blob: 38673d2860e293404a9109c88ec9b03ab6b810f1 [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090017#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <asm/uaccess.h>
19#include <linux/skbuff.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/tcp.h>
23#include <linux/udp.h>
24#include <linux/if_arp.h>
25#include <linux/mroute.h>
26#include <linux/init.h>
27#include <linux/in6.h>
28#include <linux/inetdevice.h>
29#include <linux/igmp.h>
30#include <linux/netfilter_ipv4.h>
Herbert Xue1a80002008-10-09 12:00:17 -070031#include <linux/etherdevice.h>
Kris Katterjohn46f25df2006-01-05 16:35:42 -080032#include <linux/if_ether.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
34#include <net/sock.h>
35#include <net/ip.h>
36#include <net/icmp.h>
37#include <net/protocol.h>
38#include <net/ipip.h>
39#include <net/arp.h>
40#include <net/checksum.h>
41#include <net/dsfield.h>
42#include <net/inet_ecn.h>
43#include <net/xfrm.h>
Pavel Emelyanov59a4c752008-04-16 01:08:53 -070044#include <net/net_namespace.h>
45#include <net/netns/generic.h>
Herbert Xuc19e6542008-10-09 11:59:55 -070046#include <net/rtnetlink.h>
Dmitry Kozlov00959ad2010-08-21 23:05:39 -070047#include <net/gre.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Eric Dumazetdfd56b82011-12-10 09:48:31 +000049#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <net/ipv6.h>
51#include <net/ip6_fib.h>
52#include <net/ip6_route.h>
53#endif
54
55/*
56 Problems & solutions
57 --------------------
58
59 1. The most important issue is detecting local dead loops.
60 They would cause complete host lockup in transmit, which
61 would be "resolved" by stack overflow or, if queueing is enabled,
62 with infinite looping in net_bh.
63
64 We cannot track such dead loops during route installation,
65 it is infeasible task. The most general solutions would be
66 to keep skb->encapsulation counter (sort of local ttl),
Eric Dumazet6d0722a2010-09-29 23:35:10 -070067 and silently drop packet when it expires. It is a good
stephen hemmingerbff52852012-02-24 08:08:20 +000068 solution, but it supposes maintaining new variable in ALL
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 skb, even if no tunneling is used.
70
Eric Dumazet6d0722a2010-09-29 23:35:10 -070071 Current solution: xmit_recursion breaks dead loops. This is a percpu
72 counter, since when we enter the first ndo_xmit(), cpu migration is
73 forbidden. We force an exit if this counter reaches RECURSION_LIMIT
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
75 2. Networking dead loops would not kill routers, but would really
76 kill network. IP hop limit plays role of "t->recursion" in this case,
77 if we copy it from packet being encapsulated to upper header.
78 It is very good solution, but it introduces two problems:
79
80 - Routing protocols, using packets with ttl=1 (OSPF, RIP2),
81 do not work over tunnels.
82 - traceroute does not work. I planned to relay ICMP from tunnel,
83 so that this problem would be solved and traceroute output
84 would even more informative. This idea appeared to be wrong:
85 only Linux complies to rfc1812 now (yes, guys, Linux is the only
86 true router now :-)), all routers (at least, in neighbourhood of mine)
87 return only 8 bytes of payload. It is the end.
88
89 Hence, if we want that OSPF worked or traceroute said something reasonable,
90 we should search for another solution.
91
92 One of them is to parse packet trying to detect inner encapsulation
93 made by our node. It is difficult or even impossible, especially,
stephen hemmingerbff52852012-02-24 08:08:20 +000094 taking into account fragmentation. TO be short, ttl is not solution at all.
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96 Current solution: The solution was UNEXPECTEDLY SIMPLE.
97 We force DF flag on tunnels with preconfigured hop limit,
98 that is ALL. :-) Well, it does not remove the problem completely,
99 but exponential growth of network traffic is changed to linear
100 (branches, that exceed pmtu are pruned) and tunnel mtu
stephen hemmingerbff52852012-02-24 08:08:20 +0000101 rapidly degrades to value <68, where looping stops.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 Yes, it is not good if there exists a router in the loop,
103 which does not force DF, even when encapsulating packets have DF set.
104 But it is not our problem! Nobody could accuse us, we made
105 all that we could make. Even if it is your gated who injected
106 fatal route to network, even if it were you who configured
107 fatal static route: you are innocent. :-)
108
109
110
111 3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain
112 practically identical code. It would be good to glue them
113 together, but it is not very evident, how to make them modular.
114 sit is integral part of IPv6, ipip and gre are naturally modular.
115 We could extract common parts (hash table, ioctl etc)
116 to a separate module (ip_tunnel.c).
117
118 Alexey Kuznetsov.
119 */
120
Herbert Xuc19e6542008-10-09 11:59:55 -0700121static struct rtnl_link_ops ipgre_link_ops __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122static int ipgre_tunnel_init(struct net_device *dev);
123static void ipgre_tunnel_setup(struct net_device *dev);
Herbert Xu42aa9162008-10-09 11:59:32 -0700124static int ipgre_tunnel_bind_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126/* Fallback tunnel: no source, no destination, no key, no options */
127
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700128#define HASH_SIZE 16
129
Eric Dumazetf99189b2009-11-17 10:42:49 +0000130static int ipgre_net_id __read_mostly;
Pavel Emelyanov59a4c752008-04-16 01:08:53 -0700131struct ipgre_net {
Eric Dumazet15078502010-09-15 11:07:53 +0000132 struct ip_tunnel __rcu *tunnels[4][HASH_SIZE];
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700133
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700134 struct net_device *fb_tunnel_dev;
Pavel Emelyanov59a4c752008-04-16 01:08:53 -0700135};
136
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137/* Tunnel hash table */
138
139/*
140 4 hash tables:
141
142 3: (remote,local)
143 2: (remote,*)
144 1: (*,local)
145 0: (*,*)
146
147 We require exact key match i.e. if a key is present in packet
148 it will match only tunnel with the same key; if it is not present,
149 it will match only keyless tunnel.
150
151 All keysless packets, if not matched configured keyless tunnels
152 will match fallback tunnel.
153 */
154
Al Virod5a0a1e2006-11-08 00:23:14 -0800155#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700157#define tunnels_r_l tunnels[3]
158#define tunnels_r tunnels[2]
159#define tunnels_l tunnels[1]
160#define tunnels_wc tunnels[0]
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000161/*
Eric Dumazet15078502010-09-15 11:07:53 +0000162 * Locking : hash tables are protected by RCU and RTNL
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000163 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000165#define for_each_ip_tunnel_rcu(start) \
166 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
Eric Dumazete985aad2010-09-27 03:57:11 +0000168/* often modified stats are per cpu, other are shared (netdev->stats) */
169struct pcpu_tstats {
170 unsigned long rx_packets;
171 unsigned long rx_bytes;
172 unsigned long tx_packets;
173 unsigned long tx_bytes;
Eric Dumazet8ce120f2011-11-04 23:19:28 +0000174} __attribute__((aligned(4*sizeof(unsigned long))));
Eric Dumazete985aad2010-09-27 03:57:11 +0000175
176static struct net_device_stats *ipgre_get_stats(struct net_device *dev)
177{
178 struct pcpu_tstats sum = { 0 };
179 int i;
180
181 for_each_possible_cpu(i) {
182 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
183
184 sum.rx_packets += tstats->rx_packets;
185 sum.rx_bytes += tstats->rx_bytes;
186 sum.tx_packets += tstats->tx_packets;
187 sum.tx_bytes += tstats->tx_bytes;
188 }
189 dev->stats.rx_packets = sum.rx_packets;
190 dev->stats.rx_bytes = sum.rx_bytes;
191 dev->stats.tx_packets = sum.tx_packets;
192 dev->stats.tx_bytes = sum.tx_bytes;
193 return &dev->stats;
194}
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196/* Given src, dst and key, find appropriate for input tunnel. */
197
Timo Teras749c10f2009-01-19 17:22:12 -0800198static struct ip_tunnel * ipgre_tunnel_lookup(struct net_device *dev,
Herbert Xue1a80002008-10-09 12:00:17 -0700199 __be32 remote, __be32 local,
200 __be32 key, __be16 gre_proto)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
Timo Teras749c10f2009-01-19 17:22:12 -0800202 struct net *net = dev_net(dev);
203 int link = dev->ifindex;
Eric Dumazet15078502010-09-15 11:07:53 +0000204 unsigned int h0 = HASH(remote);
205 unsigned int h1 = HASH(key);
Timo Terasafcf1242009-01-26 20:56:10 -0800206 struct ip_tunnel *t, *cand = NULL;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700207 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Herbert Xue1a80002008-10-09 12:00:17 -0700208 int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
209 ARPHRD_ETHER : ARPHRD_IPGRE;
Timo Terasafcf1242009-01-26 20:56:10 -0800210 int score, cand_score = 4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000212 for_each_ip_tunnel_rcu(ign->tunnels_r_l[h0 ^ h1]) {
Timo Teras749c10f2009-01-19 17:22:12 -0800213 if (local != t->parms.iph.saddr ||
214 remote != t->parms.iph.daddr ||
215 key != t->parms.i_key ||
216 !(t->dev->flags & IFF_UP))
217 continue;
218
219 if (t->dev->type != ARPHRD_IPGRE &&
220 t->dev->type != dev_type)
221 continue;
222
Timo Terasafcf1242009-01-26 20:56:10 -0800223 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800224 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800225 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800226 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800227 score |= 2;
228 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800229 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800230
231 if (score < cand_score) {
232 cand = t;
233 cand_score = score;
234 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 }
Herbert Xue1a80002008-10-09 12:00:17 -0700236
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000237 for_each_ip_tunnel_rcu(ign->tunnels_r[h0 ^ h1]) {
Timo Teras749c10f2009-01-19 17:22:12 -0800238 if (remote != t->parms.iph.daddr ||
239 key != t->parms.i_key ||
240 !(t->dev->flags & IFF_UP))
241 continue;
242
243 if (t->dev->type != ARPHRD_IPGRE &&
244 t->dev->type != dev_type)
245 continue;
246
Timo Terasafcf1242009-01-26 20:56:10 -0800247 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800248 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800249 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800250 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800251 score |= 2;
252 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800253 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800254
255 if (score < cand_score) {
256 cand = t;
257 cand_score = score;
258 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 }
Herbert Xue1a80002008-10-09 12:00:17 -0700260
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000261 for_each_ip_tunnel_rcu(ign->tunnels_l[h1]) {
Timo Teras749c10f2009-01-19 17:22:12 -0800262 if ((local != t->parms.iph.saddr &&
263 (local != t->parms.iph.daddr ||
264 !ipv4_is_multicast(local))) ||
265 key != t->parms.i_key ||
266 !(t->dev->flags & IFF_UP))
267 continue;
268
269 if (t->dev->type != ARPHRD_IPGRE &&
270 t->dev->type != dev_type)
271 continue;
272
Timo Terasafcf1242009-01-26 20:56:10 -0800273 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800274 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800275 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800276 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800277 score |= 2;
278 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800279 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800280
281 if (score < cand_score) {
282 cand = t;
283 cand_score = score;
284 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 }
Herbert Xue1a80002008-10-09 12:00:17 -0700286
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000287 for_each_ip_tunnel_rcu(ign->tunnels_wc[h1]) {
Timo Teras749c10f2009-01-19 17:22:12 -0800288 if (t->parms.i_key != key ||
289 !(t->dev->flags & IFF_UP))
290 continue;
291
292 if (t->dev->type != ARPHRD_IPGRE &&
293 t->dev->type != dev_type)
294 continue;
295
Timo Terasafcf1242009-01-26 20:56:10 -0800296 score = 0;
Timo Teras749c10f2009-01-19 17:22:12 -0800297 if (t->parms.link != link)
Timo Terasafcf1242009-01-26 20:56:10 -0800298 score |= 1;
Timo Teras749c10f2009-01-19 17:22:12 -0800299 if (t->dev->type != dev_type)
Timo Terasafcf1242009-01-26 20:56:10 -0800300 score |= 2;
301 if (score == 0)
Timo Teras749c10f2009-01-19 17:22:12 -0800302 return t;
Timo Terasafcf1242009-01-26 20:56:10 -0800303
304 if (score < cand_score) {
305 cand = t;
306 cand_score = score;
307 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 }
309
Timo Terasafcf1242009-01-26 20:56:10 -0800310 if (cand != NULL)
311 return cand;
Herbert Xue1a80002008-10-09 12:00:17 -0700312
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000313 dev = ign->fb_tunnel_dev;
314 if (dev->flags & IFF_UP)
315 return netdev_priv(dev);
Timo Teras749c10f2009-01-19 17:22:12 -0800316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 return NULL;
318}
319
Eric Dumazet15078502010-09-15 11:07:53 +0000320static struct ip_tunnel __rcu **__ipgre_bucket(struct ipgre_net *ign,
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700321 struct ip_tunnel_parm *parms)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322{
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900323 __be32 remote = parms->iph.daddr;
324 __be32 local = parms->iph.saddr;
325 __be32 key = parms->i_key;
Eric Dumazet15078502010-09-15 11:07:53 +0000326 unsigned int h = HASH(key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 int prio = 0;
328
329 if (local)
330 prio |= 1;
Joe Perchesf97c1e02007-12-16 13:45:43 -0800331 if (remote && !ipv4_is_multicast(remote)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 prio |= 2;
333 h ^= HASH(remote);
334 }
335
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -0700336 return &ign->tunnels[prio][h];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337}
338
Eric Dumazet15078502010-09-15 11:07:53 +0000339static inline struct ip_tunnel __rcu **ipgre_bucket(struct ipgre_net *ign,
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700340 struct ip_tunnel *t)
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900341{
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700342 return __ipgre_bucket(ign, &t->parms);
YOSHIFUJI Hideaki5056a1e2007-04-24 20:44:48 +0900343}
344
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700345static void ipgre_tunnel_link(struct ipgre_net *ign, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346{
Eric Dumazet15078502010-09-15 11:07:53 +0000347 struct ip_tunnel __rcu **tp = ipgre_bucket(ign, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
Eric Dumazet15078502010-09-15 11:07:53 +0000349 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000350 rcu_assign_pointer(*tp, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351}
352
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700353static void ipgre_tunnel_unlink(struct ipgre_net *ign, struct ip_tunnel *t)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
Eric Dumazet15078502010-09-15 11:07:53 +0000355 struct ip_tunnel __rcu **tp;
356 struct ip_tunnel *iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
Eric Dumazet15078502010-09-15 11:07:53 +0000358 for (tp = ipgre_bucket(ign, t);
359 (iter = rtnl_dereference(*tp)) != NULL;
360 tp = &iter->next) {
361 if (t == iter) {
362 rcu_assign_pointer(*tp, t->next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 break;
364 }
365 }
366}
367
Herbert Xue1a80002008-10-09 12:00:17 -0700368static struct ip_tunnel *ipgre_tunnel_find(struct net *net,
369 struct ip_tunnel_parm *parms,
370 int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
Al Virod5a0a1e2006-11-08 00:23:14 -0800372 __be32 remote = parms->iph.daddr;
373 __be32 local = parms->iph.saddr;
374 __be32 key = parms->i_key;
Timo Teras749c10f2009-01-19 17:22:12 -0800375 int link = parms->link;
Eric Dumazet15078502010-09-15 11:07:53 +0000376 struct ip_tunnel *t;
377 struct ip_tunnel __rcu **tp;
Herbert Xue1a80002008-10-09 12:00:17 -0700378 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
379
Eric Dumazet15078502010-09-15 11:07:53 +0000380 for (tp = __ipgre_bucket(ign, parms);
381 (t = rtnl_dereference(*tp)) != NULL;
382 tp = &t->next)
Herbert Xue1a80002008-10-09 12:00:17 -0700383 if (local == t->parms.iph.saddr &&
384 remote == t->parms.iph.daddr &&
385 key == t->parms.i_key &&
Timo Teras749c10f2009-01-19 17:22:12 -0800386 link == t->parms.link &&
Herbert Xue1a80002008-10-09 12:00:17 -0700387 type == t->dev->type)
388 break;
389
390 return t;
391}
392
Eric Dumazet15078502010-09-15 11:07:53 +0000393static struct ip_tunnel *ipgre_tunnel_locate(struct net *net,
Herbert Xue1a80002008-10-09 12:00:17 -0700394 struct ip_tunnel_parm *parms, int create)
395{
396 struct ip_tunnel *t, *nt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 char name[IFNAMSIZ];
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700399 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Herbert Xue1a80002008-10-09 12:00:17 -0700401 t = ipgre_tunnel_find(net, parms, ARPHRD_IPGRE);
402 if (t || !create)
403 return t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405 if (parms->name[0])
406 strlcpy(name, parms->name, IFNAMSIZ);
Pavel Emelyanov34cc7ba2008-02-23 20:19:20 -0800407 else
stephen hemminger407d6fc2010-11-29 09:47:47 +0000408 strcpy(name, "gre%d");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
410 dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup);
411 if (!dev)
stephen hemminger407d6fc2010-11-29 09:47:47 +0000412 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413
Pavel Emelyanov0b67ece2008-04-16 01:11:13 -0700414 dev_net_set(dev, net);
415
Patrick McHardy2941a482006-01-08 22:05:26 -0800416 nt = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 nt->parms = *parms;
Herbert Xuc19e6542008-10-09 11:59:55 -0700418 dev->rtnl_link_ops = &ipgre_link_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
Herbert Xu42aa9162008-10-09 11:59:32 -0700420 dev->mtu = ipgre_tunnel_bind_dev(dev);
421
Pavel Emelyanovb37d4282008-02-26 23:51:04 -0800422 if (register_netdevice(dev) < 0)
423 goto failed_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424
Willem de Bruijnf2b3ee92012-01-26 10:34:35 +0000425 /* Can use a lockless transmit, unless we generate output sequences */
426 if (!(nt->parms.o_flags & GRE_SEQ))
427 dev->features |= NETIF_F_LLTX;
428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 dev_hold(dev);
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700430 ipgre_tunnel_link(ign, nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 return nt;
432
Pavel Emelyanovb37d4282008-02-26 23:51:04 -0800433failed_free:
434 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 return NULL;
436}
437
438static void ipgre_tunnel_uninit(struct net_device *dev)
439{
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700440 struct net *net = dev_net(dev);
441 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
442
443 ipgre_tunnel_unlink(ign, netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 dev_put(dev);
445}
446
447
448static void ipgre_err(struct sk_buff *skb, u32 info)
449{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Rami Rosen071f92d2008-05-21 17:47:54 -0700451/* All the routers (except for Linux) return only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 8 bytes of packet payload. It means, that precise relaying of
453 ICMP in the real Internet is absolutely infeasible.
454
455 Moreover, Cisco "wise men" put GRE key to the third word
456 in GRE header. It makes impossible maintaining even soft state for keyed
457 GRE tunnels with enabled checksum. Tell them "thank you".
458
459 Well, I wonder, rfc1812 was written by Cisco employee,
stephen hemmingerbff52852012-02-24 08:08:20 +0000460 what the hell these idiots break standards established
461 by themselves???
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 */
463
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000464 const struct iphdr *iph = (const struct iphdr *)skb->data;
Al Virod5a0a1e2006-11-08 00:23:14 -0800465 __be16 *p = (__be16*)(skb->data+(iph->ihl<<2));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 int grehlen = (iph->ihl<<2) + 4;
Arnaldo Carvalho de Melo88c76642007-03-13 14:43:18 -0300467 const int type = icmp_hdr(skb)->type;
468 const int code = icmp_hdr(skb)->code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 struct ip_tunnel *t;
Al Virod5a0a1e2006-11-08 00:23:14 -0800470 __be16 flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 flags = p[0];
473 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
474 if (flags&(GRE_VERSION|GRE_ROUTING))
475 return;
476 if (flags&GRE_KEY) {
477 grehlen += 4;
478 if (flags&GRE_CSUM)
479 grehlen += 4;
480 }
481 }
482
483 /* If only 8 bytes returned, keyed message will be dropped here */
484 if (skb_headlen(skb) < grehlen)
485 return;
486
487 switch (type) {
488 default:
489 case ICMP_PARAMETERPROB:
490 return;
491
492 case ICMP_DEST_UNREACH:
493 switch (code) {
494 case ICMP_SR_FAILED:
495 case ICMP_PORT_UNREACH:
496 /* Impossible event. */
497 return;
498 case ICMP_FRAG_NEEDED:
499 /* Soft state for pmtu is maintained by IP core. */
500 return;
501 default:
502 /* All others are translated to HOST_UNREACH.
503 rfc2003 contains "deep thoughts" about NET_UNREACH,
504 I believe they are just ether pollution. --ANK
505 */
506 break;
507 }
508 break;
509 case ICMP_TIME_EXCEEDED:
510 if (code != ICMP_EXC_TTL)
511 return;
512 break;
513 }
514
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000515 rcu_read_lock();
Timo Teras749c10f2009-01-19 17:22:12 -0800516 t = ipgre_tunnel_lookup(skb->dev, iph->daddr, iph->saddr,
Herbert Xue1a80002008-10-09 12:00:17 -0700517 flags & GRE_KEY ?
518 *(((__be32 *)p) + (grehlen / 4) - 1) : 0,
519 p[1]);
Joe Perchesf97c1e02007-12-16 13:45:43 -0800520 if (t == NULL || t->parms.iph.daddr == 0 ||
521 ipv4_is_multicast(t->parms.iph.daddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 goto out;
523
524 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
525 goto out;
526
Wei Yongjunda6185d82009-02-24 23:34:48 -0800527 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 t->err_count++;
529 else
530 t->err_count = 1;
531 t->err_time = jiffies;
532out:
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000533 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534}
535
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000536static inline void ipgre_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
538 if (INET_ECN_is_ce(iph->tos)) {
539 if (skb->protocol == htons(ETH_P_IP)) {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700540 IP_ECN_set_ce(ip_hdr(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 } else if (skb->protocol == htons(ETH_P_IPV6)) {
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700542 IP6_ECN_set_ce(ipv6_hdr(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 }
544 }
545}
546
547static inline u8
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000548ipgre_ecn_encapsulate(u8 tos, const struct iphdr *old_iph, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549{
550 u8 inner = 0;
551 if (skb->protocol == htons(ETH_P_IP))
552 inner = old_iph->tos;
553 else if (skb->protocol == htons(ETH_P_IPV6))
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000554 inner = ipv6_get_dsfield((const struct ipv6hdr *)old_iph);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return INET_ECN_encapsulate(tos, inner);
556}
557
558static int ipgre_rcv(struct sk_buff *skb)
559{
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000560 const struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 u8 *h;
Al Virod5a0a1e2006-11-08 00:23:14 -0800562 __be16 flags;
Al Virod3bc23e2006-11-14 21:24:49 -0800563 __sum16 csum = 0;
Al Virod5a0a1e2006-11-08 00:23:14 -0800564 __be32 key = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 u32 seqno = 0;
566 struct ip_tunnel *tunnel;
567 int offset = 4;
Herbert Xue1a80002008-10-09 12:00:17 -0700568 __be16 gre_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
570 if (!pskb_may_pull(skb, 16))
571 goto drop_nolock;
572
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700573 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 h = skb->data;
Al Virod5a0a1e2006-11-08 00:23:14 -0800575 flags = *(__be16*)h;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
577 if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) {
578 /* - Version must be 0.
579 - We do not support routing headers.
580 */
581 if (flags&(GRE_VERSION|GRE_ROUTING))
582 goto drop_nolock;
583
584 if (flags&GRE_CSUM) {
Herbert Xufb286bb2005-11-10 13:01:24 -0800585 switch (skb->ip_summed) {
Patrick McHardy84fa7932006-08-29 16:44:56 -0700586 case CHECKSUM_COMPLETE:
Al Virod3bc23e2006-11-14 21:24:49 -0800587 csum = csum_fold(skb->csum);
Herbert Xufb286bb2005-11-10 13:01:24 -0800588 if (!csum)
589 break;
590 /* fall through */
591 case CHECKSUM_NONE:
592 skb->csum = 0;
593 csum = __skb_checksum_complete(skb);
Patrick McHardy84fa7932006-08-29 16:44:56 -0700594 skb->ip_summed = CHECKSUM_COMPLETE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 }
596 offset += 4;
597 }
598 if (flags&GRE_KEY) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800599 key = *(__be32*)(h + offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 offset += 4;
601 }
602 if (flags&GRE_SEQ) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800603 seqno = ntohl(*(__be32*)(h + offset));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 offset += 4;
605 }
606 }
607
Herbert Xue1a80002008-10-09 12:00:17 -0700608 gre_proto = *(__be16 *)(h + 2);
609
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000610 rcu_read_lock();
Timo Teras749c10f2009-01-19 17:22:12 -0800611 if ((tunnel = ipgre_tunnel_lookup(skb->dev,
Herbert Xue1a80002008-10-09 12:00:17 -0700612 iph->saddr, iph->daddr, key,
613 gre_proto))) {
Eric Dumazete985aad2010-09-27 03:57:11 +0000614 struct pcpu_tstats *tstats;
Pavel Emelyanovaddd68e2008-05-21 14:14:22 -0700615
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 secpath_reset(skb);
617
Herbert Xue1a80002008-10-09 12:00:17 -0700618 skb->protocol = gre_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 /* WCCP version 1 and 2 protocol decoding.
620 * - Change protocol to IP
621 * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header
622 */
Herbert Xue1a80002008-10-09 12:00:17 -0700623 if (flags == 0 && gre_proto == htons(ETH_P_WCCP)) {
YOSHIFUJI Hideaki496c98d2006-10-10 19:41:21 -0700624 skb->protocol = htons(ETH_P_IP);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900625 if ((*(h + offset) & 0xF0) != 0x40)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 offset += 4;
627 }
628
Timo Teras1d069162007-12-20 00:10:33 -0800629 skb->mac_header = skb->network_header;
Arnaldo Carvalho de Melo4209fb62007-03-10 18:42:03 -0300630 __pskb_pull(skb, offset);
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -0700631 skb_postpull_rcsum(skb, skb_transport_header(skb), offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 skb->pkt_type = PACKET_HOST;
633#ifdef CONFIG_NET_IPGRE_BROADCAST
Joe Perchesf97c1e02007-12-16 13:45:43 -0800634 if (ipv4_is_multicast(iph->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 /* Looped back packet, drop it! */
David S. Millerc7537962010-11-11 17:07:48 -0800636 if (rt_is_output_route(skb_rtable(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 goto drop;
Eric Dumazete985aad2010-09-27 03:57:11 +0000638 tunnel->dev->stats.multicast++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 skb->pkt_type = PACKET_BROADCAST;
640 }
641#endif
642
643 if (((flags&GRE_CSUM) && csum) ||
644 (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) {
Eric Dumazete985aad2010-09-27 03:57:11 +0000645 tunnel->dev->stats.rx_crc_errors++;
646 tunnel->dev->stats.rx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 goto drop;
648 }
649 if (tunnel->parms.i_flags&GRE_SEQ) {
650 if (!(flags&GRE_SEQ) ||
651 (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) {
Eric Dumazete985aad2010-09-27 03:57:11 +0000652 tunnel->dev->stats.rx_fifo_errors++;
653 tunnel->dev->stats.rx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 goto drop;
655 }
656 tunnel->i_seqno = seqno + 1;
657 }
Herbert Xue1a80002008-10-09 12:00:17 -0700658
659 /* Warning: All skb pointers will be invalidated! */
660 if (tunnel->dev->type == ARPHRD_ETHER) {
661 if (!pskb_may_pull(skb, ETH_HLEN)) {
Eric Dumazete985aad2010-09-27 03:57:11 +0000662 tunnel->dev->stats.rx_length_errors++;
663 tunnel->dev->stats.rx_errors++;
Herbert Xue1a80002008-10-09 12:00:17 -0700664 goto drop;
665 }
666
667 iph = ip_hdr(skb);
668 skb->protocol = eth_type_trans(skb, tunnel->dev);
669 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
670 }
671
Eric Dumazete985aad2010-09-27 03:57:11 +0000672 tstats = this_cpu_ptr(tunnel->dev->tstats);
673 tstats->rx_packets++;
674 tstats->rx_bytes += skb->len;
675
676 __skb_tunnel_rx(skb, tunnel->dev);
Herbert Xue1a80002008-10-09 12:00:17 -0700677
678 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 ipgre_ecn_decapsulate(iph, skb);
Herbert Xue1a80002008-10-09 12:00:17 -0700680
Eric Dumazetcaf586e2010-09-30 21:06:55 +0000681 netif_rx(skb);
Eric Dumazet8990f462010-09-20 00:12:11 +0000682
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000683 rcu_read_unlock();
Eric Dumazet8990f462010-09-20 00:12:11 +0000684 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 }
Herbert Xu45af08b2006-04-05 22:31:19 -0700686 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687
688drop:
Eric Dumazet8d5b2c02009-10-23 06:14:38 +0000689 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690drop_nolock:
691 kfree_skb(skb);
Eric Dumazeta02cec22010-09-22 20:43:57 +0000692 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693}
694
Stephen Hemminger6fef4c02009-08-31 19:50:41 +0000695static netdev_tx_t ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696{
Patrick McHardy2941a482006-01-08 22:05:26 -0800697 struct ip_tunnel *tunnel = netdev_priv(dev);
Eric Dumazete985aad2010-09-27 03:57:11 +0000698 struct pcpu_tstats *tstats;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000699 const struct iphdr *old_iph = ip_hdr(skb);
700 const struct iphdr *tiph;
David S. Millercbb1e852011-05-04 12:33:34 -0700701 struct flowi4 fl4;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 u8 tos;
Al Virod5a0a1e2006-11-08 00:23:14 -0800703 __be16 df;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 struct rtable *rt; /* Route to the other host */
Eric Dumazet15078502010-09-15 11:07:53 +0000705 struct net_device *tdev; /* Device to other host */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 struct iphdr *iph; /* Our new IP header */
Chuck Leverc2636b42007-10-23 21:07:32 -0700707 unsigned int max_headroom; /* The extra header space needed */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 int gre_hlen;
Al Virod5a0a1e2006-11-08 00:23:14 -0800709 __be32 dst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 int mtu;
711
Herbert Xue1a80002008-10-09 12:00:17 -0700712 if (dev->type == ARPHRD_ETHER)
713 IPCB(skb)->flags = 0;
714
715 if (dev->header_ops && dev->type == ARPHRD_IPGRE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 gre_hlen = 0;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000717 tiph = (const struct iphdr *)skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 } else {
719 gre_hlen = tunnel->hlen;
720 tiph = &tunnel->parms.iph;
721 }
722
723 if ((dst = tiph->daddr) == 0) {
724 /* NBMA tunnel */
725
Eric Dumazetadf30902009-06-02 05:19:30 +0000726 if (skb_dst(skb) == NULL) {
Eric Dumazete985aad2010-09-27 03:57:11 +0000727 dev->stats.tx_fifo_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 goto tx_error;
729 }
730
731 if (skb->protocol == htons(ETH_P_IP)) {
Eric Dumazet511c3f92009-06-02 05:14:27 +0000732 rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 if ((dst = rt->rt_gateway) == 0)
734 goto tx_error_icmp;
735 }
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000736#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 else if (skb->protocol == htons(ETH_P_IPV6)) {
David Miller27217452011-12-02 16:52:08 +0000738 struct neighbour *neigh = dst_get_neighbour_noref(skb_dst(skb));
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000739 const struct in6_addr *addr6;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 int addr_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742 if (neigh == NULL)
743 goto tx_error;
744
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000745 addr6 = (const struct in6_addr *)&neigh->primary_key;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 addr_type = ipv6_addr_type(addr6);
747
748 if (addr_type == IPV6_ADDR_ANY) {
Arnaldo Carvalho de Melo0660e032007-04-25 17:54:47 -0700749 addr6 = &ipv6_hdr(skb)->daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 addr_type = ipv6_addr_type(addr6);
751 }
752
753 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
754 goto tx_error_icmp;
755
756 dst = addr6->s6_addr32[3];
757 }
758#endif
759 else
760 goto tx_error;
761 }
762
763 tos = tiph->tos;
Andreas Jaggiee686ca2009-07-14 09:35:59 -0700764 if (tos == 1) {
765 tos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 if (skb->protocol == htons(ETH_P_IP))
767 tos = old_iph->tos;
Stephen Hemmingerdd4ba832010-07-08 21:35:58 -0700768 else if (skb->protocol == htons(ETH_P_IPV6))
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000769 tos = ipv6_get_dsfield((const struct ipv6hdr *)old_iph);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 }
771
David S. Millercbb1e852011-05-04 12:33:34 -0700772 rt = ip_route_output_gre(dev_net(dev), &fl4, dst, tiph->saddr,
David S. Miller78fbfd82011-03-12 00:00:52 -0500773 tunnel->parms.o_key, RT_TOS(tos),
774 tunnel->parms.link);
775 if (IS_ERR(rt)) {
776 dev->stats.tx_carrier_errors++;
777 goto tx_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 }
Changli Gaod8d1f302010-06-10 23:31:35 -0700779 tdev = rt->dst.dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
781 if (tdev == dev) {
782 ip_rt_put(rt);
Eric Dumazete985aad2010-09-27 03:57:11 +0000783 dev->stats.collisions++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 goto tx_error;
785 }
786
787 df = tiph->frag_off;
788 if (df)
Changli Gaod8d1f302010-06-10 23:31:35 -0700789 mtu = dst_mtu(&rt->dst) - dev->hard_header_len - tunnel->hlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 else
Eric Dumazetadf30902009-06-02 05:19:30 +0000791 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
Eric Dumazetadf30902009-06-02 05:19:30 +0000793 if (skb_dst(skb))
794 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
796 if (skb->protocol == htons(ETH_P_IP)) {
797 df |= (old_iph->frag_off&htons(IP_DF));
798
799 if ((old_iph->frag_off&htons(IP_DF)) &&
800 mtu < ntohs(old_iph->tot_len)) {
801 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
802 ip_rt_put(rt);
803 goto tx_error;
804 }
805 }
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000806#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 else if (skb->protocol == htons(ETH_P_IPV6)) {
Eric Dumazetadf30902009-06-02 05:19:30 +0000808 struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
Eric Dumazetadf30902009-06-02 05:19:30 +0000810 if (rt6 && mtu < dst_mtu(skb_dst(skb)) && mtu >= IPV6_MIN_MTU) {
Joe Perchesf97c1e02007-12-16 13:45:43 -0800811 if ((tunnel->parms.iph.daddr &&
812 !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 rt6->rt6i_dst.plen == 128) {
814 rt6->rt6i_flags |= RTF_MODIFIED;
David S. Millerdefb3512010-12-08 21:16:57 -0800815 dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 }
817 }
818
819 if (mtu >= IPV6_MIN_MTU && mtu < skb->len - tunnel->hlen + gre_hlen) {
Alexey Dobriyan3ffe5332010-02-18 08:25:24 +0000820 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 ip_rt_put(rt);
822 goto tx_error;
823 }
824 }
825#endif
826
827 if (tunnel->err_count > 0) {
Wei Yongjunda6185d82009-02-24 23:34:48 -0800828 if (time_before(jiffies,
829 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 tunnel->err_count--;
831
832 dst_link_failure(skb);
833 } else
834 tunnel->err_count = 0;
835 }
836
Changli Gaod8d1f302010-06-10 23:31:35 -0700837 max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen + rt->dst.header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
Patrick McHardycfbba492007-07-09 15:33:40 -0700839 if (skb_headroom(skb) < max_headroom || skb_shared(skb)||
840 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
Herbert Xu805dc1d2011-11-18 02:20:06 +0000842 if (max_headroom > dev->needed_headroom)
843 dev->needed_headroom = max_headroom;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 if (!new_skb) {
845 ip_rt_put(rt);
Eric Dumazete985aad2010-09-27 03:57:11 +0000846 dev->stats.tx_dropped++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000848 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 }
850 if (skb->sk)
851 skb_set_owner_w(new_skb, skb->sk);
852 dev_kfree_skb(skb);
853 skb = new_skb;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700854 old_iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 }
856
Herbert Xu64194c32008-10-09 12:03:17 -0700857 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloe2d1bca2007-04-10 20:46:21 -0700858 skb_push(skb, gre_hlen);
859 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy48d5cad2006-02-15 15:10:22 -0800861 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
862 IPSKB_REROUTED);
Eric Dumazetadf30902009-06-02 05:19:30 +0000863 skb_dst_drop(skb);
Changli Gaod8d1f302010-06-10 23:31:35 -0700864 skb_dst_set(skb, &rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866 /*
867 * Push down and install the IPIP header.
868 */
869
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700870 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 iph->version = 4;
872 iph->ihl = sizeof(struct iphdr) >> 2;
873 iph->frag_off = df;
874 iph->protocol = IPPROTO_GRE;
875 iph->tos = ipgre_ecn_encapsulate(tos, old_iph, skb);
David S. Millercbb1e852011-05-04 12:33:34 -0700876 iph->daddr = fl4.daddr;
877 iph->saddr = fl4.saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
879 if ((iph->ttl = tiph->ttl) == 0) {
880 if (skb->protocol == htons(ETH_P_IP))
881 iph->ttl = old_iph->ttl;
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000882#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 else if (skb->protocol == htons(ETH_P_IPV6))
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000884 iph->ttl = ((const struct ipv6hdr *)old_iph)->hop_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885#endif
886 else
David S. Miller323e1262010-12-12 21:55:08 -0800887 iph->ttl = ip4_dst_hoplimit(&rt->dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 }
889
Herbert Xue1a80002008-10-09 12:00:17 -0700890 ((__be16 *)(iph + 1))[0] = tunnel->parms.o_flags;
891 ((__be16 *)(iph + 1))[1] = (dev->type == ARPHRD_ETHER) ?
892 htons(ETH_P_TEB) : skb->protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
894 if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) {
Al Virod5a0a1e2006-11-08 00:23:14 -0800895 __be32 *ptr = (__be32*)(((u8*)iph) + tunnel->hlen - 4);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896
897 if (tunnel->parms.o_flags&GRE_SEQ) {
898 ++tunnel->o_seqno;
899 *ptr = htonl(tunnel->o_seqno);
900 ptr--;
901 }
902 if (tunnel->parms.o_flags&GRE_KEY) {
903 *ptr = tunnel->parms.o_key;
904 ptr--;
905 }
906 if (tunnel->parms.o_flags&GRE_CSUM) {
907 *ptr = 0;
Al Viro5f92a732006-11-14 21:36:54 -0800908 *(__sum16*)ptr = ip_compute_csum((void*)(iph+1), skb->len - sizeof(struct iphdr));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 }
910 }
911
912 nf_reset(skb);
Eric Dumazete985aad2010-09-27 03:57:11 +0000913 tstats = this_cpu_ptr(dev->tstats);
914 __IPTUNNEL_XMIT(tstats, &dev->stats);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000915 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
917tx_error_icmp:
918 dst_link_failure(skb);
919
920tx_error:
Eric Dumazete985aad2010-09-27 03:57:11 +0000921 dev->stats.tx_errors++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 dev_kfree_skb(skb);
Patrick McHardy6ed10652009-06-23 06:03:08 +0000923 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924}
925
Herbert Xu42aa9162008-10-09 11:59:32 -0700926static int ipgre_tunnel_bind_dev(struct net_device *dev)
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800927{
928 struct net_device *tdev = NULL;
929 struct ip_tunnel *tunnel;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000930 const struct iphdr *iph;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800931 int hlen = LL_MAX_HEADER;
932 int mtu = ETH_DATA_LEN;
933 int addend = sizeof(struct iphdr) + 4;
934
935 tunnel = netdev_priv(dev);
936 iph = &tunnel->parms.iph;
937
Herbert Xuc95b8192008-10-09 11:58:54 -0700938 /* Guess output device to choose reasonable mtu and needed_headroom */
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800939
940 if (iph->daddr) {
David S. Millercbb1e852011-05-04 12:33:34 -0700941 struct flowi4 fl4;
942 struct rtable *rt;
Eric Dumazete985aad2010-09-27 03:57:11 +0000943
David S. Millercbb1e852011-05-04 12:33:34 -0700944 rt = ip_route_output_gre(dev_net(dev), &fl4,
945 iph->daddr, iph->saddr,
946 tunnel->parms.o_key,
947 RT_TOS(iph->tos),
948 tunnel->parms.link);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800949 if (!IS_ERR(rt)) {
Changli Gaod8d1f302010-06-10 23:31:35 -0700950 tdev = rt->dst.dev;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800951 ip_rt_put(rt);
952 }
Herbert Xue1a80002008-10-09 12:00:17 -0700953
954 if (dev->type != ARPHRD_ETHER)
955 dev->flags |= IFF_POINTOPOINT;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800956 }
957
958 if (!tdev && tunnel->parms.link)
Pavel Emelyanov96635522008-04-16 01:10:44 -0700959 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800960
961 if (tdev) {
Herbert Xuc95b8192008-10-09 11:58:54 -0700962 hlen = tdev->hard_header_len + tdev->needed_headroom;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800963 mtu = tdev->mtu;
964 }
965 dev->iflink = tunnel->parms.link;
966
967 /* Precalculate GRE options length */
968 if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) {
969 if (tunnel->parms.o_flags&GRE_CSUM)
970 addend += 4;
971 if (tunnel->parms.o_flags&GRE_KEY)
972 addend += 4;
973 if (tunnel->parms.o_flags&GRE_SEQ)
974 addend += 4;
975 }
Herbert Xuc95b8192008-10-09 11:58:54 -0700976 dev->needed_headroom = addend + hlen;
Tom Goff8cdb0452009-08-14 16:33:56 -0700977 mtu -= dev->hard_header_len + addend;
Herbert Xu42aa9162008-10-09 11:59:32 -0700978
979 if (mtu < 68)
980 mtu = 68;
981
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800982 tunnel->hlen = addend;
983
Herbert Xu42aa9162008-10-09 11:59:32 -0700984 return mtu;
Michal Schmidtee34c1e2007-12-13 09:46:32 -0800985}
986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987static int
988ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
989{
990 int err = 0;
991 struct ip_tunnel_parm p;
992 struct ip_tunnel *t;
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -0700993 struct net *net = dev_net(dev);
994 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 switch (cmd) {
997 case SIOCGETTUNNEL:
998 t = NULL;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -0700999 if (dev == ign->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1001 err = -EFAULT;
1002 break;
1003 }
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001004 t = ipgre_tunnel_locate(net, &p, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 }
1006 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -08001007 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 memcpy(&p, &t->parms, sizeof(p));
1009 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1010 err = -EFAULT;
1011 break;
1012
1013 case SIOCADDTUNNEL:
1014 case SIOCCHGTUNNEL:
1015 err = -EPERM;
1016 if (!capable(CAP_NET_ADMIN))
1017 goto done;
1018
1019 err = -EFAULT;
1020 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1021 goto done;
1022
1023 err = -EINVAL;
1024 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE ||
1025 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) ||
1026 ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)))
1027 goto done;
1028 if (p.iph.ttl)
1029 p.iph.frag_off |= htons(IP_DF);
1030
1031 if (!(p.i_flags&GRE_KEY))
1032 p.i_key = 0;
1033 if (!(p.o_flags&GRE_KEY))
1034 p.o_key = 0;
1035
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001036 t = ipgre_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001038 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 if (t != NULL) {
1040 if (t->dev != dev) {
1041 err = -EEXIST;
1042 break;
1043 }
1044 } else {
Eric Dumazet15078502010-09-15 11:07:53 +00001045 unsigned int nflags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
Patrick McHardy2941a482006-01-08 22:05:26 -08001047 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Joe Perchesf97c1e02007-12-16 13:45:43 -08001049 if (ipv4_is_multicast(p.iph.daddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 nflags = IFF_BROADCAST;
1051 else if (p.iph.daddr)
1052 nflags = IFF_POINTOPOINT;
1053
1054 if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
1055 err = -EINVAL;
1056 break;
1057 }
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001058 ipgre_tunnel_unlink(ign, t);
Pavel Emelyanov74b0b852010-10-27 05:43:53 +00001059 synchronize_net();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 t->parms.iph.saddr = p.iph.saddr;
1061 t->parms.iph.daddr = p.iph.daddr;
1062 t->parms.i_key = p.i_key;
1063 t->parms.o_key = p.o_key;
1064 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1065 memcpy(dev->broadcast, &p.iph.daddr, 4);
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001066 ipgre_tunnel_link(ign, t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 netdev_state_change(dev);
1068 }
1069 }
1070
1071 if (t) {
1072 err = 0;
1073 if (cmd == SIOCCHGTUNNEL) {
1074 t->parms.iph.ttl = p.iph.ttl;
1075 t->parms.iph.tos = p.iph.tos;
1076 t->parms.iph.frag_off = p.iph.frag_off;
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001077 if (t->parms.link != p.link) {
1078 t->parms.link = p.link;
Herbert Xu42aa9162008-10-09 11:59:32 -07001079 dev->mtu = ipgre_tunnel_bind_dev(dev);
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001080 netdev_state_change(dev);
1081 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 }
1083 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1084 err = -EFAULT;
1085 } else
1086 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1087 break;
1088
1089 case SIOCDELTUNNEL:
1090 err = -EPERM;
1091 if (!capable(CAP_NET_ADMIN))
1092 goto done;
1093
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001094 if (dev == ign->fb_tunnel_dev) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 err = -EFAULT;
1096 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1097 goto done;
1098 err = -ENOENT;
Pavel Emelyanovf57e7d52008-04-16 01:09:22 -07001099 if ((t = ipgre_tunnel_locate(net, &p, 0)) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 goto done;
1101 err = -EPERM;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001102 if (t == netdev_priv(ign->fb_tunnel_dev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 goto done;
1104 dev = t->dev;
1105 }
Stephen Hemminger22f8cde2007-02-07 00:09:58 -08001106 unregister_netdevice(dev);
1107 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 break;
1109
1110 default:
1111 err = -EINVAL;
1112 }
1113
1114done:
1115 return err;
1116}
1117
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1119{
Patrick McHardy2941a482006-01-08 22:05:26 -08001120 struct ip_tunnel *tunnel = netdev_priv(dev);
Herbert Xuc95b8192008-10-09 11:58:54 -07001121 if (new_mtu < 68 ||
1122 new_mtu > 0xFFF8 - dev->hard_header_len - tunnel->hlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 return -EINVAL;
1124 dev->mtu = new_mtu;
1125 return 0;
1126}
1127
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128/* Nice toy. Unfortunately, useless in real life :-)
1129 It allows to construct virtual multiprotocol broadcast "LAN"
1130 over the Internet, provided multicast routing is tuned.
1131
1132
1133 I have no idea was this bicycle invented before me,
1134 so that I had to set ARPHRD_IPGRE to a random value.
1135 I have an impression, that Cisco could make something similar,
1136 but this feature is apparently missing in IOS<=11.2(8).
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001137
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks
1139 with broadcast 224.66.66.66. If you have access to mbone, play with me :-)
1140
1141 ping -t 255 224.66.66.66
1142
1143 If nobody answers, mbone does not work.
1144
1145 ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255
1146 ip addr add 10.66.66.<somewhat>/24 dev Universe
1147 ifconfig Universe up
1148 ifconfig Universe add fe80::<Your_real_addr>/10
1149 ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96
1150 ftp 10.66.66.66
1151 ...
1152 ftp fec0:6666:6666::193.233.7.65
1153 ...
1154
1155 */
1156
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001157static int ipgre_header(struct sk_buff *skb, struct net_device *dev,
1158 unsigned short type,
Eric Dumazet15078502010-09-15 11:07:53 +00001159 const void *daddr, const void *saddr, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160{
Patrick McHardy2941a482006-01-08 22:05:26 -08001161 struct ip_tunnel *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen);
Al Virod5a0a1e2006-11-08 00:23:14 -08001163 __be16 *p = (__be16*)(iph+1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
1165 memcpy(iph, &t->parms.iph, sizeof(struct iphdr));
1166 p[0] = t->parms.o_flags;
1167 p[1] = htons(type);
1168
1169 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001170 * Set the source hardware address.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001172
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 if (saddr)
1174 memcpy(&iph->saddr, saddr, 4);
Timo Teräs6d55cb92010-03-03 04:01:13 +00001175 if (daddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 memcpy(&iph->daddr, daddr, 4);
Timo Teräs6d55cb92010-03-03 04:01:13 +00001177 if (iph->daddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 return t->hlen;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 return -t->hlen;
1181}
1182
Timo Teras6a5f44d2007-10-23 20:31:53 -07001183static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr)
1184{
Eric Dumazetb71d1d42011-04-22 04:53:02 +00001185 const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb);
Timo Teras6a5f44d2007-10-23 20:31:53 -07001186 memcpy(haddr, &iph->saddr, 4);
1187 return 4;
1188}
1189
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001190static const struct header_ops ipgre_header_ops = {
1191 .create = ipgre_header,
Timo Teras6a5f44d2007-10-23 20:31:53 -07001192 .parse = ipgre_header_parse,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001193};
1194
Timo Teras6a5f44d2007-10-23 20:31:53 -07001195#ifdef CONFIG_NET_IPGRE_BROADCAST
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196static int ipgre_open(struct net_device *dev)
1197{
Patrick McHardy2941a482006-01-08 22:05:26 -08001198 struct ip_tunnel *t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Joe Perchesf97c1e02007-12-16 13:45:43 -08001200 if (ipv4_is_multicast(t->parms.iph.daddr)) {
David S. Millercbb1e852011-05-04 12:33:34 -07001201 struct flowi4 fl4;
1202 struct rtable *rt;
Eric Dumazete985aad2010-09-27 03:57:11 +00001203
David S. Millercbb1e852011-05-04 12:33:34 -07001204 rt = ip_route_output_gre(dev_net(dev), &fl4,
1205 t->parms.iph.daddr,
1206 t->parms.iph.saddr,
1207 t->parms.o_key,
1208 RT_TOS(t->parms.iph.tos),
1209 t->parms.link);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001210 if (IS_ERR(rt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 return -EADDRNOTAVAIL;
Changli Gaod8d1f302010-06-10 23:31:35 -07001212 dev = rt->dst.dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 ip_rt_put(rt);
Herbert Xue5ed6392005-10-03 14:35:55 -07001214 if (__in_dev_get_rtnl(dev) == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 return -EADDRNOTAVAIL;
1216 t->mlink = dev->ifindex;
Herbert Xue5ed6392005-10-03 14:35:55 -07001217 ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 }
1219 return 0;
1220}
1221
1222static int ipgre_close(struct net_device *dev)
1223{
Patrick McHardy2941a482006-01-08 22:05:26 -08001224 struct ip_tunnel *t = netdev_priv(dev);
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001225
Joe Perchesf97c1e02007-12-16 13:45:43 -08001226 if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) {
Denis V. Lunev7fee0ca2008-01-21 17:32:38 -08001227 struct in_device *in_dev;
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001228 in_dev = inetdev_by_index(dev_net(dev), t->mlink);
Eric Dumazet8723e1b2010-10-19 00:39:26 +00001229 if (in_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 ip_mc_dec_group(in_dev, t->parms.iph.daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 }
1232 return 0;
1233}
1234
1235#endif
1236
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001237static const struct net_device_ops ipgre_netdev_ops = {
1238 .ndo_init = ipgre_tunnel_init,
1239 .ndo_uninit = ipgre_tunnel_uninit,
1240#ifdef CONFIG_NET_IPGRE_BROADCAST
1241 .ndo_open = ipgre_open,
1242 .ndo_stop = ipgre_close,
1243#endif
1244 .ndo_start_xmit = ipgre_tunnel_xmit,
1245 .ndo_do_ioctl = ipgre_tunnel_ioctl,
1246 .ndo_change_mtu = ipgre_tunnel_change_mtu,
Eric Dumazete985aad2010-09-27 03:57:11 +00001247 .ndo_get_stats = ipgre_get_stats,
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001248};
1249
Eric Dumazete985aad2010-09-27 03:57:11 +00001250static void ipgre_dev_free(struct net_device *dev)
1251{
1252 free_percpu(dev->tstats);
1253 free_netdev(dev);
1254}
1255
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256static void ipgre_tunnel_setup(struct net_device *dev)
1257{
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001258 dev->netdev_ops = &ipgre_netdev_ops;
Eric Dumazete985aad2010-09-27 03:57:11 +00001259 dev->destructor = ipgre_dev_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260
1261 dev->type = ARPHRD_IPGRE;
Herbert Xuc95b8192008-10-09 11:58:54 -07001262 dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4;
Kris Katterjohn46f25df2006-01-05 16:35:42 -08001263 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 dev->flags = IFF_NOARP;
1265 dev->iflink = 0;
1266 dev->addr_len = 4;
Pavel Emelyanov0b67ece2008-04-16 01:11:13 -07001267 dev->features |= NETIF_F_NETNS_LOCAL;
Eric Dumazet108bfa82009-05-28 22:35:10 +00001268 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269}
1270
1271static int ipgre_tunnel_init(struct net_device *dev)
1272{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 struct ip_tunnel *tunnel;
1274 struct iphdr *iph;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275
Patrick McHardy2941a482006-01-08 22:05:26 -08001276 tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 iph = &tunnel->parms.iph;
1278
1279 tunnel->dev = dev;
1280 strcpy(tunnel->parms.name, dev->name);
1281
1282 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1283 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1284
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 if (iph->daddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286#ifdef CONFIG_NET_IPGRE_BROADCAST
Joe Perchesf97c1e02007-12-16 13:45:43 -08001287 if (ipv4_is_multicast(iph->daddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 if (!iph->saddr)
1289 return -EINVAL;
1290 dev->flags = IFF_BROADCAST;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001291 dev->header_ops = &ipgre_header_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 }
1293#endif
Michal Schmidtee34c1e2007-12-13 09:46:32 -08001294 } else
Timo Teras6a5f44d2007-10-23 20:31:53 -07001295 dev->header_ops = &ipgre_header_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296
Eric Dumazete985aad2010-09-27 03:57:11 +00001297 dev->tstats = alloc_percpu(struct pcpu_tstats);
1298 if (!dev->tstats)
1299 return -ENOMEM;
1300
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 return 0;
1302}
1303
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001304static void ipgre_fb_tunnel_init(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305{
Patrick McHardy2941a482006-01-08 22:05:26 -08001306 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 struct iphdr *iph = &tunnel->parms.iph;
1308
1309 tunnel->dev = dev;
1310 strcpy(tunnel->parms.name, dev->name);
1311
1312 iph->version = 4;
1313 iph->protocol = IPPROTO_GRE;
1314 iph->ihl = 5;
1315 tunnel->hlen = sizeof(struct iphdr) + 4;
1316
1317 dev_hold(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318}
1319
1320
Dmitry Kozlov00959ad2010-08-21 23:05:39 -07001321static const struct gre_protocol ipgre_protocol = {
1322 .handler = ipgre_rcv,
1323 .err_handler = ipgre_err,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324};
1325
Eric Dumazeteef6dd62009-10-27 07:07:16 +00001326static void ipgre_destroy_tunnels(struct ipgre_net *ign, struct list_head *head)
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001327{
1328 int prio;
1329
1330 for (prio = 0; prio < 4; prio++) {
1331 int h;
1332 for (h = 0; h < HASH_SIZE; h++) {
Eric Dumazet15078502010-09-15 11:07:53 +00001333 struct ip_tunnel *t;
1334
1335 t = rtnl_dereference(ign->tunnels[prio][h]);
Eric Dumazeteef6dd62009-10-27 07:07:16 +00001336
1337 while (t != NULL) {
1338 unregister_netdevice_queue(t->dev, head);
Eric Dumazet15078502010-09-15 11:07:53 +00001339 t = rtnl_dereference(t->next);
Eric Dumazeteef6dd62009-10-27 07:07:16 +00001340 }
Pavel Emelyanoveb8ce742008-04-16 01:10:26 -07001341 }
1342 }
1343}
1344
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001345static int __net_init ipgre_init_net(struct net *net)
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001346{
Eric W. Biedermancfb8fbf2009-11-29 15:46:13 +00001347 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001348 int err;
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001349
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001350 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "gre0",
1351 ipgre_tunnel_setup);
1352 if (!ign->fb_tunnel_dev) {
1353 err = -ENOMEM;
1354 goto err_alloc_dev;
1355 }
Alexey Dobriyanbe77e592008-11-23 17:26:26 -08001356 dev_net_set(ign->fb_tunnel_dev, net);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001357
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001358 ipgre_fb_tunnel_init(ign->fb_tunnel_dev);
Herbert Xuc19e6542008-10-09 11:59:55 -07001359 ign->fb_tunnel_dev->rtnl_link_ops = &ipgre_link_ops;
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001360
1361 if ((err = register_netdev(ign->fb_tunnel_dev)))
1362 goto err_reg_dev;
1363
Eric Dumazet3285ee32010-10-30 16:21:28 -07001364 rcu_assign_pointer(ign->tunnels_wc[0],
1365 netdev_priv(ign->fb_tunnel_dev));
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001366 return 0;
1367
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001368err_reg_dev:
Eric Dumazet3285ee32010-10-30 16:21:28 -07001369 ipgre_dev_free(ign->fb_tunnel_dev);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001370err_alloc_dev:
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001371 return err;
1372}
1373
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00001374static void __net_exit ipgre_exit_net(struct net *net)
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001375{
1376 struct ipgre_net *ign;
Eric Dumazeteef6dd62009-10-27 07:07:16 +00001377 LIST_HEAD(list);
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001378
1379 ign = net_generic(net, ipgre_net_id);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001380 rtnl_lock();
Eric Dumazeteef6dd62009-10-27 07:07:16 +00001381 ipgre_destroy_tunnels(ign, &list);
1382 unregister_netdevice_many(&list);
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001383 rtnl_unlock();
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001384}
1385
1386static struct pernet_operations ipgre_net_ops = {
1387 .init = ipgre_init_net,
1388 .exit = ipgre_exit_net,
Eric W. Biedermancfb8fbf2009-11-29 15:46:13 +00001389 .id = &ipgre_net_id,
1390 .size = sizeof(struct ipgre_net),
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001391};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392
Herbert Xuc19e6542008-10-09 11:59:55 -07001393static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1394{
1395 __be16 flags;
1396
1397 if (!data)
1398 return 0;
1399
1400 flags = 0;
1401 if (data[IFLA_GRE_IFLAGS])
1402 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1403 if (data[IFLA_GRE_OFLAGS])
1404 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1405 if (flags & (GRE_VERSION|GRE_ROUTING))
1406 return -EINVAL;
1407
1408 return 0;
1409}
1410
Herbert Xue1a80002008-10-09 12:00:17 -07001411static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1412{
1413 __be32 daddr;
1414
1415 if (tb[IFLA_ADDRESS]) {
1416 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1417 return -EINVAL;
1418 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1419 return -EADDRNOTAVAIL;
1420 }
1421
1422 if (!data)
1423 goto out;
1424
1425 if (data[IFLA_GRE_REMOTE]) {
1426 memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4);
1427 if (!daddr)
1428 return -EINVAL;
1429 }
1430
1431out:
1432 return ipgre_tunnel_validate(tb, data);
1433}
1434
Herbert Xuc19e6542008-10-09 11:59:55 -07001435static void ipgre_netlink_parms(struct nlattr *data[],
1436 struct ip_tunnel_parm *parms)
1437{
Herbert Xu7bb82d92008-10-11 12:20:15 -07001438 memset(parms, 0, sizeof(*parms));
Herbert Xuc19e6542008-10-09 11:59:55 -07001439
1440 parms->iph.protocol = IPPROTO_GRE;
1441
1442 if (!data)
1443 return;
1444
1445 if (data[IFLA_GRE_LINK])
1446 parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1447
1448 if (data[IFLA_GRE_IFLAGS])
1449 parms->i_flags = nla_get_be16(data[IFLA_GRE_IFLAGS]);
1450
1451 if (data[IFLA_GRE_OFLAGS])
1452 parms->o_flags = nla_get_be16(data[IFLA_GRE_OFLAGS]);
1453
1454 if (data[IFLA_GRE_IKEY])
1455 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1456
1457 if (data[IFLA_GRE_OKEY])
1458 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1459
1460 if (data[IFLA_GRE_LOCAL])
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001461 parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]);
Herbert Xuc19e6542008-10-09 11:59:55 -07001462
1463 if (data[IFLA_GRE_REMOTE])
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001464 parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]);
Herbert Xuc19e6542008-10-09 11:59:55 -07001465
1466 if (data[IFLA_GRE_TTL])
1467 parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]);
1468
1469 if (data[IFLA_GRE_TOS])
1470 parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]);
1471
1472 if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
1473 parms->iph.frag_off = htons(IP_DF);
1474}
1475
Herbert Xue1a80002008-10-09 12:00:17 -07001476static int ipgre_tap_init(struct net_device *dev)
1477{
1478 struct ip_tunnel *tunnel;
1479
1480 tunnel = netdev_priv(dev);
1481
1482 tunnel->dev = dev;
1483 strcpy(tunnel->parms.name, dev->name);
1484
1485 ipgre_tunnel_bind_dev(dev);
1486
Eric Dumazete985aad2010-09-27 03:57:11 +00001487 dev->tstats = alloc_percpu(struct pcpu_tstats);
1488 if (!dev->tstats)
1489 return -ENOMEM;
1490
Herbert Xue1a80002008-10-09 12:00:17 -07001491 return 0;
1492}
1493
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001494static const struct net_device_ops ipgre_tap_netdev_ops = {
1495 .ndo_init = ipgre_tap_init,
1496 .ndo_uninit = ipgre_tunnel_uninit,
1497 .ndo_start_xmit = ipgre_tunnel_xmit,
1498 .ndo_set_mac_address = eth_mac_addr,
1499 .ndo_validate_addr = eth_validate_addr,
1500 .ndo_change_mtu = ipgre_tunnel_change_mtu,
Eric Dumazete985aad2010-09-27 03:57:11 +00001501 .ndo_get_stats = ipgre_get_stats,
Stephen Hemmingerb8c26a32008-11-20 20:34:29 -08001502};
1503
Herbert Xue1a80002008-10-09 12:00:17 -07001504static void ipgre_tap_setup(struct net_device *dev)
1505{
1506
1507 ether_setup(dev);
1508
Herbert Xu2e9526b2009-10-30 05:51:48 +00001509 dev->netdev_ops = &ipgre_tap_netdev_ops;
Eric Dumazete985aad2010-09-27 03:57:11 +00001510 dev->destructor = ipgre_dev_free;
Herbert Xue1a80002008-10-09 12:00:17 -07001511
1512 dev->iflink = 0;
1513 dev->features |= NETIF_F_NETNS_LOCAL;
1514}
1515
Eric W. Biederman81adee42009-11-08 00:53:51 -08001516static int ipgre_newlink(struct net *src_net, struct net_device *dev, struct nlattr *tb[],
Herbert Xuc19e6542008-10-09 11:59:55 -07001517 struct nlattr *data[])
1518{
1519 struct ip_tunnel *nt;
1520 struct net *net = dev_net(dev);
1521 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1522 int mtu;
1523 int err;
1524
1525 nt = netdev_priv(dev);
1526 ipgre_netlink_parms(data, &nt->parms);
1527
Herbert Xue1a80002008-10-09 12:00:17 -07001528 if (ipgre_tunnel_find(net, &nt->parms, dev->type))
Herbert Xuc19e6542008-10-09 11:59:55 -07001529 return -EEXIST;
1530
Herbert Xue1a80002008-10-09 12:00:17 -07001531 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1532 random_ether_addr(dev->dev_addr);
1533
Herbert Xuc19e6542008-10-09 11:59:55 -07001534 mtu = ipgre_tunnel_bind_dev(dev);
1535 if (!tb[IFLA_MTU])
1536 dev->mtu = mtu;
1537
Eric Dumazetb790e012010-09-27 23:05:47 +00001538 /* Can use a lockless transmit, unless we generate output sequences */
1539 if (!(nt->parms.o_flags & GRE_SEQ))
1540 dev->features |= NETIF_F_LLTX;
1541
Herbert Xuc19e6542008-10-09 11:59:55 -07001542 err = register_netdevice(dev);
1543 if (err)
1544 goto out;
1545
1546 dev_hold(dev);
1547 ipgre_tunnel_link(ign, nt);
1548
1549out:
1550 return err;
1551}
1552
1553static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
1554 struct nlattr *data[])
1555{
1556 struct ip_tunnel *t, *nt;
1557 struct net *net = dev_net(dev);
1558 struct ipgre_net *ign = net_generic(net, ipgre_net_id);
1559 struct ip_tunnel_parm p;
1560 int mtu;
1561
1562 if (dev == ign->fb_tunnel_dev)
1563 return -EINVAL;
1564
1565 nt = netdev_priv(dev);
1566 ipgre_netlink_parms(data, &p);
1567
1568 t = ipgre_tunnel_locate(net, &p, 0);
1569
1570 if (t) {
1571 if (t->dev != dev)
1572 return -EEXIST;
1573 } else {
Herbert Xuc19e6542008-10-09 11:59:55 -07001574 t = nt;
1575
Herbert Xu2e9526b2009-10-30 05:51:48 +00001576 if (dev->type != ARPHRD_ETHER) {
Eric Dumazet15078502010-09-15 11:07:53 +00001577 unsigned int nflags = 0;
Herbert Xuc19e6542008-10-09 11:59:55 -07001578
Herbert Xu2e9526b2009-10-30 05:51:48 +00001579 if (ipv4_is_multicast(p.iph.daddr))
1580 nflags = IFF_BROADCAST;
1581 else if (p.iph.daddr)
1582 nflags = IFF_POINTOPOINT;
1583
1584 if ((dev->flags ^ nflags) &
1585 (IFF_POINTOPOINT | IFF_BROADCAST))
1586 return -EINVAL;
1587 }
Herbert Xuc19e6542008-10-09 11:59:55 -07001588
1589 ipgre_tunnel_unlink(ign, t);
1590 t->parms.iph.saddr = p.iph.saddr;
1591 t->parms.iph.daddr = p.iph.daddr;
1592 t->parms.i_key = p.i_key;
Herbert Xu2e9526b2009-10-30 05:51:48 +00001593 if (dev->type != ARPHRD_ETHER) {
1594 memcpy(dev->dev_addr, &p.iph.saddr, 4);
1595 memcpy(dev->broadcast, &p.iph.daddr, 4);
1596 }
Herbert Xuc19e6542008-10-09 11:59:55 -07001597 ipgre_tunnel_link(ign, t);
1598 netdev_state_change(dev);
1599 }
1600
1601 t->parms.o_key = p.o_key;
1602 t->parms.iph.ttl = p.iph.ttl;
1603 t->parms.iph.tos = p.iph.tos;
1604 t->parms.iph.frag_off = p.iph.frag_off;
1605
1606 if (t->parms.link != p.link) {
1607 t->parms.link = p.link;
1608 mtu = ipgre_tunnel_bind_dev(dev);
1609 if (!tb[IFLA_MTU])
1610 dev->mtu = mtu;
1611 netdev_state_change(dev);
1612 }
1613
1614 return 0;
1615}
1616
1617static size_t ipgre_get_size(const struct net_device *dev)
1618{
1619 return
1620 /* IFLA_GRE_LINK */
1621 nla_total_size(4) +
1622 /* IFLA_GRE_IFLAGS */
1623 nla_total_size(2) +
1624 /* IFLA_GRE_OFLAGS */
1625 nla_total_size(2) +
1626 /* IFLA_GRE_IKEY */
1627 nla_total_size(4) +
1628 /* IFLA_GRE_OKEY */
1629 nla_total_size(4) +
1630 /* IFLA_GRE_LOCAL */
1631 nla_total_size(4) +
1632 /* IFLA_GRE_REMOTE */
1633 nla_total_size(4) +
1634 /* IFLA_GRE_TTL */
1635 nla_total_size(1) +
1636 /* IFLA_GRE_TOS */
1637 nla_total_size(1) +
1638 /* IFLA_GRE_PMTUDISC */
1639 nla_total_size(1) +
1640 0;
1641}
1642
1643static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1644{
1645 struct ip_tunnel *t = netdev_priv(dev);
1646 struct ip_tunnel_parm *p = &t->parms;
1647
1648 NLA_PUT_U32(skb, IFLA_GRE_LINK, p->link);
1649 NLA_PUT_BE16(skb, IFLA_GRE_IFLAGS, p->i_flags);
1650 NLA_PUT_BE16(skb, IFLA_GRE_OFLAGS, p->o_flags);
Patrick McHardyba9e64b2008-10-10 12:10:30 -07001651 NLA_PUT_BE32(skb, IFLA_GRE_IKEY, p->i_key);
1652 NLA_PUT_BE32(skb, IFLA_GRE_OKEY, p->o_key);
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001653 NLA_PUT_BE32(skb, IFLA_GRE_LOCAL, p->iph.saddr);
1654 NLA_PUT_BE32(skb, IFLA_GRE_REMOTE, p->iph.daddr);
Herbert Xuc19e6542008-10-09 11:59:55 -07001655 NLA_PUT_U8(skb, IFLA_GRE_TTL, p->iph.ttl);
1656 NLA_PUT_U8(skb, IFLA_GRE_TOS, p->iph.tos);
1657 NLA_PUT_U8(skb, IFLA_GRE_PMTUDISC, !!(p->iph.frag_off & htons(IP_DF)));
1658
1659 return 0;
1660
1661nla_put_failure:
1662 return -EMSGSIZE;
1663}
1664
1665static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
1666 [IFLA_GRE_LINK] = { .type = NLA_U32 },
1667 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 },
1668 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 },
1669 [IFLA_GRE_IKEY] = { .type = NLA_U32 },
1670 [IFLA_GRE_OKEY] = { .type = NLA_U32 },
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001671 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1672 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
Herbert Xuc19e6542008-10-09 11:59:55 -07001673 [IFLA_GRE_TTL] = { .type = NLA_U8 },
1674 [IFLA_GRE_TOS] = { .type = NLA_U8 },
1675 [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 },
1676};
1677
1678static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
1679 .kind = "gre",
1680 .maxtype = IFLA_GRE_MAX,
1681 .policy = ipgre_policy,
1682 .priv_size = sizeof(struct ip_tunnel),
1683 .setup = ipgre_tunnel_setup,
1684 .validate = ipgre_tunnel_validate,
1685 .newlink = ipgre_newlink,
1686 .changelink = ipgre_changelink,
1687 .get_size = ipgre_get_size,
1688 .fill_info = ipgre_fill_info,
1689};
1690
Herbert Xue1a80002008-10-09 12:00:17 -07001691static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
1692 .kind = "gretap",
1693 .maxtype = IFLA_GRE_MAX,
1694 .policy = ipgre_policy,
1695 .priv_size = sizeof(struct ip_tunnel),
1696 .setup = ipgre_tap_setup,
1697 .validate = ipgre_tap_validate,
1698 .newlink = ipgre_newlink,
1699 .changelink = ipgre_changelink,
1700 .get_size = ipgre_get_size,
1701 .fill_info = ipgre_fill_info,
1702};
1703
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704/*
1705 * And now the modules code and kernel interface.
1706 */
1707
1708static int __init ipgre_init(void)
1709{
1710 int err;
1711
1712 printk(KERN_INFO "GRE over IPv4 tunneling driver\n");
1713
Eric W. Biedermancfb8fbf2009-11-29 15:46:13 +00001714 err = register_pernet_device(&ipgre_net_ops);
Pavel Emelyanov59a4c752008-04-16 01:08:53 -07001715 if (err < 0)
Alexey Dobriyanc2892f02010-02-16 07:57:44 +00001716 return err;
1717
Dmitry Kozlov00959ad2010-08-21 23:05:39 -07001718 err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO);
Alexey Dobriyanc2892f02010-02-16 07:57:44 +00001719 if (err < 0) {
1720 printk(KERN_INFO "ipgre init: can't add protocol\n");
1721 goto add_proto_failed;
1722 }
Pavel Emelyanov7daa0002008-04-16 01:10:05 -07001723
Herbert Xuc19e6542008-10-09 11:59:55 -07001724 err = rtnl_link_register(&ipgre_link_ops);
1725 if (err < 0)
1726 goto rtnl_link_failed;
1727
Herbert Xue1a80002008-10-09 12:00:17 -07001728 err = rtnl_link_register(&ipgre_tap_ops);
1729 if (err < 0)
1730 goto tap_ops_failed;
1731
Herbert Xuc19e6542008-10-09 11:59:55 -07001732out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 return err;
Herbert Xuc19e6542008-10-09 11:59:55 -07001734
Herbert Xue1a80002008-10-09 12:00:17 -07001735tap_ops_failed:
1736 rtnl_link_unregister(&ipgre_link_ops);
Herbert Xuc19e6542008-10-09 11:59:55 -07001737rtnl_link_failed:
Dmitry Kozlov00959ad2010-08-21 23:05:39 -07001738 gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO);
Alexey Dobriyanc2892f02010-02-16 07:57:44 +00001739add_proto_failed:
1740 unregister_pernet_device(&ipgre_net_ops);
Herbert Xuc19e6542008-10-09 11:59:55 -07001741 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742}
1743
Alexey Kuznetsovdb445752005-07-30 17:46:44 -07001744static void __exit ipgre_fini(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745{
Herbert Xue1a80002008-10-09 12:00:17 -07001746 rtnl_link_unregister(&ipgre_tap_ops);
Herbert Xuc19e6542008-10-09 11:59:55 -07001747 rtnl_link_unregister(&ipgre_link_ops);
Dmitry Kozlov00959ad2010-08-21 23:05:39 -07001748 if (gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 printk(KERN_INFO "ipgre close: can't remove protocol\n");
Alexey Dobriyanc2892f02010-02-16 07:57:44 +00001750 unregister_pernet_device(&ipgre_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751}
1752
1753module_init(ipgre_init);
1754module_exit(ipgre_fini);
1755MODULE_LICENSE("GPL");
Patrick McHardy4d74f8b2008-10-10 12:11:06 -07001756MODULE_ALIAS_RTNL_LINK("gre");
1757MODULE_ALIAS_RTNL_LINK("gretap");
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03001758MODULE_ALIAS_NETDEV("gre0");