ipv6: Fix nexthdr for reinjection
[cascardo/linux.git] / net / ipv6 / ip6_input.c
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
227         /*
228          *      Parse extension headers
229          */
230
231         rcu_read_lock();
232 resubmit:
233         idev = ip6_dst_idev(skb_dst(skb));
234         if (!pskb_pull(skb, skb_transport_offset(skb)))
235                 goto discard;
236         nhoff = IP6CB(skb)->nhoff;
237         nexthdr = skb_network_header(skb)[nhoff];
238
239 resubmit_final:
240         raw = raw6_local_deliver(skb, nexthdr);
241         ipprot = rcu_dereference(inet6_protos[nexthdr]);
242         if (ipprot) {
243                 int ret;
244
245                 if (ipprot->flags & INET6_PROTO_FINAL) {
246                         const struct ipv6hdr *hdr;
247
248                         /* Free reference early: we don't need it any more,
249                            and it may hold ip_conntrack module loaded
250                            indefinitely. */
251                         nf_reset(skb);
252
253                         skb_postpull_rcsum(skb, skb_network_header(skb),
254                                            skb_network_header_len(skb));
255                         hdr = ipv6_hdr(skb);
256                         if (ipv6_addr_is_multicast(&hdr->daddr) &&
257                             !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
258                             &hdr->saddr) &&
259                             !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb)))
260                                 goto discard;
261                 }
262                 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
263                     !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
264                         goto discard;
265
266                 ret = ipprot->handler(skb);
267                 if (ret > 0) {
268                         if (ipprot->flags & INET6_PROTO_FINAL) {
269                                 /* Not an extension header, most likely UDP
270                                  * encapsulation. Use return value as nexthdr
271                                  * protocol not nhoff (which presumably is
272                                  * not set by handler).
273                                  */
274                                 nexthdr = ret;
275                                 goto resubmit_final;
276                         } else {
277                                 goto resubmit;
278                         }
279                 } else if (ret == 0) {
280                         __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
281                 }
282         } else {
283                 if (!raw) {
284                         if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
285                                 __IP6_INC_STATS(net, idev,
286                                                 IPSTATS_MIB_INUNKNOWNPROTOS);
287                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
288                                             ICMPV6_UNK_NEXTHDR, nhoff);
289                         }
290                         kfree_skb(skb);
291                 } else {
292                         __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS);
293                         consume_skb(skb);
294                 }
295         }
296         rcu_read_unlock();
297         return 0;
298
299 discard:
300         __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS);
301         rcu_read_unlock();
302         kfree_skb(skb);
303         return 0;
304 }
305
306
307 int ip6_input(struct sk_buff *skb)
308 {
309         return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
310                        dev_net(skb->dev), NULL, skb, skb->dev, NULL,
311                        ip6_input_finish);
312 }
313
314 int ip6_mc_input(struct sk_buff *skb)
315 {
316         const struct ipv6hdr *hdr;
317         bool deliver;
318
319         __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev),
320                          ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
321                          skb->len);
322
323         hdr = ipv6_hdr(skb);
324         deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
325
326 #ifdef CONFIG_IPV6_MROUTE
327         /*
328          *      IPv6 multicast router mode is now supported ;)
329          */
330         if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
331             !(ipv6_addr_type(&hdr->daddr) &
332               (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) &&
333             likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
334                 /*
335                  * Okay, we try to forward - split and duplicate
336                  * packets.
337                  */
338                 struct sk_buff *skb2;
339                 struct inet6_skb_parm *opt = IP6CB(skb);
340
341                 /* Check for MLD */
342                 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) {
343                         /* Check if this is a mld message */
344                         u8 nexthdr = hdr->nexthdr;
345                         __be16 frag_off;
346                         int offset;
347
348                         /* Check if the value of Router Alert
349                          * is for MLD (0x0000).
350                          */
351                         if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) {
352                                 deliver = false;
353
354                                 if (!ipv6_ext_hdr(nexthdr)) {
355                                         /* BUG */
356                                         goto out;
357                                 }
358                                 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
359                                                           &nexthdr, &frag_off);
360                                 if (offset < 0)
361                                         goto out;
362
363                                 if (ipv6_is_mld(skb, nexthdr, offset))
364                                         deliver = true;
365
366                                 goto out;
367                         }
368                         /* unknown RA - process it normally */
369                 }
370
371                 if (deliver)
372                         skb2 = skb_clone(skb, GFP_ATOMIC);
373                 else {
374                         skb2 = skb;
375                         skb = NULL;
376                 }
377
378                 if (skb2) {
379                         ip6_mr_input(skb2);
380                 }
381         }
382 out:
383 #endif
384         if (likely(deliver))
385                 ip6_input(skb);
386         else {
387                 /* discard */
388                 kfree_skb(skb);
389         }
390
391         return 0;
392 }