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
ip6_rcv_finish(struct net * net,struct sock * sk,struct sk_buff * skb)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
ipv6_rcv(struct sk_buff * skb,struct net_device * dev,struct packet_type * pt,struct net_device * orig_dev)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 * The loopback address must not be used as the source address in IPv6
128 * packets that are sent outside of a single node. [..]
129 * A packet received on an interface with a destination address
130 * of loopback must be dropped.
131 */
132 if ((ipv6_addr_loopback(&hdr->saddr) ||
133 ipv6_addr_loopback(&hdr->daddr)) &&
134 !(dev->flags & IFF_LOOPBACK))
135 goto err;
136
137 /* RFC4291 Errata ID: 3480
138 * Interface-Local scope spans only a single interface on a
139 * node and is useful only for loopback transmission of
140 * multicast. Packets with interface-local scope received
141 * from another node must be discarded.
142 */
143 if (!(skb->pkt_type == PACKET_LOOPBACK ||
144 dev->flags & IFF_LOOPBACK) &&
145 ipv6_addr_is_multicast(&hdr->daddr) &&
146 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1)
147 goto err;
148
149 /* If enabled, drop unicast packets that were encapsulated in link-layer
150 * multicast or broadcast to protected against the so-called "hole-196"
151 * attack in 802.11 wireless.
152 */
153 if (!ipv6_addr_is_multicast(&hdr->daddr) &&
154 (skb->pkt_type == PACKET_BROADCAST ||
155 skb->pkt_type == PACKET_MULTICAST) &&
156 idev->cnf.drop_unicast_in_l2_multicast)
157 goto err;
158
159 /* RFC4291 2.7
160 * Nodes must not originate a packet to a multicast address whose scope
161 * field contains the reserved value 0; if such a packet is received, it
162 * must be silently dropped.
163 */
164 if (ipv6_addr_is_multicast(&hdr->daddr) &&
165 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0)
166 goto err;
167
168 /*
169 * RFC4291 2.7
170 * Multicast addresses must not be used as source addresses in IPv6
171 * packets or appear in any Routing header.
172 */
173 if (ipv6_addr_is_multicast(&hdr->saddr))
174 goto err;
175
176 /* While RFC4291 is not explicit about v4mapped addresses
177 * in IPv6 headers, it seems clear linux dual-stack
178 * model can not deal properly with these.
179 * Security models could be fooled by ::ffff:127.0.0.1 for example.
180 *
181 * https://tools.ietf.org/html/draft-itojun-v6ops-v4mapped-harmful-02
182 */
183 if (ipv6_addr_v4mapped(&hdr->saddr))
184 goto err;
185
186 skb->transport_header = skb->network_header + sizeof(*hdr);
187 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
188
189 pkt_len = ntohs(hdr->payload_len);
190
191 /* pkt_len may be zero if Jumbo payload option is present */
192 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
193 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
194 __IP6_INC_STATS(net,
195 idev, IPSTATS_MIB_INTRUNCATEDPKTS);
196 goto drop;
197 }
198 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
199 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
200 goto drop;
201 }
202 hdr = ipv6_hdr(skb);
203 }
204
205 if (hdr->nexthdr == NEXTHDR_HOP) {
206 if (ipv6_parse_hopopts(skb) < 0) {
207 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
208 rcu_read_unlock();
209 return NET_RX_DROP;
210 }
211 }
212
213 rcu_read_unlock();
214
215 /* Must drop socket now because of tproxy. */
216 skb_orphan(skb);
217
218 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
219 net, NULL, skb, dev, NULL,
220 ip6_rcv_finish);
221 err:
222 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS);
223 drop:
224 rcu_read_unlock();
225 kfree_skb(skb);
226 return NET_RX_DROP;
227 }
228
229 /*
230 * Deliver the packet to the host
231 */
232
233
ip6_input_finish(struct net * net,struct sock * sk,struct sk_buff * skb)234 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
235 {
236 const struct inet6_protocol *ipprot;
237 struct inet6_dev *idev;
238 unsigned int nhoff;
239 int nexthdr;
240 bool raw;
241 bool have_final = false;
242
243 /*
244 * Parse extension headers
245 */
246
247 rcu_read_lock();
248 resubmit:
249 idev = ip6_dst_idev(skb_dst(skb));
250 if (!pskb_pull(skb, skb_transport_offset(skb)))
251 goto discard;
252 nhoff = IP6CB(skb)->nhoff;
253 nexthdr = skb_network_header(skb)[nhoff];
254
255 resubmit_final:
256 raw = raw6_local_deliver(skb, nexthdr);
257 ipprot = rcu_dereference(inet6_protos[nexthdr]);
258 if (ipprot) {
259 int ret;
260
261 if (have_final) {
262 if (!(ipprot->flags & INET6_PROTO_FINAL)) {
263 /* Once we've seen a final protocol don't
264 * allow encapsulation on any non-final
265 * ones. This allows foo in UDP encapsulation
266 * to work.
267 */
268 goto discard;
269 }
270 } else if (ipprot->flags & INET6_PROTO_FINAL) {
271 const struct ipv6hdr *hdr;
272
273 /* Only do this once for first final protocol */
274 have_final = true;
275
276 /* Free reference early: we don't need it any more,
277 and it may hold ip_conntrack module loaded
278 indefinitely. */
279 nf_reset(skb);
280
281 skb_postpull_rcsum(skb, skb_network_header(skb),
282 skb_network_header_len(skb));
283 hdr = ipv6_hdr(skb);
284 if (ipv6_addr_is_multicast(&hdr->daddr) &&
285 !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
286 &hdr->saddr) &&
287 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
288 goto discard;
289 }
290 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
291 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
292 goto discard;
293
294 ret = ipprot->handler(skb);
295 if (ret > 0) {
296 if (ipprot->flags & INET6_PROTO_FINAL) {
297 /* Not an extension header, most likely UDP
298 * encapsulation. Use return value as nexthdr
299 * protocol not nhoff (which presumably is
300 * not set by handler).
301 */
302 nexthdr = ret;
303 goto resubmit_final;
304 } else {
305 goto resubmit;
306 }
307 } else if (ret == 0) {
308 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
309 }
310 } else {
311 if (!raw) {
312 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
313 __IP6_INC_STATS(net, idev,
314 IPSTATS_MIB_INUNKNOWNPROTOS);
315 icmpv6_send(skb, ICMPV6_PARAMPROB,
316 ICMPV6_UNK_NEXTHDR, nhoff);
317 }
318 kfree_skb(skb);
319 } else {
320 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
321 consume_skb(skb);
322 }
323 }
324 rcu_read_unlock();
325 return 0;
326
327 discard:
328 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
329 rcu_read_unlock();
330 kfree_skb(skb);
331 return 0;
332 }
333
334
ip6_input(struct sk_buff * skb)335 int ip6_input(struct sk_buff *skb)
336 {
337 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
338 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
339 ip6_input_finish);
340 }
341 EXPORT_SYMBOL_GPL(ip6_input);
342
ip6_mc_input(struct sk_buff * skb)343 int ip6_mc_input(struct sk_buff *skb)
344 {
345 const struct ipv6hdr *hdr;
346 bool deliver;
347
348 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
349 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
350 skb->len);
351
352 hdr = ipv6_hdr(skb);
353 deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
354
355 #ifdef CONFIG_IPV6_MROUTE
356 /*
357 * IPv6 multicast router mode is now supported ;)
358 */
359 if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
360 !(ipv6_addr_type(&hdr->daddr) &
361 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
362 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
363 /*
364 * Okay, we try to forward - split and duplicate
365 * packets.
366 */
367 struct sk_buff *skb2;
368 struct inet6_skb_parm *opt = IP6CB(skb);
369
370 /* Check for MLD */
371 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
372 /* Check if this is a mld message */
373 u8 nexthdr = hdr->nexthdr;
374 __be16 frag_off;
375 int offset;
376
377 /* Check if the value of Router Alert
378 * is for MLD (0x0000).
379 */
380 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
381 deliver = false;
382
383 if (!ipv6_ext_hdr(nexthdr)) {
384 /* BUG */
385 goto out;
386 }
387 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
388 &nexthdr, &frag_off);
389 if (offset < 0)
390 goto out;
391
392 if (ipv6_is_mld(skb, nexthdr, offset))
393 deliver = true;
394
395 goto out;
396 }
397 /* unknown RA - process it normally */
398 }
399
400 if (deliver)
401 skb2 = skb_clone(skb, GFP_ATOMIC);
402 else {
403 skb2 = skb;
404 skb = NULL;
405 }
406
407 if (skb2) {
408 ip6_mr_input(skb2);
409 }
410 }
411 out:
412 #endif
413 if (likely(deliver))
414 ip6_input(skb);
415 else {
416 /* discard */
417 kfree_skb(skb);
418 }
419
420 return 0;
421 }
422