blob: 168c871fcd796a4c4b411e440bb1fe3da004389d [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 *
8 * Version: $Id: ip_input.c,v 1.55 2002/01/12 07:39:45 davem Exp $
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Donald Becker, <becker@super.org>
13 * Alan Cox, <Alan.Cox@linux.org>
14 * Richard Underwood
15 * Stefan Becker, <stefanb@yello.ping.de>
16 * Jorge Cwik, <jorge@laser.satlink.net>
17 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090018 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 *
20 * Fixes:
21 * Alan Cox : Commented a couple of minor bits of surplus code
22 * Alan Cox : Undefining IP_FORWARD doesn't include the code
23 * (just stops a compiler warning).
24 * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes
25 * are junked rather than corrupting things.
26 * Alan Cox : Frames to bad broadcast subnets are dumped
27 * We used to process them non broadcast and
28 * boy could that cause havoc.
29 * Alan Cox : ip_forward sets the free flag on the
30 * new frame it queues. Still crap because
31 * it copies the frame but at least it
32 * doesn't eat memory too.
33 * Alan Cox : Generic queue code and memory fixes.
34 * Fred Van Kempen : IP fragment support (borrowed from NET2E)
35 * Gerhard Koerting: Forward fragmented frames correctly.
36 * Gerhard Koerting: Fixes to my fix of the above 8-).
37 * Gerhard Koerting: IP interface addressing fix.
38 * Linus Torvalds : More robustness checks
39 * Alan Cox : Even more checks: Still not as robust as it ought to be
40 * Alan Cox : Save IP header pointer for later
41 * Alan Cox : ip option setting
42 * Alan Cox : Use ip_tos/ip_ttl settings
43 * Alan Cox : Fragmentation bogosity removed
44 * (Thanks to Mark.Bush@prg.ox.ac.uk)
45 * Dmitry Gorodchanin : Send of a raw packet crash fix.
46 * Alan Cox : Silly ip bug when an overlength
47 * fragment turns up. Now frees the
48 * queue.
49 * Linus Torvalds/ : Memory leakage on fragmentation
50 * Alan Cox : handling.
51 * Gerhard Koerting: Forwarding uses IP priority hints
52 * Teemu Rantanen : Fragment problems.
53 * Alan Cox : General cleanup, comments and reformat
54 * Alan Cox : SNMP statistics
55 * Alan Cox : BSD address rule semantics. Also see
56 * UDP as there is a nasty checksum issue
57 * if you do things the wrong way.
58 * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file
59 * Alan Cox : IP options adjust sk->priority.
60 * Pedro Roque : Fix mtu/length error in ip_forward.
61 * Alan Cox : Avoid ip_chk_addr when possible.
62 * Richard Underwood : IP multicasting.
63 * Alan Cox : Cleaned up multicast handlers.
64 * Alan Cox : RAW sockets demultiplex in the BSD style.
65 * Gunther Mayer : Fix the SNMP reporting typo
66 * Alan Cox : Always in group 224.0.0.1
67 * Pauline Middelink : Fast ip_checksum update when forwarding
68 * Masquerading support.
69 * Alan Cox : Multicast loopback error for 224.0.0.1
70 * Alan Cox : IP_MULTICAST_LOOP option.
71 * Alan Cox : Use notifiers.
72 * Bjorn Ekwall : Removed ip_csum (from slhc.c too)
73 * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!)
74 * Stefan Becker : Send out ICMP HOST REDIRECT
75 * Arnt Gulbrandsen : ip_build_xmit
76 * Alan Cox : Per socket routing cache
77 * Alan Cox : Fixed routing cache, added header cache.
78 * Alan Cox : Loopback didn't work right in original ip_build_xmit - fixed it.
79 * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net.
80 * Alan Cox : Incoming IP option handling.
81 * Alan Cox : Set saddr on raw output frames as per BSD.
82 * Alan Cox : Stopped broadcast source route explosions.
83 * Alan Cox : Can disable source routing
84 * Takeshi Sone : Masquerading didn't work.
85 * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible.
86 * Alan Cox : Memory leaks, tramples, misc debugging.
87 * Alan Cox : Fixed multicast (by popular demand 8))
88 * Alan Cox : Fixed forwarding (by even more popular demand 8))
89 * Alan Cox : Fixed SNMP statistics [I think]
90 * Gerhard Koerting : IP fragmentation forwarding fix
91 * Alan Cox : Device lock against page fault.
92 * Alan Cox : IP_HDRINCL facility.
93 * Werner Almesberger : Zero fragment bug
94 * Alan Cox : RAW IP frame length bug
95 * Alan Cox : Outgoing firewall on build_xmit
96 * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel
97 * Alan Cox : Multicast routing hooks
98 * Jos Vos : Do accounting *before* call_in_firewall
99 * Willy Konynenberg : Transparent proxying support
100 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900101 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 *
103 * To Fix:
104 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
105 * and could be made very efficient with the addition of some virtual memory hacks to permit
106 * the allocation of a buffer that can then be 'grown' by twiddling page tables.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900107 * Output fragmentation wants updating along with the buffer management to use a single
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
109 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
110 * fragmentation anyway.
111 *
112 * This program is free software; you can redistribute it and/or
113 * modify it under the terms of the GNU General Public License
114 * as published by the Free Software Foundation; either version
115 * 2 of the License, or (at your option) any later version.
116 */
117
118#include <asm/system.h>
119#include <linux/module.h>
120#include <linux/types.h>
121#include <linux/kernel.h>
122#include <linux/string.h>
123#include <linux/errno.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/*
150 * SNMP management statistics
151 */
152
Eric Dumazetba899662005-08-26 12:05:31 -0700153DEFINE_SNMP_STAT(struct ipstats_mib, ip_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154
155/*
156 * Process Router Attention IP option
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900157 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158int ip_call_ra_chain(struct sk_buff *skb)
159{
160 struct ip_ra_chain *ra;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700161 u8 protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 struct sock *last = NULL;
163
164 read_lock(&ip_ra_lock);
165 for (ra = ip_ra_chain; ra; ra = ra->next) {
166 struct sock *sk = ra->sk;
167
168 /* If socket is bound to an interface, only report
169 * the packet if it came from that interface.
170 */
171 if (sk && inet_sk(sk)->num == protocol &&
172 (!sk->sk_bound_dev_if ||
173 sk->sk_bound_dev_if == skb->dev->ifindex)) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700174 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700175 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 read_unlock(&ip_ra_lock);
177 return 1;
178 }
179 }
180 if (last) {
181 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
182 if (skb2)
183 raw_rcv(last, skb2);
184 }
185 last = sk;
186 }
187 }
188
189 if (last) {
190 raw_rcv(last, skb);
191 read_unlock(&ip_ra_lock);
192 return 1;
193 }
194 read_unlock(&ip_ra_lock);
195 return 0;
196}
197
Patrick McHardy861d0482007-10-15 01:48:39 -0700198static int ip_local_deliver_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199{
Arnaldo Carvalho de Meloc9bdd4b2007-03-12 20:09:15 -0300200 __skb_pull(skb, ip_hdrlen(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900202 /* Point into the IP datagram, just past the header. */
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -0300203 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
205 rcu_read_lock();
206 {
207 /* Note: See raw.c and net/raw.h, RAWV4_HTABLE_SIZE==MAX_INET_PROTOS */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700208 int protocol = ip_hdr(skb)->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 int hash;
210 struct sock *raw_sk;
211 struct net_protocol *ipprot;
212
213 resubmit:
214 hash = protocol & (MAX_INET_PROTOS - 1);
215 raw_sk = sk_head(&raw_v4_htable[hash]);
216
217 /* If there maybe a raw socket we must check - if not we
218 * don't care less
219 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700220 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
Patrick McHardyd13964f2005-08-09 19:45:02 -0700221 raw_sk = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
223 if ((ipprot = rcu_dereference(inet_protos[hash])) != NULL) {
224 int ret;
225
Patrick McHardyb59c2702006-01-06 23:06:10 -0800226 if (!ipprot->no_policy) {
227 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
228 kfree_skb(skb);
229 goto out;
230 }
231 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 }
233 ret = ipprot->handler(skb);
234 if (ret < 0) {
235 protocol = -ret;
236 goto resubmit;
237 }
238 IP_INC_STATS_BH(IPSTATS_MIB_INDELIVERS);
239 } else {
240 if (!raw_sk) {
241 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
242 IP_INC_STATS_BH(IPSTATS_MIB_INUNKNOWNPROTOS);
243 icmp_send(skb, ICMP_DEST_UNREACH,
244 ICMP_PROT_UNREACH, 0);
245 }
246 } else
247 IP_INC_STATS_BH(IPSTATS_MIB_INDELIVERS);
248 kfree_skb(skb);
249 }
250 }
251 out:
252 rcu_read_unlock();
253
254 return 0;
255}
256
257/*
258 * Deliver IP Packets to the higher protocol layers.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900259 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260int ip_local_deliver(struct sk_buff *skb)
261{
262 /*
263 * Reassemble IP fragments.
264 */
265
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700266 if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
Herbert Xu776c7292007-10-14 00:38:32 -0700267 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 return 0;
269 }
270
271 return NF_HOOK(PF_INET, NF_IP_LOCAL_IN, skb, skb->dev, NULL,
272 ip_local_deliver_finish);
273}
274
Thomas Grafd2454072005-08-20 17:26:12 -0700275static inline int ip_rcv_options(struct sk_buff *skb)
276{
277 struct ip_options *opt;
278 struct iphdr *iph;
279 struct net_device *dev = skb->dev;
280
281 /* It looks as overkill, because not all
282 IP options require packet mangling.
283 But it is the easiest for now, especially taking
284 into account that combination of IP options
285 and running sniffer is extremely rare condition.
286 --ANK (980813)
287 */
288 if (skb_cow(skb, skb_headroom(skb))) {
289 IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS);
290 goto drop;
291 }
292
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700293 iph = ip_hdr(skb);
Thomas Grafd2454072005-08-20 17:26:12 -0700294
295 if (ip_options_compile(NULL, skb)) {
296 IP_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
297 goto drop;
298 }
299
300 opt = &(IPCB(skb)->opt);
301 if (unlikely(opt->srr)) {
302 struct in_device *in_dev = in_dev_get(dev);
303 if (in_dev) {
304 if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
305 if (IN_DEV_LOG_MARTIANS(in_dev) &&
306 net_ratelimit())
307 printk(KERN_INFO "source route option "
308 "%u.%u.%u.%u -> %u.%u.%u.%u\n",
309 NIPQUAD(iph->saddr),
310 NIPQUAD(iph->daddr));
311 in_dev_put(in_dev);
312 goto drop;
313 }
314
315 in_dev_put(in_dev);
316 }
317
318 if (ip_options_rcv_srr(skb))
319 goto drop;
320 }
321
322 return 0;
323drop:
324 return -1;
325}
326
Patrick McHardy861d0482007-10-15 01:48:39 -0700327static int ip_rcv_finish(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700329 const struct iphdr *iph = ip_hdr(skb);
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700330 struct rtable *rt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
332 /*
333 * Initialise the virtual path cache for the packet. It describes
334 * how the packet travels inside Linux networking.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900335 */
Hua Zhong0182bd22006-05-06 18:11:39 -0700336 if (skb->dst == NULL) {
Thomas Graf3e192be2005-08-20 17:26:30 -0700337 int err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos,
338 skb->dev);
339 if (unlikely(err)) {
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700340 if (err == -EHOSTUNREACH)
341 IP_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
Mitsuru Chinene91a47e2007-04-30 00:45:49 -0700342 else if (err == -ENETUNREACH)
343 IP_INC_STATS_BH(IPSTATS_MIB_INNOROUTES);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900344 goto drop;
Dietmar Eggemann2c2910a2005-06-28 13:06:23 -0700345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 }
347
348#ifdef CONFIG_NET_CLS_ROUTE
Thomas Graf3e192be2005-08-20 17:26:30 -0700349 if (unlikely(skb->dst->tclassid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 struct ip_rt_acct *st = ip_rt_acct + 256*smp_processor_id();
351 u32 idx = skb->dst->tclassid;
352 st[idx&0xFF].o_packets++;
353 st[idx&0xFF].o_bytes+=skb->len;
354 st[(idx>>16)&0xFF].i_packets++;
355 st[(idx>>16)&0xFF].i_bytes+=skb->len;
356 }
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
Mitsuru Chinen5506b542007-04-30 00:48:10 -0700362 rt = (struct rtable*)skb->dst;
363 if (rt->rt_type == RTN_MULTICAST)
364 IP_INC_STATS_BH(IPSTATS_MIB_INMCASTPKTS);
365 else if (rt->rt_type == RTN_BROADCAST)
366 IP_INC_STATS_BH(IPSTATS_MIB_INBCASTPKTS);
367
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{
380 struct iphdr *iph;
Thomas Graf58615242005-08-20 17:25:29 -0700381 u32 len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
Eric W. Biedermane730c152007-09-17 11:53:39 -0700383 if (dev->nd_net != &init_net)
384 goto drop;
385
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 /* When the interface is in promisc. mode, drop all the crap
387 * that it receives, do not try to analyse it.
388 */
389 if (skb->pkt_type == PACKET_OTHERHOST)
390 goto drop;
391
392 IP_INC_STATS_BH(IPSTATS_MIB_INRECEIVES);
393
394 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL) {
395 IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS);
396 goto out;
397 }
398
399 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
400 goto inhdr_error;
401
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700402 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
404 /*
405 * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the checksum.
406 *
407 * Is the datagram acceptable?
408 *
409 * 1. Length at least the size of an ip header
410 * 2. Version of 4
411 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
412 * 4. Doesn't have a bogus length
413 */
414
415 if (iph->ihl < 5 || iph->version != 4)
Thomas Graf58615242005-08-20 17:25:29 -0700416 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
418 if (!pskb_may_pull(skb, iph->ihl*4))
419 goto inhdr_error;
420
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700421 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
Thomas Grafe9c60422005-08-20 17:25:52 -0700423 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
Thomas Graf58615242005-08-20 17:25:29 -0700424 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Thomas Graf58615242005-08-20 17:25:29 -0700426 len = ntohs(iph->tot_len);
Mitsuru Chinen704aed52007-04-30 00:46:30 -0700427 if (skb->len < len) {
428 IP_INC_STATS_BH(IPSTATS_MIB_INTRUNCATEDPKTS);
429 goto drop;
430 } else if (len < (iph->ihl*4))
Thomas Graf58615242005-08-20 17:25:29 -0700431 goto inhdr_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
Thomas Graf58615242005-08-20 17:25:29 -0700433 /* Our transport medium may have padded the buffer out. Now we know it
434 * is IP we can trim to the true length of the frame.
435 * Note this now means skb->len holds ntohs(iph->tot_len).
436 */
437 if (pskb_trim_rcsum(skb, len)) {
438 IP_INC_STATS_BH(IPSTATS_MIB_INDISCARDS);
439 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 }
441
Stephen Hemminger53602f92006-07-14 14:49:32 -0700442 /* Remove any debris in the socket control block */
Guillaume Chazaraind569f1d2006-07-24 23:45:16 -0700443 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Stephen Hemminger53602f92006-07-14 14:49:32 -0700444
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 return NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, dev, NULL,
446 ip_rcv_finish);
447
448inhdr_error:
449 IP_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
450drop:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900451 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452out:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900453 return NET_RX_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454}
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456EXPORT_SYMBOL(ip_statistics);