xref: /freebsd/sys/netinet6/ip6_output.c (revision 0c76184ced739bcbe6513621b57a182f51f844c0)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip_var.h>
99 #include <netinet6/in6_var.h>
100 #include <netinet/ip6.h>
101 #include <netinet/icmp6.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/tcp_var.h>
105 #include <netinet6/nd6.h>
106 #include <netinet/in_rss.h>
107 
108 #ifdef IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netipsec/ipsec6.h>
111 #include <netipsec/key.h>
112 #include <netinet6/ip6_ipsec.h>
113 #endif /* IPSEC */
114 #ifdef SCTP
115 #include <netinet/sctp.h>
116 #include <netinet/sctp_crc32.h>
117 #endif
118 
119 #include <netinet6/ip6protosw.h>
120 #include <netinet6/scope6_var.h>
121 
122 #ifdef FLOWTABLE
123 #include <net/flowtable.h>
124 #endif
125 
126 extern int in6_mcast_loop;
127 
128 struct ip6_exthdrs {
129 	struct mbuf *ip6e_ip6;
130 	struct mbuf *ip6e_hbh;
131 	struct mbuf *ip6e_dest1;
132 	struct mbuf *ip6e_rthdr;
133 	struct mbuf *ip6e_dest2;
134 };
135 
136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
137 			   struct ucred *, int);
138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
139 	struct socket *, struct sockopt *);
140 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
142 	struct ucred *, int, int, int);
143 
144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
146 	struct ip6_frag **);
147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
149 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
150 	struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
151 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
152 
153 
154 /*
155  * Make an extension header from option data.  hp is the source, and
156  * mp is the destination.
157  */
158 #define MAKE_EXTHDR(hp, mp)						\
159     do {								\
160 	if (hp) {							\
161 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
162 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
163 		    ((eh)->ip6e_len + 1) << 3);				\
164 		if (error)						\
165 			goto freehdrs;					\
166 	}								\
167     } while (/*CONSTCOND*/ 0)
168 
169 /*
170  * Form a chain of extension headers.
171  * m is the extension header mbuf
172  * mp is the previous mbuf in the chain
173  * p is the next header
174  * i is the type of option.
175  */
176 #define MAKE_CHAIN(m, mp, p, i)\
177     do {\
178 	if (m) {\
179 		if (!hdrsplit) \
180 			panic("assumption failed: hdr not split"); \
181 		*mtod((m), u_char *) = *(p);\
182 		*(p) = (i);\
183 		p = mtod((m), u_char *);\
184 		(m)->m_next = (mp)->m_next;\
185 		(mp)->m_next = (m);\
186 		(mp) = (m);\
187 	}\
188     } while (/*CONSTCOND*/ 0)
189 
190 void
191 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
192 {
193 	u_short csum;
194 
195 	csum = in_cksum_skip(m, offset + plen, offset);
196 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
197 		csum = 0xffff;
198 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
199 
200 	if (offset + sizeof(u_short) > m->m_len) {
201 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
202 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
203 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
204 		/*
205 		 * XXX this should not happen, but if it does, the correct
206 		 * behavior may be to insert the checksum in the appropriate
207 		 * next mbuf in the chain.
208 		 */
209 		return;
210 	}
211 	*(u_short *)(m->m_data + offset) = csum;
212 }
213 
214 /*
215  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
216  * header (with pri, len, nxt, hlim, src, dst).
217  * This function may modify ver and hlim only.
218  * The mbuf chain containing the packet will be freed.
219  * The mbuf opt, if present, will not be freed.
220  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
221  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
222  * then result of route lookup is stored in ro->ro_rt.
223  *
224  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
225  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
226  * which is rt_mtu.
227  *
228  * ifpp - XXX: just for statistics
229  */
230 /*
231  * XXX TODO: no flowid is assigned for outbound flows?
232  */
233 int
234 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
235     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
236     struct ifnet **ifpp, struct inpcb *inp)
237 {
238 	struct ip6_hdr *ip6, *mhip6;
239 	struct ifnet *ifp, *origifp;
240 	struct mbuf *m = m0;
241 	struct mbuf *mprev = NULL;
242 	int hlen, tlen, len, off;
243 	struct route_in6 ip6route;
244 	struct rtentry *rt = NULL;
245 	struct sockaddr_in6 *dst, src_sa, dst_sa;
246 	struct in6_addr odst;
247 	int error = 0;
248 	struct in6_ifaddr *ia = NULL;
249 	u_long mtu;
250 	int alwaysfrag, dontfrag;
251 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
252 	struct ip6_exthdrs exthdrs;
253 	struct in6_addr finaldst, src0, dst0;
254 	u_int32_t zone;
255 	struct route_in6 *ro_pmtu = NULL;
256 	int hdrsplit = 0;
257 	int sw_csum, tso;
258 	int needfiblookup;
259 	uint32_t fibnum;
260 	struct m_tag *fwd_tag = NULL;
261 
262 	ip6 = mtod(m, struct ip6_hdr *);
263 	if (ip6 == NULL) {
264 		printf ("ip6 is NULL");
265 		goto bad;
266 	}
267 
268 	if (inp != NULL) {
269 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
270 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
271 			/* unconditionally set flowid */
272 			m->m_pkthdr.flowid = inp->inp_flowid;
273 			M_HASHTYPE_SET(m, inp->inp_flowtype);
274 		}
275 	}
276 
277 	finaldst = ip6->ip6_dst;
278 	bzero(&exthdrs, sizeof(exthdrs));
279 	if (opt) {
280 		/* Hop-by-Hop options header */
281 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
282 		/* Destination options header(1st part) */
283 		if (opt->ip6po_rthdr) {
284 			/*
285 			 * Destination options header(1st part)
286 			 * This only makes sense with a routing header.
287 			 * See Section 9.2 of RFC 3542.
288 			 * Disabling this part just for MIP6 convenience is
289 			 * a bad idea.  We need to think carefully about a
290 			 * way to make the advanced API coexist with MIP6
291 			 * options, which might automatically be inserted in
292 			 * the kernel.
293 			 */
294 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
295 		}
296 		/* Routing header */
297 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
298 		/* Destination options header(2nd part) */
299 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
300 	}
301 
302 #ifdef IPSEC
303 	/*
304 	 * IPSec checking which handles several cases.
305 	 * FAST IPSEC: We re-injected the packet.
306 	 * XXX: need scope argument.
307 	 */
308 	switch(ip6_ipsec_output(&m, inp, &error))
309 	{
310 	case 1:                 /* Bad packet */
311 		goto freehdrs;
312 	case -1:                /* IPSec done */
313 		goto done;
314 	case 0:                 /* No IPSec */
315 	default:
316 		break;
317 	}
318 #endif /* IPSEC */
319 
320 	/*
321 	 * Calculate the total length of the extension header chain.
322 	 * Keep the length of the unfragmentable part for fragmentation.
323 	 */
324 	optlen = 0;
325 	if (exthdrs.ip6e_hbh)
326 		optlen += exthdrs.ip6e_hbh->m_len;
327 	if (exthdrs.ip6e_dest1)
328 		optlen += exthdrs.ip6e_dest1->m_len;
329 	if (exthdrs.ip6e_rthdr)
330 		optlen += exthdrs.ip6e_rthdr->m_len;
331 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
332 
333 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
334 	if (exthdrs.ip6e_dest2)
335 		optlen += exthdrs.ip6e_dest2->m_len;
336 
337 	/*
338 	 * If there is at least one extension header,
339 	 * separate IP6 header from the payload.
340 	 */
341 	if (optlen && !hdrsplit) {
342 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
343 			m = NULL;
344 			goto freehdrs;
345 		}
346 		m = exthdrs.ip6e_ip6;
347 		hdrsplit++;
348 	}
349 
350 	/* adjust pointer */
351 	ip6 = mtod(m, struct ip6_hdr *);
352 
353 	/* adjust mbuf packet header length */
354 	m->m_pkthdr.len += optlen;
355 	plen = m->m_pkthdr.len - sizeof(*ip6);
356 
357 	/* If this is a jumbo payload, insert a jumbo payload option. */
358 	if (plen > IPV6_MAXPACKET) {
359 		if (!hdrsplit) {
360 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
361 				m = NULL;
362 				goto freehdrs;
363 			}
364 			m = exthdrs.ip6e_ip6;
365 			hdrsplit++;
366 		}
367 		/* adjust pointer */
368 		ip6 = mtod(m, struct ip6_hdr *);
369 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
370 			goto freehdrs;
371 		ip6->ip6_plen = 0;
372 	} else
373 		ip6->ip6_plen = htons(plen);
374 
375 	/*
376 	 * Concatenate headers and fill in next header fields.
377 	 * Here we have, on "m"
378 	 *	IPv6 payload
379 	 * and we insert headers accordingly.  Finally, we should be getting:
380 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
381 	 *
382 	 * during the header composing process, "m" points to IPv6 header.
383 	 * "mprev" points to an extension header prior to esp.
384 	 */
385 	u_char *nexthdrp = &ip6->ip6_nxt;
386 	mprev = m;
387 
388 	/*
389 	 * we treat dest2 specially.  this makes IPsec processing
390 	 * much easier.  the goal here is to make mprev point the
391 	 * mbuf prior to dest2.
392 	 *
393 	 * result: IPv6 dest2 payload
394 	 * m and mprev will point to IPv6 header.
395 	 */
396 	if (exthdrs.ip6e_dest2) {
397 		if (!hdrsplit)
398 			panic("assumption failed: hdr not split");
399 		exthdrs.ip6e_dest2->m_next = m->m_next;
400 		m->m_next = exthdrs.ip6e_dest2;
401 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
402 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
403 	}
404 
405 	/*
406 	 * result: IPv6 hbh dest1 rthdr dest2 payload
407 	 * m will point to IPv6 header.  mprev will point to the
408 	 * extension header prior to dest2 (rthdr in the above case).
409 	 */
410 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
411 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
412 		   IPPROTO_DSTOPTS);
413 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
414 		   IPPROTO_ROUTING);
415 
416 	/*
417 	 * If there is a routing header, discard the packet.
418 	 */
419 	if (exthdrs.ip6e_rthdr) {
420 		 error = EINVAL;
421 		 goto bad;
422 	}
423 
424 	/* Source address validation */
425 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
426 	    (flags & IPV6_UNSPECSRC) == 0) {
427 		error = EOPNOTSUPP;
428 		IP6STAT_INC(ip6s_badscope);
429 		goto bad;
430 	}
431 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
432 		error = EOPNOTSUPP;
433 		IP6STAT_INC(ip6s_badscope);
434 		goto bad;
435 	}
436 
437 	IP6STAT_INC(ip6s_localout);
438 
439 	/*
440 	 * Route packet.
441 	 */
442 	if (ro == 0) {
443 		ro = &ip6route;
444 		bzero((caddr_t)ro, sizeof(*ro));
445 	}
446 	ro_pmtu = ro;
447 	if (opt && opt->ip6po_rthdr)
448 		ro = &opt->ip6po_route;
449 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
450 #ifdef FLOWTABLE
451 	if (ro->ro_rt == NULL)
452 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
453 #endif
454 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
455 again:
456 	/*
457 	 * if specified, try to fill in the traffic class field.
458 	 * do not override if a non-zero value is already set.
459 	 * we check the diffserv field and the ecn field separately.
460 	 */
461 	if (opt && opt->ip6po_tclass >= 0) {
462 		int mask = 0;
463 
464 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
465 			mask |= 0xfc;
466 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
467 			mask |= 0x03;
468 		if (mask != 0)
469 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
470 	}
471 
472 	/* fill in or override the hop limit field, if necessary. */
473 	if (opt && opt->ip6po_hlim != -1)
474 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
475 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
476 		if (im6o != NULL)
477 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
478 		else
479 			ip6->ip6_hlim = V_ip6_defmcasthlim;
480 	}
481 
482 	/* adjust pointer */
483 	ip6 = mtod(m, struct ip6_hdr *);
484 
485 	if (ro->ro_rt && fwd_tag == NULL) {
486 		rt = ro->ro_rt;
487 		ifp = ro->ro_rt->rt_ifp;
488 	} else {
489 		if (fwd_tag == NULL) {
490 			bzero(&dst_sa, sizeof(dst_sa));
491 			dst_sa.sin6_family = AF_INET6;
492 			dst_sa.sin6_len = sizeof(dst_sa);
493 			dst_sa.sin6_addr = ip6->ip6_dst;
494 		}
495 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
496 		    &rt, fibnum);
497 		if (error != 0) {
498 			if (ifp != NULL)
499 				in6_ifstat_inc(ifp, ifs6_out_discard);
500 			goto bad;
501 		}
502 	}
503 	if (rt == NULL) {
504 		/*
505 		 * If in6_selectroute() does not return a route entry,
506 		 * dst may not have been updated.
507 		 */
508 		*dst = dst_sa;	/* XXX */
509 	}
510 
511 	/*
512 	 * then rt (for unicast) and ifp must be non-NULL valid values.
513 	 */
514 	if ((flags & IPV6_FORWARDING) == 0) {
515 		/* XXX: the FORWARDING flag can be set for mrouting. */
516 		in6_ifstat_inc(ifp, ifs6_out_request);
517 	}
518 	if (rt != NULL) {
519 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
520 		counter_u64_add(rt->rt_pksent, 1);
521 	}
522 
523 
524 	/*
525 	 * The outgoing interface must be in the zone of source and
526 	 * destination addresses.
527 	 */
528 	origifp = ifp;
529 
530 	src0 = ip6->ip6_src;
531 	if (in6_setscope(&src0, origifp, &zone))
532 		goto badscope;
533 	bzero(&src_sa, sizeof(src_sa));
534 	src_sa.sin6_family = AF_INET6;
535 	src_sa.sin6_len = sizeof(src_sa);
536 	src_sa.sin6_addr = ip6->ip6_src;
537 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
538 		goto badscope;
539 
540 	dst0 = ip6->ip6_dst;
541 	if (in6_setscope(&dst0, origifp, &zone))
542 		goto badscope;
543 	/* re-initialize to be sure */
544 	bzero(&dst_sa, sizeof(dst_sa));
545 	dst_sa.sin6_family = AF_INET6;
546 	dst_sa.sin6_len = sizeof(dst_sa);
547 	dst_sa.sin6_addr = ip6->ip6_dst;
548 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
549 		goto badscope;
550 	}
551 
552 	/* We should use ia_ifp to support the case of
553 	 * sending packets to an address of our own.
554 	 */
555 	if (ia != NULL && ia->ia_ifp)
556 		ifp = ia->ia_ifp;
557 
558 	/* scope check is done. */
559 	goto routefound;
560 
561   badscope:
562 	IP6STAT_INC(ip6s_badscope);
563 	in6_ifstat_inc(origifp, ifs6_out_discard);
564 	if (error == 0)
565 		error = EHOSTUNREACH; /* XXX */
566 	goto bad;
567 
568   routefound:
569 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
570 		if (opt && opt->ip6po_nextroute.ro_rt) {
571 			/*
572 			 * The nexthop is explicitly specified by the
573 			 * application.  We assume the next hop is an IPv6
574 			 * address.
575 			 */
576 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
577 		}
578 		else if ((rt->rt_flags & RTF_GATEWAY))
579 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
580 	}
581 
582 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
583 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
584 	} else {
585 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
586 		in6_ifstat_inc(ifp, ifs6_out_mcast);
587 		/*
588 		 * Confirm that the outgoing interface supports multicast.
589 		 */
590 		if (!(ifp->if_flags & IFF_MULTICAST)) {
591 			IP6STAT_INC(ip6s_noroute);
592 			in6_ifstat_inc(ifp, ifs6_out_discard);
593 			error = ENETUNREACH;
594 			goto bad;
595 		}
596 		if ((im6o == NULL && in6_mcast_loop) ||
597 		    (im6o && im6o->im6o_multicast_loop)) {
598 			/*
599 			 * Loop back multicast datagram if not expressly
600 			 * forbidden to do so, even if we have not joined
601 			 * the address; protocols will filter it later,
602 			 * thus deferring a hash lookup and lock acquisition
603 			 * at the expense of an m_copym().
604 			 */
605 			ip6_mloopback(ifp, m, dst);
606 		} else {
607 			/*
608 			 * If we are acting as a multicast router, perform
609 			 * multicast forwarding as if the packet had just
610 			 * arrived on the interface to which we are about
611 			 * to send.  The multicast forwarding function
612 			 * recursively calls this function, using the
613 			 * IPV6_FORWARDING flag to prevent infinite recursion.
614 			 *
615 			 * Multicasts that are looped back by ip6_mloopback(),
616 			 * above, will be forwarded by the ip6_input() routine,
617 			 * if necessary.
618 			 */
619 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
620 				/*
621 				 * XXX: ip6_mforward expects that rcvif is NULL
622 				 * when it is called from the originating path.
623 				 * However, it may not always be the case.
624 				 */
625 				m->m_pkthdr.rcvif = NULL;
626 				if (ip6_mforward(ip6, ifp, m) != 0) {
627 					m_freem(m);
628 					goto done;
629 				}
630 			}
631 		}
632 		/*
633 		 * Multicasts with a hoplimit of zero may be looped back,
634 		 * above, but must not be transmitted on a network.
635 		 * Also, multicasts addressed to the loopback interface
636 		 * are not sent -- the above call to ip6_mloopback() will
637 		 * loop back a copy if this host actually belongs to the
638 		 * destination group on the loopback interface.
639 		 */
640 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
641 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
642 			m_freem(m);
643 			goto done;
644 		}
645 	}
646 
647 	/*
648 	 * Fill the outgoing inteface to tell the upper layer
649 	 * to increment per-interface statistics.
650 	 */
651 	if (ifpp)
652 		*ifpp = ifp;
653 
654 	/* Determine path MTU. */
655 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
656 	    &alwaysfrag, fibnum)) != 0)
657 		goto bad;
658 
659 	/*
660 	 * The caller of this function may specify to use the minimum MTU
661 	 * in some cases.
662 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
663 	 * setting.  The logic is a bit complicated; by default, unicast
664 	 * packets will follow path MTU while multicast packets will be sent at
665 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
666 	 * including unicast ones will be sent at the minimum MTU.  Multicast
667 	 * packets will always be sent at the minimum MTU unless
668 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
669 	 * See RFC 3542 for more details.
670 	 */
671 	if (mtu > IPV6_MMTU) {
672 		if ((flags & IPV6_MINMTU))
673 			mtu = IPV6_MMTU;
674 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
675 			mtu = IPV6_MMTU;
676 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
677 			 (opt == NULL ||
678 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
679 			mtu = IPV6_MMTU;
680 		}
681 	}
682 
683 	/*
684 	 * clear embedded scope identifiers if necessary.
685 	 * in6_clearscope will touch the addresses only when necessary.
686 	 */
687 	in6_clearscope(&ip6->ip6_src);
688 	in6_clearscope(&ip6->ip6_dst);
689 
690 	/*
691 	 * If the outgoing packet contains a hop-by-hop options header,
692 	 * it must be examined and processed even by the source node.
693 	 * (RFC 2460, section 4.)
694 	 */
695 	if (exthdrs.ip6e_hbh) {
696 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
697 		u_int32_t dummy; /* XXX unused */
698 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
699 
700 #ifdef DIAGNOSTIC
701 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
702 			panic("ip6e_hbh is not contiguous");
703 #endif
704 		/*
705 		 *  XXX: if we have to send an ICMPv6 error to the sender,
706 		 *       we need the M_LOOP flag since icmp6_error() expects
707 		 *       the IPv6 and the hop-by-hop options header are
708 		 *       contiguous unless the flag is set.
709 		 */
710 		m->m_flags |= M_LOOP;
711 		m->m_pkthdr.rcvif = ifp;
712 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
713 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
714 		    &dummy, &plen) < 0) {
715 			/* m was already freed at this point */
716 			error = EINVAL;/* better error? */
717 			goto done;
718 		}
719 		m->m_flags &= ~M_LOOP; /* XXX */
720 		m->m_pkthdr.rcvif = NULL;
721 	}
722 
723 	/* Jump over all PFIL processing if hooks are not active. */
724 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
725 		goto passout;
726 
727 	odst = ip6->ip6_dst;
728 	/* Run through list of hooks for output packets. */
729 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
730 	if (error != 0 || m == NULL)
731 		goto done;
732 	ip6 = mtod(m, struct ip6_hdr *);
733 
734 	needfiblookup = 0;
735 	/* See if destination IP address was changed by packet filter. */
736 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
737 		m->m_flags |= M_SKIP_FIREWALL;
738 		/* If destination is now ourself drop to ip6_input(). */
739 		if (in6_localip(&ip6->ip6_dst)) {
740 			m->m_flags |= M_FASTFWD_OURS;
741 			if (m->m_pkthdr.rcvif == NULL)
742 				m->m_pkthdr.rcvif = V_loif;
743 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
744 				m->m_pkthdr.csum_flags |=
745 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
746 				m->m_pkthdr.csum_data = 0xffff;
747 			}
748 #ifdef SCTP
749 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
750 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
751 #endif
752 			error = netisr_queue(NETISR_IPV6, m);
753 			goto done;
754 		} else
755 			needfiblookup = 1; /* Redo the routing table lookup. */
756 	}
757 	/* See if fib was changed by packet filter. */
758 	if (fibnum != M_GETFIB(m)) {
759 		m->m_flags |= M_SKIP_FIREWALL;
760 		fibnum = M_GETFIB(m);
761 		RO_RTFREE(ro);
762 		needfiblookup = 1;
763 	}
764 	if (needfiblookup)
765 		goto again;
766 
767 	/* See if local, if yes, send it to netisr. */
768 	if (m->m_flags & M_FASTFWD_OURS) {
769 		if (m->m_pkthdr.rcvif == NULL)
770 			m->m_pkthdr.rcvif = V_loif;
771 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
772 			m->m_pkthdr.csum_flags |=
773 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
774 			m->m_pkthdr.csum_data = 0xffff;
775 		}
776 #ifdef SCTP
777 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
778 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
779 #endif
780 		error = netisr_queue(NETISR_IPV6, m);
781 		goto done;
782 	}
783 	/* Or forward to some other address? */
784 	if ((m->m_flags & M_IP6_NEXTHOP) &&
785 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
786 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
787 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
788 		m->m_flags |= M_SKIP_FIREWALL;
789 		m->m_flags &= ~M_IP6_NEXTHOP;
790 		m_tag_delete(m, fwd_tag);
791 		goto again;
792 	}
793 
794 passout:
795 	/*
796 	 * Send the packet to the outgoing interface.
797 	 * If necessary, do IPv6 fragmentation before sending.
798 	 *
799 	 * the logic here is rather complex:
800 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
801 	 * 1-a:	send as is if tlen <= path mtu
802 	 * 1-b:	fragment if tlen > path mtu
803 	 *
804 	 * 2: if user asks us not to fragment (dontfrag == 1)
805 	 * 2-a:	send as is if tlen <= interface mtu
806 	 * 2-b:	error if tlen > interface mtu
807 	 *
808 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
809 	 *	always fragment
810 	 *
811 	 * 4: if dontfrag == 1 && alwaysfrag == 1
812 	 *	error, as we cannot handle this conflicting request
813 	 */
814 	sw_csum = m->m_pkthdr.csum_flags;
815 	if (!hdrsplit) {
816 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
817 		sw_csum &= ~ifp->if_hwassist;
818 	} else
819 		tso = 0;
820 	/*
821 	 * If we added extension headers, we will not do TSO and calculate the
822 	 * checksums ourselves for now.
823 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
824 	 * with ext. hdrs.
825 	 */
826 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
827 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
828 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
829 	}
830 #ifdef SCTP
831 	if (sw_csum & CSUM_SCTP_IPV6) {
832 		sw_csum &= ~CSUM_SCTP_IPV6;
833 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
834 	}
835 #endif
836 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
837 	tlen = m->m_pkthdr.len;
838 
839 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
840 		dontfrag = 1;
841 	else
842 		dontfrag = 0;
843 	if (dontfrag && alwaysfrag) {	/* case 4 */
844 		/* conflicting request - can't transmit */
845 		error = EMSGSIZE;
846 		goto bad;
847 	}
848 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
849 		/*
850 		 * Even if the DONTFRAG option is specified, we cannot send the
851 		 * packet when the data length is larger than the MTU of the
852 		 * outgoing interface.
853 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
854 		 * well as returning an error code (the latter is not described
855 		 * in the API spec.)
856 		 */
857 		u_int32_t mtu32;
858 		struct ip6ctlparam ip6cp;
859 
860 		mtu32 = (u_int32_t)mtu;
861 		bzero(&ip6cp, sizeof(ip6cp));
862 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
863 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
864 		    (void *)&ip6cp);
865 
866 		error = EMSGSIZE;
867 		goto bad;
868 	}
869 
870 	/*
871 	 * transmit packet without fragmentation
872 	 */
873 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
874 		struct in6_ifaddr *ia6;
875 
876 		ip6 = mtod(m, struct ip6_hdr *);
877 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
878 		if (ia6) {
879 			/* Record statistics for this interface address. */
880 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
881 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
882 			    m->m_pkthdr.len);
883 			ifa_free(&ia6->ia_ifa);
884 		}
885 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
886 		goto done;
887 	}
888 
889 	/*
890 	 * try to fragment the packet.  case 1-b and 3
891 	 */
892 	if (mtu < IPV6_MMTU) {
893 		/* path MTU cannot be less than IPV6_MMTU */
894 		error = EMSGSIZE;
895 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
896 		goto bad;
897 	} else if (ip6->ip6_plen == 0) {
898 		/* jumbo payload cannot be fragmented */
899 		error = EMSGSIZE;
900 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
901 		goto bad;
902 	} else {
903 		struct mbuf **mnext, *m_frgpart;
904 		struct ip6_frag *ip6f;
905 		u_int32_t id = htonl(ip6_randomid());
906 		u_char nextproto;
907 
908 		/*
909 		 * Too large for the destination or interface;
910 		 * fragment if possible.
911 		 * Must be able to put at least 8 bytes per fragment.
912 		 */
913 		hlen = unfragpartlen;
914 		if (mtu > IPV6_MAXPACKET)
915 			mtu = IPV6_MAXPACKET;
916 
917 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
918 		if (len < 8) {
919 			error = EMSGSIZE;
920 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
921 			goto bad;
922 		}
923 
924 		/*
925 		 * If the interface will not calculate checksums on
926 		 * fragmented packets, then do it here.
927 		 * XXX-BZ handle the hw offloading case.  Need flags.
928 		 */
929 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
930 			in6_delayed_cksum(m, plen, hlen);
931 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
932 		}
933 #ifdef SCTP
934 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
935 			sctp_delayed_cksum(m, hlen);
936 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
937 		}
938 #endif
939 		mnext = &m->m_nextpkt;
940 
941 		/*
942 		 * Change the next header field of the last header in the
943 		 * unfragmentable part.
944 		 */
945 		if (exthdrs.ip6e_rthdr) {
946 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
947 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
948 		} else if (exthdrs.ip6e_dest1) {
949 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
950 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
951 		} else if (exthdrs.ip6e_hbh) {
952 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
953 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
954 		} else {
955 			nextproto = ip6->ip6_nxt;
956 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
957 		}
958 
959 		/*
960 		 * Loop through length of segment after first fragment,
961 		 * make new header and copy data of each part and link onto
962 		 * chain.
963 		 */
964 		m0 = m;
965 		for (off = hlen; off < tlen; off += len) {
966 			m = m_gethdr(M_NOWAIT, MT_DATA);
967 			if (!m) {
968 				error = ENOBUFS;
969 				IP6STAT_INC(ip6s_odropped);
970 				goto sendorfree;
971 			}
972 			m->m_flags = m0->m_flags & M_COPYFLAGS;
973 			*mnext = m;
974 			mnext = &m->m_nextpkt;
975 			m->m_data += max_linkhdr;
976 			mhip6 = mtod(m, struct ip6_hdr *);
977 			*mhip6 = *ip6;
978 			m->m_len = sizeof(*mhip6);
979 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
980 			if (error) {
981 				IP6STAT_INC(ip6s_odropped);
982 				goto sendorfree;
983 			}
984 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
985 			if (off + len >= tlen)
986 				len = tlen - off;
987 			else
988 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
989 			mhip6->ip6_plen = htons((u_short)(len + hlen +
990 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
991 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
992 				error = ENOBUFS;
993 				IP6STAT_INC(ip6s_odropped);
994 				goto sendorfree;
995 			}
996 			m_cat(m, m_frgpart);
997 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
998 			m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
999 			m->m_pkthdr.rcvif = NULL;
1000 			ip6f->ip6f_reserved = 0;
1001 			ip6f->ip6f_ident = id;
1002 			ip6f->ip6f_nxt = nextproto;
1003 			IP6STAT_INC(ip6s_ofragments);
1004 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1005 		}
1006 
1007 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1008 	}
1009 
1010 	/*
1011 	 * Remove leading garbages.
1012 	 */
1013 sendorfree:
1014 	m = m0->m_nextpkt;
1015 	m0->m_nextpkt = 0;
1016 	m_freem(m0);
1017 	for (m0 = m; m; m = m0) {
1018 		m0 = m->m_nextpkt;
1019 		m->m_nextpkt = 0;
1020 		if (error == 0) {
1021 			/* Record statistics for this interface address. */
1022 			if (ia) {
1023 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1024 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1025 				    m->m_pkthdr.len);
1026 			}
1027 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1028 		} else
1029 			m_freem(m);
1030 	}
1031 
1032 	if (error == 0)
1033 		IP6STAT_INC(ip6s_fragmented);
1034 
1035 done:
1036 	if (ro == &ip6route)
1037 		RO_RTFREE(ro);
1038 	if (ro_pmtu == &ip6route)
1039 		RO_RTFREE(ro_pmtu);
1040 	return (error);
1041 
1042 freehdrs:
1043 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1044 	m_freem(exthdrs.ip6e_dest1);
1045 	m_freem(exthdrs.ip6e_rthdr);
1046 	m_freem(exthdrs.ip6e_dest2);
1047 	/* FALLTHROUGH */
1048 bad:
1049 	if (m)
1050 		m_freem(m);
1051 	goto done;
1052 }
1053 
1054 static int
1055 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1056 {
1057 	struct mbuf *m;
1058 
1059 	if (hlen > MCLBYTES)
1060 		return (ENOBUFS); /* XXX */
1061 
1062 	if (hlen > MLEN)
1063 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1064 	else
1065 		m = m_get(M_NOWAIT, MT_DATA);
1066 	if (m == NULL)
1067 		return (ENOBUFS);
1068 	m->m_len = hlen;
1069 	if (hdr)
1070 		bcopy(hdr, mtod(m, caddr_t), hlen);
1071 
1072 	*mp = m;
1073 	return (0);
1074 }
1075 
1076 /*
1077  * Insert jumbo payload option.
1078  */
1079 static int
1080 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1081 {
1082 	struct mbuf *mopt;
1083 	u_char *optbuf;
1084 	u_int32_t v;
1085 
1086 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1087 
1088 	/*
1089 	 * If there is no hop-by-hop options header, allocate new one.
1090 	 * If there is one but it doesn't have enough space to store the
1091 	 * jumbo payload option, allocate a cluster to store the whole options.
1092 	 * Otherwise, use it to store the options.
1093 	 */
1094 	if (exthdrs->ip6e_hbh == 0) {
1095 		mopt = m_get(M_NOWAIT, MT_DATA);
1096 		if (mopt == NULL)
1097 			return (ENOBUFS);
1098 		mopt->m_len = JUMBOOPTLEN;
1099 		optbuf = mtod(mopt, u_char *);
1100 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1101 		exthdrs->ip6e_hbh = mopt;
1102 	} else {
1103 		struct ip6_hbh *hbh;
1104 
1105 		mopt = exthdrs->ip6e_hbh;
1106 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1107 			/*
1108 			 * XXX assumption:
1109 			 * - exthdrs->ip6e_hbh is not referenced from places
1110 			 *   other than exthdrs.
1111 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1112 			 */
1113 			int oldoptlen = mopt->m_len;
1114 			struct mbuf *n;
1115 
1116 			/*
1117 			 * XXX: give up if the whole (new) hbh header does
1118 			 * not fit even in an mbuf cluster.
1119 			 */
1120 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1121 				return (ENOBUFS);
1122 
1123 			/*
1124 			 * As a consequence, we must always prepare a cluster
1125 			 * at this point.
1126 			 */
1127 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1128 			if (n == NULL)
1129 				return (ENOBUFS);
1130 			n->m_len = oldoptlen + JUMBOOPTLEN;
1131 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1132 			    oldoptlen);
1133 			optbuf = mtod(n, caddr_t) + oldoptlen;
1134 			m_freem(mopt);
1135 			mopt = exthdrs->ip6e_hbh = n;
1136 		} else {
1137 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1138 			mopt->m_len += JUMBOOPTLEN;
1139 		}
1140 		optbuf[0] = IP6OPT_PADN;
1141 		optbuf[1] = 1;
1142 
1143 		/*
1144 		 * Adjust the header length according to the pad and
1145 		 * the jumbo payload option.
1146 		 */
1147 		hbh = mtod(mopt, struct ip6_hbh *);
1148 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1149 	}
1150 
1151 	/* fill in the option. */
1152 	optbuf[2] = IP6OPT_JUMBO;
1153 	optbuf[3] = 4;
1154 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1155 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1156 
1157 	/* finally, adjust the packet header length */
1158 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1159 
1160 	return (0);
1161 #undef JUMBOOPTLEN
1162 }
1163 
1164 /*
1165  * Insert fragment header and copy unfragmentable header portions.
1166  */
1167 static int
1168 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1169     struct ip6_frag **frghdrp)
1170 {
1171 	struct mbuf *n, *mlast;
1172 
1173 	if (hlen > sizeof(struct ip6_hdr)) {
1174 		n = m_copym(m0, sizeof(struct ip6_hdr),
1175 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1176 		if (n == 0)
1177 			return (ENOBUFS);
1178 		m->m_next = n;
1179 	} else
1180 		n = m;
1181 
1182 	/* Search for the last mbuf of unfragmentable part. */
1183 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1184 		;
1185 
1186 	if (M_WRITABLE(mlast) &&
1187 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1188 		/* use the trailing space of the last mbuf for the fragment hdr */
1189 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1190 		    mlast->m_len);
1191 		mlast->m_len += sizeof(struct ip6_frag);
1192 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1193 	} else {
1194 		/* allocate a new mbuf for the fragment header */
1195 		struct mbuf *mfrg;
1196 
1197 		mfrg = m_get(M_NOWAIT, MT_DATA);
1198 		if (mfrg == NULL)
1199 			return (ENOBUFS);
1200 		mfrg->m_len = sizeof(struct ip6_frag);
1201 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1202 		mlast->m_next = mfrg;
1203 	}
1204 
1205 	return (0);
1206 }
1207 
1208 static int
1209 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1210     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1211     int *alwaysfragp, u_int fibnum)
1212 {
1213 	u_int32_t mtu = 0;
1214 	int alwaysfrag = 0;
1215 	int error = 0;
1216 
1217 	if (ro_pmtu != ro) {
1218 		/* The first hop and the final destination may differ. */
1219 		struct sockaddr_in6 *sa6_dst =
1220 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1221 		if (ro_pmtu->ro_rt &&
1222 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1223 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1224 			RTFREE(ro_pmtu->ro_rt);
1225 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1226 		}
1227 		if (ro_pmtu->ro_rt == NULL) {
1228 			bzero(sa6_dst, sizeof(*sa6_dst));
1229 			sa6_dst->sin6_family = AF_INET6;
1230 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1231 			sa6_dst->sin6_addr = *dst;
1232 
1233 			in6_rtalloc(ro_pmtu, fibnum);
1234 		}
1235 	}
1236 	if (ro_pmtu->ro_rt) {
1237 		u_int32_t ifmtu;
1238 		struct in_conninfo inc;
1239 
1240 		bzero(&inc, sizeof(inc));
1241 		inc.inc_flags |= INC_ISIPV6;
1242 		inc.inc6_faddr = *dst;
1243 
1244 		if (ifp == NULL)
1245 			ifp = ro_pmtu->ro_rt->rt_ifp;
1246 		ifmtu = IN6_LINKMTU(ifp);
1247 		mtu = tcp_hc_getmtu(&inc);
1248 		if (mtu)
1249 			mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1250 		else
1251 			mtu = ro_pmtu->ro_rt->rt_mtu;
1252 		if (mtu == 0)
1253 			mtu = ifmtu;
1254 		else if (mtu < IPV6_MMTU) {
1255 			/*
1256 			 * RFC2460 section 5, last paragraph:
1257 			 * if we record ICMPv6 too big message with
1258 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1259 			 * or smaller, with framgent header attached.
1260 			 * (fragment header is needed regardless from the
1261 			 * packet size, for translators to identify packets)
1262 			 */
1263 			alwaysfrag = 1;
1264 			mtu = IPV6_MMTU;
1265 		}
1266 	} else if (ifp) {
1267 		mtu = IN6_LINKMTU(ifp);
1268 	} else
1269 		error = EHOSTUNREACH; /* XXX */
1270 
1271 	*mtup = mtu;
1272 	if (alwaysfragp)
1273 		*alwaysfragp = alwaysfrag;
1274 	return (error);
1275 }
1276 
1277 /*
1278  * IP6 socket option processing.
1279  */
1280 int
1281 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1282 {
1283 	int optdatalen, uproto;
1284 	void *optdata;
1285 	struct inpcb *in6p = sotoinpcb(so);
1286 	int error, optval;
1287 	int level, op, optname;
1288 	int optlen;
1289 	struct thread *td;
1290 #ifdef	RSS
1291 	uint32_t rss_bucket;
1292 	int retval;
1293 #endif
1294 
1295 	level = sopt->sopt_level;
1296 	op = sopt->sopt_dir;
1297 	optname = sopt->sopt_name;
1298 	optlen = sopt->sopt_valsize;
1299 	td = sopt->sopt_td;
1300 	error = 0;
1301 	optval = 0;
1302 	uproto = (int)so->so_proto->pr_protocol;
1303 
1304 	if (level != IPPROTO_IPV6) {
1305 		error = EINVAL;
1306 
1307 		if (sopt->sopt_level == SOL_SOCKET &&
1308 		    sopt->sopt_dir == SOPT_SET) {
1309 			switch (sopt->sopt_name) {
1310 			case SO_REUSEADDR:
1311 				INP_WLOCK(in6p);
1312 				if ((so->so_options & SO_REUSEADDR) != 0)
1313 					in6p->inp_flags2 |= INP_REUSEADDR;
1314 				else
1315 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1316 				INP_WUNLOCK(in6p);
1317 				error = 0;
1318 				break;
1319 			case SO_REUSEPORT:
1320 				INP_WLOCK(in6p);
1321 				if ((so->so_options & SO_REUSEPORT) != 0)
1322 					in6p->inp_flags2 |= INP_REUSEPORT;
1323 				else
1324 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1325 				INP_WUNLOCK(in6p);
1326 				error = 0;
1327 				break;
1328 			case SO_SETFIB:
1329 				INP_WLOCK(in6p);
1330 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1331 				INP_WUNLOCK(in6p);
1332 				error = 0;
1333 				break;
1334 			default:
1335 				break;
1336 			}
1337 		}
1338 	} else {		/* level == IPPROTO_IPV6 */
1339 		switch (op) {
1340 
1341 		case SOPT_SET:
1342 			switch (optname) {
1343 			case IPV6_2292PKTOPTIONS:
1344 #ifdef IPV6_PKTOPTIONS
1345 			case IPV6_PKTOPTIONS:
1346 #endif
1347 			{
1348 				struct mbuf *m;
1349 
1350 				error = soopt_getm(sopt, &m); /* XXX */
1351 				if (error != 0)
1352 					break;
1353 				error = soopt_mcopyin(sopt, m); /* XXX */
1354 				if (error != 0)
1355 					break;
1356 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1357 						    m, so, sopt);
1358 				m_freem(m); /* XXX */
1359 				break;
1360 			}
1361 
1362 			/*
1363 			 * Use of some Hop-by-Hop options or some
1364 			 * Destination options, might require special
1365 			 * privilege.  That is, normal applications
1366 			 * (without special privilege) might be forbidden
1367 			 * from setting certain options in outgoing packets,
1368 			 * and might never see certain options in received
1369 			 * packets. [RFC 2292 Section 6]
1370 			 * KAME specific note:
1371 			 *  KAME prevents non-privileged users from sending or
1372 			 *  receiving ANY hbh/dst options in order to avoid
1373 			 *  overhead of parsing options in the kernel.
1374 			 */
1375 			case IPV6_RECVHOPOPTS:
1376 			case IPV6_RECVDSTOPTS:
1377 			case IPV6_RECVRTHDRDSTOPTS:
1378 				if (td != NULL) {
1379 					error = priv_check(td,
1380 					    PRIV_NETINET_SETHDROPTS);
1381 					if (error)
1382 						break;
1383 				}
1384 				/* FALLTHROUGH */
1385 			case IPV6_UNICAST_HOPS:
1386 			case IPV6_HOPLIMIT:
1387 
1388 			case IPV6_RECVPKTINFO:
1389 			case IPV6_RECVHOPLIMIT:
1390 			case IPV6_RECVRTHDR:
1391 			case IPV6_RECVPATHMTU:
1392 			case IPV6_RECVTCLASS:
1393 			case IPV6_V6ONLY:
1394 			case IPV6_AUTOFLOWLABEL:
1395 			case IPV6_BINDANY:
1396 			case IPV6_BINDMULTI:
1397 #ifdef	RSS
1398 			case IPV6_RSS_LISTEN_BUCKET:
1399 #endif
1400 				if (optname == IPV6_BINDANY && td != NULL) {
1401 					error = priv_check(td,
1402 					    PRIV_NETINET_BINDANY);
1403 					if (error)
1404 						break;
1405 				}
1406 
1407 				if (optlen != sizeof(int)) {
1408 					error = EINVAL;
1409 					break;
1410 				}
1411 				error = sooptcopyin(sopt, &optval,
1412 					sizeof optval, sizeof optval);
1413 				if (error)
1414 					break;
1415 				switch (optname) {
1416 
1417 				case IPV6_UNICAST_HOPS:
1418 					if (optval < -1 || optval >= 256)
1419 						error = EINVAL;
1420 					else {
1421 						/* -1 = kernel default */
1422 						in6p->in6p_hops = optval;
1423 						if ((in6p->inp_vflag &
1424 						     INP_IPV4) != 0)
1425 							in6p->inp_ip_ttl = optval;
1426 					}
1427 					break;
1428 #define OPTSET(bit) \
1429 do { \
1430 	INP_WLOCK(in6p); \
1431 	if (optval) \
1432 		in6p->inp_flags |= (bit); \
1433 	else \
1434 		in6p->inp_flags &= ~(bit); \
1435 	INP_WUNLOCK(in6p); \
1436 } while (/*CONSTCOND*/ 0)
1437 #define OPTSET2292(bit) \
1438 do { \
1439 	INP_WLOCK(in6p); \
1440 	in6p->inp_flags |= IN6P_RFC2292; \
1441 	if (optval) \
1442 		in6p->inp_flags |= (bit); \
1443 	else \
1444 		in6p->inp_flags &= ~(bit); \
1445 	INP_WUNLOCK(in6p); \
1446 } while (/*CONSTCOND*/ 0)
1447 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1448 
1449 #define OPTSET2(bit, val) do {						\
1450 	INP_WLOCK(in6p);						\
1451 	if (val)							\
1452 		in6p->inp_flags2 |= bit;				\
1453 	else								\
1454 		in6p->inp_flags2 &= ~bit;				\
1455 	INP_WUNLOCK(in6p);						\
1456 } while (0)
1457 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1458 
1459 				case IPV6_RECVPKTINFO:
1460 					/* cannot mix with RFC2292 */
1461 					if (OPTBIT(IN6P_RFC2292)) {
1462 						error = EINVAL;
1463 						break;
1464 					}
1465 					OPTSET(IN6P_PKTINFO);
1466 					break;
1467 
1468 				case IPV6_HOPLIMIT:
1469 				{
1470 					struct ip6_pktopts **optp;
1471 
1472 					/* cannot mix with RFC2292 */
1473 					if (OPTBIT(IN6P_RFC2292)) {
1474 						error = EINVAL;
1475 						break;
1476 					}
1477 					optp = &in6p->in6p_outputopts;
1478 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1479 					    (u_char *)&optval, sizeof(optval),
1480 					    optp, (td != NULL) ? td->td_ucred :
1481 					    NULL, uproto);
1482 					break;
1483 				}
1484 
1485 				case IPV6_RECVHOPLIMIT:
1486 					/* cannot mix with RFC2292 */
1487 					if (OPTBIT(IN6P_RFC2292)) {
1488 						error = EINVAL;
1489 						break;
1490 					}
1491 					OPTSET(IN6P_HOPLIMIT);
1492 					break;
1493 
1494 				case IPV6_RECVHOPOPTS:
1495 					/* cannot mix with RFC2292 */
1496 					if (OPTBIT(IN6P_RFC2292)) {
1497 						error = EINVAL;
1498 						break;
1499 					}
1500 					OPTSET(IN6P_HOPOPTS);
1501 					break;
1502 
1503 				case IPV6_RECVDSTOPTS:
1504 					/* cannot mix with RFC2292 */
1505 					if (OPTBIT(IN6P_RFC2292)) {
1506 						error = EINVAL;
1507 						break;
1508 					}
1509 					OPTSET(IN6P_DSTOPTS);
1510 					break;
1511 
1512 				case IPV6_RECVRTHDRDSTOPTS:
1513 					/* cannot mix with RFC2292 */
1514 					if (OPTBIT(IN6P_RFC2292)) {
1515 						error = EINVAL;
1516 						break;
1517 					}
1518 					OPTSET(IN6P_RTHDRDSTOPTS);
1519 					break;
1520 
1521 				case IPV6_RECVRTHDR:
1522 					/* cannot mix with RFC2292 */
1523 					if (OPTBIT(IN6P_RFC2292)) {
1524 						error = EINVAL;
1525 						break;
1526 					}
1527 					OPTSET(IN6P_RTHDR);
1528 					break;
1529 
1530 				case IPV6_RECVPATHMTU:
1531 					/*
1532 					 * We ignore this option for TCP
1533 					 * sockets.
1534 					 * (RFC3542 leaves this case
1535 					 * unspecified.)
1536 					 */
1537 					if (uproto != IPPROTO_TCP)
1538 						OPTSET(IN6P_MTU);
1539 					break;
1540 
1541 				case IPV6_V6ONLY:
1542 					/*
1543 					 * make setsockopt(IPV6_V6ONLY)
1544 					 * available only prior to bind(2).
1545 					 * see ipng mailing list, Jun 22 2001.
1546 					 */
1547 					if (in6p->inp_lport ||
1548 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1549 						error = EINVAL;
1550 						break;
1551 					}
1552 					OPTSET(IN6P_IPV6_V6ONLY);
1553 					if (optval)
1554 						in6p->inp_vflag &= ~INP_IPV4;
1555 					else
1556 						in6p->inp_vflag |= INP_IPV4;
1557 					break;
1558 				case IPV6_RECVTCLASS:
1559 					/* cannot mix with RFC2292 XXX */
1560 					if (OPTBIT(IN6P_RFC2292)) {
1561 						error = EINVAL;
1562 						break;
1563 					}
1564 					OPTSET(IN6P_TCLASS);
1565 					break;
1566 				case IPV6_AUTOFLOWLABEL:
1567 					OPTSET(IN6P_AUTOFLOWLABEL);
1568 					break;
1569 
1570 				case IPV6_BINDANY:
1571 					OPTSET(INP_BINDANY);
1572 					break;
1573 
1574 				case IPV6_BINDMULTI:
1575 					OPTSET2(INP_BINDMULTI, optval);
1576 					break;
1577 #ifdef	RSS
1578 				case IPV6_RSS_LISTEN_BUCKET:
1579 					if ((optval >= 0) &&
1580 					    (optval < rss_getnumbuckets())) {
1581 						in6p->inp_rss_listen_bucket = optval;
1582 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1583 					} else {
1584 						error = EINVAL;
1585 					}
1586 					break;
1587 #endif
1588 				}
1589 				break;
1590 
1591 			case IPV6_TCLASS:
1592 			case IPV6_DONTFRAG:
1593 			case IPV6_USE_MIN_MTU:
1594 			case IPV6_PREFER_TEMPADDR:
1595 				if (optlen != sizeof(optval)) {
1596 					error = EINVAL;
1597 					break;
1598 				}
1599 				error = sooptcopyin(sopt, &optval,
1600 					sizeof optval, sizeof optval);
1601 				if (error)
1602 					break;
1603 				{
1604 					struct ip6_pktopts **optp;
1605 					optp = &in6p->in6p_outputopts;
1606 					error = ip6_pcbopt(optname,
1607 					    (u_char *)&optval, sizeof(optval),
1608 					    optp, (td != NULL) ? td->td_ucred :
1609 					    NULL, uproto);
1610 					break;
1611 				}
1612 
1613 			case IPV6_2292PKTINFO:
1614 			case IPV6_2292HOPLIMIT:
1615 			case IPV6_2292HOPOPTS:
1616 			case IPV6_2292DSTOPTS:
1617 			case IPV6_2292RTHDR:
1618 				/* RFC 2292 */
1619 				if (optlen != sizeof(int)) {
1620 					error = EINVAL;
1621 					break;
1622 				}
1623 				error = sooptcopyin(sopt, &optval,
1624 					sizeof optval, sizeof optval);
1625 				if (error)
1626 					break;
1627 				switch (optname) {
1628 				case IPV6_2292PKTINFO:
1629 					OPTSET2292(IN6P_PKTINFO);
1630 					break;
1631 				case IPV6_2292HOPLIMIT:
1632 					OPTSET2292(IN6P_HOPLIMIT);
1633 					break;
1634 				case IPV6_2292HOPOPTS:
1635 					/*
1636 					 * Check super-user privilege.
1637 					 * See comments for IPV6_RECVHOPOPTS.
1638 					 */
1639 					if (td != NULL) {
1640 						error = priv_check(td,
1641 						    PRIV_NETINET_SETHDROPTS);
1642 						if (error)
1643 							return (error);
1644 					}
1645 					OPTSET2292(IN6P_HOPOPTS);
1646 					break;
1647 				case IPV6_2292DSTOPTS:
1648 					if (td != NULL) {
1649 						error = priv_check(td,
1650 						    PRIV_NETINET_SETHDROPTS);
1651 						if (error)
1652 							return (error);
1653 					}
1654 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1655 					break;
1656 				case IPV6_2292RTHDR:
1657 					OPTSET2292(IN6P_RTHDR);
1658 					break;
1659 				}
1660 				break;
1661 			case IPV6_PKTINFO:
1662 			case IPV6_HOPOPTS:
1663 			case IPV6_RTHDR:
1664 			case IPV6_DSTOPTS:
1665 			case IPV6_RTHDRDSTOPTS:
1666 			case IPV6_NEXTHOP:
1667 			{
1668 				/* new advanced API (RFC3542) */
1669 				u_char *optbuf;
1670 				u_char optbuf_storage[MCLBYTES];
1671 				int optlen;
1672 				struct ip6_pktopts **optp;
1673 
1674 				/* cannot mix with RFC2292 */
1675 				if (OPTBIT(IN6P_RFC2292)) {
1676 					error = EINVAL;
1677 					break;
1678 				}
1679 
1680 				/*
1681 				 * We only ensure valsize is not too large
1682 				 * here.  Further validation will be done
1683 				 * later.
1684 				 */
1685 				error = sooptcopyin(sopt, optbuf_storage,
1686 				    sizeof(optbuf_storage), 0);
1687 				if (error)
1688 					break;
1689 				optlen = sopt->sopt_valsize;
1690 				optbuf = optbuf_storage;
1691 				optp = &in6p->in6p_outputopts;
1692 				error = ip6_pcbopt(optname, optbuf, optlen,
1693 				    optp, (td != NULL) ? td->td_ucred : NULL,
1694 				    uproto);
1695 				break;
1696 			}
1697 #undef OPTSET
1698 
1699 			case IPV6_MULTICAST_IF:
1700 			case IPV6_MULTICAST_HOPS:
1701 			case IPV6_MULTICAST_LOOP:
1702 			case IPV6_JOIN_GROUP:
1703 			case IPV6_LEAVE_GROUP:
1704 			case IPV6_MSFILTER:
1705 			case MCAST_BLOCK_SOURCE:
1706 			case MCAST_UNBLOCK_SOURCE:
1707 			case MCAST_JOIN_GROUP:
1708 			case MCAST_LEAVE_GROUP:
1709 			case MCAST_JOIN_SOURCE_GROUP:
1710 			case MCAST_LEAVE_SOURCE_GROUP:
1711 				error = ip6_setmoptions(in6p, sopt);
1712 				break;
1713 
1714 			case IPV6_PORTRANGE:
1715 				error = sooptcopyin(sopt, &optval,
1716 				    sizeof optval, sizeof optval);
1717 				if (error)
1718 					break;
1719 
1720 				INP_WLOCK(in6p);
1721 				switch (optval) {
1722 				case IPV6_PORTRANGE_DEFAULT:
1723 					in6p->inp_flags &= ~(INP_LOWPORT);
1724 					in6p->inp_flags &= ~(INP_HIGHPORT);
1725 					break;
1726 
1727 				case IPV6_PORTRANGE_HIGH:
1728 					in6p->inp_flags &= ~(INP_LOWPORT);
1729 					in6p->inp_flags |= INP_HIGHPORT;
1730 					break;
1731 
1732 				case IPV6_PORTRANGE_LOW:
1733 					in6p->inp_flags &= ~(INP_HIGHPORT);
1734 					in6p->inp_flags |= INP_LOWPORT;
1735 					break;
1736 
1737 				default:
1738 					error = EINVAL;
1739 					break;
1740 				}
1741 				INP_WUNLOCK(in6p);
1742 				break;
1743 
1744 #ifdef IPSEC
1745 			case IPV6_IPSEC_POLICY:
1746 			{
1747 				caddr_t req;
1748 				struct mbuf *m;
1749 
1750 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1751 					break;
1752 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1753 					break;
1754 				req = mtod(m, caddr_t);
1755 				error = ipsec_set_policy(in6p, optname, req,
1756 				    m->m_len, (sopt->sopt_td != NULL) ?
1757 				    sopt->sopt_td->td_ucred : NULL);
1758 				m_freem(m);
1759 				break;
1760 			}
1761 #endif /* IPSEC */
1762 
1763 			default:
1764 				error = ENOPROTOOPT;
1765 				break;
1766 			}
1767 			break;
1768 
1769 		case SOPT_GET:
1770 			switch (optname) {
1771 
1772 			case IPV6_2292PKTOPTIONS:
1773 #ifdef IPV6_PKTOPTIONS
1774 			case IPV6_PKTOPTIONS:
1775 #endif
1776 				/*
1777 				 * RFC3542 (effectively) deprecated the
1778 				 * semantics of the 2292-style pktoptions.
1779 				 * Since it was not reliable in nature (i.e.,
1780 				 * applications had to expect the lack of some
1781 				 * information after all), it would make sense
1782 				 * to simplify this part by always returning
1783 				 * empty data.
1784 				 */
1785 				sopt->sopt_valsize = 0;
1786 				break;
1787 
1788 			case IPV6_RECVHOPOPTS:
1789 			case IPV6_RECVDSTOPTS:
1790 			case IPV6_RECVRTHDRDSTOPTS:
1791 			case IPV6_UNICAST_HOPS:
1792 			case IPV6_RECVPKTINFO:
1793 			case IPV6_RECVHOPLIMIT:
1794 			case IPV6_RECVRTHDR:
1795 			case IPV6_RECVPATHMTU:
1796 
1797 			case IPV6_V6ONLY:
1798 			case IPV6_PORTRANGE:
1799 			case IPV6_RECVTCLASS:
1800 			case IPV6_AUTOFLOWLABEL:
1801 			case IPV6_BINDANY:
1802 			case IPV6_FLOWID:
1803 			case IPV6_FLOWTYPE:
1804 #ifdef	RSS
1805 			case IPV6_RSSBUCKETID:
1806 #endif
1807 				switch (optname) {
1808 
1809 				case IPV6_RECVHOPOPTS:
1810 					optval = OPTBIT(IN6P_HOPOPTS);
1811 					break;
1812 
1813 				case IPV6_RECVDSTOPTS:
1814 					optval = OPTBIT(IN6P_DSTOPTS);
1815 					break;
1816 
1817 				case IPV6_RECVRTHDRDSTOPTS:
1818 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1819 					break;
1820 
1821 				case IPV6_UNICAST_HOPS:
1822 					optval = in6p->in6p_hops;
1823 					break;
1824 
1825 				case IPV6_RECVPKTINFO:
1826 					optval = OPTBIT(IN6P_PKTINFO);
1827 					break;
1828 
1829 				case IPV6_RECVHOPLIMIT:
1830 					optval = OPTBIT(IN6P_HOPLIMIT);
1831 					break;
1832 
1833 				case IPV6_RECVRTHDR:
1834 					optval = OPTBIT(IN6P_RTHDR);
1835 					break;
1836 
1837 				case IPV6_RECVPATHMTU:
1838 					optval = OPTBIT(IN6P_MTU);
1839 					break;
1840 
1841 				case IPV6_V6ONLY:
1842 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1843 					break;
1844 
1845 				case IPV6_PORTRANGE:
1846 				    {
1847 					int flags;
1848 					flags = in6p->inp_flags;
1849 					if (flags & INP_HIGHPORT)
1850 						optval = IPV6_PORTRANGE_HIGH;
1851 					else if (flags & INP_LOWPORT)
1852 						optval = IPV6_PORTRANGE_LOW;
1853 					else
1854 						optval = 0;
1855 					break;
1856 				    }
1857 				case IPV6_RECVTCLASS:
1858 					optval = OPTBIT(IN6P_TCLASS);
1859 					break;
1860 
1861 				case IPV6_AUTOFLOWLABEL:
1862 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1863 					break;
1864 
1865 				case IPV6_BINDANY:
1866 					optval = OPTBIT(INP_BINDANY);
1867 					break;
1868 
1869 				case IPV6_FLOWID:
1870 					optval = in6p->inp_flowid;
1871 					break;
1872 
1873 				case IPV6_FLOWTYPE:
1874 					optval = in6p->inp_flowtype;
1875 					break;
1876 #ifdef	RSS
1877 				case IPV6_RSSBUCKETID:
1878 					retval =
1879 					    rss_hash2bucket(in6p->inp_flowid,
1880 					    in6p->inp_flowtype,
1881 					    &rss_bucket);
1882 					if (retval == 0)
1883 						optval = rss_bucket;
1884 					else
1885 						error = EINVAL;
1886 					break;
1887 #endif
1888 
1889 				case IPV6_BINDMULTI:
1890 					optval = OPTBIT2(INP_BINDMULTI);
1891 					break;
1892 
1893 				}
1894 				if (error)
1895 					break;
1896 				error = sooptcopyout(sopt, &optval,
1897 					sizeof optval);
1898 				break;
1899 
1900 			case IPV6_PATHMTU:
1901 			{
1902 				u_long pmtu = 0;
1903 				struct ip6_mtuinfo mtuinfo;
1904 				struct route_in6 sro;
1905 
1906 				bzero(&sro, sizeof(sro));
1907 
1908 				if (!(so->so_state & SS_ISCONNECTED))
1909 					return (ENOTCONN);
1910 				/*
1911 				 * XXX: we dot not consider the case of source
1912 				 * routing, or optional information to specify
1913 				 * the outgoing interface.
1914 				 */
1915 				error = ip6_getpmtu(&sro, NULL, NULL,
1916 				    &in6p->in6p_faddr, &pmtu, NULL,
1917 				    so->so_fibnum);
1918 				if (sro.ro_rt)
1919 					RTFREE(sro.ro_rt);
1920 				if (error)
1921 					break;
1922 				if (pmtu > IPV6_MAXPACKET)
1923 					pmtu = IPV6_MAXPACKET;
1924 
1925 				bzero(&mtuinfo, sizeof(mtuinfo));
1926 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1927 				optdata = (void *)&mtuinfo;
1928 				optdatalen = sizeof(mtuinfo);
1929 				error = sooptcopyout(sopt, optdata,
1930 				    optdatalen);
1931 				break;
1932 			}
1933 
1934 			case IPV6_2292PKTINFO:
1935 			case IPV6_2292HOPLIMIT:
1936 			case IPV6_2292HOPOPTS:
1937 			case IPV6_2292RTHDR:
1938 			case IPV6_2292DSTOPTS:
1939 				switch (optname) {
1940 				case IPV6_2292PKTINFO:
1941 					optval = OPTBIT(IN6P_PKTINFO);
1942 					break;
1943 				case IPV6_2292HOPLIMIT:
1944 					optval = OPTBIT(IN6P_HOPLIMIT);
1945 					break;
1946 				case IPV6_2292HOPOPTS:
1947 					optval = OPTBIT(IN6P_HOPOPTS);
1948 					break;
1949 				case IPV6_2292RTHDR:
1950 					optval = OPTBIT(IN6P_RTHDR);
1951 					break;
1952 				case IPV6_2292DSTOPTS:
1953 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1954 					break;
1955 				}
1956 				error = sooptcopyout(sopt, &optval,
1957 				    sizeof optval);
1958 				break;
1959 			case IPV6_PKTINFO:
1960 			case IPV6_HOPOPTS:
1961 			case IPV6_RTHDR:
1962 			case IPV6_DSTOPTS:
1963 			case IPV6_RTHDRDSTOPTS:
1964 			case IPV6_NEXTHOP:
1965 			case IPV6_TCLASS:
1966 			case IPV6_DONTFRAG:
1967 			case IPV6_USE_MIN_MTU:
1968 			case IPV6_PREFER_TEMPADDR:
1969 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1970 				    optname, sopt);
1971 				break;
1972 
1973 			case IPV6_MULTICAST_IF:
1974 			case IPV6_MULTICAST_HOPS:
1975 			case IPV6_MULTICAST_LOOP:
1976 			case IPV6_MSFILTER:
1977 				error = ip6_getmoptions(in6p, sopt);
1978 				break;
1979 
1980 #ifdef IPSEC
1981 			case IPV6_IPSEC_POLICY:
1982 			  {
1983 				caddr_t req = NULL;
1984 				size_t len = 0;
1985 				struct mbuf *m = NULL;
1986 				struct mbuf **mp = &m;
1987 				size_t ovalsize = sopt->sopt_valsize;
1988 				caddr_t oval = (caddr_t)sopt->sopt_val;
1989 
1990 				error = soopt_getm(sopt, &m); /* XXX */
1991 				if (error != 0)
1992 					break;
1993 				error = soopt_mcopyin(sopt, m); /* XXX */
1994 				if (error != 0)
1995 					break;
1996 				sopt->sopt_valsize = ovalsize;
1997 				sopt->sopt_val = oval;
1998 				if (m) {
1999 					req = mtod(m, caddr_t);
2000 					len = m->m_len;
2001 				}
2002 				error = ipsec_get_policy(in6p, req, len, mp);
2003 				if (error == 0)
2004 					error = soopt_mcopyout(sopt, m); /* XXX */
2005 				if (error == 0 && m)
2006 					m_freem(m);
2007 				break;
2008 			  }
2009 #endif /* IPSEC */
2010 
2011 			default:
2012 				error = ENOPROTOOPT;
2013 				break;
2014 			}
2015 			break;
2016 		}
2017 	}
2018 	return (error);
2019 }
2020 
2021 int
2022 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2023 {
2024 	int error = 0, optval, optlen;
2025 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2026 	struct inpcb *in6p = sotoinpcb(so);
2027 	int level, op, optname;
2028 
2029 	level = sopt->sopt_level;
2030 	op = sopt->sopt_dir;
2031 	optname = sopt->sopt_name;
2032 	optlen = sopt->sopt_valsize;
2033 
2034 	if (level != IPPROTO_IPV6) {
2035 		return (EINVAL);
2036 	}
2037 
2038 	switch (optname) {
2039 	case IPV6_CHECKSUM:
2040 		/*
2041 		 * For ICMPv6 sockets, no modification allowed for checksum
2042 		 * offset, permit "no change" values to help existing apps.
2043 		 *
2044 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2045 		 * for an ICMPv6 socket will fail."
2046 		 * The current behavior does not meet RFC3542.
2047 		 */
2048 		switch (op) {
2049 		case SOPT_SET:
2050 			if (optlen != sizeof(int)) {
2051 				error = EINVAL;
2052 				break;
2053 			}
2054 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2055 					    sizeof(optval));
2056 			if (error)
2057 				break;
2058 			if ((optval % 2) != 0) {
2059 				/* the API assumes even offset values */
2060 				error = EINVAL;
2061 			} else if (so->so_proto->pr_protocol ==
2062 			    IPPROTO_ICMPV6) {
2063 				if (optval != icmp6off)
2064 					error = EINVAL;
2065 			} else
2066 				in6p->in6p_cksum = optval;
2067 			break;
2068 
2069 		case SOPT_GET:
2070 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2071 				optval = icmp6off;
2072 			else
2073 				optval = in6p->in6p_cksum;
2074 
2075 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2076 			break;
2077 
2078 		default:
2079 			error = EINVAL;
2080 			break;
2081 		}
2082 		break;
2083 
2084 	default:
2085 		error = ENOPROTOOPT;
2086 		break;
2087 	}
2088 
2089 	return (error);
2090 }
2091 
2092 /*
2093  * Set up IP6 options in pcb for insertion in output packets or
2094  * specifying behavior of outgoing packets.
2095  */
2096 static int
2097 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2098     struct socket *so, struct sockopt *sopt)
2099 {
2100 	struct ip6_pktopts *opt = *pktopt;
2101 	int error = 0;
2102 	struct thread *td = sopt->sopt_td;
2103 
2104 	/* turn off any old options. */
2105 	if (opt) {
2106 #ifdef DIAGNOSTIC
2107 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2108 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2109 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2110 			printf("ip6_pcbopts: all specified options are cleared.\n");
2111 #endif
2112 		ip6_clearpktopts(opt, -1);
2113 	} else
2114 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2115 	*pktopt = NULL;
2116 
2117 	if (!m || m->m_len == 0) {
2118 		/*
2119 		 * Only turning off any previous options, regardless of
2120 		 * whether the opt is just created or given.
2121 		 */
2122 		free(opt, M_IP6OPT);
2123 		return (0);
2124 	}
2125 
2126 	/*  set options specified by user. */
2127 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2128 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2129 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2130 		free(opt, M_IP6OPT);
2131 		return (error);
2132 	}
2133 	*pktopt = opt;
2134 	return (0);
2135 }
2136 
2137 /*
2138  * initialize ip6_pktopts.  beware that there are non-zero default values in
2139  * the struct.
2140  */
2141 void
2142 ip6_initpktopts(struct ip6_pktopts *opt)
2143 {
2144 
2145 	bzero(opt, sizeof(*opt));
2146 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2147 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2148 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2149 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2150 }
2151 
2152 static int
2153 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2154     struct ucred *cred, int uproto)
2155 {
2156 	struct ip6_pktopts *opt;
2157 
2158 	if (*pktopt == NULL) {
2159 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2160 		    M_WAITOK);
2161 		ip6_initpktopts(*pktopt);
2162 	}
2163 	opt = *pktopt;
2164 
2165 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2166 }
2167 
2168 static int
2169 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2170 {
2171 	void *optdata = NULL;
2172 	int optdatalen = 0;
2173 	struct ip6_ext *ip6e;
2174 	int error = 0;
2175 	struct in6_pktinfo null_pktinfo;
2176 	int deftclass = 0, on;
2177 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2178 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2179 
2180 	switch (optname) {
2181 	case IPV6_PKTINFO:
2182 		if (pktopt && pktopt->ip6po_pktinfo)
2183 			optdata = (void *)pktopt->ip6po_pktinfo;
2184 		else {
2185 			/* XXX: we don't have to do this every time... */
2186 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2187 			optdata = (void *)&null_pktinfo;
2188 		}
2189 		optdatalen = sizeof(struct in6_pktinfo);
2190 		break;
2191 	case IPV6_TCLASS:
2192 		if (pktopt && pktopt->ip6po_tclass >= 0)
2193 			optdata = (void *)&pktopt->ip6po_tclass;
2194 		else
2195 			optdata = (void *)&deftclass;
2196 		optdatalen = sizeof(int);
2197 		break;
2198 	case IPV6_HOPOPTS:
2199 		if (pktopt && pktopt->ip6po_hbh) {
2200 			optdata = (void *)pktopt->ip6po_hbh;
2201 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2202 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2203 		}
2204 		break;
2205 	case IPV6_RTHDR:
2206 		if (pktopt && pktopt->ip6po_rthdr) {
2207 			optdata = (void *)pktopt->ip6po_rthdr;
2208 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2209 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2210 		}
2211 		break;
2212 	case IPV6_RTHDRDSTOPTS:
2213 		if (pktopt && pktopt->ip6po_dest1) {
2214 			optdata = (void *)pktopt->ip6po_dest1;
2215 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2216 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2217 		}
2218 		break;
2219 	case IPV6_DSTOPTS:
2220 		if (pktopt && pktopt->ip6po_dest2) {
2221 			optdata = (void *)pktopt->ip6po_dest2;
2222 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2223 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2224 		}
2225 		break;
2226 	case IPV6_NEXTHOP:
2227 		if (pktopt && pktopt->ip6po_nexthop) {
2228 			optdata = (void *)pktopt->ip6po_nexthop;
2229 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2230 		}
2231 		break;
2232 	case IPV6_USE_MIN_MTU:
2233 		if (pktopt)
2234 			optdata = (void *)&pktopt->ip6po_minmtu;
2235 		else
2236 			optdata = (void *)&defminmtu;
2237 		optdatalen = sizeof(int);
2238 		break;
2239 	case IPV6_DONTFRAG:
2240 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2241 			on = 1;
2242 		else
2243 			on = 0;
2244 		optdata = (void *)&on;
2245 		optdatalen = sizeof(on);
2246 		break;
2247 	case IPV6_PREFER_TEMPADDR:
2248 		if (pktopt)
2249 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2250 		else
2251 			optdata = (void *)&defpreftemp;
2252 		optdatalen = sizeof(int);
2253 		break;
2254 	default:		/* should not happen */
2255 #ifdef DIAGNOSTIC
2256 		panic("ip6_getpcbopt: unexpected option\n");
2257 #endif
2258 		return (ENOPROTOOPT);
2259 	}
2260 
2261 	error = sooptcopyout(sopt, optdata, optdatalen);
2262 
2263 	return (error);
2264 }
2265 
2266 void
2267 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2268 {
2269 	if (pktopt == NULL)
2270 		return;
2271 
2272 	if (optname == -1 || optname == IPV6_PKTINFO) {
2273 		if (pktopt->ip6po_pktinfo)
2274 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2275 		pktopt->ip6po_pktinfo = NULL;
2276 	}
2277 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2278 		pktopt->ip6po_hlim = -1;
2279 	if (optname == -1 || optname == IPV6_TCLASS)
2280 		pktopt->ip6po_tclass = -1;
2281 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2282 		if (pktopt->ip6po_nextroute.ro_rt) {
2283 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2284 			pktopt->ip6po_nextroute.ro_rt = NULL;
2285 		}
2286 		if (pktopt->ip6po_nexthop)
2287 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2288 		pktopt->ip6po_nexthop = NULL;
2289 	}
2290 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2291 		if (pktopt->ip6po_hbh)
2292 			free(pktopt->ip6po_hbh, M_IP6OPT);
2293 		pktopt->ip6po_hbh = NULL;
2294 	}
2295 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2296 		if (pktopt->ip6po_dest1)
2297 			free(pktopt->ip6po_dest1, M_IP6OPT);
2298 		pktopt->ip6po_dest1 = NULL;
2299 	}
2300 	if (optname == -1 || optname == IPV6_RTHDR) {
2301 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2302 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2303 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2304 		if (pktopt->ip6po_route.ro_rt) {
2305 			RTFREE(pktopt->ip6po_route.ro_rt);
2306 			pktopt->ip6po_route.ro_rt = NULL;
2307 		}
2308 	}
2309 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2310 		if (pktopt->ip6po_dest2)
2311 			free(pktopt->ip6po_dest2, M_IP6OPT);
2312 		pktopt->ip6po_dest2 = NULL;
2313 	}
2314 }
2315 
2316 #define PKTOPT_EXTHDRCPY(type) \
2317 do {\
2318 	if (src->type) {\
2319 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2320 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2321 		if (dst->type == NULL && canwait == M_NOWAIT)\
2322 			goto bad;\
2323 		bcopy(src->type, dst->type, hlen);\
2324 	}\
2325 } while (/*CONSTCOND*/ 0)
2326 
2327 static int
2328 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2329 {
2330 	if (dst == NULL || src == NULL)  {
2331 		printf("ip6_clearpktopts: invalid argument\n");
2332 		return (EINVAL);
2333 	}
2334 
2335 	dst->ip6po_hlim = src->ip6po_hlim;
2336 	dst->ip6po_tclass = src->ip6po_tclass;
2337 	dst->ip6po_flags = src->ip6po_flags;
2338 	dst->ip6po_minmtu = src->ip6po_minmtu;
2339 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2340 	if (src->ip6po_pktinfo) {
2341 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2342 		    M_IP6OPT, canwait);
2343 		if (dst->ip6po_pktinfo == NULL)
2344 			goto bad;
2345 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2346 	}
2347 	if (src->ip6po_nexthop) {
2348 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2349 		    M_IP6OPT, canwait);
2350 		if (dst->ip6po_nexthop == NULL)
2351 			goto bad;
2352 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2353 		    src->ip6po_nexthop->sa_len);
2354 	}
2355 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2356 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2357 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2358 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2359 	return (0);
2360 
2361   bad:
2362 	ip6_clearpktopts(dst, -1);
2363 	return (ENOBUFS);
2364 }
2365 #undef PKTOPT_EXTHDRCPY
2366 
2367 struct ip6_pktopts *
2368 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2369 {
2370 	int error;
2371 	struct ip6_pktopts *dst;
2372 
2373 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2374 	if (dst == NULL)
2375 		return (NULL);
2376 	ip6_initpktopts(dst);
2377 
2378 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2379 		free(dst, M_IP6OPT);
2380 		return (NULL);
2381 	}
2382 
2383 	return (dst);
2384 }
2385 
2386 void
2387 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2388 {
2389 	if (pktopt == NULL)
2390 		return;
2391 
2392 	ip6_clearpktopts(pktopt, -1);
2393 
2394 	free(pktopt, M_IP6OPT);
2395 }
2396 
2397 /*
2398  * Set IPv6 outgoing packet options based on advanced API.
2399  */
2400 int
2401 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2402     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2403 {
2404 	struct cmsghdr *cm = 0;
2405 
2406 	if (control == NULL || opt == NULL)
2407 		return (EINVAL);
2408 
2409 	ip6_initpktopts(opt);
2410 	if (stickyopt) {
2411 		int error;
2412 
2413 		/*
2414 		 * If stickyopt is provided, make a local copy of the options
2415 		 * for this particular packet, then override them by ancillary
2416 		 * objects.
2417 		 * XXX: copypktopts() does not copy the cached route to a next
2418 		 * hop (if any).  This is not very good in terms of efficiency,
2419 		 * but we can allow this since this option should be rarely
2420 		 * used.
2421 		 */
2422 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2423 			return (error);
2424 	}
2425 
2426 	/*
2427 	 * XXX: Currently, we assume all the optional information is stored
2428 	 * in a single mbuf.
2429 	 */
2430 	if (control->m_next)
2431 		return (EINVAL);
2432 
2433 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2434 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2435 		int error;
2436 
2437 		if (control->m_len < CMSG_LEN(0))
2438 			return (EINVAL);
2439 
2440 		cm = mtod(control, struct cmsghdr *);
2441 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2442 			return (EINVAL);
2443 		if (cm->cmsg_level != IPPROTO_IPV6)
2444 			continue;
2445 
2446 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2447 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2448 		if (error)
2449 			return (error);
2450 	}
2451 
2452 	return (0);
2453 }
2454 
2455 /*
2456  * Set a particular packet option, as a sticky option or an ancillary data
2457  * item.  "len" can be 0 only when it's a sticky option.
2458  * We have 4 cases of combination of "sticky" and "cmsg":
2459  * "sticky=0, cmsg=0": impossible
2460  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2461  * "sticky=1, cmsg=0": RFC3542 socket option
2462  * "sticky=1, cmsg=1": RFC2292 socket option
2463  */
2464 static int
2465 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2466     struct ucred *cred, int sticky, int cmsg, int uproto)
2467 {
2468 	int minmtupolicy, preftemp;
2469 	int error;
2470 
2471 	if (!sticky && !cmsg) {
2472 #ifdef DIAGNOSTIC
2473 		printf("ip6_setpktopt: impossible case\n");
2474 #endif
2475 		return (EINVAL);
2476 	}
2477 
2478 	/*
2479 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2480 	 * not be specified in the context of RFC3542.  Conversely,
2481 	 * RFC3542 types should not be specified in the context of RFC2292.
2482 	 */
2483 	if (!cmsg) {
2484 		switch (optname) {
2485 		case IPV6_2292PKTINFO:
2486 		case IPV6_2292HOPLIMIT:
2487 		case IPV6_2292NEXTHOP:
2488 		case IPV6_2292HOPOPTS:
2489 		case IPV6_2292DSTOPTS:
2490 		case IPV6_2292RTHDR:
2491 		case IPV6_2292PKTOPTIONS:
2492 			return (ENOPROTOOPT);
2493 		}
2494 	}
2495 	if (sticky && cmsg) {
2496 		switch (optname) {
2497 		case IPV6_PKTINFO:
2498 		case IPV6_HOPLIMIT:
2499 		case IPV6_NEXTHOP:
2500 		case IPV6_HOPOPTS:
2501 		case IPV6_DSTOPTS:
2502 		case IPV6_RTHDRDSTOPTS:
2503 		case IPV6_RTHDR:
2504 		case IPV6_USE_MIN_MTU:
2505 		case IPV6_DONTFRAG:
2506 		case IPV6_TCLASS:
2507 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2508 			return (ENOPROTOOPT);
2509 		}
2510 	}
2511 
2512 	switch (optname) {
2513 	case IPV6_2292PKTINFO:
2514 	case IPV6_PKTINFO:
2515 	{
2516 		struct ifnet *ifp = NULL;
2517 		struct in6_pktinfo *pktinfo;
2518 
2519 		if (len != sizeof(struct in6_pktinfo))
2520 			return (EINVAL);
2521 
2522 		pktinfo = (struct in6_pktinfo *)buf;
2523 
2524 		/*
2525 		 * An application can clear any sticky IPV6_PKTINFO option by
2526 		 * doing a "regular" setsockopt with ipi6_addr being
2527 		 * in6addr_any and ipi6_ifindex being zero.
2528 		 * [RFC 3542, Section 6]
2529 		 */
2530 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2531 		    pktinfo->ipi6_ifindex == 0 &&
2532 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2533 			ip6_clearpktopts(opt, optname);
2534 			break;
2535 		}
2536 
2537 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2538 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2539 			return (EINVAL);
2540 		}
2541 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2542 			return (EINVAL);
2543 		/* validate the interface index if specified. */
2544 		if (pktinfo->ipi6_ifindex > V_if_index)
2545 			 return (ENXIO);
2546 		if (pktinfo->ipi6_ifindex) {
2547 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2548 			if (ifp == NULL)
2549 				return (ENXIO);
2550 		}
2551 		if (ifp != NULL && (
2552 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2553 			return (ENETDOWN);
2554 
2555 		if (ifp != NULL &&
2556 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2557 			struct in6_ifaddr *ia;
2558 
2559 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2560 			if (ia == NULL)
2561 				return (EADDRNOTAVAIL);
2562 			ifa_free(&ia->ia_ifa);
2563 		}
2564 		/*
2565 		 * We store the address anyway, and let in6_selectsrc()
2566 		 * validate the specified address.  This is because ipi6_addr
2567 		 * may not have enough information about its scope zone, and
2568 		 * we may need additional information (such as outgoing
2569 		 * interface or the scope zone of a destination address) to
2570 		 * disambiguate the scope.
2571 		 * XXX: the delay of the validation may confuse the
2572 		 * application when it is used as a sticky option.
2573 		 */
2574 		if (opt->ip6po_pktinfo == NULL) {
2575 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2576 			    M_IP6OPT, M_NOWAIT);
2577 			if (opt->ip6po_pktinfo == NULL)
2578 				return (ENOBUFS);
2579 		}
2580 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2581 		break;
2582 	}
2583 
2584 	case IPV6_2292HOPLIMIT:
2585 	case IPV6_HOPLIMIT:
2586 	{
2587 		int *hlimp;
2588 
2589 		/*
2590 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2591 		 * to simplify the ordering among hoplimit options.
2592 		 */
2593 		if (optname == IPV6_HOPLIMIT && sticky)
2594 			return (ENOPROTOOPT);
2595 
2596 		if (len != sizeof(int))
2597 			return (EINVAL);
2598 		hlimp = (int *)buf;
2599 		if (*hlimp < -1 || *hlimp > 255)
2600 			return (EINVAL);
2601 
2602 		opt->ip6po_hlim = *hlimp;
2603 		break;
2604 	}
2605 
2606 	case IPV6_TCLASS:
2607 	{
2608 		int tclass;
2609 
2610 		if (len != sizeof(int))
2611 			return (EINVAL);
2612 		tclass = *(int *)buf;
2613 		if (tclass < -1 || tclass > 255)
2614 			return (EINVAL);
2615 
2616 		opt->ip6po_tclass = tclass;
2617 		break;
2618 	}
2619 
2620 	case IPV6_2292NEXTHOP:
2621 	case IPV6_NEXTHOP:
2622 		if (cred != NULL) {
2623 			error = priv_check_cred(cred,
2624 			    PRIV_NETINET_SETHDROPTS, 0);
2625 			if (error)
2626 				return (error);
2627 		}
2628 
2629 		if (len == 0) {	/* just remove the option */
2630 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2631 			break;
2632 		}
2633 
2634 		/* check if cmsg_len is large enough for sa_len */
2635 		if (len < sizeof(struct sockaddr) || len < *buf)
2636 			return (EINVAL);
2637 
2638 		switch (((struct sockaddr *)buf)->sa_family) {
2639 		case AF_INET6:
2640 		{
2641 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2642 			int error;
2643 
2644 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2645 				return (EINVAL);
2646 
2647 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2648 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2649 				return (EINVAL);
2650 			}
2651 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2652 			    != 0) {
2653 				return (error);
2654 			}
2655 			break;
2656 		}
2657 		case AF_LINK:	/* should eventually be supported */
2658 		default:
2659 			return (EAFNOSUPPORT);
2660 		}
2661 
2662 		/* turn off the previous option, then set the new option. */
2663 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2664 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2665 		if (opt->ip6po_nexthop == NULL)
2666 			return (ENOBUFS);
2667 		bcopy(buf, opt->ip6po_nexthop, *buf);
2668 		break;
2669 
2670 	case IPV6_2292HOPOPTS:
2671 	case IPV6_HOPOPTS:
2672 	{
2673 		struct ip6_hbh *hbh;
2674 		int hbhlen;
2675 
2676 		/*
2677 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2678 		 * options, since per-option restriction has too much
2679 		 * overhead.
2680 		 */
2681 		if (cred != NULL) {
2682 			error = priv_check_cred(cred,
2683 			    PRIV_NETINET_SETHDROPTS, 0);
2684 			if (error)
2685 				return (error);
2686 		}
2687 
2688 		if (len == 0) {
2689 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2690 			break;	/* just remove the option */
2691 		}
2692 
2693 		/* message length validation */
2694 		if (len < sizeof(struct ip6_hbh))
2695 			return (EINVAL);
2696 		hbh = (struct ip6_hbh *)buf;
2697 		hbhlen = (hbh->ip6h_len + 1) << 3;
2698 		if (len != hbhlen)
2699 			return (EINVAL);
2700 
2701 		/* turn off the previous option, then set the new option. */
2702 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2703 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2704 		if (opt->ip6po_hbh == NULL)
2705 			return (ENOBUFS);
2706 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2707 
2708 		break;
2709 	}
2710 
2711 	case IPV6_2292DSTOPTS:
2712 	case IPV6_DSTOPTS:
2713 	case IPV6_RTHDRDSTOPTS:
2714 	{
2715 		struct ip6_dest *dest, **newdest = NULL;
2716 		int destlen;
2717 
2718 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2719 			error = priv_check_cred(cred,
2720 			    PRIV_NETINET_SETHDROPTS, 0);
2721 			if (error)
2722 				return (error);
2723 		}
2724 
2725 		if (len == 0) {
2726 			ip6_clearpktopts(opt, optname);
2727 			break;	/* just remove the option */
2728 		}
2729 
2730 		/* message length validation */
2731 		if (len < sizeof(struct ip6_dest))
2732 			return (EINVAL);
2733 		dest = (struct ip6_dest *)buf;
2734 		destlen = (dest->ip6d_len + 1) << 3;
2735 		if (len != destlen)
2736 			return (EINVAL);
2737 
2738 		/*
2739 		 * Determine the position that the destination options header
2740 		 * should be inserted; before or after the routing header.
2741 		 */
2742 		switch (optname) {
2743 		case IPV6_2292DSTOPTS:
2744 			/*
2745 			 * The old advacned API is ambiguous on this point.
2746 			 * Our approach is to determine the position based
2747 			 * according to the existence of a routing header.
2748 			 * Note, however, that this depends on the order of the
2749 			 * extension headers in the ancillary data; the 1st
2750 			 * part of the destination options header must appear
2751 			 * before the routing header in the ancillary data,
2752 			 * too.
2753 			 * RFC3542 solved the ambiguity by introducing
2754 			 * separate ancillary data or option types.
2755 			 */
2756 			if (opt->ip6po_rthdr == NULL)
2757 				newdest = &opt->ip6po_dest1;
2758 			else
2759 				newdest = &opt->ip6po_dest2;
2760 			break;
2761 		case IPV6_RTHDRDSTOPTS:
2762 			newdest = &opt->ip6po_dest1;
2763 			break;
2764 		case IPV6_DSTOPTS:
2765 			newdest = &opt->ip6po_dest2;
2766 			break;
2767 		}
2768 
2769 		/* turn off the previous option, then set the new option. */
2770 		ip6_clearpktopts(opt, optname);
2771 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2772 		if (*newdest == NULL)
2773 			return (ENOBUFS);
2774 		bcopy(dest, *newdest, destlen);
2775 
2776 		break;
2777 	}
2778 
2779 	case IPV6_2292RTHDR:
2780 	case IPV6_RTHDR:
2781 	{
2782 		struct ip6_rthdr *rth;
2783 		int rthlen;
2784 
2785 		if (len == 0) {
2786 			ip6_clearpktopts(opt, IPV6_RTHDR);
2787 			break;	/* just remove the option */
2788 		}
2789 
2790 		/* message length validation */
2791 		if (len < sizeof(struct ip6_rthdr))
2792 			return (EINVAL);
2793 		rth = (struct ip6_rthdr *)buf;
2794 		rthlen = (rth->ip6r_len + 1) << 3;
2795 		if (len != rthlen)
2796 			return (EINVAL);
2797 
2798 		switch (rth->ip6r_type) {
2799 		case IPV6_RTHDR_TYPE_0:
2800 			if (rth->ip6r_len == 0)	/* must contain one addr */
2801 				return (EINVAL);
2802 			if (rth->ip6r_len % 2) /* length must be even */
2803 				return (EINVAL);
2804 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2805 				return (EINVAL);
2806 			break;
2807 		default:
2808 			return (EINVAL);	/* not supported */
2809 		}
2810 
2811 		/* turn off the previous option */
2812 		ip6_clearpktopts(opt, IPV6_RTHDR);
2813 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2814 		if (opt->ip6po_rthdr == NULL)
2815 			return (ENOBUFS);
2816 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2817 
2818 		break;
2819 	}
2820 
2821 	case IPV6_USE_MIN_MTU:
2822 		if (len != sizeof(int))
2823 			return (EINVAL);
2824 		minmtupolicy = *(int *)buf;
2825 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2826 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2827 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2828 			return (EINVAL);
2829 		}
2830 		opt->ip6po_minmtu = minmtupolicy;
2831 		break;
2832 
2833 	case IPV6_DONTFRAG:
2834 		if (len != sizeof(int))
2835 			return (EINVAL);
2836 
2837 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2838 			/*
2839 			 * we ignore this option for TCP sockets.
2840 			 * (RFC3542 leaves this case unspecified.)
2841 			 */
2842 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2843 		} else
2844 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2845 		break;
2846 
2847 	case IPV6_PREFER_TEMPADDR:
2848 		if (len != sizeof(int))
2849 			return (EINVAL);
2850 		preftemp = *(int *)buf;
2851 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2852 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2853 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2854 			return (EINVAL);
2855 		}
2856 		opt->ip6po_prefer_tempaddr = preftemp;
2857 		break;
2858 
2859 	default:
2860 		return (ENOPROTOOPT);
2861 	} /* end of switch */
2862 
2863 	return (0);
2864 }
2865 
2866 /*
2867  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2868  * packet to the input queue of a specified interface.  Note that this
2869  * calls the output routine of the loopback "driver", but with an interface
2870  * pointer that might NOT be &loif -- easier than replicating that code here.
2871  */
2872 void
2873 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2874 {
2875 	struct mbuf *copym;
2876 	struct ip6_hdr *ip6;
2877 
2878 	copym = m_copy(m, 0, M_COPYALL);
2879 	if (copym == NULL)
2880 		return;
2881 
2882 	/*
2883 	 * Make sure to deep-copy IPv6 header portion in case the data
2884 	 * is in an mbuf cluster, so that we can safely override the IPv6
2885 	 * header portion later.
2886 	 */
2887 	if (!M_WRITABLE(copym) ||
2888 	    copym->m_len < sizeof(struct ip6_hdr)) {
2889 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2890 		if (copym == NULL)
2891 			return;
2892 	}
2893 
2894 #ifdef DIAGNOSTIC
2895 	if (copym->m_len < sizeof(*ip6)) {
2896 		m_freem(copym);
2897 		return;
2898 	}
2899 #endif
2900 
2901 	ip6 = mtod(copym, struct ip6_hdr *);
2902 	/*
2903 	 * clear embedded scope identifiers if necessary.
2904 	 * in6_clearscope will touch the addresses only when necessary.
2905 	 */
2906 	in6_clearscope(&ip6->ip6_src);
2907 	in6_clearscope(&ip6->ip6_dst);
2908 
2909 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2910 }
2911 
2912 /*
2913  * Chop IPv6 header off from the payload.
2914  */
2915 static int
2916 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2917 {
2918 	struct mbuf *mh;
2919 	struct ip6_hdr *ip6;
2920 
2921 	ip6 = mtod(m, struct ip6_hdr *);
2922 	if (m->m_len > sizeof(*ip6)) {
2923 		mh = m_gethdr(M_NOWAIT, MT_DATA);
2924 		if (mh == NULL) {
2925 			m_freem(m);
2926 			return ENOBUFS;
2927 		}
2928 		m_move_pkthdr(mh, m);
2929 		M_ALIGN(mh, sizeof(*ip6));
2930 		m->m_len -= sizeof(*ip6);
2931 		m->m_data += sizeof(*ip6);
2932 		mh->m_next = m;
2933 		m = mh;
2934 		m->m_len = sizeof(*ip6);
2935 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2936 	}
2937 	exthdrs->ip6e_ip6 = m;
2938 	return 0;
2939 }
2940 
2941 /*
2942  * Compute IPv6 extension header length.
2943  */
2944 int
2945 ip6_optlen(struct inpcb *in6p)
2946 {
2947 	int len;
2948 
2949 	if (!in6p->in6p_outputopts)
2950 		return 0;
2951 
2952 	len = 0;
2953 #define elen(x) \
2954     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2955 
2956 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2957 	if (in6p->in6p_outputopts->ip6po_rthdr)
2958 		/* dest1 is valid with rthdr only */
2959 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2960 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2961 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2962 	return len;
2963 #undef elen
2964 }
2965