Linux Audio

Check our new training course

Loading...
v3.1
  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->priv_flags &= ~IFF_TX_SKB_SHARING;
 
 
 
267
268	dev->netdev_ops = &veth_netdev_ops;
269	dev->ethtool_ops = &veth_ethtool_ops;
270	dev->features |= NETIF_F_LLTX;
 
 
 
 
 
 
271	dev->destructor = veth_dev_free;
 
272
273	dev->hw_features = NETIF_F_NO_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
 
 
274}
275
276/*
277 * netlink interface
278 */
279
280static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
281{
282	if (tb[IFLA_ADDRESS]) {
283		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
284			return -EINVAL;
285		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
286			return -EADDRNOTAVAIL;
287	}
288	if (tb[IFLA_MTU]) {
289		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
290			return -EINVAL;
291	}
292	return 0;
293}
294
295static struct rtnl_link_ops veth_link_ops;
296
297static int veth_newlink(struct net *src_net, struct net_device *dev,
298			 struct nlattr *tb[], struct nlattr *data[])
299{
300	int err;
301	struct net_device *peer;
302	struct veth_priv *priv;
303	char ifname[IFNAMSIZ];
304	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 
305	struct ifinfomsg *ifmp;
306	struct net *net;
307
308	/*
309	 * create and register peer first
310	 */
311	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
312		struct nlattr *nla_peer;
313
314		nla_peer = data[VETH_INFO_PEER];
315		ifmp = nla_data(nla_peer);
316		err = nla_parse(peer_tb, IFLA_MAX,
317				nla_data(nla_peer) + sizeof(struct ifinfomsg),
318				nla_len(nla_peer) - sizeof(struct ifinfomsg),
319				ifla_policy);
320		if (err < 0)
321			return err;
322
323		err = veth_validate(peer_tb, NULL);
324		if (err < 0)
325			return err;
326
327		tbp = peer_tb;
328	} else {
329		ifmp = NULL;
330		tbp = tb;
331	}
332
333	if (tbp[IFLA_IFNAME])
334		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
335	else
 
336		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 
 
337
338	net = rtnl_link_get_net(src_net, tbp);
339	if (IS_ERR(net))
340		return PTR_ERR(net);
341
342	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
 
343	if (IS_ERR(peer)) {
344		put_net(net);
345		return PTR_ERR(peer);
346	}
347
348	if (tbp[IFLA_ADDRESS] == NULL)
349		random_ether_addr(peer->dev_addr);
 
 
 
350
351	err = register_netdevice(peer);
352	put_net(net);
353	net = NULL;
354	if (err < 0)
355		goto err_register_peer;
356
357	netif_carrier_off(peer);
358
359	err = rtnl_configure_link(peer, ifmp);
360	if (err < 0)
361		goto err_configure_peer;
362
363	/*
364	 * register dev last
365	 *
366	 * note, that since we've registered new device the dev's name
367	 * should be re-allocated
368	 */
369
370	if (tb[IFLA_ADDRESS] == NULL)
371		random_ether_addr(dev->dev_addr);
372
373	if (tb[IFLA_IFNAME])
374		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
375	else
376		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
377
378	if (strchr(dev->name, '%')) {
379		err = dev_alloc_name(dev, dev->name);
380		if (err < 0)
381			goto err_alloc_name;
382	}
383
384	err = register_netdevice(dev);
385	if (err < 0)
386		goto err_register_dev;
387
388	netif_carrier_off(dev);
389
390	/*
391	 * tie the deviced together
392	 */
393
394	priv = netdev_priv(dev);
395	priv->peer = peer;
396
397	priv = netdev_priv(peer);
398	priv->peer = dev;
399	return 0;
400
401err_register_dev:
402	/* nothing to do */
403err_alloc_name:
404err_configure_peer:
405	unregister_netdevice(peer);
406	return err;
407
408err_register_peer:
409	free_netdev(peer);
410	return err;
411}
412
413static void veth_dellink(struct net_device *dev, struct list_head *head)
414{
415	struct veth_priv *priv;
416	struct net_device *peer;
417
418	priv = netdev_priv(dev);
419	peer = priv->peer;
420
 
 
 
 
 
421	unregister_netdevice_queue(dev, head);
422	unregister_netdevice_queue(peer, head);
 
 
 
 
 
423}
424
425static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 
 
 
 
 
 
 
 
 
 
426
427static struct rtnl_link_ops veth_link_ops = {
428	.kind		= DRV_NAME,
429	.priv_size	= sizeof(struct veth_priv),
430	.setup		= veth_setup,
431	.validate	= veth_validate,
432	.newlink	= veth_newlink,
433	.dellink	= veth_dellink,
434	.policy		= veth_policy,
435	.maxtype	= VETH_INFO_MAX,
 
436};
437
438/*
439 * init/fini
440 */
441
442static __init int veth_init(void)
443{
444	return rtnl_link_register(&veth_link_ops);
445}
446
447static __exit void veth_exit(void)
448{
449	rtnl_link_unregister(&veth_link_ops);
450}
451
452module_init(veth_init);
453module_exit(veth_exit);
454
455MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
456MODULE_LICENSE("GPL v2");
457MODULE_ALIAS_RTNL_LINK(DRV_NAME);
v4.10.11
  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/rtnetlink.h>
 18#include <net/dst.h>
 19#include <net/xfrm.h>
 20#include <linux/veth.h>
 21#include <linux/module.h>
 22
 23#define DRV_NAME	"veth"
 24#define DRV_VERSION	"1.0"
 25
 26struct pcpu_vstats {
 27	u64			packets;
 28	u64			bytes;
 
 
 
 
 
 
 29	struct u64_stats_sync	syncp;
 30};
 31
 32struct veth_priv {
 33	struct net_device __rcu	*peer;
 34	atomic64_t		dropped;
 35	unsigned		requested_headroom;
 36};
 37
 38/*
 39 * ethtool interface
 40 */
 41
 42static struct {
 43	const char string[ETH_GSTRING_LEN];
 44} ethtool_stats_keys[] = {
 45	{ "peer_ifindex" },
 46};
 47
 48static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
 49{
 50	cmd->supported		= 0;
 51	cmd->advertising	= 0;
 52	ethtool_cmd_speed_set(cmd, SPEED_10000);
 53	cmd->duplex		= DUPLEX_FULL;
 54	cmd->port		= PORT_TP;
 55	cmd->phy_address	= 0;
 56	cmd->transceiver	= XCVR_INTERNAL;
 57	cmd->autoneg		= AUTONEG_DISABLE;
 58	cmd->maxtxpkt		= 0;
 59	cmd->maxrxpkt		= 0;
 60	return 0;
 61}
 62
 63static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 64{
 65	strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
 66	strlcpy(info->version, DRV_VERSION, sizeof(info->version));
 
 67}
 68
 69static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
 70{
 71	switch(stringset) {
 72	case ETH_SS_STATS:
 73		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
 74		break;
 75	}
 76}
 77
 78static int veth_get_sset_count(struct net_device *dev, int sset)
 79{
 80	switch (sset) {
 81	case ETH_SS_STATS:
 82		return ARRAY_SIZE(ethtool_stats_keys);
 83	default:
 84		return -EOPNOTSUPP;
 85	}
 86}
 87
 88static void veth_get_ethtool_stats(struct net_device *dev,
 89		struct ethtool_stats *stats, u64 *data)
 90{
 91	struct veth_priv *priv = netdev_priv(dev);
 92	struct net_device *peer = rtnl_dereference(priv->peer);
 93
 94	data[0] = peer ? peer->ifindex : 0;
 
 95}
 96
 97static const struct ethtool_ops veth_ethtool_ops = {
 98	.get_settings		= veth_get_settings,
 99	.get_drvinfo		= veth_get_drvinfo,
100	.get_link		= ethtool_op_get_link,
101	.get_strings		= veth_get_strings,
102	.get_sset_count		= veth_get_sset_count,
103	.get_ethtool_stats	= veth_get_ethtool_stats,
104};
105
 
 
 
 
106static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
107{
108	struct veth_priv *priv = netdev_priv(dev);
109	struct net_device *rcv;
110	int length = skb->len;
 
 
 
 
 
111
112	rcu_read_lock();
113	rcv = rcu_dereference(priv->peer);
114	if (unlikely(!rcv)) {
115		kfree_skb(skb);
116		goto drop;
117	}
118
119	if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
120		struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
121
122		u64_stats_update_begin(&stats->syncp);
123		stats->bytes += length;
124		stats->packets++;
125		u64_stats_update_end(&stats->syncp);
126	} else {
127drop:
128		atomic64_inc(&priv->dropped);
129	}
130	rcu_read_unlock();
131	return NETDEV_TX_OK;
132}
133
134/*
135 * general routines
136 */
137
138static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
 
139{
140	struct veth_priv *priv = netdev_priv(dev);
141	int cpu;
142
143	result->packets = 0;
144	result->bytes = 0;
145	for_each_possible_cpu(cpu) {
146		struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
147		u64 packets, bytes;
 
148		unsigned int start;
149
150		do {
151			start = u64_stats_fetch_begin_irq(&stats->syncp);
152			packets = stats->packets;
153			bytes = stats->bytes;
154		} while (u64_stats_fetch_retry_irq(&stats->syncp, start));
155		result->packets += packets;
156		result->bytes += bytes;
157	}
158	return atomic64_read(&priv->dropped);
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	struct net_device *peer;
166	struct pcpu_vstats one;
167
168	tot->tx_dropped = veth_stats_one(&one, dev);
169	tot->tx_bytes = one.bytes;
170	tot->tx_packets = one.packets;
171
172	rcu_read_lock();
173	peer = rcu_dereference(priv->peer);
174	if (peer) {
175		tot->rx_dropped = veth_stats_one(&one, peer);
176		tot->rx_bytes = one.bytes;
177		tot->rx_packets = one.packets;
178	}
179	rcu_read_unlock();
180
181	return tot;
182}
183
184/* fake multicast ability */
185static void veth_set_multicast_list(struct net_device *dev)
186{
187}
188
189static int veth_open(struct net_device *dev)
190{
191	struct veth_priv *priv = netdev_priv(dev);
192	struct net_device *peer = rtnl_dereference(priv->peer);
193
194	if (!peer)
 
195		return -ENOTCONN;
196
197	if (peer->flags & IFF_UP) {
198		netif_carrier_on(dev);
199		netif_carrier_on(peer);
200	}
201	return 0;
202}
203
204static int veth_close(struct net_device *dev)
205{
206	struct veth_priv *priv = netdev_priv(dev);
207	struct net_device *peer = rtnl_dereference(priv->peer);
208
209	netif_carrier_off(dev);
210	if (peer)
211		netif_carrier_off(peer);
212
213	return 0;
214}
215
216static int is_valid_veth_mtu(int mtu)
217{
218	return mtu >= ETH_MIN_MTU && mtu <= ETH_MAX_MTU;
219}
220
221static int veth_dev_init(struct net_device *dev)
222{
223	dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
224	if (!dev->vstats)
225		return -ENOMEM;
226	return 0;
227}
228
229static void veth_dev_free(struct net_device *dev)
230{
231	free_percpu(dev->vstats);
232	free_netdev(dev);
233}
234
235#ifdef CONFIG_NET_POLL_CONTROLLER
236static void veth_poll_controller(struct net_device *dev)
237{
238	/* veth only receives frames when its peer sends one
239	 * Since it's a synchronous operation, we are guaranteed
240	 * never to have pending data when we poll for it so
241	 * there is nothing to do here.
242	 *
243	 * We need this though so netpoll recognizes us as an interface that
244	 * supports polling, which enables bridge devices in virt setups to
245	 * still use netconsole
246	 */
247}
248#endif	/* CONFIG_NET_POLL_CONTROLLER */
249
250static int veth_get_iflink(const struct net_device *dev)
251{
252	struct veth_priv *priv = netdev_priv(dev);
253	struct net_device *peer;
254	int iflink;
255
256	rcu_read_lock();
257	peer = rcu_dereference(priv->peer);
258	iflink = peer ? peer->ifindex : 0;
259	rcu_read_unlock();
260
261	return iflink;
262}
263
264static void veth_set_rx_headroom(struct net_device *dev, int new_hr)
265{
266	struct veth_priv *peer_priv, *priv = netdev_priv(dev);
267	struct net_device *peer;
268
269	if (new_hr < 0)
270		new_hr = 0;
271
272	rcu_read_lock();
273	peer = rcu_dereference(priv->peer);
274	if (unlikely(!peer))
275		goto out;
276
277	peer_priv = netdev_priv(peer);
278	priv->requested_headroom = new_hr;
279	new_hr = max(priv->requested_headroom, peer_priv->requested_headroom);
280	dev->needed_headroom = new_hr;
281	peer->needed_headroom = new_hr;
282
283out:
284	rcu_read_unlock();
285}
286
287static const struct net_device_ops veth_netdev_ops = {
288	.ndo_init            = veth_dev_init,
289	.ndo_open            = veth_open,
290	.ndo_stop            = veth_close,
291	.ndo_start_xmit      = veth_xmit,
 
292	.ndo_get_stats64     = veth_get_stats64,
293	.ndo_set_rx_mode     = veth_set_multicast_list,
294	.ndo_set_mac_address = eth_mac_addr,
295#ifdef CONFIG_NET_POLL_CONTROLLER
296	.ndo_poll_controller	= veth_poll_controller,
297#endif
298	.ndo_get_iflink		= veth_get_iflink,
299	.ndo_features_check	= passthru_features_check,
300	.ndo_set_rx_headroom	= veth_set_rx_headroom,
301};
302
303#define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HW_CSUM | \
304		       NETIF_F_RXCSUM | NETIF_F_SCTP_CRC | NETIF_F_HIGHDMA | \
305		       NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL | \
306		       NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
307		       NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
308
309static void veth_setup(struct net_device *dev)
310{
311	ether_setup(dev);
312
313	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
314	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
315	dev->priv_flags |= IFF_NO_QUEUE;
316	dev->priv_flags |= IFF_PHONY_HEADROOM;
317
318	dev->netdev_ops = &veth_netdev_ops;
319	dev->ethtool_ops = &veth_ethtool_ops;
320	dev->features |= NETIF_F_LLTX;
321	dev->features |= VETH_FEATURES;
322	dev->vlan_features = dev->features &
323			     ~(NETIF_F_HW_VLAN_CTAG_TX |
324			       NETIF_F_HW_VLAN_STAG_TX |
325			       NETIF_F_HW_VLAN_CTAG_RX |
326			       NETIF_F_HW_VLAN_STAG_RX);
327	dev->destructor = veth_dev_free;
328	dev->max_mtu = ETH_MAX_MTU;
329
330	dev->hw_features = VETH_FEATURES;
331	dev->hw_enc_features = VETH_FEATURES;
332	dev->mpls_features = NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
333}
334
335/*
336 * netlink interface
337 */
338
339static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
340{
341	if (tb[IFLA_ADDRESS]) {
342		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
343			return -EINVAL;
344		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
345			return -EADDRNOTAVAIL;
346	}
347	if (tb[IFLA_MTU]) {
348		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
349			return -EINVAL;
350	}
351	return 0;
352}
353
354static struct rtnl_link_ops veth_link_ops;
355
356static int veth_newlink(struct net *src_net, struct net_device *dev,
357			 struct nlattr *tb[], struct nlattr *data[])
358{
359	int err;
360	struct net_device *peer;
361	struct veth_priv *priv;
362	char ifname[IFNAMSIZ];
363	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
364	unsigned char name_assign_type;
365	struct ifinfomsg *ifmp;
366	struct net *net;
367
368	/*
369	 * create and register peer first
370	 */
371	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
372		struct nlattr *nla_peer;
373
374		nla_peer = data[VETH_INFO_PEER];
375		ifmp = nla_data(nla_peer);
376		err = rtnl_nla_parse_ifla(peer_tb,
377					  nla_data(nla_peer) + sizeof(struct ifinfomsg),
378					  nla_len(nla_peer) - sizeof(struct ifinfomsg));
 
379		if (err < 0)
380			return err;
381
382		err = veth_validate(peer_tb, NULL);
383		if (err < 0)
384			return err;
385
386		tbp = peer_tb;
387	} else {
388		ifmp = NULL;
389		tbp = tb;
390	}
391
392	if (tbp[IFLA_IFNAME]) {
393		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
394		name_assign_type = NET_NAME_USER;
395	} else {
396		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
397		name_assign_type = NET_NAME_ENUM;
398	}
399
400	net = rtnl_link_get_net(src_net, tbp);
401	if (IS_ERR(net))
402		return PTR_ERR(net);
403
404	peer = rtnl_create_link(net, ifname, name_assign_type,
405				&veth_link_ops, tbp);
406	if (IS_ERR(peer)) {
407		put_net(net);
408		return PTR_ERR(peer);
409	}
410
411	if (tbp[IFLA_ADDRESS] == NULL)
412		eth_hw_addr_random(peer);
413
414	if (ifmp && (dev->ifindex != 0))
415		peer->ifindex = ifmp->ifi_index;
416
417	err = register_netdevice(peer);
418	put_net(net);
419	net = NULL;
420	if (err < 0)
421		goto err_register_peer;
422
423	netif_carrier_off(peer);
424
425	err = rtnl_configure_link(peer, ifmp);
426	if (err < 0)
427		goto err_configure_peer;
428
429	/*
430	 * register dev last
431	 *
432	 * note, that since we've registered new device the dev's name
433	 * should be re-allocated
434	 */
435
436	if (tb[IFLA_ADDRESS] == NULL)
437		eth_hw_addr_random(dev);
438
439	if (tb[IFLA_IFNAME])
440		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
441	else
442		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
443
 
 
 
 
 
 
444	err = register_netdevice(dev);
445	if (err < 0)
446		goto err_register_dev;
447
448	netif_carrier_off(dev);
449
450	/*
451	 * tie the deviced together
452	 */
453
454	priv = netdev_priv(dev);
455	rcu_assign_pointer(priv->peer, peer);
456
457	priv = netdev_priv(peer);
458	rcu_assign_pointer(priv->peer, dev);
459	return 0;
460
461err_register_dev:
462	/* nothing to do */
 
463err_configure_peer:
464	unregister_netdevice(peer);
465	return err;
466
467err_register_peer:
468	free_netdev(peer);
469	return err;
470}
471
472static void veth_dellink(struct net_device *dev, struct list_head *head)
473{
474	struct veth_priv *priv;
475	struct net_device *peer;
476
477	priv = netdev_priv(dev);
478	peer = rtnl_dereference(priv->peer);
479
480	/* Note : dellink() is called from default_device_exit_batch(),
481	 * before a rcu_synchronize() point. The devices are guaranteed
482	 * not being freed before one RCU grace period.
483	 */
484	RCU_INIT_POINTER(priv->peer, NULL);
485	unregister_netdevice_queue(dev, head);
486
487	if (peer) {
488		priv = netdev_priv(peer);
489		RCU_INIT_POINTER(priv->peer, NULL);
490		unregister_netdevice_queue(peer, head);
491	}
492}
493
494static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
495	[VETH_INFO_PEER]	= { .len = sizeof(struct ifinfomsg) },
496};
497
498static struct net *veth_get_link_net(const struct net_device *dev)
499{
500	struct veth_priv *priv = netdev_priv(dev);
501	struct net_device *peer = rtnl_dereference(priv->peer);
502
503	return peer ? dev_net(peer) : dev_net(dev);
504}
505
506static struct rtnl_link_ops veth_link_ops = {
507	.kind		= DRV_NAME,
508	.priv_size	= sizeof(struct veth_priv),
509	.setup		= veth_setup,
510	.validate	= veth_validate,
511	.newlink	= veth_newlink,
512	.dellink	= veth_dellink,
513	.policy		= veth_policy,
514	.maxtype	= VETH_INFO_MAX,
515	.get_link_net	= veth_get_link_net,
516};
517
518/*
519 * init/fini
520 */
521
522static __init int veth_init(void)
523{
524	return rtnl_link_register(&veth_link_ops);
525}
526
527static __exit void veth_exit(void)
528{
529	rtnl_link_unregister(&veth_link_ops);
530}
531
532module_init(veth_init);
533module_exit(veth_exit);
534
535MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
536MODULE_LICENSE("GPL v2");
537MODULE_ALIAS_RTNL_LINK(DRV_NAME);