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>
79 * Forward a packet. If some error occurs return the sender
80 * an icmp packet. Note we can't always generate a meaningful
81 * icmp message because icmp doesn't have a large enough repertoire
84 * If not forwarding, just drop the packet. This could be confusing
85 * if ipforwarding was zero but some routing protocol was advancing
86 * us as a gateway to somewhere. However, we must let the routing
87 * protocol deal with that.
91 ip6_forward(struct mbuf *m, int srcrt)
93 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
94 struct sockaddr_in6 *dst = NULL;
95 struct rtentry *rt = NULL;
96 struct route_in6 rin6;
97 int error, type = 0, code = 0;
98 struct mbuf *mcopy = NULL;
99 struct ifnet *origifp; /* maybe unnecessary */
100 u_int32_t inzone, outzone;
101 struct in6_addr src_in6, dst_in6;
103 struct secpolicy *sp = NULL;
106 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
110 * Check AH/ESP integrity.
113 * Don't increment ip6s_cantforward because this is the check
114 * before forwarding packet actually.
116 if (ipsec6_in_reject(m, NULL)) {
117 V_ipsec6stat.in_polvio++;
124 * Do not forward packets to multicast destination (should be handled
126 * Do not forward packets with unspecified source. It was discussed
127 * in July 2000, on the ipngwg mailing list.
129 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
130 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
131 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
132 V_ip6stat.ip6s_cantforward++;
133 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
134 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
135 V_ip6_log_time = time_second;
138 "from %s to %s nxt %d received on %s\n",
139 ip6_sprintf(ip6bufs, &ip6->ip6_src),
140 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
142 if_name(m->m_pkthdr.rcvif));
151 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
152 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
153 icmp6_error(m, ICMP6_TIME_EXCEEDED,
154 ICMP6_TIME_EXCEED_TRANSIT, 0);
157 ip6->ip6_hlim -= IPV6_HLIMDEC;
164 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
165 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
166 * we need to generate an ICMP6 message to the src.
167 * Thanks to M_EXT, in most cases copy will not occur.
169 * It is important to save it before IPsec processing as IPsec
170 * processing may modify the mbuf.
172 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
175 /* get a security policy for this packet */
176 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
177 IP_FORWARDING, &error);
179 V_ipsec6stat.out_inval++;
180 V_ip6stat.ip6s_cantforward++;
183 /* XXX: what icmp ? */
195 switch (sp->policy) {
196 case IPSEC_POLICY_DISCARD:
198 * This packet is just discarded.
200 V_ipsec6stat.out_polvio++;
201 V_ip6stat.ip6s_cantforward++;
205 /* XXX: what icmp ? */
213 case IPSEC_POLICY_BYPASS:
214 case IPSEC_POLICY_NONE:
215 /* no need to do IPsec. */
219 case IPSEC_POLICY_IPSEC:
220 if (sp->req == NULL) {
221 /* XXX should be panic ? */
222 printf("ip6_forward: No IPsec request specified.\n");
223 V_ip6stat.ip6s_cantforward++;
227 /* XXX: what icmp ? */
238 case IPSEC_POLICY_ENTRUST:
240 /* should be panic ?? */
241 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
247 struct ipsecrequest *isr = NULL;
248 struct ipsec_output_state state;
251 * when the kernel forwards a packet, it is not proper to apply
252 * IPsec transport mode to the packet is not proper. this check
254 * at present, if there is even a transport mode SA request in the
255 * security policy, the kernel does not apply IPsec to the packet.
256 * this check is not enough because the following case is valid.
257 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
259 for (isr = sp->req; isr; isr = isr->next) {
260 if (isr->saidx.mode == IPSEC_MODE_ANY)
262 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
267 * if there's no need for tunnel mode IPsec, skip.
274 * All the extension headers will become inaccessible
275 * (since they can be encrypted).
276 * Don't panic, we need no more updates to extension headers
277 * on inner IPv6 packet (since they are now encapsulated).
279 * IPv6 [ESP|AH] IPv6 [extension headers] payload
281 bzero(&state, sizeof(state));
283 state.ro = NULL; /* update at ipsec6_output_tunnel() */
284 state.dst = NULL; /* update at ipsec6_output_tunnel() */
286 error = ipsec6_output_tunnel(&state, sp, 0);
292 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
301 printf("ip6_output (ipsec): error code %d\n", error);
304 /* don't show these error codes to the user */
307 V_ip6stat.ip6s_cantforward++;
310 /* XXX: what icmp ? */
319 * In the FAST IPSec case we have already
320 * re-injected the packet and it has been freed
321 * by the ipsec_done() function. So, just clean
322 * up after ourselves.
328 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){
330 * now tunnel mode headers are added. we are originating
331 * packet instead of forwarding the packet.
333 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL,
339 dst = (struct sockaddr_in6 *)state.dst;
340 rt = state.ro ? state.ro->ro_rt : NULL;
341 if (dst != NULL && rt != NULL)
349 bzero(&rin6, sizeof(struct route_in6));
350 dst = (struct sockaddr_in6 *)&rin6.ro_dst;
351 dst->sin6_len = sizeof(struct sockaddr_in6);
352 dst->sin6_family = AF_INET6;
353 dst->sin6_addr = ip6->ip6_dst;
355 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
356 if (rin6.ro_rt != NULL)
357 RT_UNLOCK(rin6.ro_rt);
359 V_ip6stat.ip6s_noroute++;
360 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
362 icmp6_error(mcopy, ICMP6_DST_UNREACH,
363 ICMP6_DST_UNREACH_NOROUTE, 0);
373 * Source scope check: if a packet can't be delivered to its
374 * destination for the reason that the destination is beyond the scope
375 * of the source address, discard the packet and return an icmp6
376 * destination unreachable error with Code 2 (beyond scope of source
377 * address). We use a local copy of ip6_src, since in6_setscope()
378 * will possibly modify its first argument.
379 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
381 src_in6 = ip6->ip6_src;
382 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
383 /* XXX: this should not happen */
384 V_ip6stat.ip6s_cantforward++;
385 V_ip6stat.ip6s_badscope++;
388 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
389 V_ip6stat.ip6s_cantforward++;
390 V_ip6stat.ip6s_badscope++;
393 if (inzone != outzone
398 V_ip6stat.ip6s_cantforward++;
399 V_ip6stat.ip6s_badscope++;
400 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
402 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
403 V_ip6_log_time = time_second;
406 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
407 ip6_sprintf(ip6bufs, &ip6->ip6_src),
408 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
410 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
413 icmp6_error(mcopy, ICMP6_DST_UNREACH,
414 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
419 * Destination scope check: if a packet is going to break the scope
420 * zone of packet's destination address, discard it. This case should
421 * usually be prevented by appropriately-configured routing table, but
422 * we need an explicit check because we may mistakenly forward the
423 * packet to a different zone by (e.g.) a default route.
425 dst_in6 = ip6->ip6_dst;
426 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
427 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
429 V_ip6stat.ip6s_cantforward++;
430 V_ip6stat.ip6s_badscope++;
434 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
435 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
439 struct secpolicy *sp;
444 mtu = IN6_LINKMTU(rt->rt_ifp);
447 * When we do IPsec tunnel ingress, we need to play
448 * with the link value (decrement IPsec header size
449 * from mtu value). The code is much simpler than v4
450 * case, as we have the outgoing interface for
451 * encapsulated packet as "rt->rt_ifp".
453 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
454 IP_FORWARDING, &ipsecerror);
456 ipsechdrsiz = ipsec_hdrsiz(mcopy,
457 IPSEC_DIR_OUTBOUND, NULL);
458 if (ipsechdrsiz < mtu)
463 * if mtu becomes less than minimum MTU,
464 * tell minimum MTU (and I'll need to fragment it).
469 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
474 if (rt->rt_flags & RTF_GATEWAY)
475 dst = (struct sockaddr_in6 *)rt->rt_gateway;
478 * If we are to forward the packet using the same interface
479 * as one we got the packet from, perhaps we should send a redirect
480 * to sender to shortcut a hop.
481 * Only send redirect if source is sending directly to us,
482 * and if packet was not source routed (or has any options).
483 * Also, don't send redirect if forwarding using a route
484 * modified by a redirect.
486 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
490 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
491 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
493 * If the incoming interface is equal to the outgoing
494 * one, and the link attached to the interface is
495 * point-to-point, then it will be highly probable
496 * that a routing loop occurs. Thus, we immediately
497 * drop the packet and send an ICMPv6 error message.
499 * type/code is based on suggestion by Rich Draves.
500 * not sure if it is the best pick.
502 icmp6_error(mcopy, ICMP6_DST_UNREACH,
503 ICMP6_DST_UNREACH_ADDR, 0);
510 * Fake scoped addresses. Note that even link-local source or
511 * destinaion can appear, if the originating node just sends the
512 * packet to us (without address resolution for the destination).
513 * Since both icmp6_error and icmp6_redirect_output fill the embedded
514 * link identifiers, we can do this stuff after making a copy for
515 * returning an error.
517 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
519 * See corresponding comments in ip6_output.
520 * XXX: but is it possible that ip6_forward() sends a packet
521 * to a loopback interface? I don't think so, and thus
522 * I bark here. (jinmei@kame.net)
523 * XXX: it is common to route invalid packets to loopback.
524 * also, the codepath will be visited on use of ::1 in
530 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
533 printf("ip6_forward: outgoing interface is loopback. "
534 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
535 ip6_sprintf(ip6bufs, &ip6->ip6_src),
536 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
537 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
538 if_name(rt->rt_ifp));
541 /* we can just use rcvif in forwarding. */
542 origifp = m->m_pkthdr.rcvif;
545 origifp = rt->rt_ifp;
547 * clear embedded scope identifiers if necessary.
548 * in6_clearscope will touch the addresses only when necessary.
550 in6_clearscope(&ip6->ip6_src);
551 in6_clearscope(&ip6->ip6_dst);
553 /* Jump over all PFIL processing if hooks are not active. */
554 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
557 /* Run through list of hooks for output packets. */
558 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
563 ip6 = mtod(m, struct ip6_hdr *);
566 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
568 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
569 V_ip6stat.ip6s_cantforward++;
571 V_ip6stat.ip6s_forward++;
572 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
574 V_ip6stat.ip6s_redirectsent++;
586 if (type == ND_REDIRECT) {
587 icmp6_redirect_output(mcopy, rt);
593 /* xxx MTU is constant in PPP? */
597 /* Tell source to slow down like source quench in IP? */
600 case ENETUNREACH: /* shouldn't happen, checked above */
605 type = ICMP6_DST_UNREACH;
606 code = ICMP6_DST_UNREACH_ADDR;
609 icmp6_error(mcopy, type, code, 0);