4 * Copyright (C) 2007 OpenVZ http://openvz.org, SWsoft Inc
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>
17 #include <net/rtnetlink.h>
20 #include <linux/veth.h>
21 #include <linux/module.h>
23 #define DRV_NAME "veth"
24 #define DRV_VERSION "1.0"
26 #define MIN_MTU 68 /* Min L3 MTU */
27 #define MAX_MTU 65535 /* Max L3 MTU (arbitrary) */
32 struct u64_stats_sync syncp;
36 struct net_device __rcu *peer;
45 const char string[ETH_GSTRING_LEN];
46 } ethtool_stats_keys[] = {
50 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
54 ethtool_cmd_speed_set(cmd, SPEED_10000);
55 cmd->duplex = DUPLEX_FULL;
58 cmd->transceiver = XCVR_INTERNAL;
59 cmd->autoneg = AUTONEG_DISABLE;
65 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
67 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
68 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
71 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
75 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
80 static int veth_get_sset_count(struct net_device *dev, int sset)
84 return ARRAY_SIZE(ethtool_stats_keys);
90 static void veth_get_ethtool_stats(struct net_device *dev,
91 struct ethtool_stats *stats, u64 *data)
93 struct veth_priv *priv = netdev_priv(dev);
94 struct net_device *peer = rtnl_dereference(priv->peer);
96 data[0] = peer ? peer->ifindex : 0;
99 static 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,
108 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
110 struct veth_priv *priv = netdev_priv(dev);
111 struct net_device *rcv;
112 int length = skb->len;
115 rcv = rcu_dereference(priv->peer);
116 if (unlikely(!rcv)) {
120 /* don't change ip_summed == CHECKSUM_PARTIAL, as that
121 * will cause bad checksum on forwarded packets
123 if (skb->ip_summed == CHECKSUM_NONE &&
124 rcv->features & NETIF_F_RXCSUM)
125 skb->ip_summed = CHECKSUM_UNNECESSARY;
127 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
128 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
130 u64_stats_update_begin(&stats->syncp);
131 stats->bytes += length;
133 u64_stats_update_end(&stats->syncp);
136 atomic64_inc(&priv->dropped);
146 static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
148 struct veth_priv *priv = netdev_priv(dev);
153 for_each_possible_cpu(cpu) {
154 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
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;
166 return atomic64_read(&priv->dropped);
169 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
170 struct rtnl_link_stats64 *tot)
172 struct veth_priv *priv = netdev_priv(dev);
173 struct net_device *peer;
174 struct pcpu_vstats one;
176 tot->tx_dropped = veth_stats_one(&one, dev);
177 tot->tx_bytes = one.bytes;
178 tot->tx_packets = one.packets;
181 peer = rcu_dereference(priv->peer);
183 tot->rx_dropped = veth_stats_one(&one, peer);
184 tot->rx_bytes = one.bytes;
185 tot->rx_packets = one.packets;
192 /* fake multicast ability */
193 static void veth_set_multicast_list(struct net_device *dev)
197 static int veth_open(struct net_device *dev)
199 struct veth_priv *priv = netdev_priv(dev);
200 struct net_device *peer = rtnl_dereference(priv->peer);
205 if (peer->flags & IFF_UP) {
206 netif_carrier_on(dev);
207 netif_carrier_on(peer);
212 static int veth_close(struct net_device *dev)
214 struct veth_priv *priv = netdev_priv(dev);
215 struct net_device *peer = rtnl_dereference(priv->peer);
217 netif_carrier_off(dev);
219 netif_carrier_off(peer);
224 static int is_valid_veth_mtu(int new_mtu)
226 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
229 static int veth_change_mtu(struct net_device *dev, int new_mtu)
231 if (!is_valid_veth_mtu(new_mtu))
237 static int veth_dev_init(struct net_device *dev)
239 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
245 static void veth_dev_free(struct net_device *dev)
247 free_percpu(dev->vstats);
251 static 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,
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 )
269 static void veth_setup(struct net_device *dev)
273 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
274 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
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;
287 dev->hw_features = VETH_FEATURES;
288 dev->hw_enc_features = VETH_FEATURES;
295 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
297 if (tb[IFLA_ADDRESS]) {
298 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
300 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
301 return -EADDRNOTAVAIL;
304 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
310 static struct rtnl_link_ops veth_link_ops;
312 static int veth_newlink(struct net *src_net, struct net_device *dev,
313 struct nlattr *tb[], struct nlattr *data[])
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;
324 * create and register peer first
326 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
327 struct nlattr *nla_peer;
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));
337 err = veth_validate(peer_tb, NULL);
347 if (tbp[IFLA_IFNAME])
348 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
350 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
352 net = rtnl_link_get_net(src_net, tbp);
356 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
359 return PTR_ERR(peer);
362 if (tbp[IFLA_ADDRESS] == NULL)
363 eth_hw_addr_random(peer);
365 if (ifmp && (dev->ifindex != 0))
366 peer->ifindex = ifmp->ifi_index;
368 err = register_netdevice(peer);
372 goto err_register_peer;
374 netif_carrier_off(peer);
376 err = rtnl_configure_link(peer, ifmp);
378 goto err_configure_peer;
383 * note, that since we've registered new device the dev's name
384 * should be re-allocated
387 if (tb[IFLA_ADDRESS] == NULL)
388 eth_hw_addr_random(dev);
391 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
393 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
395 err = register_netdevice(dev);
397 goto err_register_dev;
399 netif_carrier_off(dev);
402 * tie the deviced together
405 priv = netdev_priv(dev);
406 rcu_assign_pointer(priv->peer, peer);
408 priv = netdev_priv(peer);
409 rcu_assign_pointer(priv->peer, dev);
415 unregister_netdevice(peer);
423 static void veth_dellink(struct net_device *dev, struct list_head *head)
425 struct veth_priv *priv;
426 struct net_device *peer;
428 priv = netdev_priv(dev);
429 peer = rtnl_dereference(priv->peer);
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.
435 RCU_INIT_POINTER(priv->peer, NULL);
436 unregister_netdevice_queue(dev, head);
439 priv = netdev_priv(peer);
440 RCU_INIT_POINTER(priv->peer, NULL);
441 unregister_netdevice_queue(peer, head);
445 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
446 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
449 static struct rtnl_link_ops veth_link_ops = {
451 .priv_size = sizeof(struct veth_priv),
453 .validate = veth_validate,
454 .newlink = veth_newlink,
455 .dellink = veth_dellink,
456 .policy = veth_policy,
457 .maxtype = VETH_INFO_MAX,
464 static __init int veth_init(void)
466 return rtnl_link_register(&veth_link_ops);
469 static __exit void veth_exit(void)
471 rtnl_link_unregister(&veth_link_ops);
474 module_init(veth_init);
475 module_exit(veth_exit);
477 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
478 MODULE_LICENSE("GPL v2");
479 MODULE_ALIAS_RTNL_LINK(DRV_NAME);