xref: /openbmc/linux/net/ipv4/ipip.c (revision 39b6f3aa)
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/mroute.h>
107 #include <linux/init.h>
108 #include <linux/netfilter_ipv4.h>
109 #include <linux/if_ether.h>
110 
111 #include <net/sock.h>
112 #include <net/ip.h>
113 #include <net/icmp.h>
114 #include <net/ip_tunnels.h>
115 #include <net/inet_ecn.h>
116 #include <net/xfrm.h>
117 #include <net/net_namespace.h>
118 #include <net/netns/generic.h>
119 
120 static bool log_ecn_error = true;
121 module_param(log_ecn_error, bool, 0644);
122 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
123 
124 static int ipip_net_id __read_mostly;
125 
126 static int ipip_tunnel_init(struct net_device *dev);
127 static struct rtnl_link_ops ipip_link_ops __read_mostly;
128 
129 static int ipip_err(struct sk_buff *skb, u32 info)
130 {
131 
132 /* All the routers (except for Linux) return only
133    8 bytes of packet payload. It means, that precise relaying of
134    ICMP in the real Internet is absolutely infeasible.
135  */
136 	struct net *net = dev_net(skb->dev);
137 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
138 	const struct iphdr *iph = (const struct iphdr *)skb->data;
139 	struct ip_tunnel *t;
140 	int err;
141 	const int type = icmp_hdr(skb)->type;
142 	const int code = icmp_hdr(skb)->code;
143 
144 	err = -ENOENT;
145 	t = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
146 			     iph->daddr, iph->saddr, 0);
147 	if (t == NULL)
148 		goto out;
149 
150 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
151 		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
152 				 t->dev->ifindex, 0, IPPROTO_IPIP, 0);
153 		err = 0;
154 		goto out;
155 	}
156 
157 	if (type == ICMP_REDIRECT) {
158 		ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
159 			      IPPROTO_IPIP, 0);
160 		err = 0;
161 		goto out;
162 	}
163 
164 	if (t->parms.iph.daddr == 0)
165 		goto out;
166 
167 	err = 0;
168 	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
169 		goto out;
170 
171 	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
172 		t->err_count++;
173 	else
174 		t->err_count = 1;
175 	t->err_time = jiffies;
176 
177 out:
178 	return err;
179 }
180 
181 static const struct tnl_ptk_info tpi = {
182 	/* no tunnel info required for ipip. */
183 	.proto = htons(ETH_P_IP),
184 };
185 
186 static int ipip_rcv(struct sk_buff *skb)
187 {
188 	struct net *net = dev_net(skb->dev);
189 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
190 	struct ip_tunnel *tunnel;
191 	const struct iphdr *iph = ip_hdr(skb);
192 
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 		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
199 	}
200 
201 	return -1;
202 
203 drop:
204 	kfree_skb(skb);
205 	return 0;
206 }
207 
208 /*
209  *	This function assumes it is being called from dev_queue_xmit()
210  *	and that skb is filled properly by that function.
211  */
212 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
213 {
214 	struct ip_tunnel *tunnel = netdev_priv(dev);
215 	const struct iphdr  *tiph = &tunnel->parms.iph;
216 
217 	if (unlikely(skb->protocol != htons(ETH_P_IP)))
218 		goto tx_error;
219 
220 	if (likely(!skb->encapsulation)) {
221 		skb_reset_inner_headers(skb);
222 		skb->encapsulation = 1;
223 	}
224 
225 	ip_tunnel_xmit(skb, dev, tiph);
226 	return NETDEV_TX_OK;
227 
228 tx_error:
229 	dev->stats.tx_errors++;
230 	dev_kfree_skb(skb);
231 	return NETDEV_TX_OK;
232 }
233 
234 static int
235 ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
236 {
237 	int err = 0;
238 	struct ip_tunnel_parm p;
239 
240 	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
241 		return -EFAULT;
242 
243 	if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
244 			p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
245 		return -EINVAL;
246 	if (p.i_key || p.o_key || p.i_flags || p.o_flags)
247 		return -EINVAL;
248 	if (p.iph.ttl)
249 		p.iph.frag_off |= htons(IP_DF);
250 
251 	err = ip_tunnel_ioctl(dev, &p, cmd);
252 	if (err)
253 		return err;
254 
255 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
256 		return -EFAULT;
257 
258 	return 0;
259 }
260 
261 static const struct net_device_ops ipip_netdev_ops = {
262 	.ndo_init       = ipip_tunnel_init,
263 	.ndo_uninit     = ip_tunnel_uninit,
264 	.ndo_start_xmit	= ipip_tunnel_xmit,
265 	.ndo_do_ioctl	= ipip_tunnel_ioctl,
266 	.ndo_change_mtu = ip_tunnel_change_mtu,
267 	.ndo_get_stats64 = ip_tunnel_get_stats64,
268 };
269 
270 #define IPIP_FEATURES (NETIF_F_SG |		\
271 		       NETIF_F_FRAGLIST |	\
272 		       NETIF_F_HIGHDMA |	\
273 		       NETIF_F_HW_CSUM)
274 
275 static void ipip_tunnel_setup(struct net_device *dev)
276 {
277 	dev->netdev_ops		= &ipip_netdev_ops;
278 
279 	dev->type		= ARPHRD_TUNNEL;
280 	dev->flags		= IFF_NOARP;
281 	dev->iflink		= 0;
282 	dev->addr_len		= 4;
283 	dev->features		|= NETIF_F_NETNS_LOCAL;
284 	dev->features		|= NETIF_F_LLTX;
285 	dev->priv_flags		&= ~IFF_XMIT_DST_RELEASE;
286 
287 	dev->features		|= IPIP_FEATURES;
288 	dev->hw_features	|= IPIP_FEATURES;
289 	ip_tunnel_setup(dev, ipip_net_id);
290 }
291 
292 static int ipip_tunnel_init(struct net_device *dev)
293 {
294 	struct ip_tunnel *tunnel = netdev_priv(dev);
295 
296 	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
297 	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
298 
299 	tunnel->hlen = 0;
300 	tunnel->parms.iph.protocol = IPPROTO_IPIP;
301 	return ip_tunnel_init(dev);
302 }
303 
304 static void ipip_netlink_parms(struct nlattr *data[],
305 			       struct ip_tunnel_parm *parms)
306 {
307 	memset(parms, 0, sizeof(*parms));
308 
309 	parms->iph.version = 4;
310 	parms->iph.protocol = IPPROTO_IPIP;
311 	parms->iph.ihl = 5;
312 
313 	if (!data)
314 		return;
315 
316 	if (data[IFLA_IPTUN_LINK])
317 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
318 
319 	if (data[IFLA_IPTUN_LOCAL])
320 		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
321 
322 	if (data[IFLA_IPTUN_REMOTE])
323 		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
324 
325 	if (data[IFLA_IPTUN_TTL]) {
326 		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
327 		if (parms->iph.ttl)
328 			parms->iph.frag_off = htons(IP_DF);
329 	}
330 
331 	if (data[IFLA_IPTUN_TOS])
332 		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
333 
334 	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
335 		parms->iph.frag_off = htons(IP_DF);
336 }
337 
338 static int ipip_newlink(struct net *src_net, struct net_device *dev,
339 			struct nlattr *tb[], struct nlattr *data[])
340 {
341 	struct ip_tunnel_parm p;
342 
343 	ipip_netlink_parms(data, &p);
344 	return ip_tunnel_newlink(dev, tb, &p);
345 }
346 
347 static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
348 			   struct nlattr *data[])
349 {
350 	struct ip_tunnel_parm p;
351 
352 	ipip_netlink_parms(data, &p);
353 
354 	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
355 	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
356 		return -EINVAL;
357 
358 	return ip_tunnel_changelink(dev, tb, &p);
359 }
360 
361 static size_t ipip_get_size(const struct net_device *dev)
362 {
363 	return
364 		/* IFLA_IPTUN_LINK */
365 		nla_total_size(4) +
366 		/* IFLA_IPTUN_LOCAL */
367 		nla_total_size(4) +
368 		/* IFLA_IPTUN_REMOTE */
369 		nla_total_size(4) +
370 		/* IFLA_IPTUN_TTL */
371 		nla_total_size(1) +
372 		/* IFLA_IPTUN_TOS */
373 		nla_total_size(1) +
374 		/* IFLA_IPTUN_PMTUDISC */
375 		nla_total_size(1) +
376 		0;
377 }
378 
379 static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
380 {
381 	struct ip_tunnel *tunnel = netdev_priv(dev);
382 	struct ip_tunnel_parm *parm = &tunnel->parms;
383 
384 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
385 	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
386 	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
387 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
388 	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
389 	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
390 		       !!(parm->iph.frag_off & htons(IP_DF))))
391 		goto nla_put_failure;
392 	return 0;
393 
394 nla_put_failure:
395 	return -EMSGSIZE;
396 }
397 
398 static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
399 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
400 	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
401 	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
402 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
403 	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
404 	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
405 };
406 
407 static struct rtnl_link_ops ipip_link_ops __read_mostly = {
408 	.kind		= "ipip",
409 	.maxtype	= IFLA_IPTUN_MAX,
410 	.policy		= ipip_policy,
411 	.priv_size	= sizeof(struct ip_tunnel),
412 	.setup		= ipip_tunnel_setup,
413 	.newlink	= ipip_newlink,
414 	.changelink	= ipip_changelink,
415 	.dellink	= ip_tunnel_dellink,
416 	.get_size	= ipip_get_size,
417 	.fill_info	= ipip_fill_info,
418 };
419 
420 static struct xfrm_tunnel ipip_handler __read_mostly = {
421 	.handler	=	ipip_rcv,
422 	.err_handler	=	ipip_err,
423 	.priority	=	1,
424 };
425 
426 static int __net_init ipip_init_net(struct net *net)
427 {
428 	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
429 }
430 
431 static void __net_exit ipip_exit_net(struct net *net)
432 {
433 	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
434 	ip_tunnel_delete_net(itn);
435 }
436 
437 static struct pernet_operations ipip_net_ops = {
438 	.init = ipip_init_net,
439 	.exit = ipip_exit_net,
440 	.id   = &ipip_net_id,
441 	.size = sizeof(struct ip_tunnel_net),
442 };
443 
444 static int __init ipip_init(void)
445 {
446 	int err;
447 
448 	pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
449 
450 	err = register_pernet_device(&ipip_net_ops);
451 	if (err < 0)
452 		return err;
453 	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
454 	if (err < 0) {
455 		pr_info("%s: can't register tunnel\n", __func__);
456 		goto xfrm_tunnel_failed;
457 	}
458 	err = rtnl_link_register(&ipip_link_ops);
459 	if (err < 0)
460 		goto rtnl_link_failed;
461 
462 out:
463 	return err;
464 
465 rtnl_link_failed:
466 	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
467 xfrm_tunnel_failed:
468 	unregister_pernet_device(&ipip_net_ops);
469 	goto out;
470 }
471 
472 static void __exit ipip_fini(void)
473 {
474 	rtnl_link_unregister(&ipip_link_ops);
475 	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
476 		pr_info("%s: can't deregister tunnel\n", __func__);
477 
478 	unregister_pernet_device(&ipip_net_ops);
479 }
480 
481 module_init(ipip_init);
482 module_exit(ipip_fini);
483 MODULE_LICENSE("GPL");
484 MODULE_ALIAS_NETDEV("tunl0");
485