veth.c revision 3717746ef8b5a9279281b5d318496710984ed739 
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/ethtool.h>
13#include <linux/etherdevice.h>
14
15#include <net/dst.h>
16#include <net/xfrm.h>
17#include <linux/veth.h>
18
19#define DRV_NAME	"veth"
20#define DRV_VERSION	"1.0"
21
22struct veth_net_stats {
23	unsigned long	rx_packets;
24	unsigned long	tx_packets;
25	unsigned long	rx_bytes;
26	unsigned long	tx_bytes;
27	unsigned long	tx_dropped;
28};
29
30struct veth_priv {
31	struct net_device *peer;
32	struct veth_net_stats *stats;
33	unsigned ip_summed;
34};
35
36/*
37 * ethtool interface
38 */
39
40static struct {
41	const char string[ETH_GSTRING_LEN];
42} ethtool_stats_keys[] = {
43	{ "peer_ifindex" },
44};
45
46static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
47{
48	cmd->supported		= 0;
49	cmd->advertising	= 0;
50	cmd->speed		= SPEED_10000;
51	cmd->duplex		= DUPLEX_FULL;
52	cmd->port		= PORT_TP;
53	cmd->phy_address	= 0;
54	cmd->transceiver	= XCVR_INTERNAL;
55	cmd->autoneg		= AUTONEG_DISABLE;
56	cmd->maxtxpkt		= 0;
57	cmd->maxrxpkt		= 0;
58	return 0;
59}
60
61static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
62{
63	strcpy(info->driver, DRV_NAME);
64	strcpy(info->version, DRV_VERSION);
65	strcpy(info->fw_version, "N/A");
66}
67
68static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
69{
70	switch(stringset) {
71	case ETH_SS_STATS:
72		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
73		break;
74	}
75}
76
77static int veth_get_sset_count(struct net_device *dev, int sset)
78{
79	switch (sset) {
80	case ETH_SS_STATS:
81		return ARRAY_SIZE(ethtool_stats_keys);
82	default:
83		return -EOPNOTSUPP;
84	}
85}
86
87static void veth_get_ethtool_stats(struct net_device *dev,
88		struct ethtool_stats *stats, u64 *data)
89{
90	struct veth_priv *priv;
91
92	priv = netdev_priv(dev);
93	data[0] = priv->peer->ifindex;
94}
95
96static u32 veth_get_rx_csum(struct net_device *dev)
97{
98	struct veth_priv *priv;
99
100	priv = netdev_priv(dev);
101	return priv->ip_summed == CHECKSUM_UNNECESSARY;
102}
103
104static int veth_set_rx_csum(struct net_device *dev, u32 data)
105{
106	struct veth_priv *priv;
107
108	priv = netdev_priv(dev);
109	priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
110	return 0;
111}
112
113static u32 veth_get_tx_csum(struct net_device *dev)
114{
115	return (dev->features & NETIF_F_NO_CSUM) != 0;
116}
117
118static int veth_set_tx_csum(struct net_device *dev, u32 data)
119{
120	if (data)
121		dev->features |= NETIF_F_NO_CSUM;
122	else
123		dev->features &= ~NETIF_F_NO_CSUM;
124	return 0;
125}
126
127static struct ethtool_ops veth_ethtool_ops = {
128	.get_settings		= veth_get_settings,
129	.get_drvinfo		= veth_get_drvinfo,
130	.get_link		= ethtool_op_get_link,
131	.get_rx_csum		= veth_get_rx_csum,
132	.set_rx_csum		= veth_set_rx_csum,
133	.get_tx_csum		= veth_get_tx_csum,
134	.set_tx_csum		= veth_set_tx_csum,
135	.get_sg			= ethtool_op_get_sg,
136	.set_sg			= ethtool_op_set_sg,
137	.get_strings		= veth_get_strings,
138	.get_sset_count		= veth_get_sset_count,
139	.get_ethtool_stats	= veth_get_ethtool_stats,
140};
141
142/*
143 * xmit
144 */
145
146static int veth_xmit(struct sk_buff *skb, struct net_device *dev)
147{
148	struct net_device *rcv = NULL;
149	struct veth_priv *priv, *rcv_priv;
150	struct veth_net_stats *stats;
151	int length, cpu;
152
153	skb_orphan(skb);
154
155	priv = netdev_priv(dev);
156	rcv = priv->peer;
157	rcv_priv = netdev_priv(rcv);
158
159	cpu = smp_processor_id();
160	stats = per_cpu_ptr(priv->stats, cpu);
161
162	if (!(rcv->flags & IFF_UP))
163		goto outf;
164
165	skb->pkt_type = PACKET_HOST;
166	skb->protocol = eth_type_trans(skb, rcv);
167	if (dev->features & NETIF_F_NO_CSUM)
168		skb->ip_summed = rcv_priv->ip_summed;
169
170	dst_release(skb->dst);
171	skb->dst = NULL;
172	skb->mark = 0;
173	secpath_reset(skb);
174	nf_reset(skb);
175
176	length = skb->len;
177
178	stats->tx_bytes += length;
179	stats->tx_packets++;
180
181	stats = per_cpu_ptr(rcv_priv->stats, cpu);
182	stats->rx_bytes += length;
183	stats->rx_packets++;
184
185	netif_rx(skb);
186	return 0;
187
188outf:
189	kfree_skb(skb);
190	stats->tx_dropped++;
191	return 0;
192}
193
194/*
195 * general routines
196 */
197
198static struct net_device_stats *veth_get_stats(struct net_device *dev)
199{
200	struct veth_priv *priv;
201	struct net_device_stats *dev_stats;
202	int cpu;
203	struct veth_net_stats *stats;
204
205	priv = netdev_priv(dev);
206	dev_stats = &dev->stats;
207
208	dev_stats->rx_packets = 0;
209	dev_stats->tx_packets = 0;
210	dev_stats->rx_bytes = 0;
211	dev_stats->tx_bytes = 0;
212	dev_stats->tx_dropped = 0;
213
214	for_each_online_cpu(cpu) {
215		stats = per_cpu_ptr(priv->stats, cpu);
216
217		dev_stats->rx_packets += stats->rx_packets;
218		dev_stats->tx_packets += stats->tx_packets;
219		dev_stats->rx_bytes += stats->rx_bytes;
220		dev_stats->tx_bytes += stats->tx_bytes;
221		dev_stats->tx_dropped += stats->tx_dropped;
222	}
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_dev_init(struct net_device *dev)
243{
244	struct veth_net_stats *stats;
245	struct veth_priv *priv;
246
247	stats = alloc_percpu(struct veth_net_stats);
248	if (stats == NULL)
249		return -ENOMEM;
250
251	priv = netdev_priv(dev);
252	priv->stats = stats;
253	return 0;
254}
255
256static void veth_dev_free(struct net_device *dev)
257{
258	struct veth_priv *priv;
259
260	priv = netdev_priv(dev);
261	free_percpu(priv->stats);
262	free_netdev(dev);
263}
264
265static void veth_setup(struct net_device *dev)
266{
267	ether_setup(dev);
268
269	dev->hard_start_xmit = veth_xmit;
270	dev->get_stats = veth_get_stats;
271	dev->open = veth_open;
272	dev->ethtool_ops = &veth_ethtool_ops;
273	dev->features |= NETIF_F_LLTX;
274	dev->init = veth_dev_init;
275	dev->destructor = veth_dev_free;
276}
277
278static void veth_change_state(struct net_device *dev)
279{
280	struct net_device *peer;
281	struct veth_priv *priv;
282
283	priv = netdev_priv(dev);
284	peer = priv->peer;
285
286	if (netif_carrier_ok(peer)) {
287		if (!netif_carrier_ok(dev))
288			netif_carrier_on(dev);
289	} else {
290		if (netif_carrier_ok(dev))
291			netif_carrier_off(dev);
292	}
293}
294
295static int veth_device_event(struct notifier_block *unused,
296			     unsigned long event, void *ptr)
297{
298	struct net_device *dev = ptr;
299
300	if (dev->open != veth_open)
301		goto out;
302
303	switch (event) {
304	case NETDEV_CHANGE:
305		veth_change_state(dev);
306		break;
307	}
308out:
309	return NOTIFY_DONE;
310}
311
312static struct notifier_block veth_notifier_block __read_mostly = {
313	.notifier_call	= veth_device_event,
314};
315
316/*
317 * netlink interface
318 */
319
320static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
321{
322	if (tb[IFLA_ADDRESS]) {
323		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
324			return -EINVAL;
325		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
326			return -EADDRNOTAVAIL;
327	}
328	return 0;
329}
330
331static struct rtnl_link_ops veth_link_ops;
332
333static int veth_newlink(struct net_device *dev,
334			 struct nlattr *tb[], struct nlattr *data[])
335{
336	int err;
337	struct net_device *peer;
338	struct veth_priv *priv;
339	char ifname[IFNAMSIZ];
340	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
341
342	/*
343	 * create and register peer first
344	 *
345	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
346	 * skip it since no info from it is useful yet
347	 */
348
349	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
350		struct nlattr *nla_peer;
351
352		nla_peer = data[VETH_INFO_PEER];
353		err = nla_parse(peer_tb, IFLA_MAX,
354				nla_data(nla_peer) + sizeof(struct ifinfomsg),
355				nla_len(nla_peer) - sizeof(struct ifinfomsg),
356				ifla_policy);
357		if (err < 0)
358			return err;
359
360		err = veth_validate(peer_tb, NULL);
361		if (err < 0)
362			return err;
363
364		tbp = peer_tb;
365	} else
366		tbp = tb;
367
368	if (tbp[IFLA_IFNAME])
369		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
370	else
371		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
372
373	peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
374	if (IS_ERR(peer))
375		return PTR_ERR(peer);
376
377	if (tbp[IFLA_ADDRESS] == NULL)
378		random_ether_addr(peer->dev_addr);
379
380	err = register_netdevice(peer);
381	if (err < 0)
382		goto err_register_peer;
383
384	netif_carrier_off(peer);
385
386	/*
387	 * register dev last
388	 *
389	 * note, that since we've registered new device the dev's name
390	 * should be re-allocated
391	 */
392
393	if (tb[IFLA_ADDRESS] == NULL)
394		random_ether_addr(dev->dev_addr);
395
396	if (tb[IFLA_IFNAME])
397		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
398	else
399		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
400
401	if (strchr(dev->name, '%')) {
402		err = dev_alloc_name(dev, dev->name);
403		if (err < 0)
404			goto err_alloc_name;
405	}
406
407	err = register_netdevice(dev);
408	if (err < 0)
409		goto err_register_dev;
410
411	netif_carrier_off(dev);
412
413	/*
414	 * tie the deviced together
415	 */
416
417	priv = netdev_priv(dev);
418	priv->peer = peer;
419
420	priv = netdev_priv(peer);
421	priv->peer = dev;
422	return 0;
423
424err_register_dev:
425	/* nothing to do */
426err_alloc_name:
427	unregister_netdevice(peer);
428	return err;
429
430err_register_peer:
431	free_netdev(peer);
432	return err;
433}
434
435static void veth_dellink(struct net_device *dev)
436{
437	struct veth_priv *priv;
438	struct net_device *peer;
439
440	priv = netdev_priv(dev);
441	peer = priv->peer;
442
443	unregister_netdevice(dev);
444	unregister_netdevice(peer);
445}
446
447static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
448
449static struct rtnl_link_ops veth_link_ops = {
450	.kind		= DRV_NAME,
451	.priv_size	= sizeof(struct veth_priv),
452	.setup		= veth_setup,
453	.validate	= veth_validate,
454	.newlink	= veth_newlink,
455	.dellink	= veth_dellink,
456	.policy		= veth_policy,
457	.maxtype	= VETH_INFO_MAX,
458};
459
460/*
461 * init/fini
462 */
463
464static __init int veth_init(void)
465{
466	register_netdevice_notifier(&veth_notifier_block);
467	return rtnl_link_register(&veth_link_ops);
468}
469
470static __exit void veth_exit(void)
471{
472	rtnl_link_unregister(&veth_link_ops);
473	unregister_netdevice_notifier(&veth_notifier_block);
474}
475
476module_init(veth_init);
477module_exit(veth_exit);
478
479MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
480MODULE_LICENSE("GPL v2");
481MODULE_ALIAS_RTNL_LINK(DRV_NAME);
482