xref: /openbmc/linux/net/ipv6/ip6_input.c (revision 3805e6a1) 
1 /*
2  *	IPv6 input
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *	Ian P. Morris		<I.P.Morris@soton.ac.uk>
8  *
9  *	Based in linux/net/ipv4/ip_input.c
10  *
11  *	This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  */
16 /* Changes
17  *
18  *	Mitsuru KANDA @USAGI and
19  *	YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs().
20  */
21 
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/netdevice.h>
28 #include <linux/in6.h>
29 #include <linux/icmpv6.h>
30 #include <linux/mroute6.h>
31 #include <linux/slab.h>
32 
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv6.h>
35 
36 #include <net/sock.h>
37 #include <net/snmp.h>
38 
39 #include <net/ipv6.h>
40 #include <net/protocol.h>
41 #include <net/transp_v6.h>
42 #include <net/rawv6.h>
43 #include <net/ndisc.h>
44 #include <net/ip6_route.h>
45 #include <net/addrconf.h>
46 #include <net/xfrm.h>
47 #include <net/inet_ecn.h>
48 #include <net/dst_metadata.h>
49 
50 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
51 {
52 	/* if ingress device is enslaved to an L3 master device pass the
53 	 * skb to its handler for processing
54 	 */
55 	skb = l3mdev_ip6_rcv(skb);
56 	if (!skb)
57 		return NET_RX_SUCCESS;
58 
59 	if (net->ipv4.sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
60 		const struct inet6_protocol *ipprot;
61 
62 		ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
63 		if (ipprot && ipprot->early_demux)
64 			ipprot->early_demux(skb);
65 	}
66 	if (!skb_valid_dst(skb))
67 		ip6_route_input(skb);
68 
69 	return dst_input(skb);
70 }
71 
72 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
73 {
74 	const struct ipv6hdr *hdr;
75 	u32 pkt_len;
76 	struct inet6_dev *idev;
77 	struct net *net = dev_net(skb->dev);
78 
79 	if (skb->pkt_type == PACKET_OTHERHOST) {
80 		kfree_skb(skb);
81 		return NET_RX_DROP;
82 	}
83 
84 	rcu_read_lock();
85 
86 	idev = __in6_dev_get(skb->dev);
87 
88 	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
89 
90 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
91 	    !idev || unlikely(idev->cnf.disable_ipv6)) {
92 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
93 		goto drop;
94 	}
95 
96 	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
97 
98 	/*
99 	 * Store incoming device index. When the packet will
100 	 * be queued, we cannot refer to skb->dev anymore.
101 	 *
102 	 * BTW, when we send a packet for our own local address on a
103 	 * non-loopback interface (e.g. ethX), it is being delivered
104 	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
105 	 * It, however, should be considered as if it is being
106 	 * arrived via the sending interface (ethX), because of the
107 	 * nature of scoping architecture. --yoshfuji
108 	 */
109 	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
110 
111 	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
112 		goto err;
113 
114 	hdr = ipv6_hdr(skb);
115 
116 	if (hdr->version != 6)
117 		goto err;
118 
119 	__IP6_ADD_STATS(net, idev,
120 			IPSTATS_MIB_NOECTPKTS +
121 				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
122 			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
123 	/*
124 	 * RFC4291 2.5.3
125 	 * A packet received on an interface with a destination address
126 	 * of loopback must be dropped.
127 	 */
128 	if (!(dev->flags & IFF_LOOPBACK) &&
129 	    ipv6_addr_loopback(&hdr->daddr))
130 		goto err;
131 
132 	/* RFC4291 Errata ID: 3480
133 	 * Interface-Local scope spans only a single interface on a
134 	 * node and is useful only for loopback transmission of
135 	 * multicast.  Packets with interface-local scope received
136 	 * from another node must be discarded.
137 	 */
138 	if (!(skb->pkt_type == PACKET_LOOPBACK ||
139 	      dev->flags & IFF_LOOPBACK) &&
140 	    ipv6_addr_is_multicast(&hdr->daddr) &&
141 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
142 		goto err;
143 
144 	/* If enabled, drop unicast packets that were encapsulated in link-layer
145 	 * multicast or broadcast to protected against the so-called "hole-196"
146 	 * attack in 802.11 wireless.
147 	 */
148 	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
149 	    (skb->pkt_type == PACKET_BROADCAST ||
150 	     skb->pkt_type == PACKET_MULTICAST) &&
151 	    idev->cnf.drop_unicast_in_l2_multicast)
152 		goto err;
153 
154 	/* RFC4291 2.7
155 	 * Nodes must not originate a packet to a multicast address whose scope
156 	 * field contains the reserved value 0; if such a packet is received, it
157 	 * must be silently dropped.
158 	 */
159 	if (ipv6_addr_is_multicast(&hdr->daddr) &&
160 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
161 		goto err;
162 
163 	/*
164 	 * RFC4291 2.7
165 	 * Multicast addresses must not be used as source addresses in IPv6
166 	 * packets or appear in any Routing header.
167 	 */
168 	if (ipv6_addr_is_multicast(&hdr->saddr))
169 		goto err;
170 
171 	skb->transport_header = skb->network_header + sizeof(*hdr);
172 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
173 
174 	pkt_len = ntohs(hdr->payload_len);
175 
176 	/* pkt_len may be zero if Jumbo payload option is present */
177 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
178 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
179 			__IP6_INC_STATS(net,
180 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
181 			goto drop;
182 		}
183 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
184 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
185 			goto drop;
186 		}
187 		hdr = ipv6_hdr(skb);
188 	}
189 
190 	if (hdr->nexthdr == NEXTHDR_HOP) {
191 		if (ipv6_parse_hopopts(skb) < 0) {
192 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
193 			rcu_read_unlock();
194 			return NET_RX_DROP;
195 		}
196 	}
197 
198 	rcu_read_unlock();
199 
200 	/* Must drop socket now because of tproxy. */
201 	skb_orphan(skb);
202 
203 	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
204 		       net, NULL, skb, dev, NULL,
205 		       ip6_rcv_finish);
206 err:
207 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
208 drop:
209 	rcu_read_unlock();
210 	kfree_skb(skb);
211 	return NET_RX_DROP;
212 }
213 
214 /*
215  *	Deliver the packet to the host
216  */
217 
218 
219 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
220 {
221 	const struct inet6_protocol *ipprot;
222 	struct inet6_dev *idev;
223 	unsigned int nhoff;
224 	int nexthdr;
225 	bool raw;
226 	bool have_final = false;
227 
228 	/*
229 	 *	Parse extension headers
230 	 */
231 
232 	rcu_read_lock();
233 resubmit:
234 	idev = ip6_dst_idev(skb_dst(skb));
235 	if (!pskb_pull(skb, skb_transport_offset(skb)))
236 		goto discard;
237 	nhoff = IP6CB(skb)->nhoff;
238 	nexthdr = skb_network_header(skb)[nhoff];
239 
240 resubmit_final:
241 	raw = raw6_local_deliver(skb, nexthdr);
242 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
243 	if (ipprot) {
244 		int ret;
245 
246 		if (have_final) {
247 			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
248 				/* Once we've seen a final protocol don't
249 				 * allow encapsulation on any non-final
250 				 * ones. This allows foo in UDP encapsulation
251 				 * to work.
252 				 */
253 				goto discard;
254 			}
255 		} else if (ipprot->flags & INET6_PROTO_FINAL) {
256 			const struct ipv6hdr *hdr;
257 
258 			/* Only do this once for first final protocol */
259 			have_final = true;
260 
261 			/* Free reference early: we don't need it any more,
262 			   and it may hold ip_conntrack module loaded
263 			   indefinitely. */
264 			nf_reset(skb);
265 
266 			skb_postpull_rcsum(skb, skb_network_header(skb),
267 					   skb_network_header_len(skb));
268 			hdr = ipv6_hdr(skb);
269 			if (ipv6_addr_is_multicast(&hdr->daddr) &&
270 			    !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
271 			    &hdr->saddr) &&
272 			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
273 				goto discard;
274 		}
275 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
276 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
277 			goto discard;
278 
279 		ret = ipprot->handler(skb);
280 		if (ret > 0) {
281 			if (ipprot->flags & INET6_PROTO_FINAL) {
282 				/* Not an extension header, most likely UDP
283 				 * encapsulation. Use return value as nexthdr
284 				 * protocol not nhoff (which presumably is
285 				 * not set by handler).
286 				 */
287 				nexthdr = ret;
288 				goto resubmit_final;
289 			} else {
290 				goto resubmit;
291 			}
292 		} else if (ret == 0) {
293 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
294 		}
295 	} else {
296 		if (!raw) {
297 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
298 				__IP6_INC_STATS(net, idev,
299 						IPSTATS_MIB_INUNKNOWNPROTOS);
300 				icmpv6_send(skb, ICMPV6_PARAMPROB,
301 					    ICMPV6_UNK_NEXTHDR, nhoff);
302 			}
303 			kfree_skb(skb);
304 		} else {
305 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
306 			consume_skb(skb);
307 		}
308 	}
309 	rcu_read_unlock();
310 	return 0;
311 
312 discard:
313 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
314 	rcu_read_unlock();
315 	kfree_skb(skb);
316 	return 0;
317 }
318 
319 
320 int ip6_input(struct sk_buff *skb)
321 {
322 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
323 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
324 		       ip6_input_finish);
325 }
326 
327 int ip6_mc_input(struct sk_buff *skb)
328 {
329 	const struct ipv6hdr *hdr;
330 	bool deliver;
331 
332 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
333 			 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
334 			 skb->len);
335 
336 	hdr = ipv6_hdr(skb);
337 	deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
338 
339 #ifdef CONFIG_IPV6_MROUTE
340 	/*
341 	 *      IPv6 multicast router mode is now supported ;)
342 	 */
343 	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
344 	    !(ipv6_addr_type(&hdr->daddr) &
345 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
346 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
347 		/*
348 		 * Okay, we try to forward - split and duplicate
349 		 * packets.
350 		 */
351 		struct sk_buff *skb2;
352 		struct inet6_skb_parm *opt = IP6CB(skb);
353 
354 		/* Check for MLD */
355 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
356 			/* Check if this is a mld message */
357 			u8 nexthdr = hdr->nexthdr;
358 			__be16 frag_off;
359 			int offset;
360 
361 			/* Check if the value of Router Alert
362 			 * is for MLD (0x0000).
363 			 */
364 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
365 				deliver = false;
366 
367 				if (!ipv6_ext_hdr(nexthdr)) {
368 					/* BUG */
369 					goto out;
370 				}
371 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
372 							  &nexthdr, &frag_off);
373 				if (offset < 0)
374 					goto out;
375 
376 				if (ipv6_is_mld(skb, nexthdr, offset))
377 					deliver = true;
378 
379 				goto out;
380 			}
381 			/* unknown RA - process it normally */
382 		}
383 
384 		if (deliver)
385 			skb2 = skb_clone(skb, GFP_ATOMIC);
386 		else {
387 			skb2 = skb;
388 			skb = NULL;
389 		}
390 
391 		if (skb2) {
392 			ip6_mr_input(skb2);
393 		}
394 	}
395 out:
396 #endif
397 	if (likely(deliver))
398 		ip6_input(skb);
399 	else {
400 		/* discard */
401 		kfree_skb(skb);
402 	}
403 
404 	return 0;
405 }
406