blob: d930dc5e4d85a752ddb7af15b482b47108b62241 [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +0900122#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
124#include <linux/net.h>
125#include <linux/socket.h>
126#include <linux/sockios.h>
127#include <linux/in.h>
128#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -0200129#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130#include <linux/netdevice.h>
131#include <linux/etherdevice.h>
132
133#include <net/snmp.h>
134#include <net/ip.h>
135#include <net/protocol.h>
136#include <net/route.h>
137#include <linux/skbuff.h>
138#include <net/sock.h>
139#include <net/arp.h>
140#include <net/icmp.h>
141#include <net/raw.h>
142#include <net/checksum.h>
143#include <linux/netfilter_ipv4.h>
144#include <net/xfrm.h>
145#include <linux/mroute.h>
146#include <linux/netlink.h>
147
148/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 * Process Router Attention IP option
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900150 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151int ip_call_ra_chain(struct sk_buff *skb)
152{
153 struct ip_ra_chain *ra;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700154 u8 protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 struct sock *last = NULL;
Denis V. Lunevcb846632008-03-24 15:31:00 -0700156 struct net_device *dev = skb->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158 read_lock(&ip_ra_lock);
159 for (ra = ip_ra_chain; ra; ra = ra->next) {
160 struct sock *sk = ra->sk;
161
162 /* If socket is bound to an interface, only report
163 * the packet if it came from that interface.
164 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000165 if (sk && inet_sk(sk)->inet_num == protocol &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 (!sk->sk_bound_dev_if ||
Denis V. Lunevcb846632008-03-24 15:31:00 -0700167 sk->sk_bound_dev_if == dev->ifindex) &&
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800168 net_eq(sock_net(sk), dev_net(dev))) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700169 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700170 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 read_unlock(&ip_ra_lock);
172 return 1;
173 }
174 }
175 if (last) {
176 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
177 if (skb2)
178 raw_rcv(last, skb2);
179 }
180 last = sk;
181 }
182 }
183
184 if (last) {
185 raw_rcv(last, skb);
186 read_unlock(&ip_ra_lock);
187 return 1;
188 }
189 read_unlock(&ip_ra_lock);
190 return 0;
191}
192
Patrick McHardy861d0482007-10-15 01:48:39 -0700193static int ip_local_deliver_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900195 struct net *net = dev_net(skb->dev);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700196
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300197 __skb_pull(skb, ip_hdrlen(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900199 /* Point into the IP datagram, just past the header. */
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300200 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202 rcu_read_lock();
203 {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700204 int protocol = ip_hdr(skb)->protocol;
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800205 int hash, raw;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000206 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
208 resubmit:
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800209 raw = raw_local_deliver(skb, protocol);
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 hash = protocol & (MAX_INET_PROTOS - 1);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700212 ipprot = rcu_dereference(inet_protos[hash]);
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800213 if (ipprot != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 int ret;
215
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800216 if (!net_eq(net, &init_net) && !ipprot->netns_ok) {
217 if (net_ratelimit())
218 printk("%s: proto %d isn't netns-ready\n",
219 __func__, protocol);
220 kfree_skb(skb);
221 goto out;
222 }
223
Patrick McHardyb59c2702006-01-06 23:06:10 -0800224 if (!ipprot->no_policy) {
225 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
226 kfree_skb(skb);
227 goto out;
228 }
229 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 }
231 ret = ipprot->handler(skb);
232 if (ret < 0) {
233 protocol = -ret;
234 goto resubmit;
235 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700236 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800238 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700240 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 icmp_send(skb, ICMP_DEST_UNREACH,
242 ICMP_PROT_UNREACH, 0);
243 }
244 } else
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700245 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 kfree_skb(skb);
247 }
248 }
249 out:
250 rcu_read_unlock();
251
252 return 0;
253}
254
255/*
256 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900257 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258int ip_local_deliver(struct sk_buff *skb)
259{
260 /*
261 * Reassemble IP fragments.
262 */
263
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700264 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700265 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 return 0;
267 }
268
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100269 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 ip_local_deliver_finish);
271}
272
Thomas Grafd2454072005-08-20 17:26:12 -0700273static inline int ip_rcv_options(struct sk_buff *skb)
274{
275 struct ip_options *opt;
276 struct iphdr *iph;
277 struct net_device *dev = skb->dev;
278
279 /* It looks as overkill, because not all
280 IP options require packet mangling.
281 But it is the easiest for now, especially taking
282 into account that combination of IP options
283 and running sniffer is extremely rare condition.
284 --ANK (980813)
285 */
286 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700287 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700288 goto drop;
289 }
290
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700291 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700292 opt = &(IPCB(skb)->opt);
293 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700294
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900295 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700296 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700297 goto drop;
298 }
299
Thomas Grafd2454072005-08-20 17:26:12 -0700300 if (unlikely(opt->srr)) {
301 struct in_device *in_dev = in_dev_get(dev);
302 if (in_dev) {
303 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
304 if (IN_DEV_LOG_MARTIANS(in_dev) &&
305 net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700306 printk(KERN_INFO "source route option %pI4 -> %pI4\n",
307 &iph->saddr, &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700308 in_dev_put(in_dev);
309 goto drop;
310 }
311
312 in_dev_put(in_dev);
313 }
314
315 if (ip_options_rcv_srr(skb))
316 goto drop;
317 }
318
319 return 0;
320drop:
321 return -1;
322}
323
Patrick McHardy861d0482007-10-15 01:48:39 -0700324static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700326 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700327 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
329 /*
330 * Initialise the virtual path cache for the packet. It describes
331 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900332 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000333 if (skb_dst(skb) == NULL) {
Eric Dumazet4a944452010-05-10 11:33:06 +0000334 int err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
335 iph->tos, skb->dev);
Thomas Graf3e192be2005-08-20 17:26:30 -0700336 if (unlikely(err)) {
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700337 if (err == -EHOSTUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700338 IP_INC_STATS_BH(dev_net(skb->dev),
339 IPSTATS_MIB_INADDRERRORS);
Mitsuru Chinene91a47e2007-04-30 00:45:49 -0700340 else if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700341 IP_INC_STATS_BH(dev_net(skb->dev),
342 IPSTATS_MIB_INNOROUTES);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900343 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700344 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 }
346
347#ifdef CONFIG_NET_CLS_ROUTE
Eric Dumazetadf30902009-06-02 05:19:30 +0000348 if (unlikely(skb_dst(skb)->tclassid)) {
Eric Dumazet8dbde282007-11-16 03:32:10 -0800349 struct ip_rt_acct *st = per_cpu_ptr(ip_rt_acct, smp_processor_id());
Eric Dumazetadf30902009-06-02 05:19:30 +0000350 u32 idx = skb_dst(skb)->tclassid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 st[idx&0xFF].o_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800352 st[idx&0xFF].o_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 st[(idx>>16)&0xFF].i_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800354 st[(idx>>16)&0xFF].i_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 }
356#endif
357
Thomas Grafd2454072005-08-20 17:26:12 -0700358 if (iph->ihl > 5 && ip_rcv_options(skb))
359 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
Eric Dumazet511c3f92009-06-02 05:14:27 +0000361 rt = skb_rtable(skb);
Neil Hormanedf391f2009-04-27 02:45:02 -0700362 if (rt->rt_type == RTN_MULTICAST) {
363 IP_UPD_PO_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INMCAST,
364 skb->len);
365 } else if (rt->rt_type == RTN_BROADCAST)
366 IP_UPD_PO_STATS_BH(dev_net(rt->u.dst.dev), IPSTATS_MIB_INBCAST,
367 skb->len);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700368
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 return dst_input(skb);
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900372 kfree_skb(skb);
373 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374}
375
376/*
377 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900378 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700379int 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 -0700380{
381 struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700382 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
384 /* When the interface is in promisc. mode, drop all the crap
385 * that it receives, do not try to analyse it.
386 */
387 if (skb->pkt_type == PACKET_OTHERHOST)
388 goto drop;
389
Neil Hormanedf391f2009-04-27 02:45:02 -0700390
391 IP_UPD_PO_STATS_BH(dev_net(dev), IPSTATS_MIB_IN, skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700394 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 goto out;
396 }
397
398 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
399 goto inhdr_error;
400
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700401 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
403 /*
J.H.M. Dassen (Ray)c67fa022008-05-08 01:11:04 -0700404 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 *
406 * Is the datagram acceptable?
407 *
408 * 1. Length at least the size of an ip header
409 * 2. Version of 4
410 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
411 * 4. Doesn't have a bogus length
412 */
413
414 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700415 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
417 if (!pskb_may_pull(skb, iph->ihl*4))
418 goto inhdr_error;
419
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700420 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
Thomas Grafe9c60422005-08-20 17:25:52 -0700422 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Thomas Graf58615242005-08-20 17:25:29 -0700423 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424
Thomas Graf58615242005-08-20 17:25:29 -0700425 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700426 if (skb->len < len) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700427 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700428 goto drop;
429 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700430 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
Thomas Graf58615242005-08-20 17:25:29 -0700432 /* Our transport medium may have padded the buffer out. Now we know it
433 * is IP we can trim to the true length of the frame.
434 * Note this now means skb->len holds ntohs(iph->tot_len).
435 */
436 if (pskb_trim_rcsum(skb, len)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700437 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Graf58615242005-08-20 17:25:29 -0700438 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 }
440
Stephen Hemminger53602f92006-07-14 14:49:32 -0700441 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700442 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700443
Herbert Xu71f9dac2009-06-26 19:22:37 -0700444 /* Must drop socket now because of tproxy. */
445 skb_orphan(skb);
446
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100447 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 ip_rcv_finish);
449
450inhdr_error:
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700451 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900453 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900455 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456}