veth.c revision bb7bba3d56963be59bc6764c8021290ed55205ad 
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/list.h>
12#include <linux/netdevice.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
23struct veth_net_stats {
24	unsigned long	rx_packets;
25	unsigned long	tx_packets;
26	unsigned long	rx_bytes;
27	unsigned long	tx_bytes;
28	unsigned long	tx_dropped;
29};
30
31struct veth_priv {
32	struct net_device *peer;
33	struct list_head list;
34	struct veth_net_stats *stats;
35	unsigned ip_summed;
36};
37
38static LIST_HEAD(veth_list);
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	cmd->speed		= 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 u32 veth_get_rx_csum(struct net_device *dev)
101{
102	struct veth_priv *priv;
103
104	priv = netdev_priv(dev);
105	return priv->ip_summed == CHECKSUM_UNNECESSARY;
106}
107
108static int veth_set_rx_csum(struct net_device *dev, u32 data)
109{
110	struct veth_priv *priv;
111
112	priv = netdev_priv(dev);
113	priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
114	return 0;
115}
116
117static u32 veth_get_tx_csum(struct net_device *dev)
118{
119	return (dev->features & NETIF_F_NO_CSUM) != 0;
120}
121
122static int veth_set_tx_csum(struct net_device *dev, u32 data)
123{
124	if (data)
125		dev->features |= NETIF_F_NO_CSUM;
126	else
127		dev->features &= ~NETIF_F_NO_CSUM;
128	return 0;
129}
130
131static struct ethtool_ops veth_ethtool_ops = {
132	.get_settings		= veth_get_settings,
133	.get_drvinfo		= veth_get_drvinfo,
134	.get_link		= ethtool_op_get_link,
135	.get_rx_csum		= veth_get_rx_csum,
136	.set_rx_csum		= veth_set_rx_csum,
137	.get_tx_csum		= veth_get_tx_csum,
138	.set_tx_csum		= veth_set_tx_csum,
139	.get_sg			= ethtool_op_get_sg,
140	.set_sg			= ethtool_op_set_sg,
141	.get_strings		= veth_get_strings,
142	.get_sset_count		= veth_get_sset_count,
143	.get_ethtool_stats	= veth_get_ethtool_stats,
144};
145
146/*
147 * xmit
148 */
149
150static int veth_xmit(struct sk_buff *skb, struct net_device *dev)
151{
152	struct net_device *rcv = NULL;
153	struct veth_priv *priv, *rcv_priv;
154	struct veth_net_stats *stats;
155	int length, cpu;
156
157	skb_orphan(skb);
158
159	priv = netdev_priv(dev);
160	rcv = priv->peer;
161	rcv_priv = netdev_priv(rcv);
162
163	cpu = smp_processor_id();
164	stats = per_cpu_ptr(priv->stats, cpu);
165
166	if (!(rcv->flags & IFF_UP))
167		goto outf;
168
169	skb->pkt_type = PACKET_HOST;
170	skb->protocol = eth_type_trans(skb, rcv);
171	if (dev->features & NETIF_F_NO_CSUM)
172		skb->ip_summed = rcv_priv->ip_summed;
173
174	dst_release(skb->dst);
175	skb->dst = NULL;
176	skb->mark = 0;
177	secpath_reset(skb);
178	nf_reset(skb);
179
180	length = skb->len;
181
182	stats->tx_bytes += length;
183	stats->tx_packets++;
184
185	stats = per_cpu_ptr(rcv_priv->stats, cpu);
186	stats->rx_bytes += length;
187	stats->rx_packets++;
188
189	netif_rx(skb);
190	return 0;
191
192outf:
193	kfree_skb(skb);
194	stats->tx_dropped++;
195	return 0;
196}
197
198/*
199 * general routines
200 */
201
202static struct net_device_stats *veth_get_stats(struct net_device *dev)
203{
204	struct veth_priv *priv;
205	struct net_device_stats *dev_stats;
206	int cpu;
207	struct veth_net_stats *stats;
208
209	priv = netdev_priv(dev);
210	dev_stats = &dev->stats;
211
212	dev_stats->rx_packets = 0;
213	dev_stats->tx_packets = 0;
214	dev_stats->rx_bytes = 0;
215	dev_stats->tx_bytes = 0;
216	dev_stats->tx_dropped = 0;
217
218	for_each_online_cpu(cpu) {
219		stats = per_cpu_ptr(priv->stats, cpu);
220
221		dev_stats->rx_packets += stats->rx_packets;
222		dev_stats->tx_packets += stats->tx_packets;
223		dev_stats->rx_bytes += stats->rx_bytes;
224		dev_stats->tx_bytes += stats->tx_bytes;
225		dev_stats->tx_dropped += stats->tx_dropped;
226	}
227
228	return dev_stats;
229}
230
231static int veth_open(struct net_device *dev)
232{
233	struct veth_priv *priv;
234
235	priv = netdev_priv(dev);
236	if (priv->peer == NULL)
237		return -ENOTCONN;
238
239	if (priv->peer->flags & IFF_UP) {
240		netif_carrier_on(dev);
241		netif_carrier_on(priv->peer);
242	}
243	return 0;
244}
245
246static int veth_dev_init(struct net_device *dev)
247{
248	struct veth_net_stats *stats;
249	struct veth_priv *priv;
250
251	stats = alloc_percpu(struct veth_net_stats);
252	if (stats == NULL)
253		return -ENOMEM;
254
255	priv = netdev_priv(dev);
256	priv->stats = stats;
257	return 0;
258}
259
260static void veth_dev_free(struct net_device *dev)
261{
262	struct veth_priv *priv;
263
264	priv = netdev_priv(dev);
265	free_percpu(priv->stats);
266	free_netdev(dev);
267}
268
269static void veth_setup(struct net_device *dev)
270{
271	ether_setup(dev);
272
273	dev->hard_start_xmit = veth_xmit;
274	dev->get_stats = veth_get_stats;
275	dev->open = veth_open;
276	dev->ethtool_ops = &veth_ethtool_ops;
277	dev->features |= NETIF_F_LLTX;
278	dev->init = veth_dev_init;
279	dev->destructor = veth_dev_free;
280}
281
282static void veth_change_state(struct net_device *dev)
283{
284	struct net_device *peer;
285	struct veth_priv *priv;
286
287	priv = netdev_priv(dev);
288	peer = priv->peer;
289
290	if (netif_carrier_ok(peer)) {
291		if (!netif_carrier_ok(dev))
292			netif_carrier_on(dev);
293	} else {
294		if (netif_carrier_ok(dev))
295			netif_carrier_off(dev);
296	}
297}
298
299static int veth_device_event(struct notifier_block *unused,
300			     unsigned long event, void *ptr)
301{
302	struct net_device *dev = ptr;
303
304	if (dev->open != veth_open)
305		goto out;
306
307	switch (event) {
308	case NETDEV_CHANGE:
309		veth_change_state(dev);
310		break;
311	}
312out:
313	return NOTIFY_DONE;
314}
315
316static struct notifier_block veth_notifier_block __read_mostly = {
317	.notifier_call	= veth_device_event,
318};
319
320/*
321 * netlink interface
322 */
323
324static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
325{
326	if (tb[IFLA_ADDRESS]) {
327		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
328			return -EINVAL;
329		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
330			return -EADDRNOTAVAIL;
331	}
332	return 0;
333}
334
335static struct rtnl_link_ops veth_link_ops;
336
337static int veth_newlink(struct net_device *dev,
338			 struct nlattr *tb[], struct nlattr *data[])
339{
340	int err;
341	struct net_device *peer;
342	struct veth_priv *priv;
343	char ifname[IFNAMSIZ];
344	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
345
346	/*
347	 * create and register peer first
348	 *
349	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
350	 * skip it since no info from it is useful yet
351	 */
352
353	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
354		struct nlattr *nla_peer;
355
356		nla_peer = data[VETH_INFO_PEER];
357		err = nla_parse(peer_tb, IFLA_MAX,
358				nla_data(nla_peer) + sizeof(struct ifinfomsg),
359				nla_len(nla_peer) - sizeof(struct ifinfomsg),
360				ifla_policy);
361		if (err < 0)
362			return err;
363
364		err = veth_validate(peer_tb, NULL);
365		if (err < 0)
366			return err;
367
368		tbp = peer_tb;
369	} else
370		tbp = tb;
371
372	if (tbp[IFLA_IFNAME])
373		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
374	else
375		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
376
377	peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
378	if (IS_ERR(peer))
379		return PTR_ERR(peer);
380
381	if (tbp[IFLA_ADDRESS] == NULL)
382		random_ether_addr(peer->dev_addr);
383
384	err = register_netdevice(peer);
385	if (err < 0)
386		goto err_register_peer;
387
388	netif_carrier_off(peer);
389
390	/*
391	 * register dev last
392	 *
393	 * note, that since we've registered new device the dev's name
394	 * should be re-allocated
395	 */
396
397	if (tb[IFLA_ADDRESS] == NULL)
398		random_ether_addr(dev->dev_addr);
399
400	if (tb[IFLA_IFNAME])
401		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
402	else
403		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
404
405	if (strchr(dev->name, '%')) {
406		err = dev_alloc_name(dev, dev->name);
407		if (err < 0)
408			goto err_alloc_name;
409	}
410
411	err = register_netdevice(dev);
412	if (err < 0)
413		goto err_register_dev;
414
415	netif_carrier_off(dev);
416
417	/*
418	 * tie the deviced together
419	 */
420
421	priv = netdev_priv(dev);
422	priv->peer = peer;
423	list_add(&priv->list, &veth_list);
424
425	priv = netdev_priv(peer);
426	priv->peer = dev;
427	INIT_LIST_HEAD(&priv->list);
428	return 0;
429
430err_register_dev:
431	/* nothing to do */
432err_alloc_name:
433	unregister_netdevice(peer);
434	return err;
435
436err_register_peer:
437	free_netdev(peer);
438	return err;
439}
440
441static void veth_dellink(struct net_device *dev)
442{
443	struct veth_priv *priv;
444	struct net_device *peer;
445
446	priv = netdev_priv(dev);
447	peer = priv->peer;
448
449	if (!list_empty(&priv->list))
450		list_del(&priv->list);
451
452	priv = netdev_priv(peer);
453	if (!list_empty(&priv->list))
454		list_del(&priv->list);
455
456	unregister_netdevice(dev);
457	unregister_netdevice(peer);
458}
459
460static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
461
462static struct rtnl_link_ops veth_link_ops = {
463	.kind		= DRV_NAME,
464	.priv_size	= sizeof(struct veth_priv),
465	.setup		= veth_setup,
466	.validate	= veth_validate,
467	.newlink	= veth_newlink,
468	.dellink	= veth_dellink,
469	.policy		= veth_policy,
470	.maxtype	= VETH_INFO_MAX,
471};
472
473/*
474 * init/fini
475 */
476
477static __init int veth_init(void)
478{
479	register_netdevice_notifier(&veth_notifier_block);
480	return rtnl_link_register(&veth_link_ops);
481}
482
483static __exit void veth_exit(void)
484{
485	rtnl_link_unregister(&veth_link_ops);
486	unregister_netdevice_notifier(&veth_notifier_block);
487}
488
489module_init(veth_init);
490module_exit(veth_exit);
491
492MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
493MODULE_LICENSE("GPL v2");
494MODULE_ALIAS_RTNL_LINK(DRV_NAME);
495