veth.c revision 5a0e3ad6af8660be21ca98a971cd00f331318c05 
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#define MTU_PAD (ETH_HLEN + 4)  /* Max difference between L2 and L3 size MTU */
26
27struct veth_net_stats {
28	unsigned long	rx_packets;
29	unsigned long	tx_packets;
30	unsigned long	rx_bytes;
31	unsigned long	tx_bytes;
32	unsigned long	tx_dropped;
33	unsigned long	rx_dropped;
34};
35
36struct veth_priv {
37	struct net_device *peer;
38	struct veth_net_stats __percpu *stats;
39	unsigned ip_summed;
40};
41
42/*
43 * ethtool interface
44 */
45
46static struct {
47	const char string[ETH_GSTRING_LEN];
48} ethtool_stats_keys[] = {
49	{ "peer_ifindex" },
50};
51
52static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
53{
54	cmd->supported		= 0;
55	cmd->advertising	= 0;
56	cmd->speed		= SPEED_10000;
57	cmd->duplex		= DUPLEX_FULL;
58	cmd->port		= PORT_TP;
59	cmd->phy_address	= 0;
60	cmd->transceiver	= XCVR_INTERNAL;
61	cmd->autoneg		= AUTONEG_DISABLE;
62	cmd->maxtxpkt		= 0;
63	cmd->maxrxpkt		= 0;
64	return 0;
65}
66
67static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
68{
69	strcpy(info->driver, DRV_NAME);
70	strcpy(info->version, DRV_VERSION);
71	strcpy(info->fw_version, "N/A");
72}
73
74static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
75{
76	switch(stringset) {
77	case ETH_SS_STATS:
78		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
79		break;
80	}
81}
82
83static int veth_get_sset_count(struct net_device *dev, int sset)
84{
85	switch (sset) {
86	case ETH_SS_STATS:
87		return ARRAY_SIZE(ethtool_stats_keys);
88	default:
89		return -EOPNOTSUPP;
90	}
91}
92
93static void veth_get_ethtool_stats(struct net_device *dev,
94		struct ethtool_stats *stats, u64 *data)
95{
96	struct veth_priv *priv;
97
98	priv = netdev_priv(dev);
99	data[0] = priv->peer->ifindex;
100}
101
102static u32 veth_get_rx_csum(struct net_device *dev)
103{
104	struct veth_priv *priv;
105
106	priv = netdev_priv(dev);
107	return priv->ip_summed == CHECKSUM_UNNECESSARY;
108}
109
110static int veth_set_rx_csum(struct net_device *dev, u32 data)
111{
112	struct veth_priv *priv;
113
114	priv = netdev_priv(dev);
115	priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
116	return 0;
117}
118
119static u32 veth_get_tx_csum(struct net_device *dev)
120{
121	return (dev->features & NETIF_F_NO_CSUM) != 0;
122}
123
124static int veth_set_tx_csum(struct net_device *dev, u32 data)
125{
126	if (data)
127		dev->features |= NETIF_F_NO_CSUM;
128	else
129		dev->features &= ~NETIF_F_NO_CSUM;
130	return 0;
131}
132
133static const struct ethtool_ops veth_ethtool_ops = {
134	.get_settings		= veth_get_settings,
135	.get_drvinfo		= veth_get_drvinfo,
136	.get_link		= ethtool_op_get_link,
137	.get_rx_csum		= veth_get_rx_csum,
138	.set_rx_csum		= veth_set_rx_csum,
139	.get_tx_csum		= veth_get_tx_csum,
140	.set_tx_csum		= veth_set_tx_csum,
141	.get_sg			= ethtool_op_get_sg,
142	.set_sg			= ethtool_op_set_sg,
143	.get_strings		= veth_get_strings,
144	.get_sset_count		= veth_get_sset_count,
145	.get_ethtool_stats	= veth_get_ethtool_stats,
146};
147
148/*
149 * xmit
150 */
151
152static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
153{
154	struct net_device *rcv = NULL;
155	struct veth_priv *priv, *rcv_priv;
156	struct veth_net_stats *stats, *rcv_stats;
157	int length;
158
159	priv = netdev_priv(dev);
160	rcv = priv->peer;
161	rcv_priv = netdev_priv(rcv);
162
163	stats = this_cpu_ptr(priv->stats);
164	rcv_stats = this_cpu_ptr(rcv_priv->stats);
165
166	if (!(rcv->flags & IFF_UP))
167		goto tx_drop;
168
169	if (dev->features & NETIF_F_NO_CSUM)
170		skb->ip_summed = rcv_priv->ip_summed;
171
172	length = skb->len + ETH_HLEN;
173	if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
174		goto rx_drop;
175
176	stats->tx_bytes += length;
177	stats->tx_packets++;
178
179	rcv_stats->rx_bytes += length;
180	rcv_stats->rx_packets++;
181
182	return NETDEV_TX_OK;
183
184tx_drop:
185	kfree_skb(skb);
186	stats->tx_dropped++;
187	return NETDEV_TX_OK;
188
189rx_drop:
190	kfree_skb(skb);
191	rcv_stats->rx_dropped++;
192	return NETDEV_TX_OK;
193}
194
195/*
196 * general routines
197 */
198
199static struct net_device_stats *veth_get_stats(struct net_device *dev)
200{
201	struct veth_priv *priv;
202	int cpu;
203	struct veth_net_stats *stats, total = {0};
204
205	priv = netdev_priv(dev);
206
207	for_each_possible_cpu(cpu) {
208		stats = per_cpu_ptr(priv->stats, cpu);
209
210		total.rx_packets += stats->rx_packets;
211		total.tx_packets += stats->tx_packets;
212		total.rx_bytes   += stats->rx_bytes;
213		total.tx_bytes   += stats->tx_bytes;
214		total.tx_dropped += stats->tx_dropped;
215		total.rx_dropped += stats->rx_dropped;
216	}
217	dev->stats.rx_packets = total.rx_packets;
218	dev->stats.tx_packets = total.tx_packets;
219	dev->stats.rx_bytes   = total.rx_bytes;
220	dev->stats.tx_bytes   = total.tx_bytes;
221	dev->stats.tx_dropped = total.tx_dropped;
222	dev->stats.rx_dropped = total.rx_dropped;
223
224	return &dev->stats;
225}
226
227static int veth_open(struct net_device *dev)
228{
229	struct veth_priv *priv;
230
231	priv = netdev_priv(dev);
232	if (priv->peer == NULL)
233		return -ENOTCONN;
234
235	if (priv->peer->flags & IFF_UP) {
236		netif_carrier_on(dev);
237		netif_carrier_on(priv->peer);
238	}
239	return 0;
240}
241
242static int veth_close(struct net_device *dev)
243{
244	struct veth_priv *priv = netdev_priv(dev);
245
246	netif_carrier_off(dev);
247	netif_carrier_off(priv->peer);
248
249	return 0;
250}
251
252static int is_valid_veth_mtu(int new_mtu)
253{
254	return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
255}
256
257static int veth_change_mtu(struct net_device *dev, int new_mtu)
258{
259	if (!is_valid_veth_mtu(new_mtu))
260		return -EINVAL;
261	dev->mtu = new_mtu;
262	return 0;
263}
264
265static int veth_dev_init(struct net_device *dev)
266{
267	struct veth_net_stats __percpu *stats;
268	struct veth_priv *priv;
269
270	stats = alloc_percpu(struct veth_net_stats);
271	if (stats == NULL)
272		return -ENOMEM;
273
274	priv = netdev_priv(dev);
275	priv->stats = stats;
276	return 0;
277}
278
279static void veth_dev_free(struct net_device *dev)
280{
281	struct veth_priv *priv;
282
283	priv = netdev_priv(dev);
284	free_percpu(priv->stats);
285	free_netdev(dev);
286}
287
288static const struct net_device_ops veth_netdev_ops = {
289	.ndo_init            = veth_dev_init,
290	.ndo_open            = veth_open,
291	.ndo_stop            = veth_close,
292	.ndo_start_xmit      = veth_xmit,
293	.ndo_change_mtu      = veth_change_mtu,
294	.ndo_get_stats       = veth_get_stats,
295	.ndo_set_mac_address = eth_mac_addr,
296};
297
298static void veth_setup(struct net_device *dev)
299{
300	ether_setup(dev);
301
302	dev->netdev_ops = &veth_netdev_ops;
303	dev->ethtool_ops = &veth_ethtool_ops;
304	dev->features |= NETIF_F_LLTX;
305	dev->destructor = veth_dev_free;
306}
307
308/*
309 * netlink interface
310 */
311
312static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
313{
314	if (tb[IFLA_ADDRESS]) {
315		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
316			return -EINVAL;
317		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
318			return -EADDRNOTAVAIL;
319	}
320	if (tb[IFLA_MTU]) {
321		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
322			return -EINVAL;
323	}
324	return 0;
325}
326
327static struct rtnl_link_ops veth_link_ops;
328
329static int veth_newlink(struct net *src_net, struct net_device *dev,
330			 struct nlattr *tb[], struct nlattr *data[])
331{
332	int err;
333	struct net_device *peer;
334	struct veth_priv *priv;
335	char ifname[IFNAMSIZ];
336	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
337	struct ifinfomsg *ifmp;
338	struct net *net;
339
340	/*
341	 * create and register peer first
342	 */
343	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
344		struct nlattr *nla_peer;
345
346		nla_peer = data[VETH_INFO_PEER];
347		ifmp = nla_data(nla_peer);
348		err = nla_parse(peer_tb, IFLA_MAX,
349				nla_data(nla_peer) + sizeof(struct ifinfomsg),
350				nla_len(nla_peer) - sizeof(struct ifinfomsg),
351				ifla_policy);
352		if (err < 0)
353			return err;
354
355		err = veth_validate(peer_tb, NULL);
356		if (err < 0)
357			return err;
358
359		tbp = peer_tb;
360	} else {
361		ifmp = NULL;
362		tbp = tb;
363	}
364
365	if (tbp[IFLA_IFNAME])
366		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
367	else
368		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
369
370	net = rtnl_link_get_net(src_net, tbp);
371	if (IS_ERR(net))
372		return PTR_ERR(net);
373
374	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
375	if (IS_ERR(peer)) {
376		put_net(net);
377		return PTR_ERR(peer);
378	}
379
380	if (tbp[IFLA_ADDRESS] == NULL)
381		random_ether_addr(peer->dev_addr);
382
383	err = register_netdevice(peer);
384	put_net(net);
385	net = NULL;
386	if (err < 0)
387		goto err_register_peer;
388
389	netif_carrier_off(peer);
390
391	err = rtnl_configure_link(peer, ifmp);
392	if (err < 0)
393		goto err_configure_peer;
394
395	/*
396	 * register dev last
397	 *
398	 * note, that since we've registered new device the dev's name
399	 * should be re-allocated
400	 */
401
402	if (tb[IFLA_ADDRESS] == NULL)
403		random_ether_addr(dev->dev_addr);
404
405	if (tb[IFLA_IFNAME])
406		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
407	else
408		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
409
410	if (strchr(dev->name, '%')) {
411		err = dev_alloc_name(dev, dev->name);
412		if (err < 0)
413			goto err_alloc_name;
414	}
415
416	err = register_netdevice(dev);
417	if (err < 0)
418		goto err_register_dev;
419
420	netif_carrier_off(dev);
421
422	/*
423	 * tie the deviced together
424	 */
425
426	priv = netdev_priv(dev);
427	priv->peer = peer;
428
429	priv = netdev_priv(peer);
430	priv->peer = dev;
431	return 0;
432
433err_register_dev:
434	/* nothing to do */
435err_alloc_name:
436err_configure_peer:
437	unregister_netdevice(peer);
438	return err;
439
440err_register_peer:
441	free_netdev(peer);
442	return err;
443}
444
445static void veth_dellink(struct net_device *dev, struct list_head *head)
446{
447	struct veth_priv *priv;
448	struct net_device *peer;
449
450	priv = netdev_priv(dev);
451	peer = priv->peer;
452
453	unregister_netdevice_queue(dev, head);
454	unregister_netdevice_queue(peer, head);
455}
456
457static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
458
459static struct rtnl_link_ops veth_link_ops = {
460	.kind		= DRV_NAME,
461	.priv_size	= sizeof(struct veth_priv),
462	.setup		= veth_setup,
463	.validate	= veth_validate,
464	.newlink	= veth_newlink,
465	.dellink	= veth_dellink,
466	.policy		= veth_policy,
467	.maxtype	= VETH_INFO_MAX,
468};
469
470/*
471 * init/fini
472 */
473
474static __init int veth_init(void)
475{
476	return rtnl_link_register(&veth_link_ops);
477}
478
479static __exit void veth_exit(void)
480{
481	rtnl_link_unregister(&veth_link_ops);
482}
483
484module_init(veth_init);
485module_exit(veth_exit);
486
487MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
488MODULE_LICENSE("GPL v2");
489MODULE_ALIAS_RTNL_LINK(DRV_NAME);
490