65
66#include "opt_inet.h"
67#include "opt_inet6.h"
68#include "opt_ipsec.h"
69
70#include <sys/param.h>
71#include <sys/kernel.h>
72#include <sys/malloc.h>
73#include <sys/mbuf.h>
74#include <sys/errno.h>
75#include <sys/priv.h>
76#include <sys/proc.h>
77#include <sys/protosw.h>
78#include <sys/socket.h>
79#include <sys/socketvar.h>
80#include <sys/ucred.h>
81
82#include <net/if.h>
83#include <net/netisr.h>
84#include <net/route.h>
85#include <net/pfil.h>
86
87#include <netinet/in.h>
88#include <netinet/in_var.h>
89#include <netinet6/in6_var.h>
90#include <netinet/ip6.h>
91#include <netinet/icmp6.h>
92#include <netinet6/ip6_var.h>
93#include <netinet/in_pcb.h>
94#include <netinet/tcp_var.h>
95#include <netinet6/nd6.h>
96
97#ifdef IPSEC
98#include <netipsec/ipsec.h>
99#include <netipsec/ipsec6.h>
100#include <netipsec/key.h>
101#include <netinet6/ip6_ipsec.h>
102#endif /* IPSEC */
103
104#include <netinet6/ip6protosw.h>
105#include <netinet6/scope6_var.h>
106
107static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
108
109struct 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
117static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
118 struct ucred *, int));
119static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
120 struct socket *, struct sockopt *));
121static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
122static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
123 struct ucred *, int, int, int));
124
125static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
126static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
127static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
128static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
129 struct ip6_frag **));
130static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
131static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
132static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
133 struct ifnet *, struct in6_addr *, u_long *, int *));
134static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
135
136
137/*
138 * Make an extension header from option data. hp is the source, and
139 * mp is the destination.
140 */
141#define MAKE_EXTHDR(hp, mp) \
142 do { \
143 if (hp) { \
144 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
145 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
146 ((eh)->ip6e_len + 1) << 3); \
147 if (error) \
148 goto freehdrs; \
149 } \
150 } while (/*CONSTCOND*/ 0)
151
152/*
153 * Form a chain of extension headers.
154 * m is the extension header mbuf
155 * mp is the previous mbuf in the chain
156 * p is the next header
157 * i is the type of option.
158 */
159#define MAKE_CHAIN(m, mp, p, i)\
160 do {\
161 if (m) {\
162 if (!hdrsplit) \
163 panic("assumption failed: hdr not split"); \
164 *mtod((m), u_char *) = *(p);\
165 *(p) = (i);\
166 p = mtod((m), u_char *);\
167 (m)->m_next = (mp)->m_next;\
168 (mp)->m_next = (m);\
169 (mp) = (m);\
170 }\
171 } while (/*CONSTCOND*/ 0)
172
173/*
174 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
175 * header (with pri, len, nxt, hlim, src, dst).
176 * This function may modify ver and hlim only.
177 * The mbuf chain containing the packet will be freed.
178 * The mbuf opt, if present, will not be freed.
179 *
180 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
181 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
182 * which is rt_rmx.rmx_mtu.
183 *
184 * ifpp - XXX: just for statistics
185 */
186int
187ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
188 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
189 struct ifnet **ifpp, struct inpcb *inp)
190{
191 struct ip6_hdr *ip6, *mhip6;
192 struct ifnet *ifp, *origifp;
193 struct mbuf *m = m0;
194 struct mbuf *mprev = NULL;
195 int hlen, tlen, len, off;
196 struct route_in6 ip6route;
197 struct rtentry *rt = NULL;
198 struct sockaddr_in6 *dst, src_sa, dst_sa;
199 struct in6_addr odst;
200 int error = 0;
201 struct in6_ifaddr *ia = NULL;
202 u_long mtu;
203 int alwaysfrag, dontfrag;
204 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
205 struct ip6_exthdrs exthdrs;
206 struct in6_addr finaldst, src0, dst0;
207 u_int32_t zone;
208 struct route_in6 *ro_pmtu = NULL;
209 int hdrsplit = 0;
210 int needipsec = 0;
211#ifdef IPSEC
212 struct ipsec_output_state state;
213 struct ip6_rthdr *rh = NULL;
214 int needipsectun = 0;
215 int segleft_org = 0;
216 struct secpolicy *sp = NULL;
217#endif /* IPSEC */
218
219 ip6 = mtod(m, struct ip6_hdr *);
220 if (ip6 == NULL) {
221 printf ("ip6 is NULL");
222 goto bad;
223 }
224
225 finaldst = ip6->ip6_dst;
226
227 bzero(&exthdrs, sizeof(exthdrs));
228
229 if (opt) {
230 /* Hop-by-Hop options header */
231 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
232 /* Destination options header(1st part) */
233 if (opt->ip6po_rthdr) {
234 /*
235 * Destination options header(1st part)
236 * This only makes sense with a routing header.
237 * See Section 9.2 of RFC 3542.
238 * Disabling this part just for MIP6 convenience is
239 * a bad idea. We need to think carefully about a
240 * way to make the advanced API coexist with MIP6
241 * options, which might automatically be inserted in
242 * the kernel.
243 */
244 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
245 }
246 /* Routing header */
247 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
248 /* Destination options header(2nd part) */
249 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
250 }
251
252 /*
253 * IPSec checking which handles several cases.
254 * FAST IPSEC: We re-injected the packet.
255 */
256#ifdef IPSEC
257 switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
258 {
259 case 1: /* Bad packet */
260 goto freehdrs;
261 case -1: /* Do IPSec */
262 needipsec = 1;
263 case 0: /* No IPSec */
264 default:
265 break;
266 }
267#endif /* IPSEC */
268
269 /*
270 * Calculate the total length of the extension header chain.
271 * Keep the length of the unfragmentable part for fragmentation.
272 */
273 optlen = 0;
274 if (exthdrs.ip6e_hbh)
275 optlen += exthdrs.ip6e_hbh->m_len;
276 if (exthdrs.ip6e_dest1)
277 optlen += exthdrs.ip6e_dest1->m_len;
278 if (exthdrs.ip6e_rthdr)
279 optlen += exthdrs.ip6e_rthdr->m_len;
280 unfragpartlen = optlen + sizeof(struct ip6_hdr);
281
282 /* NOTE: we don't add AH/ESP length here. do that later. */
283 if (exthdrs.ip6e_dest2)
284 optlen += exthdrs.ip6e_dest2->m_len;
285
286 /*
287 * If we need IPsec, or there is at least one extension header,
288 * separate IP6 header from the payload.
289 */
290 if ((needipsec || optlen) && !hdrsplit) {
291 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
292 m = NULL;
293 goto freehdrs;
294 }
295 m = exthdrs.ip6e_ip6;
296 hdrsplit++;
297 }
298
299 /* adjust pointer */
300 ip6 = mtod(m, struct ip6_hdr *);
301
302 /* adjust mbuf packet header length */
303 m->m_pkthdr.len += optlen;
304 plen = m->m_pkthdr.len - sizeof(*ip6);
305
306 /* If this is a jumbo payload, insert a jumbo payload option. */
307 if (plen > IPV6_MAXPACKET) {
308 if (!hdrsplit) {
309 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
310 m = NULL;
311 goto freehdrs;
312 }
313 m = exthdrs.ip6e_ip6;
314 hdrsplit++;
315 }
316 /* adjust pointer */
317 ip6 = mtod(m, struct ip6_hdr *);
318 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
319 goto freehdrs;
320 ip6->ip6_plen = 0;
321 } else
322 ip6->ip6_plen = htons(plen);
323
324 /*
325 * Concatenate headers and fill in next header fields.
326 * Here we have, on "m"
327 * IPv6 payload
328 * and we insert headers accordingly. Finally, we should be getting:
329 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
330 *
331 * during the header composing process, "m" points to IPv6 header.
332 * "mprev" points to an extension header prior to esp.
333 */
334 u_char *nexthdrp = &ip6->ip6_nxt;
335 mprev = m;
336
337 /*
338 * we treat dest2 specially. this makes IPsec processing
339 * much easier. the goal here is to make mprev point the
340 * mbuf prior to dest2.
341 *
342 * result: IPv6 dest2 payload
343 * m and mprev will point to IPv6 header.
344 */
345 if (exthdrs.ip6e_dest2) {
346 if (!hdrsplit)
347 panic("assumption failed: hdr not split");
348 exthdrs.ip6e_dest2->m_next = m->m_next;
349 m->m_next = exthdrs.ip6e_dest2;
350 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
351 ip6->ip6_nxt = IPPROTO_DSTOPTS;
352 }
353
354 /*
355 * result: IPv6 hbh dest1 rthdr dest2 payload
356 * m will point to IPv6 header. mprev will point to the
357 * extension header prior to dest2 (rthdr in the above case).
358 */
359 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
360 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
361 IPPROTO_DSTOPTS);
362 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
363 IPPROTO_ROUTING);
364
365#ifdef IPSEC
366 if (!needipsec)
367 goto skip_ipsec2;
368
369 /*
370 * pointers after IPsec headers are not valid any more.
371 * other pointers need a great care too.
372 * (IPsec routines should not mangle mbufs prior to AH/ESP)
373 */
374 exthdrs.ip6e_dest2 = NULL;
375
376 if (exthdrs.ip6e_rthdr) {
377 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
378 segleft_org = rh->ip6r_segleft;
379 rh->ip6r_segleft = 0;
380 }
381
382 bzero(&state, sizeof(state));
383 state.m = m;
384 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
385 &needipsectun);
386 m = state.m;
387 if (error) {
388 /* mbuf is already reclaimed in ipsec6_output_trans. */
389 m = NULL;
390 switch (error) {
391 case EHOSTUNREACH:
392 case ENETUNREACH:
393 case EMSGSIZE:
394 case ENOBUFS:
395 case ENOMEM:
396 break;
397 default:
398 printf("[%s:%d] (ipsec): error code %d\n",
399 __func__, __LINE__, error);
400 /* FALLTHROUGH */
401 case ENOENT:
402 /* don't show these error codes to the user */
403 error = 0;
404 break;
405 }
406 goto bad;
407 } else if (!needipsectun) {
408 /*
409 * In the FAST IPSec case we have already
410 * re-injected the packet and it has been freed
411 * by the ipsec_done() function. So, just clean
412 * up after ourselves.
413 */
414 m = NULL;
415 goto done;
416 }
417 if (exthdrs.ip6e_rthdr) {
418 /* ah6_output doesn't modify mbuf chain */
419 rh->ip6r_segleft = segleft_org;
420 }
421skip_ipsec2:;
422#endif /* IPSEC */
423
424 /*
425 * If there is a routing header, replace the destination address field
426 * with the first hop of the routing header.
427 */
428 if (exthdrs.ip6e_rthdr) {
429 struct ip6_rthdr *rh =
430 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
431 struct ip6_rthdr *));
432 struct ip6_rthdr0 *rh0;
433 struct in6_addr *addr;
434 struct sockaddr_in6 sa;
435
436 switch (rh->ip6r_type) {
437 case IPV6_RTHDR_TYPE_0:
438 rh0 = (struct ip6_rthdr0 *)rh;
439 addr = (struct in6_addr *)(rh0 + 1);
440
441 /*
442 * construct a sockaddr_in6 form of
443 * the first hop.
444 *
445 * XXX: we may not have enough
446 * information about its scope zone;
447 * there is no standard API to pass
448 * the information from the
449 * application.
450 */
451 bzero(&sa, sizeof(sa));
452 sa.sin6_family = AF_INET6;
453 sa.sin6_len = sizeof(sa);
454 sa.sin6_addr = addr[0];
455 if ((error = sa6_embedscope(&sa,
456 ip6_use_defzone)) != 0) {
457 goto bad;
458 }
459 ip6->ip6_dst = sa.sin6_addr;
460 bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
461 * (rh0->ip6r0_segleft - 1));
462 addr[rh0->ip6r0_segleft - 1] = finaldst;
463 /* XXX */
464 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
465 break;
466 default: /* is it possible? */
467 error = EINVAL;
468 goto bad;
469 }
470 }
471
472 /* Source address validation */
473 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
474 (flags & IPV6_UNSPECSRC) == 0) {
475 error = EOPNOTSUPP;
476 ip6stat.ip6s_badscope++;
477 goto bad;
478 }
479 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
480 error = EOPNOTSUPP;
481 ip6stat.ip6s_badscope++;
482 goto bad;
483 }
484
485 ip6stat.ip6s_localout++;
486
487 /*
488 * Route packet.
489 */
490 if (ro == 0) {
491 ro = &ip6route;
492 bzero((caddr_t)ro, sizeof(*ro));
493 }
494 ro_pmtu = ro;
495 if (opt && opt->ip6po_rthdr)
496 ro = &opt->ip6po_route;
497 dst = (struct sockaddr_in6 *)&ro->ro_dst;
498
499again:
500 /*
501 * if specified, try to fill in the traffic class field.
502 * do not override if a non-zero value is already set.
503 * we check the diffserv field and the ecn field separately.
504 */
505 if (opt && opt->ip6po_tclass >= 0) {
506 int mask = 0;
507
508 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
509 mask |= 0xfc;
510 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
511 mask |= 0x03;
512 if (mask != 0)
513 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
514 }
515
516 /* fill in or override the hop limit field, if necessary. */
517 if (opt && opt->ip6po_hlim != -1)
518 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
519 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
520 if (im6o != NULL)
521 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
522 else
523 ip6->ip6_hlim = ip6_defmcasthlim;
524 }
525
526#ifdef IPSEC
527 /*
528 * We may re-inject packets into the stack here.
529 */
530 if (needipsec && needipsectun) {
531 struct ipsec_output_state state;
532
533 /*
534 * All the extension headers will become inaccessible
535 * (since they can be encrypted).
536 * Don't panic, we need no more updates to extension headers
537 * on inner IPv6 packet (since they are now encapsulated).
538 *
539 * IPv6 [ESP|AH] IPv6 [extension headers] payload
540 */
541 bzero(&exthdrs, sizeof(exthdrs));
542 exthdrs.ip6e_ip6 = m;
543
544 bzero(&state, sizeof(state));
545 state.m = m;
546 state.ro = (struct route *)ro;
547 state.dst = (struct sockaddr *)dst;
548
549 error = ipsec6_output_tunnel(&state, sp, flags);
550
551 m = state.m;
552 ro = (struct route_in6 *)state.ro;
553 dst = (struct sockaddr_in6 *)state.dst;
554 if (error) {
555 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
556 m0 = m = NULL;
557 m = NULL;
558 switch (error) {
559 case EHOSTUNREACH:
560 case ENETUNREACH:
561 case EMSGSIZE:
562 case ENOBUFS:
563 case ENOMEM:
564 break;
565 default:
566 printf("[%s:%d] (ipsec): error code %d\n",
567 __func__, __LINE__, error);
568 /* FALLTHROUGH */
569 case ENOENT:
570 /* don't show these error codes to the user */
571 error = 0;
572 break;
573 }
574 goto bad;
575 } else {
576 /*
577 * In the FAST IPSec case we have already
578 * re-injected the packet and it has been freed
579 * by the ipsec_done() function. So, just clean
580 * up after ourselves.
581 */
582 m = NULL;
583 goto done;
584 }
585
586 exthdrs.ip6e_ip6 = m;
587 }
588#endif /* IPSEC */
589
590 /* adjust pointer */
591 ip6 = mtod(m, struct ip6_hdr *);
592
593 bzero(&dst_sa, sizeof(dst_sa));
594 dst_sa.sin6_family = AF_INET6;
595 dst_sa.sin6_len = sizeof(dst_sa);
596 dst_sa.sin6_addr = ip6->ip6_dst;
597 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
598 &ifp, &rt, 0)) != 0) {
599 switch (error) {
600 case EHOSTUNREACH:
601 ip6stat.ip6s_noroute++;
602 break;
603 case EADDRNOTAVAIL:
604 default:
605 break; /* XXX statistics? */
606 }
607 if (ifp != NULL)
608 in6_ifstat_inc(ifp, ifs6_out_discard);
609 goto bad;
610 }
611 if (rt == NULL) {
612 /*
613 * If in6_selectroute() does not return a route entry,
614 * dst may not have been updated.
615 */
616 *dst = dst_sa; /* XXX */
617 }
618
619 /*
620 * then rt (for unicast) and ifp must be non-NULL valid values.
621 */
622 if ((flags & IPV6_FORWARDING) == 0) {
623 /* XXX: the FORWARDING flag can be set for mrouting. */
624 in6_ifstat_inc(ifp, ifs6_out_request);
625 }
626 if (rt != NULL) {
627 ia = (struct in6_ifaddr *)(rt->rt_ifa);
628 rt->rt_use++;
629 }
630
631 /*
632 * The outgoing interface must be in the zone of source and
633 * destination addresses. We should use ia_ifp to support the
634 * case of sending packets to an address of our own.
635 */
636 if (ia != NULL && ia->ia_ifp)
637 origifp = ia->ia_ifp;
638 else
639 origifp = ifp;
640
641 src0 = ip6->ip6_src;
642 if (in6_setscope(&src0, origifp, &zone))
643 goto badscope;
644 bzero(&src_sa, sizeof(src_sa));
645 src_sa.sin6_family = AF_INET6;
646 src_sa.sin6_len = sizeof(src_sa);
647 src_sa.sin6_addr = ip6->ip6_src;
648 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
649 goto badscope;
650
651 dst0 = ip6->ip6_dst;
652 if (in6_setscope(&dst0, origifp, &zone))
653 goto badscope;
654 /* re-initialize to be sure */
655 bzero(&dst_sa, sizeof(dst_sa));
656 dst_sa.sin6_family = AF_INET6;
657 dst_sa.sin6_len = sizeof(dst_sa);
658 dst_sa.sin6_addr = ip6->ip6_dst;
659 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
660 goto badscope;
661 }
662
663 /* scope check is done. */
664 goto routefound;
665
666 badscope:
667 ip6stat.ip6s_badscope++;
668 in6_ifstat_inc(origifp, ifs6_out_discard);
669 if (error == 0)
670 error = EHOSTUNREACH; /* XXX */
671 goto bad;
672
673 routefound:
674 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
675 if (opt && opt->ip6po_nextroute.ro_rt) {
676 /*
677 * The nexthop is explicitly specified by the
678 * application. We assume the next hop is an IPv6
679 * address.
680 */
681 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
682 }
683 else if ((rt->rt_flags & RTF_GATEWAY))
684 dst = (struct sockaddr_in6 *)rt->rt_gateway;
685 }
686
687 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
688 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
689 } else {
690 struct in6_multi *in6m;
691
692 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
693
694 in6_ifstat_inc(ifp, ifs6_out_mcast);
695
696 /*
697 * Confirm that the outgoing interface supports multicast.
698 */
699 if (!(ifp->if_flags & IFF_MULTICAST)) {
700 ip6stat.ip6s_noroute++;
701 in6_ifstat_inc(ifp, ifs6_out_discard);
702 error = ENETUNREACH;
703 goto bad;
704 }
705 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
706 if (in6m != NULL &&
707 (im6o == NULL || im6o->im6o_multicast_loop)) {
708 /*
709 * If we belong to the destination multicast group
710 * on the outgoing interface, and the caller did not
711 * forbid loopback, loop back a copy.
712 */
713 ip6_mloopback(ifp, m, dst);
714 } else {
715 /*
716 * If we are acting as a multicast router, perform
717 * multicast forwarding as if the packet had just
718 * arrived on the interface to which we are about
719 * to send. The multicast forwarding function
720 * recursively calls this function, using the
721 * IPV6_FORWARDING flag to prevent infinite recursion.
722 *
723 * Multicasts that are looped back by ip6_mloopback(),
724 * above, will be forwarded by the ip6_input() routine,
725 * if necessary.
726 */
727 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
728 /*
729 * XXX: ip6_mforward expects that rcvif is NULL
730 * when it is called from the originating path.
731 * However, it is not always the case, since
732 * some versions of MGETHDR() does not
733 * initialize the field.
734 */
735 m->m_pkthdr.rcvif = NULL;
736 if (ip6_mforward(ip6, ifp, m) != 0) {
737 m_freem(m);
738 goto done;
739 }
740 }
741 }
742 /*
743 * Multicasts with a hoplimit of zero may be looped back,
744 * above, but must not be transmitted on a network.
745 * Also, multicasts addressed to the loopback interface
746 * are not sent -- the above call to ip6_mloopback() will
747 * loop back a copy if this host actually belongs to the
748 * destination group on the loopback interface.
749 */
750 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
751 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
752 m_freem(m);
753 goto done;
754 }
755 }
756
757 /*
758 * Fill the outgoing inteface to tell the upper layer
759 * to increment per-interface statistics.
760 */
761 if (ifpp)
762 *ifpp = ifp;
763
764 /* Determine path MTU. */
765 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
766 &alwaysfrag)) != 0)
767 goto bad;
768
769 /*
770 * The caller of this function may specify to use the minimum MTU
771 * in some cases.
772 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
773 * setting. The logic is a bit complicated; by default, unicast
774 * packets will follow path MTU while multicast packets will be sent at
775 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
776 * including unicast ones will be sent at the minimum MTU. Multicast
777 * packets will always be sent at the minimum MTU unless
778 * IP6PO_MINMTU_DISABLE is explicitly specified.
779 * See RFC 3542 for more details.
780 */
781 if (mtu > IPV6_MMTU) {
782 if ((flags & IPV6_MINMTU))
783 mtu = IPV6_MMTU;
784 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
785 mtu = IPV6_MMTU;
786 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
787 (opt == NULL ||
788 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
789 mtu = IPV6_MMTU;
790 }
791 }
792
793 /*
794 * clear embedded scope identifiers if necessary.
795 * in6_clearscope will touch the addresses only when necessary.
796 */
797 in6_clearscope(&ip6->ip6_src);
798 in6_clearscope(&ip6->ip6_dst);
799
800 /*
801 * If the outgoing packet contains a hop-by-hop options header,
802 * it must be examined and processed even by the source node.
803 * (RFC 2460, section 4.)
804 */
805 if (exthdrs.ip6e_hbh) {
806 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
807 u_int32_t dummy; /* XXX unused */
808 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
809
810#ifdef DIAGNOSTIC
811 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
812 panic("ip6e_hbh is not continuous");
813#endif
814 /*
815 * XXX: if we have to send an ICMPv6 error to the sender,
816 * we need the M_LOOP flag since icmp6_error() expects
817 * the IPv6 and the hop-by-hop options header are
818 * continuous unless the flag is set.
819 */
820 m->m_flags |= M_LOOP;
821 m->m_pkthdr.rcvif = ifp;
822 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
823 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
824 &dummy, &plen) < 0) {
825 /* m was already freed at this point */
826 error = EINVAL;/* better error? */
827 goto done;
828 }
829 m->m_flags &= ~M_LOOP; /* XXX */
830 m->m_pkthdr.rcvif = NULL;
831 }
832
833 /* Jump over all PFIL processing if hooks are not active. */
834 if (!PFIL_HOOKED(&inet6_pfil_hook))
835 goto passout;
836
837 odst = ip6->ip6_dst;
838 /* Run through list of hooks for output packets. */
839 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
840 if (error != 0 || m == NULL)
841 goto done;
842 ip6 = mtod(m, struct ip6_hdr *);
843
844 /* See if destination IP address was changed by packet filter. */
845 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
846 m->m_flags |= M_SKIP_FIREWALL;
847 /* If destination is now ourself drop to ip6_input(). */
848 if (in6_localaddr(&ip6->ip6_dst)) {
849 if (m->m_pkthdr.rcvif == NULL)
850 m->m_pkthdr.rcvif = loif;
851 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
852 m->m_pkthdr.csum_flags |=
853 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
854 m->m_pkthdr.csum_data = 0xffff;
855 }
856 m->m_pkthdr.csum_flags |=
857 CSUM_IP_CHECKED | CSUM_IP_VALID;
858 error = netisr_queue(NETISR_IPV6, m);
859 goto done;
860 } else
861 goto again; /* Redo the routing table lookup. */
862 }
863
864 /* XXX: IPFIREWALL_FORWARD */
865
866passout:
867 /*
868 * Send the packet to the outgoing interface.
869 * If necessary, do IPv6 fragmentation before sending.
870 *
871 * the logic here is rather complex:
872 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
873 * 1-a: send as is if tlen <= path mtu
874 * 1-b: fragment if tlen > path mtu
875 *
876 * 2: if user asks us not to fragment (dontfrag == 1)
877 * 2-a: send as is if tlen <= interface mtu
878 * 2-b: error if tlen > interface mtu
879 *
880 * 3: if we always need to attach fragment header (alwaysfrag == 1)
881 * always fragment
882 *
883 * 4: if dontfrag == 1 && alwaysfrag == 1
884 * error, as we cannot handle this conflicting request
885 */
886 tlen = m->m_pkthdr.len;
887
888 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
889 dontfrag = 1;
890 else
891 dontfrag = 0;
892 if (dontfrag && alwaysfrag) { /* case 4 */
893 /* conflicting request - can't transmit */
894 error = EMSGSIZE;
895 goto bad;
896 }
897 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */
898 /*
899 * Even if the DONTFRAG option is specified, we cannot send the
900 * packet when the data length is larger than the MTU of the
901 * outgoing interface.
902 * Notify the error by sending IPV6_PATHMTU ancillary data as
903 * well as returning an error code (the latter is not described
904 * in the API spec.)
905 */
906 u_int32_t mtu32;
907 struct ip6ctlparam ip6cp;
908
909 mtu32 = (u_int32_t)mtu;
910 bzero(&ip6cp, sizeof(ip6cp));
911 ip6cp.ip6c_cmdarg = (void *)&mtu32;
912 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
913 (void *)&ip6cp);
914
915 error = EMSGSIZE;
916 goto bad;
917 }
918
919 /*
920 * transmit packet without fragmentation
921 */
922 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
923 struct in6_ifaddr *ia6;
924
925 ip6 = mtod(m, struct ip6_hdr *);
926 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
927 if (ia6) {
928 /* Record statistics for this interface address. */
929 ia6->ia_ifa.if_opackets++;
930 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
931 }
932 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
933 goto done;
934 }
935
936 /*
937 * try to fragment the packet. case 1-b and 3
938 */
939 if (mtu < IPV6_MMTU) {
940 /* path MTU cannot be less than IPV6_MMTU */
941 error = EMSGSIZE;
942 in6_ifstat_inc(ifp, ifs6_out_fragfail);
943 goto bad;
944 } else if (ip6->ip6_plen == 0) {
945 /* jumbo payload cannot be fragmented */
946 error = EMSGSIZE;
947 in6_ifstat_inc(ifp, ifs6_out_fragfail);
948 goto bad;
949 } else {
950 struct mbuf **mnext, *m_frgpart;
951 struct ip6_frag *ip6f;
952 u_int32_t id = htonl(ip6_randomid());
953 u_char nextproto;
954
955 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
956
957 /*
958 * Too large for the destination or interface;
959 * fragment if possible.
960 * Must be able to put at least 8 bytes per fragment.
961 */
962 hlen = unfragpartlen;
963 if (mtu > IPV6_MAXPACKET)
964 mtu = IPV6_MAXPACKET;
965
966 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
967 if (len < 8) {
968 error = EMSGSIZE;
969 in6_ifstat_inc(ifp, ifs6_out_fragfail);
970 goto bad;
971 }
972
973 /*
974 * Verify that we have any chance at all of being able to queue
975 * the packet or packet fragments
976 */
977 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
978 < tlen /* - hlen */)) {
979 error = ENOBUFS;
980 ip6stat.ip6s_odropped++;
981 goto bad;
982 }
983
984 mnext = &m->m_nextpkt;
985
986 /*
987 * Change the next header field of the last header in the
988 * unfragmentable part.
989 */
990 if (exthdrs.ip6e_rthdr) {
991 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
992 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
993 } else if (exthdrs.ip6e_dest1) {
994 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
995 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
996 } else if (exthdrs.ip6e_hbh) {
997 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
998 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
999 } else {
1000 nextproto = ip6->ip6_nxt;
1001 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1002 }
1003
1004 /*
1005 * Loop through length of segment after first fragment,
1006 * make new header and copy data of each part and link onto
1007 * chain.
1008 */
1009 m0 = m;
1010 for (off = hlen; off < tlen; off += len) {
1011 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1012 if (!m) {
1013 error = ENOBUFS;
1014 ip6stat.ip6s_odropped++;
1015 goto sendorfree;
1016 }
1017 m->m_pkthdr.rcvif = NULL;
1018 m->m_flags = m0->m_flags & M_COPYFLAGS;
1019 *mnext = m;
1020 mnext = &m->m_nextpkt;
1021 m->m_data += max_linkhdr;
1022 mhip6 = mtod(m, struct ip6_hdr *);
1023 *mhip6 = *ip6;
1024 m->m_len = sizeof(*mhip6);
1025 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1026 if (error) {
1027 ip6stat.ip6s_odropped++;
1028 goto sendorfree;
1029 }
1030 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1031 if (off + len >= tlen)
1032 len = tlen - off;
1033 else
1034 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1035 mhip6->ip6_plen = htons((u_short)(len + hlen +
1036 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1037 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1038 error = ENOBUFS;
1039 ip6stat.ip6s_odropped++;
1040 goto sendorfree;
1041 }
1042 m_cat(m, m_frgpart);
1043 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1044 m->m_pkthdr.rcvif = NULL;
1045 ip6f->ip6f_reserved = 0;
1046 ip6f->ip6f_ident = id;
1047 ip6f->ip6f_nxt = nextproto;
1048 ip6stat.ip6s_ofragments++;
1049 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1050 }
1051
1052 in6_ifstat_inc(ifp, ifs6_out_fragok);
1053 }
1054
1055 /*
1056 * Remove leading garbages.
1057 */
1058sendorfree:
1059 m = m0->m_nextpkt;
1060 m0->m_nextpkt = 0;
1061 m_freem(m0);
1062 for (m0 = m; m; m = m0) {
1063 m0 = m->m_nextpkt;
1064 m->m_nextpkt = 0;
1065 if (error == 0) {
1066 /* Record statistics for this interface address. */
1067 if (ia) {
1068 ia->ia_ifa.if_opackets++;
1069 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1070 }
1071 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1072 } else
1073 m_freem(m);
1074 }
1075
1076 if (error == 0)
1077 ip6stat.ip6s_fragmented++;
1078
1079done:
1080 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1081 RTFREE(ro->ro_rt);
1082 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1083 RTFREE(ro_pmtu->ro_rt);
1084 }
| 65
66#include "opt_inet.h"
67#include "opt_inet6.h"
68#include "opt_ipsec.h"
69
70#include <sys/param.h>
71#include <sys/kernel.h>
72#include <sys/malloc.h>
73#include <sys/mbuf.h>
74#include <sys/errno.h>
75#include <sys/priv.h>
76#include <sys/proc.h>
77#include <sys/protosw.h>
78#include <sys/socket.h>
79#include <sys/socketvar.h>
80#include <sys/ucred.h>
81
82#include <net/if.h>
83#include <net/netisr.h>
84#include <net/route.h>
85#include <net/pfil.h>
86
87#include <netinet/in.h>
88#include <netinet/in_var.h>
89#include <netinet6/in6_var.h>
90#include <netinet/ip6.h>
91#include <netinet/icmp6.h>
92#include <netinet6/ip6_var.h>
93#include <netinet/in_pcb.h>
94#include <netinet/tcp_var.h>
95#include <netinet6/nd6.h>
96
97#ifdef IPSEC
98#include <netipsec/ipsec.h>
99#include <netipsec/ipsec6.h>
100#include <netipsec/key.h>
101#include <netinet6/ip6_ipsec.h>
102#endif /* IPSEC */
103
104#include <netinet6/ip6protosw.h>
105#include <netinet6/scope6_var.h>
106
107static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
108
109struct 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
117static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
118 struct ucred *, int));
119static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
120 struct socket *, struct sockopt *));
121static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
122static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
123 struct ucred *, int, int, int));
124
125static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
126static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
127static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
128static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
129 struct ip6_frag **));
130static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
131static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
132static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
133 struct ifnet *, struct in6_addr *, u_long *, int *));
134static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
135
136
137/*
138 * Make an extension header from option data. hp is the source, and
139 * mp is the destination.
140 */
141#define MAKE_EXTHDR(hp, mp) \
142 do { \
143 if (hp) { \
144 struct ip6_ext *eh = (struct ip6_ext *)(hp); \
145 error = ip6_copyexthdr((mp), (caddr_t)(hp), \
146 ((eh)->ip6e_len + 1) << 3); \
147 if (error) \
148 goto freehdrs; \
149 } \
150 } while (/*CONSTCOND*/ 0)
151
152/*
153 * Form a chain of extension headers.
154 * m is the extension header mbuf
155 * mp is the previous mbuf in the chain
156 * p is the next header
157 * i is the type of option.
158 */
159#define MAKE_CHAIN(m, mp, p, i)\
160 do {\
161 if (m) {\
162 if (!hdrsplit) \
163 panic("assumption failed: hdr not split"); \
164 *mtod((m), u_char *) = *(p);\
165 *(p) = (i);\
166 p = mtod((m), u_char *);\
167 (m)->m_next = (mp)->m_next;\
168 (mp)->m_next = (m);\
169 (mp) = (m);\
170 }\
171 } while (/*CONSTCOND*/ 0)
172
173/*
174 * IP6 output. The packet in mbuf chain m contains a skeletal IP6
175 * header (with pri, len, nxt, hlim, src, dst).
176 * This function may modify ver and hlim only.
177 * The mbuf chain containing the packet will be freed.
178 * The mbuf opt, if present, will not be freed.
179 *
180 * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
181 * nd_ifinfo.linkmtu is u_int32_t. so we use u_long to hold largest one,
182 * which is rt_rmx.rmx_mtu.
183 *
184 * ifpp - XXX: just for statistics
185 */
186int
187ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
188 struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
189 struct ifnet **ifpp, struct inpcb *inp)
190{
191 struct ip6_hdr *ip6, *mhip6;
192 struct ifnet *ifp, *origifp;
193 struct mbuf *m = m0;
194 struct mbuf *mprev = NULL;
195 int hlen, tlen, len, off;
196 struct route_in6 ip6route;
197 struct rtentry *rt = NULL;
198 struct sockaddr_in6 *dst, src_sa, dst_sa;
199 struct in6_addr odst;
200 int error = 0;
201 struct in6_ifaddr *ia = NULL;
202 u_long mtu;
203 int alwaysfrag, dontfrag;
204 u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
205 struct ip6_exthdrs exthdrs;
206 struct in6_addr finaldst, src0, dst0;
207 u_int32_t zone;
208 struct route_in6 *ro_pmtu = NULL;
209 int hdrsplit = 0;
210 int needipsec = 0;
211#ifdef IPSEC
212 struct ipsec_output_state state;
213 struct ip6_rthdr *rh = NULL;
214 int needipsectun = 0;
215 int segleft_org = 0;
216 struct secpolicy *sp = NULL;
217#endif /* IPSEC */
218
219 ip6 = mtod(m, struct ip6_hdr *);
220 if (ip6 == NULL) {
221 printf ("ip6 is NULL");
222 goto bad;
223 }
224
225 finaldst = ip6->ip6_dst;
226
227 bzero(&exthdrs, sizeof(exthdrs));
228
229 if (opt) {
230 /* Hop-by-Hop options header */
231 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
232 /* Destination options header(1st part) */
233 if (opt->ip6po_rthdr) {
234 /*
235 * Destination options header(1st part)
236 * This only makes sense with a routing header.
237 * See Section 9.2 of RFC 3542.
238 * Disabling this part just for MIP6 convenience is
239 * a bad idea. We need to think carefully about a
240 * way to make the advanced API coexist with MIP6
241 * options, which might automatically be inserted in
242 * the kernel.
243 */
244 MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
245 }
246 /* Routing header */
247 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
248 /* Destination options header(2nd part) */
249 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
250 }
251
252 /*
253 * IPSec checking which handles several cases.
254 * FAST IPSEC: We re-injected the packet.
255 */
256#ifdef IPSEC
257 switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
258 {
259 case 1: /* Bad packet */
260 goto freehdrs;
261 case -1: /* Do IPSec */
262 needipsec = 1;
263 case 0: /* No IPSec */
264 default:
265 break;
266 }
267#endif /* IPSEC */
268
269 /*
270 * Calculate the total length of the extension header chain.
271 * Keep the length of the unfragmentable part for fragmentation.
272 */
273 optlen = 0;
274 if (exthdrs.ip6e_hbh)
275 optlen += exthdrs.ip6e_hbh->m_len;
276 if (exthdrs.ip6e_dest1)
277 optlen += exthdrs.ip6e_dest1->m_len;
278 if (exthdrs.ip6e_rthdr)
279 optlen += exthdrs.ip6e_rthdr->m_len;
280 unfragpartlen = optlen + sizeof(struct ip6_hdr);
281
282 /* NOTE: we don't add AH/ESP length here. do that later. */
283 if (exthdrs.ip6e_dest2)
284 optlen += exthdrs.ip6e_dest2->m_len;
285
286 /*
287 * If we need IPsec, or there is at least one extension header,
288 * separate IP6 header from the payload.
289 */
290 if ((needipsec || optlen) && !hdrsplit) {
291 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
292 m = NULL;
293 goto freehdrs;
294 }
295 m = exthdrs.ip6e_ip6;
296 hdrsplit++;
297 }
298
299 /* adjust pointer */
300 ip6 = mtod(m, struct ip6_hdr *);
301
302 /* adjust mbuf packet header length */
303 m->m_pkthdr.len += optlen;
304 plen = m->m_pkthdr.len - sizeof(*ip6);
305
306 /* If this is a jumbo payload, insert a jumbo payload option. */
307 if (plen > IPV6_MAXPACKET) {
308 if (!hdrsplit) {
309 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
310 m = NULL;
311 goto freehdrs;
312 }
313 m = exthdrs.ip6e_ip6;
314 hdrsplit++;
315 }
316 /* adjust pointer */
317 ip6 = mtod(m, struct ip6_hdr *);
318 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
319 goto freehdrs;
320 ip6->ip6_plen = 0;
321 } else
322 ip6->ip6_plen = htons(plen);
323
324 /*
325 * Concatenate headers and fill in next header fields.
326 * Here we have, on "m"
327 * IPv6 payload
328 * and we insert headers accordingly. Finally, we should be getting:
329 * IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
330 *
331 * during the header composing process, "m" points to IPv6 header.
332 * "mprev" points to an extension header prior to esp.
333 */
334 u_char *nexthdrp = &ip6->ip6_nxt;
335 mprev = m;
336
337 /*
338 * we treat dest2 specially. this makes IPsec processing
339 * much easier. the goal here is to make mprev point the
340 * mbuf prior to dest2.
341 *
342 * result: IPv6 dest2 payload
343 * m and mprev will point to IPv6 header.
344 */
345 if (exthdrs.ip6e_dest2) {
346 if (!hdrsplit)
347 panic("assumption failed: hdr not split");
348 exthdrs.ip6e_dest2->m_next = m->m_next;
349 m->m_next = exthdrs.ip6e_dest2;
350 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
351 ip6->ip6_nxt = IPPROTO_DSTOPTS;
352 }
353
354 /*
355 * result: IPv6 hbh dest1 rthdr dest2 payload
356 * m will point to IPv6 header. mprev will point to the
357 * extension header prior to dest2 (rthdr in the above case).
358 */
359 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
360 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
361 IPPROTO_DSTOPTS);
362 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
363 IPPROTO_ROUTING);
364
365#ifdef IPSEC
366 if (!needipsec)
367 goto skip_ipsec2;
368
369 /*
370 * pointers after IPsec headers are not valid any more.
371 * other pointers need a great care too.
372 * (IPsec routines should not mangle mbufs prior to AH/ESP)
373 */
374 exthdrs.ip6e_dest2 = NULL;
375
376 if (exthdrs.ip6e_rthdr) {
377 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
378 segleft_org = rh->ip6r_segleft;
379 rh->ip6r_segleft = 0;
380 }
381
382 bzero(&state, sizeof(state));
383 state.m = m;
384 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
385 &needipsectun);
386 m = state.m;
387 if (error) {
388 /* mbuf is already reclaimed in ipsec6_output_trans. */
389 m = NULL;
390 switch (error) {
391 case EHOSTUNREACH:
392 case ENETUNREACH:
393 case EMSGSIZE:
394 case ENOBUFS:
395 case ENOMEM:
396 break;
397 default:
398 printf("[%s:%d] (ipsec): error code %d\n",
399 __func__, __LINE__, error);
400 /* FALLTHROUGH */
401 case ENOENT:
402 /* don't show these error codes to the user */
403 error = 0;
404 break;
405 }
406 goto bad;
407 } else if (!needipsectun) {
408 /*
409 * In the FAST IPSec case we have already
410 * re-injected the packet and it has been freed
411 * by the ipsec_done() function. So, just clean
412 * up after ourselves.
413 */
414 m = NULL;
415 goto done;
416 }
417 if (exthdrs.ip6e_rthdr) {
418 /* ah6_output doesn't modify mbuf chain */
419 rh->ip6r_segleft = segleft_org;
420 }
421skip_ipsec2:;
422#endif /* IPSEC */
423
424 /*
425 * If there is a routing header, replace the destination address field
426 * with the first hop of the routing header.
427 */
428 if (exthdrs.ip6e_rthdr) {
429 struct ip6_rthdr *rh =
430 (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
431 struct ip6_rthdr *));
432 struct ip6_rthdr0 *rh0;
433 struct in6_addr *addr;
434 struct sockaddr_in6 sa;
435
436 switch (rh->ip6r_type) {
437 case IPV6_RTHDR_TYPE_0:
438 rh0 = (struct ip6_rthdr0 *)rh;
439 addr = (struct in6_addr *)(rh0 + 1);
440
441 /*
442 * construct a sockaddr_in6 form of
443 * the first hop.
444 *
445 * XXX: we may not have enough
446 * information about its scope zone;
447 * there is no standard API to pass
448 * the information from the
449 * application.
450 */
451 bzero(&sa, sizeof(sa));
452 sa.sin6_family = AF_INET6;
453 sa.sin6_len = sizeof(sa);
454 sa.sin6_addr = addr[0];
455 if ((error = sa6_embedscope(&sa,
456 ip6_use_defzone)) != 0) {
457 goto bad;
458 }
459 ip6->ip6_dst = sa.sin6_addr;
460 bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
461 * (rh0->ip6r0_segleft - 1));
462 addr[rh0->ip6r0_segleft - 1] = finaldst;
463 /* XXX */
464 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
465 break;
466 default: /* is it possible? */
467 error = EINVAL;
468 goto bad;
469 }
470 }
471
472 /* Source address validation */
473 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
474 (flags & IPV6_UNSPECSRC) == 0) {
475 error = EOPNOTSUPP;
476 ip6stat.ip6s_badscope++;
477 goto bad;
478 }
479 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
480 error = EOPNOTSUPP;
481 ip6stat.ip6s_badscope++;
482 goto bad;
483 }
484
485 ip6stat.ip6s_localout++;
486
487 /*
488 * Route packet.
489 */
490 if (ro == 0) {
491 ro = &ip6route;
492 bzero((caddr_t)ro, sizeof(*ro));
493 }
494 ro_pmtu = ro;
495 if (opt && opt->ip6po_rthdr)
496 ro = &opt->ip6po_route;
497 dst = (struct sockaddr_in6 *)&ro->ro_dst;
498
499again:
500 /*
501 * if specified, try to fill in the traffic class field.
502 * do not override if a non-zero value is already set.
503 * we check the diffserv field and the ecn field separately.
504 */
505 if (opt && opt->ip6po_tclass >= 0) {
506 int mask = 0;
507
508 if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
509 mask |= 0xfc;
510 if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
511 mask |= 0x03;
512 if (mask != 0)
513 ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
514 }
515
516 /* fill in or override the hop limit field, if necessary. */
517 if (opt && opt->ip6po_hlim != -1)
518 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
519 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
520 if (im6o != NULL)
521 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
522 else
523 ip6->ip6_hlim = ip6_defmcasthlim;
524 }
525
526#ifdef IPSEC
527 /*
528 * We may re-inject packets into the stack here.
529 */
530 if (needipsec && needipsectun) {
531 struct ipsec_output_state state;
532
533 /*
534 * All the extension headers will become inaccessible
535 * (since they can be encrypted).
536 * Don't panic, we need no more updates to extension headers
537 * on inner IPv6 packet (since they are now encapsulated).
538 *
539 * IPv6 [ESP|AH] IPv6 [extension headers] payload
540 */
541 bzero(&exthdrs, sizeof(exthdrs));
542 exthdrs.ip6e_ip6 = m;
543
544 bzero(&state, sizeof(state));
545 state.m = m;
546 state.ro = (struct route *)ro;
547 state.dst = (struct sockaddr *)dst;
548
549 error = ipsec6_output_tunnel(&state, sp, flags);
550
551 m = state.m;
552 ro = (struct route_in6 *)state.ro;
553 dst = (struct sockaddr_in6 *)state.dst;
554 if (error) {
555 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
556 m0 = m = NULL;
557 m = NULL;
558 switch (error) {
559 case EHOSTUNREACH:
560 case ENETUNREACH:
561 case EMSGSIZE:
562 case ENOBUFS:
563 case ENOMEM:
564 break;
565 default:
566 printf("[%s:%d] (ipsec): error code %d\n",
567 __func__, __LINE__, error);
568 /* FALLTHROUGH */
569 case ENOENT:
570 /* don't show these error codes to the user */
571 error = 0;
572 break;
573 }
574 goto bad;
575 } else {
576 /*
577 * In the FAST IPSec case we have already
578 * re-injected the packet and it has been freed
579 * by the ipsec_done() function. So, just clean
580 * up after ourselves.
581 */
582 m = NULL;
583 goto done;
584 }
585
586 exthdrs.ip6e_ip6 = m;
587 }
588#endif /* IPSEC */
589
590 /* adjust pointer */
591 ip6 = mtod(m, struct ip6_hdr *);
592
593 bzero(&dst_sa, sizeof(dst_sa));
594 dst_sa.sin6_family = AF_INET6;
595 dst_sa.sin6_len = sizeof(dst_sa);
596 dst_sa.sin6_addr = ip6->ip6_dst;
597 if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
598 &ifp, &rt, 0)) != 0) {
599 switch (error) {
600 case EHOSTUNREACH:
601 ip6stat.ip6s_noroute++;
602 break;
603 case EADDRNOTAVAIL:
604 default:
605 break; /* XXX statistics? */
606 }
607 if (ifp != NULL)
608 in6_ifstat_inc(ifp, ifs6_out_discard);
609 goto bad;
610 }
611 if (rt == NULL) {
612 /*
613 * If in6_selectroute() does not return a route entry,
614 * dst may not have been updated.
615 */
616 *dst = dst_sa; /* XXX */
617 }
618
619 /*
620 * then rt (for unicast) and ifp must be non-NULL valid values.
621 */
622 if ((flags & IPV6_FORWARDING) == 0) {
623 /* XXX: the FORWARDING flag can be set for mrouting. */
624 in6_ifstat_inc(ifp, ifs6_out_request);
625 }
626 if (rt != NULL) {
627 ia = (struct in6_ifaddr *)(rt->rt_ifa);
628 rt->rt_use++;
629 }
630
631 /*
632 * The outgoing interface must be in the zone of source and
633 * destination addresses. We should use ia_ifp to support the
634 * case of sending packets to an address of our own.
635 */
636 if (ia != NULL && ia->ia_ifp)
637 origifp = ia->ia_ifp;
638 else
639 origifp = ifp;
640
641 src0 = ip6->ip6_src;
642 if (in6_setscope(&src0, origifp, &zone))
643 goto badscope;
644 bzero(&src_sa, sizeof(src_sa));
645 src_sa.sin6_family = AF_INET6;
646 src_sa.sin6_len = sizeof(src_sa);
647 src_sa.sin6_addr = ip6->ip6_src;
648 if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
649 goto badscope;
650
651 dst0 = ip6->ip6_dst;
652 if (in6_setscope(&dst0, origifp, &zone))
653 goto badscope;
654 /* re-initialize to be sure */
655 bzero(&dst_sa, sizeof(dst_sa));
656 dst_sa.sin6_family = AF_INET6;
657 dst_sa.sin6_len = sizeof(dst_sa);
658 dst_sa.sin6_addr = ip6->ip6_dst;
659 if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
660 goto badscope;
661 }
662
663 /* scope check is done. */
664 goto routefound;
665
666 badscope:
667 ip6stat.ip6s_badscope++;
668 in6_ifstat_inc(origifp, ifs6_out_discard);
669 if (error == 0)
670 error = EHOSTUNREACH; /* XXX */
671 goto bad;
672
673 routefound:
674 if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
675 if (opt && opt->ip6po_nextroute.ro_rt) {
676 /*
677 * The nexthop is explicitly specified by the
678 * application. We assume the next hop is an IPv6
679 * address.
680 */
681 dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
682 }
683 else if ((rt->rt_flags & RTF_GATEWAY))
684 dst = (struct sockaddr_in6 *)rt->rt_gateway;
685 }
686
687 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
688 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
689 } else {
690 struct in6_multi *in6m;
691
692 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
693
694 in6_ifstat_inc(ifp, ifs6_out_mcast);
695
696 /*
697 * Confirm that the outgoing interface supports multicast.
698 */
699 if (!(ifp->if_flags & IFF_MULTICAST)) {
700 ip6stat.ip6s_noroute++;
701 in6_ifstat_inc(ifp, ifs6_out_discard);
702 error = ENETUNREACH;
703 goto bad;
704 }
705 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
706 if (in6m != NULL &&
707 (im6o == NULL || im6o->im6o_multicast_loop)) {
708 /*
709 * If we belong to the destination multicast group
710 * on the outgoing interface, and the caller did not
711 * forbid loopback, loop back a copy.
712 */
713 ip6_mloopback(ifp, m, dst);
714 } else {
715 /*
716 * If we are acting as a multicast router, perform
717 * multicast forwarding as if the packet had just
718 * arrived on the interface to which we are about
719 * to send. The multicast forwarding function
720 * recursively calls this function, using the
721 * IPV6_FORWARDING flag to prevent infinite recursion.
722 *
723 * Multicasts that are looped back by ip6_mloopback(),
724 * above, will be forwarded by the ip6_input() routine,
725 * if necessary.
726 */
727 if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
728 /*
729 * XXX: ip6_mforward expects that rcvif is NULL
730 * when it is called from the originating path.
731 * However, it is not always the case, since
732 * some versions of MGETHDR() does not
733 * initialize the field.
734 */
735 m->m_pkthdr.rcvif = NULL;
736 if (ip6_mforward(ip6, ifp, m) != 0) {
737 m_freem(m);
738 goto done;
739 }
740 }
741 }
742 /*
743 * Multicasts with a hoplimit of zero may be looped back,
744 * above, but must not be transmitted on a network.
745 * Also, multicasts addressed to the loopback interface
746 * are not sent -- the above call to ip6_mloopback() will
747 * loop back a copy if this host actually belongs to the
748 * destination group on the loopback interface.
749 */
750 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
751 IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
752 m_freem(m);
753 goto done;
754 }
755 }
756
757 /*
758 * Fill the outgoing inteface to tell the upper layer
759 * to increment per-interface statistics.
760 */
761 if (ifpp)
762 *ifpp = ifp;
763
764 /* Determine path MTU. */
765 if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
766 &alwaysfrag)) != 0)
767 goto bad;
768
769 /*
770 * The caller of this function may specify to use the minimum MTU
771 * in some cases.
772 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
773 * setting. The logic is a bit complicated; by default, unicast
774 * packets will follow path MTU while multicast packets will be sent at
775 * the minimum MTU. If IP6PO_MINMTU_ALL is specified, all packets
776 * including unicast ones will be sent at the minimum MTU. Multicast
777 * packets will always be sent at the minimum MTU unless
778 * IP6PO_MINMTU_DISABLE is explicitly specified.
779 * See RFC 3542 for more details.
780 */
781 if (mtu > IPV6_MMTU) {
782 if ((flags & IPV6_MINMTU))
783 mtu = IPV6_MMTU;
784 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
785 mtu = IPV6_MMTU;
786 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
787 (opt == NULL ||
788 opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
789 mtu = IPV6_MMTU;
790 }
791 }
792
793 /*
794 * clear embedded scope identifiers if necessary.
795 * in6_clearscope will touch the addresses only when necessary.
796 */
797 in6_clearscope(&ip6->ip6_src);
798 in6_clearscope(&ip6->ip6_dst);
799
800 /*
801 * If the outgoing packet contains a hop-by-hop options header,
802 * it must be examined and processed even by the source node.
803 * (RFC 2460, section 4.)
804 */
805 if (exthdrs.ip6e_hbh) {
806 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
807 u_int32_t dummy; /* XXX unused */
808 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
809
810#ifdef DIAGNOSTIC
811 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
812 panic("ip6e_hbh is not continuous");
813#endif
814 /*
815 * XXX: if we have to send an ICMPv6 error to the sender,
816 * we need the M_LOOP flag since icmp6_error() expects
817 * the IPv6 and the hop-by-hop options header are
818 * continuous unless the flag is set.
819 */
820 m->m_flags |= M_LOOP;
821 m->m_pkthdr.rcvif = ifp;
822 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
823 ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
824 &dummy, &plen) < 0) {
825 /* m was already freed at this point */
826 error = EINVAL;/* better error? */
827 goto done;
828 }
829 m->m_flags &= ~M_LOOP; /* XXX */
830 m->m_pkthdr.rcvif = NULL;
831 }
832
833 /* Jump over all PFIL processing if hooks are not active. */
834 if (!PFIL_HOOKED(&inet6_pfil_hook))
835 goto passout;
836
837 odst = ip6->ip6_dst;
838 /* Run through list of hooks for output packets. */
839 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
840 if (error != 0 || m == NULL)
841 goto done;
842 ip6 = mtod(m, struct ip6_hdr *);
843
844 /* See if destination IP address was changed by packet filter. */
845 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
846 m->m_flags |= M_SKIP_FIREWALL;
847 /* If destination is now ourself drop to ip6_input(). */
848 if (in6_localaddr(&ip6->ip6_dst)) {
849 if (m->m_pkthdr.rcvif == NULL)
850 m->m_pkthdr.rcvif = loif;
851 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
852 m->m_pkthdr.csum_flags |=
853 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
854 m->m_pkthdr.csum_data = 0xffff;
855 }
856 m->m_pkthdr.csum_flags |=
857 CSUM_IP_CHECKED | CSUM_IP_VALID;
858 error = netisr_queue(NETISR_IPV6, m);
859 goto done;
860 } else
861 goto again; /* Redo the routing table lookup. */
862 }
863
864 /* XXX: IPFIREWALL_FORWARD */
865
866passout:
867 /*
868 * Send the packet to the outgoing interface.
869 * If necessary, do IPv6 fragmentation before sending.
870 *
871 * the logic here is rather complex:
872 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
873 * 1-a: send as is if tlen <= path mtu
874 * 1-b: fragment if tlen > path mtu
875 *
876 * 2: if user asks us not to fragment (dontfrag == 1)
877 * 2-a: send as is if tlen <= interface mtu
878 * 2-b: error if tlen > interface mtu
879 *
880 * 3: if we always need to attach fragment header (alwaysfrag == 1)
881 * always fragment
882 *
883 * 4: if dontfrag == 1 && alwaysfrag == 1
884 * error, as we cannot handle this conflicting request
885 */
886 tlen = m->m_pkthdr.len;
887
888 if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
889 dontfrag = 1;
890 else
891 dontfrag = 0;
892 if (dontfrag && alwaysfrag) { /* case 4 */
893 /* conflicting request - can't transmit */
894 error = EMSGSIZE;
895 goto bad;
896 }
897 if (dontfrag && tlen > IN6_LINKMTU(ifp)) { /* case 2-b */
898 /*
899 * Even if the DONTFRAG option is specified, we cannot send the
900 * packet when the data length is larger than the MTU of the
901 * outgoing interface.
902 * Notify the error by sending IPV6_PATHMTU ancillary data as
903 * well as returning an error code (the latter is not described
904 * in the API spec.)
905 */
906 u_int32_t mtu32;
907 struct ip6ctlparam ip6cp;
908
909 mtu32 = (u_int32_t)mtu;
910 bzero(&ip6cp, sizeof(ip6cp));
911 ip6cp.ip6c_cmdarg = (void *)&mtu32;
912 pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
913 (void *)&ip6cp);
914
915 error = EMSGSIZE;
916 goto bad;
917 }
918
919 /*
920 * transmit packet without fragmentation
921 */
922 if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
923 struct in6_ifaddr *ia6;
924
925 ip6 = mtod(m, struct ip6_hdr *);
926 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
927 if (ia6) {
928 /* Record statistics for this interface address. */
929 ia6->ia_ifa.if_opackets++;
930 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
931 }
932 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
933 goto done;
934 }
935
936 /*
937 * try to fragment the packet. case 1-b and 3
938 */
939 if (mtu < IPV6_MMTU) {
940 /* path MTU cannot be less than IPV6_MMTU */
941 error = EMSGSIZE;
942 in6_ifstat_inc(ifp, ifs6_out_fragfail);
943 goto bad;
944 } else if (ip6->ip6_plen == 0) {
945 /* jumbo payload cannot be fragmented */
946 error = EMSGSIZE;
947 in6_ifstat_inc(ifp, ifs6_out_fragfail);
948 goto bad;
949 } else {
950 struct mbuf **mnext, *m_frgpart;
951 struct ip6_frag *ip6f;
952 u_int32_t id = htonl(ip6_randomid());
953 u_char nextproto;
954
955 int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
956
957 /*
958 * Too large for the destination or interface;
959 * fragment if possible.
960 * Must be able to put at least 8 bytes per fragment.
961 */
962 hlen = unfragpartlen;
963 if (mtu > IPV6_MAXPACKET)
964 mtu = IPV6_MAXPACKET;
965
966 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
967 if (len < 8) {
968 error = EMSGSIZE;
969 in6_ifstat_inc(ifp, ifs6_out_fragfail);
970 goto bad;
971 }
972
973 /*
974 * Verify that we have any chance at all of being able to queue
975 * the packet or packet fragments
976 */
977 if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
978 < tlen /* - hlen */)) {
979 error = ENOBUFS;
980 ip6stat.ip6s_odropped++;
981 goto bad;
982 }
983
984 mnext = &m->m_nextpkt;
985
986 /*
987 * Change the next header field of the last header in the
988 * unfragmentable part.
989 */
990 if (exthdrs.ip6e_rthdr) {
991 nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
992 *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
993 } else if (exthdrs.ip6e_dest1) {
994 nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
995 *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
996 } else if (exthdrs.ip6e_hbh) {
997 nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
998 *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
999 } else {
1000 nextproto = ip6->ip6_nxt;
1001 ip6->ip6_nxt = IPPROTO_FRAGMENT;
1002 }
1003
1004 /*
1005 * Loop through length of segment after first fragment,
1006 * make new header and copy data of each part and link onto
1007 * chain.
1008 */
1009 m0 = m;
1010 for (off = hlen; off < tlen; off += len) {
1011 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1012 if (!m) {
1013 error = ENOBUFS;
1014 ip6stat.ip6s_odropped++;
1015 goto sendorfree;
1016 }
1017 m->m_pkthdr.rcvif = NULL;
1018 m->m_flags = m0->m_flags & M_COPYFLAGS;
1019 *mnext = m;
1020 mnext = &m->m_nextpkt;
1021 m->m_data += max_linkhdr;
1022 mhip6 = mtod(m, struct ip6_hdr *);
1023 *mhip6 = *ip6;
1024 m->m_len = sizeof(*mhip6);
1025 error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1026 if (error) {
1027 ip6stat.ip6s_odropped++;
1028 goto sendorfree;
1029 }
1030 ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1031 if (off + len >= tlen)
1032 len = tlen - off;
1033 else
1034 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1035 mhip6->ip6_plen = htons((u_short)(len + hlen +
1036 sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1037 if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1038 error = ENOBUFS;
1039 ip6stat.ip6s_odropped++;
1040 goto sendorfree;
1041 }
1042 m_cat(m, m_frgpart);
1043 m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1044 m->m_pkthdr.rcvif = NULL;
1045 ip6f->ip6f_reserved = 0;
1046 ip6f->ip6f_ident = id;
1047 ip6f->ip6f_nxt = nextproto;
1048 ip6stat.ip6s_ofragments++;
1049 in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1050 }
1051
1052 in6_ifstat_inc(ifp, ifs6_out_fragok);
1053 }
1054
1055 /*
1056 * Remove leading garbages.
1057 */
1058sendorfree:
1059 m = m0->m_nextpkt;
1060 m0->m_nextpkt = 0;
1061 m_freem(m0);
1062 for (m0 = m; m; m = m0) {
1063 m0 = m->m_nextpkt;
1064 m->m_nextpkt = 0;
1065 if (error == 0) {
1066 /* Record statistics for this interface address. */
1067 if (ia) {
1068 ia->ia_ifa.if_opackets++;
1069 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1070 }
1071 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1072 } else
1073 m_freem(m);
1074 }
1075
1076 if (error == 0)
1077 ip6stat.ip6s_fragmented++;
1078
1079done:
1080 if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1081 RTFREE(ro->ro_rt);
1082 } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1083 RTFREE(ro_pmtu->ro_rt);
1084 }
|
1085
1086 return (error);
1087
1088freehdrs:
1089 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1090 m_freem(exthdrs.ip6e_dest1);
1091 m_freem(exthdrs.ip6e_rthdr);
1092 m_freem(exthdrs.ip6e_dest2);
1093 /* FALLTHROUGH */
1094bad:
1095 if (m)
1096 m_freem(m);
1097 goto done;
1098}
1099
1100static int
1101ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1102{
1103 struct mbuf *m;
1104
1105 if (hlen > MCLBYTES)
1106 return (ENOBUFS); /* XXX */
1107
1108 MGET(m, M_DONTWAIT, MT_DATA);
1109 if (!m)
1110 return (ENOBUFS);
1111
1112 if (hlen > MLEN) {
1113 MCLGET(m, M_DONTWAIT);
1114 if ((m->m_flags & M_EXT) == 0) {
1115 m_free(m);
1116 return (ENOBUFS);
1117 }
1118 }
1119 m->m_len = hlen;
1120 if (hdr)
1121 bcopy(hdr, mtod(m, caddr_t), hlen);
1122
1123 *mp = m;
1124 return (0);
1125}
1126
1127/*
1128 * Insert jumbo payload option.
1129 */
1130static int
1131ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1132{
1133 struct mbuf *mopt;
1134 u_char *optbuf;
1135 u_int32_t v;
1136
1137#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1138
1139 /*
1140 * If there is no hop-by-hop options header, allocate new one.
1141 * If there is one but it doesn't have enough space to store the
1142 * jumbo payload option, allocate a cluster to store the whole options.
1143 * Otherwise, use it to store the options.
1144 */
1145 if (exthdrs->ip6e_hbh == 0) {
1146 MGET(mopt, M_DONTWAIT, MT_DATA);
1147 if (mopt == 0)
1148 return (ENOBUFS);
1149 mopt->m_len = JUMBOOPTLEN;
1150 optbuf = mtod(mopt, u_char *);
1151 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1152 exthdrs->ip6e_hbh = mopt;
1153 } else {
1154 struct ip6_hbh *hbh;
1155
1156 mopt = exthdrs->ip6e_hbh;
1157 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1158 /*
1159 * XXX assumption:
1160 * - exthdrs->ip6e_hbh is not referenced from places
1161 * other than exthdrs.
1162 * - exthdrs->ip6e_hbh is not an mbuf chain.
1163 */
1164 int oldoptlen = mopt->m_len;
1165 struct mbuf *n;
1166
1167 /*
1168 * XXX: give up if the whole (new) hbh header does
1169 * not fit even in an mbuf cluster.
1170 */
1171 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1172 return (ENOBUFS);
1173
1174 /*
1175 * As a consequence, we must always prepare a cluster
1176 * at this point.
1177 */
1178 MGET(n, M_DONTWAIT, MT_DATA);
1179 if (n) {
1180 MCLGET(n, M_DONTWAIT);
1181 if ((n->m_flags & M_EXT) == 0) {
1182 m_freem(n);
1183 n = NULL;
1184 }
1185 }
1186 if (!n)
1187 return (ENOBUFS);
1188 n->m_len = oldoptlen + JUMBOOPTLEN;
1189 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1190 oldoptlen);
1191 optbuf = mtod(n, caddr_t) + oldoptlen;
1192 m_freem(mopt);
1193 mopt = exthdrs->ip6e_hbh = n;
1194 } else {
1195 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1196 mopt->m_len += JUMBOOPTLEN;
1197 }
1198 optbuf[0] = IP6OPT_PADN;
1199 optbuf[1] = 1;
1200
1201 /*
1202 * Adjust the header length according to the pad and
1203 * the jumbo payload option.
1204 */
1205 hbh = mtod(mopt, struct ip6_hbh *);
1206 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1207 }
1208
1209 /* fill in the option. */
1210 optbuf[2] = IP6OPT_JUMBO;
1211 optbuf[3] = 4;
1212 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1213 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1214
1215 /* finally, adjust the packet header length */
1216 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1217
1218 return (0);
1219#undef JUMBOOPTLEN
1220}
1221
1222/*
1223 * Insert fragment header and copy unfragmentable header portions.
1224 */
1225static int
1226ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1227 struct ip6_frag **frghdrp)
1228{
1229 struct mbuf *n, *mlast;
1230
1231 if (hlen > sizeof(struct ip6_hdr)) {
1232 n = m_copym(m0, sizeof(struct ip6_hdr),
1233 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1234 if (n == 0)
1235 return (ENOBUFS);
1236 m->m_next = n;
1237 } else
1238 n = m;
1239
1240 /* Search for the last mbuf of unfragmentable part. */
1241 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1242 ;
1243
1244 if ((mlast->m_flags & M_EXT) == 0 &&
1245 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1246 /* use the trailing space of the last mbuf for the fragment hdr */
1247 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1248 mlast->m_len);
1249 mlast->m_len += sizeof(struct ip6_frag);
1250 m->m_pkthdr.len += sizeof(struct ip6_frag);
1251 } else {
1252 /* allocate a new mbuf for the fragment header */
1253 struct mbuf *mfrg;
1254
1255 MGET(mfrg, M_DONTWAIT, MT_DATA);
1256 if (mfrg == 0)
1257 return (ENOBUFS);
1258 mfrg->m_len = sizeof(struct ip6_frag);
1259 *frghdrp = mtod(mfrg, struct ip6_frag *);
1260 mlast->m_next = mfrg;
1261 }
1262
1263 return (0);
1264}
1265
1266static int
1267ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1268 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1269 int *alwaysfragp)
1270{
1271 u_int32_t mtu = 0;
1272 int alwaysfrag = 0;
1273 int error = 0;
1274
1275 if (ro_pmtu != ro) {
1276 /* The first hop and the final destination may differ. */
1277 struct sockaddr_in6 *sa6_dst =
1278 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1279 if (ro_pmtu->ro_rt &&
1280 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1281 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1282 RTFREE(ro_pmtu->ro_rt);
1283 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1284 }
1285 if (ro_pmtu->ro_rt == NULL) {
1286 bzero(sa6_dst, sizeof(*sa6_dst));
1287 sa6_dst->sin6_family = AF_INET6;
1288 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1289 sa6_dst->sin6_addr = *dst;
1290
1291 rtalloc((struct route *)ro_pmtu);
1292 }
1293 }
1294 if (ro_pmtu->ro_rt) {
1295 u_int32_t ifmtu;
1296 struct in_conninfo inc;
1297
1298 bzero(&inc, sizeof(inc));
1299 inc.inc_flags = 1; /* IPv6 */
1300 inc.inc6_faddr = *dst;
1301
1302 if (ifp == NULL)
1303 ifp = ro_pmtu->ro_rt->rt_ifp;
1304 ifmtu = IN6_LINKMTU(ifp);
1305 mtu = tcp_hc_getmtu(&inc);
1306 if (mtu)
1307 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1308 else
1309 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1310 if (mtu == 0)
1311 mtu = ifmtu;
1312 else if (mtu < IPV6_MMTU) {
1313 /*
1314 * RFC2460 section 5, last paragraph:
1315 * if we record ICMPv6 too big message with
1316 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1317 * or smaller, with framgent header attached.
1318 * (fragment header is needed regardless from the
1319 * packet size, for translators to identify packets)
1320 */
1321 alwaysfrag = 1;
1322 mtu = IPV6_MMTU;
1323 } else if (mtu > ifmtu) {
1324 /*
1325 * The MTU on the route is larger than the MTU on
1326 * the interface! This shouldn't happen, unless the
1327 * MTU of the interface has been changed after the
1328 * interface was brought up. Change the MTU in the
1329 * route to match the interface MTU (as long as the
1330 * field isn't locked).
1331 */
1332 mtu = ifmtu;
1333 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1334 }
1335 } else if (ifp) {
1336 mtu = IN6_LINKMTU(ifp);
1337 } else
1338 error = EHOSTUNREACH; /* XXX */
1339
1340 *mtup = mtu;
1341 if (alwaysfragp)
1342 *alwaysfragp = alwaysfrag;
1343 return (error);
1344}
1345
1346/*
1347 * IP6 socket option processing.
1348 */
1349int
1350ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1351{
1352 int optdatalen, uproto;
1353 void *optdata;
1354 struct inpcb *in6p = sotoinpcb(so);
1355 int error, optval;
1356 int level, op, optname;
1357 int optlen;
1358 struct thread *td;
1359
1360 if (sopt) {
1361 level = sopt->sopt_level;
1362 op = sopt->sopt_dir;
1363 optname = sopt->sopt_name;
1364 optlen = sopt->sopt_valsize;
1365 td = sopt->sopt_td;
1366 } else {
1367 panic("ip6_ctloutput: arg soopt is NULL");
1368 }
1369 error = optval = 0;
1370
1371 uproto = (int)so->so_proto->pr_protocol;
1372
1373 if (level == IPPROTO_IPV6) {
1374 switch (op) {
1375
1376 case SOPT_SET:
1377 switch (optname) {
1378 case IPV6_2292PKTOPTIONS:
1379#ifdef IPV6_PKTOPTIONS
1380 case IPV6_PKTOPTIONS:
1381#endif
1382 {
1383 struct mbuf *m;
1384
1385 error = soopt_getm(sopt, &m); /* XXX */
1386 if (error != 0)
1387 break;
1388 error = soopt_mcopyin(sopt, m); /* XXX */
1389 if (error != 0)
1390 break;
1391 error = ip6_pcbopts(&in6p->in6p_outputopts,
1392 m, so, sopt);
1393 m_freem(m); /* XXX */
1394 break;
1395 }
1396
1397 /*
1398 * Use of some Hop-by-Hop options or some
1399 * Destination options, might require special
1400 * privilege. That is, normal applications
1401 * (without special privilege) might be forbidden
1402 * from setting certain options in outgoing packets,
1403 * and might never see certain options in received
1404 * packets. [RFC 2292 Section 6]
1405 * KAME specific note:
1406 * KAME prevents non-privileged users from sending or
1407 * receiving ANY hbh/dst options in order to avoid
1408 * overhead of parsing options in the kernel.
1409 */
1410 case IPV6_RECVHOPOPTS:
1411 case IPV6_RECVDSTOPTS:
1412 case IPV6_RECVRTHDRDSTOPTS:
1413 if (td != NULL) {
1414 error = priv_check(td,
1415 PRIV_NETINET_SETHDROPTS);
1416 if (error)
1417 break;
1418 }
1419 /* FALLTHROUGH */
1420 case IPV6_UNICAST_HOPS:
1421 case IPV6_HOPLIMIT:
1422 case IPV6_FAITH:
1423
1424 case IPV6_RECVPKTINFO:
1425 case IPV6_RECVHOPLIMIT:
1426 case IPV6_RECVRTHDR:
1427 case IPV6_RECVPATHMTU:
1428 case IPV6_RECVTCLASS:
1429 case IPV6_V6ONLY:
1430 case IPV6_AUTOFLOWLABEL:
1431 if (optlen != sizeof(int)) {
1432 error = EINVAL;
1433 break;
1434 }
1435 error = sooptcopyin(sopt, &optval,
1436 sizeof optval, sizeof optval);
1437 if (error)
1438 break;
1439 switch (optname) {
1440
1441 case IPV6_UNICAST_HOPS:
1442 if (optval < -1 || optval >= 256)
1443 error = EINVAL;
1444 else {
1445 /* -1 = kernel default */
1446 in6p->in6p_hops = optval;
1447 if ((in6p->in6p_vflag &
1448 INP_IPV4) != 0)
1449 in6p->inp_ip_ttl = optval;
1450 }
1451 break;
1452#define OPTSET(bit) \
1453do { \
1454 if (optval) \
1455 in6p->in6p_flags |= (bit); \
1456 else \
1457 in6p->in6p_flags &= ~(bit); \
1458} while (/*CONSTCOND*/ 0)
1459#define OPTSET2292(bit) \
1460do { \
1461 in6p->in6p_flags |= IN6P_RFC2292; \
1462 if (optval) \
1463 in6p->in6p_flags |= (bit); \
1464 else \
1465 in6p->in6p_flags &= ~(bit); \
1466} while (/*CONSTCOND*/ 0)
1467#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1468
1469 case IPV6_RECVPKTINFO:
1470 /* cannot mix with RFC2292 */
1471 if (OPTBIT(IN6P_RFC2292)) {
1472 error = EINVAL;
1473 break;
1474 }
1475 OPTSET(IN6P_PKTINFO);
1476 break;
1477
1478 case IPV6_HOPLIMIT:
1479 {
1480 struct ip6_pktopts **optp;
1481
1482 /* cannot mix with RFC2292 */
1483 if (OPTBIT(IN6P_RFC2292)) {
1484 error = EINVAL;
1485 break;
1486 }
1487 optp = &in6p->in6p_outputopts;
1488 error = ip6_pcbopt(IPV6_HOPLIMIT,
1489 (u_char *)&optval, sizeof(optval),
1490 optp, (td != NULL) ? td->td_ucred :
1491 NULL, uproto);
1492 break;
1493 }
1494
1495 case IPV6_RECVHOPLIMIT:
1496 /* cannot mix with RFC2292 */
1497 if (OPTBIT(IN6P_RFC2292)) {
1498 error = EINVAL;
1499 break;
1500 }
1501 OPTSET(IN6P_HOPLIMIT);
1502 break;
1503
1504 case IPV6_RECVHOPOPTS:
1505 /* cannot mix with RFC2292 */
1506 if (OPTBIT(IN6P_RFC2292)) {
1507 error = EINVAL;
1508 break;
1509 }
1510 OPTSET(IN6P_HOPOPTS);
1511 break;
1512
1513 case IPV6_RECVDSTOPTS:
1514 /* cannot mix with RFC2292 */
1515 if (OPTBIT(IN6P_RFC2292)) {
1516 error = EINVAL;
1517 break;
1518 }
1519 OPTSET(IN6P_DSTOPTS);
1520 break;
1521
1522 case IPV6_RECVRTHDRDSTOPTS:
1523 /* cannot mix with RFC2292 */
1524 if (OPTBIT(IN6P_RFC2292)) {
1525 error = EINVAL;
1526 break;
1527 }
1528 OPTSET(IN6P_RTHDRDSTOPTS);
1529 break;
1530
1531 case IPV6_RECVRTHDR:
1532 /* cannot mix with RFC2292 */
1533 if (OPTBIT(IN6P_RFC2292)) {
1534 error = EINVAL;
1535 break;
1536 }
1537 OPTSET(IN6P_RTHDR);
1538 break;
1539
1540 case IPV6_FAITH:
1541 OPTSET(IN6P_FAITH);
1542 break;
1543
1544 case IPV6_RECVPATHMTU:
1545 /*
1546 * We ignore this option for TCP
1547 * sockets.
1548 * (RFC3542 leaves this case
1549 * unspecified.)
1550 */
1551 if (uproto != IPPROTO_TCP)
1552 OPTSET(IN6P_MTU);
1553 break;
1554
1555 case IPV6_V6ONLY:
1556 /*
1557 * make setsockopt(IPV6_V6ONLY)
1558 * available only prior to bind(2).
1559 * see ipng mailing list, Jun 22 2001.
1560 */
1561 if (in6p->in6p_lport ||
1562 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1563 error = EINVAL;
1564 break;
1565 }
1566 OPTSET(IN6P_IPV6_V6ONLY);
1567 if (optval)
1568 in6p->in6p_vflag &= ~INP_IPV4;
1569 else
1570 in6p->in6p_vflag |= INP_IPV4;
1571 break;
1572 case IPV6_RECVTCLASS:
1573 /* cannot mix with RFC2292 XXX */
1574 if (OPTBIT(IN6P_RFC2292)) {
1575 error = EINVAL;
1576 break;
1577 }
1578 OPTSET(IN6P_TCLASS);
1579 break;
1580 case IPV6_AUTOFLOWLABEL:
1581 OPTSET(IN6P_AUTOFLOWLABEL);
1582 break;
1583
1584 }
1585 break;
1586
1587 case IPV6_TCLASS:
1588 case IPV6_DONTFRAG:
1589 case IPV6_USE_MIN_MTU:
1590 case IPV6_PREFER_TEMPADDR:
1591 if (optlen != sizeof(optval)) {
1592 error = EINVAL;
1593 break;
1594 }
1595 error = sooptcopyin(sopt, &optval,
1596 sizeof optval, sizeof optval);
1597 if (error)
1598 break;
1599 {
1600 struct ip6_pktopts **optp;
1601 optp = &in6p->in6p_outputopts;
1602 error = ip6_pcbopt(optname,
1603 (u_char *)&optval, sizeof(optval),
1604 optp, (td != NULL) ? td->td_ucred :
1605 NULL, uproto);
1606 break;
1607 }
1608
1609 case IPV6_2292PKTINFO:
1610 case IPV6_2292HOPLIMIT:
1611 case IPV6_2292HOPOPTS:
1612 case IPV6_2292DSTOPTS:
1613 case IPV6_2292RTHDR:
1614 /* RFC 2292 */
1615 if (optlen != sizeof(int)) {
1616 error = EINVAL;
1617 break;
1618 }
1619 error = sooptcopyin(sopt, &optval,
1620 sizeof optval, sizeof optval);
1621 if (error)
1622 break;
1623 switch (optname) {
1624 case IPV6_2292PKTINFO:
1625 OPTSET2292(IN6P_PKTINFO);
1626 break;
1627 case IPV6_2292HOPLIMIT:
1628 OPTSET2292(IN6P_HOPLIMIT);
1629 break;
1630 case IPV6_2292HOPOPTS:
1631 /*
1632 * Check super-user privilege.
1633 * See comments for IPV6_RECVHOPOPTS.
1634 */
1635 if (td != NULL) {
1636 error = priv_check(td,
1637 PRIV_NETINET_SETHDROPTS);
1638 if (error)
1639 return (error);
1640 }
1641 OPTSET2292(IN6P_HOPOPTS);
1642 break;
1643 case IPV6_2292DSTOPTS:
1644 if (td != NULL) {
1645 error = priv_check(td,
1646 PRIV_NETINET_SETHDROPTS);
1647 if (error)
1648 return (error);
1649 }
1650 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1651 break;
1652 case IPV6_2292RTHDR:
1653 OPTSET2292(IN6P_RTHDR);
1654 break;
1655 }
1656 break;
1657 case IPV6_PKTINFO:
1658 case IPV6_HOPOPTS:
1659 case IPV6_RTHDR:
1660 case IPV6_DSTOPTS:
1661 case IPV6_RTHDRDSTOPTS:
1662 case IPV6_NEXTHOP:
1663 {
1664 /* new advanced API (RFC3542) */
1665 u_char *optbuf;
1666 u_char optbuf_storage[MCLBYTES];
1667 int optlen;
1668 struct ip6_pktopts **optp;
1669
1670 /* cannot mix with RFC2292 */
1671 if (OPTBIT(IN6P_RFC2292)) {
1672 error = EINVAL;
1673 break;
1674 }
1675
1676 /*
1677 * We only ensure valsize is not too large
1678 * here. Further validation will be done
1679 * later.
1680 */
1681 error = sooptcopyin(sopt, optbuf_storage,
1682 sizeof(optbuf_storage), 0);
1683 if (error)
1684 break;
1685 optlen = sopt->sopt_valsize;
1686 optbuf = optbuf_storage;
1687 optp = &in6p->in6p_outputopts;
1688 error = ip6_pcbopt(optname, optbuf, optlen,
1689 optp, (td != NULL) ? td->td_ucred : NULL,
1690 uproto);
1691 break;
1692 }
1693#undef OPTSET
1694
1695 case IPV6_MULTICAST_IF:
1696 case IPV6_MULTICAST_HOPS:
1697 case IPV6_MULTICAST_LOOP:
1698 case IPV6_JOIN_GROUP:
1699 case IPV6_LEAVE_GROUP:
1700 {
1701 if (sopt->sopt_valsize > MLEN) {
1702 error = EMSGSIZE;
1703 break;
1704 }
1705 /* XXX */
1706 }
1707 /* FALLTHROUGH */
1708 {
1709 struct mbuf *m;
1710
1711 if (sopt->sopt_valsize > MCLBYTES) {
1712 error = EMSGSIZE;
1713 break;
1714 }
1715 /* XXX */
1716 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1717 if (m == 0) {
1718 error = ENOBUFS;
1719 break;
1720 }
1721 if (sopt->sopt_valsize > MLEN) {
1722 MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1723 if ((m->m_flags & M_EXT) == 0) {
1724 m_free(m);
1725 error = ENOBUFS;
1726 break;
1727 }
1728 }
1729 m->m_len = sopt->sopt_valsize;
1730 error = sooptcopyin(sopt, mtod(m, char *),
1731 m->m_len, m->m_len);
1732 if (error) {
1733 (void)m_free(m);
1734 break;
1735 }
1736 error = ip6_setmoptions(sopt->sopt_name,
1737 &in6p->in6p_moptions,
1738 m);
1739 (void)m_free(m);
1740 }
1741 break;
1742
1743 case IPV6_PORTRANGE:
1744 error = sooptcopyin(sopt, &optval,
1745 sizeof optval, sizeof optval);
1746 if (error)
1747 break;
1748
1749 switch (optval) {
1750 case IPV6_PORTRANGE_DEFAULT:
1751 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1752 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1753 break;
1754
1755 case IPV6_PORTRANGE_HIGH:
1756 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1757 in6p->in6p_flags |= IN6P_HIGHPORT;
1758 break;
1759
1760 case IPV6_PORTRANGE_LOW:
1761 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1762 in6p->in6p_flags |= IN6P_LOWPORT;
1763 break;
1764
1765 default:
1766 error = EINVAL;
1767 break;
1768 }
1769 break;
1770
1771#ifdef IPSEC
1772 case IPV6_IPSEC_POLICY:
1773 {
1774 caddr_t req;
1775 struct mbuf *m;
1776
1777 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1778 break;
1779 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1780 break;
1781 req = mtod(m, caddr_t);
1782 error = ipsec6_set_policy(in6p, optname, req,
1783 m->m_len, (sopt->sopt_td != NULL) ?
1784 sopt->sopt_td->td_ucred : NULL);
1785 m_freem(m);
1786 break;
1787 }
1788#endif /* IPSEC */
1789
1790 default:
1791 error = ENOPROTOOPT;
1792 break;
1793 }
1794 break;
1795
1796 case SOPT_GET:
1797 switch (optname) {
1798
1799 case IPV6_2292PKTOPTIONS:
1800#ifdef IPV6_PKTOPTIONS
1801 case IPV6_PKTOPTIONS:
1802#endif
1803 /*
1804 * RFC3542 (effectively) deprecated the
1805 * semantics of the 2292-style pktoptions.
1806 * Since it was not reliable in nature (i.e.,
1807 * applications had to expect the lack of some
1808 * information after all), it would make sense
1809 * to simplify this part by always returning
1810 * empty data.
1811 */
1812 sopt->sopt_valsize = 0;
1813 break;
1814
1815 case IPV6_RECVHOPOPTS:
1816 case IPV6_RECVDSTOPTS:
1817 case IPV6_RECVRTHDRDSTOPTS:
1818 case IPV6_UNICAST_HOPS:
1819 case IPV6_RECVPKTINFO:
1820 case IPV6_RECVHOPLIMIT:
1821 case IPV6_RECVRTHDR:
1822 case IPV6_RECVPATHMTU:
1823
1824 case IPV6_FAITH:
1825 case IPV6_V6ONLY:
1826 case IPV6_PORTRANGE:
1827 case IPV6_RECVTCLASS:
1828 case IPV6_AUTOFLOWLABEL:
1829 switch (optname) {
1830
1831 case IPV6_RECVHOPOPTS:
1832 optval = OPTBIT(IN6P_HOPOPTS);
1833 break;
1834
1835 case IPV6_RECVDSTOPTS:
1836 optval = OPTBIT(IN6P_DSTOPTS);
1837 break;
1838
1839 case IPV6_RECVRTHDRDSTOPTS:
1840 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1841 break;
1842
1843 case IPV6_UNICAST_HOPS:
1844 optval = in6p->in6p_hops;
1845 break;
1846
1847 case IPV6_RECVPKTINFO:
1848 optval = OPTBIT(IN6P_PKTINFO);
1849 break;
1850
1851 case IPV6_RECVHOPLIMIT:
1852 optval = OPTBIT(IN6P_HOPLIMIT);
1853 break;
1854
1855 case IPV6_RECVRTHDR:
1856 optval = OPTBIT(IN6P_RTHDR);
1857 break;
1858
1859 case IPV6_RECVPATHMTU:
1860 optval = OPTBIT(IN6P_MTU);
1861 break;
1862
1863 case IPV6_FAITH:
1864 optval = OPTBIT(IN6P_FAITH);
1865 break;
1866
1867 case IPV6_V6ONLY:
1868 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1869 break;
1870
1871 case IPV6_PORTRANGE:
1872 {
1873 int flags;
1874 flags = in6p->in6p_flags;
1875 if (flags & IN6P_HIGHPORT)
1876 optval = IPV6_PORTRANGE_HIGH;
1877 else if (flags & IN6P_LOWPORT)
1878 optval = IPV6_PORTRANGE_LOW;
1879 else
1880 optval = 0;
1881 break;
1882 }
1883 case IPV6_RECVTCLASS:
1884 optval = OPTBIT(IN6P_TCLASS);
1885 break;
1886
1887 case IPV6_AUTOFLOWLABEL:
1888 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1889 break;
1890 }
1891 if (error)
1892 break;
1893 error = sooptcopyout(sopt, &optval,
1894 sizeof optval);
1895 break;
1896
1897 case IPV6_PATHMTU:
1898 {
1899 u_long pmtu = 0;
1900 struct ip6_mtuinfo mtuinfo;
1901 struct route_in6 sro;
1902
1903 bzero(&sro, sizeof(sro));
1904
1905 if (!(so->so_state & SS_ISCONNECTED))
1906 return (ENOTCONN);
1907 /*
1908 * XXX: we dot not consider the case of source
1909 * routing, or optional information to specify
1910 * the outgoing interface.
1911 */
1912 error = ip6_getpmtu(&sro, NULL, NULL,
1913 &in6p->in6p_faddr, &pmtu, NULL);
1914 if (sro.ro_rt)
1915 RTFREE(sro.ro_rt);
1916 if (error)
1917 break;
1918 if (pmtu > IPV6_MAXPACKET)
1919 pmtu = IPV6_MAXPACKET;
1920
1921 bzero(&mtuinfo, sizeof(mtuinfo));
1922 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1923 optdata = (void *)&mtuinfo;
1924 optdatalen = sizeof(mtuinfo);
1925 error = sooptcopyout(sopt, optdata,
1926 optdatalen);
1927 break;
1928 }
1929
1930 case IPV6_2292PKTINFO:
1931 case IPV6_2292HOPLIMIT:
1932 case IPV6_2292HOPOPTS:
1933 case IPV6_2292RTHDR:
1934 case IPV6_2292DSTOPTS:
1935 switch (optname) {
1936 case IPV6_2292PKTINFO:
1937 optval = OPTBIT(IN6P_PKTINFO);
1938 break;
1939 case IPV6_2292HOPLIMIT:
1940 optval = OPTBIT(IN6P_HOPLIMIT);
1941 break;
1942 case IPV6_2292HOPOPTS:
1943 optval = OPTBIT(IN6P_HOPOPTS);
1944 break;
1945 case IPV6_2292RTHDR:
1946 optval = OPTBIT(IN6P_RTHDR);
1947 break;
1948 case IPV6_2292DSTOPTS:
1949 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1950 break;
1951 }
1952 error = sooptcopyout(sopt, &optval,
1953 sizeof optval);
1954 break;
1955 case IPV6_PKTINFO:
1956 case IPV6_HOPOPTS:
1957 case IPV6_RTHDR:
1958 case IPV6_DSTOPTS:
1959 case IPV6_RTHDRDSTOPTS:
1960 case IPV6_NEXTHOP:
1961 case IPV6_TCLASS:
1962 case IPV6_DONTFRAG:
1963 case IPV6_USE_MIN_MTU:
1964 case IPV6_PREFER_TEMPADDR:
1965 error = ip6_getpcbopt(in6p->in6p_outputopts,
1966 optname, sopt);
1967 break;
1968
1969 case IPV6_MULTICAST_IF:
1970 case IPV6_MULTICAST_HOPS:
1971 case IPV6_MULTICAST_LOOP:
1972 case IPV6_JOIN_GROUP:
1973 case IPV6_LEAVE_GROUP:
1974 {
1975 struct mbuf *m;
1976 error = ip6_getmoptions(sopt->sopt_name,
1977 in6p->in6p_moptions, &m);
1978 if (error == 0)
1979 error = sooptcopyout(sopt,
1980 mtod(m, char *), m->m_len);
1981 m_freem(m);
1982 }
1983 break;
1984
1985#ifdef IPSEC
1986 case IPV6_IPSEC_POLICY:
1987 {
1988 caddr_t req = NULL;
1989 size_t len = 0;
1990 struct mbuf *m = NULL;
1991 struct mbuf **mp = &m;
1992 size_t ovalsize = sopt->sopt_valsize;
1993 caddr_t oval = (caddr_t)sopt->sopt_val;
1994
1995 error = soopt_getm(sopt, &m); /* XXX */
1996 if (error != 0)
1997 break;
1998 error = soopt_mcopyin(sopt, m); /* XXX */
1999 if (error != 0)
2000 break;
2001 sopt->sopt_valsize = ovalsize;
2002 sopt->sopt_val = oval;
2003 if (m) {
2004 req = mtod(m, caddr_t);
2005 len = m->m_len;
2006 }
2007 error = ipsec6_get_policy(in6p, req, len, mp);
2008 if (error == 0)
2009 error = soopt_mcopyout(sopt, m); /* XXX */
2010 if (error == 0 && m)
2011 m_freem(m);
2012 break;
2013 }
2014#endif /* IPSEC */
2015
2016 default:
2017 error = ENOPROTOOPT;
2018 break;
2019 }
2020 break;
2021 }
2022 } else { /* level != IPPROTO_IPV6 */
2023 error = EINVAL;
2024 }
2025 return (error);
2026}
2027
2028int
2029ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2030{
2031 int error = 0, optval, optlen;
2032 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2033 struct in6pcb *in6p = sotoin6pcb(so);
2034 int level, op, optname;
2035
2036 if (sopt) {
2037 level = sopt->sopt_level;
2038 op = sopt->sopt_dir;
2039 optname = sopt->sopt_name;
2040 optlen = sopt->sopt_valsize;
2041 } else
2042 panic("ip6_raw_ctloutput: arg soopt is NULL");
2043
2044 if (level != IPPROTO_IPV6) {
2045 return (EINVAL);
2046 }
2047
2048 switch (optname) {
2049 case IPV6_CHECKSUM:
2050 /*
2051 * For ICMPv6 sockets, no modification allowed for checksum
2052 * offset, permit "no change" values to help existing apps.
2053 *
2054 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2055 * for an ICMPv6 socket will fail."
2056 * The current behavior does not meet RFC3542.
2057 */
2058 switch (op) {
2059 case SOPT_SET:
2060 if (optlen != sizeof(int)) {
2061 error = EINVAL;
2062 break;
2063 }
2064 error = sooptcopyin(sopt, &optval, sizeof(optval),
2065 sizeof(optval));
2066 if (error)
2067 break;
2068 if ((optval % 2) != 0) {
2069 /* the API assumes even offset values */
2070 error = EINVAL;
2071 } else if (so->so_proto->pr_protocol ==
2072 IPPROTO_ICMPV6) {
2073 if (optval != icmp6off)
2074 error = EINVAL;
2075 } else
2076 in6p->in6p_cksum = optval;
2077 break;
2078
2079 case SOPT_GET:
2080 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2081 optval = icmp6off;
2082 else
2083 optval = in6p->in6p_cksum;
2084
2085 error = sooptcopyout(sopt, &optval, sizeof(optval));
2086 break;
2087
2088 default:
2089 error = EINVAL;
2090 break;
2091 }
2092 break;
2093
2094 default:
2095 error = ENOPROTOOPT;
2096 break;
2097 }
2098
2099 return (error);
2100}
2101
2102/*
2103 * Set up IP6 options in pcb for insertion in output packets or
2104 * specifying behavior of outgoing packets.
2105 */
2106static int
2107ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2108 struct socket *so, struct sockopt *sopt)
2109{
2110 struct ip6_pktopts *opt = *pktopt;
2111 int error = 0;
2112 struct thread *td = sopt->sopt_td;
2113
2114 /* turn off any old options. */
2115 if (opt) {
2116#ifdef DIAGNOSTIC
2117 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2118 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2119 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2120 printf("ip6_pcbopts: all specified options are cleared.\n");
2121#endif
2122 ip6_clearpktopts(opt, -1);
2123 } else
2124 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2125 *pktopt = NULL;
2126
2127 if (!m || m->m_len == 0) {
2128 /*
2129 * Only turning off any previous options, regardless of
2130 * whether the opt is just created or given.
2131 */
2132 free(opt, M_IP6OPT);
2133 return (0);
2134 }
2135
2136 /* set options specified by user. */
2137 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2138 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2139 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2140 free(opt, M_IP6OPT);
2141 return (error);
2142 }
2143 *pktopt = opt;
2144 return (0);
2145}
2146
2147/*
2148 * initialize ip6_pktopts. beware that there are non-zero default values in
2149 * the struct.
2150 */
2151void
2152ip6_initpktopts(struct ip6_pktopts *opt)
2153{
2154
2155 bzero(opt, sizeof(*opt));
2156 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2157 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2158 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2159 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2160}
2161
2162static int
2163ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2164 struct ucred *cred, int uproto)
2165{
2166 struct ip6_pktopts *opt;
2167
2168 if (*pktopt == NULL) {
2169 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2170 M_WAITOK);
2171 ip6_initpktopts(*pktopt);
2172 }
2173 opt = *pktopt;
2174
2175 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2176}
2177
2178static int
2179ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2180{
2181 void *optdata = NULL;
2182 int optdatalen = 0;
2183 struct ip6_ext *ip6e;
2184 int error = 0;
2185 struct in6_pktinfo null_pktinfo;
2186 int deftclass = 0, on;
2187 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2188 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2189
2190 switch (optname) {
2191 case IPV6_PKTINFO:
2192 if (pktopt && pktopt->ip6po_pktinfo)
2193 optdata = (void *)pktopt->ip6po_pktinfo;
2194 else {
2195 /* XXX: we don't have to do this every time... */
2196 bzero(&null_pktinfo, sizeof(null_pktinfo));
2197 optdata = (void *)&null_pktinfo;
2198 }
2199 optdatalen = sizeof(struct in6_pktinfo);
2200 break;
2201 case IPV6_TCLASS:
2202 if (pktopt && pktopt->ip6po_tclass >= 0)
2203 optdata = (void *)&pktopt->ip6po_tclass;
2204 else
2205 optdata = (void *)&deftclass;
2206 optdatalen = sizeof(int);
2207 break;
2208 case IPV6_HOPOPTS:
2209 if (pktopt && pktopt->ip6po_hbh) {
2210 optdata = (void *)pktopt->ip6po_hbh;
2211 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2212 optdatalen = (ip6e->ip6e_len + 1) << 3;
2213 }
2214 break;
2215 case IPV6_RTHDR:
2216 if (pktopt && pktopt->ip6po_rthdr) {
2217 optdata = (void *)pktopt->ip6po_rthdr;
2218 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2219 optdatalen = (ip6e->ip6e_len + 1) << 3;
2220 }
2221 break;
2222 case IPV6_RTHDRDSTOPTS:
2223 if (pktopt && pktopt->ip6po_dest1) {
2224 optdata = (void *)pktopt->ip6po_dest1;
2225 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2226 optdatalen = (ip6e->ip6e_len + 1) << 3;
2227 }
2228 break;
2229 case IPV6_DSTOPTS:
2230 if (pktopt && pktopt->ip6po_dest2) {
2231 optdata = (void *)pktopt->ip6po_dest2;
2232 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2233 optdatalen = (ip6e->ip6e_len + 1) << 3;
2234 }
2235 break;
2236 case IPV6_NEXTHOP:
2237 if (pktopt && pktopt->ip6po_nexthop) {
2238 optdata = (void *)pktopt->ip6po_nexthop;
2239 optdatalen = pktopt->ip6po_nexthop->sa_len;
2240 }
2241 break;
2242 case IPV6_USE_MIN_MTU:
2243 if (pktopt)
2244 optdata = (void *)&pktopt->ip6po_minmtu;
2245 else
2246 optdata = (void *)&defminmtu;
2247 optdatalen = sizeof(int);
2248 break;
2249 case IPV6_DONTFRAG:
2250 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2251 on = 1;
2252 else
2253 on = 0;
2254 optdata = (void *)&on;
2255 optdatalen = sizeof(on);
2256 break;
2257 case IPV6_PREFER_TEMPADDR:
2258 if (pktopt)
2259 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2260 else
2261 optdata = (void *)&defpreftemp;
2262 optdatalen = sizeof(int);
2263 break;
2264 default: /* should not happen */
2265#ifdef DIAGNOSTIC
2266 panic("ip6_getpcbopt: unexpected option\n");
2267#endif
2268 return (ENOPROTOOPT);
2269 }
2270
2271 error = sooptcopyout(sopt, optdata, optdatalen);
2272
2273 return (error);
2274}
2275
2276void
2277ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2278{
2279 if (pktopt == NULL)
2280 return;
2281
2282 if (optname == -1 || optname == IPV6_PKTINFO) {
2283 if (pktopt->ip6po_pktinfo)
2284 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2285 pktopt->ip6po_pktinfo = NULL;
2286 }
2287 if (optname == -1 || optname == IPV6_HOPLIMIT)
2288 pktopt->ip6po_hlim = -1;
2289 if (optname == -1 || optname == IPV6_TCLASS)
2290 pktopt->ip6po_tclass = -1;
2291 if (optname == -1 || optname == IPV6_NEXTHOP) {
2292 if (pktopt->ip6po_nextroute.ro_rt) {
2293 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2294 pktopt->ip6po_nextroute.ro_rt = NULL;
2295 }
2296 if (pktopt->ip6po_nexthop)
2297 free(pktopt->ip6po_nexthop, M_IP6OPT);
2298 pktopt->ip6po_nexthop = NULL;
2299 }
2300 if (optname == -1 || optname == IPV6_HOPOPTS) {
2301 if (pktopt->ip6po_hbh)
2302 free(pktopt->ip6po_hbh, M_IP6OPT);
2303 pktopt->ip6po_hbh = NULL;
2304 }
2305 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2306 if (pktopt->ip6po_dest1)
2307 free(pktopt->ip6po_dest1, M_IP6OPT);
2308 pktopt->ip6po_dest1 = NULL;
2309 }
2310 if (optname == -1 || optname == IPV6_RTHDR) {
2311 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2312 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2313 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2314 if (pktopt->ip6po_route.ro_rt) {
2315 RTFREE(pktopt->ip6po_route.ro_rt);
2316 pktopt->ip6po_route.ro_rt = NULL;
2317 }
2318 }
2319 if (optname == -1 || optname == IPV6_DSTOPTS) {
2320 if (pktopt->ip6po_dest2)
2321 free(pktopt->ip6po_dest2, M_IP6OPT);
2322 pktopt->ip6po_dest2 = NULL;
2323 }
2324}
2325
2326#define PKTOPT_EXTHDRCPY(type) \
2327do {\
2328 if (src->type) {\
2329 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2330 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2331 if (dst->type == NULL && canwait == M_NOWAIT)\
2332 goto bad;\
2333 bcopy(src->type, dst->type, hlen);\
2334 }\
2335} while (/*CONSTCOND*/ 0)
2336
2337static int
2338copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2339{
2340 if (dst == NULL || src == NULL) {
2341 printf("ip6_clearpktopts: invalid argument\n");
2342 return (EINVAL);
2343 }
2344
2345 dst->ip6po_hlim = src->ip6po_hlim;
2346 dst->ip6po_tclass = src->ip6po_tclass;
2347 dst->ip6po_flags = src->ip6po_flags;
2348 if (src->ip6po_pktinfo) {
2349 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2350 M_IP6OPT, canwait);
2351 if (dst->ip6po_pktinfo == NULL)
2352 goto bad;
2353 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2354 }
2355 if (src->ip6po_nexthop) {
2356 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2357 M_IP6OPT, canwait);
2358 if (dst->ip6po_nexthop == NULL)
2359 goto bad;
2360 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2361 src->ip6po_nexthop->sa_len);
2362 }
2363 PKTOPT_EXTHDRCPY(ip6po_hbh);
2364 PKTOPT_EXTHDRCPY(ip6po_dest1);
2365 PKTOPT_EXTHDRCPY(ip6po_dest2);
2366 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2367 return (0);
2368
2369 bad:
2370 ip6_clearpktopts(dst, -1);
2371 return (ENOBUFS);
2372}
2373#undef PKTOPT_EXTHDRCPY
2374
2375struct ip6_pktopts *
2376ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2377{
2378 int error;
2379 struct ip6_pktopts *dst;
2380
2381 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2382 if (dst == NULL)
2383 return (NULL);
2384 ip6_initpktopts(dst);
2385
2386 if ((error = copypktopts(dst, src, canwait)) != 0) {
2387 free(dst, M_IP6OPT);
2388 return (NULL);
2389 }
2390
2391 return (dst);
2392}
2393
2394void
2395ip6_freepcbopts(struct ip6_pktopts *pktopt)
2396{
2397 if (pktopt == NULL)
2398 return;
2399
2400 ip6_clearpktopts(pktopt, -1);
2401
2402 free(pktopt, M_IP6OPT);
2403}
2404
2405/*
2406 * Set the IP6 multicast options in response to user setsockopt().
2407 */
2408static int
2409ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2410{
2411 int error = 0;
2412 u_int loop, ifindex;
2413 struct ipv6_mreq *mreq;
2414 struct ifnet *ifp;
2415 struct ip6_moptions *im6o = *im6op;
2416 struct route_in6 ro;
2417 struct in6_multi_mship *imm;
2418
2419 if (im6o == NULL) {
2420 /*
2421 * No multicast option buffer attached to the pcb;
2422 * allocate one and initialize to default values.
2423 */
2424 im6o = (struct ip6_moptions *)
2425 malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2426
2427 if (im6o == NULL)
2428 return (ENOBUFS);
2429 *im6op = im6o;
2430 im6o->im6o_multicast_ifp = NULL;
2431 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2432 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2433 LIST_INIT(&im6o->im6o_memberships);
2434 }
2435
2436 switch (optname) {
2437
2438 case IPV6_MULTICAST_IF:
2439 /*
2440 * Select the interface for outgoing multicast packets.
2441 */
2442 if (m == NULL || m->m_len != sizeof(u_int)) {
2443 error = EINVAL;
2444 break;
2445 }
2446 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2447 if (ifindex < 0 || if_index < ifindex) {
2448 error = ENXIO; /* XXX EINVAL? */
2449 break;
2450 }
2451 ifp = ifnet_byindex(ifindex);
2452 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2453 error = EADDRNOTAVAIL;
2454 break;
2455 }
2456 im6o->im6o_multicast_ifp = ifp;
2457 break;
2458
2459 case IPV6_MULTICAST_HOPS:
2460 {
2461 /*
2462 * Set the IP6 hoplimit for outgoing multicast packets.
2463 */
2464 int optval;
2465 if (m == NULL || m->m_len != sizeof(int)) {
2466 error = EINVAL;
2467 break;
2468 }
2469 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2470 if (optval < -1 || optval >= 256)
2471 error = EINVAL;
2472 else if (optval == -1)
2473 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2474 else
2475 im6o->im6o_multicast_hlim = optval;
2476 break;
2477 }
2478
2479 case IPV6_MULTICAST_LOOP:
2480 /*
2481 * Set the loopback flag for outgoing multicast packets.
2482 * Must be zero or one.
2483 */
2484 if (m == NULL || m->m_len != sizeof(u_int)) {
2485 error = EINVAL;
2486 break;
2487 }
2488 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2489 if (loop > 1) {
2490 error = EINVAL;
2491 break;
2492 }
2493 im6o->im6o_multicast_loop = loop;
2494 break;
2495
2496 case IPV6_JOIN_GROUP:
2497 /*
2498 * Add a multicast group membership.
2499 * Group must be a valid IP6 multicast address.
2500 */
2501 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2502 error = EINVAL;
2503 break;
2504 }
2505 mreq = mtod(m, struct ipv6_mreq *);
2506
2507 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2508 /*
2509 * We use the unspecified address to specify to accept
2510 * all multicast addresses. Only super user is allowed
2511 * to do this.
2512 */
2513 /* XXX-BZ might need a better PRIV_NETINET_x for this */
2514 error = priv_check(curthread, PRIV_NETINET_MROUTE);
2515 if (error)
2516 break;
2517 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2518 error = EINVAL;
2519 break;
2520 }
2521
2522 /*
2523 * If no interface was explicitly specified, choose an
2524 * appropriate one according to the given multicast address.
2525 */
2526 if (mreq->ipv6mr_interface == 0) {
2527 struct sockaddr_in6 *dst;
2528
2529 /*
2530 * Look up the routing table for the
2531 * address, and choose the outgoing interface.
2532 * XXX: is it a good approach?
2533 */
2534 ro.ro_rt = NULL;
2535 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2536 bzero(dst, sizeof(*dst));
2537 dst->sin6_family = AF_INET6;
2538 dst->sin6_len = sizeof(*dst);
2539 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2540 rtalloc((struct route *)&ro);
2541 if (ro.ro_rt == NULL) {
2542 error = EADDRNOTAVAIL;
2543 break;
2544 }
2545 ifp = ro.ro_rt->rt_ifp;
2546 RTFREE(ro.ro_rt);
2547 } else {
2548 /*
2549 * If the interface is specified, validate it.
2550 */
2551 if (mreq->ipv6mr_interface < 0 ||
2552 if_index < mreq->ipv6mr_interface) {
2553 error = ENXIO; /* XXX EINVAL? */
2554 break;
2555 }
2556 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2557 if (!ifp) {
2558 error = ENXIO; /* XXX EINVAL? */
2559 break;
2560 }
2561 }
2562
2563 /*
2564 * See if we found an interface, and confirm that it
2565 * supports multicast
2566 */
2567 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2568 error = EADDRNOTAVAIL;
2569 break;
2570 }
2571
2572 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2573 error = EADDRNOTAVAIL; /* XXX: should not happen */
2574 break;
2575 }
2576
2577 /*
2578 * See if the membership already exists.
2579 */
2580 for (imm = im6o->im6o_memberships.lh_first;
2581 imm != NULL; imm = imm->i6mm_chain.le_next)
2582 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2583 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2584 &mreq->ipv6mr_multiaddr))
2585 break;
2586 if (imm != NULL) {
2587 error = EADDRINUSE;
2588 break;
2589 }
2590 /*
2591 * Everything looks good; add a new record to the multicast
2592 * address list for the given interface.
2593 */
2594 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0);
2595 if (imm == NULL)
2596 break;
2597 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2598 break;
2599
2600 case IPV6_LEAVE_GROUP:
2601 /*
2602 * Drop a multicast group membership.
2603 * Group must be a valid IP6 multicast address.
2604 */
2605 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2606 error = EINVAL;
2607 break;
2608 }
2609 mreq = mtod(m, struct ipv6_mreq *);
2610
2611 /*
2612 * If an interface address was specified, get a pointer
2613 * to its ifnet structure.
2614 */
2615 if (mreq->ipv6mr_interface < 0 ||
2616 if_index < mreq->ipv6mr_interface) {
2617 error = ENXIO; /* XXX EINVAL? */
2618 break;
2619 }
2620 if (mreq->ipv6mr_interface == 0)
2621 ifp = NULL;
2622 else
2623 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2624
2625 /* Fill in the scope zone ID */
2626 if (ifp) {
2627 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2628 /* XXX: should not happen */
2629 error = EADDRNOTAVAIL;
2630 break;
2631 }
2632 } else if (mreq->ipv6mr_interface != 0) {
2633 /*
2634 * This case happens when the (positive) index is in
2635 * the valid range, but the corresponding interface has
2636 * been detached dynamically (XXX).
2637 */
2638 error = EADDRNOTAVAIL;
2639 break;
2640 } else { /* ipv6mr_interface == 0 */
2641 struct sockaddr_in6 sa6_mc;
2642
2643 /*
2644 * The API spec says as follows:
2645 * If the interface index is specified as 0, the
2646 * system may choose a multicast group membership to
2647 * drop by matching the multicast address only.
2648 * On the other hand, we cannot disambiguate the scope
2649 * zone unless an interface is provided. Thus, we
2650 * check if there's ambiguity with the default scope
2651 * zone as the last resort.
2652 */
2653 bzero(&sa6_mc, sizeof(sa6_mc));
2654 sa6_mc.sin6_family = AF_INET6;
2655 sa6_mc.sin6_len = sizeof(sa6_mc);
2656 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2657 error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2658 if (error != 0)
2659 break;
2660 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2661 }
2662
2663 /*
2664 * Find the membership in the membership list.
2665 */
2666 for (imm = im6o->im6o_memberships.lh_first;
2667 imm != NULL; imm = imm->i6mm_chain.le_next) {
2668 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2669 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2670 &mreq->ipv6mr_multiaddr))
2671 break;
2672 }
2673 if (imm == NULL) {
2674 /* Unable to resolve interface */
2675 error = EADDRNOTAVAIL;
2676 break;
2677 }
2678 /*
2679 * Give up the multicast address record to which the
2680 * membership points.
2681 */
2682 LIST_REMOVE(imm, i6mm_chain);
2683 in6_delmulti(imm->i6mm_maddr);
2684 free(imm, M_IP6MADDR);
2685 break;
2686
2687 default:
2688 error = EOPNOTSUPP;
2689 break;
2690 }
2691
2692 /*
2693 * If all options have default values, no need to keep the mbuf.
2694 */
2695 if (im6o->im6o_multicast_ifp == NULL &&
2696 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2697 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2698 im6o->im6o_memberships.lh_first == NULL) {
2699 free(*im6op, M_IP6MOPTS);
2700 *im6op = NULL;
2701 }
2702
2703 return (error);
2704}
2705
2706/*
2707 * Return the IP6 multicast options in response to user getsockopt().
2708 */
2709static int
2710ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2711{
2712 u_int *hlim, *loop, *ifindex;
2713
2714 *mp = m_get(M_TRYWAIT, MT_HEADER); /* XXX */
2715
2716 switch (optname) {
2717
2718 case IPV6_MULTICAST_IF:
2719 ifindex = mtod(*mp, u_int *);
2720 (*mp)->m_len = sizeof(u_int);
2721 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2722 *ifindex = 0;
2723 else
2724 *ifindex = im6o->im6o_multicast_ifp->if_index;
2725 return (0);
2726
2727 case IPV6_MULTICAST_HOPS:
2728 hlim = mtod(*mp, u_int *);
2729 (*mp)->m_len = sizeof(u_int);
2730 if (im6o == NULL)
2731 *hlim = ip6_defmcasthlim;
2732 else
2733 *hlim = im6o->im6o_multicast_hlim;
2734 return (0);
2735
2736 case IPV6_MULTICAST_LOOP:
2737 loop = mtod(*mp, u_int *);
2738 (*mp)->m_len = sizeof(u_int);
2739 if (im6o == NULL)
2740 *loop = ip6_defmcasthlim;
2741 else
2742 *loop = im6o->im6o_multicast_loop;
2743 return (0);
2744
2745 default:
2746 return (EOPNOTSUPP);
2747 }
2748}
2749
2750/*
2751 * Discard the IP6 multicast options.
2752 */
2753void
2754ip6_freemoptions(struct ip6_moptions *im6o)
2755{
2756 struct in6_multi_mship *imm;
2757
2758 if (im6o == NULL)
2759 return;
2760
2761 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2762 LIST_REMOVE(imm, i6mm_chain);
2763 if (imm->i6mm_maddr)
2764 in6_delmulti(imm->i6mm_maddr);
2765 free(imm, M_IP6MADDR);
2766 }
2767 free(im6o, M_IP6MOPTS);
2768}
2769
2770/*
2771 * Set IPv6 outgoing packet options based on advanced API.
2772 */
2773int
2774ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2775 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2776{
2777 struct cmsghdr *cm = 0;
2778
2779 if (control == NULL || opt == NULL)
2780 return (EINVAL);
2781
2782 ip6_initpktopts(opt);
2783 if (stickyopt) {
2784 int error;
2785
2786 /*
2787 * If stickyopt is provided, make a local copy of the options
2788 * for this particular packet, then override them by ancillary
2789 * objects.
2790 * XXX: copypktopts() does not copy the cached route to a next
2791 * hop (if any). This is not very good in terms of efficiency,
2792 * but we can allow this since this option should be rarely
2793 * used.
2794 */
2795 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2796 return (error);
2797 }
2798
2799 /*
2800 * XXX: Currently, we assume all the optional information is stored
2801 * in a single mbuf.
2802 */
2803 if (control->m_next)
2804 return (EINVAL);
2805
2806 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2807 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2808 int error;
2809
2810 if (control->m_len < CMSG_LEN(0))
2811 return (EINVAL);
2812
2813 cm = mtod(control, struct cmsghdr *);
2814 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2815 return (EINVAL);
2816 if (cm->cmsg_level != IPPROTO_IPV6)
2817 continue;
2818
2819 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2820 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2821 if (error)
2822 return (error);
2823 }
2824
2825 return (0);
2826}
2827
2828/*
2829 * Set a particular packet option, as a sticky option or an ancillary data
2830 * item. "len" can be 0 only when it's a sticky option.
2831 * We have 4 cases of combination of "sticky" and "cmsg":
2832 * "sticky=0, cmsg=0": impossible
2833 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2834 * "sticky=1, cmsg=0": RFC3542 socket option
2835 * "sticky=1, cmsg=1": RFC2292 socket option
2836 */
2837static int
2838ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2839 struct ucred *cred, int sticky, int cmsg, int uproto)
2840{
2841 int minmtupolicy, preftemp;
2842 int error;
2843
2844 if (!sticky && !cmsg) {
2845#ifdef DIAGNOSTIC
2846 printf("ip6_setpktopt: impossible case\n");
2847#endif
2848 return (EINVAL);
2849 }
2850
2851 /*
2852 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2853 * not be specified in the context of RFC3542. Conversely,
2854 * RFC3542 types should not be specified in the context of RFC2292.
2855 */
2856 if (!cmsg) {
2857 switch (optname) {
2858 case IPV6_2292PKTINFO:
2859 case IPV6_2292HOPLIMIT:
2860 case IPV6_2292NEXTHOP:
2861 case IPV6_2292HOPOPTS:
2862 case IPV6_2292DSTOPTS:
2863 case IPV6_2292RTHDR:
2864 case IPV6_2292PKTOPTIONS:
2865 return (ENOPROTOOPT);
2866 }
2867 }
2868 if (sticky && cmsg) {
2869 switch (optname) {
2870 case IPV6_PKTINFO:
2871 case IPV6_HOPLIMIT:
2872 case IPV6_NEXTHOP:
2873 case IPV6_HOPOPTS:
2874 case IPV6_DSTOPTS:
2875 case IPV6_RTHDRDSTOPTS:
2876 case IPV6_RTHDR:
2877 case IPV6_USE_MIN_MTU:
2878 case IPV6_DONTFRAG:
2879 case IPV6_TCLASS:
2880 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2881 return (ENOPROTOOPT);
2882 }
2883 }
2884
2885 switch (optname) {
2886 case IPV6_2292PKTINFO:
2887 case IPV6_PKTINFO:
2888 {
2889 struct ifnet *ifp = NULL;
2890 struct in6_pktinfo *pktinfo;
2891
2892 if (len != sizeof(struct in6_pktinfo))
2893 return (EINVAL);
2894
2895 pktinfo = (struct in6_pktinfo *)buf;
2896
2897 /*
2898 * An application can clear any sticky IPV6_PKTINFO option by
2899 * doing a "regular" setsockopt with ipi6_addr being
2900 * in6addr_any and ipi6_ifindex being zero.
2901 * [RFC 3542, Section 6]
2902 */
2903 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2904 pktinfo->ipi6_ifindex == 0 &&
2905 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2906 ip6_clearpktopts(opt, optname);
2907 break;
2908 }
2909
2910 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2911 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2912 return (EINVAL);
2913 }
2914
2915 /* validate the interface index if specified. */
2916 if (pktinfo->ipi6_ifindex > if_index ||
2917 pktinfo->ipi6_ifindex < 0) {
2918 return (ENXIO);
2919 }
2920 if (pktinfo->ipi6_ifindex) {
2921 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2922 if (ifp == NULL)
2923 return (ENXIO);
2924 }
2925
2926 /*
2927 * We store the address anyway, and let in6_selectsrc()
2928 * validate the specified address. This is because ipi6_addr
2929 * may not have enough information about its scope zone, and
2930 * we may need additional information (such as outgoing
2931 * interface or the scope zone of a destination address) to
2932 * disambiguate the scope.
2933 * XXX: the delay of the validation may confuse the
2934 * application when it is used as a sticky option.
2935 */
2936 if (opt->ip6po_pktinfo == NULL) {
2937 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2938 M_IP6OPT, M_NOWAIT);
2939 if (opt->ip6po_pktinfo == NULL)
2940 return (ENOBUFS);
2941 }
2942 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2943 break;
2944 }
2945
2946 case IPV6_2292HOPLIMIT:
2947 case IPV6_HOPLIMIT:
2948 {
2949 int *hlimp;
2950
2951 /*
2952 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2953 * to simplify the ordering among hoplimit options.
2954 */
2955 if (optname == IPV6_HOPLIMIT && sticky)
2956 return (ENOPROTOOPT);
2957
2958 if (len != sizeof(int))
2959 return (EINVAL);
2960 hlimp = (int *)buf;
2961 if (*hlimp < -1 || *hlimp > 255)
2962 return (EINVAL);
2963
2964 opt->ip6po_hlim = *hlimp;
2965 break;
2966 }
2967
2968 case IPV6_TCLASS:
2969 {
2970 int tclass;
2971
2972 if (len != sizeof(int))
2973 return (EINVAL);
2974 tclass = *(int *)buf;
2975 if (tclass < -1 || tclass > 255)
2976 return (EINVAL);
2977
2978 opt->ip6po_tclass = tclass;
2979 break;
2980 }
2981
2982 case IPV6_2292NEXTHOP:
2983 case IPV6_NEXTHOP:
2984 if (cred != NULL) {
2985 error = priv_check_cred(cred,
2986 PRIV_NETINET_SETHDROPTS, 0);
2987 if (error)
2988 return (error);
2989 }
2990
2991 if (len == 0) { /* just remove the option */
2992 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2993 break;
2994 }
2995
2996 /* check if cmsg_len is large enough for sa_len */
2997 if (len < sizeof(struct sockaddr) || len < *buf)
2998 return (EINVAL);
2999
3000 switch (((struct sockaddr *)buf)->sa_family) {
3001 case AF_INET6:
3002 {
3003 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3004 int error;
3005
3006 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3007 return (EINVAL);
3008
3009 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3010 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3011 return (EINVAL);
3012 }
3013 if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3014 != 0) {
3015 return (error);
3016 }
3017 break;
3018 }
3019 case AF_LINK: /* should eventually be supported */
3020 default:
3021 return (EAFNOSUPPORT);
3022 }
3023
3024 /* turn off the previous option, then set the new option. */
3025 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3026 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3027 if (opt->ip6po_nexthop == NULL)
3028 return (ENOBUFS);
3029 bcopy(buf, opt->ip6po_nexthop, *buf);
3030 break;
3031
3032 case IPV6_2292HOPOPTS:
3033 case IPV6_HOPOPTS:
3034 {
3035 struct ip6_hbh *hbh;
3036 int hbhlen;
3037
3038 /*
3039 * XXX: We don't allow a non-privileged user to set ANY HbH
3040 * options, since per-option restriction has too much
3041 * overhead.
3042 */
3043 if (cred != NULL) {
3044 error = priv_check_cred(cred,
3045 PRIV_NETINET_SETHDROPTS, 0);
3046 if (error)
3047 return (error);
3048 }
3049
3050 if (len == 0) {
3051 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3052 break; /* just remove the option */
3053 }
3054
3055 /* message length validation */
3056 if (len < sizeof(struct ip6_hbh))
3057 return (EINVAL);
3058 hbh = (struct ip6_hbh *)buf;
3059 hbhlen = (hbh->ip6h_len + 1) << 3;
3060 if (len != hbhlen)
3061 return (EINVAL);
3062
3063 /* turn off the previous option, then set the new option. */
3064 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3065 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3066 if (opt->ip6po_hbh == NULL)
3067 return (ENOBUFS);
3068 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3069
3070 break;
3071 }
3072
3073 case IPV6_2292DSTOPTS:
3074 case IPV6_DSTOPTS:
3075 case IPV6_RTHDRDSTOPTS:
3076 {
3077 struct ip6_dest *dest, **newdest = NULL;
3078 int destlen;
3079
3080 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3081 error = priv_check_cred(cred,
3082 PRIV_NETINET_SETHDROPTS, 0);
3083 if (error)
3084 return (error);
3085 }
3086
3087 if (len == 0) {
3088 ip6_clearpktopts(opt, optname);
3089 break; /* just remove the option */
3090 }
3091
3092 /* message length validation */
3093 if (len < sizeof(struct ip6_dest))
3094 return (EINVAL);
3095 dest = (struct ip6_dest *)buf;
3096 destlen = (dest->ip6d_len + 1) << 3;
3097 if (len != destlen)
3098 return (EINVAL);
3099
3100 /*
3101 * Determine the position that the destination options header
3102 * should be inserted; before or after the routing header.
3103 */
3104 switch (optname) {
3105 case IPV6_2292DSTOPTS:
3106 /*
3107 * The old advacned API is ambiguous on this point.
3108 * Our approach is to determine the position based
3109 * according to the existence of a routing header.
3110 * Note, however, that this depends on the order of the
3111 * extension headers in the ancillary data; the 1st
3112 * part of the destination options header must appear
3113 * before the routing header in the ancillary data,
3114 * too.
3115 * RFC3542 solved the ambiguity by introducing
3116 * separate ancillary data or option types.
3117 */
3118 if (opt->ip6po_rthdr == NULL)
3119 newdest = &opt->ip6po_dest1;
3120 else
3121 newdest = &opt->ip6po_dest2;
3122 break;
3123 case IPV6_RTHDRDSTOPTS:
3124 newdest = &opt->ip6po_dest1;
3125 break;
3126 case IPV6_DSTOPTS:
3127 newdest = &opt->ip6po_dest2;
3128 break;
3129 }
3130
3131 /* turn off the previous option, then set the new option. */
3132 ip6_clearpktopts(opt, optname);
3133 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3134 if (*newdest == NULL)
3135 return (ENOBUFS);
3136 bcopy(dest, *newdest, destlen);
3137
3138 break;
3139 }
3140
3141 case IPV6_2292RTHDR:
3142 case IPV6_RTHDR:
3143 {
3144 struct ip6_rthdr *rth;
3145 int rthlen;
3146
3147 if (len == 0) {
3148 ip6_clearpktopts(opt, IPV6_RTHDR);
3149 break; /* just remove the option */
3150 }
3151
3152 /* message length validation */
3153 if (len < sizeof(struct ip6_rthdr))
3154 return (EINVAL);
3155 rth = (struct ip6_rthdr *)buf;
3156 rthlen = (rth->ip6r_len + 1) << 3;
3157 if (len != rthlen)
3158 return (EINVAL);
3159
3160 switch (rth->ip6r_type) {
3161 case IPV6_RTHDR_TYPE_0:
3162 if (rth->ip6r_len == 0) /* must contain one addr */
3163 return (EINVAL);
3164 if (rth->ip6r_len % 2) /* length must be even */
3165 return (EINVAL);
3166 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3167 return (EINVAL);
3168 break;
3169 default:
3170 return (EINVAL); /* not supported */
3171 }
3172
3173 /* turn off the previous option */
3174 ip6_clearpktopts(opt, IPV6_RTHDR);
3175 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3176 if (opt->ip6po_rthdr == NULL)
3177 return (ENOBUFS);
3178 bcopy(rth, opt->ip6po_rthdr, rthlen);
3179
3180 break;
3181 }
3182
3183 case IPV6_USE_MIN_MTU:
3184 if (len != sizeof(int))
3185 return (EINVAL);
3186 minmtupolicy = *(int *)buf;
3187 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3188 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3189 minmtupolicy != IP6PO_MINMTU_ALL) {
3190 return (EINVAL);
3191 }
3192 opt->ip6po_minmtu = minmtupolicy;
3193 break;
3194
3195 case IPV6_DONTFRAG:
3196 if (len != sizeof(int))
3197 return (EINVAL);
3198
3199 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3200 /*
3201 * we ignore this option for TCP sockets.
3202 * (RFC3542 leaves this case unspecified.)
3203 */
3204 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3205 } else
3206 opt->ip6po_flags |= IP6PO_DONTFRAG;
3207 break;
3208
3209 case IPV6_PREFER_TEMPADDR:
3210 if (len != sizeof(int))
3211 return (EINVAL);
3212 preftemp = *(int *)buf;
3213 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3214 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3215 preftemp != IP6PO_TEMPADDR_PREFER) {
3216 return (EINVAL);
3217 }
3218 opt->ip6po_prefer_tempaddr = preftemp;
3219 break;
3220
3221 default:
3222 return (ENOPROTOOPT);
3223 } /* end of switch */
3224
3225 return (0);
3226}
3227
3228/*
3229 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3230 * packet to the input queue of a specified interface. Note that this
3231 * calls the output routine of the loopback "driver", but with an interface
3232 * pointer that might NOT be &loif -- easier than replicating that code here.
3233 */
3234void
3235ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3236{
3237 struct mbuf *copym;
3238 struct ip6_hdr *ip6;
3239
3240 copym = m_copy(m, 0, M_COPYALL);
3241 if (copym == NULL)
3242 return;
3243
3244 /*
3245 * Make sure to deep-copy IPv6 header portion in case the data
3246 * is in an mbuf cluster, so that we can safely override the IPv6
3247 * header portion later.
3248 */
3249 if ((copym->m_flags & M_EXT) != 0 ||
3250 copym->m_len < sizeof(struct ip6_hdr)) {
3251 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3252 if (copym == NULL)
3253 return;
3254 }
3255
3256#ifdef DIAGNOSTIC
3257 if (copym->m_len < sizeof(*ip6)) {
3258 m_freem(copym);
3259 return;
3260 }
3261#endif
3262
3263 ip6 = mtod(copym, struct ip6_hdr *);
3264 /*
3265 * clear embedded scope identifiers if necessary.
3266 * in6_clearscope will touch the addresses only when necessary.
3267 */
3268 in6_clearscope(&ip6->ip6_src);
3269 in6_clearscope(&ip6->ip6_dst);
3270
3271 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3272}
3273
3274/*
3275 * Chop IPv6 header off from the payload.
3276 */
3277static int
3278ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3279{
3280 struct mbuf *mh;
3281 struct ip6_hdr *ip6;
3282
3283 ip6 = mtod(m, struct ip6_hdr *);
3284 if (m->m_len > sizeof(*ip6)) {
3285 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3286 if (mh == 0) {
3287 m_freem(m);
3288 return ENOBUFS;
3289 }
3290 M_MOVE_PKTHDR(mh, m);
3291 MH_ALIGN(mh, sizeof(*ip6));
3292 m->m_len -= sizeof(*ip6);
3293 m->m_data += sizeof(*ip6);
3294 mh->m_next = m;
3295 m = mh;
3296 m->m_len = sizeof(*ip6);
3297 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3298 }
3299 exthdrs->ip6e_ip6 = m;
3300 return 0;
3301}
3302
3303/*
3304 * Compute IPv6 extension header length.
3305 */
3306int
3307ip6_optlen(struct in6pcb *in6p)
3308{
3309 int len;
3310
3311 if (!in6p->in6p_outputopts)
3312 return 0;
3313
3314 len = 0;
3315#define elen(x) \
3316 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3317
3318 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3319 if (in6p->in6p_outputopts->ip6po_rthdr)
3320 /* dest1 is valid with rthdr only */
3321 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3322 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3323 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3324 return len;
3325#undef elen
3326}
| 1089
1090 return (error);
1091
1092freehdrs:
1093 m_freem(exthdrs.ip6e_hbh); /* m_freem will check if mbuf is 0 */
1094 m_freem(exthdrs.ip6e_dest1);
1095 m_freem(exthdrs.ip6e_rthdr);
1096 m_freem(exthdrs.ip6e_dest2);
1097 /* FALLTHROUGH */
1098bad:
1099 if (m)
1100 m_freem(m);
1101 goto done;
1102}
1103
1104static int
1105ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1106{
1107 struct mbuf *m;
1108
1109 if (hlen > MCLBYTES)
1110 return (ENOBUFS); /* XXX */
1111
1112 MGET(m, M_DONTWAIT, MT_DATA);
1113 if (!m)
1114 return (ENOBUFS);
1115
1116 if (hlen > MLEN) {
1117 MCLGET(m, M_DONTWAIT);
1118 if ((m->m_flags & M_EXT) == 0) {
1119 m_free(m);
1120 return (ENOBUFS);
1121 }
1122 }
1123 m->m_len = hlen;
1124 if (hdr)
1125 bcopy(hdr, mtod(m, caddr_t), hlen);
1126
1127 *mp = m;
1128 return (0);
1129}
1130
1131/*
1132 * Insert jumbo payload option.
1133 */
1134static int
1135ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1136{
1137 struct mbuf *mopt;
1138 u_char *optbuf;
1139 u_int32_t v;
1140
1141#define JUMBOOPTLEN 8 /* length of jumbo payload option and padding */
1142
1143 /*
1144 * If there is no hop-by-hop options header, allocate new one.
1145 * If there is one but it doesn't have enough space to store the
1146 * jumbo payload option, allocate a cluster to store the whole options.
1147 * Otherwise, use it to store the options.
1148 */
1149 if (exthdrs->ip6e_hbh == 0) {
1150 MGET(mopt, M_DONTWAIT, MT_DATA);
1151 if (mopt == 0)
1152 return (ENOBUFS);
1153 mopt->m_len = JUMBOOPTLEN;
1154 optbuf = mtod(mopt, u_char *);
1155 optbuf[1] = 0; /* = ((JUMBOOPTLEN) >> 3) - 1 */
1156 exthdrs->ip6e_hbh = mopt;
1157 } else {
1158 struct ip6_hbh *hbh;
1159
1160 mopt = exthdrs->ip6e_hbh;
1161 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1162 /*
1163 * XXX assumption:
1164 * - exthdrs->ip6e_hbh is not referenced from places
1165 * other than exthdrs.
1166 * - exthdrs->ip6e_hbh is not an mbuf chain.
1167 */
1168 int oldoptlen = mopt->m_len;
1169 struct mbuf *n;
1170
1171 /*
1172 * XXX: give up if the whole (new) hbh header does
1173 * not fit even in an mbuf cluster.
1174 */
1175 if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1176 return (ENOBUFS);
1177
1178 /*
1179 * As a consequence, we must always prepare a cluster
1180 * at this point.
1181 */
1182 MGET(n, M_DONTWAIT, MT_DATA);
1183 if (n) {
1184 MCLGET(n, M_DONTWAIT);
1185 if ((n->m_flags & M_EXT) == 0) {
1186 m_freem(n);
1187 n = NULL;
1188 }
1189 }
1190 if (!n)
1191 return (ENOBUFS);
1192 n->m_len = oldoptlen + JUMBOOPTLEN;
1193 bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1194 oldoptlen);
1195 optbuf = mtod(n, caddr_t) + oldoptlen;
1196 m_freem(mopt);
1197 mopt = exthdrs->ip6e_hbh = n;
1198 } else {
1199 optbuf = mtod(mopt, u_char *) + mopt->m_len;
1200 mopt->m_len += JUMBOOPTLEN;
1201 }
1202 optbuf[0] = IP6OPT_PADN;
1203 optbuf[1] = 1;
1204
1205 /*
1206 * Adjust the header length according to the pad and
1207 * the jumbo payload option.
1208 */
1209 hbh = mtod(mopt, struct ip6_hbh *);
1210 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1211 }
1212
1213 /* fill in the option. */
1214 optbuf[2] = IP6OPT_JUMBO;
1215 optbuf[3] = 4;
1216 v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1217 bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1218
1219 /* finally, adjust the packet header length */
1220 exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1221
1222 return (0);
1223#undef JUMBOOPTLEN
1224}
1225
1226/*
1227 * Insert fragment header and copy unfragmentable header portions.
1228 */
1229static int
1230ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1231 struct ip6_frag **frghdrp)
1232{
1233 struct mbuf *n, *mlast;
1234
1235 if (hlen > sizeof(struct ip6_hdr)) {
1236 n = m_copym(m0, sizeof(struct ip6_hdr),
1237 hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1238 if (n == 0)
1239 return (ENOBUFS);
1240 m->m_next = n;
1241 } else
1242 n = m;
1243
1244 /* Search for the last mbuf of unfragmentable part. */
1245 for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1246 ;
1247
1248 if ((mlast->m_flags & M_EXT) == 0 &&
1249 M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1250 /* use the trailing space of the last mbuf for the fragment hdr */
1251 *frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1252 mlast->m_len);
1253 mlast->m_len += sizeof(struct ip6_frag);
1254 m->m_pkthdr.len += sizeof(struct ip6_frag);
1255 } else {
1256 /* allocate a new mbuf for the fragment header */
1257 struct mbuf *mfrg;
1258
1259 MGET(mfrg, M_DONTWAIT, MT_DATA);
1260 if (mfrg == 0)
1261 return (ENOBUFS);
1262 mfrg->m_len = sizeof(struct ip6_frag);
1263 *frghdrp = mtod(mfrg, struct ip6_frag *);
1264 mlast->m_next = mfrg;
1265 }
1266
1267 return (0);
1268}
1269
1270static int
1271ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1272 struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1273 int *alwaysfragp)
1274{
1275 u_int32_t mtu = 0;
1276 int alwaysfrag = 0;
1277 int error = 0;
1278
1279 if (ro_pmtu != ro) {
1280 /* The first hop and the final destination may differ. */
1281 struct sockaddr_in6 *sa6_dst =
1282 (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1283 if (ro_pmtu->ro_rt &&
1284 ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1285 !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1286 RTFREE(ro_pmtu->ro_rt);
1287 ro_pmtu->ro_rt = (struct rtentry *)NULL;
1288 }
1289 if (ro_pmtu->ro_rt == NULL) {
1290 bzero(sa6_dst, sizeof(*sa6_dst));
1291 sa6_dst->sin6_family = AF_INET6;
1292 sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1293 sa6_dst->sin6_addr = *dst;
1294
1295 rtalloc((struct route *)ro_pmtu);
1296 }
1297 }
1298 if (ro_pmtu->ro_rt) {
1299 u_int32_t ifmtu;
1300 struct in_conninfo inc;
1301
1302 bzero(&inc, sizeof(inc));
1303 inc.inc_flags = 1; /* IPv6 */
1304 inc.inc6_faddr = *dst;
1305
1306 if (ifp == NULL)
1307 ifp = ro_pmtu->ro_rt->rt_ifp;
1308 ifmtu = IN6_LINKMTU(ifp);
1309 mtu = tcp_hc_getmtu(&inc);
1310 if (mtu)
1311 mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1312 else
1313 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1314 if (mtu == 0)
1315 mtu = ifmtu;
1316 else if (mtu < IPV6_MMTU) {
1317 /*
1318 * RFC2460 section 5, last paragraph:
1319 * if we record ICMPv6 too big message with
1320 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1321 * or smaller, with framgent header attached.
1322 * (fragment header is needed regardless from the
1323 * packet size, for translators to identify packets)
1324 */
1325 alwaysfrag = 1;
1326 mtu = IPV6_MMTU;
1327 } else if (mtu > ifmtu) {
1328 /*
1329 * The MTU on the route is larger than the MTU on
1330 * the interface! This shouldn't happen, unless the
1331 * MTU of the interface has been changed after the
1332 * interface was brought up. Change the MTU in the
1333 * route to match the interface MTU (as long as the
1334 * field isn't locked).
1335 */
1336 mtu = ifmtu;
1337 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1338 }
1339 } else if (ifp) {
1340 mtu = IN6_LINKMTU(ifp);
1341 } else
1342 error = EHOSTUNREACH; /* XXX */
1343
1344 *mtup = mtu;
1345 if (alwaysfragp)
1346 *alwaysfragp = alwaysfrag;
1347 return (error);
1348}
1349
1350/*
1351 * IP6 socket option processing.
1352 */
1353int
1354ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1355{
1356 int optdatalen, uproto;
1357 void *optdata;
1358 struct inpcb *in6p = sotoinpcb(so);
1359 int error, optval;
1360 int level, op, optname;
1361 int optlen;
1362 struct thread *td;
1363
1364 if (sopt) {
1365 level = sopt->sopt_level;
1366 op = sopt->sopt_dir;
1367 optname = sopt->sopt_name;
1368 optlen = sopt->sopt_valsize;
1369 td = sopt->sopt_td;
1370 } else {
1371 panic("ip6_ctloutput: arg soopt is NULL");
1372 }
1373 error = optval = 0;
1374
1375 uproto = (int)so->so_proto->pr_protocol;
1376
1377 if (level == IPPROTO_IPV6) {
1378 switch (op) {
1379
1380 case SOPT_SET:
1381 switch (optname) {
1382 case IPV6_2292PKTOPTIONS:
1383#ifdef IPV6_PKTOPTIONS
1384 case IPV6_PKTOPTIONS:
1385#endif
1386 {
1387 struct mbuf *m;
1388
1389 error = soopt_getm(sopt, &m); /* XXX */
1390 if (error != 0)
1391 break;
1392 error = soopt_mcopyin(sopt, m); /* XXX */
1393 if (error != 0)
1394 break;
1395 error = ip6_pcbopts(&in6p->in6p_outputopts,
1396 m, so, sopt);
1397 m_freem(m); /* XXX */
1398 break;
1399 }
1400
1401 /*
1402 * Use of some Hop-by-Hop options or some
1403 * Destination options, might require special
1404 * privilege. That is, normal applications
1405 * (without special privilege) might be forbidden
1406 * from setting certain options in outgoing packets,
1407 * and might never see certain options in received
1408 * packets. [RFC 2292 Section 6]
1409 * KAME specific note:
1410 * KAME prevents non-privileged users from sending or
1411 * receiving ANY hbh/dst options in order to avoid
1412 * overhead of parsing options in the kernel.
1413 */
1414 case IPV6_RECVHOPOPTS:
1415 case IPV6_RECVDSTOPTS:
1416 case IPV6_RECVRTHDRDSTOPTS:
1417 if (td != NULL) {
1418 error = priv_check(td,
1419 PRIV_NETINET_SETHDROPTS);
1420 if (error)
1421 break;
1422 }
1423 /* FALLTHROUGH */
1424 case IPV6_UNICAST_HOPS:
1425 case IPV6_HOPLIMIT:
1426 case IPV6_FAITH:
1427
1428 case IPV6_RECVPKTINFO:
1429 case IPV6_RECVHOPLIMIT:
1430 case IPV6_RECVRTHDR:
1431 case IPV6_RECVPATHMTU:
1432 case IPV6_RECVTCLASS:
1433 case IPV6_V6ONLY:
1434 case IPV6_AUTOFLOWLABEL:
1435 if (optlen != sizeof(int)) {
1436 error = EINVAL;
1437 break;
1438 }
1439 error = sooptcopyin(sopt, &optval,
1440 sizeof optval, sizeof optval);
1441 if (error)
1442 break;
1443 switch (optname) {
1444
1445 case IPV6_UNICAST_HOPS:
1446 if (optval < -1 || optval >= 256)
1447 error = EINVAL;
1448 else {
1449 /* -1 = kernel default */
1450 in6p->in6p_hops = optval;
1451 if ((in6p->in6p_vflag &
1452 INP_IPV4) != 0)
1453 in6p->inp_ip_ttl = optval;
1454 }
1455 break;
1456#define OPTSET(bit) \
1457do { \
1458 if (optval) \
1459 in6p->in6p_flags |= (bit); \
1460 else \
1461 in6p->in6p_flags &= ~(bit); \
1462} while (/*CONSTCOND*/ 0)
1463#define OPTSET2292(bit) \
1464do { \
1465 in6p->in6p_flags |= IN6P_RFC2292; \
1466 if (optval) \
1467 in6p->in6p_flags |= (bit); \
1468 else \
1469 in6p->in6p_flags &= ~(bit); \
1470} while (/*CONSTCOND*/ 0)
1471#define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1472
1473 case IPV6_RECVPKTINFO:
1474 /* cannot mix with RFC2292 */
1475 if (OPTBIT(IN6P_RFC2292)) {
1476 error = EINVAL;
1477 break;
1478 }
1479 OPTSET(IN6P_PKTINFO);
1480 break;
1481
1482 case IPV6_HOPLIMIT:
1483 {
1484 struct ip6_pktopts **optp;
1485
1486 /* cannot mix with RFC2292 */
1487 if (OPTBIT(IN6P_RFC2292)) {
1488 error = EINVAL;
1489 break;
1490 }
1491 optp = &in6p->in6p_outputopts;
1492 error = ip6_pcbopt(IPV6_HOPLIMIT,
1493 (u_char *)&optval, sizeof(optval),
1494 optp, (td != NULL) ? td->td_ucred :
1495 NULL, uproto);
1496 break;
1497 }
1498
1499 case IPV6_RECVHOPLIMIT:
1500 /* cannot mix with RFC2292 */
1501 if (OPTBIT(IN6P_RFC2292)) {
1502 error = EINVAL;
1503 break;
1504 }
1505 OPTSET(IN6P_HOPLIMIT);
1506 break;
1507
1508 case IPV6_RECVHOPOPTS:
1509 /* cannot mix with RFC2292 */
1510 if (OPTBIT(IN6P_RFC2292)) {
1511 error = EINVAL;
1512 break;
1513 }
1514 OPTSET(IN6P_HOPOPTS);
1515 break;
1516
1517 case IPV6_RECVDSTOPTS:
1518 /* cannot mix with RFC2292 */
1519 if (OPTBIT(IN6P_RFC2292)) {
1520 error = EINVAL;
1521 break;
1522 }
1523 OPTSET(IN6P_DSTOPTS);
1524 break;
1525
1526 case IPV6_RECVRTHDRDSTOPTS:
1527 /* cannot mix with RFC2292 */
1528 if (OPTBIT(IN6P_RFC2292)) {
1529 error = EINVAL;
1530 break;
1531 }
1532 OPTSET(IN6P_RTHDRDSTOPTS);
1533 break;
1534
1535 case IPV6_RECVRTHDR:
1536 /* cannot mix with RFC2292 */
1537 if (OPTBIT(IN6P_RFC2292)) {
1538 error = EINVAL;
1539 break;
1540 }
1541 OPTSET(IN6P_RTHDR);
1542 break;
1543
1544 case IPV6_FAITH:
1545 OPTSET(IN6P_FAITH);
1546 break;
1547
1548 case IPV6_RECVPATHMTU:
1549 /*
1550 * We ignore this option for TCP
1551 * sockets.
1552 * (RFC3542 leaves this case
1553 * unspecified.)
1554 */
1555 if (uproto != IPPROTO_TCP)
1556 OPTSET(IN6P_MTU);
1557 break;
1558
1559 case IPV6_V6ONLY:
1560 /*
1561 * make setsockopt(IPV6_V6ONLY)
1562 * available only prior to bind(2).
1563 * see ipng mailing list, Jun 22 2001.
1564 */
1565 if (in6p->in6p_lport ||
1566 !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1567 error = EINVAL;
1568 break;
1569 }
1570 OPTSET(IN6P_IPV6_V6ONLY);
1571 if (optval)
1572 in6p->in6p_vflag &= ~INP_IPV4;
1573 else
1574 in6p->in6p_vflag |= INP_IPV4;
1575 break;
1576 case IPV6_RECVTCLASS:
1577 /* cannot mix with RFC2292 XXX */
1578 if (OPTBIT(IN6P_RFC2292)) {
1579 error = EINVAL;
1580 break;
1581 }
1582 OPTSET(IN6P_TCLASS);
1583 break;
1584 case IPV6_AUTOFLOWLABEL:
1585 OPTSET(IN6P_AUTOFLOWLABEL);
1586 break;
1587
1588 }
1589 break;
1590
1591 case IPV6_TCLASS:
1592 case IPV6_DONTFRAG:
1593 case IPV6_USE_MIN_MTU:
1594 case IPV6_PREFER_TEMPADDR:
1595 if (optlen != sizeof(optval)) {
1596 error = EINVAL;
1597 break;
1598 }
1599 error = sooptcopyin(sopt, &optval,
1600 sizeof optval, sizeof optval);
1601 if (error)
1602 break;
1603 {
1604 struct ip6_pktopts **optp;
1605 optp = &in6p->in6p_outputopts;
1606 error = ip6_pcbopt(optname,
1607 (u_char *)&optval, sizeof(optval),
1608 optp, (td != NULL) ? td->td_ucred :
1609 NULL, uproto);
1610 break;
1611 }
1612
1613 case IPV6_2292PKTINFO:
1614 case IPV6_2292HOPLIMIT:
1615 case IPV6_2292HOPOPTS:
1616 case IPV6_2292DSTOPTS:
1617 case IPV6_2292RTHDR:
1618 /* RFC 2292 */
1619 if (optlen != sizeof(int)) {
1620 error = EINVAL;
1621 break;
1622 }
1623 error = sooptcopyin(sopt, &optval,
1624 sizeof optval, sizeof optval);
1625 if (error)
1626 break;
1627 switch (optname) {
1628 case IPV6_2292PKTINFO:
1629 OPTSET2292(IN6P_PKTINFO);
1630 break;
1631 case IPV6_2292HOPLIMIT:
1632 OPTSET2292(IN6P_HOPLIMIT);
1633 break;
1634 case IPV6_2292HOPOPTS:
1635 /*
1636 * Check super-user privilege.
1637 * See comments for IPV6_RECVHOPOPTS.
1638 */
1639 if (td != NULL) {
1640 error = priv_check(td,
1641 PRIV_NETINET_SETHDROPTS);
1642 if (error)
1643 return (error);
1644 }
1645 OPTSET2292(IN6P_HOPOPTS);
1646 break;
1647 case IPV6_2292DSTOPTS:
1648 if (td != NULL) {
1649 error = priv_check(td,
1650 PRIV_NETINET_SETHDROPTS);
1651 if (error)
1652 return (error);
1653 }
1654 OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1655 break;
1656 case IPV6_2292RTHDR:
1657 OPTSET2292(IN6P_RTHDR);
1658 break;
1659 }
1660 break;
1661 case IPV6_PKTINFO:
1662 case IPV6_HOPOPTS:
1663 case IPV6_RTHDR:
1664 case IPV6_DSTOPTS:
1665 case IPV6_RTHDRDSTOPTS:
1666 case IPV6_NEXTHOP:
1667 {
1668 /* new advanced API (RFC3542) */
1669 u_char *optbuf;
1670 u_char optbuf_storage[MCLBYTES];
1671 int optlen;
1672 struct ip6_pktopts **optp;
1673
1674 /* cannot mix with RFC2292 */
1675 if (OPTBIT(IN6P_RFC2292)) {
1676 error = EINVAL;
1677 break;
1678 }
1679
1680 /*
1681 * We only ensure valsize is not too large
1682 * here. Further validation will be done
1683 * later.
1684 */
1685 error = sooptcopyin(sopt, optbuf_storage,
1686 sizeof(optbuf_storage), 0);
1687 if (error)
1688 break;
1689 optlen = sopt->sopt_valsize;
1690 optbuf = optbuf_storage;
1691 optp = &in6p->in6p_outputopts;
1692 error = ip6_pcbopt(optname, optbuf, optlen,
1693 optp, (td != NULL) ? td->td_ucred : NULL,
1694 uproto);
1695 break;
1696 }
1697#undef OPTSET
1698
1699 case IPV6_MULTICAST_IF:
1700 case IPV6_MULTICAST_HOPS:
1701 case IPV6_MULTICAST_LOOP:
1702 case IPV6_JOIN_GROUP:
1703 case IPV6_LEAVE_GROUP:
1704 {
1705 if (sopt->sopt_valsize > MLEN) {
1706 error = EMSGSIZE;
1707 break;
1708 }
1709 /* XXX */
1710 }
1711 /* FALLTHROUGH */
1712 {
1713 struct mbuf *m;
1714
1715 if (sopt->sopt_valsize > MCLBYTES) {
1716 error = EMSGSIZE;
1717 break;
1718 }
1719 /* XXX */
1720 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1721 if (m == 0) {
1722 error = ENOBUFS;
1723 break;
1724 }
1725 if (sopt->sopt_valsize > MLEN) {
1726 MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1727 if ((m->m_flags & M_EXT) == 0) {
1728 m_free(m);
1729 error = ENOBUFS;
1730 break;
1731 }
1732 }
1733 m->m_len = sopt->sopt_valsize;
1734 error = sooptcopyin(sopt, mtod(m, char *),
1735 m->m_len, m->m_len);
1736 if (error) {
1737 (void)m_free(m);
1738 break;
1739 }
1740 error = ip6_setmoptions(sopt->sopt_name,
1741 &in6p->in6p_moptions,
1742 m);
1743 (void)m_free(m);
1744 }
1745 break;
1746
1747 case IPV6_PORTRANGE:
1748 error = sooptcopyin(sopt, &optval,
1749 sizeof optval, sizeof optval);
1750 if (error)
1751 break;
1752
1753 switch (optval) {
1754 case IPV6_PORTRANGE_DEFAULT:
1755 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1756 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1757 break;
1758
1759 case IPV6_PORTRANGE_HIGH:
1760 in6p->in6p_flags &= ~(IN6P_LOWPORT);
1761 in6p->in6p_flags |= IN6P_HIGHPORT;
1762 break;
1763
1764 case IPV6_PORTRANGE_LOW:
1765 in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1766 in6p->in6p_flags |= IN6P_LOWPORT;
1767 break;
1768
1769 default:
1770 error = EINVAL;
1771 break;
1772 }
1773 break;
1774
1775#ifdef IPSEC
1776 case IPV6_IPSEC_POLICY:
1777 {
1778 caddr_t req;
1779 struct mbuf *m;
1780
1781 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1782 break;
1783 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1784 break;
1785 req = mtod(m, caddr_t);
1786 error = ipsec6_set_policy(in6p, optname, req,
1787 m->m_len, (sopt->sopt_td != NULL) ?
1788 sopt->sopt_td->td_ucred : NULL);
1789 m_freem(m);
1790 break;
1791 }
1792#endif /* IPSEC */
1793
1794 default:
1795 error = ENOPROTOOPT;
1796 break;
1797 }
1798 break;
1799
1800 case SOPT_GET:
1801 switch (optname) {
1802
1803 case IPV6_2292PKTOPTIONS:
1804#ifdef IPV6_PKTOPTIONS
1805 case IPV6_PKTOPTIONS:
1806#endif
1807 /*
1808 * RFC3542 (effectively) deprecated the
1809 * semantics of the 2292-style pktoptions.
1810 * Since it was not reliable in nature (i.e.,
1811 * applications had to expect the lack of some
1812 * information after all), it would make sense
1813 * to simplify this part by always returning
1814 * empty data.
1815 */
1816 sopt->sopt_valsize = 0;
1817 break;
1818
1819 case IPV6_RECVHOPOPTS:
1820 case IPV6_RECVDSTOPTS:
1821 case IPV6_RECVRTHDRDSTOPTS:
1822 case IPV6_UNICAST_HOPS:
1823 case IPV6_RECVPKTINFO:
1824 case IPV6_RECVHOPLIMIT:
1825 case IPV6_RECVRTHDR:
1826 case IPV6_RECVPATHMTU:
1827
1828 case IPV6_FAITH:
1829 case IPV6_V6ONLY:
1830 case IPV6_PORTRANGE:
1831 case IPV6_RECVTCLASS:
1832 case IPV6_AUTOFLOWLABEL:
1833 switch (optname) {
1834
1835 case IPV6_RECVHOPOPTS:
1836 optval = OPTBIT(IN6P_HOPOPTS);
1837 break;
1838
1839 case IPV6_RECVDSTOPTS:
1840 optval = OPTBIT(IN6P_DSTOPTS);
1841 break;
1842
1843 case IPV6_RECVRTHDRDSTOPTS:
1844 optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1845 break;
1846
1847 case IPV6_UNICAST_HOPS:
1848 optval = in6p->in6p_hops;
1849 break;
1850
1851 case IPV6_RECVPKTINFO:
1852 optval = OPTBIT(IN6P_PKTINFO);
1853 break;
1854
1855 case IPV6_RECVHOPLIMIT:
1856 optval = OPTBIT(IN6P_HOPLIMIT);
1857 break;
1858
1859 case IPV6_RECVRTHDR:
1860 optval = OPTBIT(IN6P_RTHDR);
1861 break;
1862
1863 case IPV6_RECVPATHMTU:
1864 optval = OPTBIT(IN6P_MTU);
1865 break;
1866
1867 case IPV6_FAITH:
1868 optval = OPTBIT(IN6P_FAITH);
1869 break;
1870
1871 case IPV6_V6ONLY:
1872 optval = OPTBIT(IN6P_IPV6_V6ONLY);
1873 break;
1874
1875 case IPV6_PORTRANGE:
1876 {
1877 int flags;
1878 flags = in6p->in6p_flags;
1879 if (flags & IN6P_HIGHPORT)
1880 optval = IPV6_PORTRANGE_HIGH;
1881 else if (flags & IN6P_LOWPORT)
1882 optval = IPV6_PORTRANGE_LOW;
1883 else
1884 optval = 0;
1885 break;
1886 }
1887 case IPV6_RECVTCLASS:
1888 optval = OPTBIT(IN6P_TCLASS);
1889 break;
1890
1891 case IPV6_AUTOFLOWLABEL:
1892 optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1893 break;
1894 }
1895 if (error)
1896 break;
1897 error = sooptcopyout(sopt, &optval,
1898 sizeof optval);
1899 break;
1900
1901 case IPV6_PATHMTU:
1902 {
1903 u_long pmtu = 0;
1904 struct ip6_mtuinfo mtuinfo;
1905 struct route_in6 sro;
1906
1907 bzero(&sro, sizeof(sro));
1908
1909 if (!(so->so_state & SS_ISCONNECTED))
1910 return (ENOTCONN);
1911 /*
1912 * XXX: we dot not consider the case of source
1913 * routing, or optional information to specify
1914 * the outgoing interface.
1915 */
1916 error = ip6_getpmtu(&sro, NULL, NULL,
1917 &in6p->in6p_faddr, &pmtu, NULL);
1918 if (sro.ro_rt)
1919 RTFREE(sro.ro_rt);
1920 if (error)
1921 break;
1922 if (pmtu > IPV6_MAXPACKET)
1923 pmtu = IPV6_MAXPACKET;
1924
1925 bzero(&mtuinfo, sizeof(mtuinfo));
1926 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1927 optdata = (void *)&mtuinfo;
1928 optdatalen = sizeof(mtuinfo);
1929 error = sooptcopyout(sopt, optdata,
1930 optdatalen);
1931 break;
1932 }
1933
1934 case IPV6_2292PKTINFO:
1935 case IPV6_2292HOPLIMIT:
1936 case IPV6_2292HOPOPTS:
1937 case IPV6_2292RTHDR:
1938 case IPV6_2292DSTOPTS:
1939 switch (optname) {
1940 case IPV6_2292PKTINFO:
1941 optval = OPTBIT(IN6P_PKTINFO);
1942 break;
1943 case IPV6_2292HOPLIMIT:
1944 optval = OPTBIT(IN6P_HOPLIMIT);
1945 break;
1946 case IPV6_2292HOPOPTS:
1947 optval = OPTBIT(IN6P_HOPOPTS);
1948 break;
1949 case IPV6_2292RTHDR:
1950 optval = OPTBIT(IN6P_RTHDR);
1951 break;
1952 case IPV6_2292DSTOPTS:
1953 optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1954 break;
1955 }
1956 error = sooptcopyout(sopt, &optval,
1957 sizeof optval);
1958 break;
1959 case IPV6_PKTINFO:
1960 case IPV6_HOPOPTS:
1961 case IPV6_RTHDR:
1962 case IPV6_DSTOPTS:
1963 case IPV6_RTHDRDSTOPTS:
1964 case IPV6_NEXTHOP:
1965 case IPV6_TCLASS:
1966 case IPV6_DONTFRAG:
1967 case IPV6_USE_MIN_MTU:
1968 case IPV6_PREFER_TEMPADDR:
1969 error = ip6_getpcbopt(in6p->in6p_outputopts,
1970 optname, sopt);
1971 break;
1972
1973 case IPV6_MULTICAST_IF:
1974 case IPV6_MULTICAST_HOPS:
1975 case IPV6_MULTICAST_LOOP:
1976 case IPV6_JOIN_GROUP:
1977 case IPV6_LEAVE_GROUP:
1978 {
1979 struct mbuf *m;
1980 error = ip6_getmoptions(sopt->sopt_name,
1981 in6p->in6p_moptions, &m);
1982 if (error == 0)
1983 error = sooptcopyout(sopt,
1984 mtod(m, char *), m->m_len);
1985 m_freem(m);
1986 }
1987 break;
1988
1989#ifdef IPSEC
1990 case IPV6_IPSEC_POLICY:
1991 {
1992 caddr_t req = NULL;
1993 size_t len = 0;
1994 struct mbuf *m = NULL;
1995 struct mbuf **mp = &m;
1996 size_t ovalsize = sopt->sopt_valsize;
1997 caddr_t oval = (caddr_t)sopt->sopt_val;
1998
1999 error = soopt_getm(sopt, &m); /* XXX */
2000 if (error != 0)
2001 break;
2002 error = soopt_mcopyin(sopt, m); /* XXX */
2003 if (error != 0)
2004 break;
2005 sopt->sopt_valsize = ovalsize;
2006 sopt->sopt_val = oval;
2007 if (m) {
2008 req = mtod(m, caddr_t);
2009 len = m->m_len;
2010 }
2011 error = ipsec6_get_policy(in6p, req, len, mp);
2012 if (error == 0)
2013 error = soopt_mcopyout(sopt, m); /* XXX */
2014 if (error == 0 && m)
2015 m_freem(m);
2016 break;
2017 }
2018#endif /* IPSEC */
2019
2020 default:
2021 error = ENOPROTOOPT;
2022 break;
2023 }
2024 break;
2025 }
2026 } else { /* level != IPPROTO_IPV6 */
2027 error = EINVAL;
2028 }
2029 return (error);
2030}
2031
2032int
2033ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2034{
2035 int error = 0, optval, optlen;
2036 const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2037 struct in6pcb *in6p = sotoin6pcb(so);
2038 int level, op, optname;
2039
2040 if (sopt) {
2041 level = sopt->sopt_level;
2042 op = sopt->sopt_dir;
2043 optname = sopt->sopt_name;
2044 optlen = sopt->sopt_valsize;
2045 } else
2046 panic("ip6_raw_ctloutput: arg soopt is NULL");
2047
2048 if (level != IPPROTO_IPV6) {
2049 return (EINVAL);
2050 }
2051
2052 switch (optname) {
2053 case IPV6_CHECKSUM:
2054 /*
2055 * For ICMPv6 sockets, no modification allowed for checksum
2056 * offset, permit "no change" values to help existing apps.
2057 *
2058 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2059 * for an ICMPv6 socket will fail."
2060 * The current behavior does not meet RFC3542.
2061 */
2062 switch (op) {
2063 case SOPT_SET:
2064 if (optlen != sizeof(int)) {
2065 error = EINVAL;
2066 break;
2067 }
2068 error = sooptcopyin(sopt, &optval, sizeof(optval),
2069 sizeof(optval));
2070 if (error)
2071 break;
2072 if ((optval % 2) != 0) {
2073 /* the API assumes even offset values */
2074 error = EINVAL;
2075 } else if (so->so_proto->pr_protocol ==
2076 IPPROTO_ICMPV6) {
2077 if (optval != icmp6off)
2078 error = EINVAL;
2079 } else
2080 in6p->in6p_cksum = optval;
2081 break;
2082
2083 case SOPT_GET:
2084 if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2085 optval = icmp6off;
2086 else
2087 optval = in6p->in6p_cksum;
2088
2089 error = sooptcopyout(sopt, &optval, sizeof(optval));
2090 break;
2091
2092 default:
2093 error = EINVAL;
2094 break;
2095 }
2096 break;
2097
2098 default:
2099 error = ENOPROTOOPT;
2100 break;
2101 }
2102
2103 return (error);
2104}
2105
2106/*
2107 * Set up IP6 options in pcb for insertion in output packets or
2108 * specifying behavior of outgoing packets.
2109 */
2110static int
2111ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2112 struct socket *so, struct sockopt *sopt)
2113{
2114 struct ip6_pktopts *opt = *pktopt;
2115 int error = 0;
2116 struct thread *td = sopt->sopt_td;
2117
2118 /* turn off any old options. */
2119 if (opt) {
2120#ifdef DIAGNOSTIC
2121 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2122 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2123 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2124 printf("ip6_pcbopts: all specified options are cleared.\n");
2125#endif
2126 ip6_clearpktopts(opt, -1);
2127 } else
2128 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2129 *pktopt = NULL;
2130
2131 if (!m || m->m_len == 0) {
2132 /*
2133 * Only turning off any previous options, regardless of
2134 * whether the opt is just created or given.
2135 */
2136 free(opt, M_IP6OPT);
2137 return (0);
2138 }
2139
2140 /* set options specified by user. */
2141 if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2142 td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2143 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2144 free(opt, M_IP6OPT);
2145 return (error);
2146 }
2147 *pktopt = opt;
2148 return (0);
2149}
2150
2151/*
2152 * initialize ip6_pktopts. beware that there are non-zero default values in
2153 * the struct.
2154 */
2155void
2156ip6_initpktopts(struct ip6_pktopts *opt)
2157{
2158
2159 bzero(opt, sizeof(*opt));
2160 opt->ip6po_hlim = -1; /* -1 means default hop limit */
2161 opt->ip6po_tclass = -1; /* -1 means default traffic class */
2162 opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2163 opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2164}
2165
2166static int
2167ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2168 struct ucred *cred, int uproto)
2169{
2170 struct ip6_pktopts *opt;
2171
2172 if (*pktopt == NULL) {
2173 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2174 M_WAITOK);
2175 ip6_initpktopts(*pktopt);
2176 }
2177 opt = *pktopt;
2178
2179 return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2180}
2181
2182static int
2183ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2184{
2185 void *optdata = NULL;
2186 int optdatalen = 0;
2187 struct ip6_ext *ip6e;
2188 int error = 0;
2189 struct in6_pktinfo null_pktinfo;
2190 int deftclass = 0, on;
2191 int defminmtu = IP6PO_MINMTU_MCASTONLY;
2192 int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2193
2194 switch (optname) {
2195 case IPV6_PKTINFO:
2196 if (pktopt && pktopt->ip6po_pktinfo)
2197 optdata = (void *)pktopt->ip6po_pktinfo;
2198 else {
2199 /* XXX: we don't have to do this every time... */
2200 bzero(&null_pktinfo, sizeof(null_pktinfo));
2201 optdata = (void *)&null_pktinfo;
2202 }
2203 optdatalen = sizeof(struct in6_pktinfo);
2204 break;
2205 case IPV6_TCLASS:
2206 if (pktopt && pktopt->ip6po_tclass >= 0)
2207 optdata = (void *)&pktopt->ip6po_tclass;
2208 else
2209 optdata = (void *)&deftclass;
2210 optdatalen = sizeof(int);
2211 break;
2212 case IPV6_HOPOPTS:
2213 if (pktopt && pktopt->ip6po_hbh) {
2214 optdata = (void *)pktopt->ip6po_hbh;
2215 ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2216 optdatalen = (ip6e->ip6e_len + 1) << 3;
2217 }
2218 break;
2219 case IPV6_RTHDR:
2220 if (pktopt && pktopt->ip6po_rthdr) {
2221 optdata = (void *)pktopt->ip6po_rthdr;
2222 ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2223 optdatalen = (ip6e->ip6e_len + 1) << 3;
2224 }
2225 break;
2226 case IPV6_RTHDRDSTOPTS:
2227 if (pktopt && pktopt->ip6po_dest1) {
2228 optdata = (void *)pktopt->ip6po_dest1;
2229 ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2230 optdatalen = (ip6e->ip6e_len + 1) << 3;
2231 }
2232 break;
2233 case IPV6_DSTOPTS:
2234 if (pktopt && pktopt->ip6po_dest2) {
2235 optdata = (void *)pktopt->ip6po_dest2;
2236 ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2237 optdatalen = (ip6e->ip6e_len + 1) << 3;
2238 }
2239 break;
2240 case IPV6_NEXTHOP:
2241 if (pktopt && pktopt->ip6po_nexthop) {
2242 optdata = (void *)pktopt->ip6po_nexthop;
2243 optdatalen = pktopt->ip6po_nexthop->sa_len;
2244 }
2245 break;
2246 case IPV6_USE_MIN_MTU:
2247 if (pktopt)
2248 optdata = (void *)&pktopt->ip6po_minmtu;
2249 else
2250 optdata = (void *)&defminmtu;
2251 optdatalen = sizeof(int);
2252 break;
2253 case IPV6_DONTFRAG:
2254 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2255 on = 1;
2256 else
2257 on = 0;
2258 optdata = (void *)&on;
2259 optdatalen = sizeof(on);
2260 break;
2261 case IPV6_PREFER_TEMPADDR:
2262 if (pktopt)
2263 optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2264 else
2265 optdata = (void *)&defpreftemp;
2266 optdatalen = sizeof(int);
2267 break;
2268 default: /* should not happen */
2269#ifdef DIAGNOSTIC
2270 panic("ip6_getpcbopt: unexpected option\n");
2271#endif
2272 return (ENOPROTOOPT);
2273 }
2274
2275 error = sooptcopyout(sopt, optdata, optdatalen);
2276
2277 return (error);
2278}
2279
2280void
2281ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2282{
2283 if (pktopt == NULL)
2284 return;
2285
2286 if (optname == -1 || optname == IPV6_PKTINFO) {
2287 if (pktopt->ip6po_pktinfo)
2288 free(pktopt->ip6po_pktinfo, M_IP6OPT);
2289 pktopt->ip6po_pktinfo = NULL;
2290 }
2291 if (optname == -1 || optname == IPV6_HOPLIMIT)
2292 pktopt->ip6po_hlim = -1;
2293 if (optname == -1 || optname == IPV6_TCLASS)
2294 pktopt->ip6po_tclass = -1;
2295 if (optname == -1 || optname == IPV6_NEXTHOP) {
2296 if (pktopt->ip6po_nextroute.ro_rt) {
2297 RTFREE(pktopt->ip6po_nextroute.ro_rt);
2298 pktopt->ip6po_nextroute.ro_rt = NULL;
2299 }
2300 if (pktopt->ip6po_nexthop)
2301 free(pktopt->ip6po_nexthop, M_IP6OPT);
2302 pktopt->ip6po_nexthop = NULL;
2303 }
2304 if (optname == -1 || optname == IPV6_HOPOPTS) {
2305 if (pktopt->ip6po_hbh)
2306 free(pktopt->ip6po_hbh, M_IP6OPT);
2307 pktopt->ip6po_hbh = NULL;
2308 }
2309 if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2310 if (pktopt->ip6po_dest1)
2311 free(pktopt->ip6po_dest1, M_IP6OPT);
2312 pktopt->ip6po_dest1 = NULL;
2313 }
2314 if (optname == -1 || optname == IPV6_RTHDR) {
2315 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2316 free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2317 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2318 if (pktopt->ip6po_route.ro_rt) {
2319 RTFREE(pktopt->ip6po_route.ro_rt);
2320 pktopt->ip6po_route.ro_rt = NULL;
2321 }
2322 }
2323 if (optname == -1 || optname == IPV6_DSTOPTS) {
2324 if (pktopt->ip6po_dest2)
2325 free(pktopt->ip6po_dest2, M_IP6OPT);
2326 pktopt->ip6po_dest2 = NULL;
2327 }
2328}
2329
2330#define PKTOPT_EXTHDRCPY(type) \
2331do {\
2332 if (src->type) {\
2333 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2334 dst->type = malloc(hlen, M_IP6OPT, canwait);\
2335 if (dst->type == NULL && canwait == M_NOWAIT)\
2336 goto bad;\
2337 bcopy(src->type, dst->type, hlen);\
2338 }\
2339} while (/*CONSTCOND*/ 0)
2340
2341static int
2342copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2343{
2344 if (dst == NULL || src == NULL) {
2345 printf("ip6_clearpktopts: invalid argument\n");
2346 return (EINVAL);
2347 }
2348
2349 dst->ip6po_hlim = src->ip6po_hlim;
2350 dst->ip6po_tclass = src->ip6po_tclass;
2351 dst->ip6po_flags = src->ip6po_flags;
2352 if (src->ip6po_pktinfo) {
2353 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2354 M_IP6OPT, canwait);
2355 if (dst->ip6po_pktinfo == NULL)
2356 goto bad;
2357 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2358 }
2359 if (src->ip6po_nexthop) {
2360 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2361 M_IP6OPT, canwait);
2362 if (dst->ip6po_nexthop == NULL)
2363 goto bad;
2364 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2365 src->ip6po_nexthop->sa_len);
2366 }
2367 PKTOPT_EXTHDRCPY(ip6po_hbh);
2368 PKTOPT_EXTHDRCPY(ip6po_dest1);
2369 PKTOPT_EXTHDRCPY(ip6po_dest2);
2370 PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2371 return (0);
2372
2373 bad:
2374 ip6_clearpktopts(dst, -1);
2375 return (ENOBUFS);
2376}
2377#undef PKTOPT_EXTHDRCPY
2378
2379struct ip6_pktopts *
2380ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2381{
2382 int error;
2383 struct ip6_pktopts *dst;
2384
2385 dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2386 if (dst == NULL)
2387 return (NULL);
2388 ip6_initpktopts(dst);
2389
2390 if ((error = copypktopts(dst, src, canwait)) != 0) {
2391 free(dst, M_IP6OPT);
2392 return (NULL);
2393 }
2394
2395 return (dst);
2396}
2397
2398void
2399ip6_freepcbopts(struct ip6_pktopts *pktopt)
2400{
2401 if (pktopt == NULL)
2402 return;
2403
2404 ip6_clearpktopts(pktopt, -1);
2405
2406 free(pktopt, M_IP6OPT);
2407}
2408
2409/*
2410 * Set the IP6 multicast options in response to user setsockopt().
2411 */
2412static int
2413ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2414{
2415 int error = 0;
2416 u_int loop, ifindex;
2417 struct ipv6_mreq *mreq;
2418 struct ifnet *ifp;
2419 struct ip6_moptions *im6o = *im6op;
2420 struct route_in6 ro;
2421 struct in6_multi_mship *imm;
2422
2423 if (im6o == NULL) {
2424 /*
2425 * No multicast option buffer attached to the pcb;
2426 * allocate one and initialize to default values.
2427 */
2428 im6o = (struct ip6_moptions *)
2429 malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2430
2431 if (im6o == NULL)
2432 return (ENOBUFS);
2433 *im6op = im6o;
2434 im6o->im6o_multicast_ifp = NULL;
2435 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2436 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2437 LIST_INIT(&im6o->im6o_memberships);
2438 }
2439
2440 switch (optname) {
2441
2442 case IPV6_MULTICAST_IF:
2443 /*
2444 * Select the interface for outgoing multicast packets.
2445 */
2446 if (m == NULL || m->m_len != sizeof(u_int)) {
2447 error = EINVAL;
2448 break;
2449 }
2450 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2451 if (ifindex < 0 || if_index < ifindex) {
2452 error = ENXIO; /* XXX EINVAL? */
2453 break;
2454 }
2455 ifp = ifnet_byindex(ifindex);
2456 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2457 error = EADDRNOTAVAIL;
2458 break;
2459 }
2460 im6o->im6o_multicast_ifp = ifp;
2461 break;
2462
2463 case IPV6_MULTICAST_HOPS:
2464 {
2465 /*
2466 * Set the IP6 hoplimit for outgoing multicast packets.
2467 */
2468 int optval;
2469 if (m == NULL || m->m_len != sizeof(int)) {
2470 error = EINVAL;
2471 break;
2472 }
2473 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2474 if (optval < -1 || optval >= 256)
2475 error = EINVAL;
2476 else if (optval == -1)
2477 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2478 else
2479 im6o->im6o_multicast_hlim = optval;
2480 break;
2481 }
2482
2483 case IPV6_MULTICAST_LOOP:
2484 /*
2485 * Set the loopback flag for outgoing multicast packets.
2486 * Must be zero or one.
2487 */
2488 if (m == NULL || m->m_len != sizeof(u_int)) {
2489 error = EINVAL;
2490 break;
2491 }
2492 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2493 if (loop > 1) {
2494 error = EINVAL;
2495 break;
2496 }
2497 im6o->im6o_multicast_loop = loop;
2498 break;
2499
2500 case IPV6_JOIN_GROUP:
2501 /*
2502 * Add a multicast group membership.
2503 * Group must be a valid IP6 multicast address.
2504 */
2505 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2506 error = EINVAL;
2507 break;
2508 }
2509 mreq = mtod(m, struct ipv6_mreq *);
2510
2511 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2512 /*
2513 * We use the unspecified address to specify to accept
2514 * all multicast addresses. Only super user is allowed
2515 * to do this.
2516 */
2517 /* XXX-BZ might need a better PRIV_NETINET_x for this */
2518 error = priv_check(curthread, PRIV_NETINET_MROUTE);
2519 if (error)
2520 break;
2521 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2522 error = EINVAL;
2523 break;
2524 }
2525
2526 /*
2527 * If no interface was explicitly specified, choose an
2528 * appropriate one according to the given multicast address.
2529 */
2530 if (mreq->ipv6mr_interface == 0) {
2531 struct sockaddr_in6 *dst;
2532
2533 /*
2534 * Look up the routing table for the
2535 * address, and choose the outgoing interface.
2536 * XXX: is it a good approach?
2537 */
2538 ro.ro_rt = NULL;
2539 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2540 bzero(dst, sizeof(*dst));
2541 dst->sin6_family = AF_INET6;
2542 dst->sin6_len = sizeof(*dst);
2543 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2544 rtalloc((struct route *)&ro);
2545 if (ro.ro_rt == NULL) {
2546 error = EADDRNOTAVAIL;
2547 break;
2548 }
2549 ifp = ro.ro_rt->rt_ifp;
2550 RTFREE(ro.ro_rt);
2551 } else {
2552 /*
2553 * If the interface is specified, validate it.
2554 */
2555 if (mreq->ipv6mr_interface < 0 ||
2556 if_index < mreq->ipv6mr_interface) {
2557 error = ENXIO; /* XXX EINVAL? */
2558 break;
2559 }
2560 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2561 if (!ifp) {
2562 error = ENXIO; /* XXX EINVAL? */
2563 break;
2564 }
2565 }
2566
2567 /*
2568 * See if we found an interface, and confirm that it
2569 * supports multicast
2570 */
2571 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2572 error = EADDRNOTAVAIL;
2573 break;
2574 }
2575
2576 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2577 error = EADDRNOTAVAIL; /* XXX: should not happen */
2578 break;
2579 }
2580
2581 /*
2582 * See if the membership already exists.
2583 */
2584 for (imm = im6o->im6o_memberships.lh_first;
2585 imm != NULL; imm = imm->i6mm_chain.le_next)
2586 if (imm->i6mm_maddr->in6m_ifp == ifp &&
2587 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2588 &mreq->ipv6mr_multiaddr))
2589 break;
2590 if (imm != NULL) {
2591 error = EADDRINUSE;
2592 break;
2593 }
2594 /*
2595 * Everything looks good; add a new record to the multicast
2596 * address list for the given interface.
2597 */
2598 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0);
2599 if (imm == NULL)
2600 break;
2601 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2602 break;
2603
2604 case IPV6_LEAVE_GROUP:
2605 /*
2606 * Drop a multicast group membership.
2607 * Group must be a valid IP6 multicast address.
2608 */
2609 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2610 error = EINVAL;
2611 break;
2612 }
2613 mreq = mtod(m, struct ipv6_mreq *);
2614
2615 /*
2616 * If an interface address was specified, get a pointer
2617 * to its ifnet structure.
2618 */
2619 if (mreq->ipv6mr_interface < 0 ||
2620 if_index < mreq->ipv6mr_interface) {
2621 error = ENXIO; /* XXX EINVAL? */
2622 break;
2623 }
2624 if (mreq->ipv6mr_interface == 0)
2625 ifp = NULL;
2626 else
2627 ifp = ifnet_byindex(mreq->ipv6mr_interface);
2628
2629 /* Fill in the scope zone ID */
2630 if (ifp) {
2631 if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2632 /* XXX: should not happen */
2633 error = EADDRNOTAVAIL;
2634 break;
2635 }
2636 } else if (mreq->ipv6mr_interface != 0) {
2637 /*
2638 * This case happens when the (positive) index is in
2639 * the valid range, but the corresponding interface has
2640 * been detached dynamically (XXX).
2641 */
2642 error = EADDRNOTAVAIL;
2643 break;
2644 } else { /* ipv6mr_interface == 0 */
2645 struct sockaddr_in6 sa6_mc;
2646
2647 /*
2648 * The API spec says as follows:
2649 * If the interface index is specified as 0, the
2650 * system may choose a multicast group membership to
2651 * drop by matching the multicast address only.
2652 * On the other hand, we cannot disambiguate the scope
2653 * zone unless an interface is provided. Thus, we
2654 * check if there's ambiguity with the default scope
2655 * zone as the last resort.
2656 */
2657 bzero(&sa6_mc, sizeof(sa6_mc));
2658 sa6_mc.sin6_family = AF_INET6;
2659 sa6_mc.sin6_len = sizeof(sa6_mc);
2660 sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2661 error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2662 if (error != 0)
2663 break;
2664 mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2665 }
2666
2667 /*
2668 * Find the membership in the membership list.
2669 */
2670 for (imm = im6o->im6o_memberships.lh_first;
2671 imm != NULL; imm = imm->i6mm_chain.le_next) {
2672 if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2673 IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2674 &mreq->ipv6mr_multiaddr))
2675 break;
2676 }
2677 if (imm == NULL) {
2678 /* Unable to resolve interface */
2679 error = EADDRNOTAVAIL;
2680 break;
2681 }
2682 /*
2683 * Give up the multicast address record to which the
2684 * membership points.
2685 */
2686 LIST_REMOVE(imm, i6mm_chain);
2687 in6_delmulti(imm->i6mm_maddr);
2688 free(imm, M_IP6MADDR);
2689 break;
2690
2691 default:
2692 error = EOPNOTSUPP;
2693 break;
2694 }
2695
2696 /*
2697 * If all options have default values, no need to keep the mbuf.
2698 */
2699 if (im6o->im6o_multicast_ifp == NULL &&
2700 im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2701 im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2702 im6o->im6o_memberships.lh_first == NULL) {
2703 free(*im6op, M_IP6MOPTS);
2704 *im6op = NULL;
2705 }
2706
2707 return (error);
2708}
2709
2710/*
2711 * Return the IP6 multicast options in response to user getsockopt().
2712 */
2713static int
2714ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2715{
2716 u_int *hlim, *loop, *ifindex;
2717
2718 *mp = m_get(M_TRYWAIT, MT_HEADER); /* XXX */
2719
2720 switch (optname) {
2721
2722 case IPV6_MULTICAST_IF:
2723 ifindex = mtod(*mp, u_int *);
2724 (*mp)->m_len = sizeof(u_int);
2725 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2726 *ifindex = 0;
2727 else
2728 *ifindex = im6o->im6o_multicast_ifp->if_index;
2729 return (0);
2730
2731 case IPV6_MULTICAST_HOPS:
2732 hlim = mtod(*mp, u_int *);
2733 (*mp)->m_len = sizeof(u_int);
2734 if (im6o == NULL)
2735 *hlim = ip6_defmcasthlim;
2736 else
2737 *hlim = im6o->im6o_multicast_hlim;
2738 return (0);
2739
2740 case IPV6_MULTICAST_LOOP:
2741 loop = mtod(*mp, u_int *);
2742 (*mp)->m_len = sizeof(u_int);
2743 if (im6o == NULL)
2744 *loop = ip6_defmcasthlim;
2745 else
2746 *loop = im6o->im6o_multicast_loop;
2747 return (0);
2748
2749 default:
2750 return (EOPNOTSUPP);
2751 }
2752}
2753
2754/*
2755 * Discard the IP6 multicast options.
2756 */
2757void
2758ip6_freemoptions(struct ip6_moptions *im6o)
2759{
2760 struct in6_multi_mship *imm;
2761
2762 if (im6o == NULL)
2763 return;
2764
2765 while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2766 LIST_REMOVE(imm, i6mm_chain);
2767 if (imm->i6mm_maddr)
2768 in6_delmulti(imm->i6mm_maddr);
2769 free(imm, M_IP6MADDR);
2770 }
2771 free(im6o, M_IP6MOPTS);
2772}
2773
2774/*
2775 * Set IPv6 outgoing packet options based on advanced API.
2776 */
2777int
2778ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2779 struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2780{
2781 struct cmsghdr *cm = 0;
2782
2783 if (control == NULL || opt == NULL)
2784 return (EINVAL);
2785
2786 ip6_initpktopts(opt);
2787 if (stickyopt) {
2788 int error;
2789
2790 /*
2791 * If stickyopt is provided, make a local copy of the options
2792 * for this particular packet, then override them by ancillary
2793 * objects.
2794 * XXX: copypktopts() does not copy the cached route to a next
2795 * hop (if any). This is not very good in terms of efficiency,
2796 * but we can allow this since this option should be rarely
2797 * used.
2798 */
2799 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2800 return (error);
2801 }
2802
2803 /*
2804 * XXX: Currently, we assume all the optional information is stored
2805 * in a single mbuf.
2806 */
2807 if (control->m_next)
2808 return (EINVAL);
2809
2810 for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2811 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2812 int error;
2813
2814 if (control->m_len < CMSG_LEN(0))
2815 return (EINVAL);
2816
2817 cm = mtod(control, struct cmsghdr *);
2818 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2819 return (EINVAL);
2820 if (cm->cmsg_level != IPPROTO_IPV6)
2821 continue;
2822
2823 error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2824 cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2825 if (error)
2826 return (error);
2827 }
2828
2829 return (0);
2830}
2831
2832/*
2833 * Set a particular packet option, as a sticky option or an ancillary data
2834 * item. "len" can be 0 only when it's a sticky option.
2835 * We have 4 cases of combination of "sticky" and "cmsg":
2836 * "sticky=0, cmsg=0": impossible
2837 * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2838 * "sticky=1, cmsg=0": RFC3542 socket option
2839 * "sticky=1, cmsg=1": RFC2292 socket option
2840 */
2841static int
2842ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2843 struct ucred *cred, int sticky, int cmsg, int uproto)
2844{
2845 int minmtupolicy, preftemp;
2846 int error;
2847
2848 if (!sticky && !cmsg) {
2849#ifdef DIAGNOSTIC
2850 printf("ip6_setpktopt: impossible case\n");
2851#endif
2852 return (EINVAL);
2853 }
2854
2855 /*
2856 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2857 * not be specified in the context of RFC3542. Conversely,
2858 * RFC3542 types should not be specified in the context of RFC2292.
2859 */
2860 if (!cmsg) {
2861 switch (optname) {
2862 case IPV6_2292PKTINFO:
2863 case IPV6_2292HOPLIMIT:
2864 case IPV6_2292NEXTHOP:
2865 case IPV6_2292HOPOPTS:
2866 case IPV6_2292DSTOPTS:
2867 case IPV6_2292RTHDR:
2868 case IPV6_2292PKTOPTIONS:
2869 return (ENOPROTOOPT);
2870 }
2871 }
2872 if (sticky && cmsg) {
2873 switch (optname) {
2874 case IPV6_PKTINFO:
2875 case IPV6_HOPLIMIT:
2876 case IPV6_NEXTHOP:
2877 case IPV6_HOPOPTS:
2878 case IPV6_DSTOPTS:
2879 case IPV6_RTHDRDSTOPTS:
2880 case IPV6_RTHDR:
2881 case IPV6_USE_MIN_MTU:
2882 case IPV6_DONTFRAG:
2883 case IPV6_TCLASS:
2884 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2885 return (ENOPROTOOPT);
2886 }
2887 }
2888
2889 switch (optname) {
2890 case IPV6_2292PKTINFO:
2891 case IPV6_PKTINFO:
2892 {
2893 struct ifnet *ifp = NULL;
2894 struct in6_pktinfo *pktinfo;
2895
2896 if (len != sizeof(struct in6_pktinfo))
2897 return (EINVAL);
2898
2899 pktinfo = (struct in6_pktinfo *)buf;
2900
2901 /*
2902 * An application can clear any sticky IPV6_PKTINFO option by
2903 * doing a "regular" setsockopt with ipi6_addr being
2904 * in6addr_any and ipi6_ifindex being zero.
2905 * [RFC 3542, Section 6]
2906 */
2907 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2908 pktinfo->ipi6_ifindex == 0 &&
2909 IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2910 ip6_clearpktopts(opt, optname);
2911 break;
2912 }
2913
2914 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2915 sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2916 return (EINVAL);
2917 }
2918
2919 /* validate the interface index if specified. */
2920 if (pktinfo->ipi6_ifindex > if_index ||
2921 pktinfo->ipi6_ifindex < 0) {
2922 return (ENXIO);
2923 }
2924 if (pktinfo->ipi6_ifindex) {
2925 ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2926 if (ifp == NULL)
2927 return (ENXIO);
2928 }
2929
2930 /*
2931 * We store the address anyway, and let in6_selectsrc()
2932 * validate the specified address. This is because ipi6_addr
2933 * may not have enough information about its scope zone, and
2934 * we may need additional information (such as outgoing
2935 * interface or the scope zone of a destination address) to
2936 * disambiguate the scope.
2937 * XXX: the delay of the validation may confuse the
2938 * application when it is used as a sticky option.
2939 */
2940 if (opt->ip6po_pktinfo == NULL) {
2941 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2942 M_IP6OPT, M_NOWAIT);
2943 if (opt->ip6po_pktinfo == NULL)
2944 return (ENOBUFS);
2945 }
2946 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2947 break;
2948 }
2949
2950 case IPV6_2292HOPLIMIT:
2951 case IPV6_HOPLIMIT:
2952 {
2953 int *hlimp;
2954
2955 /*
2956 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2957 * to simplify the ordering among hoplimit options.
2958 */
2959 if (optname == IPV6_HOPLIMIT && sticky)
2960 return (ENOPROTOOPT);
2961
2962 if (len != sizeof(int))
2963 return (EINVAL);
2964 hlimp = (int *)buf;
2965 if (*hlimp < -1 || *hlimp > 255)
2966 return (EINVAL);
2967
2968 opt->ip6po_hlim = *hlimp;
2969 break;
2970 }
2971
2972 case IPV6_TCLASS:
2973 {
2974 int tclass;
2975
2976 if (len != sizeof(int))
2977 return (EINVAL);
2978 tclass = *(int *)buf;
2979 if (tclass < -1 || tclass > 255)
2980 return (EINVAL);
2981
2982 opt->ip6po_tclass = tclass;
2983 break;
2984 }
2985
2986 case IPV6_2292NEXTHOP:
2987 case IPV6_NEXTHOP:
2988 if (cred != NULL) {
2989 error = priv_check_cred(cred,
2990 PRIV_NETINET_SETHDROPTS, 0);
2991 if (error)
2992 return (error);
2993 }
2994
2995 if (len == 0) { /* just remove the option */
2996 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2997 break;
2998 }
2999
3000 /* check if cmsg_len is large enough for sa_len */
3001 if (len < sizeof(struct sockaddr) || len < *buf)
3002 return (EINVAL);
3003
3004 switch (((struct sockaddr *)buf)->sa_family) {
3005 case AF_INET6:
3006 {
3007 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3008 int error;
3009
3010 if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3011 return (EINVAL);
3012
3013 if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3014 IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3015 return (EINVAL);
3016 }
3017 if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3018 != 0) {
3019 return (error);
3020 }
3021 break;
3022 }
3023 case AF_LINK: /* should eventually be supported */
3024 default:
3025 return (EAFNOSUPPORT);
3026 }
3027
3028 /* turn off the previous option, then set the new option. */
3029 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3030 opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3031 if (opt->ip6po_nexthop == NULL)
3032 return (ENOBUFS);
3033 bcopy(buf, opt->ip6po_nexthop, *buf);
3034 break;
3035
3036 case IPV6_2292HOPOPTS:
3037 case IPV6_HOPOPTS:
3038 {
3039 struct ip6_hbh *hbh;
3040 int hbhlen;
3041
3042 /*
3043 * XXX: We don't allow a non-privileged user to set ANY HbH
3044 * options, since per-option restriction has too much
3045 * overhead.
3046 */
3047 if (cred != NULL) {
3048 error = priv_check_cred(cred,
3049 PRIV_NETINET_SETHDROPTS, 0);
3050 if (error)
3051 return (error);
3052 }
3053
3054 if (len == 0) {
3055 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3056 break; /* just remove the option */
3057 }
3058
3059 /* message length validation */
3060 if (len < sizeof(struct ip6_hbh))
3061 return (EINVAL);
3062 hbh = (struct ip6_hbh *)buf;
3063 hbhlen = (hbh->ip6h_len + 1) << 3;
3064 if (len != hbhlen)
3065 return (EINVAL);
3066
3067 /* turn off the previous option, then set the new option. */
3068 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3069 opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3070 if (opt->ip6po_hbh == NULL)
3071 return (ENOBUFS);
3072 bcopy(hbh, opt->ip6po_hbh, hbhlen);
3073
3074 break;
3075 }
3076
3077 case IPV6_2292DSTOPTS:
3078 case IPV6_DSTOPTS:
3079 case IPV6_RTHDRDSTOPTS:
3080 {
3081 struct ip6_dest *dest, **newdest = NULL;
3082 int destlen;
3083
3084 if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3085 error = priv_check_cred(cred,
3086 PRIV_NETINET_SETHDROPTS, 0);
3087 if (error)
3088 return (error);
3089 }
3090
3091 if (len == 0) {
3092 ip6_clearpktopts(opt, optname);
3093 break; /* just remove the option */
3094 }
3095
3096 /* message length validation */
3097 if (len < sizeof(struct ip6_dest))
3098 return (EINVAL);
3099 dest = (struct ip6_dest *)buf;
3100 destlen = (dest->ip6d_len + 1) << 3;
3101 if (len != destlen)
3102 return (EINVAL);
3103
3104 /*
3105 * Determine the position that the destination options header
3106 * should be inserted; before or after the routing header.
3107 */
3108 switch (optname) {
3109 case IPV6_2292DSTOPTS:
3110 /*
3111 * The old advacned API is ambiguous on this point.
3112 * Our approach is to determine the position based
3113 * according to the existence of a routing header.
3114 * Note, however, that this depends on the order of the
3115 * extension headers in the ancillary data; the 1st
3116 * part of the destination options header must appear
3117 * before the routing header in the ancillary data,
3118 * too.
3119 * RFC3542 solved the ambiguity by introducing
3120 * separate ancillary data or option types.
3121 */
3122 if (opt->ip6po_rthdr == NULL)
3123 newdest = &opt->ip6po_dest1;
3124 else
3125 newdest = &opt->ip6po_dest2;
3126 break;
3127 case IPV6_RTHDRDSTOPTS:
3128 newdest = &opt->ip6po_dest1;
3129 break;
3130 case IPV6_DSTOPTS:
3131 newdest = &opt->ip6po_dest2;
3132 break;
3133 }
3134
3135 /* turn off the previous option, then set the new option. */
3136 ip6_clearpktopts(opt, optname);
3137 *newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3138 if (*newdest == NULL)
3139 return (ENOBUFS);
3140 bcopy(dest, *newdest, destlen);
3141
3142 break;
3143 }
3144
3145 case IPV6_2292RTHDR:
3146 case IPV6_RTHDR:
3147 {
3148 struct ip6_rthdr *rth;
3149 int rthlen;
3150
3151 if (len == 0) {
3152 ip6_clearpktopts(opt, IPV6_RTHDR);
3153 break; /* just remove the option */
3154 }
3155
3156 /* message length validation */
3157 if (len < sizeof(struct ip6_rthdr))
3158 return (EINVAL);
3159 rth = (struct ip6_rthdr *)buf;
3160 rthlen = (rth->ip6r_len + 1) << 3;
3161 if (len != rthlen)
3162 return (EINVAL);
3163
3164 switch (rth->ip6r_type) {
3165 case IPV6_RTHDR_TYPE_0:
3166 if (rth->ip6r_len == 0) /* must contain one addr */
3167 return (EINVAL);
3168 if (rth->ip6r_len % 2) /* length must be even */
3169 return (EINVAL);
3170 if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3171 return (EINVAL);
3172 break;
3173 default:
3174 return (EINVAL); /* not supported */
3175 }
3176
3177 /* turn off the previous option */
3178 ip6_clearpktopts(opt, IPV6_RTHDR);
3179 opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3180 if (opt->ip6po_rthdr == NULL)
3181 return (ENOBUFS);
3182 bcopy(rth, opt->ip6po_rthdr, rthlen);
3183
3184 break;
3185 }
3186
3187 case IPV6_USE_MIN_MTU:
3188 if (len != sizeof(int))
3189 return (EINVAL);
3190 minmtupolicy = *(int *)buf;
3191 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3192 minmtupolicy != IP6PO_MINMTU_DISABLE &&
3193 minmtupolicy != IP6PO_MINMTU_ALL) {
3194 return (EINVAL);
3195 }
3196 opt->ip6po_minmtu = minmtupolicy;
3197 break;
3198
3199 case IPV6_DONTFRAG:
3200 if (len != sizeof(int))
3201 return (EINVAL);
3202
3203 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3204 /*
3205 * we ignore this option for TCP sockets.
3206 * (RFC3542 leaves this case unspecified.)
3207 */
3208 opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3209 } else
3210 opt->ip6po_flags |= IP6PO_DONTFRAG;
3211 break;
3212
3213 case IPV6_PREFER_TEMPADDR:
3214 if (len != sizeof(int))
3215 return (EINVAL);
3216 preftemp = *(int *)buf;
3217 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3218 preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3219 preftemp != IP6PO_TEMPADDR_PREFER) {
3220 return (EINVAL);
3221 }
3222 opt->ip6po_prefer_tempaddr = preftemp;
3223 break;
3224
3225 default:
3226 return (ENOPROTOOPT);
3227 } /* end of switch */
3228
3229 return (0);
3230}
3231
3232/*
3233 * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3234 * packet to the input queue of a specified interface. Note that this
3235 * calls the output routine of the loopback "driver", but with an interface
3236 * pointer that might NOT be &loif -- easier than replicating that code here.
3237 */
3238void
3239ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3240{
3241 struct mbuf *copym;
3242 struct ip6_hdr *ip6;
3243
3244 copym = m_copy(m, 0, M_COPYALL);
3245 if (copym == NULL)
3246 return;
3247
3248 /*
3249 * Make sure to deep-copy IPv6 header portion in case the data
3250 * is in an mbuf cluster, so that we can safely override the IPv6
3251 * header portion later.
3252 */
3253 if ((copym->m_flags & M_EXT) != 0 ||
3254 copym->m_len < sizeof(struct ip6_hdr)) {
3255 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3256 if (copym == NULL)
3257 return;
3258 }
3259
3260#ifdef DIAGNOSTIC
3261 if (copym->m_len < sizeof(*ip6)) {
3262 m_freem(copym);
3263 return;
3264 }
3265#endif
3266
3267 ip6 = mtod(copym, struct ip6_hdr *);
3268 /*
3269 * clear embedded scope identifiers if necessary.
3270 * in6_clearscope will touch the addresses only when necessary.
3271 */
3272 in6_clearscope(&ip6->ip6_src);
3273 in6_clearscope(&ip6->ip6_dst);
3274
3275 (void)if_simloop(ifp, copym, dst->sin6_family, 0);
3276}
3277
3278/*
3279 * Chop IPv6 header off from the payload.
3280 */
3281static int
3282ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3283{
3284 struct mbuf *mh;
3285 struct ip6_hdr *ip6;
3286
3287 ip6 = mtod(m, struct ip6_hdr *);
3288 if (m->m_len > sizeof(*ip6)) {
3289 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3290 if (mh == 0) {
3291 m_freem(m);
3292 return ENOBUFS;
3293 }
3294 M_MOVE_PKTHDR(mh, m);
3295 MH_ALIGN(mh, sizeof(*ip6));
3296 m->m_len -= sizeof(*ip6);
3297 m->m_data += sizeof(*ip6);
3298 mh->m_next = m;
3299 m = mh;
3300 m->m_len = sizeof(*ip6);
3301 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3302 }
3303 exthdrs->ip6e_ip6 = m;
3304 return 0;
3305}
3306
3307/*
3308 * Compute IPv6 extension header length.
3309 */
3310int
3311ip6_optlen(struct in6pcb *in6p)
3312{
3313 int len;
3314
3315 if (!in6p->in6p_outputopts)
3316 return 0;
3317
3318 len = 0;
3319#define elen(x) \
3320 (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3321
3322 len += elen(in6p->in6p_outputopts->ip6po_hbh);
3323 if (in6p->in6p_outputopts->ip6po_rthdr)
3324 /* dest1 is valid with rthdr only */
3325 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3326 len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3327 len += elen(in6p->in6p_outputopts->ip6po_dest2);
3328 return len;
3329#undef elen
3330}
|