xref: /freebsd/sys/netinet6/ip6_output.c (revision 4b63a7c678a1291c8056de3770cfa393773b1f94)
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 		(inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID))) {
272 			m->m_pkthdr.flowid = inp->inp_flowid;
273 			m->m_flags |= M_FLOWID;
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 	 */
307 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp))
308 	{
309 	case 1:                 /* Bad packet */
310 		goto freehdrs;
311 	case -1:                /* IPSec done */
312 		goto done;
313 	case 0:                 /* No IPSec */
314 	default:
315 		break;
316 	}
317 #endif /* IPSEC */
318 
319 	/*
320 	 * Calculate the total length of the extension header chain.
321 	 * Keep the length of the unfragmentable part for fragmentation.
322 	 */
323 	optlen = 0;
324 	if (exthdrs.ip6e_hbh)
325 		optlen += exthdrs.ip6e_hbh->m_len;
326 	if (exthdrs.ip6e_dest1)
327 		optlen += exthdrs.ip6e_dest1->m_len;
328 	if (exthdrs.ip6e_rthdr)
329 		optlen += exthdrs.ip6e_rthdr->m_len;
330 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
331 
332 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
333 	if (exthdrs.ip6e_dest2)
334 		optlen += exthdrs.ip6e_dest2->m_len;
335 
336 	/*
337 	 * If there is at least one extension header,
338 	 * separate IP6 header from the payload.
339 	 */
340 	if (optlen && !hdrsplit) {
341 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
342 			m = NULL;
343 			goto freehdrs;
344 		}
345 		m = exthdrs.ip6e_ip6;
346 		hdrsplit++;
347 	}
348 
349 	/* adjust pointer */
350 	ip6 = mtod(m, struct ip6_hdr *);
351 
352 	/* adjust mbuf packet header length */
353 	m->m_pkthdr.len += optlen;
354 	plen = m->m_pkthdr.len - sizeof(*ip6);
355 
356 	/* If this is a jumbo payload, insert a jumbo payload option. */
357 	if (plen > IPV6_MAXPACKET) {
358 		if (!hdrsplit) {
359 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
360 				m = NULL;
361 				goto freehdrs;
362 			}
363 			m = exthdrs.ip6e_ip6;
364 			hdrsplit++;
365 		}
366 		/* adjust pointer */
367 		ip6 = mtod(m, struct ip6_hdr *);
368 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
369 			goto freehdrs;
370 		ip6->ip6_plen = 0;
371 	} else
372 		ip6->ip6_plen = htons(plen);
373 
374 	/*
375 	 * Concatenate headers and fill in next header fields.
376 	 * Here we have, on "m"
377 	 *	IPv6 payload
378 	 * and we insert headers accordingly.  Finally, we should be getting:
379 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
380 	 *
381 	 * during the header composing process, "m" points to IPv6 header.
382 	 * "mprev" points to an extension header prior to esp.
383 	 */
384 	u_char *nexthdrp = &ip6->ip6_nxt;
385 	mprev = m;
386 
387 	/*
388 	 * we treat dest2 specially.  this makes IPsec processing
389 	 * much easier.  the goal here is to make mprev point the
390 	 * mbuf prior to dest2.
391 	 *
392 	 * result: IPv6 dest2 payload
393 	 * m and mprev will point to IPv6 header.
394 	 */
395 	if (exthdrs.ip6e_dest2) {
396 		if (!hdrsplit)
397 			panic("assumption failed: hdr not split");
398 		exthdrs.ip6e_dest2->m_next = m->m_next;
399 		m->m_next = exthdrs.ip6e_dest2;
400 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
401 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
402 	}
403 
404 	/*
405 	 * result: IPv6 hbh dest1 rthdr dest2 payload
406 	 * m will point to IPv6 header.  mprev will point to the
407 	 * extension header prior to dest2 (rthdr in the above case).
408 	 */
409 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
410 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
411 		   IPPROTO_DSTOPTS);
412 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
413 		   IPPROTO_ROUTING);
414 
415 	/*
416 	 * If there is a routing header, discard the packet.
417 	 */
418 	if (exthdrs.ip6e_rthdr) {
419 		 error = EINVAL;
420 		 goto bad;
421 	}
422 
423 	/* Source address validation */
424 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
425 	    (flags & IPV6_UNSPECSRC) == 0) {
426 		error = EOPNOTSUPP;
427 		IP6STAT_INC(ip6s_badscope);
428 		goto bad;
429 	}
430 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
431 		error = EOPNOTSUPP;
432 		IP6STAT_INC(ip6s_badscope);
433 		goto bad;
434 	}
435 
436 	IP6STAT_INC(ip6s_localout);
437 
438 	/*
439 	 * Route packet.
440 	 */
441 	if (ro == 0) {
442 		ro = &ip6route;
443 		bzero((caddr_t)ro, sizeof(*ro));
444 	}
445 	ro_pmtu = ro;
446 	if (opt && opt->ip6po_rthdr)
447 		ro = &opt->ip6po_route;
448 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
449 #ifdef FLOWTABLE
450 	if (ro->ro_rt == NULL)
451 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
452 #endif
453 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
454 again:
455 	/*
456 	 * if specified, try to fill in the traffic class field.
457 	 * do not override if a non-zero value is already set.
458 	 * we check the diffserv field and the ecn field separately.
459 	 */
460 	if (opt && opt->ip6po_tclass >= 0) {
461 		int mask = 0;
462 
463 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
464 			mask |= 0xfc;
465 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
466 			mask |= 0x03;
467 		if (mask != 0)
468 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
469 	}
470 
471 	/* fill in or override the hop limit field, if necessary. */
472 	if (opt && opt->ip6po_hlim != -1)
473 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
474 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
475 		if (im6o != NULL)
476 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
477 		else
478 			ip6->ip6_hlim = V_ip6_defmcasthlim;
479 	}
480 
481 	/* adjust pointer */
482 	ip6 = mtod(m, struct ip6_hdr *);
483 
484 	if (ro->ro_rt && fwd_tag == NULL) {
485 		rt = ro->ro_rt;
486 		ifp = ro->ro_rt->rt_ifp;
487 	} else {
488 		if (fwd_tag == NULL) {
489 			bzero(&dst_sa, sizeof(dst_sa));
490 			dst_sa.sin6_family = AF_INET6;
491 			dst_sa.sin6_len = sizeof(dst_sa);
492 			dst_sa.sin6_addr = ip6->ip6_dst;
493 		}
494 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
495 		    &rt, fibnum);
496 		if (error != 0) {
497 			if (ifp != NULL)
498 				in6_ifstat_inc(ifp, ifs6_out_discard);
499 			goto bad;
500 		}
501 	}
502 	if (rt == NULL) {
503 		/*
504 		 * If in6_selectroute() does not return a route entry,
505 		 * dst may not have been updated.
506 		 */
507 		*dst = dst_sa;	/* XXX */
508 	}
509 
510 	/*
511 	 * then rt (for unicast) and ifp must be non-NULL valid values.
512 	 */
513 	if ((flags & IPV6_FORWARDING) == 0) {
514 		/* XXX: the FORWARDING flag can be set for mrouting. */
515 		in6_ifstat_inc(ifp, ifs6_out_request);
516 	}
517 	if (rt != NULL) {
518 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
519 		counter_u64_add(rt->rt_pksent, 1);
520 	}
521 
522 
523 	/*
524 	 * The outgoing interface must be in the zone of source and
525 	 * destination addresses.
526 	 */
527 	origifp = ifp;
528 
529 	src0 = ip6->ip6_src;
530 	if (in6_setscope(&src0, origifp, &zone))
531 		goto badscope;
532 	bzero(&src_sa, sizeof(src_sa));
533 	src_sa.sin6_family = AF_INET6;
534 	src_sa.sin6_len = sizeof(src_sa);
535 	src_sa.sin6_addr = ip6->ip6_src;
536 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
537 		goto badscope;
538 
539 	dst0 = ip6->ip6_dst;
540 	if (in6_setscope(&dst0, origifp, &zone))
541 		goto badscope;
542 	/* re-initialize to be sure */
543 	bzero(&dst_sa, sizeof(dst_sa));
544 	dst_sa.sin6_family = AF_INET6;
545 	dst_sa.sin6_len = sizeof(dst_sa);
546 	dst_sa.sin6_addr = ip6->ip6_dst;
547 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
548 		goto badscope;
549 	}
550 
551 	/* We should use ia_ifp to support the case of
552 	 * sending packets to an address of our own.
553 	 */
554 	if (ia != NULL && ia->ia_ifp)
555 		ifp = ia->ia_ifp;
556 
557 	/* scope check is done. */
558 	goto routefound;
559 
560   badscope:
561 	IP6STAT_INC(ip6s_badscope);
562 	in6_ifstat_inc(origifp, ifs6_out_discard);
563 	if (error == 0)
564 		error = EHOSTUNREACH; /* XXX */
565 	goto bad;
566 
567   routefound:
568 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
569 		if (opt && opt->ip6po_nextroute.ro_rt) {
570 			/*
571 			 * The nexthop is explicitly specified by the
572 			 * application.  We assume the next hop is an IPv6
573 			 * address.
574 			 */
575 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
576 		}
577 		else if ((rt->rt_flags & RTF_GATEWAY))
578 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
579 	}
580 
581 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
582 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
583 	} else {
584 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
585 		in6_ifstat_inc(ifp, ifs6_out_mcast);
586 		/*
587 		 * Confirm that the outgoing interface supports multicast.
588 		 */
589 		if (!(ifp->if_flags & IFF_MULTICAST)) {
590 			IP6STAT_INC(ip6s_noroute);
591 			in6_ifstat_inc(ifp, ifs6_out_discard);
592 			error = ENETUNREACH;
593 			goto bad;
594 		}
595 		if ((im6o == NULL && in6_mcast_loop) ||
596 		    (im6o && im6o->im6o_multicast_loop)) {
597 			/*
598 			 * Loop back multicast datagram if not expressly
599 			 * forbidden to do so, even if we have not joined
600 			 * the address; protocols will filter it later,
601 			 * thus deferring a hash lookup and lock acquisition
602 			 * at the expense of an m_copym().
603 			 */
604 			ip6_mloopback(ifp, m, dst);
605 		} else {
606 			/*
607 			 * If we are acting as a multicast router, perform
608 			 * multicast forwarding as if the packet had just
609 			 * arrived on the interface to which we are about
610 			 * to send.  The multicast forwarding function
611 			 * recursively calls this function, using the
612 			 * IPV6_FORWARDING flag to prevent infinite recursion.
613 			 *
614 			 * Multicasts that are looped back by ip6_mloopback(),
615 			 * above, will be forwarded by the ip6_input() routine,
616 			 * if necessary.
617 			 */
618 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
619 				/*
620 				 * XXX: ip6_mforward expects that rcvif is NULL
621 				 * when it is called from the originating path.
622 				 * However, it may not always be the case.
623 				 */
624 				m->m_pkthdr.rcvif = NULL;
625 				if (ip6_mforward(ip6, ifp, m) != 0) {
626 					m_freem(m);
627 					goto done;
628 				}
629 			}
630 		}
631 		/*
632 		 * Multicasts with a hoplimit of zero may be looped back,
633 		 * above, but must not be transmitted on a network.
634 		 * Also, multicasts addressed to the loopback interface
635 		 * are not sent -- the above call to ip6_mloopback() will
636 		 * loop back a copy if this host actually belongs to the
637 		 * destination group on the loopback interface.
638 		 */
639 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
640 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
641 			m_freem(m);
642 			goto done;
643 		}
644 	}
645 
646 	/*
647 	 * Fill the outgoing inteface to tell the upper layer
648 	 * to increment per-interface statistics.
649 	 */
650 	if (ifpp)
651 		*ifpp = ifp;
652 
653 	/* Determine path MTU. */
654 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
655 	    &alwaysfrag, fibnum)) != 0)
656 		goto bad;
657 
658 	/*
659 	 * The caller of this function may specify to use the minimum MTU
660 	 * in some cases.
661 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
662 	 * setting.  The logic is a bit complicated; by default, unicast
663 	 * packets will follow path MTU while multicast packets will be sent at
664 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
665 	 * including unicast ones will be sent at the minimum MTU.  Multicast
666 	 * packets will always be sent at the minimum MTU unless
667 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
668 	 * See RFC 3542 for more details.
669 	 */
670 	if (mtu > IPV6_MMTU) {
671 		if ((flags & IPV6_MINMTU))
672 			mtu = IPV6_MMTU;
673 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
674 			mtu = IPV6_MMTU;
675 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
676 			 (opt == NULL ||
677 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
678 			mtu = IPV6_MMTU;
679 		}
680 	}
681 
682 	/*
683 	 * clear embedded scope identifiers if necessary.
684 	 * in6_clearscope will touch the addresses only when necessary.
685 	 */
686 	in6_clearscope(&ip6->ip6_src);
687 	in6_clearscope(&ip6->ip6_dst);
688 
689 	/*
690 	 * If the outgoing packet contains a hop-by-hop options header,
691 	 * it must be examined and processed even by the source node.
692 	 * (RFC 2460, section 4.)
693 	 */
694 	if (exthdrs.ip6e_hbh) {
695 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
696 		u_int32_t dummy; /* XXX unused */
697 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
698 
699 #ifdef DIAGNOSTIC
700 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
701 			panic("ip6e_hbh is not contiguous");
702 #endif
703 		/*
704 		 *  XXX: if we have to send an ICMPv6 error to the sender,
705 		 *       we need the M_LOOP flag since icmp6_error() expects
706 		 *       the IPv6 and the hop-by-hop options header are
707 		 *       contiguous unless the flag is set.
708 		 */
709 		m->m_flags |= M_LOOP;
710 		m->m_pkthdr.rcvif = ifp;
711 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
712 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
713 		    &dummy, &plen) < 0) {
714 			/* m was already freed at this point */
715 			error = EINVAL;/* better error? */
716 			goto done;
717 		}
718 		m->m_flags &= ~M_LOOP; /* XXX */
719 		m->m_pkthdr.rcvif = NULL;
720 	}
721 
722 	/* Jump over all PFIL processing if hooks are not active. */
723 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
724 		goto passout;
725 
726 	odst = ip6->ip6_dst;
727 	/* Run through list of hooks for output packets. */
728 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
729 	if (error != 0 || m == NULL)
730 		goto done;
731 	ip6 = mtod(m, struct ip6_hdr *);
732 
733 	needfiblookup = 0;
734 	/* See if destination IP address was changed by packet filter. */
735 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
736 		m->m_flags |= M_SKIP_FIREWALL;
737 		/* If destination is now ourself drop to ip6_input(). */
738 		if (in6_localip(&ip6->ip6_dst)) {
739 			m->m_flags |= M_FASTFWD_OURS;
740 			if (m->m_pkthdr.rcvif == NULL)
741 				m->m_pkthdr.rcvif = V_loif;
742 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
743 				m->m_pkthdr.csum_flags |=
744 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
745 				m->m_pkthdr.csum_data = 0xffff;
746 			}
747 #ifdef SCTP
748 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
749 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
750 #endif
751 			error = netisr_queue(NETISR_IPV6, m);
752 			goto done;
753 		} else
754 			needfiblookup = 1; /* Redo the routing table lookup. */
755 	}
756 	/* See if fib was changed by packet filter. */
757 	if (fibnum != M_GETFIB(m)) {
758 		m->m_flags |= M_SKIP_FIREWALL;
759 		fibnum = M_GETFIB(m);
760 		RO_RTFREE(ro);
761 		needfiblookup = 1;
762 	}
763 	if (needfiblookup)
764 		goto again;
765 
766 	/* See if local, if yes, send it to netisr. */
767 	if (m->m_flags & M_FASTFWD_OURS) {
768 		if (m->m_pkthdr.rcvif == NULL)
769 			m->m_pkthdr.rcvif = V_loif;
770 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
771 			m->m_pkthdr.csum_flags |=
772 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
773 			m->m_pkthdr.csum_data = 0xffff;
774 		}
775 #ifdef SCTP
776 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
777 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
778 #endif
779 		error = netisr_queue(NETISR_IPV6, m);
780 		goto done;
781 	}
782 	/* Or forward to some other address? */
783 	if ((m->m_flags & M_IP6_NEXTHOP) &&
784 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
785 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
786 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
787 		m->m_flags |= M_SKIP_FIREWALL;
788 		m->m_flags &= ~M_IP6_NEXTHOP;
789 		m_tag_delete(m, fwd_tag);
790 		goto again;
791 	}
792 
793 passout:
794 	/*
795 	 * Send the packet to the outgoing interface.
796 	 * If necessary, do IPv6 fragmentation before sending.
797 	 *
798 	 * the logic here is rather complex:
799 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
800 	 * 1-a:	send as is if tlen <= path mtu
801 	 * 1-b:	fragment if tlen > path mtu
802 	 *
803 	 * 2: if user asks us not to fragment (dontfrag == 1)
804 	 * 2-a:	send as is if tlen <= interface mtu
805 	 * 2-b:	error if tlen > interface mtu
806 	 *
807 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
808 	 *	always fragment
809 	 *
810 	 * 4: if dontfrag == 1 && alwaysfrag == 1
811 	 *	error, as we cannot handle this conflicting request
812 	 */
813 	sw_csum = m->m_pkthdr.csum_flags;
814 	if (!hdrsplit) {
815 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
816 		sw_csum &= ~ifp->if_hwassist;
817 	} else
818 		tso = 0;
819 	/*
820 	 * If we added extension headers, we will not do TSO and calculate the
821 	 * checksums ourselves for now.
822 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
823 	 * with ext. hdrs.
824 	 */
825 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
826 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
827 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
828 	}
829 #ifdef SCTP
830 	if (sw_csum & CSUM_SCTP_IPV6) {
831 		sw_csum &= ~CSUM_SCTP_IPV6;
832 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
833 	}
834 #endif
835 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
836 	tlen = m->m_pkthdr.len;
837 
838 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
839 		dontfrag = 1;
840 	else
841 		dontfrag = 0;
842 	if (dontfrag && alwaysfrag) {	/* case 4 */
843 		/* conflicting request - can't transmit */
844 		error = EMSGSIZE;
845 		goto bad;
846 	}
847 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
848 		/*
849 		 * Even if the DONTFRAG option is specified, we cannot send the
850 		 * packet when the data length is larger than the MTU of the
851 		 * outgoing interface.
852 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
853 		 * well as returning an error code (the latter is not described
854 		 * in the API spec.)
855 		 */
856 		u_int32_t mtu32;
857 		struct ip6ctlparam ip6cp;
858 
859 		mtu32 = (u_int32_t)mtu;
860 		bzero(&ip6cp, sizeof(ip6cp));
861 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
862 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
863 		    (void *)&ip6cp);
864 
865 		error = EMSGSIZE;
866 		goto bad;
867 	}
868 
869 	/*
870 	 * transmit packet without fragmentation
871 	 */
872 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
873 		struct in6_ifaddr *ia6;
874 
875 		ip6 = mtod(m, struct ip6_hdr *);
876 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
877 		if (ia6) {
878 			/* Record statistics for this interface address. */
879 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
880 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
881 			    m->m_pkthdr.len);
882 			ifa_free(&ia6->ia_ifa);
883 		}
884 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
885 		goto done;
886 	}
887 
888 	/*
889 	 * try to fragment the packet.  case 1-b and 3
890 	 */
891 	if (mtu < IPV6_MMTU) {
892 		/* path MTU cannot be less than IPV6_MMTU */
893 		error = EMSGSIZE;
894 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
895 		goto bad;
896 	} else if (ip6->ip6_plen == 0) {
897 		/* jumbo payload cannot be fragmented */
898 		error = EMSGSIZE;
899 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
900 		goto bad;
901 	} else {
902 		struct mbuf **mnext, *m_frgpart;
903 		struct ip6_frag *ip6f;
904 		u_int32_t id = htonl(ip6_randomid());
905 		u_char nextproto;
906 
907 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
908 
909 		/*
910 		 * Too large for the destination or interface;
911 		 * fragment if possible.
912 		 * Must be able to put at least 8 bytes per fragment.
913 		 */
914 		hlen = unfragpartlen;
915 		if (mtu > IPV6_MAXPACKET)
916 			mtu = IPV6_MAXPACKET;
917 
918 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
919 		if (len < 8) {
920 			error = EMSGSIZE;
921 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
922 			goto bad;
923 		}
924 
925 		/*
926 		 * Verify that we have any chance at all of being able to queue
927 		 *      the packet or packet fragments
928 		 */
929 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
930 		    < tlen  /* - hlen */)) {
931 			error = ENOBUFS;
932 			IP6STAT_INC(ip6s_odropped);
933 			goto bad;
934 		}
935 
936 
937 		/*
938 		 * If the interface will not calculate checksums on
939 		 * fragmented packets, then do it here.
940 		 * XXX-BZ handle the hw offloading case.  Need flags.
941 		 */
942 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
943 			in6_delayed_cksum(m, plen, hlen);
944 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
945 		}
946 #ifdef SCTP
947 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
948 			sctp_delayed_cksum(m, hlen);
949 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
950 		}
951 #endif
952 		mnext = &m->m_nextpkt;
953 
954 		/*
955 		 * Change the next header field of the last header in the
956 		 * unfragmentable part.
957 		 */
958 		if (exthdrs.ip6e_rthdr) {
959 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
960 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
961 		} else if (exthdrs.ip6e_dest1) {
962 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
963 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
964 		} else if (exthdrs.ip6e_hbh) {
965 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
966 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
967 		} else {
968 			nextproto = ip6->ip6_nxt;
969 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
970 		}
971 
972 		/*
973 		 * Loop through length of segment after first fragment,
974 		 * make new header and copy data of each part and link onto
975 		 * chain.
976 		 */
977 		m0 = m;
978 		for (off = hlen; off < tlen; off += len) {
979 			m = m_gethdr(M_NOWAIT, MT_DATA);
980 			if (!m) {
981 				error = ENOBUFS;
982 				IP6STAT_INC(ip6s_odropped);
983 				goto sendorfree;
984 			}
985 			m->m_flags = m0->m_flags & M_COPYFLAGS;
986 			*mnext = m;
987 			mnext = &m->m_nextpkt;
988 			m->m_data += max_linkhdr;
989 			mhip6 = mtod(m, struct ip6_hdr *);
990 			*mhip6 = *ip6;
991 			m->m_len = sizeof(*mhip6);
992 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
993 			if (error) {
994 				IP6STAT_INC(ip6s_odropped);
995 				goto sendorfree;
996 			}
997 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
998 			if (off + len >= tlen)
999 				len = tlen - off;
1000 			else
1001 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1002 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1003 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1004 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1005 				error = ENOBUFS;
1006 				IP6STAT_INC(ip6s_odropped);
1007 				goto sendorfree;
1008 			}
1009 			m_cat(m, m_frgpart);
1010 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1011 			m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
1012 			m->m_pkthdr.rcvif = NULL;
1013 			ip6f->ip6f_reserved = 0;
1014 			ip6f->ip6f_ident = id;
1015 			ip6f->ip6f_nxt = nextproto;
1016 			IP6STAT_INC(ip6s_ofragments);
1017 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1018 		}
1019 
1020 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1021 	}
1022 
1023 	/*
1024 	 * Remove leading garbages.
1025 	 */
1026 sendorfree:
1027 	m = m0->m_nextpkt;
1028 	m0->m_nextpkt = 0;
1029 	m_freem(m0);
1030 	for (m0 = m; m; m = m0) {
1031 		m0 = m->m_nextpkt;
1032 		m->m_nextpkt = 0;
1033 		if (error == 0) {
1034 			/* Record statistics for this interface address. */
1035 			if (ia) {
1036 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1037 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1038 				    m->m_pkthdr.len);
1039 			}
1040 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1041 		} else
1042 			m_freem(m);
1043 	}
1044 
1045 	if (error == 0)
1046 		IP6STAT_INC(ip6s_fragmented);
1047 
1048 done:
1049 	if (ro == &ip6route)
1050 		RO_RTFREE(ro);
1051 	if (ro_pmtu == &ip6route)
1052 		RO_RTFREE(ro_pmtu);
1053 	return (error);
1054 
1055 freehdrs:
1056 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1057 	m_freem(exthdrs.ip6e_dest1);
1058 	m_freem(exthdrs.ip6e_rthdr);
1059 	m_freem(exthdrs.ip6e_dest2);
1060 	/* FALLTHROUGH */
1061 bad:
1062 	if (m)
1063 		m_freem(m);
1064 	goto done;
1065 }
1066 
1067 static int
1068 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1069 {
1070 	struct mbuf *m;
1071 
1072 	if (hlen > MCLBYTES)
1073 		return (ENOBUFS); /* XXX */
1074 
1075 	if (hlen > MLEN)
1076 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1077 	else
1078 		m = m_get(M_NOWAIT, MT_DATA);
1079 	if (m == NULL)
1080 		return (ENOBUFS);
1081 	m->m_len = hlen;
1082 	if (hdr)
1083 		bcopy(hdr, mtod(m, caddr_t), hlen);
1084 
1085 	*mp = m;
1086 	return (0);
1087 }
1088 
1089 /*
1090  * Insert jumbo payload option.
1091  */
1092 static int
1093 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1094 {
1095 	struct mbuf *mopt;
1096 	u_char *optbuf;
1097 	u_int32_t v;
1098 
1099 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1100 
1101 	/*
1102 	 * If there is no hop-by-hop options header, allocate new one.
1103 	 * If there is one but it doesn't have enough space to store the
1104 	 * jumbo payload option, allocate a cluster to store the whole options.
1105 	 * Otherwise, use it to store the options.
1106 	 */
1107 	if (exthdrs->ip6e_hbh == 0) {
1108 		mopt = m_get(M_NOWAIT, MT_DATA);
1109 		if (mopt == NULL)
1110 			return (ENOBUFS);
1111 		mopt->m_len = JUMBOOPTLEN;
1112 		optbuf = mtod(mopt, u_char *);
1113 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1114 		exthdrs->ip6e_hbh = mopt;
1115 	} else {
1116 		struct ip6_hbh *hbh;
1117 
1118 		mopt = exthdrs->ip6e_hbh;
1119 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1120 			/*
1121 			 * XXX assumption:
1122 			 * - exthdrs->ip6e_hbh is not referenced from places
1123 			 *   other than exthdrs.
1124 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1125 			 */
1126 			int oldoptlen = mopt->m_len;
1127 			struct mbuf *n;
1128 
1129 			/*
1130 			 * XXX: give up if the whole (new) hbh header does
1131 			 * not fit even in an mbuf cluster.
1132 			 */
1133 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1134 				return (ENOBUFS);
1135 
1136 			/*
1137 			 * As a consequence, we must always prepare a cluster
1138 			 * at this point.
1139 			 */
1140 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1141 			if (n == NULL)
1142 				return (ENOBUFS);
1143 			n->m_len = oldoptlen + JUMBOOPTLEN;
1144 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1145 			    oldoptlen);
1146 			optbuf = mtod(n, caddr_t) + oldoptlen;
1147 			m_freem(mopt);
1148 			mopt = exthdrs->ip6e_hbh = n;
1149 		} else {
1150 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1151 			mopt->m_len += JUMBOOPTLEN;
1152 		}
1153 		optbuf[0] = IP6OPT_PADN;
1154 		optbuf[1] = 1;
1155 
1156 		/*
1157 		 * Adjust the header length according to the pad and
1158 		 * the jumbo payload option.
1159 		 */
1160 		hbh = mtod(mopt, struct ip6_hbh *);
1161 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1162 	}
1163 
1164 	/* fill in the option. */
1165 	optbuf[2] = IP6OPT_JUMBO;
1166 	optbuf[3] = 4;
1167 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1168 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1169 
1170 	/* finally, adjust the packet header length */
1171 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1172 
1173 	return (0);
1174 #undef JUMBOOPTLEN
1175 }
1176 
1177 /*
1178  * Insert fragment header and copy unfragmentable header portions.
1179  */
1180 static int
1181 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1182     struct ip6_frag **frghdrp)
1183 {
1184 	struct mbuf *n, *mlast;
1185 
1186 	if (hlen > sizeof(struct ip6_hdr)) {
1187 		n = m_copym(m0, sizeof(struct ip6_hdr),
1188 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1189 		if (n == 0)
1190 			return (ENOBUFS);
1191 		m->m_next = n;
1192 	} else
1193 		n = m;
1194 
1195 	/* Search for the last mbuf of unfragmentable part. */
1196 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1197 		;
1198 
1199 	if (M_WRITABLE(mlast) &&
1200 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1201 		/* use the trailing space of the last mbuf for the fragment hdr */
1202 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1203 		    mlast->m_len);
1204 		mlast->m_len += sizeof(struct ip6_frag);
1205 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1206 	} else {
1207 		/* allocate a new mbuf for the fragment header */
1208 		struct mbuf *mfrg;
1209 
1210 		mfrg = m_get(M_NOWAIT, MT_DATA);
1211 		if (mfrg == NULL)
1212 			return (ENOBUFS);
1213 		mfrg->m_len = sizeof(struct ip6_frag);
1214 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1215 		mlast->m_next = mfrg;
1216 	}
1217 
1218 	return (0);
1219 }
1220 
1221 static int
1222 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1223     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1224     int *alwaysfragp, u_int fibnum)
1225 {
1226 	u_int32_t mtu = 0;
1227 	int alwaysfrag = 0;
1228 	int error = 0;
1229 
1230 	if (ro_pmtu != ro) {
1231 		/* The first hop and the final destination may differ. */
1232 		struct sockaddr_in6 *sa6_dst =
1233 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1234 		if (ro_pmtu->ro_rt &&
1235 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1236 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1237 			RTFREE(ro_pmtu->ro_rt);
1238 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1239 		}
1240 		if (ro_pmtu->ro_rt == NULL) {
1241 			bzero(sa6_dst, sizeof(*sa6_dst));
1242 			sa6_dst->sin6_family = AF_INET6;
1243 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1244 			sa6_dst->sin6_addr = *dst;
1245 
1246 			in6_rtalloc(ro_pmtu, fibnum);
1247 		}
1248 	}
1249 	if (ro_pmtu->ro_rt) {
1250 		u_int32_t ifmtu;
1251 		struct in_conninfo inc;
1252 
1253 		bzero(&inc, sizeof(inc));
1254 		inc.inc_flags |= INC_ISIPV6;
1255 		inc.inc6_faddr = *dst;
1256 
1257 		if (ifp == NULL)
1258 			ifp = ro_pmtu->ro_rt->rt_ifp;
1259 		ifmtu = IN6_LINKMTU(ifp);
1260 		mtu = tcp_hc_getmtu(&inc);
1261 		if (mtu)
1262 			mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1263 		else
1264 			mtu = ro_pmtu->ro_rt->rt_mtu;
1265 		if (mtu == 0)
1266 			mtu = ifmtu;
1267 		else if (mtu < IPV6_MMTU) {
1268 			/*
1269 			 * RFC2460 section 5, last paragraph:
1270 			 * if we record ICMPv6 too big message with
1271 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1272 			 * or smaller, with framgent header attached.
1273 			 * (fragment header is needed regardless from the
1274 			 * packet size, for translators to identify packets)
1275 			 */
1276 			alwaysfrag = 1;
1277 			mtu = IPV6_MMTU;
1278 		} else if (mtu > ifmtu) {
1279 			/*
1280 			 * The MTU on the route is larger than the MTU on
1281 			 * the interface!  This shouldn't happen, unless the
1282 			 * MTU of the interface has been changed after the
1283 			 * interface was brought up.  Change the MTU in the
1284 			 * route to match the interface MTU (as long as the
1285 			 * field isn't locked).
1286 			 */
1287 			mtu = ifmtu;
1288 			ro_pmtu->ro_rt->rt_mtu = mtu;
1289 		}
1290 	} else if (ifp) {
1291 		mtu = IN6_LINKMTU(ifp);
1292 	} else
1293 		error = EHOSTUNREACH; /* XXX */
1294 
1295 	*mtup = mtu;
1296 	if (alwaysfragp)
1297 		*alwaysfragp = alwaysfrag;
1298 	return (error);
1299 }
1300 
1301 /*
1302  * IP6 socket option processing.
1303  */
1304 int
1305 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1306 {
1307 	int optdatalen, uproto;
1308 	void *optdata;
1309 	struct inpcb *in6p = sotoinpcb(so);
1310 	int error, optval;
1311 	int level, op, optname;
1312 	int optlen;
1313 	struct thread *td;
1314 #ifdef	RSS
1315 	uint32_t rss_bucket;
1316 	int retval;
1317 #endif
1318 
1319 	level = sopt->sopt_level;
1320 	op = sopt->sopt_dir;
1321 	optname = sopt->sopt_name;
1322 	optlen = sopt->sopt_valsize;
1323 	td = sopt->sopt_td;
1324 	error = 0;
1325 	optval = 0;
1326 	uproto = (int)so->so_proto->pr_protocol;
1327 
1328 	if (level != IPPROTO_IPV6) {
1329 		error = EINVAL;
1330 
1331 		if (sopt->sopt_level == SOL_SOCKET &&
1332 		    sopt->sopt_dir == SOPT_SET) {
1333 			switch (sopt->sopt_name) {
1334 			case SO_REUSEADDR:
1335 				INP_WLOCK(in6p);
1336 				if ((so->so_options & SO_REUSEADDR) != 0)
1337 					in6p->inp_flags2 |= INP_REUSEADDR;
1338 				else
1339 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1340 				INP_WUNLOCK(in6p);
1341 				error = 0;
1342 				break;
1343 			case SO_REUSEPORT:
1344 				INP_WLOCK(in6p);
1345 				if ((so->so_options & SO_REUSEPORT) != 0)
1346 					in6p->inp_flags2 |= INP_REUSEPORT;
1347 				else
1348 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1349 				INP_WUNLOCK(in6p);
1350 				error = 0;
1351 				break;
1352 			case SO_SETFIB:
1353 				INP_WLOCK(in6p);
1354 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1355 				INP_WUNLOCK(in6p);
1356 				error = 0;
1357 				break;
1358 			default:
1359 				break;
1360 			}
1361 		}
1362 	} else {		/* level == IPPROTO_IPV6 */
1363 		switch (op) {
1364 
1365 		case SOPT_SET:
1366 			switch (optname) {
1367 			case IPV6_2292PKTOPTIONS:
1368 #ifdef IPV6_PKTOPTIONS
1369 			case IPV6_PKTOPTIONS:
1370 #endif
1371 			{
1372 				struct mbuf *m;
1373 
1374 				error = soopt_getm(sopt, &m); /* XXX */
1375 				if (error != 0)
1376 					break;
1377 				error = soopt_mcopyin(sopt, m); /* XXX */
1378 				if (error != 0)
1379 					break;
1380 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1381 						    m, so, sopt);
1382 				m_freem(m); /* XXX */
1383 				break;
1384 			}
1385 
1386 			/*
1387 			 * Use of some Hop-by-Hop options or some
1388 			 * Destination options, might require special
1389 			 * privilege.  That is, normal applications
1390 			 * (without special privilege) might be forbidden
1391 			 * from setting certain options in outgoing packets,
1392 			 * and might never see certain options in received
1393 			 * packets. [RFC 2292 Section 6]
1394 			 * KAME specific note:
1395 			 *  KAME prevents non-privileged users from sending or
1396 			 *  receiving ANY hbh/dst options in order to avoid
1397 			 *  overhead of parsing options in the kernel.
1398 			 */
1399 			case IPV6_RECVHOPOPTS:
1400 			case IPV6_RECVDSTOPTS:
1401 			case IPV6_RECVRTHDRDSTOPTS:
1402 				if (td != NULL) {
1403 					error = priv_check(td,
1404 					    PRIV_NETINET_SETHDROPTS);
1405 					if (error)
1406 						break;
1407 				}
1408 				/* FALLTHROUGH */
1409 			case IPV6_UNICAST_HOPS:
1410 			case IPV6_HOPLIMIT:
1411 
1412 			case IPV6_RECVPKTINFO:
1413 			case IPV6_RECVHOPLIMIT:
1414 			case IPV6_RECVRTHDR:
1415 			case IPV6_RECVPATHMTU:
1416 			case IPV6_RECVTCLASS:
1417 			case IPV6_V6ONLY:
1418 			case IPV6_AUTOFLOWLABEL:
1419 			case IPV6_BINDANY:
1420 			case IPV6_BINDMULTI:
1421 #ifdef	RSS
1422 			case IPV6_RSS_LISTEN_BUCKET:
1423 #endif
1424 				if (optname == IPV6_BINDANY && td != NULL) {
1425 					error = priv_check(td,
1426 					    PRIV_NETINET_BINDANY);
1427 					if (error)
1428 						break;
1429 				}
1430 
1431 				if (optlen != sizeof(int)) {
1432 					error = EINVAL;
1433 					break;
1434 				}
1435 				error = sooptcopyin(sopt, &optval,
1436 					sizeof optval, sizeof optval);
1437 				if (error)
1438 					break;
1439 				switch (optname) {
1440 
1441 				case IPV6_UNICAST_HOPS:
1442 					if (optval < -1 || optval >= 256)
1443 						error = EINVAL;
1444 					else {
1445 						/* -1 = kernel default */
1446 						in6p->in6p_hops = optval;
1447 						if ((in6p->inp_vflag &
1448 						     INP_IPV4) != 0)
1449 							in6p->inp_ip_ttl = optval;
1450 					}
1451 					break;
1452 #define OPTSET(bit) \
1453 do { \
1454 	INP_WLOCK(in6p); \
1455 	if (optval) \
1456 		in6p->inp_flags |= (bit); \
1457 	else \
1458 		in6p->inp_flags &= ~(bit); \
1459 	INP_WUNLOCK(in6p); \
1460 } while (/*CONSTCOND*/ 0)
1461 #define OPTSET2292(bit) \
1462 do { \
1463 	INP_WLOCK(in6p); \
1464 	in6p->inp_flags |= IN6P_RFC2292; \
1465 	if (optval) \
1466 		in6p->inp_flags |= (bit); \
1467 	else \
1468 		in6p->inp_flags &= ~(bit); \
1469 	INP_WUNLOCK(in6p); \
1470 } while (/*CONSTCOND*/ 0)
1471 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1472 
1473 #define OPTSET2(bit, val) do {						\
1474 	INP_WLOCK(in6p);						\
1475 	if (val)							\
1476 		in6p->inp_flags2 |= bit;				\
1477 	else								\
1478 		in6p->inp_flags2 &= ~bit;				\
1479 	INP_WUNLOCK(in6p);						\
1480 } while (0)
1481 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1482 
1483 				case IPV6_RECVPKTINFO:
1484 					/* cannot mix with RFC2292 */
1485 					if (OPTBIT(IN6P_RFC2292)) {
1486 						error = EINVAL;
1487 						break;
1488 					}
1489 					OPTSET(IN6P_PKTINFO);
1490 					break;
1491 
1492 				case IPV6_HOPLIMIT:
1493 				{
1494 					struct ip6_pktopts **optp;
1495 
1496 					/* cannot mix with RFC2292 */
1497 					if (OPTBIT(IN6P_RFC2292)) {
1498 						error = EINVAL;
1499 						break;
1500 					}
1501 					optp = &in6p->in6p_outputopts;
1502 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1503 					    (u_char *)&optval, sizeof(optval),
1504 					    optp, (td != NULL) ? td->td_ucred :
1505 					    NULL, uproto);
1506 					break;
1507 				}
1508 
1509 				case IPV6_RECVHOPLIMIT:
1510 					/* cannot mix with RFC2292 */
1511 					if (OPTBIT(IN6P_RFC2292)) {
1512 						error = EINVAL;
1513 						break;
1514 					}
1515 					OPTSET(IN6P_HOPLIMIT);
1516 					break;
1517 
1518 				case IPV6_RECVHOPOPTS:
1519 					/* cannot mix with RFC2292 */
1520 					if (OPTBIT(IN6P_RFC2292)) {
1521 						error = EINVAL;
1522 						break;
1523 					}
1524 					OPTSET(IN6P_HOPOPTS);
1525 					break;
1526 
1527 				case IPV6_RECVDSTOPTS:
1528 					/* cannot mix with RFC2292 */
1529 					if (OPTBIT(IN6P_RFC2292)) {
1530 						error = EINVAL;
1531 						break;
1532 					}
1533 					OPTSET(IN6P_DSTOPTS);
1534 					break;
1535 
1536 				case IPV6_RECVRTHDRDSTOPTS:
1537 					/* cannot mix with RFC2292 */
1538 					if (OPTBIT(IN6P_RFC2292)) {
1539 						error = EINVAL;
1540 						break;
1541 					}
1542 					OPTSET(IN6P_RTHDRDSTOPTS);
1543 					break;
1544 
1545 				case IPV6_RECVRTHDR:
1546 					/* cannot mix with RFC2292 */
1547 					if (OPTBIT(IN6P_RFC2292)) {
1548 						error = EINVAL;
1549 						break;
1550 					}
1551 					OPTSET(IN6P_RTHDR);
1552 					break;
1553 
1554 				case IPV6_RECVPATHMTU:
1555 					/*
1556 					 * We ignore this option for TCP
1557 					 * sockets.
1558 					 * (RFC3542 leaves this case
1559 					 * unspecified.)
1560 					 */
1561 					if (uproto != IPPROTO_TCP)
1562 						OPTSET(IN6P_MTU);
1563 					break;
1564 
1565 				case IPV6_V6ONLY:
1566 					/*
1567 					 * make setsockopt(IPV6_V6ONLY)
1568 					 * available only prior to bind(2).
1569 					 * see ipng mailing list, Jun 22 2001.
1570 					 */
1571 					if (in6p->inp_lport ||
1572 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1573 						error = EINVAL;
1574 						break;
1575 					}
1576 					OPTSET(IN6P_IPV6_V6ONLY);
1577 					if (optval)
1578 						in6p->inp_vflag &= ~INP_IPV4;
1579 					else
1580 						in6p->inp_vflag |= INP_IPV4;
1581 					break;
1582 				case IPV6_RECVTCLASS:
1583 					/* cannot mix with RFC2292 XXX */
1584 					if (OPTBIT(IN6P_RFC2292)) {
1585 						error = EINVAL;
1586 						break;
1587 					}
1588 					OPTSET(IN6P_TCLASS);
1589 					break;
1590 				case IPV6_AUTOFLOWLABEL:
1591 					OPTSET(IN6P_AUTOFLOWLABEL);
1592 					break;
1593 
1594 				case IPV6_BINDANY:
1595 					OPTSET(INP_BINDANY);
1596 					break;
1597 
1598 				case IPV6_BINDMULTI:
1599 					OPTSET2(INP_BINDMULTI, optval);
1600 					break;
1601 #ifdef	RSS
1602 				case IPV6_RSS_LISTEN_BUCKET:
1603 					if ((optval >= 0) &&
1604 					    (optval < rss_getnumbuckets())) {
1605 						in6p->inp_rss_listen_bucket = optval;
1606 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1607 					} else {
1608 						error = EINVAL;
1609 					}
1610 					break;
1611 #endif
1612 				}
1613 				break;
1614 
1615 			case IPV6_TCLASS:
1616 			case IPV6_DONTFRAG:
1617 			case IPV6_USE_MIN_MTU:
1618 			case IPV6_PREFER_TEMPADDR:
1619 				if (optlen != sizeof(optval)) {
1620 					error = EINVAL;
1621 					break;
1622 				}
1623 				error = sooptcopyin(sopt, &optval,
1624 					sizeof optval, sizeof optval);
1625 				if (error)
1626 					break;
1627 				{
1628 					struct ip6_pktopts **optp;
1629 					optp = &in6p->in6p_outputopts;
1630 					error = ip6_pcbopt(optname,
1631 					    (u_char *)&optval, sizeof(optval),
1632 					    optp, (td != NULL) ? td->td_ucred :
1633 					    NULL, uproto);
1634 					break;
1635 				}
1636 
1637 			case IPV6_2292PKTINFO:
1638 			case IPV6_2292HOPLIMIT:
1639 			case IPV6_2292HOPOPTS:
1640 			case IPV6_2292DSTOPTS:
1641 			case IPV6_2292RTHDR:
1642 				/* RFC 2292 */
1643 				if (optlen != sizeof(int)) {
1644 					error = EINVAL;
1645 					break;
1646 				}
1647 				error = sooptcopyin(sopt, &optval,
1648 					sizeof optval, sizeof optval);
1649 				if (error)
1650 					break;
1651 				switch (optname) {
1652 				case IPV6_2292PKTINFO:
1653 					OPTSET2292(IN6P_PKTINFO);
1654 					break;
1655 				case IPV6_2292HOPLIMIT:
1656 					OPTSET2292(IN6P_HOPLIMIT);
1657 					break;
1658 				case IPV6_2292HOPOPTS:
1659 					/*
1660 					 * Check super-user privilege.
1661 					 * See comments for IPV6_RECVHOPOPTS.
1662 					 */
1663 					if (td != NULL) {
1664 						error = priv_check(td,
1665 						    PRIV_NETINET_SETHDROPTS);
1666 						if (error)
1667 							return (error);
1668 					}
1669 					OPTSET2292(IN6P_HOPOPTS);
1670 					break;
1671 				case IPV6_2292DSTOPTS:
1672 					if (td != NULL) {
1673 						error = priv_check(td,
1674 						    PRIV_NETINET_SETHDROPTS);
1675 						if (error)
1676 							return (error);
1677 					}
1678 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1679 					break;
1680 				case IPV6_2292RTHDR:
1681 					OPTSET2292(IN6P_RTHDR);
1682 					break;
1683 				}
1684 				break;
1685 			case IPV6_PKTINFO:
1686 			case IPV6_HOPOPTS:
1687 			case IPV6_RTHDR:
1688 			case IPV6_DSTOPTS:
1689 			case IPV6_RTHDRDSTOPTS:
1690 			case IPV6_NEXTHOP:
1691 			{
1692 				/* new advanced API (RFC3542) */
1693 				u_char *optbuf;
1694 				u_char optbuf_storage[MCLBYTES];
1695 				int optlen;
1696 				struct ip6_pktopts **optp;
1697 
1698 				/* cannot mix with RFC2292 */
1699 				if (OPTBIT(IN6P_RFC2292)) {
1700 					error = EINVAL;
1701 					break;
1702 				}
1703 
1704 				/*
1705 				 * We only ensure valsize is not too large
1706 				 * here.  Further validation will be done
1707 				 * later.
1708 				 */
1709 				error = sooptcopyin(sopt, optbuf_storage,
1710 				    sizeof(optbuf_storage), 0);
1711 				if (error)
1712 					break;
1713 				optlen = sopt->sopt_valsize;
1714 				optbuf = optbuf_storage;
1715 				optp = &in6p->in6p_outputopts;
1716 				error = ip6_pcbopt(optname, optbuf, optlen,
1717 				    optp, (td != NULL) ? td->td_ucred : NULL,
1718 				    uproto);
1719 				break;
1720 			}
1721 #undef OPTSET
1722 
1723 			case IPV6_MULTICAST_IF:
1724 			case IPV6_MULTICAST_HOPS:
1725 			case IPV6_MULTICAST_LOOP:
1726 			case IPV6_JOIN_GROUP:
1727 			case IPV6_LEAVE_GROUP:
1728 			case IPV6_MSFILTER:
1729 			case MCAST_BLOCK_SOURCE:
1730 			case MCAST_UNBLOCK_SOURCE:
1731 			case MCAST_JOIN_GROUP:
1732 			case MCAST_LEAVE_GROUP:
1733 			case MCAST_JOIN_SOURCE_GROUP:
1734 			case MCAST_LEAVE_SOURCE_GROUP:
1735 				error = ip6_setmoptions(in6p, sopt);
1736 				break;
1737 
1738 			case IPV6_PORTRANGE:
1739 				error = sooptcopyin(sopt, &optval,
1740 				    sizeof optval, sizeof optval);
1741 				if (error)
1742 					break;
1743 
1744 				INP_WLOCK(in6p);
1745 				switch (optval) {
1746 				case IPV6_PORTRANGE_DEFAULT:
1747 					in6p->inp_flags &= ~(INP_LOWPORT);
1748 					in6p->inp_flags &= ~(INP_HIGHPORT);
1749 					break;
1750 
1751 				case IPV6_PORTRANGE_HIGH:
1752 					in6p->inp_flags &= ~(INP_LOWPORT);
1753 					in6p->inp_flags |= INP_HIGHPORT;
1754 					break;
1755 
1756 				case IPV6_PORTRANGE_LOW:
1757 					in6p->inp_flags &= ~(INP_HIGHPORT);
1758 					in6p->inp_flags |= INP_LOWPORT;
1759 					break;
1760 
1761 				default:
1762 					error = EINVAL;
1763 					break;
1764 				}
1765 				INP_WUNLOCK(in6p);
1766 				break;
1767 
1768 #ifdef IPSEC
1769 			case IPV6_IPSEC_POLICY:
1770 			{
1771 				caddr_t req;
1772 				struct mbuf *m;
1773 
1774 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1775 					break;
1776 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1777 					break;
1778 				req = mtod(m, caddr_t);
1779 				error = ipsec_set_policy(in6p, optname, req,
1780 				    m->m_len, (sopt->sopt_td != NULL) ?
1781 				    sopt->sopt_td->td_ucred : NULL);
1782 				m_freem(m);
1783 				break;
1784 			}
1785 #endif /* IPSEC */
1786 
1787 			default:
1788 				error = ENOPROTOOPT;
1789 				break;
1790 			}
1791 			break;
1792 
1793 		case SOPT_GET:
1794 			switch (optname) {
1795 
1796 			case IPV6_2292PKTOPTIONS:
1797 #ifdef IPV6_PKTOPTIONS
1798 			case IPV6_PKTOPTIONS:
1799 #endif
1800 				/*
1801 				 * RFC3542 (effectively) deprecated the
1802 				 * semantics of the 2292-style pktoptions.
1803 				 * Since it was not reliable in nature (i.e.,
1804 				 * applications had to expect the lack of some
1805 				 * information after all), it would make sense
1806 				 * to simplify this part by always returning
1807 				 * empty data.
1808 				 */
1809 				sopt->sopt_valsize = 0;
1810 				break;
1811 
1812 			case IPV6_RECVHOPOPTS:
1813 			case IPV6_RECVDSTOPTS:
1814 			case IPV6_RECVRTHDRDSTOPTS:
1815 			case IPV6_UNICAST_HOPS:
1816 			case IPV6_RECVPKTINFO:
1817 			case IPV6_RECVHOPLIMIT:
1818 			case IPV6_RECVRTHDR:
1819 			case IPV6_RECVPATHMTU:
1820 
1821 			case IPV6_V6ONLY:
1822 			case IPV6_PORTRANGE:
1823 			case IPV6_RECVTCLASS:
1824 			case IPV6_AUTOFLOWLABEL:
1825 			case IPV6_BINDANY:
1826 			case IPV6_FLOWID:
1827 			case IPV6_FLOWTYPE:
1828 #ifdef	RSS
1829 			case IPV6_RSSBUCKETID:
1830 #endif
1831 				switch (optname) {
1832 
1833 				case IPV6_RECVHOPOPTS:
1834 					optval = OPTBIT(IN6P_HOPOPTS);
1835 					break;
1836 
1837 				case IPV6_RECVDSTOPTS:
1838 					optval = OPTBIT(IN6P_DSTOPTS);
1839 					break;
1840 
1841 				case IPV6_RECVRTHDRDSTOPTS:
1842 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1843 					break;
1844 
1845 				case IPV6_UNICAST_HOPS:
1846 					optval = in6p->in6p_hops;
1847 					break;
1848 
1849 				case IPV6_RECVPKTINFO:
1850 					optval = OPTBIT(IN6P_PKTINFO);
1851 					break;
1852 
1853 				case IPV6_RECVHOPLIMIT:
1854 					optval = OPTBIT(IN6P_HOPLIMIT);
1855 					break;
1856 
1857 				case IPV6_RECVRTHDR:
1858 					optval = OPTBIT(IN6P_RTHDR);
1859 					break;
1860 
1861 				case IPV6_RECVPATHMTU:
1862 					optval = OPTBIT(IN6P_MTU);
1863 					break;
1864 
1865 				case IPV6_V6ONLY:
1866 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1867 					break;
1868 
1869 				case IPV6_PORTRANGE:
1870 				    {
1871 					int flags;
1872 					flags = in6p->inp_flags;
1873 					if (flags & INP_HIGHPORT)
1874 						optval = IPV6_PORTRANGE_HIGH;
1875 					else if (flags & INP_LOWPORT)
1876 						optval = IPV6_PORTRANGE_LOW;
1877 					else
1878 						optval = 0;
1879 					break;
1880 				    }
1881 				case IPV6_RECVTCLASS:
1882 					optval = OPTBIT(IN6P_TCLASS);
1883 					break;
1884 
1885 				case IPV6_AUTOFLOWLABEL:
1886 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1887 					break;
1888 
1889 				case IPV6_BINDANY:
1890 					optval = OPTBIT(INP_BINDANY);
1891 					break;
1892 
1893 				case IPV6_FLOWID:
1894 					optval = in6p->inp_flowid;
1895 					break;
1896 
1897 				case IPV6_FLOWTYPE:
1898 					optval = in6p->inp_flowtype;
1899 					break;
1900 #ifdef	RSS
1901 				case IPV6_RSSBUCKETID:
1902 					retval =
1903 					    rss_hash2bucket(in6p->inp_flowid,
1904 					    in6p->inp_flowtype,
1905 					    &rss_bucket);
1906 					if (retval == 0)
1907 						optval = rss_bucket;
1908 					else
1909 						error = EINVAL;
1910 					break;
1911 #endif
1912 
1913 				case IPV6_BINDMULTI:
1914 					optval = OPTBIT2(INP_BINDMULTI);
1915 					break;
1916 
1917 				}
1918 				if (error)
1919 					break;
1920 				error = sooptcopyout(sopt, &optval,
1921 					sizeof optval);
1922 				break;
1923 
1924 			case IPV6_PATHMTU:
1925 			{
1926 				u_long pmtu = 0;
1927 				struct ip6_mtuinfo mtuinfo;
1928 				struct route_in6 sro;
1929 
1930 				bzero(&sro, sizeof(sro));
1931 
1932 				if (!(so->so_state & SS_ISCONNECTED))
1933 					return (ENOTCONN);
1934 				/*
1935 				 * XXX: we dot not consider the case of source
1936 				 * routing, or optional information to specify
1937 				 * the outgoing interface.
1938 				 */
1939 				error = ip6_getpmtu(&sro, NULL, NULL,
1940 				    &in6p->in6p_faddr, &pmtu, NULL,
1941 				    so->so_fibnum);
1942 				if (sro.ro_rt)
1943 					RTFREE(sro.ro_rt);
1944 				if (error)
1945 					break;
1946 				if (pmtu > IPV6_MAXPACKET)
1947 					pmtu = IPV6_MAXPACKET;
1948 
1949 				bzero(&mtuinfo, sizeof(mtuinfo));
1950 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1951 				optdata = (void *)&mtuinfo;
1952 				optdatalen = sizeof(mtuinfo);
1953 				error = sooptcopyout(sopt, optdata,
1954 				    optdatalen);
1955 				break;
1956 			}
1957 
1958 			case IPV6_2292PKTINFO:
1959 			case IPV6_2292HOPLIMIT:
1960 			case IPV6_2292HOPOPTS:
1961 			case IPV6_2292RTHDR:
1962 			case IPV6_2292DSTOPTS:
1963 				switch (optname) {
1964 				case IPV6_2292PKTINFO:
1965 					optval = OPTBIT(IN6P_PKTINFO);
1966 					break;
1967 				case IPV6_2292HOPLIMIT:
1968 					optval = OPTBIT(IN6P_HOPLIMIT);
1969 					break;
1970 				case IPV6_2292HOPOPTS:
1971 					optval = OPTBIT(IN6P_HOPOPTS);
1972 					break;
1973 				case IPV6_2292RTHDR:
1974 					optval = OPTBIT(IN6P_RTHDR);
1975 					break;
1976 				case IPV6_2292DSTOPTS:
1977 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1978 					break;
1979 				}
1980 				error = sooptcopyout(sopt, &optval,
1981 				    sizeof optval);
1982 				break;
1983 			case IPV6_PKTINFO:
1984 			case IPV6_HOPOPTS:
1985 			case IPV6_RTHDR:
1986 			case IPV6_DSTOPTS:
1987 			case IPV6_RTHDRDSTOPTS:
1988 			case IPV6_NEXTHOP:
1989 			case IPV6_TCLASS:
1990 			case IPV6_DONTFRAG:
1991 			case IPV6_USE_MIN_MTU:
1992 			case IPV6_PREFER_TEMPADDR:
1993 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1994 				    optname, sopt);
1995 				break;
1996 
1997 			case IPV6_MULTICAST_IF:
1998 			case IPV6_MULTICAST_HOPS:
1999 			case IPV6_MULTICAST_LOOP:
2000 			case IPV6_MSFILTER:
2001 				error = ip6_getmoptions(in6p, sopt);
2002 				break;
2003 
2004 #ifdef IPSEC
2005 			case IPV6_IPSEC_POLICY:
2006 			  {
2007 				caddr_t req = NULL;
2008 				size_t len = 0;
2009 				struct mbuf *m = NULL;
2010 				struct mbuf **mp = &m;
2011 				size_t ovalsize = sopt->sopt_valsize;
2012 				caddr_t oval = (caddr_t)sopt->sopt_val;
2013 
2014 				error = soopt_getm(sopt, &m); /* XXX */
2015 				if (error != 0)
2016 					break;
2017 				error = soopt_mcopyin(sopt, m); /* XXX */
2018 				if (error != 0)
2019 					break;
2020 				sopt->sopt_valsize = ovalsize;
2021 				sopt->sopt_val = oval;
2022 				if (m) {
2023 					req = mtod(m, caddr_t);
2024 					len = m->m_len;
2025 				}
2026 				error = ipsec_get_policy(in6p, req, len, mp);
2027 				if (error == 0)
2028 					error = soopt_mcopyout(sopt, m); /* XXX */
2029 				if (error == 0 && m)
2030 					m_freem(m);
2031 				break;
2032 			  }
2033 #endif /* IPSEC */
2034 
2035 			default:
2036 				error = ENOPROTOOPT;
2037 				break;
2038 			}
2039 			break;
2040 		}
2041 	}
2042 	return (error);
2043 }
2044 
2045 int
2046 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2047 {
2048 	int error = 0, optval, optlen;
2049 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2050 	struct inpcb *in6p = sotoinpcb(so);
2051 	int level, op, optname;
2052 
2053 	level = sopt->sopt_level;
2054 	op = sopt->sopt_dir;
2055 	optname = sopt->sopt_name;
2056 	optlen = sopt->sopt_valsize;
2057 
2058 	if (level != IPPROTO_IPV6) {
2059 		return (EINVAL);
2060 	}
2061 
2062 	switch (optname) {
2063 	case IPV6_CHECKSUM:
2064 		/*
2065 		 * For ICMPv6 sockets, no modification allowed for checksum
2066 		 * offset, permit "no change" values to help existing apps.
2067 		 *
2068 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2069 		 * for an ICMPv6 socket will fail."
2070 		 * The current behavior does not meet RFC3542.
2071 		 */
2072 		switch (op) {
2073 		case SOPT_SET:
2074 			if (optlen != sizeof(int)) {
2075 				error = EINVAL;
2076 				break;
2077 			}
2078 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2079 					    sizeof(optval));
2080 			if (error)
2081 				break;
2082 			if ((optval % 2) != 0) {
2083 				/* the API assumes even offset values */
2084 				error = EINVAL;
2085 			} else if (so->so_proto->pr_protocol ==
2086 			    IPPROTO_ICMPV6) {
2087 				if (optval != icmp6off)
2088 					error = EINVAL;
2089 			} else
2090 				in6p->in6p_cksum = optval;
2091 			break;
2092 
2093 		case SOPT_GET:
2094 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2095 				optval = icmp6off;
2096 			else
2097 				optval = in6p->in6p_cksum;
2098 
2099 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2100 			break;
2101 
2102 		default:
2103 			error = EINVAL;
2104 			break;
2105 		}
2106 		break;
2107 
2108 	default:
2109 		error = ENOPROTOOPT;
2110 		break;
2111 	}
2112 
2113 	return (error);
2114 }
2115 
2116 /*
2117  * Set up IP6 options in pcb for insertion in output packets or
2118  * specifying behavior of outgoing packets.
2119  */
2120 static int
2121 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2122     struct socket *so, struct sockopt *sopt)
2123 {
2124 	struct ip6_pktopts *opt = *pktopt;
2125 	int error = 0;
2126 	struct thread *td = sopt->sopt_td;
2127 
2128 	/* turn off any old options. */
2129 	if (opt) {
2130 #ifdef DIAGNOSTIC
2131 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2132 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2133 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2134 			printf("ip6_pcbopts: all specified options are cleared.\n");
2135 #endif
2136 		ip6_clearpktopts(opt, -1);
2137 	} else
2138 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2139 	*pktopt = NULL;
2140 
2141 	if (!m || m->m_len == 0) {
2142 		/*
2143 		 * Only turning off any previous options, regardless of
2144 		 * whether the opt is just created or given.
2145 		 */
2146 		free(opt, M_IP6OPT);
2147 		return (0);
2148 	}
2149 
2150 	/*  set options specified by user. */
2151 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2152 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2153 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2154 		free(opt, M_IP6OPT);
2155 		return (error);
2156 	}
2157 	*pktopt = opt;
2158 	return (0);
2159 }
2160 
2161 /*
2162  * initialize ip6_pktopts.  beware that there are non-zero default values in
2163  * the struct.
2164  */
2165 void
2166 ip6_initpktopts(struct ip6_pktopts *opt)
2167 {
2168 
2169 	bzero(opt, sizeof(*opt));
2170 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2171 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2172 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2173 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2174 }
2175 
2176 static int
2177 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2178     struct ucred *cred, int uproto)
2179 {
2180 	struct ip6_pktopts *opt;
2181 
2182 	if (*pktopt == NULL) {
2183 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2184 		    M_WAITOK);
2185 		ip6_initpktopts(*pktopt);
2186 	}
2187 	opt = *pktopt;
2188 
2189 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2190 }
2191 
2192 static int
2193 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2194 {
2195 	void *optdata = NULL;
2196 	int optdatalen = 0;
2197 	struct ip6_ext *ip6e;
2198 	int error = 0;
2199 	struct in6_pktinfo null_pktinfo;
2200 	int deftclass = 0, on;
2201 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2202 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2203 
2204 	switch (optname) {
2205 	case IPV6_PKTINFO:
2206 		if (pktopt && pktopt->ip6po_pktinfo)
2207 			optdata = (void *)pktopt->ip6po_pktinfo;
2208 		else {
2209 			/* XXX: we don't have to do this every time... */
2210 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2211 			optdata = (void *)&null_pktinfo;
2212 		}
2213 		optdatalen = sizeof(struct in6_pktinfo);
2214 		break;
2215 	case IPV6_TCLASS:
2216 		if (pktopt && pktopt->ip6po_tclass >= 0)
2217 			optdata = (void *)&pktopt->ip6po_tclass;
2218 		else
2219 			optdata = (void *)&deftclass;
2220 		optdatalen = sizeof(int);
2221 		break;
2222 	case IPV6_HOPOPTS:
2223 		if (pktopt && pktopt->ip6po_hbh) {
2224 			optdata = (void *)pktopt->ip6po_hbh;
2225 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2226 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2227 		}
2228 		break;
2229 	case IPV6_RTHDR:
2230 		if (pktopt && pktopt->ip6po_rthdr) {
2231 			optdata = (void *)pktopt->ip6po_rthdr;
2232 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2233 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2234 		}
2235 		break;
2236 	case IPV6_RTHDRDSTOPTS:
2237 		if (pktopt && pktopt->ip6po_dest1) {
2238 			optdata = (void *)pktopt->ip6po_dest1;
2239 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2240 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2241 		}
2242 		break;
2243 	case IPV6_DSTOPTS:
2244 		if (pktopt && pktopt->ip6po_dest2) {
2245 			optdata = (void *)pktopt->ip6po_dest2;
2246 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2247 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2248 		}
2249 		break;
2250 	case IPV6_NEXTHOP:
2251 		if (pktopt && pktopt->ip6po_nexthop) {
2252 			optdata = (void *)pktopt->ip6po_nexthop;
2253 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2254 		}
2255 		break;
2256 	case IPV6_USE_MIN_MTU:
2257 		if (pktopt)
2258 			optdata = (void *)&pktopt->ip6po_minmtu;
2259 		else
2260 			optdata = (void *)&defminmtu;
2261 		optdatalen = sizeof(int);
2262 		break;
2263 	case IPV6_DONTFRAG:
2264 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2265 			on = 1;
2266 		else
2267 			on = 0;
2268 		optdata = (void *)&on;
2269 		optdatalen = sizeof(on);
2270 		break;
2271 	case IPV6_PREFER_TEMPADDR:
2272 		if (pktopt)
2273 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2274 		else
2275 			optdata = (void *)&defpreftemp;
2276 		optdatalen = sizeof(int);
2277 		break;
2278 	default:		/* should not happen */
2279 #ifdef DIAGNOSTIC
2280 		panic("ip6_getpcbopt: unexpected option\n");
2281 #endif
2282 		return (ENOPROTOOPT);
2283 	}
2284 
2285 	error = sooptcopyout(sopt, optdata, optdatalen);
2286 
2287 	return (error);
2288 }
2289 
2290 void
2291 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2292 {
2293 	if (pktopt == NULL)
2294 		return;
2295 
2296 	if (optname == -1 || optname == IPV6_PKTINFO) {
2297 		if (pktopt->ip6po_pktinfo)
2298 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2299 		pktopt->ip6po_pktinfo = NULL;
2300 	}
2301 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2302 		pktopt->ip6po_hlim = -1;
2303 	if (optname == -1 || optname == IPV6_TCLASS)
2304 		pktopt->ip6po_tclass = -1;
2305 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2306 		if (pktopt->ip6po_nextroute.ro_rt) {
2307 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2308 			pktopt->ip6po_nextroute.ro_rt = NULL;
2309 		}
2310 		if (pktopt->ip6po_nexthop)
2311 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2312 		pktopt->ip6po_nexthop = NULL;
2313 	}
2314 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2315 		if (pktopt->ip6po_hbh)
2316 			free(pktopt->ip6po_hbh, M_IP6OPT);
2317 		pktopt->ip6po_hbh = NULL;
2318 	}
2319 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2320 		if (pktopt->ip6po_dest1)
2321 			free(pktopt->ip6po_dest1, M_IP6OPT);
2322 		pktopt->ip6po_dest1 = NULL;
2323 	}
2324 	if (optname == -1 || optname == IPV6_RTHDR) {
2325 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2326 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2327 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2328 		if (pktopt->ip6po_route.ro_rt) {
2329 			RTFREE(pktopt->ip6po_route.ro_rt);
2330 			pktopt->ip6po_route.ro_rt = NULL;
2331 		}
2332 	}
2333 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2334 		if (pktopt->ip6po_dest2)
2335 			free(pktopt->ip6po_dest2, M_IP6OPT);
2336 		pktopt->ip6po_dest2 = NULL;
2337 	}
2338 }
2339 
2340 #define PKTOPT_EXTHDRCPY(type) \
2341 do {\
2342 	if (src->type) {\
2343 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2344 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2345 		if (dst->type == NULL && canwait == M_NOWAIT)\
2346 			goto bad;\
2347 		bcopy(src->type, dst->type, hlen);\
2348 	}\
2349 } while (/*CONSTCOND*/ 0)
2350 
2351 static int
2352 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2353 {
2354 	if (dst == NULL || src == NULL)  {
2355 		printf("ip6_clearpktopts: invalid argument\n");
2356 		return (EINVAL);
2357 	}
2358 
2359 	dst->ip6po_hlim = src->ip6po_hlim;
2360 	dst->ip6po_tclass = src->ip6po_tclass;
2361 	dst->ip6po_flags = src->ip6po_flags;
2362 	dst->ip6po_minmtu = src->ip6po_minmtu;
2363 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2364 	if (src->ip6po_pktinfo) {
2365 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2366 		    M_IP6OPT, canwait);
2367 		if (dst->ip6po_pktinfo == NULL)
2368 			goto bad;
2369 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2370 	}
2371 	if (src->ip6po_nexthop) {
2372 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2373 		    M_IP6OPT, canwait);
2374 		if (dst->ip6po_nexthop == NULL)
2375 			goto bad;
2376 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2377 		    src->ip6po_nexthop->sa_len);
2378 	}
2379 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2380 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2381 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2382 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2383 	return (0);
2384 
2385   bad:
2386 	ip6_clearpktopts(dst, -1);
2387 	return (ENOBUFS);
2388 }
2389 #undef PKTOPT_EXTHDRCPY
2390 
2391 struct ip6_pktopts *
2392 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2393 {
2394 	int error;
2395 	struct ip6_pktopts *dst;
2396 
2397 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2398 	if (dst == NULL)
2399 		return (NULL);
2400 	ip6_initpktopts(dst);
2401 
2402 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2403 		free(dst, M_IP6OPT);
2404 		return (NULL);
2405 	}
2406 
2407 	return (dst);
2408 }
2409 
2410 void
2411 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2412 {
2413 	if (pktopt == NULL)
2414 		return;
2415 
2416 	ip6_clearpktopts(pktopt, -1);
2417 
2418 	free(pktopt, M_IP6OPT);
2419 }
2420 
2421 /*
2422  * Set IPv6 outgoing packet options based on advanced API.
2423  */
2424 int
2425 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2426     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2427 {
2428 	struct cmsghdr *cm = 0;
2429 
2430 	if (control == NULL || opt == NULL)
2431 		return (EINVAL);
2432 
2433 	ip6_initpktopts(opt);
2434 	if (stickyopt) {
2435 		int error;
2436 
2437 		/*
2438 		 * If stickyopt is provided, make a local copy of the options
2439 		 * for this particular packet, then override them by ancillary
2440 		 * objects.
2441 		 * XXX: copypktopts() does not copy the cached route to a next
2442 		 * hop (if any).  This is not very good in terms of efficiency,
2443 		 * but we can allow this since this option should be rarely
2444 		 * used.
2445 		 */
2446 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2447 			return (error);
2448 	}
2449 
2450 	/*
2451 	 * XXX: Currently, we assume all the optional information is stored
2452 	 * in a single mbuf.
2453 	 */
2454 	if (control->m_next)
2455 		return (EINVAL);
2456 
2457 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2458 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2459 		int error;
2460 
2461 		if (control->m_len < CMSG_LEN(0))
2462 			return (EINVAL);
2463 
2464 		cm = mtod(control, struct cmsghdr *);
2465 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2466 			return (EINVAL);
2467 		if (cm->cmsg_level != IPPROTO_IPV6)
2468 			continue;
2469 
2470 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2471 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2472 		if (error)
2473 			return (error);
2474 	}
2475 
2476 	return (0);
2477 }
2478 
2479 /*
2480  * Set a particular packet option, as a sticky option or an ancillary data
2481  * item.  "len" can be 0 only when it's a sticky option.
2482  * We have 4 cases of combination of "sticky" and "cmsg":
2483  * "sticky=0, cmsg=0": impossible
2484  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2485  * "sticky=1, cmsg=0": RFC3542 socket option
2486  * "sticky=1, cmsg=1": RFC2292 socket option
2487  */
2488 static int
2489 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2490     struct ucred *cred, int sticky, int cmsg, int uproto)
2491 {
2492 	int minmtupolicy, preftemp;
2493 	int error;
2494 
2495 	if (!sticky && !cmsg) {
2496 #ifdef DIAGNOSTIC
2497 		printf("ip6_setpktopt: impossible case\n");
2498 #endif
2499 		return (EINVAL);
2500 	}
2501 
2502 	/*
2503 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2504 	 * not be specified in the context of RFC3542.  Conversely,
2505 	 * RFC3542 types should not be specified in the context of RFC2292.
2506 	 */
2507 	if (!cmsg) {
2508 		switch (optname) {
2509 		case IPV6_2292PKTINFO:
2510 		case IPV6_2292HOPLIMIT:
2511 		case IPV6_2292NEXTHOP:
2512 		case IPV6_2292HOPOPTS:
2513 		case IPV6_2292DSTOPTS:
2514 		case IPV6_2292RTHDR:
2515 		case IPV6_2292PKTOPTIONS:
2516 			return (ENOPROTOOPT);
2517 		}
2518 	}
2519 	if (sticky && cmsg) {
2520 		switch (optname) {
2521 		case IPV6_PKTINFO:
2522 		case IPV6_HOPLIMIT:
2523 		case IPV6_NEXTHOP:
2524 		case IPV6_HOPOPTS:
2525 		case IPV6_DSTOPTS:
2526 		case IPV6_RTHDRDSTOPTS:
2527 		case IPV6_RTHDR:
2528 		case IPV6_USE_MIN_MTU:
2529 		case IPV6_DONTFRAG:
2530 		case IPV6_TCLASS:
2531 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2532 			return (ENOPROTOOPT);
2533 		}
2534 	}
2535 
2536 	switch (optname) {
2537 	case IPV6_2292PKTINFO:
2538 	case IPV6_PKTINFO:
2539 	{
2540 		struct ifnet *ifp = NULL;
2541 		struct in6_pktinfo *pktinfo;
2542 
2543 		if (len != sizeof(struct in6_pktinfo))
2544 			return (EINVAL);
2545 
2546 		pktinfo = (struct in6_pktinfo *)buf;
2547 
2548 		/*
2549 		 * An application can clear any sticky IPV6_PKTINFO option by
2550 		 * doing a "regular" setsockopt with ipi6_addr being
2551 		 * in6addr_any and ipi6_ifindex being zero.
2552 		 * [RFC 3542, Section 6]
2553 		 */
2554 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2555 		    pktinfo->ipi6_ifindex == 0 &&
2556 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2557 			ip6_clearpktopts(opt, optname);
2558 			break;
2559 		}
2560 
2561 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2562 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2563 			return (EINVAL);
2564 		}
2565 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2566 			return (EINVAL);
2567 		/* validate the interface index if specified. */
2568 		if (pktinfo->ipi6_ifindex > V_if_index)
2569 			 return (ENXIO);
2570 		if (pktinfo->ipi6_ifindex) {
2571 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2572 			if (ifp == NULL)
2573 				return (ENXIO);
2574 		}
2575 		if (ifp != NULL && (
2576 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2577 			return (ENETDOWN);
2578 
2579 		if (ifp != NULL &&
2580 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2581 			struct in6_ifaddr *ia;
2582 
2583 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2584 			if (ia == NULL)
2585 				return (EADDRNOTAVAIL);
2586 			ifa_free(&ia->ia_ifa);
2587 		}
2588 		/*
2589 		 * We store the address anyway, and let in6_selectsrc()
2590 		 * validate the specified address.  This is because ipi6_addr
2591 		 * may not have enough information about its scope zone, and
2592 		 * we may need additional information (such as outgoing
2593 		 * interface or the scope zone of a destination address) to
2594 		 * disambiguate the scope.
2595 		 * XXX: the delay of the validation may confuse the
2596 		 * application when it is used as a sticky option.
2597 		 */
2598 		if (opt->ip6po_pktinfo == NULL) {
2599 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2600 			    M_IP6OPT, M_NOWAIT);
2601 			if (opt->ip6po_pktinfo == NULL)
2602 				return (ENOBUFS);
2603 		}
2604 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2605 		break;
2606 	}
2607 
2608 	case IPV6_2292HOPLIMIT:
2609 	case IPV6_HOPLIMIT:
2610 	{
2611 		int *hlimp;
2612 
2613 		/*
2614 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2615 		 * to simplify the ordering among hoplimit options.
2616 		 */
2617 		if (optname == IPV6_HOPLIMIT && sticky)
2618 			return (ENOPROTOOPT);
2619 
2620 		if (len != sizeof(int))
2621 			return (EINVAL);
2622 		hlimp = (int *)buf;
2623 		if (*hlimp < -1 || *hlimp > 255)
2624 			return (EINVAL);
2625 
2626 		opt->ip6po_hlim = *hlimp;
2627 		break;
2628 	}
2629 
2630 	case IPV6_TCLASS:
2631 	{
2632 		int tclass;
2633 
2634 		if (len != sizeof(int))
2635 			return (EINVAL);
2636 		tclass = *(int *)buf;
2637 		if (tclass < -1 || tclass > 255)
2638 			return (EINVAL);
2639 
2640 		opt->ip6po_tclass = tclass;
2641 		break;
2642 	}
2643 
2644 	case IPV6_2292NEXTHOP:
2645 	case IPV6_NEXTHOP:
2646 		if (cred != NULL) {
2647 			error = priv_check_cred(cred,
2648 			    PRIV_NETINET_SETHDROPTS, 0);
2649 			if (error)
2650 				return (error);
2651 		}
2652 
2653 		if (len == 0) {	/* just remove the option */
2654 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2655 			break;
2656 		}
2657 
2658 		/* check if cmsg_len is large enough for sa_len */
2659 		if (len < sizeof(struct sockaddr) || len < *buf)
2660 			return (EINVAL);
2661 
2662 		switch (((struct sockaddr *)buf)->sa_family) {
2663 		case AF_INET6:
2664 		{
2665 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2666 			int error;
2667 
2668 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2669 				return (EINVAL);
2670 
2671 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2672 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2673 				return (EINVAL);
2674 			}
2675 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2676 			    != 0) {
2677 				return (error);
2678 			}
2679 			break;
2680 		}
2681 		case AF_LINK:	/* should eventually be supported */
2682 		default:
2683 			return (EAFNOSUPPORT);
2684 		}
2685 
2686 		/* turn off the previous option, then set the new option. */
2687 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2688 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2689 		if (opt->ip6po_nexthop == NULL)
2690 			return (ENOBUFS);
2691 		bcopy(buf, opt->ip6po_nexthop, *buf);
2692 		break;
2693 
2694 	case IPV6_2292HOPOPTS:
2695 	case IPV6_HOPOPTS:
2696 	{
2697 		struct ip6_hbh *hbh;
2698 		int hbhlen;
2699 
2700 		/*
2701 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2702 		 * options, since per-option restriction has too much
2703 		 * overhead.
2704 		 */
2705 		if (cred != NULL) {
2706 			error = priv_check_cred(cred,
2707 			    PRIV_NETINET_SETHDROPTS, 0);
2708 			if (error)
2709 				return (error);
2710 		}
2711 
2712 		if (len == 0) {
2713 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2714 			break;	/* just remove the option */
2715 		}
2716 
2717 		/* message length validation */
2718 		if (len < sizeof(struct ip6_hbh))
2719 			return (EINVAL);
2720 		hbh = (struct ip6_hbh *)buf;
2721 		hbhlen = (hbh->ip6h_len + 1) << 3;
2722 		if (len != hbhlen)
2723 			return (EINVAL);
2724 
2725 		/* turn off the previous option, then set the new option. */
2726 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2727 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2728 		if (opt->ip6po_hbh == NULL)
2729 			return (ENOBUFS);
2730 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2731 
2732 		break;
2733 	}
2734 
2735 	case IPV6_2292DSTOPTS:
2736 	case IPV6_DSTOPTS:
2737 	case IPV6_RTHDRDSTOPTS:
2738 	{
2739 		struct ip6_dest *dest, **newdest = NULL;
2740 		int destlen;
2741 
2742 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2743 			error = priv_check_cred(cred,
2744 			    PRIV_NETINET_SETHDROPTS, 0);
2745 			if (error)
2746 				return (error);
2747 		}
2748 
2749 		if (len == 0) {
2750 			ip6_clearpktopts(opt, optname);
2751 			break;	/* just remove the option */
2752 		}
2753 
2754 		/* message length validation */
2755 		if (len < sizeof(struct ip6_dest))
2756 			return (EINVAL);
2757 		dest = (struct ip6_dest *)buf;
2758 		destlen = (dest->ip6d_len + 1) << 3;
2759 		if (len != destlen)
2760 			return (EINVAL);
2761 
2762 		/*
2763 		 * Determine the position that the destination options header
2764 		 * should be inserted; before or after the routing header.
2765 		 */
2766 		switch (optname) {
2767 		case IPV6_2292DSTOPTS:
2768 			/*
2769 			 * The old advacned API is ambiguous on this point.
2770 			 * Our approach is to determine the position based
2771 			 * according to the existence of a routing header.
2772 			 * Note, however, that this depends on the order of the
2773 			 * extension headers in the ancillary data; the 1st
2774 			 * part of the destination options header must appear
2775 			 * before the routing header in the ancillary data,
2776 			 * too.
2777 			 * RFC3542 solved the ambiguity by introducing
2778 			 * separate ancillary data or option types.
2779 			 */
2780 			if (opt->ip6po_rthdr == NULL)
2781 				newdest = &opt->ip6po_dest1;
2782 			else
2783 				newdest = &opt->ip6po_dest2;
2784 			break;
2785 		case IPV6_RTHDRDSTOPTS:
2786 			newdest = &opt->ip6po_dest1;
2787 			break;
2788 		case IPV6_DSTOPTS:
2789 			newdest = &opt->ip6po_dest2;
2790 			break;
2791 		}
2792 
2793 		/* turn off the previous option, then set the new option. */
2794 		ip6_clearpktopts(opt, optname);
2795 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2796 		if (*newdest == NULL)
2797 			return (ENOBUFS);
2798 		bcopy(dest, *newdest, destlen);
2799 
2800 		break;
2801 	}
2802 
2803 	case IPV6_2292RTHDR:
2804 	case IPV6_RTHDR:
2805 	{
2806 		struct ip6_rthdr *rth;
2807 		int rthlen;
2808 
2809 		if (len == 0) {
2810 			ip6_clearpktopts(opt, IPV6_RTHDR);
2811 			break;	/* just remove the option */
2812 		}
2813 
2814 		/* message length validation */
2815 		if (len < sizeof(struct ip6_rthdr))
2816 			return (EINVAL);
2817 		rth = (struct ip6_rthdr *)buf;
2818 		rthlen = (rth->ip6r_len + 1) << 3;
2819 		if (len != rthlen)
2820 			return (EINVAL);
2821 
2822 		switch (rth->ip6r_type) {
2823 		case IPV6_RTHDR_TYPE_0:
2824 			if (rth->ip6r_len == 0)	/* must contain one addr */
2825 				return (EINVAL);
2826 			if (rth->ip6r_len % 2) /* length must be even */
2827 				return (EINVAL);
2828 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2829 				return (EINVAL);
2830 			break;
2831 		default:
2832 			return (EINVAL);	/* not supported */
2833 		}
2834 
2835 		/* turn off the previous option */
2836 		ip6_clearpktopts(opt, IPV6_RTHDR);
2837 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2838 		if (opt->ip6po_rthdr == NULL)
2839 			return (ENOBUFS);
2840 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2841 
2842 		break;
2843 	}
2844 
2845 	case IPV6_USE_MIN_MTU:
2846 		if (len != sizeof(int))
2847 			return (EINVAL);
2848 		minmtupolicy = *(int *)buf;
2849 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2850 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2851 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2852 			return (EINVAL);
2853 		}
2854 		opt->ip6po_minmtu = minmtupolicy;
2855 		break;
2856 
2857 	case IPV6_DONTFRAG:
2858 		if (len != sizeof(int))
2859 			return (EINVAL);
2860 
2861 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2862 			/*
2863 			 * we ignore this option for TCP sockets.
2864 			 * (RFC3542 leaves this case unspecified.)
2865 			 */
2866 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2867 		} else
2868 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2869 		break;
2870 
2871 	case IPV6_PREFER_TEMPADDR:
2872 		if (len != sizeof(int))
2873 			return (EINVAL);
2874 		preftemp = *(int *)buf;
2875 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2876 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2877 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2878 			return (EINVAL);
2879 		}
2880 		opt->ip6po_prefer_tempaddr = preftemp;
2881 		break;
2882 
2883 	default:
2884 		return (ENOPROTOOPT);
2885 	} /* end of switch */
2886 
2887 	return (0);
2888 }
2889 
2890 /*
2891  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2892  * packet to the input queue of a specified interface.  Note that this
2893  * calls the output routine of the loopback "driver", but with an interface
2894  * pointer that might NOT be &loif -- easier than replicating that code here.
2895  */
2896 void
2897 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2898 {
2899 	struct mbuf *copym;
2900 	struct ip6_hdr *ip6;
2901 
2902 	copym = m_copy(m, 0, M_COPYALL);
2903 	if (copym == NULL)
2904 		return;
2905 
2906 	/*
2907 	 * Make sure to deep-copy IPv6 header portion in case the data
2908 	 * is in an mbuf cluster, so that we can safely override the IPv6
2909 	 * header portion later.
2910 	 */
2911 	if (!M_WRITABLE(copym) ||
2912 	    copym->m_len < sizeof(struct ip6_hdr)) {
2913 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2914 		if (copym == NULL)
2915 			return;
2916 	}
2917 
2918 #ifdef DIAGNOSTIC
2919 	if (copym->m_len < sizeof(*ip6)) {
2920 		m_freem(copym);
2921 		return;
2922 	}
2923 #endif
2924 
2925 	ip6 = mtod(copym, struct ip6_hdr *);
2926 	/*
2927 	 * clear embedded scope identifiers if necessary.
2928 	 * in6_clearscope will touch the addresses only when necessary.
2929 	 */
2930 	in6_clearscope(&ip6->ip6_src);
2931 	in6_clearscope(&ip6->ip6_dst);
2932 
2933 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2934 }
2935 
2936 /*
2937  * Chop IPv6 header off from the payload.
2938  */
2939 static int
2940 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2941 {
2942 	struct mbuf *mh;
2943 	struct ip6_hdr *ip6;
2944 
2945 	ip6 = mtod(m, struct ip6_hdr *);
2946 	if (m->m_len > sizeof(*ip6)) {
2947 		mh = m_gethdr(M_NOWAIT, MT_DATA);
2948 		if (mh == NULL) {
2949 			m_freem(m);
2950 			return ENOBUFS;
2951 		}
2952 		m_move_pkthdr(mh, m);
2953 		MH_ALIGN(mh, sizeof(*ip6));
2954 		m->m_len -= sizeof(*ip6);
2955 		m->m_data += sizeof(*ip6);
2956 		mh->m_next = m;
2957 		m = mh;
2958 		m->m_len = sizeof(*ip6);
2959 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2960 	}
2961 	exthdrs->ip6e_ip6 = m;
2962 	return 0;
2963 }
2964 
2965 /*
2966  * Compute IPv6 extension header length.
2967  */
2968 int
2969 ip6_optlen(struct inpcb *in6p)
2970 {
2971 	int len;
2972 
2973 	if (!in6p->in6p_outputopts)
2974 		return 0;
2975 
2976 	len = 0;
2977 #define elen(x) \
2978     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2979 
2980 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2981 	if (in6p->in6p_outputopts->ip6po_rthdr)
2982 		/* dest1 is valid with rthdr only */
2983 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2984 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2985 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2986 	return len;
2987 #undef elen
2988 }
2989