xref: /netbsd-src/sys/netinet6/ip6_forward.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: ip6_forward.c,v 1.78 2015/08/24 22:21:27 pooka Exp $	*/
2 /*	$KAME: ip6_forward.c,v 1.109 2002/09/11 08:10:17 sakane Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: ip6_forward.c,v 1.78 2015/08/24 22:21:27 pooka Exp $");
35 
36 #ifdef _KERNEL_OPT
37 #include "opt_gateway.h"
38 #include "opt_ipsec.h"
39 #endif
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/errno.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/syslog.h>
51 
52 #include <net/if.h>
53 #include <net/route.h>
54 #include <net/pfil.h>
55 
56 #include <netinet/in.h>
57 #include <netinet/in_var.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip6.h>
60 #include <netinet6/ip6_var.h>
61 #include <netinet6/ip6_private.h>
62 #include <netinet6/scope6_var.h>
63 #include <netinet/icmp6.h>
64 #include <netinet6/nd6.h>
65 
66 #ifdef IPSEC
67 #include <netipsec/ipsec.h>
68 #include <netipsec/ipsec6.h>
69 #include <netipsec/key.h>
70 #include <netipsec/xform.h>
71 #endif /* IPSEC */
72 
73 #include <net/net_osdep.h>
74 
75 struct	route ip6_forward_rt;
76 
77 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
78 
79 static void __printflike(4, 5)
80 ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp,
81     const struct ifnet *dstifp, const char *fmt, ...)
82 {
83 	char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
84 	char reason[256];
85 	va_list ap;
86 	uint64_t *ip6s;
87 
88 	/* update statistics */
89 	ip6s = IP6_STAT_GETREF();
90 	ip6s[IP6_STAT_CANTFORWARD]++;
91 	if (dstifp)
92 		ip6s[IP6_STAT_BADSCOPE]++;
93 	IP6_STAT_PUTREF();
94 
95 	if (dstifp)
96 		in6_ifstat_inc(dstifp, ifs6_in_discard);
97 
98 	if (ip6_log_time + ip6_log_interval >= time_uptime)
99 		return;
100 	ip6_log_time = time_uptime;
101 
102 	va_start(ap, fmt);
103 	vsnprintf(reason, sizeof(reason), fmt, ap);
104 	va_end(ap);
105 
106 	log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n",
107 	    IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?",
108 	    IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?",
109 	    ip6->ip6_nxt, reason);
110 }
111 
112 /*
113  * Forward a packet.  If some error occurs return the sender
114  * an icmp packet.  Note we can't always generate a meaningful
115  * icmp message because icmp doesn't have a large enough repertoire
116  * of codes and types.
117  *
118  * If not forwarding, just drop the packet.  This could be confusing
119  * if ipforwarding was zero but some routing protocol was advancing
120  * us as a gateway to somewhere.  However, we must let the routing
121  * protocol deal with that.
122  *
123  */
124 
125 void
126 ip6_forward(struct mbuf *m, int srcrt)
127 {
128 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
129 	const struct sockaddr_in6 *dst;
130 	struct rtentry *rt;
131 	int error = 0, type = 0, code = 0;
132 	struct mbuf *mcopy = NULL;
133 	struct ifnet *origifp;	/* maybe unnecessary */
134 	uint32_t inzone, outzone;
135 	struct in6_addr src_in6, dst_in6;
136 #ifdef IPSEC
137 	int needipsec = 0;
138 	struct secpolicy *sp = NULL;
139 #endif
140 
141 	/*
142 	 * Clear any in-bound checksum flags for this packet.
143 	 */
144 	m->m_pkthdr.csum_flags = 0;
145 
146 	/*
147 	 * Do not forward packets to multicast destination (should be handled
148 	 * by ip6_mforward().
149 	 * Do not forward packets with unspecified source.  It was discussed
150 	 * in July 2000, on ipngwg mailing list.
151 	 */
152 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
153 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
154 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
155 		ip6_cantforward(ip6, m->m_pkthdr.rcvif, NULL,
156 		    ((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" :
157 		    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" :
158 		    "unspec/src");
159 		m_freem(m);
160 		return;
161 	}
162 
163 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
164 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
165 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
166 				ICMP6_TIME_EXCEED_TRANSIT, 0);
167 		return;
168 	}
169 	ip6->ip6_hlim -= IPV6_HLIMDEC;
170 
171 	/*
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.
176 	 *
177 	 * It is important to save it before IPsec processing as IPsec
178 	 * processing may modify the mbuf.
179 	 */
180 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
181 
182 #ifdef IPSEC
183 	if (ipsec_used) {
184 		/* Check the security policy (SP) for the packet */
185 
186 		sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error);
187 		if (error != 0) {
188 			/*
189 			 * Hack: -EINVAL is used to signal that a packet
190 			 * should be silently discarded.  This is typically
191 			 * because we asked key management for an SA and
192 			 * it was delayed (e.g. kicked up to IKE).
193 			 */
194 			if (error == -EINVAL)
195 				error = 0;
196 			goto freecopy;
197 		}
198 	}
199 #endif /* IPSEC */
200 
201 	if (srcrt) {
202 		union {
203 			struct sockaddr		dst;
204 			struct sockaddr_in6	dst6;
205 		} u;
206 
207 		sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
208 		if ((rt = rtcache_lookup(&ip6_forward_rt, &u.dst)) == NULL) {
209 			IP6_STATINC(IP6_STAT_NOROUTE);
210 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
211 			if (mcopy) {
212 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
213 					    ICMP6_DST_UNREACH_NOROUTE, 0);
214 			}
215 			m_freem(m);
216 			return;
217 		}
218 	} else if ((rt = rtcache_validate(&ip6_forward_rt)) == NULL &&
219 	           (rt = rtcache_update(&ip6_forward_rt, 1)) == NULL) {
220 		/*
221 		 * rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to
222 		 * ip6->ip6_dst
223 		 */
224 		IP6_STATINC(IP6_STAT_NOROUTE);
225 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
226 		if (mcopy) {
227 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
228 			    ICMP6_DST_UNREACH_NOROUTE, 0);
229 		}
230 		m_freem(m);
231 		return;
232 	}
233 	dst = satocsin6(rtcache_getdst(&ip6_forward_rt));
234 
235 	/*
236 	 * Source scope check: if a packet can't be delivered to its
237 	 * destination for the reason that the destination is beyond the scope
238 	 * of the source address, discard the packet and return an icmp6
239 	 * destination unreachable error with Code 2 (beyond scope of source
240 	 * address).  We use a local copy of ip6_src, since in6_setscope()
241 	 * will possibly modify its first argument.
242 	 * [draft-ietf-ipngwg-icmp-v3-07, Section 3.1]
243 	 */
244 	src_in6 = ip6->ip6_src;
245 	inzone = outzone = ~0;
246 	if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 ||
247 	    in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
248 	    inzone != outzone) {
249 		ip6_cantforward(ip6, m->m_pkthdr.rcvif, rt->rt_ifp,
250 		    "src[%s] inzone %d outzone %d",
251 		    in6_getscopename(&ip6->ip6_src), inzone, outzone);
252 		if (mcopy)
253 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
254 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
255 		m_freem(m);
256 		return;
257 	}
258 
259 #ifdef IPSEC
260 	/*
261 	 * If we need to encapsulate the packet, do it here
262 	 * ipsec6_proces_packet will send the packet using ip6_output
263 	 */
264 	if (needipsec) {
265 		int s = splsoftnet();
266 		error = ipsec6_process_packet(m, sp->req);
267 		splx(s);
268 		if (mcopy)
269 			goto freecopy;
270 	}
271 #endif
272 
273 	/*
274 	 * Destination scope check: if a packet is going to break the scope
275 	 * zone of packet's destination address, discard it.  This case should
276 	 * usually be prevented by appropriately-configured routing table, but
277 	 * we need an explicit check because we may mistakenly forward the
278 	 * packet to a different zone by (e.g.) a default route.
279 	 */
280 	dst_in6 = ip6->ip6_dst;
281 	inzone = outzone = ~0;
282 	if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
283 	    in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
284 	    inzone != outzone) {
285 		ip6_cantforward(ip6, m->m_pkthdr.rcvif, rt->rt_ifp,
286 		    "dst[%s] inzone %d outzone %d",
287 		    in6_getscopename(&ip6->ip6_dst), inzone, outzone);
288 		if (mcopy)
289 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
290 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
291 		m_freem(m);
292 		return;
293 	}
294 
295 	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
296 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
297 		if (mcopy) {
298 			u_long mtu;
299 
300 			mtu = IN6_LINKMTU(rt->rt_ifp);
301 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
302 		}
303 		m_freem(m);
304 		return;
305 	}
306 
307 	if (rt->rt_flags & RTF_GATEWAY)
308 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
309 
310 	/*
311 	 * If we are to forward the packet using the same interface
312 	 * as one we got the packet from, perhaps we should send a redirect
313 	 * to sender to shortcut a hop.
314 	 * Only send redirect if source is sending directly to us,
315 	 * and if packet was not source routed (or has any options).
316 	 * Also, don't send redirect if forwarding using a route
317 	 * modified by a redirect.
318 	 */
319 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects &&
320 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
321 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
322 		    nd6_is_addr_neighbor(
323 		        satocsin6(rtcache_getdst(&ip6_forward_rt)),
324 			rt->rt_ifp)) {
325 			/*
326 			 * If the incoming interface is equal to the outgoing
327 			 * one, the link attached to the interface is
328 			 * point-to-point, and the IPv6 destination is
329 			 * regarded as on-link on the link, then it will be
330 			 * highly probable that the destination address does
331 			 * not exist on the link and that the packet is going
332 			 * to loop.  Thus, we immediately drop the packet and
333 			 * send an ICMPv6 error message.
334 			 * For other routing loops, we dare to let the packet
335 			 * go to the loop, so that a remote diagnosing host
336 			 * can detect the loop by traceroute.
337 			 * type/code is based on suggestion by Rich Draves.
338 			 * not sure if it is the best pick.
339 			 */
340 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
341 				    ICMP6_DST_UNREACH_ADDR, 0);
342 			m_freem(m);
343 			return;
344 		}
345 		type = ND_REDIRECT;
346 	}
347 
348 	/*
349 	 * Fake scoped addresses. Note that even link-local source or
350 	 * destinaion can appear, if the originating node just sends the
351 	 * packet to us (without address resolution for the destination).
352 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
353 	 * link identifiers, we can do this stuff after making a copy for
354 	 * returning an error.
355 	 */
356 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
357 		/*
358 		 * See corresponding comments in ip6_output.
359 		 * XXX: but is it possible that ip6_forward() sends a packet
360 		 *      to a loopback interface? I don't think so, and thus
361 		 *      I bark here. (jinmei@kame.net)
362 		 * XXX: it is common to route invalid packets to loopback.
363 		 *	also, the codepath will be visited on use of ::1 in
364 		 *	rthdr. (itojun)
365 		 */
366 #if 1
367 		if (0)
368 #else
369 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
370 #endif
371 		{
372 			printf("ip6_forward: outgoing interface is loopback. "
373 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
374 			       ip6_sprintf(&ip6->ip6_src),
375 			       ip6_sprintf(&ip6->ip6_dst),
376 			       ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
377 			       if_name(rt->rt_ifp));
378 		}
379 
380 		/* we can just use rcvif in forwarding. */
381 		origifp = m->m_pkthdr.rcvif;
382 	}
383 	else
384 		origifp = rt->rt_ifp;
385 	/*
386 	 * clear embedded scope identifiers if necessary.
387 	 * in6_clearscope will touch the addresses only when necessary.
388 	 */
389 	in6_clearscope(&ip6->ip6_src);
390 	in6_clearscope(&ip6->ip6_dst);
391 
392 	/*
393 	 * Run through list of hooks for output packets.
394 	 */
395 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp,
396 	    PFIL_OUT)) != 0)
397 		goto senderr;
398 	if (m == NULL)
399 		goto freecopy;
400 	ip6 = mtod(m, struct ip6_hdr *);
401 
402 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
403 	if (error) {
404 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
405 		IP6_STATINC(IP6_STAT_CANTFORWARD);
406 	} else {
407 		IP6_STATINC(IP6_STAT_FORWARD);
408 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
409 		if (type)
410 			IP6_STATINC(IP6_STAT_REDIRECTSENT);
411 		else {
412 #ifdef GATEWAY
413 			if (m->m_flags & M_CANFASTFWD)
414 				ip6flow_create(&ip6_forward_rt, m);
415 #endif
416 			if (mcopy)
417 				goto freecopy;
418 		}
419 	}
420 
421  senderr:
422 	if (mcopy == NULL)
423 		return;
424 	switch (error) {
425 	case 0:
426 		if (type == ND_REDIRECT) {
427 			icmp6_redirect_output(mcopy, rt);
428 			return;
429 		}
430 		goto freecopy;
431 
432 	case EMSGSIZE:
433 		/* xxx MTU is constant in PPP? */
434 		goto freecopy;
435 
436 	case ENOBUFS:
437 		/* Tell source to slow down like source quench in IP? */
438 		goto freecopy;
439 
440 	case ENETUNREACH:	/* shouldn't happen, checked above */
441 	case EHOSTUNREACH:
442 	case ENETDOWN:
443 	case EHOSTDOWN:
444 	default:
445 		type = ICMP6_DST_UNREACH;
446 		code = ICMP6_DST_UNREACH_ADDR;
447 		break;
448 	}
449 	icmp6_error(mcopy, type, code, 0);
450 	return;
451 
452  freecopy:
453 	m_freem(mcopy);
454 	return;
455 }
456