blob: f1395a6fb35fcbb2f49d3a19ce6d48dcf43e4e89 [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
Joe Perchesafd465032012-03-12 07:03:32 +0000116#define pr_fmt(fmt) "IPv4: " fmt
117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118#include <linux/module.h>
119#include <linux/types.h>
120#include <linux/kernel.h>
121#include <linux/string.h>
122#include <linux/errno.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +0900123#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125#include <linux/net.h>
126#include <linux/socket.h>
127#include <linux/sockios.h>
128#include <linux/in.h>
129#include <linux/inet.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -0200130#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131#include <linux/netdevice.h>
132#include <linux/etherdevice.h>
133
134#include <net/snmp.h>
135#include <net/ip.h>
136#include <net/protocol.h>
137#include <net/route.h>
138#include <linux/skbuff.h>
139#include <net/sock.h>
140#include <net/arp.h>
141#include <net/icmp.h>
142#include <net/raw.h>
143#include <net/checksum.h>
144#include <linux/netfilter_ipv4.h>
145#include <net/xfrm.h>
146#include <linux/mroute.h>
147#include <linux/netlink.h>
148
149/*
Eric Dumazet66018502010-06-07 03:12:08 +0000150 * Process Router Attention IP option (RFC 2113)
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900151 */
David S. Millerba57b4d2012-03-07 20:45:32 -0500152bool ip_call_ra_chain(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153{
154 struct ip_ra_chain *ra;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700155 u8 protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 struct sock *last = NULL;
Denis V. Lunevcb846632008-03-24 15:31:00 -0700157 struct net_device *dev = skb->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
Eric Dumazet66018502010-06-07 03:12:08 +0000159 for (ra = rcu_dereference(ip_ra_chain); ra; ra = rcu_dereference(ra->next)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 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))) {
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700169 if (ip_is_fragment(ip_hdr(skb))) {
Eric Dumazet66018502010-06-07 03:12:08 +0000170 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN))
David S. Millerba57b4d2012-03-07 20:45:32 -0500171 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 }
173 if (last) {
174 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
175 if (skb2)
176 raw_rcv(last, skb2);
177 }
178 last = sk;
179 }
180 }
181
182 if (last) {
183 raw_rcv(last, skb);
David S. Millerba57b4d2012-03-07 20:45:32 -0500184 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 }
David S. Millerba57b4d2012-03-07 20:45:32 -0500186 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187}
188
Patrick McHardy861d0482007-10-15 01:48:39 -0700189static int ip_local_deliver_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900191 struct net *net = dev_net(skb->dev);
Denis V. Lunevf1450492008-03-24 15:33:00 -0700192
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300193 __skb_pull(skb, ip_hdrlen(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900195 /* Point into the IP datagram, just past the header. */
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300196 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
198 rcu_read_lock();
199 {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700200 int protocol = ip_hdr(skb)->protocol;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000201 const struct net_protocol *ipprot;
David S. Millerf9242b62012-06-19 18:56:21 -0700202 int raw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
204 resubmit:
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800205 raw = raw_local_deliver(skb, protocol);
206
David S. Millerf9242b62012-06-19 18:56:21 -0700207 ipprot = rcu_dereference(inet_protos[protocol]);
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) {
Joe Perchese87cc472012-05-13 21:56:26 +0000212 net_info_ratelimited("%s: proto %d isn't netns-ready\n",
213 __func__, protocol);
Alexey Dobriyan9c0188a2008-11-12 23:23:51 -0800214 kfree_skb(skb);
215 goto out;
216 }
217
Patrick McHardyb59c2702006-01-06 23:06:10 -0800218 if (!ipprot->no_policy) {
219 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
220 kfree_skb(skb);
221 goto out;
222 }
223 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 }
225 ret = ipprot->handler(skb);
226 if (ret < 0) {
227 protocol = -ret;
228 goto resubmit;
229 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700230 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800232 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700234 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 icmp_send(skb, ICMP_DEST_UNREACH,
236 ICMP_PROT_UNREACH, 0);
237 }
238 } else
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700239 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 kfree_skb(skb);
241 }
242 }
243 out:
244 rcu_read_unlock();
245
246 return 0;
247}
248
249/*
250 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900251 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252int ip_local_deliver(struct sk_buff *skb)
253{
254 /*
255 * Reassemble IP fragments.
256 */
257
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700258 if (ip_is_fragment(ip_hdr(skb))) {
Herbert Xu776c7292007-10-14 00:38:32 -0700259 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 return 0;
261 }
262
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100263 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 ip_local_deliver_finish);
265}
266
David S. Miller6a913952012-03-07 20:48:08 -0500267static inline bool ip_rcv_options(struct sk_buff *skb)
Thomas Grafd2454072005-08-20 17:26:12 -0700268{
269 struct ip_options *opt;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000270 const struct iphdr *iph;
Thomas Grafd2454072005-08-20 17:26:12 -0700271 struct net_device *dev = skb->dev;
272
273 /* It looks as overkill, because not all
274 IP options require packet mangling.
275 But it is the easiest for now, especially taking
276 into account that combination of IP options
277 and running sniffer is extremely rare condition.
278 --ANK (980813)
279 */
280 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700281 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700282 goto drop;
283 }
284
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700285 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700286 opt = &(IPCB(skb)->opt);
287 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700288
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900289 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700290 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700291 goto drop;
292 }
293
Thomas Grafd2454072005-08-20 17:26:12 -0700294 if (unlikely(opt->srr)) {
Eric Dumazet6e8b11b2010-06-07 03:54:46 +0000295 struct in_device *in_dev = __in_dev_get_rcu(dev);
296
Thomas Grafd2454072005-08-20 17:26:12 -0700297 if (in_dev) {
298 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
Joe Perchese87cc472012-05-13 21:56:26 +0000299 if (IN_DEV_LOG_MARTIANS(in_dev))
300 net_info_ratelimited("source route option %pI4 -> %pI4\n",
301 &iph->saddr,
302 &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700303 goto drop;
304 }
Thomas Grafd2454072005-08-20 17:26:12 -0700305 }
306
307 if (ip_options_rcv_srr(skb))
308 goto drop;
309 }
310
David S. Miller6a913952012-03-07 20:48:08 -0500311 return false;
Thomas Grafd2454072005-08-20 17:26:12 -0700312drop:
David S. Miller6a913952012-03-07 20:48:08 -0500313 return true;
Thomas Grafd2454072005-08-20 17:26:12 -0700314}
315
Alexander Duyck6648bd72012-06-21 13:58:31 +0000316int sysctl_ip_early_demux __read_mostly = 1;
Eric Dumazetc7109982012-07-26 12:18:11 +0000317EXPORT_SYMBOL(sysctl_ip_early_demux);
Alexander Duyck6648bd72012-06-21 13:58:31 +0000318
Patrick McHardy861d0482007-10-15 01:48:39 -0700319static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700321 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700322 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
David S. Miller160eb5a2012-06-27 22:01:22 -0700324 if (sysctl_ip_early_demux && !skb_dst(skb)) {
325 const struct net_protocol *ipprot;
326 int protocol = iph->protocol;
327
David S. Miller160eb5a2012-06-27 22:01:22 -0700328 ipprot = rcu_dereference(inet_protos[protocol]);
Eric Dumazet9cb429d2012-07-24 01:19:31 +0000329 if (ipprot && ipprot->early_demux) {
David S. Miller160eb5a2012-06-27 22:01:22 -0700330 ipprot->early_demux(skb);
Eric Dumazet9cb429d2012-07-24 01:19:31 +0000331 /* must reload iph, skb->head might have changed */
332 iph = ip_hdr(skb);
333 }
David S. Miller160eb5a2012-06-27 22:01:22 -0700334 }
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 /*
337 * Initialise the virtual path cache for the packet. It describes
338 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900339 */
David S. Miller160eb5a2012-06-27 22:01:22 -0700340 if (!skb_dst(skb)) {
David S. Millerc6cffba2012-07-26 11:14:38 +0000341 int err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
342 iph->tos, skb->dev);
David S. Miller160eb5a2012-06-27 22:01:22 -0700343 if (unlikely(err)) {
344 if (err == -EXDEV)
345 NET_INC_STATS_BH(dev_net(skb->dev),
346 LINUX_MIB_IPRPFILTER);
347 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700348 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 }
350
Patrick McHardyc7066f72011-01-14 13:36:42 +0100351#ifdef CONFIG_IP_ROUTE_CLASSID
Eric Dumazetadf30902009-06-02 05:19:30 +0000352 if (unlikely(skb_dst(skb)->tclassid)) {
Eric Dumazet7a9b2d52010-06-24 00:52:37 +0000353 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct);
Eric Dumazetadf30902009-06-02 05:19:30 +0000354 u32 idx = skb_dst(skb)->tclassid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 st[idx&0xFF].o_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800356 st[idx&0xFF].o_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 st[(idx>>16)&0xFF].i_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800358 st[(idx>>16)&0xFF].i_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 }
360#endif
361
Thomas Grafd2454072005-08-20 17:26:12 -0700362 if (iph->ihl > 5 && ip_rcv_options(skb))
363 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Eric Dumazet511c3f92009-06-02 05:14:27 +0000365 rt = skb_rtable(skb);
Neil Hormanedf391f2009-04-27 02:45:02 -0700366 if (rt->rt_type == RTN_MULTICAST) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700367 IP_UPD_PO_STATS_BH(dev_net(rt->dst.dev), IPSTATS_MIB_INMCAST,
Neil Hormanedf391f2009-04-27 02:45:02 -0700368 skb->len);
369 } else if (rt->rt_type == RTN_BROADCAST)
Changli Gaod8d1f302010-06-10 23:31:35 -0700370 IP_UPD_PO_STATS_BH(dev_net(rt->dst.dev), IPSTATS_MIB_INBCAST,
Neil Hormanedf391f2009-04-27 02:45:02 -0700371 skb->len);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700372
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 return dst_input(skb);
374
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900376 kfree_skb(skb);
377 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378}
379
380/*
381 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900382 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700383int 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 -0700384{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000385 const struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700386 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
388 /* When the interface is in promisc. mode, drop all the crap
389 * that it receives, do not try to analyse it.
390 */
391 if (skb->pkt_type == PACKET_OTHERHOST)
392 goto drop;
393
Neil Hormanedf391f2009-04-27 02:45:02 -0700394
395 IP_UPD_PO_STATS_BH(dev_net(dev), IPSTATS_MIB_IN, skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
397 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700398 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 goto out;
400 }
401
402 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
403 goto inhdr_error;
404
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700405 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
407 /*
J.H.M. Dassen (Ray)c67fa022008-05-08 01:11:04 -0700408 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 *
410 * Is the datagram acceptable?
411 *
412 * 1. Length at least the size of an ip header
413 * 2. Version of 4
414 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
415 * 4. Doesn't have a bogus length
416 */
417
418 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700419 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
421 if (!pskb_may_pull(skb, iph->ihl*4))
422 goto inhdr_error;
423
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700424 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Thomas Grafe9c60422005-08-20 17:25:52 -0700426 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Thomas Graf58615242005-08-20 17:25:29 -0700427 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
Thomas Graf58615242005-08-20 17:25:29 -0700429 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700430 if (skb->len < len) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700431 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700432 goto drop;
433 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700434 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
Thomas Graf58615242005-08-20 17:25:29 -0700436 /* Our transport medium may have padded the buffer out. Now we know it
437 * is IP we can trim to the true length of the frame.
438 * Note this now means skb->len holds ntohs(iph->tot_len).
439 */
440 if (pskb_trim_rcsum(skb, len)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700441 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Graf58615242005-08-20 17:25:29 -0700442 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 }
444
Stephen Hemminger53602f92006-07-14 14:49:32 -0700445 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700446 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700447
Herbert Xu71f9dac2009-06-26 19:22:37 -0700448 /* Must drop socket now because of tproxy. */
449 skb_orphan(skb);
450
Jan Engelhardt9bbc7682010-03-23 04:07:29 +0100451 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 ip_rcv_finish);
453
454inhdr_error:
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700455 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900457 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900459 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460}