ip6_input.c 8.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *	IPv6 input
3
 *	Linux INET6 implementation
L
Linus Torvalds 已提交
4 5 6 7 8 9 10 11 12 13 14 15 16 17
 *
 *	Authors:
 *	Pedro Roque		<roque@di.fc.ul.pt>
 *	Ian P. Morris		<I.P.Morris@soton.ac.uk>
 *
 *	Based in linux/net/ipv4/ip_input.c
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */
/* Changes
 *
18 19
 *	Mitsuru KANDA @USAGI and
 *	YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs().
L
Linus Torvalds 已提交
20 21 22 23 24 25 26 27 28 29
 */

#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/netdevice.h>
#include <linux/in6.h>
#include <linux/icmpv6.h>
30
#include <linux/mroute6.h>
31
#include <linux/slab.h>
L
Linus Torvalds 已提交
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

#include <linux/netfilter.h>
#include <linux/netfilter_ipv6.h>

#include <net/sock.h>
#include <net/snmp.h>

#include <net/ipv6.h>
#include <net/protocol.h>
#include <net/transp_v6.h>
#include <net/rawv6.h>
#include <net/ndisc.h>
#include <net/ip6_route.h>
#include <net/addrconf.h>
#include <net/xfrm.h>
47
#include <net/inet_ecn.h>
L
Linus Torvalds 已提交
48 49


E
Eric Dumazet 已提交
50
int ip6_rcv_finish(struct sk_buff *skb)
L
Linus Torvalds 已提交
51
{
52
	if (sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) {
E
Eric Dumazet 已提交
53 54 55 56 57 58 59
		const struct inet6_protocol *ipprot;

		ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]);
		if (ipprot && ipprot->early_demux)
			ipprot->early_demux(skb);
	}
	if (!skb_dst(skb))
L
Linus Torvalds 已提交
60 61 62 63 64
		ip6_route_input(skb);

	return dst_input(skb);
}

D
David S. Miller 已提交
65
int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
L
Linus Torvalds 已提交
66
{
67
	const struct ipv6hdr *hdr;
68
	u32 pkt_len;
69
	struct inet6_dev *idev;
70
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
71

72 73
	if (skb->pkt_type == PACKET_OTHERHOST) {
		kfree_skb(skb);
74
		return NET_RX_DROP;
75 76 77
	}

	rcu_read_lock();
L
Linus Torvalds 已提交
78

79 80
	idev = __in6_dev_get(skb->dev);

81
	IP6_UPD_PO_STATS_BH(net, idev, IPSTATS_MIB_IN, skb->len);
L
Linus Torvalds 已提交
82

83 84
	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
	    !idev || unlikely(idev->cnf.disable_ipv6)) {
85
		IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
86
		goto drop;
L
Linus Torvalds 已提交
87 88
	}

89 90
	memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));

L
Linus Torvalds 已提交
91 92 93 94 95 96
	/*
	 * Store incoming device index. When the packet will
	 * be queued, we cannot refer to skb->dev anymore.
	 *
	 * BTW, when we send a packet for our own local address on a
	 * non-loopback interface (e.g. ethX), it is being delivered
97
	 * via the loopback interface (lo) here; skb->dev = loopback_dev.
L
Linus Torvalds 已提交
98 99 100 101
	 * It, however, should be considered as if it is being
	 * arrived via the sending interface (ethX), because of the
	 * nature of scoping architecture. --yoshfuji
	 */
E
Eric Dumazet 已提交
102
	IP6CB(skb)->iif = skb_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
L
Linus Torvalds 已提交
103

104
	if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
L
Linus Torvalds 已提交
105 106
		goto err;

107
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
108 109 110 111

	if (hdr->version != 6)
		goto err;

112 113 114 115
	IP6_ADD_STATS_BH(dev_net(dev), idev,
			 IPSTATS_MIB_NOECTPKTS +
				(ipv6_get_dsfield(hdr) & INET_ECN_MASK),
			 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
116 117 118 119 120 121 122 123 124
	/*
	 * RFC4291 2.5.3
	 * A packet received on an interface with a destination address
	 * of loopback must be dropped.
	 */
	if (!(dev->flags & IFF_LOOPBACK) &&
	    ipv6_addr_loopback(&hdr->daddr))
		goto err;

125 126 127 128 129 130 131 132 133 134 135 136
	/* RFC4291 Errata ID: 3480
	 * Interface-Local scope spans only a single interface on a
	 * node and is useful only for loopback transmission of
	 * multicast.  Packets with interface-local scope received
	 * from another node must be discarded.
	 */
	if (!(skb->pkt_type == PACKET_LOOPBACK ||
	      dev->flags & IFF_LOOPBACK) &&
	    ipv6_addr_is_multicast(&hdr->daddr) &&
	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
		goto err;

137 138 139 140 141 142 143 144 145
	/* RFC4291 2.7
	 * Nodes must not originate a packet to a multicast address whose scope
	 * field contains the reserved value 0; if such a packet is received, it
	 * must be silently dropped.
	 */
	if (ipv6_addr_is_multicast(&hdr->daddr) &&
	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
		goto err;

146 147 148 149 150 151 152 153
	/*
	 * RFC4291 2.7
	 * Multicast addresses must not be used as source addresses in IPv6
	 * packets or appear in any Routing header.
	 */
	if (ipv6_addr_is_multicast(&hdr->saddr))
		goto err;

154
	skb->transport_header = skb->network_header + sizeof(*hdr);
155 156
	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);

L
Linus Torvalds 已提交
157 158 159 160
	pkt_len = ntohs(hdr->payload_len);

	/* pkt_len may be zero if Jumbo payload option is present */
	if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
161
		if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
162 163
			IP6_INC_STATS_BH(net,
					 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
164 165
			goto drop;
		}
L
Linus Torvalds 已提交
166
		if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
167
			IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
L
Linus Torvalds 已提交
168 169
			goto drop;
		}
170
		hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
171 172 173
	}

	if (hdr->nexthdr == NEXTHDR_HOP) {
174
		if (ipv6_parse_hopopts(skb) < 0) {
175
			IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
176
			rcu_read_unlock();
177
			return NET_RX_DROP;
L
Linus Torvalds 已提交
178 179 180
		}
	}

181 182
	rcu_read_unlock();

183 184 185
	/* Must drop socket now because of tproxy. */
	skb_orphan(skb);

186
	return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, dev, NULL,
187
		       ip6_rcv_finish);
L
Linus Torvalds 已提交
188
err:
189
	IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
L
Linus Torvalds 已提交
190
drop:
191
	rcu_read_unlock();
L
Linus Torvalds 已提交
192
	kfree_skb(skb);
193
	return NET_RX_DROP;
L
Linus Torvalds 已提交
194 195 196 197 198 199 200
}

/*
 *	Deliver the packet to the host
 */


P
Patrick McHardy 已提交
201
static int ip6_input_finish(struct sk_buff *skb)
L
Linus Torvalds 已提交
202
{
203
	struct net *net = dev_net(skb_dst(skb)->dev);
204
	const struct inet6_protocol *ipprot;
205
	struct inet6_dev *idev;
L
Linus Torvalds 已提交
206
	unsigned int nhoff;
207 208
	int nexthdr;
	bool raw;
L
Linus Torvalds 已提交
209 210 211 212 213 214 215

	/*
	 *	Parse extension headers
	 */

	rcu_read_lock();
resubmit:
E
Eric Dumazet 已提交
216
	idev = ip6_dst_idev(skb_dst(skb));
217
	if (!pskb_pull(skb, skb_transport_offset(skb)))
L
Linus Torvalds 已提交
218
		goto discard;
219
	nhoff = IP6CB(skb)->nhoff;
220
	nexthdr = skb_network_header(skb)[nhoff];
L
Linus Torvalds 已提交
221

222
	raw = raw6_local_deliver(skb, nexthdr);
223
	if ((ipprot = rcu_dereference(inet6_protos[nexthdr])) != NULL) {
L
Linus Torvalds 已提交
224
		int ret;
225

L
Linus Torvalds 已提交
226
		if (ipprot->flags & INET6_PROTO_FINAL) {
227
			const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
228

229 230 231 232 233
			/* Free reference early: we don't need it any more,
			   and it may hold ip_conntrack module loaded
			   indefinitely. */
			nf_reset(skb);

234
			skb_postpull_rcsum(skb, skb_network_header(skb),
235
					   skb_network_header_len(skb));
236
			hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
237 238 239
			if (ipv6_addr_is_multicast(&hdr->daddr) &&
			    !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
			    &hdr->saddr) &&
240
			    !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
L
Linus Torvalds 已提交
241 242 243
				goto discard;
		}
		if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
244
		    !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
L
Linus Torvalds 已提交
245
			goto discard;
246

247
		ret = ipprot->handler(skb);
L
Linus Torvalds 已提交
248 249 250
		if (ret > 0)
			goto resubmit;
		else if (ret == 0)
251
			IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
L
Linus Torvalds 已提交
252
	} else {
253
		if (!raw) {
L
Linus Torvalds 已提交
254
			if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
255 256
				IP6_INC_STATS_BH(net, idev,
						 IPSTATS_MIB_INUNKNOWNPROTOS);
257
				icmpv6_send(skb, ICMPV6_PARAMPROB,
258
					    ICMPV6_UNK_NEXTHDR, nhoff);
L
Linus Torvalds 已提交
259
			}
260 261
			kfree_skb(skb);
		} else {
262
			IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
263 264
			consume_skb(skb);
		}
L
Linus Torvalds 已提交
265 266 267 268 269
	}
	rcu_read_unlock();
	return 0;

discard:
270
	IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
L
Linus Torvalds 已提交
271 272 273 274 275 276 277 278
	rcu_read_unlock();
	kfree_skb(skb);
	return 0;
}


int ip6_input(struct sk_buff *skb)
{
279
	return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
280
		       ip6_input_finish);
L
Linus Torvalds 已提交
281 282 283 284
}

int ip6_mc_input(struct sk_buff *skb)
{
285
	const struct ipv6hdr *hdr;
286
	bool deliver;
L
Linus Torvalds 已提交
287

E
Eric Dumazet 已提交
288 289
	IP6_UPD_PO_STATS_BH(dev_net(skb_dst(skb)->dev),
			 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
290
			 skb->len);
L
Linus Torvalds 已提交
291

292
	hdr = ipv6_hdr(skb);
293
	deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
L
Linus Torvalds 已提交
294

295
#ifdef CONFIG_IPV6_MROUTE
L
Linus Torvalds 已提交
296
	/*
297
	 *      IPv6 multicast router mode is now supported ;)
L
Linus Torvalds 已提交
298
	 */
299
	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
300 301
	    !(ipv6_addr_type(&hdr->daddr) &
	      (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
302 303 304 305 306 307 308 309 310
	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
		/*
		 * Okay, we try to forward - split and duplicate
		 * packets.
		 */
		struct sk_buff *skb2;
		struct inet6_skb_parm *opt = IP6CB(skb);

		/* Check for MLD */
311
		if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
312 313
			/* Check if this is a mld message */
			u8 nexthdr = hdr->nexthdr;
314
			__be16 frag_off;
315 316 317 318 319
			int offset;

			/* Check if the value of Router Alert
			 * is for MLD (0x0000).
			 */
320
			if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
321
				deliver = false;
322

323 324
				if (!ipv6_ext_hdr(nexthdr)) {
					/* BUG */
325
					goto out;
326 327
				}
				offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
328
							  &nexthdr, &frag_off);
329
				if (offset < 0)
330
					goto out;
331

332
				if (!ipv6_is_mld(skb, nexthdr, offset))
333
					goto out;
334

335
				deliver = true;
336 337 338
			}
			/* unknown RA - process it normally */
		}
L
Linus Torvalds 已提交
339

340 341 342 343 344 345
		if (deliver)
			skb2 = skb_clone(skb, GFP_ATOMIC);
		else {
			skb2 = skb;
			skb = NULL;
		}
346

347 348
		if (skb2) {
			ip6_mr_input(skb2);
L
Linus Torvalds 已提交
349 350
		}
	}
351
out:
352 353
#endif
	if (likely(deliver))
L
Linus Torvalds 已提交
354
		ip6_input(skb);
355 356 357
	else {
		/* discard */
		kfree_skb(skb);
L
Linus Torvalds 已提交
358 359 360 361
	}

	return 0;
}