2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
29 * $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_ipstealth.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
44 #include <sys/domain.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/errno.h>
49 #include <sys/kernel.h>
50 #include <sys/syslog.h>
53 #include <net/route.h>
56 #include <netinet/in.h>
57 #include <netinet/in_var.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/ip_var.h>
61 #include <netinet6/in6_var.h>
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #include <netinet6/scope6_var.h>
65 #include <netinet/icmp6.h>
66 #include <netinet6/nd6.h>
68 #include <netinet/in_pcb.h>
71 #include <netipsec/ipsec.h>
72 #include <netipsec/ipsec6.h>
73 #include <netipsec/key.h>
76 #include <netinet6/ip6protosw.h>
78 struct route_in6 ip6_forward_rt;
81 * Forward a packet. If some error occurs return the sender
82 * an icmp packet. Note we can't always generate a meaningful
83 * icmp message because icmp doesn't have a large enough repertoire
86 * If not forwarding, just drop the packet. This could be confusing
87 * if ipforwarding was zero but some routing protocol was advancing
88 * us as a gateway to somewhere. However, we must let the routing
89 * protocol deal with that.
93 ip6_forward(struct mbuf *m, int srcrt)
95 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
96 struct sockaddr_in6 *dst = NULL;
97 struct rtentry *rt = NULL;
98 int error, type = 0, code = 0;
99 struct mbuf *mcopy = NULL;
100 struct ifnet *origifp; /* maybe unnecessary */
101 u_int32_t inzone, outzone;
102 struct in6_addr src_in6, dst_in6;
104 struct secpolicy *sp = NULL;
107 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
109 GIANT_REQUIRED; /* XXX bz: ip6_forward_rt */
113 * Check AH/ESP integrity.
116 * Don't increment ip6s_cantforward because this is the check
117 * before forwarding packet actually.
119 if (ipsec6_in_reject(m, NULL)) {
120 ipsec6stat.in_polvio++;
127 * Do not forward packets to multicast destination (should be handled
129 * Do not forward packets with unspecified source. It was discussed
130 * in July 2000, on the ipngwg mailing list.
132 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
133 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
134 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
135 ip6stat.ip6s_cantforward++;
136 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
137 if (ip6_log_time + ip6_log_interval < time_second) {
138 ip6_log_time = time_second;
141 "from %s to %s nxt %d received on %s\n",
142 ip6_sprintf(ip6bufs, &ip6->ip6_src),
143 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
145 if_name(m->m_pkthdr.rcvif));
154 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
155 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
156 icmp6_error(m, ICMP6_TIME_EXCEEDED,
157 ICMP6_TIME_EXCEED_TRANSIT, 0);
160 ip6->ip6_hlim -= IPV6_HLIMDEC;
167 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
168 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
169 * we need to generate an ICMP6 message to the src.
170 * Thanks to M_EXT, in most cases copy will not occur.
172 * It is important to save it before IPsec processing as IPsec
173 * processing may modify the mbuf.
175 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
178 /* get a security policy for this packet */
179 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
180 IP_FORWARDING, &error);
182 ipsec6stat.out_inval++;
183 ip6stat.ip6s_cantforward++;
186 /* XXX: what icmp ? */
198 switch (sp->policy) {
199 case IPSEC_POLICY_DISCARD:
201 * This packet is just discarded.
203 ipsec6stat.out_polvio++;
204 ip6stat.ip6s_cantforward++;
208 /* XXX: what icmp ? */
216 case IPSEC_POLICY_BYPASS:
217 case IPSEC_POLICY_NONE:
218 /* no need to do IPsec. */
222 case IPSEC_POLICY_IPSEC:
223 if (sp->req == NULL) {
224 /* XXX should be panic ? */
225 printf("ip6_forward: No IPsec request specified.\n");
226 ip6stat.ip6s_cantforward++;
230 /* XXX: what icmp ? */
241 case IPSEC_POLICY_ENTRUST:
243 /* should be panic ?? */
244 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
250 struct ipsecrequest *isr = NULL;
251 struct ipsec_output_state state;
254 * when the kernel forwards a packet, it is not proper to apply
255 * IPsec transport mode to the packet is not proper. this check
257 * at present, if there is even a transport mode SA request in the
258 * security policy, the kernel does not apply IPsec to the packet.
259 * this check is not enough because the following case is valid.
260 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
262 for (isr = sp->req; isr; isr = isr->next) {
263 if (isr->saidx.mode == IPSEC_MODE_ANY)
265 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
270 * if there's no need for tunnel mode IPsec, skip.
277 * All the extension headers will become inaccessible
278 * (since they can be encrypted).
279 * Don't panic, we need no more updates to extension headers
280 * on inner IPv6 packet (since they are now encapsulated).
282 * IPv6 [ESP|AH] IPv6 [extension headers] payload
284 bzero(&state, sizeof(state));
286 state.ro = NULL; /* update at ipsec6_output_tunnel() */
287 state.dst = NULL; /* update at ipsec6_output_tunnel() */
289 error = ipsec6_output_tunnel(&state, sp, 0);
295 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
304 printf("ip6_output (ipsec): error code %d\n", error);
307 /* don't show these error codes to the user */
310 ip6stat.ip6s_cantforward++;
313 /* XXX: what icmp ? */
322 * In the FAST IPSec case we have already
323 * re-injected the packet and it has been freed
324 * by the ipsec_done() function. So, just clean
325 * up after ourselves.
331 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){
333 * now tunnel mode headers are added. we are originating
334 * packet instead of forwarding the packet.
336 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL,
342 dst = (struct sockaddr_in6 *)state.dst;
343 rt = state.ro ? state.ro->ro_rt : NULL;
344 if (dst != NULL && rt != NULL)
355 dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
357 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */
358 if (ip6_forward_rt.ro_rt == 0 ||
359 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
360 if (ip6_forward_rt.ro_rt) {
361 RTFREE(ip6_forward_rt.ro_rt);
362 ip6_forward_rt.ro_rt = 0;
365 /* this probably fails but give it a try again */
366 rtalloc((struct route *)&ip6_forward_rt);
369 if (ip6_forward_rt.ro_rt == 0) {
370 ip6stat.ip6s_noroute++;
371 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
373 icmp6_error(mcopy, ICMP6_DST_UNREACH,
374 ICMP6_DST_UNREACH_NOROUTE, 0);
379 } else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
380 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
381 if (ip6_forward_rt.ro_rt) {
382 RTFREE(ip6_forward_rt.ro_rt);
383 ip6_forward_rt.ro_rt = 0;
385 bzero(dst, sizeof(*dst));
386 dst->sin6_len = sizeof(struct sockaddr_in6);
387 dst->sin6_family = AF_INET6;
388 dst->sin6_addr = ip6->ip6_dst;
390 rtalloc((struct route *)&ip6_forward_rt);
391 if (ip6_forward_rt.ro_rt == 0) {
392 ip6stat.ip6s_noroute++;
393 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
395 icmp6_error(mcopy, ICMP6_DST_UNREACH,
396 ICMP6_DST_UNREACH_NOROUTE, 0);
402 rt = ip6_forward_rt.ro_rt;
408 * Source scope check: if a packet can't be delivered to its
409 * destination for the reason that the destination is beyond the scope
410 * of the source address, discard the packet and return an icmp6
411 * destination unreachable error with Code 2 (beyond scope of source
412 * address). We use a local copy of ip6_src, since in6_setscope()
413 * will possibly modify its first argument.
414 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
416 src_in6 = ip6->ip6_src;
417 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
418 /* XXX: this should not happen */
419 ip6stat.ip6s_cantforward++;
420 ip6stat.ip6s_badscope++;
424 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
425 ip6stat.ip6s_cantforward++;
426 ip6stat.ip6s_badscope++;
430 if (inzone != outzone
435 ip6stat.ip6s_cantforward++;
436 ip6stat.ip6s_badscope++;
437 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
439 if (ip6_log_time + ip6_log_interval < time_second) {
440 ip6_log_time = time_second;
443 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
444 ip6_sprintf(ip6bufs, &ip6->ip6_src),
445 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
447 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
450 icmp6_error(mcopy, ICMP6_DST_UNREACH,
451 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
457 * Destination scope check: if a packet is going to break the scope
458 * zone of packet's destination address, discard it. This case should
459 * usually be prevented by appropriately-configured routing table, but
460 * we need an explicit check because we may mistakenly forward the
461 * packet to a different zone by (e.g.) a default route.
463 dst_in6 = ip6->ip6_dst;
464 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
465 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
467 ip6stat.ip6s_cantforward++;
468 ip6stat.ip6s_badscope++;
473 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
474 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
478 struct secpolicy *sp;
483 mtu = IN6_LINKMTU(rt->rt_ifp);
486 * When we do IPsec tunnel ingress, we need to play
487 * with the link value (decrement IPsec header size
488 * from mtu value). The code is much simpler than v4
489 * case, as we have the outgoing interface for
490 * encapsulated packet as "rt->rt_ifp".
492 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
493 IP_FORWARDING, &ipsecerror);
495 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
496 IPSEC_DIR_OUTBOUND, NULL);
497 if (ipsechdrsiz < mtu)
502 * if mtu becomes less than minimum MTU,
503 * tell minimum MTU (and I'll need to fragment it).
508 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
514 if (rt->rt_flags & RTF_GATEWAY)
515 dst = (struct sockaddr_in6 *)rt->rt_gateway;
518 * If we are to forward the packet using the same interface
519 * as one we got the packet from, perhaps we should send a redirect
520 * to sender to shortcut a hop.
521 * Only send redirect if source is sending directly to us,
522 * and if packet was not source routed (or has any options).
523 * Also, don't send redirect if forwarding using a route
524 * modified by a redirect.
526 if (ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
530 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
531 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
533 * If the incoming interface is equal to the outgoing
534 * one, and the link attached to the interface is
535 * point-to-point, then it will be highly probable
536 * that a routing loop occurs. Thus, we immediately
537 * drop the packet and send an ICMPv6 error message.
539 * type/code is based on suggestion by Rich Draves.
540 * not sure if it is the best pick.
542 icmp6_error(mcopy, ICMP6_DST_UNREACH,
543 ICMP6_DST_UNREACH_ADDR, 0);
551 * Fake scoped addresses. Note that even link-local source or
552 * destinaion can appear, if the originating node just sends the
553 * packet to us (without address resolution for the destination).
554 * Since both icmp6_error and icmp6_redirect_output fill the embedded
555 * link identifiers, we can do this stuff after making a copy for
556 * returning an error.
558 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
560 * See corresponding comments in ip6_output.
561 * XXX: but is it possible that ip6_forward() sends a packet
562 * to a loopback interface? I don't think so, and thus
563 * I bark here. (jinmei@kame.net)
564 * XXX: it is common to route invalid packets to loopback.
565 * also, the codepath will be visited on use of ::1 in
571 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
574 printf("ip6_forward: outgoing interface is loopback. "
575 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
576 ip6_sprintf(ip6bufs, &ip6->ip6_src),
577 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
578 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
579 if_name(rt->rt_ifp));
582 /* we can just use rcvif in forwarding. */
583 origifp = m->m_pkthdr.rcvif;
586 origifp = rt->rt_ifp;
588 * clear embedded scope identifiers if necessary.
589 * in6_clearscope will touch the addresses only when necessary.
591 in6_clearscope(&ip6->ip6_src);
592 in6_clearscope(&ip6->ip6_dst);
594 /* Jump over all PFIL processing if hooks are not active. */
595 if (!PFIL_HOOKED(&inet6_pfil_hook))
598 /* Run through list of hooks for output packets. */
599 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
604 ip6 = mtod(m, struct ip6_hdr *);
607 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
609 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
610 ip6stat.ip6s_cantforward++;
612 ip6stat.ip6s_forward++;
613 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
615 ip6stat.ip6s_redirectsent++;
627 if (type == ND_REDIRECT) {
628 icmp6_redirect_output(mcopy, rt);
634 /* xxx MTU is constant in PPP? */
638 /* Tell source to slow down like source quench in IP? */
641 case ENETUNREACH: /* shouldn't happen, checked above */
646 type = ICMP6_DST_UNREACH;
647 code = ICMP6_DST_UNREACH_ADDR;
650 icmp6_error(mcopy, type, code, 0);