Loading...
1/*
2 * Linux NET3: IP/IP protocol decoder.
3 *
4 * Authors:
5 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
6 *
7 * Fixes:
8 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
9 * a module taking up 2 pages).
10 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
11 * to keep ip_forward happy.
12 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
13 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
14 * David Woodhouse : Perform some basic ICMP handling.
15 * IPIP Routing without decapsulation.
16 * Carlos Picoto : GRE over IP support
17 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
18 * I do not want to merge them together.
19 *
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
24 *
25 */
26
27/* tunnel.c: an IP tunnel driver
28
29 The purpose of this driver is to provide an IP tunnel through
30 which you can tunnel network traffic transparently across subnets.
31
32 This was written by looking at Nick Holloway's dummy driver
33 Thanks for the great code!
34
35 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
36
37 Minor tweaks:
38 Cleaned up the code a little and added some pre-1.3.0 tweaks.
39 dev->hard_header/hard_header_len changed to use no headers.
40 Comments/bracketing tweaked.
41 Made the tunnels use dev->name not tunnel: when error reporting.
42 Added tx_dropped stat
43
44 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
45
46 Reworked:
47 Changed to tunnel to destination gateway in addition to the
48 tunnel's pointopoint address
49 Almost completely rewritten
50 Note: There is currently no firewall or ICMP handling done.
51
52 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
53
54*/
55
56/* Things I wish I had known when writing the tunnel driver:
57
58 When the tunnel_xmit() function is called, the skb contains the
59 packet to be sent (plus a great deal of extra info), and dev
60 contains the tunnel device that _we_ are.
61
62 When we are passed a packet, we are expected to fill in the
63 source address with our source IP address.
64
65 What is the proper way to allocate, copy and free a buffer?
66 After you allocate it, it is a "0 length" chunk of memory
67 starting at zero. If you want to add headers to the buffer
68 later, you'll have to call "skb_reserve(skb, amount)" with
69 the amount of memory you want reserved. Then, you call
70 "skb_put(skb, amount)" with the amount of space you want in
71 the buffer. skb_put() returns a pointer to the top (#0) of
72 that buffer. skb->len is set to the amount of space you have
73 "allocated" with skb_put(). You can then write up to skb->len
74 bytes to that buffer. If you need more, you can call skb_put()
75 again with the additional amount of space you need. You can
76 find out how much more space you can allocate by calling
77 "skb_tailroom(skb)".
78 Now, to add header space, call "skb_push(skb, header_len)".
79 This creates space at the beginning of the buffer and returns
80 a pointer to this new space. If later you need to strip a
81 header from a buffer, call "skb_pull(skb, header_len)".
82 skb_headroom() will return how much space is left at the top
83 of the buffer (before the main data). Remember, this headroom
84 space must be reserved before the skb_put() function is called.
85 */
86
87/*
88 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
89
90 For comments look at net/ipv4/ip_gre.c --ANK
91 */
92
93
94#include <linux/capability.h>
95#include <linux/module.h>
96#include <linux/types.h>
97#include <linux/kernel.h>
98#include <linux/slab.h>
99#include <asm/uaccess.h>
100#include <linux/skbuff.h>
101#include <linux/netdevice.h>
102#include <linux/in.h>
103#include <linux/tcp.h>
104#include <linux/udp.h>
105#include <linux/if_arp.h>
106#include <linux/init.h>
107#include <linux/netfilter_ipv4.h>
108#include <linux/if_ether.h>
109
110#include <net/sock.h>
111#include <net/ip.h>
112#include <net/icmp.h>
113#include <net/ip_tunnels.h>
114#include <net/inet_ecn.h>
115#include <net/xfrm.h>
116#include <net/net_namespace.h>
117#include <net/netns/generic.h>
118
119static bool log_ecn_error = true;
120module_param(log_ecn_error, bool, 0644);
121MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
122
123static int ipip_net_id __read_mostly;
124
125static int ipip_tunnel_init(struct net_device *dev);
126static struct rtnl_link_ops ipip_link_ops __read_mostly;
127
128static int ipip_err(struct sk_buff *skb, u32 info)
129{
130
131/* All the routers (except for Linux) return only
132 8 bytes of packet payload. It means, that precise relaying of
133 ICMP in the real Internet is absolutely infeasible.
134 */
135 struct net *net = dev_net(skb->dev);
136 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
137 const struct iphdr *iph = (const struct iphdr *)skb->data;
138 struct ip_tunnel *t;
139 int err;
140 const int type = icmp_hdr(skb)->type;
141 const int code = icmp_hdr(skb)->code;
142
143 err = -ENOENT;
144 t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
145 iph->daddr, iph->saddr, 0);
146 if (!t)
147 goto out;
148
149 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
150 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
151 t->parms.link, 0, IPPROTO_IPIP, 0);
152 err = 0;
153 goto out;
154 }
155
156 if (type == ICMP_REDIRECT) {
157 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
158 IPPROTO_IPIP, 0);
159 err = 0;
160 goto out;
161 }
162
163 if (t->parms.iph.daddr == 0)
164 goto out;
165
166 err = 0;
167 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
168 goto out;
169
170 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
171 t->err_count++;
172 else
173 t->err_count = 1;
174 t->err_time = jiffies;
175
176out:
177 return err;
178}
179
180static const struct tnl_ptk_info tpi = {
181 /* no tunnel info required for ipip. */
182 .proto = htons(ETH_P_IP),
183};
184
185static int ipip_rcv(struct sk_buff *skb)
186{
187 struct net *net = dev_net(skb->dev);
188 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
189 struct ip_tunnel *tunnel;
190 const struct iphdr *iph;
191
192 iph = ip_hdr(skb);
193 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
194 iph->saddr, iph->daddr, 0);
195 if (tunnel) {
196 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
197 goto drop;
198 if (iptunnel_pull_header(skb, 0, tpi.proto, false))
199 goto drop;
200 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
201 }
202
203 return -1;
204
205drop:
206 kfree_skb(skb);
207 return 0;
208}
209
210/*
211 * This function assumes it is being called from dev_queue_xmit()
212 * and that skb is filled properly by that function.
213 */
214static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
215{
216 struct ip_tunnel *tunnel = netdev_priv(dev);
217 const struct iphdr *tiph = &tunnel->parms.iph;
218
219 if (unlikely(skb->protocol != htons(ETH_P_IP)))
220 goto tx_error;
221
222 skb = iptunnel_handle_offloads(skb, SKB_GSO_IPIP);
223 if (IS_ERR(skb))
224 goto out;
225
226 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
227
228 ip_tunnel_xmit(skb, dev, tiph, tiph->protocol);
229 return NETDEV_TX_OK;
230
231tx_error:
232 kfree_skb(skb);
233out:
234 dev->stats.tx_errors++;
235 return NETDEV_TX_OK;
236}
237
238static int
239ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
240{
241 int err = 0;
242 struct ip_tunnel_parm p;
243
244 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
245 return -EFAULT;
246
247 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
248 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
249 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
250 return -EINVAL;
251 }
252
253 p.i_key = p.o_key = 0;
254 p.i_flags = p.o_flags = 0;
255 err = ip_tunnel_ioctl(dev, &p, cmd);
256 if (err)
257 return err;
258
259 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
260 return -EFAULT;
261
262 return 0;
263}
264
265static const struct net_device_ops ipip_netdev_ops = {
266 .ndo_init = ipip_tunnel_init,
267 .ndo_uninit = ip_tunnel_uninit,
268 .ndo_start_xmit = ipip_tunnel_xmit,
269 .ndo_do_ioctl = ipip_tunnel_ioctl,
270 .ndo_change_mtu = ip_tunnel_change_mtu,
271 .ndo_get_stats64 = ip_tunnel_get_stats64,
272 .ndo_get_iflink = ip_tunnel_get_iflink,
273};
274
275#define IPIP_FEATURES (NETIF_F_SG | \
276 NETIF_F_FRAGLIST | \
277 NETIF_F_HIGHDMA | \
278 NETIF_F_GSO_SOFTWARE | \
279 NETIF_F_HW_CSUM)
280
281static void ipip_tunnel_setup(struct net_device *dev)
282{
283 dev->netdev_ops = &ipip_netdev_ops;
284
285 dev->type = ARPHRD_TUNNEL;
286 dev->flags = IFF_NOARP;
287 dev->addr_len = 4;
288 dev->features |= NETIF_F_LLTX;
289 netif_keep_dst(dev);
290
291 dev->features |= IPIP_FEATURES;
292 dev->hw_features |= IPIP_FEATURES;
293 ip_tunnel_setup(dev, ipip_net_id);
294}
295
296static int ipip_tunnel_init(struct net_device *dev)
297{
298 struct ip_tunnel *tunnel = netdev_priv(dev);
299
300 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
301 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
302
303 tunnel->tun_hlen = 0;
304 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
305 tunnel->parms.iph.protocol = IPPROTO_IPIP;
306 return ip_tunnel_init(dev);
307}
308
309static void ipip_netlink_parms(struct nlattr *data[],
310 struct ip_tunnel_parm *parms)
311{
312 memset(parms, 0, sizeof(*parms));
313
314 parms->iph.version = 4;
315 parms->iph.protocol = IPPROTO_IPIP;
316 parms->iph.ihl = 5;
317
318 if (!data)
319 return;
320
321 if (data[IFLA_IPTUN_LINK])
322 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
323
324 if (data[IFLA_IPTUN_LOCAL])
325 parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
326
327 if (data[IFLA_IPTUN_REMOTE])
328 parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
329
330 if (data[IFLA_IPTUN_TTL]) {
331 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
332 if (parms->iph.ttl)
333 parms->iph.frag_off = htons(IP_DF);
334 }
335
336 if (data[IFLA_IPTUN_TOS])
337 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
338
339 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
340 parms->iph.frag_off = htons(IP_DF);
341}
342
343/* This function returns true when ENCAP attributes are present in the nl msg */
344static bool ipip_netlink_encap_parms(struct nlattr *data[],
345 struct ip_tunnel_encap *ipencap)
346{
347 bool ret = false;
348
349 memset(ipencap, 0, sizeof(*ipencap));
350
351 if (!data)
352 return ret;
353
354 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
355 ret = true;
356 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
357 }
358
359 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
360 ret = true;
361 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
362 }
363
364 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
365 ret = true;
366 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
367 }
368
369 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
370 ret = true;
371 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
372 }
373
374 return ret;
375}
376
377static int ipip_newlink(struct net *src_net, struct net_device *dev,
378 struct nlattr *tb[], struct nlattr *data[])
379{
380 struct ip_tunnel_parm p;
381 struct ip_tunnel_encap ipencap;
382
383 if (ipip_netlink_encap_parms(data, &ipencap)) {
384 struct ip_tunnel *t = netdev_priv(dev);
385 int err = ip_tunnel_encap_setup(t, &ipencap);
386
387 if (err < 0)
388 return err;
389 }
390
391 ipip_netlink_parms(data, &p);
392 return ip_tunnel_newlink(dev, tb, &p);
393}
394
395static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
396 struct nlattr *data[])
397{
398 struct ip_tunnel_parm p;
399 struct ip_tunnel_encap ipencap;
400
401 if (ipip_netlink_encap_parms(data, &ipencap)) {
402 struct ip_tunnel *t = netdev_priv(dev);
403 int err = ip_tunnel_encap_setup(t, &ipencap);
404
405 if (err < 0)
406 return err;
407 }
408
409 ipip_netlink_parms(data, &p);
410
411 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
412 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
413 return -EINVAL;
414
415 return ip_tunnel_changelink(dev, tb, &p);
416}
417
418static size_t ipip_get_size(const struct net_device *dev)
419{
420 return
421 /* IFLA_IPTUN_LINK */
422 nla_total_size(4) +
423 /* IFLA_IPTUN_LOCAL */
424 nla_total_size(4) +
425 /* IFLA_IPTUN_REMOTE */
426 nla_total_size(4) +
427 /* IFLA_IPTUN_TTL */
428 nla_total_size(1) +
429 /* IFLA_IPTUN_TOS */
430 nla_total_size(1) +
431 /* IFLA_IPTUN_PMTUDISC */
432 nla_total_size(1) +
433 /* IFLA_IPTUN_ENCAP_TYPE */
434 nla_total_size(2) +
435 /* IFLA_IPTUN_ENCAP_FLAGS */
436 nla_total_size(2) +
437 /* IFLA_IPTUN_ENCAP_SPORT */
438 nla_total_size(2) +
439 /* IFLA_IPTUN_ENCAP_DPORT */
440 nla_total_size(2) +
441 0;
442}
443
444static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
445{
446 struct ip_tunnel *tunnel = netdev_priv(dev);
447 struct ip_tunnel_parm *parm = &tunnel->parms;
448
449 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
450 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
451 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
452 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
453 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
454 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
455 !!(parm->iph.frag_off & htons(IP_DF))))
456 goto nla_put_failure;
457
458 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
459 tunnel->encap.type) ||
460 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
461 tunnel->encap.sport) ||
462 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
463 tunnel->encap.dport) ||
464 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
465 tunnel->encap.flags))
466 goto nla_put_failure;
467
468 return 0;
469
470nla_put_failure:
471 return -EMSGSIZE;
472}
473
474static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
475 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
476 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
477 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
478 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
479 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
480 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
481 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
482 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
483 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
484 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
485};
486
487static struct rtnl_link_ops ipip_link_ops __read_mostly = {
488 .kind = "ipip",
489 .maxtype = IFLA_IPTUN_MAX,
490 .policy = ipip_policy,
491 .priv_size = sizeof(struct ip_tunnel),
492 .setup = ipip_tunnel_setup,
493 .newlink = ipip_newlink,
494 .changelink = ipip_changelink,
495 .dellink = ip_tunnel_dellink,
496 .get_size = ipip_get_size,
497 .fill_info = ipip_fill_info,
498 .get_link_net = ip_tunnel_get_link_net,
499};
500
501static struct xfrm_tunnel ipip_handler __read_mostly = {
502 .handler = ipip_rcv,
503 .err_handler = ipip_err,
504 .priority = 1,
505};
506
507static int __net_init ipip_init_net(struct net *net)
508{
509 return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
510}
511
512static void __net_exit ipip_exit_net(struct net *net)
513{
514 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
515 ip_tunnel_delete_net(itn, &ipip_link_ops);
516}
517
518static struct pernet_operations ipip_net_ops = {
519 .init = ipip_init_net,
520 .exit = ipip_exit_net,
521 .id = &ipip_net_id,
522 .size = sizeof(struct ip_tunnel_net),
523};
524
525static int __init ipip_init(void)
526{
527 int err;
528
529 pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
530
531 err = register_pernet_device(&ipip_net_ops);
532 if (err < 0)
533 return err;
534 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
535 if (err < 0) {
536 pr_info("%s: can't register tunnel\n", __func__);
537 goto xfrm_tunnel_failed;
538 }
539 err = rtnl_link_register(&ipip_link_ops);
540 if (err < 0)
541 goto rtnl_link_failed;
542
543out:
544 return err;
545
546rtnl_link_failed:
547 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
548xfrm_tunnel_failed:
549 unregister_pernet_device(&ipip_net_ops);
550 goto out;
551}
552
553static void __exit ipip_fini(void)
554{
555 rtnl_link_unregister(&ipip_link_ops);
556 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
557 pr_info("%s: can't deregister tunnel\n", __func__);
558
559 unregister_pernet_device(&ipip_net_ops);
560}
561
562module_init(ipip_init);
563module_exit(ipip_fini);
564MODULE_LICENSE("GPL");
565MODULE_ALIAS_RTNL_LINK("ipip");
566MODULE_ALIAS_NETDEV("tunl0");
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Linux NET3: IP/IP protocol decoder.
4 *
5 * Authors:
6 * Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
7 *
8 * Fixes:
9 * Alan Cox : Merged and made usable non modular (its so tiny its silly as
10 * a module taking up 2 pages).
11 * Alan Cox : Fixed bug with 1.3.18 and IPIP not working (now needs to set skb->h.iph)
12 * to keep ip_forward happy.
13 * Alan Cox : More fixes for 1.3.21, and firewall fix. Maybe this will work soon 8).
14 * Kai Schulte : Fixed #defines for IP_FIREWALL->FIREWALL
15 * David Woodhouse : Perform some basic ICMP handling.
16 * IPIP Routing without decapsulation.
17 * Carlos Picoto : GRE over IP support
18 * Alexey Kuznetsov: Reworked. Really, now it is truncated version of ipv4/ip_gre.c.
19 * I do not want to merge them together.
20 */
21
22/* tunnel.c: an IP tunnel driver
23
24 The purpose of this driver is to provide an IP tunnel through
25 which you can tunnel network traffic transparently across subnets.
26
27 This was written by looking at Nick Holloway's dummy driver
28 Thanks for the great code!
29
30 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/01/95
31
32 Minor tweaks:
33 Cleaned up the code a little and added some pre-1.3.0 tweaks.
34 dev->hard_header/hard_header_len changed to use no headers.
35 Comments/bracketing tweaked.
36 Made the tunnels use dev->name not tunnel: when error reporting.
37 Added tx_dropped stat
38
39 -Alan Cox (alan@lxorguk.ukuu.org.uk) 21 March 95
40
41 Reworked:
42 Changed to tunnel to destination gateway in addition to the
43 tunnel's pointopoint address
44 Almost completely rewritten
45 Note: There is currently no firewall or ICMP handling done.
46
47 -Sam Lantinga (slouken@cs.ucdavis.edu) 02/13/96
48
49*/
50
51/* Things I wish I had known when writing the tunnel driver:
52
53 When the tunnel_xmit() function is called, the skb contains the
54 packet to be sent (plus a great deal of extra info), and dev
55 contains the tunnel device that _we_ are.
56
57 When we are passed a packet, we are expected to fill in the
58 source address with our source IP address.
59
60 What is the proper way to allocate, copy and free a buffer?
61 After you allocate it, it is a "0 length" chunk of memory
62 starting at zero. If you want to add headers to the buffer
63 later, you'll have to call "skb_reserve(skb, amount)" with
64 the amount of memory you want reserved. Then, you call
65 "skb_put(skb, amount)" with the amount of space you want in
66 the buffer. skb_put() returns a pointer to the top (#0) of
67 that buffer. skb->len is set to the amount of space you have
68 "allocated" with skb_put(). You can then write up to skb->len
69 bytes to that buffer. If you need more, you can call skb_put()
70 again with the additional amount of space you need. You can
71 find out how much more space you can allocate by calling
72 "skb_tailroom(skb)".
73 Now, to add header space, call "skb_push(skb, header_len)".
74 This creates space at the beginning of the buffer and returns
75 a pointer to this new space. If later you need to strip a
76 header from a buffer, call "skb_pull(skb, header_len)".
77 skb_headroom() will return how much space is left at the top
78 of the buffer (before the main data). Remember, this headroom
79 space must be reserved before the skb_put() function is called.
80 */
81
82/*
83 This version of net/ipv4/ipip.c is cloned of net/ipv4/ip_gre.c
84
85 For comments look at net/ipv4/ip_gre.c --ANK
86 */
87
88
89#include <linux/capability.h>
90#include <linux/module.h>
91#include <linux/types.h>
92#include <linux/kernel.h>
93#include <linux/slab.h>
94#include <linux/uaccess.h>
95#include <linux/skbuff.h>
96#include <linux/netdevice.h>
97#include <linux/in.h>
98#include <linux/tcp.h>
99#include <linux/udp.h>
100#include <linux/if_arp.h>
101#include <linux/init.h>
102#include <linux/netfilter_ipv4.h>
103#include <linux/if_ether.h>
104
105#include <net/sock.h>
106#include <net/ip.h>
107#include <net/icmp.h>
108#include <net/ip_tunnels.h>
109#include <net/inet_ecn.h>
110#include <net/xfrm.h>
111#include <net/net_namespace.h>
112#include <net/netns/generic.h>
113#include <net/dst_metadata.h>
114
115static bool log_ecn_error = true;
116module_param(log_ecn_error, bool, 0644);
117MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
118
119static unsigned int ipip_net_id __read_mostly;
120
121static int ipip_tunnel_init(struct net_device *dev);
122static struct rtnl_link_ops ipip_link_ops __read_mostly;
123
124static int ipip_err(struct sk_buff *skb, u32 info)
125{
126 /* All the routers (except for Linux) return only
127 * 8 bytes of packet payload. It means, that precise relaying of
128 * ICMP in the real Internet is absolutely infeasible.
129 */
130 struct net *net = dev_net(skb->dev);
131 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
132 const struct iphdr *iph = (const struct iphdr *)skb->data;
133 const int type = icmp_hdr(skb)->type;
134 const int code = icmp_hdr(skb)->code;
135 struct ip_tunnel *t;
136 int err = 0;
137
138 t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
139 iph->daddr, iph->saddr, 0);
140 if (!t) {
141 err = -ENOENT;
142 goto out;
143 }
144
145 switch (type) {
146 case ICMP_DEST_UNREACH:
147 switch (code) {
148 case ICMP_SR_FAILED:
149 /* Impossible event. */
150 goto out;
151 default:
152 /* All others are translated to HOST_UNREACH.
153 * rfc2003 contains "deep thoughts" about NET_UNREACH,
154 * I believe they are just ether pollution. --ANK
155 */
156 break;
157 }
158 break;
159
160 case ICMP_TIME_EXCEEDED:
161 if (code != ICMP_EXC_TTL)
162 goto out;
163 break;
164
165 case ICMP_REDIRECT:
166 break;
167
168 default:
169 goto out;
170 }
171
172 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
173 ipv4_update_pmtu(skb, net, info, t->parms.link, iph->protocol);
174 goto out;
175 }
176
177 if (type == ICMP_REDIRECT) {
178 ipv4_redirect(skb, net, t->parms.link, iph->protocol);
179 goto out;
180 }
181
182 if (t->parms.iph.daddr == 0) {
183 err = -ENOENT;
184 goto out;
185 }
186
187 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
188 goto out;
189
190 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
191 t->err_count++;
192 else
193 t->err_count = 1;
194 t->err_time = jiffies;
195
196out:
197 return err;
198}
199
200static const struct tnl_ptk_info ipip_tpi = {
201 /* no tunnel info required for ipip. */
202 .proto = htons(ETH_P_IP),
203};
204
205#if IS_ENABLED(CONFIG_MPLS)
206static const struct tnl_ptk_info mplsip_tpi = {
207 /* no tunnel info required for mplsip. */
208 .proto = htons(ETH_P_MPLS_UC),
209};
210#endif
211
212static int ipip_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
213{
214 struct net *net = dev_net(skb->dev);
215 struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
216 struct metadata_dst *tun_dst = NULL;
217 struct ip_tunnel *tunnel;
218 const struct iphdr *iph;
219
220 iph = ip_hdr(skb);
221 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
222 iph->saddr, iph->daddr, 0);
223 if (tunnel) {
224 const struct tnl_ptk_info *tpi;
225
226 if (tunnel->parms.iph.protocol != ipproto &&
227 tunnel->parms.iph.protocol != 0)
228 goto drop;
229
230 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
231 goto drop;
232#if IS_ENABLED(CONFIG_MPLS)
233 if (ipproto == IPPROTO_MPLS)
234 tpi = &mplsip_tpi;
235 else
236#endif
237 tpi = &ipip_tpi;
238 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
239 goto drop;
240 if (tunnel->collect_md) {
241 tun_dst = ip_tun_rx_dst(skb, 0, 0, 0);
242 if (!tun_dst)
243 return 0;
244 ip_tunnel_md_udp_encap(skb, &tun_dst->u.tun_info);
245 }
246 skb_reset_mac_header(skb);
247
248 return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
249 }
250
251 return -1;
252
253drop:
254 kfree_skb(skb);
255 return 0;
256}
257
258static int ipip_rcv(struct sk_buff *skb)
259{
260 return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
261}
262
263#if IS_ENABLED(CONFIG_MPLS)
264static int mplsip_rcv(struct sk_buff *skb)
265{
266 return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
267}
268#endif
269
270/*
271 * This function assumes it is being called from dev_queue_xmit()
272 * and that skb is filled properly by that function.
273 */
274static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
275 struct net_device *dev)
276{
277 struct ip_tunnel *tunnel = netdev_priv(dev);
278 const struct iphdr *tiph = &tunnel->parms.iph;
279 u8 ipproto;
280
281 if (!pskb_inet_may_pull(skb))
282 goto tx_error;
283
284 switch (skb->protocol) {
285 case htons(ETH_P_IP):
286 ipproto = IPPROTO_IPIP;
287 break;
288#if IS_ENABLED(CONFIG_MPLS)
289 case htons(ETH_P_MPLS_UC):
290 ipproto = IPPROTO_MPLS;
291 break;
292#endif
293 default:
294 goto tx_error;
295 }
296
297 if (tiph->protocol != ipproto && tiph->protocol != 0)
298 goto tx_error;
299
300 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
301 goto tx_error;
302
303 skb_set_inner_ipproto(skb, ipproto);
304
305 if (tunnel->collect_md)
306 ip_md_tunnel_xmit(skb, dev, ipproto, 0);
307 else
308 ip_tunnel_xmit(skb, dev, tiph, ipproto);
309 return NETDEV_TX_OK;
310
311tx_error:
312 kfree_skb(skb);
313
314 DEV_STATS_INC(dev, tx_errors);
315 return NETDEV_TX_OK;
316}
317
318static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
319{
320 switch (ipproto) {
321 case 0:
322 case IPPROTO_IPIP:
323#if IS_ENABLED(CONFIG_MPLS)
324 case IPPROTO_MPLS:
325#endif
326 return true;
327 }
328
329 return false;
330}
331
332static int
333ipip_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
334{
335 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
336 if (p->iph.version != 4 ||
337 !ipip_tunnel_ioctl_verify_protocol(p->iph.protocol) ||
338 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
339 return -EINVAL;
340 }
341
342 p->i_key = p->o_key = 0;
343 p->i_flags = p->o_flags = 0;
344 return ip_tunnel_ctl(dev, p, cmd);
345}
346
347static const struct net_device_ops ipip_netdev_ops = {
348 .ndo_init = ipip_tunnel_init,
349 .ndo_uninit = ip_tunnel_uninit,
350 .ndo_start_xmit = ipip_tunnel_xmit,
351 .ndo_siocdevprivate = ip_tunnel_siocdevprivate,
352 .ndo_change_mtu = ip_tunnel_change_mtu,
353 .ndo_get_stats64 = dev_get_tstats64,
354 .ndo_get_iflink = ip_tunnel_get_iflink,
355 .ndo_tunnel_ctl = ipip_tunnel_ctl,
356};
357
358#define IPIP_FEATURES (NETIF_F_SG | \
359 NETIF_F_FRAGLIST | \
360 NETIF_F_HIGHDMA | \
361 NETIF_F_GSO_SOFTWARE | \
362 NETIF_F_HW_CSUM)
363
364static void ipip_tunnel_setup(struct net_device *dev)
365{
366 dev->netdev_ops = &ipip_netdev_ops;
367 dev->header_ops = &ip_tunnel_header_ops;
368
369 dev->type = ARPHRD_TUNNEL;
370 dev->flags = IFF_NOARP;
371 dev->addr_len = 4;
372 dev->features |= NETIF_F_LLTX;
373 netif_keep_dst(dev);
374
375 dev->features |= IPIP_FEATURES;
376 dev->hw_features |= IPIP_FEATURES;
377 ip_tunnel_setup(dev, ipip_net_id);
378}
379
380static int ipip_tunnel_init(struct net_device *dev)
381{
382 struct ip_tunnel *tunnel = netdev_priv(dev);
383
384 __dev_addr_set(dev, &tunnel->parms.iph.saddr, 4);
385 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
386
387 tunnel->tun_hlen = 0;
388 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
389 return ip_tunnel_init(dev);
390}
391
392static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
393 struct netlink_ext_ack *extack)
394{
395 u8 proto;
396
397 if (!data || !data[IFLA_IPTUN_PROTO])
398 return 0;
399
400 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
401 if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
402 return -EINVAL;
403
404 return 0;
405}
406
407static void ipip_netlink_parms(struct nlattr *data[],
408 struct ip_tunnel_parm *parms, bool *collect_md,
409 __u32 *fwmark)
410{
411 memset(parms, 0, sizeof(*parms));
412
413 parms->iph.version = 4;
414 parms->iph.protocol = IPPROTO_IPIP;
415 parms->iph.ihl = 5;
416 *collect_md = false;
417
418 if (!data)
419 return;
420
421 ip_tunnel_netlink_parms(data, parms);
422
423 if (data[IFLA_IPTUN_COLLECT_METADATA])
424 *collect_md = true;
425
426 if (data[IFLA_IPTUN_FWMARK])
427 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
428}
429
430static int ipip_newlink(struct net *src_net, struct net_device *dev,
431 struct nlattr *tb[], struct nlattr *data[],
432 struct netlink_ext_ack *extack)
433{
434 struct ip_tunnel *t = netdev_priv(dev);
435 struct ip_tunnel_parm p;
436 struct ip_tunnel_encap ipencap;
437 __u32 fwmark = 0;
438
439 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
440 int err = ip_tunnel_encap_setup(t, &ipencap);
441
442 if (err < 0)
443 return err;
444 }
445
446 ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
447 return ip_tunnel_newlink(dev, tb, &p, fwmark);
448}
449
450static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
451 struct nlattr *data[],
452 struct netlink_ext_ack *extack)
453{
454 struct ip_tunnel *t = netdev_priv(dev);
455 struct ip_tunnel_parm p;
456 struct ip_tunnel_encap ipencap;
457 bool collect_md;
458 __u32 fwmark = t->fwmark;
459
460 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
461 int err = ip_tunnel_encap_setup(t, &ipencap);
462
463 if (err < 0)
464 return err;
465 }
466
467 ipip_netlink_parms(data, &p, &collect_md, &fwmark);
468 if (collect_md)
469 return -EINVAL;
470
471 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
472 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
473 return -EINVAL;
474
475 return ip_tunnel_changelink(dev, tb, &p, fwmark);
476}
477
478static size_t ipip_get_size(const struct net_device *dev)
479{
480 return
481 /* IFLA_IPTUN_LINK */
482 nla_total_size(4) +
483 /* IFLA_IPTUN_LOCAL */
484 nla_total_size(4) +
485 /* IFLA_IPTUN_REMOTE */
486 nla_total_size(4) +
487 /* IFLA_IPTUN_TTL */
488 nla_total_size(1) +
489 /* IFLA_IPTUN_TOS */
490 nla_total_size(1) +
491 /* IFLA_IPTUN_PROTO */
492 nla_total_size(1) +
493 /* IFLA_IPTUN_PMTUDISC */
494 nla_total_size(1) +
495 /* IFLA_IPTUN_ENCAP_TYPE */
496 nla_total_size(2) +
497 /* IFLA_IPTUN_ENCAP_FLAGS */
498 nla_total_size(2) +
499 /* IFLA_IPTUN_ENCAP_SPORT */
500 nla_total_size(2) +
501 /* IFLA_IPTUN_ENCAP_DPORT */
502 nla_total_size(2) +
503 /* IFLA_IPTUN_COLLECT_METADATA */
504 nla_total_size(0) +
505 /* IFLA_IPTUN_FWMARK */
506 nla_total_size(4) +
507 0;
508}
509
510static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
511{
512 struct ip_tunnel *tunnel = netdev_priv(dev);
513 struct ip_tunnel_parm *parm = &tunnel->parms;
514
515 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
516 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
517 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
518 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
519 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
520 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
521 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
522 !!(parm->iph.frag_off & htons(IP_DF))) ||
523 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
524 goto nla_put_failure;
525
526 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
527 tunnel->encap.type) ||
528 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
529 tunnel->encap.sport) ||
530 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
531 tunnel->encap.dport) ||
532 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
533 tunnel->encap.flags))
534 goto nla_put_failure;
535
536 if (tunnel->collect_md)
537 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
538 goto nla_put_failure;
539 return 0;
540
541nla_put_failure:
542 return -EMSGSIZE;
543}
544
545static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
546 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
547 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 },
548 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 },
549 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
550 [IFLA_IPTUN_TOS] = { .type = NLA_U8 },
551 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
552 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 },
553 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
554 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
555 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
556 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
557 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
558 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
559};
560
561static struct rtnl_link_ops ipip_link_ops __read_mostly = {
562 .kind = "ipip",
563 .maxtype = IFLA_IPTUN_MAX,
564 .policy = ipip_policy,
565 .priv_size = sizeof(struct ip_tunnel),
566 .setup = ipip_tunnel_setup,
567 .validate = ipip_tunnel_validate,
568 .newlink = ipip_newlink,
569 .changelink = ipip_changelink,
570 .dellink = ip_tunnel_dellink,
571 .get_size = ipip_get_size,
572 .fill_info = ipip_fill_info,
573 .get_link_net = ip_tunnel_get_link_net,
574};
575
576static struct xfrm_tunnel ipip_handler __read_mostly = {
577 .handler = ipip_rcv,
578 .err_handler = ipip_err,
579 .priority = 1,
580};
581
582#if IS_ENABLED(CONFIG_MPLS)
583static struct xfrm_tunnel mplsip_handler __read_mostly = {
584 .handler = mplsip_rcv,
585 .err_handler = ipip_err,
586 .priority = 1,
587};
588#endif
589
590static int __net_init ipip_init_net(struct net *net)
591{
592 return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
593}
594
595static void __net_exit ipip_exit_batch_rtnl(struct list_head *list_net,
596 struct list_head *dev_to_kill)
597{
598 ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops,
599 dev_to_kill);
600}
601
602static struct pernet_operations ipip_net_ops = {
603 .init = ipip_init_net,
604 .exit_batch_rtnl = ipip_exit_batch_rtnl,
605 .id = &ipip_net_id,
606 .size = sizeof(struct ip_tunnel_net),
607};
608
609static int __init ipip_init(void)
610{
611 int err;
612
613 pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
614
615 err = register_pernet_device(&ipip_net_ops);
616 if (err < 0)
617 return err;
618 err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
619 if (err < 0) {
620 pr_info("%s: can't register tunnel\n", __func__);
621 goto xfrm_tunnel_ipip_failed;
622 }
623#if IS_ENABLED(CONFIG_MPLS)
624 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
625 if (err < 0) {
626 pr_info("%s: can't register tunnel\n", __func__);
627 goto xfrm_tunnel_mplsip_failed;
628 }
629#endif
630 err = rtnl_link_register(&ipip_link_ops);
631 if (err < 0)
632 goto rtnl_link_failed;
633
634out:
635 return err;
636
637rtnl_link_failed:
638#if IS_ENABLED(CONFIG_MPLS)
639 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
640xfrm_tunnel_mplsip_failed:
641
642#endif
643 xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
644xfrm_tunnel_ipip_failed:
645 unregister_pernet_device(&ipip_net_ops);
646 goto out;
647}
648
649static void __exit ipip_fini(void)
650{
651 rtnl_link_unregister(&ipip_link_ops);
652 if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
653 pr_info("%s: can't deregister tunnel\n", __func__);
654#if IS_ENABLED(CONFIG_MPLS)
655 if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
656 pr_info("%s: can't deregister tunnel\n", __func__);
657#endif
658 unregister_pernet_device(&ipip_net_ops);
659}
660
661module_init(ipip_init);
662module_exit(ipip_fini);
663MODULE_DESCRIPTION("IP/IP protocol decoder library");
664MODULE_LICENSE("GPL");
665MODULE_ALIAS_RTNL_LINK("ipip");
666MODULE_ALIAS_NETDEV("tunl0");