blob: d77eb0c3b6842bc5605a882f9ad46c609dcdeb50 [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>
Eric Dumazet1f07d032013-08-06 03:32:11 -0700144#include <net/inet_ecn.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145#include <linux/netfilter_ipv4.h>
146#include <net/xfrm.h>
147#include <linux/mroute.h>
148#include <linux/netlink.h>
Thomas Graff38a9eb2015-07-21 10:43:56 +0200149#include <net/dst_metadata.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150
151/*
Eric Dumazet66018502010-06-07 03:12:08 +0000152 * Process Router Attention IP option (RFC 2113)
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900153 */
David S. Millerba57b4d2012-03-07 20:45:32 -0500154bool ip_call_ra_chain(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155{
156 struct ip_ra_chain *ra;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700157 u8 protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 struct sock *last = NULL;
Denis V. Lunevcb846632008-03-24 15:31:00 -0700159 struct net_device *dev = skb->dev;
Eric W. Biederman37fcbab2015-10-09 13:44:53 -0500160 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Eric Dumazet66018502010-06-07 03:12:08 +0000162 for (ra = rcu_dereference(ip_ra_chain); ra; ra = rcu_dereference(ra->next)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 struct sock *sk = ra->sk;
164
165 /* If socket is bound to an interface, only report
166 * the packet if it came from that interface.
167 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000168 if (sk && inet_sk(sk)->inet_num == protocol &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 (!sk->sk_bound_dev_if ||
Denis V. Lunevcb846632008-03-24 15:31:00 -0700170 sk->sk_bound_dev_if == dev->ifindex) &&
Eric W. Biederman37fcbab2015-10-09 13:44:53 -0500171 net_eq(sock_net(sk), net)) {
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700172 if (ip_is_fragment(ip_hdr(skb))) {
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500173 if (ip_defrag(net, skb, IP_DEFRAG_CALL_RA_CHAIN))
David S. Millerba57b4d2012-03-07 20:45:32 -0500174 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 }
176 if (last) {
177 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
178 if (skb2)
179 raw_rcv(last, skb2);
180 }
181 last = sk;
182 }
183 }
184
185 if (last) {
186 raw_rcv(last, skb);
David S. Millerba57b4d2012-03-07 20:45:32 -0500187 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 }
David S. Millerba57b4d2012-03-07 20:45:32 -0500189 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190}
191
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500192static int ip_local_deliver_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
Eric Dumazet21d11962013-07-15 20:03:19 -0700194 __skb_pull(skb, skb_network_header_len(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
196 rcu_read_lock();
197 {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700198 int protocol = ip_hdr(skb)->protocol;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000199 const struct net_protocol *ipprot;
David S. Millerf9242b62012-06-19 18:56:21 -0700200 int raw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202 resubmit:
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800203 raw = raw_local_deliver(skb, protocol);
204
David S. Millerf9242b62012-06-19 18:56:21 -0700205 ipprot = rcu_dereference(inet_protos[protocol]);
Ian Morris00db4122015-04-03 09:17:27 +0100206 if (ipprot) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 int ret;
208
Patrick McHardyb59c2702006-01-06 23:06:10 -0800209 if (!ipprot->no_policy) {
210 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
211 kfree_skb(skb);
212 goto out;
213 }
214 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 }
216 ret = ipprot->handler(skb);
217 if (ret < 0) {
218 protocol = -ret;
219 goto resubmit;
220 }
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700221 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 } else {
Pavel Emelyanov7bc54c92007-11-19 22:35:07 -0800223 if (!raw) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700225 IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 icmp_send(skb, ICMP_DEST_UNREACH,
227 ICMP_PROT_UNREACH, 0);
228 }
Neil Hormand8c6f4b2013-03-01 07:44:08 +0000229 kfree_skb(skb);
230 } else {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700231 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
Neil Hormand8c6f4b2013-03-01 07:44:08 +0000232 consume_skb(skb);
233 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 }
235 }
236 out:
237 rcu_read_unlock();
238
239 return 0;
240}
241
242/*
243 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900244 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245int ip_local_deliver(struct sk_buff *skb)
246{
247 /*
248 * Reassemble IP fragments.
249 */
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500250 struct net *net = dev_net(skb->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
Paul Gortmaker56f8a752011-06-21 20:33:34 -0700252 if (ip_is_fragment(ip_hdr(skb))) {
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500253 if (ip_defrag(net, skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 return 0;
255 }
256
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500257 return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN,
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500258 net, NULL, skb, skb->dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 ip_local_deliver_finish);
260}
261
David S. Miller6a913952012-03-07 20:48:08 -0500262static inline bool ip_rcv_options(struct sk_buff *skb)
Thomas Grafd2454072005-08-20 17:26:12 -0700263{
264 struct ip_options *opt;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000265 const struct iphdr *iph;
Thomas Grafd2454072005-08-20 17:26:12 -0700266 struct net_device *dev = skb->dev;
267
268 /* It looks as overkill, because not all
269 IP options require packet mangling.
270 But it is the easiest for now, especially taking
271 into account that combination of IP options
272 and running sniffer is extremely rare condition.
273 --ANK (980813)
274 */
275 if (skb_cow(skb, skb_headroom(skb))) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700276 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
Thomas Grafd2454072005-08-20 17:26:12 -0700277 goto drop;
278 }
279
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700280 iph = ip_hdr(skb);
Denis V. Lunev22aba382008-03-22 16:36:20 -0700281 opt = &(IPCB(skb)->opt);
282 opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
Thomas Grafd2454072005-08-20 17:26:12 -0700283
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +0900284 if (ip_options_compile(dev_net(dev), opt, skb)) {
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700285 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
Thomas Grafd2454072005-08-20 17:26:12 -0700286 goto drop;
287 }
288
Thomas Grafd2454072005-08-20 17:26:12 -0700289 if (unlikely(opt->srr)) {
Eric Dumazet6e8b11b2010-06-07 03:54:46 +0000290 struct in_device *in_dev = __in_dev_get_rcu(dev);
291
Thomas Grafd2454072005-08-20 17:26:12 -0700292 if (in_dev) {
293 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
Joe Perchese87cc472012-05-13 21:56:26 +0000294 if (IN_DEV_LOG_MARTIANS(in_dev))
295 net_info_ratelimited("source route option %pI4 -> %pI4\n",
296 &iph->saddr,
297 &iph->daddr);
Thomas Grafd2454072005-08-20 17:26:12 -0700298 goto drop;
299 }
Thomas Grafd2454072005-08-20 17:26:12 -0700300 }
301
302 if (ip_options_rcv_srr(skb))
303 goto drop;
304 }
305
David S. Miller6a913952012-03-07 20:48:08 -0500306 return false;
Thomas Grafd2454072005-08-20 17:26:12 -0700307drop:
David S. Miller6a913952012-03-07 20:48:08 -0500308 return true;
Thomas Grafd2454072005-08-20 17:26:12 -0700309}
310
Alexander Duyck6648bd72012-06-21 13:58:31 +0000311int sysctl_ip_early_demux __read_mostly = 1;
Eric Dumazetc7109982012-07-26 12:18:11 +0000312EXPORT_SYMBOL(sysctl_ip_early_demux);
Alexander Duyck6648bd72012-06-21 13:58:31 +0000313
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -0500314static int ip_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700316 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700317 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
Eric Dumazet63e51b62016-01-26 16:59:42 -0800319 if (sysctl_ip_early_demux &&
320 !skb_dst(skb) &&
321 !skb->sk &&
322 !ip_is_fragment(iph)) {
David S. Miller160eb5a2012-06-27 22:01:22 -0700323 const struct net_protocol *ipprot;
324 int protocol = iph->protocol;
325
David S. Miller160eb5a2012-06-27 22:01:22 -0700326 ipprot = rcu_dereference(inet_protos[protocol]);
Eric Dumazet9cb429d2012-07-24 01:19:31 +0000327 if (ipprot && ipprot->early_demux) {
David S. Miller160eb5a2012-06-27 22:01:22 -0700328 ipprot->early_demux(skb);
Eric Dumazet9cb429d2012-07-24 01:19:31 +0000329 /* must reload iph, skb->head might have changed */
330 iph = ip_hdr(skb);
331 }
David S. Miller160eb5a2012-06-27 22:01:22 -0700332 }
333
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 /*
335 * Initialise the virtual path cache for the packet. It describes
336 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900337 */
Thomas Graff38a9eb2015-07-21 10:43:56 +0200338 if (!skb_valid_dst(skb)) {
David S. Millerc6cffba2012-07-26 11:14:38 +0000339 int err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
340 iph->tos, skb->dev);
David S. Miller160eb5a2012-06-27 22:01:22 -0700341 if (unlikely(err)) {
342 if (err == -EXDEV)
Eric W. Biederman38184b32015-09-15 20:04:02 -0500343 NET_INC_STATS_BH(net, LINUX_MIB_IPRPFILTER);
David S. Miller160eb5a2012-06-27 22:01:22 -0700344 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 }
347
Patrick McHardyc7066f72011-01-14 13:36:42 +0100348#ifdef CONFIG_IP_ROUTE_CLASSID
Eric Dumazetadf30902009-06-02 05:19:30 +0000349 if (unlikely(skb_dst(skb)->tclassid)) {
Eric Dumazet7a9b2d52010-06-24 00:52:37 +0000350 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct);
Eric Dumazetadf30902009-06-02 05:19:30 +0000351 u32 idx = skb_dst(skb)->tclassid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 st[idx&0xFF].o_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800353 st[idx&0xFF].o_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 st[(idx>>16)&0xFF].i_packets++;
Jianjun Kongfd3f8c42008-11-03 02:47:38 -0800355 st[(idx>>16)&0xFF].i_bytes += skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 }
357#endif
358
Thomas Grafd2454072005-08-20 17:26:12 -0700359 if (iph->ihl > 5 && ip_rcv_options(skb))
360 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Eric Dumazet511c3f92009-06-02 05:14:27 +0000362 rt = skb_rtable(skb);
Neil Hormanedf391f2009-04-27 02:45:02 -0700363 if (rt->rt_type == RTN_MULTICAST) {
Eric W. Biederman38184b32015-09-15 20:04:02 -0500364 IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_INMCAST, skb->len);
Neil Hormanedf391f2009-04-27 02:45:02 -0700365 } else if (rt->rt_type == RTN_BROADCAST)
Eric W. Biederman38184b32015-09-15 20:04:02 -0500366 IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_INBCAST, skb->len);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700367
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 return dst_input(skb);
369
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900371 kfree_skb(skb);
372 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373}
374
375/*
376 * Main IP Receive routine.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900377 */
David S. Millerf2ccd8f2005-08-09 19:34:12 -0700378int 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 -0700379{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000380 const struct iphdr *iph;
Eric W. Biedermane7077662015-09-15 20:03:56 -0500381 struct net *net;
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
Eric W. Biedermane7077662015-09-15 20:03:56 -0500391 net = dev_net(dev);
392 IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_IN, skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
Ian Morris51456b22015-04-03 09:17:26 +0100394 skb = skb_share_check(skb, GFP_ATOMIC);
395 if (!skb) {
Eric W. Biedermane7077662015-09-15 20:03:56 -0500396 IP_INC_STATS_BH(net, IPSTATS_MIB_INDISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 goto out;
398 }
399
400 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
401 goto inhdr_error;
402
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700403 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405 /*
J.H.M. Dassen (Ray)c67fa022008-05-08 01:11:04 -0700406 * RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 *
408 * Is the datagram acceptable?
409 *
410 * 1. Length at least the size of an ip header
411 * 2. Version of 4
412 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
413 * 4. Doesn't have a bogus length
414 */
415
416 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700417 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418
Eric Dumazet1f07d032013-08-06 03:32:11 -0700419 BUILD_BUG_ON(IPSTATS_MIB_ECT1PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_1);
420 BUILD_BUG_ON(IPSTATS_MIB_ECT0PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_0);
421 BUILD_BUG_ON(IPSTATS_MIB_CEPKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_CE);
Eric W. Biedermane7077662015-09-15 20:03:56 -0500422 IP_ADD_STATS_BH(net,
Eric Dumazet1f07d032013-08-06 03:32:11 -0700423 IPSTATS_MIB_NOECTPKTS + (iph->tos & INET_ECN_MASK),
424 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
425
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 if (!pskb_may_pull(skb, iph->ihl*4))
427 goto inhdr_error;
428
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700429 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Thomas Grafe9c60422005-08-20 17:25:52 -0700431 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000432 goto csum_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
Thomas Graf58615242005-08-20 17:25:29 -0700434 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700435 if (skb->len < len) {
Eric W. Biedermane7077662015-09-15 20:03:56 -0500436 IP_INC_STATS_BH(net, IPSTATS_MIB_INTRUNCATEDPKTS);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700437 goto drop;
438 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700439 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Thomas Graf58615242005-08-20 17:25:29 -0700441 /* Our transport medium may have padded the buffer out. Now we know it
442 * is IP we can trim to the true length of the frame.
443 * Note this now means skb->len holds ntohs(iph->tot_len).
444 */
445 if (pskb_trim_rcsum(skb, len)) {
Eric W. Biedermane7077662015-09-15 20:03:56 -0500446 IP_INC_STATS_BH(net, IPSTATS_MIB_INDISCARDS);
Thomas Graf58615242005-08-20 17:25:29 -0700447 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 }
449
Eric Dumazet21d11962013-07-15 20:03:19 -0700450 skb->transport_header = skb->network_header + iph->ihl*4;
451
Stephen Hemminger53602f92006-07-14 14:49:32 -0700452 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700453 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700454
Herbert Xu71f9dac2009-06-26 19:22:37 -0700455 /* Must drop socket now because of tproxy. */
456 skb_orphan(skb);
457
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500458 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
459 net, NULL, skb, dev, NULL,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 ip_rcv_finish);
461
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +0000462csum_error:
Eric W. Biedermane7077662015-09-15 20:03:56 -0500463 IP_INC_STATS_BH(net, IPSTATS_MIB_CSUMERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464inhdr_error:
Eric W. Biedermane7077662015-09-15 20:03:56 -0500465 IP_INC_STATS_BH(net, IPSTATS_MIB_INHDRERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900467 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900469 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470}