blob: 988f52fba54a172bc1c49a692393f121314ba444 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002 * Linux NET3: IP/IP protocol decoder.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 * Authors:
5 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
6 *
7 * Fixes:
8 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
9 * a module taking up 2 pages).
10 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
11 * to keep ip_forward happy.
12 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
13 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
14 * David Woodhouse : Perform some basic ICMP handling.
15 * IPIP Routing without decapsulation.
16 * Carlos Picoto : GRE over IP support
17 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
18 * I do not want to merge them together.
19 *
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
24 *
25 */
26
27/* tunnel.c: an IP tunnel driver
28
29 The purpose of this driver is to provide an IP tunnel through
30 which you can tunnel network traffic transparently across subnets.
31
32 This was written by looking at Nick Holloway's dummy driver
33 Thanks for the great code!
34
35 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090036
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 Minor tweaks:
38 Cleaned up the code a little and added some pre-1.3.0 tweaks.
39 dev->hard_header/hard_header_len changed to use no headers.
40 Comments/bracketing tweaked.
41 Made the tunnels use dev->name not tunnel: when error reporting.
42 Added tx_dropped stat
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090043
Alan Cox113aa832008-10-13 19:01:08 -070044 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
46 Reworked:
47 Changed to tunnel to destination gateway in addition to the
48 tunnel's pointopoint address
49 Almost completely rewritten
50 Note: There is currently no firewall or ICMP handling done.
51
52 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090053
Linus Torvalds1da177e2005-04-16 15:20:36 -070054*/
55
56/* Things I wish I had known when writing the tunnel driver:
57
58 When the tunnel_xmit() function is called, the skb contains the
59 packet to be sent (plus a great deal of extra info), and dev
60 contains the tunnel device that _we_ are.
61
62 When we are passed a packet, we are expected to fill in the
63 source address with our source IP address.
64
65 What is the proper way to allocate, copy and free a buffer?
66 After you allocate it, it is a "0 length" chunk of memory
67 starting at zero. If you want to add headers to the buffer
68 later, you'll have to call "skb_reserve(skb, amount)" with
69 the amount of memory you want reserved. Then, you call
70 "skb_put(skb, amount)" with the amount of space you want in
71 the buffer. skb_put() returns a pointer to the top (#0) of
72 that buffer. skb->len is set to the amount of space you have
73 "allocated" with skb_put(). You can then write up to skb->len
74 bytes to that buffer. If you need more, you can call skb_put()
75 again with the additional amount of space you need. You can
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090076 find out how much more space you can allocate by calling
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 "skb_tailroom(skb)".
78 Now, to add header space, call "skb_push(skb, header_len)".
79 This creates space at the beginning of the buffer and returns
80 a pointer to this new space. If later you need to strip a
81 header from a buffer, call "skb_pull(skb, header_len)".
82 skb_headroom() will return how much space is left at the top
83 of the buffer (before the main data). Remember, this headroom
84 space must be reserved before the skb_put() function is called.
85 */
86
87/*
88 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
89
90 For comments look at net/ipv4/ip_gre.c --ANK
91 */
92
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090093
Randy Dunlap4fc268d2006-01-11 12:17:47 -080094#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/module.h>
96#include <linux/types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/kernel.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090098#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <asm/uaccess.h>
100#include <linux/skbuff.h>
101#include <linux/netdevice.h>
102#include <linux/in.h>
103#include <linux/tcp.h>
104#include <linux/udp.h>
105#include <linux/if_arp.h>
106#include <linux/mroute.h>
107#include <linux/init.h>
108#include <linux/netfilter_ipv4.h>
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800109#include <linux/if_ether.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
111#include <net/sock.h>
112#include <net/ip.h>
113#include <net/icmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#include <net/ipip.h>
115#include <net/inet_ecn.h>
116#include <net/xfrm.h>
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700117#include <net/net_namespace.h>
118#include <net/netns/generic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120#define HASH_SIZE 16
Al Virod5a0a1e2006-11-08 00:23:14 -0800121#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
Eric Dumazetf99189b2009-11-17 10:42:49 +0000123static int ipip_net_id __read_mostly;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700124struct ipip_net {
Eric Dumazetb7285b72010-09-15 11:07:24 +0000125 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
126 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
127 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
128 struct ip_tunnel __rcu *tunnels_wc[1];
129 struct ip_tunnel __rcu **tunnels[4];
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700130
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700131 struct net_device *fb_tunnel_dev;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700132};
133
Eric Dumazet3c97af92010-09-27 00:35:50 +0000134static int ipip_tunnel_init(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135static void ipip_tunnel_setup(struct net_device *dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000136static void ipip_dev_free(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000138/*
Eric Dumazetb7285b72010-09-15 11:07:24 +0000139 * Locking : hash tables are protected by RCU and RTNL
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000140 */
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000141
142#define for_each_ip_tunnel_rcu(start) \
143 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
Eric Dumazet3c97af92010-09-27 00:35:50 +0000145/* often modified stats are per cpu, other are shared (netdev->stats) */
146struct pcpu_tstats {
147 unsigned long rx_packets;
148 unsigned long rx_bytes;
149 unsigned long tx_packets;
150 unsigned long tx_bytes;
151};
152
153static struct net_device_stats *ipip_get_stats(struct net_device *dev)
154{
155 struct pcpu_tstats sum = { 0 };
156 int i;
157
158 for_each_possible_cpu(i) {
159 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
160
161 sum.rx_packets += tstats->rx_packets;
162 sum.rx_bytes += tstats->rx_bytes;
163 sum.tx_packets += tstats->tx_packets;
164 sum.tx_bytes += tstats->tx_bytes;
165 }
166 dev->stats.rx_packets = sum.rx_packets;
167 dev->stats.rx_bytes = sum.rx_bytes;
168 dev->stats.tx_packets = sum.tx_packets;
169 dev->stats.tx_bytes = sum.tx_bytes;
170 return &dev->stats;
171}
172
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700173static struct ip_tunnel * ipip_tunnel_lookup(struct net *net,
174 __be32 remote, __be32 local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175{
Eric Dumazetb7285b72010-09-15 11:07:24 +0000176 unsigned int h0 = HASH(remote);
177 unsigned int h1 = HASH(local);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 struct ip_tunnel *t;
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700179 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000181 for_each_ip_tunnel_rcu(ipn->tunnels_r_l[h0 ^ h1])
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 if (local == t->parms.iph.saddr &&
183 remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
184 return t;
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000185
186 for_each_ip_tunnel_rcu(ipn->tunnels_r[h0])
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
188 return t;
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000189
190 for_each_ip_tunnel_rcu(ipn->tunnels_l[h1])
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
192 return t;
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000193
194 t = rcu_dereference(ipn->tunnels_wc[0]);
195 if (t && (t->dev->flags&IFF_UP))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 return t;
197 return NULL;
198}
199
Eric Dumazetb7285b72010-09-15 11:07:24 +0000200static struct ip_tunnel __rcu **__ipip_bucket(struct ipip_net *ipn,
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700201 struct ip_tunnel_parm *parms)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
YOSHIFUJI Hideaki87d1a162007-04-24 20:44:47 +0900203 __be32 remote = parms->iph.daddr;
204 __be32 local = parms->iph.saddr;
Eric Dumazetb7285b72010-09-15 11:07:24 +0000205 unsigned int h = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 int prio = 0;
207
208 if (remote) {
209 prio |= 2;
210 h ^= HASH(remote);
211 }
212 if (local) {
213 prio |= 1;
214 h ^= HASH(local);
215 }
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700216 return &ipn->tunnels[prio][h];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217}
218
Eric Dumazetb7285b72010-09-15 11:07:24 +0000219static inline struct ip_tunnel __rcu **ipip_bucket(struct ipip_net *ipn,
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700220 struct ip_tunnel *t)
YOSHIFUJI Hideaki87d1a162007-04-24 20:44:47 +0900221{
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700222 return __ipip_bucket(ipn, &t->parms);
YOSHIFUJI Hideaki87d1a162007-04-24 20:44:47 +0900223}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700225static void ipip_tunnel_unlink(struct ipip_net *ipn, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226{
Eric Dumazetb7285b72010-09-15 11:07:24 +0000227 struct ip_tunnel __rcu **tp;
228 struct ip_tunnel *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Eric Dumazetb7285b72010-09-15 11:07:24 +0000230 for (tp = ipip_bucket(ipn, t);
231 (iter = rtnl_dereference(*tp)) != NULL;
232 tp = &iter->next) {
233 if (t == iter) {
234 rcu_assign_pointer(*tp, t->next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 break;
236 }
237 }
238}
239
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700240static void ipip_tunnel_link(struct ipip_net *ipn, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
Eric Dumazetb7285b72010-09-15 11:07:24 +0000242 struct ip_tunnel __rcu **tp = ipip_bucket(ipn, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
Eric Dumazetb7285b72010-09-15 11:07:24 +0000244 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000245 rcu_assign_pointer(*tp, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246}
247
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700248static struct ip_tunnel * ipip_tunnel_locate(struct net *net,
249 struct ip_tunnel_parm *parms, int create)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250{
Al Virod5a0a1e2006-11-08 00:23:14 -0800251 __be32 remote = parms->iph.daddr;
252 __be32 local = parms->iph.saddr;
Eric Dumazetb7285b72010-09-15 11:07:24 +0000253 struct ip_tunnel *t, *nt;
254 struct ip_tunnel __rcu **tp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 char name[IFNAMSIZ];
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700257 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
Eric Dumazetb7285b72010-09-15 11:07:24 +0000259 for (tp = __ipip_bucket(ipn, parms);
260 (t = rtnl_dereference(*tp)) != NULL;
261 tp = &t->next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
263 return t;
264 }
265 if (!create)
266 return NULL;
267
268 if (parms->name[0])
269 strlcpy(name, parms->name, IFNAMSIZ);
Pavel Emelyanov34cc7ba2008-02-23 20:19:20 -0800270 else
Eric Dumazet3c97af92010-09-27 00:35:50 +0000271 strcpy(name, "tunl%d");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
273 dev = alloc_netdev(sizeof(*t), name, ipip_tunnel_setup);
274 if (dev == NULL)
275 return NULL;
276
Pavel Emelyanov0a826402008-04-16 01:06:18 -0700277 dev_net_set(dev, net);
278
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800279 if (strchr(name, '%')) {
280 if (dev_alloc_name(dev, name) < 0)
281 goto failed_free;
282 }
283
Patrick McHardy2941a482006-01-08 22:05:26 -0800284 nt = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 nt->parms = *parms;
286
Eric Dumazet3c97af92010-09-27 00:35:50 +0000287 if (ipip_tunnel_init(dev) < 0)
288 goto failed_free;
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800289
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800290 if (register_netdevice(dev) < 0)
291 goto failed_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292
293 dev_hold(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700294 ipip_tunnel_link(ipn, nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 return nt;
296
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800297failed_free:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000298 ipip_dev_free(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 return NULL;
300}
301
Eric Dumazetb7285b72010-09-15 11:07:24 +0000302/* called with RTNL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303static void ipip_tunnel_uninit(struct net_device *dev)
304{
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700305 struct net *net = dev_net(dev);
306 struct ipip_net *ipn = net_generic(net, ipip_net_id);
307
Eric Dumazetb7285b72010-09-15 11:07:24 +0000308 if (dev == ipn->fb_tunnel_dev)
309 rcu_assign_pointer(ipn->tunnels_wc[0], NULL);
310 else
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700311 ipip_tunnel_unlink(ipn, netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 dev_put(dev);
313}
314
Herbert Xud2acc342006-03-28 01:12:13 -0800315static int ipip_err(struct sk_buff *skb, u32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Rami Rosen071f92d2008-05-21 17:47:54 -0700318/* All the routers (except for Linux) return only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 8 bytes of packet payload. It means, that precise relaying of
320 ICMP in the real Internet is absolutely infeasible.
321 */
Jianjun Kong5a5f3a82008-11-03 00:24:34 -0800322 struct iphdr *iph = (struct iphdr *)skb->data;
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300323 const int type = icmp_hdr(skb)->type;
324 const int code = icmp_hdr(skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 struct ip_tunnel *t;
Herbert Xud2acc342006-03-28 01:12:13 -0800326 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327
328 switch (type) {
329 default:
330 case ICMP_PARAMETERPROB:
Herbert Xud2acc342006-03-28 01:12:13 -0800331 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333 case ICMP_DEST_UNREACH:
334 switch (code) {
335 case ICMP_SR_FAILED:
336 case ICMP_PORT_UNREACH:
337 /* Impossible event. */
Herbert Xud2acc342006-03-28 01:12:13 -0800338 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 case ICMP_FRAG_NEEDED:
340 /* Soft state for pmtu is maintained by IP core. */
Herbert Xud2acc342006-03-28 01:12:13 -0800341 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 default:
343 /* All others are translated to HOST_UNREACH.
344 rfc2003 contains "deep thoughts" about NET_UNREACH,
345 I believe they are just ether pollution. --ANK
346 */
347 break;
348 }
349 break;
350 case ICMP_TIME_EXCEEDED:
351 if (code != ICMP_EXC_TTL)
Herbert Xud2acc342006-03-28 01:12:13 -0800352 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 break;
354 }
355
Herbert Xud2acc342006-03-28 01:12:13 -0800356 err = -ENOENT;
357
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000358 rcu_read_lock();
Pavel Emelyanovcec3ffa2008-04-16 01:05:03 -0700359 t = ipip_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 if (t == NULL || t->parms.iph.daddr == 0)
361 goto out;
Herbert Xud2acc342006-03-28 01:12:13 -0800362
363 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
365 goto out;
366
Wei Yongjun26d94b42009-02-24 23:36:47 -0800367 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 t->err_count++;
369 else
370 t->err_count = 1;
371 t->err_time = jiffies;
372out:
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000373 rcu_read_unlock();
Herbert Xud2acc342006-03-28 01:12:13 -0800374 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375}
376
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700377static inline void ipip_ecn_decapsulate(const struct iphdr *outer_iph,
378 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700380 struct iphdr *inner_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
382 if (INET_ECN_is_ce(outer_iph->tos))
383 IP_ECN_set_ce(inner_iph);
384}
385
386static int ipip_rcv(struct sk_buff *skb)
387{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 struct ip_tunnel *tunnel;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700389 const struct iphdr *iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000391 rcu_read_lock();
Eric Dumazet3c97af92010-09-27 00:35:50 +0000392 tunnel = ipip_tunnel_lookup(dev_net(skb->dev), iph->saddr, iph->daddr);
393 if (tunnel != NULL) {
394 struct pcpu_tstats *tstats;
395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000397 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 kfree_skb(skb);
399 return 0;
400 }
401
402 secpath_reset(skb);
403
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700404 skb->mac_header = skb->network_header;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700405 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 skb->protocol = htons(ETH_P_IP);
407 skb->pkt_type = PACKET_HOST;
408
Eric Dumazet3c97af92010-09-27 00:35:50 +0000409 tstats = this_cpu_ptr(tunnel->dev->tstats);
410 tstats->rx_packets++;
411 tstats->rx_bytes += skb->len;
412
413 __skb_tunnel_rx(skb, tunnel->dev);
Eric Dumazetd19d56d2010-05-17 22:36:55 -0700414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 ipip_ecn_decapsulate(iph, skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000416
Eric Dumazetcaf586e2010-09-30 21:06:55 +0000417 netif_rx(skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000418
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000419 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 return 0;
421 }
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000422 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 return -1;
425}
426
427/*
428 * This function assumes it is being called from dev_queue_xmit()
429 * and that skb is filled properly by that function.
430 */
431
Stephen Hemminger6fef4c02009-08-31 19:50:41 +0000432static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
Patrick McHardy2941a482006-01-08 22:05:26 -0800434 struct ip_tunnel *tunnel = netdev_priv(dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000435 struct pcpu_tstats *tstats;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct iphdr *tiph = &tunnel->parms.iph;
437 u8 tos = tunnel->parms.iph.tos;
Al Virod5a0a1e2006-11-08 00:23:14 -0800438 __be16 df = tiph->frag_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 struct rtable *rt; /* Route to the other host */
Eric Dumazet3c97af92010-09-27 00:35:50 +0000440 struct net_device *tdev; /* Device to other host */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700441 struct iphdr *old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 struct iphdr *iph; /* Our new IP header */
Chuck Leverc2636b42007-10-23 21:07:32 -0700443 unsigned int max_headroom; /* The extra header space needed */
Al Virod5a0a1e2006-11-08 00:23:14 -0800444 __be32 dst = tiph->daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 int mtu;
446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 if (skb->protocol != htons(ETH_P_IP))
448 goto tx_error;
449
Eric Dumazet3c97af92010-09-27 00:35:50 +0000450 if (tos & 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 tos = old_iph->tos;
452
453 if (!dst) {
454 /* NBMA tunnel */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000455 if ((rt = skb_rtable(skb)) == NULL) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000456 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 goto tx_error;
458 }
459 if ((dst = rt->rt_gateway) == 0)
460 goto tx_error_icmp;
461 }
462
463 {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000464 struct flowi fl = {
465 .oif = tunnel->parms.link,
Changli Gao58116622010-11-12 18:43:55 +0000466 .fl4_dst = dst,
467 .fl4_src= tiph->saddr,
468 .fl4_tos = RT_TOS(tos),
Eric Dumazet3c97af92010-09-27 00:35:50 +0000469 .proto = IPPROTO_IPIP
470 };
471
Pavel Emelyanovb99f0152008-04-16 01:05:57 -0700472 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000473 dev->stats.tx_carrier_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto tx_error_icmp;
475 }
476 }
Changli Gaod8d1f302010-06-10 23:31:35 -0700477 tdev = rt->dst.dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
479 if (tdev == dev) {
480 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000481 dev->stats.collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 goto tx_error;
483 }
484
Herbert Xu23ca0c92009-11-06 10:37:41 +0000485 df |= old_iph->frag_off & htons(IP_DF);
486
487 if (df) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700488 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
Herbert Xu23ca0c92009-11-06 10:37:41 +0000490 if (mtu < 68) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000491 dev->stats.collisions++;
Herbert Xu23ca0c92009-11-06 10:37:41 +0000492 ip_rt_put(rt);
493 goto tx_error;
494 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Herbert Xu23ca0c92009-11-06 10:37:41 +0000496 if (skb_dst(skb))
497 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Herbert Xu23ca0c92009-11-06 10:37:41 +0000499 if ((old_iph->frag_off & htons(IP_DF)) &&
500 mtu < ntohs(old_iph->tot_len)) {
501 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
502 htonl(mtu));
503 ip_rt_put(rt);
504 goto tx_error;
505 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 }
507
508 if (tunnel->err_count > 0) {
Wei Yongjun26d94b42009-02-24 23:36:47 -0800509 if (time_before(jiffies,
510 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 tunnel->err_count--;
512 dst_link_failure(skb);
513 } else
514 tunnel->err_count = 0;
515 }
516
517 /*
518 * Okay, now see if we can stuff it in the buffer as-is.
519 */
520 max_headroom = (LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr));
521
Patrick McHardycfbba492007-07-09 15:33:40 -0700522 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
523 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
525 if (!new_skb) {
526 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000527 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000529 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 }
531 if (skb->sk)
532 skb_set_owner_w(new_skb, skb->sk);
533 dev_kfree_skb(skb);
534 skb = new_skb;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700535 old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 }
537
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700538 skb->transport_header = skb->network_header;
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700539 skb_push(skb, sizeof(struct iphdr));
540 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy48d5cad2006-02-15 15:10:22 -0800542 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
543 IPSKB_REROUTED);
Eric Dumazetadf30902009-06-02 05:19:30 +0000544 skb_dst_drop(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700545 skb_dst_set(skb, &rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
547 /*
548 * Push down and install the IPIP header.
549 */
550
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700551 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 iph->version = 4;
553 iph->ihl = sizeof(struct iphdr)>>2;
554 iph->frag_off = df;
555 iph->protocol = IPPROTO_IPIP;
556 iph->tos = INET_ECN_encapsulate(tos, old_iph->tos);
557 iph->daddr = rt->rt_dst;
558 iph->saddr = rt->rt_src;
559
560 if ((iph->ttl = tiph->ttl) == 0)
561 iph->ttl = old_iph->ttl;
562
563 nf_reset(skb);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000564 tstats = this_cpu_ptr(dev->tstats);
565 __IPTUNNEL_XMIT(tstats, &dev->stats);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000566 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
568tx_error_icmp:
569 dst_link_failure(skb);
570tx_error:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000571 dev->stats.tx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000573 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574}
575
Michal Schmidt55339952007-12-12 11:01:43 -0800576static void ipip_tunnel_bind_dev(struct net_device *dev)
577{
578 struct net_device *tdev = NULL;
579 struct ip_tunnel *tunnel;
580 struct iphdr *iph;
581
582 tunnel = netdev_priv(dev);
583 iph = &tunnel->parms.iph;
584
585 if (iph->daddr) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000586 struct flowi fl = {
587 .oif = tunnel->parms.link,
Changli Gao58116622010-11-12 18:43:55 +0000588 .fl4_dst = iph->daddr,
589 .fl4_src = iph->saddr,
590 .fl4_tos = RT_TOS(iph->tos),
Eric Dumazet3c97af92010-09-27 00:35:50 +0000591 .proto = IPPROTO_IPIP
592 };
Michal Schmidt55339952007-12-12 11:01:43 -0800593 struct rtable *rt;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000594
Pavel Emelyanovb99f0152008-04-16 01:05:57 -0700595 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700596 tdev = rt->dst.dev;
Michal Schmidt55339952007-12-12 11:01:43 -0800597 ip_rt_put(rt);
598 }
599 dev->flags |= IFF_POINTOPOINT;
600 }
601
602 if (!tdev && tunnel->parms.link)
Pavel Emelyanovb99f0152008-04-16 01:05:57 -0700603 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
Michal Schmidt55339952007-12-12 11:01:43 -0800604
605 if (tdev) {
606 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
607 dev->mtu = tdev->mtu - sizeof(struct iphdr);
608 }
609 dev->iflink = tunnel->parms.link;
610}
611
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612static int
613ipip_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
614{
615 int err = 0;
616 struct ip_tunnel_parm p;
617 struct ip_tunnel *t;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700618 struct net *net = dev_net(dev);
619 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
621 switch (cmd) {
622 case SIOCGETTUNNEL:
623 t = NULL;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700624 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
626 err = -EFAULT;
627 break;
628 }
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700629 t = ipip_tunnel_locate(net, &p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 }
631 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -0800632 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 memcpy(&p, &t->parms, sizeof(p));
634 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
635 err = -EFAULT;
636 break;
637
638 case SIOCADDTUNNEL:
639 case SIOCCHGTUNNEL:
640 err = -EPERM;
641 if (!capable(CAP_NET_ADMIN))
642 goto done;
643
644 err = -EFAULT;
645 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
646 goto done;
647
648 err = -EINVAL;
649 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
650 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
651 goto done;
652 if (p.iph.ttl)
653 p.iph.frag_off |= htons(IP_DF);
654
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700655 t = ipip_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700657 if (dev != ipn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 if (t != NULL) {
659 if (t->dev != dev) {
660 err = -EEXIST;
661 break;
662 }
663 } else {
664 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
665 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
666 err = -EINVAL;
667 break;
668 }
Patrick McHardy2941a482006-01-08 22:05:26 -0800669 t = netdev_priv(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700670 ipip_tunnel_unlink(ipn, t);
Pavel Emelyanov74b0b852010-10-27 05:43:53 +0000671 synchronize_net();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 t->parms.iph.saddr = p.iph.saddr;
673 t->parms.iph.daddr = p.iph.daddr;
674 memcpy(dev->dev_addr, &p.iph.saddr, 4);
675 memcpy(dev->broadcast, &p.iph.daddr, 4);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700676 ipip_tunnel_link(ipn, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 netdev_state_change(dev);
678 }
679 }
680
681 if (t) {
682 err = 0;
683 if (cmd == SIOCCHGTUNNEL) {
684 t->parms.iph.ttl = p.iph.ttl;
685 t->parms.iph.tos = p.iph.tos;
686 t->parms.iph.frag_off = p.iph.frag_off;
Michal Schmidt55339952007-12-12 11:01:43 -0800687 if (t->parms.link != p.link) {
688 t->parms.link = p.link;
689 ipip_tunnel_bind_dev(dev);
690 netdev_state_change(dev);
691 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 }
693 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
694 err = -EFAULT;
695 } else
696 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
697 break;
698
699 case SIOCDELTUNNEL:
700 err = -EPERM;
701 if (!capable(CAP_NET_ADMIN))
702 goto done;
703
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700704 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 err = -EFAULT;
706 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
707 goto done;
708 err = -ENOENT;
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700709 if ((t = ipip_tunnel_locate(net, &p, 0)) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 goto done;
711 err = -EPERM;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700712 if (t->dev == ipn->fb_tunnel_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 goto done;
714 dev = t->dev;
715 }
Stephen Hemminger22f8cde2007-02-07 00:09:58 -0800716 unregister_netdevice(dev);
717 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 break;
719
720 default:
721 err = -EINVAL;
722 }
723
724done:
725 return err;
726}
727
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728static int ipip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
729{
730 if (new_mtu < 68 || new_mtu > 0xFFF8 - sizeof(struct iphdr))
731 return -EINVAL;
732 dev->mtu = new_mtu;
733 return 0;
734}
735
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800736static const struct net_device_ops ipip_netdev_ops = {
737 .ndo_uninit = ipip_tunnel_uninit,
738 .ndo_start_xmit = ipip_tunnel_xmit,
739 .ndo_do_ioctl = ipip_tunnel_ioctl,
740 .ndo_change_mtu = ipip_tunnel_change_mtu,
Eric Dumazet3c97af92010-09-27 00:35:50 +0000741 .ndo_get_stats = ipip_get_stats,
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800742};
743
Eric Dumazet3c97af92010-09-27 00:35:50 +0000744static void ipip_dev_free(struct net_device *dev)
745{
746 free_percpu(dev->tstats);
747 free_netdev(dev);
748}
749
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750static void ipip_tunnel_setup(struct net_device *dev)
751{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800752 dev->netdev_ops = &ipip_netdev_ops;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000753 dev->destructor = ipip_dev_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
755 dev->type = ARPHRD_TUNNEL;
756 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800757 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 dev->flags = IFF_NOARP;
759 dev->iflink = 0;
760 dev->addr_len = 4;
Pavel Emelyanov0a826402008-04-16 01:06:18 -0700761 dev->features |= NETIF_F_NETNS_LOCAL;
Eric Dumazet153f0942010-09-28 00:17:17 +0000762 dev->features |= NETIF_F_LLTX;
Eric Dumazet28e72212009-05-28 10:44:30 +0000763 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764}
765
Eric Dumazet3c97af92010-09-27 00:35:50 +0000766static int ipip_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800768 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
770 tunnel->dev = dev;
771 strcpy(tunnel->parms.name, dev->name);
772
773 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
774 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
775
Michal Schmidt55339952007-12-12 11:01:43 -0800776 ipip_tunnel_bind_dev(dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000777
778 dev->tstats = alloc_percpu(struct pcpu_tstats);
779 if (!dev->tstats)
780 return -ENOMEM;
781
782 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783}
784
Eric Dumazetfada5632010-09-27 23:56:46 +0000785static int __net_init ipip_fb_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786{
Patrick McHardy2941a482006-01-08 22:05:26 -0800787 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 struct iphdr *iph = &tunnel->parms.iph;
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700789 struct ipip_net *ipn = net_generic(dev_net(dev), ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
791 tunnel->dev = dev;
792 strcpy(tunnel->parms.name, dev->name);
793
794 iph->version = 4;
795 iph->protocol = IPPROTO_IPIP;
796 iph->ihl = 5;
797
Eric Dumazetfada5632010-09-27 23:56:46 +0000798 dev->tstats = alloc_percpu(struct pcpu_tstats);
799 if (!dev->tstats)
800 return -ENOMEM;
801
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 dev_hold(dev);
Eric Dumazetb7285b72010-09-15 11:07:24 +0000803 rcu_assign_pointer(ipn->tunnels_wc[0], tunnel);
Eric Dumazetfada5632010-09-27 23:56:46 +0000804 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805}
806
Eric Dumazet6dcd8142010-08-30 07:04:14 +0000807static struct xfrm_tunnel ipip_handler __read_mostly = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 .handler = ipip_rcv,
809 .err_handler = ipip_err,
Herbert Xud2acc342006-03-28 01:12:13 -0800810 .priority = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811};
812
Stephen Hemminger5747a1a2009-02-22 00:02:08 -0800813static const char banner[] __initconst =
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 KERN_INFO "IPv4 over IPv4 tunneling driver\n";
815
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000816static void ipip_destroy_tunnels(struct ipip_net *ipn, struct list_head *head)
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700817{
818 int prio;
819
820 for (prio = 1; prio < 4; prio++) {
821 int h;
822 for (h = 0; h < HASH_SIZE; h++) {
Eric Dumazetb7285b72010-09-15 11:07:24 +0000823 struct ip_tunnel *t;
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000824
Eric Dumazetb7285b72010-09-15 11:07:24 +0000825 t = rtnl_dereference(ipn->tunnels[prio][h]);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000826 while (t != NULL) {
827 unregister_netdevice_queue(t->dev, head);
Eric Dumazetb7285b72010-09-15 11:07:24 +0000828 t = rtnl_dereference(t->next);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000829 }
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700830 }
831 }
832}
833
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000834static int __net_init ipip_init_net(struct net *net)
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700835{
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000836 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700837 int err;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700838
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700839 ipn->tunnels[0] = ipn->tunnels_wc;
840 ipn->tunnels[1] = ipn->tunnels_l;
841 ipn->tunnels[2] = ipn->tunnels_r;
842 ipn->tunnels[3] = ipn->tunnels_r_l;
843
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700844 ipn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
845 "tunl0",
846 ipip_tunnel_setup);
847 if (!ipn->fb_tunnel_dev) {
848 err = -ENOMEM;
849 goto err_alloc_dev;
850 }
Alexey Dobriyanbe77e592008-11-23 17:26:26 -0800851 dev_net_set(ipn->fb_tunnel_dev, net);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700852
Eric Dumazetfada5632010-09-27 23:56:46 +0000853 err = ipip_fb_tunnel_init(ipn->fb_tunnel_dev);
854 if (err)
855 goto err_reg_dev;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700856
857 if ((err = register_netdev(ipn->fb_tunnel_dev)))
858 goto err_reg_dev;
859
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700860 return 0;
861
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700862err_reg_dev:
Eric Dumazetfada5632010-09-27 23:56:46 +0000863 ipip_dev_free(ipn->fb_tunnel_dev);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700864err_alloc_dev:
865 /* nothing */
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700866 return err;
867}
868
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000869static void __net_exit ipip_exit_net(struct net *net)
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700870{
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000871 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000872 LIST_HEAD(list);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700873
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700874 rtnl_lock();
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000875 ipip_destroy_tunnels(ipn, &list);
876 unregister_netdevice_queue(ipn->fb_tunnel_dev, &list);
877 unregister_netdevice_many(&list);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700878 rtnl_unlock();
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700879}
880
881static struct pernet_operations ipip_net_ops = {
882 .init = ipip_init_net,
883 .exit = ipip_exit_net,
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000884 .id = &ipip_net_id,
885 .size = sizeof(struct ipip_net),
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700886};
887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888static int __init ipip_init(void)
889{
890 int err;
891
892 printk(banner);
893
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000894 err = register_pernet_device(&ipip_net_ops);
Alexey Dobriyand5aa4072010-02-16 09:05:04 +0000895 if (err < 0)
896 return err;
897 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
898 if (err < 0) {
899 unregister_pernet_device(&ipip_net_ops);
900 printk(KERN_INFO "ipip init: can't register tunnel\n");
901 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903}
904
905static void __exit ipip_fini(void)
906{
Kazunori MIYAZAWAc0d56402007-02-13 12:54:47 -0800907 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 printk(KERN_INFO "ipip close: can't deregister tunnel\n");
909
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000910 unregister_pernet_device(&ipip_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911}
912
913module_init(ipip_init);
914module_exit(ipip_fini);
915MODULE_LICENSE("GPL");
stephen hemminger8afe7c82010-11-29 09:47:49 +0000916MODULE_ALIAS("tunl0");