blob: ae1413e3f2f8bd7a16ed73bdd327f908a6012a28 [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;
Eric Dumazet8ce120f2011-11-04 23:19:28 +0000151} __attribute__((aligned(4*sizeof(unsigned long))));
Eric Dumazet3c97af92010-09-27 00:35:50 +0000152
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) {
Eric Dumazetcf778b02012-01-12 04:41:32 +0000234 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 Dumazetcf778b02012-01-12 04:41:32 +0000244 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
245 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
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 Emelyanovb37d428b2008-02-26 23:51:04 -0800285 if (register_netdevice(dev) < 0)
286 goto failed_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Ted Feng72b36012011-12-08 00:46:21 +0000288 strcpy(nt->parms.name, dev->name);
289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 dev_hold(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700291 ipip_tunnel_link(ipn, nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 return nt;
293
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800294failed_free:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000295 ipip_dev_free(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 return NULL;
297}
298
Eric Dumazetb7285b72010-09-15 11:07:24 +0000299/* called with RTNL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300static void ipip_tunnel_uninit(struct net_device *dev)
301{
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700302 struct net *net = dev_net(dev);
303 struct ipip_net *ipn = net_generic(net, ipip_net_id);
304
Eric Dumazetb7285b72010-09-15 11:07:24 +0000305 if (dev == ipn->fb_tunnel_dev)
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000306 RCU_INIT_POINTER(ipn->tunnels_wc[0], NULL);
Eric Dumazetb7285b72010-09-15 11:07:24 +0000307 else
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700308 ipip_tunnel_unlink(ipn, netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 dev_put(dev);
310}
311
Herbert Xud2acc342006-03-28 01:12:13 -0800312static int ipip_err(struct sk_buff *skb, u32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
Rami Rosen071f92d2008-05-21 17:47:54 -0700315/* All the routers (except for Linux) return only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 8 bytes of packet payload. It means, that precise relaying of
317 ICMP in the real Internet is absolutely infeasible.
318 */
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000319 const struct iphdr *iph = (const struct iphdr *)skb->data;
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300320 const int type = icmp_hdr(skb)->type;
321 const int code = icmp_hdr(skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 struct ip_tunnel *t;
Herbert Xud2acc342006-03-28 01:12:13 -0800323 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
325 switch (type) {
326 default:
327 case ICMP_PARAMETERPROB:
Herbert Xud2acc342006-03-28 01:12:13 -0800328 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
330 case ICMP_DEST_UNREACH:
331 switch (code) {
332 case ICMP_SR_FAILED:
333 case ICMP_PORT_UNREACH:
334 /* Impossible event. */
Herbert Xud2acc342006-03-28 01:12:13 -0800335 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 case ICMP_FRAG_NEEDED:
337 /* Soft state for pmtu is maintained by IP core. */
Herbert Xud2acc342006-03-28 01:12:13 -0800338 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 default:
340 /* All others are translated to HOST_UNREACH.
341 rfc2003 contains "deep thoughts" about NET_UNREACH,
342 I believe they are just ether pollution. --ANK
343 */
344 break;
345 }
346 break;
347 case ICMP_TIME_EXCEEDED:
348 if (code != ICMP_EXC_TTL)
Herbert Xud2acc342006-03-28 01:12:13 -0800349 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 break;
351 }
352
Herbert Xud2acc342006-03-28 01:12:13 -0800353 err = -ENOENT;
354
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000355 rcu_read_lock();
Pavel Emelyanovcec3ffa2008-04-16 01:05:03 -0700356 t = ipip_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 if (t == NULL || t->parms.iph.daddr == 0)
358 goto out;
Herbert Xud2acc342006-03-28 01:12:13 -0800359
360 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
362 goto out;
363
Wei Yongjun26d94b42009-02-24 23:36:47 -0800364 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 t->err_count++;
366 else
367 t->err_count = 1;
368 t->err_time = jiffies;
369out:
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000370 rcu_read_unlock();
Herbert Xud2acc342006-03-28 01:12:13 -0800371 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700374static inline void ipip_ecn_decapsulate(const struct iphdr *outer_iph,
375 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376{
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700377 struct iphdr *inner_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
379 if (INET_ECN_is_ce(outer_iph->tos))
380 IP_ECN_set_ce(inner_iph);
381}
382
383static int ipip_rcv(struct sk_buff *skb)
384{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 struct ip_tunnel *tunnel;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700386 const struct iphdr *iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000388 rcu_read_lock();
Eric Dumazet3c97af92010-09-27 00:35:50 +0000389 tunnel = ipip_tunnel_lookup(dev_net(skb->dev), iph->saddr, iph->daddr);
390 if (tunnel != NULL) {
391 struct pcpu_tstats *tstats;
392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000394 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 kfree_skb(skb);
396 return 0;
397 }
398
399 secpath_reset(skb);
400
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700401 skb->mac_header = skb->network_header;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700402 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 skb->protocol = htons(ETH_P_IP);
404 skb->pkt_type = PACKET_HOST;
405
Eric Dumazet3c97af92010-09-27 00:35:50 +0000406 tstats = this_cpu_ptr(tunnel->dev->tstats);
407 tstats->rx_packets++;
408 tstats->rx_bytes += skb->len;
409
410 __skb_tunnel_rx(skb, tunnel->dev);
Eric Dumazetd19d56d2010-05-17 22:36:55 -0700411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 ipip_ecn_decapsulate(iph, skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000413
Eric Dumazetcaf586e2010-09-30 21:06:55 +0000414 netif_rx(skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000415
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000416 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 return 0;
418 }
Eric Dumazet8f95dd62009-10-23 05:42:02 +0000419 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 return -1;
422}
423
424/*
425 * This function assumes it is being called from dev_queue_xmit()
426 * and that skb is filled properly by that function.
427 */
428
Stephen Hemminger6fef4c02009-08-31 19:50:41 +0000429static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430{
Patrick McHardy2941a482006-01-08 22:05:26 -0800431 struct ip_tunnel *tunnel = netdev_priv(dev);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000432 struct pcpu_tstats *tstats;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000433 const struct iphdr *tiph = &tunnel->parms.iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 u8 tos = tunnel->parms.iph.tos;
Al Virod5a0a1e2006-11-08 00:23:14 -0800435 __be16 df = tiph->frag_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct rtable *rt; /* Route to the other host */
Eric Dumazet3c97af92010-09-27 00:35:50 +0000437 struct net_device *tdev; /* Device to other host */
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000438 const struct iphdr *old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 struct iphdr *iph; /* Our new IP header */
Chuck Leverc2636b42007-10-23 21:07:32 -0700440 unsigned int max_headroom; /* The extra header space needed */
Al Virod5a0a1e2006-11-08 00:23:14 -0800441 __be32 dst = tiph->daddr;
David S. Miller31e45432011-05-03 20:25:42 -0700442 struct flowi4 fl4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 int mtu;
444
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 if (skb->protocol != htons(ETH_P_IP))
446 goto tx_error;
447
Eric Dumazet3c97af92010-09-27 00:35:50 +0000448 if (tos & 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 tos = old_iph->tos;
450
451 if (!dst) {
452 /* NBMA tunnel */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000453 if ((rt = skb_rtable(skb)) == NULL) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000454 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 goto tx_error;
456 }
David S. Miller658c8d92012-01-24 18:18:05 -0500457 dst = rt->rt_gateway;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 }
459
David S. Miller31e45432011-05-03 20:25:42 -0700460 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
David S. Miller78fbfd82011-03-12 00:00:52 -0500461 dst, tiph->saddr,
462 0, 0,
463 IPPROTO_IPIP, RT_TOS(tos),
464 tunnel->parms.link);
465 if (IS_ERR(rt)) {
466 dev->stats.tx_carrier_errors++;
467 goto tx_error_icmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 }
Changli Gaod8d1f302010-06-10 23:31:35 -0700469 tdev = rt->dst.dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471 if (tdev == dev) {
472 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000473 dev->stats.collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto tx_error;
475 }
476
Herbert Xu23ca0c92009-11-06 10:37:41 +0000477 df |= old_iph->frag_off & htons(IP_DF);
478
479 if (df) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700480 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
Herbert Xu23ca0c92009-11-06 10:37:41 +0000482 if (mtu < 68) {
Eric Dumazet3c97af92010-09-27 00:35:50 +0000483 dev->stats.collisions++;
Herbert Xu23ca0c92009-11-06 10:37:41 +0000484 ip_rt_put(rt);
485 goto tx_error;
486 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Herbert Xu23ca0c92009-11-06 10:37:41 +0000488 if (skb_dst(skb))
489 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Herbert Xu23ca0c92009-11-06 10:37:41 +0000491 if ((old_iph->frag_off & htons(IP_DF)) &&
492 mtu < ntohs(old_iph->tot_len)) {
493 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
494 htonl(mtu));
495 ip_rt_put(rt);
496 goto tx_error;
497 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 }
499
500 if (tunnel->err_count > 0) {
Wei Yongjun26d94b42009-02-24 23:36:47 -0800501 if (time_before(jiffies,
502 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 tunnel->err_count--;
504 dst_link_failure(skb);
505 } else
506 tunnel->err_count = 0;
507 }
508
509 /*
510 * Okay, now see if we can stuff it in the buffer as-is.
511 */
512 max_headroom = (LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr));
513
Patrick McHardycfbba492007-07-09 15:33:40 -0700514 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
515 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
517 if (!new_skb) {
518 ip_rt_put(rt);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000519 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000521 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 }
523 if (skb->sk)
524 skb_set_owner_w(new_skb, skb->sk);
525 dev_kfree_skb(skb);
526 skb = new_skb;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700527 old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
529
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700530 skb->transport_header = skb->network_header;
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700531 skb_push(skb, sizeof(struct iphdr));
532 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy48d5cad2006-02-15 15:10:22 -0800534 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
535 IPSKB_REROUTED);
Eric Dumazetadf30902009-06-02 05:19:30 +0000536 skb_dst_drop(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700537 skb_dst_set(skb, &rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
539 /*
540 * Push down and install the IPIP header.
541 */
542
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700543 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 iph->version = 4;
545 iph->ihl = sizeof(struct iphdr)>>2;
546 iph->frag_off = df;
547 iph->protocol = IPPROTO_IPIP;
548 iph->tos = INET_ECN_encapsulate(tos, old_iph->tos);
David S. Miller69458cb2011-05-04 11:10:28 -0700549 iph->daddr = fl4.daddr;
550 iph->saddr = fl4.saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
552 if ((iph->ttl = tiph->ttl) == 0)
553 iph->ttl = old_iph->ttl;
554
555 nf_reset(skb);
Eric Dumazet3c97af92010-09-27 00:35:50 +0000556 tstats = this_cpu_ptr(dev->tstats);
557 __IPTUNNEL_XMIT(tstats, &dev->stats);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000558 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
560tx_error_icmp:
561 dst_link_failure(skb);
562tx_error:
Eric Dumazet3c97af92010-09-27 00:35:50 +0000563 dev->stats.tx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000565 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566}
567
Michal Schmidt55339952007-12-12 11:01:43 -0800568static void ipip_tunnel_bind_dev(struct net_device *dev)
569{
570 struct net_device *tdev = NULL;
571 struct ip_tunnel *tunnel;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000572 const struct iphdr *iph;
Michal Schmidt55339952007-12-12 11:01:43 -0800573
574 tunnel = netdev_priv(dev);
575 iph = &tunnel->parms.iph;
576
577 if (iph->daddr) {
David S. Miller31e45432011-05-03 20:25:42 -0700578 struct rtable *rt;
579 struct flowi4 fl4;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000580
David S. Miller31e45432011-05-03 20:25:42 -0700581 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
582 iph->daddr, iph->saddr,
583 0, 0,
584 IPPROTO_IPIP,
585 RT_TOS(iph->tos),
586 tunnel->parms.link);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800587 if (!IS_ERR(rt)) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700588 tdev = rt->dst.dev;
Michal Schmidt55339952007-12-12 11:01:43 -0800589 ip_rt_put(rt);
590 }
591 dev->flags |= IFF_POINTOPOINT;
592 }
593
594 if (!tdev && tunnel->parms.link)
Pavel Emelyanovb99f0152008-04-16 01:05:57 -0700595 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
Michal Schmidt55339952007-12-12 11:01:43 -0800596
597 if (tdev) {
598 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
599 dev->mtu = tdev->mtu - sizeof(struct iphdr);
600 }
601 dev->iflink = tunnel->parms.link;
602}
603
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604static int
605ipip_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
606{
607 int err = 0;
608 struct ip_tunnel_parm p;
609 struct ip_tunnel *t;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700610 struct net *net = dev_net(dev);
611 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 switch (cmd) {
614 case SIOCGETTUNNEL:
615 t = NULL;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700616 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
618 err = -EFAULT;
619 break;
620 }
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700621 t = ipip_tunnel_locate(net, &p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 }
623 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -0800624 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 memcpy(&p, &t->parms, sizeof(p));
626 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
627 err = -EFAULT;
628 break;
629
630 case SIOCADDTUNNEL:
631 case SIOCCHGTUNNEL:
632 err = -EPERM;
633 if (!capable(CAP_NET_ADMIN))
634 goto done;
635
636 err = -EFAULT;
637 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
638 goto done;
639
640 err = -EINVAL;
641 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
642 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
643 goto done;
644 if (p.iph.ttl)
645 p.iph.frag_off |= htons(IP_DF);
646
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700647 t = ipip_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700649 if (dev != ipn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 if (t != NULL) {
651 if (t->dev != dev) {
652 err = -EEXIST;
653 break;
654 }
655 } else {
656 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
657 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
658 err = -EINVAL;
659 break;
660 }
Patrick McHardy2941a482006-01-08 22:05:26 -0800661 t = netdev_priv(dev);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700662 ipip_tunnel_unlink(ipn, t);
Pavel Emelyanov74b0b852010-10-27 05:43:53 +0000663 synchronize_net();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 t->parms.iph.saddr = p.iph.saddr;
665 t->parms.iph.daddr = p.iph.daddr;
666 memcpy(dev->dev_addr, &p.iph.saddr, 4);
667 memcpy(dev->broadcast, &p.iph.daddr, 4);
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700668 ipip_tunnel_link(ipn, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 netdev_state_change(dev);
670 }
671 }
672
673 if (t) {
674 err = 0;
675 if (cmd == SIOCCHGTUNNEL) {
676 t->parms.iph.ttl = p.iph.ttl;
677 t->parms.iph.tos = p.iph.tos;
678 t->parms.iph.frag_off = p.iph.frag_off;
Michal Schmidt55339952007-12-12 11:01:43 -0800679 if (t->parms.link != p.link) {
680 t->parms.link = p.link;
681 ipip_tunnel_bind_dev(dev);
682 netdev_state_change(dev);
683 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 }
685 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
686 err = -EFAULT;
687 } else
688 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
689 break;
690
691 case SIOCDELTUNNEL:
692 err = -EPERM;
693 if (!capable(CAP_NET_ADMIN))
694 goto done;
695
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700696 if (dev == ipn->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 err = -EFAULT;
698 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
699 goto done;
700 err = -ENOENT;
Pavel Emelyanovb9fae5c2008-04-16 01:04:35 -0700701 if ((t = ipip_tunnel_locate(net, &p, 0)) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 goto done;
703 err = -EPERM;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700704 if (t->dev == ipn->fb_tunnel_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 goto done;
706 dev = t->dev;
707 }
Stephen Hemminger22f8cde2007-02-07 00:09:58 -0800708 unregister_netdevice(dev);
709 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 break;
711
712 default:
713 err = -EINVAL;
714 }
715
716done:
717 return err;
718}
719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720static int ipip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
721{
722 if (new_mtu < 68 || new_mtu > 0xFFF8 - sizeof(struct iphdr))
723 return -EINVAL;
724 dev->mtu = new_mtu;
725 return 0;
726}
727
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800728static const struct net_device_ops ipip_netdev_ops = {
729 .ndo_uninit = ipip_tunnel_uninit,
730 .ndo_start_xmit = ipip_tunnel_xmit,
731 .ndo_do_ioctl = ipip_tunnel_ioctl,
732 .ndo_change_mtu = ipip_tunnel_change_mtu,
Eric Dumazet3c97af92010-09-27 00:35:50 +0000733 .ndo_get_stats = ipip_get_stats,
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800734};
735
Eric Dumazet3c97af92010-09-27 00:35:50 +0000736static void ipip_dev_free(struct net_device *dev)
737{
738 free_percpu(dev->tstats);
739 free_netdev(dev);
740}
741
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742static void ipip_tunnel_setup(struct net_device *dev)
743{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800744 dev->netdev_ops = &ipip_netdev_ops;
Eric Dumazet3c97af92010-09-27 00:35:50 +0000745 dev->destructor = ipip_dev_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
747 dev->type = ARPHRD_TUNNEL;
748 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800749 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 dev->flags = IFF_NOARP;
751 dev->iflink = 0;
752 dev->addr_len = 4;
Pavel Emelyanov0a826402008-04-16 01:06:18 -0700753 dev->features |= NETIF_F_NETNS_LOCAL;
Eric Dumazet153f0942010-09-28 00:17:17 +0000754 dev->features |= NETIF_F_LLTX;
Eric Dumazet28e72212009-05-28 10:44:30 +0000755 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756}
757
Eric Dumazet3c97af92010-09-27 00:35:50 +0000758static int ipip_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759{
Stephen Hemminger23a12b12008-11-20 20:33:21 -0800760 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 tunnel->dev = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
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);
Eric Dumazetcf778b02012-01-12 04:41:32 +0000794 rcu_assign_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);
Ted Feng72b36012011-12-08 00:46:21 +0000828 struct ip_tunnel *t;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700829 int err;
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700830
Pavel Emelyanov44d3c292008-04-16 01:05:32 -0700831 ipn->tunnels[0] = ipn->tunnels_wc;
832 ipn->tunnels[1] = ipn->tunnels_l;
833 ipn->tunnels[2] = ipn->tunnels_r;
834 ipn->tunnels[3] = ipn->tunnels_r_l;
835
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700836 ipn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
837 "tunl0",
838 ipip_tunnel_setup);
839 if (!ipn->fb_tunnel_dev) {
840 err = -ENOMEM;
841 goto err_alloc_dev;
842 }
Alexey Dobriyanbe77e592008-11-23 17:26:26 -0800843 dev_net_set(ipn->fb_tunnel_dev, net);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700844
Eric Dumazetfada5632010-09-27 23:56:46 +0000845 err = ipip_fb_tunnel_init(ipn->fb_tunnel_dev);
846 if (err)
847 goto err_reg_dev;
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700848
849 if ((err = register_netdev(ipn->fb_tunnel_dev)))
850 goto err_reg_dev;
851
Ted Feng72b36012011-12-08 00:46:21 +0000852 t = netdev_priv(ipn->fb_tunnel_dev);
853
854 strcpy(t->parms.name, ipn->fb_tunnel_dev->name);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700855 return 0;
856
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700857err_reg_dev:
Eric Dumazetfada5632010-09-27 23:56:46 +0000858 ipip_dev_free(ipn->fb_tunnel_dev);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700859err_alloc_dev:
860 /* nothing */
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700861 return err;
862}
863
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +0000864static void __net_exit ipip_exit_net(struct net *net)
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700865{
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000866 struct ipip_net *ipn = net_generic(net, ipip_net_id);
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000867 LIST_HEAD(list);
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700868
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700869 rtnl_lock();
Eric Dumazet0694c4c2009-10-27 07:06:59 +0000870 ipip_destroy_tunnels(ipn, &list);
871 unregister_netdevice_queue(ipn->fb_tunnel_dev, &list);
872 unregister_netdevice_many(&list);
Pavel Emelyanovb9855c52008-04-16 01:04:13 -0700873 rtnl_unlock();
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700874}
875
876static struct pernet_operations ipip_net_ops = {
877 .init = ipip_init_net,
878 .exit = ipip_exit_net,
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000879 .id = &ipip_net_id,
880 .size = sizeof(struct ipip_net),
Pavel Emelyanov10dc4c72008-04-16 01:03:13 -0700881};
882
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883static int __init ipip_init(void)
884{
885 int err;
886
887 printk(banner);
888
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000889 err = register_pernet_device(&ipip_net_ops);
Alexey Dobriyand5aa4072010-02-16 09:05:04 +0000890 if (err < 0)
891 return err;
892 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
893 if (err < 0) {
894 unregister_pernet_device(&ipip_net_ops);
Joe Perches058bd4d2012-03-11 18:36:11 +0000895 pr_info("%s: can't register tunnel\n", __func__);
Alexey Dobriyand5aa4072010-02-16 09:05:04 +0000896 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
899
900static void __exit ipip_fini(void)
901{
Kazunori MIYAZAWAc0d56402007-02-13 12:54:47 -0800902 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
Joe Perches058bd4d2012-03-11 18:36:11 +0000903 pr_info("%s: can't deregister tunnel\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
Eric W. Biederman86de8a62009-11-29 15:46:14 +0000905 unregister_pernet_device(&ipip_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906}
907
908module_init(ipip_init);
909module_exit(ipip_fini);
910MODULE_LICENSE("GPL");
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +0300911MODULE_ALIAS_NETDEV("tunl0");