veth.c revision 81b16ba2f1cc93a1ee1dda48be2ea2d91a0cb72e 
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#include <linux/u64_stats_sync.h>
16
17#include <net/dst.h>
18#include <net/xfrm.h>
19#include <linux/veth.h>
20
21#define DRV_NAME	"veth"
22#define DRV_VERSION	"1.0"
23
24#define MIN_MTU 68		/* Min L3 MTU */
25#define MAX_MTU 65535		/* Max L3 MTU (arbitrary) */
26
27struct veth_net_stats {
28	u64			rx_packets;
29	u64			tx_packets;
30	u64			rx_bytes;
31	u64			tx_bytes;
32	u64			rx_dropped;
33	struct u64_stats_sync	syncp;
34};
35
36struct veth_priv {
37	struct net_device *peer;
38	struct veth_net_stats __percpu *stats;
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	ethtool_cmd_speed_set(cmd, 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 const struct ethtool_ops veth_ethtool_ops = {
102	.get_settings		= veth_get_settings,
103	.get_drvinfo		= veth_get_drvinfo,
104	.get_link		= ethtool_op_get_link,
105	.get_strings		= veth_get_strings,
106	.get_sset_count		= veth_get_sset_count,
107	.get_ethtool_stats	= veth_get_ethtool_stats,
108};
109
110/*
111 * xmit
112 */
113
114static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
115{
116	struct net_device *rcv = NULL;
117	struct veth_priv *priv, *rcv_priv;
118	struct veth_net_stats *stats, *rcv_stats;
119	int length;
120
121	priv = netdev_priv(dev);
122	rcv = priv->peer;
123	rcv_priv = netdev_priv(rcv);
124
125	stats = this_cpu_ptr(priv->stats);
126	rcv_stats = this_cpu_ptr(rcv_priv->stats);
127
128	/* don't change ip_summed == CHECKSUM_PARTIAL, as that
129	   will cause bad checksum on forwarded packets */
130	if (skb->ip_summed == CHECKSUM_NONE &&
131	    rcv->features & NETIF_F_RXCSUM)
132		skb->ip_summed = CHECKSUM_UNNECESSARY;
133
134	length = skb->len;
135	if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
136		goto rx_drop;
137
138	u64_stats_update_begin(&stats->syncp);
139	stats->tx_bytes += length;
140	stats->tx_packets++;
141	u64_stats_update_end(&stats->syncp);
142
143	u64_stats_update_begin(&rcv_stats->syncp);
144	rcv_stats->rx_bytes += length;
145	rcv_stats->rx_packets++;
146	u64_stats_update_end(&rcv_stats->syncp);
147
148	return NETDEV_TX_OK;
149
150rx_drop:
151	u64_stats_update_begin(&rcv_stats->syncp);
152	rcv_stats->rx_dropped++;
153	u64_stats_update_end(&rcv_stats->syncp);
154	return NETDEV_TX_OK;
155}
156
157/*
158 * general routines
159 */
160
161static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
162						  struct rtnl_link_stats64 *tot)
163{
164	struct veth_priv *priv = netdev_priv(dev);
165	int cpu;
166
167	for_each_possible_cpu(cpu) {
168		struct veth_net_stats *stats = per_cpu_ptr(priv->stats, cpu);
169		u64 rx_packets, rx_bytes, rx_dropped;
170		u64 tx_packets, tx_bytes;
171		unsigned int start;
172
173		do {
174			start = u64_stats_fetch_begin_bh(&stats->syncp);
175			rx_packets = stats->rx_packets;
176			tx_packets = stats->tx_packets;
177			rx_bytes = stats->rx_bytes;
178			tx_bytes = stats->tx_bytes;
179			rx_dropped = stats->rx_dropped;
180		} while (u64_stats_fetch_retry_bh(&stats->syncp, start));
181		tot->rx_packets += rx_packets;
182		tot->tx_packets += tx_packets;
183		tot->rx_bytes   += rx_bytes;
184		tot->tx_bytes   += tx_bytes;
185		tot->rx_dropped += rx_dropped;
186	}
187
188	return tot;
189}
190
191static int veth_open(struct net_device *dev)
192{
193	struct veth_priv *priv;
194
195	priv = netdev_priv(dev);
196	if (priv->peer == NULL)
197		return -ENOTCONN;
198
199	if (priv->peer->flags & IFF_UP) {
200		netif_carrier_on(dev);
201		netif_carrier_on(priv->peer);
202	}
203	return 0;
204}
205
206static int veth_close(struct net_device *dev)
207{
208	struct veth_priv *priv = netdev_priv(dev);
209
210	netif_carrier_off(dev);
211	netif_carrier_off(priv->peer);
212
213	return 0;
214}
215
216static int is_valid_veth_mtu(int new_mtu)
217{
218	return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
219}
220
221static int veth_change_mtu(struct net_device *dev, int new_mtu)
222{
223	if (!is_valid_veth_mtu(new_mtu))
224		return -EINVAL;
225	dev->mtu = new_mtu;
226	return 0;
227}
228
229static int veth_dev_init(struct net_device *dev)
230{
231	struct veth_net_stats __percpu *stats;
232	struct veth_priv *priv;
233
234	stats = alloc_percpu(struct veth_net_stats);
235	if (stats == NULL)
236		return -ENOMEM;
237
238	priv = netdev_priv(dev);
239	priv->stats = stats;
240	return 0;
241}
242
243static void veth_dev_free(struct net_device *dev)
244{
245	struct veth_priv *priv;
246
247	priv = netdev_priv(dev);
248	free_percpu(priv->stats);
249	free_netdev(dev);
250}
251
252static const struct net_device_ops veth_netdev_ops = {
253	.ndo_init            = veth_dev_init,
254	.ndo_open            = veth_open,
255	.ndo_stop            = veth_close,
256	.ndo_start_xmit      = veth_xmit,
257	.ndo_change_mtu      = veth_change_mtu,
258	.ndo_get_stats64     = veth_get_stats64,
259	.ndo_set_mac_address = eth_mac_addr,
260};
261
262static void veth_setup(struct net_device *dev)
263{
264	ether_setup(dev);
265
266	dev->netdev_ops = &veth_netdev_ops;
267	dev->ethtool_ops = &veth_ethtool_ops;
268	dev->features |= NETIF_F_LLTX;
269	dev->destructor = veth_dev_free;
270
271	dev->hw_features = NETIF_F_NO_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
272}
273
274/*
275 * netlink interface
276 */
277
278static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
279{
280	if (tb[IFLA_ADDRESS]) {
281		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
282			return -EINVAL;
283		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
284			return -EADDRNOTAVAIL;
285	}
286	if (tb[IFLA_MTU]) {
287		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
288			return -EINVAL;
289	}
290	return 0;
291}
292
293static struct rtnl_link_ops veth_link_ops;
294
295static int veth_newlink(struct net *src_net, struct net_device *dev,
296			 struct nlattr *tb[], struct nlattr *data[])
297{
298	int err;
299	struct net_device *peer;
300	struct veth_priv *priv;
301	char ifname[IFNAMSIZ];
302	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
303	struct ifinfomsg *ifmp;
304	struct net *net;
305
306	/*
307	 * create and register peer first
308	 */
309	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
310		struct nlattr *nla_peer;
311
312		nla_peer = data[VETH_INFO_PEER];
313		ifmp = nla_data(nla_peer);
314		err = nla_parse(peer_tb, IFLA_MAX,
315				nla_data(nla_peer) + sizeof(struct ifinfomsg),
316				nla_len(nla_peer) - sizeof(struct ifinfomsg),
317				ifla_policy);
318		if (err < 0)
319			return err;
320
321		err = veth_validate(peer_tb, NULL);
322		if (err < 0)
323			return err;
324
325		tbp = peer_tb;
326	} else {
327		ifmp = NULL;
328		tbp = tb;
329	}
330
331	if (tbp[IFLA_IFNAME])
332		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
333	else
334		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
335
336	net = rtnl_link_get_net(src_net, tbp);
337	if (IS_ERR(net))
338		return PTR_ERR(net);
339
340	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
341	if (IS_ERR(peer)) {
342		put_net(net);
343		return PTR_ERR(peer);
344	}
345
346	if (tbp[IFLA_ADDRESS] == NULL)
347		random_ether_addr(peer->dev_addr);
348
349	err = register_netdevice(peer);
350	put_net(net);
351	net = NULL;
352	if (err < 0)
353		goto err_register_peer;
354
355	netif_carrier_off(peer);
356
357	err = rtnl_configure_link(peer, ifmp);
358	if (err < 0)
359		goto err_configure_peer;
360
361	/*
362	 * register dev last
363	 *
364	 * note, that since we've registered new device the dev's name
365	 * should be re-allocated
366	 */
367
368	if (tb[IFLA_ADDRESS] == NULL)
369		random_ether_addr(dev->dev_addr);
370
371	if (tb[IFLA_IFNAME])
372		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
373	else
374		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
375
376	if (strchr(dev->name, '%')) {
377		err = dev_alloc_name(dev, dev->name);
378		if (err < 0)
379			goto err_alloc_name;
380	}
381
382	err = register_netdevice(dev);
383	if (err < 0)
384		goto err_register_dev;
385
386	netif_carrier_off(dev);
387
388	/*
389	 * tie the deviced together
390	 */
391
392	priv = netdev_priv(dev);
393	priv->peer = peer;
394
395	priv = netdev_priv(peer);
396	priv->peer = dev;
397	return 0;
398
399err_register_dev:
400	/* nothing to do */
401err_alloc_name:
402err_configure_peer:
403	unregister_netdevice(peer);
404	return err;
405
406err_register_peer:
407	free_netdev(peer);
408	return err;
409}
410
411static void veth_dellink(struct net_device *dev, struct list_head *head)
412{
413	struct veth_priv *priv;
414	struct net_device *peer;
415
416	priv = netdev_priv(dev);
417	peer = priv->peer;
418
419	unregister_netdevice_queue(dev, head);
420	unregister_netdevice_queue(peer, head);
421}
422
423static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
424
425static struct rtnl_link_ops veth_link_ops = {
426	.kind		= DRV_NAME,
427	.priv_size	= sizeof(struct veth_priv),
428	.setup		= veth_setup,
429	.validate	= veth_validate,
430	.newlink	= veth_newlink,
431	.dellink	= veth_dellink,
432	.policy		= veth_policy,
433	.maxtype	= VETH_INFO_MAX,
434};
435
436/*
437 * init/fini
438 */
439
440static __init int veth_init(void)
441{
442	return rtnl_link_register(&veth_link_ops);
443}
444
445static __exit void veth_exit(void)
446{
447	rtnl_link_unregister(&veth_link_ops);
448}
449
450module_init(veth_init);
451module_exit(veth_exit);
452
453MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
454MODULE_LICENSE("GPL v2");
455MODULE_ALIAS_RTNL_LINK(DRV_NAME);
456