xref: /freebsd/sys/netinet6/ip6_output.c (revision c8dfaf382fa6df9dc6fd1e1c3356e0c8bf607e6a)
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 	struct m_tag *fwd_tag = NULL;
259 
260 	ip6 = mtod(m, struct ip6_hdr *);
261 	if (ip6 == NULL) {
262 		printf ("ip6 is NULL");
263 		goto bad;
264 	}
265 
266 	if (inp != NULL) {
267 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
268 		if (((flags & IP_NODEFAULTFLOWID) == 0) &&
269 		(inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID))) {
270 			m->m_pkthdr.flowid = inp->inp_flowid;
271 			m->m_flags |= M_FLOWID;
272 		}
273 	}
274 
275 	finaldst = ip6->ip6_dst;
276 	bzero(&exthdrs, sizeof(exthdrs));
277 	if (opt) {
278 		/* Hop-by-Hop options header */
279 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
280 		/* Destination options header(1st part) */
281 		if (opt->ip6po_rthdr) {
282 			/*
283 			 * Destination options header(1st part)
284 			 * This only makes sense with a routing header.
285 			 * See Section 9.2 of RFC 3542.
286 			 * Disabling this part just for MIP6 convenience is
287 			 * a bad idea.  We need to think carefully about a
288 			 * way to make the advanced API coexist with MIP6
289 			 * options, which might automatically be inserted in
290 			 * the kernel.
291 			 */
292 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
293 		}
294 		/* Routing header */
295 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
296 		/* Destination options header(2nd part) */
297 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
298 	}
299 
300 #ifdef IPSEC
301 	/*
302 	 * IPSec checking which handles several cases.
303 	 * FAST IPSEC: We re-injected the packet.
304 	 */
305 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp))
306 	{
307 	case 1:                 /* Bad packet */
308 		goto freehdrs;
309 	case -1:                /* IPSec done */
310 		goto done;
311 	case 0:                 /* No IPSec */
312 	default:
313 		break;
314 	}
315 #endif /* IPSEC */
316 
317 	/*
318 	 * Calculate the total length of the extension header chain.
319 	 * Keep the length of the unfragmentable part for fragmentation.
320 	 */
321 	optlen = 0;
322 	if (exthdrs.ip6e_hbh)
323 		optlen += exthdrs.ip6e_hbh->m_len;
324 	if (exthdrs.ip6e_dest1)
325 		optlen += exthdrs.ip6e_dest1->m_len;
326 	if (exthdrs.ip6e_rthdr)
327 		optlen += exthdrs.ip6e_rthdr->m_len;
328 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
329 
330 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
331 	if (exthdrs.ip6e_dest2)
332 		optlen += exthdrs.ip6e_dest2->m_len;
333 
334 	/*
335 	 * If there is at least one extension header,
336 	 * separate IP6 header from the payload.
337 	 */
338 	if (optlen && !hdrsplit) {
339 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
340 			m = NULL;
341 			goto freehdrs;
342 		}
343 		m = exthdrs.ip6e_ip6;
344 		hdrsplit++;
345 	}
346 
347 	/* adjust pointer */
348 	ip6 = mtod(m, struct ip6_hdr *);
349 
350 	/* adjust mbuf packet header length */
351 	m->m_pkthdr.len += optlen;
352 	plen = m->m_pkthdr.len - sizeof(*ip6);
353 
354 	/* If this is a jumbo payload, insert a jumbo payload option. */
355 	if (plen > IPV6_MAXPACKET) {
356 		if (!hdrsplit) {
357 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
358 				m = NULL;
359 				goto freehdrs;
360 			}
361 			m = exthdrs.ip6e_ip6;
362 			hdrsplit++;
363 		}
364 		/* adjust pointer */
365 		ip6 = mtod(m, struct ip6_hdr *);
366 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
367 			goto freehdrs;
368 		ip6->ip6_plen = 0;
369 	} else
370 		ip6->ip6_plen = htons(plen);
371 
372 	/*
373 	 * Concatenate headers and fill in next header fields.
374 	 * Here we have, on "m"
375 	 *	IPv6 payload
376 	 * and we insert headers accordingly.  Finally, we should be getting:
377 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
378 	 *
379 	 * during the header composing process, "m" points to IPv6 header.
380 	 * "mprev" points to an extension header prior to esp.
381 	 */
382 	u_char *nexthdrp = &ip6->ip6_nxt;
383 	mprev = m;
384 
385 	/*
386 	 * we treat dest2 specially.  this makes IPsec processing
387 	 * much easier.  the goal here is to make mprev point the
388 	 * mbuf prior to dest2.
389 	 *
390 	 * result: IPv6 dest2 payload
391 	 * m and mprev will point to IPv6 header.
392 	 */
393 	if (exthdrs.ip6e_dest2) {
394 		if (!hdrsplit)
395 			panic("assumption failed: hdr not split");
396 		exthdrs.ip6e_dest2->m_next = m->m_next;
397 		m->m_next = exthdrs.ip6e_dest2;
398 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
399 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
400 	}
401 
402 	/*
403 	 * result: IPv6 hbh dest1 rthdr dest2 payload
404 	 * m will point to IPv6 header.  mprev will point to the
405 	 * extension header prior to dest2 (rthdr in the above case).
406 	 */
407 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
408 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
409 		   IPPROTO_DSTOPTS);
410 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
411 		   IPPROTO_ROUTING);
412 
413 	/*
414 	 * If there is a routing header, discard the packet.
415 	 */
416 	if (exthdrs.ip6e_rthdr) {
417 		 error = EINVAL;
418 		 goto bad;
419 	}
420 
421 	/* Source address validation */
422 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
423 	    (flags & IPV6_UNSPECSRC) == 0) {
424 		error = EOPNOTSUPP;
425 		IP6STAT_INC(ip6s_badscope);
426 		goto bad;
427 	}
428 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
429 		error = EOPNOTSUPP;
430 		IP6STAT_INC(ip6s_badscope);
431 		goto bad;
432 	}
433 
434 	IP6STAT_INC(ip6s_localout);
435 
436 	/*
437 	 * Route packet.
438 	 */
439 	if (ro == 0) {
440 		ro = &ip6route;
441 		bzero((caddr_t)ro, sizeof(*ro));
442 	}
443 	ro_pmtu = ro;
444 	if (opt && opt->ip6po_rthdr)
445 		ro = &opt->ip6po_route;
446 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
447 #ifdef FLOWTABLE
448 	if (ro->ro_rt == NULL)
449 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
450 #endif
451 again:
452 	/*
453 	 * if specified, try to fill in the traffic class field.
454 	 * do not override if a non-zero value is already set.
455 	 * we check the diffserv field and the ecn field separately.
456 	 */
457 	if (opt && opt->ip6po_tclass >= 0) {
458 		int mask = 0;
459 
460 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
461 			mask |= 0xfc;
462 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
463 			mask |= 0x03;
464 		if (mask != 0)
465 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
466 	}
467 
468 	/* fill in or override the hop limit field, if necessary. */
469 	if (opt && opt->ip6po_hlim != -1)
470 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
471 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
472 		if (im6o != NULL)
473 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
474 		else
475 			ip6->ip6_hlim = V_ip6_defmcasthlim;
476 	}
477 
478 	/* adjust pointer */
479 	ip6 = mtod(m, struct ip6_hdr *);
480 
481 	if (ro->ro_rt && fwd_tag == NULL) {
482 		rt = ro->ro_rt;
483 		ifp = ro->ro_rt->rt_ifp;
484 	} else {
485 		if (fwd_tag == NULL) {
486 			bzero(&dst_sa, sizeof(dst_sa));
487 			dst_sa.sin6_family = AF_INET6;
488 			dst_sa.sin6_len = sizeof(dst_sa);
489 			dst_sa.sin6_addr = ip6->ip6_dst;
490 		}
491 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
492 		    &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
493 		if (error != 0) {
494 			if (ifp != NULL)
495 				in6_ifstat_inc(ifp, ifs6_out_discard);
496 			goto bad;
497 		}
498 	}
499 	if (rt == NULL) {
500 		/*
501 		 * If in6_selectroute() does not return a route entry,
502 		 * dst may not have been updated.
503 		 */
504 		*dst = dst_sa;	/* XXX */
505 	}
506 
507 	/*
508 	 * then rt (for unicast) and ifp must be non-NULL valid values.
509 	 */
510 	if ((flags & IPV6_FORWARDING) == 0) {
511 		/* XXX: the FORWARDING flag can be set for mrouting. */
512 		in6_ifstat_inc(ifp, ifs6_out_request);
513 	}
514 	if (rt != NULL) {
515 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
516 		counter_u64_add(rt->rt_pksent, 1);
517 	}
518 
519 
520 	/*
521 	 * The outgoing interface must be in the zone of source and
522 	 * destination addresses.
523 	 */
524 	origifp = ifp;
525 
526 	src0 = ip6->ip6_src;
527 	if (in6_setscope(&src0, origifp, &zone))
528 		goto badscope;
529 	bzero(&src_sa, sizeof(src_sa));
530 	src_sa.sin6_family = AF_INET6;
531 	src_sa.sin6_len = sizeof(src_sa);
532 	src_sa.sin6_addr = ip6->ip6_src;
533 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
534 		goto badscope;
535 
536 	dst0 = ip6->ip6_dst;
537 	if (in6_setscope(&dst0, origifp, &zone))
538 		goto badscope;
539 	/* re-initialize to be sure */
540 	bzero(&dst_sa, sizeof(dst_sa));
541 	dst_sa.sin6_family = AF_INET6;
542 	dst_sa.sin6_len = sizeof(dst_sa);
543 	dst_sa.sin6_addr = ip6->ip6_dst;
544 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
545 		goto badscope;
546 	}
547 
548 	/* We should use ia_ifp to support the case of
549 	 * sending packets to an address of our own.
550 	 */
551 	if (ia != NULL && ia->ia_ifp)
552 		ifp = ia->ia_ifp;
553 
554 	/* scope check is done. */
555 	goto routefound;
556 
557   badscope:
558 	IP6STAT_INC(ip6s_badscope);
559 	in6_ifstat_inc(origifp, ifs6_out_discard);
560 	if (error == 0)
561 		error = EHOSTUNREACH; /* XXX */
562 	goto bad;
563 
564   routefound:
565 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
566 		if (opt && opt->ip6po_nextroute.ro_rt) {
567 			/*
568 			 * The nexthop is explicitly specified by the
569 			 * application.  We assume the next hop is an IPv6
570 			 * address.
571 			 */
572 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
573 		}
574 		else if ((rt->rt_flags & RTF_GATEWAY))
575 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
576 	}
577 
578 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
579 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
580 	} else {
581 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
582 		in6_ifstat_inc(ifp, ifs6_out_mcast);
583 		/*
584 		 * Confirm that the outgoing interface supports multicast.
585 		 */
586 		if (!(ifp->if_flags & IFF_MULTICAST)) {
587 			IP6STAT_INC(ip6s_noroute);
588 			in6_ifstat_inc(ifp, ifs6_out_discard);
589 			error = ENETUNREACH;
590 			goto bad;
591 		}
592 		if ((im6o == NULL && in6_mcast_loop) ||
593 		    (im6o && im6o->im6o_multicast_loop)) {
594 			/*
595 			 * Loop back multicast datagram if not expressly
596 			 * forbidden to do so, even if we have not joined
597 			 * the address; protocols will filter it later,
598 			 * thus deferring a hash lookup and lock acquisition
599 			 * at the expense of an m_copym().
600 			 */
601 			ip6_mloopback(ifp, m, dst);
602 		} else {
603 			/*
604 			 * If we are acting as a multicast router, perform
605 			 * multicast forwarding as if the packet had just
606 			 * arrived on the interface to which we are about
607 			 * to send.  The multicast forwarding function
608 			 * recursively calls this function, using the
609 			 * IPV6_FORWARDING flag to prevent infinite recursion.
610 			 *
611 			 * Multicasts that are looped back by ip6_mloopback(),
612 			 * above, will be forwarded by the ip6_input() routine,
613 			 * if necessary.
614 			 */
615 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
616 				/*
617 				 * XXX: ip6_mforward expects that rcvif is NULL
618 				 * when it is called from the originating path.
619 				 * However, it may not always be the case.
620 				 */
621 				m->m_pkthdr.rcvif = NULL;
622 				if (ip6_mforward(ip6, ifp, m) != 0) {
623 					m_freem(m);
624 					goto done;
625 				}
626 			}
627 		}
628 		/*
629 		 * Multicasts with a hoplimit of zero may be looped back,
630 		 * above, but must not be transmitted on a network.
631 		 * Also, multicasts addressed to the loopback interface
632 		 * are not sent -- the above call to ip6_mloopback() will
633 		 * loop back a copy if this host actually belongs to the
634 		 * destination group on the loopback interface.
635 		 */
636 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
637 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
638 			m_freem(m);
639 			goto done;
640 		}
641 	}
642 
643 	/*
644 	 * Fill the outgoing inteface to tell the upper layer
645 	 * to increment per-interface statistics.
646 	 */
647 	if (ifpp)
648 		*ifpp = ifp;
649 
650 	/* Determine path MTU. */
651 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
652 	    &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
653 		goto bad;
654 
655 	/*
656 	 * The caller of this function may specify to use the minimum MTU
657 	 * in some cases.
658 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
659 	 * setting.  The logic is a bit complicated; by default, unicast
660 	 * packets will follow path MTU while multicast packets will be sent at
661 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
662 	 * including unicast ones will be sent at the minimum MTU.  Multicast
663 	 * packets will always be sent at the minimum MTU unless
664 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
665 	 * See RFC 3542 for more details.
666 	 */
667 	if (mtu > IPV6_MMTU) {
668 		if ((flags & IPV6_MINMTU))
669 			mtu = IPV6_MMTU;
670 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
671 			mtu = IPV6_MMTU;
672 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
673 			 (opt == NULL ||
674 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
675 			mtu = IPV6_MMTU;
676 		}
677 	}
678 
679 	/*
680 	 * clear embedded scope identifiers if necessary.
681 	 * in6_clearscope will touch the addresses only when necessary.
682 	 */
683 	in6_clearscope(&ip6->ip6_src);
684 	in6_clearscope(&ip6->ip6_dst);
685 
686 	/*
687 	 * If the outgoing packet contains a hop-by-hop options header,
688 	 * it must be examined and processed even by the source node.
689 	 * (RFC 2460, section 4.)
690 	 */
691 	if (exthdrs.ip6e_hbh) {
692 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
693 		u_int32_t dummy; /* XXX unused */
694 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
695 
696 #ifdef DIAGNOSTIC
697 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
698 			panic("ip6e_hbh is not contiguous");
699 #endif
700 		/*
701 		 *  XXX: if we have to send an ICMPv6 error to the sender,
702 		 *       we need the M_LOOP flag since icmp6_error() expects
703 		 *       the IPv6 and the hop-by-hop options header are
704 		 *       contiguous unless the flag is set.
705 		 */
706 		m->m_flags |= M_LOOP;
707 		m->m_pkthdr.rcvif = ifp;
708 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
709 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
710 		    &dummy, &plen) < 0) {
711 			/* m was already freed at this point */
712 			error = EINVAL;/* better error? */
713 			goto done;
714 		}
715 		m->m_flags &= ~M_LOOP; /* XXX */
716 		m->m_pkthdr.rcvif = NULL;
717 	}
718 
719 	/* Jump over all PFIL processing if hooks are not active. */
720 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
721 		goto passout;
722 
723 	odst = ip6->ip6_dst;
724 	/* Run through list of hooks for output packets. */
725 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
726 	if (error != 0 || m == NULL)
727 		goto done;
728 	ip6 = mtod(m, struct ip6_hdr *);
729 
730 	/* See if destination IP address was changed by packet filter. */
731 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
732 		m->m_flags |= M_SKIP_FIREWALL;
733 		/* If destination is now ourself drop to ip6_input(). */
734 		if (in6_localip(&ip6->ip6_dst)) {
735 			m->m_flags |= M_FASTFWD_OURS;
736 			if (m->m_pkthdr.rcvif == NULL)
737 				m->m_pkthdr.rcvif = V_loif;
738 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
739 				m->m_pkthdr.csum_flags |=
740 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
741 				m->m_pkthdr.csum_data = 0xffff;
742 			}
743 #ifdef SCTP
744 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
745 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
746 #endif
747 			error = netisr_queue(NETISR_IPV6, m);
748 			goto done;
749 		} else
750 			goto again;	/* Redo the routing table lookup. */
751 	}
752 
753 	/* See if local, if yes, send it to netisr. */
754 	if (m->m_flags & M_FASTFWD_OURS) {
755 		if (m->m_pkthdr.rcvif == NULL)
756 			m->m_pkthdr.rcvif = V_loif;
757 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
758 			m->m_pkthdr.csum_flags |=
759 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
760 			m->m_pkthdr.csum_data = 0xffff;
761 		}
762 #ifdef SCTP
763 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
764 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
765 #endif
766 		error = netisr_queue(NETISR_IPV6, m);
767 		goto done;
768 	}
769 	/* Or forward to some other address? */
770 	if ((m->m_flags & M_IP6_NEXTHOP) &&
771 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
772 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
773 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
774 		m->m_flags |= M_SKIP_FIREWALL;
775 		m->m_flags &= ~M_IP6_NEXTHOP;
776 		m_tag_delete(m, fwd_tag);
777 		goto again;
778 	}
779 
780 passout:
781 	/*
782 	 * Send the packet to the outgoing interface.
783 	 * If necessary, do IPv6 fragmentation before sending.
784 	 *
785 	 * the logic here is rather complex:
786 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
787 	 * 1-a:	send as is if tlen <= path mtu
788 	 * 1-b:	fragment if tlen > path mtu
789 	 *
790 	 * 2: if user asks us not to fragment (dontfrag == 1)
791 	 * 2-a:	send as is if tlen <= interface mtu
792 	 * 2-b:	error if tlen > interface mtu
793 	 *
794 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
795 	 *	always fragment
796 	 *
797 	 * 4: if dontfrag == 1 && alwaysfrag == 1
798 	 *	error, as we cannot handle this conflicting request
799 	 */
800 	sw_csum = m->m_pkthdr.csum_flags;
801 	if (!hdrsplit) {
802 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
803 		sw_csum &= ~ifp->if_hwassist;
804 	} else
805 		tso = 0;
806 	/*
807 	 * If we added extension headers, we will not do TSO and calculate the
808 	 * checksums ourselves for now.
809 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
810 	 * with ext. hdrs.
811 	 */
812 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
813 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
814 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
815 	}
816 #ifdef SCTP
817 	if (sw_csum & CSUM_SCTP_IPV6) {
818 		sw_csum &= ~CSUM_SCTP_IPV6;
819 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
820 	}
821 #endif
822 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
823 	tlen = m->m_pkthdr.len;
824 
825 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
826 		dontfrag = 1;
827 	else
828 		dontfrag = 0;
829 	if (dontfrag && alwaysfrag) {	/* case 4 */
830 		/* conflicting request - can't transmit */
831 		error = EMSGSIZE;
832 		goto bad;
833 	}
834 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
835 		/*
836 		 * Even if the DONTFRAG option is specified, we cannot send the
837 		 * packet when the data length is larger than the MTU of the
838 		 * outgoing interface.
839 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
840 		 * well as returning an error code (the latter is not described
841 		 * in the API spec.)
842 		 */
843 		u_int32_t mtu32;
844 		struct ip6ctlparam ip6cp;
845 
846 		mtu32 = (u_int32_t)mtu;
847 		bzero(&ip6cp, sizeof(ip6cp));
848 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
849 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
850 		    (void *)&ip6cp);
851 
852 		error = EMSGSIZE;
853 		goto bad;
854 	}
855 
856 	/*
857 	 * transmit packet without fragmentation
858 	 */
859 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
860 		struct in6_ifaddr *ia6;
861 
862 		ip6 = mtod(m, struct ip6_hdr *);
863 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
864 		if (ia6) {
865 			/* Record statistics for this interface address. */
866 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
867 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
868 			    m->m_pkthdr.len);
869 			ifa_free(&ia6->ia_ifa);
870 		}
871 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
872 		goto done;
873 	}
874 
875 	/*
876 	 * try to fragment the packet.  case 1-b and 3
877 	 */
878 	if (mtu < IPV6_MMTU) {
879 		/* path MTU cannot be less than IPV6_MMTU */
880 		error = EMSGSIZE;
881 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
882 		goto bad;
883 	} else if (ip6->ip6_plen == 0) {
884 		/* jumbo payload cannot be fragmented */
885 		error = EMSGSIZE;
886 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
887 		goto bad;
888 	} else {
889 		struct mbuf **mnext, *m_frgpart;
890 		struct ip6_frag *ip6f;
891 		u_int32_t id = htonl(ip6_randomid());
892 		u_char nextproto;
893 
894 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
895 
896 		/*
897 		 * Too large for the destination or interface;
898 		 * fragment if possible.
899 		 * Must be able to put at least 8 bytes per fragment.
900 		 */
901 		hlen = unfragpartlen;
902 		if (mtu > IPV6_MAXPACKET)
903 			mtu = IPV6_MAXPACKET;
904 
905 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
906 		if (len < 8) {
907 			error = EMSGSIZE;
908 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
909 			goto bad;
910 		}
911 
912 		/*
913 		 * Verify that we have any chance at all of being able to queue
914 		 *      the packet or packet fragments
915 		 */
916 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
917 		    < tlen  /* - hlen */)) {
918 			error = ENOBUFS;
919 			IP6STAT_INC(ip6s_odropped);
920 			goto bad;
921 		}
922 
923 
924 		/*
925 		 * If the interface will not calculate checksums on
926 		 * fragmented packets, then do it here.
927 		 * XXX-BZ handle the hw offloading case.  Need flags.
928 		 */
929 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
930 			in6_delayed_cksum(m, plen, hlen);
931 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
932 		}
933 #ifdef SCTP
934 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
935 			sctp_delayed_cksum(m, hlen);
936 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
937 		}
938 #endif
939 		mnext = &m->m_nextpkt;
940 
941 		/*
942 		 * Change the next header field of the last header in the
943 		 * unfragmentable part.
944 		 */
945 		if (exthdrs.ip6e_rthdr) {
946 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
947 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
948 		} else if (exthdrs.ip6e_dest1) {
949 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
950 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
951 		} else if (exthdrs.ip6e_hbh) {
952 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
953 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
954 		} else {
955 			nextproto = ip6->ip6_nxt;
956 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
957 		}
958 
959 		/*
960 		 * Loop through length of segment after first fragment,
961 		 * make new header and copy data of each part and link onto
962 		 * chain.
963 		 */
964 		m0 = m;
965 		for (off = hlen; off < tlen; off += len) {
966 			m = m_gethdr(M_NOWAIT, MT_DATA);
967 			if (!m) {
968 				error = ENOBUFS;
969 				IP6STAT_INC(ip6s_odropped);
970 				goto sendorfree;
971 			}
972 			m->m_flags = m0->m_flags & M_COPYFLAGS;
973 			*mnext = m;
974 			mnext = &m->m_nextpkt;
975 			m->m_data += max_linkhdr;
976 			mhip6 = mtod(m, struct ip6_hdr *);
977 			*mhip6 = *ip6;
978 			m->m_len = sizeof(*mhip6);
979 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
980 			if (error) {
981 				IP6STAT_INC(ip6s_odropped);
982 				goto sendorfree;
983 			}
984 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
985 			if (off + len >= tlen)
986 				len = tlen - off;
987 			else
988 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
989 			mhip6->ip6_plen = htons((u_short)(len + hlen +
990 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
991 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
992 				error = ENOBUFS;
993 				IP6STAT_INC(ip6s_odropped);
994 				goto sendorfree;
995 			}
996 			m_cat(m, m_frgpart);
997 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
998 			m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
999 			m->m_pkthdr.rcvif = NULL;
1000 			ip6f->ip6f_reserved = 0;
1001 			ip6f->ip6f_ident = id;
1002 			ip6f->ip6f_nxt = nextproto;
1003 			IP6STAT_INC(ip6s_ofragments);
1004 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1005 		}
1006 
1007 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1008 	}
1009 
1010 	/*
1011 	 * Remove leading garbages.
1012 	 */
1013 sendorfree:
1014 	m = m0->m_nextpkt;
1015 	m0->m_nextpkt = 0;
1016 	m_freem(m0);
1017 	for (m0 = m; m; m = m0) {
1018 		m0 = m->m_nextpkt;
1019 		m->m_nextpkt = 0;
1020 		if (error == 0) {
1021 			/* Record statistics for this interface address. */
1022 			if (ia) {
1023 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1024 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1025 				    m->m_pkthdr.len);
1026 			}
1027 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1028 		} else
1029 			m_freem(m);
1030 	}
1031 
1032 	if (error == 0)
1033 		IP6STAT_INC(ip6s_fragmented);
1034 
1035 done:
1036 	if (ro == &ip6route)
1037 		RO_RTFREE(ro);
1038 	if (ro_pmtu == &ip6route)
1039 		RO_RTFREE(ro_pmtu);
1040 	return (error);
1041 
1042 freehdrs:
1043 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1044 	m_freem(exthdrs.ip6e_dest1);
1045 	m_freem(exthdrs.ip6e_rthdr);
1046 	m_freem(exthdrs.ip6e_dest2);
1047 	/* FALLTHROUGH */
1048 bad:
1049 	if (m)
1050 		m_freem(m);
1051 	goto done;
1052 }
1053 
1054 static int
1055 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1056 {
1057 	struct mbuf *m;
1058 
1059 	if (hlen > MCLBYTES)
1060 		return (ENOBUFS); /* XXX */
1061 
1062 	if (hlen > MLEN)
1063 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1064 	else
1065 		m = m_get(M_NOWAIT, MT_DATA);
1066 	if (m == NULL)
1067 		return (ENOBUFS);
1068 	m->m_len = hlen;
1069 	if (hdr)
1070 		bcopy(hdr, mtod(m, caddr_t), hlen);
1071 
1072 	*mp = m;
1073 	return (0);
1074 }
1075 
1076 /*
1077  * Insert jumbo payload option.
1078  */
1079 static int
1080 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1081 {
1082 	struct mbuf *mopt;
1083 	u_char *optbuf;
1084 	u_int32_t v;
1085 
1086 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1087 
1088 	/*
1089 	 * If there is no hop-by-hop options header, allocate new one.
1090 	 * If there is one but it doesn't have enough space to store the
1091 	 * jumbo payload option, allocate a cluster to store the whole options.
1092 	 * Otherwise, use it to store the options.
1093 	 */
1094 	if (exthdrs->ip6e_hbh == 0) {
1095 		mopt = m_get(M_NOWAIT, MT_DATA);
1096 		if (mopt == NULL)
1097 			return (ENOBUFS);
1098 		mopt->m_len = JUMBOOPTLEN;
1099 		optbuf = mtod(mopt, u_char *);
1100 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1101 		exthdrs->ip6e_hbh = mopt;
1102 	} else {
1103 		struct ip6_hbh *hbh;
1104 
1105 		mopt = exthdrs->ip6e_hbh;
1106 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1107 			/*
1108 			 * XXX assumption:
1109 			 * - exthdrs->ip6e_hbh is not referenced from places
1110 			 *   other than exthdrs.
1111 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1112 			 */
1113 			int oldoptlen = mopt->m_len;
1114 			struct mbuf *n;
1115 
1116 			/*
1117 			 * XXX: give up if the whole (new) hbh header does
1118 			 * not fit even in an mbuf cluster.
1119 			 */
1120 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1121 				return (ENOBUFS);
1122 
1123 			/*
1124 			 * As a consequence, we must always prepare a cluster
1125 			 * at this point.
1126 			 */
1127 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1128 			if (n == NULL)
1129 				return (ENOBUFS);
1130 			n->m_len = oldoptlen + JUMBOOPTLEN;
1131 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1132 			    oldoptlen);
1133 			optbuf = mtod(n, caddr_t) + oldoptlen;
1134 			m_freem(mopt);
1135 			mopt = exthdrs->ip6e_hbh = n;
1136 		} else {
1137 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1138 			mopt->m_len += JUMBOOPTLEN;
1139 		}
1140 		optbuf[0] = IP6OPT_PADN;
1141 		optbuf[1] = 1;
1142 
1143 		/*
1144 		 * Adjust the header length according to the pad and
1145 		 * the jumbo payload option.
1146 		 */
1147 		hbh = mtod(mopt, struct ip6_hbh *);
1148 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1149 	}
1150 
1151 	/* fill in the option. */
1152 	optbuf[2] = IP6OPT_JUMBO;
1153 	optbuf[3] = 4;
1154 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1155 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1156 
1157 	/* finally, adjust the packet header length */
1158 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1159 
1160 	return (0);
1161 #undef JUMBOOPTLEN
1162 }
1163 
1164 /*
1165  * Insert fragment header and copy unfragmentable header portions.
1166  */
1167 static int
1168 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1169     struct ip6_frag **frghdrp)
1170 {
1171 	struct mbuf *n, *mlast;
1172 
1173 	if (hlen > sizeof(struct ip6_hdr)) {
1174 		n = m_copym(m0, sizeof(struct ip6_hdr),
1175 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1176 		if (n == 0)
1177 			return (ENOBUFS);
1178 		m->m_next = n;
1179 	} else
1180 		n = m;
1181 
1182 	/* Search for the last mbuf of unfragmentable part. */
1183 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1184 		;
1185 
1186 	if ((mlast->m_flags & M_EXT) == 0 &&
1187 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1188 		/* use the trailing space of the last mbuf for the fragment hdr */
1189 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1190 		    mlast->m_len);
1191 		mlast->m_len += sizeof(struct ip6_frag);
1192 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1193 	} else {
1194 		/* allocate a new mbuf for the fragment header */
1195 		struct mbuf *mfrg;
1196 
1197 		mfrg = m_get(M_NOWAIT, MT_DATA);
1198 		if (mfrg == NULL)
1199 			return (ENOBUFS);
1200 		mfrg->m_len = sizeof(struct ip6_frag);
1201 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1202 		mlast->m_next = mfrg;
1203 	}
1204 
1205 	return (0);
1206 }
1207 
1208 static int
1209 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1210     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1211     int *alwaysfragp, u_int fibnum)
1212 {
1213 	u_int32_t mtu = 0;
1214 	int alwaysfrag = 0;
1215 	int error = 0;
1216 
1217 	if (ro_pmtu != ro) {
1218 		/* The first hop and the final destination may differ. */
1219 		struct sockaddr_in6 *sa6_dst =
1220 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1221 		if (ro_pmtu->ro_rt &&
1222 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1223 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1224 			RTFREE(ro_pmtu->ro_rt);
1225 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1226 		}
1227 		if (ro_pmtu->ro_rt == NULL) {
1228 			bzero(sa6_dst, sizeof(*sa6_dst));
1229 			sa6_dst->sin6_family = AF_INET6;
1230 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1231 			sa6_dst->sin6_addr = *dst;
1232 
1233 			in6_rtalloc(ro_pmtu, fibnum);
1234 		}
1235 	}
1236 	if (ro_pmtu->ro_rt) {
1237 		u_int32_t ifmtu;
1238 		struct in_conninfo inc;
1239 
1240 		bzero(&inc, sizeof(inc));
1241 		inc.inc_flags |= INC_ISIPV6;
1242 		inc.inc6_faddr = *dst;
1243 
1244 		if (ifp == NULL)
1245 			ifp = ro_pmtu->ro_rt->rt_ifp;
1246 		ifmtu = IN6_LINKMTU(ifp);
1247 		mtu = tcp_hc_getmtu(&inc);
1248 		if (mtu)
1249 			mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1250 		else
1251 			mtu = ro_pmtu->ro_rt->rt_mtu;
1252 		if (mtu == 0)
1253 			mtu = ifmtu;
1254 		else if (mtu < IPV6_MMTU) {
1255 			/*
1256 			 * RFC2460 section 5, last paragraph:
1257 			 * if we record ICMPv6 too big message with
1258 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1259 			 * or smaller, with framgent header attached.
1260 			 * (fragment header is needed regardless from the
1261 			 * packet size, for translators to identify packets)
1262 			 */
1263 			alwaysfrag = 1;
1264 			mtu = IPV6_MMTU;
1265 		} else if (mtu > ifmtu) {
1266 			/*
1267 			 * The MTU on the route is larger than the MTU on
1268 			 * the interface!  This shouldn't happen, unless the
1269 			 * MTU of the interface has been changed after the
1270 			 * interface was brought up.  Change the MTU in the
1271 			 * route to match the interface MTU (as long as the
1272 			 * field isn't locked).
1273 			 */
1274 			mtu = ifmtu;
1275 			ro_pmtu->ro_rt->rt_mtu = mtu;
1276 		}
1277 	} else if (ifp) {
1278 		mtu = IN6_LINKMTU(ifp);
1279 	} else
1280 		error = EHOSTUNREACH; /* XXX */
1281 
1282 	*mtup = mtu;
1283 	if (alwaysfragp)
1284 		*alwaysfragp = alwaysfrag;
1285 	return (error);
1286 }
1287 
1288 /*
1289  * IP6 socket option processing.
1290  */
1291 int
1292 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1293 {
1294 	int optdatalen, uproto;
1295 	void *optdata;
1296 	struct inpcb *in6p = sotoinpcb(so);
1297 	int error, optval;
1298 	int level, op, optname;
1299 	int optlen;
1300 	struct thread *td;
1301 #ifdef	RSS
1302 	uint32_t rss_bucket;
1303 	int retval;
1304 #endif
1305 
1306 	level = sopt->sopt_level;
1307 	op = sopt->sopt_dir;
1308 	optname = sopt->sopt_name;
1309 	optlen = sopt->sopt_valsize;
1310 	td = sopt->sopt_td;
1311 	error = 0;
1312 	optval = 0;
1313 	uproto = (int)so->so_proto->pr_protocol;
1314 
1315 	if (level != IPPROTO_IPV6) {
1316 		error = EINVAL;
1317 
1318 		if (sopt->sopt_level == SOL_SOCKET &&
1319 		    sopt->sopt_dir == SOPT_SET) {
1320 			switch (sopt->sopt_name) {
1321 			case SO_REUSEADDR:
1322 				INP_WLOCK(in6p);
1323 				if ((so->so_options & SO_REUSEADDR) != 0)
1324 					in6p->inp_flags2 |= INP_REUSEADDR;
1325 				else
1326 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1327 				INP_WUNLOCK(in6p);
1328 				error = 0;
1329 				break;
1330 			case SO_REUSEPORT:
1331 				INP_WLOCK(in6p);
1332 				if ((so->so_options & SO_REUSEPORT) != 0)
1333 					in6p->inp_flags2 |= INP_REUSEPORT;
1334 				else
1335 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1336 				INP_WUNLOCK(in6p);
1337 				error = 0;
1338 				break;
1339 			case SO_SETFIB:
1340 				INP_WLOCK(in6p);
1341 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1342 				INP_WUNLOCK(in6p);
1343 				error = 0;
1344 				break;
1345 			default:
1346 				break;
1347 			}
1348 		}
1349 	} else {		/* level == IPPROTO_IPV6 */
1350 		switch (op) {
1351 
1352 		case SOPT_SET:
1353 			switch (optname) {
1354 			case IPV6_2292PKTOPTIONS:
1355 #ifdef IPV6_PKTOPTIONS
1356 			case IPV6_PKTOPTIONS:
1357 #endif
1358 			{
1359 				struct mbuf *m;
1360 
1361 				error = soopt_getm(sopt, &m); /* XXX */
1362 				if (error != 0)
1363 					break;
1364 				error = soopt_mcopyin(sopt, m); /* XXX */
1365 				if (error != 0)
1366 					break;
1367 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1368 						    m, so, sopt);
1369 				m_freem(m); /* XXX */
1370 				break;
1371 			}
1372 
1373 			/*
1374 			 * Use of some Hop-by-Hop options or some
1375 			 * Destination options, might require special
1376 			 * privilege.  That is, normal applications
1377 			 * (without special privilege) might be forbidden
1378 			 * from setting certain options in outgoing packets,
1379 			 * and might never see certain options in received
1380 			 * packets. [RFC 2292 Section 6]
1381 			 * KAME specific note:
1382 			 *  KAME prevents non-privileged users from sending or
1383 			 *  receiving ANY hbh/dst options in order to avoid
1384 			 *  overhead of parsing options in the kernel.
1385 			 */
1386 			case IPV6_RECVHOPOPTS:
1387 			case IPV6_RECVDSTOPTS:
1388 			case IPV6_RECVRTHDRDSTOPTS:
1389 				if (td != NULL) {
1390 					error = priv_check(td,
1391 					    PRIV_NETINET_SETHDROPTS);
1392 					if (error)
1393 						break;
1394 				}
1395 				/* FALLTHROUGH */
1396 			case IPV6_UNICAST_HOPS:
1397 			case IPV6_HOPLIMIT:
1398 			case IPV6_FAITH:
1399 
1400 			case IPV6_RECVPKTINFO:
1401 			case IPV6_RECVHOPLIMIT:
1402 			case IPV6_RECVRTHDR:
1403 			case IPV6_RECVPATHMTU:
1404 			case IPV6_RECVTCLASS:
1405 			case IPV6_V6ONLY:
1406 			case IPV6_AUTOFLOWLABEL:
1407 			case IPV6_BINDANY:
1408 			case IPV6_BINDMULTI:
1409 #ifdef	RSS
1410 			case IPV6_RSS_LISTEN_BUCKET:
1411 #endif
1412 				if (optname == IPV6_BINDANY && td != NULL) {
1413 					error = priv_check(td,
1414 					    PRIV_NETINET_BINDANY);
1415 					if (error)
1416 						break;
1417 				}
1418 
1419 				if (optlen != sizeof(int)) {
1420 					error = EINVAL;
1421 					break;
1422 				}
1423 				error = sooptcopyin(sopt, &optval,
1424 					sizeof optval, sizeof optval);
1425 				if (error)
1426 					break;
1427 				switch (optname) {
1428 
1429 				case IPV6_UNICAST_HOPS:
1430 					if (optval < -1 || optval >= 256)
1431 						error = EINVAL;
1432 					else {
1433 						/* -1 = kernel default */
1434 						in6p->in6p_hops = optval;
1435 						if ((in6p->inp_vflag &
1436 						     INP_IPV4) != 0)
1437 							in6p->inp_ip_ttl = optval;
1438 					}
1439 					break;
1440 #define OPTSET(bit) \
1441 do { \
1442 	INP_WLOCK(in6p); \
1443 	if (optval) \
1444 		in6p->inp_flags |= (bit); \
1445 	else \
1446 		in6p->inp_flags &= ~(bit); \
1447 	INP_WUNLOCK(in6p); \
1448 } while (/*CONSTCOND*/ 0)
1449 #define OPTSET2292(bit) \
1450 do { \
1451 	INP_WLOCK(in6p); \
1452 	in6p->inp_flags |= IN6P_RFC2292; \
1453 	if (optval) \
1454 		in6p->inp_flags |= (bit); \
1455 	else \
1456 		in6p->inp_flags &= ~(bit); \
1457 	INP_WUNLOCK(in6p); \
1458 } while (/*CONSTCOND*/ 0)
1459 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1460 
1461 #define OPTSET2(bit, val) do {						\
1462 	INP_WLOCK(in6p);						\
1463 	if (val)							\
1464 		in6p->inp_flags2 |= bit;				\
1465 	else								\
1466 		in6p->inp_flags2 &= ~bit;				\
1467 	INP_WUNLOCK(in6p);						\
1468 } while (0)
1469 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1470 
1471 				case IPV6_RECVPKTINFO:
1472 					/* cannot mix with RFC2292 */
1473 					if (OPTBIT(IN6P_RFC2292)) {
1474 						error = EINVAL;
1475 						break;
1476 					}
1477 					OPTSET(IN6P_PKTINFO);
1478 					break;
1479 
1480 				case IPV6_HOPLIMIT:
1481 				{
1482 					struct ip6_pktopts **optp;
1483 
1484 					/* cannot mix with RFC2292 */
1485 					if (OPTBIT(IN6P_RFC2292)) {
1486 						error = EINVAL;
1487 						break;
1488 					}
1489 					optp = &in6p->in6p_outputopts;
1490 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1491 					    (u_char *)&optval, sizeof(optval),
1492 					    optp, (td != NULL) ? td->td_ucred :
1493 					    NULL, uproto);
1494 					break;
1495 				}
1496 
1497 				case IPV6_RECVHOPLIMIT:
1498 					/* cannot mix with RFC2292 */
1499 					if (OPTBIT(IN6P_RFC2292)) {
1500 						error = EINVAL;
1501 						break;
1502 					}
1503 					OPTSET(IN6P_HOPLIMIT);
1504 					break;
1505 
1506 				case IPV6_RECVHOPOPTS:
1507 					/* cannot mix with RFC2292 */
1508 					if (OPTBIT(IN6P_RFC2292)) {
1509 						error = EINVAL;
1510 						break;
1511 					}
1512 					OPTSET(IN6P_HOPOPTS);
1513 					break;
1514 
1515 				case IPV6_RECVDSTOPTS:
1516 					/* cannot mix with RFC2292 */
1517 					if (OPTBIT(IN6P_RFC2292)) {
1518 						error = EINVAL;
1519 						break;
1520 					}
1521 					OPTSET(IN6P_DSTOPTS);
1522 					break;
1523 
1524 				case IPV6_RECVRTHDRDSTOPTS:
1525 					/* cannot mix with RFC2292 */
1526 					if (OPTBIT(IN6P_RFC2292)) {
1527 						error = EINVAL;
1528 						break;
1529 					}
1530 					OPTSET(IN6P_RTHDRDSTOPTS);
1531 					break;
1532 
1533 				case IPV6_RECVRTHDR:
1534 					/* cannot mix with RFC2292 */
1535 					if (OPTBIT(IN6P_RFC2292)) {
1536 						error = EINVAL;
1537 						break;
1538 					}
1539 					OPTSET(IN6P_RTHDR);
1540 					break;
1541 
1542 				case IPV6_FAITH:
1543 					OPTSET(INP_FAITH);
1544 					break;
1545 
1546 				case IPV6_RECVPATHMTU:
1547 					/*
1548 					 * We ignore this option for TCP
1549 					 * sockets.
1550 					 * (RFC3542 leaves this case
1551 					 * unspecified.)
1552 					 */
1553 					if (uproto != IPPROTO_TCP)
1554 						OPTSET(IN6P_MTU);
1555 					break;
1556 
1557 				case IPV6_V6ONLY:
1558 					/*
1559 					 * make setsockopt(IPV6_V6ONLY)
1560 					 * available only prior to bind(2).
1561 					 * see ipng mailing list, Jun 22 2001.
1562 					 */
1563 					if (in6p->inp_lport ||
1564 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1565 						error = EINVAL;
1566 						break;
1567 					}
1568 					OPTSET(IN6P_IPV6_V6ONLY);
1569 					if (optval)
1570 						in6p->inp_vflag &= ~INP_IPV4;
1571 					else
1572 						in6p->inp_vflag |= INP_IPV4;
1573 					break;
1574 				case IPV6_RECVTCLASS:
1575 					/* cannot mix with RFC2292 XXX */
1576 					if (OPTBIT(IN6P_RFC2292)) {
1577 						error = EINVAL;
1578 						break;
1579 					}
1580 					OPTSET(IN6P_TCLASS);
1581 					break;
1582 				case IPV6_AUTOFLOWLABEL:
1583 					OPTSET(IN6P_AUTOFLOWLABEL);
1584 					break;
1585 
1586 				case IPV6_BINDANY:
1587 					OPTSET(INP_BINDANY);
1588 					break;
1589 
1590 				case IPV6_BINDMULTI:
1591 					OPTSET2(INP_BINDMULTI, optval);
1592 					break;
1593 #ifdef	RSS
1594 				case IPV6_RSS_LISTEN_BUCKET:
1595 					if ((optval >= 0) &&
1596 					    (optval < rss_getnumbuckets())) {
1597 						in6p->inp_rss_listen_bucket = optval;
1598 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1599 					} else {
1600 						error = EINVAL;
1601 					}
1602 					break;
1603 #endif
1604 				}
1605 				break;
1606 
1607 			case IPV6_TCLASS:
1608 			case IPV6_DONTFRAG:
1609 			case IPV6_USE_MIN_MTU:
1610 			case IPV6_PREFER_TEMPADDR:
1611 				if (optlen != sizeof(optval)) {
1612 					error = EINVAL;
1613 					break;
1614 				}
1615 				error = sooptcopyin(sopt, &optval,
1616 					sizeof optval, sizeof optval);
1617 				if (error)
1618 					break;
1619 				{
1620 					struct ip6_pktopts **optp;
1621 					optp = &in6p->in6p_outputopts;
1622 					error = ip6_pcbopt(optname,
1623 					    (u_char *)&optval, sizeof(optval),
1624 					    optp, (td != NULL) ? td->td_ucred :
1625 					    NULL, uproto);
1626 					break;
1627 				}
1628 
1629 			case IPV6_2292PKTINFO:
1630 			case IPV6_2292HOPLIMIT:
1631 			case IPV6_2292HOPOPTS:
1632 			case IPV6_2292DSTOPTS:
1633 			case IPV6_2292RTHDR:
1634 				/* RFC 2292 */
1635 				if (optlen != sizeof(int)) {
1636 					error = EINVAL;
1637 					break;
1638 				}
1639 				error = sooptcopyin(sopt, &optval,
1640 					sizeof optval, sizeof optval);
1641 				if (error)
1642 					break;
1643 				switch (optname) {
1644 				case IPV6_2292PKTINFO:
1645 					OPTSET2292(IN6P_PKTINFO);
1646 					break;
1647 				case IPV6_2292HOPLIMIT:
1648 					OPTSET2292(IN6P_HOPLIMIT);
1649 					break;
1650 				case IPV6_2292HOPOPTS:
1651 					/*
1652 					 * Check super-user privilege.
1653 					 * See comments for IPV6_RECVHOPOPTS.
1654 					 */
1655 					if (td != NULL) {
1656 						error = priv_check(td,
1657 						    PRIV_NETINET_SETHDROPTS);
1658 						if (error)
1659 							return (error);
1660 					}
1661 					OPTSET2292(IN6P_HOPOPTS);
1662 					break;
1663 				case IPV6_2292DSTOPTS:
1664 					if (td != NULL) {
1665 						error = priv_check(td,
1666 						    PRIV_NETINET_SETHDROPTS);
1667 						if (error)
1668 							return (error);
1669 					}
1670 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1671 					break;
1672 				case IPV6_2292RTHDR:
1673 					OPTSET2292(IN6P_RTHDR);
1674 					break;
1675 				}
1676 				break;
1677 			case IPV6_PKTINFO:
1678 			case IPV6_HOPOPTS:
1679 			case IPV6_RTHDR:
1680 			case IPV6_DSTOPTS:
1681 			case IPV6_RTHDRDSTOPTS:
1682 			case IPV6_NEXTHOP:
1683 			{
1684 				/* new advanced API (RFC3542) */
1685 				u_char *optbuf;
1686 				u_char optbuf_storage[MCLBYTES];
1687 				int optlen;
1688 				struct ip6_pktopts **optp;
1689 
1690 				/* cannot mix with RFC2292 */
1691 				if (OPTBIT(IN6P_RFC2292)) {
1692 					error = EINVAL;
1693 					break;
1694 				}
1695 
1696 				/*
1697 				 * We only ensure valsize is not too large
1698 				 * here.  Further validation will be done
1699 				 * later.
1700 				 */
1701 				error = sooptcopyin(sopt, optbuf_storage,
1702 				    sizeof(optbuf_storage), 0);
1703 				if (error)
1704 					break;
1705 				optlen = sopt->sopt_valsize;
1706 				optbuf = optbuf_storage;
1707 				optp = &in6p->in6p_outputopts;
1708 				error = ip6_pcbopt(optname, optbuf, optlen,
1709 				    optp, (td != NULL) ? td->td_ucred : NULL,
1710 				    uproto);
1711 				break;
1712 			}
1713 #undef OPTSET
1714 
1715 			case IPV6_MULTICAST_IF:
1716 			case IPV6_MULTICAST_HOPS:
1717 			case IPV6_MULTICAST_LOOP:
1718 			case IPV6_JOIN_GROUP:
1719 			case IPV6_LEAVE_GROUP:
1720 			case IPV6_MSFILTER:
1721 			case MCAST_BLOCK_SOURCE:
1722 			case MCAST_UNBLOCK_SOURCE:
1723 			case MCAST_JOIN_GROUP:
1724 			case MCAST_LEAVE_GROUP:
1725 			case MCAST_JOIN_SOURCE_GROUP:
1726 			case MCAST_LEAVE_SOURCE_GROUP:
1727 				error = ip6_setmoptions(in6p, sopt);
1728 				break;
1729 
1730 			case IPV6_PORTRANGE:
1731 				error = sooptcopyin(sopt, &optval,
1732 				    sizeof optval, sizeof optval);
1733 				if (error)
1734 					break;
1735 
1736 				INP_WLOCK(in6p);
1737 				switch (optval) {
1738 				case IPV6_PORTRANGE_DEFAULT:
1739 					in6p->inp_flags &= ~(INP_LOWPORT);
1740 					in6p->inp_flags &= ~(INP_HIGHPORT);
1741 					break;
1742 
1743 				case IPV6_PORTRANGE_HIGH:
1744 					in6p->inp_flags &= ~(INP_LOWPORT);
1745 					in6p->inp_flags |= INP_HIGHPORT;
1746 					break;
1747 
1748 				case IPV6_PORTRANGE_LOW:
1749 					in6p->inp_flags &= ~(INP_HIGHPORT);
1750 					in6p->inp_flags |= INP_LOWPORT;
1751 					break;
1752 
1753 				default:
1754 					error = EINVAL;
1755 					break;
1756 				}
1757 				INP_WUNLOCK(in6p);
1758 				break;
1759 
1760 #ifdef IPSEC
1761 			case IPV6_IPSEC_POLICY:
1762 			{
1763 				caddr_t req;
1764 				struct mbuf *m;
1765 
1766 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1767 					break;
1768 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1769 					break;
1770 				req = mtod(m, caddr_t);
1771 				error = ipsec_set_policy(in6p, optname, req,
1772 				    m->m_len, (sopt->sopt_td != NULL) ?
1773 				    sopt->sopt_td->td_ucred : NULL);
1774 				m_freem(m);
1775 				break;
1776 			}
1777 #endif /* IPSEC */
1778 
1779 			default:
1780 				error = ENOPROTOOPT;
1781 				break;
1782 			}
1783 			break;
1784 
1785 		case SOPT_GET:
1786 			switch (optname) {
1787 
1788 			case IPV6_2292PKTOPTIONS:
1789 #ifdef IPV6_PKTOPTIONS
1790 			case IPV6_PKTOPTIONS:
1791 #endif
1792 				/*
1793 				 * RFC3542 (effectively) deprecated the
1794 				 * semantics of the 2292-style pktoptions.
1795 				 * Since it was not reliable in nature (i.e.,
1796 				 * applications had to expect the lack of some
1797 				 * information after all), it would make sense
1798 				 * to simplify this part by always returning
1799 				 * empty data.
1800 				 */
1801 				sopt->sopt_valsize = 0;
1802 				break;
1803 
1804 			case IPV6_RECVHOPOPTS:
1805 			case IPV6_RECVDSTOPTS:
1806 			case IPV6_RECVRTHDRDSTOPTS:
1807 			case IPV6_UNICAST_HOPS:
1808 			case IPV6_RECVPKTINFO:
1809 			case IPV6_RECVHOPLIMIT:
1810 			case IPV6_RECVRTHDR:
1811 			case IPV6_RECVPATHMTU:
1812 
1813 			case IPV6_FAITH:
1814 			case IPV6_V6ONLY:
1815 			case IPV6_PORTRANGE:
1816 			case IPV6_RECVTCLASS:
1817 			case IPV6_AUTOFLOWLABEL:
1818 			case IPV6_BINDANY:
1819 			case IPV6_FLOWID:
1820 			case IPV6_FLOWTYPE:
1821 #ifdef	RSS
1822 			case IPV6_RSSBUCKETID:
1823 #endif
1824 				switch (optname) {
1825 
1826 				case IPV6_RECVHOPOPTS:
1827 					optval = OPTBIT(IN6P_HOPOPTS);
1828 					break;
1829 
1830 				case IPV6_RECVDSTOPTS:
1831 					optval = OPTBIT(IN6P_DSTOPTS);
1832 					break;
1833 
1834 				case IPV6_RECVRTHDRDSTOPTS:
1835 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1836 					break;
1837 
1838 				case IPV6_UNICAST_HOPS:
1839 					optval = in6p->in6p_hops;
1840 					break;
1841 
1842 				case IPV6_RECVPKTINFO:
1843 					optval = OPTBIT(IN6P_PKTINFO);
1844 					break;
1845 
1846 				case IPV6_RECVHOPLIMIT:
1847 					optval = OPTBIT(IN6P_HOPLIMIT);
1848 					break;
1849 
1850 				case IPV6_RECVRTHDR:
1851 					optval = OPTBIT(IN6P_RTHDR);
1852 					break;
1853 
1854 				case IPV6_RECVPATHMTU:
1855 					optval = OPTBIT(IN6P_MTU);
1856 					break;
1857 
1858 				case IPV6_FAITH:
1859 					optval = OPTBIT(INP_FAITH);
1860 					break;
1861 
1862 				case IPV6_V6ONLY:
1863 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1864 					break;
1865 
1866 				case IPV6_PORTRANGE:
1867 				    {
1868 					int flags;
1869 					flags = in6p->inp_flags;
1870 					if (flags & INP_HIGHPORT)
1871 						optval = IPV6_PORTRANGE_HIGH;
1872 					else if (flags & INP_LOWPORT)
1873 						optval = IPV6_PORTRANGE_LOW;
1874 					else
1875 						optval = 0;
1876 					break;
1877 				    }
1878 				case IPV6_RECVTCLASS:
1879 					optval = OPTBIT(IN6P_TCLASS);
1880 					break;
1881 
1882 				case IPV6_AUTOFLOWLABEL:
1883 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1884 					break;
1885 
1886 				case IPV6_BINDANY:
1887 					optval = OPTBIT(INP_BINDANY);
1888 					break;
1889 
1890 				case IPV6_FLOWID:
1891 					optval = in6p->inp_flowid;
1892 					break;
1893 
1894 				case IPV6_FLOWTYPE:
1895 					optval = in6p->inp_flowtype;
1896 					break;
1897 #ifdef	RSS
1898 				case IPV6_RSSBUCKETID:
1899 					retval =
1900 					    rss_hash2bucket(in6p->inp_flowid,
1901 					    in6p->inp_flowtype,
1902 					    &rss_bucket);
1903 					if (retval == 0)
1904 						optval = rss_bucket;
1905 					else
1906 						error = EINVAL;
1907 					break;
1908 #endif
1909 
1910 				case IPV6_BINDMULTI:
1911 					optval = OPTBIT2(INP_BINDMULTI);
1912 					break;
1913 
1914 				}
1915 				if (error)
1916 					break;
1917 				error = sooptcopyout(sopt, &optval,
1918 					sizeof optval);
1919 				break;
1920 
1921 			case IPV6_PATHMTU:
1922 			{
1923 				u_long pmtu = 0;
1924 				struct ip6_mtuinfo mtuinfo;
1925 				struct route_in6 sro;
1926 
1927 				bzero(&sro, sizeof(sro));
1928 
1929 				if (!(so->so_state & SS_ISCONNECTED))
1930 					return (ENOTCONN);
1931 				/*
1932 				 * XXX: we dot not consider the case of source
1933 				 * routing, or optional information to specify
1934 				 * the outgoing interface.
1935 				 */
1936 				error = ip6_getpmtu(&sro, NULL, NULL,
1937 				    &in6p->in6p_faddr, &pmtu, NULL,
1938 				    so->so_fibnum);
1939 				if (sro.ro_rt)
1940 					RTFREE(sro.ro_rt);
1941 				if (error)
1942 					break;
1943 				if (pmtu > IPV6_MAXPACKET)
1944 					pmtu = IPV6_MAXPACKET;
1945 
1946 				bzero(&mtuinfo, sizeof(mtuinfo));
1947 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1948 				optdata = (void *)&mtuinfo;
1949 				optdatalen = sizeof(mtuinfo);
1950 				error = sooptcopyout(sopt, optdata,
1951 				    optdatalen);
1952 				break;
1953 			}
1954 
1955 			case IPV6_2292PKTINFO:
1956 			case IPV6_2292HOPLIMIT:
1957 			case IPV6_2292HOPOPTS:
1958 			case IPV6_2292RTHDR:
1959 			case IPV6_2292DSTOPTS:
1960 				switch (optname) {
1961 				case IPV6_2292PKTINFO:
1962 					optval = OPTBIT(IN6P_PKTINFO);
1963 					break;
1964 				case IPV6_2292HOPLIMIT:
1965 					optval = OPTBIT(IN6P_HOPLIMIT);
1966 					break;
1967 				case IPV6_2292HOPOPTS:
1968 					optval = OPTBIT(IN6P_HOPOPTS);
1969 					break;
1970 				case IPV6_2292RTHDR:
1971 					optval = OPTBIT(IN6P_RTHDR);
1972 					break;
1973 				case IPV6_2292DSTOPTS:
1974 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1975 					break;
1976 				}
1977 				error = sooptcopyout(sopt, &optval,
1978 				    sizeof optval);
1979 				break;
1980 			case IPV6_PKTINFO:
1981 			case IPV6_HOPOPTS:
1982 			case IPV6_RTHDR:
1983 			case IPV6_DSTOPTS:
1984 			case IPV6_RTHDRDSTOPTS:
1985 			case IPV6_NEXTHOP:
1986 			case IPV6_TCLASS:
1987 			case IPV6_DONTFRAG:
1988 			case IPV6_USE_MIN_MTU:
1989 			case IPV6_PREFER_TEMPADDR:
1990 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1991 				    optname, sopt);
1992 				break;
1993 
1994 			case IPV6_MULTICAST_IF:
1995 			case IPV6_MULTICAST_HOPS:
1996 			case IPV6_MULTICAST_LOOP:
1997 			case IPV6_MSFILTER:
1998 				error = ip6_getmoptions(in6p, sopt);
1999 				break;
2000 
2001 #ifdef IPSEC
2002 			case IPV6_IPSEC_POLICY:
2003 			  {
2004 				caddr_t req = NULL;
2005 				size_t len = 0;
2006 				struct mbuf *m = NULL;
2007 				struct mbuf **mp = &m;
2008 				size_t ovalsize = sopt->sopt_valsize;
2009 				caddr_t oval = (caddr_t)sopt->sopt_val;
2010 
2011 				error = soopt_getm(sopt, &m); /* XXX */
2012 				if (error != 0)
2013 					break;
2014 				error = soopt_mcopyin(sopt, m); /* XXX */
2015 				if (error != 0)
2016 					break;
2017 				sopt->sopt_valsize = ovalsize;
2018 				sopt->sopt_val = oval;
2019 				if (m) {
2020 					req = mtod(m, caddr_t);
2021 					len = m->m_len;
2022 				}
2023 				error = ipsec_get_policy(in6p, req, len, mp);
2024 				if (error == 0)
2025 					error = soopt_mcopyout(sopt, m); /* XXX */
2026 				if (error == 0 && m)
2027 					m_freem(m);
2028 				break;
2029 			  }
2030 #endif /* IPSEC */
2031 
2032 			default:
2033 				error = ENOPROTOOPT;
2034 				break;
2035 			}
2036 			break;
2037 		}
2038 	}
2039 	return (error);
2040 }
2041 
2042 int
2043 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2044 {
2045 	int error = 0, optval, optlen;
2046 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2047 	struct inpcb *in6p = sotoinpcb(so);
2048 	int level, op, optname;
2049 
2050 	level = sopt->sopt_level;
2051 	op = sopt->sopt_dir;
2052 	optname = sopt->sopt_name;
2053 	optlen = sopt->sopt_valsize;
2054 
2055 	if (level != IPPROTO_IPV6) {
2056 		return (EINVAL);
2057 	}
2058 
2059 	switch (optname) {
2060 	case IPV6_CHECKSUM:
2061 		/*
2062 		 * For ICMPv6 sockets, no modification allowed for checksum
2063 		 * offset, permit "no change" values to help existing apps.
2064 		 *
2065 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2066 		 * for an ICMPv6 socket will fail."
2067 		 * The current behavior does not meet RFC3542.
2068 		 */
2069 		switch (op) {
2070 		case SOPT_SET:
2071 			if (optlen != sizeof(int)) {
2072 				error = EINVAL;
2073 				break;
2074 			}
2075 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2076 					    sizeof(optval));
2077 			if (error)
2078 				break;
2079 			if ((optval % 2) != 0) {
2080 				/* the API assumes even offset values */
2081 				error = EINVAL;
2082 			} else if (so->so_proto->pr_protocol ==
2083 			    IPPROTO_ICMPV6) {
2084 				if (optval != icmp6off)
2085 					error = EINVAL;
2086 			} else
2087 				in6p->in6p_cksum = optval;
2088 			break;
2089 
2090 		case SOPT_GET:
2091 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2092 				optval = icmp6off;
2093 			else
2094 				optval = in6p->in6p_cksum;
2095 
2096 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2097 			break;
2098 
2099 		default:
2100 			error = EINVAL;
2101 			break;
2102 		}
2103 		break;
2104 
2105 	default:
2106 		error = ENOPROTOOPT;
2107 		break;
2108 	}
2109 
2110 	return (error);
2111 }
2112 
2113 /*
2114  * Set up IP6 options in pcb for insertion in output packets or
2115  * specifying behavior of outgoing packets.
2116  */
2117 static int
2118 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2119     struct socket *so, struct sockopt *sopt)
2120 {
2121 	struct ip6_pktopts *opt = *pktopt;
2122 	int error = 0;
2123 	struct thread *td = sopt->sopt_td;
2124 
2125 	/* turn off any old options. */
2126 	if (opt) {
2127 #ifdef DIAGNOSTIC
2128 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2129 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2130 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2131 			printf("ip6_pcbopts: all specified options are cleared.\n");
2132 #endif
2133 		ip6_clearpktopts(opt, -1);
2134 	} else
2135 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2136 	*pktopt = NULL;
2137 
2138 	if (!m || m->m_len == 0) {
2139 		/*
2140 		 * Only turning off any previous options, regardless of
2141 		 * whether the opt is just created or given.
2142 		 */
2143 		free(opt, M_IP6OPT);
2144 		return (0);
2145 	}
2146 
2147 	/*  set options specified by user. */
2148 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2149 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2150 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2151 		free(opt, M_IP6OPT);
2152 		return (error);
2153 	}
2154 	*pktopt = opt;
2155 	return (0);
2156 }
2157 
2158 /*
2159  * initialize ip6_pktopts.  beware that there are non-zero default values in
2160  * the struct.
2161  */
2162 void
2163 ip6_initpktopts(struct ip6_pktopts *opt)
2164 {
2165 
2166 	bzero(opt, sizeof(*opt));
2167 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2168 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2169 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2170 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2171 }
2172 
2173 static int
2174 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2175     struct ucred *cred, int uproto)
2176 {
2177 	struct ip6_pktopts *opt;
2178 
2179 	if (*pktopt == NULL) {
2180 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2181 		    M_WAITOK);
2182 		ip6_initpktopts(*pktopt);
2183 	}
2184 	opt = *pktopt;
2185 
2186 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2187 }
2188 
2189 static int
2190 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2191 {
2192 	void *optdata = NULL;
2193 	int optdatalen = 0;
2194 	struct ip6_ext *ip6e;
2195 	int error = 0;
2196 	struct in6_pktinfo null_pktinfo;
2197 	int deftclass = 0, on;
2198 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2199 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2200 
2201 	switch (optname) {
2202 	case IPV6_PKTINFO:
2203 		if (pktopt && pktopt->ip6po_pktinfo)
2204 			optdata = (void *)pktopt->ip6po_pktinfo;
2205 		else {
2206 			/* XXX: we don't have to do this every time... */
2207 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2208 			optdata = (void *)&null_pktinfo;
2209 		}
2210 		optdatalen = sizeof(struct in6_pktinfo);
2211 		break;
2212 	case IPV6_TCLASS:
2213 		if (pktopt && pktopt->ip6po_tclass >= 0)
2214 			optdata = (void *)&pktopt->ip6po_tclass;
2215 		else
2216 			optdata = (void *)&deftclass;
2217 		optdatalen = sizeof(int);
2218 		break;
2219 	case IPV6_HOPOPTS:
2220 		if (pktopt && pktopt->ip6po_hbh) {
2221 			optdata = (void *)pktopt->ip6po_hbh;
2222 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2223 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2224 		}
2225 		break;
2226 	case IPV6_RTHDR:
2227 		if (pktopt && pktopt->ip6po_rthdr) {
2228 			optdata = (void *)pktopt->ip6po_rthdr;
2229 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2230 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2231 		}
2232 		break;
2233 	case IPV6_RTHDRDSTOPTS:
2234 		if (pktopt && pktopt->ip6po_dest1) {
2235 			optdata = (void *)pktopt->ip6po_dest1;
2236 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2237 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2238 		}
2239 		break;
2240 	case IPV6_DSTOPTS:
2241 		if (pktopt && pktopt->ip6po_dest2) {
2242 			optdata = (void *)pktopt->ip6po_dest2;
2243 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2244 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2245 		}
2246 		break;
2247 	case IPV6_NEXTHOP:
2248 		if (pktopt && pktopt->ip6po_nexthop) {
2249 			optdata = (void *)pktopt->ip6po_nexthop;
2250 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2251 		}
2252 		break;
2253 	case IPV6_USE_MIN_MTU:
2254 		if (pktopt)
2255 			optdata = (void *)&pktopt->ip6po_minmtu;
2256 		else
2257 			optdata = (void *)&defminmtu;
2258 		optdatalen = sizeof(int);
2259 		break;
2260 	case IPV6_DONTFRAG:
2261 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2262 			on = 1;
2263 		else
2264 			on = 0;
2265 		optdata = (void *)&on;
2266 		optdatalen = sizeof(on);
2267 		break;
2268 	case IPV6_PREFER_TEMPADDR:
2269 		if (pktopt)
2270 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2271 		else
2272 			optdata = (void *)&defpreftemp;
2273 		optdatalen = sizeof(int);
2274 		break;
2275 	default:		/* should not happen */
2276 #ifdef DIAGNOSTIC
2277 		panic("ip6_getpcbopt: unexpected option\n");
2278 #endif
2279 		return (ENOPROTOOPT);
2280 	}
2281 
2282 	error = sooptcopyout(sopt, optdata, optdatalen);
2283 
2284 	return (error);
2285 }
2286 
2287 void
2288 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2289 {
2290 	if (pktopt == NULL)
2291 		return;
2292 
2293 	if (optname == -1 || optname == IPV6_PKTINFO) {
2294 		if (pktopt->ip6po_pktinfo)
2295 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2296 		pktopt->ip6po_pktinfo = NULL;
2297 	}
2298 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2299 		pktopt->ip6po_hlim = -1;
2300 	if (optname == -1 || optname == IPV6_TCLASS)
2301 		pktopt->ip6po_tclass = -1;
2302 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2303 		if (pktopt->ip6po_nextroute.ro_rt) {
2304 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2305 			pktopt->ip6po_nextroute.ro_rt = NULL;
2306 		}
2307 		if (pktopt->ip6po_nexthop)
2308 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2309 		pktopt->ip6po_nexthop = NULL;
2310 	}
2311 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2312 		if (pktopt->ip6po_hbh)
2313 			free(pktopt->ip6po_hbh, M_IP6OPT);
2314 		pktopt->ip6po_hbh = NULL;
2315 	}
2316 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2317 		if (pktopt->ip6po_dest1)
2318 			free(pktopt->ip6po_dest1, M_IP6OPT);
2319 		pktopt->ip6po_dest1 = NULL;
2320 	}
2321 	if (optname == -1 || optname == IPV6_RTHDR) {
2322 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2323 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2324 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2325 		if (pktopt->ip6po_route.ro_rt) {
2326 			RTFREE(pktopt->ip6po_route.ro_rt);
2327 			pktopt->ip6po_route.ro_rt = NULL;
2328 		}
2329 	}
2330 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2331 		if (pktopt->ip6po_dest2)
2332 			free(pktopt->ip6po_dest2, M_IP6OPT);
2333 		pktopt->ip6po_dest2 = NULL;
2334 	}
2335 }
2336 
2337 #define PKTOPT_EXTHDRCPY(type) \
2338 do {\
2339 	if (src->type) {\
2340 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2341 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2342 		if (dst->type == NULL && canwait == M_NOWAIT)\
2343 			goto bad;\
2344 		bcopy(src->type, dst->type, hlen);\
2345 	}\
2346 } while (/*CONSTCOND*/ 0)
2347 
2348 static int
2349 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2350 {
2351 	if (dst == NULL || src == NULL)  {
2352 		printf("ip6_clearpktopts: invalid argument\n");
2353 		return (EINVAL);
2354 	}
2355 
2356 	dst->ip6po_hlim = src->ip6po_hlim;
2357 	dst->ip6po_tclass = src->ip6po_tclass;
2358 	dst->ip6po_flags = src->ip6po_flags;
2359 	dst->ip6po_minmtu = src->ip6po_minmtu;
2360 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2361 	if (src->ip6po_pktinfo) {
2362 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2363 		    M_IP6OPT, canwait);
2364 		if (dst->ip6po_pktinfo == NULL)
2365 			goto bad;
2366 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2367 	}
2368 	if (src->ip6po_nexthop) {
2369 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2370 		    M_IP6OPT, canwait);
2371 		if (dst->ip6po_nexthop == NULL)
2372 			goto bad;
2373 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2374 		    src->ip6po_nexthop->sa_len);
2375 	}
2376 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2377 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2378 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2379 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2380 	return (0);
2381 
2382   bad:
2383 	ip6_clearpktopts(dst, -1);
2384 	return (ENOBUFS);
2385 }
2386 #undef PKTOPT_EXTHDRCPY
2387 
2388 struct ip6_pktopts *
2389 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2390 {
2391 	int error;
2392 	struct ip6_pktopts *dst;
2393 
2394 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2395 	if (dst == NULL)
2396 		return (NULL);
2397 	ip6_initpktopts(dst);
2398 
2399 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2400 		free(dst, M_IP6OPT);
2401 		return (NULL);
2402 	}
2403 
2404 	return (dst);
2405 }
2406 
2407 void
2408 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2409 {
2410 	if (pktopt == NULL)
2411 		return;
2412 
2413 	ip6_clearpktopts(pktopt, -1);
2414 
2415 	free(pktopt, M_IP6OPT);
2416 }
2417 
2418 /*
2419  * Set IPv6 outgoing packet options based on advanced API.
2420  */
2421 int
2422 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2423     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2424 {
2425 	struct cmsghdr *cm = 0;
2426 
2427 	if (control == NULL || opt == NULL)
2428 		return (EINVAL);
2429 
2430 	ip6_initpktopts(opt);
2431 	if (stickyopt) {
2432 		int error;
2433 
2434 		/*
2435 		 * If stickyopt is provided, make a local copy of the options
2436 		 * for this particular packet, then override them by ancillary
2437 		 * objects.
2438 		 * XXX: copypktopts() does not copy the cached route to a next
2439 		 * hop (if any).  This is not very good in terms of efficiency,
2440 		 * but we can allow this since this option should be rarely
2441 		 * used.
2442 		 */
2443 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2444 			return (error);
2445 	}
2446 
2447 	/*
2448 	 * XXX: Currently, we assume all the optional information is stored
2449 	 * in a single mbuf.
2450 	 */
2451 	if (control->m_next)
2452 		return (EINVAL);
2453 
2454 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2455 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2456 		int error;
2457 
2458 		if (control->m_len < CMSG_LEN(0))
2459 			return (EINVAL);
2460 
2461 		cm = mtod(control, struct cmsghdr *);
2462 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2463 			return (EINVAL);
2464 		if (cm->cmsg_level != IPPROTO_IPV6)
2465 			continue;
2466 
2467 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2468 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2469 		if (error)
2470 			return (error);
2471 	}
2472 
2473 	return (0);
2474 }
2475 
2476 /*
2477  * Set a particular packet option, as a sticky option or an ancillary data
2478  * item.  "len" can be 0 only when it's a sticky option.
2479  * We have 4 cases of combination of "sticky" and "cmsg":
2480  * "sticky=0, cmsg=0": impossible
2481  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2482  * "sticky=1, cmsg=0": RFC3542 socket option
2483  * "sticky=1, cmsg=1": RFC2292 socket option
2484  */
2485 static int
2486 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2487     struct ucred *cred, int sticky, int cmsg, int uproto)
2488 {
2489 	int minmtupolicy, preftemp;
2490 	int error;
2491 
2492 	if (!sticky && !cmsg) {
2493 #ifdef DIAGNOSTIC
2494 		printf("ip6_setpktopt: impossible case\n");
2495 #endif
2496 		return (EINVAL);
2497 	}
2498 
2499 	/*
2500 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2501 	 * not be specified in the context of RFC3542.  Conversely,
2502 	 * RFC3542 types should not be specified in the context of RFC2292.
2503 	 */
2504 	if (!cmsg) {
2505 		switch (optname) {
2506 		case IPV6_2292PKTINFO:
2507 		case IPV6_2292HOPLIMIT:
2508 		case IPV6_2292NEXTHOP:
2509 		case IPV6_2292HOPOPTS:
2510 		case IPV6_2292DSTOPTS:
2511 		case IPV6_2292RTHDR:
2512 		case IPV6_2292PKTOPTIONS:
2513 			return (ENOPROTOOPT);
2514 		}
2515 	}
2516 	if (sticky && cmsg) {
2517 		switch (optname) {
2518 		case IPV6_PKTINFO:
2519 		case IPV6_HOPLIMIT:
2520 		case IPV6_NEXTHOP:
2521 		case IPV6_HOPOPTS:
2522 		case IPV6_DSTOPTS:
2523 		case IPV6_RTHDRDSTOPTS:
2524 		case IPV6_RTHDR:
2525 		case IPV6_USE_MIN_MTU:
2526 		case IPV6_DONTFRAG:
2527 		case IPV6_TCLASS:
2528 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2529 			return (ENOPROTOOPT);
2530 		}
2531 	}
2532 
2533 	switch (optname) {
2534 	case IPV6_2292PKTINFO:
2535 	case IPV6_PKTINFO:
2536 	{
2537 		struct ifnet *ifp = NULL;
2538 		struct in6_pktinfo *pktinfo;
2539 
2540 		if (len != sizeof(struct in6_pktinfo))
2541 			return (EINVAL);
2542 
2543 		pktinfo = (struct in6_pktinfo *)buf;
2544 
2545 		/*
2546 		 * An application can clear any sticky IPV6_PKTINFO option by
2547 		 * doing a "regular" setsockopt with ipi6_addr being
2548 		 * in6addr_any and ipi6_ifindex being zero.
2549 		 * [RFC 3542, Section 6]
2550 		 */
2551 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2552 		    pktinfo->ipi6_ifindex == 0 &&
2553 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2554 			ip6_clearpktopts(opt, optname);
2555 			break;
2556 		}
2557 
2558 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2559 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2560 			return (EINVAL);
2561 		}
2562 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2563 			return (EINVAL);
2564 		/* validate the interface index if specified. */
2565 		if (pktinfo->ipi6_ifindex > V_if_index)
2566 			 return (ENXIO);
2567 		if (pktinfo->ipi6_ifindex) {
2568 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2569 			if (ifp == NULL)
2570 				return (ENXIO);
2571 		}
2572 		if (ifp != NULL && (
2573 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2574 			return (ENETDOWN);
2575 
2576 		if (ifp != NULL &&
2577 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2578 			struct in6_ifaddr *ia;
2579 
2580 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2581 			if (ia == NULL)
2582 				return (EADDRNOTAVAIL);
2583 			ifa_free(&ia->ia_ifa);
2584 		}
2585 		/*
2586 		 * We store the address anyway, and let in6_selectsrc()
2587 		 * validate the specified address.  This is because ipi6_addr
2588 		 * may not have enough information about its scope zone, and
2589 		 * we may need additional information (such as outgoing
2590 		 * interface or the scope zone of a destination address) to
2591 		 * disambiguate the scope.
2592 		 * XXX: the delay of the validation may confuse the
2593 		 * application when it is used as a sticky option.
2594 		 */
2595 		if (opt->ip6po_pktinfo == NULL) {
2596 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2597 			    M_IP6OPT, M_NOWAIT);
2598 			if (opt->ip6po_pktinfo == NULL)
2599 				return (ENOBUFS);
2600 		}
2601 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2602 		break;
2603 	}
2604 
2605 	case IPV6_2292HOPLIMIT:
2606 	case IPV6_HOPLIMIT:
2607 	{
2608 		int *hlimp;
2609 
2610 		/*
2611 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2612 		 * to simplify the ordering among hoplimit options.
2613 		 */
2614 		if (optname == IPV6_HOPLIMIT && sticky)
2615 			return (ENOPROTOOPT);
2616 
2617 		if (len != sizeof(int))
2618 			return (EINVAL);
2619 		hlimp = (int *)buf;
2620 		if (*hlimp < -1 || *hlimp > 255)
2621 			return (EINVAL);
2622 
2623 		opt->ip6po_hlim = *hlimp;
2624 		break;
2625 	}
2626 
2627 	case IPV6_TCLASS:
2628 	{
2629 		int tclass;
2630 
2631 		if (len != sizeof(int))
2632 			return (EINVAL);
2633 		tclass = *(int *)buf;
2634 		if (tclass < -1 || tclass > 255)
2635 			return (EINVAL);
2636 
2637 		opt->ip6po_tclass = tclass;
2638 		break;
2639 	}
2640 
2641 	case IPV6_2292NEXTHOP:
2642 	case IPV6_NEXTHOP:
2643 		if (cred != NULL) {
2644 			error = priv_check_cred(cred,
2645 			    PRIV_NETINET_SETHDROPTS, 0);
2646 			if (error)
2647 				return (error);
2648 		}
2649 
2650 		if (len == 0) {	/* just remove the option */
2651 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2652 			break;
2653 		}
2654 
2655 		/* check if cmsg_len is large enough for sa_len */
2656 		if (len < sizeof(struct sockaddr) || len < *buf)
2657 			return (EINVAL);
2658 
2659 		switch (((struct sockaddr *)buf)->sa_family) {
2660 		case AF_INET6:
2661 		{
2662 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2663 			int error;
2664 
2665 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2666 				return (EINVAL);
2667 
2668 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2669 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2670 				return (EINVAL);
2671 			}
2672 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2673 			    != 0) {
2674 				return (error);
2675 			}
2676 			break;
2677 		}
2678 		case AF_LINK:	/* should eventually be supported */
2679 		default:
2680 			return (EAFNOSUPPORT);
2681 		}
2682 
2683 		/* turn off the previous option, then set the new option. */
2684 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2685 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2686 		if (opt->ip6po_nexthop == NULL)
2687 			return (ENOBUFS);
2688 		bcopy(buf, opt->ip6po_nexthop, *buf);
2689 		break;
2690 
2691 	case IPV6_2292HOPOPTS:
2692 	case IPV6_HOPOPTS:
2693 	{
2694 		struct ip6_hbh *hbh;
2695 		int hbhlen;
2696 
2697 		/*
2698 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2699 		 * options, since per-option restriction has too much
2700 		 * overhead.
2701 		 */
2702 		if (cred != NULL) {
2703 			error = priv_check_cred(cred,
2704 			    PRIV_NETINET_SETHDROPTS, 0);
2705 			if (error)
2706 				return (error);
2707 		}
2708 
2709 		if (len == 0) {
2710 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2711 			break;	/* just remove the option */
2712 		}
2713 
2714 		/* message length validation */
2715 		if (len < sizeof(struct ip6_hbh))
2716 			return (EINVAL);
2717 		hbh = (struct ip6_hbh *)buf;
2718 		hbhlen = (hbh->ip6h_len + 1) << 3;
2719 		if (len != hbhlen)
2720 			return (EINVAL);
2721 
2722 		/* turn off the previous option, then set the new option. */
2723 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2724 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2725 		if (opt->ip6po_hbh == NULL)
2726 			return (ENOBUFS);
2727 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2728 
2729 		break;
2730 	}
2731 
2732 	case IPV6_2292DSTOPTS:
2733 	case IPV6_DSTOPTS:
2734 	case IPV6_RTHDRDSTOPTS:
2735 	{
2736 		struct ip6_dest *dest, **newdest = NULL;
2737 		int destlen;
2738 
2739 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2740 			error = priv_check_cred(cred,
2741 			    PRIV_NETINET_SETHDROPTS, 0);
2742 			if (error)
2743 				return (error);
2744 		}
2745 
2746 		if (len == 0) {
2747 			ip6_clearpktopts(opt, optname);
2748 			break;	/* just remove the option */
2749 		}
2750 
2751 		/* message length validation */
2752 		if (len < sizeof(struct ip6_dest))
2753 			return (EINVAL);
2754 		dest = (struct ip6_dest *)buf;
2755 		destlen = (dest->ip6d_len + 1) << 3;
2756 		if (len != destlen)
2757 			return (EINVAL);
2758 
2759 		/*
2760 		 * Determine the position that the destination options header
2761 		 * should be inserted; before or after the routing header.
2762 		 */
2763 		switch (optname) {
2764 		case IPV6_2292DSTOPTS:
2765 			/*
2766 			 * The old advacned API is ambiguous on this point.
2767 			 * Our approach is to determine the position based
2768 			 * according to the existence of a routing header.
2769 			 * Note, however, that this depends on the order of the
2770 			 * extension headers in the ancillary data; the 1st
2771 			 * part of the destination options header must appear
2772 			 * before the routing header in the ancillary data,
2773 			 * too.
2774 			 * RFC3542 solved the ambiguity by introducing
2775 			 * separate ancillary data or option types.
2776 			 */
2777 			if (opt->ip6po_rthdr == NULL)
2778 				newdest = &opt->ip6po_dest1;
2779 			else
2780 				newdest = &opt->ip6po_dest2;
2781 			break;
2782 		case IPV6_RTHDRDSTOPTS:
2783 			newdest = &opt->ip6po_dest1;
2784 			break;
2785 		case IPV6_DSTOPTS:
2786 			newdest = &opt->ip6po_dest2;
2787 			break;
2788 		}
2789 
2790 		/* turn off the previous option, then set the new option. */
2791 		ip6_clearpktopts(opt, optname);
2792 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2793 		if (*newdest == NULL)
2794 			return (ENOBUFS);
2795 		bcopy(dest, *newdest, destlen);
2796 
2797 		break;
2798 	}
2799 
2800 	case IPV6_2292RTHDR:
2801 	case IPV6_RTHDR:
2802 	{
2803 		struct ip6_rthdr *rth;
2804 		int rthlen;
2805 
2806 		if (len == 0) {
2807 			ip6_clearpktopts(opt, IPV6_RTHDR);
2808 			break;	/* just remove the option */
2809 		}
2810 
2811 		/* message length validation */
2812 		if (len < sizeof(struct ip6_rthdr))
2813 			return (EINVAL);
2814 		rth = (struct ip6_rthdr *)buf;
2815 		rthlen = (rth->ip6r_len + 1) << 3;
2816 		if (len != rthlen)
2817 			return (EINVAL);
2818 
2819 		switch (rth->ip6r_type) {
2820 		case IPV6_RTHDR_TYPE_0:
2821 			if (rth->ip6r_len == 0)	/* must contain one addr */
2822 				return (EINVAL);
2823 			if (rth->ip6r_len % 2) /* length must be even */
2824 				return (EINVAL);
2825 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2826 				return (EINVAL);
2827 			break;
2828 		default:
2829 			return (EINVAL);	/* not supported */
2830 		}
2831 
2832 		/* turn off the previous option */
2833 		ip6_clearpktopts(opt, IPV6_RTHDR);
2834 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2835 		if (opt->ip6po_rthdr == NULL)
2836 			return (ENOBUFS);
2837 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2838 
2839 		break;
2840 	}
2841 
2842 	case IPV6_USE_MIN_MTU:
2843 		if (len != sizeof(int))
2844 			return (EINVAL);
2845 		minmtupolicy = *(int *)buf;
2846 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2847 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2848 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2849 			return (EINVAL);
2850 		}
2851 		opt->ip6po_minmtu = minmtupolicy;
2852 		break;
2853 
2854 	case IPV6_DONTFRAG:
2855 		if (len != sizeof(int))
2856 			return (EINVAL);
2857 
2858 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2859 			/*
2860 			 * we ignore this option for TCP sockets.
2861 			 * (RFC3542 leaves this case unspecified.)
2862 			 */
2863 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2864 		} else
2865 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2866 		break;
2867 
2868 	case IPV6_PREFER_TEMPADDR:
2869 		if (len != sizeof(int))
2870 			return (EINVAL);
2871 		preftemp = *(int *)buf;
2872 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2873 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2874 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2875 			return (EINVAL);
2876 		}
2877 		opt->ip6po_prefer_tempaddr = preftemp;
2878 		break;
2879 
2880 	default:
2881 		return (ENOPROTOOPT);
2882 	} /* end of switch */
2883 
2884 	return (0);
2885 }
2886 
2887 /*
2888  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2889  * packet to the input queue of a specified interface.  Note that this
2890  * calls the output routine of the loopback "driver", but with an interface
2891  * pointer that might NOT be &loif -- easier than replicating that code here.
2892  */
2893 void
2894 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2895 {
2896 	struct mbuf *copym;
2897 	struct ip6_hdr *ip6;
2898 
2899 	copym = m_copy(m, 0, M_COPYALL);
2900 	if (copym == NULL)
2901 		return;
2902 
2903 	/*
2904 	 * Make sure to deep-copy IPv6 header portion in case the data
2905 	 * is in an mbuf cluster, so that we can safely override the IPv6
2906 	 * header portion later.
2907 	 */
2908 	if ((copym->m_flags & M_EXT) != 0 ||
2909 	    copym->m_len < sizeof(struct ip6_hdr)) {
2910 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2911 		if (copym == NULL)
2912 			return;
2913 	}
2914 
2915 #ifdef DIAGNOSTIC
2916 	if (copym->m_len < sizeof(*ip6)) {
2917 		m_freem(copym);
2918 		return;
2919 	}
2920 #endif
2921 
2922 	ip6 = mtod(copym, struct ip6_hdr *);
2923 	/*
2924 	 * clear embedded scope identifiers if necessary.
2925 	 * in6_clearscope will touch the addresses only when necessary.
2926 	 */
2927 	in6_clearscope(&ip6->ip6_src);
2928 	in6_clearscope(&ip6->ip6_dst);
2929 
2930 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2931 }
2932 
2933 /*
2934  * Chop IPv6 header off from the payload.
2935  */
2936 static int
2937 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2938 {
2939 	struct mbuf *mh;
2940 	struct ip6_hdr *ip6;
2941 
2942 	ip6 = mtod(m, struct ip6_hdr *);
2943 	if (m->m_len > sizeof(*ip6)) {
2944 		mh = m_gethdr(M_NOWAIT, MT_DATA);
2945 		if (mh == NULL) {
2946 			m_freem(m);
2947 			return ENOBUFS;
2948 		}
2949 		m_move_pkthdr(mh, m);
2950 		MH_ALIGN(mh, sizeof(*ip6));
2951 		m->m_len -= sizeof(*ip6);
2952 		m->m_data += sizeof(*ip6);
2953 		mh->m_next = m;
2954 		m = mh;
2955 		m->m_len = sizeof(*ip6);
2956 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2957 	}
2958 	exthdrs->ip6e_ip6 = m;
2959 	return 0;
2960 }
2961 
2962 /*
2963  * Compute IPv6 extension header length.
2964  */
2965 int
2966 ip6_optlen(struct inpcb *in6p)
2967 {
2968 	int len;
2969 
2970 	if (!in6p->in6p_outputopts)
2971 		return 0;
2972 
2973 	len = 0;
2974 #define elen(x) \
2975     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2976 
2977 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2978 	if (in6p->in6p_outputopts->ip6po_rthdr)
2979 		/* dest1 is valid with rthdr only */
2980 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2981 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2982 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2983 	return len;
2984 #undef elen
2985 }
2986