xref: /netbsd-src/sys/netinet6/ip6_output.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: ip6_output.c,v 1.105 2006/11/23 19:41:58 yamt 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 1.105 2006/11/23 19:41:58 yamt 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/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/ip6protosw.h>
99 #include <netinet6/scope6_var.h>
100 
101 #ifdef IPSEC
102 #include <netinet6/ipsec.h>
103 #include <netkey/key.h>
104 #endif /* IPSEC */
105 
106 #include <net/net_osdep.h>
107 
108 #ifdef PFIL_HOOKS
109 extern struct pfil_head inet6_pfil_hook;	/* XXX */
110 #endif
111 
112 struct ip6_exthdrs {
113 	struct mbuf *ip6e_ip6;
114 	struct mbuf *ip6e_hbh;
115 	struct mbuf *ip6e_dest1;
116 	struct mbuf *ip6e_rthdr;
117 	struct mbuf *ip6e_dest2;
118 };
119 
120 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
121 	int, int));
122 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **));
123 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int,
124 	int, int, int));
125 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
126 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
127 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
128 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
129 	struct ip6_frag **));
130 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
131 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
132 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
133 	struct ifnet *, struct in6_addr *, u_long *, int *));
134 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int));
135 
136 #ifdef RFC2292
137 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
138 	struct socket *));
139 #endif
140 
141 #define	IN6_NEED_CHECKSUM(ifp, csum_flags) \
142 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
143 	(((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \
144 	(((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum)))
145 
146 struct ip6_tso_output_args {
147 	struct ifnet *ifp;
148 	struct ifnet *origifp;
149 	struct sockaddr_in6 *dst;
150 	struct rtentry *rt;
151 };
152 
153 static int ip6_tso_output_callback(void *, struct mbuf *);
154 static int ip6_tso_output(struct ifnet *, struct ifnet *, struct mbuf *,
155     struct sockaddr_in6 *, struct rtentry *);
156 
157 static int
158 ip6_tso_output_callback(void *vp, struct mbuf *m)
159 {
160 	struct ip6_tso_output_args *args = vp;
161 
162 	return nd6_output(args->ifp, args->origifp, m, args->dst, args->rt);
163 }
164 
165 static int
166 ip6_tso_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
167     struct sockaddr_in6 *dst, struct rtentry *rt)
168 {
169 	struct ip6_tso_output_args args;
170 
171 	args.ifp = ifp;
172 	args.origifp = origifp;
173 	args.dst = dst;
174 	args.rt = rt;
175 
176 	return tcp6_segment(m, ip6_tso_output_callback, &args);
177 }
178 
179 /*
180  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
181  * header (with pri, len, nxt, hlim, src, dst).
182  * This function may modify ver and hlim only.
183  * The mbuf chain containing the packet will be freed.
184  * The mbuf opt, if present, will not be freed.
185  *
186  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
187  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
188  * which is rt_rmx.rmx_mtu.
189  */
190 int
191 ip6_output(
192     struct mbuf *m0,
193     struct ip6_pktopts *opt,
194     struct route_in6 *ro,
195     int flags,
196     struct ip6_moptions *im6o,
197     struct socket *so,
198     struct ifnet **ifpp		/* XXX: just for statistics */
199 )
200 {
201 	struct ip6_hdr *ip6, *mhip6;
202 	struct ifnet *ifp, *origifp;
203 	struct mbuf *m = m0;
204 	int hlen, tlen, len, off;
205 	boolean_t tso;
206 	struct route_in6 ip6route;
207 	struct rtentry *rt = NULL;
208 	struct sockaddr_in6 *dst, src_sa, dst_sa;
209 	int error = 0;
210 	struct in6_ifaddr *ia = NULL;
211 	u_long mtu;
212 	int alwaysfrag, dontfrag;
213 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
214 	struct ip6_exthdrs exthdrs;
215 	struct in6_addr finaldst, src0, dst0;
216 	u_int32_t zone;
217 	struct route_in6 *ro_pmtu = NULL;
218 	int hdrsplit = 0;
219 	int needipsec = 0;
220 #ifdef IPSEC
221 	int needipsectun = 0;
222 	struct secpolicy *sp = NULL;
223 
224 	ip6 = mtod(m, struct ip6_hdr *);
225 #endif /* IPSEC */
226 
227 #ifdef  DIAGNOSTIC
228 	if ((m->m_flags & M_PKTHDR) == 0)
229 		panic("ip6_output: no HDR");
230 
231 	if ((m->m_pkthdr.csum_flags &
232 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
233 		panic("ip6_output: IPv4 checksum offload flags: %d",
234 		    m->m_pkthdr.csum_flags);
235 	}
236 
237 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
238 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
239 		panic("ip6_output: conflicting checksum offload flags: %d",
240 		    m->m_pkthdr.csum_flags);
241 	}
242 #endif
243 
244 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
245 
246 #define MAKE_EXTHDR(hp, mp)						\
247     do {								\
248 	if (hp) {							\
249 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
250 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
251 		    ((eh)->ip6e_len + 1) << 3);				\
252 		if (error)						\
253 			goto freehdrs;					\
254 	}								\
255     } while (/*CONSTCOND*/ 0)
256 
257 	bzero(&exthdrs, sizeof(exthdrs));
258 	if (opt) {
259 		/* Hop-by-Hop options header */
260 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
261 		/* Destination options header(1st part) */
262 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
263 		/* Routing header */
264 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
265 		/* Destination options header(2nd part) */
266 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
267 	}
268 
269 #ifdef IPSEC
270 	if ((flags & IPV6_FORWARDING) != 0) {
271 		needipsec = 0;
272 		goto skippolicycheck;
273 	}
274 
275 	/* get a security policy for this packet */
276 	if (so == NULL)
277 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
278 	else {
279 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
280 					 IPSEC_DIR_OUTBOUND)) {
281 			needipsec = 0;
282 			goto skippolicycheck;
283 		}
284 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
285 	}
286 
287 	if (sp == NULL) {
288 		ipsec6stat.out_inval++;
289 		goto freehdrs;
290 	}
291 
292 	error = 0;
293 
294 	/* check policy */
295 	switch (sp->policy) {
296 	case IPSEC_POLICY_DISCARD:
297 		/*
298 		 * This packet is just discarded.
299 		 */
300 		ipsec6stat.out_polvio++;
301 		goto freehdrs;
302 
303 	case IPSEC_POLICY_BYPASS:
304 	case IPSEC_POLICY_NONE:
305 		/* no need to do IPsec. */
306 		needipsec = 0;
307 		break;
308 
309 	case IPSEC_POLICY_IPSEC:
310 		if (sp->req == NULL) {
311 			/* XXX should be panic ? */
312 			printf("ip6_output: No IPsec request specified.\n");
313 			error = EINVAL;
314 			goto freehdrs;
315 		}
316 		needipsec = 1;
317 		break;
318 
319 	case IPSEC_POLICY_ENTRUST:
320 	default:
321 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
322 	}
323 
324   skippolicycheck:;
325 #endif /* IPSEC */
326 
327 	if (needipsec &&
328 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
329 		in6_delayed_cksum(m);
330 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
331 	}
332 
333 	/*
334 	 * Calculate the total length of the extension header chain.
335 	 * Keep the length of the unfragmentable part for fragmentation.
336 	 */
337 	optlen = 0;
338 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
339 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
340 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
341 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
342 	/* NOTE: we don't add AH/ESP length here. do that later. */
343 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
344 
345 	/*
346 	 * If we need IPsec, or there is at least one extension header,
347 	 * separate IP6 header from the payload.
348 	 */
349 	if ((needipsec || optlen) && !hdrsplit) {
350 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
351 			m = NULL;
352 			goto freehdrs;
353 		}
354 		m = exthdrs.ip6e_ip6;
355 		hdrsplit++;
356 	}
357 
358 	/* adjust pointer */
359 	ip6 = mtod(m, struct ip6_hdr *);
360 
361 	/* adjust mbuf packet header length */
362 	m->m_pkthdr.len += optlen;
363 	plen = m->m_pkthdr.len - sizeof(*ip6);
364 
365 	/* If this is a jumbo payload, insert a jumbo payload option. */
366 	if (plen > IPV6_MAXPACKET) {
367 		if (!hdrsplit) {
368 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
369 				m = NULL;
370 				goto freehdrs;
371 			}
372 			m = exthdrs.ip6e_ip6;
373 			hdrsplit++;
374 		}
375 		/* adjust pointer */
376 		ip6 = mtod(m, struct ip6_hdr *);
377 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
378 			goto freehdrs;
379 		optlen += 8; /* XXX JUMBOOPTLEN */
380 		ip6->ip6_plen = 0;
381 	} else
382 		ip6->ip6_plen = htons(plen);
383 
384 	/*
385 	 * Concatenate headers and fill in next header fields.
386 	 * Here we have, on "m"
387 	 *	IPv6 payload
388 	 * and we insert headers accordingly.  Finally, we should be getting:
389 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
390 	 *
391 	 * during the header composing process, "m" points to IPv6 header.
392 	 * "mprev" points to an extension header prior to esp.
393 	 */
394 	{
395 		u_char *nexthdrp = &ip6->ip6_nxt;
396 		struct mbuf *mprev = m;
397 
398 		/*
399 		 * we treat dest2 specially.  this makes IPsec processing
400 		 * much easier.  the goal here is to make mprev point the
401 		 * mbuf prior to dest2.
402 		 *
403 		 * result: IPv6 dest2 payload
404 		 * m and mprev will point to IPv6 header.
405 		 */
406 		if (exthdrs.ip6e_dest2) {
407 			if (!hdrsplit)
408 				panic("assumption failed: hdr not split");
409 			exthdrs.ip6e_dest2->m_next = m->m_next;
410 			m->m_next = exthdrs.ip6e_dest2;
411 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
412 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
413 		}
414 
415 #define MAKE_CHAIN(m, mp, p, i)\
416     do {\
417 	if (m) {\
418 		if (!hdrsplit) \
419 			panic("assumption failed: hdr not split"); \
420 		*mtod((m), u_char *) = *(p);\
421 		*(p) = (i);\
422 		p = mtod((m), u_char *);\
423 		(m)->m_next = (mp)->m_next;\
424 		(mp)->m_next = (m);\
425 		(mp) = (m);\
426 	}\
427     } while (/*CONSTCOND*/ 0)
428 		/*
429 		 * result: IPv6 hbh dest1 rthdr dest2 payload
430 		 * m will point to IPv6 header.  mprev will point to the
431 		 * extension header prior to dest2 (rthdr in the above case).
432 		 */
433 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
434 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
435 		    IPPROTO_DSTOPTS);
436 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
437 		    IPPROTO_ROUTING);
438 
439 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
440 		    sizeof(struct ip6_hdr) + optlen);
441 
442 #ifdef IPSEC
443 		if (!needipsec)
444 			goto skip_ipsec2;
445 
446 		/*
447 		 * pointers after IPsec headers are not valid any more.
448 		 * other pointers need a great care too.
449 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
450 		 */
451 		exthdrs.ip6e_dest2 = NULL;
452 
453 	    {
454 		struct ip6_rthdr *rh = NULL;
455 		int segleft_org = 0;
456 		struct ipsec_output_state state;
457 
458 		if (exthdrs.ip6e_rthdr) {
459 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
460 			segleft_org = rh->ip6r_segleft;
461 			rh->ip6r_segleft = 0;
462 		}
463 
464 		bzero(&state, sizeof(state));
465 		state.m = m;
466 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
467 		    &needipsectun);
468 		m = state.m;
469 		if (error) {
470 			/* mbuf is already reclaimed in ipsec6_output_trans. */
471 			m = NULL;
472 			switch (error) {
473 			case EHOSTUNREACH:
474 			case ENETUNREACH:
475 			case EMSGSIZE:
476 			case ENOBUFS:
477 			case ENOMEM:
478 				break;
479 			default:
480 				printf("ip6_output (ipsec): error code %d\n", error);
481 				/* FALLTHROUGH */
482 			case ENOENT:
483 				/* don't show these error codes to the user */
484 				error = 0;
485 				break;
486 			}
487 			goto bad;
488 		}
489 		if (exthdrs.ip6e_rthdr) {
490 			/* ah6_output doesn't modify mbuf chain */
491 			rh->ip6r_segleft = segleft_org;
492 		}
493 	    }
494 skip_ipsec2:;
495 #endif
496 	}
497 
498 	/*
499 	 * If there is a routing header, replace destination address field
500 	 * with the first hop of the routing header.
501 	 */
502 	if (exthdrs.ip6e_rthdr) {
503 		struct ip6_rthdr *rh;
504 		struct ip6_rthdr0 *rh0;
505 		struct in6_addr *addr;
506 		struct sockaddr_in6 sa;
507 
508 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
509 		    struct ip6_rthdr *));
510 		finaldst = ip6->ip6_dst;
511 		switch (rh->ip6r_type) {
512 		case IPV6_RTHDR_TYPE_0:
513 			 rh0 = (struct ip6_rthdr0 *)rh;
514 			 addr = (struct in6_addr *)(rh0 + 1);
515 
516 			 /*
517 			  * construct a sockaddr_in6 form of
518 			  * the first hop.
519 			  *
520 			  * XXX: we may not have enough
521 			  * information about its scope zone;
522 			  * there is no standard API to pass
523 			  * the information from the
524 			  * application.
525 			  */
526 			 bzero(&sa, sizeof(sa));
527 			 sa.sin6_family = AF_INET6;
528 			 sa.sin6_len = sizeof(sa);
529 			 sa.sin6_addr = addr[0];
530 			 if ((error = sa6_embedscope(&sa,
531 			     ip6_use_defzone)) != 0) {
532 				 goto bad;
533 			 }
534 			 ip6->ip6_dst = sa.sin6_addr;
535 			 (void)memmove(&addr[0], &addr[1],
536 			     sizeof(struct in6_addr) *
537 			     (rh0->ip6r0_segleft - 1));
538 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
539 			 /* XXX */
540 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
541 			 break;
542 		default:	/* is it possible? */
543 			 error = EINVAL;
544 			 goto bad;
545 		}
546 	}
547 
548 	/* Source address validation */
549 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
550 	    (flags & IPV6_UNSPECSRC) == 0) {
551 		error = EOPNOTSUPP;
552 		ip6stat.ip6s_badscope++;
553 		goto bad;
554 	}
555 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
556 		error = EOPNOTSUPP;
557 		ip6stat.ip6s_badscope++;
558 		goto bad;
559 	}
560 
561 	ip6stat.ip6s_localout++;
562 
563 	/*
564 	 * Route packet.
565 	 */
566 	/* initialize cached route */
567 	if (ro == 0) {
568 		ro = &ip6route;
569 		bzero((caddr_t)ro, sizeof(*ro));
570 	}
571 	ro_pmtu = ro;
572 	if (opt && opt->ip6po_rthdr)
573 		ro = &opt->ip6po_route;
574 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
575 
576  	/*
577 	 * if specified, try to fill in the traffic class field.
578 	 * do not override if a non-zero value is already set.
579 	 * we check the diffserv field and the ecn field separately.
580 	 */
581 	if (opt && opt->ip6po_tclass >= 0) {
582 		int mask = 0;
583 
584 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
585 			mask |= 0xfc;
586 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
587 			mask |= 0x03;
588 		if (mask != 0)
589 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
590 	}
591 
592 	/* fill in or override the hop limit field, if necessary. */
593 	if (opt && opt->ip6po_hlim != -1)
594 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
595 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
596 		if (im6o != NULL)
597 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
598 		else
599 			ip6->ip6_hlim = ip6_defmcasthlim;
600 	}
601 
602 #ifdef IPSEC
603 	if (needipsec && needipsectun) {
604 		struct ipsec_output_state state;
605 
606 		/*
607 		 * All the extension headers will become inaccessible
608 		 * (since they can be encrypted).
609 		 * Don't panic, we need no more updates to extension headers
610 		 * on inner IPv6 packet (since they are now encapsulated).
611 		 *
612 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
613 		 */
614 		bzero(&exthdrs, sizeof(exthdrs));
615 		exthdrs.ip6e_ip6 = m;
616 
617 		bzero(&state, sizeof(state));
618 		state.m = m;
619 		state.ro = (struct route *)ro;
620 		state.dst = (struct sockaddr *)dst;
621 
622 		error = ipsec6_output_tunnel(&state, sp, flags);
623 
624 		m = state.m;
625 		ro_pmtu = ro = (struct route_in6 *)state.ro;
626 		dst = (struct sockaddr_in6 *)state.dst;
627 		if (error) {
628 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
629 			m0 = m = NULL;
630 			m = NULL;
631 			switch (error) {
632 			case EHOSTUNREACH:
633 			case ENETUNREACH:
634 			case EMSGSIZE:
635 			case ENOBUFS:
636 			case ENOMEM:
637 				break;
638 			default:
639 				printf("ip6_output (ipsec): error code %d\n", error);
640 				/* FALLTHROUGH */
641 			case ENOENT:
642 				/* don't show these error codes to the user */
643 				error = 0;
644 				break;
645 			}
646 			goto bad;
647 		}
648 
649 		exthdrs.ip6e_ip6 = m;
650 	}
651 #endif /* IPSEC */
652 
653 	/* adjust pointer */
654 	ip6 = mtod(m, struct ip6_hdr *);
655 
656 	bzero(&dst_sa, sizeof(dst_sa));
657 	dst_sa.sin6_family = AF_INET6;
658 	dst_sa.sin6_len = sizeof(dst_sa);
659 	dst_sa.sin6_addr = ip6->ip6_dst;
660 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp, &rt, 0))
661 	    != 0) {
662 		switch (error) {
663 		case EHOSTUNREACH:
664 			ip6stat.ip6s_noroute++;
665 			break;
666 		case EADDRNOTAVAIL:
667 		default:
668 			break; /* XXX statistics? */
669 		}
670 		if (ifp != NULL)
671 			in6_ifstat_inc(ifp, ifs6_out_discard);
672 		goto bad;
673 	}
674 	if (rt == NULL) {
675 		/*
676 		 * If in6_selectroute() does not return a route entry,
677 		 * dst may not have been updated.
678 		 */
679 		*dst = dst_sa;	/* XXX */
680 	}
681 
682 	/*
683 	 * then rt (for unicast) and ifp must be non-NULL valid values.
684 	 */
685 	if ((flags & IPV6_FORWARDING) == 0) {
686 		/* XXX: the FORWARDING flag can be set for mrouting. */
687 		in6_ifstat_inc(ifp, ifs6_out_request);
688 	}
689 	if (rt != NULL) {
690 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
691 		rt->rt_use++;
692 	}
693 
694 	/*
695 	 * The outgoing interface must be in the zone of source and
696 	 * destination addresses.  We should use ia_ifp to support the
697 	 * case of sending packets to an address of our own.
698 	 */
699 	if (ia != NULL && ia->ia_ifp)
700 		origifp = ia->ia_ifp;
701 	else
702 		origifp = ifp;
703 
704 	src0 = ip6->ip6_src;
705 	if (in6_setscope(&src0, origifp, &zone))
706 		goto badscope;
707 	bzero(&src_sa, sizeof(src_sa));
708 	src_sa.sin6_family = AF_INET6;
709 	src_sa.sin6_len = sizeof(src_sa);
710 	src_sa.sin6_addr = ip6->ip6_src;
711 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
712 		goto badscope;
713 
714 	dst0 = ip6->ip6_dst;
715 	if (in6_setscope(&dst0, origifp, &zone))
716 		goto badscope;
717 	/* re-initialize to be sure */
718 	bzero(&dst_sa, sizeof(dst_sa));
719 	dst_sa.sin6_family = AF_INET6;
720 	dst_sa.sin6_len = sizeof(dst_sa);
721 	dst_sa.sin6_addr = ip6->ip6_dst;
722 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
723 		goto badscope;
724 
725 	/* scope check is done. */
726 	goto routefound;
727 
728   badscope:
729 	ip6stat.ip6s_badscope++;
730 	in6_ifstat_inc(origifp, ifs6_out_discard);
731 	if (error == 0)
732 		error = EHOSTUNREACH; /* XXX */
733 	goto bad;
734 
735   routefound:
736 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
737 		if (opt && opt->ip6po_nextroute.ro_rt) {
738 			/*
739 			 * The nexthop is explicitly specified by the
740 			 * application.  We assume the next hop is an IPv6
741 			 * address.
742 			 */
743 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
744 		} else if ((rt->rt_flags & RTF_GATEWAY))
745 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
746 	}
747 
748 	/*
749 	 * XXXXXX: original code follows:
750 	 */
751 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
752 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
753 	else {
754 		struct	in6_multi *in6m;
755 
756 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
757 
758 		in6_ifstat_inc(ifp, ifs6_out_mcast);
759 
760 		/*
761 		 * Confirm that the outgoing interface supports multicast.
762 		 */
763 		if (!(ifp->if_flags & IFF_MULTICAST)) {
764 			ip6stat.ip6s_noroute++;
765 			in6_ifstat_inc(ifp, ifs6_out_discard);
766 			error = ENETUNREACH;
767 			goto bad;
768 		}
769 
770 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
771 		if (in6m != NULL &&
772 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
773 			/*
774 			 * If we belong to the destination multicast group
775 			 * on the outgoing interface, and the caller did not
776 			 * forbid loopback, loop back a copy.
777 			 */
778 			ip6_mloopback(ifp, m, dst);
779 		} else {
780 			/*
781 			 * If we are acting as a multicast router, perform
782 			 * multicast forwarding as if the packet had just
783 			 * arrived on the interface to which we are about
784 			 * to send.  The multicast forwarding function
785 			 * recursively calls this function, using the
786 			 * IPV6_FORWARDING flag to prevent infinite recursion.
787 			 *
788 			 * Multicasts that are looped back by ip6_mloopback(),
789 			 * above, will be forwarded by the ip6_input() routine,
790 			 * if necessary.
791 			 */
792 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
793 				if (ip6_mforward(ip6, ifp, m) != 0) {
794 					m_freem(m);
795 					goto done;
796 				}
797 			}
798 		}
799 		/*
800 		 * Multicasts with a hoplimit of zero may be looped back,
801 		 * above, but must not be transmitted on a network.
802 		 * Also, multicasts addressed to the loopback interface
803 		 * are not sent -- the above call to ip6_mloopback() will
804 		 * loop back a copy if this host actually belongs to the
805 		 * destination group on the loopback interface.
806 		 */
807 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
808 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
809 			m_freem(m);
810 			goto done;
811 		}
812 	}
813 
814 	/*
815 	 * Fill the outgoing inteface to tell the upper layer
816 	 * to increment per-interface statistics.
817 	 */
818 	if (ifpp)
819 		*ifpp = ifp;
820 
821 	/* Determine path MTU. */
822 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
823 	    &alwaysfrag)) != 0)
824 		goto bad;
825 #ifdef IPSEC
826 	if (needipsectun)
827 		mtu = IPV6_MMTU;
828 #endif
829 
830 	/*
831 	 * The caller of this function may specify to use the minimum MTU
832 	 * in some cases.
833 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
834 	 * setting.  The logic is a bit complicated; by default, unicast
835 	 * packets will follow path MTU while multicast packets will be sent at
836 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
837 	 * including unicast ones will be sent at the minimum MTU.  Multicast
838 	 * packets will always be sent at the minimum MTU unless
839 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
840 	 * See RFC 3542 for more details.
841 	 */
842 	if (mtu > IPV6_MMTU) {
843 		if ((flags & IPV6_MINMTU))
844 			mtu = IPV6_MMTU;
845 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
846 			mtu = IPV6_MMTU;
847 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
848 			 (opt == NULL ||
849 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
850 			mtu = IPV6_MMTU;
851 		}
852 	}
853 
854 	/*
855 	 * clear embedded scope identifiers if necessary.
856 	 * in6_clearscope will touch the addresses only when necessary.
857 	 */
858 	in6_clearscope(&ip6->ip6_src);
859 	in6_clearscope(&ip6->ip6_dst);
860 
861 	/*
862 	 * If the outgoing packet contains a hop-by-hop options header,
863 	 * it must be examined and processed even by the source node.
864 	 * (RFC 2460, section 4.)
865 	 */
866 	if (exthdrs.ip6e_hbh) {
867 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
868 		u_int32_t dummy1; /* XXX unused */
869 		u_int32_t dummy2; /* XXX unused */
870 
871 		/*
872 		 *  XXX: if we have to send an ICMPv6 error to the sender,
873 		 *       we need the M_LOOP flag since icmp6_error() expects
874 		 *       the IPv6 and the hop-by-hop options header are
875 		 *       continuous unless the flag is set.
876 		 */
877 		m->m_flags |= M_LOOP;
878 		m->m_pkthdr.rcvif = ifp;
879 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
880 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
881 		    &dummy1, &dummy2) < 0) {
882 			/* m was already freed at this point */
883 			error = EINVAL;/* better error? */
884 			goto done;
885 		}
886 		m->m_flags &= ~M_LOOP; /* XXX */
887 		m->m_pkthdr.rcvif = NULL;
888 	}
889 
890 #ifdef PFIL_HOOKS
891 	/*
892 	 * Run through list of hooks for output packets.
893 	 */
894 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
895 		goto done;
896 	if (m == NULL)
897 		goto done;
898 	ip6 = mtod(m, struct ip6_hdr *);
899 #endif /* PFIL_HOOKS */
900 	/*
901 	 * Send the packet to the outgoing interface.
902 	 * If necessary, do IPv6 fragmentation before sending.
903 	 *
904 	 * the logic here is rather complex:
905 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
906 	 * 1-a:	send as is if tlen <= path mtu
907 	 * 1-b:	fragment if tlen > path mtu
908 	 *
909 	 * 2: if user asks us not to fragment (dontfrag == 1)
910 	 * 2-a:	send as is if tlen <= interface mtu
911 	 * 2-b:	error if tlen > interface mtu
912 	 *
913 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
914 	 *	always fragment
915 	 *
916 	 * 4: if dontfrag == 1 && alwaysfrag == 1
917 	 *	error, as we cannot handle this conflicting request
918 	 */
919 	tlen = m->m_pkthdr.len;
920 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
921 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
922 		dontfrag = 1;
923 	else
924 		dontfrag = 0;
925 
926 	if (dontfrag && alwaysfrag) {	/* case 4 */
927 		/* conflicting request - can't transmit */
928 		error = EMSGSIZE;
929 		goto bad;
930 	}
931 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
932 		/*
933 		 * Even if the DONTFRAG option is specified, we cannot send the
934 		 * packet when the data length is larger than the MTU of the
935 		 * outgoing interface.
936 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
937 		 * well as returning an error code (the latter is not described
938 		 * in the API spec.)
939 		 */
940 		u_int32_t mtu32;
941 		struct ip6ctlparam ip6cp;
942 
943 		mtu32 = (u_int32_t)mtu;
944 		bzero(&ip6cp, sizeof(ip6cp));
945 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
946 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
947 		    (void *)&ip6cp);
948 
949 		error = EMSGSIZE;
950 		goto bad;
951 	}
952 
953 	/*
954 	 * transmit packet without fragmentation
955 	 */
956 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
957 		/* case 1-a and 2-a */
958 		struct in6_ifaddr *ia6;
959 		int sw_csum;
960 
961 		ip6 = mtod(m, struct ip6_hdr *);
962 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
963 		if (ia6) {
964 			/* Record statistics for this interface address. */
965 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
966 		}
967 #ifdef IPSEC
968 		/* clean ipsec history once it goes out of the node */
969 		ipsec_delaux(m);
970 #endif
971 
972 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
973 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
974 			if (IN6_NEED_CHECKSUM(ifp,
975 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
976 				in6_delayed_cksum(m);
977 			}
978 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
979 		}
980 
981 		if (__predict_true(!tso ||
982 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
983 			error = nd6_output(ifp, origifp, m, dst, rt);
984 		} else {
985 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
986 		}
987 		goto done;
988 	}
989 
990 	if (tso) {
991 		error = EINVAL; /* XXX */
992 		goto bad;
993 	}
994 
995 	/*
996 	 * try to fragment the packet.  case 1-b and 3
997 	 */
998 	if (mtu < IPV6_MMTU) {
999 		/* path MTU cannot be less than IPV6_MMTU */
1000 		error = EMSGSIZE;
1001 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1002 		goto bad;
1003 	} else if (ip6->ip6_plen == 0) {
1004 		/* jumbo payload cannot be fragmented */
1005 		error = EMSGSIZE;
1006 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1007 		goto bad;
1008 	} else {
1009 		struct mbuf **mnext, *m_frgpart;
1010 		struct ip6_frag *ip6f;
1011 		u_int32_t id = htonl(ip6_randomid());
1012 		u_char nextproto;
1013 #if 0				/* see below */
1014 		struct ip6ctlparam ip6cp;
1015 		u_int32_t mtu32;
1016 #endif
1017 
1018 		/*
1019 		 * Too large for the destination or interface;
1020 		 * fragment if possible.
1021 		 * Must be able to put at least 8 bytes per fragment.
1022 		 */
1023 		hlen = unfragpartlen;
1024 		if (mtu > IPV6_MAXPACKET)
1025 			mtu = IPV6_MAXPACKET;
1026 
1027 #if 0
1028 		/*
1029 		 * It is believed this code is a leftover from the
1030 		 * development of the IPV6_RECVPATHMTU sockopt and
1031 		 * associated work to implement RFC3542.
1032 		 * It's not entirely clear what the intent of the API
1033 		 * is at this point, so disable this code for now.
1034 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1035 		 * will send notifications if the application requests.
1036 		 */
1037 
1038 		/* Notify a proper path MTU to applications. */
1039 		mtu32 = (u_int32_t)mtu;
1040 		bzero(&ip6cp, sizeof(ip6cp));
1041 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1042 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1043 		    (void *)&ip6cp);
1044 #endif
1045 
1046 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1047 		if (len < 8) {
1048 			error = EMSGSIZE;
1049 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1050 			goto bad;
1051 		}
1052 
1053 		mnext = &m->m_nextpkt;
1054 
1055 		/*
1056 		 * Change the next header field of the last header in the
1057 		 * unfragmentable part.
1058 		 */
1059 		if (exthdrs.ip6e_rthdr) {
1060 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1061 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1062 		} else if (exthdrs.ip6e_dest1) {
1063 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1064 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1065 		} else if (exthdrs.ip6e_hbh) {
1066 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1067 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1068 		} else {
1069 			nextproto = ip6->ip6_nxt;
1070 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1071 		}
1072 
1073 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
1074 		    != 0) {
1075 			if (IN6_NEED_CHECKSUM(ifp,
1076 			    m->m_pkthdr.csum_flags &
1077 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
1078 				in6_delayed_cksum(m);
1079 			}
1080 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
1081 		}
1082 
1083 		/*
1084 		 * Loop through length of segment after first fragment,
1085 		 * make new header and copy data of each part and link onto
1086 		 * chain.
1087 		 */
1088 		m0 = m;
1089 		for (off = hlen; off < tlen; off += len) {
1090 			struct mbuf *mlast;
1091 
1092 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1093 			if (!m) {
1094 				error = ENOBUFS;
1095 				ip6stat.ip6s_odropped++;
1096 				goto sendorfree;
1097 			}
1098 			m->m_pkthdr.rcvif = NULL;
1099 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1100 			*mnext = m;
1101 			mnext = &m->m_nextpkt;
1102 			m->m_data += max_linkhdr;
1103 			mhip6 = mtod(m, struct ip6_hdr *);
1104 			*mhip6 = *ip6;
1105 			m->m_len = sizeof(*mhip6);
1106 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1107 			if (error) {
1108 				ip6stat.ip6s_odropped++;
1109 				goto sendorfree;
1110 			}
1111 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
1112 			if (off + len >= tlen)
1113 				len = tlen - off;
1114 			else
1115 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1116 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
1117 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1118 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1119 				error = ENOBUFS;
1120 				ip6stat.ip6s_odropped++;
1121 				goto sendorfree;
1122 			}
1123 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
1124 				;
1125 			mlast->m_next = m_frgpart;
1126 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1127 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1128 			ip6f->ip6f_reserved = 0;
1129 			ip6f->ip6f_ident = id;
1130 			ip6f->ip6f_nxt = nextproto;
1131 			ip6stat.ip6s_ofragments++;
1132 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1133 		}
1134 
1135 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1136 	}
1137 
1138 	/*
1139 	 * Remove leading garbages.
1140 	 */
1141 sendorfree:
1142 	m = m0->m_nextpkt;
1143 	m0->m_nextpkt = 0;
1144 	m_freem(m0);
1145 	for (m0 = m; m; m = m0) {
1146 		m0 = m->m_nextpkt;
1147 		m->m_nextpkt = 0;
1148 		if (error == 0) {
1149 			struct in6_ifaddr *ia6;
1150 			ip6 = mtod(m, struct ip6_hdr *);
1151 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1152 			if (ia6) {
1153 				/*
1154 				 * Record statistics for this interface
1155 				 * address.
1156 				 */
1157 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1158 				    m->m_pkthdr.len;
1159 			}
1160 #ifdef IPSEC
1161 			/* clean ipsec history once it goes out of the node */
1162 			ipsec_delaux(m);
1163 #endif
1164 			error = nd6_output(ifp, origifp, m, dst, rt);
1165 		} else
1166 			m_freem(m);
1167 	}
1168 
1169 	if (error == 0)
1170 		ip6stat.ip6s_fragmented++;
1171 
1172 done:
1173 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1174 		RTFREE(ro->ro_rt);
1175 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1176 		RTFREE(ro_pmtu->ro_rt);
1177 	}
1178 
1179 #ifdef IPSEC
1180 	if (sp != NULL)
1181 		key_freesp(sp);
1182 #endif /* IPSEC */
1183 
1184 	return (error);
1185 
1186 freehdrs:
1187 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1188 	m_freem(exthdrs.ip6e_dest1);
1189 	m_freem(exthdrs.ip6e_rthdr);
1190 	m_freem(exthdrs.ip6e_dest2);
1191 	/* FALLTHROUGH */
1192 bad:
1193 	m_freem(m);
1194 	goto done;
1195 }
1196 
1197 static int
1198 ip6_copyexthdr(mp, hdr, hlen)
1199 	struct mbuf **mp;
1200 	caddr_t hdr;
1201 	int hlen;
1202 {
1203 	struct mbuf *m;
1204 
1205 	if (hlen > MCLBYTES)
1206 		return (ENOBUFS); /* XXX */
1207 
1208 	MGET(m, M_DONTWAIT, MT_DATA);
1209 	if (!m)
1210 		return (ENOBUFS);
1211 
1212 	if (hlen > MLEN) {
1213 		MCLGET(m, M_DONTWAIT);
1214 		if ((m->m_flags & M_EXT) == 0) {
1215 			m_free(m);
1216 			return (ENOBUFS);
1217 		}
1218 	}
1219 	m->m_len = hlen;
1220 	if (hdr)
1221 		bcopy(hdr, mtod(m, caddr_t), hlen);
1222 
1223 	*mp = m;
1224 	return (0);
1225 }
1226 
1227 /*
1228  * Process a delayed payload checksum calculation.
1229  */
1230 void
1231 in6_delayed_cksum(struct mbuf *m)
1232 {
1233 	uint16_t csum, offset;
1234 
1235 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1236 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1237 	KASSERT((m->m_pkthdr.csum_flags
1238 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1239 
1240 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1241 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1242 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1243 		csum = 0xffff;
1244 	}
1245 
1246 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1247 	if ((offset + sizeof(csum)) > m->m_len) {
1248 		m_copyback(m, offset, sizeof(csum), &csum);
1249 	} else {
1250 		*(uint16_t *)(mtod(m, caddr_t) + offset) = csum;
1251 	}
1252 }
1253 
1254 /*
1255  * Insert jumbo payload option.
1256  */
1257 static int
1258 ip6_insert_jumboopt(exthdrs, plen)
1259 	struct ip6_exthdrs *exthdrs;
1260 	u_int32_t plen;
1261 {
1262 	struct mbuf *mopt;
1263 	u_int8_t *optbuf;
1264 	u_int32_t v;
1265 
1266 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1267 
1268 	/*
1269 	 * If there is no hop-by-hop options header, allocate new one.
1270 	 * If there is one but it doesn't have enough space to store the
1271 	 * jumbo payload option, allocate a cluster to store the whole options.
1272 	 * Otherwise, use it to store the options.
1273 	 */
1274 	if (exthdrs->ip6e_hbh == 0) {
1275 		MGET(mopt, M_DONTWAIT, MT_DATA);
1276 		if (mopt == 0)
1277 			return (ENOBUFS);
1278 		mopt->m_len = JUMBOOPTLEN;
1279 		optbuf = mtod(mopt, u_int8_t *);
1280 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1281 		exthdrs->ip6e_hbh = mopt;
1282 	} else {
1283 		struct ip6_hbh *hbh;
1284 
1285 		mopt = exthdrs->ip6e_hbh;
1286 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1287 			/*
1288 			 * XXX assumption:
1289 			 * - exthdrs->ip6e_hbh is not referenced from places
1290 			 *   other than exthdrs.
1291 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1292 			 */
1293 			int oldoptlen = mopt->m_len;
1294 			struct mbuf *n;
1295 
1296 			/*
1297 			 * XXX: give up if the whole (new) hbh header does
1298 			 * not fit even in an mbuf cluster.
1299 			 */
1300 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1301 				return (ENOBUFS);
1302 
1303 			/*
1304 			 * As a consequence, we must always prepare a cluster
1305 			 * at this point.
1306 			 */
1307 			MGET(n, M_DONTWAIT, MT_DATA);
1308 			if (n) {
1309 				MCLGET(n, M_DONTWAIT);
1310 				if ((n->m_flags & M_EXT) == 0) {
1311 					m_freem(n);
1312 					n = NULL;
1313 				}
1314 			}
1315 			if (!n)
1316 				return (ENOBUFS);
1317 			n->m_len = oldoptlen + JUMBOOPTLEN;
1318 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1319 			    oldoptlen);
1320 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1321 			m_freem(mopt);
1322 			mopt = exthdrs->ip6e_hbh = n;
1323 		} else {
1324 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1325 			mopt->m_len += JUMBOOPTLEN;
1326 		}
1327 		optbuf[0] = IP6OPT_PADN;
1328 		optbuf[1] = 0;
1329 
1330 		/*
1331 		 * Adjust the header length according to the pad and
1332 		 * the jumbo payload option.
1333 		 */
1334 		hbh = mtod(mopt, struct ip6_hbh *);
1335 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1336 	}
1337 
1338 	/* fill in the option. */
1339 	optbuf[2] = IP6OPT_JUMBO;
1340 	optbuf[3] = 4;
1341 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1342 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1343 
1344 	/* finally, adjust the packet header length */
1345 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1346 
1347 	return (0);
1348 #undef JUMBOOPTLEN
1349 }
1350 
1351 /*
1352  * Insert fragment header and copy unfragmentable header portions.
1353  */
1354 static int
1355 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1356 	struct mbuf *m0, *m;
1357 	int hlen;
1358 	struct ip6_frag **frghdrp;
1359 {
1360 	struct mbuf *n, *mlast;
1361 
1362 	if (hlen > sizeof(struct ip6_hdr)) {
1363 		n = m_copym(m0, sizeof(struct ip6_hdr),
1364 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1365 		if (n == 0)
1366 			return (ENOBUFS);
1367 		m->m_next = n;
1368 	} else
1369 		n = m;
1370 
1371 	/* Search for the last mbuf of unfragmentable part. */
1372 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1373 		;
1374 
1375 	if ((mlast->m_flags & M_EXT) == 0 &&
1376 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1377 		/* use the trailing space of the last mbuf for the fragment hdr */
1378 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1379 		    mlast->m_len);
1380 		mlast->m_len += sizeof(struct ip6_frag);
1381 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1382 	} else {
1383 		/* allocate a new mbuf for the fragment header */
1384 		struct mbuf *mfrg;
1385 
1386 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1387 		if (mfrg == 0)
1388 			return (ENOBUFS);
1389 		mfrg->m_len = sizeof(struct ip6_frag);
1390 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1391 		mlast->m_next = mfrg;
1392 	}
1393 
1394 	return (0);
1395 }
1396 
1397 static int
1398 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1399 	struct route_in6 *ro_pmtu, *ro;
1400 	struct ifnet *ifp;
1401 	struct in6_addr *dst;
1402 	u_long *mtup;
1403 	int *alwaysfragp;
1404 {
1405 	u_int32_t mtu = 0;
1406 	int alwaysfrag = 0;
1407 	int error = 0;
1408 
1409 	if (ro_pmtu != ro) {
1410 		/* The first hop and the final destination may differ. */
1411 		struct sockaddr_in6 *sa6_dst =
1412 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1413 		if (ro_pmtu->ro_rt &&
1414 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1415 		      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1416 			RTFREE(ro_pmtu->ro_rt);
1417 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1418 		}
1419 		if (ro_pmtu->ro_rt == NULL) {
1420 			bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */
1421 			sa6_dst->sin6_family = AF_INET6;
1422 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1423 			sa6_dst->sin6_addr = *dst;
1424 
1425 			rtalloc((struct route *)ro_pmtu);
1426 		}
1427 	}
1428 	if (ro_pmtu->ro_rt) {
1429 		u_int32_t ifmtu;
1430 
1431 		if (ifp == NULL)
1432 			ifp = ro_pmtu->ro_rt->rt_ifp;
1433 		ifmtu = IN6_LINKMTU(ifp);
1434 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1435 		if (mtu == 0)
1436 			mtu = ifmtu;
1437 		else if (mtu < IPV6_MMTU) {
1438 			/*
1439 			 * RFC2460 section 5, last paragraph:
1440 			 * if we record ICMPv6 too big message with
1441 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1442 			 * or smaller, with fragment header attached.
1443 			 * (fragment header is needed regardless from the
1444 			 * packet size, for translators to identify packets)
1445 			 */
1446 			alwaysfrag = 1;
1447 			mtu = IPV6_MMTU;
1448 		} else if (mtu > ifmtu) {
1449 			/*
1450 			 * The MTU on the route is larger than the MTU on
1451 			 * the interface!  This shouldn't happen, unless the
1452 			 * MTU of the interface has been changed after the
1453 			 * interface was brought up.  Change the MTU in the
1454 			 * route to match the interface MTU (as long as the
1455 			 * field isn't locked).
1456 			 */
1457 			mtu = ifmtu;
1458 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1459 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1460 		}
1461 	} else if (ifp) {
1462 		mtu = IN6_LINKMTU(ifp);
1463 	} else
1464 		error = EHOSTUNREACH; /* XXX */
1465 
1466 	*mtup = mtu;
1467 	if (alwaysfragp)
1468 		*alwaysfragp = alwaysfrag;
1469 	return (error);
1470 }
1471 
1472 /*
1473  * IP6 socket option processing.
1474  */
1475 int
1476 ip6_ctloutput(op, so, level, optname, mp)
1477 	int op;
1478 	struct socket *so;
1479 	int level, optname;
1480 	struct mbuf **mp;
1481 {
1482 	int privileged, optdatalen, uproto;
1483 	void *optdata;
1484 	struct in6pcb *in6p = sotoin6pcb(so);
1485 	struct mbuf *m = *mp;
1486 	int error, optval;
1487 	int optlen;
1488 	struct lwp *l = curlwp;	/* XXX */
1489 
1490 	optlen = m ? m->m_len : 0;
1491 	error = optval = 0;
1492 	privileged = (l == 0 || kauth_authorize_generic(l->l_cred,
1493 	    KAUTH_GENERIC_ISSUSER, &l->l_acflag)) ? 0 : 1;
1494 	uproto = (int)so->so_proto->pr_protocol;
1495 
1496 	if (level == IPPROTO_IPV6) {
1497 		switch (op) {
1498 		case PRCO_SETOPT:
1499 			switch (optname) {
1500 #ifdef RFC2292
1501 			case IPV6_2292PKTOPTIONS:
1502 				/* m is freed in ip6_pcbopts */
1503 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1504 				    m, so);
1505 				break;
1506 #endif
1507 
1508 			/*
1509 			 * Use of some Hop-by-Hop options or some
1510 			 * Destination options, might require special
1511 			 * privilege.  That is, normal applications
1512 			 * (without special privilege) might be forbidden
1513 			 * from setting certain options in outgoing packets,
1514 			 * and might never see certain options in received
1515 			 * packets. [RFC 2292 Section 6]
1516 			 * KAME specific note:
1517 			 *  KAME prevents non-privileged users from sending or
1518 			 *  receiving ANY hbh/dst options in order to avoid
1519 			 *  overhead of parsing options in the kernel.
1520 			 */
1521 			case IPV6_RECVHOPOPTS:
1522 			case IPV6_RECVDSTOPTS:
1523 			case IPV6_RECVRTHDRDSTOPTS:
1524 				if (!privileged) {
1525 					error = EPERM;
1526 					break;
1527 				}
1528 				/* FALLTHROUGH */
1529 			case IPV6_UNICAST_HOPS:
1530 			case IPV6_HOPLIMIT:
1531 			case IPV6_FAITH:
1532 
1533 			case IPV6_RECVPKTINFO:
1534 			case IPV6_RECVHOPLIMIT:
1535 			case IPV6_RECVRTHDR:
1536 			case IPV6_RECVPATHMTU:
1537 			case IPV6_RECVTCLASS:
1538 			case IPV6_V6ONLY:
1539 				if (optlen != sizeof(int)) {
1540 					error = EINVAL;
1541 					break;
1542 				}
1543 				optval = *mtod(m, int *);
1544 				switch (optname) {
1545 
1546 				case IPV6_UNICAST_HOPS:
1547 					if (optval < -1 || optval >= 256)
1548 						error = EINVAL;
1549 					else {
1550 						/* -1 = kernel default */
1551 						in6p->in6p_hops = optval;
1552 					}
1553 					break;
1554 #define OPTSET(bit) \
1555 do { \
1556 	if (optval) \
1557 		in6p->in6p_flags |= (bit); \
1558 	else \
1559 		in6p->in6p_flags &= ~(bit); \
1560 } while (/*CONSTCOND*/ 0)
1561 
1562 #ifdef RFC2292
1563 #define OPTSET2292(bit) 			\
1564 do { 						\
1565 	in6p->in6p_flags |= IN6P_RFC2292; 	\
1566 	if (optval) 				\
1567 		in6p->in6p_flags |= (bit); 	\
1568 	else 					\
1569 		in6p->in6p_flags &= ~(bit); 	\
1570 } while (/*CONSTCOND*/ 0)
1571 #endif
1572 
1573 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1574 
1575 				case IPV6_RECVPKTINFO:
1576 #ifdef RFC2292
1577 					/* cannot mix with RFC2292 */
1578 					if (OPTBIT(IN6P_RFC2292)) {
1579 						error = EINVAL;
1580 						break;
1581 					}
1582 #endif
1583 					OPTSET(IN6P_PKTINFO);
1584 					break;
1585 
1586 				case IPV6_HOPLIMIT:
1587 				{
1588 					struct ip6_pktopts **optp;
1589 
1590 #ifdef RFC2292
1591 					/* cannot mix with RFC2292 */
1592 					if (OPTBIT(IN6P_RFC2292)) {
1593 						error = EINVAL;
1594 						break;
1595 					}
1596 #endif
1597 					optp = &in6p->in6p_outputopts;
1598 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1599 							   (u_char *)&optval,
1600 							   sizeof(optval),
1601 							   optp,
1602 							   privileged, uproto);
1603 					break;
1604 				}
1605 
1606 				case IPV6_RECVHOPLIMIT:
1607 #ifdef RFC2292
1608 					/* cannot mix with RFC2292 */
1609 					if (OPTBIT(IN6P_RFC2292)) {
1610 						error = EINVAL;
1611 						break;
1612 					}
1613 #endif
1614 					OPTSET(IN6P_HOPLIMIT);
1615 					break;
1616 
1617 				case IPV6_RECVHOPOPTS:
1618 #ifdef RFC2292
1619 					/* cannot mix with RFC2292 */
1620 					if (OPTBIT(IN6P_RFC2292)) {
1621 						error = EINVAL;
1622 						break;
1623 					}
1624 #endif
1625 					OPTSET(IN6P_HOPOPTS);
1626 					break;
1627 
1628 				case IPV6_RECVDSTOPTS:
1629 #ifdef RFC2292
1630 					/* cannot mix with RFC2292 */
1631 					if (OPTBIT(IN6P_RFC2292)) {
1632 						error = EINVAL;
1633 						break;
1634 					}
1635 #endif
1636 					OPTSET(IN6P_DSTOPTS);
1637 					break;
1638 
1639 				case IPV6_RECVRTHDRDSTOPTS:
1640 #ifdef RFC2292
1641 					/* cannot mix with RFC2292 */
1642 					if (OPTBIT(IN6P_RFC2292)) {
1643 						error = EINVAL;
1644 						break;
1645 					}
1646 #endif
1647 					OPTSET(IN6P_RTHDRDSTOPTS);
1648 					break;
1649 
1650 				case IPV6_RECVRTHDR:
1651 #ifdef RFC2292
1652 					/* cannot mix with RFC2292 */
1653 					if (OPTBIT(IN6P_RFC2292)) {
1654 						error = EINVAL;
1655 						break;
1656 					}
1657 #endif
1658 					OPTSET(IN6P_RTHDR);
1659 					break;
1660 
1661 				case IPV6_FAITH:
1662 					OPTSET(IN6P_FAITH);
1663 					break;
1664 
1665 				case IPV6_RECVPATHMTU:
1666 					/*
1667 					 * We ignore this option for TCP
1668 					 * sockets.
1669 					 * (RFC3542 leaves this case
1670 					 * unspecified.)
1671 					 */
1672 					if (uproto != IPPROTO_TCP)
1673 						OPTSET(IN6P_MTU);
1674 					break;
1675 
1676 				case IPV6_V6ONLY:
1677 					/*
1678 					 * make setsockopt(IPV6_V6ONLY)
1679 					 * available only prior to bind(2).
1680 					 * see ipng mailing list, Jun 22 2001.
1681 					 */
1682 					if (in6p->in6p_lport ||
1683 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1684 						error = EINVAL;
1685 						break;
1686 					}
1687 #ifdef INET6_BINDV6ONLY
1688 					if (!optval)
1689 						error = EINVAL;
1690 #else
1691 					OPTSET(IN6P_IPV6_V6ONLY);
1692 #endif
1693 					break;
1694 				case IPV6_RECVTCLASS:
1695 #ifdef RFC2292
1696 					/* cannot mix with RFC2292 XXX */
1697 					if (OPTBIT(IN6P_RFC2292)) {
1698 						error = EINVAL;
1699 						break;
1700 					}
1701 #endif
1702 					OPTSET(IN6P_TCLASS);
1703 					break;
1704 
1705 				}
1706 				break;
1707 
1708 			case IPV6_OTCLASS:
1709 			{
1710 				struct ip6_pktopts **optp;
1711 				u_int8_t tclass;
1712 
1713 				if (optlen != sizeof(tclass)) {
1714 					error = EINVAL;
1715 					break;
1716 				}
1717 				tclass = *mtod(m, u_int8_t *);
1718 				optp = &in6p->in6p_outputopts;
1719 				error = ip6_pcbopt(optname,
1720 						   (u_char *)&tclass,
1721 						   sizeof(tclass),
1722 						   optp,
1723 						   privileged, uproto);
1724 				break;
1725 			}
1726 
1727 			case IPV6_TCLASS:
1728 			case IPV6_DONTFRAG:
1729 			case IPV6_USE_MIN_MTU:
1730 				if (optlen != sizeof(optval)) {
1731 					error = EINVAL;
1732 					break;
1733 				}
1734 				optval = *mtod(m, int *);
1735 				{
1736 					struct ip6_pktopts **optp;
1737 					optp = &in6p->in6p_outputopts;
1738 					error = ip6_pcbopt(optname,
1739 							   (u_char *)&optval,
1740 							   sizeof(optval),
1741 							   optp,
1742 							   privileged, uproto);
1743 					break;
1744 				}
1745 
1746 #ifdef RFC2292
1747 			case IPV6_2292PKTINFO:
1748 			case IPV6_2292HOPLIMIT:
1749 			case IPV6_2292HOPOPTS:
1750 			case IPV6_2292DSTOPTS:
1751 			case IPV6_2292RTHDR:
1752 				/* RFC 2292 */
1753 				if (optlen != sizeof(int)) {
1754 					error = EINVAL;
1755 					break;
1756 				}
1757 				optval = *mtod(m, int *);
1758 				switch (optname) {
1759 				case IPV6_2292PKTINFO:
1760 					OPTSET2292(IN6P_PKTINFO);
1761 					break;
1762 				case IPV6_2292HOPLIMIT:
1763 					OPTSET2292(IN6P_HOPLIMIT);
1764 					break;
1765 				case IPV6_2292HOPOPTS:
1766 					/*
1767 					 * Check super-user privilege.
1768 					 * See comments for IPV6_RECVHOPOPTS.
1769 					 */
1770 					if (!privileged)
1771 						return (EPERM);
1772 					OPTSET2292(IN6P_HOPOPTS);
1773 					break;
1774 				case IPV6_2292DSTOPTS:
1775 					if (!privileged)
1776 						return (EPERM);
1777 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1778 					break;
1779 				case IPV6_2292RTHDR:
1780 					OPTSET2292(IN6P_RTHDR);
1781 					break;
1782 				}
1783 				break;
1784 #endif
1785 			case IPV6_PKTINFO:
1786 			case IPV6_HOPOPTS:
1787 			case IPV6_RTHDR:
1788 			case IPV6_DSTOPTS:
1789 			case IPV6_RTHDRDSTOPTS:
1790 			case IPV6_NEXTHOP:
1791 			{
1792 				/* new advanced API (RFC3542) */
1793 				u_char *optbuf;
1794 				int optbuflen;
1795 				struct ip6_pktopts **optp;
1796 
1797 #ifdef RFC2292
1798 				/* cannot mix with RFC2292 */
1799 				if (OPTBIT(IN6P_RFC2292)) {
1800 					error = EINVAL;
1801 					break;
1802 				}
1803 #endif
1804 
1805 				if (m && m->m_next) {
1806 					error = EINVAL;	/* XXX */
1807 					break;
1808 				}
1809 				if (m) {
1810 					optbuf = mtod(m, u_char *);
1811 					optbuflen = m->m_len;
1812 				} else {
1813 					optbuf = NULL;
1814 					optbuflen = 0;
1815 				}
1816 				optp = &in6p->in6p_outputopts;
1817 				error = ip6_pcbopt(optname,
1818 						   optbuf, optbuflen,
1819 						   optp, privileged, uproto);
1820 				break;
1821 			}
1822 #undef OPTSET
1823 
1824 			case IPV6_MULTICAST_IF:
1825 			case IPV6_MULTICAST_HOPS:
1826 			case IPV6_MULTICAST_LOOP:
1827 			case IPV6_JOIN_GROUP:
1828 			case IPV6_LEAVE_GROUP:
1829                                 error = ip6_setmoptions(optname,
1830 				    &in6p->in6p_moptions, m);
1831 				break;
1832 
1833 			case IPV6_PORTRANGE:
1834 				optval = *mtod(m, int *);
1835 
1836 				switch (optval) {
1837 				case IPV6_PORTRANGE_DEFAULT:
1838 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1839 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1840 					break;
1841 
1842 				case IPV6_PORTRANGE_HIGH:
1843 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1844 					in6p->in6p_flags |= IN6P_HIGHPORT;
1845 					break;
1846 
1847 				case IPV6_PORTRANGE_LOW:
1848 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1849 					in6p->in6p_flags |= IN6P_LOWPORT;
1850 					break;
1851 
1852 				default:
1853 					error = EINVAL;
1854 					break;
1855 				}
1856 				break;
1857 
1858 #ifdef IPSEC
1859 			case IPV6_IPSEC_POLICY:
1860 			{
1861 				caddr_t req = NULL;
1862 				size_t len = 0;
1863 				if (m) {
1864 					req = mtod(m, caddr_t);
1865 					len = m->m_len;
1866 				}
1867 				error = ipsec6_set_policy(in6p, optname, req,
1868 							  len, privileged);
1869 			}
1870 				break;
1871 #endif /* IPSEC */
1872 
1873 			default:
1874 				error = ENOPROTOOPT;
1875 				break;
1876 			}
1877 			if (m)
1878 				(void)m_free(m);
1879 			break;
1880 
1881 		case PRCO_GETOPT:
1882 			switch (optname) {
1883 #ifdef RFC2292
1884 			case IPV6_2292PKTOPTIONS:
1885 				/*
1886 				 * RFC3542 (effectively) deprecated the
1887 				 * semantics of the 2292-style pktoptions.
1888 				 * Since it was not reliable in nature (i.e.,
1889 				 * applications had to expect the lack of some
1890 				 * information after all), it would make sense
1891 				 * to simplify this part by always returning
1892 				 * empty data.
1893 				 */
1894 				*mp = m_get(M_WAIT, MT_SOOPTS);
1895 				(*mp)->m_len = 0;
1896 				break;
1897 #endif
1898 
1899 			case IPV6_RECVHOPOPTS:
1900 			case IPV6_RECVDSTOPTS:
1901 			case IPV6_RECVRTHDRDSTOPTS:
1902 			case IPV6_UNICAST_HOPS:
1903 			case IPV6_RECVPKTINFO:
1904 			case IPV6_RECVHOPLIMIT:
1905 			case IPV6_RECVRTHDR:
1906 			case IPV6_RECVPATHMTU:
1907 
1908 			case IPV6_FAITH:
1909 			case IPV6_V6ONLY:
1910 			case IPV6_PORTRANGE:
1911 			case IPV6_RECVTCLASS:
1912 				switch (optname) {
1913 
1914 				case IPV6_RECVHOPOPTS:
1915 					optval = OPTBIT(IN6P_HOPOPTS);
1916 					break;
1917 
1918 				case IPV6_RECVDSTOPTS:
1919 					optval = OPTBIT(IN6P_DSTOPTS);
1920 					break;
1921 
1922 				case IPV6_RECVRTHDRDSTOPTS:
1923 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1924 					break;
1925 
1926 				case IPV6_UNICAST_HOPS:
1927 					optval = in6p->in6p_hops;
1928 					break;
1929 
1930 				case IPV6_RECVPKTINFO:
1931 					optval = OPTBIT(IN6P_PKTINFO);
1932 					break;
1933 
1934 				case IPV6_RECVHOPLIMIT:
1935 					optval = OPTBIT(IN6P_HOPLIMIT);
1936 					break;
1937 
1938 				case IPV6_RECVRTHDR:
1939 					optval = OPTBIT(IN6P_RTHDR);
1940 					break;
1941 
1942 				case IPV6_RECVPATHMTU:
1943 					optval = OPTBIT(IN6P_MTU);
1944 					break;
1945 
1946 				case IPV6_FAITH:
1947 					optval = OPTBIT(IN6P_FAITH);
1948 					break;
1949 
1950 				case IPV6_V6ONLY:
1951 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1952 					break;
1953 
1954 				case IPV6_PORTRANGE:
1955 				    {
1956 					int flags;
1957 					flags = in6p->in6p_flags;
1958 					if (flags & IN6P_HIGHPORT)
1959 						optval = IPV6_PORTRANGE_HIGH;
1960 					else if (flags & IN6P_LOWPORT)
1961 						optval = IPV6_PORTRANGE_LOW;
1962 					else
1963 						optval = 0;
1964 					break;
1965 				    }
1966 				case IPV6_RECVTCLASS:
1967 					optval = OPTBIT(IN6P_TCLASS);
1968 					break;
1969 
1970 				}
1971 				if (error)
1972 					break;
1973 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1974 				m->m_len = sizeof(int);
1975 				*mtod(m, int *) = optval;
1976 				break;
1977 
1978 			case IPV6_PATHMTU:
1979 			    {
1980 				u_long pmtu = 0;
1981 				struct ip6_mtuinfo mtuinfo;
1982 				struct route_in6 *ro = (struct route_in6 *)&in6p
1983 ->in6p_route;
1984 
1985 				if (!(so->so_state & SS_ISCONNECTED))
1986 					return (ENOTCONN);
1987 				/*
1988 				 * XXX: we dot not consider the case of source
1989 				 * routing, or optional information to specify
1990 				 * the outgoing interface.
1991 				 */
1992 				error = ip6_getpmtu(ro, NULL, NULL,
1993 				    &in6p->in6p_faddr, &pmtu, NULL);
1994 				if (error)
1995 					break;
1996 				if (pmtu > IPV6_MAXPACKET)
1997 					pmtu = IPV6_MAXPACKET;
1998 
1999 				memset(&mtuinfo, 0, sizeof(mtuinfo));
2000 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2001 				optdata = (void *)&mtuinfo;
2002 				optdatalen = sizeof(mtuinfo);
2003 				if (optdatalen > MCLBYTES)
2004 					return (EMSGSIZE); /* XXX */
2005 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
2006 				if (optdatalen > MLEN)
2007 					MCLGET(m, M_WAIT);
2008 				m->m_len = optdatalen;
2009 				memcpy(mtod(m, void *), optdata, optdatalen);
2010 				break;
2011 			    }
2012 
2013 #ifdef RFC2292
2014 			case IPV6_2292PKTINFO:
2015 			case IPV6_2292HOPLIMIT:
2016 			case IPV6_2292HOPOPTS:
2017 			case IPV6_2292RTHDR:
2018 			case IPV6_2292DSTOPTS:
2019 				switch (optname) {
2020 				case IPV6_2292PKTINFO:
2021 					optval = OPTBIT(IN6P_PKTINFO);
2022 					break;
2023 				case IPV6_2292HOPLIMIT:
2024 					optval = OPTBIT(IN6P_HOPLIMIT);
2025 					break;
2026 				case IPV6_2292HOPOPTS:
2027 					optval = OPTBIT(IN6P_HOPOPTS);
2028 					break;
2029 				case IPV6_2292RTHDR:
2030 					optval = OPTBIT(IN6P_RTHDR);
2031 					break;
2032 				case IPV6_2292DSTOPTS:
2033 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2034 					break;
2035 				}
2036 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
2037 				m->m_len = sizeof(int);
2038 				*mtod(m, int *) = optval;
2039 				break;
2040 #endif
2041 			case IPV6_PKTINFO:
2042 			case IPV6_HOPOPTS:
2043 			case IPV6_RTHDR:
2044 			case IPV6_DSTOPTS:
2045 			case IPV6_RTHDRDSTOPTS:
2046 			case IPV6_NEXTHOP:
2047 			case IPV6_OTCLASS:
2048 			case IPV6_TCLASS:
2049 			case IPV6_DONTFRAG:
2050 			case IPV6_USE_MIN_MTU:
2051 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2052 				    optname, mp);
2053 				break;
2054 
2055 			case IPV6_MULTICAST_IF:
2056 			case IPV6_MULTICAST_HOPS:
2057 			case IPV6_MULTICAST_LOOP:
2058 			case IPV6_JOIN_GROUP:
2059 			case IPV6_LEAVE_GROUP:
2060 				error = ip6_getmoptions(optname,
2061 				    in6p->in6p_moptions, mp);
2062 				break;
2063 
2064 #ifdef IPSEC
2065 			case IPV6_IPSEC_POLICY:
2066 			    {
2067 				caddr_t req = NULL;
2068 				size_t len = 0;
2069 				if (m) {
2070 					req = mtod(m, caddr_t);
2071 					len = m->m_len;
2072 				}
2073 				error = ipsec6_get_policy(in6p, req, len, mp);
2074 				break;
2075 			    }
2076 #endif /* IPSEC */
2077 
2078 
2079 
2080 
2081 			default:
2082 				error = ENOPROTOOPT;
2083 				break;
2084 			}
2085 			break;
2086 		}
2087 	} else {
2088 		error = EINVAL;
2089 		if (op == PRCO_SETOPT && *mp)
2090 			(void)m_free(*mp);
2091 	}
2092 	return (error);
2093 }
2094 
2095 int
2096 ip6_raw_ctloutput(op, so, level, optname, mp)
2097 	int op;
2098 	struct socket *so;
2099 	int level, optname;
2100 	struct mbuf **mp;
2101 {
2102 	int error = 0, optval, optlen;
2103 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2104 	struct in6pcb *in6p = sotoin6pcb(so);
2105 	struct mbuf *m = *mp;
2106 
2107 	optlen = m ? m->m_len : 0;
2108 
2109 	if (level != IPPROTO_IPV6) {
2110 		if (op == PRCO_SETOPT && *mp)
2111 			(void)m_free(*mp);
2112 		return (EINVAL);
2113 	}
2114 
2115 	switch (optname) {
2116 	case IPV6_CHECKSUM:
2117 		/*
2118 		 * For ICMPv6 sockets, no modification allowed for checksum
2119 		 * offset, permit "no change" values to help existing apps.
2120 		 *
2121 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
2122 		 * for an ICMPv6 socket will fail."  The current
2123 		 * behavior does not meet RFC3542.
2124 		 */
2125 		switch (op) {
2126 		case PRCO_SETOPT:
2127 			if (optlen != sizeof(int)) {
2128 				error = EINVAL;
2129 				break;
2130 			}
2131 			optval = *mtod(m, int *);
2132 			if ((optval % 2) != 0) {
2133 				/* the API assumes even offset values */
2134 				error = EINVAL;
2135 			} else if (so->so_proto->pr_protocol ==
2136 			    IPPROTO_ICMPV6) {
2137 				if (optval != icmp6off)
2138 					error = EINVAL;
2139 			} else
2140 				in6p->in6p_cksum = optval;
2141 			break;
2142 
2143 		case PRCO_GETOPT:
2144 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2145 				optval = icmp6off;
2146 			else
2147 				optval = in6p->in6p_cksum;
2148 
2149 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2150 			m->m_len = sizeof(int);
2151 			*mtod(m, int *) = optval;
2152 			break;
2153 
2154 		default:
2155 			error = EINVAL;
2156 			break;
2157 		}
2158 		break;
2159 
2160 	default:
2161 		error = ENOPROTOOPT;
2162 		break;
2163 	}
2164 
2165 	if (op == PRCO_SETOPT && m)
2166 		(void)m_free(m);
2167 
2168 	return (error);
2169 }
2170 
2171 #ifdef RFC2292
2172 /*
2173  * Set up IP6 options in pcb for insertion in output packets or
2174  * specifying behavior of outgoing packets.
2175  */
2176 static int
2177 ip6_pcbopts(pktopt, m, so)
2178 	struct ip6_pktopts **pktopt;
2179 	struct mbuf *m;
2180 	struct socket *so;
2181 {
2182 	struct ip6_pktopts *opt = *pktopt;
2183 	int error = 0;
2184 	struct lwp *l = curlwp;	/* XXX */
2185 	int priv = 0;
2186 
2187 	/* turn off any old options. */
2188 	if (opt) {
2189 #ifdef DIAGNOSTIC
2190 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2191 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2192 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2193 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2194 #endif
2195 		ip6_clearpktopts(opt, -1);
2196 	} else
2197 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2198 	*pktopt = NULL;
2199 
2200 	if (!m || m->m_len == 0) {
2201 		/*
2202 		 * Only turning off any previous options, regardless of
2203 		 * whether the opt is just created or given.
2204 		 */
2205 		free(opt, M_IP6OPT);
2206 		return (0);
2207 	}
2208 
2209 	/*  set options specified by user. */
2210 	if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2211 	    &l->l_acflag))
2212 		priv = 1;
2213 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
2214 	    so->so_proto->pr_protocol)) != 0) {
2215 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2216 		free(opt, M_IP6OPT);
2217 		return (error);
2218 	}
2219 	*pktopt = opt;
2220 	return (0);
2221 }
2222 #endif
2223 
2224 /*
2225  * initialize ip6_pktopts.  beware that there are non-zero default values in
2226  * the struct.
2227  */
2228 void
2229 ip6_initpktopts(struct ip6_pktopts *opt)
2230 {
2231 
2232 	memset(opt, 0, sizeof(*opt));
2233 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2234 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2235 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2236 }
2237 
2238 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2239 static int
2240 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2241     int priv, int uproto)
2242 {
2243 	struct ip6_pktopts *opt;
2244 
2245 	if (*pktopt == NULL) {
2246 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2247 		    M_WAITOK);
2248 		ip6_initpktopts(*pktopt);
2249 	}
2250 	opt = *pktopt;
2251 
2252 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2253 }
2254 
2255 static int
2256 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
2257 {
2258 	void *optdata = NULL;
2259 	int optdatalen = 0;
2260 	struct ip6_ext *ip6e;
2261 	int error = 0;
2262 	struct in6_pktinfo null_pktinfo;
2263 	int deftclass = 0, on;
2264 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2265 	struct mbuf *m;
2266 
2267 	switch (optname) {
2268 	case IPV6_PKTINFO:
2269 		if (pktopt && pktopt->ip6po_pktinfo)
2270 			optdata = (void *)pktopt->ip6po_pktinfo;
2271 		else {
2272 			/* XXX: we don't have to do this every time... */
2273 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2274 			optdata = (void *)&null_pktinfo;
2275 		}
2276 		optdatalen = sizeof(struct in6_pktinfo);
2277 		break;
2278 	case IPV6_OTCLASS:
2279 		/* XXX */
2280 		return (EINVAL);
2281 	case IPV6_TCLASS:
2282 		if (pktopt && pktopt->ip6po_tclass >= 0)
2283 			optdata = (void *)&pktopt->ip6po_tclass;
2284 		else
2285 			optdata = (void *)&deftclass;
2286 		optdatalen = sizeof(int);
2287 		break;
2288 	case IPV6_HOPOPTS:
2289 		if (pktopt && pktopt->ip6po_hbh) {
2290 			optdata = (void *)pktopt->ip6po_hbh;
2291 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2292 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2293 		}
2294 		break;
2295 	case IPV6_RTHDR:
2296 		if (pktopt && pktopt->ip6po_rthdr) {
2297 			optdata = (void *)pktopt->ip6po_rthdr;
2298 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2299 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2300 		}
2301 		break;
2302 	case IPV6_RTHDRDSTOPTS:
2303 		if (pktopt && pktopt->ip6po_dest1) {
2304 			optdata = (void *)pktopt->ip6po_dest1;
2305 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2306 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2307 		}
2308 		break;
2309 	case IPV6_DSTOPTS:
2310 		if (pktopt && pktopt->ip6po_dest2) {
2311 			optdata = (void *)pktopt->ip6po_dest2;
2312 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2313 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2314 		}
2315 		break;
2316 	case IPV6_NEXTHOP:
2317 		if (pktopt && pktopt->ip6po_nexthop) {
2318 			optdata = (void *)pktopt->ip6po_nexthop;
2319 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2320 		}
2321 		break;
2322 	case IPV6_USE_MIN_MTU:
2323 		if (pktopt)
2324 			optdata = (void *)&pktopt->ip6po_minmtu;
2325 		else
2326 			optdata = (void *)&defminmtu;
2327 		optdatalen = sizeof(int);
2328 		break;
2329 	case IPV6_DONTFRAG:
2330 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2331 			on = 1;
2332 		else
2333 			on = 0;
2334 		optdata = (void *)&on;
2335 		optdatalen = sizeof(on);
2336 		break;
2337 	default:		/* should not happen */
2338 #ifdef DIAGNOSTIC
2339 		panic("ip6_getpcbopt: unexpected option\n");
2340 #endif
2341 		return (ENOPROTOOPT);
2342 	}
2343 
2344 	if (optdatalen > MCLBYTES)
2345 		return (EMSGSIZE); /* XXX */
2346 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
2347 	if (optdatalen > MLEN)
2348 		MCLGET(m, M_WAIT);
2349 	m->m_len = optdatalen;
2350 	if (optdatalen)
2351 		memcpy(mtod(m, void *), optdata, optdatalen);
2352 
2353 	return (error);
2354 }
2355 
2356 void
2357 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2358 {
2359 	if (optname == -1 || optname == IPV6_PKTINFO) {
2360 		if (pktopt->ip6po_pktinfo)
2361 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2362 		pktopt->ip6po_pktinfo = NULL;
2363 	}
2364 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2365 		pktopt->ip6po_hlim = -1;
2366 	if (optname == -1 || optname == IPV6_TCLASS)
2367 		pktopt->ip6po_tclass = -1;
2368 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2369 		if (pktopt->ip6po_nextroute.ro_rt) {
2370 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2371 			pktopt->ip6po_nextroute.ro_rt = NULL;
2372 		}
2373 		if (pktopt->ip6po_nexthop)
2374 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2375 		pktopt->ip6po_nexthop = NULL;
2376 	}
2377 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2378 		if (pktopt->ip6po_hbh)
2379 			free(pktopt->ip6po_hbh, M_IP6OPT);
2380 		pktopt->ip6po_hbh = NULL;
2381 	}
2382 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2383 		if (pktopt->ip6po_dest1)
2384 			free(pktopt->ip6po_dest1, M_IP6OPT);
2385 		pktopt->ip6po_dest1 = NULL;
2386 	}
2387 	if (optname == -1 || optname == IPV6_RTHDR) {
2388 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2389 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2390 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2391 		if (pktopt->ip6po_route.ro_rt) {
2392 			RTFREE(pktopt->ip6po_route.ro_rt);
2393 			pktopt->ip6po_route.ro_rt = NULL;
2394 		}
2395 	}
2396 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2397 		if (pktopt->ip6po_dest2)
2398 			free(pktopt->ip6po_dest2, M_IP6OPT);
2399 		pktopt->ip6po_dest2 = NULL;
2400 	}
2401 }
2402 
2403 #define PKTOPT_EXTHDRCPY(type) 					\
2404 do {								\
2405 	if (src->type) {					\
2406 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2407 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2408 		if (dst->type == NULL && canwait == M_NOWAIT)	\
2409 			goto bad;				\
2410 		memcpy(dst->type, src->type, hlen);		\
2411 	}							\
2412 } while (/*CONSTCOND*/ 0)
2413 
2414 static int
2415 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2416 {
2417 	dst->ip6po_hlim = src->ip6po_hlim;
2418 	dst->ip6po_tclass = src->ip6po_tclass;
2419 	dst->ip6po_flags = src->ip6po_flags;
2420 	if (src->ip6po_pktinfo) {
2421 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2422 		    M_IP6OPT, canwait);
2423 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2424 			goto bad;
2425 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2426 	}
2427 	if (src->ip6po_nexthop) {
2428 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2429 		    M_IP6OPT, canwait);
2430 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
2431 			goto bad;
2432 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2433 		    src->ip6po_nexthop->sa_len);
2434 	}
2435 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2436 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2437 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2438 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2439 	return (0);
2440 
2441   bad:
2442 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2443 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2444 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2445 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2446 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2447 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2448 
2449 	return (ENOBUFS);
2450 }
2451 #undef PKTOPT_EXTHDRCPY
2452 
2453 struct ip6_pktopts *
2454 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2455 {
2456 	int error;
2457 	struct ip6_pktopts *dst;
2458 
2459 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2460 	if (dst == NULL && canwait == M_NOWAIT)
2461 		return (NULL);
2462 	ip6_initpktopts(dst);
2463 
2464 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2465 		free(dst, M_IP6OPT);
2466 		return (NULL);
2467 	}
2468 
2469 	return (dst);
2470 }
2471 
2472 void
2473 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2474 {
2475 	if (pktopt == NULL)
2476 		return;
2477 
2478 	ip6_clearpktopts(pktopt, -1);
2479 
2480 	free(pktopt, M_IP6OPT);
2481 }
2482 
2483 /*
2484  * Set the IP6 multicast options in response to user setsockopt().
2485  */
2486 static int
2487 ip6_setmoptions(optname, im6op, m)
2488 	int optname;
2489 	struct ip6_moptions **im6op;
2490 	struct mbuf *m;
2491 {
2492 	int error = 0;
2493 	u_int loop, ifindex;
2494 	struct ipv6_mreq *mreq;
2495 	struct ifnet *ifp;
2496 	struct ip6_moptions *im6o = *im6op;
2497 	struct route_in6 ro;
2498 	struct in6_multi_mship *imm;
2499 	struct lwp *l = curlwp;	/* XXX */
2500 
2501 	if (im6o == NULL) {
2502 		/*
2503 		 * No multicast option buffer attached to the pcb;
2504 		 * allocate one and initialize to default values.
2505 		 */
2506 		im6o = (struct ip6_moptions *)
2507 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2508 
2509 		if (im6o == NULL)
2510 			return (ENOBUFS);
2511 		*im6op = im6o;
2512 		im6o->im6o_multicast_ifp = NULL;
2513 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2514 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2515 		LIST_INIT(&im6o->im6o_memberships);
2516 	}
2517 
2518 	switch (optname) {
2519 
2520 	case IPV6_MULTICAST_IF:
2521 		/*
2522 		 * Select the interface for outgoing multicast packets.
2523 		 */
2524 		if (m == NULL || m->m_len != sizeof(u_int)) {
2525 			error = EINVAL;
2526 			break;
2527 		}
2528 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2529 		if (ifindex != 0) {
2530 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2531 				error = ENXIO;	/* XXX EINVAL? */
2532 				break;
2533 			}
2534 			ifp = ifindex2ifnet[ifindex];
2535 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2536 				error = EADDRNOTAVAIL;
2537 				break;
2538 			}
2539 		} else
2540 			ifp = NULL;
2541 		im6o->im6o_multicast_ifp = ifp;
2542 		break;
2543 
2544 	case IPV6_MULTICAST_HOPS:
2545 	    {
2546 		/*
2547 		 * Set the IP6 hoplimit for outgoing multicast packets.
2548 		 */
2549 		int optval;
2550 		if (m == NULL || m->m_len != sizeof(int)) {
2551 			error = EINVAL;
2552 			break;
2553 		}
2554 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2555 		if (optval < -1 || optval >= 256)
2556 			error = EINVAL;
2557 		else if (optval == -1)
2558 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2559 		else
2560 			im6o->im6o_multicast_hlim = optval;
2561 		break;
2562 	    }
2563 
2564 	case IPV6_MULTICAST_LOOP:
2565 		/*
2566 		 * Set the loopback flag for outgoing multicast packets.
2567 		 * Must be zero or one.
2568 		 */
2569 		if (m == NULL || m->m_len != sizeof(u_int)) {
2570 			error = EINVAL;
2571 			break;
2572 		}
2573 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2574 		if (loop > 1) {
2575 			error = EINVAL;
2576 			break;
2577 		}
2578 		im6o->im6o_multicast_loop = loop;
2579 		break;
2580 
2581 	case IPV6_JOIN_GROUP:
2582 		/*
2583 		 * Add a multicast group membership.
2584 		 * Group must be a valid IP6 multicast address.
2585 		 */
2586 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2587 			error = EINVAL;
2588 			break;
2589 		}
2590 		mreq = mtod(m, struct ipv6_mreq *);
2591 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2592 			/*
2593 			 * We use the unspecified address to specify to accept
2594 			 * all multicast addresses. Only super user is allowed
2595 			 * to do this.
2596 			 */
2597 			if (kauth_authorize_generic(l->l_cred,
2598 			    KAUTH_GENERIC_ISSUSER, &l->l_acflag))
2599 			{
2600 				error = EACCES;
2601 				break;
2602 			}
2603 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2604 			error = EINVAL;
2605 			break;
2606 		}
2607 
2608 		/*
2609 		 * If no interface was explicitly specified, choose an
2610 		 * appropriate one according to the given multicast address.
2611 		 */
2612 		if (mreq->ipv6mr_interface == 0) {
2613 			struct sockaddr_in6 *dst;
2614 
2615 			/*
2616 			 * Look up the routing table for the
2617 			 * address, and choose the outgoing interface.
2618 			 *   XXX: is it a good approach?
2619 			 */
2620 			ro.ro_rt = NULL;
2621 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2622 			bzero(dst, sizeof(*dst));
2623 			dst->sin6_family = AF_INET6;
2624 			dst->sin6_len = sizeof(*dst);
2625 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2626 			rtalloc((struct route *)&ro);
2627 			if (ro.ro_rt == NULL) {
2628 				error = EADDRNOTAVAIL;
2629 				break;
2630 			}
2631 			ifp = ro.ro_rt->rt_ifp;
2632 			rtfree(ro.ro_rt);
2633 		} else {
2634 			/*
2635 			 * If the interface is specified, validate it.
2636 			 */
2637 			if (if_indexlim <= mreq->ipv6mr_interface ||
2638 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2639 				error = ENXIO;	/* XXX EINVAL? */
2640 				break;
2641 			}
2642 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2643 		}
2644 
2645 		/*
2646 		 * See if we found an interface, and confirm that it
2647 		 * supports multicast
2648 		 */
2649 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2650 			error = EADDRNOTAVAIL;
2651 			break;
2652 		}
2653 
2654 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2655 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2656 			break;
2657 		}
2658 
2659 		/*
2660 		 * See if the membership already exists.
2661 		 */
2662 		for (imm = im6o->im6o_memberships.lh_first;
2663 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2664 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2665 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2666 			    &mreq->ipv6mr_multiaddr))
2667 				break;
2668 		if (imm != NULL) {
2669 			error = EADDRINUSE;
2670 			break;
2671 		}
2672 		/*
2673 		 * Everything looks good; add a new record to the multicast
2674 		 * address list for the given interface.
2675 		 */
2676 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0);
2677 		if (imm == NULL)
2678 			break;
2679 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2680 		break;
2681 
2682 	case IPV6_LEAVE_GROUP:
2683 		/*
2684 		 * Drop a multicast group membership.
2685 		 * Group must be a valid IP6 multicast address.
2686 		 */
2687 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2688 			error = EINVAL;
2689 			break;
2690 		}
2691 		mreq = mtod(m, struct ipv6_mreq *);
2692 
2693 		/*
2694 		 * If an interface address was specified, get a pointer
2695 		 * to its ifnet structure.
2696 		 */
2697 		if (mreq->ipv6mr_interface != 0) {
2698 			if (if_indexlim <= mreq->ipv6mr_interface ||
2699 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2700 				error = ENXIO;	/* XXX EINVAL? */
2701 				break;
2702 			}
2703 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2704 		} else
2705 			ifp = NULL;
2706 
2707 		/* Fill in the scope zone ID */
2708 		if (ifp) {
2709 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2710 				/* XXX: should not happen */
2711 				error = EADDRNOTAVAIL;
2712 				break;
2713 			}
2714 		} else if (mreq->ipv6mr_interface != 0) {
2715 			/*
2716 			 * XXX: This case would happens when the (positive)
2717 			 * index is in the valid range, but the corresponding
2718 			 * interface has been detached dynamically.  The above
2719 			 * check probably avoids such case to happen here, but
2720 			 * we check it explicitly for safety.
2721 			 */
2722 			error = EADDRNOTAVAIL;
2723 			break;
2724 		} else {	/* ipv6mr_interface == 0 */
2725 			struct sockaddr_in6 sa6_mc;
2726 
2727 			/*
2728 			 * The API spec says as follows:
2729 			 *  If the interface index is specified as 0, the
2730 			 *  system may choose a multicast group membership to
2731 			 *  drop by matching the multicast address only.
2732 			 * On the other hand, we cannot disambiguate the scope
2733 			 * zone unless an interface is provided.  Thus, we
2734 			 * check if there's ambiguity with the default scope
2735 			 * zone as the last resort.
2736 			 */
2737 			bzero(&sa6_mc, sizeof(sa6_mc));
2738 			sa6_mc.sin6_family = AF_INET6;
2739 			sa6_mc.sin6_len = sizeof(sa6_mc);
2740 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2741 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2742 			if (error != 0)
2743 				break;
2744 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2745 		}
2746 
2747 		/*
2748 		 * Find the membership in the membership list.
2749 		 */
2750 		for (imm = im6o->im6o_memberships.lh_first;
2751 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2752 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2753 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2754 			    &mreq->ipv6mr_multiaddr))
2755 				break;
2756 		}
2757 		if (imm == NULL) {
2758 			/* Unable to resolve interface */
2759 			error = EADDRNOTAVAIL;
2760 			break;
2761 		}
2762 		/*
2763 		 * Give up the multicast address record to which the
2764 		 * membership points.
2765 		 */
2766 		LIST_REMOVE(imm, i6mm_chain);
2767 		in6_leavegroup(imm);
2768 		break;
2769 
2770 	default:
2771 		error = EOPNOTSUPP;
2772 		break;
2773 	}
2774 
2775 	/*
2776 	 * If all options have default values, no need to keep the mbuf.
2777 	 */
2778 	if (im6o->im6o_multicast_ifp == NULL &&
2779 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2780 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2781 	    im6o->im6o_memberships.lh_first == NULL) {
2782 		free(*im6op, M_IPMOPTS);
2783 		*im6op = NULL;
2784 	}
2785 
2786 	return (error);
2787 }
2788 
2789 /*
2790  * Return the IP6 multicast options in response to user getsockopt().
2791  */
2792 static int
2793 ip6_getmoptions(optname, im6o, mp)
2794 	int optname;
2795 	struct ip6_moptions *im6o;
2796 	struct mbuf **mp;
2797 {
2798 	u_int *hlim, *loop, *ifindex;
2799 
2800 	*mp = m_get(M_WAIT, MT_SOOPTS);
2801 
2802 	switch (optname) {
2803 
2804 	case IPV6_MULTICAST_IF:
2805 		ifindex = mtod(*mp, u_int *);
2806 		(*mp)->m_len = sizeof(u_int);
2807 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2808 			*ifindex = 0;
2809 		else
2810 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2811 		return (0);
2812 
2813 	case IPV6_MULTICAST_HOPS:
2814 		hlim = mtod(*mp, u_int *);
2815 		(*mp)->m_len = sizeof(u_int);
2816 		if (im6o == NULL)
2817 			*hlim = ip6_defmcasthlim;
2818 		else
2819 			*hlim = im6o->im6o_multicast_hlim;
2820 		return (0);
2821 
2822 	case IPV6_MULTICAST_LOOP:
2823 		loop = mtod(*mp, u_int *);
2824 		(*mp)->m_len = sizeof(u_int);
2825 		if (im6o == NULL)
2826 			*loop = ip6_defmcasthlim;
2827 		else
2828 			*loop = im6o->im6o_multicast_loop;
2829 		return (0);
2830 
2831 	default:
2832 		return (EOPNOTSUPP);
2833 	}
2834 }
2835 
2836 /*
2837  * Discard the IP6 multicast options.
2838  */
2839 void
2840 ip6_freemoptions(im6o)
2841 	struct ip6_moptions *im6o;
2842 {
2843 	struct in6_multi_mship *imm;
2844 
2845 	if (im6o == NULL)
2846 		return;
2847 
2848 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2849 		LIST_REMOVE(imm, i6mm_chain);
2850 		in6_leavegroup(imm);
2851 	}
2852 	free(im6o, M_IPMOPTS);
2853 }
2854 
2855 /*
2856  * Set IPv6 outgoing packet options based on advanced API.
2857  */
2858 int
2859 ip6_setpktopts(control, opt, stickyopt, priv, uproto)
2860 	struct mbuf *control;
2861 	struct ip6_pktopts *opt, *stickyopt;
2862 	int priv, uproto;
2863 {
2864 	struct cmsghdr *cm = 0;
2865 
2866 	if (control == NULL || opt == NULL)
2867 		return (EINVAL);
2868 
2869 	ip6_initpktopts(opt);
2870 	if (stickyopt) {
2871 		int error;
2872 
2873 		/*
2874 		 * If stickyopt is provided, make a local copy of the options
2875 		 * for this particular packet, then override them by ancillary
2876 		 * objects.
2877 		 * XXX: copypktopts() does not copy the cached route to a next
2878 		 * hop (if any).  This is not very good in terms of efficiency,
2879 		 * but we can allow this since this option should be rarely
2880 		 * used.
2881 		 */
2882 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2883 			return (error);
2884 	}
2885 
2886 	/*
2887 	 * XXX: Currently, we assume all the optional information is stored
2888 	 * in a single mbuf.
2889 	 */
2890 	if (control->m_next)
2891 		return (EINVAL);
2892 
2893 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2894 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2895 		int error;
2896 
2897 		if (control->m_len < CMSG_LEN(0))
2898 			return (EINVAL);
2899 
2900 		cm = mtod(control, struct cmsghdr *);
2901 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2902 			return (EINVAL);
2903 		if (cm->cmsg_level != IPPROTO_IPV6)
2904 			continue;
2905 
2906 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2907 		    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2908 		if (error)
2909 			return (error);
2910 	}
2911 
2912 	return (0);
2913 }
2914 
2915 /*
2916  * Set a particular packet option, as a sticky option or an ancillary data
2917  * item.  "len" can be 0 only when it's a sticky option.
2918  * We have 4 cases of combination of "sticky" and "cmsg":
2919  * "sticky=0, cmsg=0": impossible
2920  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2921  * "sticky=1, cmsg=0": RFC3542 socket option
2922  * "sticky=1, cmsg=1": RFC2292 socket option
2923  */
2924 static int
2925 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2926     int priv, int sticky, int cmsg, int uproto)
2927 {
2928 	int minmtupolicy;
2929 
2930 	if (!sticky && !cmsg) {
2931 #ifdef DIAGNOSTIC
2932 		printf("ip6_setpktopt: impossible case\n");
2933 #endif
2934 		return (EINVAL);
2935 	}
2936 
2937 	/*
2938 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2939 	 * not be specified in the context of RFC3542.  Conversely,
2940 	 * RFC3542 types should not be specified in the context of RFC2292.
2941 	 */
2942 	if (!cmsg) {
2943 		switch (optname) {
2944 		case IPV6_2292PKTINFO:
2945 		case IPV6_2292HOPLIMIT:
2946 		case IPV6_2292NEXTHOP:
2947 		case IPV6_2292HOPOPTS:
2948 		case IPV6_2292DSTOPTS:
2949 		case IPV6_2292RTHDR:
2950 		case IPV6_2292PKTOPTIONS:
2951 			return (ENOPROTOOPT);
2952 		}
2953 	}
2954 	if (sticky && cmsg) {
2955 		switch (optname) {
2956 		case IPV6_PKTINFO:
2957 		case IPV6_HOPLIMIT:
2958 		case IPV6_NEXTHOP:
2959 		case IPV6_HOPOPTS:
2960 		case IPV6_DSTOPTS:
2961 		case IPV6_RTHDRDSTOPTS:
2962 		case IPV6_RTHDR:
2963 		case IPV6_USE_MIN_MTU:
2964 		case IPV6_DONTFRAG:
2965 		case IPV6_OTCLASS:
2966 		case IPV6_TCLASS:
2967 			return (ENOPROTOOPT);
2968 		}
2969 	}
2970 
2971 	switch (optname) {
2972 #ifdef RFC2292
2973 	case IPV6_2292PKTINFO:
2974 #endif
2975 	case IPV6_PKTINFO:
2976 	{
2977 		struct ifnet *ifp = NULL;
2978 		struct in6_pktinfo *pktinfo;
2979 
2980 		if (len != sizeof(struct in6_pktinfo))
2981 			return (EINVAL);
2982 
2983 		pktinfo = (struct in6_pktinfo *)buf;
2984 
2985 		/*
2986 		 * An application can clear any sticky IPV6_PKTINFO option by
2987 		 * doing a "regular" setsockopt with ipi6_addr being
2988 		 * in6addr_any and ipi6_ifindex being zero.
2989 		 * [RFC 3542, Section 6]
2990 		 */
2991 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2992 		    pktinfo->ipi6_ifindex == 0 &&
2993 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2994 			ip6_clearpktopts(opt, optname);
2995 			break;
2996 		}
2997 
2998 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2999 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3000 			return (EINVAL);
3001 		}
3002 
3003 		/* validate the interface index if specified. */
3004 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
3005 			 return (ENXIO);
3006 		}
3007 		if (pktinfo->ipi6_ifindex) {
3008 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
3009 			if (ifp == NULL)
3010 				return (ENXIO);
3011 		}
3012 
3013 		/*
3014 		 * We store the address anyway, and let in6_selectsrc()
3015 		 * validate the specified address.  This is because ipi6_addr
3016 		 * may not have enough information about its scope zone, and
3017 		 * we may need additional information (such as outgoing
3018 		 * interface or the scope zone of a destination address) to
3019 		 * disambiguate the scope.
3020 		 * XXX: the delay of the validation may confuse the
3021 		 * application when it is used as a sticky option.
3022 		 */
3023 		if (opt->ip6po_pktinfo == NULL) {
3024 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3025 			    M_IP6OPT, M_NOWAIT);
3026 			if (opt->ip6po_pktinfo == NULL)
3027 				return (ENOBUFS);
3028 		}
3029 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
3030 		break;
3031 	}
3032 
3033 #ifdef RFC2292
3034 	case IPV6_2292HOPLIMIT:
3035 #endif
3036 	case IPV6_HOPLIMIT:
3037 	{
3038 		int *hlimp;
3039 
3040 		/*
3041 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3042 		 * to simplify the ordering among hoplimit options.
3043 		 */
3044 		if (optname == IPV6_HOPLIMIT && sticky)
3045 			return (ENOPROTOOPT);
3046 
3047 		if (len != sizeof(int))
3048 			return (EINVAL);
3049 		hlimp = (int *)buf;
3050 		if (*hlimp < -1 || *hlimp > 255)
3051 			return (EINVAL);
3052 
3053 		opt->ip6po_hlim = *hlimp;
3054 		break;
3055 	}
3056 
3057 	case IPV6_OTCLASS:
3058 		if (len != sizeof(u_int8_t))
3059 			return (EINVAL);
3060 
3061 		opt->ip6po_tclass = *(u_int8_t *)buf;
3062 		break;
3063 
3064 	case IPV6_TCLASS:
3065 	{
3066 		int tclass;
3067 
3068 		if (len != sizeof(int))
3069 			return (EINVAL);
3070 		tclass = *(int *)buf;
3071 		if (tclass < -1 || tclass > 255)
3072 			return (EINVAL);
3073 
3074 		opt->ip6po_tclass = tclass;
3075 		break;
3076 	}
3077 
3078 #ifdef RFC2292
3079 	case IPV6_2292NEXTHOP:
3080 #endif
3081 	case IPV6_NEXTHOP:
3082 		if (!priv)
3083 			return (EPERM);
3084 
3085 		if (len == 0) {	/* just remove the option */
3086 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3087 			break;
3088 		}
3089 
3090 		/* check if cmsg_len is large enough for sa_len */
3091 		if (len < sizeof(struct sockaddr) || len < *buf)
3092 			return (EINVAL);
3093 
3094 		switch (((struct sockaddr *)buf)->sa_family) {
3095 		case AF_INET6:
3096 		{
3097 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3098 			int error;
3099 
3100 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3101 				return (EINVAL);
3102 
3103 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3104 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3105 				return (EINVAL);
3106 			}
3107 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3108 			    != 0) {
3109 				return (error);
3110 			}
3111 			break;
3112 		}
3113 		case AF_LINK:	/* eventually be supported? */
3114 		default:
3115 			return (EAFNOSUPPORT);
3116 		}
3117 
3118 		/* turn off the previous option, then set the new option. */
3119 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3120 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3121 		if (opt->ip6po_nexthop == NULL)
3122 			return (ENOBUFS);
3123 		memcpy(opt->ip6po_nexthop, buf, *buf);
3124 		break;
3125 
3126 #ifdef RFC2292
3127 	case IPV6_2292HOPOPTS:
3128 #endif
3129 	case IPV6_HOPOPTS:
3130 	{
3131 		struct ip6_hbh *hbh;
3132 		int hbhlen;
3133 
3134 		/*
3135 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3136 		 * options, since per-option restriction has too much
3137 		 * overhead.
3138 		 */
3139 		if (!priv)
3140 			return (EPERM);
3141 
3142 		if (len == 0) {
3143 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3144 			break;	/* just remove the option */
3145 		}
3146 
3147 		/* message length validation */
3148 		if (len < sizeof(struct ip6_hbh))
3149 			return (EINVAL);
3150 		hbh = (struct ip6_hbh *)buf;
3151 		hbhlen = (hbh->ip6h_len + 1) << 3;
3152 		if (len != hbhlen)
3153 			return (EINVAL);
3154 
3155 		/* turn off the previous option, then set the new option. */
3156 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3157 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3158 		if (opt->ip6po_hbh == NULL)
3159 			return (ENOBUFS);
3160 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3161 
3162 		break;
3163 	}
3164 
3165 #ifdef RFC2292
3166 	case IPV6_2292DSTOPTS:
3167 #endif
3168 	case IPV6_DSTOPTS:
3169 	case IPV6_RTHDRDSTOPTS:
3170 	{
3171 		struct ip6_dest *dest, **newdest = NULL;
3172 		int destlen;
3173 
3174 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
3175 			return (EPERM);
3176 
3177 		if (len == 0) {
3178 			ip6_clearpktopts(opt, optname);
3179 			break;	/* just remove the option */
3180 		}
3181 
3182 		/* message length validation */
3183 		if (len < sizeof(struct ip6_dest))
3184 			return (EINVAL);
3185 		dest = (struct ip6_dest *)buf;
3186 		destlen = (dest->ip6d_len + 1) << 3;
3187 		if (len != destlen)
3188 			return (EINVAL);
3189 		/*
3190 		 * Determine the position that the destination options header
3191 		 * should be inserted; before or after the routing header.
3192 		 */
3193 		switch (optname) {
3194 		case IPV6_2292DSTOPTS:
3195 			/*
3196 			 * The old advanced API is ambiguous on this point.
3197 			 * Our approach is to determine the position based
3198 			 * according to the existence of a routing header.
3199 			 * Note, however, that this depends on the order of the
3200 			 * extension headers in the ancillary data; the 1st
3201 			 * part of the destination options header must appear
3202 			 * before the routing header in the ancillary data,
3203 			 * too.
3204 			 * RFC3542 solved the ambiguity by introducing
3205 			 * separate ancillary data or option types.
3206 			 */
3207 			if (opt->ip6po_rthdr == NULL)
3208 				newdest = &opt->ip6po_dest1;
3209 			else
3210 				newdest = &opt->ip6po_dest2;
3211 			break;
3212 		case IPV6_RTHDRDSTOPTS:
3213 			newdest = &opt->ip6po_dest1;
3214 			break;
3215 		case IPV6_DSTOPTS:
3216 			newdest = &opt->ip6po_dest2;
3217 			break;
3218 		}
3219 
3220 		/* turn off the previous option, then set the new option. */
3221 		ip6_clearpktopts(opt, optname);
3222 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3223 		if (*newdest == NULL)
3224 			return (ENOBUFS);
3225 		memcpy(*newdest, dest, destlen);
3226 
3227 		break;
3228 	}
3229 
3230 #ifdef RFC2292
3231 	case IPV6_2292RTHDR:
3232 #endif
3233 	case IPV6_RTHDR:
3234 	{
3235 		struct ip6_rthdr *rth;
3236 		int rthlen;
3237 
3238 		if (len == 0) {
3239 			ip6_clearpktopts(opt, IPV6_RTHDR);
3240 			break;	/* just remove the option */
3241 		}
3242 
3243 		/* message length validation */
3244 		if (len < sizeof(struct ip6_rthdr))
3245 			return (EINVAL);
3246 		rth = (struct ip6_rthdr *)buf;
3247 		rthlen = (rth->ip6r_len + 1) << 3;
3248 		if (len != rthlen)
3249 			return (EINVAL);
3250 		switch (rth->ip6r_type) {
3251 		case IPV6_RTHDR_TYPE_0:
3252 			if (rth->ip6r_len == 0)	/* must contain one addr */
3253 				return (EINVAL);
3254 			if (rth->ip6r_len % 2) /* length must be even */
3255 				return (EINVAL);
3256 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3257 				return (EINVAL);
3258 			break;
3259 		default:
3260 			return (EINVAL);	/* not supported */
3261 		}
3262 		/* turn off the previous option */
3263 		ip6_clearpktopts(opt, IPV6_RTHDR);
3264 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3265 		if (opt->ip6po_rthdr == NULL)
3266 			return (ENOBUFS);
3267 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3268 		break;
3269 	}
3270 
3271 	case IPV6_USE_MIN_MTU:
3272 		if (len != sizeof(int))
3273 			return (EINVAL);
3274 		minmtupolicy = *(int *)buf;
3275 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3276 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3277 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3278 			return (EINVAL);
3279 		}
3280 		opt->ip6po_minmtu = minmtupolicy;
3281 		break;
3282 
3283 	case IPV6_DONTFRAG:
3284 		if (len != sizeof(int))
3285 			return (EINVAL);
3286 
3287 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3288 			/*
3289 			 * we ignore this option for TCP sockets.
3290 			 * (RFC3542 leaves this case unspecified.)
3291 			 */
3292 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3293 		} else
3294 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3295 		break;
3296 
3297 	default:
3298 		return (ENOPROTOOPT);
3299 	} /* end of switch */
3300 
3301 	return (0);
3302 }
3303 
3304 /*
3305  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3306  * packet to the input queue of a specified interface.  Note that this
3307  * calls the output routine of the loopback "driver", but with an interface
3308  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3309  */
3310 void
3311 ip6_mloopback(ifp, m, dst)
3312 	struct ifnet *ifp;
3313 	struct mbuf *m;
3314 	struct sockaddr_in6 *dst;
3315 {
3316 	struct mbuf *copym;
3317 	struct ip6_hdr *ip6;
3318 
3319 	copym = m_copy(m, 0, M_COPYALL);
3320 	if (copym == NULL)
3321 		return;
3322 
3323 	/*
3324 	 * Make sure to deep-copy IPv6 header portion in case the data
3325 	 * is in an mbuf cluster, so that we can safely override the IPv6
3326 	 * header portion later.
3327 	 */
3328 	if ((copym->m_flags & M_EXT) != 0 ||
3329 	    copym->m_len < sizeof(struct ip6_hdr)) {
3330 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3331 		if (copym == NULL)
3332 			return;
3333 	}
3334 
3335 #ifdef DIAGNOSTIC
3336 	if (copym->m_len < sizeof(*ip6)) {
3337 		m_freem(copym);
3338 		return;
3339 	}
3340 #endif
3341 
3342 	ip6 = mtod(copym, struct ip6_hdr *);
3343 	/*
3344 	 * clear embedded scope identifiers if necessary.
3345 	 * in6_clearscope will touch the addresses only when necessary.
3346 	 */
3347 	in6_clearscope(&ip6->ip6_src);
3348 	in6_clearscope(&ip6->ip6_dst);
3349 
3350 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
3351 }
3352 
3353 /*
3354  * Chop IPv6 header off from the payload.
3355  */
3356 static int
3357 ip6_splithdr(m, exthdrs)
3358 	struct mbuf *m;
3359 	struct ip6_exthdrs *exthdrs;
3360 {
3361 	struct mbuf *mh;
3362 	struct ip6_hdr *ip6;
3363 
3364 	ip6 = mtod(m, struct ip6_hdr *);
3365 	if (m->m_len > sizeof(*ip6)) {
3366 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3367 		if (mh == 0) {
3368 			m_freem(m);
3369 			return ENOBUFS;
3370 		}
3371 		M_MOVE_PKTHDR(mh, m);
3372 		MH_ALIGN(mh, sizeof(*ip6));
3373 		m->m_len -= sizeof(*ip6);
3374 		m->m_data += sizeof(*ip6);
3375 		mh->m_next = m;
3376 		m = mh;
3377 		m->m_len = sizeof(*ip6);
3378 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3379 	}
3380 	exthdrs->ip6e_ip6 = m;
3381 	return 0;
3382 }
3383 
3384 /*
3385  * Compute IPv6 extension header length.
3386  */
3387 int
3388 ip6_optlen(in6p)
3389 	struct in6pcb *in6p;
3390 {
3391 	int len;
3392 
3393 	if (!in6p->in6p_outputopts)
3394 		return 0;
3395 
3396 	len = 0;
3397 #define elen(x) \
3398     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3399 
3400 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3401 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3402 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3403 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3404 	return len;
3405 #undef elen
3406 }
3407