blob: 065effd8349a81689828927c84142a778aebb650 [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) {
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000234 RCU_INIT_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
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000244 RCU_INIT_POINTER(t->next, rtnl_dereference(*tp));
245 RCU_INIT_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
Patrick McHardy2941a482006-01-08 22:05:26 -0800279 nt = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 nt->parms = *parms;
281
Eric Dumazet3c97af92010-09-27 00:35:50 +0000282 if (ipip_tunnel_init(dev) < 0)
283 goto failed_free;
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800284
Pavel Emelyanovb37d4282008-02-26 23:51:04 -0800285 if (register_netdevice(dev) < 0)
286 goto failed_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
288 dev_hold(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700289 ipip_tunnel_link(ipn, nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 return nt;
291
Pavel Emelyanovb37d4282008-02-26 23:51:04 -0800292failed_free:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000293 ipip_dev_free(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 return NULL;
295}
296
Eric Dumazetb7285b72010-09-15 11:07:24 +0000297/* called with RTNL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298static void ipip_tunnel_uninit(struct net_device *dev)
299{
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700300 struct net *net = dev_net(dev);
301 struct ipip_net *ipn = net_generic(net, ipip_net_id);
302
Eric Dumazetb7285b72010-09-15 11:07:24 +0000303 if (dev == ipn->fb_tunnel_dev)
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000304 RCU_INIT_POINTER(ipn->tunnels_wc[0], NULL);
Eric Dumazetb7285b72010-09-15 11:07:24 +0000305 else
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700306 ipip_tunnel_unlink(ipn, netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 dev_put(dev);
308}
309
Herbert Xud2acc342006-03-28 01:12:13 -0800310static int ipip_err(struct sk_buff *skb, u32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Rami Rosen071f92d2008-05-21 17:47:54 -0700313/* All the routers (except for Linux) return only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 8 bytes of packet payload. It means, that precise relaying of
315 ICMP in the real Internet is absolutely infeasible.
316 */
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000317 const struct iphdr *iph = (const struct iphdr *)skb->data;
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300318 const int type = icmp_hdr(skb)->type;
319 const int code = icmp_hdr(skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 struct ip_tunnel *t;
Herbert Xud2acc342006-03-28 01:12:13 -0800321 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
323 switch (type) {
324 default:
325 case ICMP_PARAMETERPROB:
Herbert Xud2acc342006-03-28 01:12:13 -0800326 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327
328 case ICMP_DEST_UNREACH:
329 switch (code) {
330 case ICMP_SR_FAILED:
331 case ICMP_PORT_UNREACH:
332 /* Impossible event. */
Herbert Xud2acc342006-03-28 01:12:13 -0800333 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 case ICMP_FRAG_NEEDED:
335 /* Soft state for pmtu is maintained by IP core. */
Herbert Xud2acc342006-03-28 01:12:13 -0800336 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 default:
338 /* All others are translated to HOST_UNREACH.
339 rfc2003 contains "deep thoughts" about NET_UNREACH,
340 I believe they are just ether pollution. --ANK
341 */
342 break;
343 }
344 break;
345 case ICMP_TIME_EXCEEDED:
346 if (code != ICMP_EXC_TTL)
Herbert Xud2acc342006-03-28 01:12:13 -0800347 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 break;
349 }
350
Herbert Xud2acc342006-03-28 01:12:13 -0800351 err = -ENOENT;
352
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000353 rcu_read_lock();
Pavel Emelyanovcec3ffa2008-04-16 01:05:03 -0700354 t = ipip_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 if (t == NULL || t->parms.iph.daddr == 0)
356 goto out;
Herbert Xud2acc342006-03-28 01:12:13 -0800357
358 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
360 goto out;
361
Wei Yongjun26d94b42009-02-24 23:36:47 -0800362 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 t->err_count++;
364 else
365 t->err_count = 1;
366 t->err_time = jiffies;
367out:
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000368 rcu_read_unlock();
Herbert Xud2acc342006-03-28 01:12:13 -0800369 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370}
371
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700372static inline void ipip_ecn_decapsulate(const struct iphdr *outer_iph,
373 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700375 struct iphdr *inner_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
377 if (INET_ECN_is_ce(outer_iph->tos))
378 IP_ECN_set_ce(inner_iph);
379}
380
381static int ipip_rcv(struct sk_buff *skb)
382{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 struct ip_tunnel *tunnel;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700384 const struct iphdr *iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000386 rcu_read_lock();
Eric Dumazet3c97af92010-09-27 00:35:50 +0000387 tunnel = ipip_tunnel_lookup(dev_net(skb->dev), iph->saddr, iph->daddr);
388 if (tunnel != NULL) {
389 struct pcpu_tstats *tstats;
390
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000392 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 kfree_skb(skb);
394 return 0;
395 }
396
397 secpath_reset(skb);
398
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700399 skb->mac_header = skb->network_header;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700400 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 skb->protocol = htons(ETH_P_IP);
402 skb->pkt_type = PACKET_HOST;
403
Eric Dumazet3c97af92010-09-27 00:35:50 +0000404 tstats = this_cpu_ptr(tunnel->dev->tstats);
405 tstats->rx_packets++;
406 tstats->rx_bytes += skb->len;
407
408 __skb_tunnel_rx(skb, tunnel->dev);
Eric Dumazetd19d56d2010-05-17 22:36:55 -0700409
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 ipip_ecn_decapsulate(iph, skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000411
Eric Dumazetcaf586e2010-09-30 21:06:55 +0000412 netif_rx(skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000413
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000414 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 return 0;
416 }
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000417 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 return -1;
420}
421
422/*
423 * This function assumes it is being called from dev_queue_xmit()
424 * and that skb is filled properly by that function.
425 */
426
Stephen Hemminger6fef4c02009-08-31 19:50:41 +0000427static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428{
Patrick McHardy2941a482006-01-08 22:05:26 -0800429 struct ip_tunnel *tunnel = netdev_priv(dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000430 struct pcpu_tstats *tstats;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000431 const struct iphdr *tiph = &tunnel->parms.iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 u8 tos = tunnel->parms.iph.tos;
Al Virod5a0a1e2006-11-08 00:23:14 -0800433 __be16 df = tiph->frag_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 struct rtable *rt; /* Route to the other host */
Eric Dumazet3c97af92010-09-27 00:35:50 +0000435 struct net_device *tdev; /* Device to other host */
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000436 const struct iphdr *old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 struct iphdr *iph; /* Our new IP header */
Chuck Leverc2636b42007-10-23 21:07:32 -0700438 unsigned int max_headroom; /* The extra header space needed */
Al Virod5a0a1e2006-11-08 00:23:14 -0800439 __be32 dst = tiph->daddr;
David S. Miller31e45432011-05-03 20:25:42 -0700440 struct flowi4 fl4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 int mtu;
442
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 if (skb->protocol != htons(ETH_P_IP))
444 goto tx_error;
445
Eric Dumazet3c97af92010-09-27 00:35:50 +0000446 if (tos & 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 tos = old_iph->tos;
448
449 if (!dst) {
450 /* NBMA tunnel */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000451 if ((rt = skb_rtable(skb)) == NULL) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000452 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 goto tx_error;
454 }
455 if ((dst = rt->rt_gateway) == 0)
456 goto tx_error_icmp;
457 }
458
David S. Miller31e45432011-05-03 20:25:42 -0700459 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
David S. Miller78fbfd82011-03-12 00:00:52 -0500460 dst, tiph->saddr,
461 0, 0,
462 IPPROTO_IPIP, RT_TOS(tos),
463 tunnel->parms.link);
464 if (IS_ERR(rt)) {
465 dev->stats.tx_carrier_errors++;
466 goto tx_error_icmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 }
Changli Gaod8d1f302010-06-10 23:31:35 -0700468 tdev = rt->dst.dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470 if (tdev == dev) {
471 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000472 dev->stats.collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 goto tx_error;
474 }
475
Herbert Xu23ca0c92009-11-06 10:37:41 +0000476 df |= old_iph->frag_off & htons(IP_DF);
477
478 if (df) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700479 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
Herbert Xu23ca0c92009-11-06 10:37:41 +0000481 if (mtu < 68) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000482 dev->stats.collisions++;
Herbert Xu23ca0c92009-11-06 10:37:41 +0000483 ip_rt_put(rt);
484 goto tx_error;
485 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Herbert Xu23ca0c92009-11-06 10:37:41 +0000487 if (skb_dst(skb))
488 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489
Herbert Xu23ca0c92009-11-06 10:37:41 +0000490 if ((old_iph->frag_off & htons(IP_DF)) &&
491 mtu < ntohs(old_iph->tot_len)) {
492 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
493 htonl(mtu));
494 ip_rt_put(rt);
495 goto tx_error;
496 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 }
498
499 if (tunnel->err_count > 0) {
Wei Yongjun26d94b42009-02-24 23:36:47 -0800500 if (time_before(jiffies,
501 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 tunnel->err_count--;
503 dst_link_failure(skb);
504 } else
505 tunnel->err_count = 0;
506 }
507
508 /*
509 * Okay, now see if we can stuff it in the buffer as-is.
510 */
511 max_headroom = (LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr));
512
Patrick McHardycfbba492007-07-09 15:33:40 -0700513 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
514 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
516 if (!new_skb) {
517 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000518 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000520 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 }
522 if (skb->sk)
523 skb_set_owner_w(new_skb, skb->sk);
524 dev_kfree_skb(skb);
525 skb = new_skb;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700526 old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 }
528
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700529 skb->transport_header = skb->network_header;
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700530 skb_push(skb, sizeof(struct iphdr));
531 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy48d5cad2006-02-15 15:10:22 -0800533 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
534 IPSKB_REROUTED);
Eric Dumazetadf30902009-06-02 05:19:30 +0000535 skb_dst_drop(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700536 skb_dst_set(skb, &rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
538 /*
539 * Push down and install the IPIP header.
540 */
541
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700542 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 iph->version = 4;
544 iph->ihl = sizeof(struct iphdr)>>2;
545 iph->frag_off = df;
546 iph->protocol = IPPROTO_IPIP;
547 iph->tos = INET_ECN_encapsulate(tos, old_iph->tos);
David S. Miller69458cb2011-05-04 11:10:28 -0700548 iph->daddr = fl4.daddr;
549 iph->saddr = fl4.saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550
551 if ((iph->ttl = tiph->ttl) == 0)
552 iph->ttl = old_iph->ttl;
553
554 nf_reset(skb);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000555 tstats = this_cpu_ptr(dev->tstats);
556 __IPTUNNEL_XMIT(tstats, &dev->stats);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000557 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
559tx_error_icmp:
560 dst_link_failure(skb);
561tx_error:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000562 dev->stats.tx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000564 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565}
566
Michal Schmidt55339952007-12-12 11:01:43 -0800567static void ipip_tunnel_bind_dev(struct net_device *dev)
568{
569 struct net_device *tdev = NULL;
570 struct ip_tunnel *tunnel;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000571 const struct iphdr *iph;
Michal Schmidt55339952007-12-12 11:01:43 -0800572
573 tunnel = netdev_priv(dev);
574 iph = &tunnel->parms.iph;
575
576 if (iph->daddr) {
David S. Miller31e45432011-05-03 20:25:42 -0700577 struct rtable *rt;
578 struct flowi4 fl4;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000579
David S. Miller31e45432011-05-03 20:25:42 -0700580 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
581 iph->daddr, iph->saddr,
582 0, 0,
583 IPPROTO_IPIP,
584 RT_TOS(iph->tos),
585 tunnel->parms.link);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800586 if (!IS_ERR(rt)) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700587 tdev = rt->dst.dev;
Michal Schmidt55339952007-12-12 11:01:43 -0800588 ip_rt_put(rt);
589 }
590 dev->flags |= IFF_POINTOPOINT;
591 }
592
593 if (!tdev && tunnel->parms.link)
Pavel Emelyanovb99f0152008-04-16 01:05:57 -0700594 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
Michal Schmidt55339952007-12-12 11:01:43 -0800595
596 if (tdev) {
597 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
598 dev->mtu = tdev->mtu - sizeof(struct iphdr);
599 }
600 dev->iflink = tunnel->parms.link;
601}
602
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603static int
604ipip_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
605{
606 int err = 0;
607 struct ip_tunnel_parm p;
608 struct ip_tunnel *t;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700609 struct net *net = dev_net(dev);
610 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
612 switch (cmd) {
613 case SIOCGETTUNNEL:
614 t = NULL;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700615 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
617 err = -EFAULT;
618 break;
619 }
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700620 t = ipip_tunnel_locate(net, &p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 }
622 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -0800623 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 memcpy(&p, &t->parms, sizeof(p));
625 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
626 err = -EFAULT;
627 break;
628
629 case SIOCADDTUNNEL:
630 case SIOCCHGTUNNEL:
631 err = -EPERM;
632 if (!capable(CAP_NET_ADMIN))
633 goto done;
634
635 err = -EFAULT;
636 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
637 goto done;
638
639 err = -EINVAL;
640 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
641 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
642 goto done;
643 if (p.iph.ttl)
644 p.iph.frag_off |= htons(IP_DF);
645
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700646 t = ipip_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700648 if (dev != ipn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 if (t != NULL) {
650 if (t->dev != dev) {
651 err = -EEXIST;
652 break;
653 }
654 } else {
655 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
656 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
657 err = -EINVAL;
658 break;
659 }
Patrick McHardy2941a482006-01-08 22:05:26 -0800660 t = netdev_priv(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700661 ipip_tunnel_unlink(ipn, t);
Pavel Emelyanov74b0b852010-10-27 05:43:53 +0000662 synchronize_net();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 t->parms.iph.saddr = p.iph.saddr;
664 t->parms.iph.daddr = p.iph.daddr;
665 memcpy(dev->dev_addr, &p.iph.saddr, 4);
666 memcpy(dev->broadcast, &p.iph.daddr, 4);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700667 ipip_tunnel_link(ipn, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 netdev_state_change(dev);
669 }
670 }
671
672 if (t) {
673 err = 0;
674 if (cmd == SIOCCHGTUNNEL) {
675 t->parms.iph.ttl = p.iph.ttl;
676 t->parms.iph.tos = p.iph.tos;
677 t->parms.iph.frag_off = p.iph.frag_off;
Michal Schmidt55339952007-12-12 11:01:43 -0800678 if (t->parms.link != p.link) {
679 t->parms.link = p.link;
680 ipip_tunnel_bind_dev(dev);
681 netdev_state_change(dev);
682 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 }
684 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
685 err = -EFAULT;
686 } else
687 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
688 break;
689
690 case SIOCDELTUNNEL:
691 err = -EPERM;
692 if (!capable(CAP_NET_ADMIN))
693 goto done;
694
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700695 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 err = -EFAULT;
697 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
698 goto done;
699 err = -ENOENT;
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700700 if ((t = ipip_tunnel_locate(net, &p, 0)) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 goto done;
702 err = -EPERM;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700703 if (t->dev == ipn->fb_tunnel_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 goto done;
705 dev = t->dev;
706 }
Stephen Hemminger22f8cde2007-02-07 00:09:58 -0800707 unregister_netdevice(dev);
708 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 break;
710
711 default:
712 err = -EINVAL;
713 }
714
715done:
716 return err;
717}
718
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719static int ipip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
720{
721 if (new_mtu < 68 || new_mtu > 0xFFF8 - sizeof(struct iphdr))
722 return -EINVAL;
723 dev->mtu = new_mtu;
724 return 0;
725}
726
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800727static const struct net_device_ops ipip_netdev_ops = {
728 .ndo_uninit = ipip_tunnel_uninit,
729 .ndo_start_xmit = ipip_tunnel_xmit,
730 .ndo_do_ioctl = ipip_tunnel_ioctl,
731 .ndo_change_mtu = ipip_tunnel_change_mtu,
Eric Dumazet3c97af92010-09-27 00:35:50 +0000732 .ndo_get_stats = ipip_get_stats,
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800733};
734
Eric Dumazet3c97af92010-09-27 00:35:50 +0000735static void ipip_dev_free(struct net_device *dev)
736{
737 free_percpu(dev->tstats);
738 free_netdev(dev);
739}
740
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741static void ipip_tunnel_setup(struct net_device *dev)
742{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800743 dev->netdev_ops = &ipip_netdev_ops;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000744 dev->destructor = ipip_dev_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
746 dev->type = ARPHRD_TUNNEL;
747 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800748 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 dev->flags = IFF_NOARP;
750 dev->iflink = 0;
751 dev->addr_len = 4;
Pavel Emelyanov0a826402008-04-16 01:06:18 -0700752 dev->features |= NETIF_F_NETNS_LOCAL;
Eric Dumazet153f0942010-09-28 00:17:17 +0000753 dev->features |= NETIF_F_LLTX;
Eric Dumazet28e72212009-05-28 10:44:30 +0000754 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755}
756
Eric Dumazet3c97af92010-09-27 00:35:50 +0000757static int ipip_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800759 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
761 tunnel->dev = dev;
762 strcpy(tunnel->parms.name, dev->name);
763
764 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
765 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
766
Michal Schmidt55339952007-12-12 11:01:43 -0800767 ipip_tunnel_bind_dev(dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000768
769 dev->tstats = alloc_percpu(struct pcpu_tstats);
770 if (!dev->tstats)
771 return -ENOMEM;
772
773 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774}
775
Eric Dumazetfada5632010-09-27 23:56:46 +0000776static int __net_init ipip_fb_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777{
Patrick McHardy2941a482006-01-08 22:05:26 -0800778 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 struct iphdr *iph = &tunnel->parms.iph;
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700780 struct ipip_net *ipn = net_generic(dev_net(dev), ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782 tunnel->dev = dev;
783 strcpy(tunnel->parms.name, dev->name);
784
785 iph->version = 4;
786 iph->protocol = IPPROTO_IPIP;
787 iph->ihl = 5;
788
Eric Dumazetfada5632010-09-27 23:56:46 +0000789 dev->tstats = alloc_percpu(struct pcpu_tstats);
790 if (!dev->tstats)
791 return -ENOMEM;
792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 dev_hold(dev);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000794 RCU_INIT_POINTER(ipn->tunnels_wc[0], tunnel);
Eric Dumazetfada5632010-09-27 23:56:46 +0000795 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796}
797
Eric Dumazet6dcd8142010-08-30 07:04:14 +0000798static struct xfrm_tunnel ipip_handler __read_mostly = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 .handler = ipip_rcv,
800 .err_handler = ipip_err,
Herbert Xud2acc342006-03-28 01:12:13 -0800801 .priority = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802};
803
Stephen Hemminger5747a1a2009-02-22 00:02:08 -0800804static const char banner[] __initconst =
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 KERN_INFO "IPv4 over IPv4 tunneling driver\n";
806
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000807static void ipip_destroy_tunnels(struct ipip_net *ipn, struct list_head *head)
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700808{
809 int prio;
810
811 for (prio = 1; prio < 4; prio++) {
812 int h;
813 for (h = 0; h < HASH_SIZE; h++) {
Eric Dumazetb7285b72010-09-15 11:07:24 +0000814 struct ip_tunnel *t;
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000815
Eric Dumazetb7285b72010-09-15 11:07:24 +0000816 t = rtnl_dereference(ipn->tunnels[prio][h]);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000817 while (t != NULL) {
818 unregister_netdevice_queue(t->dev, head);
Eric Dumazetb7285b72010-09-15 11:07:24 +0000819 t = rtnl_dereference(t->next);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000820 }
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700821 }
822 }
823}
824
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000825static int __net_init ipip_init_net(struct net *net)
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700826{
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000827 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700828 int err;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700829
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700830 ipn->tunnels[0] = ipn->tunnels_wc;
831 ipn->tunnels[1] = ipn->tunnels_l;
832 ipn->tunnels[2] = ipn->tunnels_r;
833 ipn->tunnels[3] = ipn->tunnels_r_l;
834
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700835 ipn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
836 "tunl0",
837 ipip_tunnel_setup);
838 if (!ipn->fb_tunnel_dev) {
839 err = -ENOMEM;
840 goto err_alloc_dev;
841 }
Alexey Dobriyanbe77e592008-11-23 17:26:26 -0800842 dev_net_set(ipn->fb_tunnel_dev, net);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700843
Eric Dumazetfada5632010-09-27 23:56:46 +0000844 err = ipip_fb_tunnel_init(ipn->fb_tunnel_dev);
845 if (err)
846 goto err_reg_dev;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700847
848 if ((err = register_netdev(ipn->fb_tunnel_dev)))
849 goto err_reg_dev;
850
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700851 return 0;
852
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700853err_reg_dev:
Eric Dumazetfada5632010-09-27 23:56:46 +0000854 ipip_dev_free(ipn->fb_tunnel_dev);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700855err_alloc_dev:
856 /* nothing */
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700857 return err;
858}
859
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000860static void __net_exit ipip_exit_net(struct net *net)
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700861{
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000862 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000863 LIST_HEAD(list);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700864
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700865 rtnl_lock();
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000866 ipip_destroy_tunnels(ipn, &list);
867 unregister_netdevice_queue(ipn->fb_tunnel_dev, &list);
868 unregister_netdevice_many(&list);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700869 rtnl_unlock();
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700870}
871
872static struct pernet_operations ipip_net_ops = {
873 .init = ipip_init_net,
874 .exit = ipip_exit_net,
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000875 .id = &ipip_net_id,
876 .size = sizeof(struct ipip_net),
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700877};
878
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879static int __init ipip_init(void)
880{
881 int err;
882
883 printk(banner);
884
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000885 err = register_pernet_device(&ipip_net_ops);
Alexey Dobriyand5aa4072010-02-16 09:05:04 +0000886 if (err < 0)
887 return err;
888 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
889 if (err < 0) {
890 unregister_pernet_device(&ipip_net_ops);
891 printk(KERN_INFO "ipip init: can't register tunnel\n");
892 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894}
895
896static void __exit ipip_fini(void)
897{
Kazunori MIYAZAWAc0d56402007-02-13 12:54:47 -0800898 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 printk(KERN_INFO "ipip close: can't deregister tunnel\n");
900
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000901 unregister_pernet_device(&ipip_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902}
903
904module_init(ipip_init);
905module_exit(ipip_fini);
906MODULE_LICENSE("GPL");
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +0300907MODULE_ALIAS_NETDEV("tunl0");