ip6_output.c revision 1.164 1 /* $NetBSD: ip6_output.c,v 1.164 2015/04/24 08:53:06 ozaki-r Exp $ */
2 /* $KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $ */
3
4 /*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 /*
34 * Copyright (c) 1982, 1986, 1988, 1990, 1993
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
62 */
63
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.164 2015/04/24 08:53:06 ozaki-r Exp $");
66
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/kauth.h>
81
82 #include <net/if.h>
83 #include <net/route.h>
84 #include <net/pfil.h>
85
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet/ip_var.h>
90 #include <netinet/icmp6.h>
91 #include <netinet/in_offload.h>
92 #include <netinet/portalgo.h>
93 #include <netinet6/in6_offload.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/ip6_private.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/ip6protosw.h>
99 #include <netinet6/scope6_var.h>
100
101 #ifdef IPSEC
102 #include <netipsec/ipsec.h>
103 #include <netipsec/ipsec6.h>
104 #include <netipsec/key.h>
105 #include <netipsec/xform.h>
106 #endif
107
108
109 #include <net/net_osdep.h>
110
111 extern pfil_head_t *inet6_pfil_hook; /* XXX */
112
113 struct ip6_exthdrs {
114 struct mbuf *ip6e_ip6;
115 struct mbuf *ip6e_hbh;
116 struct mbuf *ip6e_dest1;
117 struct mbuf *ip6e_rthdr;
118 struct mbuf *ip6e_dest2;
119 };
120
121 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
122 kauth_cred_t, int);
123 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
124 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
125 int, int, int);
126 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *);
127 static int ip6_getmoptions(struct sockopt *, struct in6pcb *);
128 static int ip6_copyexthdr(struct mbuf **, void *, int);
129 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
130 struct ip6_frag **);
131 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
132 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
133 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
134 const struct in6_addr *, u_long *, int *);
135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
136
137 #ifdef RFC2292
138 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
139 #endif
140
141 /*
142 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
143 * header (with pri, len, nxt, hlim, src, dst).
144 * This function may modify ver and hlim only.
145 * The mbuf chain containing the packet will be freed.
146 * The mbuf opt, if present, will not be freed.
147 *
148 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
149 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
150 * which is rt_rmx.rmx_mtu.
151 */
152 int
153 ip6_output(
154 struct mbuf *m0,
155 struct ip6_pktopts *opt,
156 struct route *ro,
157 int flags,
158 struct ip6_moptions *im6o,
159 struct socket *so,
160 struct ifnet **ifpp /* XXX: just for statistics */
161 )
162 {
163 struct ip6_hdr *ip6, *mhip6;
164 struct ifnet *ifp, *origifp;
165 struct mbuf *m = m0;
166 int hlen, tlen, len, off;
167 bool tso;
168 struct route ip6route;
169 struct rtentry *rt = NULL;
170 const struct sockaddr_in6 *dst;
171 struct sockaddr_in6 src_sa, dst_sa;
172 int error = 0;
173 struct in6_ifaddr *ia = NULL;
174 u_long mtu;
175 int alwaysfrag, dontfrag;
176 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
177 struct ip6_exthdrs exthdrs;
178 struct in6_addr finaldst, src0, dst0;
179 u_int32_t zone;
180 struct route *ro_pmtu = NULL;
181 int hdrsplit = 0;
182 int needipsec = 0;
183 #ifdef IPSEC
184 struct secpolicy *sp = NULL;
185 #endif
186
187 memset(&ip6route, 0, sizeof(ip6route));
188
189 #ifdef DIAGNOSTIC
190 if ((m->m_flags & M_PKTHDR) == 0)
191 panic("ip6_output: no HDR");
192
193 if ((m->m_pkthdr.csum_flags &
194 (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
195 panic("ip6_output: IPv4 checksum offload flags: %d",
196 m->m_pkthdr.csum_flags);
197 }
198
199 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
200 (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
201 panic("ip6_output: conflicting checksum offload flags: %d",
202 m->m_pkthdr.csum_flags);
203 }
204 #endif
205
206 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
207
208 #define MAKE_EXTHDR(hp, mp) \
209 do { \
210 if (hp) { \
211 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
212 error = ip6_copyexthdr((mp), (void *)(hp), \
213 ((eh)->ip6e_len + 1) << 3); \
214 if (error) \
215 goto freehdrs; \
216 } \
217 } while (/*CONSTCOND*/ 0)
218
219 memset(&exthdrs, 0, sizeof(exthdrs));
220 if (opt) {
221 /* Hop-by-Hop options header */
222 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
223 /* Destination options header(1st part) */
224 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
225 /* Routing header */
226 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
227 /* Destination options header(2nd part) */
228 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
229 }
230
231 /*
232 * Calculate the total length of the extension header chain.
233 * Keep the length of the unfragmentable part for fragmentation.
234 */
235 optlen = 0;
236 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
237 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
238 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
239 unfragpartlen = optlen + sizeof(struct ip6_hdr);
240 /* NOTE: we don't add AH/ESP length here. do that later. */
241 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
242
243 #ifdef IPSEC
244 if (ipsec_used) {
245 /* Check the security policy (SP) for the packet */
246
247 sp = ipsec6_check_policy(m, so, flags, &needipsec, &error);
248 if (error != 0) {
249 /*
250 * Hack: -EINVAL is used to signal that a packet
251 * should be silently discarded. This is typically
252 * because we asked key management for an SA and
253 * it was delayed (e.g. kicked up to IKE).
254 */
255 if (error == -EINVAL)
256 error = 0;
257 goto freehdrs;
258 }
259 }
260 #endif /* IPSEC */
261
262
263 if (needipsec &&
264 (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
265 in6_delayed_cksum(m);
266 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
267 }
268
269
270 /*
271 * If we need IPsec, or there is at least one extension header,
272 * separate IP6 header from the payload.
273 */
274 if ((needipsec || optlen) && !hdrsplit) {
275 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
276 m = NULL;
277 goto freehdrs;
278 }
279 m = exthdrs.ip6e_ip6;
280 hdrsplit++;
281 }
282
283 /* adjust pointer */
284 ip6 = mtod(m, struct ip6_hdr *);
285
286 /* adjust mbuf packet header length */
287 m->m_pkthdr.len += optlen;
288 plen = m->m_pkthdr.len - sizeof(*ip6);
289
290 /* If this is a jumbo payload, insert a jumbo payload option. */
291 if (plen > IPV6_MAXPACKET) {
292 if (!hdrsplit) {
293 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
294 m = NULL;
295 goto freehdrs;
296 }
297 m = exthdrs.ip6e_ip6;
298 hdrsplit++;
299 }
300 /* adjust pointer */
301 ip6 = mtod(m, struct ip6_hdr *);
302 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
303 goto freehdrs;
304 optlen += 8; /* XXX JUMBOOPTLEN */
305 ip6->ip6_plen = 0;
306 } else
307 ip6->ip6_plen = htons(plen);
308
309 /*
310 * Concatenate headers and fill in next header fields.
311 * Here we have, on "m"
312 * IPv6 payload
313 * and we insert headers accordingly. Finally, we should be getting:
314 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
315 *
316 * during the header composing process, "m" points to IPv6 header.
317 * "mprev" points to an extension header prior to esp.
318 */
319 {
320 u_char *nexthdrp = &ip6->ip6_nxt;
321 struct mbuf *mprev = m;
322
323 /*
324 * we treat dest2 specially. this makes IPsec processing
325 * much easier. the goal here is to make mprev point the
326 * mbuf prior to dest2.
327 *
328 * result: IPv6 dest2 payload
329 * m and mprev will point to IPv6 header.
330 */
331 if (exthdrs.ip6e_dest2) {
332 if (!hdrsplit)
333 panic("assumption failed: hdr not split");
334 exthdrs.ip6e_dest2->m_next = m->m_next;
335 m->m_next = exthdrs.ip6e_dest2;
336 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
337 ip6->ip6_nxt = IPPROTO_DSTOPTS;
338 }
339
340 #define MAKE_CHAIN(m, mp, p, i)\
341 do {\
342 if (m) {\
343 if (!hdrsplit) \
344 panic("assumption failed: hdr not split"); \
345 *mtod((m), u_char *) = *(p);\
346 *(p) = (i);\
347 p = mtod((m), u_char *);\
348 (m)->m_next = (mp)->m_next;\
349 (mp)->m_next = (m);\
350 (mp) = (m);\
351 }\
352 } while (/*CONSTCOND*/ 0)
353 /*
354 * result: IPv6 hbh dest1 rthdr dest2 payload
355 * m will point to IPv6 header. mprev will point to the
356 * extension header prior to dest2 (rthdr in the above case).
357 */
358 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
359 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
360 IPPROTO_DSTOPTS);
361 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
362 IPPROTO_ROUTING);
363
364 M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
365 sizeof(struct ip6_hdr) + optlen);
366 }
367
368 /*
369 * If there is a routing header, replace destination address field
370 * with the first hop of the routing header.
371 */
372 if (exthdrs.ip6e_rthdr) {
373 struct ip6_rthdr *rh;
374 struct ip6_rthdr0 *rh0;
375 struct in6_addr *addr;
376 struct sockaddr_in6 sa;
377
378 rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
379 struct ip6_rthdr *));
380 finaldst = ip6->ip6_dst;
381 switch (rh->ip6r_type) {
382 case IPV6_RTHDR_TYPE_0:
383 rh0 = (struct ip6_rthdr0 *)rh;
384 addr = (struct in6_addr *)(rh0 + 1);
385
386 /*
387 * construct a sockaddr_in6 form of
388 * the first hop.
389 *
390 * XXX: we may not have enough
391 * information about its scope zone;
392 * there is no standard API to pass
393 * the information from the
394 * application.
395 */
396 sockaddr_in6_init(&sa, addr, 0, 0, 0);
397 if ((error = sa6_embedscope(&sa,
398 ip6_use_defzone)) != 0) {
399 goto bad;
400 }
401 ip6->ip6_dst = sa.sin6_addr;
402 (void)memmove(&addr[0], &addr[1],
403 sizeof(struct in6_addr) *
404 (rh0->ip6r0_segleft - 1));
405 addr[rh0->ip6r0_segleft - 1] = finaldst;
406 /* XXX */
407 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
408 break;
409 default: /* is it possible? */
410 error = EINVAL;
411 goto bad;
412 }
413 }
414
415 /* Source address validation */
416 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
417 (flags & IPV6_UNSPECSRC) == 0) {
418 error = EOPNOTSUPP;
419 IP6_STATINC(IP6_STAT_BADSCOPE);
420 goto bad;
421 }
422 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
423 error = EOPNOTSUPP;
424 IP6_STATINC(IP6_STAT_BADSCOPE);
425 goto bad;
426 }
427
428 IP6_STATINC(IP6_STAT_LOCALOUT);
429
430 /*
431 * Route packet.
432 */
433 /* initialize cached route */
434 if (ro == NULL) {
435 ro = &ip6route;
436 }
437 ro_pmtu = ro;
438 if (opt && opt->ip6po_rthdr)
439 ro = &opt->ip6po_route;
440
441 /*
442 * if specified, try to fill in the traffic class field.
443 * do not override if a non-zero value is already set.
444 * we check the diffserv field and the ecn field separately.
445 */
446 if (opt && opt->ip6po_tclass >= 0) {
447 int mask = 0;
448
449 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
450 mask |= 0xfc;
451 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
452 mask |= 0x03;
453 if (mask != 0)
454 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
455 }
456
457 /* fill in or override the hop limit field, if necessary. */
458 if (opt && opt->ip6po_hlim != -1)
459 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
460 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
461 if (im6o != NULL)
462 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
463 else
464 ip6->ip6_hlim = ip6_defmcasthlim;
465 }
466
467 #ifdef IPSEC
468 if (needipsec) {
469 int s = splsoftnet();
470 error = ipsec6_process_packet(m, sp->req);
471
472 /*
473 * Preserve KAME behaviour: ENOENT can be returned
474 * when an SA acquire is in progress. Don't propagate
475 * this to user-level; it confuses applications.
476 * XXX this will go away when the SADB is redone.
477 */
478 if (error == ENOENT)
479 error = 0;
480 splx(s);
481 goto done;
482 }
483 #endif /* IPSEC */
484
485 /* adjust pointer */
486 ip6 = mtod(m, struct ip6_hdr *);
487
488 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
489 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
490 &ifp, &rt, 0)) != 0) {
491 if (ifp != NULL)
492 in6_ifstat_inc(ifp, ifs6_out_discard);
493 goto bad;
494 }
495 if (rt == NULL) {
496 /*
497 * If in6_selectroute() does not return a route entry,
498 * dst may not have been updated.
499 */
500 error = rtcache_setdst(ro, sin6tosa(&dst_sa));
501 if (error) {
502 goto bad;
503 }
504 }
505
506 /*
507 * then rt (for unicast) and ifp must be non-NULL valid values.
508 */
509 if ((flags & IPV6_FORWARDING) == 0) {
510 /* XXX: the FORWARDING flag can be set for mrouting. */
511 in6_ifstat_inc(ifp, ifs6_out_request);
512 }
513 if (rt != NULL) {
514 ia = (struct in6_ifaddr *)(rt->rt_ifa);
515 rt->rt_use++;
516 }
517
518 /*
519 * The outgoing interface must be in the zone of source and
520 * destination addresses. We should use ia_ifp to support the
521 * case of sending packets to an address of our own.
522 */
523 if (ia != NULL && ia->ia_ifp)
524 origifp = ia->ia_ifp;
525 else
526 origifp = ifp;
527
528 src0 = ip6->ip6_src;
529 if (in6_setscope(&src0, origifp, &zone))
530 goto badscope;
531 sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
532 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
533 goto badscope;
534
535 dst0 = ip6->ip6_dst;
536 if (in6_setscope(&dst0, origifp, &zone))
537 goto badscope;
538 /* re-initialize to be sure */
539 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
540 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
541 goto badscope;
542
543 /* scope check is done. */
544
545 if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
546 dst = satocsin6(rtcache_getdst(ro));
547 KASSERT(dst != NULL);
548 } else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
549 /*
550 * The nexthop is explicitly specified by the
551 * application. We assume the next hop is an IPv6
552 * address.
553 */
554 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
555 } else if ((rt->rt_flags & RTF_GATEWAY))
556 dst = (struct sockaddr_in6 *)rt->rt_gateway;
557 else
558 dst = satocsin6(rtcache_getdst(ro));
559
560 /*
561 * XXXXXX: original code follows:
562 */
563 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
564 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
565 else {
566 struct in6_multi *in6m;
567
568 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
569
570 in6_ifstat_inc(ifp, ifs6_out_mcast);
571
572 /*
573 * Confirm that the outgoing interface supports multicast.
574 */
575 if (!(ifp->if_flags & IFF_MULTICAST)) {
576 IP6_STATINC(IP6_STAT_NOROUTE);
577 in6_ifstat_inc(ifp, ifs6_out_discard);
578 error = ENETUNREACH;
579 goto bad;
580 }
581
582 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
583 if (in6m != NULL &&
584 (im6o == NULL || im6o->im6o_multicast_loop)) {
585 /*
586 * If we belong to the destination multicast group
587 * on the outgoing interface, and the caller did not
588 * forbid loopback, loop back a copy.
589 */
590 KASSERT(dst != NULL);
591 ip6_mloopback(ifp, m, dst);
592 } else {
593 /*
594 * If we are acting as a multicast router, perform
595 * multicast forwarding as if the packet had just
596 * arrived on the interface to which we are about
597 * to send. The multicast forwarding function
598 * recursively calls this function, using the
599 * IPV6_FORWARDING flag to prevent infinite recursion.
600 *
601 * Multicasts that are looped back by ip6_mloopback(),
602 * above, will be forwarded by the ip6_input() routine,
603 * if necessary.
604 */
605 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
606 if (ip6_mforward(ip6, ifp, m) != 0) {
607 m_freem(m);
608 goto done;
609 }
610 }
611 }
612 /*
613 * Multicasts with a hoplimit of zero may be looped back,
614 * above, but must not be transmitted on a network.
615 * Also, multicasts addressed to the loopback interface
616 * are not sent -- the above call to ip6_mloopback() will
617 * loop back a copy if this host actually belongs to the
618 * destination group on the loopback interface.
619 */
620 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
621 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
622 m_freem(m);
623 goto done;
624 }
625 }
626
627 /*
628 * Fill the outgoing inteface to tell the upper layer
629 * to increment per-interface statistics.
630 */
631 if (ifpp)
632 *ifpp = ifp;
633
634 /* Determine path MTU. */
635 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
636 &alwaysfrag)) != 0)
637 goto bad;
638
639 /*
640 * The caller of this function may specify to use the minimum MTU
641 * in some cases.
642 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
643 * setting. The logic is a bit complicated; by default, unicast
644 * packets will follow path MTU while multicast packets will be sent at
645 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
646 * including unicast ones will be sent at the minimum MTU. Multicast
647 * packets will always be sent at the minimum MTU unless
648 * IP6PO_MINMTU_DISABLE is explicitly specified.
649 * See RFC 3542 for more details.
650 */
651 if (mtu > IPV6_MMTU) {
652 if ((flags & IPV6_MINMTU))
653 mtu = IPV6_MMTU;
654 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
655 mtu = IPV6_MMTU;
656 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
657 (opt == NULL ||
658 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
659 mtu = IPV6_MMTU;
660 }
661 }
662
663 /*
664 * clear embedded scope identifiers if necessary.
665 * in6_clearscope will touch the addresses only when necessary.
666 */
667 in6_clearscope(&ip6->ip6_src);
668 in6_clearscope(&ip6->ip6_dst);
669
670 /*
671 * If the outgoing packet contains a hop-by-hop options header,
672 * it must be examined and processed even by the source node.
673 * (RFC 2460, section 4.)
674 */
675 if (ip6->ip6_nxt == IPV6_HOPOPTS) {
676 u_int32_t dummy1; /* XXX unused */
677 u_int32_t dummy2; /* XXX unused */
678 int hoff = sizeof(struct ip6_hdr);
679
680 if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
681 /* m was already freed at this point */
682 error = EINVAL;/* better error? */
683 goto done;
684 }
685
686 ip6 = mtod(m, struct ip6_hdr *);
687 }
688
689 /*
690 * Run through list of hooks for output packets.
691 */
692 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
693 goto done;
694 if (m == NULL)
695 goto done;
696 ip6 = mtod(m, struct ip6_hdr *);
697
698 /*
699 * Send the packet to the outgoing interface.
700 * If necessary, do IPv6 fragmentation before sending.
701 *
702 * the logic here is rather complex:
703 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
704 * 1-a: send as is if tlen <= path mtu
705 * 1-b: fragment if tlen > path mtu
706 *
707 * 2: if user asks us not to fragment (dontfrag == 1)
708 * 2-a: send as is if tlen <= interface mtu
709 * 2-b: error if tlen > interface mtu
710 *
711 * 3: if we always need to attach fragment header (alwaysfrag == 1)
712 * always fragment
713 *
714 * 4: if dontfrag == 1 && alwaysfrag == 1
715 * error, as we cannot handle this conflicting request
716 */
717 tlen = m->m_pkthdr.len;
718 tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
719 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
720 dontfrag = 1;
721 else
722 dontfrag = 0;
723
724 if (dontfrag && alwaysfrag) { /* case 4 */
725 /* conflicting request - can't transmit */
726 error = EMSGSIZE;
727 goto bad;
728 }
729 if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) { /* case 2-b */
730 /*
731 * Even if the DONTFRAG option is specified, we cannot send the
732 * packet when the data length is larger than the MTU of the
733 * outgoing interface.
734 * Notify the error by sending IPV6_PATHMTU ancillary data as
735 * well as returning an error code (the latter is not described
736 * in the API spec.)
737 */
738 u_int32_t mtu32;
739 struct ip6ctlparam ip6cp;
740
741 mtu32 = (u_int32_t)mtu;
742 memset(&ip6cp, 0, sizeof(ip6cp));
743 ip6cp.ip6c_cmdarg = (void *)&mtu32;
744 pfctlinput2(PRC_MSGSIZE,
745 rtcache_getdst(ro_pmtu), &ip6cp);
746
747 error = EMSGSIZE;
748 goto bad;
749 }
750
751 /*
752 * transmit packet without fragmentation
753 */
754 if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
755 /* case 1-a and 2-a */
756 struct in6_ifaddr *ia6;
757 int sw_csum;
758
759 ip6 = mtod(m, struct ip6_hdr *);
760 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
761 if (ia6) {
762 /* Record statistics for this interface address. */
763 ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
764 }
765
766 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
767 if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
768 if (IN6_NEED_CHECKSUM(ifp,
769 sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
770 in6_delayed_cksum(m);
771 }
772 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
773 }
774
775 KASSERT(dst != NULL);
776 if (__predict_true(!tso ||
777 (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
778 error = nd6_output(ifp, origifp, m, dst, rt);
779 } else {
780 error = ip6_tso_output(ifp, origifp, m, dst, rt);
781 }
782 goto done;
783 }
784
785 if (tso) {
786 error = EINVAL; /* XXX */
787 goto bad;
788 }
789
790 /*
791 * try to fragment the packet. case 1-b and 3
792 */
793 if (mtu < IPV6_MMTU) {
794 /* path MTU cannot be less than IPV6_MMTU */
795 error = EMSGSIZE;
796 in6_ifstat_inc(ifp, ifs6_out_fragfail);
797 goto bad;
798 } else if (ip6->ip6_plen == 0) {
799 /* jumbo payload cannot be fragmented */
800 error = EMSGSIZE;
801 in6_ifstat_inc(ifp, ifs6_out_fragfail);
802 goto bad;
803 } else {
804 struct mbuf **mnext, *m_frgpart;
805 struct ip6_frag *ip6f;
806 u_int32_t id = htonl(ip6_randomid());
807 u_char nextproto;
808 #if 0 /* see below */
809 struct ip6ctlparam ip6cp;
810 u_int32_t mtu32;
811 #endif
812
813 /*
814 * Too large for the destination or interface;
815 * fragment if possible.
816 * Must be able to put at least 8 bytes per fragment.
817 */
818 hlen = unfragpartlen;
819 if (mtu > IPV6_MAXPACKET)
820 mtu = IPV6_MAXPACKET;
821
822 #if 0
823 /*
824 * It is believed this code is a leftover from the
825 * development of the IPV6_RECVPATHMTU sockopt and
826 * associated work to implement RFC3542.
827 * It's not entirely clear what the intent of the API
828 * is at this point, so disable this code for now.
829 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
830 * will send notifications if the application requests.
831 */
832
833 /* Notify a proper path MTU to applications. */
834 mtu32 = (u_int32_t)mtu;
835 memset(&ip6cp, 0, sizeof(ip6cp));
836 ip6cp.ip6c_cmdarg = (void *)&mtu32;
837 pfctlinput2(PRC_MSGSIZE,
838 rtcache_getdst(ro_pmtu), &ip6cp);
839 #endif
840
841 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
842 if (len < 8) {
843 error = EMSGSIZE;
844 in6_ifstat_inc(ifp, ifs6_out_fragfail);
845 goto bad;
846 }
847
848 mnext = &m->m_nextpkt;
849
850 /*
851 * Change the next header field of the last header in the
852 * unfragmentable part.
853 */
854 if (exthdrs.ip6e_rthdr) {
855 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
856 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
857 } else if (exthdrs.ip6e_dest1) {
858 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
859 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
860 } else if (exthdrs.ip6e_hbh) {
861 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
862 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
863 } else {
864 nextproto = ip6->ip6_nxt;
865 ip6->ip6_nxt = IPPROTO_FRAGMENT;
866 }
867
868 if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
869 != 0) {
870 if (IN6_NEED_CHECKSUM(ifp,
871 m->m_pkthdr.csum_flags &
872 (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
873 in6_delayed_cksum(m);
874 }
875 m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
876 }
877
878 /*
879 * Loop through length of segment after first fragment,
880 * make new header and copy data of each part and link onto
881 * chain.
882 */
883 m0 = m;
884 for (off = hlen; off < tlen; off += len) {
885 struct mbuf *mlast;
886
887 MGETHDR(m, M_DONTWAIT, MT_HEADER);
888 if (!m) {
889 error = ENOBUFS;
890 IP6_STATINC(IP6_STAT_ODROPPED);
891 goto sendorfree;
892 }
893 m->m_pkthdr.rcvif = NULL;
894 m->m_flags = m0->m_flags & M_COPYFLAGS;
895 *mnext = m;
896 mnext = &m->m_nextpkt;
897 m->m_data += max_linkhdr;
898 mhip6 = mtod(m, struct ip6_hdr *);
899 *mhip6 = *ip6;
900 m->m_len = sizeof(*mhip6);
901 /*
902 * ip6f must be valid if error is 0. But how
903 * can a compiler be expected to infer this?
904 */
905 ip6f = NULL;
906 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
907 if (error) {
908 IP6_STATINC(IP6_STAT_ODROPPED);
909 goto sendorfree;
910 }
911 ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
912 if (off + len >= tlen)
913 len = tlen - off;
914 else
915 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
916 mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
917 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
918 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
919 error = ENOBUFS;
920 IP6_STATINC(IP6_STAT_ODROPPED);
921 goto sendorfree;
922 }
923 for (mlast = m; mlast->m_next; mlast = mlast->m_next)
924 ;
925 mlast->m_next = m_frgpart;
926 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
927 m->m_pkthdr.rcvif = NULL;
928 ip6f->ip6f_reserved = 0;
929 ip6f->ip6f_ident = id;
930 ip6f->ip6f_nxt = nextproto;
931 IP6_STATINC(IP6_STAT_OFRAGMENTS);
932 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
933 }
934
935 in6_ifstat_inc(ifp, ifs6_out_fragok);
936 }
937
938 /*
939 * Remove leading garbages.
940 */
941 sendorfree:
942 m = m0->m_nextpkt;
943 m0->m_nextpkt = 0;
944 m_freem(m0);
945 for (m0 = m; m; m = m0) {
946 m0 = m->m_nextpkt;
947 m->m_nextpkt = 0;
948 if (error == 0) {
949 struct in6_ifaddr *ia6;
950 ip6 = mtod(m, struct ip6_hdr *);
951 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
952 if (ia6) {
953 /*
954 * Record statistics for this interface
955 * address.
956 */
957 ia6->ia_ifa.ifa_data.ifad_outbytes +=
958 m->m_pkthdr.len;
959 }
960 KASSERT(dst != NULL);
961 error = nd6_output(ifp, origifp, m, dst, rt);
962 } else
963 m_freem(m);
964 }
965
966 if (error == 0)
967 IP6_STATINC(IP6_STAT_FRAGMENTED);
968
969 done:
970 rtcache_free(&ip6route);
971
972 #ifdef IPSEC
973 if (sp != NULL)
974 KEY_FREESP(&sp);
975 #endif /* IPSEC */
976
977
978 return (error);
979
980 freehdrs:
981 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
982 m_freem(exthdrs.ip6e_dest1);
983 m_freem(exthdrs.ip6e_rthdr);
984 m_freem(exthdrs.ip6e_dest2);
985 /* FALLTHROUGH */
986 bad:
987 m_freem(m);
988 goto done;
989 badscope:
990 IP6_STATINC(IP6_STAT_BADSCOPE);
991 in6_ifstat_inc(origifp, ifs6_out_discard);
992 if (error == 0)
993 error = EHOSTUNREACH; /* XXX */
994 goto bad;
995 }
996
997 static int
998 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
999 {
1000 struct mbuf *m;
1001
1002 if (hlen > MCLBYTES)
1003 return (ENOBUFS); /* XXX */
1004
1005 MGET(m, M_DONTWAIT, MT_DATA);
1006 if (!m)
1007 return (ENOBUFS);
1008
1009 if (hlen > MLEN) {
1010 MCLGET(m, M_DONTWAIT);
1011 if ((m->m_flags & M_EXT) == 0) {
1012 m_free(m);
1013 return (ENOBUFS);
1014 }
1015 }
1016 m->m_len = hlen;
1017 if (hdr)
1018 bcopy(hdr, mtod(m, void *), hlen);
1019
1020 *mp = m;
1021 return (0);
1022 }
1023
1024 /*
1025 * Process a delayed payload checksum calculation.
1026 */
1027 void
1028 in6_delayed_cksum(struct mbuf *m)
1029 {
1030 uint16_t csum, offset;
1031
1032 KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1033 KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1034 KASSERT((m->m_pkthdr.csum_flags
1035 & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1036
1037 offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1038 csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1039 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1040 csum = 0xffff;
1041 }
1042
1043 offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1044 if ((offset + sizeof(csum)) > m->m_len) {
1045 m_copyback(m, offset, sizeof(csum), &csum);
1046 } else {
1047 *(uint16_t *)(mtod(m, char *) + offset) = csum;
1048 }
1049 }
1050
1051 /*
1052 * Insert jumbo payload option.
1053 */
1054 static int
1055 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1056 {
1057 struct mbuf *mopt;
1058 u_int8_t *optbuf;
1059 u_int32_t v;
1060
1061 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1062
1063 /*
1064 * If there is no hop-by-hop options header, allocate new one.
1065 * If there is one but it doesn't have enough space to store the
1066 * jumbo payload option, allocate a cluster to store the whole options.
1067 * Otherwise, use it to store the options.
1068 */
1069 if (exthdrs->ip6e_hbh == 0) {
1070 MGET(mopt, M_DONTWAIT, MT_DATA);
1071 if (mopt == 0)
1072 return (ENOBUFS);
1073 mopt->m_len = JUMBOOPTLEN;
1074 optbuf = mtod(mopt, u_int8_t *);
1075 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1076 exthdrs->ip6e_hbh = mopt;
1077 } else {
1078 struct ip6_hbh *hbh;
1079
1080 mopt = exthdrs->ip6e_hbh;
1081 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1082 /*
1083 * XXX assumption:
1084 * - exthdrs->ip6e_hbh is not referenced from places
1085 * other than exthdrs.
1086 * - exthdrs->ip6e_hbh is not an mbuf chain.
1087 */
1088 int oldoptlen = mopt->m_len;
1089 struct mbuf *n;
1090
1091 /*
1092 * XXX: give up if the whole (new) hbh header does
1093 * not fit even in an mbuf cluster.
1094 */
1095 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1096 return (ENOBUFS);
1097
1098 /*
1099 * As a consequence, we must always prepare a cluster
1100 * at this point.
1101 */
1102 MGET(n, M_DONTWAIT, MT_DATA);
1103 if (n) {
1104 MCLGET(n, M_DONTWAIT);
1105 if ((n->m_flags & M_EXT) == 0) {
1106 m_freem(n);
1107 n = NULL;
1108 }
1109 }
1110 if (!n)
1111 return (ENOBUFS);
1112 n->m_len = oldoptlen + JUMBOOPTLEN;
1113 bcopy(mtod(mopt, void *), mtod(n, void *),
1114 oldoptlen);
1115 optbuf = mtod(n, u_int8_t *) + oldoptlen;
1116 m_freem(mopt);
1117 mopt = exthdrs->ip6e_hbh = n;
1118 } else {
1119 optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1120 mopt->m_len += JUMBOOPTLEN;
1121 }
1122 optbuf[0] = IP6OPT_PADN;
1123 optbuf[1] = 0;
1124
1125 /*
1126 * Adjust the header length according to the pad and
1127 * the jumbo payload option.
1128 */
1129 hbh = mtod(mopt, struct ip6_hbh *);
1130 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1131 }
1132
1133 /* fill in the option. */
1134 optbuf[2] = IP6OPT_JUMBO;
1135 optbuf[3] = 4;
1136 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1137 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1138
1139 /* finally, adjust the packet header length */
1140 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1141
1142 return (0);
1143 #undef JUMBOOPTLEN
1144 }
1145
1146 /*
1147 * Insert fragment header and copy unfragmentable header portions.
1148 *
1149 * *frghdrp will not be read, and it is guaranteed that either an
1150 * error is returned or that *frghdrp will point to space allocated
1151 * for the fragment header.
1152 */
1153 static int
1154 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1155 struct ip6_frag **frghdrp)
1156 {
1157 struct mbuf *n, *mlast;
1158
1159 if (hlen > sizeof(struct ip6_hdr)) {
1160 n = m_copym(m0, sizeof(struct ip6_hdr),
1161 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1162 if (n == 0)
1163 return (ENOBUFS);
1164 m->m_next = n;
1165 } else
1166 n = m;
1167
1168 /* Search for the last mbuf of unfragmentable part. */
1169 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1170 ;
1171
1172 if ((mlast->m_flags & M_EXT) == 0 &&
1173 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1174 /* use the trailing space of the last mbuf for the fragment hdr */
1175 *frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1176 mlast->m_len);
1177 mlast->m_len += sizeof(struct ip6_frag);
1178 m->m_pkthdr.len += sizeof(struct ip6_frag);
1179 } else {
1180 /* allocate a new mbuf for the fragment header */
1181 struct mbuf *mfrg;
1182
1183 MGET(mfrg, M_DONTWAIT, MT_DATA);
1184 if (mfrg == 0)
1185 return (ENOBUFS);
1186 mfrg->m_len = sizeof(struct ip6_frag);
1187 *frghdrp = mtod(mfrg, struct ip6_frag *);
1188 mlast->m_next = mfrg;
1189 }
1190
1191 return (0);
1192 }
1193
1194 static int
1195 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1196 const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1197 {
1198 struct rtentry *rt;
1199 u_int32_t mtu = 0;
1200 int alwaysfrag = 0;
1201 int error = 0;
1202
1203 if (ro_pmtu != ro) {
1204 union {
1205 struct sockaddr dst;
1206 struct sockaddr_in6 dst6;
1207 } u;
1208
1209 /* The first hop and the final destination may differ. */
1210 sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1211 rt = rtcache_lookup(ro_pmtu, &u.dst);
1212 } else
1213 rt = rtcache_validate(ro_pmtu);
1214 if (rt != NULL) {
1215 u_int32_t ifmtu;
1216
1217 if (ifp == NULL)
1218 ifp = rt->rt_ifp;
1219 ifmtu = IN6_LINKMTU(ifp);
1220 mtu = rt->rt_rmx.rmx_mtu;
1221 if (mtu == 0)
1222 mtu = ifmtu;
1223 else if (mtu < IPV6_MMTU) {
1224 /*
1225 * RFC2460 section 5, last paragraph:
1226 * if we record ICMPv6 too big message with
1227 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1228 * or smaller, with fragment header attached.
1229 * (fragment header is needed regardless from the
1230 * packet size, for translators to identify packets)
1231 */
1232 alwaysfrag = 1;
1233 mtu = IPV6_MMTU;
1234 } else if (mtu > ifmtu) {
1235 /*
1236 * The MTU on the route is larger than the MTU on
1237 * the interface! This shouldn't happen, unless the
1238 * MTU of the interface has been changed after the
1239 * interface was brought up. Change the MTU in the
1240 * route to match the interface MTU (as long as the
1241 * field isn't locked).
1242 */
1243 mtu = ifmtu;
1244 if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1245 rt->rt_rmx.rmx_mtu = mtu;
1246 }
1247 } else if (ifp) {
1248 mtu = IN6_LINKMTU(ifp);
1249 } else
1250 error = EHOSTUNREACH; /* XXX */
1251
1252 *mtup = mtu;
1253 if (alwaysfragp)
1254 *alwaysfragp = alwaysfrag;
1255 return (error);
1256 }
1257
1258 /*
1259 * IP6 socket option processing.
1260 */
1261 int
1262 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1263 {
1264 int optdatalen, uproto;
1265 void *optdata;
1266 struct in6pcb *in6p = sotoin6pcb(so);
1267 struct ip_moptions **mopts;
1268 int error, optval;
1269 int level, optname;
1270
1271 KASSERT(sopt != NULL);
1272
1273 level = sopt->sopt_level;
1274 optname = sopt->sopt_name;
1275
1276 error = optval = 0;
1277 uproto = (int)so->so_proto->pr_protocol;
1278
1279 switch (level) {
1280 case IPPROTO_IP:
1281 switch (optname) {
1282 case IP_ADD_MEMBERSHIP:
1283 case IP_DROP_MEMBERSHIP:
1284 case IP_MULTICAST_IF:
1285 case IP_MULTICAST_LOOP:
1286 case IP_MULTICAST_TTL:
1287 mopts = &in6p->in6p_v4moptions;
1288 switch (op) {
1289 case PRCO_GETOPT:
1290 return ip_getmoptions(*mopts, sopt);
1291 case PRCO_SETOPT:
1292 return ip_setmoptions(mopts, sopt);
1293 default:
1294 return EINVAL;
1295 }
1296 default:
1297 return ENOPROTOOPT;
1298 }
1299 case IPPROTO_IPV6:
1300 break;
1301 default:
1302 return ENOPROTOOPT;
1303 }
1304 switch (op) {
1305 case PRCO_SETOPT:
1306 switch (optname) {
1307 #ifdef RFC2292
1308 case IPV6_2292PKTOPTIONS:
1309 error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1310 break;
1311 #endif
1312
1313 /*
1314 * Use of some Hop-by-Hop options or some
1315 * Destination options, might require special
1316 * privilege. That is, normal applications
1317 * (without special privilege) might be forbidden
1318 * from setting certain options in outgoing packets,
1319 * and might never see certain options in received
1320 * packets. [RFC 2292 Section 6]
1321 * KAME specific note:
1322 * KAME prevents non-privileged users from sending or
1323 * receiving ANY hbh/dst options in order to avoid
1324 * overhead of parsing options in the kernel.
1325 */
1326 case IPV6_RECVHOPOPTS:
1327 case IPV6_RECVDSTOPTS:
1328 case IPV6_RECVRTHDRDSTOPTS:
1329 error = kauth_authorize_network(kauth_cred_get(),
1330 KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1331 NULL, NULL, NULL);
1332 if (error)
1333 break;
1334 /* FALLTHROUGH */
1335 case IPV6_UNICAST_HOPS:
1336 case IPV6_HOPLIMIT:
1337 case IPV6_FAITH:
1338
1339 case IPV6_RECVPKTINFO:
1340 case IPV6_RECVHOPLIMIT:
1341 case IPV6_RECVRTHDR:
1342 case IPV6_RECVPATHMTU:
1343 case IPV6_RECVTCLASS:
1344 case IPV6_V6ONLY:
1345 error = sockopt_getint(sopt, &optval);
1346 if (error)
1347 break;
1348 switch (optname) {
1349 case IPV6_UNICAST_HOPS:
1350 if (optval < -1 || optval >= 256)
1351 error = EINVAL;
1352 else {
1353 /* -1 = kernel default */
1354 in6p->in6p_hops = optval;
1355 }
1356 break;
1357 #define OPTSET(bit) \
1358 do { \
1359 if (optval) \
1360 in6p->in6p_flags |= (bit); \
1361 else \
1362 in6p->in6p_flags &= ~(bit); \
1363 } while (/*CONSTCOND*/ 0)
1364
1365 #ifdef RFC2292
1366 #define OPTSET2292(bit) \
1367 do { \
1368 in6p->in6p_flags |= IN6P_RFC2292; \
1369 if (optval) \
1370 in6p->in6p_flags |= (bit); \
1371 else \
1372 in6p->in6p_flags &= ~(bit); \
1373 } while (/*CONSTCOND*/ 0)
1374 #endif
1375
1376 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1377
1378 case IPV6_RECVPKTINFO:
1379 #ifdef RFC2292
1380 /* cannot mix with RFC2292 */
1381 if (OPTBIT(IN6P_RFC2292)) {
1382 error = EINVAL;
1383 break;
1384 }
1385 #endif
1386 OPTSET(IN6P_PKTINFO);
1387 break;
1388
1389 case IPV6_HOPLIMIT:
1390 {
1391 struct ip6_pktopts **optp;
1392
1393 #ifdef RFC2292
1394 /* cannot mix with RFC2292 */
1395 if (OPTBIT(IN6P_RFC2292)) {
1396 error = EINVAL;
1397 break;
1398 }
1399 #endif
1400 optp = &in6p->in6p_outputopts;
1401 error = ip6_pcbopt(IPV6_HOPLIMIT,
1402 (u_char *)&optval,
1403 sizeof(optval),
1404 optp,
1405 kauth_cred_get(), uproto);
1406 break;
1407 }
1408
1409 case IPV6_RECVHOPLIMIT:
1410 #ifdef RFC2292
1411 /* cannot mix with RFC2292 */
1412 if (OPTBIT(IN6P_RFC2292)) {
1413 error = EINVAL;
1414 break;
1415 }
1416 #endif
1417 OPTSET(IN6P_HOPLIMIT);
1418 break;
1419
1420 case IPV6_RECVHOPOPTS:
1421 #ifdef RFC2292
1422 /* cannot mix with RFC2292 */
1423 if (OPTBIT(IN6P_RFC2292)) {
1424 error = EINVAL;
1425 break;
1426 }
1427 #endif
1428 OPTSET(IN6P_HOPOPTS);
1429 break;
1430
1431 case IPV6_RECVDSTOPTS:
1432 #ifdef RFC2292
1433 /* cannot mix with RFC2292 */
1434 if (OPTBIT(IN6P_RFC2292)) {
1435 error = EINVAL;
1436 break;
1437 }
1438 #endif
1439 OPTSET(IN6P_DSTOPTS);
1440 break;
1441
1442 case IPV6_RECVRTHDRDSTOPTS:
1443 #ifdef RFC2292
1444 /* cannot mix with RFC2292 */
1445 if (OPTBIT(IN6P_RFC2292)) {
1446 error = EINVAL;
1447 break;
1448 }
1449 #endif
1450 OPTSET(IN6P_RTHDRDSTOPTS);
1451 break;
1452
1453 case IPV6_RECVRTHDR:
1454 #ifdef RFC2292
1455 /* cannot mix with RFC2292 */
1456 if (OPTBIT(IN6P_RFC2292)) {
1457 error = EINVAL;
1458 break;
1459 }
1460 #endif
1461 OPTSET(IN6P_RTHDR);
1462 break;
1463
1464 case IPV6_FAITH:
1465 OPTSET(IN6P_FAITH);
1466 break;
1467
1468 case IPV6_RECVPATHMTU:
1469 /*
1470 * We ignore this option for TCP
1471 * sockets.
1472 * (RFC3542 leaves this case
1473 * unspecified.)
1474 */
1475 if (uproto != IPPROTO_TCP)
1476 OPTSET(IN6P_MTU);
1477 break;
1478
1479 case IPV6_V6ONLY:
1480 /*
1481 * make setsockopt(IPV6_V6ONLY)
1482 * available only prior to bind(2).
1483 * see ipng mailing list, Jun 22 2001.
1484 */
1485 if (in6p->in6p_lport ||
1486 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1487 error = EINVAL;
1488 break;
1489 }
1490 #ifdef INET6_BINDV6ONLY
1491 if (!optval)
1492 error = EINVAL;
1493 #else
1494 OPTSET(IN6P_IPV6_V6ONLY);
1495 #endif
1496 break;
1497 case IPV6_RECVTCLASS:
1498 #ifdef RFC2292
1499 /* cannot mix with RFC2292 XXX */
1500 if (OPTBIT(IN6P_RFC2292)) {
1501 error = EINVAL;
1502 break;
1503 }
1504 #endif
1505 OPTSET(IN6P_TCLASS);
1506 break;
1507
1508 }
1509 break;
1510
1511 case IPV6_OTCLASS:
1512 {
1513 struct ip6_pktopts **optp;
1514 u_int8_t tclass;
1515
1516 error = sockopt_get(sopt, &tclass, sizeof(tclass));
1517 if (error)
1518 break;
1519 optp = &in6p->in6p_outputopts;
1520 error = ip6_pcbopt(optname,
1521 (u_char *)&tclass,
1522 sizeof(tclass),
1523 optp,
1524 kauth_cred_get(), uproto);
1525 break;
1526 }
1527
1528 case IPV6_TCLASS:
1529 case IPV6_DONTFRAG:
1530 case IPV6_USE_MIN_MTU:
1531 case IPV6_PREFER_TEMPADDR:
1532 error = sockopt_getint(sopt, &optval);
1533 if (error)
1534 break;
1535 {
1536 struct ip6_pktopts **optp;
1537 optp = &in6p->in6p_outputopts;
1538 error = ip6_pcbopt(optname,
1539 (u_char *)&optval,
1540 sizeof(optval),
1541 optp,
1542 kauth_cred_get(), uproto);
1543 break;
1544 }
1545
1546 #ifdef RFC2292
1547 case IPV6_2292PKTINFO:
1548 case IPV6_2292HOPLIMIT:
1549 case IPV6_2292HOPOPTS:
1550 case IPV6_2292DSTOPTS:
1551 case IPV6_2292RTHDR:
1552 /* RFC 2292 */
1553 error = sockopt_getint(sopt, &optval);
1554 if (error)
1555 break;
1556
1557 switch (optname) {
1558 case IPV6_2292PKTINFO:
1559 OPTSET2292(IN6P_PKTINFO);
1560 break;
1561 case IPV6_2292HOPLIMIT:
1562 OPTSET2292(IN6P_HOPLIMIT);
1563 break;
1564 case IPV6_2292HOPOPTS:
1565 /*
1566 * Check super-user privilege.
1567 * See comments for IPV6_RECVHOPOPTS.
1568 */
1569 error =
1570 kauth_authorize_network(kauth_cred_get(),
1571 KAUTH_NETWORK_IPV6,
1572 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1573 NULL, NULL);
1574 if (error)
1575 return (error);
1576 OPTSET2292(IN6P_HOPOPTS);
1577 break;
1578 case IPV6_2292DSTOPTS:
1579 error =
1580 kauth_authorize_network(kauth_cred_get(),
1581 KAUTH_NETWORK_IPV6,
1582 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1583 NULL, NULL);
1584 if (error)
1585 return (error);
1586 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1587 break;
1588 case IPV6_2292RTHDR:
1589 OPTSET2292(IN6P_RTHDR);
1590 break;
1591 }
1592 break;
1593 #endif
1594 case IPV6_PKTINFO:
1595 case IPV6_HOPOPTS:
1596 case IPV6_RTHDR:
1597 case IPV6_DSTOPTS:
1598 case IPV6_RTHDRDSTOPTS:
1599 case IPV6_NEXTHOP: {
1600 /* new advanced API (RFC3542) */
1601 void *optbuf;
1602 int optbuflen;
1603 struct ip6_pktopts **optp;
1604
1605 #ifdef RFC2292
1606 /* cannot mix with RFC2292 */
1607 if (OPTBIT(IN6P_RFC2292)) {
1608 error = EINVAL;
1609 break;
1610 }
1611 #endif
1612
1613 optbuflen = sopt->sopt_size;
1614 optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1615 if (optbuf == NULL) {
1616 error = ENOBUFS;
1617 break;
1618 }
1619
1620 error = sockopt_get(sopt, optbuf, optbuflen);
1621 if (error) {
1622 free(optbuf, M_IP6OPT);
1623 break;
1624 }
1625 optp = &in6p->in6p_outputopts;
1626 error = ip6_pcbopt(optname, optbuf, optbuflen,
1627 optp, kauth_cred_get(), uproto);
1628
1629 free(optbuf, M_IP6OPT);
1630 break;
1631 }
1632 #undef OPTSET
1633
1634 case IPV6_MULTICAST_IF:
1635 case IPV6_MULTICAST_HOPS:
1636 case IPV6_MULTICAST_LOOP:
1637 case IPV6_JOIN_GROUP:
1638 case IPV6_LEAVE_GROUP:
1639 error = ip6_setmoptions(sopt, in6p);
1640 break;
1641
1642 case IPV6_PORTRANGE:
1643 error = sockopt_getint(sopt, &optval);
1644 if (error)
1645 break;
1646
1647 switch (optval) {
1648 case IPV6_PORTRANGE_DEFAULT:
1649 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1650 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1651 break;
1652
1653 case IPV6_PORTRANGE_HIGH:
1654 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1655 in6p->in6p_flags |= IN6P_HIGHPORT;
1656 break;
1657
1658 case IPV6_PORTRANGE_LOW:
1659 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1660 in6p->in6p_flags |= IN6P_LOWPORT;
1661 break;
1662
1663 default:
1664 error = EINVAL;
1665 break;
1666 }
1667 break;
1668
1669 case IPV6_PORTALGO:
1670 error = sockopt_getint(sopt, &optval);
1671 if (error)
1672 break;
1673
1674 error = portalgo_algo_index_select(
1675 (struct inpcb_hdr *)in6p, optval);
1676 break;
1677
1678 #if defined(IPSEC)
1679 case IPV6_IPSEC_POLICY:
1680 if (ipsec_enabled) {
1681 error = ipsec6_set_policy(in6p, optname,
1682 sopt->sopt_data, sopt->sopt_size,
1683 kauth_cred_get());
1684 break;
1685 }
1686 /*FALLTHROUGH*/
1687 #endif /* IPSEC */
1688
1689 default:
1690 error = ENOPROTOOPT;
1691 break;
1692 }
1693 break;
1694
1695 case PRCO_GETOPT:
1696 switch (optname) {
1697 #ifdef RFC2292
1698 case IPV6_2292PKTOPTIONS:
1699 /*
1700 * RFC3542 (effectively) deprecated the
1701 * semantics of the 2292-style pktoptions.
1702 * Since it was not reliable in nature (i.e.,
1703 * applications had to expect the lack of some
1704 * information after all), it would make sense
1705 * to simplify this part by always returning
1706 * empty data.
1707 */
1708 break;
1709 #endif
1710
1711 case IPV6_RECVHOPOPTS:
1712 case IPV6_RECVDSTOPTS:
1713 case IPV6_RECVRTHDRDSTOPTS:
1714 case IPV6_UNICAST_HOPS:
1715 case IPV6_RECVPKTINFO:
1716 case IPV6_RECVHOPLIMIT:
1717 case IPV6_RECVRTHDR:
1718 case IPV6_RECVPATHMTU:
1719
1720 case IPV6_FAITH:
1721 case IPV6_V6ONLY:
1722 case IPV6_PORTRANGE:
1723 case IPV6_RECVTCLASS:
1724 switch (optname) {
1725
1726 case IPV6_RECVHOPOPTS:
1727 optval = OPTBIT(IN6P_HOPOPTS);
1728 break;
1729
1730 case IPV6_RECVDSTOPTS:
1731 optval = OPTBIT(IN6P_DSTOPTS);
1732 break;
1733
1734 case IPV6_RECVRTHDRDSTOPTS:
1735 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1736 break;
1737
1738 case IPV6_UNICAST_HOPS:
1739 optval = in6p->in6p_hops;
1740 break;
1741
1742 case IPV6_RECVPKTINFO:
1743 optval = OPTBIT(IN6P_PKTINFO);
1744 break;
1745
1746 case IPV6_RECVHOPLIMIT:
1747 optval = OPTBIT(IN6P_HOPLIMIT);
1748 break;
1749
1750 case IPV6_RECVRTHDR:
1751 optval = OPTBIT(IN6P_RTHDR);
1752 break;
1753
1754 case IPV6_RECVPATHMTU:
1755 optval = OPTBIT(IN6P_MTU);
1756 break;
1757
1758 case IPV6_FAITH:
1759 optval = OPTBIT(IN6P_FAITH);
1760 break;
1761
1762 case IPV6_V6ONLY:
1763 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1764 break;
1765
1766 case IPV6_PORTRANGE:
1767 {
1768 int flags;
1769 flags = in6p->in6p_flags;
1770 if (flags & IN6P_HIGHPORT)
1771 optval = IPV6_PORTRANGE_HIGH;
1772 else if (flags & IN6P_LOWPORT)
1773 optval = IPV6_PORTRANGE_LOW;
1774 else
1775 optval = 0;
1776 break;
1777 }
1778 case IPV6_RECVTCLASS:
1779 optval = OPTBIT(IN6P_TCLASS);
1780 break;
1781
1782 }
1783 if (error)
1784 break;
1785 error = sockopt_setint(sopt, optval);
1786 break;
1787
1788 case IPV6_PATHMTU:
1789 {
1790 u_long pmtu = 0;
1791 struct ip6_mtuinfo mtuinfo;
1792 struct route *ro = &in6p->in6p_route;
1793
1794 if (!(so->so_state & SS_ISCONNECTED))
1795 return (ENOTCONN);
1796 /*
1797 * XXX: we dot not consider the case of source
1798 * routing, or optional information to specify
1799 * the outgoing interface.
1800 */
1801 error = ip6_getpmtu(ro, NULL, NULL,
1802 &in6p->in6p_faddr, &pmtu, NULL);
1803 if (error)
1804 break;
1805 if (pmtu > IPV6_MAXPACKET)
1806 pmtu = IPV6_MAXPACKET;
1807
1808 memset(&mtuinfo, 0, sizeof(mtuinfo));
1809 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1810 optdata = (void *)&mtuinfo;
1811 optdatalen = sizeof(mtuinfo);
1812 if (optdatalen > MCLBYTES)
1813 return (EMSGSIZE); /* XXX */
1814 error = sockopt_set(sopt, optdata, optdatalen);
1815 break;
1816 }
1817
1818 #ifdef RFC2292
1819 case IPV6_2292PKTINFO:
1820 case IPV6_2292HOPLIMIT:
1821 case IPV6_2292HOPOPTS:
1822 case IPV6_2292RTHDR:
1823 case IPV6_2292DSTOPTS:
1824 switch (optname) {
1825 case IPV6_2292PKTINFO:
1826 optval = OPTBIT(IN6P_PKTINFO);
1827 break;
1828 case IPV6_2292HOPLIMIT:
1829 optval = OPTBIT(IN6P_HOPLIMIT);
1830 break;
1831 case IPV6_2292HOPOPTS:
1832 optval = OPTBIT(IN6P_HOPOPTS);
1833 break;
1834 case IPV6_2292RTHDR:
1835 optval = OPTBIT(IN6P_RTHDR);
1836 break;
1837 case IPV6_2292DSTOPTS:
1838 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1839 break;
1840 }
1841 error = sockopt_setint(sopt, optval);
1842 break;
1843 #endif
1844 case IPV6_PKTINFO:
1845 case IPV6_HOPOPTS:
1846 case IPV6_RTHDR:
1847 case IPV6_DSTOPTS:
1848 case IPV6_RTHDRDSTOPTS:
1849 case IPV6_NEXTHOP:
1850 case IPV6_OTCLASS:
1851 case IPV6_TCLASS:
1852 case IPV6_DONTFRAG:
1853 case IPV6_USE_MIN_MTU:
1854 case IPV6_PREFER_TEMPADDR:
1855 error = ip6_getpcbopt(in6p->in6p_outputopts,
1856 optname, sopt);
1857 break;
1858
1859 case IPV6_MULTICAST_IF:
1860 case IPV6_MULTICAST_HOPS:
1861 case IPV6_MULTICAST_LOOP:
1862 case IPV6_JOIN_GROUP:
1863 case IPV6_LEAVE_GROUP:
1864 error = ip6_getmoptions(sopt, in6p);
1865 break;
1866
1867 case IPV6_PORTALGO:
1868 optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1869 error = sockopt_setint(sopt, optval);
1870 break;
1871
1872 #if defined(IPSEC)
1873 case IPV6_IPSEC_POLICY:
1874 if (ipsec_used) {
1875 struct mbuf *m = NULL;
1876
1877 /*
1878 * XXX: this will return EINVAL as sopt is
1879 * empty
1880 */
1881 error = ipsec6_get_policy(in6p, sopt->sopt_data,
1882 sopt->sopt_size, &m);
1883 if (!error)
1884 error = sockopt_setmbuf(sopt, m);
1885 break;
1886 }
1887 /*FALLTHROUGH*/
1888 #endif /* IPSEC */
1889
1890 default:
1891 error = ENOPROTOOPT;
1892 break;
1893 }
1894 break;
1895 }
1896 return (error);
1897 }
1898
1899 int
1900 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1901 {
1902 int error = 0, optval;
1903 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1904 struct in6pcb *in6p = sotoin6pcb(so);
1905 int level, optname;
1906
1907 KASSERT(sopt != NULL);
1908
1909 level = sopt->sopt_level;
1910 optname = sopt->sopt_name;
1911
1912 if (level != IPPROTO_IPV6) {
1913 return ENOPROTOOPT;
1914 }
1915
1916 switch (optname) {
1917 case IPV6_CHECKSUM:
1918 /*
1919 * For ICMPv6 sockets, no modification allowed for checksum
1920 * offset, permit "no change" values to help existing apps.
1921 *
1922 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1923 * for an ICMPv6 socket will fail." The current
1924 * behavior does not meet RFC3542.
1925 */
1926 switch (op) {
1927 case PRCO_SETOPT:
1928 error = sockopt_getint(sopt, &optval);
1929 if (error)
1930 break;
1931 if ((optval % 2) != 0) {
1932 /* the API assumes even offset values */
1933 error = EINVAL;
1934 } else if (so->so_proto->pr_protocol ==
1935 IPPROTO_ICMPV6) {
1936 if (optval != icmp6off)
1937 error = EINVAL;
1938 } else
1939 in6p->in6p_cksum = optval;
1940 break;
1941
1942 case PRCO_GETOPT:
1943 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1944 optval = icmp6off;
1945 else
1946 optval = in6p->in6p_cksum;
1947
1948 error = sockopt_setint(sopt, optval);
1949 break;
1950
1951 default:
1952 error = EINVAL;
1953 break;
1954 }
1955 break;
1956
1957 default:
1958 error = ENOPROTOOPT;
1959 break;
1960 }
1961
1962 return (error);
1963 }
1964
1965 #ifdef RFC2292
1966 /*
1967 * Set up IP6 options in pcb for insertion in output packets or
1968 * specifying behavior of outgoing packets.
1969 */
1970 static int
1971 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1972 struct sockopt *sopt)
1973 {
1974 struct ip6_pktopts *opt = *pktopt;
1975 struct mbuf *m;
1976 int error = 0;
1977
1978 /* turn off any old options. */
1979 if (opt) {
1980 #ifdef DIAGNOSTIC
1981 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1982 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1983 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1984 printf("ip6_pcbopts: all specified options are cleared.\n");
1985 #endif
1986 ip6_clearpktopts(opt, -1);
1987 } else {
1988 opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1989 if (opt == NULL)
1990 return (ENOBUFS);
1991 }
1992 *pktopt = NULL;
1993
1994 if (sopt == NULL || sopt->sopt_size == 0) {
1995 /*
1996 * Only turning off any previous options, regardless of
1997 * whether the opt is just created or given.
1998 */
1999 free(opt, M_IP6OPT);
2000 return (0);
2001 }
2002
2003 /* set options specified by user. */
2004 m = sockopt_getmbuf(sopt);
2005 if (m == NULL) {
2006 free(opt, M_IP6OPT);
2007 return (ENOBUFS);
2008 }
2009
2010 error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2011 so->so_proto->pr_protocol);
2012 m_freem(m);
2013 if (error != 0) {
2014 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2015 free(opt, M_IP6OPT);
2016 return (error);
2017 }
2018 *pktopt = opt;
2019 return (0);
2020 }
2021 #endif
2022
2023 /*
2024 * initialize ip6_pktopts. beware that there are non-zero default values in
2025 * the struct.
2026 */
2027 void
2028 ip6_initpktopts(struct ip6_pktopts *opt)
2029 {
2030
2031 memset(opt, 0, sizeof(*opt));
2032 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2033 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2034 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2035 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2036 }
2037
2038 #define sin6tosa(sin6) ((struct sockaddr *)(sin6)) /* XXX */
2039 static int
2040 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2041 kauth_cred_t cred, int uproto)
2042 {
2043 struct ip6_pktopts *opt;
2044
2045 if (*pktopt == NULL) {
2046 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2047 M_NOWAIT);
2048 if (*pktopt == NULL)
2049 return (ENOBUFS);
2050
2051 ip6_initpktopts(*pktopt);
2052 }
2053 opt = *pktopt;
2054
2055 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2056 }
2057
2058 static int
2059 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2060 {
2061 void *optdata = NULL;
2062 int optdatalen = 0;
2063 struct ip6_ext *ip6e;
2064 int error = 0;
2065 struct in6_pktinfo null_pktinfo;
2066 int deftclass = 0, on;
2067 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2068 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2069
2070 switch (optname) {
2071 case IPV6_PKTINFO:
2072 if (pktopt && pktopt->ip6po_pktinfo)
2073 optdata = (void *)pktopt->ip6po_pktinfo;
2074 else {
2075 /* XXX: we don't have to do this every time... */
2076 memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2077 optdata = (void *)&null_pktinfo;
2078 }
2079 optdatalen = sizeof(struct in6_pktinfo);
2080 break;
2081 case IPV6_OTCLASS:
2082 /* XXX */
2083 return (EINVAL);
2084 case IPV6_TCLASS:
2085 if (pktopt && pktopt->ip6po_tclass >= 0)
2086 optdata = (void *)&pktopt->ip6po_tclass;
2087 else
2088 optdata = (void *)&deftclass;
2089 optdatalen = sizeof(int);
2090 break;
2091 case IPV6_HOPOPTS:
2092 if (pktopt && pktopt->ip6po_hbh) {
2093 optdata = (void *)pktopt->ip6po_hbh;
2094 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2095 optdatalen = (ip6e->ip6e_len + 1) << 3;
2096 }
2097 break;
2098 case IPV6_RTHDR:
2099 if (pktopt && pktopt->ip6po_rthdr) {
2100 optdata = (void *)pktopt->ip6po_rthdr;
2101 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2102 optdatalen = (ip6e->ip6e_len + 1) << 3;
2103 }
2104 break;
2105 case IPV6_RTHDRDSTOPTS:
2106 if (pktopt && pktopt->ip6po_dest1) {
2107 optdata = (void *)pktopt->ip6po_dest1;
2108 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2109 optdatalen = (ip6e->ip6e_len + 1) << 3;
2110 }
2111 break;
2112 case IPV6_DSTOPTS:
2113 if (pktopt && pktopt->ip6po_dest2) {
2114 optdata = (void *)pktopt->ip6po_dest2;
2115 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2116 optdatalen = (ip6e->ip6e_len + 1) << 3;
2117 }
2118 break;
2119 case IPV6_NEXTHOP:
2120 if (pktopt && pktopt->ip6po_nexthop) {
2121 optdata = (void *)pktopt->ip6po_nexthop;
2122 optdatalen = pktopt->ip6po_nexthop->sa_len;
2123 }
2124 break;
2125 case IPV6_USE_MIN_MTU:
2126 if (pktopt)
2127 optdata = (void *)&pktopt->ip6po_minmtu;
2128 else
2129 optdata = (void *)&defminmtu;
2130 optdatalen = sizeof(int);
2131 break;
2132 case IPV6_DONTFRAG:
2133 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2134 on = 1;
2135 else
2136 on = 0;
2137 optdata = (void *)&on;
2138 optdatalen = sizeof(on);
2139 break;
2140 case IPV6_PREFER_TEMPADDR:
2141 if (pktopt)
2142 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2143 else
2144 optdata = (void *)&defpreftemp;
2145 optdatalen = sizeof(int);
2146 break;
2147 default: /* should not happen */
2148 #ifdef DIAGNOSTIC
2149 panic("ip6_getpcbopt: unexpected option\n");
2150 #endif
2151 return (ENOPROTOOPT);
2152 }
2153
2154 error = sockopt_set(sopt, optdata, optdatalen);
2155
2156 return (error);
2157 }
2158
2159 void
2160 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2161 {
2162 if (optname == -1 || optname == IPV6_PKTINFO) {
2163 if (pktopt->ip6po_pktinfo)
2164 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2165 pktopt->ip6po_pktinfo = NULL;
2166 }
2167 if (optname == -1 || optname == IPV6_HOPLIMIT)
2168 pktopt->ip6po_hlim = -1;
2169 if (optname == -1 || optname == IPV6_TCLASS)
2170 pktopt->ip6po_tclass = -1;
2171 if (optname == -1 || optname == IPV6_NEXTHOP) {
2172 rtcache_free(&pktopt->ip6po_nextroute);
2173 if (pktopt->ip6po_nexthop)
2174 free(pktopt->ip6po_nexthop, M_IP6OPT);
2175 pktopt->ip6po_nexthop = NULL;
2176 }
2177 if (optname == -1 || optname == IPV6_HOPOPTS) {
2178 if (pktopt->ip6po_hbh)
2179 free(pktopt->ip6po_hbh, M_IP6OPT);
2180 pktopt->ip6po_hbh = NULL;
2181 }
2182 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2183 if (pktopt->ip6po_dest1)
2184 free(pktopt->ip6po_dest1, M_IP6OPT);
2185 pktopt->ip6po_dest1 = NULL;
2186 }
2187 if (optname == -1 || optname == IPV6_RTHDR) {
2188 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2189 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2190 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2191 rtcache_free(&pktopt->ip6po_route);
2192 }
2193 if (optname == -1 || optname == IPV6_DSTOPTS) {
2194 if (pktopt->ip6po_dest2)
2195 free(pktopt->ip6po_dest2, M_IP6OPT);
2196 pktopt->ip6po_dest2 = NULL;
2197 }
2198 }
2199
2200 #define PKTOPT_EXTHDRCPY(type) \
2201 do { \
2202 if (src->type) { \
2203 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2204 dst->type = malloc(hlen, M_IP6OPT, canwait); \
2205 if (dst->type == NULL) \
2206 goto bad; \
2207 memcpy(dst->type, src->type, hlen); \
2208 } \
2209 } while (/*CONSTCOND*/ 0)
2210
2211 static int
2212 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2213 {
2214 dst->ip6po_hlim = src->ip6po_hlim;
2215 dst->ip6po_tclass = src->ip6po_tclass;
2216 dst->ip6po_flags = src->ip6po_flags;
2217 dst->ip6po_minmtu = src->ip6po_minmtu;
2218 dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2219 if (src->ip6po_pktinfo) {
2220 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2221 M_IP6OPT, canwait);
2222 if (dst->ip6po_pktinfo == NULL)
2223 goto bad;
2224 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2225 }
2226 if (src->ip6po_nexthop) {
2227 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2228 M_IP6OPT, canwait);
2229 if (dst->ip6po_nexthop == NULL)
2230 goto bad;
2231 memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2232 src->ip6po_nexthop->sa_len);
2233 }
2234 PKTOPT_EXTHDRCPY(ip6po_hbh);
2235 PKTOPT_EXTHDRCPY(ip6po_dest1);
2236 PKTOPT_EXTHDRCPY(ip6po_dest2);
2237 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2238 return (0);
2239
2240 bad:
2241 if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2242 if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2243 if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2244 if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2245 if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2246 if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2247
2248 return (ENOBUFS);
2249 }
2250 #undef PKTOPT_EXTHDRCPY
2251
2252 struct ip6_pktopts *
2253 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2254 {
2255 int error;
2256 struct ip6_pktopts *dst;
2257
2258 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2259 if (dst == NULL)
2260 return (NULL);
2261 ip6_initpktopts(dst);
2262
2263 if ((error = copypktopts(dst, src, canwait)) != 0) {
2264 free(dst, M_IP6OPT);
2265 return (NULL);
2266 }
2267
2268 return (dst);
2269 }
2270
2271 void
2272 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2273 {
2274 if (pktopt == NULL)
2275 return;
2276
2277 ip6_clearpktopts(pktopt, -1);
2278
2279 free(pktopt, M_IP6OPT);
2280 }
2281
2282 int
2283 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, void *v,
2284 size_t l)
2285 {
2286 struct ipv6_mreq mreq;
2287 int error;
2288 struct in6_addr *ia = &mreq.ipv6mr_multiaddr;
2289 struct in_addr *ia4 = (void *)&ia->s6_addr32[3];
2290 error = sockopt_get(sopt, &mreq, sizeof(mreq));
2291 if (error != 0)
2292 return error;
2293
2294 if (IN6_IS_ADDR_UNSPECIFIED(ia)) {
2295 /*
2296 * We use the unspecified address to specify to accept
2297 * all multicast addresses. Only super user is allowed
2298 * to do this.
2299 */
2300 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6,
2301 KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2302 return EACCES;
2303 } else if (IN6_IS_ADDR_V4MAPPED(ia)) {
2304 // Don't bother if we are not going to use ifp.
2305 if (l == sizeof(*ia)) {
2306 memcpy(v, ia, l);
2307 return 0;
2308 }
2309 } else if (!IN6_IS_ADDR_MULTICAST(ia)) {
2310 return EINVAL;
2311 }
2312
2313 /*
2314 * If no interface was explicitly specified, choose an
2315 * appropriate one according to the given multicast address.
2316 */
2317 if (mreq.ipv6mr_interface == 0) {
2318 struct rtentry *rt;
2319 union {
2320 struct sockaddr dst;
2321 struct sockaddr_in dst4;
2322 struct sockaddr_in6 dst6;
2323 } u;
2324 struct route ro;
2325
2326 /*
2327 * Look up the routing table for the
2328 * address, and choose the outgoing interface.
2329 * XXX: is it a good approach?
2330 */
2331 memset(&ro, 0, sizeof(ro));
2332 if (IN6_IS_ADDR_V4MAPPED(ia))
2333 sockaddr_in_init(&u.dst4, ia4, 0);
2334 else
2335 sockaddr_in6_init(&u.dst6, ia, 0, 0, 0);
2336 rtcache_setdst(&ro, &u.dst);
2337 *ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
2338 rtcache_free(&ro);
2339 } else {
2340 /*
2341 * If the interface is specified, validate it.
2342 */
2343 if ((*ifp = if_byindex(mreq.ipv6mr_interface)) == NULL)
2344 return ENXIO; /* XXX EINVAL? */
2345 }
2346 if (sizeof(*ia) == l)
2347 memcpy(v, ia, l);
2348 else
2349 memcpy(v, ia4, l);
2350 return 0;
2351 }
2352
2353 /*
2354 * Set the IP6 multicast options in response to user setsockopt().
2355 */
2356 static int
2357 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p)
2358 {
2359 int error = 0;
2360 u_int loop, ifindex;
2361 struct ipv6_mreq mreq;
2362 struct in6_addr ia;
2363 struct ifnet *ifp;
2364 struct ip6_moptions *im6o = in6p->in6p_moptions;
2365 struct in6_multi_mship *imm;
2366
2367 if (im6o == NULL) {
2368 /*
2369 * No multicast option buffer attached to the pcb;
2370 * allocate one and initialize to default values.
2371 */
2372 im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2373 if (im6o == NULL)
2374 return (ENOBUFS);
2375 in6p->in6p_moptions = im6o;
2376 im6o->im6o_multicast_ifp = NULL;
2377 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2378 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2379 LIST_INIT(&im6o->im6o_memberships);
2380 }
2381
2382 switch (sopt->sopt_name) {
2383
2384 case IPV6_MULTICAST_IF:
2385 /*
2386 * Select the interface for outgoing multicast packets.
2387 */
2388 error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2389 if (error != 0)
2390 break;
2391
2392 if (ifindex != 0) {
2393 if ((ifp = if_byindex(ifindex)) == NULL) {
2394 error = ENXIO; /* XXX EINVAL? */
2395 break;
2396 }
2397 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2398 error = EADDRNOTAVAIL;
2399 break;
2400 }
2401 } else
2402 ifp = NULL;
2403 im6o->im6o_multicast_ifp = ifp;
2404 break;
2405
2406 case IPV6_MULTICAST_HOPS:
2407 {
2408 /*
2409 * Set the IP6 hoplimit for outgoing multicast packets.
2410 */
2411 int optval;
2412
2413 error = sockopt_getint(sopt, &optval);
2414 if (error != 0)
2415 break;
2416
2417 if (optval < -1 || optval >= 256)
2418 error = EINVAL;
2419 else if (optval == -1)
2420 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2421 else
2422 im6o->im6o_multicast_hlim = optval;
2423 break;
2424 }
2425
2426 case IPV6_MULTICAST_LOOP:
2427 /*
2428 * Set the loopback flag for outgoing multicast packets.
2429 * Must be zero or one.
2430 */
2431 error = sockopt_get(sopt, &loop, sizeof(loop));
2432 if (error != 0)
2433 break;
2434 if (loop > 1) {
2435 error = EINVAL;
2436 break;
2437 }
2438 im6o->im6o_multicast_loop = loop;
2439 break;
2440
2441 case IPV6_JOIN_GROUP:
2442 /*
2443 * Add a multicast group membership.
2444 * Group must be a valid IP6 multicast address.
2445 */
2446 if ((error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia))))
2447 return error;
2448
2449 if (IN6_IS_ADDR_V4MAPPED(&ia)) {
2450 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2451 break;
2452 }
2453 /*
2454 * See if we found an interface, and confirm that it
2455 * supports multicast
2456 */
2457 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2458 error = EADDRNOTAVAIL;
2459 break;
2460 }
2461
2462 if (in6_setscope(&ia, ifp, NULL)) {
2463 error = EADDRNOTAVAIL; /* XXX: should not happen */
2464 break;
2465 }
2466
2467 /*
2468 * See if the membership already exists.
2469 */
2470 for (imm = im6o->im6o_memberships.lh_first;
2471 imm != NULL; imm = imm->i6mm_chain.le_next)
2472 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2473 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2474 &ia))
2475 break;
2476 if (imm != NULL) {
2477 error = EADDRINUSE;
2478 break;
2479 }
2480 /*
2481 * Everything looks good; add a new record to the multicast
2482 * address list for the given interface.
2483 */
2484 imm = in6_joingroup(ifp, &ia, &error, 0);
2485 if (imm == NULL)
2486 break;
2487 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2488 break;
2489
2490 case IPV6_LEAVE_GROUP:
2491 /*
2492 * Drop a multicast group membership.
2493 * Group must be a valid IP6 multicast address.
2494 */
2495 error = sockopt_get(sopt, &mreq, sizeof(mreq));
2496 if (error != 0)
2497 break;
2498
2499 if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) {
2500 error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2501 break;
2502 }
2503 /*
2504 * If an interface address was specified, get a pointer
2505 * to its ifnet structure.
2506 */
2507 if (mreq.ipv6mr_interface != 0) {
2508 if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2509 error = ENXIO; /* XXX EINVAL? */
2510 break;
2511 }
2512 } else
2513 ifp = NULL;
2514
2515 /* Fill in the scope zone ID */
2516 if (ifp) {
2517 if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2518 /* XXX: should not happen */
2519 error = EADDRNOTAVAIL;
2520 break;
2521 }
2522 } else if (mreq.ipv6mr_interface != 0) {
2523 /*
2524 * XXX: This case would happens when the (positive)
2525 * index is in the valid range, but the corresponding
2526 * interface has been detached dynamically. The above
2527 * check probably avoids such case to happen here, but
2528 * we check it explicitly for safety.
2529 */
2530 error = EADDRNOTAVAIL;
2531 break;
2532 } else { /* ipv6mr_interface == 0 */
2533 struct sockaddr_in6 sa6_mc;
2534
2535 /*
2536 * The API spec says as follows:
2537 * If the interface index is specified as 0, the
2538 * system may choose a multicast group membership to
2539 * drop by matching the multicast address only.
2540 * On the other hand, we cannot disambiguate the scope
2541 * zone unless an interface is provided. Thus, we
2542 * check if there's ambiguity with the default scope
2543 * zone as the last resort.
2544 */
2545 sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2546 0, 0, 0);
2547 error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2548 if (error != 0)
2549 break;
2550 mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2551 }
2552
2553 /*
2554 * Find the membership in the membership list.
2555 */
2556 for (imm = im6o->im6o_memberships.lh_first;
2557 imm != NULL; imm = imm->i6mm_chain.le_next) {
2558 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2559 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2560 &mreq.ipv6mr_multiaddr))
2561 break;
2562 }
2563 if (imm == NULL) {
2564 /* Unable to resolve interface */
2565 error = EADDRNOTAVAIL;
2566 break;
2567 }
2568 /*
2569 * Give up the multicast address record to which the
2570 * membership points.
2571 */
2572 LIST_REMOVE(imm, i6mm_chain);
2573 in6_leavegroup(imm);
2574 break;
2575
2576 default:
2577 error = EOPNOTSUPP;
2578 break;
2579 }
2580
2581 /*
2582 * If all options have default values, no need to keep the mbuf.
2583 */
2584 if (im6o->im6o_multicast_ifp == NULL &&
2585 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2586 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2587 im6o->im6o_memberships.lh_first == NULL) {
2588 free(in6p->in6p_moptions, M_IPMOPTS);
2589 in6p->in6p_moptions = NULL;
2590 }
2591
2592 return (error);
2593 }
2594
2595 /*
2596 * Return the IP6 multicast options in response to user getsockopt().
2597 */
2598 static int
2599 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p)
2600 {
2601 u_int optval;
2602 int error;
2603 struct ip6_moptions *im6o = in6p->in6p_moptions;
2604
2605 switch (sopt->sopt_name) {
2606 case IPV6_MULTICAST_IF:
2607 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2608 optval = 0;
2609 else
2610 optval = im6o->im6o_multicast_ifp->if_index;
2611
2612 error = sockopt_set(sopt, &optval, sizeof(optval));
2613 break;
2614
2615 case IPV6_MULTICAST_HOPS:
2616 if (im6o == NULL)
2617 optval = ip6_defmcasthlim;
2618 else
2619 optval = im6o->im6o_multicast_hlim;
2620
2621 error = sockopt_set(sopt, &optval, sizeof(optval));
2622 break;
2623
2624 case IPV6_MULTICAST_LOOP:
2625 if (im6o == NULL)
2626 optval = IPV6_DEFAULT_MULTICAST_LOOP;
2627 else
2628 optval = im6o->im6o_multicast_loop;
2629
2630 error = sockopt_set(sopt, &optval, sizeof(optval));
2631 break;
2632
2633 default:
2634 error = EOPNOTSUPP;
2635 }
2636
2637 return (error);
2638 }
2639
2640 /*
2641 * Discard the IP6 multicast options.
2642 */
2643 void
2644 ip6_freemoptions(struct ip6_moptions *im6o)
2645 {
2646 struct in6_multi_mship *imm;
2647
2648 if (im6o == NULL)
2649 return;
2650
2651 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2652 LIST_REMOVE(imm, i6mm_chain);
2653 in6_leavegroup(imm);
2654 }
2655 free(im6o, M_IPMOPTS);
2656 }
2657
2658 /*
2659 * Set IPv6 outgoing packet options based on advanced API.
2660 */
2661 int
2662 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2663 struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2664 {
2665 struct cmsghdr *cm = 0;
2666
2667 if (control == NULL || opt == NULL)
2668 return (EINVAL);
2669
2670 ip6_initpktopts(opt);
2671 if (stickyopt) {
2672 int error;
2673
2674 /*
2675 * If stickyopt is provided, make a local copy of the options
2676 * for this particular packet, then override them by ancillary
2677 * objects.
2678 * XXX: copypktopts() does not copy the cached route to a next
2679 * hop (if any). This is not very good in terms of efficiency,
2680 * but we can allow this since this option should be rarely
2681 * used.
2682 */
2683 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2684 return (error);
2685 }
2686
2687 /*
2688 * XXX: Currently, we assume all the optional information is stored
2689 * in a single mbuf.
2690 */
2691 if (control->m_next)
2692 return (EINVAL);
2693
2694 /* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2695 for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2696 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2697 int error;
2698
2699 if (control->m_len < CMSG_LEN(0))
2700 return (EINVAL);
2701
2702 cm = mtod(control, struct cmsghdr *);
2703 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2704 return (EINVAL);
2705 if (cm->cmsg_level != IPPROTO_IPV6)
2706 continue;
2707
2708 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2709 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2710 if (error)
2711 return (error);
2712 }
2713
2714 return (0);
2715 }
2716
2717 /*
2718 * Set a particular packet option, as a sticky option or an ancillary data
2719 * item. "len" can be 0 only when it's a sticky option.
2720 * We have 4 cases of combination of "sticky" and "cmsg":
2721 * "sticky=0, cmsg=0": impossible
2722 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2723 * "sticky=1, cmsg=0": RFC3542 socket option
2724 * "sticky=1, cmsg=1": RFC2292 socket option
2725 */
2726 static int
2727 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2728 kauth_cred_t cred, int sticky, int cmsg, int uproto)
2729 {
2730 int minmtupolicy;
2731 int error;
2732
2733 if (!sticky && !cmsg) {
2734 #ifdef DIAGNOSTIC
2735 printf("ip6_setpktopt: impossible case\n");
2736 #endif
2737 return (EINVAL);
2738 }
2739
2740 /*
2741 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2742 * not be specified in the context of RFC3542. Conversely,
2743 * RFC3542 types should not be specified in the context of RFC2292.
2744 */
2745 if (!cmsg) {
2746 switch (optname) {
2747 case IPV6_2292PKTINFO:
2748 case IPV6_2292HOPLIMIT:
2749 case IPV6_2292NEXTHOP:
2750 case IPV6_2292HOPOPTS:
2751 case IPV6_2292DSTOPTS:
2752 case IPV6_2292RTHDR:
2753 case IPV6_2292PKTOPTIONS:
2754 return (ENOPROTOOPT);
2755 }
2756 }
2757 if (sticky && cmsg) {
2758 switch (optname) {
2759 case IPV6_PKTINFO:
2760 case IPV6_HOPLIMIT:
2761 case IPV6_NEXTHOP:
2762 case IPV6_HOPOPTS:
2763 case IPV6_DSTOPTS:
2764 case IPV6_RTHDRDSTOPTS:
2765 case IPV6_RTHDR:
2766 case IPV6_USE_MIN_MTU:
2767 case IPV6_DONTFRAG:
2768 case IPV6_OTCLASS:
2769 case IPV6_TCLASS:
2770 case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */
2771 return (ENOPROTOOPT);
2772 }
2773 }
2774
2775 switch (optname) {
2776 #ifdef RFC2292
2777 case IPV6_2292PKTINFO:
2778 #endif
2779 case IPV6_PKTINFO:
2780 {
2781 struct ifnet *ifp = NULL;
2782 struct in6_pktinfo *pktinfo;
2783
2784 if (len != sizeof(struct in6_pktinfo))
2785 return (EINVAL);
2786
2787 pktinfo = (struct in6_pktinfo *)buf;
2788
2789 /*
2790 * An application can clear any sticky IPV6_PKTINFO option by
2791 * doing a "regular" setsockopt with ipi6_addr being
2792 * in6addr_any and ipi6_ifindex being zero.
2793 * [RFC 3542, Section 6]
2794 */
2795 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2796 pktinfo->ipi6_ifindex == 0 &&
2797 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2798 ip6_clearpktopts(opt, optname);
2799 break;
2800 }
2801
2802 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2803 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2804 return (EINVAL);
2805 }
2806
2807 /* Validate the interface index if specified. */
2808 if (pktinfo->ipi6_ifindex) {
2809 ifp = if_byindex(pktinfo->ipi6_ifindex);
2810 if (ifp == NULL)
2811 return (ENXIO);
2812 }
2813
2814 /*
2815 * We store the address anyway, and let in6_selectsrc()
2816 * validate the specified address. This is because ipi6_addr
2817 * may not have enough information about its scope zone, and
2818 * we may need additional information (such as outgoing
2819 * interface or the scope zone of a destination address) to
2820 * disambiguate the scope.
2821 * XXX: the delay of the validation may confuse the
2822 * application when it is used as a sticky option.
2823 */
2824 if (opt->ip6po_pktinfo == NULL) {
2825 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2826 M_IP6OPT, M_NOWAIT);
2827 if (opt->ip6po_pktinfo == NULL)
2828 return (ENOBUFS);
2829 }
2830 memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2831 break;
2832 }
2833
2834 #ifdef RFC2292
2835 case IPV6_2292HOPLIMIT:
2836 #endif
2837 case IPV6_HOPLIMIT:
2838 {
2839 int *hlimp;
2840
2841 /*
2842 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2843 * to simplify the ordering among hoplimit options.
2844 */
2845 if (optname == IPV6_HOPLIMIT && sticky)
2846 return (ENOPROTOOPT);
2847
2848 if (len != sizeof(int))
2849 return (EINVAL);
2850 hlimp = (int *)buf;
2851 if (*hlimp < -1 || *hlimp > 255)
2852 return (EINVAL);
2853
2854 opt->ip6po_hlim = *hlimp;
2855 break;
2856 }
2857
2858 case IPV6_OTCLASS:
2859 if (len != sizeof(u_int8_t))
2860 return (EINVAL);
2861
2862 opt->ip6po_tclass = *(u_int8_t *)buf;
2863 break;
2864
2865 case IPV6_TCLASS:
2866 {
2867 int tclass;
2868
2869 if (len != sizeof(int))
2870 return (EINVAL);
2871 tclass = *(int *)buf;
2872 if (tclass < -1 || tclass > 255)
2873 return (EINVAL);
2874
2875 opt->ip6po_tclass = tclass;
2876 break;
2877 }
2878
2879 #ifdef RFC2292
2880 case IPV6_2292NEXTHOP:
2881 #endif
2882 case IPV6_NEXTHOP:
2883 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2884 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2885 if (error)
2886 return (error);
2887
2888 if (len == 0) { /* just remove the option */
2889 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2890 break;
2891 }
2892
2893 /* check if cmsg_len is large enough for sa_len */
2894 if (len < sizeof(struct sockaddr) || len < *buf)
2895 return (EINVAL);
2896
2897 switch (((struct sockaddr *)buf)->sa_family) {
2898 case AF_INET6:
2899 {
2900 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2901
2902 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2903 return (EINVAL);
2904
2905 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2906 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2907 return (EINVAL);
2908 }
2909 if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2910 != 0) {
2911 return (error);
2912 }
2913 break;
2914 }
2915 case AF_LINK: /* eventually be supported? */
2916 default:
2917 return (EAFNOSUPPORT);
2918 }
2919
2920 /* turn off the previous option, then set the new option. */
2921 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2922 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2923 if (opt->ip6po_nexthop == NULL)
2924 return (ENOBUFS);
2925 memcpy(opt->ip6po_nexthop, buf, *buf);
2926 break;
2927
2928 #ifdef RFC2292
2929 case IPV6_2292HOPOPTS:
2930 #endif
2931 case IPV6_HOPOPTS:
2932 {
2933 struct ip6_hbh *hbh;
2934 int hbhlen;
2935
2936 /*
2937 * XXX: We don't allow a non-privileged user to set ANY HbH
2938 * options, since per-option restriction has too much
2939 * overhead.
2940 */
2941 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2942 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2943 if (error)
2944 return (error);
2945
2946 if (len == 0) {
2947 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2948 break; /* just remove the option */
2949 }
2950
2951 /* message length validation */
2952 if (len < sizeof(struct ip6_hbh))
2953 return (EINVAL);
2954 hbh = (struct ip6_hbh *)buf;
2955 hbhlen = (hbh->ip6h_len + 1) << 3;
2956 if (len != hbhlen)
2957 return (EINVAL);
2958
2959 /* turn off the previous option, then set the new option. */
2960 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2961 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2962 if (opt->ip6po_hbh == NULL)
2963 return (ENOBUFS);
2964 memcpy(opt->ip6po_hbh, hbh, hbhlen);
2965
2966 break;
2967 }
2968
2969 #ifdef RFC2292
2970 case IPV6_2292DSTOPTS:
2971 #endif
2972 case IPV6_DSTOPTS:
2973 case IPV6_RTHDRDSTOPTS:
2974 {
2975 struct ip6_dest *dest, **newdest = NULL;
2976 int destlen;
2977
2978 /* XXX: see the comment for IPV6_HOPOPTS */
2979 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2980 KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2981 if (error)
2982 return (error);
2983
2984 if (len == 0) {
2985 ip6_clearpktopts(opt, optname);
2986 break; /* just remove the option */
2987 }
2988
2989 /* message length validation */
2990 if (len < sizeof(struct ip6_dest))
2991 return (EINVAL);
2992 dest = (struct ip6_dest *)buf;
2993 destlen = (dest->ip6d_len + 1) << 3;
2994 if (len != destlen)
2995 return (EINVAL);
2996 /*
2997 * Determine the position that the destination options header
2998 * should be inserted; before or after the routing header.
2999 */
3000 switch (optname) {
3001 case IPV6_2292DSTOPTS:
3002 /*
3003 * The old advanced API is ambiguous on this point.
3004 * Our approach is to determine the position based
3005 * according to the existence of a routing header.
3006 * Note, however, that this depends on the order of the
3007 * extension headers in the ancillary data; the 1st
3008 * part of the destination options header must appear
3009 * before the routing header in the ancillary data,
3010 * too.
3011 * RFC3542 solved the ambiguity by introducing
3012 * separate ancillary data or option types.
3013 */
3014 if (opt->ip6po_rthdr == NULL)
3015 newdest = &opt->ip6po_dest1;
3016 else
3017 newdest = &opt->ip6po_dest2;
3018 break;
3019 case IPV6_RTHDRDSTOPTS:
3020 newdest = &opt->ip6po_dest1;
3021 break;
3022 case IPV6_DSTOPTS:
3023 newdest = &opt->ip6po_dest2;
3024 break;
3025 }
3026
3027 /* turn off the previous option, then set the new option. */
3028 ip6_clearpktopts(opt, optname);
3029 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3030 if (*newdest == NULL)
3031 return (ENOBUFS);
3032 memcpy(*newdest, dest, destlen);
3033
3034 break;
3035 }
3036
3037 #ifdef RFC2292
3038 case IPV6_2292RTHDR:
3039 #endif
3040 case IPV6_RTHDR:
3041 {
3042 struct ip6_rthdr *rth;
3043 int rthlen;
3044
3045 if (len == 0) {
3046 ip6_clearpktopts(opt, IPV6_RTHDR);
3047 break; /* just remove the option */
3048 }
3049
3050 /* message length validation */
3051 if (len < sizeof(struct ip6_rthdr))
3052 return (EINVAL);
3053 rth = (struct ip6_rthdr *)buf;
3054 rthlen = (rth->ip6r_len + 1) << 3;
3055 if (len != rthlen)
3056 return (EINVAL);
3057 switch (rth->ip6r_type) {
3058 case IPV6_RTHDR_TYPE_0:
3059 if (rth->ip6r_len == 0) /* must contain one addr */
3060 return (EINVAL);
3061 if (rth->ip6r_len % 2) /* length must be even */
3062 return (EINVAL);
3063 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3064 return (EINVAL);
3065 break;
3066 default:
3067 return (EINVAL); /* not supported */
3068 }
3069 /* turn off the previous option */
3070 ip6_clearpktopts(opt, IPV6_RTHDR);
3071 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3072 if (opt->ip6po_rthdr == NULL)
3073 return (ENOBUFS);
3074 memcpy(opt->ip6po_rthdr, rth, rthlen);
3075 break;
3076 }
3077
3078 case IPV6_USE_MIN_MTU:
3079 if (len != sizeof(int))
3080 return (EINVAL);
3081 minmtupolicy = *(int *)buf;
3082 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3083 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3084 minmtupolicy != IP6PO_MINMTU_ALL) {
3085 return (EINVAL);
3086 }
3087 opt->ip6po_minmtu = minmtupolicy;
3088 break;
3089
3090 case IPV6_DONTFRAG:
3091 if (len != sizeof(int))
3092 return (EINVAL);
3093
3094 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3095 /*
3096 * we ignore this option for TCP sockets.
3097 * (RFC3542 leaves this case unspecified.)
3098 */
3099 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3100 } else
3101 opt->ip6po_flags |= IP6PO_DONTFRAG;
3102 break;
3103
3104 case IPV6_PREFER_TEMPADDR:
3105 {
3106 int preftemp;
3107
3108 if (len != sizeof(int))
3109 return (EINVAL);
3110 preftemp = *(int *)buf;
3111 switch (preftemp) {
3112 case IP6PO_TEMPADDR_SYSTEM:
3113 case IP6PO_TEMPADDR_NOTPREFER:
3114 case IP6PO_TEMPADDR_PREFER:
3115 break;
3116 default:
3117 return (EINVAL);
3118 }
3119 opt->ip6po_prefer_tempaddr = preftemp;
3120 break;
3121 }
3122
3123 default:
3124 return (ENOPROTOOPT);
3125 } /* end of switch */
3126
3127 return (0);
3128 }
3129
3130 /*
3131 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3132 * packet to the input queue of a specified interface. Note that this
3133 * calls the output routine of the loopback "driver", but with an interface
3134 * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3135 */
3136 void
3137 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3138 const struct sockaddr_in6 *dst)
3139 {
3140 struct mbuf *copym;
3141 struct ip6_hdr *ip6;
3142
3143 copym = m_copy(m, 0, M_COPYALL);
3144 if (copym == NULL)
3145 return;
3146
3147 /*
3148 * Make sure to deep-copy IPv6 header portion in case the data
3149 * is in an mbuf cluster, so that we can safely override the IPv6
3150 * header portion later.
3151 */
3152 if ((copym->m_flags & M_EXT) != 0 ||
3153 copym->m_len < sizeof(struct ip6_hdr)) {
3154 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3155 if (copym == NULL)
3156 return;
3157 }
3158
3159 #ifdef DIAGNOSTIC
3160 if (copym->m_len < sizeof(*ip6)) {
3161 m_freem(copym);
3162 return;
3163 }
3164 #endif
3165
3166 ip6 = mtod(copym, struct ip6_hdr *);
3167 /*
3168 * clear embedded scope identifiers if necessary.
3169 * in6_clearscope will touch the addresses only when necessary.
3170 */
3171 in6_clearscope(&ip6->ip6_src);
3172 in6_clearscope(&ip6->ip6_dst);
3173
3174 (void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3175 }
3176
3177 /*
3178 * Chop IPv6 header off from the payload.
3179 */
3180 static int
3181 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3182 {
3183 struct mbuf *mh;
3184 struct ip6_hdr *ip6;
3185
3186 ip6 = mtod(m, struct ip6_hdr *);
3187 if (m->m_len > sizeof(*ip6)) {
3188 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3189 if (mh == 0) {
3190 m_freem(m);
3191 return ENOBUFS;
3192 }
3193 M_MOVE_PKTHDR(mh, m);
3194 MH_ALIGN(mh, sizeof(*ip6));
3195 m->m_len -= sizeof(*ip6);
3196 m->m_data += sizeof(*ip6);
3197 mh->m_next = m;
3198 m = mh;
3199 m->m_len = sizeof(*ip6);
3200 bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3201 }
3202 exthdrs->ip6e_ip6 = m;
3203 return 0;
3204 }
3205
3206 /*
3207 * Compute IPv6 extension header length.
3208 */
3209 int
3210 ip6_optlen(struct in6pcb *in6p)
3211 {
3212 int len;
3213
3214 if (!in6p->in6p_outputopts)
3215 return 0;
3216
3217 len = 0;
3218 #define elen(x) \
3219 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3220
3221 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3222 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3223 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3224 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3225 return len;
3226 #undef elen
3227 }
3228