Linux Audio

Check our new training course

Linux BSP development engineering services

Need help to port Linux and bootloaders to your hardware?
Loading...
v3.15
 
  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
120static bool log_ecn_error = true;
121module_param(log_ecn_error, bool, 0644);
122MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
123
124static int ipip_net_id __read_mostly;
125
126static int ipip_tunnel_init(struct net_device *dev);
127static struct rtnl_link_ops ipip_link_ops __read_mostly;
128
129static 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
177out:
178	return err;
179}
180
181static const struct tnl_ptk_info tpi = {
182	/* no tunnel info required for ipip. */
183	.proto = htons(ETH_P_IP),
184};
185
186static 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;
192
193	iph = ip_hdr(skb);
194	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
195			iph->saddr, iph->daddr, 0);
196	if (tunnel) {
 
 
 
 
 
 
197		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
198			goto drop;
199		if (iptunnel_pull_header(skb, 0, tpi.proto))
 
 
 
 
 
 
200			goto drop;
201		return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 
 
 
 
 
202	}
203
204	return -1;
205
206drop:
207	kfree_skb(skb);
208	return 0;
209}
210
 
 
 
 
 
 
 
 
 
 
 
 
211/*
212 *	This function assumes it is being called from dev_queue_xmit()
213 *	and that skb is filled properly by that function.
214 */
215static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 
216{
217	struct ip_tunnel *tunnel = netdev_priv(dev);
218	const struct iphdr  *tiph = &tunnel->parms.iph;
 
219
220	if (unlikely(skb->protocol != htons(ETH_P_IP)))
221		goto tx_error;
222
223	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
224	if (IS_ERR(skb))
225		goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
226
227	ip_tunnel_xmit(skb, dev, tiph, tiph->protocol);
 
 
 
228	return NETDEV_TX_OK;
229
230tx_error:
231	kfree_skb(skb);
232out:
233	dev->stats.tx_errors++;
234	return NETDEV_TX_OK;
235}
236
 
 
 
 
 
 
 
 
 
 
 
 
 
 
237static int
238ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
239{
240	int err = 0;
241	struct ip_tunnel_parm p;
242
243	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
244		return -EFAULT;
245
246	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
247		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
 
248		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
249			return -EINVAL;
250	}
251
252	p.i_key = p.o_key = p.i_flags = p.o_flags = 0;
253	if (p.iph.ttl)
254		p.iph.frag_off |= htons(IP_DF);
255
256	err = ip_tunnel_ioctl(dev, &p, cmd);
257	if (err)
258		return err;
259
260	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
261		return -EFAULT;
262
263	return 0;
264}
265
266static const struct net_device_ops ipip_netdev_ops = {
267	.ndo_init       = ipip_tunnel_init,
268	.ndo_uninit     = ip_tunnel_uninit,
269	.ndo_start_xmit	= ipip_tunnel_xmit,
270	.ndo_do_ioctl	= ipip_tunnel_ioctl,
271	.ndo_change_mtu = ip_tunnel_change_mtu,
272	.ndo_get_stats64 = ip_tunnel_get_stats64,
 
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->iflink		= 0;
288	dev->addr_len		= 4;
289	dev->features		|= NETIF_F_LLTX;
290	dev->priv_flags		&= ~IFF_XMIT_DST_RELEASE;
291
292	dev->features		|= IPIP_FEATURES;
293	dev->hw_features	|= IPIP_FEATURES;
294	ip_tunnel_setup(dev, ipip_net_id);
295}
296
297static int ipip_tunnel_init(struct net_device *dev)
298{
299	struct ip_tunnel *tunnel = netdev_priv(dev);
300
301	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
302	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
303
304	tunnel->hlen = 0;
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_be32(data[IFLA_IPTUN_LOCAL]);
326
327	if (data[IFLA_IPTUN_REMOTE])
328		parms->iph.daddr = nla_get_be32(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
343static int ipip_newlink(struct net *src_net, struct net_device *dev,
344			struct nlattr *tb[], struct nlattr *data[])
 
345{
 
346	struct ip_tunnel_parm p;
 
 
347
348	ipip_netlink_parms(data, &p);
349	return ip_tunnel_newlink(dev, tb, &p);
 
 
 
 
 
 
 
350}
351
352static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
353			   struct nlattr *data[])
 
354{
 
355	struct ip_tunnel_parm p;
 
 
 
 
 
 
 
 
 
 
356
357	ipip_netlink_parms(data, &p);
 
 
358
359	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
360	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
361		return -EINVAL;
362
363	return ip_tunnel_changelink(dev, tb, &p);
364}
365
366static size_t ipip_get_size(const struct net_device *dev)
367{
368	return
369		/* IFLA_IPTUN_LINK */
370		nla_total_size(4) +
371		/* IFLA_IPTUN_LOCAL */
372		nla_total_size(4) +
373		/* IFLA_IPTUN_REMOTE */
374		nla_total_size(4) +
375		/* IFLA_IPTUN_TTL */
376		nla_total_size(1) +
377		/* IFLA_IPTUN_TOS */
378		nla_total_size(1) +
 
 
379		/* IFLA_IPTUN_PMTUDISC */
380		nla_total_size(1) +
 
 
 
 
 
 
 
 
 
 
 
 
381		0;
382}
383
384static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
385{
386	struct ip_tunnel *tunnel = netdev_priv(dev);
387	struct ip_tunnel_parm *parm = &tunnel->parms;
388
389	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
390	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
391	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
392	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
393	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
 
394	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
395		       !!(parm->iph.frag_off & htons(IP_DF))))
 
396		goto nla_put_failure;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
397	return 0;
398
399nla_put_failure:
400	return -EMSGSIZE;
401}
402
403static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
404	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
405	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
406	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
407	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
408	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
 
409	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
 
 
 
 
 
 
410};
411
412static struct rtnl_link_ops ipip_link_ops __read_mostly = {
413	.kind		= "ipip",
414	.maxtype	= IFLA_IPTUN_MAX,
415	.policy		= ipip_policy,
416	.priv_size	= sizeof(struct ip_tunnel),
417	.setup		= ipip_tunnel_setup,
 
418	.newlink	= ipip_newlink,
419	.changelink	= ipip_changelink,
420	.dellink	= ip_tunnel_dellink,
421	.get_size	= ipip_get_size,
422	.fill_info	= ipip_fill_info,
 
423};
424
425static struct xfrm_tunnel ipip_handler __read_mostly = {
426	.handler	=	ipip_rcv,
427	.err_handler	=	ipip_err,
428	.priority	=	1,
429};
430
 
 
 
 
 
 
 
 
431static int __net_init ipip_init_net(struct net *net)
432{
433	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
434}
435
436static void __net_exit ipip_exit_net(struct net *net)
437{
438	struct ip_tunnel_net *itn = net_generic(net, ipip_net_id);
439	ip_tunnel_delete_net(itn, &ipip_link_ops);
440}
441
442static struct pernet_operations ipip_net_ops = {
443	.init = ipip_init_net,
444	.exit = ipip_exit_net,
445	.id   = &ipip_net_id,
446	.size = sizeof(struct ip_tunnel_net),
447};
448
449static int __init ipip_init(void)
450{
451	int err;
452
453	pr_info("ipip: IPv4 over IPv4 tunneling driver\n");
454
455	err = register_pernet_device(&ipip_net_ops);
456	if (err < 0)
457		return err;
458	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
459	if (err < 0) {
460		pr_info("%s: can't register tunnel\n", __func__);
461		goto xfrm_tunnel_failed;
462	}
 
 
 
 
 
 
 
463	err = rtnl_link_register(&ipip_link_ops);
464	if (err < 0)
465		goto rtnl_link_failed;
466
467out:
468	return err;
469
470rtnl_link_failed:
 
 
 
 
 
471	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
472xfrm_tunnel_failed:
473	unregister_pernet_device(&ipip_net_ops);
474	goto out;
475}
476
477static void __exit ipip_fini(void)
478{
479	rtnl_link_unregister(&ipip_link_ops);
480	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
481		pr_info("%s: can't deregister tunnel\n", __func__);
482
 
 
 
483	unregister_pernet_device(&ipip_net_ops);
484}
485
486module_init(ipip_init);
487module_exit(ipip_fini);
488MODULE_LICENSE("GPL");
 
489MODULE_ALIAS_NETDEV("tunl0");
v5.4
  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		}
245		return ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
246	}
247
248	return -1;
249
250drop:
251	kfree_skb(skb);
252	return 0;
253}
254
255static int ipip_rcv(struct sk_buff *skb)
256{
257	return ipip_tunnel_rcv(skb, IPPROTO_IPIP);
258}
259
260#if IS_ENABLED(CONFIG_MPLS)
261static int mplsip_rcv(struct sk_buff *skb)
262{
263	return ipip_tunnel_rcv(skb, IPPROTO_MPLS);
264}
265#endif
266
267/*
268 *	This function assumes it is being called from dev_queue_xmit()
269 *	and that skb is filled properly by that function.
270 */
271static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb,
272				    struct net_device *dev)
273{
274	struct ip_tunnel *tunnel = netdev_priv(dev);
275	const struct iphdr  *tiph = &tunnel->parms.iph;
276	u8 ipproto;
277
278	if (!pskb_inet_may_pull(skb))
279		goto tx_error;
280
281	switch (skb->protocol) {
282	case htons(ETH_P_IP):
283		ipproto = IPPROTO_IPIP;
284		break;
285#if IS_ENABLED(CONFIG_MPLS)
286	case htons(ETH_P_MPLS_UC):
287		ipproto = IPPROTO_MPLS;
288		break;
289#endif
290	default:
291		goto tx_error;
292	}
293
294	if (tiph->protocol != ipproto && tiph->protocol != 0)
295		goto tx_error;
296
297	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
298		goto tx_error;
299
300	skb_set_inner_ipproto(skb, ipproto);
301
302	if (tunnel->collect_md)
303		ip_md_tunnel_xmit(skb, dev, ipproto, 0);
304	else
305		ip_tunnel_xmit(skb, dev, tiph, ipproto);
306	return NETDEV_TX_OK;
307
308tx_error:
309	kfree_skb(skb);
310
311	dev->stats.tx_errors++;
312	return NETDEV_TX_OK;
313}
314
315static bool ipip_tunnel_ioctl_verify_protocol(u8 ipproto)
316{
317	switch (ipproto) {
318	case 0:
319	case IPPROTO_IPIP:
320#if IS_ENABLED(CONFIG_MPLS)
321	case IPPROTO_MPLS:
322#endif
323		return true;
324	}
325
326	return false;
327}
328
329static int
330ipip_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
331{
332	int err = 0;
333	struct ip_tunnel_parm p;
334
335	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
336		return -EFAULT;
337
338	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
339		if (p.iph.version != 4 ||
340		    !ipip_tunnel_ioctl_verify_protocol(p.iph.protocol) ||
341		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
342			return -EINVAL;
343	}
344
345	p.i_key = p.o_key = 0;
346	p.i_flags = p.o_flags = 0;
 
 
347	err = ip_tunnel_ioctl(dev, &p, cmd);
348	if (err)
349		return err;
350
351	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
352		return -EFAULT;
353
354	return 0;
355}
356
357static const struct net_device_ops ipip_netdev_ops = {
358	.ndo_init       = ipip_tunnel_init,
359	.ndo_uninit     = ip_tunnel_uninit,
360	.ndo_start_xmit	= ipip_tunnel_xmit,
361	.ndo_do_ioctl	= ipip_tunnel_ioctl,
362	.ndo_change_mtu = ip_tunnel_change_mtu,
363	.ndo_get_stats64 = ip_tunnel_get_stats64,
364	.ndo_get_iflink = ip_tunnel_get_iflink,
365};
366
367#define IPIP_FEATURES (NETIF_F_SG |		\
368		       NETIF_F_FRAGLIST |	\
369		       NETIF_F_HIGHDMA |	\
370		       NETIF_F_GSO_SOFTWARE |	\
371		       NETIF_F_HW_CSUM)
372
373static void ipip_tunnel_setup(struct net_device *dev)
374{
375	dev->netdev_ops		= &ipip_netdev_ops;
376
377	dev->type		= ARPHRD_TUNNEL;
378	dev->flags		= IFF_NOARP;
 
379	dev->addr_len		= 4;
380	dev->features		|= NETIF_F_LLTX;
381	netif_keep_dst(dev);
382
383	dev->features		|= IPIP_FEATURES;
384	dev->hw_features	|= IPIP_FEATURES;
385	ip_tunnel_setup(dev, ipip_net_id);
386}
387
388static int ipip_tunnel_init(struct net_device *dev)
389{
390	struct ip_tunnel *tunnel = netdev_priv(dev);
391
392	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
393	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
394
395	tunnel->tun_hlen = 0;
396	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
397	return ip_tunnel_init(dev);
398}
399
400static int ipip_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
401				struct netlink_ext_ack *extack)
402{
403	u8 proto;
404
405	if (!data || !data[IFLA_IPTUN_PROTO])
406		return 0;
407
408	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
409	if (proto != IPPROTO_IPIP && proto != IPPROTO_MPLS && proto != 0)
410		return -EINVAL;
411
412	return 0;
413}
414
415static void ipip_netlink_parms(struct nlattr *data[],
416			       struct ip_tunnel_parm *parms, bool *collect_md,
417			       __u32 *fwmark)
418{
419	memset(parms, 0, sizeof(*parms));
420
421	parms->iph.version = 4;
422	parms->iph.protocol = IPPROTO_IPIP;
423	parms->iph.ihl = 5;
424	*collect_md = false;
425
426	if (!data)
427		return;
428
429	if (data[IFLA_IPTUN_LINK])
430		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
431
432	if (data[IFLA_IPTUN_LOCAL])
433		parms->iph.saddr = nla_get_in_addr(data[IFLA_IPTUN_LOCAL]);
434
435	if (data[IFLA_IPTUN_REMOTE])
436		parms->iph.daddr = nla_get_in_addr(data[IFLA_IPTUN_REMOTE]);
437
438	if (data[IFLA_IPTUN_TTL]) {
439		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
440		if (parms->iph.ttl)
441			parms->iph.frag_off = htons(IP_DF);
442	}
443
444	if (data[IFLA_IPTUN_TOS])
445		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
446
447	if (data[IFLA_IPTUN_PROTO])
448		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
449
450	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
451		parms->iph.frag_off = htons(IP_DF);
452
453	if (data[IFLA_IPTUN_COLLECT_METADATA])
454		*collect_md = true;
455
456	if (data[IFLA_IPTUN_FWMARK])
457		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
458}
459
460/* This function returns true when ENCAP attributes are present in the nl msg */
461static bool ipip_netlink_encap_parms(struct nlattr *data[],
462				     struct ip_tunnel_encap *ipencap)
463{
464	bool ret = false;
465
466	memset(ipencap, 0, sizeof(*ipencap));
467
468	if (!data)
469		return ret;
470
471	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
472		ret = true;
473		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
474	}
475
476	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
477		ret = true;
478		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
479	}
480
481	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
482		ret = true;
483		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
484	}
485
486	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
487		ret = true;
488		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
489	}
490
491	return ret;
492}
493
494static int ipip_newlink(struct net *src_net, struct net_device *dev,
495			struct nlattr *tb[], struct nlattr *data[],
496			struct netlink_ext_ack *extack)
497{
498	struct ip_tunnel *t = netdev_priv(dev);
499	struct ip_tunnel_parm p;
500	struct ip_tunnel_encap ipencap;
501	__u32 fwmark = 0;
502
503	if (ipip_netlink_encap_parms(data, &ipencap)) {
504		int err = ip_tunnel_encap_setup(t, &ipencap);
505
506		if (err < 0)
507			return err;
508	}
509
510	ipip_netlink_parms(data, &p, &t->collect_md, &fwmark);
511	return ip_tunnel_newlink(dev, tb, &p, fwmark);
512}
513
514static int ipip_changelink(struct net_device *dev, struct nlattr *tb[],
515			   struct nlattr *data[],
516			   struct netlink_ext_ack *extack)
517{
518	struct ip_tunnel *t = netdev_priv(dev);
519	struct ip_tunnel_parm p;
520	struct ip_tunnel_encap ipencap;
521	bool collect_md;
522	__u32 fwmark = t->fwmark;
523
524	if (ipip_netlink_encap_parms(data, &ipencap)) {
525		int err = ip_tunnel_encap_setup(t, &ipencap);
526
527		if (err < 0)
528			return err;
529	}
530
531	ipip_netlink_parms(data, &p, &collect_md, &fwmark);
532	if (collect_md)
533		return -EINVAL;
534
535	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
536	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
537		return -EINVAL;
538
539	return ip_tunnel_changelink(dev, tb, &p, fwmark);
540}
541
542static size_t ipip_get_size(const struct net_device *dev)
543{
544	return
545		/* IFLA_IPTUN_LINK */
546		nla_total_size(4) +
547		/* IFLA_IPTUN_LOCAL */
548		nla_total_size(4) +
549		/* IFLA_IPTUN_REMOTE */
550		nla_total_size(4) +
551		/* IFLA_IPTUN_TTL */
552		nla_total_size(1) +
553		/* IFLA_IPTUN_TOS */
554		nla_total_size(1) +
555		/* IFLA_IPTUN_PROTO */
556		nla_total_size(1) +
557		/* IFLA_IPTUN_PMTUDISC */
558		nla_total_size(1) +
559		/* IFLA_IPTUN_ENCAP_TYPE */
560		nla_total_size(2) +
561		/* IFLA_IPTUN_ENCAP_FLAGS */
562		nla_total_size(2) +
563		/* IFLA_IPTUN_ENCAP_SPORT */
564		nla_total_size(2) +
565		/* IFLA_IPTUN_ENCAP_DPORT */
566		nla_total_size(2) +
567		/* IFLA_IPTUN_COLLECT_METADATA */
568		nla_total_size(0) +
569		/* IFLA_IPTUN_FWMARK */
570		nla_total_size(4) +
571		0;
572}
573
574static int ipip_fill_info(struct sk_buff *skb, const struct net_device *dev)
575{
576	struct ip_tunnel *tunnel = netdev_priv(dev);
577	struct ip_tunnel_parm *parm = &tunnel->parms;
578
579	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
580	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
581	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
582	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
583	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
584	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
585	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
586		       !!(parm->iph.frag_off & htons(IP_DF))) ||
587	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
588		goto nla_put_failure;
589
590	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
591			tunnel->encap.type) ||
592	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
593			 tunnel->encap.sport) ||
594	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
595			 tunnel->encap.dport) ||
596	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
597			tunnel->encap.flags))
598		goto nla_put_failure;
599
600	if (tunnel->collect_md)
601		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
602			goto nla_put_failure;
603	return 0;
604
605nla_put_failure:
606	return -EMSGSIZE;
607}
608
609static const struct nla_policy ipip_policy[IFLA_IPTUN_MAX + 1] = {
610	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
611	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
612	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
613	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
614	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
615	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
616	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
617	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
618	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
619	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
620	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
621	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
622	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
623};
624
625static struct rtnl_link_ops ipip_link_ops __read_mostly = {
626	.kind		= "ipip",
627	.maxtype	= IFLA_IPTUN_MAX,
628	.policy		= ipip_policy,
629	.priv_size	= sizeof(struct ip_tunnel),
630	.setup		= ipip_tunnel_setup,
631	.validate	= ipip_tunnel_validate,
632	.newlink	= ipip_newlink,
633	.changelink	= ipip_changelink,
634	.dellink	= ip_tunnel_dellink,
635	.get_size	= ipip_get_size,
636	.fill_info	= ipip_fill_info,
637	.get_link_net	= ip_tunnel_get_link_net,
638};
639
640static struct xfrm_tunnel ipip_handler __read_mostly = {
641	.handler	=	ipip_rcv,
642	.err_handler	=	ipip_err,
643	.priority	=	1,
644};
645
646#if IS_ENABLED(CONFIG_MPLS)
647static struct xfrm_tunnel mplsip_handler __read_mostly = {
648	.handler	=	mplsip_rcv,
649	.err_handler	=	ipip_err,
650	.priority	=	1,
651};
652#endif
653
654static int __net_init ipip_init_net(struct net *net)
655{
656	return ip_tunnel_init_net(net, ipip_net_id, &ipip_link_ops, "tunl0");
657}
658
659static void __net_exit ipip_exit_batch_net(struct list_head *list_net)
660{
661	ip_tunnel_delete_nets(list_net, ipip_net_id, &ipip_link_ops);
 
662}
663
664static struct pernet_operations ipip_net_ops = {
665	.init = ipip_init_net,
666	.exit_batch = ipip_exit_batch_net,
667	.id   = &ipip_net_id,
668	.size = sizeof(struct ip_tunnel_net),
669};
670
671static int __init ipip_init(void)
672{
673	int err;
674
675	pr_info("ipip: IPv4 and MPLS over IPv4 tunneling driver\n");
676
677	err = register_pernet_device(&ipip_net_ops);
678	if (err < 0)
679		return err;
680	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
681	if (err < 0) {
682		pr_info("%s: can't register tunnel\n", __func__);
683		goto xfrm_tunnel_ipip_failed;
684	}
685#if IS_ENABLED(CONFIG_MPLS)
686	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
687	if (err < 0) {
688		pr_info("%s: can't register tunnel\n", __func__);
689		goto xfrm_tunnel_mplsip_failed;
690	}
691#endif
692	err = rtnl_link_register(&ipip_link_ops);
693	if (err < 0)
694		goto rtnl_link_failed;
695
696out:
697	return err;
698
699rtnl_link_failed:
700#if IS_ENABLED(CONFIG_MPLS)
701	xfrm4_tunnel_deregister(&mplsip_handler, AF_INET);
702xfrm_tunnel_mplsip_failed:
703
704#endif
705	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
706xfrm_tunnel_ipip_failed:
707	unregister_pernet_device(&ipip_net_ops);
708	goto out;
709}
710
711static void __exit ipip_fini(void)
712{
713	rtnl_link_unregister(&ipip_link_ops);
714	if (xfrm4_tunnel_deregister(&ipip_handler, AF_INET))
715		pr_info("%s: can't deregister tunnel\n", __func__);
716#if IS_ENABLED(CONFIG_MPLS)
717	if (xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS))
718		pr_info("%s: can't deregister tunnel\n", __func__);
719#endif
720	unregister_pernet_device(&ipip_net_ops);
721}
722
723module_init(ipip_init);
724module_exit(ipip_fini);
725MODULE_LICENSE("GPL");
726MODULE_ALIAS_RTNL_LINK("ipip");
727MODULE_ALIAS_NETDEV("tunl0");