xref: /freebsd/sys/netinet6/ip6_output.c (revision 7b71f57f4e514a2ab7308ce4147e14d90e099ad0)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  */
62 
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include "opt_ipsec.h"
66 #include "opt_kern_tls.h"
67 #include "opt_ratelimit.h"
68 #include "opt_route.h"
69 #include "opt_rss.h"
70 #include "opt_sctp.h"
71 
72 #include <sys/param.h>
73 #include <sys/kernel.h>
74 #include <sys/ktls.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_private.h>
91 #include <net/if_vlan_var.h>
92 #include <net/if_llatbl.h>
93 #include <net/ethernet.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/route/nhop.h>
97 #include <net/pfil.h>
98 #include <net/rss_config.h>
99 #include <net/vnet.h>
100 
101 #include <netinet/in.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip_var.h>
104 #include <netinet6/in6_fib.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet/ip6.h>
107 #include <netinet/icmp6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/tcp_var.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_rss.h>
113 
114 #include <netipsec/ipsec_support.h>
115 #if defined(SCTP) || defined(SCTP_SUPPORT)
116 #include <netinet/sctp.h>
117 #include <netinet/sctp_crc32.h>
118 #endif
119 
120 #include <netinet6/scope6_var.h>
121 
122 extern int in6_mcast_loop;
123 
124 struct ip6_exthdrs {
125 	struct mbuf *ip6e_ip6;
126 	struct mbuf *ip6e_hbh;
127 	struct mbuf *ip6e_dest1;
128 	struct mbuf *ip6e_rthdr;
129 	struct mbuf *ip6e_dest2;
130 };
131 
132 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
133 
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 			   struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 	struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct inpcb *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 	struct ucred *, int, int, int);
141 
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 	struct ip6_frag **);
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static void ip6_getpmtu(struct route_in6 *, int,
148 	struct ifnet *, const struct in6_addr *, u_long *, u_int, u_int);
149 static void ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
150 	u_long *, u_int);
151 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
152 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
153 
154 /*
155  * Make an extension header from option data.  hp is the source,
156  * mp is the destination, and _ol is the optlen.
157  */
158 #define	MAKE_EXTHDR(hp, mp, _ol)					\
159     do {								\
160 	struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
161 	error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
162 	    ((eh)->ip6e_len + 1) << 3);				\
163 	if (error)						\
164 		goto freehdrs;					\
165 	(_ol) += (*(mp))->m_len;				\
166     } while (/*CONSTCOND*/ 0)
167 
168 /*
169  * Form a chain of extension headers.
170  * m is the extension header mbuf
171  * mp is the previous mbuf in the chain
172  * p is the next header
173  * i is the type of option.
174  */
175 #define MAKE_CHAIN(m, mp, p, i)\
176     do {\
177 	if (m) {\
178 		if (!hdrsplit) \
179 			panic("%s:%d: assumption failed: "\
180 			    "hdr not split: hdrsplit %d exthdrs %p",\
181 			    __func__, __LINE__, hdrsplit, &exthdrs);\
182 		*mtod((m), u_char *) = *(p);\
183 		*(p) = (i);\
184 		p = mtod((m), u_char *);\
185 		(m)->m_next = (mp)->m_next;\
186 		(mp)->m_next = (m);\
187 		(mp) = (m);\
188 	}\
189     } while (/*CONSTCOND*/ 0)
190 
191 void
in6_delayed_cksum(struct mbuf * m,uint32_t plen,u_short offset)192 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
193 {
194 	u_short csum;
195 
196 	csum = in_cksum_skip(m, offset + plen, offset);
197 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
198 		csum = 0xffff;
199 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
200 
201 	if (offset + sizeof(csum) > m->m_len)
202 		m_copyback(m, offset, sizeof(csum), (caddr_t)&csum);
203 	else
204 		*(u_short *)mtodo(m, offset) = csum;
205 }
206 
207 static void
ip6_output_delayed_csum(struct mbuf * m,struct ifnet * ifp,int csum_flags,int plen,int optlen)208 ip6_output_delayed_csum(struct mbuf *m, struct ifnet *ifp, int csum_flags,
209     int plen, int optlen)
210 {
211 
212 	KASSERT((plen >= optlen), ("%s:%d: plen %d < optlen %d, m %p, ifp %p "
213 	    "csum_flags %#x",
214 	    __func__, __LINE__, plen, optlen, m, ifp, csum_flags));
215 
216 	if (csum_flags & CSUM_DELAY_DATA_IPV6) {
217 		in6_delayed_cksum(m, plen - optlen,
218 		    sizeof(struct ip6_hdr) + optlen);
219 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
220 	}
221 #if defined(SCTP) || defined(SCTP_SUPPORT)
222 	if (csum_flags & CSUM_SCTP_IPV6) {
223 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr) + optlen);
224 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
225 	}
226 #endif
227 }
228 
229 int
ip6_fragment(struct ifnet * ifp,struct mbuf * m0,int hlen,u_char nextproto,int fraglen,uint32_t id)230 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
231     int fraglen , uint32_t id)
232 {
233 	struct mbuf *m, **mnext, *m_frgpart;
234 	struct ip6_hdr *ip6, *mhip6;
235 	struct ip6_frag *ip6f;
236 	int off;
237 	int error;
238 	int tlen = m0->m_pkthdr.len;
239 
240 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
241 
242 	m = m0;
243 	ip6 = mtod(m, struct ip6_hdr *);
244 	mnext = &m->m_nextpkt;
245 
246 	for (off = hlen; off < tlen; off += fraglen) {
247 		m = m_gethdr(M_NOWAIT, MT_DATA);
248 		if (!m) {
249 			IP6STAT_INC(ip6s_odropped);
250 			return (ENOBUFS);
251 		}
252 
253 		/*
254 		 * Make sure the complete packet header gets copied
255 		 * from the originating mbuf to the newly created
256 		 * mbuf. This also ensures that existing firewall
257 		 * classification(s), VLAN tags and so on get copied
258 		 * to the resulting fragmented packet(s):
259 		 */
260 		if (m_dup_pkthdr(m, m0, M_NOWAIT) == 0) {
261 			m_free(m);
262 			IP6STAT_INC(ip6s_odropped);
263 			return (ENOBUFS);
264 		}
265 
266 		*mnext = m;
267 		mnext = &m->m_nextpkt;
268 		m->m_data += max_linkhdr;
269 		mhip6 = mtod(m, struct ip6_hdr *);
270 		*mhip6 = *ip6;
271 		m->m_len = sizeof(*mhip6);
272 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
273 		if (error) {
274 			IP6STAT_INC(ip6s_odropped);
275 			return (error);
276 		}
277 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
278 		if (off + fraglen >= tlen)
279 			fraglen = tlen - off;
280 		else
281 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
282 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
283 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
284 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
285 			IP6STAT_INC(ip6s_odropped);
286 			return (ENOBUFS);
287 		}
288 		m_cat(m, m_frgpart);
289 		m->m_pkthdr.len = fraglen + hlen + sizeof(*ip6f);
290 		ip6f->ip6f_reserved = 0;
291 		ip6f->ip6f_ident = id;
292 		ip6f->ip6f_nxt = nextproto;
293 		IP6STAT_INC(ip6s_ofragments);
294 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
295 	}
296 
297 	return (0);
298 }
299 
300 static int
ip6_output_send(struct inpcb * inp,struct ifnet * ifp,struct ifnet * origifp,struct mbuf * m,struct sockaddr_in6 * dst,struct route_in6 * ro,bool stamp_tag)301 ip6_output_send(struct inpcb *inp, struct ifnet *ifp, struct ifnet *origifp,
302     struct mbuf *m, struct sockaddr_in6 *dst, struct route_in6 *ro,
303     bool stamp_tag)
304 {
305 #ifdef KERN_TLS
306 	struct ktls_session *tls = NULL;
307 #endif
308 	struct m_snd_tag *mst;
309 	int error;
310 
311 	MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
312 	mst = NULL;
313 
314 #ifdef KERN_TLS
315 	/*
316 	 * If this is an unencrypted TLS record, save a reference to
317 	 * the record.  This local reference is used to call
318 	 * ktls_output_eagain after the mbuf has been freed (thus
319 	 * dropping the mbuf's reference) in if_output.
320 	 */
321 	if (m->m_next != NULL && mbuf_has_tls_session(m->m_next)) {
322 		tls = ktls_hold(m->m_next->m_epg_tls);
323 		mst = tls->snd_tag;
324 
325 		/*
326 		 * If a TLS session doesn't have a valid tag, it must
327 		 * have had an earlier ifp mismatch, so drop this
328 		 * packet.
329 		 */
330 		if (mst == NULL) {
331 			m_freem(m);
332 			error = EAGAIN;
333 			goto done;
334 		}
335 		/*
336 		 * Always stamp tags that include NIC ktls.
337 		 */
338 		stamp_tag = true;
339 	}
340 #endif
341 #ifdef RATELIMIT
342 	if (inp != NULL && mst == NULL) {
343 		if ((inp->inp_flags2 & INP_RATE_LIMIT_CHANGED) != 0 ||
344 		    (inp->inp_snd_tag != NULL &&
345 		    inp->inp_snd_tag->ifp != ifp))
346 			in_pcboutput_txrtlmt(inp, ifp, m);
347 
348 		if (inp->inp_snd_tag != NULL)
349 			mst = inp->inp_snd_tag;
350 	}
351 #endif
352 	if (stamp_tag && mst != NULL) {
353 		KASSERT(m->m_pkthdr.rcvif == NULL,
354 		    ("trying to add a send tag to a forwarded packet"));
355 		if (mst->ifp != ifp) {
356 			m_freem(m);
357 			error = EAGAIN;
358 			goto done;
359 		}
360 
361 		/* stamp send tag on mbuf */
362 		m->m_pkthdr.snd_tag = m_snd_tag_ref(mst);
363 		m->m_pkthdr.csum_flags |= CSUM_SND_TAG;
364 	}
365 
366 	error = nd6_output_ifp(ifp, origifp, m, dst, (struct route *)ro);
367 
368 done:
369 	/* Check for route change invalidating send tags. */
370 #ifdef KERN_TLS
371 	if (tls != NULL) {
372 		if (error == EAGAIN)
373 			error = ktls_output_eagain(inp, tls);
374 		ktls_free(tls);
375 	}
376 #endif
377 #ifdef RATELIMIT
378 	if (error == EAGAIN)
379 		in_pcboutput_eagain(inp);
380 #endif
381 	return (error);
382 }
383 
384 /*
385  * IP6 output.
386  * The packet in mbuf chain m contains a skeletal IP6 header (with pri, len,
387  * nxt, hlim, src, dst).
388  * This function may modify ver and hlim only.
389  * The mbuf chain containing the packet will be freed.
390  * The mbuf opt, if present, will not be freed.
391  * If route_in6 ro is present and has ro_nh initialized, route lookup would be
392  * skipped and ro->ro_nh would be used. If ro is present but ro->ro_nh is NULL,
393  * then result of route lookup is stored in ro->ro_nh.
394  *
395  * Type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and nd_ifinfo.linkmtu
396  * is uint32_t.  So we use u_long to hold largest one, which is rt_mtu.
397  *
398  * ifpp - XXX: just for statistics
399  */
400 int
ip6_output(struct mbuf * m0,struct ip6_pktopts * opt,struct route_in6 * ro,int flags,struct ip6_moptions * im6o,struct ifnet ** ifpp,struct inpcb * inp)401 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
402     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
403     struct ifnet **ifpp, struct inpcb *inp)
404 {
405 	struct ip6_hdr *ip6;
406 	struct ifnet *ifp, *origifp;
407 	struct mbuf *m = m0;
408 	struct mbuf *mprev;
409 	struct route_in6 *ro_pmtu;
410 	struct nhop_object *nh;
411 	struct sockaddr_in6 *dst, sin6, src_sa, dst_sa;
412 	struct in6_addr odst;
413 	u_char *nexthdrp;
414 	int tlen, len;
415 	int error = 0;
416 	int vlan_pcp = -1;
417 	struct in6_ifaddr *ia = NULL;
418 	u_long mtu;
419 	int dontfrag;
420 	u_int32_t optlen, plen = 0, unfragpartlen;
421 	struct ip6_exthdrs exthdrs;
422 	struct in6_addr src0, dst0;
423 	u_int32_t zone;
424 	bool hdrsplit;
425 	int sw_csum, tso;
426 	int needfiblookup;
427 	uint32_t fibnum;
428 	struct m_tag *fwd_tag = NULL;
429 	uint32_t id;
430 	uint32_t optvalid;
431 
432 	NET_EPOCH_ASSERT();
433 
434 	if (inp != NULL) {
435 		INP_LOCK_ASSERT(inp);
436 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
437 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
438 			/* Unconditionally set flowid. */
439 			m->m_pkthdr.flowid = inp->inp_flowid;
440 			M_HASHTYPE_SET(m, inp->inp_flowtype);
441 		}
442 		if ((inp->inp_flags2 & INP_2PCP_SET) != 0)
443 			vlan_pcp = (inp->inp_flags2 & INP_2PCP_MASK) >>
444 			    INP_2PCP_SHIFT;
445 #ifdef NUMA
446 		m->m_pkthdr.numa_domain = inp->inp_numa_domain;
447 #endif
448 	}
449 
450 	/* Source address validation. */
451 	ip6 = mtod(m, struct ip6_hdr *);
452 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
453 	    (flags & IPV6_UNSPECSRC) == 0) {
454 		error = EOPNOTSUPP;
455 		IP6STAT_INC(ip6s_badscope);
456 		goto bad;
457 	}
458 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
459 		error = EOPNOTSUPP;
460 		IP6STAT_INC(ip6s_badscope);
461 		goto bad;
462 	}
463 
464 	/*
465 	 * If we are given packet options to add extension headers prepare them.
466 	 * Calculate the total length of the extension header chain.
467 	 * Keep the length of the unfragmentable part for fragmentation.
468 	 */
469 	bzero(&exthdrs, sizeof(exthdrs));
470 	optlen = optvalid = 0;
471 	unfragpartlen = sizeof(struct ip6_hdr);
472 	if (opt) {
473 		optvalid = opt->ip6po_valid;
474 
475 		/* Hop-by-Hop options header. */
476 		if ((optvalid & IP6PO_VALID_HBH) != 0)
477 			MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh, optlen);
478 
479 		/* Destination options header (1st part). */
480 		if ((optvalid & IP6PO_VALID_RHINFO) != 0) {
481 #ifndef RTHDR_SUPPORT_IMPLEMENTED
482 			/*
483 			 * If there is a routing header, discard the packet
484 			 * right away here. RH0/1 are obsolete and we do not
485 			 * currently support RH2/3/4.
486 			 * People trying to use RH253/254 may want to disable
487 			 * this check.
488 			 * The moment we do support any routing header (again)
489 			 * this block should check the routing type more
490 			 * selectively.
491 			 */
492 			error = EINVAL;
493 			goto bad;
494 #endif
495 
496 			/*
497 			 * Destination options header (1st part).
498 			 * This only makes sense with a routing header.
499 			 * See Section 9.2 of RFC 3542.
500 			 * Disabling this part just for MIP6 convenience is
501 			 * a bad idea.  We need to think carefully about a
502 			 * way to make the advanced API coexist with MIP6
503 			 * options, which might automatically be inserted in
504 			 * the kernel.
505 			 */
506 			if ((optvalid & IP6PO_VALID_DEST1) != 0)
507 				MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1,
508 				    optlen);
509 		}
510 		/* Routing header. */
511 		if ((optvalid & IP6PO_VALID_RHINFO) != 0)
512 			MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr, optlen);
513 
514 		unfragpartlen += optlen;
515 
516 		/*
517 		 * NOTE: we don't add AH/ESP length here (done in
518 		 * ip6_ipsec_output()).
519 		 */
520 
521 		/* Destination options header (2nd part). */
522 		if ((optvalid & IP6PO_VALID_DEST2) != 0)
523 			MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2, optlen);
524 	}
525 
526 	/*
527 	 * If there is at least one extension header,
528 	 * separate IP6 header from the payload.
529 	 */
530 	hdrsplit = false;
531 	if (optlen) {
532 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
533 			m = NULL;
534 			goto freehdrs;
535 		}
536 		m = exthdrs.ip6e_ip6;
537 		ip6 = mtod(m, struct ip6_hdr *);
538 		hdrsplit = true;
539 	}
540 
541 	/* Adjust mbuf packet header length. */
542 	m->m_pkthdr.len += optlen;
543 	plen = m->m_pkthdr.len - sizeof(*ip6);
544 
545 	/* If this is a jumbo payload, insert a jumbo payload option. */
546 	if (plen > IPV6_MAXPACKET) {
547 		if (!hdrsplit) {
548 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
549 				m = NULL;
550 				goto freehdrs;
551 			}
552 			m = exthdrs.ip6e_ip6;
553 			ip6 = mtod(m, struct ip6_hdr *);
554 			hdrsplit = true;
555 		}
556 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
557 			goto freehdrs;
558 		ip6->ip6_plen = 0;
559 	} else
560 		ip6->ip6_plen = htons(plen);
561 	nexthdrp = &ip6->ip6_nxt;
562 
563 	if (optlen) {
564 		/*
565 		 * Concatenate headers and fill in next header fields.
566 		 * Here we have, on "m"
567 		 *	IPv6 payload
568 		 * and we insert headers accordingly.
569 		 * Finally, we should be getting:
570 		 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload].
571 		 *
572 		 * During the header composing process "m" points to IPv6
573 		 * header.  "mprev" points to an extension header prior to esp.
574 		 */
575 		mprev = m;
576 
577 		/*
578 		 * We treat dest2 specially.  This makes IPsec processing
579 		 * much easier.  The goal here is to make mprev point the
580 		 * mbuf prior to dest2.
581 		 *
582 		 * Result: IPv6 dest2 payload.
583 		 * m and mprev will point to IPv6 header.
584 		 */
585 		if (exthdrs.ip6e_dest2) {
586 			if (!hdrsplit)
587 				panic("%s:%d: assumption failed: "
588 				    "hdr not split: hdrsplit %d exthdrs %p",
589 				    __func__, __LINE__, hdrsplit, &exthdrs);
590 			exthdrs.ip6e_dest2->m_next = m->m_next;
591 			m->m_next = exthdrs.ip6e_dest2;
592 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
593 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
594 		}
595 
596 		/*
597 		 * Result: IPv6 hbh dest1 rthdr dest2 payload.
598 		 * m will point to IPv6 header.  mprev will point to the
599 		 * extension header prior to dest2 (rthdr in the above case).
600 		 */
601 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
602 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
603 			   IPPROTO_DSTOPTS);
604 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
605 			   IPPROTO_ROUTING);
606 	}
607 
608 	IP6STAT_INC(ip6s_localout);
609 
610 	/* Route packet. */
611 	ro_pmtu = ro;
612 	if ((optvalid & IP6PO_VALID_RHINFO) != 0)
613 		ro = &opt->ip6po_route;
614 	if (ro != NULL)
615 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
616 	else
617 		dst = &sin6;
618 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
619 
620 again:
621 	/*
622 	 * If specified, try to fill in the traffic class field.
623 	 * Do not override if a non-zero value is already set.
624 	 * We check the diffserv field and the ECN field separately.
625 	 */
626 	if ((optvalid & IP6PO_VALID_TC) != 0){
627 		int mask = 0;
628 
629 		if (IPV6_DSCP(ip6) == 0)
630 			mask |= 0xfc;
631 		if (IPV6_ECN(ip6) == 0)
632 			mask |= 0x03;
633 		if (mask != 0)
634 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
635 	}
636 
637 	/* Fill in or override the hop limit field, if necessary. */
638 	if ((optvalid & IP6PO_VALID_HLIM) != 0)
639 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
640 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
641 		if (im6o != NULL)
642 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
643 		else
644 			ip6->ip6_hlim = V_ip6_defmcasthlim;
645 	}
646 
647 	if (ro == NULL || ro->ro_nh == NULL) {
648 		bzero(dst, sizeof(*dst));
649 		dst->sin6_family = AF_INET6;
650 		dst->sin6_len = sizeof(*dst);
651 		dst->sin6_addr = ip6->ip6_dst;
652 	}
653 	/*
654 	 * Validate route against routing table changes.
655 	 * Make sure that the address family is set in route.
656 	 */
657 	nh = NULL;
658 	ifp = NULL;
659 	mtu = 0;
660 	if (ro != NULL) {
661 		if (ro->ro_nh != NULL && inp != NULL) {
662 			ro->ro_dst.sin6_family = AF_INET6; /* XXX KASSERT? */
663 			NH_VALIDATE((struct route *)ro, &inp->inp_rt_cookie,
664 			    fibnum);
665 		}
666 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
667 		    (!NH_IS_VALID(ro->ro_nh) ||
668 		    ro->ro_dst.sin6_family != AF_INET6 ||
669 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)))
670 			RO_INVALIDATE_CACHE(ro);
671 
672 		if (ro->ro_nh != NULL && fwd_tag == NULL &&
673 		    ro->ro_dst.sin6_family == AF_INET6 &&
674 		    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
675 			/* Nexthop is valid and contains valid ifp */
676 			nh = ro->ro_nh;
677 		} else {
678 			if (ro->ro_lle)
679 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
680 			ro->ro_lle = NULL;
681 			if (fwd_tag == NULL) {
682 				bzero(&dst_sa, sizeof(dst_sa));
683 				dst_sa.sin6_family = AF_INET6;
684 				dst_sa.sin6_len = sizeof(dst_sa);
685 				dst_sa.sin6_addr = ip6->ip6_dst;
686 			}
687 			error = in6_selectroute(&dst_sa, opt, im6o, ro, &ifp,
688 			    &nh, fibnum, m->m_pkthdr.flowid);
689 			if (error != 0) {
690 				IP6STAT_INC(ip6s_noroute);
691 				if (ifp != NULL)
692 					in6_ifstat_inc(ifp, ifs6_out_discard);
693 				goto bad;
694 			}
695 			/*
696 			 * At this point at least @ifp is not NULL
697 			 * Can be the case when dst is multicast, link-local or
698 			 * interface is explicitly specificed by the caller.
699 			 */
700 		}
701 		if (nh == NULL) {
702 			/*
703 			 * If in6_selectroute() does not return a nexthop
704 			 * dst may not have been updated.
705 			 */
706 			*dst = dst_sa;	/* XXX */
707 			origifp = ifp;
708 			mtu = ifp->if_mtu;
709 		} else {
710 			ifp = nh->nh_ifp;
711 			origifp = nh->nh_aifp;
712 			ia = (struct in6_ifaddr *)(nh->nh_ifa);
713 			counter_u64_add(nh->nh_pksent, 1);
714 		}
715 	} else {
716 		struct nhop_object *nh;
717 		struct in6_addr kdst;
718 		uint32_t scopeid;
719 
720 		if (fwd_tag == NULL) {
721 			bzero(&dst_sa, sizeof(dst_sa));
722 			dst_sa.sin6_family = AF_INET6;
723 			dst_sa.sin6_len = sizeof(dst_sa);
724 			dst_sa.sin6_addr = ip6->ip6_dst;
725 		}
726 
727 		if (IN6_IS_ADDR_MULTICAST(&dst_sa.sin6_addr) &&
728 		    im6o != NULL &&
729 		    (ifp = im6o->im6o_multicast_ifp) != NULL) {
730 			/* We do not need a route lookup. */
731 			*dst = dst_sa;	/* XXX */
732 			origifp = ifp;
733 			goto nonh6lookup;
734 		}
735 
736 		in6_splitscope(&dst_sa.sin6_addr, &kdst, &scopeid);
737 
738 		if (IN6_IS_ADDR_MC_LINKLOCAL(&dst_sa.sin6_addr) ||
739 		    IN6_IS_ADDR_MC_NODELOCAL(&dst_sa.sin6_addr)) {
740 			if (scopeid > 0) {
741 				ifp = in6_getlinkifnet(scopeid);
742 				if (ifp == NULL) {
743 					error = EHOSTUNREACH;
744 					goto bad;
745 				}
746 				*dst = dst_sa;	/* XXX */
747 				origifp = ifp;
748 				goto nonh6lookup;
749 			}
750 		}
751 
752 		nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE,
753 		    m->m_pkthdr.flowid);
754 		if (nh == NULL) {
755 			IP6STAT_INC(ip6s_noroute);
756 			/* No ifp in6_ifstat_inc(ifp, ifs6_out_discard); */
757 			error = EHOSTUNREACH;
758 			goto bad;
759 		}
760 
761 		ifp = nh->nh_ifp;
762 		origifp = nh->nh_aifp;
763 		ia = ifatoia6(nh->nh_ifa);
764 		if (nh->nh_flags & NHF_GATEWAY)
765 			dst->sin6_addr = nh->gw6_sa.sin6_addr;
766 		else if (fwd_tag != NULL)
767 			dst->sin6_addr = dst_sa.sin6_addr;
768 nonh6lookup:
769 		;
770 	}
771 	/*
772 	 * At this point ifp MUST be pointing to the valid transmit ifp.
773 	 * origifp MUST be valid and pointing to either the same ifp or,
774 	 * in case of loopback output, to the interface which ip6_src
775 	 * belongs to.
776 	 * Examples:
777 	 *  fe80::1%em0 -> fe80::2%em0 -> ifp=em0, origifp=em0
778 	 *  fe80::1%em0 -> fe80::1%em0 -> ifp=lo0, origifp=em0
779 	 *  ::1 -> ::1 -> ifp=lo0, origifp=lo0
780 	 *
781 	 * mtu can be 0 and will be refined later.
782 	 */
783 	KASSERT((ifp != NULL), ("output interface must not be NULL"));
784 	KASSERT((origifp != NULL), ("output address interface must not be NULL"));
785 
786 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
787 	/*
788 	 * IPSec checking which handles several cases.
789 	 * FAST IPSEC: We re-injected the packet.
790 	 * XXX: need scope argument.
791 	 */
792 	if (IPSEC_ENABLED(ipv6)) {
793 		if ((error = IPSEC_OUTPUT(ipv6, ifp, m, inp, mtu == 0 ?
794 		    ifp->if_mtu : mtu)) != 0) {
795 			if (error == EINPROGRESS)
796 				error = 0;
797 			goto done;
798 		}
799 	}
800 #endif /* IPSEC */
801 
802 	if ((flags & IPV6_FORWARDING) == 0) {
803 		/* XXX: the FORWARDING flag can be set for mrouting. */
804 		in6_ifstat_inc(ifp, ifs6_out_request);
805 	}
806 
807 	/* Setup data structures for scope ID checks. */
808 	src0 = ip6->ip6_src;
809 	bzero(&src_sa, sizeof(src_sa));
810 	src_sa.sin6_family = AF_INET6;
811 	src_sa.sin6_len = sizeof(src_sa);
812 	src_sa.sin6_addr = ip6->ip6_src;
813 
814 	dst0 = ip6->ip6_dst;
815 	/* Re-initialize to be sure. */
816 	bzero(&dst_sa, sizeof(dst_sa));
817 	dst_sa.sin6_family = AF_INET6;
818 	dst_sa.sin6_len = sizeof(dst_sa);
819 	dst_sa.sin6_addr = ip6->ip6_dst;
820 
821 	/* Check for valid scope ID. */
822 	if (in6_setscope(&src0, origifp, &zone) == 0 &&
823 	    sa6_recoverscope(&src_sa) == 0 && zone == src_sa.sin6_scope_id &&
824 	    in6_setscope(&dst0, origifp, &zone) == 0 &&
825 	    sa6_recoverscope(&dst_sa) == 0 && zone == dst_sa.sin6_scope_id) {
826 		/*
827 		 * The outgoing interface is in the zone of the source
828 		 * and destination addresses.
829 		 *
830 		 */
831 	} else if ((origifp->if_flags & IFF_LOOPBACK) == 0 ||
832 	    sa6_recoverscope(&src_sa) != 0 ||
833 	    sa6_recoverscope(&dst_sa) != 0 ||
834 	    dst_sa.sin6_scope_id == 0 ||
835 	    (src_sa.sin6_scope_id != 0 &&
836 	    src_sa.sin6_scope_id != dst_sa.sin6_scope_id) ||
837 	    ifnet_byindex(dst_sa.sin6_scope_id) == NULL) {
838 		/*
839 		 * If the destination network interface is not a
840 		 * loopback interface, or the destination network
841 		 * address has no scope ID, or the source address has
842 		 * a scope ID set which is different from the
843 		 * destination address one, or there is no network
844 		 * interface representing this scope ID, the address
845 		 * pair is considered invalid.
846 		 */
847 		IP6STAT_INC(ip6s_badscope);
848 		in6_ifstat_inc(origifp, ifs6_out_discard);
849 		if (error == 0)
850 			error = EHOSTUNREACH; /* XXX */
851 		goto bad;
852 	}
853 	/* All scope ID checks are successful. */
854 
855 	if (nh && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
856 		if ((optvalid & IP6PO_VALID_NHINFO) != 0) {
857 			/*
858 			 * The nexthop is explicitly specified by the
859 			 * application.  We assume the next hop is an IPv6
860 			 * address.
861 			 */
862 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
863 		}
864 		else if ((nh->nh_flags & NHF_GATEWAY))
865 			dst = &nh->gw6_sa;
866 	}
867 
868 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
869 		m->m_flags &= ~(M_BCAST | M_MCAST); /* Just in case. */
870 	} else {
871 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
872 		in6_ifstat_inc(ifp, ifs6_out_mcast);
873 
874 		/* Confirm that the outgoing interface supports multicast. */
875 		if (!(ifp->if_flags & IFF_MULTICAST)) {
876 			IP6STAT_INC(ip6s_noroute);
877 			in6_ifstat_inc(ifp, ifs6_out_discard);
878 			error = ENETUNREACH;
879 			goto bad;
880 		}
881 		if ((im6o == NULL && in6_mcast_loop) ||
882 		    (im6o && im6o->im6o_multicast_loop)) {
883 			/*
884 			 * Loop back multicast datagram if not expressly
885 			 * forbidden to do so, even if we have not joined
886 			 * the address; protocols will filter it later,
887 			 * thus deferring a hash lookup and lock acquisition
888 			 * at the expense of an m_copym().
889 			 */
890 			ip6_mloopback(ifp, m);
891 		} else {
892 			/*
893 			 * If we are acting as a multicast router, perform
894 			 * multicast forwarding as if the packet had just
895 			 * arrived on the interface to which we are about
896 			 * to send.  The multicast forwarding function
897 			 * recursively calls this function, using the
898 			 * IPV6_FORWARDING flag to prevent infinite recursion.
899 			 *
900 			 * Multicasts that are looped back by ip6_mloopback(),
901 			 * above, will be forwarded by the ip6_input() routine,
902 			 * if necessary.
903 			 */
904 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
905 				/*
906 				 * XXX: ip6_mforward expects that rcvif is NULL
907 				 * when it is called from the originating path.
908 				 * However, it may not always be the case.
909 				 */
910 				m->m_pkthdr.rcvif = NULL;
911 				if (ip6_mforward(ip6, ifp, m) != 0) {
912 					m_freem(m);
913 					goto done;
914 				}
915 			}
916 		}
917 		/*
918 		 * Multicasts with a hoplimit of zero may be looped back,
919 		 * above, but must not be transmitted on a network.
920 		 * Also, multicasts addressed to the loopback interface
921 		 * are not sent -- the above call to ip6_mloopback() will
922 		 * loop back a copy if this host actually belongs to the
923 		 * destination group on the loopback interface.
924 		 */
925 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
926 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
927 			m_freem(m);
928 			goto done;
929 		}
930 	}
931 
932 	/*
933 	 * Fill the outgoing inteface to tell the upper layer
934 	 * to increment per-interface statistics.
935 	 */
936 	if (ifpp)
937 		*ifpp = ifp;
938 
939 	/* Determine path MTU. */
940 	ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst, &mtu, fibnum,
941 	    *nexthdrp);
942 	KASSERT(mtu > 0, ("%s:%d: mtu %ld, ro_pmtu %p ro %p ifp %p fibnum %u",
943 	    __func__, __LINE__, mtu, ro_pmtu, ro, ifp, fibnum));
944 
945 	/*
946 	 * The caller of this function may specify to use the minimum MTU
947 	 * in some cases.
948 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
949 	 * setting.  The logic is a bit complicated; by default, unicast
950 	 * packets will follow path MTU while multicast packets will be sent at
951 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
952 	 * including unicast ones will be sent at the minimum MTU.  Multicast
953 	 * packets will always be sent at the minimum MTU unless
954 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
955 	 * See RFC 3542 for more details.
956 	 */
957 	if (mtu > IPV6_MMTU) {
958 		if ((flags & IPV6_MINMTU))
959 			mtu = IPV6_MMTU;
960 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
961 			mtu = IPV6_MMTU;
962 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
963 			 (opt == NULL ||
964 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
965 			mtu = IPV6_MMTU;
966 		}
967 	}
968 
969 	/*
970 	 * Clear embedded scope identifiers if necessary.
971 	 * in6_clearscope() will touch the addresses only when necessary.
972 	 */
973 	in6_clearscope(&ip6->ip6_src);
974 	in6_clearscope(&ip6->ip6_dst);
975 
976 	/*
977 	 * If the outgoing packet contains a hop-by-hop options header,
978 	 * it must be examined and processed even by the source node.
979 	 * (RFC 2460, section 4.)
980 	 */
981 	if (exthdrs.ip6e_hbh) {
982 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
983 		u_int32_t dummy; /* XXX unused */
984 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
985 
986 #ifdef DIAGNOSTIC
987 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
988 			panic("ip6e_hbh is not contiguous");
989 #endif
990 		/*
991 		 *  XXX: if we have to send an ICMPv6 error to the sender,
992 		 *       we need the M_LOOP flag since icmp6_error() expects
993 		 *       the IPv6 and the hop-by-hop options header are
994 		 *       contiguous unless the flag is set.
995 		 */
996 		m->m_flags |= M_LOOP;
997 		m->m_pkthdr.rcvif = ifp;
998 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
999 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
1000 		    &dummy, &plen) < 0) {
1001 			/* m was already freed at this point. */
1002 			error = EINVAL;/* better error? */
1003 			goto done;
1004 		}
1005 		m->m_flags &= ~M_LOOP; /* XXX */
1006 		m->m_pkthdr.rcvif = NULL;
1007 	}
1008 
1009 	/* Jump over all PFIL processing if hooks are not active. */
1010 	if (!PFIL_HOOKED_OUT(V_inet6_pfil_head))
1011 		goto passout;
1012 
1013 	odst = ip6->ip6_dst;
1014 	/* Run through list of hooks for output packets. */
1015 	switch (pfil_mbuf_out(V_inet6_pfil_head, &m, ifp, inp)) {
1016 	case PFIL_PASS:
1017 		ip6 = mtod(m, struct ip6_hdr *);
1018 		break;
1019 	case PFIL_DROPPED:
1020 		error = EACCES;
1021 		/* FALLTHROUGH */
1022 	case PFIL_CONSUMED:
1023 		goto done;
1024 	}
1025 
1026 	needfiblookup = 0;
1027 	/* See if destination IP address was changed by packet filter. */
1028 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
1029 		m->m_flags |= M_SKIP_FIREWALL;
1030 		/* If destination is now ourself drop to ip6_input(). */
1031 		if (in6_localip(&ip6->ip6_dst)) {
1032 			m->m_flags |= M_FASTFWD_OURS;
1033 			if (m->m_pkthdr.rcvif == NULL)
1034 				m->m_pkthdr.rcvif = V_loif;
1035 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1036 				m->m_pkthdr.csum_flags |=
1037 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1038 				m->m_pkthdr.csum_data = 0xffff;
1039 			}
1040 #if defined(SCTP) || defined(SCTP_SUPPORT)
1041 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1042 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1043 #endif
1044 			error = netisr_queue(NETISR_IPV6, m);
1045 			goto done;
1046 		} else {
1047 			if (ro != NULL)
1048 				RO_INVALIDATE_CACHE(ro);
1049 			needfiblookup = 1; /* Redo the routing table lookup. */
1050 		}
1051 	}
1052 	/* See if fib was changed by packet filter. */
1053 	if (fibnum != M_GETFIB(m)) {
1054 		m->m_flags |= M_SKIP_FIREWALL;
1055 		fibnum = M_GETFIB(m);
1056 		if (ro != NULL)
1057 			RO_INVALIDATE_CACHE(ro);
1058 		needfiblookup = 1;
1059 	}
1060 	if (needfiblookup)
1061 		goto again;
1062 
1063 	/* See if local, if yes, send it to netisr. */
1064 	if (m->m_flags & M_FASTFWD_OURS) {
1065 		if (m->m_pkthdr.rcvif == NULL)
1066 			m->m_pkthdr.rcvif = V_loif;
1067 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1068 			m->m_pkthdr.csum_flags |=
1069 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
1070 			m->m_pkthdr.csum_data = 0xffff;
1071 		}
1072 #if defined(SCTP) || defined(SCTP_SUPPORT)
1073 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
1074 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
1075 #endif
1076 		error = netisr_queue(NETISR_IPV6, m);
1077 		goto done;
1078 	}
1079 	/* Or forward to some other address? */
1080 	if ((m->m_flags & M_IP6_NEXTHOP) &&
1081 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
1082 		if (ro != NULL)
1083 			dst = (struct sockaddr_in6 *)&ro->ro_dst;
1084 		else
1085 			dst = &sin6;
1086 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
1087 		m->m_flags |= M_SKIP_FIREWALL;
1088 		m->m_flags &= ~M_IP6_NEXTHOP;
1089 		m_tag_delete(m, fwd_tag);
1090 		goto again;
1091 	}
1092 
1093 passout:
1094 	if (vlan_pcp > -1)
1095 		EVL_APPLY_PRI(m, vlan_pcp);
1096 
1097 	/* Ensure the packet data is mapped if the interface requires it. */
1098 	if ((ifp->if_capenable & IFCAP_MEXTPG) == 0) {
1099 		struct mbuf *m1;
1100 
1101 		error = mb_unmapped_to_ext(m, &m1);
1102 		if (error != 0) {
1103 			if (error == EINVAL) {
1104 				if_printf(ifp, "TLS packet\n");
1105 				/* XXXKIB */
1106 			} else if (error == ENOMEM) {
1107 				error = ENOBUFS;
1108 			}
1109 			IP6STAT_INC(ip6s_odropped);
1110 			return (error);
1111 		} else {
1112 			m = m1;
1113 		}
1114 	}
1115 
1116 	/*
1117 	 * Send the packet to the outgoing interface.
1118 	 * If necessary, do IPv6 fragmentation before sending.
1119 	 *
1120 	 * 1: normal case (dontfrag == 0)
1121 	 * 1-a:	send as is if tlen <= path mtu
1122 	 * 1-b:	fragment if tlen > path mtu
1123 	 *
1124 	 * 2: if user asks us not to fragment (dontfrag == 1)
1125 	 * 2-a:	send as is if tlen <= interface mtu
1126 	 * 2-b:	error if tlen > interface mtu
1127 	 */
1128 	sw_csum = m->m_pkthdr.csum_flags;
1129 	if (!hdrsplit) {
1130 		tso = ((sw_csum & ifp->if_hwassist &
1131 		    (CSUM_TSO | CSUM_INNER_TSO)) != 0) ? 1 : 0;
1132 		sw_csum &= ~ifp->if_hwassist;
1133 	} else
1134 		tso = 0;
1135 	/*
1136 	 * If we added extension headers, we will not do TSO and calculate the
1137 	 * checksums ourselves for now.
1138 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
1139 	 * with ext. hdrs.
1140 	 */
1141 	ip6_output_delayed_csum(m, ifp, sw_csum, plen, optlen);
1142 	/* XXX-BZ m->m_pkthdr.csum_flags &= ~ifp->if_hwassist; */
1143 	tlen = m->m_pkthdr.len;
1144 
1145 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
1146 		dontfrag = 1;
1147 	else
1148 		dontfrag = 0;
1149 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* Case 2-b. */
1150 		/*
1151 		 * If the DONTFRAG option is specified, we cannot send the
1152 		 * packet when the data length is larger than the MTU of the
1153 		 * outgoing interface.
1154 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
1155 		 * application wanted to know the MTU value. Also return an
1156 		 * error code (this is not described in the API spec).
1157 		 */
1158 		if (inp != NULL)
1159 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
1160 		error = EMSGSIZE;
1161 		goto bad;
1162 	}
1163 
1164 	/* Transmit packet without fragmentation. */
1165 	if (dontfrag || tlen <= mtu) {	/* Cases 1-a and 2-a. */
1166 		struct in6_ifaddr *ia6;
1167 
1168 		ip6 = mtod(m, struct ip6_hdr *);
1169 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1170 		if (ia6) {
1171 			/* Record statistics for this interface address. */
1172 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
1173 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
1174 			    m->m_pkthdr.len);
1175 		}
1176 		error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1177 		    (flags & IP_NO_SND_TAG_RL) ? false : true);
1178 		goto done;
1179 	}
1180 
1181 	/* Try to fragment the packet.  Case 1-b. */
1182 	if (mtu < IPV6_MMTU) {
1183 		/* Path MTU cannot be less than IPV6_MMTU. */
1184 		error = EMSGSIZE;
1185 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1186 		goto bad;
1187 	} else if (ip6->ip6_plen == 0) {
1188 		/* Jumbo payload cannot be fragmented. */
1189 		error = EMSGSIZE;
1190 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1191 		goto bad;
1192 	} else {
1193 		u_char nextproto;
1194 
1195 		/*
1196 		 * Too large for the destination or interface;
1197 		 * fragment if possible.
1198 		 * Must be able to put at least 8 bytes per fragment.
1199 		 */
1200 		if (mtu > IPV6_MAXPACKET)
1201 			mtu = IPV6_MAXPACKET;
1202 
1203 		len = (mtu - unfragpartlen - sizeof(struct ip6_frag)) & ~7;
1204 		if (len < 8) {
1205 			error = EMSGSIZE;
1206 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1207 			goto bad;
1208 		}
1209 
1210 		/*
1211 		 * If the interface will not calculate checksums on
1212 		 * fragmented packets, then do it here.
1213 		 * XXX-BZ handle the hw offloading case.  Need flags.
1214 		 */
1215 		ip6_output_delayed_csum(m, ifp, m->m_pkthdr.csum_flags, plen,
1216 		    optlen);
1217 
1218 		/*
1219 		 * Change the next header field of the last header in the
1220 		 * unfragmentable part.
1221 		 */
1222 		if (exthdrs.ip6e_rthdr) {
1223 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1224 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1225 		} else if (exthdrs.ip6e_dest1) {
1226 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1227 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1228 		} else if (exthdrs.ip6e_hbh) {
1229 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1230 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1231 		} else {
1232 			ip6 = mtod(m, struct ip6_hdr *);
1233 			nextproto = ip6->ip6_nxt;
1234 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1235 		}
1236 
1237 		/*
1238 		 * Loop through length of segment after first fragment,
1239 		 * make new header and copy data of each part and link onto
1240 		 * chain.
1241 		 */
1242 		m0 = m;
1243 		id = htonl(ip6_randomid());
1244 		error = ip6_fragment(ifp, m, unfragpartlen, nextproto,len, id);
1245 		if (error != 0)
1246 			goto sendorfree;
1247 
1248 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1249 	}
1250 
1251 	/* Remove leading garbage. */
1252 sendorfree:
1253 	m = m0->m_nextpkt;
1254 	m0->m_nextpkt = 0;
1255 	m_freem(m0);
1256 	for (; m; m = m0) {
1257 		m0 = m->m_nextpkt;
1258 		m->m_nextpkt = 0;
1259 		if (error == 0) {
1260 			/* Record statistics for this interface address. */
1261 			if (ia) {
1262 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1263 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1264 				    m->m_pkthdr.len);
1265 			}
1266 			if (vlan_pcp > -1)
1267 				EVL_APPLY_PRI(m, vlan_pcp);
1268 			error = ip6_output_send(inp, ifp, origifp, m, dst, ro,
1269 			    true);
1270 		} else
1271 			m_freem(m);
1272 	}
1273 
1274 	if (error == 0)
1275 		IP6STAT_INC(ip6s_fragmented);
1276 
1277 done:
1278 	return (error);
1279 
1280 freehdrs:
1281 	m_freem(exthdrs.ip6e_hbh);	/* m_freem() checks if mbuf is NULL. */
1282 	m_freem(exthdrs.ip6e_dest1);
1283 	m_freem(exthdrs.ip6e_rthdr);
1284 	m_freem(exthdrs.ip6e_dest2);
1285 	/* FALLTHROUGH */
1286 bad:
1287 	if (m)
1288 		m_freem(m);
1289 	goto done;
1290 }
1291 
1292 static int
ip6_copyexthdr(struct mbuf ** mp,caddr_t hdr,int hlen)1293 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1294 {
1295 	struct mbuf *m;
1296 
1297 	if (hlen > MCLBYTES)
1298 		return (ENOBUFS); /* XXX */
1299 
1300 	if (hlen > MLEN)
1301 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1302 	else
1303 		m = m_get(M_NOWAIT, MT_DATA);
1304 	if (m == NULL)
1305 		return (ENOBUFS);
1306 	m->m_len = hlen;
1307 	if (hdr)
1308 		bcopy(hdr, mtod(m, caddr_t), hlen);
1309 
1310 	*mp = m;
1311 	return (0);
1312 }
1313 
1314 /*
1315  * Insert jumbo payload option.
1316  */
1317 static int
ip6_insert_jumboopt(struct ip6_exthdrs * exthdrs,u_int32_t plen)1318 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1319 {
1320 	struct mbuf *mopt;
1321 	u_char *optbuf;
1322 	u_int32_t v;
1323 
1324 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1325 
1326 	/*
1327 	 * If there is no hop-by-hop options header, allocate new one.
1328 	 * If there is one but it doesn't have enough space to store the
1329 	 * jumbo payload option, allocate a cluster to store the whole options.
1330 	 * Otherwise, use it to store the options.
1331 	 */
1332 	if (exthdrs->ip6e_hbh == NULL) {
1333 		mopt = m_get(M_NOWAIT, MT_DATA);
1334 		if (mopt == NULL)
1335 			return (ENOBUFS);
1336 		mopt->m_len = JUMBOOPTLEN;
1337 		optbuf = mtod(mopt, u_char *);
1338 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1339 		exthdrs->ip6e_hbh = mopt;
1340 	} else {
1341 		struct ip6_hbh *hbh;
1342 
1343 		mopt = exthdrs->ip6e_hbh;
1344 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1345 			/*
1346 			 * XXX assumption:
1347 			 * - exthdrs->ip6e_hbh is not referenced from places
1348 			 *   other than exthdrs.
1349 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1350 			 */
1351 			int oldoptlen = mopt->m_len;
1352 			struct mbuf *n;
1353 
1354 			/*
1355 			 * XXX: give up if the whole (new) hbh header does
1356 			 * not fit even in an mbuf cluster.
1357 			 */
1358 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1359 				return (ENOBUFS);
1360 
1361 			/*
1362 			 * As a consequence, we must always prepare a cluster
1363 			 * at this point.
1364 			 */
1365 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1366 			if (n == NULL)
1367 				return (ENOBUFS);
1368 			n->m_len = oldoptlen + JUMBOOPTLEN;
1369 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1370 			    oldoptlen);
1371 			optbuf = mtod(n, caddr_t) + oldoptlen;
1372 			m_freem(mopt);
1373 			mopt = exthdrs->ip6e_hbh = n;
1374 		} else {
1375 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1376 			mopt->m_len += JUMBOOPTLEN;
1377 		}
1378 		optbuf[0] = IP6OPT_PADN;
1379 		optbuf[1] = 1;
1380 
1381 		/*
1382 		 * Adjust the header length according to the pad and
1383 		 * the jumbo payload option.
1384 		 */
1385 		hbh = mtod(mopt, struct ip6_hbh *);
1386 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1387 	}
1388 
1389 	/* fill in the option. */
1390 	optbuf[2] = IP6OPT_JUMBO;
1391 	optbuf[3] = 4;
1392 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1393 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1394 
1395 	/* finally, adjust the packet header length */
1396 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1397 
1398 	return (0);
1399 #undef JUMBOOPTLEN
1400 }
1401 
1402 /*
1403  * Insert fragment header and copy unfragmentable header portions.
1404  */
1405 static int
ip6_insertfraghdr(struct mbuf * m0,struct mbuf * m,int hlen,struct ip6_frag ** frghdrp)1406 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1407     struct ip6_frag **frghdrp)
1408 {
1409 	struct mbuf *n, *mlast;
1410 
1411 	if (hlen > sizeof(struct ip6_hdr)) {
1412 		n = m_copym(m0, sizeof(struct ip6_hdr),
1413 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1414 		if (n == NULL)
1415 			return (ENOBUFS);
1416 		m->m_next = n;
1417 	} else
1418 		n = m;
1419 
1420 	/* Search for the last mbuf of unfragmentable part. */
1421 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1422 		;
1423 
1424 	if (M_WRITABLE(mlast) &&
1425 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1426 		/* use the trailing space of the last mbuf for the fragment hdr */
1427 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1428 		    mlast->m_len);
1429 		mlast->m_len += sizeof(struct ip6_frag);
1430 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1431 	} else {
1432 		/* allocate a new mbuf for the fragment header */
1433 		struct mbuf *mfrg;
1434 
1435 		mfrg = m_get(M_NOWAIT, MT_DATA);
1436 		if (mfrg == NULL)
1437 			return (ENOBUFS);
1438 		mfrg->m_len = sizeof(struct ip6_frag);
1439 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1440 		mlast->m_next = mfrg;
1441 	}
1442 
1443 	return (0);
1444 }
1445 
1446 /*
1447  * Calculates IPv6 path mtu for destination @dst.
1448  * Resulting MTU is stored in @mtup.
1449  *
1450  * Returns 0 on success.
1451  */
1452 static int
ip6_getpmtu_ctl(u_int fibnum,const struct in6_addr * dst,u_long * mtup)1453 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1454 {
1455 	struct epoch_tracker et;
1456 	struct nhop_object *nh;
1457 	struct in6_addr kdst;
1458 	uint32_t scopeid;
1459 	int error;
1460 
1461 	in6_splitscope(dst, &kdst, &scopeid);
1462 
1463 	NET_EPOCH_ENTER(et);
1464 	nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1465 	if (nh != NULL) {
1466 		ip6_calcmtu(nh->nh_ifp, dst, nh->nh_mtu, mtup, 0);
1467 		error = 0;
1468 	} else
1469 		error = EHOSTUNREACH;
1470 	NET_EPOCH_EXIT(et);
1471 
1472 	return (error);
1473 }
1474 
1475 /*
1476  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1477  * and cached data in @ro_pmtu.
1478  * MTU from (successful) route lookup is saved (along with dst)
1479  * inside @ro_pmtu to avoid subsequent route lookups after packet
1480  * filter processing.
1481  *
1482  * Stores mtu into @mtup.
1483  */
1484 static void
ip6_getpmtu(struct route_in6 * ro_pmtu,int do_lookup,struct ifnet * ifp,const struct in6_addr * dst,u_long * mtup,u_int fibnum,u_int proto)1485 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1486     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1487     u_int fibnum, u_int proto)
1488 {
1489 	struct nhop_object *nh;
1490 	struct in6_addr kdst;
1491 	uint32_t scopeid;
1492 	struct sockaddr_in6 *sa6_dst, sin6;
1493 	u_long mtu;
1494 
1495 	NET_EPOCH_ASSERT();
1496 
1497 	mtu = 0;
1498 	if (ro_pmtu == NULL || do_lookup) {
1499 		/*
1500 		 * Here ro_pmtu has final destination address, while
1501 		 * ro might represent immediate destination.
1502 		 * Use ro_pmtu destination since mtu might differ.
1503 		 */
1504 		if (ro_pmtu != NULL) {
1505 			sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1506 			if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1507 				ro_pmtu->ro_mtu = 0;
1508 		} else
1509 			sa6_dst = &sin6;
1510 
1511 		if (ro_pmtu == NULL || ro_pmtu->ro_mtu == 0) {
1512 			bzero(sa6_dst, sizeof(*sa6_dst));
1513 			sa6_dst->sin6_family = AF_INET6;
1514 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1515 			sa6_dst->sin6_addr = *dst;
1516 
1517 			in6_splitscope(dst, &kdst, &scopeid);
1518 			nh = fib6_lookup(fibnum, &kdst, scopeid, NHR_NONE, 0);
1519 			if (nh != NULL) {
1520 				mtu = nh->nh_mtu;
1521 				if (ro_pmtu != NULL)
1522 					ro_pmtu->ro_mtu = mtu;
1523 			}
1524 		} else
1525 			mtu = ro_pmtu->ro_mtu;
1526 	}
1527 
1528 	if (ro_pmtu != NULL && ro_pmtu->ro_nh != NULL)
1529 		mtu = ro_pmtu->ro_nh->nh_mtu;
1530 
1531 	ip6_calcmtu(ifp, dst, mtu, mtup, proto);
1532 }
1533 
1534 /*
1535  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1536  * hostcache data for @dst.
1537  * Stores mtu into @mtup.
1538  */
1539 static void
ip6_calcmtu(struct ifnet * ifp,const struct in6_addr * dst,u_long rt_mtu,u_long * mtup,u_int proto)1540 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1541     u_long *mtup, u_int proto)
1542 {
1543 	u_long mtu = 0;
1544 
1545 	if (rt_mtu > 0) {
1546 		/* Skip the hostcache if the protocol handles PMTU changes. */
1547 		if (proto != IPPROTO_TCP && proto != IPPROTO_SCTP) {
1548 			struct in_conninfo inc = {
1549 			    .inc_flags = INC_ISIPV6,
1550 			    .inc6_faddr = *dst,
1551 			};
1552 
1553 			mtu = tcp_hc_getmtu(&inc);
1554 		}
1555 
1556 		if (mtu)
1557 			mtu = min(mtu, rt_mtu);
1558 		else
1559 			mtu = rt_mtu;
1560 	}
1561 
1562 	if (mtu == 0)
1563 		mtu = IN6_LINKMTU(ifp);
1564 
1565 	*mtup = mtu;
1566 }
1567 
1568 /*
1569  * IP6 socket option processing.
1570  */
1571 int
ip6_ctloutput(struct socket * so,struct sockopt * sopt)1572 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1573 {
1574 	int optdatalen, uproto;
1575 	void *optdata;
1576 	struct inpcb *inp = sotoinpcb(so);
1577 	int error, optval;
1578 	int level, op, optname;
1579 	int optlen;
1580 	struct thread *td;
1581 #ifdef	RSS
1582 	uint32_t rss_bucket;
1583 	int retval;
1584 #endif
1585 
1586 /*
1587  * Don't use more than a quarter of mbuf clusters.  N.B.:
1588  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1589  * on LP64 architectures, so cast to u_long to avoid undefined
1590  * behavior.  ILP32 architectures cannot have nmbclusters
1591  * large enough to overflow for other reasons.
1592  */
1593 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1594 
1595 	level = sopt->sopt_level;
1596 	op = sopt->sopt_dir;
1597 	optname = sopt->sopt_name;
1598 	optlen = sopt->sopt_valsize;
1599 	td = sopt->sopt_td;
1600 	error = 0;
1601 	optval = 0;
1602 	uproto = (int)so->so_proto->pr_protocol;
1603 
1604 	if (level != IPPROTO_IPV6) {
1605 		error = EINVAL;
1606 
1607 		if (sopt->sopt_level == SOL_SOCKET &&
1608 		    sopt->sopt_dir == SOPT_SET) {
1609 			switch (sopt->sopt_name) {
1610 			case SO_SETFIB:
1611 				error = sooptcopyin(sopt, &optval,
1612 				    sizeof(optval), sizeof(optval));
1613 				if (error != 0)
1614 					break;
1615 
1616 				INP_WLOCK(inp);
1617 				if ((inp->inp_flags & INP_BOUNDFIB) != 0 &&
1618 				    optval != so->so_fibnum) {
1619 					INP_WUNLOCK(inp);
1620 					error = EISCONN;
1621 					break;
1622 				}
1623 				error = sosetfib(inp->inp_socket, optval);
1624 				if (error == 0)
1625 					inp->inp_inc.inc_fibnum = optval;
1626 				INP_WUNLOCK(inp);
1627 				break;
1628 			case SO_MAX_PACING_RATE:
1629 #ifdef RATELIMIT
1630 				INP_WLOCK(inp);
1631 				inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1632 				INP_WUNLOCK(inp);
1633 				error = 0;
1634 #else
1635 				error = EOPNOTSUPP;
1636 #endif
1637 				break;
1638 			default:
1639 				break;
1640 			}
1641 		}
1642 	} else {		/* level == IPPROTO_IPV6 */
1643 		switch (op) {
1644 		case SOPT_SET:
1645 			switch (optname) {
1646 			case IPV6_2292PKTOPTIONS:
1647 #ifdef IPV6_PKTOPTIONS
1648 			case IPV6_PKTOPTIONS:
1649 #endif
1650 			{
1651 				struct mbuf *m;
1652 
1653 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1654 					printf("ip6_ctloutput: mbuf limit hit\n");
1655 					error = ENOBUFS;
1656 					break;
1657 				}
1658 
1659 				error = soopt_getm(sopt, &m); /* XXX */
1660 				if (error != 0)
1661 					break;
1662 				error = soopt_mcopyin(sopt, m); /* XXX */
1663 				if (error != 0)
1664 					break;
1665 				INP_WLOCK(inp);
1666 				error = ip6_pcbopts(&inp->in6p_outputopts, m,
1667 				    so, sopt);
1668 				INP_WUNLOCK(inp);
1669 				m_freem(m); /* XXX */
1670 				break;
1671 			}
1672 
1673 			/*
1674 			 * Use of some Hop-by-Hop options or some
1675 			 * Destination options, might require special
1676 			 * privilege.  That is, normal applications
1677 			 * (without special privilege) might be forbidden
1678 			 * from setting certain options in outgoing packets,
1679 			 * and might never see certain options in received
1680 			 * packets. [RFC 2292 Section 6]
1681 			 * KAME specific note:
1682 			 *  KAME prevents non-privileged users from sending or
1683 			 *  receiving ANY hbh/dst options in order to avoid
1684 			 *  overhead of parsing options in the kernel.
1685 			 */
1686 			case IPV6_RECVHOPOPTS:
1687 			case IPV6_RECVDSTOPTS:
1688 			case IPV6_RECVRTHDRDSTOPTS:
1689 				if (td != NULL) {
1690 					error = priv_check(td,
1691 					    PRIV_NETINET_SETHDROPTS);
1692 					if (error)
1693 						break;
1694 				}
1695 				/* FALLTHROUGH */
1696 			case IPV6_UNICAST_HOPS:
1697 			case IPV6_HOPLIMIT:
1698 
1699 			case IPV6_RECVPKTINFO:
1700 			case IPV6_RECVHOPLIMIT:
1701 			case IPV6_RECVRTHDR:
1702 			case IPV6_RECVPATHMTU:
1703 			case IPV6_RECVTCLASS:
1704 			case IPV6_RECVFLOWID:
1705 #ifdef	RSS
1706 			case IPV6_RECVRSSBUCKETID:
1707 #endif
1708 			case IPV6_V6ONLY:
1709 			case IPV6_AUTOFLOWLABEL:
1710 			case IPV6_ORIGDSTADDR:
1711 			case IPV6_BINDANY:
1712 			case IPV6_VLAN_PCP:
1713 				if (optname == IPV6_BINDANY && td != NULL) {
1714 					error = priv_check(td,
1715 					    PRIV_NETINET_BINDANY);
1716 					if (error)
1717 						break;
1718 				}
1719 
1720 				if (optlen != sizeof(int)) {
1721 					error = EINVAL;
1722 					break;
1723 				}
1724 				error = sooptcopyin(sopt, &optval,
1725 					sizeof optval, sizeof optval);
1726 				if (error)
1727 					break;
1728 				switch (optname) {
1729 				case IPV6_UNICAST_HOPS:
1730 					if (optval < -1 || optval >= 256)
1731 						error = EINVAL;
1732 					else {
1733 						/* -1 = kernel default */
1734 						inp->in6p_hops = optval;
1735 						if ((inp->inp_vflag &
1736 						     INP_IPV4) != 0)
1737 							inp->inp_ip_ttl = optval;
1738 					}
1739 					break;
1740 #define OPTSET(bit) \
1741 do { \
1742 	INP_WLOCK(inp); \
1743 	if (optval) \
1744 		inp->inp_flags |= (bit); \
1745 	else \
1746 		inp->inp_flags &= ~(bit); \
1747 	INP_WUNLOCK(inp); \
1748 } while (/*CONSTCOND*/ 0)
1749 #define OPTSET2292(bit) \
1750 do { \
1751 	INP_WLOCK(inp); \
1752 	inp->inp_flags |= IN6P_RFC2292; \
1753 	if (optval) \
1754 		inp->inp_flags |= (bit); \
1755 	else \
1756 		inp->inp_flags &= ~(bit); \
1757 	INP_WUNLOCK(inp); \
1758 } while (/*CONSTCOND*/ 0)
1759 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1760 
1761 #define OPTSET2_N(bit, val) do {					\
1762 	if (val)							\
1763 		inp->inp_flags2 |= bit;					\
1764 	else								\
1765 		inp->inp_flags2 &= ~bit;				\
1766 } while (0)
1767 #define OPTSET2(bit, val) do {						\
1768 	INP_WLOCK(inp);							\
1769 	OPTSET2_N(bit, val);						\
1770 	INP_WUNLOCK(inp);						\
1771 } while (0)
1772 #define OPTBIT2(bit) (inp->inp_flags2 & (bit) ? 1 : 0)
1773 #define OPTSET2292_EXCLUSIVE(bit)					\
1774 do {									\
1775 	INP_WLOCK(inp);							\
1776 	if (OPTBIT(IN6P_RFC2292)) {					\
1777 		error = EINVAL;						\
1778 	} else {							\
1779 		if (optval)						\
1780 			inp->inp_flags |= (bit);			\
1781 		else							\
1782 			inp->inp_flags &= ~(bit);			\
1783 	}								\
1784 	INP_WUNLOCK(inp);						\
1785 } while (/*CONSTCOND*/ 0)
1786 
1787 				case IPV6_RECVPKTINFO:
1788 					OPTSET2292_EXCLUSIVE(IN6P_PKTINFO);
1789 					break;
1790 
1791 				case IPV6_HOPLIMIT:
1792 				{
1793 					struct ip6_pktopts **optp;
1794 
1795 					/* cannot mix with RFC2292 */
1796 					if (OPTBIT(IN6P_RFC2292)) {
1797 						error = EINVAL;
1798 						break;
1799 					}
1800 					INP_WLOCK(inp);
1801 					if (inp->inp_flags & INP_DROPPED) {
1802 						INP_WUNLOCK(inp);
1803 						return (ECONNRESET);
1804 					}
1805 					optp = &inp->in6p_outputopts;
1806 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1807 					    (u_char *)&optval, sizeof(optval),
1808 					    optp, (td != NULL) ? td->td_ucred :
1809 					    NULL, uproto);
1810 					INP_WUNLOCK(inp);
1811 					break;
1812 				}
1813 
1814 				case IPV6_RECVHOPLIMIT:
1815 					OPTSET2292_EXCLUSIVE(IN6P_HOPLIMIT);
1816 					break;
1817 
1818 				case IPV6_RECVHOPOPTS:
1819 					OPTSET2292_EXCLUSIVE(IN6P_HOPOPTS);
1820 					break;
1821 
1822 				case IPV6_RECVDSTOPTS:
1823 					OPTSET2292_EXCLUSIVE(IN6P_DSTOPTS);
1824 					break;
1825 
1826 				case IPV6_RECVRTHDRDSTOPTS:
1827 					OPTSET2292_EXCLUSIVE(IN6P_RTHDRDSTOPTS);
1828 					break;
1829 
1830 				case IPV6_RECVRTHDR:
1831 					OPTSET2292_EXCLUSIVE(IN6P_RTHDR);
1832 					break;
1833 
1834 				case IPV6_RECVPATHMTU:
1835 					/*
1836 					 * We ignore this option for TCP
1837 					 * sockets.
1838 					 * (RFC3542 leaves this case
1839 					 * unspecified.)
1840 					 */
1841 					if (uproto != IPPROTO_TCP)
1842 						OPTSET(IN6P_MTU);
1843 					break;
1844 
1845 				case IPV6_RECVFLOWID:
1846 					OPTSET2(INP_RECVFLOWID, optval);
1847 					break;
1848 
1849 #ifdef	RSS
1850 				case IPV6_RECVRSSBUCKETID:
1851 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1852 					break;
1853 #endif
1854 
1855 				case IPV6_V6ONLY:
1856 					INP_WLOCK(inp);
1857 					if (inp->inp_lport ||
1858 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1859 						/*
1860 						 * The socket is already bound.
1861 						 */
1862 						INP_WUNLOCK(inp);
1863 						error = EINVAL;
1864 						break;
1865 					}
1866 					if (optval) {
1867 						inp->inp_flags |= IN6P_IPV6_V6ONLY;
1868 						inp->inp_vflag &= ~INP_IPV4;
1869 					} else {
1870 						inp->inp_flags &= ~IN6P_IPV6_V6ONLY;
1871 						inp->inp_vflag |= INP_IPV4;
1872 					}
1873 					INP_WUNLOCK(inp);
1874 					break;
1875 				case IPV6_RECVTCLASS:
1876 					/* cannot mix with RFC2292 XXX */
1877 					OPTSET2292_EXCLUSIVE(IN6P_TCLASS);
1878 					break;
1879 				case IPV6_AUTOFLOWLABEL:
1880 					OPTSET(IN6P_AUTOFLOWLABEL);
1881 					break;
1882 
1883 				case IPV6_ORIGDSTADDR:
1884 					OPTSET2(INP_ORIGDSTADDR, optval);
1885 					break;
1886 				case IPV6_BINDANY:
1887 					OPTSET(INP_BINDANY);
1888 					break;
1889 				case IPV6_VLAN_PCP:
1890 					if ((optval >= -1) && (optval <=
1891 					    (INP_2PCP_MASK >> INP_2PCP_SHIFT))) {
1892 						if (optval == -1) {
1893 							INP_WLOCK(inp);
1894 							inp->inp_flags2 &=
1895 							    ~(INP_2PCP_SET |
1896 							    INP_2PCP_MASK);
1897 							INP_WUNLOCK(inp);
1898 						} else {
1899 							INP_WLOCK(inp);
1900 							inp->inp_flags2 |=
1901 							    INP_2PCP_SET;
1902 							inp->inp_flags2 &=
1903 							    ~INP_2PCP_MASK;
1904 							inp->inp_flags2 |=
1905 							    optval <<
1906 							    INP_2PCP_SHIFT;
1907 							INP_WUNLOCK(inp);
1908 						}
1909 					} else
1910 						error = EINVAL;
1911 					break;
1912 				}
1913 				break;
1914 
1915 			case IPV6_TCLASS:
1916 			case IPV6_DONTFRAG:
1917 			case IPV6_USE_MIN_MTU:
1918 			case IPV6_PREFER_TEMPADDR:
1919 				if (optlen != sizeof(optval)) {
1920 					error = EINVAL;
1921 					break;
1922 				}
1923 				error = sooptcopyin(sopt, &optval,
1924 					sizeof optval, sizeof optval);
1925 				if (error)
1926 					break;
1927 				{
1928 					struct ip6_pktopts **optp;
1929 					INP_WLOCK(inp);
1930 					if (inp->inp_flags & INP_DROPPED) {
1931 						INP_WUNLOCK(inp);
1932 						return (ECONNRESET);
1933 					}
1934 					optp = &inp->in6p_outputopts;
1935 					error = ip6_pcbopt(optname,
1936 					    (u_char *)&optval, sizeof(optval),
1937 					    optp, (td != NULL) ? td->td_ucred :
1938 					    NULL, uproto);
1939 					INP_WUNLOCK(inp);
1940 					break;
1941 				}
1942 
1943 			case IPV6_2292PKTINFO:
1944 			case IPV6_2292HOPLIMIT:
1945 			case IPV6_2292HOPOPTS:
1946 			case IPV6_2292DSTOPTS:
1947 			case IPV6_2292RTHDR:
1948 				/* RFC 2292 */
1949 				if (optlen != sizeof(int)) {
1950 					error = EINVAL;
1951 					break;
1952 				}
1953 				error = sooptcopyin(sopt, &optval,
1954 					sizeof optval, sizeof optval);
1955 				if (error)
1956 					break;
1957 				switch (optname) {
1958 				case IPV6_2292PKTINFO:
1959 					OPTSET2292(IN6P_PKTINFO);
1960 					break;
1961 				case IPV6_2292HOPLIMIT:
1962 					OPTSET2292(IN6P_HOPLIMIT);
1963 					break;
1964 				case IPV6_2292HOPOPTS:
1965 					/*
1966 					 * Check super-user privilege.
1967 					 * See comments for IPV6_RECVHOPOPTS.
1968 					 */
1969 					if (td != NULL) {
1970 						error = priv_check(td,
1971 						    PRIV_NETINET_SETHDROPTS);
1972 						if (error)
1973 							return (error);
1974 					}
1975 					OPTSET2292(IN6P_HOPOPTS);
1976 					break;
1977 				case IPV6_2292DSTOPTS:
1978 					if (td != NULL) {
1979 						error = priv_check(td,
1980 						    PRIV_NETINET_SETHDROPTS);
1981 						if (error)
1982 							return (error);
1983 					}
1984 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1985 					break;
1986 				case IPV6_2292RTHDR:
1987 					OPTSET2292(IN6P_RTHDR);
1988 					break;
1989 				}
1990 				break;
1991 			case IPV6_PKTINFO:
1992 			case IPV6_HOPOPTS:
1993 			case IPV6_RTHDR:
1994 			case IPV6_DSTOPTS:
1995 			case IPV6_RTHDRDSTOPTS:
1996 			case IPV6_NEXTHOP:
1997 			{
1998 				/* new advanced API (RFC3542) */
1999 				u_char *optbuf;
2000 				u_char optbuf_storage[MCLBYTES];
2001 				int optlen;
2002 				struct ip6_pktopts **optp;
2003 
2004 				/* cannot mix with RFC2292 */
2005 				if (OPTBIT(IN6P_RFC2292)) {
2006 					error = EINVAL;
2007 					break;
2008 				}
2009 
2010 				/*
2011 				 * We only ensure valsize is not too large
2012 				 * here.  Further validation will be done
2013 				 * later.
2014 				 */
2015 				error = sooptcopyin(sopt, optbuf_storage,
2016 				    sizeof(optbuf_storage), 0);
2017 				if (error)
2018 					break;
2019 				optlen = sopt->sopt_valsize;
2020 				optbuf = optbuf_storage;
2021 				INP_WLOCK(inp);
2022 				if (inp->inp_flags & INP_DROPPED) {
2023 					INP_WUNLOCK(inp);
2024 					return (ECONNRESET);
2025 				}
2026 				optp = &inp->in6p_outputopts;
2027 				error = ip6_pcbopt(optname, optbuf, optlen,
2028 				    optp, (td != NULL) ? td->td_ucred : NULL,
2029 				    uproto);
2030 				INP_WUNLOCK(inp);
2031 				break;
2032 			}
2033 #undef OPTSET
2034 
2035 			case IPV6_MULTICAST_IF:
2036 			case IPV6_MULTICAST_HOPS:
2037 			case IPV6_MULTICAST_LOOP:
2038 			case IPV6_JOIN_GROUP:
2039 			case IPV6_LEAVE_GROUP:
2040 			case IPV6_MSFILTER:
2041 			case MCAST_BLOCK_SOURCE:
2042 			case MCAST_UNBLOCK_SOURCE:
2043 			case MCAST_JOIN_GROUP:
2044 			case MCAST_LEAVE_GROUP:
2045 			case MCAST_JOIN_SOURCE_GROUP:
2046 			case MCAST_LEAVE_SOURCE_GROUP:
2047 				error = ip6_setmoptions(inp, sopt);
2048 				break;
2049 
2050 			case IPV6_PORTRANGE:
2051 				error = sooptcopyin(sopt, &optval,
2052 				    sizeof optval, sizeof optval);
2053 				if (error)
2054 					break;
2055 
2056 				INP_WLOCK(inp);
2057 				switch (optval) {
2058 				case IPV6_PORTRANGE_DEFAULT:
2059 					inp->inp_flags &= ~(INP_LOWPORT);
2060 					inp->inp_flags &= ~(INP_HIGHPORT);
2061 					break;
2062 
2063 				case IPV6_PORTRANGE_HIGH:
2064 					inp->inp_flags &= ~(INP_LOWPORT);
2065 					inp->inp_flags |= INP_HIGHPORT;
2066 					break;
2067 
2068 				case IPV6_PORTRANGE_LOW:
2069 					inp->inp_flags &= ~(INP_HIGHPORT);
2070 					inp->inp_flags |= INP_LOWPORT;
2071 					break;
2072 
2073 				default:
2074 					error = EINVAL;
2075 					break;
2076 				}
2077 				INP_WUNLOCK(inp);
2078 				break;
2079 
2080 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2081 			case IPV6_IPSEC_POLICY:
2082 				if (IPSEC_ENABLED(ipv6)) {
2083 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2084 					break;
2085 				}
2086 				/* FALLTHROUGH */
2087 #endif /* IPSEC */
2088 
2089 			default:
2090 				error = ENOPROTOOPT;
2091 				break;
2092 			}
2093 			break;
2094 
2095 		case SOPT_GET:
2096 			switch (optname) {
2097 			case IPV6_2292PKTOPTIONS:
2098 #ifdef IPV6_PKTOPTIONS
2099 			case IPV6_PKTOPTIONS:
2100 #endif
2101 				/*
2102 				 * RFC3542 (effectively) deprecated the
2103 				 * semantics of the 2292-style pktoptions.
2104 				 * Since it was not reliable in nature (i.e.,
2105 				 * applications had to expect the lack of some
2106 				 * information after all), it would make sense
2107 				 * to simplify this part by always returning
2108 				 * empty data.
2109 				 */
2110 				sopt->sopt_valsize = 0;
2111 				break;
2112 
2113 			case IPV6_RECVHOPOPTS:
2114 			case IPV6_RECVDSTOPTS:
2115 			case IPV6_RECVRTHDRDSTOPTS:
2116 			case IPV6_UNICAST_HOPS:
2117 			case IPV6_RECVPKTINFO:
2118 			case IPV6_RECVHOPLIMIT:
2119 			case IPV6_RECVRTHDR:
2120 			case IPV6_RECVPATHMTU:
2121 
2122 			case IPV6_V6ONLY:
2123 			case IPV6_PORTRANGE:
2124 			case IPV6_RECVTCLASS:
2125 			case IPV6_AUTOFLOWLABEL:
2126 			case IPV6_BINDANY:
2127 			case IPV6_FLOWID:
2128 			case IPV6_FLOWTYPE:
2129 			case IPV6_RECVFLOWID:
2130 #ifdef	RSS
2131 			case IPV6_RSSBUCKETID:
2132 			case IPV6_RECVRSSBUCKETID:
2133 #endif
2134 			case IPV6_VLAN_PCP:
2135 				switch (optname) {
2136 				case IPV6_RECVHOPOPTS:
2137 					optval = OPTBIT(IN6P_HOPOPTS);
2138 					break;
2139 
2140 				case IPV6_RECVDSTOPTS:
2141 					optval = OPTBIT(IN6P_DSTOPTS);
2142 					break;
2143 
2144 				case IPV6_RECVRTHDRDSTOPTS:
2145 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2146 					break;
2147 
2148 				case IPV6_UNICAST_HOPS:
2149 					optval = inp->in6p_hops;
2150 					break;
2151 
2152 				case IPV6_RECVPKTINFO:
2153 					optval = OPTBIT(IN6P_PKTINFO);
2154 					break;
2155 
2156 				case IPV6_RECVHOPLIMIT:
2157 					optval = OPTBIT(IN6P_HOPLIMIT);
2158 					break;
2159 
2160 				case IPV6_RECVRTHDR:
2161 					optval = OPTBIT(IN6P_RTHDR);
2162 					break;
2163 
2164 				case IPV6_RECVPATHMTU:
2165 					optval = OPTBIT(IN6P_MTU);
2166 					break;
2167 
2168 				case IPV6_V6ONLY:
2169 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2170 					break;
2171 
2172 				case IPV6_PORTRANGE:
2173 				    {
2174 					int flags;
2175 					flags = inp->inp_flags;
2176 					if (flags & INP_HIGHPORT)
2177 						optval = IPV6_PORTRANGE_HIGH;
2178 					else if (flags & INP_LOWPORT)
2179 						optval = IPV6_PORTRANGE_LOW;
2180 					else
2181 						optval = 0;
2182 					break;
2183 				    }
2184 				case IPV6_RECVTCLASS:
2185 					optval = OPTBIT(IN6P_TCLASS);
2186 					break;
2187 
2188 				case IPV6_AUTOFLOWLABEL:
2189 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2190 					break;
2191 
2192 				case IPV6_ORIGDSTADDR:
2193 					optval = OPTBIT2(INP_ORIGDSTADDR);
2194 					break;
2195 
2196 				case IPV6_BINDANY:
2197 					optval = OPTBIT(INP_BINDANY);
2198 					break;
2199 
2200 				case IPV6_FLOWID:
2201 					optval = inp->inp_flowid;
2202 					break;
2203 
2204 				case IPV6_FLOWTYPE:
2205 					optval = inp->inp_flowtype;
2206 					break;
2207 
2208 				case IPV6_RECVFLOWID:
2209 					optval = OPTBIT2(INP_RECVFLOWID);
2210 					break;
2211 #ifdef	RSS
2212 				case IPV6_RSSBUCKETID:
2213 					retval =
2214 					    rss_hash2bucket(inp->inp_flowid,
2215 					    inp->inp_flowtype,
2216 					    &rss_bucket);
2217 					if (retval == 0)
2218 						optval = rss_bucket;
2219 					else
2220 						error = EINVAL;
2221 					break;
2222 
2223 				case IPV6_RECVRSSBUCKETID:
2224 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2225 					break;
2226 #endif
2227 
2228 
2229 				case IPV6_VLAN_PCP:
2230 					if (OPTBIT2(INP_2PCP_SET)) {
2231 						optval = (inp->inp_flags2 &
2232 							    INP_2PCP_MASK) >>
2233 							    INP_2PCP_SHIFT;
2234 					} else {
2235 						optval = -1;
2236 					}
2237 					break;
2238 				}
2239 
2240 				if (error)
2241 					break;
2242 				error = sooptcopyout(sopt, &optval,
2243 					sizeof optval);
2244 				break;
2245 
2246 			case IPV6_PATHMTU:
2247 			{
2248 				u_long pmtu = 0;
2249 				struct ip6_mtuinfo mtuinfo;
2250 				struct in6_addr addr;
2251 
2252 				if (!(so->so_state & SS_ISCONNECTED))
2253 					return (ENOTCONN);
2254 				/*
2255 				 * XXX: we dot not consider the case of source
2256 				 * routing, or optional information to specify
2257 				 * the outgoing interface.
2258 				 * Copy faddr out of inp to avoid holding lock
2259 				 * on inp during route lookup.
2260 				 */
2261 				INP_RLOCK(inp);
2262 				bcopy(&inp->in6p_faddr, &addr, sizeof(addr));
2263 				INP_RUNLOCK(inp);
2264 				error = ip6_getpmtu_ctl(so->so_fibnum,
2265 				    &addr, &pmtu);
2266 				if (error)
2267 					break;
2268 				if (pmtu > IPV6_MAXPACKET)
2269 					pmtu = IPV6_MAXPACKET;
2270 
2271 				bzero(&mtuinfo, sizeof(mtuinfo));
2272 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2273 				optdata = (void *)&mtuinfo;
2274 				optdatalen = sizeof(mtuinfo);
2275 				error = sooptcopyout(sopt, optdata,
2276 				    optdatalen);
2277 				break;
2278 			}
2279 
2280 			case IPV6_2292PKTINFO:
2281 			case IPV6_2292HOPLIMIT:
2282 			case IPV6_2292HOPOPTS:
2283 			case IPV6_2292RTHDR:
2284 			case IPV6_2292DSTOPTS:
2285 				switch (optname) {
2286 				case IPV6_2292PKTINFO:
2287 					optval = OPTBIT(IN6P_PKTINFO);
2288 					break;
2289 				case IPV6_2292HOPLIMIT:
2290 					optval = OPTBIT(IN6P_HOPLIMIT);
2291 					break;
2292 				case IPV6_2292HOPOPTS:
2293 					optval = OPTBIT(IN6P_HOPOPTS);
2294 					break;
2295 				case IPV6_2292RTHDR:
2296 					optval = OPTBIT(IN6P_RTHDR);
2297 					break;
2298 				case IPV6_2292DSTOPTS:
2299 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2300 					break;
2301 				}
2302 				error = sooptcopyout(sopt, &optval,
2303 				    sizeof optval);
2304 				break;
2305 			case IPV6_PKTINFO:
2306 			case IPV6_HOPOPTS:
2307 			case IPV6_RTHDR:
2308 			case IPV6_DSTOPTS:
2309 			case IPV6_RTHDRDSTOPTS:
2310 			case IPV6_NEXTHOP:
2311 			case IPV6_TCLASS:
2312 			case IPV6_DONTFRAG:
2313 			case IPV6_USE_MIN_MTU:
2314 			case IPV6_PREFER_TEMPADDR:
2315 				error = ip6_getpcbopt(inp, optname, sopt);
2316 				break;
2317 
2318 			case IPV6_MULTICAST_IF:
2319 			case IPV6_MULTICAST_HOPS:
2320 			case IPV6_MULTICAST_LOOP:
2321 			case IPV6_MSFILTER:
2322 				error = ip6_getmoptions(inp, sopt);
2323 				break;
2324 
2325 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2326 			case IPV6_IPSEC_POLICY:
2327 				if (IPSEC_ENABLED(ipv6)) {
2328 					error = IPSEC_PCBCTL(ipv6, inp, sopt);
2329 					break;
2330 				}
2331 				/* FALLTHROUGH */
2332 #endif /* IPSEC */
2333 			default:
2334 				error = ENOPROTOOPT;
2335 				break;
2336 			}
2337 			break;
2338 		}
2339 	}
2340 	return (error);
2341 }
2342 
2343 int
ip6_raw_ctloutput(struct socket * so,struct sockopt * sopt)2344 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2345 {
2346 	int error = 0, optval, optlen;
2347 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2348 	struct inpcb *inp = sotoinpcb(so);
2349 	int level, op, optname;
2350 
2351 	level = sopt->sopt_level;
2352 	op = sopt->sopt_dir;
2353 	optname = sopt->sopt_name;
2354 	optlen = sopt->sopt_valsize;
2355 
2356 	if (level != IPPROTO_IPV6) {
2357 		return (EINVAL);
2358 	}
2359 
2360 	switch (optname) {
2361 	case IPV6_CHECKSUM:
2362 		/*
2363 		 * For ICMPv6 sockets, no modification allowed for checksum
2364 		 * offset, permit "no change" values to help existing apps.
2365 		 *
2366 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2367 		 * for an ICMPv6 socket will fail."
2368 		 * The current behavior does not meet RFC3542.
2369 		 */
2370 		switch (op) {
2371 		case SOPT_SET:
2372 			if (optlen != sizeof(int)) {
2373 				error = EINVAL;
2374 				break;
2375 			}
2376 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2377 					    sizeof(optval));
2378 			if (error)
2379 				break;
2380 			if (optval < -1 || (optval % 2) != 0) {
2381 				/*
2382 				 * The API assumes non-negative even offset
2383 				 * values or -1 as a special value.
2384 				 */
2385 				error = EINVAL;
2386 			} else if (inp->inp_ip_p == IPPROTO_ICMPV6) {
2387 				if (optval != icmp6off)
2388 					error = EINVAL;
2389 			} else
2390 				inp->in6p_cksum = optval;
2391 			break;
2392 
2393 		case SOPT_GET:
2394 			if (inp->inp_ip_p == IPPROTO_ICMPV6)
2395 				optval = icmp6off;
2396 			else
2397 				optval = inp->in6p_cksum;
2398 
2399 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2400 			break;
2401 
2402 		default:
2403 			error = EINVAL;
2404 			break;
2405 		}
2406 		break;
2407 
2408 	default:
2409 		error = ENOPROTOOPT;
2410 		break;
2411 	}
2412 
2413 	return (error);
2414 }
2415 
2416 /*
2417  * Set up IP6 options in pcb for insertion in output packets or
2418  * specifying behavior of outgoing packets.
2419  */
2420 static int
ip6_pcbopts(struct ip6_pktopts ** pktopt,struct mbuf * m,struct socket * so,struct sockopt * sopt)2421 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2422     struct socket *so, struct sockopt *sopt)
2423 {
2424 	struct ip6_pktopts *opt = *pktopt;
2425 	int error = 0;
2426 	struct thread *td = sopt->sopt_td;
2427 	struct epoch_tracker et;
2428 
2429 	/* turn off any old options. */
2430 	if (opt) {
2431 #ifdef DIAGNOSTIC
2432 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2433 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2434 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2435 			printf("ip6_pcbopts: all specified options are cleared.\n");
2436 #endif
2437 		ip6_clearpktopts(opt, -1);
2438 	} else {
2439 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2440 		if (opt == NULL)
2441 			return (ENOMEM);
2442 	}
2443 	*pktopt = NULL;
2444 
2445 	if (!m || m->m_len == 0) {
2446 		/*
2447 		 * Only turning off any previous options, regardless of
2448 		 * whether the opt is just created or given.
2449 		 */
2450 		free(opt, M_IP6OPT);
2451 		return (0);
2452 	}
2453 
2454 	/*  set options specified by user. */
2455 	NET_EPOCH_ENTER(et);
2456 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2457 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2458 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2459 		free(opt, M_IP6OPT);
2460 		NET_EPOCH_EXIT(et);
2461 		return (error);
2462 	}
2463 	NET_EPOCH_EXIT(et);
2464 	*pktopt = opt;
2465 	return (0);
2466 }
2467 
2468 /*
2469  * initialize ip6_pktopts.  beware that there are non-zero default values in
2470  * the struct.
2471  */
2472 void
ip6_initpktopts(struct ip6_pktopts * opt)2473 ip6_initpktopts(struct ip6_pktopts *opt)
2474 {
2475 
2476 	bzero(opt, sizeof(*opt));
2477 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2478 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2479 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2480 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2481 }
2482 
2483 static int
ip6_pcbopt(int optname,u_char * buf,int len,struct ip6_pktopts ** pktopt,struct ucred * cred,int uproto)2484 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2485     struct ucred *cred, int uproto)
2486 {
2487 	struct epoch_tracker et;
2488 	struct ip6_pktopts *opt;
2489 	int ret;
2490 
2491 	if (*pktopt == NULL) {
2492 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2493 		    M_NOWAIT);
2494 		if (*pktopt == NULL)
2495 			return (ENOBUFS);
2496 		ip6_initpktopts(*pktopt);
2497 	}
2498 	opt = *pktopt;
2499 
2500 	NET_EPOCH_ENTER(et);
2501 	ret = ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto);
2502 	NET_EPOCH_EXIT(et);
2503 
2504 	return (ret);
2505 }
2506 
2507 #define GET_PKTOPT_VAR(field, lenexpr) do {				\
2508 	if (pktopt && pktopt->field) {					\
2509 		INP_RUNLOCK(inp);					\
2510 		optdata = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK);	\
2511 		malloc_optdata = true;					\
2512 		INP_RLOCK(inp);						\
2513 		if (inp->inp_flags & INP_DROPPED) {			\
2514 			INP_RUNLOCK(inp);				\
2515 			free(optdata, M_TEMP);				\
2516 			return (ECONNRESET);				\
2517 		}							\
2518 		pktopt = inp->in6p_outputopts;				\
2519 		if (pktopt && pktopt->field) {				\
2520 			optdatalen = min(lenexpr, sopt->sopt_valsize);	\
2521 			bcopy(pktopt->field, optdata, optdatalen);	\
2522 		} else {						\
2523 			free(optdata, M_TEMP);				\
2524 			optdata = NULL;					\
2525 			malloc_optdata = false;				\
2526 		}							\
2527 	}								\
2528 } while(0)
2529 
2530 #define GET_PKTOPT_EXT_HDR(field) GET_PKTOPT_VAR(field,			\
2531 	(((struct ip6_ext *)pktopt->field)->ip6e_len + 1) << 3)
2532 
2533 #define GET_PKTOPT_SOCKADDR(field) GET_PKTOPT_VAR(field,		\
2534 	pktopt->field->sa_len)
2535 
2536 static int
ip6_getpcbopt(struct inpcb * inp,int optname,struct sockopt * sopt)2537 ip6_getpcbopt(struct inpcb *inp, int optname, struct sockopt *sopt)
2538 {
2539 	void *optdata = NULL;
2540 	bool malloc_optdata = false;
2541 	int optdatalen = 0;
2542 	int error = 0;
2543 	struct in6_pktinfo null_pktinfo;
2544 	int deftclass = 0, on;
2545 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2546 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2547 	struct ip6_pktopts *pktopt;
2548 
2549 	INP_RLOCK(inp);
2550 	pktopt = inp->in6p_outputopts;
2551 
2552 	switch (optname) {
2553 	case IPV6_PKTINFO:
2554 		optdata = (void *)&null_pktinfo;
2555 		if (pktopt && pktopt->ip6po_pktinfo) {
2556 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2557 			    sizeof(null_pktinfo));
2558 			in6_clearscope(&null_pktinfo.ipi6_addr);
2559 		} else {
2560 			/* XXX: we don't have to do this every time... */
2561 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2562 		}
2563 		optdatalen = sizeof(struct in6_pktinfo);
2564 		break;
2565 	case IPV6_TCLASS:
2566 		if (pktopt && pktopt->ip6po_tclass >= 0)
2567 			deftclass = pktopt->ip6po_tclass;
2568 		optdata = (void *)&deftclass;
2569 		optdatalen = sizeof(int);
2570 		break;
2571 	case IPV6_HOPOPTS:
2572 		GET_PKTOPT_EXT_HDR(ip6po_hbh);
2573 		break;
2574 	case IPV6_RTHDR:
2575 		GET_PKTOPT_EXT_HDR(ip6po_rthdr);
2576 		break;
2577 	case IPV6_RTHDRDSTOPTS:
2578 		GET_PKTOPT_EXT_HDR(ip6po_dest1);
2579 		break;
2580 	case IPV6_DSTOPTS:
2581 		GET_PKTOPT_EXT_HDR(ip6po_dest2);
2582 		break;
2583 	case IPV6_NEXTHOP:
2584 		GET_PKTOPT_SOCKADDR(ip6po_nexthop);
2585 		break;
2586 	case IPV6_USE_MIN_MTU:
2587 		if (pktopt)
2588 			defminmtu = pktopt->ip6po_minmtu;
2589 		optdata = (void *)&defminmtu;
2590 		optdatalen = sizeof(int);
2591 		break;
2592 	case IPV6_DONTFRAG:
2593 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2594 			on = 1;
2595 		else
2596 			on = 0;
2597 		optdata = (void *)&on;
2598 		optdatalen = sizeof(on);
2599 		break;
2600 	case IPV6_PREFER_TEMPADDR:
2601 		if (pktopt)
2602 			defpreftemp = pktopt->ip6po_prefer_tempaddr;
2603 		optdata = (void *)&defpreftemp;
2604 		optdatalen = sizeof(int);
2605 		break;
2606 	default:		/* should not happen */
2607 #ifdef DIAGNOSTIC
2608 		panic("ip6_getpcbopt: unexpected option\n");
2609 #endif
2610 		INP_RUNLOCK(inp);
2611 		return (ENOPROTOOPT);
2612 	}
2613 	INP_RUNLOCK(inp);
2614 
2615 	error = sooptcopyout(sopt, optdata, optdatalen);
2616 	if (malloc_optdata)
2617 		free(optdata, M_TEMP);
2618 
2619 	return (error);
2620 }
2621 
2622 void
ip6_clearpktopts(struct ip6_pktopts * pktopt,int optname)2623 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2624 {
2625 	if (pktopt == NULL)
2626 		return;
2627 
2628 	if (optname == -1 || optname == IPV6_PKTINFO) {
2629 		if (pktopt->ip6po_pktinfo)
2630 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2631 		pktopt->ip6po_pktinfo = NULL;
2632 	}
2633 	if (optname == -1 || optname == IPV6_HOPLIMIT) {
2634 		pktopt->ip6po_hlim = -1;
2635 		pktopt->ip6po_valid &= ~IP6PO_VALID_HLIM;
2636 	}
2637 	if (optname == -1 || optname == IPV6_TCLASS) {
2638 		pktopt->ip6po_tclass = -1;
2639 		pktopt->ip6po_valid &= ~IP6PO_VALID_TC;
2640 	}
2641 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2642 		if (pktopt->ip6po_nextroute.ro_nh) {
2643 			NH_FREE(pktopt->ip6po_nextroute.ro_nh);
2644 			pktopt->ip6po_nextroute.ro_nh = NULL;
2645 		}
2646 		if (pktopt->ip6po_nexthop)
2647 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2648 		pktopt->ip6po_nexthop = NULL;
2649 		pktopt->ip6po_valid &= ~IP6PO_VALID_NHINFO;
2650 	}
2651 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2652 		if (pktopt->ip6po_hbh)
2653 			free(pktopt->ip6po_hbh, M_IP6OPT);
2654 		pktopt->ip6po_hbh = NULL;
2655 		pktopt->ip6po_valid &= ~IP6PO_VALID_HBH;
2656 	}
2657 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2658 		if (pktopt->ip6po_dest1)
2659 			free(pktopt->ip6po_dest1, M_IP6OPT);
2660 		pktopt->ip6po_dest1 = NULL;
2661 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST1;
2662 	}
2663 	if (optname == -1 || optname == IPV6_RTHDR) {
2664 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2665 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2666 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2667 		if (pktopt->ip6po_route.ro_nh) {
2668 			NH_FREE(pktopt->ip6po_route.ro_nh);
2669 			pktopt->ip6po_route.ro_nh = NULL;
2670 		}
2671 		pktopt->ip6po_valid &= ~IP6PO_VALID_RHINFO;
2672 	}
2673 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2674 		if (pktopt->ip6po_dest2)
2675 			free(pktopt->ip6po_dest2, M_IP6OPT);
2676 		pktopt->ip6po_dest2 = NULL;
2677 		pktopt->ip6po_valid &= ~IP6PO_VALID_DEST2;
2678 	}
2679 }
2680 
2681 #define PKTOPT_EXTHDRCPY(type) \
2682 do {\
2683 	if (src->type) {\
2684 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2685 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2686 		if (dst->type == NULL)\
2687 			goto bad;\
2688 		bcopy(src->type, dst->type, hlen);\
2689 	}\
2690 } while (/*CONSTCOND*/ 0)
2691 
2692 static int
copypktopts(struct ip6_pktopts * dst,struct ip6_pktopts * src,int canwait)2693 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2694 {
2695 	if (dst == NULL || src == NULL)  {
2696 		printf("ip6_clearpktopts: invalid argument\n");
2697 		return (EINVAL);
2698 	}
2699 
2700 	dst->ip6po_hlim = src->ip6po_hlim;
2701 	dst->ip6po_tclass = src->ip6po_tclass;
2702 	dst->ip6po_flags = src->ip6po_flags;
2703 	dst->ip6po_minmtu = src->ip6po_minmtu;
2704 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2705 	if (src->ip6po_pktinfo) {
2706 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2707 		    M_IP6OPT, canwait);
2708 		if (dst->ip6po_pktinfo == NULL)
2709 			goto bad;
2710 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2711 	}
2712 	if (src->ip6po_nexthop) {
2713 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2714 		    M_IP6OPT, canwait);
2715 		if (dst->ip6po_nexthop == NULL)
2716 			goto bad;
2717 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2718 		    src->ip6po_nexthop->sa_len);
2719 	}
2720 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2721 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2722 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2723 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2724 	dst->ip6po_valid = src->ip6po_valid;
2725 	return (0);
2726 
2727   bad:
2728 	ip6_clearpktopts(dst, -1);
2729 	return (ENOBUFS);
2730 }
2731 #undef PKTOPT_EXTHDRCPY
2732 
2733 struct ip6_pktopts *
ip6_copypktopts(struct ip6_pktopts * src,int canwait)2734 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2735 {
2736 	int error;
2737 	struct ip6_pktopts *dst;
2738 
2739 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2740 	if (dst == NULL)
2741 		return (NULL);
2742 	ip6_initpktopts(dst);
2743 
2744 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2745 		free(dst, M_IP6OPT);
2746 		return (NULL);
2747 	}
2748 
2749 	return (dst);
2750 }
2751 
2752 void
ip6_freepcbopts(struct ip6_pktopts * pktopt)2753 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2754 {
2755 	if (pktopt == NULL)
2756 		return;
2757 
2758 	ip6_clearpktopts(pktopt, -1);
2759 
2760 	free(pktopt, M_IP6OPT);
2761 }
2762 
2763 /*
2764  * Set IPv6 outgoing packet options based on advanced API.
2765  */
2766 int
ip6_setpktopts(struct mbuf * control,struct ip6_pktopts * opt,struct ip6_pktopts * stickyopt,struct ucred * cred,int uproto)2767 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2768     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2769 {
2770 	struct cmsghdr *cm = NULL;
2771 
2772 	if (control == NULL || opt == NULL)
2773 		return (EINVAL);
2774 
2775 	/*
2776 	 * ip6_setpktopt can call ifnet_byindex(), so it's imperative that we
2777 	 * are in the network epoch here.
2778 	 */
2779 	NET_EPOCH_ASSERT();
2780 
2781 	ip6_initpktopts(opt);
2782 	if (stickyopt) {
2783 		int error;
2784 
2785 		/*
2786 		 * If stickyopt is provided, make a local copy of the options
2787 		 * for this particular packet, then override them by ancillary
2788 		 * objects.
2789 		 * XXX: copypktopts() does not copy the cached route to a next
2790 		 * hop (if any).  This is not very good in terms of efficiency,
2791 		 * but we can allow this since this option should be rarely
2792 		 * used.
2793 		 */
2794 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2795 			return (error);
2796 	}
2797 
2798 	/*
2799 	 * XXX: Currently, we assume all the optional information is stored
2800 	 * in a single mbuf.
2801 	 */
2802 	if (control->m_next)
2803 		return (EINVAL);
2804 
2805 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2806 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2807 		int error;
2808 
2809 		if (control->m_len < CMSG_LEN(0))
2810 			return (EINVAL);
2811 
2812 		cm = mtod(control, struct cmsghdr *);
2813 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2814 			return (EINVAL);
2815 		if (cm->cmsg_level != IPPROTO_IPV6)
2816 			continue;
2817 
2818 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2819 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2820 		if (error)
2821 			return (error);
2822 	}
2823 
2824 	return (0);
2825 }
2826 
2827 /*
2828  * Set a particular packet option, as a sticky option or an ancillary data
2829  * item.  "len" can be 0 only when it's a sticky option.
2830  * We have 4 cases of combination of "sticky" and "cmsg":
2831  * "sticky=0, cmsg=0": impossible
2832  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2833  * "sticky=1, cmsg=0": RFC3542 socket option
2834  * "sticky=1, cmsg=1": RFC2292 socket option
2835  */
2836 static int
ip6_setpktopt(int optname,u_char * buf,int len,struct ip6_pktopts * opt,struct ucred * cred,int sticky,int cmsg,int uproto)2837 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2838     struct ucred *cred, int sticky, int cmsg, int uproto)
2839 {
2840 	int minmtupolicy, preftemp;
2841 	int error;
2842 
2843 	NET_EPOCH_ASSERT();
2844 
2845 	if (!sticky && !cmsg) {
2846 #ifdef DIAGNOSTIC
2847 		printf("ip6_setpktopt: impossible case\n");
2848 #endif
2849 		return (EINVAL);
2850 	}
2851 
2852 	/*
2853 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2854 	 * not be specified in the context of RFC3542.  Conversely,
2855 	 * RFC3542 types should not be specified in the context of RFC2292.
2856 	 */
2857 	if (!cmsg) {
2858 		switch (optname) {
2859 		case IPV6_2292PKTINFO:
2860 		case IPV6_2292HOPLIMIT:
2861 		case IPV6_2292NEXTHOP:
2862 		case IPV6_2292HOPOPTS:
2863 		case IPV6_2292DSTOPTS:
2864 		case IPV6_2292RTHDR:
2865 		case IPV6_2292PKTOPTIONS:
2866 			return (ENOPROTOOPT);
2867 		}
2868 	}
2869 	if (sticky && cmsg) {
2870 		switch (optname) {
2871 		case IPV6_PKTINFO:
2872 		case IPV6_HOPLIMIT:
2873 		case IPV6_NEXTHOP:
2874 		case IPV6_HOPOPTS:
2875 		case IPV6_DSTOPTS:
2876 		case IPV6_RTHDRDSTOPTS:
2877 		case IPV6_RTHDR:
2878 		case IPV6_USE_MIN_MTU:
2879 		case IPV6_DONTFRAG:
2880 		case IPV6_TCLASS:
2881 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2882 			return (ENOPROTOOPT);
2883 		}
2884 	}
2885 
2886 	switch (optname) {
2887 	case IPV6_2292PKTINFO:
2888 	case IPV6_PKTINFO:
2889 	{
2890 		struct ifnet *ifp = NULL;
2891 		struct in6_pktinfo *pktinfo;
2892 
2893 		if (len != sizeof(struct in6_pktinfo))
2894 			return (EINVAL);
2895 
2896 		pktinfo = (struct in6_pktinfo *)buf;
2897 
2898 		/*
2899 		 * An application can clear any sticky IPV6_PKTINFO option by
2900 		 * doing a "regular" setsockopt with ipi6_addr being
2901 		 * in6addr_any and ipi6_ifindex being zero.
2902 		 * [RFC 3542, Section 6]
2903 		 */
2904 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2905 		    pktinfo->ipi6_ifindex == 0 &&
2906 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2907 			ip6_clearpktopts(opt, optname);
2908 			break;
2909 		}
2910 
2911 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2912 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2913 			return (EINVAL);
2914 		}
2915 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2916 			return (EINVAL);
2917 		/* validate the interface index if specified. */
2918 		if (pktinfo->ipi6_ifindex) {
2919 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2920 			if (ifp == NULL)
2921 				return (ENXIO);
2922 		}
2923 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2924 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2925 			return (ENETDOWN);
2926 
2927 		if (ifp != NULL &&
2928 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2929 			struct in6_ifaddr *ia;
2930 
2931 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2932 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2933 			if (ia == NULL)
2934 				return (EADDRNOTAVAIL);
2935 			ifa_free(&ia->ia_ifa);
2936 		}
2937 		/*
2938 		 * We store the address anyway, and let in6_selectsrc()
2939 		 * validate the specified address.  This is because ipi6_addr
2940 		 * may not have enough information about its scope zone, and
2941 		 * we may need additional information (such as outgoing
2942 		 * interface or the scope zone of a destination address) to
2943 		 * disambiguate the scope.
2944 		 * XXX: the delay of the validation may confuse the
2945 		 * application when it is used as a sticky option.
2946 		 */
2947 		if (opt->ip6po_pktinfo == NULL) {
2948 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2949 			    M_IP6OPT, M_NOWAIT);
2950 			if (opt->ip6po_pktinfo == NULL)
2951 				return (ENOBUFS);
2952 		}
2953 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2954 		opt->ip6po_valid |= IP6PO_VALID_PKTINFO;
2955 		break;
2956 	}
2957 
2958 	case IPV6_2292HOPLIMIT:
2959 	case IPV6_HOPLIMIT:
2960 	{
2961 		int *hlimp;
2962 
2963 		/*
2964 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2965 		 * to simplify the ordering among hoplimit options.
2966 		 */
2967 		if (optname == IPV6_HOPLIMIT && sticky)
2968 			return (ENOPROTOOPT);
2969 
2970 		if (len != sizeof(int))
2971 			return (EINVAL);
2972 		hlimp = (int *)buf;
2973 		if (*hlimp < -1 || *hlimp > 255)
2974 			return (EINVAL);
2975 
2976 		opt->ip6po_hlim = *hlimp;
2977 		opt->ip6po_valid |= IP6PO_VALID_HLIM;
2978 		break;
2979 	}
2980 
2981 	case IPV6_TCLASS:
2982 	{
2983 		int tclass;
2984 
2985 		if (len != sizeof(int))
2986 			return (EINVAL);
2987 		tclass = *(int *)buf;
2988 		if (tclass < -1 || tclass > 255)
2989 			return (EINVAL);
2990 
2991 		opt->ip6po_tclass = tclass;
2992 		opt->ip6po_valid |= IP6PO_VALID_TC;
2993 		break;
2994 	}
2995 
2996 	case IPV6_2292NEXTHOP:
2997 	case IPV6_NEXTHOP:
2998 		if (cred != NULL) {
2999 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3000 			if (error)
3001 				return (error);
3002 		}
3003 
3004 		if (len == 0) {	/* just remove the option */
3005 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3006 			break;
3007 		}
3008 
3009 		/* check if cmsg_len is large enough for sa_len */
3010 		if (len < sizeof(struct sockaddr) || len < *buf)
3011 			return (EINVAL);
3012 
3013 		switch (((struct sockaddr *)buf)->sa_family) {
3014 		case AF_INET6:
3015 		{
3016 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3017 			int error;
3018 
3019 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3020 				return (EINVAL);
3021 
3022 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3023 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3024 				return (EINVAL);
3025 			}
3026 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3027 			    != 0) {
3028 				return (error);
3029 			}
3030 			break;
3031 		}
3032 		case AF_LINK:	/* should eventually be supported */
3033 		default:
3034 			return (EAFNOSUPPORT);
3035 		}
3036 
3037 		/* turn off the previous option, then set the new option. */
3038 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3039 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3040 		if (opt->ip6po_nexthop == NULL)
3041 			return (ENOBUFS);
3042 		bcopy(buf, opt->ip6po_nexthop, *buf);
3043 		opt->ip6po_valid |= IP6PO_VALID_NHINFO;
3044 		break;
3045 
3046 	case IPV6_2292HOPOPTS:
3047 	case IPV6_HOPOPTS:
3048 	{
3049 		struct ip6_hbh *hbh;
3050 		int hbhlen;
3051 
3052 		/*
3053 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3054 		 * options, since per-option restriction has too much
3055 		 * overhead.
3056 		 */
3057 		if (cred != NULL) {
3058 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3059 			if (error)
3060 				return (error);
3061 		}
3062 
3063 		if (len == 0) {
3064 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3065 			break;	/* just remove the option */
3066 		}
3067 
3068 		/* message length validation */
3069 		if (len < sizeof(struct ip6_hbh))
3070 			return (EINVAL);
3071 		hbh = (struct ip6_hbh *)buf;
3072 		hbhlen = (hbh->ip6h_len + 1) << 3;
3073 		if (len != hbhlen)
3074 			return (EINVAL);
3075 
3076 		/* turn off the previous option, then set the new option. */
3077 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3078 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3079 		if (opt->ip6po_hbh == NULL)
3080 			return (ENOBUFS);
3081 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3082 		opt->ip6po_valid |= IP6PO_VALID_HBH;
3083 
3084 		break;
3085 	}
3086 
3087 	case IPV6_2292DSTOPTS:
3088 	case IPV6_DSTOPTS:
3089 	case IPV6_RTHDRDSTOPTS:
3090 	{
3091 		struct ip6_dest *dest, **newdest = NULL;
3092 		int destlen;
3093 
3094 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3095 			error = priv_check_cred(cred, PRIV_NETINET_SETHDROPTS);
3096 			if (error)
3097 				return (error);
3098 		}
3099 
3100 		if (len == 0) {
3101 			ip6_clearpktopts(opt, optname);
3102 			break;	/* just remove the option */
3103 		}
3104 
3105 		/* message length validation */
3106 		if (len < sizeof(struct ip6_dest))
3107 			return (EINVAL);
3108 		dest = (struct ip6_dest *)buf;
3109 		destlen = (dest->ip6d_len + 1) << 3;
3110 		if (len != destlen)
3111 			return (EINVAL);
3112 
3113 		/*
3114 		 * Determine the position that the destination options header
3115 		 * should be inserted; before or after the routing header.
3116 		 */
3117 		switch (optname) {
3118 		case IPV6_2292DSTOPTS:
3119 			/*
3120 			 * The old advacned API is ambiguous on this point.
3121 			 * Our approach is to determine the position based
3122 			 * according to the existence of a routing header.
3123 			 * Note, however, that this depends on the order of the
3124 			 * extension headers in the ancillary data; the 1st
3125 			 * part of the destination options header must appear
3126 			 * before the routing header in the ancillary data,
3127 			 * too.
3128 			 * RFC3542 solved the ambiguity by introducing
3129 			 * separate ancillary data or option types.
3130 			 */
3131 			if (opt->ip6po_rthdr == NULL)
3132 				newdest = &opt->ip6po_dest1;
3133 			else
3134 				newdest = &opt->ip6po_dest2;
3135 			break;
3136 		case IPV6_RTHDRDSTOPTS:
3137 			newdest = &opt->ip6po_dest1;
3138 			break;
3139 		case IPV6_DSTOPTS:
3140 			newdest = &opt->ip6po_dest2;
3141 			break;
3142 		}
3143 
3144 		/* turn off the previous option, then set the new option. */
3145 		ip6_clearpktopts(opt, optname);
3146 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3147 		if (*newdest == NULL)
3148 			return (ENOBUFS);
3149 		bcopy(dest, *newdest, destlen);
3150 		if (newdest == &opt->ip6po_dest1)
3151 			opt->ip6po_valid |= IP6PO_VALID_DEST1;
3152 		else
3153 			opt->ip6po_valid |= IP6PO_VALID_DEST2;
3154 
3155 		break;
3156 	}
3157 
3158 	case IPV6_2292RTHDR:
3159 	case IPV6_RTHDR:
3160 	{
3161 		struct ip6_rthdr *rth;
3162 		int rthlen;
3163 
3164 		if (len == 0) {
3165 			ip6_clearpktopts(opt, IPV6_RTHDR);
3166 			break;	/* just remove the option */
3167 		}
3168 
3169 		/* message length validation */
3170 		if (len < sizeof(struct ip6_rthdr))
3171 			return (EINVAL);
3172 		rth = (struct ip6_rthdr *)buf;
3173 		rthlen = (rth->ip6r_len + 1) << 3;
3174 		if (len != rthlen)
3175 			return (EINVAL);
3176 
3177 		switch (rth->ip6r_type) {
3178 		case IPV6_RTHDR_TYPE_0:
3179 			if (rth->ip6r_len == 0)	/* must contain one addr */
3180 				return (EINVAL);
3181 			if (rth->ip6r_len % 2) /* length must be even */
3182 				return (EINVAL);
3183 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3184 				return (EINVAL);
3185 			break;
3186 		default:
3187 			return (EINVAL);	/* not supported */
3188 		}
3189 
3190 		/* turn off the previous option */
3191 		ip6_clearpktopts(opt, IPV6_RTHDR);
3192 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3193 		if (opt->ip6po_rthdr == NULL)
3194 			return (ENOBUFS);
3195 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3196 		opt->ip6po_valid |= IP6PO_VALID_RHINFO;
3197 
3198 		break;
3199 	}
3200 
3201 	case IPV6_USE_MIN_MTU:
3202 		if (len != sizeof(int))
3203 			return (EINVAL);
3204 		minmtupolicy = *(int *)buf;
3205 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3206 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3207 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3208 			return (EINVAL);
3209 		}
3210 		opt->ip6po_minmtu = minmtupolicy;
3211 		break;
3212 
3213 	case IPV6_DONTFRAG:
3214 		if (len != sizeof(int))
3215 			return (EINVAL);
3216 
3217 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3218 			/*
3219 			 * we ignore this option for TCP sockets.
3220 			 * (RFC3542 leaves this case unspecified.)
3221 			 */
3222 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3223 		} else
3224 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3225 		break;
3226 
3227 	case IPV6_PREFER_TEMPADDR:
3228 		if (len != sizeof(int))
3229 			return (EINVAL);
3230 		preftemp = *(int *)buf;
3231 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3232 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3233 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3234 			return (EINVAL);
3235 		}
3236 		opt->ip6po_prefer_tempaddr = preftemp;
3237 		break;
3238 
3239 	default:
3240 		return (ENOPROTOOPT);
3241 	} /* end of switch */
3242 
3243 	return (0);
3244 }
3245 
3246 /*
3247  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3248  * packet to the input queue of a specified interface.  Note that this
3249  * calls the output routine of the loopback "driver", but with an interface
3250  * pointer that might NOT be &loif -- easier than replicating that code here.
3251  */
3252 void
ip6_mloopback(struct ifnet * ifp,struct mbuf * m)3253 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3254 {
3255 	struct mbuf *copym;
3256 	struct ip6_hdr *ip6;
3257 
3258 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3259 	if (copym == NULL)
3260 		return;
3261 
3262 	/*
3263 	 * Make sure to deep-copy IPv6 header portion in case the data
3264 	 * is in an mbuf cluster, so that we can safely override the IPv6
3265 	 * header portion later.
3266 	 */
3267 	if (!M_WRITABLE(copym) ||
3268 	    copym->m_len < sizeof(struct ip6_hdr)) {
3269 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3270 		if (copym == NULL)
3271 			return;
3272 	}
3273 	ip6 = mtod(copym, struct ip6_hdr *);
3274 	/*
3275 	 * clear embedded scope identifiers if necessary.
3276 	 * in6_clearscope will touch the addresses only when necessary.
3277 	 */
3278 	in6_clearscope(&ip6->ip6_src);
3279 	in6_clearscope(&ip6->ip6_dst);
3280 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3281 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3282 		    CSUM_PSEUDO_HDR;
3283 		copym->m_pkthdr.csum_data = 0xffff;
3284 	}
3285 	if_simloop(ifp, copym, AF_INET6, 0);
3286 }
3287 
3288 /*
3289  * Chop IPv6 header off from the payload.
3290  */
3291 static int
ip6_splithdr(struct mbuf * m,struct ip6_exthdrs * exthdrs)3292 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3293 {
3294 	struct mbuf *mh;
3295 	struct ip6_hdr *ip6;
3296 
3297 	ip6 = mtod(m, struct ip6_hdr *);
3298 	if (m->m_len > sizeof(*ip6)) {
3299 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3300 		if (mh == NULL) {
3301 			m_freem(m);
3302 			return ENOBUFS;
3303 		}
3304 		m_move_pkthdr(mh, m);
3305 		M_ALIGN(mh, sizeof(*ip6));
3306 		m->m_len -= sizeof(*ip6);
3307 		m->m_data += sizeof(*ip6);
3308 		mh->m_next = m;
3309 		m = mh;
3310 		m->m_len = sizeof(*ip6);
3311 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3312 	}
3313 	exthdrs->ip6e_ip6 = m;
3314 	return 0;
3315 }
3316 
3317 /*
3318  * Compute IPv6 extension header length.
3319  */
3320 int
ip6_optlen(struct inpcb * inp)3321 ip6_optlen(struct inpcb *inp)
3322 {
3323 	int len;
3324 
3325 	if (!inp->in6p_outputopts)
3326 		return 0;
3327 
3328 	len = 0;
3329 #define elen(x) \
3330     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3331 
3332 	len += elen(inp->in6p_outputopts->ip6po_hbh);
3333 	if (inp->in6p_outputopts->ip6po_rthdr)
3334 		/* dest1 is valid with rthdr only */
3335 		len += elen(inp->in6p_outputopts->ip6po_dest1);
3336 	len += elen(inp->in6p_outputopts->ip6po_rthdr);
3337 	len += elen(inp->in6p_outputopts->ip6po_dest2);
3338 	return len;
3339 #undef elen
3340 }
3341