blob: 1a58a6fa1dc01d1c9eefd48dbb56c5017cec0689 [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]);
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800212 if (ipprot != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 int ret;
214
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800215 if (!net_eq(net, &init_net) && !ipprot->netns_ok) {
216 if (net_ratelimit())
217 printk("%s: proto %d isn't netns-ready\n",
218 __func__, protocol);
219 kfree_skb(skb);
220 goto out;
221 }
222
Patrick McHardyb59c2702006-01-06 23:06:10 -0800223 if (!ipprot->no_policy) {
224 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
225 kfree_skb(skb);
226 goto out;
227 }
228 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 }
230 ret = ipprot->handler(skb);
231 if (ret < 0) {
232 protocol = -ret;
233 goto resubmit;
234 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700235 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800237 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700239 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 icmp_send(skb, ICMP_DEST_UNREACH,
241 ICMP_PROT_UNREACH, 0);
242 }
243 } else
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700244 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 kfree_skb(skb);
246 }
247 }
248 out:
249 rcu_read_unlock();
250
251 return 0;
252}
253
254/*
255 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900256 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257int ip_local_deliver(struct sk_buff *skb)
258{
259 /*
260 * Reassemble IP fragments.
261 */
262
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700263 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700264 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 return 0;
266 }
267
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800268 return NF_HOOK(PF_INET, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 ip_local_deliver_finish);
270}
271
Thomas Grafd2454072005-08-20 17:26:12 -0700272static inline int ip_rcv_options(struct sk_buff *skb)
273{
274 struct ip_options *opt;
275 struct iphdr *iph;
276 struct net_device *dev = skb->dev;
277
278 /* It looks as overkill, because not all
279 IP options require packet mangling.
280 But it is the easiest for now, especially taking
281 into account that combination of IP options
282 and running sniffer is extremely rare condition.
283 --ANK (980813)
284 */
285 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700286 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700287 goto drop;
288 }
289
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700290 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700291 opt = &(IPCB(skb)->opt);
292 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700293
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900294 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700295 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700296 goto drop;
297 }
298
Thomas Grafd2454072005-08-20 17:26:12 -0700299 if (unlikely(opt->srr)) {
300 struct in_device *in_dev = in_dev_get(dev);
301 if (in_dev) {
302 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
303 if (IN_DEV_LOG_MARTIANS(in_dev) &&
304 net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700305 printk(KERN_INFO "source route option %pI4 -> %pI4\n",
306 &iph->saddr, &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700307 in_dev_put(in_dev);
308 goto drop;
309 }
310
311 in_dev_put(in_dev);
312 }
313
314 if (ip_options_rcv_srr(skb))
315 goto drop;
316 }
317
318 return 0;
319drop:
320 return -1;
321}
322
Patrick McHardy861d0482007-10-15 01:48:39 -0700323static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700325 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700326 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327
328 /*
329 * Initialise the virtual path cache for the packet. It describes
330 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900331 */
Hua Zhong0182bd22006-05-06 18:11:39 -0700332 if (skb->dst == NULL) {
Thomas Graf3e192be2005-08-20 17:26:30 -0700333 int err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos,
334 skb->dev);
335 if (unlikely(err)) {
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700336 if (err == -EHOSTUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700337 IP_INC_STATS_BH(dev_net(skb->dev),
338 IPSTATS_MIB_INADDRERRORS);
Mitsuru Chinene91a47e2007-04-30 00:45:49 -0700339 else if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700340 IP_INC_STATS_BH(dev_net(skb->dev),
341 IPSTATS_MIB_INNOROUTES);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900342 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 }
345
346#ifdef CONFIG_NET_CLS_ROUTE
Thomas Graf3e192be2005-08-20 17:26:30 -0700347 if (unlikely(skb->dst->tclassid)) {
Eric Dumazet8dbde282007-11-16 03:32:10 -0800348 struct ip_rt_acct *st = per_cpu_ptr(ip_rt_acct, smp_processor_id());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 u32 idx = skb->dst->tclassid;
350 st[idx&0xFF].o_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800351 st[idx&0xFF].o_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 st[(idx>>16)&0xFF].i_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800353 st[(idx>>16)&0xFF].i_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 }
355#endif
356
Thomas Grafd2454072005-08-20 17:26:12 -0700357 if (iph->ihl > 5 && ip_rcv_options(skb))
358 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
Eric Dumazetee6b9672008-03-05 18:30:47 -0800360 rt = skb->rtable;
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700361 if (rt->rt_type == RTN_MULTICAST)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700362 IP_INC_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INMCASTPKTS);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700363 else if (rt->rt_type == RTN_BROADCAST)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700364 IP_INC_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INBCASTPKTS);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700365
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 return dst_input(skb);
367
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900369 kfree_skb(skb);
370 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371}
372
373/*
374 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900375 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700376int 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 -0700377{
378 struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700379 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
381 /* When the interface is in promisc. mode, drop all the crap
382 * that it receives, do not try to analyse it.
383 */
384 if (skb->pkt_type == PACKET_OTHERHOST)
385 goto drop;
386
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700387 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INRECEIVES);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700390 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 goto out;
392 }
393
394 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
395 goto inhdr_error;
396
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700397 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398
399 /*
J.H.M. Dassen (Ray)c67fa022008-05-08 01:11:04 -0700400 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 *
402 * Is the datagram acceptable?
403 *
404 * 1. Length at least the size of an ip header
405 * 2. Version of 4
406 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
407 * 4. Doesn't have a bogus length
408 */
409
410 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700411 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
413 if (!pskb_may_pull(skb, iph->ihl*4))
414 goto inhdr_error;
415
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700416 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
Thomas Grafe9c60422005-08-20 17:25:52 -0700418 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Thomas Graf58615242005-08-20 17:25:29 -0700419 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Thomas Graf58615242005-08-20 17:25:29 -0700421 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700422 if (skb->len < len) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700423 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700424 goto drop;
425 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700426 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
Thomas Graf58615242005-08-20 17:25:29 -0700428 /* Our transport medium may have padded the buffer out. Now we know it
429 * is IP we can trim to the true length of the frame.
430 * Note this now means skb->len holds ntohs(iph->tot_len).
431 */
432 if (pskb_trim_rcsum(skb, len)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700433 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Graf58615242005-08-20 17:25:29 -0700434 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 }
436
Stephen Hemminger53602f92006-07-14 14:49:32 -0700437 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700438 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700439
Patrick McHardy6e23ae22007-11-19 18:53:30 -0800440 return NF_HOOK(PF_INET, NF_INET_PRE_ROUTING, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 ip_rcv_finish);
442
443inhdr_error:
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700444 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900446 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900448 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449}