xref: /linux/net/ipv4/ip_vti.c (revision d458cdf712e0c671e8e819abb16ecd6e44f9daec)
1 /*
2  *	Linux NET3: IP/IP protocol decoder modified to support
3  *		    virtual tunnel interface
4  *
5  *	Authors:
6  *		Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
7  *
8  *	This program is free software; you can redistribute it and/or
9  *	modify it under the terms of the GNU General Public License
10  *	as published by the Free Software Foundation; either version
11  *	2 of the License, or (at your option) any later version.
12  *
13  */
14 
15 /*
16    This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
17 
18    For comments look at net/ipv4/ip_gre.c --ANK
19  */
20 
21 
22 #include <linux/capability.h>
23 #include <linux/module.h>
24 #include <linux/types.h>
25 #include <linux/kernel.h>
26 #include <linux/uaccess.h>
27 #include <linux/skbuff.h>
28 #include <linux/netdevice.h>
29 #include <linux/in.h>
30 #include <linux/tcp.h>
31 #include <linux/udp.h>
32 #include <linux/if_arp.h>
33 #include <linux/mroute.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.h>
36 #include <linux/if_ether.h>
37 
38 #include <net/sock.h>
39 #include <net/ip.h>
40 #include <net/icmp.h>
41 #include <net/ip_tunnels.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 
47 static struct rtnl_link_ops vti_link_ops __read_mostly;
48 
49 static int vti_net_id __read_mostly;
50 static int vti_tunnel_init(struct net_device *dev);
51 
52 /* We dont digest the packet therefore let the packet pass */
53 static int vti_rcv(struct sk_buff *skb)
54 {
55 	struct ip_tunnel *tunnel;
56 	const struct iphdr *iph = ip_hdr(skb);
57 	struct net *net = dev_net(skb->dev);
58 	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
59 
60 	tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
61 				  iph->saddr, iph->daddr, 0);
62 	if (tunnel != NULL) {
63 		struct pcpu_tstats *tstats;
64 
65 		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
66 			return -1;
67 
68 		tstats = this_cpu_ptr(tunnel->dev->tstats);
69 		u64_stats_update_begin(&tstats->syncp);
70 		tstats->rx_packets++;
71 		tstats->rx_bytes += skb->len;
72 		u64_stats_update_end(&tstats->syncp);
73 
74 		skb->mark = 0;
75 		secpath_reset(skb);
76 		skb->dev = tunnel->dev;
77 		return 1;
78 	}
79 
80 	return -1;
81 }
82 
83 /* This function assumes it is being called from dev_queue_xmit()
84  * and that skb is filled properly by that function.
85  */
86 
87 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
88 {
89 	struct ip_tunnel *tunnel = netdev_priv(dev);
90 	struct iphdr  *tiph = &tunnel->parms.iph;
91 	u8     tos;
92 	struct rtable *rt;		/* Route to the other host */
93 	struct net_device *tdev;	/* Device to other host */
94 	struct iphdr  *old_iph = ip_hdr(skb);
95 	__be32 dst = tiph->daddr;
96 	struct flowi4 fl4;
97 	int err;
98 
99 	if (skb->protocol != htons(ETH_P_IP))
100 		goto tx_error;
101 
102 	tos = old_iph->tos;
103 
104 	memset(&fl4, 0, sizeof(fl4));
105 	flowi4_init_output(&fl4, tunnel->parms.link,
106 			   be32_to_cpu(tunnel->parms.i_key), RT_TOS(tos),
107 			   RT_SCOPE_UNIVERSE,
108 			   IPPROTO_IPIP, 0,
109 			   dst, tiph->saddr, 0, 0);
110 	rt = ip_route_output_key(dev_net(dev), &fl4);
111 	if (IS_ERR(rt)) {
112 		dev->stats.tx_carrier_errors++;
113 		goto tx_error_icmp;
114 	}
115 	/* if there is no transform then this tunnel is not functional.
116 	 * Or if the xfrm is not mode tunnel.
117 	 */
118 	if (!rt->dst.xfrm ||
119 	    rt->dst.xfrm->props.mode != XFRM_MODE_TUNNEL) {
120 		dev->stats.tx_carrier_errors++;
121 		goto tx_error_icmp;
122 	}
123 	tdev = rt->dst.dev;
124 
125 	if (tdev == dev) {
126 		ip_rt_put(rt);
127 		dev->stats.collisions++;
128 		goto tx_error;
129 	}
130 
131 	if (tunnel->err_count > 0) {
132 		if (time_before(jiffies,
133 				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
134 			tunnel->err_count--;
135 			dst_link_failure(skb);
136 		} else
137 			tunnel->err_count = 0;
138 	}
139 
140 	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
141 	skb_dst_drop(skb);
142 	skb_dst_set(skb, &rt->dst);
143 	nf_reset(skb);
144 	skb->dev = skb_dst(skb)->dev;
145 
146 	err = dst_output(skb);
147 	if (net_xmit_eval(err) == 0)
148 		err = skb->len;
149 	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
150 	return NETDEV_TX_OK;
151 
152 tx_error_icmp:
153 	dst_link_failure(skb);
154 tx_error:
155 	dev->stats.tx_errors++;
156 	dev_kfree_skb(skb);
157 	return NETDEV_TX_OK;
158 }
159 
160 static int
161 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
162 {
163 	int err = 0;
164 	struct ip_tunnel_parm p;
165 
166 	if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
167 		return -EFAULT;
168 
169 	if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
170 		if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
171 		    p.iph.ihl != 5)
172 			return -EINVAL;
173 	}
174 
175 	err = ip_tunnel_ioctl(dev, &p, cmd);
176 	if (err)
177 		return err;
178 
179 	if (cmd != SIOCDELTUNNEL) {
180 		p.i_flags |= GRE_KEY | VTI_ISVTI;
181 		p.o_flags |= GRE_KEY;
182 	}
183 
184 	if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
185 		return -EFAULT;
186 	return 0;
187 }
188 
189 static const struct net_device_ops vti_netdev_ops = {
190 	.ndo_init	= vti_tunnel_init,
191 	.ndo_uninit	= ip_tunnel_uninit,
192 	.ndo_start_xmit	= vti_tunnel_xmit,
193 	.ndo_do_ioctl	= vti_tunnel_ioctl,
194 	.ndo_change_mtu	= ip_tunnel_change_mtu,
195 	.ndo_get_stats64 = ip_tunnel_get_stats64,
196 };
197 
198 static void vti_tunnel_setup(struct net_device *dev)
199 {
200 	dev->netdev_ops		= &vti_netdev_ops;
201 	ip_tunnel_setup(dev, vti_net_id);
202 }
203 
204 static int vti_tunnel_init(struct net_device *dev)
205 {
206 	struct ip_tunnel *tunnel = netdev_priv(dev);
207 	struct iphdr *iph = &tunnel->parms.iph;
208 
209 	memcpy(dev->dev_addr, &iph->saddr, 4);
210 	memcpy(dev->broadcast, &iph->daddr, 4);
211 
212 	dev->type		= ARPHRD_TUNNEL;
213 	dev->hard_header_len	= LL_MAX_HEADER + sizeof(struct iphdr);
214 	dev->mtu		= ETH_DATA_LEN;
215 	dev->flags		= IFF_NOARP;
216 	dev->iflink		= 0;
217 	dev->addr_len		= 4;
218 	dev->features		|= NETIF_F_NETNS_LOCAL;
219 	dev->features		|= NETIF_F_LLTX;
220 	dev->priv_flags		&= ~IFF_XMIT_DST_RELEASE;
221 
222 	return ip_tunnel_init(dev);
223 }
224 
225 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
226 {
227 	struct ip_tunnel *tunnel = netdev_priv(dev);
228 	struct iphdr *iph = &tunnel->parms.iph;
229 
230 	iph->version		= 4;
231 	iph->protocol		= IPPROTO_IPIP;
232 	iph->ihl		= 5;
233 }
234 
235 static struct xfrm_tunnel_notifier vti_handler __read_mostly = {
236 	.handler	=	vti_rcv,
237 	.priority	=	1,
238 };
239 
240 static int __net_init vti_init_net(struct net *net)
241 {
242 	int err;
243 	struct ip_tunnel_net *itn;
244 
245 	err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
246 	if (err)
247 		return err;
248 	itn = net_generic(net, vti_net_id);
249 	vti_fb_tunnel_init(itn->fb_tunnel_dev);
250 	return 0;
251 }
252 
253 static void __net_exit vti_exit_net(struct net *net)
254 {
255 	struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
256 	ip_tunnel_delete_net(itn, &vti_link_ops);
257 }
258 
259 static struct pernet_operations vti_net_ops = {
260 	.init = vti_init_net,
261 	.exit = vti_exit_net,
262 	.id   = &vti_net_id,
263 	.size = sizeof(struct ip_tunnel_net),
264 };
265 
266 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
267 {
268 	return 0;
269 }
270 
271 static void vti_netlink_parms(struct nlattr *data[],
272 			      struct ip_tunnel_parm *parms)
273 {
274 	memset(parms, 0, sizeof(*parms));
275 
276 	parms->iph.protocol = IPPROTO_IPIP;
277 
278 	if (!data)
279 		return;
280 
281 	if (data[IFLA_VTI_LINK])
282 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
283 
284 	if (data[IFLA_VTI_IKEY])
285 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
286 
287 	if (data[IFLA_VTI_OKEY])
288 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
289 
290 	if (data[IFLA_VTI_LOCAL])
291 		parms->iph.saddr = nla_get_be32(data[IFLA_VTI_LOCAL]);
292 
293 	if (data[IFLA_VTI_REMOTE])
294 		parms->iph.daddr = nla_get_be32(data[IFLA_VTI_REMOTE]);
295 
296 }
297 
298 static int vti_newlink(struct net *src_net, struct net_device *dev,
299 		       struct nlattr *tb[], struct nlattr *data[])
300 {
301 	struct ip_tunnel_parm parms;
302 
303 	vti_netlink_parms(data, &parms);
304 	return ip_tunnel_newlink(dev, tb, &parms);
305 }
306 
307 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
308 			  struct nlattr *data[])
309 {
310 	struct ip_tunnel_parm p;
311 
312 	vti_netlink_parms(data, &p);
313 	return ip_tunnel_changelink(dev, tb, &p);
314 }
315 
316 static size_t vti_get_size(const struct net_device *dev)
317 {
318 	return
319 		/* IFLA_VTI_LINK */
320 		nla_total_size(4) +
321 		/* IFLA_VTI_IKEY */
322 		nla_total_size(4) +
323 		/* IFLA_VTI_OKEY */
324 		nla_total_size(4) +
325 		/* IFLA_VTI_LOCAL */
326 		nla_total_size(4) +
327 		/* IFLA_VTI_REMOTE */
328 		nla_total_size(4) +
329 		0;
330 }
331 
332 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
333 {
334 	struct ip_tunnel *t = netdev_priv(dev);
335 	struct ip_tunnel_parm *p = &t->parms;
336 
337 	nla_put_u32(skb, IFLA_VTI_LINK, p->link);
338 	nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
339 	nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
340 	nla_put_be32(skb, IFLA_VTI_LOCAL, p->iph.saddr);
341 	nla_put_be32(skb, IFLA_VTI_REMOTE, p->iph.daddr);
342 
343 	return 0;
344 }
345 
346 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
347 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
348 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
349 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
350 	[IFLA_VTI_LOCAL]	= { .len = FIELD_SIZEOF(struct iphdr, saddr) },
351 	[IFLA_VTI_REMOTE]	= { .len = FIELD_SIZEOF(struct iphdr, daddr) },
352 };
353 
354 static struct rtnl_link_ops vti_link_ops __read_mostly = {
355 	.kind		= "vti",
356 	.maxtype	= IFLA_VTI_MAX,
357 	.policy		= vti_policy,
358 	.priv_size	= sizeof(struct ip_tunnel),
359 	.setup		= vti_tunnel_setup,
360 	.validate	= vti_tunnel_validate,
361 	.newlink	= vti_newlink,
362 	.changelink	= vti_changelink,
363 	.get_size	= vti_get_size,
364 	.fill_info	= vti_fill_info,
365 };
366 
367 static int __init vti_init(void)
368 {
369 	int err;
370 
371 	pr_info("IPv4 over IPSec tunneling driver\n");
372 
373 	err = register_pernet_device(&vti_net_ops);
374 	if (err < 0)
375 		return err;
376 	err = xfrm4_mode_tunnel_input_register(&vti_handler);
377 	if (err < 0) {
378 		unregister_pernet_device(&vti_net_ops);
379 		pr_info("vti init: can't register tunnel\n");
380 	}
381 
382 	err = rtnl_link_register(&vti_link_ops);
383 	if (err < 0)
384 		goto rtnl_link_failed;
385 
386 	return err;
387 
388 rtnl_link_failed:
389 	xfrm4_mode_tunnel_input_deregister(&vti_handler);
390 	unregister_pernet_device(&vti_net_ops);
391 	return err;
392 }
393 
394 static void __exit vti_fini(void)
395 {
396 	rtnl_link_unregister(&vti_link_ops);
397 	if (xfrm4_mode_tunnel_input_deregister(&vti_handler))
398 		pr_info("vti close: can't deregister tunnel\n");
399 
400 	unregister_pernet_device(&vti_net_ops);
401 }
402 
403 module_init(vti_init);
404 module_exit(vti_fini);
405 MODULE_LICENSE("GPL");
406 MODULE_ALIAS_RTNL_LINK("vti");
407 MODULE_ALIAS_NETDEV("ip_vti0");
408