]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
e905a9ed | 2 | * Linux NET3: GRE over IP protocol decoder. |
1da177e4 LT |
3 | * |
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 | ||
4fc268d2 | 13 | #include <linux/capability.h> |
1da177e4 LT |
14 | #include <linux/module.h> |
15 | #include <linux/types.h> | |
1da177e4 LT |
16 | #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> | |
46f25dff | 30 | #include <linux/if_ether.h> |
1da177e4 LT |
31 | |
32 | #include <net/sock.h> | |
33 | #include <net/ip.h> | |
34 | #include <net/icmp.h> | |
35 | #include <net/protocol.h> | |
36 | #include <net/ipip.h> | |
37 | #include <net/arp.h> | |
38 | #include <net/checksum.h> | |
39 | #include <net/dsfield.h> | |
40 | #include <net/inet_ecn.h> | |
41 | #include <net/xfrm.h> | |
59a4c759 PE |
42 | #include <net/net_namespace.h> |
43 | #include <net/netns/generic.h> | |
1da177e4 LT |
44 | |
45 | #ifdef CONFIG_IPV6 | |
46 | #include <net/ipv6.h> | |
47 | #include <net/ip6_fib.h> | |
48 | #include <net/ip6_route.h> | |
49 | #endif | |
50 | ||
51 | /* | |
52 | Problems & solutions | |
53 | -------------------- | |
54 | ||
55 | 1. The most important issue is detecting local dead loops. | |
56 | They would cause complete host lockup in transmit, which | |
57 | would be "resolved" by stack overflow or, if queueing is enabled, | |
58 | with infinite looping in net_bh. | |
59 | ||
60 | We cannot track such dead loops during route installation, | |
61 | it is infeasible task. The most general solutions would be | |
62 | to keep skb->encapsulation counter (sort of local ttl), | |
63 | and silently drop packet when it expires. It is the best | |
64 | solution, but it supposes maintaing new variable in ALL | |
65 | skb, even if no tunneling is used. | |
66 | ||
e905a9ed | 67 | Current solution: t->recursion lock breaks dead loops. It looks |
1da177e4 LT |
68 | like dev->tbusy flag, but I preferred new variable, because |
69 | the semantics is different. One day, when hard_start_xmit | |
70 | will be multithreaded we will have to use skb->encapsulation. | |
71 | ||
72 | ||
73 | ||
74 | 2. Networking dead loops would not kill routers, but would really | |
75 | kill network. IP hop limit plays role of "t->recursion" in this case, | |
76 | if we copy it from packet being encapsulated to upper header. | |
77 | It is very good solution, but it introduces two problems: | |
78 | ||
79 | - Routing protocols, using packets with ttl=1 (OSPF, RIP2), | |
80 | do not work over tunnels. | |
81 | - traceroute does not work. I planned to relay ICMP from tunnel, | |
82 | so that this problem would be solved and traceroute output | |
83 | would even more informative. This idea appeared to be wrong: | |
84 | only Linux complies to rfc1812 now (yes, guys, Linux is the only | |
85 | true router now :-)), all routers (at least, in neighbourhood of mine) | |
86 | return only 8 bytes of payload. It is the end. | |
87 | ||
88 | Hence, if we want that OSPF worked or traceroute said something reasonable, | |
89 | we should search for another solution. | |
90 | ||
91 | One of them is to parse packet trying to detect inner encapsulation | |
92 | made by our node. It is difficult or even impossible, especially, | |
93 | taking into account fragmentation. TO be short, tt is not solution at all. | |
94 | ||
95 | Current solution: The solution was UNEXPECTEDLY SIMPLE. | |
96 | We force DF flag on tunnels with preconfigured hop limit, | |
97 | that is ALL. :-) Well, it does not remove the problem completely, | |
98 | but exponential growth of network traffic is changed to linear | |
99 | (branches, that exceed pmtu are pruned) and tunnel mtu | |
100 | fastly degrades to value <68, where looping stops. | |
101 | Yes, it is not good if there exists a router in the loop, | |
102 | which does not force DF, even when encapsulating packets have DF set. | |
103 | But it is not our problem! Nobody could accuse us, we made | |
104 | all that we could make. Even if it is your gated who injected | |
105 | fatal route to network, even if it were you who configured | |
106 | fatal static route: you are innocent. :-) | |
107 | ||
108 | ||
109 | ||
110 | 3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain | |
111 | practically identical code. It would be good to glue them | |
112 | together, but it is not very evident, how to make them modular. | |
113 | sit is integral part of IPv6, ipip and gre are naturally modular. | |
114 | We could extract common parts (hash table, ioctl etc) | |
115 | to a separate module (ip_tunnel.c). | |
116 | ||
117 | Alexey Kuznetsov. | |
118 | */ | |
119 | ||
120 | static int ipgre_tunnel_init(struct net_device *dev); | |
121 | static void ipgre_tunnel_setup(struct net_device *dev); | |
122 | ||
123 | /* Fallback tunnel: no source, no destination, no key, no options */ | |
124 | ||
125 | static int ipgre_fb_tunnel_init(struct net_device *dev); | |
126 | ||
59a4c759 PE |
127 | static int ipgre_net_id; |
128 | struct ipgre_net { | |
7daa0004 | 129 | struct net_device *fb_tunnel_dev; |
59a4c759 PE |
130 | }; |
131 | ||
1da177e4 LT |
132 | /* Tunnel hash table */ |
133 | ||
134 | /* | |
135 | 4 hash tables: | |
136 | ||
137 | 3: (remote,local) | |
138 | 2: (remote,*) | |
139 | 1: (*,local) | |
140 | 0: (*,*) | |
141 | ||
142 | We require exact key match i.e. if a key is present in packet | |
143 | it will match only tunnel with the same key; if it is not present, | |
144 | it will match only keyless tunnel. | |
145 | ||
146 | All keysless packets, if not matched configured keyless tunnels | |
147 | will match fallback tunnel. | |
148 | */ | |
149 | ||
150 | #define HASH_SIZE 16 | |
d5a0a1e3 | 151 | #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) |
1da177e4 LT |
152 | |
153 | static struct ip_tunnel *tunnels[4][HASH_SIZE]; | |
154 | ||
155 | #define tunnels_r_l (tunnels[3]) | |
156 | #define tunnels_r (tunnels[2]) | |
157 | #define tunnels_l (tunnels[1]) | |
158 | #define tunnels_wc (tunnels[0]) | |
159 | ||
160 | static DEFINE_RWLOCK(ipgre_lock); | |
161 | ||
162 | /* Given src, dst and key, find appropriate for input tunnel. */ | |
163 | ||
f57e7d5a PE |
164 | static struct ip_tunnel * ipgre_tunnel_lookup(struct net *net, |
165 | __be32 remote, __be32 local, __be32 key) | |
1da177e4 LT |
166 | { |
167 | unsigned h0 = HASH(remote); | |
168 | unsigned h1 = HASH(key); | |
169 | struct ip_tunnel *t; | |
7daa0004 | 170 | struct ipgre_net *ign = net_generic(net, ipgre_net_id); |
1da177e4 LT |
171 | |
172 | for (t = tunnels_r_l[h0^h1]; t; t = t->next) { | |
173 | if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) { | |
174 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
175 | return t; | |
176 | } | |
177 | } | |
178 | for (t = tunnels_r[h0^h1]; t; t = t->next) { | |
179 | if (remote == t->parms.iph.daddr) { | |
180 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
181 | return t; | |
182 | } | |
183 | } | |
184 | for (t = tunnels_l[h1]; t; t = t->next) { | |
185 | if (local == t->parms.iph.saddr || | |
f97c1e0c JP |
186 | (local == t->parms.iph.daddr && |
187 | ipv4_is_multicast(local))) { | |
1da177e4 LT |
188 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) |
189 | return t; | |
190 | } | |
191 | } | |
192 | for (t = tunnels_wc[h1]; t; t = t->next) { | |
193 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
194 | return t; | |
195 | } | |
196 | ||
7daa0004 PE |
197 | if (ign->fb_tunnel_dev->flags&IFF_UP) |
198 | return netdev_priv(ign->fb_tunnel_dev); | |
1da177e4 LT |
199 | return NULL; |
200 | } | |
201 | ||
f57e7d5a PE |
202 | static struct ip_tunnel **__ipgre_bucket(struct ipgre_net *ign, |
203 | struct ip_tunnel_parm *parms) | |
1da177e4 | 204 | { |
5056a1ef YH |
205 | __be32 remote = parms->iph.daddr; |
206 | __be32 local = parms->iph.saddr; | |
207 | __be32 key = parms->i_key; | |
1da177e4 LT |
208 | unsigned h = HASH(key); |
209 | int prio = 0; | |
210 | ||
211 | if (local) | |
212 | prio |= 1; | |
f97c1e0c | 213 | if (remote && !ipv4_is_multicast(remote)) { |
1da177e4 LT |
214 | prio |= 2; |
215 | h ^= HASH(remote); | |
216 | } | |
217 | ||
218 | return &tunnels[prio][h]; | |
219 | } | |
220 | ||
f57e7d5a PE |
221 | static inline struct ip_tunnel **ipgre_bucket(struct ipgre_net *ign, |
222 | struct ip_tunnel *t) | |
5056a1ef | 223 | { |
f57e7d5a | 224 | return __ipgre_bucket(ign, &t->parms); |
5056a1ef YH |
225 | } |
226 | ||
f57e7d5a | 227 | static void ipgre_tunnel_link(struct ipgre_net *ign, struct ip_tunnel *t) |
1da177e4 | 228 | { |
f57e7d5a | 229 | struct ip_tunnel **tp = ipgre_bucket(ign, t); |
1da177e4 LT |
230 | |
231 | t->next = *tp; | |
232 | write_lock_bh(&ipgre_lock); | |
233 | *tp = t; | |
234 | write_unlock_bh(&ipgre_lock); | |
235 | } | |
236 | ||
f57e7d5a | 237 | static void ipgre_tunnel_unlink(struct ipgre_net *ign, struct ip_tunnel *t) |
1da177e4 LT |
238 | { |
239 | struct ip_tunnel **tp; | |
240 | ||
f57e7d5a | 241 | for (tp = ipgre_bucket(ign, t); *tp; tp = &(*tp)->next) { |
1da177e4 LT |
242 | if (t == *tp) { |
243 | write_lock_bh(&ipgre_lock); | |
244 | *tp = t->next; | |
245 | write_unlock_bh(&ipgre_lock); | |
246 | break; | |
247 | } | |
248 | } | |
249 | } | |
250 | ||
f57e7d5a PE |
251 | static struct ip_tunnel * ipgre_tunnel_locate(struct net *net, |
252 | struct ip_tunnel_parm *parms, int create) | |
1da177e4 | 253 | { |
d5a0a1e3 AV |
254 | __be32 remote = parms->iph.daddr; |
255 | __be32 local = parms->iph.saddr; | |
256 | __be32 key = parms->i_key; | |
1da177e4 LT |
257 | struct ip_tunnel *t, **tp, *nt; |
258 | struct net_device *dev; | |
1da177e4 | 259 | char name[IFNAMSIZ]; |
f57e7d5a | 260 | struct ipgre_net *ign = net_generic(net, ipgre_net_id); |
1da177e4 | 261 | |
f57e7d5a | 262 | for (tp = __ipgre_bucket(ign, parms); (t = *tp) != NULL; tp = &t->next) { |
1da177e4 LT |
263 | if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) { |
264 | if (key == t->parms.i_key) | |
265 | return t; | |
266 | } | |
267 | } | |
268 | if (!create) | |
269 | return NULL; | |
270 | ||
271 | if (parms->name[0]) | |
272 | strlcpy(name, parms->name, IFNAMSIZ); | |
34cc7ba6 PE |
273 | else |
274 | sprintf(name, "gre%%d"); | |
1da177e4 LT |
275 | |
276 | dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup); | |
277 | if (!dev) | |
278 | return NULL; | |
279 | ||
b37d428b PE |
280 | if (strchr(name, '%')) { |
281 | if (dev_alloc_name(dev, name) < 0) | |
282 | goto failed_free; | |
283 | } | |
284 | ||
1da177e4 | 285 | dev->init = ipgre_tunnel_init; |
2941a486 | 286 | nt = netdev_priv(dev); |
1da177e4 LT |
287 | nt->parms = *parms; |
288 | ||
b37d428b PE |
289 | if (register_netdevice(dev) < 0) |
290 | goto failed_free; | |
1da177e4 | 291 | |
1da177e4 | 292 | dev_hold(dev); |
f57e7d5a | 293 | ipgre_tunnel_link(ign, nt); |
1da177e4 LT |
294 | return nt; |
295 | ||
b37d428b PE |
296 | failed_free: |
297 | free_netdev(dev); | |
1da177e4 LT |
298 | return NULL; |
299 | } | |
300 | ||
301 | static void ipgre_tunnel_uninit(struct net_device *dev) | |
302 | { | |
f57e7d5a PE |
303 | struct net *net = dev_net(dev); |
304 | struct ipgre_net *ign = net_generic(net, ipgre_net_id); | |
305 | ||
306 | ipgre_tunnel_unlink(ign, netdev_priv(dev)); | |
1da177e4 LT |
307 | dev_put(dev); |
308 | } | |
309 | ||
310 | ||
311 | static void ipgre_err(struct sk_buff *skb, u32 info) | |
312 | { | |
313 | #ifndef I_WISH_WORLD_WERE_PERFECT | |
314 | ||
315 | /* It is not :-( All the routers (except for Linux) return only | |
316 | 8 bytes of packet payload. It means, that precise relaying of | |
317 | ICMP in the real Internet is absolutely infeasible. | |
318 | ||
319 | Moreover, Cisco "wise men" put GRE key to the third word | |
320 | in GRE header. It makes impossible maintaining even soft state for keyed | |
321 | GRE tunnels with enabled checksum. Tell them "thank you". | |
322 | ||
323 | Well, I wonder, rfc1812 was written by Cisco employee, | |
324 | what the hell these idiots break standrads established | |
325 | by themself??? | |
326 | */ | |
327 | ||
328 | struct iphdr *iph = (struct iphdr*)skb->data; | |
d5a0a1e3 | 329 | __be16 *p = (__be16*)(skb->data+(iph->ihl<<2)); |
1da177e4 | 330 | int grehlen = (iph->ihl<<2) + 4; |
88c7664f ACM |
331 | const int type = icmp_hdr(skb)->type; |
332 | const int code = icmp_hdr(skb)->code; | |
1da177e4 | 333 | struct ip_tunnel *t; |
d5a0a1e3 | 334 | __be16 flags; |
1da177e4 LT |
335 | |
336 | flags = p[0]; | |
337 | if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { | |
338 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
339 | return; | |
340 | if (flags&GRE_KEY) { | |
341 | grehlen += 4; | |
342 | if (flags&GRE_CSUM) | |
343 | grehlen += 4; | |
344 | } | |
345 | } | |
346 | ||
347 | /* If only 8 bytes returned, keyed message will be dropped here */ | |
348 | if (skb_headlen(skb) < grehlen) | |
349 | return; | |
350 | ||
351 | switch (type) { | |
352 | default: | |
353 | case ICMP_PARAMETERPROB: | |
354 | return; | |
355 | ||
356 | case ICMP_DEST_UNREACH: | |
357 | switch (code) { | |
358 | case ICMP_SR_FAILED: | |
359 | case ICMP_PORT_UNREACH: | |
360 | /* Impossible event. */ | |
361 | return; | |
362 | case ICMP_FRAG_NEEDED: | |
363 | /* Soft state for pmtu is maintained by IP core. */ | |
364 | return; | |
365 | default: | |
366 | /* All others are translated to HOST_UNREACH. | |
367 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
368 | I believe they are just ether pollution. --ANK | |
369 | */ | |
370 | break; | |
371 | } | |
372 | break; | |
373 | case ICMP_TIME_EXCEEDED: | |
374 | if (code != ICMP_EXC_TTL) | |
375 | return; | |
376 | break; | |
377 | } | |
378 | ||
379 | read_lock(&ipgre_lock); | |
3b4667f3 | 380 | t = ipgre_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr, |
f57e7d5a PE |
381 | (flags&GRE_KEY) ? |
382 | *(((__be32*)p) + (grehlen>>2) - 1) : 0); | |
f97c1e0c JP |
383 | if (t == NULL || t->parms.iph.daddr == 0 || |
384 | ipv4_is_multicast(t->parms.iph.daddr)) | |
1da177e4 LT |
385 | goto out; |
386 | ||
387 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) | |
388 | goto out; | |
389 | ||
390 | if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO) | |
391 | t->err_count++; | |
392 | else | |
393 | t->err_count = 1; | |
394 | t->err_time = jiffies; | |
395 | out: | |
396 | read_unlock(&ipgre_lock); | |
397 | return; | |
398 | #else | |
399 | struct iphdr *iph = (struct iphdr*)dp; | |
400 | struct iphdr *eiph; | |
d5a0a1e3 | 401 | __be16 *p = (__be16*)(dp+(iph->ihl<<2)); |
88c7664f ACM |
402 | const int type = icmp_hdr(skb)->type; |
403 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
404 | int rel_type = 0; |
405 | int rel_code = 0; | |
c55e2f49 AV |
406 | __be32 rel_info = 0; |
407 | __u32 n = 0; | |
d5a0a1e3 | 408 | __be16 flags; |
1da177e4 LT |
409 | int grehlen = (iph->ihl<<2) + 4; |
410 | struct sk_buff *skb2; | |
411 | struct flowi fl; | |
412 | struct rtable *rt; | |
413 | ||
414 | if (p[1] != htons(ETH_P_IP)) | |
415 | return; | |
416 | ||
417 | flags = p[0]; | |
418 | if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { | |
419 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
420 | return; | |
421 | if (flags&GRE_CSUM) | |
422 | grehlen += 4; | |
423 | if (flags&GRE_KEY) | |
424 | grehlen += 4; | |
425 | if (flags&GRE_SEQ) | |
426 | grehlen += 4; | |
427 | } | |
428 | if (len < grehlen + sizeof(struct iphdr)) | |
429 | return; | |
430 | eiph = (struct iphdr*)(dp + grehlen); | |
431 | ||
432 | switch (type) { | |
433 | default: | |
434 | return; | |
435 | case ICMP_PARAMETERPROB: | |
88c7664f | 436 | n = ntohl(icmp_hdr(skb)->un.gateway) >> 24; |
c55e2f49 | 437 | if (n < (iph->ihl<<2)) |
1da177e4 LT |
438 | return; |
439 | ||
440 | /* So... This guy found something strange INSIDE encapsulated | |
441 | packet. Well, he is fool, but what can we do ? | |
442 | */ | |
443 | rel_type = ICMP_PARAMETERPROB; | |
c55e2f49 AV |
444 | n -= grehlen; |
445 | rel_info = htonl(n << 24); | |
1da177e4 LT |
446 | break; |
447 | ||
448 | case ICMP_DEST_UNREACH: | |
449 | switch (code) { | |
450 | case ICMP_SR_FAILED: | |
451 | case ICMP_PORT_UNREACH: | |
452 | /* Impossible event. */ | |
453 | return; | |
454 | case ICMP_FRAG_NEEDED: | |
455 | /* And it is the only really necessary thing :-) */ | |
88c7664f | 456 | n = ntohs(icmp_hdr(skb)->un.frag.mtu); |
c55e2f49 | 457 | if (n < grehlen+68) |
1da177e4 | 458 | return; |
c55e2f49 | 459 | n -= grehlen; |
1da177e4 | 460 | /* BSD 4.2 MORE DOES NOT EXIST IN NATURE. */ |
c55e2f49 | 461 | if (n > ntohs(eiph->tot_len)) |
1da177e4 | 462 | return; |
c55e2f49 | 463 | rel_info = htonl(n); |
1da177e4 LT |
464 | break; |
465 | default: | |
466 | /* All others are translated to HOST_UNREACH. | |
467 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
468 | I believe, it is just ether pollution. --ANK | |
469 | */ | |
470 | rel_type = ICMP_DEST_UNREACH; | |
471 | rel_code = ICMP_HOST_UNREACH; | |
472 | break; | |
473 | } | |
474 | break; | |
475 | case ICMP_TIME_EXCEEDED: | |
476 | if (code != ICMP_EXC_TTL) | |
477 | return; | |
478 | break; | |
479 | } | |
480 | ||
481 | /* Prepare fake skb to feed it to icmp_send */ | |
482 | skb2 = skb_clone(skb, GFP_ATOMIC); | |
483 | if (skb2 == NULL) | |
484 | return; | |
485 | dst_release(skb2->dst); | |
486 | skb2->dst = NULL; | |
487 | skb_pull(skb2, skb->data - (u8*)eiph); | |
c1d2bbe1 | 488 | skb_reset_network_header(skb2); |
1da177e4 LT |
489 | |
490 | /* Try to guess incoming interface */ | |
491 | memset(&fl, 0, sizeof(fl)); | |
492 | fl.fl4_dst = eiph->saddr; | |
493 | fl.fl4_tos = RT_TOS(eiph->tos); | |
494 | fl.proto = IPPROTO_GRE; | |
f206351a | 495 | if (ip_route_output_key(&init_net, &rt, &fl)) { |
1da177e4 LT |
496 | kfree_skb(skb2); |
497 | return; | |
498 | } | |
499 | skb2->dev = rt->u.dst.dev; | |
500 | ||
501 | /* route "incoming" packet */ | |
502 | if (rt->rt_flags&RTCF_LOCAL) { | |
503 | ip_rt_put(rt); | |
504 | rt = NULL; | |
505 | fl.fl4_dst = eiph->daddr; | |
506 | fl.fl4_src = eiph->saddr; | |
507 | fl.fl4_tos = eiph->tos; | |
f206351a | 508 | if (ip_route_output_key(&init_net, &rt, &fl) || |
1da177e4 LT |
509 | rt->u.dst.dev->type != ARPHRD_IPGRE) { |
510 | ip_rt_put(rt); | |
511 | kfree_skb(skb2); | |
512 | return; | |
513 | } | |
514 | } else { | |
515 | ip_rt_put(rt); | |
516 | if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, skb2->dev) || | |
517 | skb2->dst->dev->type != ARPHRD_IPGRE) { | |
518 | kfree_skb(skb2); | |
519 | return; | |
520 | } | |
521 | } | |
522 | ||
523 | /* change mtu on this route */ | |
524 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { | |
c55e2f49 | 525 | if (n > dst_mtu(skb2->dst)) { |
1da177e4 LT |
526 | kfree_skb(skb2); |
527 | return; | |
528 | } | |
c55e2f49 | 529 | skb2->dst->ops->update_pmtu(skb2->dst, n); |
1da177e4 | 530 | } else if (type == ICMP_TIME_EXCEEDED) { |
2941a486 | 531 | struct ip_tunnel *t = netdev_priv(skb2->dev); |
1da177e4 LT |
532 | if (t->parms.iph.ttl) { |
533 | rel_type = ICMP_DEST_UNREACH; | |
534 | rel_code = ICMP_HOST_UNREACH; | |
535 | } | |
536 | } | |
537 | ||
538 | icmp_send(skb2, rel_type, rel_code, rel_info); | |
539 | kfree_skb(skb2); | |
540 | #endif | |
541 | } | |
542 | ||
543 | static inline void ipgre_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb) | |
544 | { | |
545 | if (INET_ECN_is_ce(iph->tos)) { | |
546 | if (skb->protocol == htons(ETH_P_IP)) { | |
eddc9ec5 | 547 | IP_ECN_set_ce(ip_hdr(skb)); |
1da177e4 | 548 | } else if (skb->protocol == htons(ETH_P_IPV6)) { |
0660e03f | 549 | IP6_ECN_set_ce(ipv6_hdr(skb)); |
1da177e4 LT |
550 | } |
551 | } | |
552 | } | |
553 | ||
554 | static inline u8 | |
555 | ipgre_ecn_encapsulate(u8 tos, struct iphdr *old_iph, struct sk_buff *skb) | |
556 | { | |
557 | u8 inner = 0; | |
558 | if (skb->protocol == htons(ETH_P_IP)) | |
559 | inner = old_iph->tos; | |
560 | else if (skb->protocol == htons(ETH_P_IPV6)) | |
561 | inner = ipv6_get_dsfield((struct ipv6hdr *)old_iph); | |
562 | return INET_ECN_encapsulate(tos, inner); | |
563 | } | |
564 | ||
565 | static int ipgre_rcv(struct sk_buff *skb) | |
566 | { | |
567 | struct iphdr *iph; | |
568 | u8 *h; | |
d5a0a1e3 | 569 | __be16 flags; |
d3bc23e7 | 570 | __sum16 csum = 0; |
d5a0a1e3 | 571 | __be32 key = 0; |
1da177e4 LT |
572 | u32 seqno = 0; |
573 | struct ip_tunnel *tunnel; | |
574 | int offset = 4; | |
575 | ||
576 | if (!pskb_may_pull(skb, 16)) | |
577 | goto drop_nolock; | |
578 | ||
eddc9ec5 | 579 | iph = ip_hdr(skb); |
1da177e4 | 580 | h = skb->data; |
d5a0a1e3 | 581 | flags = *(__be16*)h; |
1da177e4 LT |
582 | |
583 | if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) { | |
584 | /* - Version must be 0. | |
585 | - We do not support routing headers. | |
586 | */ | |
587 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
588 | goto drop_nolock; | |
589 | ||
590 | if (flags&GRE_CSUM) { | |
fb286bb2 | 591 | switch (skb->ip_summed) { |
84fa7933 | 592 | case CHECKSUM_COMPLETE: |
d3bc23e7 | 593 | csum = csum_fold(skb->csum); |
fb286bb2 HX |
594 | if (!csum) |
595 | break; | |
596 | /* fall through */ | |
597 | case CHECKSUM_NONE: | |
598 | skb->csum = 0; | |
599 | csum = __skb_checksum_complete(skb); | |
84fa7933 | 600 | skb->ip_summed = CHECKSUM_COMPLETE; |
1da177e4 LT |
601 | } |
602 | offset += 4; | |
603 | } | |
604 | if (flags&GRE_KEY) { | |
d5a0a1e3 | 605 | key = *(__be32*)(h + offset); |
1da177e4 LT |
606 | offset += 4; |
607 | } | |
608 | if (flags&GRE_SEQ) { | |
d5a0a1e3 | 609 | seqno = ntohl(*(__be32*)(h + offset)); |
1da177e4 LT |
610 | offset += 4; |
611 | } | |
612 | } | |
613 | ||
614 | read_lock(&ipgre_lock); | |
3b4667f3 | 615 | if ((tunnel = ipgre_tunnel_lookup(dev_net(skb->dev), |
f57e7d5a | 616 | iph->saddr, iph->daddr, key)) != NULL) { |
1da177e4 LT |
617 | secpath_reset(skb); |
618 | ||
d5a0a1e3 | 619 | skb->protocol = *(__be16*)(h + 2); |
1da177e4 LT |
620 | /* WCCP version 1 and 2 protocol decoding. |
621 | * - Change protocol to IP | |
622 | * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header | |
623 | */ | |
624 | if (flags == 0 && | |
496c98df YH |
625 | skb->protocol == htons(ETH_P_WCCP)) { |
626 | skb->protocol = htons(ETH_P_IP); | |
e905a9ed | 627 | if ((*(h + offset) & 0xF0) != 0x40) |
1da177e4 LT |
628 | offset += 4; |
629 | } | |
630 | ||
1d069167 | 631 | skb->mac_header = skb->network_header; |
4209fb60 ACM |
632 | __pskb_pull(skb, offset); |
633 | skb_reset_network_header(skb); | |
9c70220b | 634 | skb_postpull_rcsum(skb, skb_transport_header(skb), offset); |
1da177e4 LT |
635 | skb->pkt_type = PACKET_HOST; |
636 | #ifdef CONFIG_NET_IPGRE_BROADCAST | |
f97c1e0c | 637 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 | 638 | /* Looped back packet, drop it! */ |
ee6b9673 | 639 | if (skb->rtable->fl.iif == 0) |
1da177e4 LT |
640 | goto drop; |
641 | tunnel->stat.multicast++; | |
642 | skb->pkt_type = PACKET_BROADCAST; | |
643 | } | |
644 | #endif | |
645 | ||
646 | if (((flags&GRE_CSUM) && csum) || | |
647 | (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) { | |
648 | tunnel->stat.rx_crc_errors++; | |
649 | tunnel->stat.rx_errors++; | |
650 | goto drop; | |
651 | } | |
652 | if (tunnel->parms.i_flags&GRE_SEQ) { | |
653 | if (!(flags&GRE_SEQ) || | |
654 | (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) { | |
655 | tunnel->stat.rx_fifo_errors++; | |
656 | tunnel->stat.rx_errors++; | |
657 | goto drop; | |
658 | } | |
659 | tunnel->i_seqno = seqno + 1; | |
660 | } | |
661 | tunnel->stat.rx_packets++; | |
662 | tunnel->stat.rx_bytes += skb->len; | |
663 | skb->dev = tunnel->dev; | |
664 | dst_release(skb->dst); | |
665 | skb->dst = NULL; | |
666 | nf_reset(skb); | |
667 | ipgre_ecn_decapsulate(iph, skb); | |
668 | netif_rx(skb); | |
669 | read_unlock(&ipgre_lock); | |
670 | return(0); | |
671 | } | |
45af08be | 672 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1da177e4 LT |
673 | |
674 | drop: | |
675 | read_unlock(&ipgre_lock); | |
676 | drop_nolock: | |
677 | kfree_skb(skb); | |
678 | return(0); | |
679 | } | |
680 | ||
681 | static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) | |
682 | { | |
2941a486 | 683 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 | 684 | struct net_device_stats *stats = &tunnel->stat; |
eddc9ec5 | 685 | struct iphdr *old_iph = ip_hdr(skb); |
1da177e4 LT |
686 | struct iphdr *tiph; |
687 | u8 tos; | |
d5a0a1e3 | 688 | __be16 df; |
1da177e4 LT |
689 | struct rtable *rt; /* Route to the other host */ |
690 | struct net_device *tdev; /* Device to other host */ | |
691 | struct iphdr *iph; /* Our new IP header */ | |
c2636b4d | 692 | unsigned int max_headroom; /* The extra header space needed */ |
1da177e4 | 693 | int gre_hlen; |
d5a0a1e3 | 694 | __be32 dst; |
1da177e4 LT |
695 | int mtu; |
696 | ||
697 | if (tunnel->recursion++) { | |
698 | tunnel->stat.collisions++; | |
699 | goto tx_error; | |
700 | } | |
701 | ||
3b04ddde | 702 | if (dev->header_ops) { |
1da177e4 LT |
703 | gre_hlen = 0; |
704 | tiph = (struct iphdr*)skb->data; | |
705 | } else { | |
706 | gre_hlen = tunnel->hlen; | |
707 | tiph = &tunnel->parms.iph; | |
708 | } | |
709 | ||
710 | if ((dst = tiph->daddr) == 0) { | |
711 | /* NBMA tunnel */ | |
712 | ||
713 | if (skb->dst == NULL) { | |
714 | tunnel->stat.tx_fifo_errors++; | |
715 | goto tx_error; | |
716 | } | |
717 | ||
718 | if (skb->protocol == htons(ETH_P_IP)) { | |
ee6b9673 | 719 | rt = skb->rtable; |
1da177e4 LT |
720 | if ((dst = rt->rt_gateway) == 0) |
721 | goto tx_error_icmp; | |
722 | } | |
723 | #ifdef CONFIG_IPV6 | |
724 | else if (skb->protocol == htons(ETH_P_IPV6)) { | |
725 | struct in6_addr *addr6; | |
726 | int addr_type; | |
727 | struct neighbour *neigh = skb->dst->neighbour; | |
728 | ||
729 | if (neigh == NULL) | |
730 | goto tx_error; | |
731 | ||
732 | addr6 = (struct in6_addr*)&neigh->primary_key; | |
733 | addr_type = ipv6_addr_type(addr6); | |
734 | ||
735 | if (addr_type == IPV6_ADDR_ANY) { | |
0660e03f | 736 | addr6 = &ipv6_hdr(skb)->daddr; |
1da177e4 LT |
737 | addr_type = ipv6_addr_type(addr6); |
738 | } | |
739 | ||
740 | if ((addr_type & IPV6_ADDR_COMPATv4) == 0) | |
741 | goto tx_error_icmp; | |
742 | ||
743 | dst = addr6->s6_addr32[3]; | |
744 | } | |
745 | #endif | |
746 | else | |
747 | goto tx_error; | |
748 | } | |
749 | ||
750 | tos = tiph->tos; | |
751 | if (tos&1) { | |
752 | if (skb->protocol == htons(ETH_P_IP)) | |
753 | tos = old_iph->tos; | |
754 | tos &= ~1; | |
755 | } | |
756 | ||
757 | { | |
758 | struct flowi fl = { .oif = tunnel->parms.link, | |
759 | .nl_u = { .ip4_u = | |
760 | { .daddr = dst, | |
761 | .saddr = tiph->saddr, | |
762 | .tos = RT_TOS(tos) } }, | |
763 | .proto = IPPROTO_GRE }; | |
f206351a | 764 | if (ip_route_output_key(&init_net, &rt, &fl)) { |
1da177e4 LT |
765 | tunnel->stat.tx_carrier_errors++; |
766 | goto tx_error; | |
767 | } | |
768 | } | |
769 | tdev = rt->u.dst.dev; | |
770 | ||
771 | if (tdev == dev) { | |
772 | ip_rt_put(rt); | |
773 | tunnel->stat.collisions++; | |
774 | goto tx_error; | |
775 | } | |
776 | ||
777 | df = tiph->frag_off; | |
778 | if (df) | |
779 | mtu = dst_mtu(&rt->u.dst) - tunnel->hlen; | |
780 | else | |
781 | mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu; | |
782 | ||
783 | if (skb->dst) | |
784 | skb->dst->ops->update_pmtu(skb->dst, mtu); | |
785 | ||
786 | if (skb->protocol == htons(ETH_P_IP)) { | |
787 | df |= (old_iph->frag_off&htons(IP_DF)); | |
788 | ||
789 | if ((old_iph->frag_off&htons(IP_DF)) && | |
790 | mtu < ntohs(old_iph->tot_len)) { | |
791 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu)); | |
792 | ip_rt_put(rt); | |
793 | goto tx_error; | |
794 | } | |
795 | } | |
796 | #ifdef CONFIG_IPV6 | |
797 | else if (skb->protocol == htons(ETH_P_IPV6)) { | |
798 | struct rt6_info *rt6 = (struct rt6_info*)skb->dst; | |
799 | ||
800 | if (rt6 && mtu < dst_mtu(skb->dst) && mtu >= IPV6_MIN_MTU) { | |
f97c1e0c JP |
801 | if ((tunnel->parms.iph.daddr && |
802 | !ipv4_is_multicast(tunnel->parms.iph.daddr)) || | |
1da177e4 LT |
803 | rt6->rt6i_dst.plen == 128) { |
804 | rt6->rt6i_flags |= RTF_MODIFIED; | |
805 | skb->dst->metrics[RTAX_MTU-1] = mtu; | |
806 | } | |
807 | } | |
808 | ||
809 | if (mtu >= IPV6_MIN_MTU && mtu < skb->len - tunnel->hlen + gre_hlen) { | |
810 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); | |
811 | ip_rt_put(rt); | |
812 | goto tx_error; | |
813 | } | |
814 | } | |
815 | #endif | |
816 | ||
817 | if (tunnel->err_count > 0) { | |
818 | if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) { | |
819 | tunnel->err_count--; | |
820 | ||
821 | dst_link_failure(skb); | |
822 | } else | |
823 | tunnel->err_count = 0; | |
824 | } | |
825 | ||
826 | max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen; | |
827 | ||
cfbba49d PM |
828 | if (skb_headroom(skb) < max_headroom || skb_shared(skb)|| |
829 | (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { | |
1da177e4 LT |
830 | struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); |
831 | if (!new_skb) { | |
832 | ip_rt_put(rt); | |
e905a9ed | 833 | stats->tx_dropped++; |
1da177e4 LT |
834 | dev_kfree_skb(skb); |
835 | tunnel->recursion--; | |
836 | return 0; | |
837 | } | |
838 | if (skb->sk) | |
839 | skb_set_owner_w(new_skb, skb->sk); | |
840 | dev_kfree_skb(skb); | |
841 | skb = new_skb; | |
eddc9ec5 | 842 | old_iph = ip_hdr(skb); |
1da177e4 LT |
843 | } |
844 | ||
b0e380b1 | 845 | skb->transport_header = skb->network_header; |
e2d1bca7 ACM |
846 | skb_push(skb, gre_hlen); |
847 | skb_reset_network_header(skb); | |
1da177e4 | 848 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
48d5cad8 PM |
849 | IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | |
850 | IPSKB_REROUTED); | |
1da177e4 LT |
851 | dst_release(skb->dst); |
852 | skb->dst = &rt->u.dst; | |
853 | ||
854 | /* | |
855 | * Push down and install the IPIP header. | |
856 | */ | |
857 | ||
eddc9ec5 | 858 | iph = ip_hdr(skb); |
1da177e4 LT |
859 | iph->version = 4; |
860 | iph->ihl = sizeof(struct iphdr) >> 2; | |
861 | iph->frag_off = df; | |
862 | iph->protocol = IPPROTO_GRE; | |
863 | iph->tos = ipgre_ecn_encapsulate(tos, old_iph, skb); | |
864 | iph->daddr = rt->rt_dst; | |
865 | iph->saddr = rt->rt_src; | |
866 | ||
867 | if ((iph->ttl = tiph->ttl) == 0) { | |
868 | if (skb->protocol == htons(ETH_P_IP)) | |
869 | iph->ttl = old_iph->ttl; | |
870 | #ifdef CONFIG_IPV6 | |
871 | else if (skb->protocol == htons(ETH_P_IPV6)) | |
872 | iph->ttl = ((struct ipv6hdr*)old_iph)->hop_limit; | |
873 | #endif | |
874 | else | |
875 | iph->ttl = dst_metric(&rt->u.dst, RTAX_HOPLIMIT); | |
876 | } | |
877 | ||
d5a0a1e3 AV |
878 | ((__be16*)(iph+1))[0] = tunnel->parms.o_flags; |
879 | ((__be16*)(iph+1))[1] = skb->protocol; | |
1da177e4 LT |
880 | |
881 | if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) { | |
d5a0a1e3 | 882 | __be32 *ptr = (__be32*)(((u8*)iph) + tunnel->hlen - 4); |
1da177e4 LT |
883 | |
884 | if (tunnel->parms.o_flags&GRE_SEQ) { | |
885 | ++tunnel->o_seqno; | |
886 | *ptr = htonl(tunnel->o_seqno); | |
887 | ptr--; | |
888 | } | |
889 | if (tunnel->parms.o_flags&GRE_KEY) { | |
890 | *ptr = tunnel->parms.o_key; | |
891 | ptr--; | |
892 | } | |
893 | if (tunnel->parms.o_flags&GRE_CSUM) { | |
894 | *ptr = 0; | |
5f92a738 | 895 | *(__sum16*)ptr = ip_compute_csum((void*)(iph+1), skb->len - sizeof(struct iphdr)); |
1da177e4 LT |
896 | } |
897 | } | |
898 | ||
899 | nf_reset(skb); | |
900 | ||
901 | IPTUNNEL_XMIT(); | |
902 | tunnel->recursion--; | |
903 | return 0; | |
904 | ||
905 | tx_error_icmp: | |
906 | dst_link_failure(skb); | |
907 | ||
908 | tx_error: | |
909 | stats->tx_errors++; | |
910 | dev_kfree_skb(skb); | |
911 | tunnel->recursion--; | |
912 | return 0; | |
913 | } | |
914 | ||
ee34c1eb MS |
915 | static void ipgre_tunnel_bind_dev(struct net_device *dev) |
916 | { | |
917 | struct net_device *tdev = NULL; | |
918 | struct ip_tunnel *tunnel; | |
919 | struct iphdr *iph; | |
920 | int hlen = LL_MAX_HEADER; | |
921 | int mtu = ETH_DATA_LEN; | |
922 | int addend = sizeof(struct iphdr) + 4; | |
923 | ||
924 | tunnel = netdev_priv(dev); | |
925 | iph = &tunnel->parms.iph; | |
926 | ||
927 | /* Guess output device to choose reasonable mtu and hard_header_len */ | |
928 | ||
929 | if (iph->daddr) { | |
930 | struct flowi fl = { .oif = tunnel->parms.link, | |
931 | .nl_u = { .ip4_u = | |
932 | { .daddr = iph->daddr, | |
933 | .saddr = iph->saddr, | |
934 | .tos = RT_TOS(iph->tos) } }, | |
935 | .proto = IPPROTO_GRE }; | |
936 | struct rtable *rt; | |
f206351a | 937 | if (!ip_route_output_key(&init_net, &rt, &fl)) { |
ee34c1eb MS |
938 | tdev = rt->u.dst.dev; |
939 | ip_rt_put(rt); | |
940 | } | |
941 | dev->flags |= IFF_POINTOPOINT; | |
942 | } | |
943 | ||
944 | if (!tdev && tunnel->parms.link) | |
945 | tdev = __dev_get_by_index(&init_net, tunnel->parms.link); | |
946 | ||
947 | if (tdev) { | |
948 | hlen = tdev->hard_header_len; | |
949 | mtu = tdev->mtu; | |
950 | } | |
951 | dev->iflink = tunnel->parms.link; | |
952 | ||
953 | /* Precalculate GRE options length */ | |
954 | if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) { | |
955 | if (tunnel->parms.o_flags&GRE_CSUM) | |
956 | addend += 4; | |
957 | if (tunnel->parms.o_flags&GRE_KEY) | |
958 | addend += 4; | |
959 | if (tunnel->parms.o_flags&GRE_SEQ) | |
960 | addend += 4; | |
961 | } | |
962 | dev->hard_header_len = hlen + addend; | |
963 | dev->mtu = mtu - addend; | |
964 | tunnel->hlen = addend; | |
965 | ||
966 | } | |
967 | ||
1da177e4 LT |
968 | static int |
969 | ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) | |
970 | { | |
971 | int err = 0; | |
972 | struct ip_tunnel_parm p; | |
973 | struct ip_tunnel *t; | |
f57e7d5a PE |
974 | struct net *net = dev_net(dev); |
975 | struct ipgre_net *ign = net_generic(net, ipgre_net_id); | |
1da177e4 LT |
976 | |
977 | switch (cmd) { | |
978 | case SIOCGETTUNNEL: | |
979 | t = NULL; | |
7daa0004 | 980 | if (dev == ign->fb_tunnel_dev) { |
1da177e4 LT |
981 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { |
982 | err = -EFAULT; | |
983 | break; | |
984 | } | |
f57e7d5a | 985 | t = ipgre_tunnel_locate(net, &p, 0); |
1da177e4 LT |
986 | } |
987 | if (t == NULL) | |
2941a486 | 988 | t = netdev_priv(dev); |
1da177e4 LT |
989 | memcpy(&p, &t->parms, sizeof(p)); |
990 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
991 | err = -EFAULT; | |
992 | break; | |
993 | ||
994 | case SIOCADDTUNNEL: | |
995 | case SIOCCHGTUNNEL: | |
996 | err = -EPERM; | |
997 | if (!capable(CAP_NET_ADMIN)) | |
998 | goto done; | |
999 | ||
1000 | err = -EFAULT; | |
1001 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
1002 | goto done; | |
1003 | ||
1004 | err = -EINVAL; | |
1005 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE || | |
1006 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) || | |
1007 | ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) | |
1008 | goto done; | |
1009 | if (p.iph.ttl) | |
1010 | p.iph.frag_off |= htons(IP_DF); | |
1011 | ||
1012 | if (!(p.i_flags&GRE_KEY)) | |
1013 | p.i_key = 0; | |
1014 | if (!(p.o_flags&GRE_KEY)) | |
1015 | p.o_key = 0; | |
1016 | ||
f57e7d5a | 1017 | t = ipgre_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); |
1da177e4 | 1018 | |
7daa0004 | 1019 | if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { |
1da177e4 LT |
1020 | if (t != NULL) { |
1021 | if (t->dev != dev) { | |
1022 | err = -EEXIST; | |
1023 | break; | |
1024 | } | |
1025 | } else { | |
1026 | unsigned nflags=0; | |
1027 | ||
2941a486 | 1028 | t = netdev_priv(dev); |
1da177e4 | 1029 | |
f97c1e0c | 1030 | if (ipv4_is_multicast(p.iph.daddr)) |
1da177e4 LT |
1031 | nflags = IFF_BROADCAST; |
1032 | else if (p.iph.daddr) | |
1033 | nflags = IFF_POINTOPOINT; | |
1034 | ||
1035 | if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) { | |
1036 | err = -EINVAL; | |
1037 | break; | |
1038 | } | |
f57e7d5a | 1039 | ipgre_tunnel_unlink(ign, t); |
1da177e4 LT |
1040 | t->parms.iph.saddr = p.iph.saddr; |
1041 | t->parms.iph.daddr = p.iph.daddr; | |
1042 | t->parms.i_key = p.i_key; | |
1043 | t->parms.o_key = p.o_key; | |
1044 | memcpy(dev->dev_addr, &p.iph.saddr, 4); | |
1045 | memcpy(dev->broadcast, &p.iph.daddr, 4); | |
f57e7d5a | 1046 | ipgre_tunnel_link(ign, t); |
1da177e4 LT |
1047 | netdev_state_change(dev); |
1048 | } | |
1049 | } | |
1050 | ||
1051 | if (t) { | |
1052 | err = 0; | |
1053 | if (cmd == SIOCCHGTUNNEL) { | |
1054 | t->parms.iph.ttl = p.iph.ttl; | |
1055 | t->parms.iph.tos = p.iph.tos; | |
1056 | t->parms.iph.frag_off = p.iph.frag_off; | |
ee34c1eb MS |
1057 | if (t->parms.link != p.link) { |
1058 | t->parms.link = p.link; | |
1059 | ipgre_tunnel_bind_dev(dev); | |
1060 | netdev_state_change(dev); | |
1061 | } | |
1da177e4 LT |
1062 | } |
1063 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) | |
1064 | err = -EFAULT; | |
1065 | } else | |
1066 | err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); | |
1067 | break; | |
1068 | ||
1069 | case SIOCDELTUNNEL: | |
1070 | err = -EPERM; | |
1071 | if (!capable(CAP_NET_ADMIN)) | |
1072 | goto done; | |
1073 | ||
7daa0004 | 1074 | if (dev == ign->fb_tunnel_dev) { |
1da177e4 LT |
1075 | err = -EFAULT; |
1076 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
1077 | goto done; | |
1078 | err = -ENOENT; | |
f57e7d5a | 1079 | if ((t = ipgre_tunnel_locate(net, &p, 0)) == NULL) |
1da177e4 LT |
1080 | goto done; |
1081 | err = -EPERM; | |
7daa0004 | 1082 | if (t == netdev_priv(ign->fb_tunnel_dev)) |
1da177e4 LT |
1083 | goto done; |
1084 | dev = t->dev; | |
1085 | } | |
22f8cde5 SH |
1086 | unregister_netdevice(dev); |
1087 | err = 0; | |
1da177e4 LT |
1088 | break; |
1089 | ||
1090 | default: | |
1091 | err = -EINVAL; | |
1092 | } | |
1093 | ||
1094 | done: | |
1095 | return err; | |
1096 | } | |
1097 | ||
1098 | static struct net_device_stats *ipgre_tunnel_get_stats(struct net_device *dev) | |
1099 | { | |
2941a486 | 1100 | return &(((struct ip_tunnel*)netdev_priv(dev))->stat); |
1da177e4 LT |
1101 | } |
1102 | ||
1103 | static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu) | |
1104 | { | |
2941a486 | 1105 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 LT |
1106 | if (new_mtu < 68 || new_mtu > 0xFFF8 - tunnel->hlen) |
1107 | return -EINVAL; | |
1108 | dev->mtu = new_mtu; | |
1109 | return 0; | |
1110 | } | |
1111 | ||
1da177e4 LT |
1112 | /* Nice toy. Unfortunately, useless in real life :-) |
1113 | It allows to construct virtual multiprotocol broadcast "LAN" | |
1114 | over the Internet, provided multicast routing is tuned. | |
1115 | ||
1116 | ||
1117 | I have no idea was this bicycle invented before me, | |
1118 | so that I had to set ARPHRD_IPGRE to a random value. | |
1119 | I have an impression, that Cisco could make something similar, | |
1120 | but this feature is apparently missing in IOS<=11.2(8). | |
e905a9ed | 1121 | |
1da177e4 LT |
1122 | I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks |
1123 | with broadcast 224.66.66.66. If you have access to mbone, play with me :-) | |
1124 | ||
1125 | ping -t 255 224.66.66.66 | |
1126 | ||
1127 | If nobody answers, mbone does not work. | |
1128 | ||
1129 | ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255 | |
1130 | ip addr add 10.66.66.<somewhat>/24 dev Universe | |
1131 | ifconfig Universe up | |
1132 | ifconfig Universe add fe80::<Your_real_addr>/10 | |
1133 | ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96 | |
1134 | ftp 10.66.66.66 | |
1135 | ... | |
1136 | ftp fec0:6666:6666::193.233.7.65 | |
1137 | ... | |
1138 | ||
1139 | */ | |
1140 | ||
3b04ddde SH |
1141 | static int ipgre_header(struct sk_buff *skb, struct net_device *dev, |
1142 | unsigned short type, | |
1143 | const void *daddr, const void *saddr, unsigned len) | |
1da177e4 | 1144 | { |
2941a486 | 1145 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 1146 | struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen); |
d5a0a1e3 | 1147 | __be16 *p = (__be16*)(iph+1); |
1da177e4 LT |
1148 | |
1149 | memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); | |
1150 | p[0] = t->parms.o_flags; | |
1151 | p[1] = htons(type); | |
1152 | ||
1153 | /* | |
e905a9ed | 1154 | * Set the source hardware address. |
1da177e4 | 1155 | */ |
e905a9ed | 1156 | |
1da177e4 LT |
1157 | if (saddr) |
1158 | memcpy(&iph->saddr, saddr, 4); | |
1159 | ||
1160 | if (daddr) { | |
1161 | memcpy(&iph->daddr, daddr, 4); | |
1162 | return t->hlen; | |
1163 | } | |
f97c1e0c | 1164 | if (iph->daddr && !ipv4_is_multicast(iph->daddr)) |
1da177e4 | 1165 | return t->hlen; |
e905a9ed | 1166 | |
1da177e4 LT |
1167 | return -t->hlen; |
1168 | } | |
1169 | ||
6a5f44d7 TT |
1170 | static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) |
1171 | { | |
1172 | struct iphdr *iph = (struct iphdr*) skb_mac_header(skb); | |
1173 | memcpy(haddr, &iph->saddr, 4); | |
1174 | return 4; | |
1175 | } | |
1176 | ||
3b04ddde SH |
1177 | static const struct header_ops ipgre_header_ops = { |
1178 | .create = ipgre_header, | |
6a5f44d7 | 1179 | .parse = ipgre_header_parse, |
3b04ddde SH |
1180 | }; |
1181 | ||
6a5f44d7 | 1182 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
1da177e4 LT |
1183 | static int ipgre_open(struct net_device *dev) |
1184 | { | |
2941a486 | 1185 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 1186 | |
f97c1e0c | 1187 | if (ipv4_is_multicast(t->parms.iph.daddr)) { |
1da177e4 LT |
1188 | struct flowi fl = { .oif = t->parms.link, |
1189 | .nl_u = { .ip4_u = | |
1190 | { .daddr = t->parms.iph.daddr, | |
1191 | .saddr = t->parms.iph.saddr, | |
1192 | .tos = RT_TOS(t->parms.iph.tos) } }, | |
1193 | .proto = IPPROTO_GRE }; | |
1194 | struct rtable *rt; | |
f206351a | 1195 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1196 | return -EADDRNOTAVAIL; |
1197 | dev = rt->u.dst.dev; | |
1198 | ip_rt_put(rt); | |
e5ed6399 | 1199 | if (__in_dev_get_rtnl(dev) == NULL) |
1da177e4 LT |
1200 | return -EADDRNOTAVAIL; |
1201 | t->mlink = dev->ifindex; | |
e5ed6399 | 1202 | ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr); |
1da177e4 LT |
1203 | } |
1204 | return 0; | |
1205 | } | |
1206 | ||
1207 | static int ipgre_close(struct net_device *dev) | |
1208 | { | |
2941a486 | 1209 | struct ip_tunnel *t = netdev_priv(dev); |
f97c1e0c | 1210 | if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) { |
7fee0ca2 | 1211 | struct in_device *in_dev; |
c346dca1 | 1212 | in_dev = inetdev_by_index(dev_net(dev), t->mlink); |
1da177e4 LT |
1213 | if (in_dev) { |
1214 | ip_mc_dec_group(in_dev, t->parms.iph.daddr); | |
1215 | in_dev_put(in_dev); | |
1216 | } | |
1217 | } | |
1218 | return 0; | |
1219 | } | |
1220 | ||
1221 | #endif | |
1222 | ||
1223 | static void ipgre_tunnel_setup(struct net_device *dev) | |
1224 | { | |
1da177e4 LT |
1225 | dev->uninit = ipgre_tunnel_uninit; |
1226 | dev->destructor = free_netdev; | |
1227 | dev->hard_start_xmit = ipgre_tunnel_xmit; | |
1228 | dev->get_stats = ipgre_tunnel_get_stats; | |
1229 | dev->do_ioctl = ipgre_tunnel_ioctl; | |
1230 | dev->change_mtu = ipgre_tunnel_change_mtu; | |
1231 | ||
1232 | dev->type = ARPHRD_IPGRE; | |
1233 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr) + 4; | |
46f25dff | 1234 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4; |
1da177e4 LT |
1235 | dev->flags = IFF_NOARP; |
1236 | dev->iflink = 0; | |
1237 | dev->addr_len = 4; | |
1238 | } | |
1239 | ||
1240 | static int ipgre_tunnel_init(struct net_device *dev) | |
1241 | { | |
1da177e4 LT |
1242 | struct ip_tunnel *tunnel; |
1243 | struct iphdr *iph; | |
1da177e4 | 1244 | |
2941a486 | 1245 | tunnel = netdev_priv(dev); |
1da177e4 LT |
1246 | iph = &tunnel->parms.iph; |
1247 | ||
1248 | tunnel->dev = dev; | |
1249 | strcpy(tunnel->parms.name, dev->name); | |
1250 | ||
1251 | memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); | |
1252 | memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); | |
1253 | ||
ee34c1eb | 1254 | ipgre_tunnel_bind_dev(dev); |
1da177e4 LT |
1255 | |
1256 | if (iph->daddr) { | |
1da177e4 | 1257 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
f97c1e0c | 1258 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 LT |
1259 | if (!iph->saddr) |
1260 | return -EINVAL; | |
1261 | dev->flags = IFF_BROADCAST; | |
3b04ddde | 1262 | dev->header_ops = &ipgre_header_ops; |
1da177e4 LT |
1263 | dev->open = ipgre_open; |
1264 | dev->stop = ipgre_close; | |
1265 | } | |
1266 | #endif | |
ee34c1eb | 1267 | } else |
6a5f44d7 | 1268 | dev->header_ops = &ipgre_header_ops; |
1da177e4 | 1269 | |
1da177e4 LT |
1270 | return 0; |
1271 | } | |
1272 | ||
7daa0004 | 1273 | static int ipgre_fb_tunnel_init(struct net_device *dev) |
1da177e4 | 1274 | { |
2941a486 | 1275 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 LT |
1276 | struct iphdr *iph = &tunnel->parms.iph; |
1277 | ||
1278 | tunnel->dev = dev; | |
1279 | strcpy(tunnel->parms.name, dev->name); | |
1280 | ||
1281 | iph->version = 4; | |
1282 | iph->protocol = IPPROTO_GRE; | |
1283 | iph->ihl = 5; | |
1284 | tunnel->hlen = sizeof(struct iphdr) + 4; | |
1285 | ||
1286 | dev_hold(dev); | |
1287 | tunnels_wc[0] = tunnel; | |
1288 | return 0; | |
1289 | } | |
1290 | ||
1291 | ||
1292 | static struct net_protocol ipgre_protocol = { | |
1293 | .handler = ipgre_rcv, | |
1294 | .err_handler = ipgre_err, | |
1295 | }; | |
1296 | ||
59a4c759 PE |
1297 | static int ipgre_init_net(struct net *net) |
1298 | { | |
1299 | int err; | |
1300 | struct ipgre_net *ign; | |
1301 | ||
1302 | err = -ENOMEM; | |
1303 | ign = kmalloc(sizeof(struct ipgre_net), GFP_KERNEL); | |
1304 | if (ign == NULL) | |
1305 | goto err_alloc; | |
1306 | ||
1307 | err = net_assign_generic(net, ipgre_net_id, ign); | |
1308 | if (err < 0) | |
1309 | goto err_assign; | |
1310 | ||
7daa0004 PE |
1311 | ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "gre0", |
1312 | ipgre_tunnel_setup); | |
1313 | if (!ign->fb_tunnel_dev) { | |
1314 | err = -ENOMEM; | |
1315 | goto err_alloc_dev; | |
1316 | } | |
1317 | ||
1318 | ign->fb_tunnel_dev->init = ipgre_fb_tunnel_init; | |
1319 | dev_net_set(ign->fb_tunnel_dev, net); | |
1320 | ||
1321 | if ((err = register_netdev(ign->fb_tunnel_dev))) | |
1322 | goto err_reg_dev; | |
1323 | ||
59a4c759 PE |
1324 | return 0; |
1325 | ||
7daa0004 PE |
1326 | err_reg_dev: |
1327 | free_netdev(ign->fb_tunnel_dev); | |
1328 | err_alloc_dev: | |
1329 | /* nothing */ | |
59a4c759 PE |
1330 | err_assign: |
1331 | kfree(ign); | |
1332 | err_alloc: | |
1333 | return err; | |
1334 | } | |
1335 | ||
1336 | static void ipgre_exit_net(struct net *net) | |
1337 | { | |
1338 | struct ipgre_net *ign; | |
1339 | ||
1340 | ign = net_generic(net, ipgre_net_id); | |
7daa0004 PE |
1341 | rtnl_lock(); |
1342 | if (net != &init_net) | |
1343 | unregister_netdevice(ign->fb_tunnel_dev); | |
1344 | rtnl_unlock(); | |
59a4c759 PE |
1345 | kfree(ign); |
1346 | } | |
1347 | ||
1348 | static struct pernet_operations ipgre_net_ops = { | |
1349 | .init = ipgre_init_net, | |
1350 | .exit = ipgre_exit_net, | |
1351 | }; | |
1da177e4 LT |
1352 | |
1353 | /* | |
1354 | * And now the modules code and kernel interface. | |
1355 | */ | |
1356 | ||
1357 | static int __init ipgre_init(void) | |
1358 | { | |
1359 | int err; | |
1360 | ||
1361 | printk(KERN_INFO "GRE over IPv4 tunneling driver\n"); | |
1362 | ||
1363 | if (inet_add_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) { | |
1364 | printk(KERN_INFO "ipgre init: can't add protocol\n"); | |
1365 | return -EAGAIN; | |
1366 | } | |
1367 | ||
59a4c759 PE |
1368 | err = register_pernet_gen_device(&ipgre_net_id, &ipgre_net_ops); |
1369 | if (err < 0) | |
7daa0004 PE |
1370 | inet_del_protocol(&ipgre_protocol, IPPROTO_GRE); |
1371 | ||
1da177e4 | 1372 | return err; |
1da177e4 LT |
1373 | } |
1374 | ||
db44575f AK |
1375 | static void __exit ipgre_destroy_tunnels(void) |
1376 | { | |
1377 | int prio; | |
1378 | ||
1379 | for (prio = 0; prio < 4; prio++) { | |
1380 | int h; | |
1381 | for (h = 0; h < HASH_SIZE; h++) { | |
1382 | struct ip_tunnel *t; | |
1383 | while ((t = tunnels[prio][h]) != NULL) | |
1384 | unregister_netdevice(t->dev); | |
1385 | } | |
1386 | } | |
1387 | } | |
1388 | ||
1389 | static void __exit ipgre_fini(void) | |
1da177e4 LT |
1390 | { |
1391 | if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) | |
1392 | printk(KERN_INFO "ipgre close: can't remove protocol\n"); | |
1393 | ||
db44575f AK |
1394 | rtnl_lock(); |
1395 | ipgre_destroy_tunnels(); | |
1396 | rtnl_unlock(); | |
59a4c759 PE |
1397 | |
1398 | unregister_pernet_gen_device(ipgre_net_id, &ipgre_net_ops); | |
1da177e4 LT |
1399 | } |
1400 | ||
1401 | module_init(ipgre_init); | |
1402 | module_exit(ipgre_fini); | |
1403 | MODULE_LICENSE("GPL"); |