1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $
30 */
31
32#include <sys/cdefs.h>
|
69
70#include <netinet/in_pcb.h>
71
72#ifdef IPSEC
73#include <netipsec/ipsec.h>
74#include <netipsec/ipsec6.h>
75#include <netipsec/key.h>
76#endif /* IPSEC */
77
78#include <netinet6/ip6protosw.h>
79
80#ifdef VIMAGE_GLOBALS
81struct route_in6 ip6_forward_rt;
82#endif
83
84/*
85 * Forward a packet. If some error occurs return the sender
86 * an icmp packet. Note we can't always generate a meaningful
87 * icmp message because icmp doesn't have a large enough repertoire
88 * of codes and types.
89 *
90 * If not forwarding, just drop the packet. This could be confusing
91 * if ipforwarding was zero but some routing protocol was advancing
92 * us as a gateway to somewhere. However, we must let the routing
93 * protocol deal with that.
94 *
95 */
96void
97ip6_forward(struct mbuf *m, int srcrt)
98{
99 INIT_VNET_INET6(curvnet);
100 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
101 struct sockaddr_in6 *dst = NULL;
102 struct rtentry *rt = NULL;
103 int error, type = 0, code = 0;
104 struct mbuf *mcopy = NULL;
105 struct ifnet *origifp; /* maybe unnecessary */
106 u_int32_t inzone, outzone;
107 struct in6_addr src_in6, dst_in6;
108#ifdef IPSEC
109 INIT_VNET_IPSEC(curvnet);
110 struct secpolicy *sp = NULL;
111 int ipsecrt = 0;
112#endif
113 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
114
115 /* GIANT_REQUIRED; */ /* XXX bz: ip6_forward_rt */
116
117#ifdef IPSEC
118 /*
119 * Check AH/ESP integrity.
120 */
121 /*
122 * Don't increment ip6s_cantforward because this is the check
123 * before forwarding packet actually.
124 */
125 if (ipsec6_in_reject(m, NULL)) {
126 V_ipsec6stat.in_polvio++;
127 m_freem(m);
128 return;
129 }
130#endif /* IPSEC */
131
132 /*
133 * Do not forward packets to multicast destination (should be handled
134 * by ip6_mforward().
135 * Do not forward packets with unspecified source. It was discussed
136 * in July 2000, on the ipngwg mailing list.
137 */
138 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
139 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
140 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
141 V_ip6stat.ip6s_cantforward++;
142 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
143 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
144 V_ip6_log_time = time_second;
145 log(LOG_DEBUG,
146 "cannot forward "
147 "from %s to %s nxt %d received on %s\n",
148 ip6_sprintf(ip6bufs, &ip6->ip6_src),
149 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
150 ip6->ip6_nxt,
151 if_name(m->m_pkthdr.rcvif));
152 }
153 m_freem(m);
154 return;
155 }
156
157#ifdef IPSTEALTH
158 if (!V_ip6stealth) {
159#endif
160 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
161 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
162 icmp6_error(m, ICMP6_TIME_EXCEEDED,
163 ICMP6_TIME_EXCEED_TRANSIT, 0);
164 return;
165 }
166 ip6->ip6_hlim -= IPV6_HLIMDEC;
167
168#ifdef IPSTEALTH
169 }
170#endif
171
172 /*
173 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
174 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
175 * we need to generate an ICMP6 message to the src.
176 * Thanks to M_EXT, in most cases copy will not occur.
177 *
178 * It is important to save it before IPsec processing as IPsec
179 * processing may modify the mbuf.
180 */
181 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
182
183#ifdef IPSEC
184 /* get a security policy for this packet */
185 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
186 IP_FORWARDING, &error);
187 if (sp == NULL) {
188 V_ipsec6stat.out_inval++;
189 V_ip6stat.ip6s_cantforward++;
190 if (mcopy) {
191#if 0
192 /* XXX: what icmp ? */
193#else
194 m_freem(mcopy);
195#endif
196 }
197 m_freem(m);
198 return;
199 }
200
201 error = 0;
202
203 /* check policy */
204 switch (sp->policy) {
205 case IPSEC_POLICY_DISCARD:
206 /*
207 * This packet is just discarded.
208 */
209 V_ipsec6stat.out_polvio++;
210 V_ip6stat.ip6s_cantforward++;
211 KEY_FREESP(&sp);
212 if (mcopy) {
213#if 0
214 /* XXX: what icmp ? */
215#else
216 m_freem(mcopy);
217#endif
218 }
219 m_freem(m);
220 return;
221
222 case IPSEC_POLICY_BYPASS:
223 case IPSEC_POLICY_NONE:
224 /* no need to do IPsec. */
225 KEY_FREESP(&sp);
226 goto skip_ipsec;
227
228 case IPSEC_POLICY_IPSEC:
229 if (sp->req == NULL) {
230 /* XXX should be panic ? */
231 printf("ip6_forward: No IPsec request specified.\n");
232 V_ip6stat.ip6s_cantforward++;
233 KEY_FREESP(&sp);
234 if (mcopy) {
235#if 0
236 /* XXX: what icmp ? */
237#else
238 m_freem(mcopy);
239#endif
240 }
241 m_freem(m);
242 return;
243 }
244 /* do IPsec */
245 break;
246
247 case IPSEC_POLICY_ENTRUST:
248 default:
249 /* should be panic ?? */
250 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
251 KEY_FREESP(&sp);
252 goto skip_ipsec;
253 }
254
255 {
256 struct ipsecrequest *isr = NULL;
257 struct ipsec_output_state state;
258
259 /*
260 * when the kernel forwards a packet, it is not proper to apply
261 * IPsec transport mode to the packet is not proper. this check
262 * avoid from this.
263 * at present, if there is even a transport mode SA request in the
264 * security policy, the kernel does not apply IPsec to the packet.
265 * this check is not enough because the following case is valid.
266 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
267 */
268 for (isr = sp->req; isr; isr = isr->next) {
269 if (isr->saidx.mode == IPSEC_MODE_ANY)
270 goto doipsectunnel;
271 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
272 goto doipsectunnel;
273 }
274
275 /*
276 * if there's no need for tunnel mode IPsec, skip.
277 */
278 if (!isr)
279 goto skip_ipsec;
280
281 doipsectunnel:
282 /*
283 * All the extension headers will become inaccessible
284 * (since they can be encrypted).
285 * Don't panic, we need no more updates to extension headers
286 * on inner IPv6 packet (since they are now encapsulated).
287 *
288 * IPv6 [ESP|AH] IPv6 [extension headers] payload
289 */
290 bzero(&state, sizeof(state));
291 state.m = m;
292 state.ro = NULL; /* update at ipsec6_output_tunnel() */
293 state.dst = NULL; /* update at ipsec6_output_tunnel() */
294
295 error = ipsec6_output_tunnel(&state, sp, 0);
296
297 m = state.m;
298 KEY_FREESP(&sp);
299
300 if (error) {
301 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
302 switch (error) {
303 case EHOSTUNREACH:
304 case ENETUNREACH:
305 case EMSGSIZE:
306 case ENOBUFS:
307 case ENOMEM:
308 break;
309 default:
310 printf("ip6_output (ipsec): error code %d\n", error);
311 /* FALLTHROUGH */
312 case ENOENT:
313 /* don't show these error codes to the user */
314 break;
315 }
316 V_ip6stat.ip6s_cantforward++;
317 if (mcopy) {
318#if 0
319 /* XXX: what icmp ? */
320#else
321 m_freem(mcopy);
322#endif
323 }
324 m_freem(m);
325 return;
326 } else {
327 /*
328 * In the FAST IPSec case we have already
329 * re-injected the packet and it has been freed
330 * by the ipsec_done() function. So, just clean
331 * up after ourselves.
332 */
333 m = NULL;
334 goto freecopy;
335 }
336
337 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){
338 /*
339 * now tunnel mode headers are added. we are originating
340 * packet instead of forwarding the packet.
341 */
342 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL,
343 NULL);
344 goto freecopy;
345 }
346
347 /* adjust pointer */
348 dst = (struct sockaddr_in6 *)state.dst;
349 rt = state.ro ? state.ro->ro_rt : NULL;
350 if (dst != NULL && rt != NULL)
351 ipsecrt = 1;
352 }
353 skip_ipsec:
354#endif /* IPSEC */
355
356#ifdef IPSEC
357 if (ipsecrt)
358 goto skip_routing;
359#endif
360
361 dst = (struct sockaddr_in6 *)&V_ip6_forward_rt.ro_dst;
362 if (!srcrt) {
363 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */
364 if (V_ip6_forward_rt.ro_rt == 0 ||
365 (V_ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
366 if (V_ip6_forward_rt.ro_rt) {
367 RTFREE(V_ip6_forward_rt.ro_rt);
368 V_ip6_forward_rt.ro_rt = 0;
369 }
370
371 /* this probably fails but give it a try again */
372 rtalloc((struct route *)&V_ip6_forward_rt);
373 }
374
375 if (V_ip6_forward_rt.ro_rt == 0) {
376 V_ip6stat.ip6s_noroute++;
377 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
378 if (mcopy) {
379 icmp6_error(mcopy, ICMP6_DST_UNREACH,
380 ICMP6_DST_UNREACH_NOROUTE, 0);
381 }
382 m_freem(m);
383 return;
384 }
385 } else if ((rt = V_ip6_forward_rt.ro_rt) == 0 ||
386 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
387 if (V_ip6_forward_rt.ro_rt) {
388 RTFREE(V_ip6_forward_rt.ro_rt);
389 V_ip6_forward_rt.ro_rt = 0;
390 }
391 bzero(dst, sizeof(*dst));
392 dst->sin6_len = sizeof(struct sockaddr_in6);
393 dst->sin6_family = AF_INET6;
394 dst->sin6_addr = ip6->ip6_dst;
395
396 rtalloc((struct route *)&V_ip6_forward_rt);
397 if (V_ip6_forward_rt.ro_rt == 0) {
398 V_ip6stat.ip6s_noroute++;
399 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
400 if (mcopy) {
401 icmp6_error(mcopy, ICMP6_DST_UNREACH,
402 ICMP6_DST_UNREACH_NOROUTE, 0);
403 }
404 m_freem(m);
405 return;
406 }
407 }
408 rt = V_ip6_forward_rt.ro_rt;
409#ifdef IPSEC
410 skip_routing:;
411#endif
412
413 /*
414 * Source scope check: if a packet can't be delivered to its
415 * destination for the reason that the destination is beyond the scope
416 * of the source address, discard the packet and return an icmp6
417 * destination unreachable error with Code 2 (beyond scope of source
418 * address). We use a local copy of ip6_src, since in6_setscope()
419 * will possibly modify its first argument.
420 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
421 */
422 src_in6 = ip6->ip6_src;
423 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
424 /* XXX: this should not happen */
425 V_ip6stat.ip6s_cantforward++;
426 V_ip6stat.ip6s_badscope++;
427 m_freem(m);
428 return;
429 }
430 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
431 V_ip6stat.ip6s_cantforward++;
432 V_ip6stat.ip6s_badscope++;
433 m_freem(m);
434 return;
435 }
436 if (inzone != outzone
437#ifdef IPSEC
438 && !ipsecrt
439#endif
440 ) {
441 V_ip6stat.ip6s_cantforward++;
442 V_ip6stat.ip6s_badscope++;
443 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
444
445 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
446 V_ip6_log_time = time_second;
447 log(LOG_DEBUG,
448 "cannot forward "
449 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
450 ip6_sprintf(ip6bufs, &ip6->ip6_src),
451 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
452 ip6->ip6_nxt,
453 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
454 }
455 if (mcopy)
456 icmp6_error(mcopy, ICMP6_DST_UNREACH,
457 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
458 m_freem(m);
459 return;
460 }
461
462 /*
463 * Destination scope check: if a packet is going to break the scope
464 * zone of packet's destination address, discard it. This case should
465 * usually be prevented by appropriately-configured routing table, but
466 * we need an explicit check because we may mistakenly forward the
467 * packet to a different zone by (e.g.) a default route.
468 */
469 dst_in6 = ip6->ip6_dst;
470 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
471 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
472 inzone != outzone) {
473 V_ip6stat.ip6s_cantforward++;
474 V_ip6stat.ip6s_badscope++;
475 m_freem(m);
476 return;
477 }
478
479 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
480 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
481 if (mcopy) {
482 u_long mtu;
483#ifdef IPSEC
484 struct secpolicy *sp;
485 int ipsecerror;
486 size_t ipsechdrsiz;
487#endif /* IPSEC */
488
489 mtu = IN6_LINKMTU(rt->rt_ifp);
490#ifdef IPSEC
491 /*
492 * When we do IPsec tunnel ingress, we need to play
493 * with the link value (decrement IPsec header size
494 * from mtu value). The code is much simpler than v4
495 * case, as we have the outgoing interface for
496 * encapsulated packet as "rt->rt_ifp".
497 */
498 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
499 IP_FORWARDING, &ipsecerror);
500 if (sp) {
501 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
502 IPSEC_DIR_OUTBOUND, NULL);
503 if (ipsechdrsiz < mtu)
504 mtu -= ipsechdrsiz;
505 }
506
507 /*
508 * if mtu becomes less than minimum MTU,
509 * tell minimum MTU (and I'll need to fragment it).
510 */
511 if (mtu < IPV6_MMTU)
512 mtu = IPV6_MMTU;
513#endif /* IPSEC */
514 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
515 }
516 m_freem(m);
517 return;
518 }
519
520 if (rt->rt_flags & RTF_GATEWAY)
521 dst = (struct sockaddr_in6 *)rt->rt_gateway;
522
523 /*
524 * If we are to forward the packet using the same interface
525 * as one we got the packet from, perhaps we should send a redirect
526 * to sender to shortcut a hop.
527 * Only send redirect if source is sending directly to us,
528 * and if packet was not source routed (or has any options).
529 * Also, don't send redirect if forwarding using a route
530 * modified by a redirect.
531 */
532 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
533#ifdef IPSEC
534 !ipsecrt &&
535#endif /* IPSEC */
536 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
537 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
538 /*
539 * If the incoming interface is equal to the outgoing
540 * one, and the link attached to the interface is
541 * point-to-point, then it will be highly probable
542 * that a routing loop occurs. Thus, we immediately
543 * drop the packet and send an ICMPv6 error message.
544 *
545 * type/code is based on suggestion by Rich Draves.
546 * not sure if it is the best pick.
547 */
548 icmp6_error(mcopy, ICMP6_DST_UNREACH,
549 ICMP6_DST_UNREACH_ADDR, 0);
550 m_freem(m);
551 return;
552 }
553 type = ND_REDIRECT;
554 }
555
556 /*
557 * Fake scoped addresses. Note that even link-local source or
558 * destinaion can appear, if the originating node just sends the
559 * packet to us (without address resolution for the destination).
560 * Since both icmp6_error and icmp6_redirect_output fill the embedded
561 * link identifiers, we can do this stuff after making a copy for
562 * returning an error.
563 */
564 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
565 /*
566 * See corresponding comments in ip6_output.
567 * XXX: but is it possible that ip6_forward() sends a packet
568 * to a loopback interface? I don't think so, and thus
569 * I bark here. (jinmei@kame.net)
570 * XXX: it is common to route invalid packets to loopback.
571 * also, the codepath will be visited on use of ::1 in
572 * rthdr. (itojun)
573 */
574#if 1
575 if (0)
576#else
577 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
578#endif
579 {
580 printf("ip6_forward: outgoing interface is loopback. "
581 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
582 ip6_sprintf(ip6bufs, &ip6->ip6_src),
583 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
584 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
585 if_name(rt->rt_ifp));
586 }
587
588 /* we can just use rcvif in forwarding. */
589 origifp = m->m_pkthdr.rcvif;
590 }
591 else
592 origifp = rt->rt_ifp;
593 /*
594 * clear embedded scope identifiers if necessary.
595 * in6_clearscope will touch the addresses only when necessary.
596 */
597 in6_clearscope(&ip6->ip6_src);
598 in6_clearscope(&ip6->ip6_dst);
599
600 /* Jump over all PFIL processing if hooks are not active. */
601 if (!PFIL_HOOKED(&inet6_pfil_hook))
602 goto pass;
603
604 /* Run through list of hooks for output packets. */
605 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
606 if (error != 0)
607 goto senderr;
608 if (m == NULL)
609 goto freecopy;
610 ip6 = mtod(m, struct ip6_hdr *);
611
612pass:
613 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
614 if (error) {
615 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
616 V_ip6stat.ip6s_cantforward++;
617 } else {
618 V_ip6stat.ip6s_forward++;
619 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
620 if (type)
621 V_ip6stat.ip6s_redirectsent++;
622 else {
623 if (mcopy)
624 goto freecopy;
625 }
626 }
627
628senderr:
629 if (mcopy == NULL)
630 return;
631 switch (error) {
632 case 0:
633 if (type == ND_REDIRECT) {
634 icmp6_redirect_output(mcopy, rt);
635 return;
636 }
637 goto freecopy;
638
639 case EMSGSIZE:
640 /* xxx MTU is constant in PPP? */
641 goto freecopy;
642
643 case ENOBUFS:
644 /* Tell source to slow down like source quench in IP? */
645 goto freecopy;
646
647 case ENETUNREACH: /* shouldn't happen, checked above */
648 case EHOSTUNREACH:
649 case ENETDOWN:
650 case EHOSTDOWN:
651 default:
652 type = ICMP6_DST_UNREACH;
653 code = ICMP6_DST_UNREACH_ADDR;
654 break;
655 }
656 icmp6_error(mcopy, type, code, 0);
657 return;
658
659 freecopy:
660 m_freem(mcopy);
661 return;
662}
|