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"
38 #include "opt_ipsec.h"
39 #include "opt_ipstealth.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/errno.h>
50 #include <sys/kernel.h>
51 #include <sys/syslog.h>
54 #include <net/netisr.h>
55 #include <net/route.h>
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/ip.h>
62 #include <netinet/ip_var.h>
63 #include <netinet6/in6_var.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet6/scope6_var.h>
67 #include <netinet/icmp6.h>
68 #include <netinet6/nd6.h>
70 #include <netinet/in_pcb.h>
73 #include <netipsec/ipsec.h>
74 #include <netipsec/ipsec6.h>
75 #include <netipsec/key.h>
78 #include <netinet6/ip6protosw.h>
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 struct route_in6 rin6;
99 int error, type = 0, code = 0;
100 struct mbuf *mcopy = NULL;
101 struct ifnet *origifp; /* maybe unnecessary */
102 u_int32_t inzone, outzone;
103 struct in6_addr src_in6, dst_in6, odst;
105 struct secpolicy *sp = NULL;
111 #ifdef IPFIREWALL_FORWARD
112 struct m_tag *fwd_tag;
114 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
118 * Check AH/ESP integrity.
121 * Don't increment ip6s_cantforward because this is the check
122 * before forwarding packet actually.
124 if (ipsec6_in_reject(m, NULL)) {
125 V_ipsec6stat.in_polvio++;
132 * Do not forward packets to multicast destination (should be handled
134 * Do not forward packets with unspecified source. It was discussed
135 * in July 2000, on the ipngwg mailing list.
137 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
138 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
139 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
140 V_ip6stat.ip6s_cantforward++;
141 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
142 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
143 V_ip6_log_time = time_second;
146 "from %s to %s nxt %d received on %s\n",
147 ip6_sprintf(ip6bufs, &ip6->ip6_src),
148 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
150 if_name(m->m_pkthdr.rcvif));
159 if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
160 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
161 icmp6_error(m, ICMP6_TIME_EXCEEDED,
162 ICMP6_TIME_EXCEED_TRANSIT, 0);
165 ip6->ip6_hlim -= IPV6_HLIMDEC;
172 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
173 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
174 * we need to generate an ICMP6 message to the src.
175 * Thanks to M_EXT, in most cases copy will not occur.
177 * It is important to save it before IPsec processing as IPsec
178 * processing may modify the mbuf.
180 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
183 /* get a security policy for this packet */
184 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
185 IP_FORWARDING, &error);
187 V_ipsec6stat.out_inval++;
188 V_ip6stat.ip6s_cantforward++;
191 /* XXX: what icmp ? */
203 switch (sp->policy) {
204 case IPSEC_POLICY_DISCARD:
206 * This packet is just discarded.
208 V_ipsec6stat.out_polvio++;
209 V_ip6stat.ip6s_cantforward++;
213 /* XXX: what icmp ? */
221 case IPSEC_POLICY_BYPASS:
222 case IPSEC_POLICY_NONE:
223 /* no need to do IPsec. */
227 case IPSEC_POLICY_IPSEC:
228 if (sp->req == NULL) {
229 /* XXX should be panic ? */
230 printf("ip6_forward: No IPsec request specified.\n");
231 V_ip6stat.ip6s_cantforward++;
235 /* XXX: what icmp ? */
246 case IPSEC_POLICY_ENTRUST:
248 /* should be panic ?? */
249 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
255 struct ipsecrequest *isr = NULL;
256 struct ipsec_output_state state;
259 * when the kernel forwards a packet, it is not proper to apply
260 * IPsec transport mode to the packet is not proper. this check
262 * at present, if there is even a transport mode SA request in the
263 * security policy, the kernel does not apply IPsec to the packet.
264 * this check is not enough because the following case is valid.
265 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require;
267 for (isr = sp->req; isr; isr = isr->next) {
268 if (isr->saidx.mode == IPSEC_MODE_ANY)
270 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
275 * if there's no need for tunnel mode IPsec, skip.
282 * All the extension headers will become inaccessible
283 * (since they can be encrypted).
284 * Don't panic, we need no more updates to extension headers
285 * on inner IPv6 packet (since they are now encapsulated).
287 * IPv6 [ESP|AH] IPv6 [extension headers] payload
289 bzero(&state, sizeof(state));
291 state.ro = NULL; /* update at ipsec6_output_tunnel() */
292 state.dst = NULL; /* update at ipsec6_output_tunnel() */
294 error = ipsec6_output_tunnel(&state, sp, 0);
300 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
309 printf("ip6_output (ipsec): error code %d\n", error);
312 /* don't show these error codes to the user */
315 V_ip6stat.ip6s_cantforward++;
318 /* XXX: what icmp ? */
327 * In the FAST IPSec case we have already
328 * re-injected the packet and it has been freed
329 * by the ipsec_done() function. So, just clean
330 * up after ourselves.
336 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){
338 * now tunnel mode headers are added. we are originating
339 * packet instead of forwarding the packet.
341 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL,
347 dst = (struct sockaddr_in6 *)state.dst;
348 rt = state.ro ? state.ro->ro_rt : NULL;
349 if (dst != NULL && rt != NULL)
357 bzero(&rin6, sizeof(struct route_in6));
358 dst = (struct sockaddr_in6 *)&rin6.ro_dst;
359 dst->sin6_len = sizeof(struct sockaddr_in6);
360 dst->sin6_family = AF_INET6;
361 dst->sin6_addr = ip6->ip6_dst;
362 #ifdef IPFIREWALL_FORWARD
365 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
366 if (rin6.ro_rt != NULL)
367 RT_UNLOCK(rin6.ro_rt);
369 V_ip6stat.ip6s_noroute++;
370 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
372 icmp6_error(mcopy, ICMP6_DST_UNREACH,
373 ICMP6_DST_UNREACH_NOROUTE, 0);
383 * Source scope check: if a packet can't be delivered to its
384 * destination for the reason that the destination is beyond the scope
385 * of the source address, discard the packet and return an icmp6
386 * destination unreachable error with Code 2 (beyond scope of source
387 * address). We use a local copy of ip6_src, since in6_setscope()
388 * will possibly modify its first argument.
389 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
391 src_in6 = ip6->ip6_src;
392 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
393 /* XXX: this should not happen */
394 V_ip6stat.ip6s_cantforward++;
395 V_ip6stat.ip6s_badscope++;
398 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
399 V_ip6stat.ip6s_cantforward++;
400 V_ip6stat.ip6s_badscope++;
403 if (inzone != outzone
408 V_ip6stat.ip6s_cantforward++;
409 V_ip6stat.ip6s_badscope++;
410 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
412 if (V_ip6_log_time + V_ip6_log_interval < time_second) {
413 V_ip6_log_time = time_second;
416 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
417 ip6_sprintf(ip6bufs, &ip6->ip6_src),
418 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
420 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
423 icmp6_error(mcopy, ICMP6_DST_UNREACH,
424 ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
429 * Destination scope check: if a packet is going to break the scope
430 * zone of packet's destination address, discard it. This case should
431 * usually be prevented by appropriately-configured routing table, but
432 * we need an explicit check because we may mistakenly forward the
433 * packet to a different zone by (e.g.) a default route.
435 dst_in6 = ip6->ip6_dst;
436 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
437 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
439 V_ip6stat.ip6s_cantforward++;
440 V_ip6stat.ip6s_badscope++;
444 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
445 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
449 struct secpolicy *sp;
454 mtu = IN6_LINKMTU(rt->rt_ifp);
457 * When we do IPsec tunnel ingress, we need to play
458 * with the link value (decrement IPsec header size
459 * from mtu value). The code is much simpler than v4
460 * case, as we have the outgoing interface for
461 * encapsulated packet as "rt->rt_ifp".
463 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
464 IP_FORWARDING, &ipsecerror);
466 ipsechdrsiz = ipsec_hdrsiz(mcopy,
467 IPSEC_DIR_OUTBOUND, NULL);
468 if (ipsechdrsiz < mtu)
473 * if mtu becomes less than minimum MTU,
474 * tell minimum MTU (and I'll need to fragment it).
479 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
484 if (rt->rt_flags & RTF_GATEWAY)
485 dst = (struct sockaddr_in6 *)rt->rt_gateway;
488 * If we are to forward the packet using the same interface
489 * as one we got the packet from, perhaps we should send a redirect
490 * to sender to shortcut a hop.
491 * Only send redirect if source is sending directly to us,
492 * and if packet was not source routed (or has any options).
493 * Also, don't send redirect if forwarding using a route
494 * modified by a redirect.
496 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
500 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
501 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
503 * If the incoming interface is equal to the outgoing
504 * one, and the link attached to the interface is
505 * point-to-point, then it will be highly probable
506 * that a routing loop occurs. Thus, we immediately
507 * drop the packet and send an ICMPv6 error message.
509 * type/code is based on suggestion by Rich Draves.
510 * not sure if it is the best pick.
512 icmp6_error(mcopy, ICMP6_DST_UNREACH,
513 ICMP6_DST_UNREACH_ADDR, 0);
520 * Fake scoped addresses. Note that even link-local source or
521 * destinaion can appear, if the originating node just sends the
522 * packet to us (without address resolution for the destination).
523 * Since both icmp6_error and icmp6_redirect_output fill the embedded
524 * link identifiers, we can do this stuff after making a copy for
525 * returning an error.
527 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
529 * See corresponding comments in ip6_output.
530 * XXX: but is it possible that ip6_forward() sends a packet
531 * to a loopback interface? I don't think so, and thus
532 * I bark here. (jinmei@kame.net)
533 * XXX: it is common to route invalid packets to loopback.
534 * also, the codepath will be visited on use of ::1 in
540 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
543 printf("ip6_forward: outgoing interface is loopback. "
544 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
545 ip6_sprintf(ip6bufs, &ip6->ip6_src),
546 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
547 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
548 if_name(rt->rt_ifp));
551 /* we can just use rcvif in forwarding. */
552 origifp = m->m_pkthdr.rcvif;
555 origifp = rt->rt_ifp;
557 * clear embedded scope identifiers if necessary.
558 * in6_clearscope will touch the addresses only when necessary.
560 in6_clearscope(&ip6->ip6_src);
561 in6_clearscope(&ip6->ip6_dst);
563 /* Jump over all PFIL processing if hooks are not active. */
564 if (!PFIL_HOOKED(&V_inet6_pfil_hook))
568 /* Run through list of hooks for output packets. */
569 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL);
574 ip6 = mtod(m, struct ip6_hdr *);
576 /* See if destination IP address was changed by packet filter. */
577 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
578 m->m_flags |= M_SKIP_FIREWALL;
579 /* If destination is now ourself drop to ip6_input(). */
580 if (in6_localip(&ip6->ip6_dst)) {
581 m->m_flags |= M_FASTFWD_OURS;
582 if (m->m_pkthdr.rcvif == NULL)
583 m->m_pkthdr.rcvif = V_loif;
584 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
585 m->m_pkthdr.csum_flags |=
586 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
587 m->m_pkthdr.csum_data = 0xffff;
589 m->m_pkthdr.csum_flags |=
590 CSUM_IP_CHECKED | CSUM_IP_VALID;
592 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
593 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
595 error = netisr_queue(NETISR_IPV6, m);
598 goto again; /* Redo the routing table lookup. */
601 #ifdef IPFIREWALL_FORWARD
602 /* See if local, if yes, send it to netisr. */
603 if (m->m_flags & M_FASTFWD_OURS) {
604 if (m->m_pkthdr.rcvif == NULL)
605 m->m_pkthdr.rcvif = V_loif;
606 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
607 m->m_pkthdr.csum_flags |=
608 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
609 m->m_pkthdr.csum_data = 0xffff;
612 if (m->m_pkthdr.csum_flags & CSUM_SCTP)
613 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
615 error = netisr_queue(NETISR_IPV6, m);
618 /* Or forward to some other address? */
619 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
621 dst = (struct sockaddr_in6 *)&rin6.ro_dst;
622 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
623 m->m_flags |= M_SKIP_FIREWALL;
624 m_tag_delete(m, fwd_tag);
627 #endif /* IPFIREWALL_FORWARD */
630 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
632 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
633 V_ip6stat.ip6s_cantforward++;
635 V_ip6stat.ip6s_forward++;
636 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
638 V_ip6stat.ip6s_redirectsent++;
650 if (type == ND_REDIRECT) {
651 icmp6_redirect_output(mcopy, rt);
657 /* xxx MTU is constant in PPP? */
661 /* Tell source to slow down like source quench in IP? */
664 case ENETUNREACH: /* shouldn't happen, checked above */
669 type = ICMP6_DST_UNREACH;
670 code = ICMP6_DST_UNREACH_ADDR;
673 icmp6_error(mcopy, type, code, 0);