blob: e5cbe72c6b80bd0e858fb3af17a674e315e3a517 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Linux NET3: IP/IP protocol decoder.
3 *
4 * Version: $Id: ipip.c,v 1.50 2001/10/02 02:22:36 davem Exp $
5 *
6 * Authors:
7 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
8 *
9 * Fixes:
10 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
11 * a module taking up 2 pages).
12 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
13 * to keep ip_forward happy.
14 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
15 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
16 * David Woodhouse : Perform some basic ICMP handling.
17 * IPIP Routing without decapsulation.
18 * Carlos Picoto : GRE over IP support
19 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
20 * I do not want to merge them together.
21 *
22 * This program is free software; you can redistribute it and/or
23 * modify it under the terms of the GNU General Public License
24 * as published by the Free Software Foundation; either version
25 * 2 of the License, or (at your option) any later version.
26 *
27 */
28
29/* tunnel.c: an IP tunnel driver
30
31 The purpose of this driver is to provide an IP tunnel through
32 which you can tunnel network traffic transparently across subnets.
33
34 This was written by looking at Nick Holloway's dummy driver
35 Thanks for the great code!
36
37 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
38
39 Minor tweaks:
40 Cleaned up the code a little and added some pre-1.3.0 tweaks.
41 dev->hard_header/hard_header_len changed to use no headers.
42 Comments/bracketing tweaked.
43 Made the tunnels use dev->name not tunnel: when error reporting.
44 Added tx_dropped stat
45
46 -Alan Cox (Alan.Cox@linux.org) 21 March 95
47
48 Reworked:
49 Changed to tunnel to destination gateway in addition to the
50 tunnel's pointopoint address
51 Almost completely rewritten
52 Note: There is currently no firewall or ICMP handling done.
53
54 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
55
56*/
57
58/* Things I wish I had known when writing the tunnel driver:
59
60 When the tunnel_xmit() function is called, the skb contains the
61 packet to be sent (plus a great deal of extra info), and dev
62 contains the tunnel device that _we_ are.
63
64 When we are passed a packet, we are expected to fill in the
65 source address with our source IP address.
66
67 What is the proper way to allocate, copy and free a buffer?
68 After you allocate it, it is a "0 length" chunk of memory
69 starting at zero. If you want to add headers to the buffer
70 later, you'll have to call "skb_reserve(skb, amount)" with
71 the amount of memory you want reserved. Then, you call
72 "skb_put(skb, amount)" with the amount of space you want in
73 the buffer. skb_put() returns a pointer to the top (#0) of
74 that buffer. skb->len is set to the amount of space you have
75 "allocated" with skb_put(). You can then write up to skb->len
76 bytes to that buffer. If you need more, you can call skb_put()
77 again with the additional amount of space you need. You can
78 find out how much more space you can allocate by calling
79 "skb_tailroom(skb)".
80 Now, to add header space, call "skb_push(skb, header_len)".
81 This creates space at the beginning of the buffer and returns
82 a pointer to this new space. If later you need to strip a
83 header from a buffer, call "skb_pull(skb, header_len)".
84 skb_headroom() will return how much space is left at the top
85 of the buffer (before the main data). Remember, this headroom
86 space must be reserved before the skb_put() function is called.
87 */
88
89/*
90 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
91
92 For comments look at net/ipv4/ip_gre.c --ANK
93 */
94
95
Randy Dunlap4fc268d2006-01-11 12:17:47 -080096#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/config.h>
98#include <linux/module.h>
99#include <linux/types.h>
100#include <linux/sched.h>
101#include <linux/kernel.h>
102#include <asm/uaccess.h>
103#include <linux/skbuff.h>
104#include <linux/netdevice.h>
105#include <linux/in.h>
106#include <linux/tcp.h>
107#include <linux/udp.h>
108#include <linux/if_arp.h>
109#include <linux/mroute.h>
110#include <linux/init.h>
111#include <linux/netfilter_ipv4.h>
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800112#include <linux/if_ether.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
114#include <net/sock.h>
115#include <net/ip.h>
116#include <net/icmp.h>
117#include <net/protocol.h>
118#include <net/ipip.h>
119#include <net/inet_ecn.h>
120#include <net/xfrm.h>
121
122#define HASH_SIZE 16
123#define HASH(addr) ((addr^(addr>>4))&0xF)
124
125static int ipip_fb_tunnel_init(struct net_device *dev);
126static int ipip_tunnel_init(struct net_device *dev);
127static void ipip_tunnel_setup(struct net_device *dev);
128
129static struct net_device *ipip_fb_tunnel_dev;
130
131static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
132static struct ip_tunnel *tunnels_r[HASH_SIZE];
133static struct ip_tunnel *tunnels_l[HASH_SIZE];
134static struct ip_tunnel *tunnels_wc[1];
135static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
136
137static DEFINE_RWLOCK(ipip_lock);
138
139static struct ip_tunnel * ipip_tunnel_lookup(u32 remote, u32 local)
140{
141 unsigned h0 = HASH(remote);
142 unsigned h1 = HASH(local);
143 struct ip_tunnel *t;
144
145 for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
146 if (local == t->parms.iph.saddr &&
147 remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
148 return t;
149 }
150 for (t = tunnels_r[h0]; t; t = t->next) {
151 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
152 return t;
153 }
154 for (t = tunnels_l[h1]; t; t = t->next) {
155 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
156 return t;
157 }
158 if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
159 return t;
160 return NULL;
161}
162
163static struct ip_tunnel **ipip_bucket(struct ip_tunnel *t)
164{
165 u32 remote = t->parms.iph.daddr;
166 u32 local = t->parms.iph.saddr;
167 unsigned h = 0;
168 int prio = 0;
169
170 if (remote) {
171 prio |= 2;
172 h ^= HASH(remote);
173 }
174 if (local) {
175 prio |= 1;
176 h ^= HASH(local);
177 }
178 return &tunnels[prio][h];
179}
180
181
182static void ipip_tunnel_unlink(struct ip_tunnel *t)
183{
184 struct ip_tunnel **tp;
185
186 for (tp = ipip_bucket(t); *tp; tp = &(*tp)->next) {
187 if (t == *tp) {
188 write_lock_bh(&ipip_lock);
189 *tp = t->next;
190 write_unlock_bh(&ipip_lock);
191 break;
192 }
193 }
194}
195
196static void ipip_tunnel_link(struct ip_tunnel *t)
197{
198 struct ip_tunnel **tp = ipip_bucket(t);
199
200 t->next = *tp;
201 write_lock_bh(&ipip_lock);
202 *tp = t;
203 write_unlock_bh(&ipip_lock);
204}
205
206static struct ip_tunnel * ipip_tunnel_locate(struct ip_tunnel_parm *parms, int create)
207{
208 u32 remote = parms->iph.daddr;
209 u32 local = parms->iph.saddr;
210 struct ip_tunnel *t, **tp, *nt;
211 struct net_device *dev;
212 unsigned h = 0;
213 int prio = 0;
214 char name[IFNAMSIZ];
215
216 if (remote) {
217 prio |= 2;
218 h ^= HASH(remote);
219 }
220 if (local) {
221 prio |= 1;
222 h ^= HASH(local);
223 }
224 for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
225 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
226 return t;
227 }
228 if (!create)
229 return NULL;
230
231 if (parms->name[0])
232 strlcpy(name, parms->name, IFNAMSIZ);
233 else {
234 int i;
235 for (i=1; i<100; i++) {
236 sprintf(name, "tunl%d", i);
237 if (__dev_get_by_name(name) == NULL)
238 break;
239 }
240 if (i==100)
241 goto failed;
242 }
243
244 dev = alloc_netdev(sizeof(*t), name, ipip_tunnel_setup);
245 if (dev == NULL)
246 return NULL;
247
Patrick McHardy2941a482006-01-08 22:05:26 -0800248 nt = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 SET_MODULE_OWNER(dev);
250 dev->init = ipip_tunnel_init;
251 nt->parms = *parms;
252
253 if (register_netdevice(dev) < 0) {
254 free_netdev(dev);
255 goto failed;
256 }
257
258 dev_hold(dev);
259 ipip_tunnel_link(nt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 return nt;
261
262failed:
263 return NULL;
264}
265
266static void ipip_tunnel_uninit(struct net_device *dev)
267{
268 if (dev == ipip_fb_tunnel_dev) {
269 write_lock_bh(&ipip_lock);
270 tunnels_wc[0] = NULL;
271 write_unlock_bh(&ipip_lock);
272 } else
Patrick McHardy2941a482006-01-08 22:05:26 -0800273 ipip_tunnel_unlink(netdev_priv(dev));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 dev_put(dev);
275}
276
Patrick McHardy03037702005-07-19 14:03:34 -0700277static void ipip_err(struct sk_buff *skb, u32 info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
279#ifndef I_WISH_WORLD_WERE_PERFECT
280
281/* It is not :-( All the routers (except for Linux) return only
282 8 bytes of packet payload. It means, that precise relaying of
283 ICMP in the real Internet is absolutely infeasible.
284 */
285 struct iphdr *iph = (struct iphdr*)skb->data;
286 int type = skb->h.icmph->type;
287 int code = skb->h.icmph->code;
288 struct ip_tunnel *t;
289
290 switch (type) {
291 default:
292 case ICMP_PARAMETERPROB:
293 return;
294
295 case ICMP_DEST_UNREACH:
296 switch (code) {
297 case ICMP_SR_FAILED:
298 case ICMP_PORT_UNREACH:
299 /* Impossible event. */
300 return;
301 case ICMP_FRAG_NEEDED:
302 /* Soft state for pmtu is maintained by IP core. */
303 return;
304 default:
305 /* All others are translated to HOST_UNREACH.
306 rfc2003 contains "deep thoughts" about NET_UNREACH,
307 I believe they are just ether pollution. --ANK
308 */
309 break;
310 }
311 break;
312 case ICMP_TIME_EXCEEDED:
313 if (code != ICMP_EXC_TTL)
314 return;
315 break;
316 }
317
318 read_lock(&ipip_lock);
319 t = ipip_tunnel_lookup(iph->daddr, iph->saddr);
320 if (t == NULL || t->parms.iph.daddr == 0)
321 goto out;
322 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
323 goto out;
324
325 if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
326 t->err_count++;
327 else
328 t->err_count = 1;
329 t->err_time = jiffies;
330out:
331 read_unlock(&ipip_lock);
332 return;
333#else
334 struct iphdr *iph = (struct iphdr*)dp;
335 int hlen = iph->ihl<<2;
336 struct iphdr *eiph;
337 int type = skb->h.icmph->type;
338 int code = skb->h.icmph->code;
339 int rel_type = 0;
340 int rel_code = 0;
341 int rel_info = 0;
342 struct sk_buff *skb2;
343 struct flowi fl;
344 struct rtable *rt;
345
346 if (len < hlen + sizeof(struct iphdr))
347 return;
348 eiph = (struct iphdr*)(dp + hlen);
349
350 switch (type) {
351 default:
352 return;
353 case ICMP_PARAMETERPROB:
354 if (skb->h.icmph->un.gateway < hlen)
355 return;
356
357 /* So... This guy found something strange INSIDE encapsulated
358 packet. Well, he is fool, but what can we do ?
359 */
360 rel_type = ICMP_PARAMETERPROB;
361 rel_info = skb->h.icmph->un.gateway - hlen;
362 break;
363
364 case ICMP_DEST_UNREACH:
365 switch (code) {
366 case ICMP_SR_FAILED:
367 case ICMP_PORT_UNREACH:
368 /* Impossible event. */
369 return;
370 case ICMP_FRAG_NEEDED:
371 /* And it is the only really necessary thing :-) */
372 rel_info = ntohs(skb->h.icmph->un.frag.mtu);
373 if (rel_info < hlen+68)
374 return;
375 rel_info -= hlen;
376 /* BSD 4.2 MORE DOES NOT EXIST IN NATURE. */
377 if (rel_info > ntohs(eiph->tot_len))
378 return;
379 break;
380 default:
381 /* All others are translated to HOST_UNREACH.
382 rfc2003 contains "deep thoughts" about NET_UNREACH,
383 I believe, it is just ether pollution. --ANK
384 */
385 rel_type = ICMP_DEST_UNREACH;
386 rel_code = ICMP_HOST_UNREACH;
387 break;
388 }
389 break;
390 case ICMP_TIME_EXCEEDED:
391 if (code != ICMP_EXC_TTL)
392 return;
393 break;
394 }
395
396 /* Prepare fake skb to feed it to icmp_send */
397 skb2 = skb_clone(skb, GFP_ATOMIC);
398 if (skb2 == NULL)
399 return;
400 dst_release(skb2->dst);
401 skb2->dst = NULL;
402 skb_pull(skb2, skb->data - (u8*)eiph);
403 skb2->nh.raw = skb2->data;
404
405 /* Try to guess incoming interface */
406 memset(&fl, 0, sizeof(fl));
407 fl.fl4_daddr = eiph->saddr;
408 fl.fl4_tos = RT_TOS(eiph->tos);
409 fl.proto = IPPROTO_IPIP;
410 if (ip_route_output_key(&rt, &key)) {
411 kfree_skb(skb2);
412 return;
413 }
414 skb2->dev = rt->u.dst.dev;
415
416 /* route "incoming" packet */
417 if (rt->rt_flags&RTCF_LOCAL) {
418 ip_rt_put(rt);
419 rt = NULL;
420 fl.fl4_daddr = eiph->daddr;
421 fl.fl4_src = eiph->saddr;
422 fl.fl4_tos = eiph->tos;
423 if (ip_route_output_key(&rt, &fl) ||
424 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
425 ip_rt_put(rt);
426 kfree_skb(skb2);
427 return;
428 }
429 } else {
430 ip_rt_put(rt);
431 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, skb2->dev) ||
432 skb2->dst->dev->type != ARPHRD_TUNNEL) {
433 kfree_skb(skb2);
434 return;
435 }
436 }
437
438 /* change mtu on this route */
439 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
440 if (rel_info > dst_mtu(skb2->dst)) {
441 kfree_skb(skb2);
442 return;
443 }
444 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
445 rel_info = htonl(rel_info);
446 } else if (type == ICMP_TIME_EXCEEDED) {
Patrick McHardy2941a482006-01-08 22:05:26 -0800447 struct ip_tunnel *t = netdev_priv(skb2->dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 if (t->parms.iph.ttl) {
449 rel_type = ICMP_DEST_UNREACH;
450 rel_code = ICMP_HOST_UNREACH;
451 }
452 }
453
454 icmp_send(skb2, rel_type, rel_code, rel_info);
455 kfree_skb(skb2);
456 return;
457#endif
458}
459
460static inline void ipip_ecn_decapsulate(struct iphdr *outer_iph, struct sk_buff *skb)
461{
462 struct iphdr *inner_iph = skb->nh.iph;
463
464 if (INET_ECN_is_ce(outer_iph->tos))
465 IP_ECN_set_ce(inner_iph);
466}
467
468static int ipip_rcv(struct sk_buff *skb)
469{
470 struct iphdr *iph;
471 struct ip_tunnel *tunnel;
472
473 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
474 goto out;
475
476 iph = skb->nh.iph;
477
478 read_lock(&ipip_lock);
479 if ((tunnel = ipip_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
480 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
481 read_unlock(&ipip_lock);
482 kfree_skb(skb);
483 return 0;
484 }
485
486 secpath_reset(skb);
487
488 skb->mac.raw = skb->nh.raw;
489 skb->nh.raw = skb->data;
490 memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
491 skb->protocol = htons(ETH_P_IP);
492 skb->pkt_type = PACKET_HOST;
493
494 tunnel->stat.rx_packets++;
495 tunnel->stat.rx_bytes += skb->len;
496 skb->dev = tunnel->dev;
497 dst_release(skb->dst);
498 skb->dst = NULL;
499 nf_reset(skb);
500 ipip_ecn_decapsulate(iph, skb);
501 netif_rx(skb);
502 read_unlock(&ipip_lock);
503 return 0;
504 }
505 read_unlock(&ipip_lock);
506
507out:
508 return -1;
509}
510
511/*
512 * This function assumes it is being called from dev_queue_xmit()
513 * and that skb is filled properly by that function.
514 */
515
516static int ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
517{
Patrick McHardy2941a482006-01-08 22:05:26 -0800518 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 struct net_device_stats *stats = &tunnel->stat;
520 struct iphdr *tiph = &tunnel->parms.iph;
521 u8 tos = tunnel->parms.iph.tos;
522 u16 df = tiph->frag_off;
523 struct rtable *rt; /* Route to the other host */
524 struct net_device *tdev; /* Device to other host */
525 struct iphdr *old_iph = skb->nh.iph;
526 struct iphdr *iph; /* Our new IP header */
527 int max_headroom; /* The extra header space needed */
528 u32 dst = tiph->daddr;
529 int mtu;
530
531 if (tunnel->recursion++) {
532 tunnel->stat.collisions++;
533 goto tx_error;
534 }
535
536 if (skb->protocol != htons(ETH_P_IP))
537 goto tx_error;
538
539 if (tos&1)
540 tos = old_iph->tos;
541
542 if (!dst) {
543 /* NBMA tunnel */
544 if ((rt = (struct rtable*)skb->dst) == NULL) {
545 tunnel->stat.tx_fifo_errors++;
546 goto tx_error;
547 }
548 if ((dst = rt->rt_gateway) == 0)
549 goto tx_error_icmp;
550 }
551
552 {
553 struct flowi fl = { .oif = tunnel->parms.link,
554 .nl_u = { .ip4_u =
555 { .daddr = dst,
556 .saddr = tiph->saddr,
557 .tos = RT_TOS(tos) } },
558 .proto = IPPROTO_IPIP };
559 if (ip_route_output_key(&rt, &fl)) {
560 tunnel->stat.tx_carrier_errors++;
561 goto tx_error_icmp;
562 }
563 }
564 tdev = rt->u.dst.dev;
565
566 if (tdev == dev) {
567 ip_rt_put(rt);
568 tunnel->stat.collisions++;
569 goto tx_error;
570 }
571
572 if (tiph->frag_off)
573 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
574 else
575 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
576
577 if (mtu < 68) {
578 tunnel->stat.collisions++;
579 ip_rt_put(rt);
580 goto tx_error;
581 }
582 if (skb->dst)
583 skb->dst->ops->update_pmtu(skb->dst, mtu);
584
585 df |= (old_iph->frag_off&htons(IP_DF));
586
587 if ((old_iph->frag_off&htons(IP_DF)) && mtu < ntohs(old_iph->tot_len)) {
588 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
589 ip_rt_put(rt);
590 goto tx_error;
591 }
592
593 if (tunnel->err_count > 0) {
594 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
595 tunnel->err_count--;
596 dst_link_failure(skb);
597 } else
598 tunnel->err_count = 0;
599 }
600
601 /*
602 * Okay, now see if we can stuff it in the buffer as-is.
603 */
604 max_headroom = (LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr));
605
606 if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
607 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
608 if (!new_skb) {
609 ip_rt_put(rt);
610 stats->tx_dropped++;
611 dev_kfree_skb(skb);
612 tunnel->recursion--;
613 return 0;
614 }
615 if (skb->sk)
616 skb_set_owner_w(new_skb, skb->sk);
617 dev_kfree_skb(skb);
618 skb = new_skb;
619 old_iph = skb->nh.iph;
620 }
621
622 skb->h.raw = skb->nh.raw;
623 skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
624 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
Patrick McHardy3e3850e2006-01-06 23:04:54 -0800625 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE|IPSKB_XFRM_TRANSFORMED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 dst_release(skb->dst);
627 skb->dst = &rt->u.dst;
628
629 /*
630 * Push down and install the IPIP header.
631 */
632
633 iph = skb->nh.iph;
634 iph->version = 4;
635 iph->ihl = sizeof(struct iphdr)>>2;
636 iph->frag_off = df;
637 iph->protocol = IPPROTO_IPIP;
638 iph->tos = INET_ECN_encapsulate(tos, old_iph->tos);
639 iph->daddr = rt->rt_dst;
640 iph->saddr = rt->rt_src;
641
642 if ((iph->ttl = tiph->ttl) == 0)
643 iph->ttl = old_iph->ttl;
644
645 nf_reset(skb);
646
647 IPTUNNEL_XMIT();
648 tunnel->recursion--;
649 return 0;
650
651tx_error_icmp:
652 dst_link_failure(skb);
653tx_error:
654 stats->tx_errors++;
655 dev_kfree_skb(skb);
656 tunnel->recursion--;
657 return 0;
658}
659
660static int
661ipip_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
662{
663 int err = 0;
664 struct ip_tunnel_parm p;
665 struct ip_tunnel *t;
666
667 switch (cmd) {
668 case SIOCGETTUNNEL:
669 t = NULL;
670 if (dev == ipip_fb_tunnel_dev) {
671 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
672 err = -EFAULT;
673 break;
674 }
675 t = ipip_tunnel_locate(&p, 0);
676 }
677 if (t == NULL)
Patrick McHardy2941a482006-01-08 22:05:26 -0800678 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 memcpy(&p, &t->parms, sizeof(p));
680 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
681 err = -EFAULT;
682 break;
683
684 case SIOCADDTUNNEL:
685 case SIOCCHGTUNNEL:
686 err = -EPERM;
687 if (!capable(CAP_NET_ADMIN))
688 goto done;
689
690 err = -EFAULT;
691 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
692 goto done;
693
694 err = -EINVAL;
695 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
696 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
697 goto done;
698 if (p.iph.ttl)
699 p.iph.frag_off |= htons(IP_DF);
700
701 t = ipip_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
702
703 if (dev != ipip_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
704 if (t != NULL) {
705 if (t->dev != dev) {
706 err = -EEXIST;
707 break;
708 }
709 } else {
710 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
711 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
712 err = -EINVAL;
713 break;
714 }
Patrick McHardy2941a482006-01-08 22:05:26 -0800715 t = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 ipip_tunnel_unlink(t);
717 t->parms.iph.saddr = p.iph.saddr;
718 t->parms.iph.daddr = p.iph.daddr;
719 memcpy(dev->dev_addr, &p.iph.saddr, 4);
720 memcpy(dev->broadcast, &p.iph.daddr, 4);
721 ipip_tunnel_link(t);
722 netdev_state_change(dev);
723 }
724 }
725
726 if (t) {
727 err = 0;
728 if (cmd == SIOCCHGTUNNEL) {
729 t->parms.iph.ttl = p.iph.ttl;
730 t->parms.iph.tos = p.iph.tos;
731 t->parms.iph.frag_off = p.iph.frag_off;
732 }
733 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
734 err = -EFAULT;
735 } else
736 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
737 break;
738
739 case SIOCDELTUNNEL:
740 err = -EPERM;
741 if (!capable(CAP_NET_ADMIN))
742 goto done;
743
744 if (dev == ipip_fb_tunnel_dev) {
745 err = -EFAULT;
746 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
747 goto done;
748 err = -ENOENT;
749 if ((t = ipip_tunnel_locate(&p, 0)) == NULL)
750 goto done;
751 err = -EPERM;
752 if (t->dev == ipip_fb_tunnel_dev)
753 goto done;
754 dev = t->dev;
755 }
756 err = unregister_netdevice(dev);
757 break;
758
759 default:
760 err = -EINVAL;
761 }
762
763done:
764 return err;
765}
766
767static struct net_device_stats *ipip_tunnel_get_stats(struct net_device *dev)
768{
Patrick McHardy2941a482006-01-08 22:05:26 -0800769 return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770}
771
772static int ipip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
773{
774 if (new_mtu < 68 || new_mtu > 0xFFF8 - sizeof(struct iphdr))
775 return -EINVAL;
776 dev->mtu = new_mtu;
777 return 0;
778}
779
780static void ipip_tunnel_setup(struct net_device *dev)
781{
782 SET_MODULE_OWNER(dev);
783 dev->uninit = ipip_tunnel_uninit;
784 dev->hard_start_xmit = ipip_tunnel_xmit;
785 dev->get_stats = ipip_tunnel_get_stats;
786 dev->do_ioctl = ipip_tunnel_ioctl;
787 dev->change_mtu = ipip_tunnel_change_mtu;
788 dev->destructor = free_netdev;
789
790 dev->type = ARPHRD_TUNNEL;
791 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
Kris Katterjohn46f25df2006-01-05 16:35:42 -0800792 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 dev->flags = IFF_NOARP;
794 dev->iflink = 0;
795 dev->addr_len = 4;
796}
797
798static int ipip_tunnel_init(struct net_device *dev)
799{
800 struct net_device *tdev = NULL;
801 struct ip_tunnel *tunnel;
802 struct iphdr *iph;
803
Patrick McHardy2941a482006-01-08 22:05:26 -0800804 tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 iph = &tunnel->parms.iph;
806
807 tunnel->dev = dev;
808 strcpy(tunnel->parms.name, dev->name);
809
810 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
811 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
812
813 if (iph->daddr) {
814 struct flowi fl = { .oif = tunnel->parms.link,
815 .nl_u = { .ip4_u =
816 { .daddr = iph->daddr,
817 .saddr = iph->saddr,
818 .tos = RT_TOS(iph->tos) } },
819 .proto = IPPROTO_IPIP };
820 struct rtable *rt;
821 if (!ip_route_output_key(&rt, &fl)) {
822 tdev = rt->u.dst.dev;
823 ip_rt_put(rt);
824 }
825 dev->flags |= IFF_POINTOPOINT;
826 }
827
828 if (!tdev && tunnel->parms.link)
829 tdev = __dev_get_by_index(tunnel->parms.link);
830
831 if (tdev) {
832 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
833 dev->mtu = tdev->mtu - sizeof(struct iphdr);
834 }
835 dev->iflink = tunnel->parms.link;
836
837 return 0;
838}
839
840static int __init ipip_fb_tunnel_init(struct net_device *dev)
841{
Patrick McHardy2941a482006-01-08 22:05:26 -0800842 struct ip_tunnel *tunnel = netdev_priv(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 struct iphdr *iph = &tunnel->parms.iph;
844
845 tunnel->dev = dev;
846 strcpy(tunnel->parms.name, dev->name);
847
848 iph->version = 4;
849 iph->protocol = IPPROTO_IPIP;
850 iph->ihl = 5;
851
852 dev_hold(dev);
853 tunnels_wc[0] = tunnel;
854 return 0;
855}
856
Patrick McHardy03037702005-07-19 14:03:34 -0700857#ifdef CONFIG_INET_TUNNEL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858static struct xfrm_tunnel ipip_handler = {
859 .handler = ipip_rcv,
860 .err_handler = ipip_err,
861};
862
Patrick McHardy03037702005-07-19 14:03:34 -0700863static inline int ipip_register(void)
864{
865 return xfrm4_tunnel_register(&ipip_handler);
866}
867
868static inline int ipip_unregister(void)
869{
870 return xfrm4_tunnel_deregister(&ipip_handler);
871}
872#else
873static struct net_protocol ipip_protocol = {
874 .handler = ipip_rcv,
875 .err_handler = ipip_err,
876 .no_policy = 1,
877};
878
879static inline int ipip_register(void)
880{
881 return inet_add_protocol(&ipip_protocol, IPPROTO_IPIP);
882}
883
884static inline int ipip_unregister(void)
885{
886 return inet_del_protocol(&ipip_protocol, IPPROTO_IPIP);
887}
888#endif
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890static char banner[] __initdata =
891 KERN_INFO "IPv4 over IPv4 tunneling driver\n";
892
893static int __init ipip_init(void)
894{
895 int err;
896
897 printk(banner);
898
Patrick McHardy03037702005-07-19 14:03:34 -0700899 if (ipip_register() < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 printk(KERN_INFO "ipip init: can't register tunnel\n");
901 return -EAGAIN;
902 }
903
904 ipip_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
905 "tunl0",
906 ipip_tunnel_setup);
907 if (!ipip_fb_tunnel_dev) {
908 err = -ENOMEM;
909 goto err1;
910 }
911
912 ipip_fb_tunnel_dev->init = ipip_fb_tunnel_init;
913
914 if ((err = register_netdev(ipip_fb_tunnel_dev)))
915 goto err2;
916 out:
917 return err;
918 err2:
919 free_netdev(ipip_fb_tunnel_dev);
920 err1:
Patrick McHardy03037702005-07-19 14:03:34 -0700921 ipip_unregister();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 goto out;
923}
924
Alexey Kuznetsovdb445752005-07-30 17:46:44 -0700925static void __exit ipip_destroy_tunnels(void)
926{
927 int prio;
928
929 for (prio = 1; prio < 4; prio++) {
930 int h;
931 for (h = 0; h < HASH_SIZE; h++) {
932 struct ip_tunnel *t;
933 while ((t = tunnels[prio][h]) != NULL)
934 unregister_netdevice(t->dev);
935 }
936 }
937}
938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939static void __exit ipip_fini(void)
940{
Patrick McHardy03037702005-07-19 14:03:34 -0700941 if (ipip_unregister() < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 printk(KERN_INFO "ipip close: can't deregister tunnel\n");
943
Alexey Kuznetsovdb445752005-07-30 17:46:44 -0700944 rtnl_lock();
945 ipip_destroy_tunnels();
946 unregister_netdevice(ipip_fb_tunnel_dev);
947 rtnl_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948}
949
950module_init(ipip_init);
951module_exit(ipip_fini);
952MODULE_LICENSE("GPL");