ip6_output.c revision 1.38 1 /* $NetBSD: ip6_output.c,v 1.38 2001/10/17 08:23:06 itojun 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. All advertising materials mentioning features or use of this software
46 * must display the following acknowledgement:
47 * This product includes software developed by the University of
48 * California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
66 */
67
68 #include "opt_inet.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81
82 #include <net/if.h>
83 #include <net/route.h>
84 #ifdef PFIL_HOOKS
85 #include <net/pfil.h>
86 #endif
87
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet6/in6_pcb.h>
94 #include <netinet6/nd6.h>
95
96 #ifdef IPSEC
97 #include <netinet6/ipsec.h>
98 #include <netkey/key.h>
99 #endif /* IPSEC */
100
101 #include "loop.h"
102
103 #include <net/net_osdep.h>
104
105 #ifdef PFIL_HOOKS
106 extern struct pfil_head inet6_pfil_hook; /* XXX */
107 #endif
108
109 struct ip6_exthdrs {
110 struct mbuf *ip6e_ip6;
111 struct mbuf *ip6e_hbh;
112 struct mbuf *ip6e_dest1;
113 struct mbuf *ip6e_rthdr;
114 struct mbuf *ip6e_dest2;
115 };
116
117 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
118 struct socket *));
119 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
120 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
121 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
122 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
123 struct ip6_frag **));
124 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
125 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
126
127 extern struct ifnet loif[NLOOP];
128
129 /*
130 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
131 * header (with pri, len, nxt, hlim, src, dst).
132 * This function may modify ver and hlim only.
133 * The mbuf chain containing the packet will be freed.
134 * The mbuf opt, if present, will not be freed.
135 */
136 int
137 ip6_output(m0, opt, ro, flags, im6o, ifpp)
138 struct mbuf *m0;
139 struct ip6_pktopts *opt;
140 struct route_in6 *ro;
141 int flags;
142 struct ip6_moptions *im6o;
143 struct ifnet **ifpp; /* XXX: just for statistics */
144 {
145 struct ip6_hdr *ip6, *mhip6;
146 struct ifnet *ifp, *origifp;
147 struct mbuf *m = m0;
148 int hlen, tlen, len, off;
149 struct route_in6 ip6route;
150 struct sockaddr_in6 *dst;
151 int error = 0;
152 struct in6_ifaddr *ia;
153 u_long mtu;
154 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
155 struct ip6_exthdrs exthdrs;
156 struct in6_addr finaldst;
157 struct route_in6 *ro_pmtu = NULL;
158 int hdrsplit = 0;
159 int needipsec = 0;
160 #ifdef IPSEC
161 int needipsectun = 0;
162 struct socket *so;
163 struct secpolicy *sp = NULL;
164
165 /* for AH processing. stupid to have "socket" variable in IP layer... */
166 so = ipsec_getsocket(m);
167 (void)ipsec_setsocket(m, NULL);
168 ip6 = mtod(m, struct ip6_hdr *);
169 #endif /* IPSEC */
170
171 #define MAKE_EXTHDR(hp, mp) \
172 do { \
173 if (hp) { \
174 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
175 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
176 ((eh)->ip6e_len + 1) << 3); \
177 if (error) \
178 goto freehdrs; \
179 } \
180 } while (0)
181
182 bzero(&exthdrs, sizeof(exthdrs));
183 if (opt) {
184 /* Hop-by-Hop options header */
185 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
186 /* Destination options header(1st part) */
187 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
188 /* Routing header */
189 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
190 /* Destination options header(2nd part) */
191 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
192 }
193
194 #ifdef IPSEC
195 /* get a security policy for this packet */
196 if (so == NULL)
197 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
198 else
199 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
200
201 if (sp == NULL) {
202 ipsec6stat.out_inval++;
203 goto freehdrs;
204 }
205
206 error = 0;
207
208 /* check policy */
209 switch (sp->policy) {
210 case IPSEC_POLICY_DISCARD:
211 /*
212 * This packet is just discarded.
213 */
214 ipsec6stat.out_polvio++;
215 goto freehdrs;
216
217 case IPSEC_POLICY_BYPASS:
218 case IPSEC_POLICY_NONE:
219 /* no need to do IPsec. */
220 needipsec = 0;
221 break;
222
223 case IPSEC_POLICY_IPSEC:
224 if (sp->req == NULL) {
225 /* XXX should be panic ? */
226 printf("ip6_output: No IPsec request specified.\n");
227 error = EINVAL;
228 goto freehdrs;
229 }
230 needipsec = 1;
231 break;
232
233 case IPSEC_POLICY_ENTRUST:
234 default:
235 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
236 }
237 #endif /* IPSEC */
238
239 /*
240 * Calculate the total length of the extension header chain.
241 * Keep the length of the unfragmentable part for fragmentation.
242 */
243 optlen = 0;
244 if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
245 if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
246 if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
247 unfragpartlen = optlen + sizeof(struct ip6_hdr);
248 /* NOTE: we don't add AH/ESP length here. do that later. */
249 if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
250
251 /*
252 * If we need IPsec, or there is at least one extension header,
253 * separate IP6 header from the payload.
254 */
255 if ((needipsec || optlen) && !hdrsplit) {
256 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
257 m = NULL;
258 goto freehdrs;
259 }
260 m = exthdrs.ip6e_ip6;
261 hdrsplit++;
262 }
263
264 /* adjust pointer */
265 ip6 = mtod(m, struct ip6_hdr *);
266
267 /* adjust mbuf packet header length */
268 m->m_pkthdr.len += optlen;
269 plen = m->m_pkthdr.len - sizeof(*ip6);
270
271 /* If this is a jumbo payload, insert a jumbo payload option. */
272 if (plen > IPV6_MAXPACKET) {
273 if (!hdrsplit) {
274 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
275 m = NULL;
276 goto freehdrs;
277 }
278 m = exthdrs.ip6e_ip6;
279 hdrsplit++;
280 }
281 /* adjust pointer */
282 ip6 = mtod(m, struct ip6_hdr *);
283 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
284 goto freehdrs;
285 ip6->ip6_plen = 0;
286 } else
287 ip6->ip6_plen = htons(plen);
288
289 /*
290 * Concatenate headers and fill in next header fields.
291 * Here we have, on "m"
292 * IPv6 payload
293 * and we insert headers accordingly. Finally, we should be getting:
294 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
295 *
296 * during the header composing process, "m" points to IPv6 header.
297 * "mprev" points to an extension header prior to esp.
298 */
299 {
300 u_char *nexthdrp = &ip6->ip6_nxt;
301 struct mbuf *mprev = m;
302
303 /*
304 * we treat dest2 specially. this makes IPsec processing
305 * much easier.
306 *
307 * result: IPv6 dest2 payload
308 * m and mprev will point to IPv6 header.
309 */
310 if (exthdrs.ip6e_dest2) {
311 if (!hdrsplit)
312 panic("assumption failed: hdr not split");
313 exthdrs.ip6e_dest2->m_next = m->m_next;
314 m->m_next = exthdrs.ip6e_dest2;
315 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
316 ip6->ip6_nxt = IPPROTO_DSTOPTS;
317 }
318
319 #define MAKE_CHAIN(m, mp, p, i)\
320 do {\
321 if (m) {\
322 if (!hdrsplit) \
323 panic("assumption failed: hdr not split"); \
324 *mtod((m), u_char *) = *(p);\
325 *(p) = (i);\
326 p = mtod((m), u_char *);\
327 (m)->m_next = (mp)->m_next;\
328 (mp)->m_next = (m);\
329 (mp) = (m);\
330 }\
331 } while (0)
332 /*
333 * result: IPv6 hbh dest1 rthdr dest2 payload
334 * m will point to IPv6 header. mprev will point to the
335 * extension header prior to dest2 (rthdr in the above case).
336 */
337 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
338 nexthdrp, IPPROTO_HOPOPTS);
339 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
340 nexthdrp, IPPROTO_DSTOPTS);
341 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
342 nexthdrp, IPPROTO_ROUTING);
343
344 #ifdef IPSEC
345 if (!needipsec)
346 goto skip_ipsec2;
347
348 /*
349 * pointers after IPsec headers are not valid any more.
350 * other pointers need a great care too.
351 * (IPsec routines should not mangle mbufs prior to AH/ESP)
352 */
353 exthdrs.ip6e_dest2 = NULL;
354
355 {
356 struct ip6_rthdr *rh = NULL;
357 int segleft_org = 0;
358 struct ipsec_output_state state;
359
360 if (exthdrs.ip6e_rthdr) {
361 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
362 segleft_org = rh->ip6r_segleft;
363 rh->ip6r_segleft = 0;
364 }
365
366 bzero(&state, sizeof(state));
367 state.m = m;
368 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
369 &needipsectun);
370 m = state.m;
371 if (error) {
372 /* mbuf is already reclaimed in ipsec6_output_trans. */
373 m = NULL;
374 switch (error) {
375 case EHOSTUNREACH:
376 case ENETUNREACH:
377 case EMSGSIZE:
378 case ENOBUFS:
379 case ENOMEM:
380 break;
381 default:
382 printf("ip6_output (ipsec): error code %d\n", error);
383 /*fall through*/
384 case ENOENT:
385 /* don't show these error codes to the user */
386 error = 0;
387 break;
388 }
389 goto bad;
390 }
391 if (exthdrs.ip6e_rthdr) {
392 /* ah6_output doesn't modify mbuf chain */
393 rh->ip6r_segleft = segleft_org;
394 }
395 }
396 skip_ipsec2:;
397 #endif
398 }
399
400 /*
401 * If there is a routing header, replace destination address field
402 * with the first hop of the routing header.
403 */
404 if (exthdrs.ip6e_rthdr) {
405 struct ip6_rthdr *rh =
406 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
407 struct ip6_rthdr *));
408 struct ip6_rthdr0 *rh0;
409
410 finaldst = ip6->ip6_dst;
411 switch (rh->ip6r_type) {
412 case IPV6_RTHDR_TYPE_0:
413 rh0 = (struct ip6_rthdr0 *)rh;
414 ip6->ip6_dst = rh0->ip6r0_addr[0];
415 bcopy((caddr_t)&rh0->ip6r0_addr[1],
416 (caddr_t)&rh0->ip6r0_addr[0],
417 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)
418 );
419 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
420 break;
421 default: /* is it possible? */
422 error = EINVAL;
423 goto bad;
424 }
425 }
426
427 /* Source address validation */
428 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
429 (flags & IPV6_DADOUTPUT) == 0) {
430 error = EOPNOTSUPP;
431 ip6stat.ip6s_badscope++;
432 goto bad;
433 }
434 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
435 error = EOPNOTSUPP;
436 ip6stat.ip6s_badscope++;
437 goto bad;
438 }
439
440 ip6stat.ip6s_localout++;
441
442 /*
443 * Route packet.
444 */
445 if (ro == 0) {
446 ro = &ip6route;
447 bzero((caddr_t)ro, sizeof(*ro));
448 }
449 ro_pmtu = ro;
450 if (opt && opt->ip6po_rthdr)
451 ro = &opt->ip6po_route;
452 dst = (struct sockaddr_in6 *)&ro->ro_dst;
453 /*
454 * If there is a cached route,
455 * check that it is to the same destination
456 * and is still up. If not, free it and try again.
457 */
458 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
459 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
460 RTFREE(ro->ro_rt);
461 ro->ro_rt = (struct rtentry *)0;
462 }
463 if (ro->ro_rt == 0) {
464 bzero(dst, sizeof(*dst));
465 dst->sin6_family = AF_INET6;
466 dst->sin6_len = sizeof(struct sockaddr_in6);
467 dst->sin6_addr = ip6->ip6_dst;
468 }
469 #ifdef IPSEC
470 if (needipsec && needipsectun) {
471 struct ipsec_output_state state;
472
473 /*
474 * All the extension headers will become inaccessible
475 * (since they can be encrypted).
476 * Don't panic, we need no more updates to extension headers
477 * on inner IPv6 packet (since they are now encapsulated).
478 *
479 * IPv6 [ESP|AH] IPv6 [extension headers] payload
480 */
481 bzero(&exthdrs, sizeof(exthdrs));
482 exthdrs.ip6e_ip6 = m;
483
484 bzero(&state, sizeof(state));
485 state.m = m;
486 state.ro = (struct route *)ro;
487 state.dst = (struct sockaddr *)dst;
488
489 error = ipsec6_output_tunnel(&state, sp, flags);
490
491 m = state.m;
492 ro = (struct route_in6 *)state.ro;
493 dst = (struct sockaddr_in6 *)state.dst;
494 if (error) {
495 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
496 m0 = m = NULL;
497 m = NULL;
498 switch (error) {
499 case EHOSTUNREACH:
500 case ENETUNREACH:
501 case EMSGSIZE:
502 case ENOBUFS:
503 case ENOMEM:
504 break;
505 default:
506 printf("ip6_output (ipsec): error code %d\n", error);
507 /*fall through*/
508 case ENOENT:
509 /* don't show these error codes to the user */
510 error = 0;
511 break;
512 }
513 goto bad;
514 }
515
516 exthdrs.ip6e_ip6 = m;
517 }
518 #endif /*IPSEC*/
519
520 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
521 /* Unicast */
522
523 #define ifatoia6(ifa) ((struct in6_ifaddr *)(ifa))
524 #define sin6tosa(sin6) ((struct sockaddr *)(sin6))
525 /* xxx
526 * interface selection comes here
527 * if an interface is specified from an upper layer,
528 * ifp must point it.
529 */
530 if (ro->ro_rt == 0) {
531 /*
532 * non-bsdi always clone routes, if parent is
533 * PRF_CLONING.
534 */
535 rtalloc((struct route *)ro);
536 }
537 if (ro->ro_rt == 0) {
538 ip6stat.ip6s_noroute++;
539 error = EHOSTUNREACH;
540 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
541 goto bad;
542 }
543 ia = ifatoia6(ro->ro_rt->rt_ifa);
544 ifp = ro->ro_rt->rt_ifp;
545 ro->ro_rt->rt_use++;
546 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
547 dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
548 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
549
550 in6_ifstat_inc(ifp, ifs6_out_request);
551
552 /*
553 * Check if the outgoing interface conflicts with
554 * the interface specified by ifi6_ifindex (if specified).
555 * Note that loopback interface is always okay.
556 * (this may happen when we are sending a packet to one of
557 * our own addresses.)
558 */
559 if (opt && opt->ip6po_pktinfo
560 && opt->ip6po_pktinfo->ipi6_ifindex) {
561 if (!(ifp->if_flags & IFF_LOOPBACK)
562 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
563 ip6stat.ip6s_noroute++;
564 in6_ifstat_inc(ifp, ifs6_out_discard);
565 error = EHOSTUNREACH;
566 goto bad;
567 }
568 }
569
570 if (opt && opt->ip6po_hlim != -1)
571 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
572 } else {
573 /* Multicast */
574 struct in6_multi *in6m;
575
576 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
577
578 /*
579 * See if the caller provided any multicast options
580 */
581 ifp = NULL;
582 if (im6o != NULL) {
583 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
584 if (im6o->im6o_multicast_ifp != NULL)
585 ifp = im6o->im6o_multicast_ifp;
586 } else
587 ip6->ip6_hlim = ip6_defmcasthlim;
588
589 /*
590 * See if the caller provided the outgoing interface
591 * as an ancillary data.
592 * Boundary check for ifindex is assumed to be already done.
593 */
594 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
595 ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
596
597 /*
598 * If the destination is a node-local scope multicast,
599 * the packet should be loop-backed only.
600 */
601 if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
602 /*
603 * If the outgoing interface is already specified,
604 * it should be a loopback interface.
605 */
606 if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
607 ip6stat.ip6s_badscope++;
608 error = ENETUNREACH; /* XXX: better error? */
609 /* XXX correct ifp? */
610 in6_ifstat_inc(ifp, ifs6_out_discard);
611 goto bad;
612 } else {
613 ifp = &loif[0];
614 }
615 }
616
617 if (opt && opt->ip6po_hlim != -1)
618 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
619
620 /*
621 * If caller did not provide an interface lookup a
622 * default in the routing table. This is either a
623 * default for the speicfied group (i.e. a host
624 * route), or a multicast default (a route for the
625 * ``net'' ff00::/8).
626 */
627 if (ifp == NULL) {
628 if (ro->ro_rt == 0) {
629 ro->ro_rt = rtalloc1((struct sockaddr *)
630 &ro->ro_dst, 0
631 );
632 }
633 if (ro->ro_rt == 0) {
634 ip6stat.ip6s_noroute++;
635 error = EHOSTUNREACH;
636 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
637 goto bad;
638 }
639 ia = ifatoia6(ro->ro_rt->rt_ifa);
640 ifp = ro->ro_rt->rt_ifp;
641 ro->ro_rt->rt_use++;
642 }
643
644 if ((flags & IPV6_FORWARDING) == 0)
645 in6_ifstat_inc(ifp, ifs6_out_request);
646 in6_ifstat_inc(ifp, ifs6_out_mcast);
647
648 /*
649 * Confirm that the outgoing interface supports multicast.
650 */
651 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
652 ip6stat.ip6s_noroute++;
653 in6_ifstat_inc(ifp, ifs6_out_discard);
654 error = ENETUNREACH;
655 goto bad;
656 }
657 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
658 if (in6m != NULL &&
659 (im6o == NULL || im6o->im6o_multicast_loop)) {
660 /*
661 * If we belong to the destination multicast group
662 * on the outgoing interface, and the caller did not
663 * forbid loopback, loop back a copy.
664 */
665 ip6_mloopback(ifp, m, dst);
666 } else {
667 /*
668 * If we are acting as a multicast router, perform
669 * multicast forwarding as if the packet had just
670 * arrived on the interface to which we are about
671 * to send. The multicast forwarding function
672 * recursively calls this function, using the
673 * IPV6_FORWARDING flag to prevent infinite recursion.
674 *
675 * Multicasts that are looped back by ip6_mloopback(),
676 * above, will be forwarded by the ip6_input() routine,
677 * if necessary.
678 */
679 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
680 if (ip6_mforward(ip6, ifp, m) != 0) {
681 m_freem(m);
682 goto done;
683 }
684 }
685 }
686 /*
687 * Multicasts with a hoplimit of zero may be looped back,
688 * above, but must not be transmitted on a network.
689 * Also, multicasts addressed to the loopback interface
690 * are not sent -- the above call to ip6_mloopback() will
691 * loop back a copy if this host actually belongs to the
692 * destination group on the loopback interface.
693 */
694 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
695 m_freem(m);
696 goto done;
697 }
698 }
699
700 /*
701 * Fill the outgoing inteface to tell the upper layer
702 * to increment per-interface statistics.
703 */
704 if (ifpp)
705 *ifpp = ifp;
706
707 /*
708 * Determine path MTU.
709 */
710 if (ro_pmtu != ro) {
711 /* The first hop and the final destination may differ. */
712 struct sockaddr_in6 *sin6_fin =
713 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
714 if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
715 !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
716 &finaldst))) {
717 RTFREE(ro_pmtu->ro_rt);
718 ro_pmtu->ro_rt = (struct rtentry *)0;
719 }
720 if (ro_pmtu->ro_rt == 0) {
721 bzero(sin6_fin, sizeof(*sin6_fin));
722 sin6_fin->sin6_family = AF_INET6;
723 sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
724 sin6_fin->sin6_addr = finaldst;
725
726 rtalloc((struct route *)ro_pmtu);
727 }
728 }
729 if (ro_pmtu->ro_rt != NULL) {
730 u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
731
732 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
733 if (mtu > ifmtu || mtu == 0) {
734 /*
735 * The MTU on the route is larger than the MTU on
736 * the interface! This shouldn't happen, unless the
737 * MTU of the interface has been changed after the
738 * interface was brought up. Change the MTU in the
739 * route to match the interface MTU (as long as the
740 * field isn't locked).
741 *
742 * if MTU on the route is 0, we need to fix the MTU.
743 * this case happens with path MTU discovery timeouts.
744 */
745 mtu = ifmtu;
746 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
747 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
748 }
749 } else {
750 mtu = nd_ifinfo[ifp->if_index].linkmtu;
751 }
752
753 /* Fake scoped addresses */
754 if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
755 /*
756 * If source or destination address is a scoped address, and
757 * the packet is going to be sent to a loopback interface,
758 * we should keep the original interface.
759 */
760
761 /*
762 * XXX: this is a very experimental and temporary solution.
763 * We eventually have sockaddr_in6 and use the sin6_scope_id
764 * field of the structure here.
765 * We rely on the consistency between two scope zone ids
766 * of source add destination, which should already be assured
767 * Larger scopes than link will be supported in the near
768 * future.
769 */
770 origifp = NULL;
771 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
772 origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
773 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
774 origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
775 /*
776 * XXX: origifp can be NULL even in those two cases above.
777 * For example, if we remove the (only) link-local address
778 * from the loopback interface, and try to send a link-local
779 * address without link-id information. Then the source
780 * address is ::1, and the destination address is the
781 * link-local address with its s6_addr16[1] being zero.
782 * What is worse, if the packet goes to the loopback interface
783 * by a default rejected route, the null pointer would be
784 * passed to looutput, and the kernel would hang.
785 * The following last resort would prevent such disaster.
786 */
787 if (origifp == NULL)
788 origifp = ifp;
789 }
790 else
791 origifp = ifp;
792 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
793 ip6->ip6_src.s6_addr16[1] = 0;
794 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
795 ip6->ip6_dst.s6_addr16[1] = 0;
796
797 /*
798 * If the outgoing packet contains a hop-by-hop options header,
799 * it must be examined and processed even by the source node.
800 * (RFC 2460, section 4.)
801 */
802 if (exthdrs.ip6e_hbh) {
803 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
804 u_int32_t dummy1; /* XXX unused */
805 u_int32_t dummy2; /* XXX unused */
806
807 /*
808 * XXX: if we have to send an ICMPv6 error to the sender,
809 * we need the M_LOOP flag since icmp6_error() expects
810 * the IPv6 and the hop-by-hop options header are
811 * continuous unless the flag is set.
812 */
813 m->m_flags |= M_LOOP;
814 m->m_pkthdr.rcvif = ifp;
815 if (ip6_process_hopopts(m,
816 (u_int8_t *)(hbh + 1),
817 ((hbh->ip6h_len + 1) << 3) -
818 sizeof(struct ip6_hbh),
819 &dummy1, &dummy2) < 0) {
820 /* m was already freed at this point */
821 error = EINVAL;/* better error? */
822 goto done;
823 }
824 m->m_flags &= ~M_LOOP; /* XXX */
825 m->m_pkthdr.rcvif = NULL;
826 }
827
828 #ifdef PFIL_HOOKS
829 /*
830 * Run through list of hooks for output packets.
831 */
832 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp,
833 PFIL_OUT)) != 0)
834 goto done;
835 if (m == NULL)
836 goto done;
837 ip6 = mtod(m, struct ip6_hdr *);
838 #endif /* PFIL_HOOKS */
839 /*
840 * Send the packet to the outgoing interface.
841 * If necessary, do IPv6 fragmentation before sending.
842 */
843 tlen = m->m_pkthdr.len;
844 if (tlen <= mtu
845 #ifdef notyet
846 /*
847 * On any link that cannot convey a 1280-octet packet in one piece,
848 * link-specific fragmentation and reassembly must be provided at
849 * a layer below IPv6. [RFC 2460, sec.5]
850 * Thus if the interface has ability of link-level fragmentation,
851 * we can just send the packet even if the packet size is
852 * larger than the link's MTU.
853 * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
854 */
855
856 || ifp->if_flags & IFF_FRAGMENTABLE
857 #endif
858 )
859 {
860 #ifdef IFA_STATS
861 struct in6_ifaddr *ia6;
862 ip6 = mtod(m, struct ip6_hdr *);
863 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
864 if (ia6) {
865 ia6->ia_ifa.ifa_data.ifad_outbytes +=
866 m->m_pkthdr.len;
867 }
868 #endif
869 #ifdef IPSEC
870 /* clean ipsec history once it goes out of the node */
871 ipsec_delaux(m);
872 #endif
873 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
874 goto done;
875 } else if (mtu < IPV6_MMTU) {
876 /*
877 * note that path MTU is never less than IPV6_MMTU
878 * (see icmp6_input).
879 */
880 error = EMSGSIZE;
881 in6_ifstat_inc(ifp, ifs6_out_fragfail);
882 goto bad;
883 } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
884 error = EMSGSIZE;
885 in6_ifstat_inc(ifp, ifs6_out_fragfail);
886 goto bad;
887 } else {
888 struct mbuf **mnext, *m_frgpart;
889 struct ip6_frag *ip6f;
890 u_int32_t id = htonl(ip6_id++);
891 u_char nextproto;
892
893 /*
894 * Too large for the destination or interface;
895 * fragment if possible.
896 * Must be able to put at least 8 bytes per fragment.
897 */
898 hlen = unfragpartlen;
899 if (mtu > IPV6_MAXPACKET)
900 mtu = IPV6_MAXPACKET;
901 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
902 if (len < 8) {
903 error = EMSGSIZE;
904 in6_ifstat_inc(ifp, ifs6_out_fragfail);
905 goto bad;
906 }
907
908 mnext = &m->m_nextpkt;
909
910 /*
911 * Change the next header field of the last header in the
912 * unfragmentable part.
913 */
914 if (exthdrs.ip6e_rthdr) {
915 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
916 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
917 } else if (exthdrs.ip6e_dest1) {
918 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
919 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
920 } else if (exthdrs.ip6e_hbh) {
921 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
922 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
923 } else {
924 nextproto = ip6->ip6_nxt;
925 ip6->ip6_nxt = IPPROTO_FRAGMENT;
926 }
927
928 /*
929 * Loop through length of segment after first fragment,
930 * make new header and copy data of each part and link onto chain.
931 */
932 m0 = m;
933 for (off = hlen; off < tlen; off += len) {
934 MGETHDR(m, M_DONTWAIT, MT_HEADER);
935 if (!m) {
936 error = ENOBUFS;
937 ip6stat.ip6s_odropped++;
938 goto sendorfree;
939 }
940 m->m_flags = m0->m_flags & M_COPYFLAGS;
941 *mnext = m;
942 mnext = &m->m_nextpkt;
943 m->m_data += max_linkhdr;
944 mhip6 = mtod(m, struct ip6_hdr *);
945 *mhip6 = *ip6;
946 m->m_len = sizeof(*mhip6);
947 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
948 if (error) {
949 ip6stat.ip6s_odropped++;
950 goto sendorfree;
951 }
952 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
953 if (off + len >= tlen)
954 len = tlen - off;
955 else
956 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
957 mhip6->ip6_plen = htons((u_short)(len + hlen +
958 sizeof(*ip6f) -
959 sizeof(struct ip6_hdr)));
960 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
961 error = ENOBUFS;
962 ip6stat.ip6s_odropped++;
963 goto sendorfree;
964 }
965 m_cat(m, m_frgpart);
966 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
967 m->m_pkthdr.rcvif = (struct ifnet *)0;
968 ip6f->ip6f_reserved = 0;
969 ip6f->ip6f_ident = id;
970 ip6f->ip6f_nxt = nextproto;
971 ip6stat.ip6s_ofragments++;
972 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
973 }
974
975 in6_ifstat_inc(ifp, ifs6_out_fragok);
976 }
977
978 /*
979 * Remove leading garbages.
980 */
981 sendorfree:
982 m = m0->m_nextpkt;
983 m0->m_nextpkt = 0;
984 m_freem(m0);
985 for (m0 = m; m; m = m0) {
986 m0 = m->m_nextpkt;
987 m->m_nextpkt = 0;
988 if (error == 0) {
989 #ifdef IFA_STATS
990 struct in6_ifaddr *ia6;
991 ip6 = mtod(m, struct ip6_hdr *);
992 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
993 if (ia6) {
994 ia6->ia_ifa.ifa_data.ifad_outbytes +=
995 m->m_pkthdr.len;
996 }
997 #endif
998 #ifdef IPSEC
999 /* clean ipsec history once it goes out of the node */
1000 ipsec_delaux(m);
1001 #endif
1002 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1003 } else
1004 m_freem(m);
1005 }
1006
1007 if (error == 0)
1008 ip6stat.ip6s_fragmented++;
1009
1010 done:
1011 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1012 RTFREE(ro->ro_rt);
1013 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1014 RTFREE(ro_pmtu->ro_rt);
1015 }
1016
1017 #ifdef IPSEC
1018 if (sp != NULL)
1019 key_freesp(sp);
1020 #endif /* IPSEC */
1021
1022 return(error);
1023
1024 freehdrs:
1025 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1026 m_freem(exthdrs.ip6e_dest1);
1027 m_freem(exthdrs.ip6e_rthdr);
1028 m_freem(exthdrs.ip6e_dest2);
1029 /* fall through */
1030 bad:
1031 m_freem(m);
1032 goto done;
1033 }
1034
1035 static int
1036 ip6_copyexthdr(mp, hdr, hlen)
1037 struct mbuf **mp;
1038 caddr_t hdr;
1039 int hlen;
1040 {
1041 struct mbuf *m;
1042
1043 if (hlen > MCLBYTES)
1044 return(ENOBUFS); /* XXX */
1045
1046 MGET(m, M_DONTWAIT, MT_DATA);
1047 if (!m)
1048 return(ENOBUFS);
1049
1050 if (hlen > MLEN) {
1051 MCLGET(m, M_DONTWAIT);
1052 if ((m->m_flags & M_EXT) == 0) {
1053 m_free(m);
1054 return(ENOBUFS);
1055 }
1056 }
1057 m->m_len = hlen;
1058 if (hdr)
1059 bcopy(hdr, mtod(m, caddr_t), hlen);
1060
1061 *mp = m;
1062 return(0);
1063 }
1064
1065 /*
1066 * Insert jumbo payload option.
1067 */
1068 static int
1069 ip6_insert_jumboopt(exthdrs, plen)
1070 struct ip6_exthdrs *exthdrs;
1071 u_int32_t plen;
1072 {
1073 struct mbuf *mopt;
1074 u_char *optbuf;
1075 u_int32_t v;
1076
1077 #define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1078
1079 /*
1080 * If there is no hop-by-hop options header, allocate new one.
1081 * If there is one but it doesn't have enough space to store the
1082 * jumbo payload option, allocate a cluster to store the whole options.
1083 * Otherwise, use it to store the options.
1084 */
1085 if (exthdrs->ip6e_hbh == 0) {
1086 MGET(mopt, M_DONTWAIT, MT_DATA);
1087 if (mopt == 0)
1088 return(ENOBUFS);
1089 mopt->m_len = JUMBOOPTLEN;
1090 optbuf = mtod(mopt, u_char *);
1091 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1092 exthdrs->ip6e_hbh = mopt;
1093 } else {
1094 struct ip6_hbh *hbh;
1095
1096 mopt = exthdrs->ip6e_hbh;
1097 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1098 /*
1099 * XXX assumption:
1100 * - exthdrs->ip6e_hbh is not referenced from places
1101 * other than exthdrs.
1102 * - exthdrs->ip6e_hbh is not an mbuf chain.
1103 */
1104 int oldoptlen = mopt->m_len;
1105 struct mbuf *n;
1106
1107 /*
1108 * XXX: give up if the whole (new) hbh header does
1109 * not fit even in an mbuf cluster.
1110 */
1111 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1112 return(ENOBUFS);
1113
1114 /*
1115 * As a consequence, we must always prepare a cluster
1116 * at this point.
1117 */
1118 MGET(n, M_DONTWAIT, MT_DATA);
1119 if (n) {
1120 MCLGET(n, M_DONTWAIT);
1121 if ((n->m_flags & M_EXT) == 0) {
1122 m_freem(n);
1123 n = NULL;
1124 }
1125 }
1126 if (!n)
1127 return(ENOBUFS);
1128 n->m_len = oldoptlen + JUMBOOPTLEN;
1129 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1130 oldoptlen);
1131 optbuf = mtod(n, caddr_t) + oldoptlen;
1132 m_freem(mopt);
1133 mopt = exthdrs->ip6e_hbh = n;
1134 } else {
1135 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1136 mopt->m_len += JUMBOOPTLEN;
1137 }
1138 optbuf[0] = IP6OPT_PADN;
1139 optbuf[1] = 1;
1140
1141 /*
1142 * Adjust the header length according to the pad and
1143 * the jumbo payload option.
1144 */
1145 hbh = mtod(mopt, struct ip6_hbh *);
1146 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1147 }
1148
1149 /* fill in the option. */
1150 optbuf[2] = IP6OPT_JUMBO;
1151 optbuf[3] = 4;
1152 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1153 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1154
1155 /* finally, adjust the packet header length */
1156 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1157
1158 return(0);
1159 #undef JUMBOOPTLEN
1160 }
1161
1162 /*
1163 * Insert fragment header and copy unfragmentable header portions.
1164 */
1165 static int
1166 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1167 struct mbuf *m0, *m;
1168 int hlen;
1169 struct ip6_frag **frghdrp;
1170 {
1171 struct mbuf *n, *mlast;
1172
1173 if (hlen > sizeof(struct ip6_hdr)) {
1174 n = m_copym(m0, sizeof(struct ip6_hdr),
1175 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1176 if (n == 0)
1177 return(ENOBUFS);
1178 m->m_next = n;
1179 } else
1180 n = m;
1181
1182 /* Search for the last mbuf of unfragmentable part. */
1183 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1184 ;
1185
1186 if ((mlast->m_flags & M_EXT) == 0 &&
1187 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1188 /* use the trailing space of the last mbuf for the fragment hdr */
1189 *frghdrp =
1190 (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1191 mlast->m_len += sizeof(struct ip6_frag);
1192 m->m_pkthdr.len += sizeof(struct ip6_frag);
1193 } else {
1194 /* allocate a new mbuf for the fragment header */
1195 struct mbuf *mfrg;
1196
1197 MGET(mfrg, M_DONTWAIT, MT_DATA);
1198 if (mfrg == 0)
1199 return(ENOBUFS);
1200 mfrg->m_len = sizeof(struct ip6_frag);
1201 *frghdrp = mtod(mfrg, struct ip6_frag *);
1202 mlast->m_next = mfrg;
1203 }
1204
1205 return(0);
1206 }
1207
1208 /*
1209 * IP6 socket option processing.
1210 */
1211 int
1212 ip6_ctloutput(op, so, level, optname, mp)
1213 int op;
1214 struct socket *so;
1215 int level, optname;
1216 struct mbuf **mp;
1217 {
1218 struct in6pcb *in6p = sotoin6pcb(so);
1219 struct mbuf *m = *mp;
1220 int optval = 0;
1221 int error = 0;
1222 struct proc *p = curproc; /* XXX */
1223
1224 if (level == IPPROTO_IPV6) {
1225 switch (op) {
1226
1227 case PRCO_SETOPT:
1228 switch (optname) {
1229 case IPV6_PKTOPTIONS:
1230 /* m is freed in ip6_pcbopts */
1231 return(ip6_pcbopts(&in6p->in6p_outputopts,
1232 m, so));
1233 case IPV6_HOPOPTS:
1234 case IPV6_DSTOPTS:
1235 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1236 error = EPERM;
1237 break;
1238 }
1239 /* fall through */
1240 case IPV6_UNICAST_HOPS:
1241 case IPV6_RECVOPTS:
1242 case IPV6_RECVRETOPTS:
1243 case IPV6_RECVDSTADDR:
1244 case IPV6_PKTINFO:
1245 case IPV6_HOPLIMIT:
1246 case IPV6_RTHDR:
1247 case IPV6_CHECKSUM:
1248 case IPV6_FAITH:
1249 case IPV6_V6ONLY:
1250 if (!m || m->m_len != sizeof(int)) {
1251 error = EINVAL;
1252 break;
1253 }
1254 optval = *mtod(m, int *);
1255 switch (optname) {
1256
1257 case IPV6_UNICAST_HOPS:
1258 if (optval < -1 || optval >= 256)
1259 error = EINVAL;
1260 else {
1261 /* -1 = kernel default */
1262 in6p->in6p_hops = optval;
1263 }
1264 break;
1265 #define OPTSET(bit) \
1266 if (optval) \
1267 in6p->in6p_flags |= bit; \
1268 else \
1269 in6p->in6p_flags &= ~bit;
1270
1271 case IPV6_RECVOPTS:
1272 OPTSET(IN6P_RECVOPTS);
1273 break;
1274
1275 case IPV6_RECVRETOPTS:
1276 OPTSET(IN6P_RECVRETOPTS);
1277 break;
1278
1279 case IPV6_RECVDSTADDR:
1280 OPTSET(IN6P_RECVDSTADDR);
1281 break;
1282
1283 case IPV6_PKTINFO:
1284 OPTSET(IN6P_PKTINFO);
1285 break;
1286
1287 case IPV6_HOPLIMIT:
1288 OPTSET(IN6P_HOPLIMIT);
1289 break;
1290
1291 case IPV6_HOPOPTS:
1292 OPTSET(IN6P_HOPOPTS);
1293 break;
1294
1295 case IPV6_DSTOPTS:
1296 OPTSET(IN6P_DSTOPTS);
1297 break;
1298
1299 case IPV6_RTHDR:
1300 OPTSET(IN6P_RTHDR);
1301 break;
1302
1303 case IPV6_CHECKSUM:
1304 in6p->in6p_cksum = optval;
1305 break;
1306
1307 case IPV6_FAITH:
1308 OPTSET(IN6P_FAITH);
1309 break;
1310
1311 case IPV6_V6ONLY:
1312 /*
1313 * make setsockopt(IPV6_V6ONLY)
1314 * available only prior to bind(2).
1315 * see ipng mailing list, Jun 22 2001.
1316 */
1317 if (in6p->in6p_lport ||
1318 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1319 {
1320 error = EINVAL;
1321 break;
1322 }
1323 #ifdef INET6_BINDV6ONLY
1324 if (!optval)
1325 error = EINVAL;
1326 #else
1327 OPTSET(IN6P_IPV6_V6ONLY);
1328 #endif
1329 break;
1330 }
1331 break;
1332 #undef OPTSET
1333
1334 case IPV6_MULTICAST_IF:
1335 case IPV6_MULTICAST_HOPS:
1336 case IPV6_MULTICAST_LOOP:
1337 case IPV6_JOIN_GROUP:
1338 case IPV6_LEAVE_GROUP:
1339 error = ip6_setmoptions(optname, &in6p->in6p_moptions, m);
1340 break;
1341
1342 case IPV6_PORTRANGE:
1343 optval = *mtod(m, int *);
1344
1345 switch (optval) {
1346 case IPV6_PORTRANGE_DEFAULT:
1347 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1348 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1349 break;
1350
1351 case IPV6_PORTRANGE_HIGH:
1352 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1353 in6p->in6p_flags |= IN6P_HIGHPORT;
1354 break;
1355
1356 case IPV6_PORTRANGE_LOW:
1357 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1358 in6p->in6p_flags |= IN6P_LOWPORT;
1359 break;
1360
1361 default:
1362 error = EINVAL;
1363 break;
1364 }
1365 break;
1366
1367 #ifdef IPSEC
1368 case IPV6_IPSEC_POLICY:
1369 {
1370 caddr_t req = NULL;
1371 size_t len = 0;
1372
1373 int priv = 0;
1374 if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1375 priv = 0;
1376 else
1377 priv = 1;
1378 if (m) {
1379 req = mtod(m, caddr_t);
1380 len = m->m_len;
1381 }
1382 error = ipsec6_set_policy(in6p,
1383 optname, req, len, priv);
1384 }
1385 break;
1386 #endif /* IPSEC */
1387
1388 default:
1389 error = ENOPROTOOPT;
1390 break;
1391 }
1392 if (m)
1393 (void)m_free(m);
1394 break;
1395
1396 case PRCO_GETOPT:
1397 switch (optname) {
1398
1399 case IPV6_OPTIONS:
1400 case IPV6_RETOPTS:
1401 #if 0
1402 *mp = m = m_get(M_WAIT, MT_SOOPTS);
1403 if (in6p->in6p_options) {
1404 m->m_len = in6p->in6p_options->m_len;
1405 bcopy(mtod(in6p->in6p_options, caddr_t),
1406 mtod(m, caddr_t),
1407 (unsigned)m->m_len);
1408 } else
1409 m->m_len = 0;
1410 break;
1411 #else
1412 error = ENOPROTOOPT;
1413 break;
1414 #endif
1415
1416 case IPV6_PKTOPTIONS:
1417 if (in6p->in6p_options) {
1418 *mp = m_copym(in6p->in6p_options, 0,
1419 M_COPYALL, M_WAIT);
1420 } else {
1421 *mp = m_get(M_WAIT, MT_SOOPTS);
1422 (*mp)->m_len = 0;
1423 }
1424 break;
1425
1426 case IPV6_HOPOPTS:
1427 case IPV6_DSTOPTS:
1428 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1429 error = EPERM;
1430 break;
1431 }
1432 /* fall through */
1433 case IPV6_UNICAST_HOPS:
1434 case IPV6_RECVOPTS:
1435 case IPV6_RECVRETOPTS:
1436 case IPV6_RECVDSTADDR:
1437 case IPV6_PORTRANGE:
1438 case IPV6_PKTINFO:
1439 case IPV6_HOPLIMIT:
1440 case IPV6_RTHDR:
1441 case IPV6_CHECKSUM:
1442 case IPV6_FAITH:
1443 case IPV6_V6ONLY:
1444 *mp = m = m_get(M_WAIT, MT_SOOPTS);
1445 m->m_len = sizeof(int);
1446 switch (optname) {
1447
1448 case IPV6_UNICAST_HOPS:
1449 optval = in6p->in6p_hops;
1450 break;
1451
1452 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1453
1454 case IPV6_RECVOPTS:
1455 optval = OPTBIT(IN6P_RECVOPTS);
1456 break;
1457
1458 case IPV6_RECVRETOPTS:
1459 optval = OPTBIT(IN6P_RECVRETOPTS);
1460 break;
1461
1462 case IPV6_RECVDSTADDR:
1463 optval = OPTBIT(IN6P_RECVDSTADDR);
1464 break;
1465
1466 case IPV6_PORTRANGE:
1467 {
1468 int flags;
1469 flags = in6p->in6p_flags;
1470 if (flags & IN6P_HIGHPORT)
1471 optval = IPV6_PORTRANGE_HIGH;
1472 else if (flags & IN6P_LOWPORT)
1473 optval = IPV6_PORTRANGE_LOW;
1474 else
1475 optval = 0;
1476 break;
1477 }
1478
1479 case IPV6_PKTINFO:
1480 optval = OPTBIT(IN6P_PKTINFO);
1481 break;
1482
1483 case IPV6_HOPLIMIT:
1484 optval = OPTBIT(IN6P_HOPLIMIT);
1485 break;
1486
1487 case IPV6_HOPOPTS:
1488 optval = OPTBIT(IN6P_HOPOPTS);
1489 break;
1490
1491 case IPV6_DSTOPTS:
1492 optval = OPTBIT(IN6P_DSTOPTS);
1493 break;
1494
1495 case IPV6_RTHDR:
1496 optval = OPTBIT(IN6P_RTHDR);
1497 break;
1498
1499 case IPV6_CHECKSUM:
1500 optval = in6p->in6p_cksum;
1501 break;
1502
1503 case IPV6_FAITH:
1504 optval = OPTBIT(IN6P_FAITH);
1505 break;
1506
1507 case IPV6_V6ONLY:
1508 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1509 break;
1510 }
1511 *mtod(m, int *) = optval;
1512 break;
1513
1514 case IPV6_MULTICAST_IF:
1515 case IPV6_MULTICAST_HOPS:
1516 case IPV6_MULTICAST_LOOP:
1517 case IPV6_JOIN_GROUP:
1518 case IPV6_LEAVE_GROUP:
1519 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1520 break;
1521
1522 #ifdef IPSEC
1523 case IPV6_IPSEC_POLICY:
1524 {
1525 caddr_t req = NULL;
1526 size_t len = 0;
1527
1528 if (m) {
1529 req = mtod(m, caddr_t);
1530 len = m->m_len;
1531 }
1532 error = ipsec6_get_policy(in6p, req, len, mp);
1533 break;
1534 }
1535 #endif /* IPSEC */
1536
1537 default:
1538 error = ENOPROTOOPT;
1539 break;
1540 }
1541 break;
1542 }
1543 } else {
1544 error = EINVAL;
1545 if (op == PRCO_SETOPT && *mp)
1546 (void)m_free(*mp);
1547 }
1548 return(error);
1549 }
1550
1551 /*
1552 * Set up IP6 options in pcb for insertion in output packets.
1553 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1554 * with destination address if source routed.
1555 */
1556 static int
1557 ip6_pcbopts(pktopt, m, so)
1558 struct ip6_pktopts **pktopt;
1559 struct mbuf *m;
1560 struct socket *so;
1561 {
1562 struct ip6_pktopts *opt = *pktopt;
1563 int error = 0;
1564 struct proc *p = curproc; /* XXX */
1565 int priv = 0;
1566
1567 /* turn off any old options. */
1568 if (opt) {
1569 if (opt->ip6po_m)
1570 (void)m_free(opt->ip6po_m);
1571 } else
1572 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1573 *pktopt = 0;
1574
1575 if (!m || m->m_len == 0) {
1576 /*
1577 * Only turning off any previous options.
1578 */
1579 if (opt)
1580 free(opt, M_IP6OPT);
1581 if (m)
1582 (void)m_free(m);
1583 return(0);
1584 }
1585
1586 /* set options specified by user. */
1587 if (p && !suser(p->p_ucred, &p->p_acflag))
1588 priv = 1;
1589 if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1590 (void)m_free(m);
1591 return(error);
1592 }
1593 *pktopt = opt;
1594 return(0);
1595 }
1596
1597 /*
1598 * Set the IP6 multicast options in response to user setsockopt().
1599 */
1600 static int
1601 ip6_setmoptions(optname, im6op, m)
1602 int optname;
1603 struct ip6_moptions **im6op;
1604 struct mbuf *m;
1605 {
1606 int error = 0;
1607 u_int loop, ifindex;
1608 struct ipv6_mreq *mreq;
1609 struct ifnet *ifp;
1610 struct ip6_moptions *im6o = *im6op;
1611 struct route_in6 ro;
1612 struct sockaddr_in6 *dst;
1613 struct in6_multi_mship *imm;
1614 struct proc *p = curproc; /* XXX */
1615
1616 if (im6o == NULL) {
1617 /*
1618 * No multicast option buffer attached to the pcb;
1619 * allocate one and initialize to default values.
1620 */
1621 im6o = (struct ip6_moptions *)
1622 malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1623
1624 if (im6o == NULL)
1625 return(ENOBUFS);
1626 *im6op = im6o;
1627 im6o->im6o_multicast_ifp = NULL;
1628 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1629 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1630 LIST_INIT(&im6o->im6o_memberships);
1631 }
1632
1633 switch (optname) {
1634
1635 case IPV6_MULTICAST_IF:
1636 /*
1637 * Select the interface for outgoing multicast packets.
1638 */
1639 if (m == NULL || m->m_len != sizeof(u_int)) {
1640 error = EINVAL;
1641 break;
1642 }
1643 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1644 if (ifindex < 0 || if_index < ifindex) {
1645 error = ENXIO; /* XXX EINVAL? */
1646 break;
1647 }
1648 ifp = ifindex2ifnet[ifindex];
1649 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1650 error = EADDRNOTAVAIL;
1651 break;
1652 }
1653 im6o->im6o_multicast_ifp = ifp;
1654 break;
1655
1656 case IPV6_MULTICAST_HOPS:
1657 {
1658 /*
1659 * Set the IP6 hoplimit for outgoing multicast packets.
1660 */
1661 int optval;
1662 if (m == NULL || m->m_len != sizeof(int)) {
1663 error = EINVAL;
1664 break;
1665 }
1666 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1667 if (optval < -1 || optval >= 256)
1668 error = EINVAL;
1669 else if (optval == -1)
1670 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1671 else
1672 im6o->im6o_multicast_hlim = optval;
1673 break;
1674 }
1675
1676 case IPV6_MULTICAST_LOOP:
1677 /*
1678 * Set the loopback flag for outgoing multicast packets.
1679 * Must be zero or one.
1680 */
1681 if (m == NULL || m->m_len != sizeof(u_int)) {
1682 error = EINVAL;
1683 break;
1684 }
1685 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1686 if (loop > 1) {
1687 error = EINVAL;
1688 break;
1689 }
1690 im6o->im6o_multicast_loop = loop;
1691 break;
1692
1693 case IPV6_JOIN_GROUP:
1694 /*
1695 * Add a multicast group membership.
1696 * Group must be a valid IP6 multicast address.
1697 */
1698 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1699 error = EINVAL;
1700 break;
1701 }
1702 mreq = mtod(m, struct ipv6_mreq *);
1703 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1704 /*
1705 * We use the unspecified address to specify to accept
1706 * all multicast addresses. Only super user is allowed
1707 * to do this.
1708 */
1709 if (suser(p->p_ucred, &p->p_acflag))
1710 {
1711 error = EACCES;
1712 break;
1713 }
1714 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1715 error = EINVAL;
1716 break;
1717 }
1718
1719 /*
1720 * If the interface is specified, validate it.
1721 */
1722 if (mreq->ipv6mr_interface < 0
1723 || if_index < mreq->ipv6mr_interface) {
1724 error = ENXIO; /* XXX EINVAL? */
1725 break;
1726 }
1727 /*
1728 * If no interface was explicitly specified, choose an
1729 * appropriate one according to the given multicast address.
1730 */
1731 if (mreq->ipv6mr_interface == 0) {
1732 /*
1733 * If the multicast address is in node-local scope,
1734 * the interface should be a loopback interface.
1735 * Otherwise, look up the routing table for the
1736 * address, and choose the outgoing interface.
1737 * XXX: is it a good approach?
1738 */
1739 if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1740 ifp = &loif[0];
1741 } else {
1742 ro.ro_rt = NULL;
1743 dst = (struct sockaddr_in6 *)&ro.ro_dst;
1744 bzero(dst, sizeof(*dst));
1745 dst->sin6_len = sizeof(struct sockaddr_in6);
1746 dst->sin6_family = AF_INET6;
1747 dst->sin6_addr = mreq->ipv6mr_multiaddr;
1748 rtalloc((struct route *)&ro);
1749 if (ro.ro_rt == NULL) {
1750 error = EADDRNOTAVAIL;
1751 break;
1752 }
1753 ifp = ro.ro_rt->rt_ifp;
1754 rtfree(ro.ro_rt);
1755 }
1756 } else
1757 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1758
1759 /*
1760 * See if we found an interface, and confirm that it
1761 * supports multicast
1762 */
1763 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1764 error = EADDRNOTAVAIL;
1765 break;
1766 }
1767 /*
1768 * Put interface index into the multicast address,
1769 * if the address has link-local scope.
1770 */
1771 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1772 mreq->ipv6mr_multiaddr.s6_addr16[1]
1773 = htons(mreq->ipv6mr_interface);
1774 }
1775 /*
1776 * See if the membership already exists.
1777 */
1778 for (imm = im6o->im6o_memberships.lh_first;
1779 imm != NULL; imm = imm->i6mm_chain.le_next)
1780 if (imm->i6mm_maddr->in6m_ifp == ifp &&
1781 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1782 &mreq->ipv6mr_multiaddr))
1783 break;
1784 if (imm != NULL) {
1785 error = EADDRINUSE;
1786 break;
1787 }
1788 /*
1789 * Everything looks good; add a new record to the multicast
1790 * address list for the given interface.
1791 */
1792 imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
1793 if (imm == NULL) {
1794 error = ENOBUFS;
1795 break;
1796 }
1797 if ((imm->i6mm_maddr =
1798 in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
1799 free(imm, M_IPMADDR);
1800 break;
1801 }
1802 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1803 break;
1804
1805 case IPV6_LEAVE_GROUP:
1806 /*
1807 * Drop a multicast group membership.
1808 * Group must be a valid IP6 multicast address.
1809 */
1810 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1811 error = EINVAL;
1812 break;
1813 }
1814 mreq = mtod(m, struct ipv6_mreq *);
1815 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1816 if (suser(p->p_ucred, &p->p_acflag))
1817 {
1818 error = EACCES;
1819 break;
1820 }
1821 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1822 error = EINVAL;
1823 break;
1824 }
1825 /*
1826 * If an interface address was specified, get a pointer
1827 * to its ifnet structure.
1828 */
1829 if (mreq->ipv6mr_interface < 0
1830 || if_index < mreq->ipv6mr_interface) {
1831 error = ENXIO; /* XXX EINVAL? */
1832 break;
1833 }
1834 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1835 /*
1836 * Put interface index into the multicast address,
1837 * if the address has link-local scope.
1838 */
1839 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1840 mreq->ipv6mr_multiaddr.s6_addr16[1]
1841 = htons(mreq->ipv6mr_interface);
1842 }
1843 /*
1844 * Find the membership in the membership list.
1845 */
1846 for (imm = im6o->im6o_memberships.lh_first;
1847 imm != NULL; imm = imm->i6mm_chain.le_next) {
1848 if ((ifp == NULL ||
1849 imm->i6mm_maddr->in6m_ifp == ifp) &&
1850 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1851 &mreq->ipv6mr_multiaddr))
1852 break;
1853 }
1854 if (imm == NULL) {
1855 /* Unable to resolve interface */
1856 error = EADDRNOTAVAIL;
1857 break;
1858 }
1859 /*
1860 * Give up the multicast address record to which the
1861 * membership points.
1862 */
1863 LIST_REMOVE(imm, i6mm_chain);
1864 in6_delmulti(imm->i6mm_maddr);
1865 free(imm, M_IPMADDR);
1866 break;
1867
1868 default:
1869 error = EOPNOTSUPP;
1870 break;
1871 }
1872
1873 /*
1874 * If all options have default values, no need to keep the mbuf.
1875 */
1876 if (im6o->im6o_multicast_ifp == NULL &&
1877 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1878 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1879 im6o->im6o_memberships.lh_first == NULL) {
1880 free(*im6op, M_IPMOPTS);
1881 *im6op = NULL;
1882 }
1883
1884 return(error);
1885 }
1886
1887 /*
1888 * Return the IP6 multicast options in response to user getsockopt().
1889 */
1890 static int
1891 ip6_getmoptions(optname, im6o, mp)
1892 int optname;
1893 struct ip6_moptions *im6o;
1894 struct mbuf **mp;
1895 {
1896 u_int *hlim, *loop, *ifindex;
1897
1898 *mp = m_get(M_WAIT, MT_SOOPTS);
1899
1900 switch (optname) {
1901
1902 case IPV6_MULTICAST_IF:
1903 ifindex = mtod(*mp, u_int *);
1904 (*mp)->m_len = sizeof(u_int);
1905 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1906 *ifindex = 0;
1907 else
1908 *ifindex = im6o->im6o_multicast_ifp->if_index;
1909 return(0);
1910
1911 case IPV6_MULTICAST_HOPS:
1912 hlim = mtod(*mp, u_int *);
1913 (*mp)->m_len = sizeof(u_int);
1914 if (im6o == NULL)
1915 *hlim = ip6_defmcasthlim;
1916 else
1917 *hlim = im6o->im6o_multicast_hlim;
1918 return(0);
1919
1920 case IPV6_MULTICAST_LOOP:
1921 loop = mtod(*mp, u_int *);
1922 (*mp)->m_len = sizeof(u_int);
1923 if (im6o == NULL)
1924 *loop = ip6_defmcasthlim;
1925 else
1926 *loop = im6o->im6o_multicast_loop;
1927 return(0);
1928
1929 default:
1930 return(EOPNOTSUPP);
1931 }
1932 }
1933
1934 /*
1935 * Discard the IP6 multicast options.
1936 */
1937 void
1938 ip6_freemoptions(im6o)
1939 struct ip6_moptions *im6o;
1940 {
1941 struct in6_multi_mship *imm;
1942
1943 if (im6o == NULL)
1944 return;
1945
1946 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
1947 LIST_REMOVE(imm, i6mm_chain);
1948 if (imm->i6mm_maddr)
1949 in6_delmulti(imm->i6mm_maddr);
1950 free(imm, M_IPMADDR);
1951 }
1952 free(im6o, M_IPMOPTS);
1953 }
1954
1955 /*
1956 * Set IPv6 outgoing packet options based on advanced API.
1957 */
1958 int
1959 ip6_setpktoptions(control, opt, priv)
1960 struct mbuf *control;
1961 struct ip6_pktopts *opt;
1962 int priv;
1963 {
1964 struct cmsghdr *cm = 0;
1965
1966 if (control == 0 || opt == 0)
1967 return(EINVAL);
1968
1969 bzero(opt, sizeof(*opt));
1970 opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
1971
1972 /*
1973 * XXX: Currently, we assume all the optional information is stored
1974 * in a single mbuf.
1975 */
1976 if (control->m_next)
1977 return(EINVAL);
1978
1979 opt->ip6po_m = control;
1980
1981 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
1982 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1983 cm = mtod(control, struct cmsghdr *);
1984 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
1985 return(EINVAL);
1986 if (cm->cmsg_level != IPPROTO_IPV6)
1987 continue;
1988
1989 switch (cm->cmsg_type) {
1990 case IPV6_PKTINFO:
1991 if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
1992 return(EINVAL);
1993 opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
1994 if (opt->ip6po_pktinfo->ipi6_ifindex &&
1995 IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
1996 opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
1997 htons(opt->ip6po_pktinfo->ipi6_ifindex);
1998
1999 if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2000 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2001 return(ENXIO);
2002 }
2003
2004 /*
2005 * Check if the requested source address is indeed a
2006 * unicast address assigned to the node, and can be
2007 * used as the packet's source address.
2008 */
2009 if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2010 struct ifaddr *ia;
2011 struct in6_ifaddr *ia6;
2012 struct sockaddr_in6 sin6;
2013
2014 bzero(&sin6, sizeof(sin6));
2015 sin6.sin6_len = sizeof(sin6);
2016 sin6.sin6_family = AF_INET6;
2017 sin6.sin6_addr =
2018 opt->ip6po_pktinfo->ipi6_addr;
2019 ia = ifa_ifwithaddr(sin6tosa(&sin6));
2020 if (ia == NULL ||
2021 (opt->ip6po_pktinfo->ipi6_ifindex &&
2022 (ia->ifa_ifp->if_index !=
2023 opt->ip6po_pktinfo->ipi6_ifindex))) {
2024 return(EADDRNOTAVAIL);
2025 }
2026 ia6 = (struct in6_ifaddr *)ia;
2027 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
2028 return(EADDRNOTAVAIL);
2029 }
2030
2031 /*
2032 * Check if the requested source address is
2033 * indeed a unicast address assigned to the
2034 * node.
2035 */
2036 if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2037 return(EADDRNOTAVAIL);
2038 }
2039 break;
2040
2041 case IPV6_HOPLIMIT:
2042 if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2043 return(EINVAL);
2044
2045 bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
2046 sizeof(opt->ip6po_hlim));
2047 if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2048 return(EINVAL);
2049 break;
2050
2051 case IPV6_NEXTHOP:
2052 if (!priv)
2053 return(EPERM);
2054
2055 if (cm->cmsg_len < sizeof(u_char) ||
2056 /* check if cmsg_len is large enough for sa_len */
2057 cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2058 return(EINVAL);
2059
2060 opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2061
2062 break;
2063
2064 case IPV6_HOPOPTS:
2065 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2066 return(EINVAL);
2067 opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2068 if (cm->cmsg_len !=
2069 CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2070 return(EINVAL);
2071 break;
2072
2073 case IPV6_DSTOPTS:
2074 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2075 return(EINVAL);
2076
2077 /*
2078 * If there is no routing header yet, the destination
2079 * options header should be put on the 1st part.
2080 * Otherwise, the header should be on the 2nd part.
2081 * (See RFC 2460, section 4.1)
2082 */
2083 if (opt->ip6po_rthdr == NULL) {
2084 opt->ip6po_dest1 =
2085 (struct ip6_dest *)CMSG_DATA(cm);
2086 if (cm->cmsg_len !=
2087 CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
2088 << 3))
2089 return(EINVAL);
2090 }
2091 else {
2092 opt->ip6po_dest2 =
2093 (struct ip6_dest *)CMSG_DATA(cm);
2094 if (cm->cmsg_len !=
2095 CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
2096 << 3))
2097 return(EINVAL);
2098 }
2099 break;
2100
2101 case IPV6_RTHDR:
2102 if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2103 return(EINVAL);
2104 opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2105 if (cm->cmsg_len !=
2106 CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2107 return(EINVAL);
2108 switch (opt->ip6po_rthdr->ip6r_type) {
2109 case IPV6_RTHDR_TYPE_0:
2110 if (opt->ip6po_rthdr->ip6r_segleft == 0)
2111 return(EINVAL);
2112 break;
2113 default:
2114 return(EINVAL);
2115 }
2116 break;
2117
2118 default:
2119 return(ENOPROTOOPT);
2120 }
2121 }
2122
2123 return(0);
2124 }
2125
2126 /*
2127 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2128 * packet to the input queue of a specified interface. Note that this
2129 * calls the output routine of the loopback "driver", but with an interface
2130 * pointer that might NOT be &loif -- easier than replicating that code here.
2131 */
2132 void
2133 ip6_mloopback(ifp, m, dst)
2134 struct ifnet *ifp;
2135 struct mbuf *m;
2136 struct sockaddr_in6 *dst;
2137 {
2138 struct mbuf *copym;
2139 struct ip6_hdr *ip6;
2140
2141 copym = m_copy(m, 0, M_COPYALL);
2142 if (copym == NULL)
2143 return;
2144
2145 /*
2146 * Make sure to deep-copy IPv6 header portion in case the data
2147 * is in an mbuf cluster, so that we can safely override the IPv6
2148 * header portion later.
2149 */
2150 if ((copym->m_flags & M_EXT) != 0 ||
2151 copym->m_len < sizeof(struct ip6_hdr)) {
2152 copym = m_pullup(copym, sizeof(struct ip6_hdr));
2153 if (copym == NULL)
2154 return;
2155 }
2156
2157 #ifdef DIAGNOSTIC
2158 if (copym->m_len < sizeof(*ip6)) {
2159 m_freem(copym);
2160 return;
2161 }
2162 #endif
2163
2164 ip6 = mtod(copym, struct ip6_hdr *);
2165 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2166 ip6->ip6_src.s6_addr16[1] = 0;
2167 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2168 ip6->ip6_dst.s6_addr16[1] = 0;
2169
2170 (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2171 }
2172
2173 /*
2174 * Chop IPv6 header off from the payload.
2175 */
2176 static int
2177 ip6_splithdr(m, exthdrs)
2178 struct mbuf *m;
2179 struct ip6_exthdrs *exthdrs;
2180 {
2181 struct mbuf *mh;
2182 struct ip6_hdr *ip6;
2183
2184 ip6 = mtod(m, struct ip6_hdr *);
2185 if (m->m_len > sizeof(*ip6)) {
2186 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2187 if (mh == 0) {
2188 m_freem(m);
2189 return ENOBUFS;
2190 }
2191 M_COPY_PKTHDR(mh, m);
2192 MH_ALIGN(mh, sizeof(*ip6));
2193 m->m_flags &= ~M_PKTHDR;
2194 m->m_len -= sizeof(*ip6);
2195 m->m_data += sizeof(*ip6);
2196 mh->m_next = m;
2197 m = mh;
2198 m->m_len = sizeof(*ip6);
2199 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2200 }
2201 exthdrs->ip6e_ip6 = m;
2202 return 0;
2203 }
2204
2205 /*
2206 * Compute IPv6 extension header length.
2207 */
2208 int
2209 ip6_optlen(in6p)
2210 struct in6pcb *in6p;
2211 {
2212 int len;
2213
2214 if (!in6p->in6p_outputopts)
2215 return 0;
2216
2217 len = 0;
2218 #define elen(x) \
2219 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2220
2221 len += elen(in6p->in6p_outputopts->ip6po_hbh);
2222 len += elen(in6p->in6p_outputopts->ip6po_dest1);
2223 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2224 len += elen(in6p->in6p_outputopts->ip6po_dest2);
2225 return len;
2226 #undef elen
2227 }
2228