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
 26#define MIN_MTU 68		/* Min L3 MTU */
 27#define MAX_MTU 65535		/* Max L3 MTU (arbitrary) */
 28
 29struct pcpu_vstats {
 30	u64			packets;
 31	u64			bytes;
 32	struct u64_stats_sync	syncp;
 33};
 34
 35struct veth_priv {
 36	struct net_device __rcu	*peer;
 37	atomic64_t		dropped;
 
 38};
 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	ethtool_cmd_speed_set(cmd, 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	strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
 68	strlcpy(info->version, DRV_VERSION, sizeof(info->version));
 69}
 70
 71static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
 72{
 73	switch(stringset) {
 74	case ETH_SS_STATS:
 75		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
 76		break;
 77	}
 78}
 79
 80static int veth_get_sset_count(struct net_device *dev, int sset)
 81{
 82	switch (sset) {
 83	case ETH_SS_STATS:
 84		return ARRAY_SIZE(ethtool_stats_keys);
 85	default:
 86		return -EOPNOTSUPP;
 87	}
 88}
 89
 90static void veth_get_ethtool_stats(struct net_device *dev,
 91		struct ethtool_stats *stats, u64 *data)
 92{
 93	struct veth_priv *priv = netdev_priv(dev);
 94	struct net_device *peer = rtnl_dereference(priv->peer);
 95
 96	data[0] = peer ? peer->ifindex : 0;
 97}
 98
 99static const struct ethtool_ops veth_ethtool_ops = {
100	.get_settings		= veth_get_settings,
101	.get_drvinfo		= veth_get_drvinfo,
102	.get_link		= ethtool_op_get_link,
103	.get_strings		= veth_get_strings,
104	.get_sset_count		= veth_get_sset_count,
105	.get_ethtool_stats	= veth_get_ethtool_stats,
106};
107
108static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
109{
110	struct veth_priv *priv = netdev_priv(dev);
111	struct net_device *rcv;
112	int length = skb->len;
113
114	rcu_read_lock();
115	rcv = rcu_dereference(priv->peer);
116	if (unlikely(!rcv)) {
117		kfree_skb(skb);
118		goto drop;
119	}
120	/* don't change ip_summed == CHECKSUM_PARTIAL, as that
121	 * will cause bad checksum on forwarded packets
122	 */
123	if (skb->ip_summed == CHECKSUM_NONE &&
124	    rcv->features & NETIF_F_RXCSUM)
125		skb->ip_summed = CHECKSUM_UNNECESSARY;
126
127	if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
128		struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
129
130		u64_stats_update_begin(&stats->syncp);
131		stats->bytes += length;
132		stats->packets++;
133		u64_stats_update_end(&stats->syncp);
134	} else {
135drop:
136		atomic64_inc(&priv->dropped);
137	}
138	rcu_read_unlock();
139	return NETDEV_TX_OK;
140}
141
142/*
143 * general routines
144 */
145
146static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
147{
148	struct veth_priv *priv = netdev_priv(dev);
149	int cpu;
150
151	result->packets = 0;
152	result->bytes = 0;
153	for_each_possible_cpu(cpu) {
154		struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
155		u64 packets, bytes;
156		unsigned int start;
157
158		do {
159			start = u64_stats_fetch_begin_irq(&stats->syncp);
160			packets = stats->packets;
161			bytes = stats->bytes;
162		} while (u64_stats_fetch_retry_irq(&stats->syncp, start));
163		result->packets += packets;
164		result->bytes += bytes;
165	}
166	return atomic64_read(&priv->dropped);
167}
168
169static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
170						  struct rtnl_link_stats64 *tot)
171{
172	struct veth_priv *priv = netdev_priv(dev);
173	struct net_device *peer;
174	struct pcpu_vstats one;
175
176	tot->tx_dropped = veth_stats_one(&one, dev);
177	tot->tx_bytes = one.bytes;
178	tot->tx_packets = one.packets;
179
180	rcu_read_lock();
181	peer = rcu_dereference(priv->peer);
182	if (peer) {
183		tot->rx_dropped = veth_stats_one(&one, peer);
184		tot->rx_bytes = one.bytes;
185		tot->rx_packets = one.packets;
186	}
187	rcu_read_unlock();
188
189	return tot;
190}
191
192/* fake multicast ability */
193static void veth_set_multicast_list(struct net_device *dev)
194{
195}
196
197static int veth_open(struct net_device *dev)
198{
199	struct veth_priv *priv = netdev_priv(dev);
200	struct net_device *peer = rtnl_dereference(priv->peer);
201
202	if (!peer)
203		return -ENOTCONN;
204
205	if (peer->flags & IFF_UP) {
206		netif_carrier_on(dev);
207		netif_carrier_on(peer);
208	}
209	return 0;
210}
211
212static int veth_close(struct net_device *dev)
213{
214	struct veth_priv *priv = netdev_priv(dev);
215	struct net_device *peer = rtnl_dereference(priv->peer);
216
217	netif_carrier_off(dev);
218	if (peer)
219		netif_carrier_off(peer);
220
221	return 0;
222}
223
224static int is_valid_veth_mtu(int new_mtu)
225{
226	return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
227}
228
229static int veth_change_mtu(struct net_device *dev, int new_mtu)
230{
231	if (!is_valid_veth_mtu(new_mtu))
232		return -EINVAL;
233	dev->mtu = new_mtu;
234	return 0;
235}
236
237static int veth_dev_init(struct net_device *dev)
238{
239	dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
240	if (!dev->vstats)
241		return -ENOMEM;
242	return 0;
243}
244
245static void veth_dev_free(struct net_device *dev)
246{
247	free_percpu(dev->vstats);
248	free_netdev(dev);
249}
250
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
251static const struct net_device_ops veth_netdev_ops = {
252	.ndo_init            = veth_dev_init,
253	.ndo_open            = veth_open,
254	.ndo_stop            = veth_close,
255	.ndo_start_xmit      = veth_xmit,
256	.ndo_change_mtu      = veth_change_mtu,
257	.ndo_get_stats64     = veth_get_stats64,
258	.ndo_set_rx_mode     = veth_set_multicast_list,
259	.ndo_set_mac_address = eth_mac_addr,
 
 
 
 
 
 
260};
261
262#define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO |    \
263		       NETIF_F_HW_CSUM | NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
264		       NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL |	    \
265		       NETIF_F_GSO_IPIP | NETIF_F_GSO_SIT | NETIF_F_UFO	|   \
266		       NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
267		       NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
268
269static void veth_setup(struct net_device *dev)
270{
271	ether_setup(dev);
272
273	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
274	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
 
 
275
276	dev->netdev_ops = &veth_netdev_ops;
277	dev->ethtool_ops = &veth_ethtool_ops;
278	dev->features |= NETIF_F_LLTX;
279	dev->features |= VETH_FEATURES;
280	dev->vlan_features = dev->features &
281			     ~(NETIF_F_HW_VLAN_CTAG_TX |
282			       NETIF_F_HW_VLAN_STAG_TX |
283			       NETIF_F_HW_VLAN_CTAG_RX |
284			       NETIF_F_HW_VLAN_STAG_RX);
285	dev->destructor = veth_dev_free;
 
286
287	dev->hw_features = VETH_FEATURES;
288	dev->hw_enc_features = VETH_FEATURES;
 
289}
290
291/*
292 * netlink interface
293 */
294
295static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
296{
297	if (tb[IFLA_ADDRESS]) {
298		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
299			return -EINVAL;
300		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
301			return -EADDRNOTAVAIL;
302	}
303	if (tb[IFLA_MTU]) {
304		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
305			return -EINVAL;
306	}
307	return 0;
308}
309
310static struct rtnl_link_ops veth_link_ops;
311
312static int veth_newlink(struct net *src_net, struct net_device *dev,
313			 struct nlattr *tb[], struct nlattr *data[])
314{
315	int err;
316	struct net_device *peer;
317	struct veth_priv *priv;
318	char ifname[IFNAMSIZ];
319	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 
320	struct ifinfomsg *ifmp;
321	struct net *net;
322
323	/*
324	 * create and register peer first
325	 */
326	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
327		struct nlattr *nla_peer;
328
329		nla_peer = data[VETH_INFO_PEER];
330		ifmp = nla_data(nla_peer);
331		err = rtnl_nla_parse_ifla(peer_tb,
332					  nla_data(nla_peer) + sizeof(struct ifinfomsg),
333					  nla_len(nla_peer) - sizeof(struct ifinfomsg));
334		if (err < 0)
335			return err;
336
337		err = veth_validate(peer_tb, NULL);
338		if (err < 0)
339			return err;
340
341		tbp = peer_tb;
342	} else {
343		ifmp = NULL;
344		tbp = tb;
345	}
346
347	if (tbp[IFLA_IFNAME])
348		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
349	else
 
350		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 
 
351
352	net = rtnl_link_get_net(src_net, tbp);
353	if (IS_ERR(net))
354		return PTR_ERR(net);
355
356	peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
 
357	if (IS_ERR(peer)) {
358		put_net(net);
359		return PTR_ERR(peer);
360	}
361
362	if (tbp[IFLA_ADDRESS] == NULL)
363		eth_hw_addr_random(peer);
364
365	if (ifmp && (dev->ifindex != 0))
366		peer->ifindex = ifmp->ifi_index;
367
368	err = register_netdevice(peer);
369	put_net(net);
370	net = NULL;
371	if (err < 0)
372		goto err_register_peer;
373
374	netif_carrier_off(peer);
375
376	err = rtnl_configure_link(peer, ifmp);
377	if (err < 0)
378		goto err_configure_peer;
379
380	/*
381	 * register dev last
382	 *
383	 * note, that since we've registered new device the dev's name
384	 * should be re-allocated
385	 */
386
387	if (tb[IFLA_ADDRESS] == NULL)
388		eth_hw_addr_random(dev);
389
390	if (tb[IFLA_IFNAME])
391		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
392	else
393		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
394
395	err = register_netdevice(dev);
396	if (err < 0)
397		goto err_register_dev;
398
399	netif_carrier_off(dev);
400
401	/*
402	 * tie the deviced together
403	 */
404
405	priv = netdev_priv(dev);
406	rcu_assign_pointer(priv->peer, peer);
407
408	priv = netdev_priv(peer);
409	rcu_assign_pointer(priv->peer, dev);
410	return 0;
411
412err_register_dev:
413	/* nothing to do */
414err_configure_peer:
415	unregister_netdevice(peer);
416	return err;
417
418err_register_peer:
419	free_netdev(peer);
420	return err;
421}
422
423static void veth_dellink(struct net_device *dev, struct list_head *head)
424{
425	struct veth_priv *priv;
426	struct net_device *peer;
427
428	priv = netdev_priv(dev);
429	peer = rtnl_dereference(priv->peer);
430
431	/* Note : dellink() is called from default_device_exit_batch(),
432	 * before a rcu_synchronize() point. The devices are guaranteed
433	 * not being freed before one RCU grace period.
434	 */
435	RCU_INIT_POINTER(priv->peer, NULL);
436	unregister_netdevice_queue(dev, head);
437
438	if (peer) {
439		priv = netdev_priv(peer);
440		RCU_INIT_POINTER(priv->peer, NULL);
441		unregister_netdevice_queue(peer, head);
442	}
443}
444
445static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
446	[VETH_INFO_PEER]	= { .len = sizeof(struct ifinfomsg) },
447};
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	return rtnl_link_register(&veth_link_ops);
467}
468
469static __exit void veth_exit(void)
470{
471	rtnl_link_unregister(&veth_link_ops);
472}
473
474module_init(veth_init);
475module_exit(veth_exit);
476
477MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
478MODULE_LICENSE("GPL v2");
479MODULE_ALIAS_RTNL_LINK(DRV_NAME);

  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);