blob: 1b6475c05eb4045129b61da28972203e4a3780b3 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The Internet Protocol (IP) module.
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Donald Becker, <becker@super.org>
Alan Cox113aa832008-10-13 19:01:08 -070011 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Richard Underwood
13 * Stefan Becker, <stefanb@yello.ping.de>
14 * Jorge Cwik, <jorge@laser.satlink.net>
15 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090016 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 *
18 * Fixes:
19 * Alan Cox : Commented a couple of minor bits of surplus code
20 * Alan Cox : Undefining IP_FORWARD doesn't include the code
21 * (just stops a compiler warning).
22 * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes
23 * are junked rather than corrupting things.
24 * Alan Cox : Frames to bad broadcast subnets are dumped
25 * We used to process them non broadcast and
26 * boy could that cause havoc.
27 * Alan Cox : ip_forward sets the free flag on the
28 * new frame it queues. Still crap because
29 * it copies the frame but at least it
30 * doesn't eat memory too.
31 * Alan Cox : Generic queue code and memory fixes.
32 * Fred Van Kempen : IP fragment support (borrowed from NET2E)
33 * Gerhard Koerting: Forward fragmented frames correctly.
34 * Gerhard Koerting: Fixes to my fix of the above 8-).
35 * Gerhard Koerting: IP interface addressing fix.
36 * Linus Torvalds : More robustness checks
37 * Alan Cox : Even more checks: Still not as robust as it ought to be
38 * Alan Cox : Save IP header pointer for later
39 * Alan Cox : ip option setting
40 * Alan Cox : Use ip_tos/ip_ttl settings
41 * Alan Cox : Fragmentation bogosity removed
42 * (Thanks to Mark.Bush@prg.ox.ac.uk)
43 * Dmitry Gorodchanin : Send of a raw packet crash fix.
44 * Alan Cox : Silly ip bug when an overlength
45 * fragment turns up. Now frees the
46 * queue.
47 * Linus Torvalds/ : Memory leakage on fragmentation
48 * Alan Cox : handling.
49 * Gerhard Koerting: Forwarding uses IP priority hints
50 * Teemu Rantanen : Fragment problems.
51 * Alan Cox : General cleanup, comments and reformat
52 * Alan Cox : SNMP statistics
53 * Alan Cox : BSD address rule semantics. Also see
54 * UDP as there is a nasty checksum issue
55 * if you do things the wrong way.
56 * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file
57 * Alan Cox : IP options adjust sk->priority.
58 * Pedro Roque : Fix mtu/length error in ip_forward.
59 * Alan Cox : Avoid ip_chk_addr when possible.
60 * Richard Underwood : IP multicasting.
61 * Alan Cox : Cleaned up multicast handlers.
62 * Alan Cox : RAW sockets demultiplex in the BSD style.
63 * Gunther Mayer : Fix the SNMP reporting typo
64 * Alan Cox : Always in group 224.0.0.1
65 * Pauline Middelink : Fast ip_checksum update when forwarding
66 * Masquerading support.
67 * Alan Cox : Multicast loopback error for 224.0.0.1
68 * Alan Cox : IP_MULTICAST_LOOP option.
69 * Alan Cox : Use notifiers.
70 * Bjorn Ekwall : Removed ip_csum (from slhc.c too)
71 * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!)
72 * Stefan Becker : Send out ICMP HOST REDIRECT
73 * Arnt Gulbrandsen : ip_build_xmit
74 * Alan Cox : Per socket routing cache
75 * Alan Cox : Fixed routing cache, added header cache.
76 * Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it.
77 * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net.
78 * Alan Cox : Incoming IP option handling.
79 * Alan Cox : Set saddr on raw output frames as per BSD.
80 * Alan Cox : Stopped broadcast source route explosions.
81 * Alan Cox : Can disable source routing
82 * Takeshi Sone : Masquerading didn't work.
83 * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible.
84 * Alan Cox : Memory leaks, tramples, misc debugging.
85 * Alan Cox : Fixed multicast (by popular demand 8))
86 * Alan Cox : Fixed forwarding (by even more popular demand 8))
87 * Alan Cox : Fixed SNMP statistics [I think]
88 * Gerhard Koerting : IP fragmentation forwarding fix
89 * Alan Cox : Device lock against page fault.
90 * Alan Cox : IP_HDRINCL facility.
91 * Werner Almesberger : Zero fragment bug
92 * Alan Cox : RAW IP frame length bug
93 * Alan Cox : Outgoing firewall on build_xmit
94 * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel
95 * Alan Cox : Multicast routing hooks
96 * Jos Vos : Do accounting *before* call_in_firewall
97 * Willy Konynenberg : Transparent proxying support
98 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090099 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 *
101 * To Fix:
102 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
103 * and could be made very efficient with the addition of some virtual memory hacks to permit
104 * the allocation of a buffer that can then be 'grown' by twiddling page tables.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900105 * Output fragmentation wants updating along with the buffer management to use a single
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
107 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
108 * fragmentation anyway.
109 *
110 * This program is free software; you can redistribute it and/or
111 * modify it under the terms of the GNU General Public License
112 * as published by the Free Software Foundation; either version
113 * 2 of the License, or (at your option) any later version.
114 */
115
116#include <asm/system.h>
117#include <linux/module.h>
118#include <linux/types.h>
119#include <linux/kernel.h>
120#include <linux/string.h>
121#include <linux/errno.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123#include <linux/net.h>
124#include <linux/socket.h>
125#include <linux/sockios.h>
126#include <linux/in.h>
127#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -0200128#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129#include <linux/netdevice.h>
130#include <linux/etherdevice.h>
131
132#include <net/snmp.h>
133#include <net/ip.h>
134#include <net/protocol.h>
135#include <net/route.h>
136#include <linux/skbuff.h>
137#include <net/sock.h>
138#include <net/arp.h>
139#include <net/icmp.h>
140#include <net/raw.h>
141#include <net/checksum.h>
142#include <linux/netfilter_ipv4.h>
143#include <net/xfrm.h>
144#include <linux/mroute.h>
145#include <linux/netlink.h>
146
147/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 * Process Router Attention IP option
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900149 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150int ip_call_ra_chain(struct sk_buff *skb)
151{
152 struct ip_ra_chain *ra;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700153 u8 protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 struct sock *last = NULL;
Denis V. Lunevcb846632008-03-24 15:31:00 -0700155 struct net_device *dev = skb->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
157 read_lock(&ip_ra_lock);
158 for (ra = ip_ra_chain; ra; ra = ra->next) {
159 struct sock *sk = ra->sk;
160
161 /* If socket is bound to an interface, only report
162 * the packet if it came from that interface.
163 */
164 if (sk && inet_sk(sk)->num == protocol &&
165 (!sk->sk_bound_dev_if ||
Denis V. Lunevcb846632008-03-24 15:31:00 -0700166 sk->sk_bound_dev_if == dev->ifindex) &&
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900167 sock_net(sk) == dev_net(dev)) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700168 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700169 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 read_unlock(&ip_ra_lock);
171 return 1;
172 }
173 }
174 if (last) {
175 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
176 if (skb2)
177 raw_rcv(last, skb2);
178 }
179 last = sk;
180 }
181 }
182
183 if (last) {
184 raw_rcv(last, skb);
185 read_unlock(&ip_ra_lock);
186 return 1;
187 }
188 read_unlock(&ip_ra_lock);
189 return 0;
190}
191
Patrick McHardy861d0482007-10-15 01:48:39 -0700192static int ip_local_deliver_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900194 struct net *net = dev_net(skb->dev);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700195
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300196 __skb_pull(skb, ip_hdrlen(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900198 /* Point into the IP datagram, just past the header. */
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300199 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
201 rcu_read_lock();
202 {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700203 int protocol = ip_hdr(skb)->protocol;
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800204 int hash, raw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 struct net_protocol *ipprot;
206
207 resubmit:
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800208 raw = raw_local_deliver(skb, protocol);
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 hash = protocol & (MAX_INET_PROTOS - 1);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700211 ipprot = rcu_dereference(inet_protos[hash]);
212 if (ipprot != NULL && (net == &init_net || ipprot->netns_ok)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 int ret;
214
Patrick McHardyb59c2702006-01-06 23:06:10 -0800215 if (!ipprot->no_policy) {
216 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
217 kfree_skb(skb);
218 goto out;
219 }
220 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 }
222 ret = ipprot->handler(skb);
223 if (ret < 0) {
224 protocol = -ret;
225 goto resubmit;
226 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700227 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800229 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700231 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 icmp_send(skb, ICMP_DEST_UNREACH,
233 ICMP_PROT_UNREACH, 0);
234 }
235 } else
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700236 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 kfree_skb(skb);
238 }
239 }
240 out:
241 rcu_read_unlock();
242
243 return 0;
244}
245
246/*
247 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900248 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249int ip_local_deliver(struct sk_buff *skb)
250{
251 /*
252 * Reassemble IP fragments.
253 */
254
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700255 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700256 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 return 0;
258 }
259
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800260 return NF_HOOK(PF_INET, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 ip_local_deliver_finish);
262}
263
Thomas Grafd2454072005-08-20 17:26:12 -0700264static inline int ip_rcv_options(struct sk_buff *skb)
265{
266 struct ip_options *opt;
267 struct iphdr *iph;
268 struct net_device *dev = skb->dev;
269
270 /* It looks as overkill, because not all
271 IP options require packet mangling.
272 But it is the easiest for now, especially taking
273 into account that combination of IP options
274 and running sniffer is extremely rare condition.
275 --ANK (980813)
276 */
277 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700278 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700279 goto drop;
280 }
281
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700282 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700283 opt = &(IPCB(skb)->opt);
284 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700285
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900286 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700287 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700288 goto drop;
289 }
290
Thomas Grafd2454072005-08-20 17:26:12 -0700291 if (unlikely(opt->srr)) {
292 struct in_device *in_dev = in_dev_get(dev);
293 if (in_dev) {
294 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
295 if (IN_DEV_LOG_MARTIANS(in_dev) &&
296 net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700297 printk(KERN_INFO "source route option %pI4 -> %pI4\n",
298 &iph->saddr, &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700299 in_dev_put(in_dev);
300 goto drop;
301 }
302
303 in_dev_put(in_dev);
304 }
305
306 if (ip_options_rcv_srr(skb))
307 goto drop;
308 }
309
310 return 0;
311drop:
312 return -1;
313}
314
Patrick McHardy861d0482007-10-15 01:48:39 -0700315static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700317 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700318 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
320 /*
321 * Initialise the virtual path cache for the packet. It describes
322 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900323 */
Hua Zhong0182bd22006-05-06 18:11:39 -0700324 if (skb->dst == NULL) {
Thomas Graf3e192be2005-08-20 17:26:30 -0700325 int err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos,
326 skb->dev);
327 if (unlikely(err)) {
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700328 if (err == -EHOSTUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700329 IP_INC_STATS_BH(dev_net(skb->dev),
330 IPSTATS_MIB_INADDRERRORS);
Mitsuru Chinene91a47e2007-04-30 00:45:49 -0700331 else if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700332 IP_INC_STATS_BH(dev_net(skb->dev),
333 IPSTATS_MIB_INNOROUTES);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900334 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700335 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 }
337
338#ifdef CONFIG_NET_CLS_ROUTE
Thomas Graf3e192be2005-08-20 17:26:30 -0700339 if (unlikely(skb->dst->tclassid)) {
Eric Dumazet8dbde282007-11-16 03:32:10 -0800340 struct ip_rt_acct *st = per_cpu_ptr(ip_rt_acct, smp_processor_id());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 u32 idx = skb->dst->tclassid;
342 st[idx&0xFF].o_packets++;
343 st[idx&0xFF].o_bytes+=skb->len;
344 st[(idx>>16)&0xFF].i_packets++;
345 st[(idx>>16)&0xFF].i_bytes+=skb->len;
346 }
347#endif
348
Thomas Grafd2454072005-08-20 17:26:12 -0700349 if (iph->ihl > 5 && ip_rcv_options(skb))
350 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
Eric Dumazetee6b9672008-03-05 18:30:47 -0800352 rt = skb->rtable;
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700353 if (rt->rt_type == RTN_MULTICAST)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700354 IP_INC_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INMCASTPKTS);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700355 else if (rt->rt_type == RTN_BROADCAST)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700356 IP_INC_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INBCASTPKTS);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700357
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 return dst_input(skb);
359
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900361 kfree_skb(skb);
362 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363}
364
365/*
366 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900367 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700368int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369{
370 struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700371 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373 /* When the interface is in promisc. mode, drop all the crap
374 * that it receives, do not try to analyse it.
375 */
376 if (skb->pkt_type == PACKET_OTHERHOST)
377 goto drop;
378
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700379 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INRECEIVES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
381 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700382 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 goto out;
384 }
385
386 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
387 goto inhdr_error;
388
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700389 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390
391 /*
J.H.M. Dassen (Ray)c67fa022008-05-08 01:11:04 -0700392 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 *
394 * Is the datagram acceptable?
395 *
396 * 1. Length at least the size of an ip header
397 * 2. Version of 4
398 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
399 * 4. Doesn't have a bogus length
400 */
401
402 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700403 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405 if (!pskb_may_pull(skb, iph->ihl*4))
406 goto inhdr_error;
407
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700408 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
Thomas Grafe9c60422005-08-20 17:25:52 -0700410 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Thomas Graf58615242005-08-20 17:25:29 -0700411 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
Thomas Graf58615242005-08-20 17:25:29 -0700413 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700414 if (skb->len < len) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700415 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700416 goto drop;
417 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700418 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
Thomas Graf58615242005-08-20 17:25:29 -0700420 /* Our transport medium may have padded the buffer out. Now we know it
421 * is IP we can trim to the true length of the frame.
422 * Note this now means skb->len holds ntohs(iph->tot_len).
423 */
424 if (pskb_trim_rcsum(skb, len)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700425 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Graf58615242005-08-20 17:25:29 -0700426 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 }
428
Stephen Hemminger53602f92006-07-14 14:49:32 -0700429 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700430 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700431
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800432 return NF_HOOK(PF_INET, NF_INET_PRE_ROUTING, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 ip_rcv_finish);
434
435inhdr_error:
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700436 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900438 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900440 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441}