veth.c revision 534ea99b063de7c30262a8e22f0ab44dd7d11a71 
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	unsigned long	rx_packets;
28	unsigned long	tx_packets;
29	unsigned long	rx_bytes;
30	unsigned long	tx_bytes;
31	unsigned long	tx_dropped;
32	unsigned long	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 net_device_stats *veth_get_stats(struct net_device *dev)
163{
164	struct veth_priv *priv;
165	int cpu;
166	struct veth_net_stats *stats, total = {0};
167
168	priv = netdev_priv(dev);
169
170	for_each_possible_cpu(cpu) {
171		stats = per_cpu_ptr(priv->stats, cpu);
172
173		total.rx_packets += stats->rx_packets;
174		total.tx_packets += stats->tx_packets;
175		total.rx_bytes   += stats->rx_bytes;
176		total.tx_bytes   += stats->tx_bytes;
177		total.tx_dropped += stats->tx_dropped;
178		total.rx_dropped += stats->rx_dropped;
179	}
180	dev->stats.rx_packets = total.rx_packets;
181	dev->stats.tx_packets = total.tx_packets;
182	dev->stats.rx_bytes   = total.rx_bytes;
183	dev->stats.tx_bytes   = total.tx_bytes;
184	dev->stats.tx_dropped = total.tx_dropped;
185	dev->stats.rx_dropped = total.rx_dropped;
186
187	return &dev->stats;
188}
189
190static int veth_open(struct net_device *dev)
191{
192	struct veth_priv *priv;
193
194	priv = netdev_priv(dev);
195	if (priv->peer == NULL)
196		return -ENOTCONN;
197
198	if (priv->peer->flags & IFF_UP) {
199		netif_carrier_on(dev);
200		netif_carrier_on(priv->peer);
201	}
202	return 0;
203}
204
205static int veth_close(struct net_device *dev)
206{
207	struct veth_priv *priv = netdev_priv(dev);
208
209	netif_carrier_off(dev);
210	netif_carrier_off(priv->peer);
211
212	return 0;
213}
214
215static int is_valid_veth_mtu(int new_mtu)
216{
217	return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
218}
219
220static int veth_change_mtu(struct net_device *dev, int new_mtu)
221{
222	if (!is_valid_veth_mtu(new_mtu))
223		return -EINVAL;
224	dev->mtu = new_mtu;
225	return 0;
226}
227
228static int veth_dev_init(struct net_device *dev)
229{
230	struct veth_net_stats __percpu *stats;
231	struct veth_priv *priv;
232
233	stats = alloc_percpu(struct veth_net_stats);
234	if (stats == NULL)
235		return -ENOMEM;
236
237	priv = netdev_priv(dev);
238	priv->stats = stats;
239	return 0;
240}
241
242static void veth_dev_free(struct net_device *dev)
243{
244	struct veth_priv *priv;
245
246	priv = netdev_priv(dev);
247	free_percpu(priv->stats);
248	free_netdev(dev);
249}
250
251static const struct net_device_ops veth_netdev_ops = {
252	.ndo_init            = veth_dev_init,
253	.ndo_open            = veth_open,
254	.ndo_stop            = veth_close,
255	.ndo_start_xmit      = veth_xmit,
256	.ndo_change_mtu      = veth_change_mtu,
257	.ndo_get_stats       = veth_get_stats,
258	.ndo_set_mac_address = eth_mac_addr,
259};
260
261static void veth_setup(struct net_device *dev)
262{
263	ether_setup(dev);
264
265	dev->netdev_ops = &veth_netdev_ops;
266	dev->ethtool_ops = &veth_ethtool_ops;
267	dev->features |= NETIF_F_LLTX;
268	dev->destructor = veth_dev_free;
269
270	dev->hw_features = NETIF_F_NO_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
271}
272
273/*
274 * netlink interface
275 */
276
277static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
278{
279	if (tb[IFLA_ADDRESS]) {
280		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
281			return -EINVAL;
282		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
283			return -EADDRNOTAVAIL;
284	}
285	if (tb[IFLA_MTU]) {
286		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
287			return -EINVAL;
288	}
289	return 0;
290}
291
292static struct rtnl_link_ops veth_link_ops;
293
294static int veth_newlink(struct net *src_net, struct net_device *dev,
295			 struct nlattr *tb[], struct nlattr *data[])
296{
297	int err;
298	struct net_device *peer;
299	struct veth_priv *priv;
300	char ifname[IFNAMSIZ];
301	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
302	struct ifinfomsg *ifmp;
303	struct net *net;
304
305	/*
306	 * create and register peer first
307	 */
308	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
309		struct nlattr *nla_peer;
310
311		nla_peer = data[VETH_INFO_PEER];
312		ifmp = nla_data(nla_peer);
313		err = nla_parse(peer_tb, IFLA_MAX,
314				nla_data(nla_peer) + sizeof(struct ifinfomsg),
315				nla_len(nla_peer) - sizeof(struct ifinfomsg),
316				ifla_policy);
317		if (err < 0)
318			return err;
319
320		err = veth_validate(peer_tb, NULL);
321		if (err < 0)
322			return err;
323
324		tbp = peer_tb;
325	} else {
326		ifmp = NULL;
327		tbp = tb;
328	}
329
330	if (tbp[IFLA_IFNAME])
331		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
332	else
333		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
334
335	net = rtnl_link_get_net(src_net, tbp);
336	if (IS_ERR(net))
337		return PTR_ERR(net);
338
339	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
340	if (IS_ERR(peer)) {
341		put_net(net);
342		return PTR_ERR(peer);
343	}
344
345	if (tbp[IFLA_ADDRESS] == NULL)
346		random_ether_addr(peer->dev_addr);
347
348	err = register_netdevice(peer);
349	put_net(net);
350	net = NULL;
351	if (err < 0)
352		goto err_register_peer;
353
354	netif_carrier_off(peer);
355
356	err = rtnl_configure_link(peer, ifmp);
357	if (err < 0)
358		goto err_configure_peer;
359
360	/*
361	 * register dev last
362	 *
363	 * note, that since we've registered new device the dev's name
364	 * should be re-allocated
365	 */
366
367	if (tb[IFLA_ADDRESS] == NULL)
368		random_ether_addr(dev->dev_addr);
369
370	if (tb[IFLA_IFNAME])
371		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
372	else
373		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
374
375	if (strchr(dev->name, '%')) {
376		err = dev_alloc_name(dev, dev->name);
377		if (err < 0)
378			goto err_alloc_name;
379	}
380
381	err = register_netdevice(dev);
382	if (err < 0)
383		goto err_register_dev;
384
385	netif_carrier_off(dev);
386
387	/*
388	 * tie the deviced together
389	 */
390
391	priv = netdev_priv(dev);
392	priv->peer = peer;
393
394	priv = netdev_priv(peer);
395	priv->peer = dev;
396	return 0;
397
398err_register_dev:
399	/* nothing to do */
400err_alloc_name:
401err_configure_peer:
402	unregister_netdevice(peer);
403	return err;
404
405err_register_peer:
406	free_netdev(peer);
407	return err;
408}
409
410static void veth_dellink(struct net_device *dev, struct list_head *head)
411{
412	struct veth_priv *priv;
413	struct net_device *peer;
414
415	priv = netdev_priv(dev);
416	peer = priv->peer;
417
418	unregister_netdevice_queue(dev, head);
419	unregister_netdevice_queue(peer, head);
420}
421
422static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
423
424static struct rtnl_link_ops veth_link_ops = {
425	.kind		= DRV_NAME,
426	.priv_size	= sizeof(struct veth_priv),
427	.setup		= veth_setup,
428	.validate	= veth_validate,
429	.newlink	= veth_newlink,
430	.dellink	= veth_dellink,
431	.policy		= veth_policy,
432	.maxtype	= VETH_INFO_MAX,
433};
434
435/*
436 * init/fini
437 */
438
439static __init int veth_init(void)
440{
441	return rtnl_link_register(&veth_link_ops);
442}
443
444static __exit void veth_exit(void)
445{
446	rtnl_link_unregister(&veth_link_ops);
447}
448
449module_init(veth_init);
450module_exit(veth_exit);
451
452MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
453MODULE_LICENSE("GPL v2");
454MODULE_ALIAS_RTNL_LINK(DRV_NAME);
455