blob: 073a9b01c40c9c37400ce9c320ffb355b0ae6e76 [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/*
Eric Dumazet66018502010-06-07 03:12:08 +0000149 * Process Router Attention IP option (RFC 2113)
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
Eric Dumazet66018502010-06-07 03:12:08 +0000158 for (ra = rcu_dereference(ip_ra_chain); ra; ra = rcu_dereference(ra->next)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 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 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000164 if (sk && inet_sk(sk)->inet_num == protocol &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 (!sk->sk_bound_dev_if ||
Denis V. Lunevcb846632008-03-24 15:31:00 -0700166 sk->sk_bound_dev_if == dev->ifindex) &&
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800167 net_eq(sock_net(sk), dev_net(dev))) {
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700168 if (ip_is_fragment(ip_hdr(skb))) {
Eric Dumazet66018502010-06-07 03:12:08 +0000169 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 }
172 if (last) {
173 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
174 if (skb2)
175 raw_rcv(last, skb2);
176 }
177 last = sk;
178 }
179 }
180
181 if (last) {
182 raw_rcv(last, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 return 1;
184 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 return 0;
186}
187
Patrick McHardy861d0482007-10-15 01:48:39 -0700188static int ip_local_deliver_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900190 struct net *net = dev_net(skb->dev);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700191
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300192 __skb_pull(skb, ip_hdrlen(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900194 /* Point into the IP datagram, just past the header. */
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300195 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196
197 rcu_read_lock();
198 {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700199 int protocol = ip_hdr(skb)->protocol;
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800200 int hash, raw;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000201 const struct net_protocol *ipprot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202
203 resubmit:
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800204 raw = raw_local_deliver(skb, protocol);
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 hash = protocol & (MAX_INET_PROTOS - 1);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700207 ipprot = rcu_dereference(inet_protos[hash]);
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800208 if (ipprot != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 int ret;
210
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800211 if (!net_eq(net, &init_net) && !ipprot->netns_ok) {
212 if (net_ratelimit())
213 printk("%s: proto %d isn't netns-ready\n",
214 __func__, protocol);
215 kfree_skb(skb);
216 goto out;
217 }
218
Patrick McHardyb59c2702006-01-06 23:06:10 -0800219 if (!ipprot->no_policy) {
220 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
221 kfree_skb(skb);
222 goto out;
223 }
224 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 }
226 ret = ipprot->handler(skb);
227 if (ret < 0) {
228 protocol = -ret;
229 goto resubmit;
230 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700231 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800233 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700235 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 icmp_send(skb, ICMP_DEST_UNREACH,
237 ICMP_PROT_UNREACH, 0);
238 }
239 } else
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700240 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 kfree_skb(skb);
242 }
243 }
244 out:
245 rcu_read_unlock();
246
247 return 0;
248}
249
250/*
251 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900252 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253int ip_local_deliver(struct sk_buff *skb)
254{
255 /*
256 * Reassemble IP fragments.
257 */
258
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700259 if (ip_is_fragment(ip_hdr(skb))) {
Herbert Xu776c7292007-10-14 00:38:32 -0700260 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 return 0;
262 }
263
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100264 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 ip_local_deliver_finish);
266}
267
Thomas Grafd2454072005-08-20 17:26:12 -0700268static inline int ip_rcv_options(struct sk_buff *skb)
269{
270 struct ip_options *opt;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000271 const struct iphdr *iph;
Thomas Grafd2454072005-08-20 17:26:12 -0700272 struct net_device *dev = skb->dev;
273
274 /* It looks as overkill, because not all
275 IP options require packet mangling.
276 But it is the easiest for now, especially taking
277 into account that combination of IP options
278 and running sniffer is extremely rare condition.
279 --ANK (980813)
280 */
281 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700282 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700283 goto drop;
284 }
285
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700286 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700287 opt = &(IPCB(skb)->opt);
288 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700289
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900290 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700291 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700292 goto drop;
293 }
294
Thomas Grafd2454072005-08-20 17:26:12 -0700295 if (unlikely(opt->srr)) {
Eric Dumazet6e8b11b2010-06-07 03:54:46 +0000296 struct in_device *in_dev = __in_dev_get_rcu(dev);
297
Thomas Grafd2454072005-08-20 17:26:12 -0700298 if (in_dev) {
299 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
300 if (IN_DEV_LOG_MARTIANS(in_dev) &&
301 net_ratelimit())
Harvey Harrison673d57e2008-10-31 00:53:57 -0700302 printk(KERN_INFO "source route option %pI4 -> %pI4\n",
303 &iph->saddr, &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700304 goto drop;
305 }
Thomas Grafd2454072005-08-20 17:26:12 -0700306 }
307
308 if (ip_options_rcv_srr(skb))
309 goto drop;
310 }
311
312 return 0;
313drop:
314 return -1;
315}
316
Patrick McHardy861d0482007-10-15 01:48:39 -0700317static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700319 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700320 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
322 /*
323 * Initialise the virtual path cache for the packet. It describes
324 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900325 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000326 if (skb_dst(skb) == NULL) {
Eric Dumazet4a944452010-05-10 11:33:06 +0000327 int err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
328 iph->tos, skb->dev);
Thomas Graf3e192be2005-08-20 17:26:30 -0700329 if (unlikely(err)) {
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700330 if (err == -EHOSTUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700331 IP_INC_STATS_BH(dev_net(skb->dev),
332 IPSTATS_MIB_INADDRERRORS);
Mitsuru Chinene91a47e2007-04-30 00:45:49 -0700333 else if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700334 IP_INC_STATS_BH(dev_net(skb->dev),
335 IPSTATS_MIB_INNOROUTES);
Eric Dumazetb5f7e752010-06-02 12:05:27 +0000336 else if (err == -EXDEV)
337 NET_INC_STATS_BH(dev_net(skb->dev),
338 LINUX_MIB_IPRPFILTER);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900339 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700340 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 }
342
Patrick McHardyc7066f72011-01-14 13:36:42 +0100343#ifdef CONFIG_IP_ROUTE_CLASSID
Eric Dumazetadf30902009-06-02 05:19:30 +0000344 if (unlikely(skb_dst(skb)->tclassid)) {
Eric Dumazet7a9b2d52010-06-24 00:52:37 +0000345 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct);
Eric Dumazetadf30902009-06-02 05:19:30 +0000346 u32 idx = skb_dst(skb)->tclassid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 st[idx&0xFF].o_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800348 st[idx&0xFF].o_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 st[(idx>>16)&0xFF].i_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800350 st[(idx>>16)&0xFF].i_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 }
352#endif
353
Thomas Grafd2454072005-08-20 17:26:12 -0700354 if (iph->ihl > 5 && ip_rcv_options(skb))
355 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356
Eric Dumazet511c3f92009-06-02 05:14:27 +0000357 rt = skb_rtable(skb);
Neil Hormanedf391f2009-04-27 02:45:02 -0700358 if (rt->rt_type == RTN_MULTICAST) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700359 IP_UPD_PO_STATS_BH(dev_net(rt->dst.dev), IPSTATS_MIB_INMCAST,
Neil Hormanedf391f2009-04-27 02:45:02 -0700360 skb->len);
361 } else if (rt->rt_type == RTN_BROADCAST)
Changli Gaod8d1f302010-06-10 23:31:35 -0700362 IP_UPD_PO_STATS_BH(dev_net(rt->dst.dev), IPSTATS_MIB_INBCAST,
Neil Hormanedf391f2009-04-27 02:45:02 -0700363 skb->len);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700364
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 return dst_input(skb);
366
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900368 kfree_skb(skb);
369 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370}
371
372/*
373 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900374 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700375int 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 -0700376{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000377 const struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700378 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379
380 /* When the interface is in promisc. mode, drop all the crap
381 * that it receives, do not try to analyse it.
382 */
383 if (skb->pkt_type == PACKET_OTHERHOST)
384 goto drop;
385
Neil Hormanedf391f2009-04-27 02:45:02 -0700386
387 IP_UPD_PO_STATS_BH(dev_net(dev), IPSTATS_MIB_IN, skb->len);
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
Herbert Xu71f9dac2009-06-26 19:22:37 -0700440 /* Must drop socket now because of tproxy. */
441 skb_orphan(skb);
442
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100443 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 ip_rcv_finish);
445
446inhdr_error:
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700447 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900449 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900451 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452}