veth.c revision 6311cc44a23bb42636f5076fef0a67859d0a0102 
1/*
2 *  drivers/net/veth.c
3 *
4 *  Copyright (C) 2007 OpenVZ http://openvz.org, SWsoft Inc
5 *
6 * Author: Pavel Emelianov <xemul@openvz.org>
7 * Ethtool interface from: Eric W. Biederman <ebiederm@xmission.com>
8 *
9 */
10
11#include <linux/netdevice.h>
12#include <linux/slab.h>
13#include <linux/ethtool.h>
14#include <linux/etherdevice.h>
15
16#include <net/dst.h>
17#include <net/xfrm.h>
18#include <linux/veth.h>
19
20#define DRV_NAME	"veth"
21#define DRV_VERSION	"1.0"
22
23#define MIN_MTU 68		/* Min L3 MTU */
24#define MAX_MTU 65535		/* Max L3 MTU (arbitrary) */
25
26struct veth_net_stats {
27	u64	rx_packets;
28	u64	tx_packets;
29	u64	rx_bytes;
30	u64	tx_bytes;
31	u64	tx_dropped;
32	u64	rx_dropped;
33};
34
35struct veth_priv {
36	struct net_device *peer;
37	struct veth_net_stats __percpu *stats;
38};
39
40/*
41 * ethtool interface
42 */
43
44static struct {
45	const char string[ETH_GSTRING_LEN];
46} ethtool_stats_keys[] = {
47	{ "peer_ifindex" },
48};
49
50static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
51{
52	cmd->supported		= 0;
53	cmd->advertising	= 0;
54	ethtool_cmd_speed_set(cmd, SPEED_10000);
55	cmd->duplex		= DUPLEX_FULL;
56	cmd->port		= PORT_TP;
57	cmd->phy_address	= 0;
58	cmd->transceiver	= XCVR_INTERNAL;
59	cmd->autoneg		= AUTONEG_DISABLE;
60	cmd->maxtxpkt		= 0;
61	cmd->maxrxpkt		= 0;
62	return 0;
63}
64
65static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
66{
67	strcpy(info->driver, DRV_NAME);
68	strcpy(info->version, DRV_VERSION);
69	strcpy(info->fw_version, "N/A");
70}
71
72static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
73{
74	switch(stringset) {
75	case ETH_SS_STATS:
76		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
77		break;
78	}
79}
80
81static int veth_get_sset_count(struct net_device *dev, int sset)
82{
83	switch (sset) {
84	case ETH_SS_STATS:
85		return ARRAY_SIZE(ethtool_stats_keys);
86	default:
87		return -EOPNOTSUPP;
88	}
89}
90
91static void veth_get_ethtool_stats(struct net_device *dev,
92		struct ethtool_stats *stats, u64 *data)
93{
94	struct veth_priv *priv;
95
96	priv = netdev_priv(dev);
97	data[0] = priv->peer->ifindex;
98}
99
100static const struct ethtool_ops veth_ethtool_ops = {
101	.get_settings		= veth_get_settings,
102	.get_drvinfo		= veth_get_drvinfo,
103	.get_link		= ethtool_op_get_link,
104	.get_strings		= veth_get_strings,
105	.get_sset_count		= veth_get_sset_count,
106	.get_ethtool_stats	= veth_get_ethtool_stats,
107};
108
109/*
110 * xmit
111 */
112
113static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
114{
115	struct net_device *rcv = NULL;
116	struct veth_priv *priv, *rcv_priv;
117	struct veth_net_stats *stats, *rcv_stats;
118	int length;
119
120	priv = netdev_priv(dev);
121	rcv = priv->peer;
122	rcv_priv = netdev_priv(rcv);
123
124	stats = this_cpu_ptr(priv->stats);
125	rcv_stats = this_cpu_ptr(rcv_priv->stats);
126
127	if (!(rcv->flags & IFF_UP))
128		goto tx_drop;
129
130	/* don't change ip_summed == CHECKSUM_PARTIAL, as that
131	   will cause bad checksum on forwarded packets */
132	if (skb->ip_summed == CHECKSUM_NONE &&
133	    rcv->features & NETIF_F_RXCSUM)
134		skb->ip_summed = CHECKSUM_UNNECESSARY;
135
136	length = skb->len;
137	if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
138		goto rx_drop;
139
140	stats->tx_bytes += length;
141	stats->tx_packets++;
142
143	rcv_stats->rx_bytes += length;
144	rcv_stats->rx_packets++;
145
146	return NETDEV_TX_OK;
147
148tx_drop:
149	kfree_skb(skb);
150	stats->tx_dropped++;
151	return NETDEV_TX_OK;
152
153rx_drop:
154	rcv_stats->rx_dropped++;
155	return NETDEV_TX_OK;
156}
157
158/*
159 * general routines
160 */
161
162static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
163						  struct rtnl_link_stats64 *tot)
164{
165	struct veth_priv *priv;
166	int cpu;
167	struct veth_net_stats *stats;
168
169	priv = netdev_priv(dev);
170
171	for_each_possible_cpu(cpu) {
172		stats = per_cpu_ptr(priv->stats, cpu);
173
174		tot->rx_packets += stats->rx_packets;
175		tot->tx_packets += stats->tx_packets;
176		tot->rx_bytes   += stats->rx_bytes;
177		tot->tx_bytes   += stats->tx_bytes;
178		tot->tx_dropped += stats->tx_dropped;
179		tot->rx_dropped += stats->rx_dropped;
180	}
181
182	return tot;
183}
184
185static int veth_open(struct net_device *dev)
186{
187	struct veth_priv *priv;
188
189	priv = netdev_priv(dev);
190	if (priv->peer == NULL)
191		return -ENOTCONN;
192
193	if (priv->peer->flags & IFF_UP) {
194		netif_carrier_on(dev);
195		netif_carrier_on(priv->peer);
196	}
197	return 0;
198}
199
200static int veth_close(struct net_device *dev)
201{
202	struct veth_priv *priv = netdev_priv(dev);
203
204	netif_carrier_off(dev);
205	netif_carrier_off(priv->peer);
206
207	return 0;
208}
209
210static int is_valid_veth_mtu(int new_mtu)
211{
212	return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
213}
214
215static int veth_change_mtu(struct net_device *dev, int new_mtu)
216{
217	if (!is_valid_veth_mtu(new_mtu))
218		return -EINVAL;
219	dev->mtu = new_mtu;
220	return 0;
221}
222
223static int veth_dev_init(struct net_device *dev)
224{
225	struct veth_net_stats __percpu *stats;
226	struct veth_priv *priv;
227
228	stats = alloc_percpu(struct veth_net_stats);
229	if (stats == NULL)
230		return -ENOMEM;
231
232	priv = netdev_priv(dev);
233	priv->stats = stats;
234	return 0;
235}
236
237static void veth_dev_free(struct net_device *dev)
238{
239	struct veth_priv *priv;
240
241	priv = netdev_priv(dev);
242	free_percpu(priv->stats);
243	free_netdev(dev);
244}
245
246static const struct net_device_ops veth_netdev_ops = {
247	.ndo_init            = veth_dev_init,
248	.ndo_open            = veth_open,
249	.ndo_stop            = veth_close,
250	.ndo_start_xmit      = veth_xmit,
251	.ndo_change_mtu      = veth_change_mtu,
252	.ndo_get_stats64     = veth_get_stats64,
253	.ndo_set_mac_address = eth_mac_addr,
254};
255
256static void veth_setup(struct net_device *dev)
257{
258	ether_setup(dev);
259
260	dev->netdev_ops = &veth_netdev_ops;
261	dev->ethtool_ops = &veth_ethtool_ops;
262	dev->features |= NETIF_F_LLTX;
263	dev->destructor = veth_dev_free;
264
265	dev->hw_features = NETIF_F_NO_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
266}
267
268/*
269 * netlink interface
270 */
271
272static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
273{
274	if (tb[IFLA_ADDRESS]) {
275		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
276			return -EINVAL;
277		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
278			return -EADDRNOTAVAIL;
279	}
280	if (tb[IFLA_MTU]) {
281		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
282			return -EINVAL;
283	}
284	return 0;
285}
286
287static struct rtnl_link_ops veth_link_ops;
288
289static int veth_newlink(struct net *src_net, struct net_device *dev,
290			 struct nlattr *tb[], struct nlattr *data[])
291{
292	int err;
293	struct net_device *peer;
294	struct veth_priv *priv;
295	char ifname[IFNAMSIZ];
296	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
297	struct ifinfomsg *ifmp;
298	struct net *net;
299
300	/*
301	 * create and register peer first
302	 */
303	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
304		struct nlattr *nla_peer;
305
306		nla_peer = data[VETH_INFO_PEER];
307		ifmp = nla_data(nla_peer);
308		err = nla_parse(peer_tb, IFLA_MAX,
309				nla_data(nla_peer) + sizeof(struct ifinfomsg),
310				nla_len(nla_peer) - sizeof(struct ifinfomsg),
311				ifla_policy);
312		if (err < 0)
313			return err;
314
315		err = veth_validate(peer_tb, NULL);
316		if (err < 0)
317			return err;
318
319		tbp = peer_tb;
320	} else {
321		ifmp = NULL;
322		tbp = tb;
323	}
324
325	if (tbp[IFLA_IFNAME])
326		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
327	else
328		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
329
330	net = rtnl_link_get_net(src_net, tbp);
331	if (IS_ERR(net))
332		return PTR_ERR(net);
333
334	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
335	if (IS_ERR(peer)) {
336		put_net(net);
337		return PTR_ERR(peer);
338	}
339
340	if (tbp[IFLA_ADDRESS] == NULL)
341		random_ether_addr(peer->dev_addr);
342
343	err = register_netdevice(peer);
344	put_net(net);
345	net = NULL;
346	if (err < 0)
347		goto err_register_peer;
348
349	netif_carrier_off(peer);
350
351	err = rtnl_configure_link(peer, ifmp);
352	if (err < 0)
353		goto err_configure_peer;
354
355	/*
356	 * register dev last
357	 *
358	 * note, that since we've registered new device the dev's name
359	 * should be re-allocated
360	 */
361
362	if (tb[IFLA_ADDRESS] == NULL)
363		random_ether_addr(dev->dev_addr);
364
365	if (tb[IFLA_IFNAME])
366		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
367	else
368		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
369
370	if (strchr(dev->name, '%')) {
371		err = dev_alloc_name(dev, dev->name);
372		if (err < 0)
373			goto err_alloc_name;
374	}
375
376	err = register_netdevice(dev);
377	if (err < 0)
378		goto err_register_dev;
379
380	netif_carrier_off(dev);
381
382	/*
383	 * tie the deviced together
384	 */
385
386	priv = netdev_priv(dev);
387	priv->peer = peer;
388
389	priv = netdev_priv(peer);
390	priv->peer = dev;
391	return 0;
392
393err_register_dev:
394	/* nothing to do */
395err_alloc_name:
396err_configure_peer:
397	unregister_netdevice(peer);
398	return err;
399
400err_register_peer:
401	free_netdev(peer);
402	return err;
403}
404
405static void veth_dellink(struct net_device *dev, struct list_head *head)
406{
407	struct veth_priv *priv;
408	struct net_device *peer;
409
410	priv = netdev_priv(dev);
411	peer = priv->peer;
412
413	unregister_netdevice_queue(dev, head);
414	unregister_netdevice_queue(peer, head);
415}
416
417static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
418
419static struct rtnl_link_ops veth_link_ops = {
420	.kind		= DRV_NAME,
421	.priv_size	= sizeof(struct veth_priv),
422	.setup		= veth_setup,
423	.validate	= veth_validate,
424	.newlink	= veth_newlink,
425	.dellink	= veth_dellink,
426	.policy		= veth_policy,
427	.maxtype	= VETH_INFO_MAX,
428};
429
430/*
431 * init/fini
432 */
433
434static __init int veth_init(void)
435{
436	return rtnl_link_register(&veth_link_ops);
437}
438
439static __exit void veth_exit(void)
440{
441	rtnl_link_unregister(&veth_link_ops);
442}
443
444module_init(veth_init);
445module_exit(veth_exit);
446
447MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
448MODULE_LICENSE("GPL v2");
449MODULE_ALIAS_RTNL_LINK(DRV_NAME);
450