blob: 07a188afb3ac7af64650d8eb7f641f27b8eb09b6 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002 * Linux NET3: GRE over IP protocol decoder.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 *
4 * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.ru)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
Randy Dunlap4fc268d2006-01-11 12:17:47 -080013#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/module.h>
15#include <linux/types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/kernel.h>
17#include <asm/uaccess.h>
18#include <linux/skbuff.h>
19#include <linux/netdevice.h>
20#include <linux/in.h>
21#include <linux/tcp.h>
22#include <linux/udp.h>
23#include <linux/if_arp.h>
24#include <linux/mroute.h>
25#include <linux/init.h>
26#include <linux/in6.h>
27#include <linux/inetdevice.h>
28#include <linux/igmp.h>
29#include <linux/netfilter_ipv4.h>
Herbert Xue1a80002008-10-09 12:00:17 -070030#include <linux/etherdevice.h>
Kris Katterjohn46f25df2006-01-05 16:35:42 -080031#include <linux/if_ether.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
33#include <net/sock.h>
34#include <net/ip.h>
35#include <net/icmp.h>
36#include <net/protocol.h>
37#include <net/ipip.h>
38#include <net/arp.h>
39#include <net/checksum.h>
40#include <net/dsfield.h>
41#include <net/inet_ecn.h>
42#include <net/xfrm.h>
Pavel Emelyanov59a4c752008-04-16 01:08:53 -070043#include <net/net_namespace.h>
44#include <net/netns/generic.h>
Herbert Xuc19e6542008-10-09 11:59:55 -070045#include <net/rtnetlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47#ifdef CONFIG_IPV6
48#include <net/ipv6.h>
49#include <net/ip6_fib.h>
50#include <net/ip6_route.h>
51#endif
52
53/*
54 Problems & solutions
55 --------------------
56
57 1. The most important issue is detecting local dead loops.
58 They would cause complete host lockup in transmit, which
59 would be "resolved" by stack overflow or, if queueing is enabled,
60 with infinite looping in net_bh.
61
62 We cannot track such dead loops during route installation,
63 it is infeasible task. The most general solutions would be
64 to keep skb->encapsulation counter (sort of local ttl),
65 and silently drop packet when it expires. It is the best
66 solution, but it supposes maintaing new variable in ALL
67 skb, even if no tunneling is used.
68
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090069 Current solution: t->recursion lock breaks dead loops. It looks
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 like dev->tbusy flag, but I preferred new variable, because
71 the semantics is different. One day, when hard_start_xmit
72 will be multithreaded we will have to use skb->encapsulation.
73
74
75
76 2. Networking dead loops would not kill routers, but would really
77 kill network. IP hop limit plays role of "t->recursion" in this case,
78 if we copy it from packet being encapsulated to upper header.
79 It is very good solution, but it introduces two problems:
80
81 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
82 do not work over tunnels.
83 - traceroute does not work. I planned to relay ICMP from tunnel,
84 so that this problem would be solved and traceroute output
85 would even more informative. This idea appeared to be wrong:
86 only Linux complies to rfc1812 now (yes, guys, Linux is the only
87 true router now :-)), all routers (at least, in neighbourhood of mine)
88 return only 8 bytes of payload. It is the end.
89
90 Hence, if we want that OSPF worked or traceroute said something reasonable,
91 we should search for another solution.
92
93 One of them is to parse packet trying to detect inner encapsulation
94 made by our node. It is difficult or even impossible, especially,
95 taking into account fragmentation. TO be short, tt is not solution at all.
96
97 Current solution: The solution was UNEXPECTEDLY SIMPLE.
98 We force DF flag on tunnels with preconfigured hop limit,
99 that is ALL. :-) Well, it does not remove the problem completely,
100 but exponential growth of network traffic is changed to linear
101 (branches, that exceed pmtu are pruned) and tunnel mtu
102 fastly degrades to value <68, where looping stops.
103 Yes, it is not good if there exists a router in the loop,
104 which does not force DF, even when encapsulating packets have DF set.
105 But it is not our problem! Nobody could accuse us, we made
106 all that we could make. Even if it is your gated who injected
107 fatal route to network, even if it were you who configured
108 fatal static route: you are innocent. :-)
109
110
111
112 3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain
113 practically identical code. It would be good to glue them
114 together, but it is not very evident, how to make them modular.
115 sit is integral part of IPv6, ipip and gre are naturally modular.
116 We could extract common parts (hash table, ioctl etc)
117 to a separate module (ip_tunnel.c).
118
119 Alexey Kuznetsov.
120 */
121
Herbert Xuc19e6542008-10-09 11:59:55 -0700122static struct rtnl_link_ops ipgre_link_ops __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123static int ipgre_tunnel_init(struct net_device *dev);
124static void ipgre_tunnel_setup(struct net_device *dev);
Herbert Xu42aa9162008-10-09 11:59:32 -0700125static int ipgre_tunnel_bind_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
127/* Fallback tunnel: no source, no destination, no key, no options */
128
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700129#define HASH_SIZE 16
130
Pavel Emelyanov59a4c752008-04-16 01:08:53 -0700131static int ipgre_net_id;
132struct ipgre_net {
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700133 struct ip_tunnel *tunnels[4][HASH_SIZE];
134
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700135 struct net_device *fb_tunnel_dev;
Pavel Emelyanov59a4c752008-04-16 01:08:53 -0700136};
137
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138/* Tunnel hash table */
139
140/*
141 4 hash tables:
142
143 3: (remote,local)
144 2: (remote,*)
145 1: (*,local)
146 0: (*,*)
147
148 We require exact key match i.e. if a key is present in packet
149 it will match only tunnel with the same key; if it is not present,
150 it will match only keyless tunnel.
151
152 All keysless packets, if not matched configured keyless tunnels
153 will match fallback tunnel.
154 */
155
Al Virod5a0a1e2006-11-08 00:23:14 -0800156#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700158#define tunnels_r_l tunnels[3]
159#define tunnels_r tunnels[2]
160#define tunnels_l tunnels[1]
161#define tunnels_wc tunnels[0]
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
163static DEFINE_RWLOCK(ipgre_lock);
164
165/* Given src, dst and key, find appropriate for input tunnel. */
166
Timo Teras749c10f2009-01-19 17:22:12 -0800167static struct ip_tunnel * ipgre_tunnel_lookup(struct net_device *dev,
Herbert Xue1a80002008-10-09 12:00:17 -0700168 __be32 remote, __be32 local,
169 __be32 key, __be16 gre_proto)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170{
Timo Teras749c10f2009-01-19 17:22:12 -0800171 struct net *net = dev_net(dev);
172 int link = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 unsigned h0 = HASH(remote);
174 unsigned h1 = HASH(key);
Timo Terasafcf1242009-01-26 20:56:10 -0800175 struct ip_tunnel *t, *cand = NULL;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700176 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Herbert Xue1a80002008-10-09 12:00:17 -0700177 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
178 ARPHRD_ETHER : ARPHRD_IPGRE;
Timo Terasafcf1242009-01-26 20:56:10 -0800179 int score, cand_score = 4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700181 for (t = ign->tunnels_r_l[h0^h1]; t; t = t->next) {
Timo Teras749c10f2009-01-19 17:22:12 -0800182 if (local != t->parms.iph.saddr ||
183 remote != t->parms.iph.daddr ||
184 key != t->parms.i_key ||
185 !(t->dev->flags & IFF_UP))
186 continue;
187
188 if (t->dev->type != ARPHRD_IPGRE &&
189 t->dev->type != dev_type)
190 continue;
191
Timo Terasafcf1242009-01-26 20:56:10 -0800192 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800193 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800194 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800195 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800196 score |= 2;
197 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800198 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800199
200 if (score < cand_score) {
201 cand = t;
202 cand_score = score;
203 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 }
Herbert Xue1a80002008-10-09 12:00:17 -0700205
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700206 for (t = ign->tunnels_r[h0^h1]; t; t = t->next) {
Timo Teras749c10f2009-01-19 17:22:12 -0800207 if (remote != t->parms.iph.daddr ||
208 key != t->parms.i_key ||
209 !(t->dev->flags & IFF_UP))
210 continue;
211
212 if (t->dev->type != ARPHRD_IPGRE &&
213 t->dev->type != dev_type)
214 continue;
215
Timo Terasafcf1242009-01-26 20:56:10 -0800216 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800217 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800218 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800219 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800220 score |= 2;
221 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800222 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800223
224 if (score < cand_score) {
225 cand = t;
226 cand_score = score;
227 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 }
Herbert Xue1a80002008-10-09 12:00:17 -0700229
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700230 for (t = ign->tunnels_l[h1]; t; t = t->next) {
Timo Teras749c10f2009-01-19 17:22:12 -0800231 if ((local != t->parms.iph.saddr &&
232 (local != t->parms.iph.daddr ||
233 !ipv4_is_multicast(local))) ||
234 key != t->parms.i_key ||
235 !(t->dev->flags & IFF_UP))
236 continue;
237
238 if (t->dev->type != ARPHRD_IPGRE &&
239 t->dev->type != dev_type)
240 continue;
241
Timo Terasafcf1242009-01-26 20:56:10 -0800242 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800243 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800244 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800245 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800246 score |= 2;
247 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800248 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800249
250 if (score < cand_score) {
251 cand = t;
252 cand_score = score;
253 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 }
Herbert Xue1a80002008-10-09 12:00:17 -0700255
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700256 for (t = ign->tunnels_wc[h1]; t; t = t->next) {
Timo Teras749c10f2009-01-19 17:22:12 -0800257 if (t->parms.i_key != key ||
258 !(t->dev->flags & IFF_UP))
259 continue;
260
261 if (t->dev->type != ARPHRD_IPGRE &&
262 t->dev->type != dev_type)
263 continue;
264
Timo Terasafcf1242009-01-26 20:56:10 -0800265 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800266 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800267 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800268 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800269 score |= 2;
270 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800271 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800272
273 if (score < cand_score) {
274 cand = t;
275 cand_score = score;
276 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 }
278
Timo Terasafcf1242009-01-26 20:56:10 -0800279 if (cand != NULL)
280 return cand;
Herbert Xue1a80002008-10-09 12:00:17 -0700281
Timo Teras749c10f2009-01-19 17:22:12 -0800282 if (ign->fb_tunnel_dev->flags & IFF_UP)
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700283 return netdev_priv(ign->fb_tunnel_dev);
Timo Teras749c10f2009-01-19 17:22:12 -0800284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 return NULL;
286}
287
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700288static struct ip_tunnel **__ipgre_bucket(struct ipgre_net *ign,
289 struct ip_tunnel_parm *parms)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290{
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900291 __be32 remote = parms->iph.daddr;
292 __be32 local = parms->iph.saddr;
293 __be32 key = parms->i_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 unsigned h = HASH(key);
295 int prio = 0;
296
297 if (local)
298 prio |= 1;
Joe Perchesf97c1e02007-12-16 13:45:43 -0800299 if (remote && !ipv4_is_multicast(remote)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 prio |= 2;
301 h ^= HASH(remote);
302 }
303
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700304 return &ign->tunnels[prio][h];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305}
306
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700307static inline struct ip_tunnel **ipgre_bucket(struct ipgre_net *ign,
308 struct ip_tunnel *t)
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900309{
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700310 return __ipgre_bucket(ign, &t->parms);
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900311}
312
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700313static void ipgre_tunnel_link(struct ipgre_net *ign, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314{
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700315 struct ip_tunnel **tp = ipgre_bucket(ign, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
317 t->next = *tp;
318 write_lock_bh(&ipgre_lock);
319 *tp = t;
320 write_unlock_bh(&ipgre_lock);
321}
322
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700323static void ipgre_tunnel_unlink(struct ipgre_net *ign, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324{
325 struct ip_tunnel **tp;
326
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700327 for (tp = ipgre_bucket(ign, t); *tp; tp = &(*tp)->next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 if (t == *tp) {
329 write_lock_bh(&ipgre_lock);
330 *tp = t->next;
331 write_unlock_bh(&ipgre_lock);
332 break;
333 }
334 }
335}
336
Herbert Xue1a80002008-10-09 12:00:17 -0700337static struct ip_tunnel *ipgre_tunnel_find(struct net *net,
338 struct ip_tunnel_parm *parms,
339 int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
Al Virod5a0a1e2006-11-08 00:23:14 -0800341 __be32 remote = parms->iph.daddr;
342 __be32 local = parms->iph.saddr;
343 __be32 key = parms->i_key;
Timo Teras749c10f2009-01-19 17:22:12 -0800344 int link = parms->link;
Herbert Xue1a80002008-10-09 12:00:17 -0700345 struct ip_tunnel *t, **tp;
346 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
347
348 for (tp = __ipgre_bucket(ign, parms); (t = *tp) != NULL; tp = &t->next)
349 if (local == t->parms.iph.saddr &&
350 remote == t->parms.iph.daddr &&
351 key == t->parms.i_key &&
Timo Teras749c10f2009-01-19 17:22:12 -0800352 link == t->parms.link &&
Herbert Xue1a80002008-10-09 12:00:17 -0700353 type == t->dev->type)
354 break;
355
356 return t;
357}
358
359static struct ip_tunnel * ipgre_tunnel_locate(struct net *net,
360 struct ip_tunnel_parm *parms, int create)
361{
362 struct ip_tunnel *t, *nt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 char name[IFNAMSIZ];
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700365 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
Herbert Xue1a80002008-10-09 12:00:17 -0700367 t = ipgre_tunnel_find(net, parms, ARPHRD_IPGRE);
368 if (t || !create)
369 return t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
371 if (parms->name[0])
372 strlcpy(name, parms->name, IFNAMSIZ);
Pavel Emelyanov34cc7ba2008-02-23 20:19:20 -0800373 else
374 sprintf(name, "gre%%d");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
376 dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup);
377 if (!dev)
378 return NULL;
379
Pavel Emelyanov0b67ece2008-04-16 01:11:13 -0700380 dev_net_set(dev, net);
381
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800382 if (strchr(name, '%')) {
383 if (dev_alloc_name(dev, name) < 0)
384 goto failed_free;
385 }
386
Patrick McHardy2941a482006-01-08 22:05:26 -0800387 nt = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 nt->parms = *parms;
Herbert Xuc19e6542008-10-09 11:59:55 -0700389 dev->rtnl_link_ops = &ipgre_link_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390
Herbert Xu42aa9162008-10-09 11:59:32 -0700391 dev->mtu = ipgre_tunnel_bind_dev(dev);
392
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800393 if (register_netdevice(dev) < 0)
394 goto failed_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 dev_hold(dev);
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700397 ipgre_tunnel_link(ign, nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 return nt;
399
Pavel Emelyanovb37d428b2008-02-26 23:51:04 -0800400failed_free:
401 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 return NULL;
403}
404
405static void ipgre_tunnel_uninit(struct net_device *dev)
406{
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700407 struct net *net = dev_net(dev);
408 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
409
410 ipgre_tunnel_unlink(ign, netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 dev_put(dev);
412}
413
414
415static void ipgre_err(struct sk_buff *skb, u32 info)
416{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
Rami Rosen071f92d2008-05-21 17:47:54 -0700418/* All the routers (except for Linux) return only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 8 bytes of packet payload. It means, that precise relaying of
420 ICMP in the real Internet is absolutely infeasible.
421
422 Moreover, Cisco "wise men" put GRE key to the third word
423 in GRE header. It makes impossible maintaining even soft state for keyed
424 GRE tunnels with enabled checksum. Tell them "thank you".
425
426 Well, I wonder, rfc1812 was written by Cisco employee,
427 what the hell these idiots break standrads established
428 by themself???
429 */
430
Jianjun Kong6ed25332008-11-03 00:25:16 -0800431 struct iphdr *iph = (struct iphdr *)skb->data;
Al Virod5a0a1e2006-11-08 00:23:14 -0800432 __be16 *p = (__be16*)(skb->data+(iph->ihl<<2));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 int grehlen = (iph->ihl<<2) + 4;
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300434 const int type = icmp_hdr(skb)->type;
435 const int code = icmp_hdr(skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct ip_tunnel *t;
Al Virod5a0a1e2006-11-08 00:23:14 -0800437 __be16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 flags = p[0];
440 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
441 if (flags&(GRE_VERSION|GRE_ROUTING))
442 return;
443 if (flags&GRE_KEY) {
444 grehlen += 4;
445 if (flags&GRE_CSUM)
446 grehlen += 4;
447 }
448 }
449
450 /* If only 8 bytes returned, keyed message will be dropped here */
451 if (skb_headlen(skb) < grehlen)
452 return;
453
454 switch (type) {
455 default:
456 case ICMP_PARAMETERPROB:
457 return;
458
459 case ICMP_DEST_UNREACH:
460 switch (code) {
461 case ICMP_SR_FAILED:
462 case ICMP_PORT_UNREACH:
463 /* Impossible event. */
464 return;
465 case ICMP_FRAG_NEEDED:
466 /* Soft state for pmtu is maintained by IP core. */
467 return;
468 default:
469 /* All others are translated to HOST_UNREACH.
470 rfc2003 contains "deep thoughts" about NET_UNREACH,
471 I believe they are just ether pollution. --ANK
472 */
473 break;
474 }
475 break;
476 case ICMP_TIME_EXCEEDED:
477 if (code != ICMP_EXC_TTL)
478 return;
479 break;
480 }
481
482 read_lock(&ipgre_lock);
Timo Teras749c10f2009-01-19 17:22:12 -0800483 t = ipgre_tunnel_lookup(skb->dev, iph->daddr, iph->saddr,
Herbert Xue1a80002008-10-09 12:00:17 -0700484 flags & GRE_KEY ?
485 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
486 p[1]);
Joe Perchesf97c1e02007-12-16 13:45:43 -0800487 if (t == NULL || t->parms.iph.daddr == 0 ||
488 ipv4_is_multicast(t->parms.iph.daddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 goto out;
490
491 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
492 goto out;
493
494 if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
495 t->err_count++;
496 else
497 t->err_count = 1;
498 t->err_time = jiffies;
499out:
500 read_unlock(&ipgre_lock);
501 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
504static inline void ipgre_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
505{
506 if (INET_ECN_is_ce(iph->tos)) {
507 if (skb->protocol == htons(ETH_P_IP)) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700508 IP_ECN_set_ce(ip_hdr(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 } else if (skb->protocol == htons(ETH_P_IPV6)) {
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700510 IP6_ECN_set_ce(ipv6_hdr(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 }
512 }
513}
514
515static inline u8
516ipgre_ecn_encapsulate(u8 tos, struct iphdr *old_iph, struct sk_buff *skb)
517{
518 u8 inner = 0;
519 if (skb->protocol == htons(ETH_P_IP))
520 inner = old_iph->tos;
521 else if (skb->protocol == htons(ETH_P_IPV6))
522 inner = ipv6_get_dsfield((struct ipv6hdr *)old_iph);
523 return INET_ECN_encapsulate(tos, inner);
524}
525
526static int ipgre_rcv(struct sk_buff *skb)
527{
528 struct iphdr *iph;
529 u8 *h;
Al Virod5a0a1e2006-11-08 00:23:14 -0800530 __be16 flags;
Al Virod3bc23e2006-11-14 21:24:49 -0800531 __sum16 csum = 0;
Al Virod5a0a1e2006-11-08 00:23:14 -0800532 __be32 key = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 u32 seqno = 0;
534 struct ip_tunnel *tunnel;
535 int offset = 4;
Herbert Xue1a80002008-10-09 12:00:17 -0700536 __be16 gre_proto;
Herbert Xu64194c32008-10-09 12:03:17 -0700537 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
539 if (!pskb_may_pull(skb, 16))
540 goto drop_nolock;
541
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700542 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 h = skb->data;
Al Virod5a0a1e2006-11-08 00:23:14 -0800544 flags = *(__be16*)h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
547 /* - Version must be 0.
548 - We do not support routing headers.
549 */
550 if (flags&(GRE_VERSION|GRE_ROUTING))
551 goto drop_nolock;
552
553 if (flags&GRE_CSUM) {
Herbert Xufb286bb2005-11-10 13:01:24 -0800554 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700555 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800556 csum = csum_fold(skb->csum);
Herbert Xufb286bb2005-11-10 13:01:24 -0800557 if (!csum)
558 break;
559 /* fall through */
560 case CHECKSUM_NONE:
561 skb->csum = 0;
562 csum = __skb_checksum_complete(skb);
Patrick McHardy84fa7932006-08-29 16:44:56 -0700563 skb->ip_summed = CHECKSUM_COMPLETE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 }
565 offset += 4;
566 }
567 if (flags&GRE_KEY) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800568 key = *(__be32*)(h + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 offset += 4;
570 }
571 if (flags&GRE_SEQ) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800572 seqno = ntohl(*(__be32*)(h + offset));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 offset += 4;
574 }
575 }
576
Herbert Xue1a80002008-10-09 12:00:17 -0700577 gre_proto = *(__be16 *)(h + 2);
578
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 read_lock(&ipgre_lock);
Timo Teras749c10f2009-01-19 17:22:12 -0800580 if ((tunnel = ipgre_tunnel_lookup(skb->dev,
Herbert Xue1a80002008-10-09 12:00:17 -0700581 iph->saddr, iph->daddr, key,
582 gre_proto))) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700583 struct net_device_stats *stats = &tunnel->dev->stats;
584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 secpath_reset(skb);
586
Herbert Xue1a80002008-10-09 12:00:17 -0700587 skb->protocol = gre_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 /* WCCP version 1 and 2 protocol decoding.
589 * - Change protocol to IP
590 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
591 */
Herbert Xue1a80002008-10-09 12:00:17 -0700592 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
YOSHIFUJI Hideaki496c98d2006-10-10 19:41:21 -0700593 skb->protocol = htons(ETH_P_IP);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900594 if ((*(h + offset) & 0xF0) != 0x40)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 offset += 4;
596 }
597
Timo Teras1d069162007-12-20 00:10:33 -0800598 skb->mac_header = skb->network_header;
Arnaldo Carvalho de Melo4209fb62007-03-10 18:42:03 -0300599 __pskb_pull(skb, offset);
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -0700600 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 skb->pkt_type = PACKET_HOST;
602#ifdef CONFIG_NET_IPGRE_BROADCAST
Joe Perchesf97c1e02007-12-16 13:45:43 -0800603 if (ipv4_is_multicast(iph->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 /* Looped back packet, drop it! */
Eric Dumazetee6b9672008-03-05 18:30:47 -0800605 if (skb->rtable->fl.iif == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 goto drop;
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700607 stats->multicast++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 skb->pkt_type = PACKET_BROADCAST;
609 }
610#endif
611
612 if (((flags&GRE_CSUM) && csum) ||
613 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700614 stats->rx_crc_errors++;
615 stats->rx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 goto drop;
617 }
618 if (tunnel->parms.i_flags&GRE_SEQ) {
619 if (!(flags&GRE_SEQ) ||
620 (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700621 stats->rx_fifo_errors++;
622 stats->rx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 goto drop;
624 }
625 tunnel->i_seqno = seqno + 1;
626 }
Herbert Xue1a80002008-10-09 12:00:17 -0700627
Herbert Xu64194c32008-10-09 12:03:17 -0700628 len = skb->len;
629
Herbert Xue1a80002008-10-09 12:00:17 -0700630 /* Warning: All skb pointers will be invalidated! */
631 if (tunnel->dev->type == ARPHRD_ETHER) {
632 if (!pskb_may_pull(skb, ETH_HLEN)) {
633 stats->rx_length_errors++;
634 stats->rx_errors++;
635 goto drop;
636 }
637
638 iph = ip_hdr(skb);
639 skb->protocol = eth_type_trans(skb, tunnel->dev);
640 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
641 }
642
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700643 stats->rx_packets++;
Herbert Xu64194c32008-10-09 12:03:17 -0700644 stats->rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 skb->dev = tunnel->dev;
646 dst_release(skb->dst);
647 skb->dst = NULL;
648 nf_reset(skb);
Herbert Xue1a80002008-10-09 12:00:17 -0700649
650 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 ipgre_ecn_decapsulate(iph, skb);
Herbert Xue1a80002008-10-09 12:00:17 -0700652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 netif_rx(skb);
654 read_unlock(&ipgre_lock);
655 return(0);
656 }
Herbert Xu45af08b2006-04-05 22:31:19 -0700657 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659drop:
660 read_unlock(&ipgre_lock);
661drop_nolock:
662 kfree_skb(skb);
663 return(0);
664}
665
666static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
667{
Patrick McHardy2941a482006-01-08 22:05:26 -0800668 struct ip_tunnel *tunnel = netdev_priv(dev);
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700669 struct net_device_stats *stats = &tunnel->dev->stats;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700670 struct iphdr *old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 struct iphdr *tiph;
672 u8 tos;
Al Virod5a0a1e2006-11-08 00:23:14 -0800673 __be16 df;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 struct rtable *rt; /* Route to the other host */
675 struct net_device *tdev; /* Device to other host */
676 struct iphdr *iph; /* Our new IP header */
Chuck Leverc2636b42007-10-23 21:07:32 -0700677 unsigned int max_headroom; /* The extra header space needed */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 int gre_hlen;
Al Virod5a0a1e2006-11-08 00:23:14 -0800679 __be32 dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 int mtu;
681
682 if (tunnel->recursion++) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700683 stats->collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 goto tx_error;
685 }
686
Herbert Xue1a80002008-10-09 12:00:17 -0700687 if (dev->type == ARPHRD_ETHER)
688 IPCB(skb)->flags = 0;
689
690 if (dev->header_ops && dev->type == ARPHRD_IPGRE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 gre_hlen = 0;
Jianjun Kong6ed25332008-11-03 00:25:16 -0800692 tiph = (struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 } else {
694 gre_hlen = tunnel->hlen;
695 tiph = &tunnel->parms.iph;
696 }
697
698 if ((dst = tiph->daddr) == 0) {
699 /* NBMA tunnel */
700
701 if (skb->dst == NULL) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700702 stats->tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 goto tx_error;
704 }
705
706 if (skb->protocol == htons(ETH_P_IP)) {
Eric Dumazetee6b9672008-03-05 18:30:47 -0800707 rt = skb->rtable;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 if ((dst = rt->rt_gateway) == 0)
709 goto tx_error_icmp;
710 }
711#ifdef CONFIG_IPV6
712 else if (skb->protocol == htons(ETH_P_IPV6)) {
713 struct in6_addr *addr6;
714 int addr_type;
715 struct neighbour *neigh = skb->dst->neighbour;
716
717 if (neigh == NULL)
718 goto tx_error;
719
Jianjun Kong6ed25332008-11-03 00:25:16 -0800720 addr6 = (struct in6_addr *)&neigh->primary_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 addr_type = ipv6_addr_type(addr6);
722
723 if (addr_type == IPV6_ADDR_ANY) {
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700724 addr6 = &ipv6_hdr(skb)->daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 addr_type = ipv6_addr_type(addr6);
726 }
727
728 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
729 goto tx_error_icmp;
730
731 dst = addr6->s6_addr32[3];
732 }
733#endif
734 else
735 goto tx_error;
736 }
737
738 tos = tiph->tos;
739 if (tos&1) {
740 if (skb->protocol == htons(ETH_P_IP))
741 tos = old_iph->tos;
742 tos &= ~1;
743 }
744
745 {
746 struct flowi fl = { .oif = tunnel->parms.link,
747 .nl_u = { .ip4_u =
748 { .daddr = dst,
749 .saddr = tiph->saddr,
750 .tos = RT_TOS(tos) } },
751 .proto = IPPROTO_GRE };
Pavel Emelyanov96635522008-04-16 01:10:44 -0700752 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700753 stats->tx_carrier_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 goto tx_error;
755 }
756 }
757 tdev = rt->u.dst.dev;
758
759 if (tdev == dev) {
760 ip_rt_put(rt);
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700761 stats->collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 goto tx_error;
763 }
764
765 df = tiph->frag_off;
766 if (df)
Herbert Xuc95b8192008-10-09 11:58:54 -0700767 mtu = dst_mtu(&rt->u.dst) - dev->hard_header_len - tunnel->hlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 else
769 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
770
771 if (skb->dst)
772 skb->dst->ops->update_pmtu(skb->dst, mtu);
773
774 if (skb->protocol == htons(ETH_P_IP)) {
775 df |= (old_iph->frag_off&htons(IP_DF));
776
777 if ((old_iph->frag_off&htons(IP_DF)) &&
778 mtu < ntohs(old_iph->tot_len)) {
779 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
780 ip_rt_put(rt);
781 goto tx_error;
782 }
783 }
784#ifdef CONFIG_IPV6
785 else if (skb->protocol == htons(ETH_P_IPV6)) {
Jianjun Kong6ed25332008-11-03 00:25:16 -0800786 struct rt6_info *rt6 = (struct rt6_info *)skb->dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788 if (rt6 && mtu < dst_mtu(skb->dst) && mtu >= IPV6_MIN_MTU) {
Joe Perchesf97c1e02007-12-16 13:45:43 -0800789 if ((tunnel->parms.iph.daddr &&
790 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 rt6->rt6i_dst.plen == 128) {
792 rt6->rt6i_flags |= RTF_MODIFIED;
793 skb->dst->metrics[RTAX_MTU-1] = mtu;
794 }
795 }
796
797 if (mtu >= IPV6_MIN_MTU && mtu < skb->len - tunnel->hlen + gre_hlen) {
798 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
799 ip_rt_put(rt);
800 goto tx_error;
801 }
802 }
803#endif
804
805 if (tunnel->err_count > 0) {
806 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
807 tunnel->err_count--;
808
809 dst_link_failure(skb);
810 } else
811 tunnel->err_count = 0;
812 }
813
814 max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen;
815
Patrick McHardycfbba492007-07-09 15:33:40 -0700816 if (skb_headroom(skb) < max_headroom || skb_shared(skb)||
817 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
819 if (!new_skb) {
820 ip_rt_put(rt);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900821 stats->tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 dev_kfree_skb(skb);
823 tunnel->recursion--;
824 return 0;
825 }
826 if (skb->sk)
827 skb_set_owner_w(new_skb, skb->sk);
828 dev_kfree_skb(skb);
829 skb = new_skb;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700830 old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 }
832
Herbert Xu64194c32008-10-09 12:03:17 -0700833 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700834 skb_push(skb, gre_hlen);
835 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy48d5cad2006-02-15 15:10:22 -0800837 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
838 IPSKB_REROUTED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 dst_release(skb->dst);
840 skb->dst = &rt->u.dst;
841
842 /*
843 * Push down and install the IPIP header.
844 */
845
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700846 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 iph->version = 4;
848 iph->ihl = sizeof(struct iphdr) >> 2;
849 iph->frag_off = df;
850 iph->protocol = IPPROTO_GRE;
851 iph->tos = ipgre_ecn_encapsulate(tos, old_iph, skb);
852 iph->daddr = rt->rt_dst;
853 iph->saddr = rt->rt_src;
854
855 if ((iph->ttl = tiph->ttl) == 0) {
856 if (skb->protocol == htons(ETH_P_IP))
857 iph->ttl = old_iph->ttl;
858#ifdef CONFIG_IPV6
859 else if (skb->protocol == htons(ETH_P_IPV6))
Jianjun Kong6ed25332008-11-03 00:25:16 -0800860 iph->ttl = ((struct ipv6hdr *)old_iph)->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861#endif
862 else
863 iph->ttl = dst_metric(&rt->u.dst, RTAX_HOPLIMIT);
864 }
865
Herbert Xue1a80002008-10-09 12:00:17 -0700866 ((__be16 *)(iph + 1))[0] = tunnel->parms.o_flags;
867 ((__be16 *)(iph + 1))[1] = (dev->type == ARPHRD_ETHER) ?
868 htons(ETH_P_TEB) : skb->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
870 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800871 __be32 *ptr = (__be32*)(((u8*)iph) + tunnel->hlen - 4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
873 if (tunnel->parms.o_flags&GRE_SEQ) {
874 ++tunnel->o_seqno;
875 *ptr = htonl(tunnel->o_seqno);
876 ptr--;
877 }
878 if (tunnel->parms.o_flags&GRE_KEY) {
879 *ptr = tunnel->parms.o_key;
880 ptr--;
881 }
882 if (tunnel->parms.o_flags&GRE_CSUM) {
883 *ptr = 0;
Al Viro5f92a732006-11-14 21:36:54 -0800884 *(__sum16*)ptr = ip_compute_csum((void*)(iph+1), skb->len - sizeof(struct iphdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 }
886 }
887
888 nf_reset(skb);
889
890 IPTUNNEL_XMIT();
891 tunnel->recursion--;
892 return 0;
893
894tx_error_icmp:
895 dst_link_failure(skb);
896
897tx_error:
898 stats->tx_errors++;
899 dev_kfree_skb(skb);
900 tunnel->recursion--;
901 return 0;
902}
903
Herbert Xu42aa9162008-10-09 11:59:32 -0700904static int ipgre_tunnel_bind_dev(struct net_device *dev)
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800905{
906 struct net_device *tdev = NULL;
907 struct ip_tunnel *tunnel;
908 struct iphdr *iph;
909 int hlen = LL_MAX_HEADER;
910 int mtu = ETH_DATA_LEN;
911 int addend = sizeof(struct iphdr) + 4;
912
913 tunnel = netdev_priv(dev);
914 iph = &tunnel->parms.iph;
915
Herbert Xuc95b8192008-10-09 11:58:54 -0700916 /* Guess output device to choose reasonable mtu and needed_headroom */
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800917
918 if (iph->daddr) {
919 struct flowi fl = { .oif = tunnel->parms.link,
920 .nl_u = { .ip4_u =
921 { .daddr = iph->daddr,
922 .saddr = iph->saddr,
923 .tos = RT_TOS(iph->tos) } },
924 .proto = IPPROTO_GRE };
925 struct rtable *rt;
Pavel Emelyanov96635522008-04-16 01:10:44 -0700926 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800927 tdev = rt->u.dst.dev;
928 ip_rt_put(rt);
929 }
Herbert Xue1a80002008-10-09 12:00:17 -0700930
931 if (dev->type != ARPHRD_ETHER)
932 dev->flags |= IFF_POINTOPOINT;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800933 }
934
935 if (!tdev && tunnel->parms.link)
Pavel Emelyanov96635522008-04-16 01:10:44 -0700936 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800937
938 if (tdev) {
Herbert Xuc95b8192008-10-09 11:58:54 -0700939 hlen = tdev->hard_header_len + tdev->needed_headroom;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800940 mtu = tdev->mtu;
941 }
942 dev->iflink = tunnel->parms.link;
943
944 /* Precalculate GRE options length */
945 if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
946 if (tunnel->parms.o_flags&GRE_CSUM)
947 addend += 4;
948 if (tunnel->parms.o_flags&GRE_KEY)
949 addend += 4;
950 if (tunnel->parms.o_flags&GRE_SEQ)
951 addend += 4;
952 }
Herbert Xuc95b8192008-10-09 11:58:54 -0700953 dev->needed_headroom = addend + hlen;
Herbert Xu42aa9162008-10-09 11:59:32 -0700954 mtu -= dev->hard_header_len - addend;
955
956 if (mtu < 68)
957 mtu = 68;
958
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800959 tunnel->hlen = addend;
960
Herbert Xu42aa9162008-10-09 11:59:32 -0700961 return mtu;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800962}
963
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964static int
965ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
966{
967 int err = 0;
968 struct ip_tunnel_parm p;
969 struct ip_tunnel *t;
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700970 struct net *net = dev_net(dev);
971 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
973 switch (cmd) {
974 case SIOCGETTUNNEL:
975 t = NULL;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700976 if (dev == ign->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
978 err = -EFAULT;
979 break;
980 }
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700981 t = ipgre_tunnel_locate(net, &p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 }
983 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -0800984 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 memcpy(&p, &t->parms, sizeof(p));
986 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
987 err = -EFAULT;
988 break;
989
990 case SIOCADDTUNNEL:
991 case SIOCCHGTUNNEL:
992 err = -EPERM;
993 if (!capable(CAP_NET_ADMIN))
994 goto done;
995
996 err = -EFAULT;
997 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
998 goto done;
999
1000 err = -EINVAL;
1001 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
1002 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
1003 ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)))
1004 goto done;
1005 if (p.iph.ttl)
1006 p.iph.frag_off |= htons(IP_DF);
1007
1008 if (!(p.i_flags&GRE_KEY))
1009 p.i_key = 0;
1010 if (!(p.o_flags&GRE_KEY))
1011 p.o_key = 0;
1012
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001013 t = ipgre_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001015 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 if (t != NULL) {
1017 if (t->dev != dev) {
1018 err = -EEXIST;
1019 break;
1020 }
1021 } else {
Jianjun Kong6ed25332008-11-03 00:25:16 -08001022 unsigned nflags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023
Patrick McHardy2941a482006-01-08 22:05:26 -08001024 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
Joe Perchesf97c1e02007-12-16 13:45:43 -08001026 if (ipv4_is_multicast(p.iph.daddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 nflags = IFF_BROADCAST;
1028 else if (p.iph.daddr)
1029 nflags = IFF_POINTOPOINT;
1030
1031 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
1032 err = -EINVAL;
1033 break;
1034 }
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001035 ipgre_tunnel_unlink(ign, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 t->parms.iph.saddr = p.iph.saddr;
1037 t->parms.iph.daddr = p.iph.daddr;
1038 t->parms.i_key = p.i_key;
1039 t->parms.o_key = p.o_key;
1040 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1041 memcpy(dev->broadcast, &p.iph.daddr, 4);
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001042 ipgre_tunnel_link(ign, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 netdev_state_change(dev);
1044 }
1045 }
1046
1047 if (t) {
1048 err = 0;
1049 if (cmd == SIOCCHGTUNNEL) {
1050 t->parms.iph.ttl = p.iph.ttl;
1051 t->parms.iph.tos = p.iph.tos;
1052 t->parms.iph.frag_off = p.iph.frag_off;
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001053 if (t->parms.link != p.link) {
1054 t->parms.link = p.link;
Herbert Xu42aa9162008-10-09 11:59:32 -07001055 dev->mtu = ipgre_tunnel_bind_dev(dev);
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001056 netdev_state_change(dev);
1057 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 }
1059 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1060 err = -EFAULT;
1061 } else
1062 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1063 break;
1064
1065 case SIOCDELTUNNEL:
1066 err = -EPERM;
1067 if (!capable(CAP_NET_ADMIN))
1068 goto done;
1069
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001070 if (dev == ign->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 err = -EFAULT;
1072 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1073 goto done;
1074 err = -ENOENT;
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001075 if ((t = ipgre_tunnel_locate(net, &p, 0)) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 goto done;
1077 err = -EPERM;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001078 if (t == netdev_priv(ign->fb_tunnel_dev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 goto done;
1080 dev = t->dev;
1081 }
Stephen Hemminger22f8cde2007-02-07 00:09:58 -08001082 unregister_netdevice(dev);
1083 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 break;
1085
1086 default:
1087 err = -EINVAL;
1088 }
1089
1090done:
1091 return err;
1092}
1093
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1095{
Patrick McHardy2941a482006-01-08 22:05:26 -08001096 struct ip_tunnel *tunnel = netdev_priv(dev);
Herbert Xuc95b8192008-10-09 11:58:54 -07001097 if (new_mtu < 68 ||
1098 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 return -EINVAL;
1100 dev->mtu = new_mtu;
1101 return 0;
1102}
1103
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104/* Nice toy. Unfortunately, useless in real life :-)
1105 It allows to construct virtual multiprotocol broadcast "LAN"
1106 over the Internet, provided multicast routing is tuned.
1107
1108
1109 I have no idea was this bicycle invented before me,
1110 so that I had to set ARPHRD_IPGRE to a random value.
1111 I have an impression, that Cisco could make something similar,
1112 but this feature is apparently missing in IOS<=11.2(8).
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001113
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
1115 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
1116
1117 ping -t 255 224.66.66.66
1118
1119 If nobody answers, mbone does not work.
1120
1121 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
1122 ip addr add 10.66.66.<somewhat>/24 dev Universe
1123 ifconfig Universe up
1124 ifconfig Universe add fe80::<Your_real_addr>/10
1125 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
1126 ftp 10.66.66.66
1127 ...
1128 ftp fec0:6666:6666::193.233.7.65
1129 ...
1130
1131 */
1132
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001133static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
1134 unsigned short type,
1135 const void *daddr, const void *saddr, unsigned len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136{
Patrick McHardy2941a482006-01-08 22:05:26 -08001137 struct ip_tunnel *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen);
Al Virod5a0a1e2006-11-08 00:23:14 -08001139 __be16 *p = (__be16*)(iph+1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
1141 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
1142 p[0] = t->parms.o_flags;
1143 p[1] = htons(type);
1144
1145 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001146 * Set the source hardware address.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001148
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 if (saddr)
1150 memcpy(&iph->saddr, saddr, 4);
1151
1152 if (daddr) {
1153 memcpy(&iph->daddr, daddr, 4);
1154 return t->hlen;
1155 }
Joe Perchesf97c1e02007-12-16 13:45:43 -08001156 if (iph->daddr && !ipv4_is_multicast(iph->daddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 return t->hlen;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001158
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 return -t->hlen;
1160}
1161
Timo Teras6a5f44d2007-10-23 20:31:53 -07001162static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
1163{
Jianjun Kong6ed25332008-11-03 00:25:16 -08001164 struct iphdr *iph = (struct iphdr *) skb_mac_header(skb);
Timo Teras6a5f44d2007-10-23 20:31:53 -07001165 memcpy(haddr, &iph->saddr, 4);
1166 return 4;
1167}
1168
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001169static const struct header_ops ipgre_header_ops = {
1170 .create = ipgre_header,
Timo Teras6a5f44d2007-10-23 20:31:53 -07001171 .parse = ipgre_header_parse,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001172};
1173
Timo Teras6a5f44d2007-10-23 20:31:53 -07001174#ifdef CONFIG_NET_IPGRE_BROADCAST
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175static int ipgre_open(struct net_device *dev)
1176{
Patrick McHardy2941a482006-01-08 22:05:26 -08001177 struct ip_tunnel *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178
Joe Perchesf97c1e02007-12-16 13:45:43 -08001179 if (ipv4_is_multicast(t->parms.iph.daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 struct flowi fl = { .oif = t->parms.link,
1181 .nl_u = { .ip4_u =
1182 { .daddr = t->parms.iph.daddr,
1183 .saddr = t->parms.iph.saddr,
1184 .tos = RT_TOS(t->parms.iph.tos) } },
1185 .proto = IPPROTO_GRE };
1186 struct rtable *rt;
Pavel Emelyanov96635522008-04-16 01:10:44 -07001187 if (ip_route_output_key(dev_net(dev), &rt, &fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 return -EADDRNOTAVAIL;
1189 dev = rt->u.dst.dev;
1190 ip_rt_put(rt);
Herbert Xue5ed6392005-10-03 14:35:55 -07001191 if (__in_dev_get_rtnl(dev) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 return -EADDRNOTAVAIL;
1193 t->mlink = dev->ifindex;
Herbert Xue5ed6392005-10-03 14:35:55 -07001194 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 }
1196 return 0;
1197}
1198
1199static int ipgre_close(struct net_device *dev)
1200{
Patrick McHardy2941a482006-01-08 22:05:26 -08001201 struct ip_tunnel *t = netdev_priv(dev);
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001202
Joe Perchesf97c1e02007-12-16 13:45:43 -08001203 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
Denis V. Lunev7fee0ca2008-01-21 17:32:38 -08001204 struct in_device *in_dev;
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001205 in_dev = inetdev_by_index(dev_net(dev), t->mlink);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 if (in_dev) {
1207 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
1208 in_dev_put(in_dev);
1209 }
1210 }
1211 return 0;
1212}
1213
1214#endif
1215
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001216static const struct net_device_ops ipgre_netdev_ops = {
1217 .ndo_init = ipgre_tunnel_init,
1218 .ndo_uninit = ipgre_tunnel_uninit,
1219#ifdef CONFIG_NET_IPGRE_BROADCAST
1220 .ndo_open = ipgre_open,
1221 .ndo_stop = ipgre_close,
1222#endif
1223 .ndo_start_xmit = ipgre_tunnel_xmit,
1224 .ndo_do_ioctl = ipgre_tunnel_ioctl,
1225 .ndo_change_mtu = ipgre_tunnel_change_mtu,
1226};
1227
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228static void ipgre_tunnel_setup(struct net_device *dev)
1229{
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001230 dev->netdev_ops = &ipgre_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 dev->destructor = free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232
1233 dev->type = ARPHRD_IPGRE;
Herbert Xuc95b8192008-10-09 11:58:54 -07001234 dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
Kris Katterjohn46f25df2006-01-05 16:35:42 -08001235 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 dev->flags = IFF_NOARP;
1237 dev->iflink = 0;
1238 dev->addr_len = 4;
Pavel Emelyanov0b67ece2008-04-16 01:11:13 -07001239 dev->features |= NETIF_F_NETNS_LOCAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240}
1241
1242static int ipgre_tunnel_init(struct net_device *dev)
1243{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 struct ip_tunnel *tunnel;
1245 struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
Patrick McHardy2941a482006-01-08 22:05:26 -08001247 tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 iph = &tunnel->parms.iph;
1249
1250 tunnel->dev = dev;
1251 strcpy(tunnel->parms.name, dev->name);
1252
1253 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1254 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1255
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 if (iph->daddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257#ifdef CONFIG_NET_IPGRE_BROADCAST
Joe Perchesf97c1e02007-12-16 13:45:43 -08001258 if (ipv4_is_multicast(iph->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 if (!iph->saddr)
1260 return -EINVAL;
1261 dev->flags = IFF_BROADCAST;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001262 dev->header_ops = &ipgre_header_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 }
1264#endif
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001265 } else
Timo Teras6a5f44d2007-10-23 20:31:53 -07001266 dev->header_ops = &ipgre_header_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 return 0;
1269}
1270
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001271static void ipgre_fb_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272{
Patrick McHardy2941a482006-01-08 22:05:26 -08001273 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 struct iphdr *iph = &tunnel->parms.iph;
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001275 struct ipgre_net *ign = net_generic(dev_net(dev), ipgre_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
1277 tunnel->dev = dev;
1278 strcpy(tunnel->parms.name, dev->name);
1279
1280 iph->version = 4;
1281 iph->protocol = IPPROTO_GRE;
1282 iph->ihl = 5;
1283 tunnel->hlen = sizeof(struct iphdr) + 4;
1284
1285 dev_hold(dev);
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001286 ign->tunnels_wc[0] = tunnel;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287}
1288
1289
1290static struct net_protocol ipgre_protocol = {
1291 .handler = ipgre_rcv,
1292 .err_handler = ipgre_err,
Pavel Emelyanovf96c1482008-04-16 01:11:36 -07001293 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294};
1295
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001296static void ipgre_destroy_tunnels(struct ipgre_net *ign)
1297{
1298 int prio;
1299
1300 for (prio = 0; prio < 4; prio++) {
1301 int h;
1302 for (h = 0; h < HASH_SIZE; h++) {
1303 struct ip_tunnel *t;
1304 while ((t = ign->tunnels[prio][h]) != NULL)
1305 unregister_netdevice(t->dev);
1306 }
1307 }
1308}
1309
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001310static int ipgre_init_net(struct net *net)
1311{
1312 int err;
1313 struct ipgre_net *ign;
1314
1315 err = -ENOMEM;
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001316 ign = kzalloc(sizeof(struct ipgre_net), GFP_KERNEL);
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001317 if (ign == NULL)
1318 goto err_alloc;
1319
1320 err = net_assign_generic(net, ipgre_net_id, ign);
1321 if (err < 0)
1322 goto err_assign;
1323
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001324 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "gre0",
1325 ipgre_tunnel_setup);
1326 if (!ign->fb_tunnel_dev) {
1327 err = -ENOMEM;
1328 goto err_alloc_dev;
1329 }
Alexey Dobriyanbe77e592008-11-23 17:26:26 -08001330 dev_net_set(ign->fb_tunnel_dev, net);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001331
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001332 ipgre_fb_tunnel_init(ign->fb_tunnel_dev);
Herbert Xuc19e6542008-10-09 11:59:55 -07001333 ign->fb_tunnel_dev->rtnl_link_ops = &ipgre_link_ops;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001334
1335 if ((err = register_netdev(ign->fb_tunnel_dev)))
1336 goto err_reg_dev;
1337
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001338 return 0;
1339
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001340err_reg_dev:
1341 free_netdev(ign->fb_tunnel_dev);
1342err_alloc_dev:
1343 /* nothing */
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001344err_assign:
1345 kfree(ign);
1346err_alloc:
1347 return err;
1348}
1349
1350static void ipgre_exit_net(struct net *net)
1351{
1352 struct ipgre_net *ign;
1353
1354 ign = net_generic(net, ipgre_net_id);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001355 rtnl_lock();
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001356 ipgre_destroy_tunnels(ign);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001357 rtnl_unlock();
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001358 kfree(ign);
1359}
1360
1361static struct pernet_operations ipgre_net_ops = {
1362 .init = ipgre_init_net,
1363 .exit = ipgre_exit_net,
1364};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Herbert Xuc19e6542008-10-09 11:59:55 -07001366static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1367{
1368 __be16 flags;
1369
1370 if (!data)
1371 return 0;
1372
1373 flags = 0;
1374 if (data[IFLA_GRE_IFLAGS])
1375 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1376 if (data[IFLA_GRE_OFLAGS])
1377 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1378 if (flags & (GRE_VERSION|GRE_ROUTING))
1379 return -EINVAL;
1380
1381 return 0;
1382}
1383
Herbert Xue1a80002008-10-09 12:00:17 -07001384static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1385{
1386 __be32 daddr;
1387
1388 if (tb[IFLA_ADDRESS]) {
1389 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1390 return -EINVAL;
1391 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1392 return -EADDRNOTAVAIL;
1393 }
1394
1395 if (!data)
1396 goto out;
1397
1398 if (data[IFLA_GRE_REMOTE]) {
1399 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1400 if (!daddr)
1401 return -EINVAL;
1402 }
1403
1404out:
1405 return ipgre_tunnel_validate(tb, data);
1406}
1407
Herbert Xuc19e6542008-10-09 11:59:55 -07001408static void ipgre_netlink_parms(struct nlattr *data[],
1409 struct ip_tunnel_parm *parms)
1410{
Herbert Xu7bb82d92008-10-11 12:20:15 -07001411 memset(parms, 0, sizeof(*parms));
Herbert Xuc19e6542008-10-09 11:59:55 -07001412
1413 parms->iph.protocol = IPPROTO_GRE;
1414
1415 if (!data)
1416 return;
1417
1418 if (data[IFLA_GRE_LINK])
1419 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1420
1421 if (data[IFLA_GRE_IFLAGS])
1422 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1423
1424 if (data[IFLA_GRE_OFLAGS])
1425 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1426
1427 if (data[IFLA_GRE_IKEY])
1428 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1429
1430 if (data[IFLA_GRE_OKEY])
1431 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1432
1433 if (data[IFLA_GRE_LOCAL])
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001434 parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]);
Herbert Xuc19e6542008-10-09 11:59:55 -07001435
1436 if (data[IFLA_GRE_REMOTE])
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001437 parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]);
Herbert Xuc19e6542008-10-09 11:59:55 -07001438
1439 if (data[IFLA_GRE_TTL])
1440 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1441
1442 if (data[IFLA_GRE_TOS])
1443 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1444
1445 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
1446 parms->iph.frag_off = htons(IP_DF);
1447}
1448
Herbert Xue1a80002008-10-09 12:00:17 -07001449static int ipgre_tap_init(struct net_device *dev)
1450{
1451 struct ip_tunnel *tunnel;
1452
1453 tunnel = netdev_priv(dev);
1454
1455 tunnel->dev = dev;
1456 strcpy(tunnel->parms.name, dev->name);
1457
1458 ipgre_tunnel_bind_dev(dev);
1459
1460 return 0;
1461}
1462
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001463static const struct net_device_ops ipgre_tap_netdev_ops = {
1464 .ndo_init = ipgre_tap_init,
1465 .ndo_uninit = ipgre_tunnel_uninit,
1466 .ndo_start_xmit = ipgre_tunnel_xmit,
1467 .ndo_set_mac_address = eth_mac_addr,
1468 .ndo_validate_addr = eth_validate_addr,
1469 .ndo_change_mtu = ipgre_tunnel_change_mtu,
1470};
1471
Herbert Xue1a80002008-10-09 12:00:17 -07001472static void ipgre_tap_setup(struct net_device *dev)
1473{
1474
1475 ether_setup(dev);
1476
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001477 dev->netdev_ops = &ipgre_netdev_ops;
Herbert Xue1a80002008-10-09 12:00:17 -07001478 dev->destructor = free_netdev;
Herbert Xue1a80002008-10-09 12:00:17 -07001479
1480 dev->iflink = 0;
1481 dev->features |= NETIF_F_NETNS_LOCAL;
1482}
1483
Herbert Xuc19e6542008-10-09 11:59:55 -07001484static int ipgre_newlink(struct net_device *dev, struct nlattr *tb[],
1485 struct nlattr *data[])
1486{
1487 struct ip_tunnel *nt;
1488 struct net *net = dev_net(dev);
1489 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1490 int mtu;
1491 int err;
1492
1493 nt = netdev_priv(dev);
1494 ipgre_netlink_parms(data, &nt->parms);
1495
Herbert Xue1a80002008-10-09 12:00:17 -07001496 if (ipgre_tunnel_find(net, &nt->parms, dev->type))
Herbert Xuc19e6542008-10-09 11:59:55 -07001497 return -EEXIST;
1498
Herbert Xue1a80002008-10-09 12:00:17 -07001499 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1500 random_ether_addr(dev->dev_addr);
1501
Herbert Xuc19e6542008-10-09 11:59:55 -07001502 mtu = ipgre_tunnel_bind_dev(dev);
1503 if (!tb[IFLA_MTU])
1504 dev->mtu = mtu;
1505
1506 err = register_netdevice(dev);
1507 if (err)
1508 goto out;
1509
1510 dev_hold(dev);
1511 ipgre_tunnel_link(ign, nt);
1512
1513out:
1514 return err;
1515}
1516
1517static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1518 struct nlattr *data[])
1519{
1520 struct ip_tunnel *t, *nt;
1521 struct net *net = dev_net(dev);
1522 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1523 struct ip_tunnel_parm p;
1524 int mtu;
1525
1526 if (dev == ign->fb_tunnel_dev)
1527 return -EINVAL;
1528
1529 nt = netdev_priv(dev);
1530 ipgre_netlink_parms(data, &p);
1531
1532 t = ipgre_tunnel_locate(net, &p, 0);
1533
1534 if (t) {
1535 if (t->dev != dev)
1536 return -EEXIST;
1537 } else {
1538 unsigned nflags = 0;
1539
1540 t = nt;
1541
1542 if (ipv4_is_multicast(p.iph.daddr))
1543 nflags = IFF_BROADCAST;
1544 else if (p.iph.daddr)
1545 nflags = IFF_POINTOPOINT;
1546
1547 if ((dev->flags ^ nflags) &
1548 (IFF_POINTOPOINT | IFF_BROADCAST))
1549 return -EINVAL;
1550
1551 ipgre_tunnel_unlink(ign, t);
1552 t->parms.iph.saddr = p.iph.saddr;
1553 t->parms.iph.daddr = p.iph.daddr;
1554 t->parms.i_key = p.i_key;
1555 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1556 memcpy(dev->broadcast, &p.iph.daddr, 4);
1557 ipgre_tunnel_link(ign, t);
1558 netdev_state_change(dev);
1559 }
1560
1561 t->parms.o_key = p.o_key;
1562 t->parms.iph.ttl = p.iph.ttl;
1563 t->parms.iph.tos = p.iph.tos;
1564 t->parms.iph.frag_off = p.iph.frag_off;
1565
1566 if (t->parms.link != p.link) {
1567 t->parms.link = p.link;
1568 mtu = ipgre_tunnel_bind_dev(dev);
1569 if (!tb[IFLA_MTU])
1570 dev->mtu = mtu;
1571 netdev_state_change(dev);
1572 }
1573
1574 return 0;
1575}
1576
1577static size_t ipgre_get_size(const struct net_device *dev)
1578{
1579 return
1580 /* IFLA_GRE_LINK */
1581 nla_total_size(4) +
1582 /* IFLA_GRE_IFLAGS */
1583 nla_total_size(2) +
1584 /* IFLA_GRE_OFLAGS */
1585 nla_total_size(2) +
1586 /* IFLA_GRE_IKEY */
1587 nla_total_size(4) +
1588 /* IFLA_GRE_OKEY */
1589 nla_total_size(4) +
1590 /* IFLA_GRE_LOCAL */
1591 nla_total_size(4) +
1592 /* IFLA_GRE_REMOTE */
1593 nla_total_size(4) +
1594 /* IFLA_GRE_TTL */
1595 nla_total_size(1) +
1596 /* IFLA_GRE_TOS */
1597 nla_total_size(1) +
1598 /* IFLA_GRE_PMTUDISC */
1599 nla_total_size(1) +
1600 0;
1601}
1602
1603static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1604{
1605 struct ip_tunnel *t = netdev_priv(dev);
1606 struct ip_tunnel_parm *p = &t->parms;
1607
1608 NLA_PUT_U32(skb, IFLA_GRE_LINK, p->link);
1609 NLA_PUT_BE16(skb, IFLA_GRE_IFLAGS, p->i_flags);
1610 NLA_PUT_BE16(skb, IFLA_GRE_OFLAGS, p->o_flags);
Patrick McHardyba9e64b2008-10-10 12:10:30 -07001611 NLA_PUT_BE32(skb, IFLA_GRE_IKEY, p->i_key);
1612 NLA_PUT_BE32(skb, IFLA_GRE_OKEY, p->o_key);
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001613 NLA_PUT_BE32(skb, IFLA_GRE_LOCAL, p->iph.saddr);
1614 NLA_PUT_BE32(skb, IFLA_GRE_REMOTE, p->iph.daddr);
Herbert Xuc19e6542008-10-09 11:59:55 -07001615 NLA_PUT_U8(skb, IFLA_GRE_TTL, p->iph.ttl);
1616 NLA_PUT_U8(skb, IFLA_GRE_TOS, p->iph.tos);
1617 NLA_PUT_U8(skb, IFLA_GRE_PMTUDISC, !!(p->iph.frag_off & htons(IP_DF)));
1618
1619 return 0;
1620
1621nla_put_failure:
1622 return -EMSGSIZE;
1623}
1624
1625static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1626 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1627 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1628 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1629 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1630 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001631 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1632 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
Herbert Xuc19e6542008-10-09 11:59:55 -07001633 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1634 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1635 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1636};
1637
1638static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1639 .kind = "gre",
1640 .maxtype = IFLA_GRE_MAX,
1641 .policy = ipgre_policy,
1642 .priv_size = sizeof(struct ip_tunnel),
1643 .setup = ipgre_tunnel_setup,
1644 .validate = ipgre_tunnel_validate,
1645 .newlink = ipgre_newlink,
1646 .changelink = ipgre_changelink,
1647 .get_size = ipgre_get_size,
1648 .fill_info = ipgre_fill_info,
1649};
1650
Herbert Xue1a80002008-10-09 12:00:17 -07001651static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1652 .kind = "gretap",
1653 .maxtype = IFLA_GRE_MAX,
1654 .policy = ipgre_policy,
1655 .priv_size = sizeof(struct ip_tunnel),
1656 .setup = ipgre_tap_setup,
1657 .validate = ipgre_tap_validate,
1658 .newlink = ipgre_newlink,
1659 .changelink = ipgre_changelink,
1660 .get_size = ipgre_get_size,
1661 .fill_info = ipgre_fill_info,
1662};
1663
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664/*
1665 * And now the modules code and kernel interface.
1666 */
1667
1668static int __init ipgre_init(void)
1669{
1670 int err;
1671
1672 printk(KERN_INFO "GRE over IPv4 tunneling driver\n");
1673
1674 if (inet_add_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) {
1675 printk(KERN_INFO "ipgre init: can't add protocol\n");
1676 return -EAGAIN;
1677 }
1678
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001679 err = register_pernet_gen_device(&ipgre_net_id, &ipgre_net_ops);
1680 if (err < 0)
Herbert Xuc19e6542008-10-09 11:59:55 -07001681 goto gen_device_failed;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001682
Herbert Xuc19e6542008-10-09 11:59:55 -07001683 err = rtnl_link_register(&ipgre_link_ops);
1684 if (err < 0)
1685 goto rtnl_link_failed;
1686
Herbert Xue1a80002008-10-09 12:00:17 -07001687 err = rtnl_link_register(&ipgre_tap_ops);
1688 if (err < 0)
1689 goto tap_ops_failed;
1690
Herbert Xuc19e6542008-10-09 11:59:55 -07001691out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 return err;
Herbert Xuc19e6542008-10-09 11:59:55 -07001693
Herbert Xue1a80002008-10-09 12:00:17 -07001694tap_ops_failed:
1695 rtnl_link_unregister(&ipgre_link_ops);
Herbert Xuc19e6542008-10-09 11:59:55 -07001696rtnl_link_failed:
1697 unregister_pernet_gen_device(ipgre_net_id, &ipgre_net_ops);
1698gen_device_failed:
1699 inet_del_protocol(&ipgre_protocol, IPPROTO_GRE);
1700 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701}
1702
Alexey Kuznetsovdb445752005-07-30 17:46:44 -07001703static void __exit ipgre_fini(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704{
Herbert Xue1a80002008-10-09 12:00:17 -07001705 rtnl_link_unregister(&ipgre_tap_ops);
Herbert Xuc19e6542008-10-09 11:59:55 -07001706 rtnl_link_unregister(&ipgre_link_ops);
1707 unregister_pernet_gen_device(ipgre_net_id, &ipgre_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0)
1709 printk(KERN_INFO "ipgre close: can't remove protocol\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710}
1711
1712module_init(ipgre_init);
1713module_exit(ipgre_fini);
1714MODULE_LICENSE("GPL");
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001715MODULE_ALIAS_RTNL_LINK("gre");
1716MODULE_ALIAS_RTNL_LINK("gretap");