xref: /freebsd/sys/netinet6/ip6_output.c (revision 6b129086dcee14496517fae085b448e3edc69bc7)
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 		}
1279 	} else if (ifp) {
1280 		mtu = IN6_LINKMTU(ifp);
1281 	} else
1282 		error = EHOSTUNREACH; /* XXX */
1283 
1284 	*mtup = mtu;
1285 	if (alwaysfragp)
1286 		*alwaysfragp = alwaysfrag;
1287 	return (error);
1288 }
1289 
1290 /*
1291  * IP6 socket option processing.
1292  */
1293 int
1294 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1295 {
1296 	int optdatalen, uproto;
1297 	void *optdata;
1298 	struct inpcb *in6p = sotoinpcb(so);
1299 	int error, optval;
1300 	int level, op, optname;
1301 	int optlen;
1302 	struct thread *td;
1303 #ifdef	RSS
1304 	uint32_t rss_bucket;
1305 	int retval;
1306 #endif
1307 
1308 	level = sopt->sopt_level;
1309 	op = sopt->sopt_dir;
1310 	optname = sopt->sopt_name;
1311 	optlen = sopt->sopt_valsize;
1312 	td = sopt->sopt_td;
1313 	error = 0;
1314 	optval = 0;
1315 	uproto = (int)so->so_proto->pr_protocol;
1316 
1317 	if (level != IPPROTO_IPV6) {
1318 		error = EINVAL;
1319 
1320 		if (sopt->sopt_level == SOL_SOCKET &&
1321 		    sopt->sopt_dir == SOPT_SET) {
1322 			switch (sopt->sopt_name) {
1323 			case SO_REUSEADDR:
1324 				INP_WLOCK(in6p);
1325 				if ((so->so_options & SO_REUSEADDR) != 0)
1326 					in6p->inp_flags2 |= INP_REUSEADDR;
1327 				else
1328 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1329 				INP_WUNLOCK(in6p);
1330 				error = 0;
1331 				break;
1332 			case SO_REUSEPORT:
1333 				INP_WLOCK(in6p);
1334 				if ((so->so_options & SO_REUSEPORT) != 0)
1335 					in6p->inp_flags2 |= INP_REUSEPORT;
1336 				else
1337 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1338 				INP_WUNLOCK(in6p);
1339 				error = 0;
1340 				break;
1341 			case SO_SETFIB:
1342 				INP_WLOCK(in6p);
1343 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1344 				INP_WUNLOCK(in6p);
1345 				error = 0;
1346 				break;
1347 			default:
1348 				break;
1349 			}
1350 		}
1351 	} else {		/* level == IPPROTO_IPV6 */
1352 		switch (op) {
1353 
1354 		case SOPT_SET:
1355 			switch (optname) {
1356 			case IPV6_2292PKTOPTIONS:
1357 #ifdef IPV6_PKTOPTIONS
1358 			case IPV6_PKTOPTIONS:
1359 #endif
1360 			{
1361 				struct mbuf *m;
1362 
1363 				error = soopt_getm(sopt, &m); /* XXX */
1364 				if (error != 0)
1365 					break;
1366 				error = soopt_mcopyin(sopt, m); /* XXX */
1367 				if (error != 0)
1368 					break;
1369 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1370 						    m, so, sopt);
1371 				m_freem(m); /* XXX */
1372 				break;
1373 			}
1374 
1375 			/*
1376 			 * Use of some Hop-by-Hop options or some
1377 			 * Destination options, might require special
1378 			 * privilege.  That is, normal applications
1379 			 * (without special privilege) might be forbidden
1380 			 * from setting certain options in outgoing packets,
1381 			 * and might never see certain options in received
1382 			 * packets. [RFC 2292 Section 6]
1383 			 * KAME specific note:
1384 			 *  KAME prevents non-privileged users from sending or
1385 			 *  receiving ANY hbh/dst options in order to avoid
1386 			 *  overhead of parsing options in the kernel.
1387 			 */
1388 			case IPV6_RECVHOPOPTS:
1389 			case IPV6_RECVDSTOPTS:
1390 			case IPV6_RECVRTHDRDSTOPTS:
1391 				if (td != NULL) {
1392 					error = priv_check(td,
1393 					    PRIV_NETINET_SETHDROPTS);
1394 					if (error)
1395 						break;
1396 				}
1397 				/* FALLTHROUGH */
1398 			case IPV6_UNICAST_HOPS:
1399 			case IPV6_HOPLIMIT:
1400 
1401 			case IPV6_RECVPKTINFO:
1402 			case IPV6_RECVHOPLIMIT:
1403 			case IPV6_RECVRTHDR:
1404 			case IPV6_RECVPATHMTU:
1405 			case IPV6_RECVTCLASS:
1406 			case IPV6_V6ONLY:
1407 			case IPV6_AUTOFLOWLABEL:
1408 			case IPV6_BINDANY:
1409 			case IPV6_BINDMULTI:
1410 #ifdef	RSS
1411 			case IPV6_RSS_LISTEN_BUCKET:
1412 #endif
1413 				if (optname == IPV6_BINDANY && td != NULL) {
1414 					error = priv_check(td,
1415 					    PRIV_NETINET_BINDANY);
1416 					if (error)
1417 						break;
1418 				}
1419 
1420 				if (optlen != sizeof(int)) {
1421 					error = EINVAL;
1422 					break;
1423 				}
1424 				error = sooptcopyin(sopt, &optval,
1425 					sizeof optval, sizeof optval);
1426 				if (error)
1427 					break;
1428 				switch (optname) {
1429 
1430 				case IPV6_UNICAST_HOPS:
1431 					if (optval < -1 || optval >= 256)
1432 						error = EINVAL;
1433 					else {
1434 						/* -1 = kernel default */
1435 						in6p->in6p_hops = optval;
1436 						if ((in6p->inp_vflag &
1437 						     INP_IPV4) != 0)
1438 							in6p->inp_ip_ttl = optval;
1439 					}
1440 					break;
1441 #define OPTSET(bit) \
1442 do { \
1443 	INP_WLOCK(in6p); \
1444 	if (optval) \
1445 		in6p->inp_flags |= (bit); \
1446 	else \
1447 		in6p->inp_flags &= ~(bit); \
1448 	INP_WUNLOCK(in6p); \
1449 } while (/*CONSTCOND*/ 0)
1450 #define OPTSET2292(bit) \
1451 do { \
1452 	INP_WLOCK(in6p); \
1453 	in6p->inp_flags |= IN6P_RFC2292; \
1454 	if (optval) \
1455 		in6p->inp_flags |= (bit); \
1456 	else \
1457 		in6p->inp_flags &= ~(bit); \
1458 	INP_WUNLOCK(in6p); \
1459 } while (/*CONSTCOND*/ 0)
1460 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1461 
1462 #define OPTSET2(bit, val) do {						\
1463 	INP_WLOCK(in6p);						\
1464 	if (val)							\
1465 		in6p->inp_flags2 |= bit;				\
1466 	else								\
1467 		in6p->inp_flags2 &= ~bit;				\
1468 	INP_WUNLOCK(in6p);						\
1469 } while (0)
1470 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1471 
1472 				case IPV6_RECVPKTINFO:
1473 					/* cannot mix with RFC2292 */
1474 					if (OPTBIT(IN6P_RFC2292)) {
1475 						error = EINVAL;
1476 						break;
1477 					}
1478 					OPTSET(IN6P_PKTINFO);
1479 					break;
1480 
1481 				case IPV6_HOPLIMIT:
1482 				{
1483 					struct ip6_pktopts **optp;
1484 
1485 					/* cannot mix with RFC2292 */
1486 					if (OPTBIT(IN6P_RFC2292)) {
1487 						error = EINVAL;
1488 						break;
1489 					}
1490 					optp = &in6p->in6p_outputopts;
1491 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1492 					    (u_char *)&optval, sizeof(optval),
1493 					    optp, (td != NULL) ? td->td_ucred :
1494 					    NULL, uproto);
1495 					break;
1496 				}
1497 
1498 				case IPV6_RECVHOPLIMIT:
1499 					/* cannot mix with RFC2292 */
1500 					if (OPTBIT(IN6P_RFC2292)) {
1501 						error = EINVAL;
1502 						break;
1503 					}
1504 					OPTSET(IN6P_HOPLIMIT);
1505 					break;
1506 
1507 				case IPV6_RECVHOPOPTS:
1508 					/* cannot mix with RFC2292 */
1509 					if (OPTBIT(IN6P_RFC2292)) {
1510 						error = EINVAL;
1511 						break;
1512 					}
1513 					OPTSET(IN6P_HOPOPTS);
1514 					break;
1515 
1516 				case IPV6_RECVDSTOPTS:
1517 					/* cannot mix with RFC2292 */
1518 					if (OPTBIT(IN6P_RFC2292)) {
1519 						error = EINVAL;
1520 						break;
1521 					}
1522 					OPTSET(IN6P_DSTOPTS);
1523 					break;
1524 
1525 				case IPV6_RECVRTHDRDSTOPTS:
1526 					/* cannot mix with RFC2292 */
1527 					if (OPTBIT(IN6P_RFC2292)) {
1528 						error = EINVAL;
1529 						break;
1530 					}
1531 					OPTSET(IN6P_RTHDRDSTOPTS);
1532 					break;
1533 
1534 				case IPV6_RECVRTHDR:
1535 					/* cannot mix with RFC2292 */
1536 					if (OPTBIT(IN6P_RFC2292)) {
1537 						error = EINVAL;
1538 						break;
1539 					}
1540 					OPTSET(IN6P_RTHDR);
1541 					break;
1542 
1543 				case IPV6_RECVPATHMTU:
1544 					/*
1545 					 * We ignore this option for TCP
1546 					 * sockets.
1547 					 * (RFC3542 leaves this case
1548 					 * unspecified.)
1549 					 */
1550 					if (uproto != IPPROTO_TCP)
1551 						OPTSET(IN6P_MTU);
1552 					break;
1553 
1554 				case IPV6_V6ONLY:
1555 					/*
1556 					 * make setsockopt(IPV6_V6ONLY)
1557 					 * available only prior to bind(2).
1558 					 * see ipng mailing list, Jun 22 2001.
1559 					 */
1560 					if (in6p->inp_lport ||
1561 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1562 						error = EINVAL;
1563 						break;
1564 					}
1565 					OPTSET(IN6P_IPV6_V6ONLY);
1566 					if (optval)
1567 						in6p->inp_vflag &= ~INP_IPV4;
1568 					else
1569 						in6p->inp_vflag |= INP_IPV4;
1570 					break;
1571 				case IPV6_RECVTCLASS:
1572 					/* cannot mix with RFC2292 XXX */
1573 					if (OPTBIT(IN6P_RFC2292)) {
1574 						error = EINVAL;
1575 						break;
1576 					}
1577 					OPTSET(IN6P_TCLASS);
1578 					break;
1579 				case IPV6_AUTOFLOWLABEL:
1580 					OPTSET(IN6P_AUTOFLOWLABEL);
1581 					break;
1582 
1583 				case IPV6_BINDANY:
1584 					OPTSET(INP_BINDANY);
1585 					break;
1586 
1587 				case IPV6_BINDMULTI:
1588 					OPTSET2(INP_BINDMULTI, optval);
1589 					break;
1590 #ifdef	RSS
1591 				case IPV6_RSS_LISTEN_BUCKET:
1592 					if ((optval >= 0) &&
1593 					    (optval < rss_getnumbuckets())) {
1594 						in6p->inp_rss_listen_bucket = optval;
1595 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1596 					} else {
1597 						error = EINVAL;
1598 					}
1599 					break;
1600 #endif
1601 				}
1602 				break;
1603 
1604 			case IPV6_TCLASS:
1605 			case IPV6_DONTFRAG:
1606 			case IPV6_USE_MIN_MTU:
1607 			case IPV6_PREFER_TEMPADDR:
1608 				if (optlen != sizeof(optval)) {
1609 					error = EINVAL;
1610 					break;
1611 				}
1612 				error = sooptcopyin(sopt, &optval,
1613 					sizeof optval, sizeof optval);
1614 				if (error)
1615 					break;
1616 				{
1617 					struct ip6_pktopts **optp;
1618 					optp = &in6p->in6p_outputopts;
1619 					error = ip6_pcbopt(optname,
1620 					    (u_char *)&optval, sizeof(optval),
1621 					    optp, (td != NULL) ? td->td_ucred :
1622 					    NULL, uproto);
1623 					break;
1624 				}
1625 
1626 			case IPV6_2292PKTINFO:
1627 			case IPV6_2292HOPLIMIT:
1628 			case IPV6_2292HOPOPTS:
1629 			case IPV6_2292DSTOPTS:
1630 			case IPV6_2292RTHDR:
1631 				/* RFC 2292 */
1632 				if (optlen != sizeof(int)) {
1633 					error = EINVAL;
1634 					break;
1635 				}
1636 				error = sooptcopyin(sopt, &optval,
1637 					sizeof optval, sizeof optval);
1638 				if (error)
1639 					break;
1640 				switch (optname) {
1641 				case IPV6_2292PKTINFO:
1642 					OPTSET2292(IN6P_PKTINFO);
1643 					break;
1644 				case IPV6_2292HOPLIMIT:
1645 					OPTSET2292(IN6P_HOPLIMIT);
1646 					break;
1647 				case IPV6_2292HOPOPTS:
1648 					/*
1649 					 * Check super-user privilege.
1650 					 * See comments for IPV6_RECVHOPOPTS.
1651 					 */
1652 					if (td != NULL) {
1653 						error = priv_check(td,
1654 						    PRIV_NETINET_SETHDROPTS);
1655 						if (error)
1656 							return (error);
1657 					}
1658 					OPTSET2292(IN6P_HOPOPTS);
1659 					break;
1660 				case IPV6_2292DSTOPTS:
1661 					if (td != NULL) {
1662 						error = priv_check(td,
1663 						    PRIV_NETINET_SETHDROPTS);
1664 						if (error)
1665 							return (error);
1666 					}
1667 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1668 					break;
1669 				case IPV6_2292RTHDR:
1670 					OPTSET2292(IN6P_RTHDR);
1671 					break;
1672 				}
1673 				break;
1674 			case IPV6_PKTINFO:
1675 			case IPV6_HOPOPTS:
1676 			case IPV6_RTHDR:
1677 			case IPV6_DSTOPTS:
1678 			case IPV6_RTHDRDSTOPTS:
1679 			case IPV6_NEXTHOP:
1680 			{
1681 				/* new advanced API (RFC3542) */
1682 				u_char *optbuf;
1683 				u_char optbuf_storage[MCLBYTES];
1684 				int optlen;
1685 				struct ip6_pktopts **optp;
1686 
1687 				/* cannot mix with RFC2292 */
1688 				if (OPTBIT(IN6P_RFC2292)) {
1689 					error = EINVAL;
1690 					break;
1691 				}
1692 
1693 				/*
1694 				 * We only ensure valsize is not too large
1695 				 * here.  Further validation will be done
1696 				 * later.
1697 				 */
1698 				error = sooptcopyin(sopt, optbuf_storage,
1699 				    sizeof(optbuf_storage), 0);
1700 				if (error)
1701 					break;
1702 				optlen = sopt->sopt_valsize;
1703 				optbuf = optbuf_storage;
1704 				optp = &in6p->in6p_outputopts;
1705 				error = ip6_pcbopt(optname, optbuf, optlen,
1706 				    optp, (td != NULL) ? td->td_ucred : NULL,
1707 				    uproto);
1708 				break;
1709 			}
1710 #undef OPTSET
1711 
1712 			case IPV6_MULTICAST_IF:
1713 			case IPV6_MULTICAST_HOPS:
1714 			case IPV6_MULTICAST_LOOP:
1715 			case IPV6_JOIN_GROUP:
1716 			case IPV6_LEAVE_GROUP:
1717 			case IPV6_MSFILTER:
1718 			case MCAST_BLOCK_SOURCE:
1719 			case MCAST_UNBLOCK_SOURCE:
1720 			case MCAST_JOIN_GROUP:
1721 			case MCAST_LEAVE_GROUP:
1722 			case MCAST_JOIN_SOURCE_GROUP:
1723 			case MCAST_LEAVE_SOURCE_GROUP:
1724 				error = ip6_setmoptions(in6p, sopt);
1725 				break;
1726 
1727 			case IPV6_PORTRANGE:
1728 				error = sooptcopyin(sopt, &optval,
1729 				    sizeof optval, sizeof optval);
1730 				if (error)
1731 					break;
1732 
1733 				INP_WLOCK(in6p);
1734 				switch (optval) {
1735 				case IPV6_PORTRANGE_DEFAULT:
1736 					in6p->inp_flags &= ~(INP_LOWPORT);
1737 					in6p->inp_flags &= ~(INP_HIGHPORT);
1738 					break;
1739 
1740 				case IPV6_PORTRANGE_HIGH:
1741 					in6p->inp_flags &= ~(INP_LOWPORT);
1742 					in6p->inp_flags |= INP_HIGHPORT;
1743 					break;
1744 
1745 				case IPV6_PORTRANGE_LOW:
1746 					in6p->inp_flags &= ~(INP_HIGHPORT);
1747 					in6p->inp_flags |= INP_LOWPORT;
1748 					break;
1749 
1750 				default:
1751 					error = EINVAL;
1752 					break;
1753 				}
1754 				INP_WUNLOCK(in6p);
1755 				break;
1756 
1757 #ifdef IPSEC
1758 			case IPV6_IPSEC_POLICY:
1759 			{
1760 				caddr_t req;
1761 				struct mbuf *m;
1762 
1763 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1764 					break;
1765 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1766 					break;
1767 				req = mtod(m, caddr_t);
1768 				error = ipsec_set_policy(in6p, optname, req,
1769 				    m->m_len, (sopt->sopt_td != NULL) ?
1770 				    sopt->sopt_td->td_ucred : NULL);
1771 				m_freem(m);
1772 				break;
1773 			}
1774 #endif /* IPSEC */
1775 
1776 			default:
1777 				error = ENOPROTOOPT;
1778 				break;
1779 			}
1780 			break;
1781 
1782 		case SOPT_GET:
1783 			switch (optname) {
1784 
1785 			case IPV6_2292PKTOPTIONS:
1786 #ifdef IPV6_PKTOPTIONS
1787 			case IPV6_PKTOPTIONS:
1788 #endif
1789 				/*
1790 				 * RFC3542 (effectively) deprecated the
1791 				 * semantics of the 2292-style pktoptions.
1792 				 * Since it was not reliable in nature (i.e.,
1793 				 * applications had to expect the lack of some
1794 				 * information after all), it would make sense
1795 				 * to simplify this part by always returning
1796 				 * empty data.
1797 				 */
1798 				sopt->sopt_valsize = 0;
1799 				break;
1800 
1801 			case IPV6_RECVHOPOPTS:
1802 			case IPV6_RECVDSTOPTS:
1803 			case IPV6_RECVRTHDRDSTOPTS:
1804 			case IPV6_UNICAST_HOPS:
1805 			case IPV6_RECVPKTINFO:
1806 			case IPV6_RECVHOPLIMIT:
1807 			case IPV6_RECVRTHDR:
1808 			case IPV6_RECVPATHMTU:
1809 
1810 			case IPV6_V6ONLY:
1811 			case IPV6_PORTRANGE:
1812 			case IPV6_RECVTCLASS:
1813 			case IPV6_AUTOFLOWLABEL:
1814 			case IPV6_BINDANY:
1815 			case IPV6_FLOWID:
1816 			case IPV6_FLOWTYPE:
1817 #ifdef	RSS
1818 			case IPV6_RSSBUCKETID:
1819 #endif
1820 				switch (optname) {
1821 
1822 				case IPV6_RECVHOPOPTS:
1823 					optval = OPTBIT(IN6P_HOPOPTS);
1824 					break;
1825 
1826 				case IPV6_RECVDSTOPTS:
1827 					optval = OPTBIT(IN6P_DSTOPTS);
1828 					break;
1829 
1830 				case IPV6_RECVRTHDRDSTOPTS:
1831 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1832 					break;
1833 
1834 				case IPV6_UNICAST_HOPS:
1835 					optval = in6p->in6p_hops;
1836 					break;
1837 
1838 				case IPV6_RECVPKTINFO:
1839 					optval = OPTBIT(IN6P_PKTINFO);
1840 					break;
1841 
1842 				case IPV6_RECVHOPLIMIT:
1843 					optval = OPTBIT(IN6P_HOPLIMIT);
1844 					break;
1845 
1846 				case IPV6_RECVRTHDR:
1847 					optval = OPTBIT(IN6P_RTHDR);
1848 					break;
1849 
1850 				case IPV6_RECVPATHMTU:
1851 					optval = OPTBIT(IN6P_MTU);
1852 					break;
1853 
1854 				case IPV6_V6ONLY:
1855 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1856 					break;
1857 
1858 				case IPV6_PORTRANGE:
1859 				    {
1860 					int flags;
1861 					flags = in6p->inp_flags;
1862 					if (flags & INP_HIGHPORT)
1863 						optval = IPV6_PORTRANGE_HIGH;
1864 					else if (flags & INP_LOWPORT)
1865 						optval = IPV6_PORTRANGE_LOW;
1866 					else
1867 						optval = 0;
1868 					break;
1869 				    }
1870 				case IPV6_RECVTCLASS:
1871 					optval = OPTBIT(IN6P_TCLASS);
1872 					break;
1873 
1874 				case IPV6_AUTOFLOWLABEL:
1875 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1876 					break;
1877 
1878 				case IPV6_BINDANY:
1879 					optval = OPTBIT(INP_BINDANY);
1880 					break;
1881 
1882 				case IPV6_FLOWID:
1883 					optval = in6p->inp_flowid;
1884 					break;
1885 
1886 				case IPV6_FLOWTYPE:
1887 					optval = in6p->inp_flowtype;
1888 					break;
1889 #ifdef	RSS
1890 				case IPV6_RSSBUCKETID:
1891 					retval =
1892 					    rss_hash2bucket(in6p->inp_flowid,
1893 					    in6p->inp_flowtype,
1894 					    &rss_bucket);
1895 					if (retval == 0)
1896 						optval = rss_bucket;
1897 					else
1898 						error = EINVAL;
1899 					break;
1900 #endif
1901 
1902 				case IPV6_BINDMULTI:
1903 					optval = OPTBIT2(INP_BINDMULTI);
1904 					break;
1905 
1906 				}
1907 				if (error)
1908 					break;
1909 				error = sooptcopyout(sopt, &optval,
1910 					sizeof optval);
1911 				break;
1912 
1913 			case IPV6_PATHMTU:
1914 			{
1915 				u_long pmtu = 0;
1916 				struct ip6_mtuinfo mtuinfo;
1917 				struct route_in6 sro;
1918 
1919 				bzero(&sro, sizeof(sro));
1920 
1921 				if (!(so->so_state & SS_ISCONNECTED))
1922 					return (ENOTCONN);
1923 				/*
1924 				 * XXX: we dot not consider the case of source
1925 				 * routing, or optional information to specify
1926 				 * the outgoing interface.
1927 				 */
1928 				error = ip6_getpmtu(&sro, NULL, NULL,
1929 				    &in6p->in6p_faddr, &pmtu, NULL,
1930 				    so->so_fibnum);
1931 				if (sro.ro_rt)
1932 					RTFREE(sro.ro_rt);
1933 				if (error)
1934 					break;
1935 				if (pmtu > IPV6_MAXPACKET)
1936 					pmtu = IPV6_MAXPACKET;
1937 
1938 				bzero(&mtuinfo, sizeof(mtuinfo));
1939 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1940 				optdata = (void *)&mtuinfo;
1941 				optdatalen = sizeof(mtuinfo);
1942 				error = sooptcopyout(sopt, optdata,
1943 				    optdatalen);
1944 				break;
1945 			}
1946 
1947 			case IPV6_2292PKTINFO:
1948 			case IPV6_2292HOPLIMIT:
1949 			case IPV6_2292HOPOPTS:
1950 			case IPV6_2292RTHDR:
1951 			case IPV6_2292DSTOPTS:
1952 				switch (optname) {
1953 				case IPV6_2292PKTINFO:
1954 					optval = OPTBIT(IN6P_PKTINFO);
1955 					break;
1956 				case IPV6_2292HOPLIMIT:
1957 					optval = OPTBIT(IN6P_HOPLIMIT);
1958 					break;
1959 				case IPV6_2292HOPOPTS:
1960 					optval = OPTBIT(IN6P_HOPOPTS);
1961 					break;
1962 				case IPV6_2292RTHDR:
1963 					optval = OPTBIT(IN6P_RTHDR);
1964 					break;
1965 				case IPV6_2292DSTOPTS:
1966 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1967 					break;
1968 				}
1969 				error = sooptcopyout(sopt, &optval,
1970 				    sizeof optval);
1971 				break;
1972 			case IPV6_PKTINFO:
1973 			case IPV6_HOPOPTS:
1974 			case IPV6_RTHDR:
1975 			case IPV6_DSTOPTS:
1976 			case IPV6_RTHDRDSTOPTS:
1977 			case IPV6_NEXTHOP:
1978 			case IPV6_TCLASS:
1979 			case IPV6_DONTFRAG:
1980 			case IPV6_USE_MIN_MTU:
1981 			case IPV6_PREFER_TEMPADDR:
1982 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1983 				    optname, sopt);
1984 				break;
1985 
1986 			case IPV6_MULTICAST_IF:
1987 			case IPV6_MULTICAST_HOPS:
1988 			case IPV6_MULTICAST_LOOP:
1989 			case IPV6_MSFILTER:
1990 				error = ip6_getmoptions(in6p, sopt);
1991 				break;
1992 
1993 #ifdef IPSEC
1994 			case IPV6_IPSEC_POLICY:
1995 			  {
1996 				caddr_t req = NULL;
1997 				size_t len = 0;
1998 				struct mbuf *m = NULL;
1999 				struct mbuf **mp = &m;
2000 				size_t ovalsize = sopt->sopt_valsize;
2001 				caddr_t oval = (caddr_t)sopt->sopt_val;
2002 
2003 				error = soopt_getm(sopt, &m); /* XXX */
2004 				if (error != 0)
2005 					break;
2006 				error = soopt_mcopyin(sopt, m); /* XXX */
2007 				if (error != 0)
2008 					break;
2009 				sopt->sopt_valsize = ovalsize;
2010 				sopt->sopt_val = oval;
2011 				if (m) {
2012 					req = mtod(m, caddr_t);
2013 					len = m->m_len;
2014 				}
2015 				error = ipsec_get_policy(in6p, req, len, mp);
2016 				if (error == 0)
2017 					error = soopt_mcopyout(sopt, m); /* XXX */
2018 				if (error == 0 && m)
2019 					m_freem(m);
2020 				break;
2021 			  }
2022 #endif /* IPSEC */
2023 
2024 			default:
2025 				error = ENOPROTOOPT;
2026 				break;
2027 			}
2028 			break;
2029 		}
2030 	}
2031 	return (error);
2032 }
2033 
2034 int
2035 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2036 {
2037 	int error = 0, optval, optlen;
2038 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2039 	struct inpcb *in6p = sotoinpcb(so);
2040 	int level, op, optname;
2041 
2042 	level = sopt->sopt_level;
2043 	op = sopt->sopt_dir;
2044 	optname = sopt->sopt_name;
2045 	optlen = sopt->sopt_valsize;
2046 
2047 	if (level != IPPROTO_IPV6) {
2048 		return (EINVAL);
2049 	}
2050 
2051 	switch (optname) {
2052 	case IPV6_CHECKSUM:
2053 		/*
2054 		 * For ICMPv6 sockets, no modification allowed for checksum
2055 		 * offset, permit "no change" values to help existing apps.
2056 		 *
2057 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2058 		 * for an ICMPv6 socket will fail."
2059 		 * The current behavior does not meet RFC3542.
2060 		 */
2061 		switch (op) {
2062 		case SOPT_SET:
2063 			if (optlen != sizeof(int)) {
2064 				error = EINVAL;
2065 				break;
2066 			}
2067 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2068 					    sizeof(optval));
2069 			if (error)
2070 				break;
2071 			if ((optval % 2) != 0) {
2072 				/* the API assumes even offset values */
2073 				error = EINVAL;
2074 			} else if (so->so_proto->pr_protocol ==
2075 			    IPPROTO_ICMPV6) {
2076 				if (optval != icmp6off)
2077 					error = EINVAL;
2078 			} else
2079 				in6p->in6p_cksum = optval;
2080 			break;
2081 
2082 		case SOPT_GET:
2083 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2084 				optval = icmp6off;
2085 			else
2086 				optval = in6p->in6p_cksum;
2087 
2088 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2089 			break;
2090 
2091 		default:
2092 			error = EINVAL;
2093 			break;
2094 		}
2095 		break;
2096 
2097 	default:
2098 		error = ENOPROTOOPT;
2099 		break;
2100 	}
2101 
2102 	return (error);
2103 }
2104 
2105 /*
2106  * Set up IP6 options in pcb for insertion in output packets or
2107  * specifying behavior of outgoing packets.
2108  */
2109 static int
2110 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2111     struct socket *so, struct sockopt *sopt)
2112 {
2113 	struct ip6_pktopts *opt = *pktopt;
2114 	int error = 0;
2115 	struct thread *td = sopt->sopt_td;
2116 
2117 	/* turn off any old options. */
2118 	if (opt) {
2119 #ifdef DIAGNOSTIC
2120 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2121 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2122 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2123 			printf("ip6_pcbopts: all specified options are cleared.\n");
2124 #endif
2125 		ip6_clearpktopts(opt, -1);
2126 	} else
2127 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2128 	*pktopt = NULL;
2129 
2130 	if (!m || m->m_len == 0) {
2131 		/*
2132 		 * Only turning off any previous options, regardless of
2133 		 * whether the opt is just created or given.
2134 		 */
2135 		free(opt, M_IP6OPT);
2136 		return (0);
2137 	}
2138 
2139 	/*  set options specified by user. */
2140 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2141 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2142 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2143 		free(opt, M_IP6OPT);
2144 		return (error);
2145 	}
2146 	*pktopt = opt;
2147 	return (0);
2148 }
2149 
2150 /*
2151  * initialize ip6_pktopts.  beware that there are non-zero default values in
2152  * the struct.
2153  */
2154 void
2155 ip6_initpktopts(struct ip6_pktopts *opt)
2156 {
2157 
2158 	bzero(opt, sizeof(*opt));
2159 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2160 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2161 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2162 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2163 }
2164 
2165 static int
2166 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2167     struct ucred *cred, int uproto)
2168 {
2169 	struct ip6_pktopts *opt;
2170 
2171 	if (*pktopt == NULL) {
2172 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2173 		    M_WAITOK);
2174 		ip6_initpktopts(*pktopt);
2175 	}
2176 	opt = *pktopt;
2177 
2178 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2179 }
2180 
2181 static int
2182 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2183 {
2184 	void *optdata = NULL;
2185 	int optdatalen = 0;
2186 	struct ip6_ext *ip6e;
2187 	int error = 0;
2188 	struct in6_pktinfo null_pktinfo;
2189 	int deftclass = 0, on;
2190 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2191 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2192 
2193 	switch (optname) {
2194 	case IPV6_PKTINFO:
2195 		if (pktopt && pktopt->ip6po_pktinfo)
2196 			optdata = (void *)pktopt->ip6po_pktinfo;
2197 		else {
2198 			/* XXX: we don't have to do this every time... */
2199 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2200 			optdata = (void *)&null_pktinfo;
2201 		}
2202 		optdatalen = sizeof(struct in6_pktinfo);
2203 		break;
2204 	case IPV6_TCLASS:
2205 		if (pktopt && pktopt->ip6po_tclass >= 0)
2206 			optdata = (void *)&pktopt->ip6po_tclass;
2207 		else
2208 			optdata = (void *)&deftclass;
2209 		optdatalen = sizeof(int);
2210 		break;
2211 	case IPV6_HOPOPTS:
2212 		if (pktopt && pktopt->ip6po_hbh) {
2213 			optdata = (void *)pktopt->ip6po_hbh;
2214 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2215 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2216 		}
2217 		break;
2218 	case IPV6_RTHDR:
2219 		if (pktopt && pktopt->ip6po_rthdr) {
2220 			optdata = (void *)pktopt->ip6po_rthdr;
2221 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2222 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2223 		}
2224 		break;
2225 	case IPV6_RTHDRDSTOPTS:
2226 		if (pktopt && pktopt->ip6po_dest1) {
2227 			optdata = (void *)pktopt->ip6po_dest1;
2228 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2229 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2230 		}
2231 		break;
2232 	case IPV6_DSTOPTS:
2233 		if (pktopt && pktopt->ip6po_dest2) {
2234 			optdata = (void *)pktopt->ip6po_dest2;
2235 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2236 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2237 		}
2238 		break;
2239 	case IPV6_NEXTHOP:
2240 		if (pktopt && pktopt->ip6po_nexthop) {
2241 			optdata = (void *)pktopt->ip6po_nexthop;
2242 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2243 		}
2244 		break;
2245 	case IPV6_USE_MIN_MTU:
2246 		if (pktopt)
2247 			optdata = (void *)&pktopt->ip6po_minmtu;
2248 		else
2249 			optdata = (void *)&defminmtu;
2250 		optdatalen = sizeof(int);
2251 		break;
2252 	case IPV6_DONTFRAG:
2253 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2254 			on = 1;
2255 		else
2256 			on = 0;
2257 		optdata = (void *)&on;
2258 		optdatalen = sizeof(on);
2259 		break;
2260 	case IPV6_PREFER_TEMPADDR:
2261 		if (pktopt)
2262 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2263 		else
2264 			optdata = (void *)&defpreftemp;
2265 		optdatalen = sizeof(int);
2266 		break;
2267 	default:		/* should not happen */
2268 #ifdef DIAGNOSTIC
2269 		panic("ip6_getpcbopt: unexpected option\n");
2270 #endif
2271 		return (ENOPROTOOPT);
2272 	}
2273 
2274 	error = sooptcopyout(sopt, optdata, optdatalen);
2275 
2276 	return (error);
2277 }
2278 
2279 void
2280 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2281 {
2282 	if (pktopt == NULL)
2283 		return;
2284 
2285 	if (optname == -1 || optname == IPV6_PKTINFO) {
2286 		if (pktopt->ip6po_pktinfo)
2287 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2288 		pktopt->ip6po_pktinfo = NULL;
2289 	}
2290 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2291 		pktopt->ip6po_hlim = -1;
2292 	if (optname == -1 || optname == IPV6_TCLASS)
2293 		pktopt->ip6po_tclass = -1;
2294 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2295 		if (pktopt->ip6po_nextroute.ro_rt) {
2296 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2297 			pktopt->ip6po_nextroute.ro_rt = NULL;
2298 		}
2299 		if (pktopt->ip6po_nexthop)
2300 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2301 		pktopt->ip6po_nexthop = NULL;
2302 	}
2303 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2304 		if (pktopt->ip6po_hbh)
2305 			free(pktopt->ip6po_hbh, M_IP6OPT);
2306 		pktopt->ip6po_hbh = NULL;
2307 	}
2308 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2309 		if (pktopt->ip6po_dest1)
2310 			free(pktopt->ip6po_dest1, M_IP6OPT);
2311 		pktopt->ip6po_dest1 = NULL;
2312 	}
2313 	if (optname == -1 || optname == IPV6_RTHDR) {
2314 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2315 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2316 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2317 		if (pktopt->ip6po_route.ro_rt) {
2318 			RTFREE(pktopt->ip6po_route.ro_rt);
2319 			pktopt->ip6po_route.ro_rt = NULL;
2320 		}
2321 	}
2322 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2323 		if (pktopt->ip6po_dest2)
2324 			free(pktopt->ip6po_dest2, M_IP6OPT);
2325 		pktopt->ip6po_dest2 = NULL;
2326 	}
2327 }
2328 
2329 #define PKTOPT_EXTHDRCPY(type) \
2330 do {\
2331 	if (src->type) {\
2332 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2333 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2334 		if (dst->type == NULL && canwait == M_NOWAIT)\
2335 			goto bad;\
2336 		bcopy(src->type, dst->type, hlen);\
2337 	}\
2338 } while (/*CONSTCOND*/ 0)
2339 
2340 static int
2341 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2342 {
2343 	if (dst == NULL || src == NULL)  {
2344 		printf("ip6_clearpktopts: invalid argument\n");
2345 		return (EINVAL);
2346 	}
2347 
2348 	dst->ip6po_hlim = src->ip6po_hlim;
2349 	dst->ip6po_tclass = src->ip6po_tclass;
2350 	dst->ip6po_flags = src->ip6po_flags;
2351 	dst->ip6po_minmtu = src->ip6po_minmtu;
2352 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2353 	if (src->ip6po_pktinfo) {
2354 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2355 		    M_IP6OPT, canwait);
2356 		if (dst->ip6po_pktinfo == NULL)
2357 			goto bad;
2358 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2359 	}
2360 	if (src->ip6po_nexthop) {
2361 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2362 		    M_IP6OPT, canwait);
2363 		if (dst->ip6po_nexthop == NULL)
2364 			goto bad;
2365 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2366 		    src->ip6po_nexthop->sa_len);
2367 	}
2368 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2369 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2370 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2371 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2372 	return (0);
2373 
2374   bad:
2375 	ip6_clearpktopts(dst, -1);
2376 	return (ENOBUFS);
2377 }
2378 #undef PKTOPT_EXTHDRCPY
2379 
2380 struct ip6_pktopts *
2381 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2382 {
2383 	int error;
2384 	struct ip6_pktopts *dst;
2385 
2386 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2387 	if (dst == NULL)
2388 		return (NULL);
2389 	ip6_initpktopts(dst);
2390 
2391 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2392 		free(dst, M_IP6OPT);
2393 		return (NULL);
2394 	}
2395 
2396 	return (dst);
2397 }
2398 
2399 void
2400 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2401 {
2402 	if (pktopt == NULL)
2403 		return;
2404 
2405 	ip6_clearpktopts(pktopt, -1);
2406 
2407 	free(pktopt, M_IP6OPT);
2408 }
2409 
2410 /*
2411  * Set IPv6 outgoing packet options based on advanced API.
2412  */
2413 int
2414 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2415     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2416 {
2417 	struct cmsghdr *cm = 0;
2418 
2419 	if (control == NULL || opt == NULL)
2420 		return (EINVAL);
2421 
2422 	ip6_initpktopts(opt);
2423 	if (stickyopt) {
2424 		int error;
2425 
2426 		/*
2427 		 * If stickyopt is provided, make a local copy of the options
2428 		 * for this particular packet, then override them by ancillary
2429 		 * objects.
2430 		 * XXX: copypktopts() does not copy the cached route to a next
2431 		 * hop (if any).  This is not very good in terms of efficiency,
2432 		 * but we can allow this since this option should be rarely
2433 		 * used.
2434 		 */
2435 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2436 			return (error);
2437 	}
2438 
2439 	/*
2440 	 * XXX: Currently, we assume all the optional information is stored
2441 	 * in a single mbuf.
2442 	 */
2443 	if (control->m_next)
2444 		return (EINVAL);
2445 
2446 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2447 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2448 		int error;
2449 
2450 		if (control->m_len < CMSG_LEN(0))
2451 			return (EINVAL);
2452 
2453 		cm = mtod(control, struct cmsghdr *);
2454 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2455 			return (EINVAL);
2456 		if (cm->cmsg_level != IPPROTO_IPV6)
2457 			continue;
2458 
2459 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2460 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2461 		if (error)
2462 			return (error);
2463 	}
2464 
2465 	return (0);
2466 }
2467 
2468 /*
2469  * Set a particular packet option, as a sticky option or an ancillary data
2470  * item.  "len" can be 0 only when it's a sticky option.
2471  * We have 4 cases of combination of "sticky" and "cmsg":
2472  * "sticky=0, cmsg=0": impossible
2473  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2474  * "sticky=1, cmsg=0": RFC3542 socket option
2475  * "sticky=1, cmsg=1": RFC2292 socket option
2476  */
2477 static int
2478 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2479     struct ucred *cred, int sticky, int cmsg, int uproto)
2480 {
2481 	int minmtupolicy, preftemp;
2482 	int error;
2483 
2484 	if (!sticky && !cmsg) {
2485 #ifdef DIAGNOSTIC
2486 		printf("ip6_setpktopt: impossible case\n");
2487 #endif
2488 		return (EINVAL);
2489 	}
2490 
2491 	/*
2492 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2493 	 * not be specified in the context of RFC3542.  Conversely,
2494 	 * RFC3542 types should not be specified in the context of RFC2292.
2495 	 */
2496 	if (!cmsg) {
2497 		switch (optname) {
2498 		case IPV6_2292PKTINFO:
2499 		case IPV6_2292HOPLIMIT:
2500 		case IPV6_2292NEXTHOP:
2501 		case IPV6_2292HOPOPTS:
2502 		case IPV6_2292DSTOPTS:
2503 		case IPV6_2292RTHDR:
2504 		case IPV6_2292PKTOPTIONS:
2505 			return (ENOPROTOOPT);
2506 		}
2507 	}
2508 	if (sticky && cmsg) {
2509 		switch (optname) {
2510 		case IPV6_PKTINFO:
2511 		case IPV6_HOPLIMIT:
2512 		case IPV6_NEXTHOP:
2513 		case IPV6_HOPOPTS:
2514 		case IPV6_DSTOPTS:
2515 		case IPV6_RTHDRDSTOPTS:
2516 		case IPV6_RTHDR:
2517 		case IPV6_USE_MIN_MTU:
2518 		case IPV6_DONTFRAG:
2519 		case IPV6_TCLASS:
2520 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2521 			return (ENOPROTOOPT);
2522 		}
2523 	}
2524 
2525 	switch (optname) {
2526 	case IPV6_2292PKTINFO:
2527 	case IPV6_PKTINFO:
2528 	{
2529 		struct ifnet *ifp = NULL;
2530 		struct in6_pktinfo *pktinfo;
2531 
2532 		if (len != sizeof(struct in6_pktinfo))
2533 			return (EINVAL);
2534 
2535 		pktinfo = (struct in6_pktinfo *)buf;
2536 
2537 		/*
2538 		 * An application can clear any sticky IPV6_PKTINFO option by
2539 		 * doing a "regular" setsockopt with ipi6_addr being
2540 		 * in6addr_any and ipi6_ifindex being zero.
2541 		 * [RFC 3542, Section 6]
2542 		 */
2543 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2544 		    pktinfo->ipi6_ifindex == 0 &&
2545 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2546 			ip6_clearpktopts(opt, optname);
2547 			break;
2548 		}
2549 
2550 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2551 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2552 			return (EINVAL);
2553 		}
2554 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2555 			return (EINVAL);
2556 		/* validate the interface index if specified. */
2557 		if (pktinfo->ipi6_ifindex > V_if_index)
2558 			 return (ENXIO);
2559 		if (pktinfo->ipi6_ifindex) {
2560 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2561 			if (ifp == NULL)
2562 				return (ENXIO);
2563 		}
2564 		if (ifp != NULL && (
2565 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2566 			return (ENETDOWN);
2567 
2568 		if (ifp != NULL &&
2569 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2570 			struct in6_ifaddr *ia;
2571 
2572 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2573 			if (ia == NULL)
2574 				return (EADDRNOTAVAIL);
2575 			ifa_free(&ia->ia_ifa);
2576 		}
2577 		/*
2578 		 * We store the address anyway, and let in6_selectsrc()
2579 		 * validate the specified address.  This is because ipi6_addr
2580 		 * may not have enough information about its scope zone, and
2581 		 * we may need additional information (such as outgoing
2582 		 * interface or the scope zone of a destination address) to
2583 		 * disambiguate the scope.
2584 		 * XXX: the delay of the validation may confuse the
2585 		 * application when it is used as a sticky option.
2586 		 */
2587 		if (opt->ip6po_pktinfo == NULL) {
2588 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2589 			    M_IP6OPT, M_NOWAIT);
2590 			if (opt->ip6po_pktinfo == NULL)
2591 				return (ENOBUFS);
2592 		}
2593 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2594 		break;
2595 	}
2596 
2597 	case IPV6_2292HOPLIMIT:
2598 	case IPV6_HOPLIMIT:
2599 	{
2600 		int *hlimp;
2601 
2602 		/*
2603 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2604 		 * to simplify the ordering among hoplimit options.
2605 		 */
2606 		if (optname == IPV6_HOPLIMIT && sticky)
2607 			return (ENOPROTOOPT);
2608 
2609 		if (len != sizeof(int))
2610 			return (EINVAL);
2611 		hlimp = (int *)buf;
2612 		if (*hlimp < -1 || *hlimp > 255)
2613 			return (EINVAL);
2614 
2615 		opt->ip6po_hlim = *hlimp;
2616 		break;
2617 	}
2618 
2619 	case IPV6_TCLASS:
2620 	{
2621 		int tclass;
2622 
2623 		if (len != sizeof(int))
2624 			return (EINVAL);
2625 		tclass = *(int *)buf;
2626 		if (tclass < -1 || tclass > 255)
2627 			return (EINVAL);
2628 
2629 		opt->ip6po_tclass = tclass;
2630 		break;
2631 	}
2632 
2633 	case IPV6_2292NEXTHOP:
2634 	case IPV6_NEXTHOP:
2635 		if (cred != NULL) {
2636 			error = priv_check_cred(cred,
2637 			    PRIV_NETINET_SETHDROPTS, 0);
2638 			if (error)
2639 				return (error);
2640 		}
2641 
2642 		if (len == 0) {	/* just remove the option */
2643 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2644 			break;
2645 		}
2646 
2647 		/* check if cmsg_len is large enough for sa_len */
2648 		if (len < sizeof(struct sockaddr) || len < *buf)
2649 			return (EINVAL);
2650 
2651 		switch (((struct sockaddr *)buf)->sa_family) {
2652 		case AF_INET6:
2653 		{
2654 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2655 			int error;
2656 
2657 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2658 				return (EINVAL);
2659 
2660 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2661 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2662 				return (EINVAL);
2663 			}
2664 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2665 			    != 0) {
2666 				return (error);
2667 			}
2668 			break;
2669 		}
2670 		case AF_LINK:	/* should eventually be supported */
2671 		default:
2672 			return (EAFNOSUPPORT);
2673 		}
2674 
2675 		/* turn off the previous option, then set the new option. */
2676 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2677 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2678 		if (opt->ip6po_nexthop == NULL)
2679 			return (ENOBUFS);
2680 		bcopy(buf, opt->ip6po_nexthop, *buf);
2681 		break;
2682 
2683 	case IPV6_2292HOPOPTS:
2684 	case IPV6_HOPOPTS:
2685 	{
2686 		struct ip6_hbh *hbh;
2687 		int hbhlen;
2688 
2689 		/*
2690 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2691 		 * options, since per-option restriction has too much
2692 		 * overhead.
2693 		 */
2694 		if (cred != NULL) {
2695 			error = priv_check_cred(cred,
2696 			    PRIV_NETINET_SETHDROPTS, 0);
2697 			if (error)
2698 				return (error);
2699 		}
2700 
2701 		if (len == 0) {
2702 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2703 			break;	/* just remove the option */
2704 		}
2705 
2706 		/* message length validation */
2707 		if (len < sizeof(struct ip6_hbh))
2708 			return (EINVAL);
2709 		hbh = (struct ip6_hbh *)buf;
2710 		hbhlen = (hbh->ip6h_len + 1) << 3;
2711 		if (len != hbhlen)
2712 			return (EINVAL);
2713 
2714 		/* turn off the previous option, then set the new option. */
2715 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2716 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2717 		if (opt->ip6po_hbh == NULL)
2718 			return (ENOBUFS);
2719 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2720 
2721 		break;
2722 	}
2723 
2724 	case IPV6_2292DSTOPTS:
2725 	case IPV6_DSTOPTS:
2726 	case IPV6_RTHDRDSTOPTS:
2727 	{
2728 		struct ip6_dest *dest, **newdest = NULL;
2729 		int destlen;
2730 
2731 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2732 			error = priv_check_cred(cred,
2733 			    PRIV_NETINET_SETHDROPTS, 0);
2734 			if (error)
2735 				return (error);
2736 		}
2737 
2738 		if (len == 0) {
2739 			ip6_clearpktopts(opt, optname);
2740 			break;	/* just remove the option */
2741 		}
2742 
2743 		/* message length validation */
2744 		if (len < sizeof(struct ip6_dest))
2745 			return (EINVAL);
2746 		dest = (struct ip6_dest *)buf;
2747 		destlen = (dest->ip6d_len + 1) << 3;
2748 		if (len != destlen)
2749 			return (EINVAL);
2750 
2751 		/*
2752 		 * Determine the position that the destination options header
2753 		 * should be inserted; before or after the routing header.
2754 		 */
2755 		switch (optname) {
2756 		case IPV6_2292DSTOPTS:
2757 			/*
2758 			 * The old advacned API is ambiguous on this point.
2759 			 * Our approach is to determine the position based
2760 			 * according to the existence of a routing header.
2761 			 * Note, however, that this depends on the order of the
2762 			 * extension headers in the ancillary data; the 1st
2763 			 * part of the destination options header must appear
2764 			 * before the routing header in the ancillary data,
2765 			 * too.
2766 			 * RFC3542 solved the ambiguity by introducing
2767 			 * separate ancillary data or option types.
2768 			 */
2769 			if (opt->ip6po_rthdr == NULL)
2770 				newdest = &opt->ip6po_dest1;
2771 			else
2772 				newdest = &opt->ip6po_dest2;
2773 			break;
2774 		case IPV6_RTHDRDSTOPTS:
2775 			newdest = &opt->ip6po_dest1;
2776 			break;
2777 		case IPV6_DSTOPTS:
2778 			newdest = &opt->ip6po_dest2;
2779 			break;
2780 		}
2781 
2782 		/* turn off the previous option, then set the new option. */
2783 		ip6_clearpktopts(opt, optname);
2784 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2785 		if (*newdest == NULL)
2786 			return (ENOBUFS);
2787 		bcopy(dest, *newdest, destlen);
2788 
2789 		break;
2790 	}
2791 
2792 	case IPV6_2292RTHDR:
2793 	case IPV6_RTHDR:
2794 	{
2795 		struct ip6_rthdr *rth;
2796 		int rthlen;
2797 
2798 		if (len == 0) {
2799 			ip6_clearpktopts(opt, IPV6_RTHDR);
2800 			break;	/* just remove the option */
2801 		}
2802 
2803 		/* message length validation */
2804 		if (len < sizeof(struct ip6_rthdr))
2805 			return (EINVAL);
2806 		rth = (struct ip6_rthdr *)buf;
2807 		rthlen = (rth->ip6r_len + 1) << 3;
2808 		if (len != rthlen)
2809 			return (EINVAL);
2810 
2811 		switch (rth->ip6r_type) {
2812 		case IPV6_RTHDR_TYPE_0:
2813 			if (rth->ip6r_len == 0)	/* must contain one addr */
2814 				return (EINVAL);
2815 			if (rth->ip6r_len % 2) /* length must be even */
2816 				return (EINVAL);
2817 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2818 				return (EINVAL);
2819 			break;
2820 		default:
2821 			return (EINVAL);	/* not supported */
2822 		}
2823 
2824 		/* turn off the previous option */
2825 		ip6_clearpktopts(opt, IPV6_RTHDR);
2826 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2827 		if (opt->ip6po_rthdr == NULL)
2828 			return (ENOBUFS);
2829 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2830 
2831 		break;
2832 	}
2833 
2834 	case IPV6_USE_MIN_MTU:
2835 		if (len != sizeof(int))
2836 			return (EINVAL);
2837 		minmtupolicy = *(int *)buf;
2838 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2839 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2840 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2841 			return (EINVAL);
2842 		}
2843 		opt->ip6po_minmtu = minmtupolicy;
2844 		break;
2845 
2846 	case IPV6_DONTFRAG:
2847 		if (len != sizeof(int))
2848 			return (EINVAL);
2849 
2850 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2851 			/*
2852 			 * we ignore this option for TCP sockets.
2853 			 * (RFC3542 leaves this case unspecified.)
2854 			 */
2855 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2856 		} else
2857 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2858 		break;
2859 
2860 	case IPV6_PREFER_TEMPADDR:
2861 		if (len != sizeof(int))
2862 			return (EINVAL);
2863 		preftemp = *(int *)buf;
2864 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2865 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2866 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2867 			return (EINVAL);
2868 		}
2869 		opt->ip6po_prefer_tempaddr = preftemp;
2870 		break;
2871 
2872 	default:
2873 		return (ENOPROTOOPT);
2874 	} /* end of switch */
2875 
2876 	return (0);
2877 }
2878 
2879 /*
2880  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2881  * packet to the input queue of a specified interface.  Note that this
2882  * calls the output routine of the loopback "driver", but with an interface
2883  * pointer that might NOT be &loif -- easier than replicating that code here.
2884  */
2885 void
2886 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2887 {
2888 	struct mbuf *copym;
2889 	struct ip6_hdr *ip6;
2890 
2891 	copym = m_copy(m, 0, M_COPYALL);
2892 	if (copym == NULL)
2893 		return;
2894 
2895 	/*
2896 	 * Make sure to deep-copy IPv6 header portion in case the data
2897 	 * is in an mbuf cluster, so that we can safely override the IPv6
2898 	 * header portion later.
2899 	 */
2900 	if (!M_WRITABLE(copym) ||
2901 	    copym->m_len < sizeof(struct ip6_hdr)) {
2902 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2903 		if (copym == NULL)
2904 			return;
2905 	}
2906 
2907 #ifdef DIAGNOSTIC
2908 	if (copym->m_len < sizeof(*ip6)) {
2909 		m_freem(copym);
2910 		return;
2911 	}
2912 #endif
2913 
2914 	ip6 = mtod(copym, struct ip6_hdr *);
2915 	/*
2916 	 * clear embedded scope identifiers if necessary.
2917 	 * in6_clearscope will touch the addresses only when necessary.
2918 	 */
2919 	in6_clearscope(&ip6->ip6_src);
2920 	in6_clearscope(&ip6->ip6_dst);
2921 
2922 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2923 }
2924 
2925 /*
2926  * Chop IPv6 header off from the payload.
2927  */
2928 static int
2929 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2930 {
2931 	struct mbuf *mh;
2932 	struct ip6_hdr *ip6;
2933 
2934 	ip6 = mtod(m, struct ip6_hdr *);
2935 	if (m->m_len > sizeof(*ip6)) {
2936 		mh = m_gethdr(M_NOWAIT, MT_DATA);
2937 		if (mh == NULL) {
2938 			m_freem(m);
2939 			return ENOBUFS;
2940 		}
2941 		m_move_pkthdr(mh, m);
2942 		MH_ALIGN(mh, sizeof(*ip6));
2943 		m->m_len -= sizeof(*ip6);
2944 		m->m_data += sizeof(*ip6);
2945 		mh->m_next = m;
2946 		m = mh;
2947 		m->m_len = sizeof(*ip6);
2948 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2949 	}
2950 	exthdrs->ip6e_ip6 = m;
2951 	return 0;
2952 }
2953 
2954 /*
2955  * Compute IPv6 extension header length.
2956  */
2957 int
2958 ip6_optlen(struct inpcb *in6p)
2959 {
2960 	int len;
2961 
2962 	if (!in6p->in6p_outputopts)
2963 		return 0;
2964 
2965 	len = 0;
2966 #define elen(x) \
2967     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2968 
2969 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2970 	if (in6p->in6p_outputopts->ip6po_rthdr)
2971 		/* dest1 is valid with rthdr only */
2972 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2973 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2974 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2975 	return len;
2976 #undef elen
2977 }
2978