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> | |
42 | ||
43 | #ifdef CONFIG_IPV6 | |
44 | #include <net/ipv6.h> | |
45 | #include <net/ip6_fib.h> | |
46 | #include <net/ip6_route.h> | |
47 | #endif | |
48 | ||
49 | /* | |
50 | Problems & solutions | |
51 | -------------------- | |
52 | ||
53 | 1. The most important issue is detecting local dead loops. | |
54 | They would cause complete host lockup in transmit, which | |
55 | would be "resolved" by stack overflow or, if queueing is enabled, | |
56 | with infinite looping in net_bh. | |
57 | ||
58 | We cannot track such dead loops during route installation, | |
59 | it is infeasible task. The most general solutions would be | |
60 | to keep skb->encapsulation counter (sort of local ttl), | |
61 | and silently drop packet when it expires. It is the best | |
62 | solution, but it supposes maintaing new variable in ALL | |
63 | skb, even if no tunneling is used. | |
64 | ||
e905a9ed | 65 | Current solution: t->recursion lock breaks dead loops. It looks |
1da177e4 LT |
66 | like dev->tbusy flag, but I preferred new variable, because |
67 | the semantics is different. One day, when hard_start_xmit | |
68 | will be multithreaded we will have to use skb->encapsulation. | |
69 | ||
70 | ||
71 | ||
72 | 2. Networking dead loops would not kill routers, but would really | |
73 | kill network. IP hop limit plays role of "t->recursion" in this case, | |
74 | if we copy it from packet being encapsulated to upper header. | |
75 | It is very good solution, but it introduces two problems: | |
76 | ||
77 | - Routing protocols, using packets with ttl=1 (OSPF, RIP2), | |
78 | do not work over tunnels. | |
79 | - traceroute does not work. I planned to relay ICMP from tunnel, | |
80 | so that this problem would be solved and traceroute output | |
81 | would even more informative. This idea appeared to be wrong: | |
82 | only Linux complies to rfc1812 now (yes, guys, Linux is the only | |
83 | true router now :-)), all routers (at least, in neighbourhood of mine) | |
84 | return only 8 bytes of payload. It is the end. | |
85 | ||
86 | Hence, if we want that OSPF worked or traceroute said something reasonable, | |
87 | we should search for another solution. | |
88 | ||
89 | One of them is to parse packet trying to detect inner encapsulation | |
90 | made by our node. It is difficult or even impossible, especially, | |
91 | taking into account fragmentation. TO be short, tt is not solution at all. | |
92 | ||
93 | Current solution: The solution was UNEXPECTEDLY SIMPLE. | |
94 | We force DF flag on tunnels with preconfigured hop limit, | |
95 | that is ALL. :-) Well, it does not remove the problem completely, | |
96 | but exponential growth of network traffic is changed to linear | |
97 | (branches, that exceed pmtu are pruned) and tunnel mtu | |
98 | fastly degrades to value <68, where looping stops. | |
99 | Yes, it is not good if there exists a router in the loop, | |
100 | which does not force DF, even when encapsulating packets have DF set. | |
101 | But it is not our problem! Nobody could accuse us, we made | |
102 | all that we could make. Even if it is your gated who injected | |
103 | fatal route to network, even if it were you who configured | |
104 | fatal static route: you are innocent. :-) | |
105 | ||
106 | ||
107 | ||
108 | 3. Really, ipv4/ipip.c, ipv4/ip_gre.c and ipv6/sit.c contain | |
109 | practically identical code. It would be good to glue them | |
110 | together, but it is not very evident, how to make them modular. | |
111 | sit is integral part of IPv6, ipip and gre are naturally modular. | |
112 | We could extract common parts (hash table, ioctl etc) | |
113 | to a separate module (ip_tunnel.c). | |
114 | ||
115 | Alexey Kuznetsov. | |
116 | */ | |
117 | ||
118 | static int ipgre_tunnel_init(struct net_device *dev); | |
119 | static void ipgre_tunnel_setup(struct net_device *dev); | |
120 | ||
121 | /* Fallback tunnel: no source, no destination, no key, no options */ | |
122 | ||
123 | static int ipgre_fb_tunnel_init(struct net_device *dev); | |
124 | ||
125 | static struct net_device *ipgre_fb_tunnel_dev; | |
126 | ||
127 | /* Tunnel hash table */ | |
128 | ||
129 | /* | |
130 | 4 hash tables: | |
131 | ||
132 | 3: (remote,local) | |
133 | 2: (remote,*) | |
134 | 1: (*,local) | |
135 | 0: (*,*) | |
136 | ||
137 | We require exact key match i.e. if a key is present in packet | |
138 | it will match only tunnel with the same key; if it is not present, | |
139 | it will match only keyless tunnel. | |
140 | ||
141 | All keysless packets, if not matched configured keyless tunnels | |
142 | will match fallback tunnel. | |
143 | */ | |
144 | ||
145 | #define HASH_SIZE 16 | |
d5a0a1e3 | 146 | #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) |
1da177e4 LT |
147 | |
148 | static struct ip_tunnel *tunnels[4][HASH_SIZE]; | |
149 | ||
150 | #define tunnels_r_l (tunnels[3]) | |
151 | #define tunnels_r (tunnels[2]) | |
152 | #define tunnels_l (tunnels[1]) | |
153 | #define tunnels_wc (tunnels[0]) | |
154 | ||
155 | static DEFINE_RWLOCK(ipgre_lock); | |
156 | ||
157 | /* Given src, dst and key, find appropriate for input tunnel. */ | |
158 | ||
d5a0a1e3 | 159 | static struct ip_tunnel * ipgre_tunnel_lookup(__be32 remote, __be32 local, __be32 key) |
1da177e4 LT |
160 | { |
161 | unsigned h0 = HASH(remote); | |
162 | unsigned h1 = HASH(key); | |
163 | struct ip_tunnel *t; | |
164 | ||
165 | for (t = tunnels_r_l[h0^h1]; t; t = t->next) { | |
166 | if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) { | |
167 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
168 | return t; | |
169 | } | |
170 | } | |
171 | for (t = tunnels_r[h0^h1]; t; t = t->next) { | |
172 | if (remote == t->parms.iph.daddr) { | |
173 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
174 | return t; | |
175 | } | |
176 | } | |
177 | for (t = tunnels_l[h1]; t; t = t->next) { | |
178 | if (local == t->parms.iph.saddr || | |
f97c1e0c JP |
179 | (local == t->parms.iph.daddr && |
180 | ipv4_is_multicast(local))) { | |
1da177e4 LT |
181 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) |
182 | return t; | |
183 | } | |
184 | } | |
185 | for (t = tunnels_wc[h1]; t; t = t->next) { | |
186 | if (t->parms.i_key == key && (t->dev->flags&IFF_UP)) | |
187 | return t; | |
188 | } | |
189 | ||
190 | if (ipgre_fb_tunnel_dev->flags&IFF_UP) | |
2941a486 | 191 | return netdev_priv(ipgre_fb_tunnel_dev); |
1da177e4 LT |
192 | return NULL; |
193 | } | |
194 | ||
5056a1ef | 195 | static struct ip_tunnel **__ipgre_bucket(struct ip_tunnel_parm *parms) |
1da177e4 | 196 | { |
5056a1ef YH |
197 | __be32 remote = parms->iph.daddr; |
198 | __be32 local = parms->iph.saddr; | |
199 | __be32 key = parms->i_key; | |
1da177e4 LT |
200 | unsigned h = HASH(key); |
201 | int prio = 0; | |
202 | ||
203 | if (local) | |
204 | prio |= 1; | |
f97c1e0c | 205 | if (remote && !ipv4_is_multicast(remote)) { |
1da177e4 LT |
206 | prio |= 2; |
207 | h ^= HASH(remote); | |
208 | } | |
209 | ||
210 | return &tunnels[prio][h]; | |
211 | } | |
212 | ||
5056a1ef YH |
213 | static inline struct ip_tunnel **ipgre_bucket(struct ip_tunnel *t) |
214 | { | |
215 | return __ipgre_bucket(&t->parms); | |
216 | } | |
217 | ||
1da177e4 LT |
218 | static void ipgre_tunnel_link(struct ip_tunnel *t) |
219 | { | |
220 | struct ip_tunnel **tp = ipgre_bucket(t); | |
221 | ||
222 | t->next = *tp; | |
223 | write_lock_bh(&ipgre_lock); | |
224 | *tp = t; | |
225 | write_unlock_bh(&ipgre_lock); | |
226 | } | |
227 | ||
228 | static void ipgre_tunnel_unlink(struct ip_tunnel *t) | |
229 | { | |
230 | struct ip_tunnel **tp; | |
231 | ||
232 | for (tp = ipgre_bucket(t); *tp; tp = &(*tp)->next) { | |
233 | if (t == *tp) { | |
234 | write_lock_bh(&ipgre_lock); | |
235 | *tp = t->next; | |
236 | write_unlock_bh(&ipgre_lock); | |
237 | break; | |
238 | } | |
239 | } | |
240 | } | |
241 | ||
242 | static struct ip_tunnel * ipgre_tunnel_locate(struct ip_tunnel_parm *parms, int create) | |
243 | { | |
d5a0a1e3 AV |
244 | __be32 remote = parms->iph.daddr; |
245 | __be32 local = parms->iph.saddr; | |
246 | __be32 key = parms->i_key; | |
1da177e4 LT |
247 | struct ip_tunnel *t, **tp, *nt; |
248 | struct net_device *dev; | |
1da177e4 LT |
249 | char name[IFNAMSIZ]; |
250 | ||
5056a1ef | 251 | for (tp = __ipgre_bucket(parms); (t = *tp) != NULL; tp = &t->next) { |
1da177e4 LT |
252 | if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) { |
253 | if (key == t->parms.i_key) | |
254 | return t; | |
255 | } | |
256 | } | |
257 | if (!create) | |
258 | return NULL; | |
259 | ||
260 | if (parms->name[0]) | |
261 | strlcpy(name, parms->name, IFNAMSIZ); | |
34cc7ba6 PE |
262 | else |
263 | sprintf(name, "gre%%d"); | |
1da177e4 LT |
264 | |
265 | dev = alloc_netdev(sizeof(*t), name, ipgre_tunnel_setup); | |
266 | if (!dev) | |
267 | return NULL; | |
268 | ||
269 | dev->init = ipgre_tunnel_init; | |
2941a486 | 270 | nt = netdev_priv(dev); |
1da177e4 LT |
271 | nt->parms = *parms; |
272 | ||
273 | if (register_netdevice(dev) < 0) { | |
274 | free_netdev(dev); | |
275 | goto failed; | |
276 | } | |
277 | ||
1da177e4 LT |
278 | dev_hold(dev); |
279 | ipgre_tunnel_link(nt); | |
1da177e4 LT |
280 | return nt; |
281 | ||
282 | failed: | |
283 | return NULL; | |
284 | } | |
285 | ||
286 | static void ipgre_tunnel_uninit(struct net_device *dev) | |
287 | { | |
2941a486 | 288 | ipgre_tunnel_unlink(netdev_priv(dev)); |
1da177e4 LT |
289 | dev_put(dev); |
290 | } | |
291 | ||
292 | ||
293 | static void ipgre_err(struct sk_buff *skb, u32 info) | |
294 | { | |
295 | #ifndef I_WISH_WORLD_WERE_PERFECT | |
296 | ||
297 | /* It is not :-( All the routers (except for Linux) return only | |
298 | 8 bytes of packet payload. It means, that precise relaying of | |
299 | ICMP in the real Internet is absolutely infeasible. | |
300 | ||
301 | Moreover, Cisco "wise men" put GRE key to the third word | |
302 | in GRE header. It makes impossible maintaining even soft state for keyed | |
303 | GRE tunnels with enabled checksum. Tell them "thank you". | |
304 | ||
305 | Well, I wonder, rfc1812 was written by Cisco employee, | |
306 | what the hell these idiots break standrads established | |
307 | by themself??? | |
308 | */ | |
309 | ||
310 | struct iphdr *iph = (struct iphdr*)skb->data; | |
d5a0a1e3 | 311 | __be16 *p = (__be16*)(skb->data+(iph->ihl<<2)); |
1da177e4 | 312 | int grehlen = (iph->ihl<<2) + 4; |
88c7664f ACM |
313 | const int type = icmp_hdr(skb)->type; |
314 | const int code = icmp_hdr(skb)->code; | |
1da177e4 | 315 | struct ip_tunnel *t; |
d5a0a1e3 | 316 | __be16 flags; |
1da177e4 LT |
317 | |
318 | flags = p[0]; | |
319 | if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { | |
320 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
321 | return; | |
322 | if (flags&GRE_KEY) { | |
323 | grehlen += 4; | |
324 | if (flags&GRE_CSUM) | |
325 | grehlen += 4; | |
326 | } | |
327 | } | |
328 | ||
329 | /* If only 8 bytes returned, keyed message will be dropped here */ | |
330 | if (skb_headlen(skb) < grehlen) | |
331 | return; | |
332 | ||
333 | switch (type) { | |
334 | default: | |
335 | case ICMP_PARAMETERPROB: | |
336 | return; | |
337 | ||
338 | case ICMP_DEST_UNREACH: | |
339 | switch (code) { | |
340 | case ICMP_SR_FAILED: | |
341 | case ICMP_PORT_UNREACH: | |
342 | /* Impossible event. */ | |
343 | return; | |
344 | case ICMP_FRAG_NEEDED: | |
345 | /* Soft state for pmtu is maintained by IP core. */ | |
346 | return; | |
347 | default: | |
348 | /* All others are translated to HOST_UNREACH. | |
349 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
350 | I believe they are just ether pollution. --ANK | |
351 | */ | |
352 | break; | |
353 | } | |
354 | break; | |
355 | case ICMP_TIME_EXCEEDED: | |
356 | if (code != ICMP_EXC_TTL) | |
357 | return; | |
358 | break; | |
359 | } | |
360 | ||
361 | read_lock(&ipgre_lock); | |
d5a0a1e3 | 362 | t = ipgre_tunnel_lookup(iph->daddr, iph->saddr, (flags&GRE_KEY) ? *(((__be32*)p) + (grehlen>>2) - 1) : 0); |
f97c1e0c JP |
363 | if (t == NULL || t->parms.iph.daddr == 0 || |
364 | ipv4_is_multicast(t->parms.iph.daddr)) | |
1da177e4 LT |
365 | goto out; |
366 | ||
367 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) | |
368 | goto out; | |
369 | ||
370 | if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO) | |
371 | t->err_count++; | |
372 | else | |
373 | t->err_count = 1; | |
374 | t->err_time = jiffies; | |
375 | out: | |
376 | read_unlock(&ipgre_lock); | |
377 | return; | |
378 | #else | |
379 | struct iphdr *iph = (struct iphdr*)dp; | |
380 | struct iphdr *eiph; | |
d5a0a1e3 | 381 | __be16 *p = (__be16*)(dp+(iph->ihl<<2)); |
88c7664f ACM |
382 | const int type = icmp_hdr(skb)->type; |
383 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
384 | int rel_type = 0; |
385 | int rel_code = 0; | |
c55e2f49 AV |
386 | __be32 rel_info = 0; |
387 | __u32 n = 0; | |
d5a0a1e3 | 388 | __be16 flags; |
1da177e4 LT |
389 | int grehlen = (iph->ihl<<2) + 4; |
390 | struct sk_buff *skb2; | |
391 | struct flowi fl; | |
392 | struct rtable *rt; | |
393 | ||
394 | if (p[1] != htons(ETH_P_IP)) | |
395 | return; | |
396 | ||
397 | flags = p[0]; | |
398 | if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { | |
399 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
400 | return; | |
401 | if (flags&GRE_CSUM) | |
402 | grehlen += 4; | |
403 | if (flags&GRE_KEY) | |
404 | grehlen += 4; | |
405 | if (flags&GRE_SEQ) | |
406 | grehlen += 4; | |
407 | } | |
408 | if (len < grehlen + sizeof(struct iphdr)) | |
409 | return; | |
410 | eiph = (struct iphdr*)(dp + grehlen); | |
411 | ||
412 | switch (type) { | |
413 | default: | |
414 | return; | |
415 | case ICMP_PARAMETERPROB: | |
88c7664f | 416 | n = ntohl(icmp_hdr(skb)->un.gateway) >> 24; |
c55e2f49 | 417 | if (n < (iph->ihl<<2)) |
1da177e4 LT |
418 | return; |
419 | ||
420 | /* So... This guy found something strange INSIDE encapsulated | |
421 | packet. Well, he is fool, but what can we do ? | |
422 | */ | |
423 | rel_type = ICMP_PARAMETERPROB; | |
c55e2f49 AV |
424 | n -= grehlen; |
425 | rel_info = htonl(n << 24); | |
1da177e4 LT |
426 | break; |
427 | ||
428 | case ICMP_DEST_UNREACH: | |
429 | switch (code) { | |
430 | case ICMP_SR_FAILED: | |
431 | case ICMP_PORT_UNREACH: | |
432 | /* Impossible event. */ | |
433 | return; | |
434 | case ICMP_FRAG_NEEDED: | |
435 | /* And it is the only really necessary thing :-) */ | |
88c7664f | 436 | n = ntohs(icmp_hdr(skb)->un.frag.mtu); |
c55e2f49 | 437 | if (n < grehlen+68) |
1da177e4 | 438 | return; |
c55e2f49 | 439 | n -= grehlen; |
1da177e4 | 440 | /* BSD 4.2 MORE DOES NOT EXIST IN NATURE. */ |
c55e2f49 | 441 | if (n > ntohs(eiph->tot_len)) |
1da177e4 | 442 | return; |
c55e2f49 | 443 | rel_info = htonl(n); |
1da177e4 LT |
444 | break; |
445 | default: | |
446 | /* All others are translated to HOST_UNREACH. | |
447 | rfc2003 contains "deep thoughts" about NET_UNREACH, | |
448 | I believe, it is just ether pollution. --ANK | |
449 | */ | |
450 | rel_type = ICMP_DEST_UNREACH; | |
451 | rel_code = ICMP_HOST_UNREACH; | |
452 | break; | |
453 | } | |
454 | break; | |
455 | case ICMP_TIME_EXCEEDED: | |
456 | if (code != ICMP_EXC_TTL) | |
457 | return; | |
458 | break; | |
459 | } | |
460 | ||
461 | /* Prepare fake skb to feed it to icmp_send */ | |
462 | skb2 = skb_clone(skb, GFP_ATOMIC); | |
463 | if (skb2 == NULL) | |
464 | return; | |
465 | dst_release(skb2->dst); | |
466 | skb2->dst = NULL; | |
467 | skb_pull(skb2, skb->data - (u8*)eiph); | |
c1d2bbe1 | 468 | skb_reset_network_header(skb2); |
1da177e4 LT |
469 | |
470 | /* Try to guess incoming interface */ | |
471 | memset(&fl, 0, sizeof(fl)); | |
472 | fl.fl4_dst = eiph->saddr; | |
473 | fl.fl4_tos = RT_TOS(eiph->tos); | |
474 | fl.proto = IPPROTO_GRE; | |
f206351a | 475 | if (ip_route_output_key(&init_net, &rt, &fl)) { |
1da177e4 LT |
476 | kfree_skb(skb2); |
477 | return; | |
478 | } | |
479 | skb2->dev = rt->u.dst.dev; | |
480 | ||
481 | /* route "incoming" packet */ | |
482 | if (rt->rt_flags&RTCF_LOCAL) { | |
483 | ip_rt_put(rt); | |
484 | rt = NULL; | |
485 | fl.fl4_dst = eiph->daddr; | |
486 | fl.fl4_src = eiph->saddr; | |
487 | fl.fl4_tos = eiph->tos; | |
f206351a | 488 | if (ip_route_output_key(&init_net, &rt, &fl) || |
1da177e4 LT |
489 | rt->u.dst.dev->type != ARPHRD_IPGRE) { |
490 | ip_rt_put(rt); | |
491 | kfree_skb(skb2); | |
492 | return; | |
493 | } | |
494 | } else { | |
495 | ip_rt_put(rt); | |
496 | if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, skb2->dev) || | |
497 | skb2->dst->dev->type != ARPHRD_IPGRE) { | |
498 | kfree_skb(skb2); | |
499 | return; | |
500 | } | |
501 | } | |
502 | ||
503 | /* change mtu on this route */ | |
504 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { | |
c55e2f49 | 505 | if (n > dst_mtu(skb2->dst)) { |
1da177e4 LT |
506 | kfree_skb(skb2); |
507 | return; | |
508 | } | |
c55e2f49 | 509 | skb2->dst->ops->update_pmtu(skb2->dst, n); |
1da177e4 | 510 | } else if (type == ICMP_TIME_EXCEEDED) { |
2941a486 | 511 | struct ip_tunnel *t = netdev_priv(skb2->dev); |
1da177e4 LT |
512 | if (t->parms.iph.ttl) { |
513 | rel_type = ICMP_DEST_UNREACH; | |
514 | rel_code = ICMP_HOST_UNREACH; | |
515 | } | |
516 | } | |
517 | ||
518 | icmp_send(skb2, rel_type, rel_code, rel_info); | |
519 | kfree_skb(skb2); | |
520 | #endif | |
521 | } | |
522 | ||
523 | static inline void ipgre_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb) | |
524 | { | |
525 | if (INET_ECN_is_ce(iph->tos)) { | |
526 | if (skb->protocol == htons(ETH_P_IP)) { | |
eddc9ec5 | 527 | IP_ECN_set_ce(ip_hdr(skb)); |
1da177e4 | 528 | } else if (skb->protocol == htons(ETH_P_IPV6)) { |
0660e03f | 529 | IP6_ECN_set_ce(ipv6_hdr(skb)); |
1da177e4 LT |
530 | } |
531 | } | |
532 | } | |
533 | ||
534 | static inline u8 | |
535 | ipgre_ecn_encapsulate(u8 tos, struct iphdr *old_iph, struct sk_buff *skb) | |
536 | { | |
537 | u8 inner = 0; | |
538 | if (skb->protocol == htons(ETH_P_IP)) | |
539 | inner = old_iph->tos; | |
540 | else if (skb->protocol == htons(ETH_P_IPV6)) | |
541 | inner = ipv6_get_dsfield((struct ipv6hdr *)old_iph); | |
542 | return INET_ECN_encapsulate(tos, inner); | |
543 | } | |
544 | ||
545 | static int ipgre_rcv(struct sk_buff *skb) | |
546 | { | |
547 | struct iphdr *iph; | |
548 | u8 *h; | |
d5a0a1e3 | 549 | __be16 flags; |
d3bc23e7 | 550 | __sum16 csum = 0; |
d5a0a1e3 | 551 | __be32 key = 0; |
1da177e4 LT |
552 | u32 seqno = 0; |
553 | struct ip_tunnel *tunnel; | |
554 | int offset = 4; | |
555 | ||
556 | if (!pskb_may_pull(skb, 16)) | |
557 | goto drop_nolock; | |
558 | ||
eddc9ec5 | 559 | iph = ip_hdr(skb); |
1da177e4 | 560 | h = skb->data; |
d5a0a1e3 | 561 | flags = *(__be16*)h; |
1da177e4 LT |
562 | |
563 | if (flags&(GRE_CSUM|GRE_KEY|GRE_ROUTING|GRE_SEQ|GRE_VERSION)) { | |
564 | /* - Version must be 0. | |
565 | - We do not support routing headers. | |
566 | */ | |
567 | if (flags&(GRE_VERSION|GRE_ROUTING)) | |
568 | goto drop_nolock; | |
569 | ||
570 | if (flags&GRE_CSUM) { | |
fb286bb2 | 571 | switch (skb->ip_summed) { |
84fa7933 | 572 | case CHECKSUM_COMPLETE: |
d3bc23e7 | 573 | csum = csum_fold(skb->csum); |
fb286bb2 HX |
574 | if (!csum) |
575 | break; | |
576 | /* fall through */ | |
577 | case CHECKSUM_NONE: | |
578 | skb->csum = 0; | |
579 | csum = __skb_checksum_complete(skb); | |
84fa7933 | 580 | skb->ip_summed = CHECKSUM_COMPLETE; |
1da177e4 LT |
581 | } |
582 | offset += 4; | |
583 | } | |
584 | if (flags&GRE_KEY) { | |
d5a0a1e3 | 585 | key = *(__be32*)(h + offset); |
1da177e4 LT |
586 | offset += 4; |
587 | } | |
588 | if (flags&GRE_SEQ) { | |
d5a0a1e3 | 589 | seqno = ntohl(*(__be32*)(h + offset)); |
1da177e4 LT |
590 | offset += 4; |
591 | } | |
592 | } | |
593 | ||
594 | read_lock(&ipgre_lock); | |
595 | if ((tunnel = ipgre_tunnel_lookup(iph->saddr, iph->daddr, key)) != NULL) { | |
596 | secpath_reset(skb); | |
597 | ||
d5a0a1e3 | 598 | skb->protocol = *(__be16*)(h + 2); |
1da177e4 LT |
599 | /* WCCP version 1 and 2 protocol decoding. |
600 | * - Change protocol to IP | |
601 | * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header | |
602 | */ | |
603 | if (flags == 0 && | |
496c98df YH |
604 | skb->protocol == htons(ETH_P_WCCP)) { |
605 | skb->protocol = htons(ETH_P_IP); | |
e905a9ed | 606 | if ((*(h + offset) & 0xF0) != 0x40) |
1da177e4 LT |
607 | offset += 4; |
608 | } | |
609 | ||
1d069167 | 610 | skb->mac_header = skb->network_header; |
4209fb60 ACM |
611 | __pskb_pull(skb, offset); |
612 | skb_reset_network_header(skb); | |
9c70220b | 613 | skb_postpull_rcsum(skb, skb_transport_header(skb), offset); |
1da177e4 LT |
614 | skb->pkt_type = PACKET_HOST; |
615 | #ifdef CONFIG_NET_IPGRE_BROADCAST | |
f97c1e0c | 616 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 LT |
617 | /* Looped back packet, drop it! */ |
618 | if (((struct rtable*)skb->dst)->fl.iif == 0) | |
619 | goto drop; | |
620 | tunnel->stat.multicast++; | |
621 | skb->pkt_type = PACKET_BROADCAST; | |
622 | } | |
623 | #endif | |
624 | ||
625 | if (((flags&GRE_CSUM) && csum) || | |
626 | (!(flags&GRE_CSUM) && tunnel->parms.i_flags&GRE_CSUM)) { | |
627 | tunnel->stat.rx_crc_errors++; | |
628 | tunnel->stat.rx_errors++; | |
629 | goto drop; | |
630 | } | |
631 | if (tunnel->parms.i_flags&GRE_SEQ) { | |
632 | if (!(flags&GRE_SEQ) || | |
633 | (tunnel->i_seqno && (s32)(seqno - tunnel->i_seqno) < 0)) { | |
634 | tunnel->stat.rx_fifo_errors++; | |
635 | tunnel->stat.rx_errors++; | |
636 | goto drop; | |
637 | } | |
638 | tunnel->i_seqno = seqno + 1; | |
639 | } | |
640 | tunnel->stat.rx_packets++; | |
641 | tunnel->stat.rx_bytes += skb->len; | |
642 | skb->dev = tunnel->dev; | |
643 | dst_release(skb->dst); | |
644 | skb->dst = NULL; | |
645 | nf_reset(skb); | |
646 | ipgre_ecn_decapsulate(iph, skb); | |
647 | netif_rx(skb); | |
648 | read_unlock(&ipgre_lock); | |
649 | return(0); | |
650 | } | |
45af08be | 651 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1da177e4 LT |
652 | |
653 | drop: | |
654 | read_unlock(&ipgre_lock); | |
655 | drop_nolock: | |
656 | kfree_skb(skb); | |
657 | return(0); | |
658 | } | |
659 | ||
660 | static int ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) | |
661 | { | |
2941a486 | 662 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 | 663 | struct net_device_stats *stats = &tunnel->stat; |
eddc9ec5 | 664 | struct iphdr *old_iph = ip_hdr(skb); |
1da177e4 LT |
665 | struct iphdr *tiph; |
666 | u8 tos; | |
d5a0a1e3 | 667 | __be16 df; |
1da177e4 LT |
668 | struct rtable *rt; /* Route to the other host */ |
669 | struct net_device *tdev; /* Device to other host */ | |
670 | struct iphdr *iph; /* Our new IP header */ | |
c2636b4d | 671 | unsigned int max_headroom; /* The extra header space needed */ |
1da177e4 | 672 | int gre_hlen; |
d5a0a1e3 | 673 | __be32 dst; |
1da177e4 LT |
674 | int mtu; |
675 | ||
676 | if (tunnel->recursion++) { | |
677 | tunnel->stat.collisions++; | |
678 | goto tx_error; | |
679 | } | |
680 | ||
3b04ddde | 681 | if (dev->header_ops) { |
1da177e4 LT |
682 | gre_hlen = 0; |
683 | tiph = (struct iphdr*)skb->data; | |
684 | } else { | |
685 | gre_hlen = tunnel->hlen; | |
686 | tiph = &tunnel->parms.iph; | |
687 | } | |
688 | ||
689 | if ((dst = tiph->daddr) == 0) { | |
690 | /* NBMA tunnel */ | |
691 | ||
692 | if (skb->dst == NULL) { | |
693 | tunnel->stat.tx_fifo_errors++; | |
694 | goto tx_error; | |
695 | } | |
696 | ||
697 | if (skb->protocol == htons(ETH_P_IP)) { | |
698 | rt = (struct rtable*)skb->dst; | |
699 | if ((dst = rt->rt_gateway) == 0) | |
700 | goto tx_error_icmp; | |
701 | } | |
702 | #ifdef CONFIG_IPV6 | |
703 | else if (skb->protocol == htons(ETH_P_IPV6)) { | |
704 | struct in6_addr *addr6; | |
705 | int addr_type; | |
706 | struct neighbour *neigh = skb->dst->neighbour; | |
707 | ||
708 | if (neigh == NULL) | |
709 | goto tx_error; | |
710 | ||
711 | addr6 = (struct in6_addr*)&neigh->primary_key; | |
712 | addr_type = ipv6_addr_type(addr6); | |
713 | ||
714 | if (addr_type == IPV6_ADDR_ANY) { | |
0660e03f | 715 | addr6 = &ipv6_hdr(skb)->daddr; |
1da177e4 LT |
716 | addr_type = ipv6_addr_type(addr6); |
717 | } | |
718 | ||
719 | if ((addr_type & IPV6_ADDR_COMPATv4) == 0) | |
720 | goto tx_error_icmp; | |
721 | ||
722 | dst = addr6->s6_addr32[3]; | |
723 | } | |
724 | #endif | |
725 | else | |
726 | goto tx_error; | |
727 | } | |
728 | ||
729 | tos = tiph->tos; | |
730 | if (tos&1) { | |
731 | if (skb->protocol == htons(ETH_P_IP)) | |
732 | tos = old_iph->tos; | |
733 | tos &= ~1; | |
734 | } | |
735 | ||
736 | { | |
737 | struct flowi fl = { .oif = tunnel->parms.link, | |
738 | .nl_u = { .ip4_u = | |
739 | { .daddr = dst, | |
740 | .saddr = tiph->saddr, | |
741 | .tos = RT_TOS(tos) } }, | |
742 | .proto = IPPROTO_GRE }; | |
f206351a | 743 | if (ip_route_output_key(&init_net, &rt, &fl)) { |
1da177e4 LT |
744 | tunnel->stat.tx_carrier_errors++; |
745 | goto tx_error; | |
746 | } | |
747 | } | |
748 | tdev = rt->u.dst.dev; | |
749 | ||
750 | if (tdev == dev) { | |
751 | ip_rt_put(rt); | |
752 | tunnel->stat.collisions++; | |
753 | goto tx_error; | |
754 | } | |
755 | ||
756 | df = tiph->frag_off; | |
757 | if (df) | |
758 | mtu = dst_mtu(&rt->u.dst) - tunnel->hlen; | |
759 | else | |
760 | mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu; | |
761 | ||
762 | if (skb->dst) | |
763 | skb->dst->ops->update_pmtu(skb->dst, mtu); | |
764 | ||
765 | if (skb->protocol == htons(ETH_P_IP)) { | |
766 | df |= (old_iph->frag_off&htons(IP_DF)); | |
767 | ||
768 | if ((old_iph->frag_off&htons(IP_DF)) && | |
769 | mtu < ntohs(old_iph->tot_len)) { | |
770 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu)); | |
771 | ip_rt_put(rt); | |
772 | goto tx_error; | |
773 | } | |
774 | } | |
775 | #ifdef CONFIG_IPV6 | |
776 | else if (skb->protocol == htons(ETH_P_IPV6)) { | |
777 | struct rt6_info *rt6 = (struct rt6_info*)skb->dst; | |
778 | ||
779 | if (rt6 && mtu < dst_mtu(skb->dst) && mtu >= IPV6_MIN_MTU) { | |
f97c1e0c JP |
780 | if ((tunnel->parms.iph.daddr && |
781 | !ipv4_is_multicast(tunnel->parms.iph.daddr)) || | |
1da177e4 LT |
782 | rt6->rt6i_dst.plen == 128) { |
783 | rt6->rt6i_flags |= RTF_MODIFIED; | |
784 | skb->dst->metrics[RTAX_MTU-1] = mtu; | |
785 | } | |
786 | } | |
787 | ||
788 | if (mtu >= IPV6_MIN_MTU && mtu < skb->len - tunnel->hlen + gre_hlen) { | |
789 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); | |
790 | ip_rt_put(rt); | |
791 | goto tx_error; | |
792 | } | |
793 | } | |
794 | #endif | |
795 | ||
796 | if (tunnel->err_count > 0) { | |
797 | if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) { | |
798 | tunnel->err_count--; | |
799 | ||
800 | dst_link_failure(skb); | |
801 | } else | |
802 | tunnel->err_count = 0; | |
803 | } | |
804 | ||
805 | max_headroom = LL_RESERVED_SPACE(tdev) + gre_hlen; | |
806 | ||
cfbba49d PM |
807 | if (skb_headroom(skb) < max_headroom || skb_shared(skb)|| |
808 | (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { | |
1da177e4 LT |
809 | struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); |
810 | if (!new_skb) { | |
811 | ip_rt_put(rt); | |
e905a9ed | 812 | stats->tx_dropped++; |
1da177e4 LT |
813 | dev_kfree_skb(skb); |
814 | tunnel->recursion--; | |
815 | return 0; | |
816 | } | |
817 | if (skb->sk) | |
818 | skb_set_owner_w(new_skb, skb->sk); | |
819 | dev_kfree_skb(skb); | |
820 | skb = new_skb; | |
eddc9ec5 | 821 | old_iph = ip_hdr(skb); |
1da177e4 LT |
822 | } |
823 | ||
b0e380b1 | 824 | skb->transport_header = skb->network_header; |
e2d1bca7 ACM |
825 | skb_push(skb, gre_hlen); |
826 | skb_reset_network_header(skb); | |
1da177e4 | 827 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
48d5cad8 PM |
828 | IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | |
829 | IPSKB_REROUTED); | |
1da177e4 LT |
830 | dst_release(skb->dst); |
831 | skb->dst = &rt->u.dst; | |
832 | ||
833 | /* | |
834 | * Push down and install the IPIP header. | |
835 | */ | |
836 | ||
eddc9ec5 | 837 | iph = ip_hdr(skb); |
1da177e4 LT |
838 | iph->version = 4; |
839 | iph->ihl = sizeof(struct iphdr) >> 2; | |
840 | iph->frag_off = df; | |
841 | iph->protocol = IPPROTO_GRE; | |
842 | iph->tos = ipgre_ecn_encapsulate(tos, old_iph, skb); | |
843 | iph->daddr = rt->rt_dst; | |
844 | iph->saddr = rt->rt_src; | |
845 | ||
846 | if ((iph->ttl = tiph->ttl) == 0) { | |
847 | if (skb->protocol == htons(ETH_P_IP)) | |
848 | iph->ttl = old_iph->ttl; | |
849 | #ifdef CONFIG_IPV6 | |
850 | else if (skb->protocol == htons(ETH_P_IPV6)) | |
851 | iph->ttl = ((struct ipv6hdr*)old_iph)->hop_limit; | |
852 | #endif | |
853 | else | |
854 | iph->ttl = dst_metric(&rt->u.dst, RTAX_HOPLIMIT); | |
855 | } | |
856 | ||
d5a0a1e3 AV |
857 | ((__be16*)(iph+1))[0] = tunnel->parms.o_flags; |
858 | ((__be16*)(iph+1))[1] = skb->protocol; | |
1da177e4 LT |
859 | |
860 | if (tunnel->parms.o_flags&(GRE_KEY|GRE_CSUM|GRE_SEQ)) { | |
d5a0a1e3 | 861 | __be32 *ptr = (__be32*)(((u8*)iph) + tunnel->hlen - 4); |
1da177e4 LT |
862 | |
863 | if (tunnel->parms.o_flags&GRE_SEQ) { | |
864 | ++tunnel->o_seqno; | |
865 | *ptr = htonl(tunnel->o_seqno); | |
866 | ptr--; | |
867 | } | |
868 | if (tunnel->parms.o_flags&GRE_KEY) { | |
869 | *ptr = tunnel->parms.o_key; | |
870 | ptr--; | |
871 | } | |
872 | if (tunnel->parms.o_flags&GRE_CSUM) { | |
873 | *ptr = 0; | |
5f92a738 | 874 | *(__sum16*)ptr = ip_compute_csum((void*)(iph+1), skb->len - sizeof(struct iphdr)); |
1da177e4 LT |
875 | } |
876 | } | |
877 | ||
878 | nf_reset(skb); | |
879 | ||
880 | IPTUNNEL_XMIT(); | |
881 | tunnel->recursion--; | |
882 | return 0; | |
883 | ||
884 | tx_error_icmp: | |
885 | dst_link_failure(skb); | |
886 | ||
887 | tx_error: | |
888 | stats->tx_errors++; | |
889 | dev_kfree_skb(skb); | |
890 | tunnel->recursion--; | |
891 | return 0; | |
892 | } | |
893 | ||
ee34c1eb MS |
894 | static void ipgre_tunnel_bind_dev(struct net_device *dev) |
895 | { | |
896 | struct net_device *tdev = NULL; | |
897 | struct ip_tunnel *tunnel; | |
898 | struct iphdr *iph; | |
899 | int hlen = LL_MAX_HEADER; | |
900 | int mtu = ETH_DATA_LEN; | |
901 | int addend = sizeof(struct iphdr) + 4; | |
902 | ||
903 | tunnel = netdev_priv(dev); | |
904 | iph = &tunnel->parms.iph; | |
905 | ||
906 | /* Guess output device to choose reasonable mtu and hard_header_len */ | |
907 | ||
908 | if (iph->daddr) { | |
909 | struct flowi fl = { .oif = tunnel->parms.link, | |
910 | .nl_u = { .ip4_u = | |
911 | { .daddr = iph->daddr, | |
912 | .saddr = iph->saddr, | |
913 | .tos = RT_TOS(iph->tos) } }, | |
914 | .proto = IPPROTO_GRE }; | |
915 | struct rtable *rt; | |
f206351a | 916 | if (!ip_route_output_key(&init_net, &rt, &fl)) { |
ee34c1eb MS |
917 | tdev = rt->u.dst.dev; |
918 | ip_rt_put(rt); | |
919 | } | |
920 | dev->flags |= IFF_POINTOPOINT; | |
921 | } | |
922 | ||
923 | if (!tdev && tunnel->parms.link) | |
924 | tdev = __dev_get_by_index(&init_net, tunnel->parms.link); | |
925 | ||
926 | if (tdev) { | |
927 | hlen = tdev->hard_header_len; | |
928 | mtu = tdev->mtu; | |
929 | } | |
930 | dev->iflink = tunnel->parms.link; | |
931 | ||
932 | /* Precalculate GRE options length */ | |
933 | if (tunnel->parms.o_flags&(GRE_CSUM|GRE_KEY|GRE_SEQ)) { | |
934 | if (tunnel->parms.o_flags&GRE_CSUM) | |
935 | addend += 4; | |
936 | if (tunnel->parms.o_flags&GRE_KEY) | |
937 | addend += 4; | |
938 | if (tunnel->parms.o_flags&GRE_SEQ) | |
939 | addend += 4; | |
940 | } | |
941 | dev->hard_header_len = hlen + addend; | |
942 | dev->mtu = mtu - addend; | |
943 | tunnel->hlen = addend; | |
944 | ||
945 | } | |
946 | ||
1da177e4 LT |
947 | static int |
948 | ipgre_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) | |
949 | { | |
950 | int err = 0; | |
951 | struct ip_tunnel_parm p; | |
952 | struct ip_tunnel *t; | |
953 | ||
954 | switch (cmd) { | |
955 | case SIOCGETTUNNEL: | |
956 | t = NULL; | |
957 | if (dev == ipgre_fb_tunnel_dev) { | |
958 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { | |
959 | err = -EFAULT; | |
960 | break; | |
961 | } | |
962 | t = ipgre_tunnel_locate(&p, 0); | |
963 | } | |
964 | if (t == NULL) | |
2941a486 | 965 | t = netdev_priv(dev); |
1da177e4 LT |
966 | memcpy(&p, &t->parms, sizeof(p)); |
967 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) | |
968 | err = -EFAULT; | |
969 | break; | |
970 | ||
971 | case SIOCADDTUNNEL: | |
972 | case SIOCCHGTUNNEL: | |
973 | err = -EPERM; | |
974 | if (!capable(CAP_NET_ADMIN)) | |
975 | goto done; | |
976 | ||
977 | err = -EFAULT; | |
978 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
979 | goto done; | |
980 | ||
981 | err = -EINVAL; | |
982 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE || | |
983 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) || | |
984 | ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) | |
985 | goto done; | |
986 | if (p.iph.ttl) | |
987 | p.iph.frag_off |= htons(IP_DF); | |
988 | ||
989 | if (!(p.i_flags&GRE_KEY)) | |
990 | p.i_key = 0; | |
991 | if (!(p.o_flags&GRE_KEY)) | |
992 | p.o_key = 0; | |
993 | ||
994 | t = ipgre_tunnel_locate(&p, cmd == SIOCADDTUNNEL); | |
995 | ||
996 | if (dev != ipgre_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { | |
997 | if (t != NULL) { | |
998 | if (t->dev != dev) { | |
999 | err = -EEXIST; | |
1000 | break; | |
1001 | } | |
1002 | } else { | |
1003 | unsigned nflags=0; | |
1004 | ||
2941a486 | 1005 | t = netdev_priv(dev); |
1da177e4 | 1006 | |
f97c1e0c | 1007 | if (ipv4_is_multicast(p.iph.daddr)) |
1da177e4 LT |
1008 | nflags = IFF_BROADCAST; |
1009 | else if (p.iph.daddr) | |
1010 | nflags = IFF_POINTOPOINT; | |
1011 | ||
1012 | if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) { | |
1013 | err = -EINVAL; | |
1014 | break; | |
1015 | } | |
1016 | ipgre_tunnel_unlink(t); | |
1017 | t->parms.iph.saddr = p.iph.saddr; | |
1018 | t->parms.iph.daddr = p.iph.daddr; | |
1019 | t->parms.i_key = p.i_key; | |
1020 | t->parms.o_key = p.o_key; | |
1021 | memcpy(dev->dev_addr, &p.iph.saddr, 4); | |
1022 | memcpy(dev->broadcast, &p.iph.daddr, 4); | |
1023 | ipgre_tunnel_link(t); | |
1024 | netdev_state_change(dev); | |
1025 | } | |
1026 | } | |
1027 | ||
1028 | if (t) { | |
1029 | err = 0; | |
1030 | if (cmd == SIOCCHGTUNNEL) { | |
1031 | t->parms.iph.ttl = p.iph.ttl; | |
1032 | t->parms.iph.tos = p.iph.tos; | |
1033 | t->parms.iph.frag_off = p.iph.frag_off; | |
ee34c1eb MS |
1034 | if (t->parms.link != p.link) { |
1035 | t->parms.link = p.link; | |
1036 | ipgre_tunnel_bind_dev(dev); | |
1037 | netdev_state_change(dev); | |
1038 | } | |
1da177e4 LT |
1039 | } |
1040 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) | |
1041 | err = -EFAULT; | |
1042 | } else | |
1043 | err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); | |
1044 | break; | |
1045 | ||
1046 | case SIOCDELTUNNEL: | |
1047 | err = -EPERM; | |
1048 | if (!capable(CAP_NET_ADMIN)) | |
1049 | goto done; | |
1050 | ||
1051 | if (dev == ipgre_fb_tunnel_dev) { | |
1052 | err = -EFAULT; | |
1053 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) | |
1054 | goto done; | |
1055 | err = -ENOENT; | |
1056 | if ((t = ipgre_tunnel_locate(&p, 0)) == NULL) | |
1057 | goto done; | |
1058 | err = -EPERM; | |
2941a486 | 1059 | if (t == netdev_priv(ipgre_fb_tunnel_dev)) |
1da177e4 LT |
1060 | goto done; |
1061 | dev = t->dev; | |
1062 | } | |
22f8cde5 SH |
1063 | unregister_netdevice(dev); |
1064 | err = 0; | |
1da177e4 LT |
1065 | break; |
1066 | ||
1067 | default: | |
1068 | err = -EINVAL; | |
1069 | } | |
1070 | ||
1071 | done: | |
1072 | return err; | |
1073 | } | |
1074 | ||
1075 | static struct net_device_stats *ipgre_tunnel_get_stats(struct net_device *dev) | |
1076 | { | |
2941a486 | 1077 | return &(((struct ip_tunnel*)netdev_priv(dev))->stat); |
1da177e4 LT |
1078 | } |
1079 | ||
1080 | static int ipgre_tunnel_change_mtu(struct net_device *dev, int new_mtu) | |
1081 | { | |
2941a486 | 1082 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 LT |
1083 | if (new_mtu < 68 || new_mtu > 0xFFF8 - tunnel->hlen) |
1084 | return -EINVAL; | |
1085 | dev->mtu = new_mtu; | |
1086 | return 0; | |
1087 | } | |
1088 | ||
1da177e4 LT |
1089 | /* Nice toy. Unfortunately, useless in real life :-) |
1090 | It allows to construct virtual multiprotocol broadcast "LAN" | |
1091 | over the Internet, provided multicast routing is tuned. | |
1092 | ||
1093 | ||
1094 | I have no idea was this bicycle invented before me, | |
1095 | so that I had to set ARPHRD_IPGRE to a random value. | |
1096 | I have an impression, that Cisco could make something similar, | |
1097 | but this feature is apparently missing in IOS<=11.2(8). | |
e905a9ed | 1098 | |
1da177e4 LT |
1099 | I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks |
1100 | with broadcast 224.66.66.66. If you have access to mbone, play with me :-) | |
1101 | ||
1102 | ping -t 255 224.66.66.66 | |
1103 | ||
1104 | If nobody answers, mbone does not work. | |
1105 | ||
1106 | ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255 | |
1107 | ip addr add 10.66.66.<somewhat>/24 dev Universe | |
1108 | ifconfig Universe up | |
1109 | ifconfig Universe add fe80::<Your_real_addr>/10 | |
1110 | ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96 | |
1111 | ftp 10.66.66.66 | |
1112 | ... | |
1113 | ftp fec0:6666:6666::193.233.7.65 | |
1114 | ... | |
1115 | ||
1116 | */ | |
1117 | ||
3b04ddde SH |
1118 | static int ipgre_header(struct sk_buff *skb, struct net_device *dev, |
1119 | unsigned short type, | |
1120 | const void *daddr, const void *saddr, unsigned len) | |
1da177e4 | 1121 | { |
2941a486 | 1122 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 1123 | struct iphdr *iph = (struct iphdr *)skb_push(skb, t->hlen); |
d5a0a1e3 | 1124 | __be16 *p = (__be16*)(iph+1); |
1da177e4 LT |
1125 | |
1126 | memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); | |
1127 | p[0] = t->parms.o_flags; | |
1128 | p[1] = htons(type); | |
1129 | ||
1130 | /* | |
e905a9ed | 1131 | * Set the source hardware address. |
1da177e4 | 1132 | */ |
e905a9ed | 1133 | |
1da177e4 LT |
1134 | if (saddr) |
1135 | memcpy(&iph->saddr, saddr, 4); | |
1136 | ||
1137 | if (daddr) { | |
1138 | memcpy(&iph->daddr, daddr, 4); | |
1139 | return t->hlen; | |
1140 | } | |
f97c1e0c | 1141 | if (iph->daddr && !ipv4_is_multicast(iph->daddr)) |
1da177e4 | 1142 | return t->hlen; |
e905a9ed | 1143 | |
1da177e4 LT |
1144 | return -t->hlen; |
1145 | } | |
1146 | ||
6a5f44d7 TT |
1147 | static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) |
1148 | { | |
1149 | struct iphdr *iph = (struct iphdr*) skb_mac_header(skb); | |
1150 | memcpy(haddr, &iph->saddr, 4); | |
1151 | return 4; | |
1152 | } | |
1153 | ||
3b04ddde SH |
1154 | static const struct header_ops ipgre_header_ops = { |
1155 | .create = ipgre_header, | |
6a5f44d7 | 1156 | .parse = ipgre_header_parse, |
3b04ddde SH |
1157 | }; |
1158 | ||
6a5f44d7 | 1159 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
1da177e4 LT |
1160 | static int ipgre_open(struct net_device *dev) |
1161 | { | |
2941a486 | 1162 | struct ip_tunnel *t = netdev_priv(dev); |
1da177e4 | 1163 | |
f97c1e0c | 1164 | if (ipv4_is_multicast(t->parms.iph.daddr)) { |
1da177e4 LT |
1165 | struct flowi fl = { .oif = t->parms.link, |
1166 | .nl_u = { .ip4_u = | |
1167 | { .daddr = t->parms.iph.daddr, | |
1168 | .saddr = t->parms.iph.saddr, | |
1169 | .tos = RT_TOS(t->parms.iph.tos) } }, | |
1170 | .proto = IPPROTO_GRE }; | |
1171 | struct rtable *rt; | |
f206351a | 1172 | if (ip_route_output_key(&init_net, &rt, &fl)) |
1da177e4 LT |
1173 | return -EADDRNOTAVAIL; |
1174 | dev = rt->u.dst.dev; | |
1175 | ip_rt_put(rt); | |
e5ed6399 | 1176 | if (__in_dev_get_rtnl(dev) == NULL) |
1da177e4 LT |
1177 | return -EADDRNOTAVAIL; |
1178 | t->mlink = dev->ifindex; | |
e5ed6399 | 1179 | ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr); |
1da177e4 LT |
1180 | } |
1181 | return 0; | |
1182 | } | |
1183 | ||
1184 | static int ipgre_close(struct net_device *dev) | |
1185 | { | |
2941a486 | 1186 | struct ip_tunnel *t = netdev_priv(dev); |
f97c1e0c | 1187 | if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) { |
7fee0ca2 DL |
1188 | struct in_device *in_dev; |
1189 | in_dev = inetdev_by_index(dev->nd_net, t->mlink); | |
1da177e4 LT |
1190 | if (in_dev) { |
1191 | ip_mc_dec_group(in_dev, t->parms.iph.daddr); | |
1192 | in_dev_put(in_dev); | |
1193 | } | |
1194 | } | |
1195 | return 0; | |
1196 | } | |
1197 | ||
1198 | #endif | |
1199 | ||
1200 | static void ipgre_tunnel_setup(struct net_device *dev) | |
1201 | { | |
1da177e4 LT |
1202 | dev->uninit = ipgre_tunnel_uninit; |
1203 | dev->destructor = free_netdev; | |
1204 | dev->hard_start_xmit = ipgre_tunnel_xmit; | |
1205 | dev->get_stats = ipgre_tunnel_get_stats; | |
1206 | dev->do_ioctl = ipgre_tunnel_ioctl; | |
1207 | dev->change_mtu = ipgre_tunnel_change_mtu; | |
1208 | ||
1209 | dev->type = ARPHRD_IPGRE; | |
1210 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr) + 4; | |
46f25dff | 1211 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4; |
1da177e4 LT |
1212 | dev->flags = IFF_NOARP; |
1213 | dev->iflink = 0; | |
1214 | dev->addr_len = 4; | |
1215 | } | |
1216 | ||
1217 | static int ipgre_tunnel_init(struct net_device *dev) | |
1218 | { | |
1da177e4 LT |
1219 | struct ip_tunnel *tunnel; |
1220 | struct iphdr *iph; | |
1da177e4 | 1221 | |
2941a486 | 1222 | tunnel = netdev_priv(dev); |
1da177e4 LT |
1223 | iph = &tunnel->parms.iph; |
1224 | ||
1225 | tunnel->dev = dev; | |
1226 | strcpy(tunnel->parms.name, dev->name); | |
1227 | ||
1228 | memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); | |
1229 | memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); | |
1230 | ||
ee34c1eb | 1231 | ipgre_tunnel_bind_dev(dev); |
1da177e4 LT |
1232 | |
1233 | if (iph->daddr) { | |
1da177e4 | 1234 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
f97c1e0c | 1235 | if (ipv4_is_multicast(iph->daddr)) { |
1da177e4 LT |
1236 | if (!iph->saddr) |
1237 | return -EINVAL; | |
1238 | dev->flags = IFF_BROADCAST; | |
3b04ddde | 1239 | dev->header_ops = &ipgre_header_ops; |
1da177e4 LT |
1240 | dev->open = ipgre_open; |
1241 | dev->stop = ipgre_close; | |
1242 | } | |
1243 | #endif | |
ee34c1eb | 1244 | } else |
6a5f44d7 | 1245 | dev->header_ops = &ipgre_header_ops; |
1da177e4 | 1246 | |
1da177e4 LT |
1247 | return 0; |
1248 | } | |
1249 | ||
4b30b1c6 | 1250 | static int __init ipgre_fb_tunnel_init(struct net_device *dev) |
1da177e4 | 1251 | { |
2941a486 | 1252 | struct ip_tunnel *tunnel = netdev_priv(dev); |
1da177e4 LT |
1253 | struct iphdr *iph = &tunnel->parms.iph; |
1254 | ||
1255 | tunnel->dev = dev; | |
1256 | strcpy(tunnel->parms.name, dev->name); | |
1257 | ||
1258 | iph->version = 4; | |
1259 | iph->protocol = IPPROTO_GRE; | |
1260 | iph->ihl = 5; | |
1261 | tunnel->hlen = sizeof(struct iphdr) + 4; | |
1262 | ||
1263 | dev_hold(dev); | |
1264 | tunnels_wc[0] = tunnel; | |
1265 | return 0; | |
1266 | } | |
1267 | ||
1268 | ||
1269 | static struct net_protocol ipgre_protocol = { | |
1270 | .handler = ipgre_rcv, | |
1271 | .err_handler = ipgre_err, | |
1272 | }; | |
1273 | ||
1274 | ||
1275 | /* | |
1276 | * And now the modules code and kernel interface. | |
1277 | */ | |
1278 | ||
1279 | static int __init ipgre_init(void) | |
1280 | { | |
1281 | int err; | |
1282 | ||
1283 | printk(KERN_INFO "GRE over IPv4 tunneling driver\n"); | |
1284 | ||
1285 | if (inet_add_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) { | |
1286 | printk(KERN_INFO "ipgre init: can't add protocol\n"); | |
1287 | return -EAGAIN; | |
1288 | } | |
1289 | ||
1290 | ipgre_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "gre0", | |
1291 | ipgre_tunnel_setup); | |
1292 | if (!ipgre_fb_tunnel_dev) { | |
1293 | err = -ENOMEM; | |
1294 | goto err1; | |
1295 | } | |
1296 | ||
1297 | ipgre_fb_tunnel_dev->init = ipgre_fb_tunnel_init; | |
1298 | ||
1299 | if ((err = register_netdev(ipgre_fb_tunnel_dev))) | |
1300 | goto err2; | |
1301 | out: | |
1302 | return err; | |
1303 | err2: | |
1304 | free_netdev(ipgre_fb_tunnel_dev); | |
1305 | err1: | |
1306 | inet_del_protocol(&ipgre_protocol, IPPROTO_GRE); | |
1307 | goto out; | |
1308 | } | |
1309 | ||
db44575f AK |
1310 | static void __exit ipgre_destroy_tunnels(void) |
1311 | { | |
1312 | int prio; | |
1313 | ||
1314 | for (prio = 0; prio < 4; prio++) { | |
1315 | int h; | |
1316 | for (h = 0; h < HASH_SIZE; h++) { | |
1317 | struct ip_tunnel *t; | |
1318 | while ((t = tunnels[prio][h]) != NULL) | |
1319 | unregister_netdevice(t->dev); | |
1320 | } | |
1321 | } | |
1322 | } | |
1323 | ||
1324 | static void __exit ipgre_fini(void) | |
1da177e4 LT |
1325 | { |
1326 | if (inet_del_protocol(&ipgre_protocol, IPPROTO_GRE) < 0) | |
1327 | printk(KERN_INFO "ipgre close: can't remove protocol\n"); | |
1328 | ||
db44575f AK |
1329 | rtnl_lock(); |
1330 | ipgre_destroy_tunnels(); | |
1331 | rtnl_unlock(); | |
1da177e4 LT |
1332 | } |
1333 | ||
1334 | module_init(ipgre_init); | |
1335 | module_exit(ipgre_fini); | |
1336 | MODULE_LICENSE("GPL"); |