xref: /netbsd-src/sys/netinet6/ip6_forward.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: ip6_forward.c,v 1.85 2017/01/16 15:44:47 christos 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 1.85 2017/01/16 15:44:47 christos 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/errno.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/syslog.h>
48 #include <sys/percpu.h>
49 
50 #include <net/if.h>
51 #include <net/route.h>
52 #include <net/pfil.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/ip6.h>
58 #include <netinet6/ip6_var.h>
59 #include <netinet6/ip6_private.h>
60 #include <netinet6/scope6_var.h>
61 #include <netinet/icmp6.h>
62 #include <netinet6/nd6.h>
63 
64 #ifdef IPSEC
65 #include <netipsec/ipsec.h>
66 #include <netipsec/ipsec6.h>
67 #include <netipsec/key.h>
68 #include <netipsec/xform.h>
69 #endif /* IPSEC */
70 
71 #include <net/net_osdep.h>
72 
73 extern percpu_t *ip6_forward_rt_percpu;
74 
75 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
76 
77 static void __printflike(4, 5)
78 ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp,
79     const struct ifnet *dstifp, const char *fmt, ...)
80 {
81 	char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
82 	char reason[256];
83 	va_list ap;
84 	uint64_t *ip6s;
85 
86 	/* update statistics */
87 	ip6s = IP6_STAT_GETREF();
88 	ip6s[IP6_STAT_CANTFORWARD]++;
89 	if (dstifp)
90 		ip6s[IP6_STAT_BADSCOPE]++;
91 	IP6_STAT_PUTREF();
92 
93 	if (dstifp)
94 		in6_ifstat_inc(dstifp, ifs6_in_discard);
95 
96 	if (ip6_log_time + ip6_log_interval >= time_uptime)
97 		return;
98 	ip6_log_time = time_uptime;
99 
100 	va_start(ap, fmt);
101 	vsnprintf(reason, sizeof(reason), fmt, ap);
102 	va_end(ap);
103 
104 	log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n",
105 	    IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?",
106 	    IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?",
107 	    ip6->ip6_nxt, reason);
108 }
109 
110 /*
111  * Forward a packet.  If some error occurs return the sender
112  * an icmp packet.  Note we can't always generate a meaningful
113  * icmp message because icmp doesn't have a large enough repertoire
114  * of codes and types.
115  *
116  * If not forwarding, just drop the packet.  This could be confusing
117  * if ipforwarding was zero but some routing protocol was advancing
118  * us as a gateway to somewhere.  However, we must let the routing
119  * protocol deal with that.
120  *
121  */
122 
123 void
124 ip6_forward(struct mbuf *m, int srcrt)
125 {
126 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
127 	const struct sockaddr_in6 *dst;
128 	struct rtentry *rt = NULL;
129 	int error = 0, type = 0, code = 0;
130 	struct mbuf *mcopy = NULL;
131 	struct ifnet *origifp;	/* maybe unnecessary */
132 	uint32_t inzone, outzone;
133 	struct in6_addr src_in6, dst_in6;
134 	struct ifnet *rcvif = NULL;
135 	struct psref psref;
136 	struct route *ro = NULL;
137 #ifdef IPSEC
138 	int needipsec = 0;
139 	struct secpolicy *sp = NULL;
140 #endif
141 
142 	/*
143 	 * Clear any in-bound checksum flags for this packet.
144 	 */
145 	m->m_pkthdr.csum_flags = 0;
146 
147 	rcvif = m_get_rcvif_psref(m, &psref);
148 	if (__predict_false(rcvif == NULL))
149 		goto drop;
150 
151 	/*
152 	 * Do not forward packets to multicast destination (should be handled
153 	 * by ip6_mforward().
154 	 * Do not forward packets with unspecified source.  It was discussed
155 	 * in July 2000, on ipngwg mailing list.
156 	 */
157 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
158 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
159 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
160 		ip6_cantforward(ip6, rcvif, NULL,
161 		    ((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" :
162 		    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" :
163 		    "unspec/src");
164 		goto drop;
165 	}
166 
167 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
168 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
169 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
170 				ICMP6_TIME_EXCEED_TRANSIT, 0);
171 		goto out;
172 	}
173 	ip6->ip6_hlim -= IPV6_HLIMDEC;
174 
175 	/*
176 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
177 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
178 	 * we need to generate an ICMP6 message to the src.
179 	 * Thanks to M_EXT, in most cases copy will not occur.
180 	 *
181 	 * It is important to save it before IPsec processing as IPsec
182 	 * processing may modify the mbuf.
183 	 */
184 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
185 
186 #ifdef IPSEC
187 	if (ipsec_used) {
188 		/* Check the security policy (SP) for the packet */
189 
190 		sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error);
191 		if (error != 0) {
192 			/*
193 			 * Hack: -EINVAL is used to signal that a packet
194 			 * should be silently discarded.  This is typically
195 			 * because we asked key management for an SA and
196 			 * it was delayed (e.g. kicked up to IKE).
197 			 */
198 			if (error == -EINVAL)
199 				error = 0;
200 			goto freecopy;
201 		}
202 	}
203 #endif /* IPSEC */
204 
205 	ro = percpu_getref(ip6_forward_rt_percpu);
206 	if (srcrt) {
207 		union {
208 			struct sockaddr		dst;
209 			struct sockaddr_in6	dst6;
210 		} u;
211 
212 		sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0);
213 		rt = rtcache_lookup(ro, &u.dst);
214 		if (rt == NULL) {
215 			IP6_STATINC(IP6_STAT_NOROUTE);
216 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
217 			if (mcopy) {
218 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
219 					    ICMP6_DST_UNREACH_NOROUTE, 0);
220 			}
221 			goto drop;
222 		}
223 	} else if ((rt = rtcache_validate(ro)) == NULL &&
224 	           (rt = rtcache_update(ro, 1)) == NULL) {
225 		/*
226 		 * rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to
227 		 * ip6->ip6_dst
228 		 */
229 		IP6_STATINC(IP6_STAT_NOROUTE);
230 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
231 		if (mcopy) {
232 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
233 			    ICMP6_DST_UNREACH_NOROUTE, 0);
234 		}
235 		goto drop;
236 	}
237 	dst = satocsin6(rtcache_getdst(ro));
238 
239 	/*
240 	 * Source scope check: if a packet can't be delivered to its
241 	 * destination for the reason that the destination is beyond the scope
242 	 * of the source address, discard the packet and return an icmp6
243 	 * destination unreachable error with Code 2 (beyond scope of source
244 	 * address).  We use a local copy of ip6_src, since in6_setscope()
245 	 * will possibly modify its first argument.
246 	 * [draft-ietf-ipngwg-icmp-v3-07, Section 3.1]
247 	 */
248 	src_in6 = ip6->ip6_src;
249 	inzone = outzone = ~0;
250 	if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 ||
251 	    in6_setscope(&src_in6, rcvif, &inzone) != 0 ||
252 	    inzone != outzone) {
253 		ip6_cantforward(ip6, rcvif, rt->rt_ifp,
254 		    "src[%s] inzone %d outzone %d",
255 		    in6_getscopename(&ip6->ip6_src), inzone, outzone);
256 		if (mcopy)
257 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
258 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
259 		goto drop;
260 	}
261 
262 #ifdef IPSEC
263 	/*
264 	 * If we need to encapsulate the packet, do it here
265 	 * ipsec6_proces_packet will send the packet using ip6_output
266 	 */
267 	if (needipsec) {
268 		int s = splsoftnet();
269 		error = ipsec6_process_packet(m, sp->req);
270 		splx(s);
271 		if (mcopy)
272 			goto freecopy;
273 	}
274 #endif
275 
276 	/*
277 	 * Destination scope check: if a packet is going to break the scope
278 	 * zone of packet's destination address, discard it.  This case should
279 	 * usually be prevented by appropriately-configured routing table, but
280 	 * we need an explicit check because we may mistakenly forward the
281 	 * packet to a different zone by (e.g.) a default route.
282 	 */
283 	dst_in6 = ip6->ip6_dst;
284 	inzone = outzone = ~0;
285 	if (in6_setscope(&dst_in6, rcvif, &inzone) != 0 ||
286 	    in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
287 	    inzone != outzone) {
288 		ip6_cantforward(ip6, rcvif, rt->rt_ifp,
289 		    "dst[%s] inzone %d outzone %d",
290 		    in6_getscopename(&ip6->ip6_dst), inzone, outzone);
291 		if (mcopy)
292 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
293 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
294 		goto drop;
295 	}
296 
297 	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
298 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
299 		if (mcopy) {
300 			u_long mtu;
301 
302 			mtu = IN6_LINKMTU(rt->rt_ifp);
303 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
304 		}
305 		goto drop;
306 	}
307 
308 	if (rt->rt_flags & RTF_GATEWAY)
309 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
310 
311 	/*
312 	 * If we are to forward the packet using the same interface
313 	 * as one we got the packet from, perhaps we should send a redirect
314 	 * to sender to shortcut a hop.
315 	 * Only send redirect if source is sending directly to us,
316 	 * and if packet was not source routed (or has any options).
317 	 * Also, don't send redirect if forwarding using a route
318 	 * modified by a redirect.
319 	 */
320 	if (rt->rt_ifp == rcvif && !srcrt && ip6_sendredirects &&
321 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
322 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
323 		    nd6_is_addr_neighbor(satocsin6(rtcache_getdst(ro)),
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 			goto drop;
343 		}
344 		type = ND_REDIRECT;
345 	}
346 
347 	/*
348 	 * Fake scoped addresses. Note that even link-local source or
349 	 * destinaion can appear, if the originating node just sends the
350 	 * packet to us (without address resolution for the destination).
351 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
352 	 * link identifiers, we can do this stuff after making a copy for
353 	 * returning an error.
354 	 */
355 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
356 		/*
357 		 * See corresponding comments in ip6_output.
358 		 * XXX: but is it possible that ip6_forward() sends a packet
359 		 *      to a loopback interface? I don't think so, and thus
360 		 *      I bark here. (jinmei@kame.net)
361 		 * XXX: it is common to route invalid packets to loopback.
362 		 *	also, the codepath will be visited on use of ::1 in
363 		 *	rthdr. (itojun)
364 		 */
365 #if 1
366 		if (0)
367 #else
368 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
369 #endif
370 		{
371 			char ip6bufs[INET6_ADDRSTRLEN];
372 			char ip6bufd[INET6_ADDRSTRLEN];
373 
374 			printf("ip6_forward: outgoing interface is loopback. "
375 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
376 			       IN6_PRINT(ip6bufs, &ip6->ip6_src),
377 			       IN6_PRINT(ip6bufd, &ip6->ip6_dst),
378 			       ip6->ip6_nxt, if_name(rcvif),
379 			       if_name(rt->rt_ifp));
380 		}
381 
382 		/* we can just use rcvif in forwarding. */
383 		origifp = rcvif;
384 	}
385 	else
386 		origifp = rt->rt_ifp;
387 	/*
388 	 * clear embedded scope identifiers if necessary.
389 	 * in6_clearscope will touch the addresses only when necessary.
390 	 */
391 	in6_clearscope(&ip6->ip6_src);
392 	in6_clearscope(&ip6->ip6_dst);
393 
394 	/*
395 	 * Run through list of hooks for output packets.
396 	 */
397 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp,
398 	    PFIL_OUT)) != 0)
399 		goto senderr;
400 	if (m == NULL)
401 		goto freecopy;
402 	ip6 = mtod(m, struct ip6_hdr *);
403 
404 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
405 	if (error) {
406 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
407 		IP6_STATINC(IP6_STAT_CANTFORWARD);
408 	} else {
409 		IP6_STATINC(IP6_STAT_FORWARD);
410 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
411 		if (type)
412 			IP6_STATINC(IP6_STAT_REDIRECTSENT);
413 		else {
414 #ifdef GATEWAY
415 			/* Need to release rt here */
416 			rtcache_unref(rt, ro);
417 			rt = NULL;
418 			if (m->m_flags & M_CANFASTFWD)
419 				ip6flow_create(ro, m);
420 #endif
421 			if (mcopy)
422 				goto freecopy;
423 		}
424 	}
425 
426  senderr:
427 	if (mcopy == NULL)
428 		goto out;
429 	switch (error) {
430 	case 0:
431 		if (type == ND_REDIRECT) {
432 			icmp6_redirect_output(mcopy, rt);
433 			goto out;
434 		}
435 		goto freecopy;
436 
437 	case EMSGSIZE:
438 		/* xxx MTU is constant in PPP? */
439 		goto freecopy;
440 
441 	case ENOBUFS:
442 		/* Tell source to slow down like source quench in IP? */
443 		goto freecopy;
444 
445 	case ENETUNREACH:	/* shouldn't happen, checked above */
446 	case EHOSTUNREACH:
447 	case ENETDOWN:
448 	case EHOSTDOWN:
449 	default:
450 		type = ICMP6_DST_UNREACH;
451 		code = ICMP6_DST_UNREACH_ADDR;
452 		break;
453 	}
454 	icmp6_error(mcopy, type, code, 0);
455 	goto out;
456 
457  freecopy:
458 	m_freem(mcopy);
459 	goto out;
460  drop:
461  	m_freem(m);
462  out:
463 	rtcache_unref(rt, ro);
464 	if (ro != NULL)
465 		percpu_putref(ip6_forward_rt_percpu);
466 	if (rcvif != NULL)
467 		m_put_rcvif_psref(rcvif, &psref);
468 	return;
469 }
470