sctp: Do not account for sizeof(struct sk_buff) in estimated rwnd
[linux-2.6.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
48
49
50 inline int ip6_rcv_finish( struct sk_buff *skb)
51 {
52         if (skb_dst(skb) == NULL)
53                 ip6_route_input(skb);
54
55         return dst_input(skb);
56 }
57
58 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
59 {
60         const struct ipv6hdr *hdr;
61         u32             pkt_len;
62         struct inet6_dev *idev;
63         struct net *net = dev_net(skb->dev);
64
65         if (skb->pkt_type == PACKET_OTHERHOST) {
66                 kfree_skb(skb);
67                 return NET_RX_DROP;
68         }
69
70         rcu_read_lock();
71
72         idev = __in6_dev_get(skb->dev);
73
74         IP6_UPD_PO_STATS_BH(net, idev, IPSTATS_MIB_IN, skb->len);
75
76         if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL ||
77             !idev || unlikely(idev->cnf.disable_ipv6)) {
78                 IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
79                 goto drop;
80         }
81
82         memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
83
84         /*
85          * Store incoming device index. When the packet will
86          * be queued, we cannot refer to skb->dev anymore.
87          *
88          * BTW, when we send a packet for our own local address on a
89          * non-loopback interface (e.g. ethX), it is being delivered
90          * via the loopback interface (lo) here; skb->dev = loopback_dev.
91          * It, however, should be considered as if it is being
92          * arrived via the sending interface (ethX), because of the
93          * nature of scoping architecture. --yoshfuji
94          */
95         IP6CB(skb)->iif = skb_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex;
96
97         if (unlikely(!pskb_may_pull(skb, sizeof(*hdr))))
98                 goto err;
99
100         hdr = ipv6_hdr(skb);
101
102         if (hdr->version != 6)
103                 goto err;
104
105         /*
106          * RFC4291 2.5.3
107          * A packet received on an interface with a destination address
108          * of loopback must be dropped.
109          */
110         if (!(dev->flags & IFF_LOOPBACK) &&
111             ipv6_addr_loopback(&hdr->daddr))
112                 goto err;
113
114         skb->transport_header = skb->network_header + sizeof(*hdr);
115         IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
116
117         pkt_len = ntohs(hdr->payload_len);
118
119         /* pkt_len may be zero if Jumbo payload option is present */
120         if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
121                 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) {
122                         IP6_INC_STATS_BH(net,
123                                          idev, IPSTATS_MIB_INTRUNCATEDPKTS);
124                         goto drop;
125                 }
126                 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) {
127                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
128                         goto drop;
129                 }
130                 hdr = ipv6_hdr(skb);
131         }
132
133         if (hdr->nexthdr == NEXTHDR_HOP) {
134                 if (ipv6_parse_hopopts(skb) < 0) {
135                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
136                         rcu_read_unlock();
137                         return NET_RX_DROP;
138                 }
139         }
140
141         rcu_read_unlock();
142
143         /* Must drop socket now because of tproxy. */
144         skb_orphan(skb);
145
146         return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, dev, NULL,
147                        ip6_rcv_finish);
148 err:
149         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INHDRERRORS);
150 drop:
151         rcu_read_unlock();
152         kfree_skb(skb);
153         return NET_RX_DROP;
154 }
155
156 /*
157  *      Deliver the packet to the host
158  */
159
160
161 static int ip6_input_finish(struct sk_buff *skb)
162 {
163         const struct inet6_protocol *ipprot;
164         unsigned int nhoff;
165         int nexthdr, raw;
166         u8 hash;
167         struct inet6_dev *idev;
168         struct net *net = dev_net(skb_dst(skb)->dev);
169
170         /*
171          *      Parse extension headers
172          */
173
174         rcu_read_lock();
175 resubmit:
176         idev = ip6_dst_idev(skb_dst(skb));
177         if (!pskb_pull(skb, skb_transport_offset(skb)))
178                 goto discard;
179         nhoff = IP6CB(skb)->nhoff;
180         nexthdr = skb_network_header(skb)[nhoff];
181
182         raw = raw6_local_deliver(skb, nexthdr);
183
184         hash = nexthdr & (MAX_INET_PROTOS - 1);
185         if ((ipprot = rcu_dereference(inet6_protos[hash])) != NULL) {
186                 int ret;
187
188                 if (ipprot->flags & INET6_PROTO_FINAL) {
189                         const struct ipv6hdr *hdr;
190
191                         /* Free reference early: we don't need it any more,
192                            and it may hold ip_conntrack module loaded
193                            indefinitely. */
194                         nf_reset(skb);
195
196                         skb_postpull_rcsum(skb, skb_network_header(skb),
197                                            skb_network_header_len(skb));
198                         hdr = ipv6_hdr(skb);
199                         if (ipv6_addr_is_multicast(&hdr->daddr) &&
200                             !ipv6_chk_mcast_addr(skb->dev, &hdr->daddr,
201                             &hdr->saddr) &&
202                             !ipv6_is_mld(skb, nexthdr))
203                                 goto discard;
204                 }
205                 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) &&
206                     !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
207                         goto discard;
208
209                 ret = ipprot->handler(skb);
210                 if (ret > 0)
211                         goto resubmit;
212                 else if (ret == 0)
213                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
214         } else {
215                 if (!raw) {
216                         if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
217                                 IP6_INC_STATS_BH(net, idev,
218                                                  IPSTATS_MIB_INUNKNOWNPROTOS);
219                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
220                                             ICMPV6_UNK_NEXTHDR, nhoff);
221                         }
222                 } else
223                         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDELIVERS);
224                 kfree_skb(skb);
225         }
226         rcu_read_unlock();
227         return 0;
228
229 discard:
230         IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_INDISCARDS);
231         rcu_read_unlock();
232         kfree_skb(skb);
233         return 0;
234 }
235
236
237 int ip6_input(struct sk_buff *skb)
238 {
239         return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
240                        ip6_input_finish);
241 }
242
243 int ip6_mc_input(struct sk_buff *skb)
244 {
245         const struct ipv6hdr *hdr;
246         int deliver;
247
248         IP6_UPD_PO_STATS_BH(dev_net(skb_dst(skb)->dev),
249                          ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INMCAST,
250                          skb->len);
251
252         hdr = ipv6_hdr(skb);
253         deliver = ipv6_chk_mcast_addr(skb->dev, &hdr->daddr, NULL);
254
255 #ifdef CONFIG_IPV6_MROUTE
256         /*
257          *      IPv6 multicast router mode is now supported ;)
258          */
259         if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
260             !(ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) &&
261             likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
262                 /*
263                  * Okay, we try to forward - split and duplicate
264                  * packets.
265                  */
266                 struct sk_buff *skb2;
267                 struct inet6_skb_parm *opt = IP6CB(skb);
268
269                 /* Check for MLD */
270                 if (unlikely(opt->ra)) {
271                         /* Check if this is a mld message */
272                         u8 *ptr = skb_network_header(skb) + opt->ra;
273                         struct icmp6hdr *icmp6;
274                         u8 nexthdr = hdr->nexthdr;
275                         int offset;
276
277                         /* Check if the value of Router Alert
278                          * is for MLD (0x0000).
279                          */
280                         if ((ptr[2] | ptr[3]) == 0) {
281                                 deliver = 0;
282
283                                 if (!ipv6_ext_hdr(nexthdr)) {
284                                         /* BUG */
285                                         goto out;
286                                 }
287                                 offset = ipv6_skip_exthdr(skb, sizeof(*hdr),
288                                                           &nexthdr);
289                                 if (offset < 0)
290                                         goto out;
291
292                                 if (nexthdr != IPPROTO_ICMPV6)
293                                         goto out;
294
295                                 if (!pskb_may_pull(skb, (skb_network_header(skb) +
296                                                    offset + 1 - skb->data)))
297                                         goto out;
298
299                                 icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
300
301                                 switch (icmp6->icmp6_type) {
302                                 case ICMPV6_MGM_QUERY:
303                                 case ICMPV6_MGM_REPORT:
304                                 case ICMPV6_MGM_REDUCTION:
305                                 case ICMPV6_MLD2_REPORT:
306                                         deliver = 1;
307                                         break;
308                                 }
309                                 goto out;
310                         }
311                         /* unknown RA - process it normally */
312                 }
313
314                 if (deliver)
315                         skb2 = skb_clone(skb, GFP_ATOMIC);
316                 else {
317                         skb2 = skb;
318                         skb = NULL;
319                 }
320
321                 if (skb2) {
322                         ip6_mr_input(skb2);
323                 }
324         }
325 out:
326 #endif
327         if (likely(deliver))
328                 ip6_input(skb);
329         else {
330                 /* discard */
331                 kfree_skb(skb);
332         }
333
334         return 0;
335 }