xref: /netbsd-src/sys/netinet6/ip6_output.c (revision e39ef1d61eee3ccba837ee281f1e098c864487aa)
1 /*	$NetBSD: ip6_output.c,v 1.144 2012/01/10 20:05:37 drochner Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.144 2012/01/10 20:05:37 drochner Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/kauth.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #ifdef PFIL_HOOKS
86 #include <net/pfil.h>
87 #endif
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet6/in6_offload.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/ip6_private.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/scope6_var.h>
101 
102 #ifdef KAME_IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netinet6/ipsec_private.h>
105 #include <netkey/key.h>
106 #endif /* KAME_IPSEC */
107 
108 #ifdef FAST_IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netipsec/ipsec6.h>
111 #include <netipsec/key.h>
112 #include <netipsec/xform.h>
113 #endif
114 
115 
116 #include <net/net_osdep.h>
117 
118 #ifdef PFIL_HOOKS
119 extern struct pfil_head inet6_pfil_hook;	/* XXX */
120 #endif
121 
122 struct ip6_exthdrs {
123 	struct mbuf *ip6e_ip6;
124 	struct mbuf *ip6e_hbh;
125 	struct mbuf *ip6e_dest1;
126 	struct mbuf *ip6e_rthdr;
127 	struct mbuf *ip6e_dest2;
128 };
129 
130 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
131 	kauth_cred_t, int);
132 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
133 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
134 	int, int, int);
135 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
136 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
137 static int ip6_copyexthdr(struct mbuf **, void *, int);
138 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
139 	struct ip6_frag **);
140 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
141 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
142 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
143     const struct in6_addr *, u_long *, int *);
144 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
145 
146 #ifdef RFC2292
147 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
148 #endif
149 
150 /*
151  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
152  * header (with pri, len, nxt, hlim, src, dst).
153  * This function may modify ver and hlim only.
154  * The mbuf chain containing the packet will be freed.
155  * The mbuf opt, if present, will not be freed.
156  *
157  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
158  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
159  * which is rt_rmx.rmx_mtu.
160  */
161 int
162 ip6_output(
163     struct mbuf *m0,
164     struct ip6_pktopts *opt,
165     struct route *ro,
166     int flags,
167     struct ip6_moptions *im6o,
168     struct socket *so,
169     struct ifnet **ifpp		/* XXX: just for statistics */
170 )
171 {
172 	struct ip6_hdr *ip6, *mhip6;
173 	struct ifnet *ifp, *origifp;
174 	struct mbuf *m = m0;
175 	int hlen, tlen, len, off;
176 	bool tso;
177 	struct route ip6route;
178 	struct rtentry *rt = NULL;
179 	const struct sockaddr_in6 *dst = NULL;
180 	struct sockaddr_in6 src_sa, dst_sa;
181 	int error = 0;
182 	struct in6_ifaddr *ia = NULL;
183 	u_long mtu;
184 	int alwaysfrag, dontfrag;
185 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
186 	struct ip6_exthdrs exthdrs;
187 	struct in6_addr finaldst, src0, dst0;
188 	u_int32_t zone;
189 	struct route *ro_pmtu = NULL;
190 	int hdrsplit = 0;
191 	int needipsec = 0;
192 #ifdef KAME_IPSEC
193 	int needipsectun = 0;
194 	struct secpolicy *sp = NULL;
195 
196 	ip6 = mtod(m, struct ip6_hdr *);
197 #endif /* KAME_IPSEC */
198 #ifdef FAST_IPSEC
199 	struct secpolicy *sp = NULL;
200 	int s;
201 #endif
202 
203 	memset(&ip6route, 0, sizeof(ip6route));
204 
205 #ifdef  DIAGNOSTIC
206 	if ((m->m_flags & M_PKTHDR) == 0)
207 		panic("ip6_output: no HDR");
208 
209 	if ((m->m_pkthdr.csum_flags &
210 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
211 		panic("ip6_output: IPv4 checksum offload flags: %d",
212 		    m->m_pkthdr.csum_flags);
213 	}
214 
215 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
216 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
217 		panic("ip6_output: conflicting checksum offload flags: %d",
218 		    m->m_pkthdr.csum_flags);
219 	}
220 #endif
221 
222 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
223 
224 #define MAKE_EXTHDR(hp, mp)						\
225     do {								\
226 	if (hp) {							\
227 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
228 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
229 		    ((eh)->ip6e_len + 1) << 3);				\
230 		if (error)						\
231 			goto freehdrs;					\
232 	}								\
233     } while (/*CONSTCOND*/ 0)
234 
235 	memset(&exthdrs, 0, sizeof(exthdrs));
236 	if (opt) {
237 		/* Hop-by-Hop options header */
238 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
239 		/* Destination options header(1st part) */
240 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
241 		/* Routing header */
242 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
243 		/* Destination options header(2nd part) */
244 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
245 	}
246 
247 #ifdef KAME_IPSEC
248 	if ((flags & IPV6_FORWARDING) != 0) {
249 		needipsec = 0;
250 		goto skippolicycheck;
251 	}
252 
253 	/* get a security policy for this packet */
254 	if (so == NULL)
255 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
256 	else {
257 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
258 					 IPSEC_DIR_OUTBOUND)) {
259 			needipsec = 0;
260 			goto skippolicycheck;
261 		}
262 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
263 	}
264 
265 	if (sp == NULL) {
266 		IPSEC6_STATINC(IPSEC_STAT_OUT_INVAL);
267 		goto freehdrs;
268 	}
269 
270 	error = 0;
271 
272 	/* check policy */
273 	switch (sp->policy) {
274 	case IPSEC_POLICY_DISCARD:
275 		/*
276 		 * This packet is just discarded.
277 		 */
278 		IPSEC6_STATINC(IPSEC_STAT_OUT_POLVIO);
279 		goto freehdrs;
280 
281 	case IPSEC_POLICY_BYPASS:
282 	case IPSEC_POLICY_NONE:
283 		/* no need to do IPsec. */
284 		needipsec = 0;
285 		break;
286 
287 	case IPSEC_POLICY_IPSEC:
288 		if (sp->req == NULL) {
289 			/* XXX should be panic ? */
290 			printf("ip6_output: No IPsec request specified.\n");
291 			error = EINVAL;
292 			goto freehdrs;
293 		}
294 		needipsec = 1;
295 		break;
296 
297 	case IPSEC_POLICY_ENTRUST:
298 	default:
299 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
300 	}
301 
302   skippolicycheck:;
303 #endif /* KAME_IPSEC */
304 
305 	/*
306 	 * Calculate the total length of the extension header chain.
307 	 * Keep the length of the unfragmentable part for fragmentation.
308 	 */
309 	optlen = 0;
310 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
311 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
312 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
313 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
314 	/* NOTE: we don't add AH/ESP length here. do that later. */
315 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
316 
317 #ifdef FAST_IPSEC
318 	/* Check the security policy (SP) for the packet */
319 
320 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
321 	if (error != 0) {
322 		/*
323 		 * Hack: -EINVAL is used to signal that a packet
324 		 * should be silently discarded.  This is typically
325 		 * because we asked key management for an SA and
326 		 * it was delayed (e.g. kicked up to IKE).
327 		 */
328 	if (error == -EINVAL)
329 		error = 0;
330 	goto freehdrs;
331     }
332 #endif /* FAST_IPSEC */
333 
334 
335 	if (needipsec &&
336 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
337 		in6_delayed_cksum(m);
338 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
339 	}
340 
341 
342 	/*
343 	 * If we need IPsec, or there is at least one extension header,
344 	 * separate IP6 header from the payload.
345 	 */
346 	if ((needipsec || optlen) && !hdrsplit) {
347 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
348 			m = NULL;
349 			goto freehdrs;
350 		}
351 		m = exthdrs.ip6e_ip6;
352 		hdrsplit++;
353 	}
354 
355 	/* adjust pointer */
356 	ip6 = mtod(m, struct ip6_hdr *);
357 
358 	/* adjust mbuf packet header length */
359 	m->m_pkthdr.len += optlen;
360 	plen = m->m_pkthdr.len - sizeof(*ip6);
361 
362 	/* If this is a jumbo payload, insert a jumbo payload option. */
363 	if (plen > IPV6_MAXPACKET) {
364 		if (!hdrsplit) {
365 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
366 				m = NULL;
367 				goto freehdrs;
368 			}
369 			m = exthdrs.ip6e_ip6;
370 			hdrsplit++;
371 		}
372 		/* adjust pointer */
373 		ip6 = mtod(m, struct ip6_hdr *);
374 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
375 			goto freehdrs;
376 		optlen += 8; /* XXX JUMBOOPTLEN */
377 		ip6->ip6_plen = 0;
378 	} else
379 		ip6->ip6_plen = htons(plen);
380 
381 	/*
382 	 * Concatenate headers and fill in next header fields.
383 	 * Here we have, on "m"
384 	 *	IPv6 payload
385 	 * and we insert headers accordingly.  Finally, we should be getting:
386 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
387 	 *
388 	 * during the header composing process, "m" points to IPv6 header.
389 	 * "mprev" points to an extension header prior to esp.
390 	 */
391 	{
392 		u_char *nexthdrp = &ip6->ip6_nxt;
393 		struct mbuf *mprev = m;
394 
395 		/*
396 		 * we treat dest2 specially.  this makes IPsec processing
397 		 * much easier.  the goal here is to make mprev point the
398 		 * mbuf prior to dest2.
399 		 *
400 		 * result: IPv6 dest2 payload
401 		 * m and mprev will point to IPv6 header.
402 		 */
403 		if (exthdrs.ip6e_dest2) {
404 			if (!hdrsplit)
405 				panic("assumption failed: hdr not split");
406 			exthdrs.ip6e_dest2->m_next = m->m_next;
407 			m->m_next = exthdrs.ip6e_dest2;
408 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
409 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
410 		}
411 
412 #define MAKE_CHAIN(m, mp, p, i)\
413     do {\
414 	if (m) {\
415 		if (!hdrsplit) \
416 			panic("assumption failed: hdr not split"); \
417 		*mtod((m), u_char *) = *(p);\
418 		*(p) = (i);\
419 		p = mtod((m), u_char *);\
420 		(m)->m_next = (mp)->m_next;\
421 		(mp)->m_next = (m);\
422 		(mp) = (m);\
423 	}\
424     } while (/*CONSTCOND*/ 0)
425 		/*
426 		 * result: IPv6 hbh dest1 rthdr dest2 payload
427 		 * m will point to IPv6 header.  mprev will point to the
428 		 * extension header prior to dest2 (rthdr in the above case).
429 		 */
430 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
431 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
432 		    IPPROTO_DSTOPTS);
433 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
434 		    IPPROTO_ROUTING);
435 
436 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
437 		    sizeof(struct ip6_hdr) + optlen);
438 
439 #ifdef KAME_IPSEC
440 		if (!needipsec)
441 			goto skip_ipsec2;
442 
443 		/*
444 		 * pointers after IPsec headers are not valid any more.
445 		 * other pointers need a great care too.
446 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
447 		 */
448 		exthdrs.ip6e_dest2 = NULL;
449 
450 	    {
451 		struct ip6_rthdr *rh = NULL;
452 		int segleft_org = 0;
453 		struct ipsec_output_state state;
454 
455 		if (exthdrs.ip6e_rthdr) {
456 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
457 			segleft_org = rh->ip6r_segleft;
458 			rh->ip6r_segleft = 0;
459 		}
460 
461 		memset(&state, 0, sizeof(state));
462 		state.m = m;
463 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
464 		    &needipsectun);
465 		m = state.m;
466 		if (error) {
467 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
468 			/* mbuf is already reclaimed in ipsec6_output_trans. */
469 			m = NULL;
470 			switch (error) {
471 			case EHOSTUNREACH:
472 			case ENETUNREACH:
473 			case EMSGSIZE:
474 			case ENOBUFS:
475 			case ENOMEM:
476 				break;
477 			default:
478 				printf("ip6_output (ipsec): error code %d\n", error);
479 				/* FALLTHROUGH */
480 			case ENOENT:
481 				/* don't show these error codes to the user */
482 				error = 0;
483 				break;
484 			}
485 			goto bad;
486 		}
487 		if (exthdrs.ip6e_rthdr) {
488 			/* ah6_output doesn't modify mbuf chain */
489 			rh->ip6r_segleft = segleft_org;
490 		}
491 	    }
492 skip_ipsec2:;
493 #endif
494 	}
495 
496 	/*
497 	 * If there is a routing header, replace destination address field
498 	 * with the first hop of the routing header.
499 	 */
500 	if (exthdrs.ip6e_rthdr) {
501 		struct ip6_rthdr *rh;
502 		struct ip6_rthdr0 *rh0;
503 		struct in6_addr *addr;
504 		struct sockaddr_in6 sa;
505 
506 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
507 		    struct ip6_rthdr *));
508 		finaldst = ip6->ip6_dst;
509 		switch (rh->ip6r_type) {
510 		case IPV6_RTHDR_TYPE_0:
511 			 rh0 = (struct ip6_rthdr0 *)rh;
512 			 addr = (struct in6_addr *)(rh0 + 1);
513 
514 			 /*
515 			  * construct a sockaddr_in6 form of
516 			  * the first hop.
517 			  *
518 			  * XXX: we may not have enough
519 			  * information about its scope zone;
520 			  * there is no standard API to pass
521 			  * the information from the
522 			  * application.
523 			  */
524 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
525 			 if ((error = sa6_embedscope(&sa,
526 			     ip6_use_defzone)) != 0) {
527 				 goto bad;
528 			 }
529 			 ip6->ip6_dst = sa.sin6_addr;
530 			 (void)memmove(&addr[0], &addr[1],
531 			     sizeof(struct in6_addr) *
532 			     (rh0->ip6r0_segleft - 1));
533 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
534 			 /* XXX */
535 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
536 			 break;
537 		default:	/* is it possible? */
538 			 error = EINVAL;
539 			 goto bad;
540 		}
541 	}
542 
543 	/* Source address validation */
544 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
545 	    (flags & IPV6_UNSPECSRC) == 0) {
546 		error = EOPNOTSUPP;
547 		IP6_STATINC(IP6_STAT_BADSCOPE);
548 		goto bad;
549 	}
550 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
551 		error = EOPNOTSUPP;
552 		IP6_STATINC(IP6_STAT_BADSCOPE);
553 		goto bad;
554 	}
555 
556 	IP6_STATINC(IP6_STAT_LOCALOUT);
557 
558 	/*
559 	 * Route packet.
560 	 */
561 	/* initialize cached route */
562 	if (ro == NULL) {
563 		ro = &ip6route;
564 	}
565 	ro_pmtu = ro;
566 	if (opt && opt->ip6po_rthdr)
567 		ro = &opt->ip6po_route;
568 
569  	/*
570 	 * if specified, try to fill in the traffic class field.
571 	 * do not override if a non-zero value is already set.
572 	 * we check the diffserv field and the ecn field separately.
573 	 */
574 	if (opt && opt->ip6po_tclass >= 0) {
575 		int mask = 0;
576 
577 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
578 			mask |= 0xfc;
579 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
580 			mask |= 0x03;
581 		if (mask != 0)
582 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
583 	}
584 
585 	/* fill in or override the hop limit field, if necessary. */
586 	if (opt && opt->ip6po_hlim != -1)
587 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
588 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
589 		if (im6o != NULL)
590 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
591 		else
592 			ip6->ip6_hlim = ip6_defmcasthlim;
593 	}
594 
595 #ifdef KAME_IPSEC
596 	if (needipsec && needipsectun) {
597 		struct ipsec_output_state state;
598 
599 		/*
600 		 * All the extension headers will become inaccessible
601 		 * (since they can be encrypted).
602 		 * Don't panic, we need no more updates to extension headers
603 		 * on inner IPv6 packet (since they are now encapsulated).
604 		 *
605 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
606 		 */
607 		memset(&exthdrs, 0, sizeof(exthdrs));
608 		exthdrs.ip6e_ip6 = m;
609 
610 		memset(&state, 0, sizeof(state));
611 		state.m = m;
612 		state.ro = ro;
613 		state.dst = rtcache_getdst(ro);
614 
615 		error = ipsec6_output_tunnel(&state, sp, flags);
616 
617 		m = state.m;
618 		ro_pmtu = ro = state.ro;
619 		dst = satocsin6(state.dst);
620 		if (error) {
621 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
622 			m0 = m = NULL;
623 			m = NULL;
624 			switch (error) {
625 			case EHOSTUNREACH:
626 			case ENETUNREACH:
627 			case EMSGSIZE:
628 			case ENOBUFS:
629 			case ENOMEM:
630 				break;
631 			default:
632 				printf("ip6_output (ipsec): error code %d\n", error);
633 				/* FALLTHROUGH */
634 			case ENOENT:
635 				/* don't show these error codes to the user */
636 				error = 0;
637 				break;
638 			}
639 			goto bad;
640 		}
641 
642 		exthdrs.ip6e_ip6 = m;
643 	}
644 #endif /* KAME_IPSEC */
645 #ifdef FAST_IPSEC
646 	if (needipsec) {
647 		s = splsoftnet();
648 		error = ipsec6_process_packet(m,sp->req);
649 
650 		/*
651 		 * Preserve KAME behaviour: ENOENT can be returned
652 		 * when an SA acquire is in progress.  Don't propagate
653 		 * this to user-level; it confuses applications.
654 		 * XXX this will go away when the SADB is redone.
655 		 */
656 		if (error == ENOENT)
657 			error = 0;
658 		splx(s);
659 		goto done;
660 	}
661 #endif /* FAST_IPSEC */
662 
663 
664 
665 	/* adjust pointer */
666 	ip6 = mtod(m, struct ip6_hdr *);
667 
668 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
669 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
670 	    &ifp, &rt, 0)) != 0) {
671 		if (ifp != NULL)
672 			in6_ifstat_inc(ifp, ifs6_out_discard);
673 		goto bad;
674 	}
675 	if (rt == NULL) {
676 		/*
677 		 * If in6_selectroute() does not return a route entry,
678 		 * dst may not have been updated.
679 		 */
680 		rtcache_setdst(ro, sin6tosa(&dst_sa));
681 	}
682 
683 	/*
684 	 * then rt (for unicast) and ifp must be non-NULL valid values.
685 	 */
686 	if ((flags & IPV6_FORWARDING) == 0) {
687 		/* XXX: the FORWARDING flag can be set for mrouting. */
688 		in6_ifstat_inc(ifp, ifs6_out_request);
689 	}
690 	if (rt != NULL) {
691 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
692 		rt->rt_use++;
693 	}
694 
695 	/*
696 	 * The outgoing interface must be in the zone of source and
697 	 * destination addresses.  We should use ia_ifp to support the
698 	 * case of sending packets to an address of our own.
699 	 */
700 	if (ia != NULL && ia->ia_ifp)
701 		origifp = ia->ia_ifp;
702 	else
703 		origifp = ifp;
704 
705 	src0 = ip6->ip6_src;
706 	if (in6_setscope(&src0, origifp, &zone))
707 		goto badscope;
708 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
709 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
710 		goto badscope;
711 
712 	dst0 = ip6->ip6_dst;
713 	if (in6_setscope(&dst0, origifp, &zone))
714 		goto badscope;
715 	/* re-initialize to be sure */
716 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
717 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
718 		goto badscope;
719 
720 	/* scope check is done. */
721 
722 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
723 		if (dst == NULL)
724 			dst = satocsin6(rtcache_getdst(ro));
725 		KASSERT(dst != NULL);
726 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
727 		/*
728 		 * The nexthop is explicitly specified by the
729 		 * application.  We assume the next hop is an IPv6
730 		 * address.
731 		 */
732 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
733 	} else if ((rt->rt_flags & RTF_GATEWAY))
734 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
735 	else if (dst == NULL)
736 		dst = satocsin6(rtcache_getdst(ro));
737 
738 	/*
739 	 * XXXXXX: original code follows:
740 	 */
741 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
742 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
743 	else {
744 		struct	in6_multi *in6m;
745 
746 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
747 
748 		in6_ifstat_inc(ifp, ifs6_out_mcast);
749 
750 		/*
751 		 * Confirm that the outgoing interface supports multicast.
752 		 */
753 		if (!(ifp->if_flags & IFF_MULTICAST)) {
754 			IP6_STATINC(IP6_STAT_NOROUTE);
755 			in6_ifstat_inc(ifp, ifs6_out_discard);
756 			error = ENETUNREACH;
757 			goto bad;
758 		}
759 
760 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
761 		if (in6m != NULL &&
762 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
763 			/*
764 			 * If we belong to the destination multicast group
765 			 * on the outgoing interface, and the caller did not
766 			 * forbid loopback, loop back a copy.
767 			 */
768 			KASSERT(dst != NULL);
769 			ip6_mloopback(ifp, m, dst);
770 		} else {
771 			/*
772 			 * If we are acting as a multicast router, perform
773 			 * multicast forwarding as if the packet had just
774 			 * arrived on the interface to which we are about
775 			 * to send.  The multicast forwarding function
776 			 * recursively calls this function, using the
777 			 * IPV6_FORWARDING flag to prevent infinite recursion.
778 			 *
779 			 * Multicasts that are looped back by ip6_mloopback(),
780 			 * above, will be forwarded by the ip6_input() routine,
781 			 * if necessary.
782 			 */
783 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
784 				if (ip6_mforward(ip6, ifp, m) != 0) {
785 					m_freem(m);
786 					goto done;
787 				}
788 			}
789 		}
790 		/*
791 		 * Multicasts with a hoplimit of zero may be looped back,
792 		 * above, but must not be transmitted on a network.
793 		 * Also, multicasts addressed to the loopback interface
794 		 * are not sent -- the above call to ip6_mloopback() will
795 		 * loop back a copy if this host actually belongs to the
796 		 * destination group on the loopback interface.
797 		 */
798 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
799 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
800 			m_freem(m);
801 			goto done;
802 		}
803 	}
804 
805 	/*
806 	 * Fill the outgoing inteface to tell the upper layer
807 	 * to increment per-interface statistics.
808 	 */
809 	if (ifpp)
810 		*ifpp = ifp;
811 
812 	/* Determine path MTU. */
813 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
814 	    &alwaysfrag)) != 0)
815 		goto bad;
816 #ifdef KAME_IPSEC
817 	if (needipsectun)
818 		mtu = IPV6_MMTU;
819 #endif
820 
821 	/*
822 	 * The caller of this function may specify to use the minimum MTU
823 	 * in some cases.
824 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
825 	 * setting.  The logic is a bit complicated; by default, unicast
826 	 * packets will follow path MTU while multicast packets will be sent at
827 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
828 	 * including unicast ones will be sent at the minimum MTU.  Multicast
829 	 * packets will always be sent at the minimum MTU unless
830 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
831 	 * See RFC 3542 for more details.
832 	 */
833 	if (mtu > IPV6_MMTU) {
834 		if ((flags & IPV6_MINMTU))
835 			mtu = IPV6_MMTU;
836 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
837 			mtu = IPV6_MMTU;
838 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
839 			 (opt == NULL ||
840 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
841 			mtu = IPV6_MMTU;
842 		}
843 	}
844 
845 	/*
846 	 * clear embedded scope identifiers if necessary.
847 	 * in6_clearscope will touch the addresses only when necessary.
848 	 */
849 	in6_clearscope(&ip6->ip6_src);
850 	in6_clearscope(&ip6->ip6_dst);
851 
852 	/*
853 	 * If the outgoing packet contains a hop-by-hop options header,
854 	 * it must be examined and processed even by the source node.
855 	 * (RFC 2460, section 4.)
856 	 */
857 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
858 		u_int32_t dummy1; /* XXX unused */
859 		u_int32_t dummy2; /* XXX unused */
860 		int hoff = sizeof(struct ip6_hdr);
861 
862 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
863 			/* m was already freed at this point */
864 			error = EINVAL;/* better error? */
865 			goto done;
866 		}
867 
868 		ip6 = mtod(m, struct ip6_hdr *);
869 	}
870 
871 #ifdef PFIL_HOOKS
872 	/*
873 	 * Run through list of hooks for output packets.
874 	 */
875 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
876 		goto done;
877 	if (m == NULL)
878 		goto done;
879 	ip6 = mtod(m, struct ip6_hdr *);
880 #endif /* PFIL_HOOKS */
881 	/*
882 	 * Send the packet to the outgoing interface.
883 	 * If necessary, do IPv6 fragmentation before sending.
884 	 *
885 	 * the logic here is rather complex:
886 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
887 	 * 1-a:	send as is if tlen <= path mtu
888 	 * 1-b:	fragment if tlen > path mtu
889 	 *
890 	 * 2: if user asks us not to fragment (dontfrag == 1)
891 	 * 2-a:	send as is if tlen <= interface mtu
892 	 * 2-b:	error if tlen > interface mtu
893 	 *
894 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
895 	 *	always fragment
896 	 *
897 	 * 4: if dontfrag == 1 && alwaysfrag == 1
898 	 *	error, as we cannot handle this conflicting request
899 	 */
900 	tlen = m->m_pkthdr.len;
901 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
902 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
903 		dontfrag = 1;
904 	else
905 		dontfrag = 0;
906 
907 	if (dontfrag && alwaysfrag) {	/* case 4 */
908 		/* conflicting request - can't transmit */
909 		error = EMSGSIZE;
910 		goto bad;
911 	}
912 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
913 		/*
914 		 * Even if the DONTFRAG option is specified, we cannot send the
915 		 * packet when the data length is larger than the MTU of the
916 		 * outgoing interface.
917 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
918 		 * well as returning an error code (the latter is not described
919 		 * in the API spec.)
920 		 */
921 		u_int32_t mtu32;
922 		struct ip6ctlparam ip6cp;
923 
924 		mtu32 = (u_int32_t)mtu;
925 		memset(&ip6cp, 0, sizeof(ip6cp));
926 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
927 		pfctlinput2(PRC_MSGSIZE,
928 		    rtcache_getdst(ro_pmtu), &ip6cp);
929 
930 		error = EMSGSIZE;
931 		goto bad;
932 	}
933 
934 	/*
935 	 * transmit packet without fragmentation
936 	 */
937 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
938 		/* case 1-a and 2-a */
939 		struct in6_ifaddr *ia6;
940 		int sw_csum;
941 
942 		ip6 = mtod(m, struct ip6_hdr *);
943 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
944 		if (ia6) {
945 			/* Record statistics for this interface address. */
946 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
947 		}
948 #ifdef KAME_IPSEC
949 		/* clean ipsec history once it goes out of the node */
950 		ipsec_delaux(m);
951 #endif
952 
953 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
954 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
955 			if (IN6_NEED_CHECKSUM(ifp,
956 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
957 				in6_delayed_cksum(m);
958 			}
959 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
960 		}
961 
962 		KASSERT(dst != NULL);
963 		if (__predict_true(!tso ||
964 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
965 			error = nd6_output(ifp, origifp, m, dst, rt);
966 		} else {
967 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
968 		}
969 		goto done;
970 	}
971 
972 	if (tso) {
973 		error = EINVAL; /* XXX */
974 		goto bad;
975 	}
976 
977 	/*
978 	 * try to fragment the packet.  case 1-b and 3
979 	 */
980 	if (mtu < IPV6_MMTU) {
981 		/* path MTU cannot be less than IPV6_MMTU */
982 		error = EMSGSIZE;
983 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
984 		goto bad;
985 	} else if (ip6->ip6_plen == 0) {
986 		/* jumbo payload cannot be fragmented */
987 		error = EMSGSIZE;
988 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
989 		goto bad;
990 	} else {
991 		struct mbuf **mnext, *m_frgpart;
992 		struct ip6_frag *ip6f;
993 		u_int32_t id = htonl(ip6_randomid());
994 		u_char nextproto;
995 #if 0				/* see below */
996 		struct ip6ctlparam ip6cp;
997 		u_int32_t mtu32;
998 #endif
999 
1000 		/*
1001 		 * Too large for the destination or interface;
1002 		 * fragment if possible.
1003 		 * Must be able to put at least 8 bytes per fragment.
1004 		 */
1005 		hlen = unfragpartlen;
1006 		if (mtu > IPV6_MAXPACKET)
1007 			mtu = IPV6_MAXPACKET;
1008 
1009 #if 0
1010 		/*
1011 		 * It is believed this code is a leftover from the
1012 		 * development of the IPV6_RECVPATHMTU sockopt and
1013 		 * associated work to implement RFC3542.
1014 		 * It's not entirely clear what the intent of the API
1015 		 * is at this point, so disable this code for now.
1016 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1017 		 * will send notifications if the application requests.
1018 		 */
1019 
1020 		/* Notify a proper path MTU to applications. */
1021 		mtu32 = (u_int32_t)mtu;
1022 		memset(&ip6cp, 0, sizeof(ip6cp));
1023 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1024 		pfctlinput2(PRC_MSGSIZE,
1025 		    rtcache_getdst(ro_pmtu), &ip6cp);
1026 #endif
1027 
1028 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1029 		if (len < 8) {
1030 			error = EMSGSIZE;
1031 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1032 			goto bad;
1033 		}
1034 
1035 		mnext = &m->m_nextpkt;
1036 
1037 		/*
1038 		 * Change the next header field of the last header in the
1039 		 * unfragmentable part.
1040 		 */
1041 		if (exthdrs.ip6e_rthdr) {
1042 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1043 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1044 		} else if (exthdrs.ip6e_dest1) {
1045 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1046 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1047 		} else if (exthdrs.ip6e_hbh) {
1048 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1049 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1050 		} else {
1051 			nextproto = ip6->ip6_nxt;
1052 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1053 		}
1054 
1055 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
1056 		    != 0) {
1057 			if (IN6_NEED_CHECKSUM(ifp,
1058 			    m->m_pkthdr.csum_flags &
1059 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
1060 				in6_delayed_cksum(m);
1061 			}
1062 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
1063 		}
1064 
1065 		/*
1066 		 * Loop through length of segment after first fragment,
1067 		 * make new header and copy data of each part and link onto
1068 		 * chain.
1069 		 */
1070 		m0 = m;
1071 		for (off = hlen; off < tlen; off += len) {
1072 			struct mbuf *mlast;
1073 
1074 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1075 			if (!m) {
1076 				error = ENOBUFS;
1077 				IP6_STATINC(IP6_STAT_ODROPPED);
1078 				goto sendorfree;
1079 			}
1080 			m->m_pkthdr.rcvif = NULL;
1081 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1082 			*mnext = m;
1083 			mnext = &m->m_nextpkt;
1084 			m->m_data += max_linkhdr;
1085 			mhip6 = mtod(m, struct ip6_hdr *);
1086 			*mhip6 = *ip6;
1087 			m->m_len = sizeof(*mhip6);
1088 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1089 			if (error) {
1090 				IP6_STATINC(IP6_STAT_ODROPPED);
1091 				goto sendorfree;
1092 			}
1093 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
1094 			if (off + len >= tlen)
1095 				len = tlen - off;
1096 			else
1097 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1098 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
1099 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1100 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1101 				error = ENOBUFS;
1102 				IP6_STATINC(IP6_STAT_ODROPPED);
1103 				goto sendorfree;
1104 			}
1105 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
1106 				;
1107 			mlast->m_next = m_frgpart;
1108 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1109 			m->m_pkthdr.rcvif = NULL;
1110 			ip6f->ip6f_reserved = 0;
1111 			ip6f->ip6f_ident = id;
1112 			ip6f->ip6f_nxt = nextproto;
1113 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
1114 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1115 		}
1116 
1117 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1118 	}
1119 
1120 	/*
1121 	 * Remove leading garbages.
1122 	 */
1123 sendorfree:
1124 	m = m0->m_nextpkt;
1125 	m0->m_nextpkt = 0;
1126 	m_freem(m0);
1127 	for (m0 = m; m; m = m0) {
1128 		m0 = m->m_nextpkt;
1129 		m->m_nextpkt = 0;
1130 		if (error == 0) {
1131 			struct in6_ifaddr *ia6;
1132 			ip6 = mtod(m, struct ip6_hdr *);
1133 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1134 			if (ia6) {
1135 				/*
1136 				 * Record statistics for this interface
1137 				 * address.
1138 				 */
1139 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1140 				    m->m_pkthdr.len;
1141 			}
1142 #ifdef KAME_IPSEC
1143 			/* clean ipsec history once it goes out of the node */
1144 			ipsec_delaux(m);
1145 #endif
1146 			KASSERT(dst != NULL);
1147 			error = nd6_output(ifp, origifp, m, dst, rt);
1148 		} else
1149 			m_freem(m);
1150 	}
1151 
1152 	if (error == 0)
1153 		IP6_STATINC(IP6_STAT_FRAGMENTED);
1154 
1155 done:
1156 	rtcache_free(&ip6route);
1157 
1158 #ifdef KAME_IPSEC
1159 	if (sp != NULL)
1160 		key_freesp(sp);
1161 #endif /* KAME_IPSEC */
1162 #ifdef FAST_IPSEC
1163 	if (sp != NULL)
1164 		KEY_FREESP(&sp);
1165 #endif /* FAST_IPSEC */
1166 
1167 
1168 	return (error);
1169 
1170 freehdrs:
1171 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1172 	m_freem(exthdrs.ip6e_dest1);
1173 	m_freem(exthdrs.ip6e_rthdr);
1174 	m_freem(exthdrs.ip6e_dest2);
1175 	/* FALLTHROUGH */
1176 bad:
1177 	m_freem(m);
1178 	goto done;
1179 badscope:
1180 	IP6_STATINC(IP6_STAT_BADSCOPE);
1181 	in6_ifstat_inc(origifp, ifs6_out_discard);
1182 	if (error == 0)
1183 		error = EHOSTUNREACH; /* XXX */
1184 	goto bad;
1185 }
1186 
1187 static int
1188 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1189 {
1190 	struct mbuf *m;
1191 
1192 	if (hlen > MCLBYTES)
1193 		return (ENOBUFS); /* XXX */
1194 
1195 	MGET(m, M_DONTWAIT, MT_DATA);
1196 	if (!m)
1197 		return (ENOBUFS);
1198 
1199 	if (hlen > MLEN) {
1200 		MCLGET(m, M_DONTWAIT);
1201 		if ((m->m_flags & M_EXT) == 0) {
1202 			m_free(m);
1203 			return (ENOBUFS);
1204 		}
1205 	}
1206 	m->m_len = hlen;
1207 	if (hdr)
1208 		bcopy(hdr, mtod(m, void *), hlen);
1209 
1210 	*mp = m;
1211 	return (0);
1212 }
1213 
1214 /*
1215  * Process a delayed payload checksum calculation.
1216  */
1217 void
1218 in6_delayed_cksum(struct mbuf *m)
1219 {
1220 	uint16_t csum, offset;
1221 
1222 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1223 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1224 	KASSERT((m->m_pkthdr.csum_flags
1225 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1226 
1227 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1228 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1229 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1230 		csum = 0xffff;
1231 	}
1232 
1233 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1234 	if ((offset + sizeof(csum)) > m->m_len) {
1235 		m_copyback(m, offset, sizeof(csum), &csum);
1236 	} else {
1237 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1238 	}
1239 }
1240 
1241 /*
1242  * Insert jumbo payload option.
1243  */
1244 static int
1245 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1246 {
1247 	struct mbuf *mopt;
1248 	u_int8_t *optbuf;
1249 	u_int32_t v;
1250 
1251 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1252 
1253 	/*
1254 	 * If there is no hop-by-hop options header, allocate new one.
1255 	 * If there is one but it doesn't have enough space to store the
1256 	 * jumbo payload option, allocate a cluster to store the whole options.
1257 	 * Otherwise, use it to store the options.
1258 	 */
1259 	if (exthdrs->ip6e_hbh == 0) {
1260 		MGET(mopt, M_DONTWAIT, MT_DATA);
1261 		if (mopt == 0)
1262 			return (ENOBUFS);
1263 		mopt->m_len = JUMBOOPTLEN;
1264 		optbuf = mtod(mopt, u_int8_t *);
1265 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1266 		exthdrs->ip6e_hbh = mopt;
1267 	} else {
1268 		struct ip6_hbh *hbh;
1269 
1270 		mopt = exthdrs->ip6e_hbh;
1271 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1272 			/*
1273 			 * XXX assumption:
1274 			 * - exthdrs->ip6e_hbh is not referenced from places
1275 			 *   other than exthdrs.
1276 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1277 			 */
1278 			int oldoptlen = mopt->m_len;
1279 			struct mbuf *n;
1280 
1281 			/*
1282 			 * XXX: give up if the whole (new) hbh header does
1283 			 * not fit even in an mbuf cluster.
1284 			 */
1285 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1286 				return (ENOBUFS);
1287 
1288 			/*
1289 			 * As a consequence, we must always prepare a cluster
1290 			 * at this point.
1291 			 */
1292 			MGET(n, M_DONTWAIT, MT_DATA);
1293 			if (n) {
1294 				MCLGET(n, M_DONTWAIT);
1295 				if ((n->m_flags & M_EXT) == 0) {
1296 					m_freem(n);
1297 					n = NULL;
1298 				}
1299 			}
1300 			if (!n)
1301 				return (ENOBUFS);
1302 			n->m_len = oldoptlen + JUMBOOPTLEN;
1303 			bcopy(mtod(mopt, void *), mtod(n, void *),
1304 			    oldoptlen);
1305 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1306 			m_freem(mopt);
1307 			mopt = exthdrs->ip6e_hbh = n;
1308 		} else {
1309 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1310 			mopt->m_len += JUMBOOPTLEN;
1311 		}
1312 		optbuf[0] = IP6OPT_PADN;
1313 		optbuf[1] = 0;
1314 
1315 		/*
1316 		 * Adjust the header length according to the pad and
1317 		 * the jumbo payload option.
1318 		 */
1319 		hbh = mtod(mopt, struct ip6_hbh *);
1320 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1321 	}
1322 
1323 	/* fill in the option. */
1324 	optbuf[2] = IP6OPT_JUMBO;
1325 	optbuf[3] = 4;
1326 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1327 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1328 
1329 	/* finally, adjust the packet header length */
1330 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1331 
1332 	return (0);
1333 #undef JUMBOOPTLEN
1334 }
1335 
1336 /*
1337  * Insert fragment header and copy unfragmentable header portions.
1338  */
1339 static int
1340 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1341 	struct ip6_frag **frghdrp)
1342 {
1343 	struct mbuf *n, *mlast;
1344 
1345 	if (hlen > sizeof(struct ip6_hdr)) {
1346 		n = m_copym(m0, sizeof(struct ip6_hdr),
1347 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1348 		if (n == 0)
1349 			return (ENOBUFS);
1350 		m->m_next = n;
1351 	} else
1352 		n = m;
1353 
1354 	/* Search for the last mbuf of unfragmentable part. */
1355 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1356 		;
1357 
1358 	if ((mlast->m_flags & M_EXT) == 0 &&
1359 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1360 		/* use the trailing space of the last mbuf for the fragment hdr */
1361 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1362 		    mlast->m_len);
1363 		mlast->m_len += sizeof(struct ip6_frag);
1364 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1365 	} else {
1366 		/* allocate a new mbuf for the fragment header */
1367 		struct mbuf *mfrg;
1368 
1369 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1370 		if (mfrg == 0)
1371 			return (ENOBUFS);
1372 		mfrg->m_len = sizeof(struct ip6_frag);
1373 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1374 		mlast->m_next = mfrg;
1375 	}
1376 
1377 	return (0);
1378 }
1379 
1380 static int
1381 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1382     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1383 {
1384 	struct rtentry *rt;
1385 	u_int32_t mtu = 0;
1386 	int alwaysfrag = 0;
1387 	int error = 0;
1388 
1389 	if (ro_pmtu != ro) {
1390 		union {
1391 			struct sockaddr		dst;
1392 			struct sockaddr_in6	dst6;
1393 		} u;
1394 
1395 		/* The first hop and the final destination may differ. */
1396 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1397 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1398 	} else
1399 		rt = rtcache_validate(ro_pmtu);
1400 	if (rt != NULL) {
1401 		u_int32_t ifmtu;
1402 
1403 		if (ifp == NULL)
1404 			ifp = rt->rt_ifp;
1405 		ifmtu = IN6_LINKMTU(ifp);
1406 		mtu = rt->rt_rmx.rmx_mtu;
1407 		if (mtu == 0)
1408 			mtu = ifmtu;
1409 		else if (mtu < IPV6_MMTU) {
1410 			/*
1411 			 * RFC2460 section 5, last paragraph:
1412 			 * if we record ICMPv6 too big message with
1413 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1414 			 * or smaller, with fragment header attached.
1415 			 * (fragment header is needed regardless from the
1416 			 * packet size, for translators to identify packets)
1417 			 */
1418 			alwaysfrag = 1;
1419 			mtu = IPV6_MMTU;
1420 		} else if (mtu > ifmtu) {
1421 			/*
1422 			 * The MTU on the route is larger than the MTU on
1423 			 * the interface!  This shouldn't happen, unless the
1424 			 * MTU of the interface has been changed after the
1425 			 * interface was brought up.  Change the MTU in the
1426 			 * route to match the interface MTU (as long as the
1427 			 * field isn't locked).
1428 			 */
1429 			mtu = ifmtu;
1430 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1431 				rt->rt_rmx.rmx_mtu = mtu;
1432 		}
1433 	} else if (ifp) {
1434 		mtu = IN6_LINKMTU(ifp);
1435 	} else
1436 		error = EHOSTUNREACH; /* XXX */
1437 
1438 	*mtup = mtu;
1439 	if (alwaysfragp)
1440 		*alwaysfragp = alwaysfrag;
1441 	return (error);
1442 }
1443 
1444 /*
1445  * IP6 socket option processing.
1446  */
1447 int
1448 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1449 {
1450 	int optdatalen, uproto;
1451 	void *optdata;
1452 	struct in6pcb *in6p = sotoin6pcb(so);
1453 	int error, optval;
1454 	int level, optname;
1455 
1456 	KASSERT(sopt != NULL);
1457 
1458 	level = sopt->sopt_level;
1459 	optname = sopt->sopt_name;
1460 
1461 	error = optval = 0;
1462 	uproto = (int)so->so_proto->pr_protocol;
1463 
1464 	if (level != IPPROTO_IPV6) {
1465 		return ENOPROTOOPT;
1466 	}
1467 	switch (op) {
1468 	case PRCO_SETOPT:
1469 		switch (optname) {
1470 #ifdef RFC2292
1471 		case IPV6_2292PKTOPTIONS:
1472 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1473 			break;
1474 #endif
1475 
1476 		/*
1477 		 * Use of some Hop-by-Hop options or some
1478 		 * Destination options, might require special
1479 		 * privilege.  That is, normal applications
1480 		 * (without special privilege) might be forbidden
1481 		 * from setting certain options in outgoing packets,
1482 		 * and might never see certain options in received
1483 		 * packets. [RFC 2292 Section 6]
1484 		 * KAME specific note:
1485 		 *  KAME prevents non-privileged users from sending or
1486 		 *  receiving ANY hbh/dst options in order to avoid
1487 		 *  overhead of parsing options in the kernel.
1488 		 */
1489 		case IPV6_RECVHOPOPTS:
1490 		case IPV6_RECVDSTOPTS:
1491 		case IPV6_RECVRTHDRDSTOPTS:
1492 			error = kauth_authorize_generic(kauth_cred_get(),
1493 			    KAUTH_GENERIC_ISSUSER, NULL);
1494 			if (error)
1495 				break;
1496 			/* FALLTHROUGH */
1497 		case IPV6_UNICAST_HOPS:
1498 		case IPV6_HOPLIMIT:
1499 		case IPV6_FAITH:
1500 
1501 		case IPV6_RECVPKTINFO:
1502 		case IPV6_RECVHOPLIMIT:
1503 		case IPV6_RECVRTHDR:
1504 		case IPV6_RECVPATHMTU:
1505 		case IPV6_RECVTCLASS:
1506 		case IPV6_V6ONLY:
1507 			error = sockopt_getint(sopt, &optval);
1508 			if (error)
1509 				break;
1510 			switch (optname) {
1511 			case IPV6_UNICAST_HOPS:
1512 				if (optval < -1 || optval >= 256)
1513 					error = EINVAL;
1514 				else {
1515 					/* -1 = kernel default */
1516 					in6p->in6p_hops = optval;
1517 				}
1518 				break;
1519 #define OPTSET(bit) \
1520 do { \
1521 if (optval) \
1522 	in6p->in6p_flags |= (bit); \
1523 else \
1524 	in6p->in6p_flags &= ~(bit); \
1525 } while (/*CONSTCOND*/ 0)
1526 
1527 #ifdef RFC2292
1528 #define OPTSET2292(bit) 			\
1529 do { 						\
1530 in6p->in6p_flags |= IN6P_RFC2292; 	\
1531 if (optval) 				\
1532 	in6p->in6p_flags |= (bit); 	\
1533 else 					\
1534 	in6p->in6p_flags &= ~(bit); 	\
1535 } while (/*CONSTCOND*/ 0)
1536 #endif
1537 
1538 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1539 
1540 			case IPV6_RECVPKTINFO:
1541 #ifdef RFC2292
1542 				/* cannot mix with RFC2292 */
1543 				if (OPTBIT(IN6P_RFC2292)) {
1544 					error = EINVAL;
1545 					break;
1546 				}
1547 #endif
1548 				OPTSET(IN6P_PKTINFO);
1549 				break;
1550 
1551 			case IPV6_HOPLIMIT:
1552 			{
1553 				struct ip6_pktopts **optp;
1554 
1555 #ifdef RFC2292
1556 				/* cannot mix with RFC2292 */
1557 				if (OPTBIT(IN6P_RFC2292)) {
1558 					error = EINVAL;
1559 					break;
1560 				}
1561 #endif
1562 				optp = &in6p->in6p_outputopts;
1563 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1564 						   (u_char *)&optval,
1565 						   sizeof(optval),
1566 						   optp,
1567 						   kauth_cred_get(), uproto);
1568 				break;
1569 			}
1570 
1571 			case IPV6_RECVHOPLIMIT:
1572 #ifdef RFC2292
1573 				/* cannot mix with RFC2292 */
1574 				if (OPTBIT(IN6P_RFC2292)) {
1575 					error = EINVAL;
1576 					break;
1577 				}
1578 #endif
1579 				OPTSET(IN6P_HOPLIMIT);
1580 				break;
1581 
1582 			case IPV6_RECVHOPOPTS:
1583 #ifdef RFC2292
1584 				/* cannot mix with RFC2292 */
1585 				if (OPTBIT(IN6P_RFC2292)) {
1586 					error = EINVAL;
1587 					break;
1588 				}
1589 #endif
1590 				OPTSET(IN6P_HOPOPTS);
1591 				break;
1592 
1593 			case IPV6_RECVDSTOPTS:
1594 #ifdef RFC2292
1595 				/* cannot mix with RFC2292 */
1596 				if (OPTBIT(IN6P_RFC2292)) {
1597 					error = EINVAL;
1598 					break;
1599 				}
1600 #endif
1601 				OPTSET(IN6P_DSTOPTS);
1602 				break;
1603 
1604 			case IPV6_RECVRTHDRDSTOPTS:
1605 #ifdef RFC2292
1606 				/* cannot mix with RFC2292 */
1607 				if (OPTBIT(IN6P_RFC2292)) {
1608 					error = EINVAL;
1609 					break;
1610 				}
1611 #endif
1612 				OPTSET(IN6P_RTHDRDSTOPTS);
1613 				break;
1614 
1615 			case IPV6_RECVRTHDR:
1616 #ifdef RFC2292
1617 				/* cannot mix with RFC2292 */
1618 				if (OPTBIT(IN6P_RFC2292)) {
1619 					error = EINVAL;
1620 					break;
1621 				}
1622 #endif
1623 				OPTSET(IN6P_RTHDR);
1624 				break;
1625 
1626 			case IPV6_FAITH:
1627 				OPTSET(IN6P_FAITH);
1628 				break;
1629 
1630 			case IPV6_RECVPATHMTU:
1631 				/*
1632 				 * We ignore this option for TCP
1633 				 * sockets.
1634 				 * (RFC3542 leaves this case
1635 				 * unspecified.)
1636 				 */
1637 				if (uproto != IPPROTO_TCP)
1638 					OPTSET(IN6P_MTU);
1639 				break;
1640 
1641 			case IPV6_V6ONLY:
1642 				/*
1643 				 * make setsockopt(IPV6_V6ONLY)
1644 				 * available only prior to bind(2).
1645 				 * see ipng mailing list, Jun 22 2001.
1646 				 */
1647 				if (in6p->in6p_lport ||
1648 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1649 					error = EINVAL;
1650 					break;
1651 				}
1652 #ifdef INET6_BINDV6ONLY
1653 				if (!optval)
1654 					error = EINVAL;
1655 #else
1656 				OPTSET(IN6P_IPV6_V6ONLY);
1657 #endif
1658 				break;
1659 			case IPV6_RECVTCLASS:
1660 #ifdef RFC2292
1661 				/* cannot mix with RFC2292 XXX */
1662 				if (OPTBIT(IN6P_RFC2292)) {
1663 					error = EINVAL;
1664 					break;
1665 				}
1666 #endif
1667 				OPTSET(IN6P_TCLASS);
1668 				break;
1669 
1670 			}
1671 			break;
1672 
1673 		case IPV6_OTCLASS:
1674 		{
1675 			struct ip6_pktopts **optp;
1676 			u_int8_t tclass;
1677 
1678 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1679 			if (error)
1680 				break;
1681 			optp = &in6p->in6p_outputopts;
1682 			error = ip6_pcbopt(optname,
1683 					   (u_char *)&tclass,
1684 					   sizeof(tclass),
1685 					   optp,
1686 					   kauth_cred_get(), uproto);
1687 			break;
1688 		}
1689 
1690 		case IPV6_TCLASS:
1691 		case IPV6_DONTFRAG:
1692 		case IPV6_USE_MIN_MTU:
1693 			error = sockopt_getint(sopt, &optval);
1694 			if (error)
1695 				break;
1696 			{
1697 				struct ip6_pktopts **optp;
1698 				optp = &in6p->in6p_outputopts;
1699 				error = ip6_pcbopt(optname,
1700 						   (u_char *)&optval,
1701 						   sizeof(optval),
1702 						   optp,
1703 						   kauth_cred_get(), uproto);
1704 				break;
1705 			}
1706 
1707 #ifdef RFC2292
1708 		case IPV6_2292PKTINFO:
1709 		case IPV6_2292HOPLIMIT:
1710 		case IPV6_2292HOPOPTS:
1711 		case IPV6_2292DSTOPTS:
1712 		case IPV6_2292RTHDR:
1713 			/* RFC 2292 */
1714 			error = sockopt_getint(sopt, &optval);
1715 			if (error)
1716 				break;
1717 
1718 			switch (optname) {
1719 			case IPV6_2292PKTINFO:
1720 				OPTSET2292(IN6P_PKTINFO);
1721 				break;
1722 			case IPV6_2292HOPLIMIT:
1723 				OPTSET2292(IN6P_HOPLIMIT);
1724 				break;
1725 			case IPV6_2292HOPOPTS:
1726 				/*
1727 				 * Check super-user privilege.
1728 				 * See comments for IPV6_RECVHOPOPTS.
1729 				 */
1730 				error =
1731 				    kauth_authorize_generic(kauth_cred_get(),
1732 				    KAUTH_GENERIC_ISSUSER, NULL);
1733 				if (error)
1734 					return (error);
1735 				OPTSET2292(IN6P_HOPOPTS);
1736 				break;
1737 			case IPV6_2292DSTOPTS:
1738 				error =
1739 				    kauth_authorize_generic(kauth_cred_get(),
1740 				    KAUTH_GENERIC_ISSUSER, NULL);
1741 				if (error)
1742 					return (error);
1743 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1744 				break;
1745 			case IPV6_2292RTHDR:
1746 				OPTSET2292(IN6P_RTHDR);
1747 				break;
1748 			}
1749 			break;
1750 #endif
1751 		case IPV6_PKTINFO:
1752 		case IPV6_HOPOPTS:
1753 		case IPV6_RTHDR:
1754 		case IPV6_DSTOPTS:
1755 		case IPV6_RTHDRDSTOPTS:
1756 		case IPV6_NEXTHOP: {
1757 			/* new advanced API (RFC3542) */
1758 			void *optbuf;
1759 			int optbuflen;
1760 			struct ip6_pktopts **optp;
1761 
1762 #ifdef RFC2292
1763 			/* cannot mix with RFC2292 */
1764 			if (OPTBIT(IN6P_RFC2292)) {
1765 				error = EINVAL;
1766 				break;
1767 			}
1768 #endif
1769 
1770 			optbuflen = sopt->sopt_size;
1771 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1772 			if (optbuf == NULL) {
1773 				error = ENOBUFS;
1774 				break;
1775 			}
1776 
1777 			sockopt_get(sopt, optbuf, optbuflen);
1778 			optp = &in6p->in6p_outputopts;
1779 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1780 			    optp, kauth_cred_get(), uproto);
1781 			break;
1782 			}
1783 #undef OPTSET
1784 
1785 		case IPV6_MULTICAST_IF:
1786 		case IPV6_MULTICAST_HOPS:
1787 		case IPV6_MULTICAST_LOOP:
1788 		case IPV6_JOIN_GROUP:
1789 		case IPV6_LEAVE_GROUP:
1790 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1791 			break;
1792 
1793 		case IPV6_PORTRANGE:
1794 			error = sockopt_getint(sopt, &optval);
1795 			if (error)
1796 				break;
1797 
1798 			switch (optval) {
1799 			case IPV6_PORTRANGE_DEFAULT:
1800 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1801 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1802 				break;
1803 
1804 			case IPV6_PORTRANGE_HIGH:
1805 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1806 				in6p->in6p_flags |= IN6P_HIGHPORT;
1807 				break;
1808 
1809 			case IPV6_PORTRANGE_LOW:
1810 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1811 				in6p->in6p_flags |= IN6P_LOWPORT;
1812 				break;
1813 
1814 			default:
1815 				error = EINVAL;
1816 				break;
1817 			}
1818 			break;
1819 
1820 
1821 #if defined(KAME_IPSEC) || defined(FAST_IPSEC)
1822 		case IPV6_IPSEC_POLICY:
1823 			error = ipsec6_set_policy(in6p, optname,
1824 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1825 			break;
1826 #endif /* IPSEC */
1827 
1828 		default:
1829 			error = ENOPROTOOPT;
1830 			break;
1831 		}
1832 		break;
1833 
1834 	case PRCO_GETOPT:
1835 		switch (optname) {
1836 #ifdef RFC2292
1837 		case IPV6_2292PKTOPTIONS:
1838 			/*
1839 			 * RFC3542 (effectively) deprecated the
1840 			 * semantics of the 2292-style pktoptions.
1841 			 * Since it was not reliable in nature (i.e.,
1842 			 * applications had to expect the lack of some
1843 			 * information after all), it would make sense
1844 			 * to simplify this part by always returning
1845 			 * empty data.
1846 			 */
1847 			break;
1848 #endif
1849 
1850 		case IPV6_RECVHOPOPTS:
1851 		case IPV6_RECVDSTOPTS:
1852 		case IPV6_RECVRTHDRDSTOPTS:
1853 		case IPV6_UNICAST_HOPS:
1854 		case IPV6_RECVPKTINFO:
1855 		case IPV6_RECVHOPLIMIT:
1856 		case IPV6_RECVRTHDR:
1857 		case IPV6_RECVPATHMTU:
1858 
1859 		case IPV6_FAITH:
1860 		case IPV6_V6ONLY:
1861 		case IPV6_PORTRANGE:
1862 		case IPV6_RECVTCLASS:
1863 			switch (optname) {
1864 
1865 			case IPV6_RECVHOPOPTS:
1866 				optval = OPTBIT(IN6P_HOPOPTS);
1867 				break;
1868 
1869 			case IPV6_RECVDSTOPTS:
1870 				optval = OPTBIT(IN6P_DSTOPTS);
1871 				break;
1872 
1873 			case IPV6_RECVRTHDRDSTOPTS:
1874 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1875 				break;
1876 
1877 			case IPV6_UNICAST_HOPS:
1878 				optval = in6p->in6p_hops;
1879 				break;
1880 
1881 			case IPV6_RECVPKTINFO:
1882 				optval = OPTBIT(IN6P_PKTINFO);
1883 				break;
1884 
1885 			case IPV6_RECVHOPLIMIT:
1886 				optval = OPTBIT(IN6P_HOPLIMIT);
1887 				break;
1888 
1889 			case IPV6_RECVRTHDR:
1890 				optval = OPTBIT(IN6P_RTHDR);
1891 				break;
1892 
1893 			case IPV6_RECVPATHMTU:
1894 				optval = OPTBIT(IN6P_MTU);
1895 				break;
1896 
1897 			case IPV6_FAITH:
1898 				optval = OPTBIT(IN6P_FAITH);
1899 				break;
1900 
1901 			case IPV6_V6ONLY:
1902 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1903 				break;
1904 
1905 			case IPV6_PORTRANGE:
1906 			    {
1907 				int flags;
1908 				flags = in6p->in6p_flags;
1909 				if (flags & IN6P_HIGHPORT)
1910 					optval = IPV6_PORTRANGE_HIGH;
1911 				else if (flags & IN6P_LOWPORT)
1912 					optval = IPV6_PORTRANGE_LOW;
1913 				else
1914 					optval = 0;
1915 				break;
1916 			    }
1917 			case IPV6_RECVTCLASS:
1918 				optval = OPTBIT(IN6P_TCLASS);
1919 				break;
1920 
1921 			}
1922 			if (error)
1923 				break;
1924 			error = sockopt_setint(sopt, optval);
1925 			break;
1926 
1927 		case IPV6_PATHMTU:
1928 		    {
1929 			u_long pmtu = 0;
1930 			struct ip6_mtuinfo mtuinfo;
1931 			struct route *ro = &in6p->in6p_route;
1932 
1933 			if (!(so->so_state & SS_ISCONNECTED))
1934 				return (ENOTCONN);
1935 			/*
1936 			 * XXX: we dot not consider the case of source
1937 			 * routing, or optional information to specify
1938 			 * the outgoing interface.
1939 			 */
1940 			error = ip6_getpmtu(ro, NULL, NULL,
1941 			    &in6p->in6p_faddr, &pmtu, NULL);
1942 			if (error)
1943 				break;
1944 			if (pmtu > IPV6_MAXPACKET)
1945 				pmtu = IPV6_MAXPACKET;
1946 
1947 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1948 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1949 			optdata = (void *)&mtuinfo;
1950 			optdatalen = sizeof(mtuinfo);
1951 			if (optdatalen > MCLBYTES)
1952 				return (EMSGSIZE); /* XXX */
1953 			error = sockopt_set(sopt, optdata, optdatalen);
1954 			break;
1955 		    }
1956 
1957 #ifdef RFC2292
1958 		case IPV6_2292PKTINFO:
1959 		case IPV6_2292HOPLIMIT:
1960 		case IPV6_2292HOPOPTS:
1961 		case IPV6_2292RTHDR:
1962 		case IPV6_2292DSTOPTS:
1963 			switch (optname) {
1964 			case IPV6_2292PKTINFO:
1965 				optval = OPTBIT(IN6P_PKTINFO);
1966 				break;
1967 			case IPV6_2292HOPLIMIT:
1968 				optval = OPTBIT(IN6P_HOPLIMIT);
1969 				break;
1970 			case IPV6_2292HOPOPTS:
1971 				optval = OPTBIT(IN6P_HOPOPTS);
1972 				break;
1973 			case IPV6_2292RTHDR:
1974 				optval = OPTBIT(IN6P_RTHDR);
1975 				break;
1976 			case IPV6_2292DSTOPTS:
1977 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1978 				break;
1979 			}
1980 			error = sockopt_setint(sopt, optval);
1981 			break;
1982 #endif
1983 		case IPV6_PKTINFO:
1984 		case IPV6_HOPOPTS:
1985 		case IPV6_RTHDR:
1986 		case IPV6_DSTOPTS:
1987 		case IPV6_RTHDRDSTOPTS:
1988 		case IPV6_NEXTHOP:
1989 		case IPV6_OTCLASS:
1990 		case IPV6_TCLASS:
1991 		case IPV6_DONTFRAG:
1992 		case IPV6_USE_MIN_MTU:
1993 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1994 			    optname, sopt);
1995 			break;
1996 
1997 		case IPV6_MULTICAST_IF:
1998 		case IPV6_MULTICAST_HOPS:
1999 		case IPV6_MULTICAST_LOOP:
2000 		case IPV6_JOIN_GROUP:
2001 		case IPV6_LEAVE_GROUP:
2002 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
2003 			break;
2004 
2005 #if defined(KAME_IPSEC) || defined(FAST_IPSEC)
2006 		case IPV6_IPSEC_POLICY:
2007 		    {
2008 			struct mbuf *m = NULL;
2009 
2010 			/* XXX this will return EINVAL as sopt is empty */
2011 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
2012 			    sopt->sopt_size, &m);
2013 			if (!error)
2014 				error = sockopt_setmbuf(sopt, m);
2015 
2016 			break;
2017 		    }
2018 #endif /* IPSEC */
2019 
2020 		default:
2021 			error = ENOPROTOOPT;
2022 			break;
2023 		}
2024 		break;
2025 	}
2026 	return (error);
2027 }
2028 
2029 int
2030 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2031 {
2032 	int error = 0, optval;
2033 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2034 	struct in6pcb *in6p = sotoin6pcb(so);
2035 	int level, optname;
2036 
2037 	KASSERT(sopt != NULL);
2038 
2039 	level = sopt->sopt_level;
2040 	optname = sopt->sopt_name;
2041 
2042 	if (level != IPPROTO_IPV6) {
2043 		return ENOPROTOOPT;
2044 	}
2045 
2046 	switch (optname) {
2047 	case IPV6_CHECKSUM:
2048 		/*
2049 		 * For ICMPv6 sockets, no modification allowed for checksum
2050 		 * offset, permit "no change" values to help existing apps.
2051 		 *
2052 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
2053 		 * for an ICMPv6 socket will fail."  The current
2054 		 * behavior does not meet RFC3542.
2055 		 */
2056 		switch (op) {
2057 		case PRCO_SETOPT:
2058 			error = sockopt_getint(sopt, &optval);
2059 			if (error)
2060 				break;
2061 			if ((optval % 2) != 0) {
2062 				/* the API assumes even offset values */
2063 				error = EINVAL;
2064 			} else if (so->so_proto->pr_protocol ==
2065 			    IPPROTO_ICMPV6) {
2066 				if (optval != icmp6off)
2067 					error = EINVAL;
2068 			} else
2069 				in6p->in6p_cksum = optval;
2070 			break;
2071 
2072 		case PRCO_GETOPT:
2073 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2074 				optval = icmp6off;
2075 			else
2076 				optval = in6p->in6p_cksum;
2077 
2078 			error = sockopt_setint(sopt, optval);
2079 			break;
2080 
2081 		default:
2082 			error = EINVAL;
2083 			break;
2084 		}
2085 		break;
2086 
2087 	default:
2088 		error = ENOPROTOOPT;
2089 		break;
2090 	}
2091 
2092 	return (error);
2093 }
2094 
2095 #ifdef RFC2292
2096 /*
2097  * Set up IP6 options in pcb for insertion in output packets or
2098  * specifying behavior of outgoing packets.
2099  */
2100 static int
2101 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
2102     struct sockopt *sopt)
2103 {
2104 	struct ip6_pktopts *opt = *pktopt;
2105 	struct mbuf *m;
2106 	int error = 0;
2107 
2108 	/* turn off any old options. */
2109 	if (opt) {
2110 #ifdef DIAGNOSTIC
2111 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2112 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2113 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2114 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2115 #endif
2116 		ip6_clearpktopts(opt, -1);
2117 	} else {
2118 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2119 		if (opt == NULL)
2120 			return (ENOBUFS);
2121 	}
2122 	*pktopt = NULL;
2123 
2124 	if (sopt == NULL || sopt->sopt_size == 0) {
2125 		/*
2126 		 * Only turning off any previous options, regardless of
2127 		 * whether the opt is just created or given.
2128 		 */
2129 		free(opt, M_IP6OPT);
2130 		return (0);
2131 	}
2132 
2133 	/*  set options specified by user. */
2134 	m = sockopt_getmbuf(sopt);
2135 	if (m == NULL) {
2136 		free(opt, M_IP6OPT);
2137 		return (ENOBUFS);
2138 	}
2139 
2140 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2141 	    so->so_proto->pr_protocol);
2142 	m_freem(m);
2143 	if (error != 0) {
2144 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2145 		free(opt, M_IP6OPT);
2146 		return (error);
2147 	}
2148 	*pktopt = opt;
2149 	return (0);
2150 }
2151 #endif
2152 
2153 /*
2154  * initialize ip6_pktopts.  beware that there are non-zero default values in
2155  * the struct.
2156  */
2157 void
2158 ip6_initpktopts(struct ip6_pktopts *opt)
2159 {
2160 
2161 	memset(opt, 0, sizeof(*opt));
2162 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2163 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2164 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2165 }
2166 
2167 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2168 static int
2169 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2170     kauth_cred_t cred, int uproto)
2171 {
2172 	struct ip6_pktopts *opt;
2173 
2174 	if (*pktopt == NULL) {
2175 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2176 		    M_NOWAIT);
2177 		if (*pktopt == NULL)
2178 			return (ENOBUFS);
2179 
2180 		ip6_initpktopts(*pktopt);
2181 	}
2182 	opt = *pktopt;
2183 
2184 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2185 }
2186 
2187 static int
2188 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2189 {
2190 	void *optdata = NULL;
2191 	int optdatalen = 0;
2192 	struct ip6_ext *ip6e;
2193 	int error = 0;
2194 	struct in6_pktinfo null_pktinfo;
2195 	int deftclass = 0, on;
2196 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2197 
2198 	switch (optname) {
2199 	case IPV6_PKTINFO:
2200 		if (pktopt && pktopt->ip6po_pktinfo)
2201 			optdata = (void *)pktopt->ip6po_pktinfo;
2202 		else {
2203 			/* XXX: we don't have to do this every time... */
2204 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2205 			optdata = (void *)&null_pktinfo;
2206 		}
2207 		optdatalen = sizeof(struct in6_pktinfo);
2208 		break;
2209 	case IPV6_OTCLASS:
2210 		/* XXX */
2211 		return (EINVAL);
2212 	case IPV6_TCLASS:
2213 		if (pktopt && pktopt->ip6po_tclass >= 0)
2214 			optdata = (void *)&pktopt->ip6po_tclass;
2215 		else
2216 			optdata = (void *)&deftclass;
2217 		optdatalen = sizeof(int);
2218 		break;
2219 	case IPV6_HOPOPTS:
2220 		if (pktopt && pktopt->ip6po_hbh) {
2221 			optdata = (void *)pktopt->ip6po_hbh;
2222 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2223 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2224 		}
2225 		break;
2226 	case IPV6_RTHDR:
2227 		if (pktopt && pktopt->ip6po_rthdr) {
2228 			optdata = (void *)pktopt->ip6po_rthdr;
2229 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2230 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2231 		}
2232 		break;
2233 	case IPV6_RTHDRDSTOPTS:
2234 		if (pktopt && pktopt->ip6po_dest1) {
2235 			optdata = (void *)pktopt->ip6po_dest1;
2236 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2237 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2238 		}
2239 		break;
2240 	case IPV6_DSTOPTS:
2241 		if (pktopt && pktopt->ip6po_dest2) {
2242 			optdata = (void *)pktopt->ip6po_dest2;
2243 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2244 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2245 		}
2246 		break;
2247 	case IPV6_NEXTHOP:
2248 		if (pktopt && pktopt->ip6po_nexthop) {
2249 			optdata = (void *)pktopt->ip6po_nexthop;
2250 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2251 		}
2252 		break;
2253 	case IPV6_USE_MIN_MTU:
2254 		if (pktopt)
2255 			optdata = (void *)&pktopt->ip6po_minmtu;
2256 		else
2257 			optdata = (void *)&defminmtu;
2258 		optdatalen = sizeof(int);
2259 		break;
2260 	case IPV6_DONTFRAG:
2261 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2262 			on = 1;
2263 		else
2264 			on = 0;
2265 		optdata = (void *)&on;
2266 		optdatalen = sizeof(on);
2267 		break;
2268 	default:		/* should not happen */
2269 #ifdef DIAGNOSTIC
2270 		panic("ip6_getpcbopt: unexpected option\n");
2271 #endif
2272 		return (ENOPROTOOPT);
2273 	}
2274 
2275 	error = sockopt_set(sopt, optdata, optdatalen);
2276 
2277 	return (error);
2278 }
2279 
2280 void
2281 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2282 {
2283 	if (optname == -1 || optname == IPV6_PKTINFO) {
2284 		if (pktopt->ip6po_pktinfo)
2285 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2286 		pktopt->ip6po_pktinfo = NULL;
2287 	}
2288 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2289 		pktopt->ip6po_hlim = -1;
2290 	if (optname == -1 || optname == IPV6_TCLASS)
2291 		pktopt->ip6po_tclass = -1;
2292 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2293 		rtcache_free(&pktopt->ip6po_nextroute);
2294 		if (pktopt->ip6po_nexthop)
2295 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2296 		pktopt->ip6po_nexthop = NULL;
2297 	}
2298 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2299 		if (pktopt->ip6po_hbh)
2300 			free(pktopt->ip6po_hbh, M_IP6OPT);
2301 		pktopt->ip6po_hbh = NULL;
2302 	}
2303 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2304 		if (pktopt->ip6po_dest1)
2305 			free(pktopt->ip6po_dest1, M_IP6OPT);
2306 		pktopt->ip6po_dest1 = NULL;
2307 	}
2308 	if (optname == -1 || optname == IPV6_RTHDR) {
2309 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2310 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2311 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2312 		rtcache_free(&pktopt->ip6po_route);
2313 	}
2314 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2315 		if (pktopt->ip6po_dest2)
2316 			free(pktopt->ip6po_dest2, M_IP6OPT);
2317 		pktopt->ip6po_dest2 = NULL;
2318 	}
2319 }
2320 
2321 #define PKTOPT_EXTHDRCPY(type) 					\
2322 do {								\
2323 	if (src->type) {					\
2324 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2325 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2326 		if (dst->type == NULL)				\
2327 			goto bad;				\
2328 		memcpy(dst->type, src->type, hlen);		\
2329 	}							\
2330 } while (/*CONSTCOND*/ 0)
2331 
2332 static int
2333 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2334 {
2335 	dst->ip6po_hlim = src->ip6po_hlim;
2336 	dst->ip6po_tclass = src->ip6po_tclass;
2337 	dst->ip6po_flags = src->ip6po_flags;
2338 	if (src->ip6po_pktinfo) {
2339 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2340 		    M_IP6OPT, canwait);
2341 		if (dst->ip6po_pktinfo == NULL)
2342 			goto bad;
2343 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2344 	}
2345 	if (src->ip6po_nexthop) {
2346 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2347 		    M_IP6OPT, canwait);
2348 		if (dst->ip6po_nexthop == NULL)
2349 			goto bad;
2350 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2351 		    src->ip6po_nexthop->sa_len);
2352 	}
2353 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2354 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2355 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2356 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2357 	return (0);
2358 
2359   bad:
2360 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2361 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2362 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2363 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2364 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2365 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2366 
2367 	return (ENOBUFS);
2368 }
2369 #undef PKTOPT_EXTHDRCPY
2370 
2371 struct ip6_pktopts *
2372 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2373 {
2374 	int error;
2375 	struct ip6_pktopts *dst;
2376 
2377 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2378 	if (dst == NULL)
2379 		return (NULL);
2380 	ip6_initpktopts(dst);
2381 
2382 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2383 		free(dst, M_IP6OPT);
2384 		return (NULL);
2385 	}
2386 
2387 	return (dst);
2388 }
2389 
2390 void
2391 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2392 {
2393 	if (pktopt == NULL)
2394 		return;
2395 
2396 	ip6_clearpktopts(pktopt, -1);
2397 
2398 	free(pktopt, M_IP6OPT);
2399 }
2400 
2401 /*
2402  * Set the IP6 multicast options in response to user setsockopt().
2403  */
2404 static int
2405 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2406 {
2407 	int error = 0;
2408 	u_int loop, ifindex;
2409 	struct ipv6_mreq mreq;
2410 	struct ifnet *ifp;
2411 	struct ip6_moptions *im6o = *im6op;
2412 	struct route ro;
2413 	struct in6_multi_mship *imm;
2414 	struct lwp *l = curlwp;	/* XXX */
2415 
2416 	if (im6o == NULL) {
2417 		/*
2418 		 * No multicast option buffer attached to the pcb;
2419 		 * allocate one and initialize to default values.
2420 		 */
2421 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2422 		if (im6o == NULL)
2423 			return (ENOBUFS);
2424 
2425 		*im6op = im6o;
2426 		im6o->im6o_multicast_ifp = NULL;
2427 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2428 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2429 		LIST_INIT(&im6o->im6o_memberships);
2430 	}
2431 
2432 	switch (sopt->sopt_name) {
2433 
2434 	case IPV6_MULTICAST_IF:
2435 		/*
2436 		 * Select the interface for outgoing multicast packets.
2437 		 */
2438 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2439 		if (error != 0)
2440 			break;
2441 
2442 		if (ifindex != 0) {
2443 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2444 				error = ENXIO;	/* XXX EINVAL? */
2445 				break;
2446 			}
2447 			ifp = ifindex2ifnet[ifindex];
2448 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2449 				error = EADDRNOTAVAIL;
2450 				break;
2451 			}
2452 		} else
2453 			ifp = NULL;
2454 		im6o->im6o_multicast_ifp = ifp;
2455 		break;
2456 
2457 	case IPV6_MULTICAST_HOPS:
2458 	    {
2459 		/*
2460 		 * Set the IP6 hoplimit for outgoing multicast packets.
2461 		 */
2462 		int optval;
2463 
2464 		error = sockopt_getint(sopt, &optval);
2465 		if (error != 0)
2466 			break;
2467 
2468 		if (optval < -1 || optval >= 256)
2469 			error = EINVAL;
2470 		else if (optval == -1)
2471 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2472 		else
2473 			im6o->im6o_multicast_hlim = optval;
2474 		break;
2475 	    }
2476 
2477 	case IPV6_MULTICAST_LOOP:
2478 		/*
2479 		 * Set the loopback flag for outgoing multicast packets.
2480 		 * Must be zero or one.
2481 		 */
2482 		error = sockopt_get(sopt, &loop, sizeof(loop));
2483 		if (error != 0)
2484 			break;
2485 		if (loop > 1) {
2486 			error = EINVAL;
2487 			break;
2488 		}
2489 		im6o->im6o_multicast_loop = loop;
2490 		break;
2491 
2492 	case IPV6_JOIN_GROUP:
2493 		/*
2494 		 * Add a multicast group membership.
2495 		 * Group must be a valid IP6 multicast address.
2496 		 */
2497 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2498 		if (error != 0)
2499 			break;
2500 
2501 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2502 			/*
2503 			 * We use the unspecified address to specify to accept
2504 			 * all multicast addresses. Only super user is allowed
2505 			 * to do this.
2506 			 */
2507 			if (kauth_authorize_generic(l->l_cred,
2508 			    KAUTH_GENERIC_ISSUSER, NULL))
2509 			{
2510 				error = EACCES;
2511 				break;
2512 			}
2513 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2514 			error = EINVAL;
2515 			break;
2516 		}
2517 
2518 		/*
2519 		 * If no interface was explicitly specified, choose an
2520 		 * appropriate one according to the given multicast address.
2521 		 */
2522 		if (mreq.ipv6mr_interface == 0) {
2523 			struct rtentry *rt;
2524 			union {
2525 				struct sockaddr		dst;
2526 				struct sockaddr_in6	dst6;
2527 			} u;
2528 
2529 			/*
2530 			 * Look up the routing table for the
2531 			 * address, and choose the outgoing interface.
2532 			 *   XXX: is it a good approach?
2533 			 */
2534 			memset(&ro, 0, sizeof(ro));
2535 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2536 			    0, 0);
2537 			rtcache_setdst(&ro, &u.dst);
2538 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2539 			                                        : NULL;
2540 			rtcache_free(&ro);
2541 		} else {
2542 			/*
2543 			 * If the interface is specified, validate it.
2544 			 */
2545 			if (if_indexlim <= mreq.ipv6mr_interface ||
2546 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2547 				error = ENXIO;	/* XXX EINVAL? */
2548 				break;
2549 			}
2550 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2551 		}
2552 
2553 		/*
2554 		 * See if we found an interface, and confirm that it
2555 		 * supports multicast
2556 		 */
2557 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2558 			error = EADDRNOTAVAIL;
2559 			break;
2560 		}
2561 
2562 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2563 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2564 			break;
2565 		}
2566 
2567 		/*
2568 		 * See if the membership already exists.
2569 		 */
2570 		for (imm = im6o->im6o_memberships.lh_first;
2571 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2572 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2573 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2574 			    &mreq.ipv6mr_multiaddr))
2575 				break;
2576 		if (imm != NULL) {
2577 			error = EADDRINUSE;
2578 			break;
2579 		}
2580 		/*
2581 		 * Everything looks good; add a new record to the multicast
2582 		 * address list for the given interface.
2583 		 */
2584 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2585 		if (imm == NULL)
2586 			break;
2587 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2588 		break;
2589 
2590 	case IPV6_LEAVE_GROUP:
2591 		/*
2592 		 * Drop a multicast group membership.
2593 		 * Group must be a valid IP6 multicast address.
2594 		 */
2595 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2596 		if (error != 0)
2597 			break;
2598 
2599 		/*
2600 		 * If an interface address was specified, get a pointer
2601 		 * to its ifnet structure.
2602 		 */
2603 		if (mreq.ipv6mr_interface != 0) {
2604 			if (if_indexlim <= mreq.ipv6mr_interface ||
2605 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2606 				error = ENXIO;	/* XXX EINVAL? */
2607 				break;
2608 			}
2609 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2610 		} else
2611 			ifp = NULL;
2612 
2613 		/* Fill in the scope zone ID */
2614 		if (ifp) {
2615 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2616 				/* XXX: should not happen */
2617 				error = EADDRNOTAVAIL;
2618 				break;
2619 			}
2620 		} else if (mreq.ipv6mr_interface != 0) {
2621 			/*
2622 			 * XXX: This case would happens when the (positive)
2623 			 * index is in the valid range, but the corresponding
2624 			 * interface has been detached dynamically.  The above
2625 			 * check probably avoids such case to happen here, but
2626 			 * we check it explicitly for safety.
2627 			 */
2628 			error = EADDRNOTAVAIL;
2629 			break;
2630 		} else {	/* ipv6mr_interface == 0 */
2631 			struct sockaddr_in6 sa6_mc;
2632 
2633 			/*
2634 			 * The API spec says as follows:
2635 			 *  If the interface index is specified as 0, the
2636 			 *  system may choose a multicast group membership to
2637 			 *  drop by matching the multicast address only.
2638 			 * On the other hand, we cannot disambiguate the scope
2639 			 * zone unless an interface is provided.  Thus, we
2640 			 * check if there's ambiguity with the default scope
2641 			 * zone as the last resort.
2642 			 */
2643 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2644 			    0, 0, 0);
2645 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2646 			if (error != 0)
2647 				break;
2648 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2649 		}
2650 
2651 		/*
2652 		 * Find the membership in the membership list.
2653 		 */
2654 		for (imm = im6o->im6o_memberships.lh_first;
2655 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2656 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2657 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2658 			    &mreq.ipv6mr_multiaddr))
2659 				break;
2660 		}
2661 		if (imm == NULL) {
2662 			/* Unable to resolve interface */
2663 			error = EADDRNOTAVAIL;
2664 			break;
2665 		}
2666 		/*
2667 		 * Give up the multicast address record to which the
2668 		 * membership points.
2669 		 */
2670 		LIST_REMOVE(imm, i6mm_chain);
2671 		in6_leavegroup(imm);
2672 		break;
2673 
2674 	default:
2675 		error = EOPNOTSUPP;
2676 		break;
2677 	}
2678 
2679 	/*
2680 	 * If all options have default values, no need to keep the mbuf.
2681 	 */
2682 	if (im6o->im6o_multicast_ifp == NULL &&
2683 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2684 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2685 	    im6o->im6o_memberships.lh_first == NULL) {
2686 		free(*im6op, M_IPMOPTS);
2687 		*im6op = NULL;
2688 	}
2689 
2690 	return (error);
2691 }
2692 
2693 /*
2694  * Return the IP6 multicast options in response to user getsockopt().
2695  */
2696 static int
2697 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2698 {
2699 	u_int optval;
2700 	int error;
2701 
2702 	switch (sopt->sopt_name) {
2703 	case IPV6_MULTICAST_IF:
2704 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2705 			optval = 0;
2706 		else
2707 			optval = im6o->im6o_multicast_ifp->if_index;
2708 
2709 		error = sockopt_set(sopt, &optval, sizeof(optval));
2710 		break;
2711 
2712 	case IPV6_MULTICAST_HOPS:
2713 		if (im6o == NULL)
2714 			optval = ip6_defmcasthlim;
2715 		else
2716 			optval = im6o->im6o_multicast_hlim;
2717 
2718 		error = sockopt_set(sopt, &optval, sizeof(optval));
2719 		break;
2720 
2721 	case IPV6_MULTICAST_LOOP:
2722 		if (im6o == NULL)
2723 			optval = ip6_defmcasthlim;
2724 		else
2725 			optval = im6o->im6o_multicast_loop;
2726 
2727 		error = sockopt_set(sopt, &optval, sizeof(optval));
2728 		break;
2729 
2730 	default:
2731 		error = EOPNOTSUPP;
2732 	}
2733 
2734 	return (error);
2735 }
2736 
2737 /*
2738  * Discard the IP6 multicast options.
2739  */
2740 void
2741 ip6_freemoptions(struct ip6_moptions *im6o)
2742 {
2743 	struct in6_multi_mship *imm;
2744 
2745 	if (im6o == NULL)
2746 		return;
2747 
2748 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2749 		LIST_REMOVE(imm, i6mm_chain);
2750 		in6_leavegroup(imm);
2751 	}
2752 	free(im6o, M_IPMOPTS);
2753 }
2754 
2755 /*
2756  * Set IPv6 outgoing packet options based on advanced API.
2757  */
2758 int
2759 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2760 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2761 {
2762 	struct cmsghdr *cm = 0;
2763 
2764 	if (control == NULL || opt == NULL)
2765 		return (EINVAL);
2766 
2767 	ip6_initpktopts(opt);
2768 	if (stickyopt) {
2769 		int error;
2770 
2771 		/*
2772 		 * If stickyopt is provided, make a local copy of the options
2773 		 * for this particular packet, then override them by ancillary
2774 		 * objects.
2775 		 * XXX: copypktopts() does not copy the cached route to a next
2776 		 * hop (if any).  This is not very good in terms of efficiency,
2777 		 * but we can allow this since this option should be rarely
2778 		 * used.
2779 		 */
2780 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2781 			return (error);
2782 	}
2783 
2784 	/*
2785 	 * XXX: Currently, we assume all the optional information is stored
2786 	 * in a single mbuf.
2787 	 */
2788 	if (control->m_next)
2789 		return (EINVAL);
2790 
2791 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2792 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2793 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2794 		int error;
2795 
2796 		if (control->m_len < CMSG_LEN(0))
2797 			return (EINVAL);
2798 
2799 		cm = mtod(control, struct cmsghdr *);
2800 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2801 			return (EINVAL);
2802 		if (cm->cmsg_level != IPPROTO_IPV6)
2803 			continue;
2804 
2805 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2806 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2807 		if (error)
2808 			return (error);
2809 	}
2810 
2811 	return (0);
2812 }
2813 
2814 /*
2815  * Set a particular packet option, as a sticky option or an ancillary data
2816  * item.  "len" can be 0 only when it's a sticky option.
2817  * We have 4 cases of combination of "sticky" and "cmsg":
2818  * "sticky=0, cmsg=0": impossible
2819  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2820  * "sticky=1, cmsg=0": RFC3542 socket option
2821  * "sticky=1, cmsg=1": RFC2292 socket option
2822  */
2823 static int
2824 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2825     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2826 {
2827 	int minmtupolicy;
2828 	int error;
2829 
2830 	if (!sticky && !cmsg) {
2831 #ifdef DIAGNOSTIC
2832 		printf("ip6_setpktopt: impossible case\n");
2833 #endif
2834 		return (EINVAL);
2835 	}
2836 
2837 	/*
2838 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2839 	 * not be specified in the context of RFC3542.  Conversely,
2840 	 * RFC3542 types should not be specified in the context of RFC2292.
2841 	 */
2842 	if (!cmsg) {
2843 		switch (optname) {
2844 		case IPV6_2292PKTINFO:
2845 		case IPV6_2292HOPLIMIT:
2846 		case IPV6_2292NEXTHOP:
2847 		case IPV6_2292HOPOPTS:
2848 		case IPV6_2292DSTOPTS:
2849 		case IPV6_2292RTHDR:
2850 		case IPV6_2292PKTOPTIONS:
2851 			return (ENOPROTOOPT);
2852 		}
2853 	}
2854 	if (sticky && cmsg) {
2855 		switch (optname) {
2856 		case IPV6_PKTINFO:
2857 		case IPV6_HOPLIMIT:
2858 		case IPV6_NEXTHOP:
2859 		case IPV6_HOPOPTS:
2860 		case IPV6_DSTOPTS:
2861 		case IPV6_RTHDRDSTOPTS:
2862 		case IPV6_RTHDR:
2863 		case IPV6_USE_MIN_MTU:
2864 		case IPV6_DONTFRAG:
2865 		case IPV6_OTCLASS:
2866 		case IPV6_TCLASS:
2867 			return (ENOPROTOOPT);
2868 		}
2869 	}
2870 
2871 	switch (optname) {
2872 #ifdef RFC2292
2873 	case IPV6_2292PKTINFO:
2874 #endif
2875 	case IPV6_PKTINFO:
2876 	{
2877 		struct ifnet *ifp = NULL;
2878 		struct in6_pktinfo *pktinfo;
2879 
2880 		if (len != sizeof(struct in6_pktinfo))
2881 			return (EINVAL);
2882 
2883 		pktinfo = (struct in6_pktinfo *)buf;
2884 
2885 		/*
2886 		 * An application can clear any sticky IPV6_PKTINFO option by
2887 		 * doing a "regular" setsockopt with ipi6_addr being
2888 		 * in6addr_any and ipi6_ifindex being zero.
2889 		 * [RFC 3542, Section 6]
2890 		 */
2891 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2892 		    pktinfo->ipi6_ifindex == 0 &&
2893 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2894 			ip6_clearpktopts(opt, optname);
2895 			break;
2896 		}
2897 
2898 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2899 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2900 			return (EINVAL);
2901 		}
2902 
2903 		/* validate the interface index if specified. */
2904 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2905 			 return (ENXIO);
2906 		}
2907 		if (pktinfo->ipi6_ifindex) {
2908 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2909 			if (ifp == NULL)
2910 				return (ENXIO);
2911 		}
2912 
2913 		/*
2914 		 * We store the address anyway, and let in6_selectsrc()
2915 		 * validate the specified address.  This is because ipi6_addr
2916 		 * may not have enough information about its scope zone, and
2917 		 * we may need additional information (such as outgoing
2918 		 * interface or the scope zone of a destination address) to
2919 		 * disambiguate the scope.
2920 		 * XXX: the delay of the validation may confuse the
2921 		 * application when it is used as a sticky option.
2922 		 */
2923 		if (opt->ip6po_pktinfo == NULL) {
2924 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2925 			    M_IP6OPT, M_NOWAIT);
2926 			if (opt->ip6po_pktinfo == NULL)
2927 				return (ENOBUFS);
2928 		}
2929 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2930 		break;
2931 	}
2932 
2933 #ifdef RFC2292
2934 	case IPV6_2292HOPLIMIT:
2935 #endif
2936 	case IPV6_HOPLIMIT:
2937 	{
2938 		int *hlimp;
2939 
2940 		/*
2941 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2942 		 * to simplify the ordering among hoplimit options.
2943 		 */
2944 		if (optname == IPV6_HOPLIMIT && sticky)
2945 			return (ENOPROTOOPT);
2946 
2947 		if (len != sizeof(int))
2948 			return (EINVAL);
2949 		hlimp = (int *)buf;
2950 		if (*hlimp < -1 || *hlimp > 255)
2951 			return (EINVAL);
2952 
2953 		opt->ip6po_hlim = *hlimp;
2954 		break;
2955 	}
2956 
2957 	case IPV6_OTCLASS:
2958 		if (len != sizeof(u_int8_t))
2959 			return (EINVAL);
2960 
2961 		opt->ip6po_tclass = *(u_int8_t *)buf;
2962 		break;
2963 
2964 	case IPV6_TCLASS:
2965 	{
2966 		int tclass;
2967 
2968 		if (len != sizeof(int))
2969 			return (EINVAL);
2970 		tclass = *(int *)buf;
2971 		if (tclass < -1 || tclass > 255)
2972 			return (EINVAL);
2973 
2974 		opt->ip6po_tclass = tclass;
2975 		break;
2976 	}
2977 
2978 #ifdef RFC2292
2979 	case IPV6_2292NEXTHOP:
2980 #endif
2981 	case IPV6_NEXTHOP:
2982 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
2983 		    NULL);
2984 		if (error)
2985 			return (error);
2986 
2987 		if (len == 0) {	/* just remove the option */
2988 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2989 			break;
2990 		}
2991 
2992 		/* check if cmsg_len is large enough for sa_len */
2993 		if (len < sizeof(struct sockaddr) || len < *buf)
2994 			return (EINVAL);
2995 
2996 		switch (((struct sockaddr *)buf)->sa_family) {
2997 		case AF_INET6:
2998 		{
2999 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3000 
3001 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3002 				return (EINVAL);
3003 
3004 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3005 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3006 				return (EINVAL);
3007 			}
3008 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3009 			    != 0) {
3010 				return (error);
3011 			}
3012 			break;
3013 		}
3014 		case AF_LINK:	/* eventually be supported? */
3015 		default:
3016 			return (EAFNOSUPPORT);
3017 		}
3018 
3019 		/* turn off the previous option, then set the new option. */
3020 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3021 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3022 		if (opt->ip6po_nexthop == NULL)
3023 			return (ENOBUFS);
3024 		memcpy(opt->ip6po_nexthop, buf, *buf);
3025 		break;
3026 
3027 #ifdef RFC2292
3028 	case IPV6_2292HOPOPTS:
3029 #endif
3030 	case IPV6_HOPOPTS:
3031 	{
3032 		struct ip6_hbh *hbh;
3033 		int hbhlen;
3034 
3035 		/*
3036 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3037 		 * options, since per-option restriction has too much
3038 		 * overhead.
3039 		 */
3040 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
3041 		    NULL);
3042 		if (error)
3043 			return (error);
3044 
3045 		if (len == 0) {
3046 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3047 			break;	/* just remove the option */
3048 		}
3049 
3050 		/* message length validation */
3051 		if (len < sizeof(struct ip6_hbh))
3052 			return (EINVAL);
3053 		hbh = (struct ip6_hbh *)buf;
3054 		hbhlen = (hbh->ip6h_len + 1) << 3;
3055 		if (len != hbhlen)
3056 			return (EINVAL);
3057 
3058 		/* turn off the previous option, then set the new option. */
3059 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3060 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3061 		if (opt->ip6po_hbh == NULL)
3062 			return (ENOBUFS);
3063 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3064 
3065 		break;
3066 	}
3067 
3068 #ifdef RFC2292
3069 	case IPV6_2292DSTOPTS:
3070 #endif
3071 	case IPV6_DSTOPTS:
3072 	case IPV6_RTHDRDSTOPTS:
3073 	{
3074 		struct ip6_dest *dest, **newdest = NULL;
3075 		int destlen;
3076 
3077 		/* XXX: see the comment for IPV6_HOPOPTS */
3078 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
3079 		    NULL);
3080 		if (error)
3081 			return (error);
3082 
3083 		if (len == 0) {
3084 			ip6_clearpktopts(opt, optname);
3085 			break;	/* just remove the option */
3086 		}
3087 
3088 		/* message length validation */
3089 		if (len < sizeof(struct ip6_dest))
3090 			return (EINVAL);
3091 		dest = (struct ip6_dest *)buf;
3092 		destlen = (dest->ip6d_len + 1) << 3;
3093 		if (len != destlen)
3094 			return (EINVAL);
3095 		/*
3096 		 * Determine the position that the destination options header
3097 		 * should be inserted; before or after the routing header.
3098 		 */
3099 		switch (optname) {
3100 		case IPV6_2292DSTOPTS:
3101 			/*
3102 			 * The old advanced API is ambiguous on this point.
3103 			 * Our approach is to determine the position based
3104 			 * according to the existence of a routing header.
3105 			 * Note, however, that this depends on the order of the
3106 			 * extension headers in the ancillary data; the 1st
3107 			 * part of the destination options header must appear
3108 			 * before the routing header in the ancillary data,
3109 			 * too.
3110 			 * RFC3542 solved the ambiguity by introducing
3111 			 * separate ancillary data or option types.
3112 			 */
3113 			if (opt->ip6po_rthdr == NULL)
3114 				newdest = &opt->ip6po_dest1;
3115 			else
3116 				newdest = &opt->ip6po_dest2;
3117 			break;
3118 		case IPV6_RTHDRDSTOPTS:
3119 			newdest = &opt->ip6po_dest1;
3120 			break;
3121 		case IPV6_DSTOPTS:
3122 			newdest = &opt->ip6po_dest2;
3123 			break;
3124 		}
3125 
3126 		/* turn off the previous option, then set the new option. */
3127 		ip6_clearpktopts(opt, optname);
3128 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3129 		if (*newdest == NULL)
3130 			return (ENOBUFS);
3131 		memcpy(*newdest, dest, destlen);
3132 
3133 		break;
3134 	}
3135 
3136 #ifdef RFC2292
3137 	case IPV6_2292RTHDR:
3138 #endif
3139 	case IPV6_RTHDR:
3140 	{
3141 		struct ip6_rthdr *rth;
3142 		int rthlen;
3143 
3144 		if (len == 0) {
3145 			ip6_clearpktopts(opt, IPV6_RTHDR);
3146 			break;	/* just remove the option */
3147 		}
3148 
3149 		/* message length validation */
3150 		if (len < sizeof(struct ip6_rthdr))
3151 			return (EINVAL);
3152 		rth = (struct ip6_rthdr *)buf;
3153 		rthlen = (rth->ip6r_len + 1) << 3;
3154 		if (len != rthlen)
3155 			return (EINVAL);
3156 		switch (rth->ip6r_type) {
3157 		case IPV6_RTHDR_TYPE_0:
3158 			if (rth->ip6r_len == 0)	/* must contain one addr */
3159 				return (EINVAL);
3160 			if (rth->ip6r_len % 2) /* length must be even */
3161 				return (EINVAL);
3162 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3163 				return (EINVAL);
3164 			break;
3165 		default:
3166 			return (EINVAL);	/* not supported */
3167 		}
3168 		/* turn off the previous option */
3169 		ip6_clearpktopts(opt, IPV6_RTHDR);
3170 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3171 		if (opt->ip6po_rthdr == NULL)
3172 			return (ENOBUFS);
3173 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3174 		break;
3175 	}
3176 
3177 	case IPV6_USE_MIN_MTU:
3178 		if (len != sizeof(int))
3179 			return (EINVAL);
3180 		minmtupolicy = *(int *)buf;
3181 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3182 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3183 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3184 			return (EINVAL);
3185 		}
3186 		opt->ip6po_minmtu = minmtupolicy;
3187 		break;
3188 
3189 	case IPV6_DONTFRAG:
3190 		if (len != sizeof(int))
3191 			return (EINVAL);
3192 
3193 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3194 			/*
3195 			 * we ignore this option for TCP sockets.
3196 			 * (RFC3542 leaves this case unspecified.)
3197 			 */
3198 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3199 		} else
3200 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3201 		break;
3202 
3203 	default:
3204 		return (ENOPROTOOPT);
3205 	} /* end of switch */
3206 
3207 	return (0);
3208 }
3209 
3210 /*
3211  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3212  * packet to the input queue of a specified interface.  Note that this
3213  * calls the output routine of the loopback "driver", but with an interface
3214  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3215  */
3216 void
3217 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3218 	const struct sockaddr_in6 *dst)
3219 {
3220 	struct mbuf *copym;
3221 	struct ip6_hdr *ip6;
3222 
3223 	copym = m_copy(m, 0, M_COPYALL);
3224 	if (copym == NULL)
3225 		return;
3226 
3227 	/*
3228 	 * Make sure to deep-copy IPv6 header portion in case the data
3229 	 * is in an mbuf cluster, so that we can safely override the IPv6
3230 	 * header portion later.
3231 	 */
3232 	if ((copym->m_flags & M_EXT) != 0 ||
3233 	    copym->m_len < sizeof(struct ip6_hdr)) {
3234 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3235 		if (copym == NULL)
3236 			return;
3237 	}
3238 
3239 #ifdef DIAGNOSTIC
3240 	if (copym->m_len < sizeof(*ip6)) {
3241 		m_freem(copym);
3242 		return;
3243 	}
3244 #endif
3245 
3246 	ip6 = mtod(copym, struct ip6_hdr *);
3247 	/*
3248 	 * clear embedded scope identifiers if necessary.
3249 	 * in6_clearscope will touch the addresses only when necessary.
3250 	 */
3251 	in6_clearscope(&ip6->ip6_src);
3252 	in6_clearscope(&ip6->ip6_dst);
3253 
3254 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3255 }
3256 
3257 /*
3258  * Chop IPv6 header off from the payload.
3259  */
3260 static int
3261 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3262 {
3263 	struct mbuf *mh;
3264 	struct ip6_hdr *ip6;
3265 
3266 	ip6 = mtod(m, struct ip6_hdr *);
3267 	if (m->m_len > sizeof(*ip6)) {
3268 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3269 		if (mh == 0) {
3270 			m_freem(m);
3271 			return ENOBUFS;
3272 		}
3273 		M_MOVE_PKTHDR(mh, m);
3274 		MH_ALIGN(mh, sizeof(*ip6));
3275 		m->m_len -= sizeof(*ip6);
3276 		m->m_data += sizeof(*ip6);
3277 		mh->m_next = m;
3278 		m = mh;
3279 		m->m_len = sizeof(*ip6);
3280 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3281 	}
3282 	exthdrs->ip6e_ip6 = m;
3283 	return 0;
3284 }
3285 
3286 /*
3287  * Compute IPv6 extension header length.
3288  */
3289 int
3290 ip6_optlen(struct in6pcb *in6p)
3291 {
3292 	int len;
3293 
3294 	if (!in6p->in6p_outputopts)
3295 		return 0;
3296 
3297 	len = 0;
3298 #define elen(x) \
3299     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3300 
3301 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3302 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3303 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3304 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3305 	return len;
3306 #undef elen
3307 }
3308