veth.c revision 38d408152a86598a50680a82fe3353b506630409 
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
22#define MIN_MTU 68		/* Min L3 MTU */
23#define MAX_MTU 65535		/* Max L3 MTU (arbitrary) */
24#define MTU_PAD (ETH_HLEN + 4)  /* Max difference between L2 and L3 size MTU */
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 *stats;
38	unsigned ip_summed;
39};
40
41/*
42 * ethtool interface
43 */
44
45static struct {
46	const char string[ETH_GSTRING_LEN];
47} ethtool_stats_keys[] = {
48	{ "peer_ifindex" },
49};
50
51static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
52{
53	cmd->supported		= 0;
54	cmd->advertising	= 0;
55	cmd->speed		= SPEED_10000;
56	cmd->duplex		= DUPLEX_FULL;
57	cmd->port		= PORT_TP;
58	cmd->phy_address	= 0;
59	cmd->transceiver	= XCVR_INTERNAL;
60	cmd->autoneg		= AUTONEG_DISABLE;
61	cmd->maxtxpkt		= 0;
62	cmd->maxrxpkt		= 0;
63	return 0;
64}
65
66static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
67{
68	strcpy(info->driver, DRV_NAME);
69	strcpy(info->version, DRV_VERSION);
70	strcpy(info->fw_version, "N/A");
71}
72
73static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
74{
75	switch(stringset) {
76	case ETH_SS_STATS:
77		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
78		break;
79	}
80}
81
82static int veth_get_sset_count(struct net_device *dev, int sset)
83{
84	switch (sset) {
85	case ETH_SS_STATS:
86		return ARRAY_SIZE(ethtool_stats_keys);
87	default:
88		return -EOPNOTSUPP;
89	}
90}
91
92static void veth_get_ethtool_stats(struct net_device *dev,
93		struct ethtool_stats *stats, u64 *data)
94{
95	struct veth_priv *priv;
96
97	priv = netdev_priv(dev);
98	data[0] = priv->peer->ifindex;
99}
100
101static u32 veth_get_rx_csum(struct net_device *dev)
102{
103	struct veth_priv *priv;
104
105	priv = netdev_priv(dev);
106	return priv->ip_summed == CHECKSUM_UNNECESSARY;
107}
108
109static int veth_set_rx_csum(struct net_device *dev, u32 data)
110{
111	struct veth_priv *priv;
112
113	priv = netdev_priv(dev);
114	priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
115	return 0;
116}
117
118static u32 veth_get_tx_csum(struct net_device *dev)
119{
120	return (dev->features & NETIF_F_NO_CSUM) != 0;
121}
122
123static int veth_set_tx_csum(struct net_device *dev, u32 data)
124{
125	if (data)
126		dev->features |= NETIF_F_NO_CSUM;
127	else
128		dev->features &= ~NETIF_F_NO_CSUM;
129	return 0;
130}
131
132static struct ethtool_ops veth_ethtool_ops = {
133	.get_settings		= veth_get_settings,
134	.get_drvinfo		= veth_get_drvinfo,
135	.get_link		= ethtool_op_get_link,
136	.get_rx_csum		= veth_get_rx_csum,
137	.set_rx_csum		= veth_set_rx_csum,
138	.get_tx_csum		= veth_get_tx_csum,
139	.set_tx_csum		= veth_set_tx_csum,
140	.get_sg			= ethtool_op_get_sg,
141	.set_sg			= ethtool_op_set_sg,
142	.get_strings		= veth_get_strings,
143	.get_sset_count		= veth_get_sset_count,
144	.get_ethtool_stats	= veth_get_ethtool_stats,
145};
146
147/*
148 * xmit
149 */
150
151static int veth_xmit(struct sk_buff *skb, struct net_device *dev)
152{
153	struct net_device *rcv = NULL;
154	struct veth_priv *priv, *rcv_priv;
155	struct veth_net_stats *stats, *rcv_stats;
156	int length, cpu;
157
158	skb_orphan(skb);
159
160	priv = netdev_priv(dev);
161	rcv = priv->peer;
162	rcv_priv = netdev_priv(rcv);
163
164	cpu = smp_processor_id();
165	stats = per_cpu_ptr(priv->stats, cpu);
166	rcv_stats = per_cpu_ptr(rcv_priv->stats, cpu);
167
168	if (!(rcv->flags & IFF_UP))
169		goto tx_drop;
170
171	if (skb->len > (rcv->mtu + MTU_PAD))
172		goto rx_drop;
173
174	skb->pkt_type = PACKET_HOST;
175	skb->protocol = eth_type_trans(skb, rcv);
176	if (dev->features & NETIF_F_NO_CSUM)
177		skb->ip_summed = rcv_priv->ip_summed;
178
179	dst_release(skb->dst);
180	skb->dst = NULL;
181	skb->mark = 0;
182	secpath_reset(skb);
183	nf_reset(skb);
184
185	length = skb->len;
186
187	stats->tx_bytes += length;
188	stats->tx_packets++;
189
190	rcv_stats->rx_bytes += length;
191	rcv_stats->rx_packets++;
192
193	netif_rx(skb);
194	return 0;
195
196tx_drop:
197	kfree_skb(skb);
198	stats->tx_dropped++;
199	return 0;
200
201rx_drop:
202	kfree_skb(skb);
203	rcv_stats->rx_dropped++;
204	return 0;
205}
206
207/*
208 * general routines
209 */
210
211static struct net_device_stats *veth_get_stats(struct net_device *dev)
212{
213	struct veth_priv *priv;
214	struct net_device_stats *dev_stats;
215	int cpu;
216	struct veth_net_stats *stats;
217
218	priv = netdev_priv(dev);
219	dev_stats = &dev->stats;
220
221	dev_stats->rx_packets = 0;
222	dev_stats->tx_packets = 0;
223	dev_stats->rx_bytes = 0;
224	dev_stats->tx_bytes = 0;
225	dev_stats->tx_dropped = 0;
226	dev_stats->rx_dropped = 0;
227
228	for_each_online_cpu(cpu) {
229		stats = per_cpu_ptr(priv->stats, cpu);
230
231		dev_stats->rx_packets += stats->rx_packets;
232		dev_stats->tx_packets += stats->tx_packets;
233		dev_stats->rx_bytes += stats->rx_bytes;
234		dev_stats->tx_bytes += stats->tx_bytes;
235		dev_stats->tx_dropped += stats->tx_dropped;
236		dev_stats->rx_dropped += stats->rx_dropped;
237	}
238
239	return dev_stats;
240}
241
242static int veth_open(struct net_device *dev)
243{
244	struct veth_priv *priv;
245
246	priv = netdev_priv(dev);
247	if (priv->peer == NULL)
248		return -ENOTCONN;
249
250	if (priv->peer->flags & IFF_UP) {
251		netif_carrier_on(dev);
252		netif_carrier_on(priv->peer);
253	}
254	return 0;
255}
256
257static int veth_close(struct net_device *dev)
258{
259	struct veth_priv *priv = netdev_priv(dev);
260
261	netif_carrier_off(dev);
262	netif_carrier_off(priv->peer);
263
264	return 0;
265}
266
267static int is_valid_veth_mtu(int new_mtu)
268{
269	return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
270}
271
272static int veth_change_mtu(struct net_device *dev, int new_mtu)
273{
274	if (!is_valid_veth_mtu(new_mtu))
275		return -EINVAL;
276	dev->mtu = new_mtu;
277	return 0;
278}
279
280static int veth_dev_init(struct net_device *dev)
281{
282	struct veth_net_stats *stats;
283	struct veth_priv *priv;
284
285	stats = alloc_percpu(struct veth_net_stats);
286	if (stats == NULL)
287		return -ENOMEM;
288
289	priv = netdev_priv(dev);
290	priv->stats = stats;
291	return 0;
292}
293
294static void veth_dev_free(struct net_device *dev)
295{
296	struct veth_priv *priv;
297
298	priv = netdev_priv(dev);
299	free_percpu(priv->stats);
300	free_netdev(dev);
301}
302
303static const struct net_device_ops veth_netdev_ops = {
304	.ndo_init            = veth_dev_init,
305	.ndo_open            = veth_open,
306	.ndo_stop            = veth_close,
307	.ndo_start_xmit      = veth_xmit,
308	.ndo_change_mtu      = veth_change_mtu,
309	.ndo_get_stats       = veth_get_stats,
310	.ndo_set_mac_address = eth_mac_addr,
311};
312
313static void veth_setup(struct net_device *dev)
314{
315	ether_setup(dev);
316
317	dev->netdev_ops = &veth_netdev_ops;
318	dev->ethtool_ops = &veth_ethtool_ops;
319	dev->features |= NETIF_F_LLTX;
320	dev->destructor = veth_dev_free;
321}
322
323/*
324 * netlink interface
325 */
326
327static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
328{
329	if (tb[IFLA_ADDRESS]) {
330		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
331			return -EINVAL;
332		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
333			return -EADDRNOTAVAIL;
334	}
335	if (tb[IFLA_MTU]) {
336		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
337			return -EINVAL;
338	}
339	return 0;
340}
341
342static struct rtnl_link_ops veth_link_ops;
343
344static int veth_newlink(struct net_device *dev,
345			 struct nlattr *tb[], struct nlattr *data[])
346{
347	int err;
348	struct net_device *peer;
349	struct veth_priv *priv;
350	char ifname[IFNAMSIZ];
351	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
352
353	/*
354	 * create and register peer first
355	 *
356	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
357	 * skip it since no info from it is useful yet
358	 */
359
360	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
361		struct nlattr *nla_peer;
362
363		nla_peer = data[VETH_INFO_PEER];
364		err = nla_parse(peer_tb, IFLA_MAX,
365				nla_data(nla_peer) + sizeof(struct ifinfomsg),
366				nla_len(nla_peer) - sizeof(struct ifinfomsg),
367				ifla_policy);
368		if (err < 0)
369			return err;
370
371		err = veth_validate(peer_tb, NULL);
372		if (err < 0)
373			return err;
374
375		tbp = peer_tb;
376	} else
377		tbp = tb;
378
379	if (tbp[IFLA_IFNAME])
380		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
381	else
382		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
383
384	peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
385	if (IS_ERR(peer))
386		return PTR_ERR(peer);
387
388	if (tbp[IFLA_ADDRESS] == NULL)
389		random_ether_addr(peer->dev_addr);
390
391	err = register_netdevice(peer);
392	if (err < 0)
393		goto err_register_peer;
394
395	netif_carrier_off(peer);
396
397	/*
398	 * register dev last
399	 *
400	 * note, that since we've registered new device the dev's name
401	 * should be re-allocated
402	 */
403
404	if (tb[IFLA_ADDRESS] == NULL)
405		random_ether_addr(dev->dev_addr);
406
407	if (tb[IFLA_IFNAME])
408		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
409	else
410		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
411
412	if (strchr(dev->name, '%')) {
413		err = dev_alloc_name(dev, dev->name);
414		if (err < 0)
415			goto err_alloc_name;
416	}
417
418	err = register_netdevice(dev);
419	if (err < 0)
420		goto err_register_dev;
421
422	netif_carrier_off(dev);
423
424	/*
425	 * tie the deviced together
426	 */
427
428	priv = netdev_priv(dev);
429	priv->peer = peer;
430
431	priv = netdev_priv(peer);
432	priv->peer = dev;
433	return 0;
434
435err_register_dev:
436	/* nothing to do */
437err_alloc_name:
438	unregister_netdevice(peer);
439	return err;
440
441err_register_peer:
442	free_netdev(peer);
443	return err;
444}
445
446static void veth_dellink(struct net_device *dev)
447{
448	struct veth_priv *priv;
449	struct net_device *peer;
450
451	priv = netdev_priv(dev);
452	peer = priv->peer;
453
454	unregister_netdevice(dev);
455	unregister_netdevice(peer);
456}
457
458static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
459
460static struct rtnl_link_ops veth_link_ops = {
461	.kind		= DRV_NAME,
462	.priv_size	= sizeof(struct veth_priv),
463	.setup		= veth_setup,
464	.validate	= veth_validate,
465	.newlink	= veth_newlink,
466	.dellink	= veth_dellink,
467	.policy		= veth_policy,
468	.maxtype	= VETH_INFO_MAX,
469};
470
471/*
472 * init/fini
473 */
474
475static __init int veth_init(void)
476{
477	return rtnl_link_register(&veth_link_ops);
478}
479
480static __exit void veth_exit(void)
481{
482	rtnl_link_unregister(&veth_link_ops);
483}
484
485module_init(veth_init);
486module_exit(veth_exit);
487
488MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
489MODULE_LICENSE("GPL v2");
490MODULE_ALIAS_RTNL_LINK(DRV_NAME);
491