xref: /openbmc/linux/net/ipv6/ip6_input.c (revision 4f139972b489f8bc2c821aa25ac65018d92af3f7) 
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 	void (*edemux)(struct sk_buff *skb);
53 
54 	/* if ingress device is enslaved to an L3 master device pass the
55 	 * skb to its handler for processing
56 	 */
57 	skb = l3mdev_ip6_rcv(skb);
58 	if (!skb)
59 		return NET_RX_SUCCESS;
60 
61 	if (net->ipv4.sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
62 		const struct inet6_protocol *ipprot;
63 
64 		ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
65 		if (ipprot && (edemux = READ_ONCE(ipprot->early_demux)))
66 			edemux(skb);
67 	}
68 	if (!skb_valid_dst(skb))
69 		ip6_route_input(skb);
70 
71 	return dst_input(skb);
72 }
73 
74 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
75 {
76 	const struct ipv6hdr *hdr;
77 	u32 pkt_len;
78 	struct inet6_dev *idev;
79 	struct net *net = dev_net(skb->dev);
80 
81 	if (skb->pkt_type == PACKET_OTHERHOST) {
82 		kfree_skb(skb);
83 		return NET_RX_DROP;
84 	}
85 
86 	rcu_read_lock();
87 
88 	idev = __in6_dev_get(skb->dev);
89 
90 	__IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len);
91 
92 	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
93 	    !idev || unlikely(idev->cnf.disable_ipv6)) {
94 		__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
95 		goto drop;
96 	}
97 
98 	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
99 
100 	/*
101 	 * Store incoming device index. When the packet will
102 	 * be queued, we cannot refer to skb->dev anymore.
103 	 *
104 	 * BTW, when we send a packet for our own local address on a
105 	 * non-loopback interface (e.g. ethX), it is being delivered
106 	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
107 	 * It, however, should be considered as if it is being
108 	 * arrived via the sending interface (ethX), because of the
109 	 * nature of scoping architecture. --yoshfuji
110 	 */
111 	IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
112 
113 	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
114 		goto err;
115 
116 	hdr = ipv6_hdr(skb);
117 
118 	if (hdr->version != 6)
119 		goto err;
120 
121 	__IP6_ADD_STATS(net, idev,
122 			IPSTATS_MIB_NOECTPKTS +
123 				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
124 			max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
125 	/*
126 	 * RFC4291 2.5.3
127 	 * A packet received on an interface with a destination address
128 	 * of loopback must be dropped.
129 	 */
130 	if (!(dev->flags & IFF_LOOPBACK) &&
131 	    ipv6_addr_loopback(&hdr->daddr))
132 		goto err;
133 
134 	/* RFC4291 Errata ID: 3480
135 	 * Interface-Local scope spans only a single interface on a
136 	 * node and is useful only for loopback transmission of
137 	 * multicast.  Packets with interface-local scope received
138 	 * from another node must be discarded.
139 	 */
140 	if (!(skb->pkt_type == PACKET_LOOPBACK ||
141 	      dev->flags & IFF_LOOPBACK) &&
142 	    ipv6_addr_is_multicast(&hdr->daddr) &&
143 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
144 		goto err;
145 
146 	/* If enabled, drop unicast packets that were encapsulated in link-layer
147 	 * multicast or broadcast to protected against the so-called "hole-196"
148 	 * attack in 802.11 wireless.
149 	 */
150 	if (!ipv6_addr_is_multicast(&hdr->daddr) &&
151 	    (skb->pkt_type == PACKET_BROADCAST ||
152 	     skb->pkt_type == PACKET_MULTICAST) &&
153 	    idev->cnf.drop_unicast_in_l2_multicast)
154 		goto err;
155 
156 	/* RFC4291 2.7
157 	 * Nodes must not originate a packet to a multicast address whose scope
158 	 * field contains the reserved value 0; if such a packet is received, it
159 	 * must be silently dropped.
160 	 */
161 	if (ipv6_addr_is_multicast(&hdr->daddr) &&
162 	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
163 		goto err;
164 
165 	/*
166 	 * RFC4291 2.7
167 	 * Multicast addresses must not be used as source addresses in IPv6
168 	 * packets or appear in any Routing header.
169 	 */
170 	if (ipv6_addr_is_multicast(&hdr->saddr))
171 		goto err;
172 
173 	skb->transport_header = skb->network_header + sizeof(*hdr);
174 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
175 
176 	pkt_len = ntohs(hdr->payload_len);
177 
178 	/* pkt_len may be zero if Jumbo payload option is present */
179 	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
180 		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
181 			__IP6_INC_STATS(net,
182 					idev, IPSTATS_MIB_INTRUNCATEDPKTS);
183 			goto drop;
184 		}
185 		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
186 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
187 			goto drop;
188 		}
189 		hdr = ipv6_hdr(skb);
190 	}
191 
192 	if (hdr->nexthdr == NEXTHDR_HOP) {
193 		if (ipv6_parse_hopopts(skb) < 0) {
194 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
195 			rcu_read_unlock();
196 			return NET_RX_DROP;
197 		}
198 	}
199 
200 	rcu_read_unlock();
201 
202 	/* Must drop socket now because of tproxy. */
203 	skb_orphan(skb);
204 
205 	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
206 		       net, NULL, skb, dev, NULL,
207 		       ip6_rcv_finish);
208 err:
209 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
210 drop:
211 	rcu_read_unlock();
212 	kfree_skb(skb);
213 	return NET_RX_DROP;
214 }
215 
216 /*
217  *	Deliver the packet to the host
218  */
219 
220 
221 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
222 {
223 	const struct inet6_protocol *ipprot;
224 	struct inet6_dev *idev;
225 	unsigned int nhoff;
226 	int nexthdr;
227 	bool raw;
228 	bool have_final = false;
229 
230 	/*
231 	 *	Parse extension headers
232 	 */
233 
234 	rcu_read_lock();
235 resubmit:
236 	idev = ip6_dst_idev(skb_dst(skb));
237 	if (!pskb_pull(skb, skb_transport_offset(skb)))
238 		goto discard;
239 	nhoff = IP6CB(skb)->nhoff;
240 	nexthdr = skb_network_header(skb)[nhoff];
241 
242 resubmit_final:
243 	raw = raw6_local_deliver(skb, nexthdr);
244 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
245 	if (ipprot) {
246 		int ret;
247 
248 		if (have_final) {
249 			if (!(ipprot->flags & INET6_PROTO_FINAL)) {
250 				/* Once we've seen a final protocol don't
251 				 * allow encapsulation on any non-final
252 				 * ones. This allows foo in UDP encapsulation
253 				 * to work.
254 				 */
255 				goto discard;
256 			}
257 		} else if (ipprot->flags & INET6_PROTO_FINAL) {
258 			const struct ipv6hdr *hdr;
259 
260 			/* Only do this once for first final protocol */
261 			have_final = true;
262 
263 			/* Free reference early: we don't need it any more,
264 			   and it may hold ip_conntrack module loaded
265 			   indefinitely. */
266 			nf_reset(skb);
267 
268 			skb_postpull_rcsum(skb, skb_network_header(skb),
269 					   skb_network_header_len(skb));
270 			hdr = ipv6_hdr(skb);
271 			if (ipv6_addr_is_multicast(&hdr->daddr) &&
272 			    !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
273 			    &hdr->saddr) &&
274 			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
275 				goto discard;
276 		}
277 		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
278 		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
279 			goto discard;
280 
281 		ret = ipprot->handler(skb);
282 		if (ret > 0) {
283 			if (ipprot->flags & INET6_PROTO_FINAL) {
284 				/* Not an extension header, most likely UDP
285 				 * encapsulation. Use return value as nexthdr
286 				 * protocol not nhoff (which presumably is
287 				 * not set by handler).
288 				 */
289 				nexthdr = ret;
290 				goto resubmit_final;
291 			} else {
292 				goto resubmit;
293 			}
294 		} else if (ret == 0) {
295 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
296 		}
297 	} else {
298 		if (!raw) {
299 			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
300 				__IP6_INC_STATS(net, idev,
301 						IPSTATS_MIB_INUNKNOWNPROTOS);
302 				icmpv6_send(skb, ICMPV6_PARAMPROB,
303 					    ICMPV6_UNK_NEXTHDR, nhoff);
304 			}
305 			kfree_skb(skb);
306 		} else {
307 			__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
308 			consume_skb(skb);
309 		}
310 	}
311 	rcu_read_unlock();
312 	return 0;
313 
314 discard:
315 	__IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
316 	rcu_read_unlock();
317 	kfree_skb(skb);
318 	return 0;
319 }
320 
321 
322 int ip6_input(struct sk_buff *skb)
323 {
324 	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
325 		       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
326 		       ip6_input_finish);
327 }
328 EXPORT_SYMBOL_GPL(ip6_input);
329 
330 int ip6_mc_input(struct sk_buff *skb)
331 {
332 	const struct ipv6hdr *hdr;
333 	bool deliver;
334 
335 	__IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
336 			 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
337 			 skb->len);
338 
339 	hdr = ipv6_hdr(skb);
340 	deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
341 
342 #ifdef CONFIG_IPV6_MROUTE
343 	/*
344 	 *      IPv6 multicast router mode is now supported ;)
345 	 */
346 	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
347 	    !(ipv6_addr_type(&hdr->daddr) &
348 	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
349 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
350 		/*
351 		 * Okay, we try to forward - split and duplicate
352 		 * packets.
353 		 */
354 		struct sk_buff *skb2;
355 		struct inet6_skb_parm *opt = IP6CB(skb);
356 
357 		/* Check for MLD */
358 		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
359 			/* Check if this is a mld message */
360 			u8 nexthdr = hdr->nexthdr;
361 			__be16 frag_off;
362 			int offset;
363 
364 			/* Check if the value of Router Alert
365 			 * is for MLD (0x0000).
366 			 */
367 			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
368 				deliver = false;
369 
370 				if (!ipv6_ext_hdr(nexthdr)) {
371 					/* BUG */
372 					goto out;
373 				}
374 				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
375 							  &nexthdr, &frag_off);
376 				if (offset < 0)
377 					goto out;
378 
379 				if (ipv6_is_mld(skb, nexthdr, offset))
380 					deliver = true;
381 
382 				goto out;
383 			}
384 			/* unknown RA - process it normally */
385 		}
386 
387 		if (deliver)
388 			skb2 = skb_clone(skb, GFP_ATOMIC);
389 		else {
390 			skb2 = skb;
391 			skb = NULL;
392 		}
393 
394 		if (skb2) {
395 			ip6_mr_input(skb2);
396 		}
397 	}
398 out:
399 #endif
400 	if (likely(deliver))
401 		ip6_input(skb);
402 	else {
403 		/* discard */
404 		kfree_skb(skb);
405 	}
406 
407 	return 0;
408 }
409