4 * Copyright (C) 2007 OpenVZ http://openvz.org, SWsoft Inc
6 * Author: Pavel Emelianov <xemul@openvz.org>
7 * Ethtool interface from: Eric W. Biederman <ebiederm@xmission.com>
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)) {
121 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
122 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
124 u64_stats_update_begin(&stats->syncp);
125 stats->bytes += length;
127 u64_stats_update_end(&stats->syncp);
130 atomic64_inc(&priv->dropped);
140 static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
142 struct veth_priv *priv = netdev_priv(dev);
147 for_each_possible_cpu(cpu) {
148 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
153 start = u64_stats_fetch_begin_irq(&stats->syncp);
154 packets = stats->packets;
155 bytes = stats->bytes;
156 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
157 result->packets += packets;
158 result->bytes += bytes;
160 return atomic64_read(&priv->dropped);
163 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
164 struct rtnl_link_stats64 *tot)
166 struct veth_priv *priv = netdev_priv(dev);
167 struct net_device *peer;
168 struct pcpu_vstats one;
170 tot->tx_dropped = veth_stats_one(&one, dev);
171 tot->tx_bytes = one.bytes;
172 tot->tx_packets = one.packets;
175 peer = rcu_dereference(priv->peer);
177 tot->rx_dropped = veth_stats_one(&one, peer);
178 tot->rx_bytes = one.bytes;
179 tot->rx_packets = one.packets;
186 /* fake multicast ability */
187 static void veth_set_multicast_list(struct net_device *dev)
191 static int veth_open(struct net_device *dev)
193 struct veth_priv *priv = netdev_priv(dev);
194 struct net_device *peer = rtnl_dereference(priv->peer);
199 if (peer->flags & IFF_UP) {
200 netif_carrier_on(dev);
201 netif_carrier_on(peer);
206 static int veth_close(struct net_device *dev)
208 struct veth_priv *priv = netdev_priv(dev);
209 struct net_device *peer = rtnl_dereference(priv->peer);
211 netif_carrier_off(dev);
213 netif_carrier_off(peer);
218 static int is_valid_veth_mtu(int new_mtu)
220 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
223 static int veth_change_mtu(struct net_device *dev, int new_mtu)
225 if (!is_valid_veth_mtu(new_mtu))
231 static int veth_dev_init(struct net_device *dev)
233 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
239 static void veth_dev_free(struct net_device *dev)
241 free_percpu(dev->vstats);
245 #ifdef CONFIG_NET_POLL_CONTROLLER
246 static void veth_poll_controller(struct net_device *dev)
248 /* veth only receives frames when its peer sends one
249 * Since it's a synchronous operation, we are guaranteed
250 * never to have pending data when we poll for it so
251 * there is nothing to do here.
253 * We need this though so netpoll recognizes us as an interface that
254 * supports polling, which enables bridge devices in virt setups to
255 * still use netconsole
258 #endif /* CONFIG_NET_POLL_CONTROLLER */
260 static int veth_get_iflink(const struct net_device *dev)
262 struct veth_priv *priv = netdev_priv(dev);
263 struct net_device *peer;
267 peer = rcu_dereference(priv->peer);
268 iflink = peer ? peer->ifindex : 0;
274 static const struct net_device_ops veth_netdev_ops = {
275 .ndo_init = veth_dev_init,
276 .ndo_open = veth_open,
277 .ndo_stop = veth_close,
278 .ndo_start_xmit = veth_xmit,
279 .ndo_change_mtu = veth_change_mtu,
280 .ndo_get_stats64 = veth_get_stats64,
281 .ndo_set_rx_mode = veth_set_multicast_list,
282 .ndo_set_mac_address = eth_mac_addr,
283 #ifdef CONFIG_NET_POLL_CONTROLLER
284 .ndo_poll_controller = veth_poll_controller,
286 .ndo_get_iflink = veth_get_iflink,
287 .ndo_features_check = passthru_features_check,
290 #define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
291 NETIF_F_HW_CSUM | NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
292 NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL | \
293 NETIF_F_GSO_IPIP | NETIF_F_GSO_SIT | NETIF_F_UFO | \
294 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
295 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
297 static void veth_setup(struct net_device *dev)
301 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
302 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
303 dev->priv_flags |= IFF_NO_QUEUE;
305 dev->netdev_ops = &veth_netdev_ops;
306 dev->ethtool_ops = &veth_ethtool_ops;
307 dev->features |= NETIF_F_LLTX;
308 dev->features |= VETH_FEATURES;
309 dev->vlan_features = dev->features &
310 ~(NETIF_F_HW_VLAN_CTAG_TX |
311 NETIF_F_HW_VLAN_STAG_TX |
312 NETIF_F_HW_VLAN_CTAG_RX |
313 NETIF_F_HW_VLAN_STAG_RX);
314 dev->destructor = veth_dev_free;
316 dev->hw_features = VETH_FEATURES;
317 dev->hw_enc_features = VETH_FEATURES;
324 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
326 if (tb[IFLA_ADDRESS]) {
327 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
329 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
330 return -EADDRNOTAVAIL;
333 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
339 static struct rtnl_link_ops veth_link_ops;
341 static int veth_newlink(struct net *src_net, struct net_device *dev,
342 struct nlattr *tb[], struct nlattr *data[])
345 struct net_device *peer;
346 struct veth_priv *priv;
347 char ifname[IFNAMSIZ];
348 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
349 unsigned char name_assign_type;
350 struct ifinfomsg *ifmp;
354 * create and register peer first
356 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
357 struct nlattr *nla_peer;
359 nla_peer = data[VETH_INFO_PEER];
360 ifmp = nla_data(nla_peer);
361 err = rtnl_nla_parse_ifla(peer_tb,
362 nla_data(nla_peer) + sizeof(struct ifinfomsg),
363 nla_len(nla_peer) - sizeof(struct ifinfomsg));
367 err = veth_validate(peer_tb, NULL);
377 if (tbp[IFLA_IFNAME]) {
378 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
379 name_assign_type = NET_NAME_USER;
381 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
382 name_assign_type = NET_NAME_ENUM;
385 net = rtnl_link_get_net(src_net, tbp);
389 peer = rtnl_create_link(net, ifname, name_assign_type,
390 &veth_link_ops, tbp);
393 return PTR_ERR(peer);
396 if (tbp[IFLA_ADDRESS] == NULL)
397 eth_hw_addr_random(peer);
399 if (ifmp && (dev->ifindex != 0))
400 peer->ifindex = ifmp->ifi_index;
402 peer->gso_max_size = dev->gso_max_size;
403 peer->gso_max_segs = dev->gso_max_segs;
405 err = register_netdevice(peer);
409 goto err_register_peer;
411 netif_carrier_off(peer);
413 err = rtnl_configure_link(peer, ifmp);
415 goto err_configure_peer;
420 * note, that since we've registered new device the dev's name
421 * should be re-allocated
424 if (tb[IFLA_ADDRESS] == NULL)
425 eth_hw_addr_random(dev);
428 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
430 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
432 err = register_netdevice(dev);
434 goto err_register_dev;
436 netif_carrier_off(dev);
439 * tie the deviced together
442 priv = netdev_priv(dev);
443 rcu_assign_pointer(priv->peer, peer);
445 priv = netdev_priv(peer);
446 rcu_assign_pointer(priv->peer, dev);
452 unregister_netdevice(peer);
460 static void veth_dellink(struct net_device *dev, struct list_head *head)
462 struct veth_priv *priv;
463 struct net_device *peer;
465 priv = netdev_priv(dev);
466 peer = rtnl_dereference(priv->peer);
468 /* Note : dellink() is called from default_device_exit_batch(),
469 * before a rcu_synchronize() point. The devices are guaranteed
470 * not being freed before one RCU grace period.
472 RCU_INIT_POINTER(priv->peer, NULL);
473 unregister_netdevice_queue(dev, head);
476 priv = netdev_priv(peer);
477 RCU_INIT_POINTER(priv->peer, NULL);
478 unregister_netdevice_queue(peer, head);
482 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
483 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
486 static struct net *veth_get_link_net(const struct net_device *dev)
488 struct veth_priv *priv = netdev_priv(dev);
489 struct net_device *peer = rtnl_dereference(priv->peer);
491 return peer ? dev_net(peer) : dev_net(dev);
494 static struct rtnl_link_ops veth_link_ops = {
496 .priv_size = sizeof(struct veth_priv),
498 .validate = veth_validate,
499 .newlink = veth_newlink,
500 .dellink = veth_dellink,
501 .policy = veth_policy,
502 .maxtype = VETH_INFO_MAX,
503 .get_link_net = veth_get_link_net,
510 static __init int veth_init(void)
512 return rtnl_link_register(&veth_link_ops);
515 static __exit void veth_exit(void)
517 rtnl_link_unregister(&veth_link_ops);
520 module_init(veth_init);
521 module_exit(veth_exit);
523 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
524 MODULE_LICENSE("GPL v2");
525 MODULE_ALIAS_RTNL_LINK(DRV_NAME);