2861 		/* turn off the previous option */
2862 		ip6_clearpktopts(opt, IPV6_RTHDR);
2863 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2864 		if (opt->ip6po_rthdr == NULL)
2865 			return (ENOBUFS);
2866 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2867 
2868 		break;
2869 	}
2870 
2871 	case IPV6_USE_MIN_MTU:
2872 		if (len != sizeof(int))
2873 			return (EINVAL);
2874 		minmtupolicy = *(int *)buf;
2875 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2876 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2877 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2878 			return (EINVAL);
2879 		}
2880 		opt->ip6po_minmtu = minmtupolicy;
2881 		break;
2882 
2883 	case IPV6_DONTFRAG:
2884 		if (len != sizeof(int))
2885 			return (EINVAL);
2886 
2887 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2888 			/*
2889 			 * we ignore this option for TCP sockets.
2890 			 * (RFC3542 leaves this case unspecified.)
2891 			 */
2892 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2893 		} else
2894 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2895 		break;
2896 
2897 	case IPV6_PREFER_TEMPADDR:
2898 		if (len != sizeof(int))
2899 			return (EINVAL);
2900 		preftemp = *(int *)buf;
2901 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2902 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2903 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2904 			return (EINVAL);
2905 		}
2906 		opt->ip6po_prefer_tempaddr = preftemp;
2907 		break;
2908 
2909 	default:
2910 		return (ENOPROTOOPT);
2911 	} /* end of switch */
2912 
2913 	return (0);
2914 }
2915 
2916 /*
2917  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2918  * packet to the input queue of a specified interface.  Note that this
2919  * calls the output routine of the loopback "driver", but with an interface
2920  * pointer that might NOT be &loif -- easier than replicating that code here.
2921  */
2922 void
2923 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2924 {
2925 	struct mbuf *copym;
2926 	struct ip6_hdr *ip6;
2927 
2928 	copym = m_copy(m, 0, M_COPYALL);
2929 	if (copym == NULL)
2930 		return;
2931 
2932 	/*
2933 	 * Make sure to deep-copy IPv6 header portion in case the data
2934 	 * is in an mbuf cluster, so that we can safely override the IPv6
2935 	 * header portion later.
2936 	 */
2937 	if ((copym->m_flags & M_EXT) != 0 ||
2938 	    copym->m_len < sizeof(struct ip6_hdr)) {
2939 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2940 		if (copym == NULL)
2941 			return;
2942 	}
2943 
2944 #ifdef DIAGNOSTIC
2945 	if (copym->m_len < sizeof(*ip6)) {
2946 		m_freem(copym);
2947 		return;
2948 	}
2949 #endif
2950 
2951 	ip6 = mtod(copym, struct ip6_hdr *);
2952 	/*
2953 	 * clear embedded scope identifiers if necessary.
2954 	 * in6_clearscope will touch the addresses only when necessary.
2955 	 */
2956 	in6_clearscope(&ip6->ip6_src);
2957 	in6_clearscope(&ip6->ip6_dst);
2958 
2959 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2960 }
2961 
2962 /*
2963  * Chop IPv6 header off from the payload.
2964  */
2965 static int
2966 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2967 {
2968 	struct mbuf *mh;
2969 	struct ip6_hdr *ip6;
2970 
2971 	ip6 = mtod(m, struct ip6_hdr *);
2972 	if (m->m_len > sizeof(*ip6)) {
2973 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2974 		if (mh == 0) {
2975 			m_freem(m);
2976 			return ENOBUFS;
2977 		}
2978 		M_MOVE_PKTHDR(mh, m);
2979 		MH_ALIGN(mh, sizeof(*ip6));
2980 		m->m_len -= sizeof(*ip6);
2981 		m->m_data += sizeof(*ip6);
2982 		mh->m_next = m;
2983 		m = mh;
2984 		m->m_len = sizeof(*ip6);
2985 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2986 	}
2987 	exthdrs->ip6e_ip6 = m;
2988 	return 0;
2989 }
2990 
2991 /*
2992  * Compute IPv6 extension header length.
2993  */
2994 int
2995 ip6_optlen(struct inpcb *in6p)
2996 {
2997 	int len;
2998 
2999 	if (!in6p->in6p_outputopts)
3000 		return 0;
3001 
3002 	len = 0;
3003 #define elen(x) \
3004     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3005 
3006 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3007 	if (in6p->in6p_outputopts->ip6po_rthdr)
3008 		/* dest1 is valid with rthdr only */
3009 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3010 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3011 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3012 	return len;
3013 #undef elen
3014 }
3015