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"
29 struct u64_stats_sync syncp;
33 struct net_device __rcu *peer;
35 unsigned requested_headroom;
43 const char string[ETH_GSTRING_LEN];
44 } ethtool_stats_keys[] = {
48 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
52 ethtool_cmd_speed_set(cmd, SPEED_10000);
53 cmd->duplex = DUPLEX_FULL;
56 cmd->transceiver = XCVR_INTERNAL;
57 cmd->autoneg = AUTONEG_DISABLE;
63 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
65 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
66 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
69 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
73 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
78 static int veth_get_sset_count(struct net_device *dev, int sset)
82 return ARRAY_SIZE(ethtool_stats_keys);
88 static void veth_get_ethtool_stats(struct net_device *dev,
89 struct ethtool_stats *stats, u64 *data)
91 struct veth_priv *priv = netdev_priv(dev);
92 struct net_device *peer = rtnl_dereference(priv->peer);
94 data[0] = peer ? peer->ifindex : 0;
97 static 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,
106 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
108 struct veth_priv *priv = netdev_priv(dev);
109 struct net_device *rcv;
110 int length = skb->len;
113 rcv = rcu_dereference(priv->peer);
114 if (unlikely(!rcv)) {
119 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
120 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
122 u64_stats_update_begin(&stats->syncp);
123 stats->bytes += length;
125 u64_stats_update_end(&stats->syncp);
128 atomic64_inc(&priv->dropped);
138 static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
140 struct veth_priv *priv = netdev_priv(dev);
145 for_each_possible_cpu(cpu) {
146 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
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;
158 return atomic64_read(&priv->dropped);
161 static void veth_get_stats64(struct net_device *dev,
162 struct rtnl_link_stats64 *tot)
164 struct veth_priv *priv = netdev_priv(dev);
165 struct net_device *peer;
166 struct pcpu_vstats one;
168 tot->tx_dropped = veth_stats_one(&one, dev);
169 tot->tx_bytes = one.bytes;
170 tot->tx_packets = one.packets;
173 peer = rcu_dereference(priv->peer);
175 tot->rx_dropped = veth_stats_one(&one, peer);
176 tot->rx_bytes = one.bytes;
177 tot->rx_packets = one.packets;
182 /* fake multicast ability */
183 static void veth_set_multicast_list(struct net_device *dev)
187 static int veth_open(struct net_device *dev)
189 struct veth_priv *priv = netdev_priv(dev);
190 struct net_device *peer = rtnl_dereference(priv->peer);
195 if (peer->flags & IFF_UP) {
196 netif_carrier_on(dev);
197 netif_carrier_on(peer);
202 static int veth_close(struct net_device *dev)
204 struct veth_priv *priv = netdev_priv(dev);
205 struct net_device *peer = rtnl_dereference(priv->peer);
207 netif_carrier_off(dev);
209 netif_carrier_off(peer);
214 static int is_valid_veth_mtu(int mtu)
216 return mtu >= ETH_MIN_MTU && mtu <= ETH_MAX_MTU;
219 static int veth_dev_init(struct net_device *dev)
221 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
227 static void veth_dev_free(struct net_device *dev)
229 free_percpu(dev->vstats);
233 #ifdef CONFIG_NET_POLL_CONTROLLER
234 static void veth_poll_controller(struct net_device *dev)
236 /* veth only receives frames when its peer sends one
237 * Since it's a synchronous operation, we are guaranteed
238 * never to have pending data when we poll for it so
239 * there is nothing to do here.
241 * We need this though so netpoll recognizes us as an interface that
242 * supports polling, which enables bridge devices in virt setups to
243 * still use netconsole
246 #endif /* CONFIG_NET_POLL_CONTROLLER */
248 static int veth_get_iflink(const struct net_device *dev)
250 struct veth_priv *priv = netdev_priv(dev);
251 struct net_device *peer;
255 peer = rcu_dereference(priv->peer);
256 iflink = peer ? peer->ifindex : 0;
262 static void veth_set_rx_headroom(struct net_device *dev, int new_hr)
264 struct veth_priv *peer_priv, *priv = netdev_priv(dev);
265 struct net_device *peer;
271 peer = rcu_dereference(priv->peer);
275 peer_priv = netdev_priv(peer);
276 priv->requested_headroom = new_hr;
277 new_hr = max(priv->requested_headroom, peer_priv->requested_headroom);
278 dev->needed_headroom = new_hr;
279 peer->needed_headroom = new_hr;
285 static const struct net_device_ops veth_netdev_ops = {
286 .ndo_init = veth_dev_init,
287 .ndo_open = veth_open,
288 .ndo_stop = veth_close,
289 .ndo_start_xmit = veth_xmit,
290 .ndo_get_stats64 = veth_get_stats64,
291 .ndo_set_rx_mode = veth_set_multicast_list,
292 .ndo_set_mac_address = eth_mac_addr,
293 #ifdef CONFIG_NET_POLL_CONTROLLER
294 .ndo_poll_controller = veth_poll_controller,
296 .ndo_get_iflink = veth_get_iflink,
297 .ndo_features_check = passthru_features_check,
298 .ndo_set_rx_headroom = veth_set_rx_headroom,
301 #define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HW_CSUM | \
302 NETIF_F_RXCSUM | NETIF_F_SCTP_CRC | NETIF_F_HIGHDMA | \
303 NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL | \
304 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
305 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
307 static void veth_setup(struct net_device *dev)
311 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
312 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
313 dev->priv_flags |= IFF_NO_QUEUE;
314 dev->priv_flags |= IFF_PHONY_HEADROOM;
316 dev->netdev_ops = &veth_netdev_ops;
317 dev->ethtool_ops = &veth_ethtool_ops;
318 dev->features |= NETIF_F_LLTX;
319 dev->features |= VETH_FEATURES;
320 dev->vlan_features = dev->features &
321 ~(NETIF_F_HW_VLAN_CTAG_TX |
322 NETIF_F_HW_VLAN_STAG_TX |
323 NETIF_F_HW_VLAN_CTAG_RX |
324 NETIF_F_HW_VLAN_STAG_RX);
325 dev->destructor = veth_dev_free;
326 dev->max_mtu = ETH_MAX_MTU;
328 dev->hw_features = VETH_FEATURES;
329 dev->hw_enc_features = VETH_FEATURES;
330 dev->mpls_features = NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
337 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
339 if (tb[IFLA_ADDRESS]) {
340 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
342 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
343 return -EADDRNOTAVAIL;
346 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
352 static struct rtnl_link_ops veth_link_ops;
354 static int veth_newlink(struct net *src_net, struct net_device *dev,
355 struct nlattr *tb[], struct nlattr *data[])
358 struct net_device *peer;
359 struct veth_priv *priv;
360 char ifname[IFNAMSIZ];
361 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
362 unsigned char name_assign_type;
363 struct ifinfomsg *ifmp;
367 * create and register peer first
369 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
370 struct nlattr *nla_peer;
372 nla_peer = data[VETH_INFO_PEER];
373 ifmp = nla_data(nla_peer);
374 err = rtnl_nla_parse_ifla(peer_tb,
375 nla_data(nla_peer) + sizeof(struct ifinfomsg),
376 nla_len(nla_peer) - sizeof(struct ifinfomsg));
380 err = veth_validate(peer_tb, NULL);
390 if (tbp[IFLA_IFNAME]) {
391 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
392 name_assign_type = NET_NAME_USER;
394 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
395 name_assign_type = NET_NAME_ENUM;
398 net = rtnl_link_get_net(src_net, tbp);
402 peer = rtnl_create_link(net, ifname, name_assign_type,
403 &veth_link_ops, tbp);
406 return PTR_ERR(peer);
409 if (tbp[IFLA_ADDRESS] == NULL)
410 eth_hw_addr_random(peer);
412 if (ifmp && (dev->ifindex != 0))
413 peer->ifindex = ifmp->ifi_index;
415 err = register_netdevice(peer);
419 goto err_register_peer;
421 netif_carrier_off(peer);
423 err = rtnl_configure_link(peer, ifmp);
425 goto err_configure_peer;
430 * note, that since we've registered new device the dev's name
431 * should be re-allocated
434 if (tb[IFLA_ADDRESS] == NULL)
435 eth_hw_addr_random(dev);
438 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
440 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
442 err = register_netdevice(dev);
444 goto err_register_dev;
446 netif_carrier_off(dev);
449 * tie the deviced together
452 priv = netdev_priv(dev);
453 rcu_assign_pointer(priv->peer, peer);
455 priv = netdev_priv(peer);
456 rcu_assign_pointer(priv->peer, dev);
462 unregister_netdevice(peer);
470 static void veth_dellink(struct net_device *dev, struct list_head *head)
472 struct veth_priv *priv;
473 struct net_device *peer;
475 priv = netdev_priv(dev);
476 peer = rtnl_dereference(priv->peer);
478 /* Note : dellink() is called from default_device_exit_batch(),
479 * before a rcu_synchronize() point. The devices are guaranteed
480 * not being freed before one RCU grace period.
482 RCU_INIT_POINTER(priv->peer, NULL);
483 unregister_netdevice_queue(dev, head);
486 priv = netdev_priv(peer);
487 RCU_INIT_POINTER(priv->peer, NULL);
488 unregister_netdevice_queue(peer, head);
492 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
493 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
496 static struct net *veth_get_link_net(const struct net_device *dev)
498 struct veth_priv *priv = netdev_priv(dev);
499 struct net_device *peer = rtnl_dereference(priv->peer);
501 return peer ? dev_net(peer) : dev_net(dev);
504 static struct rtnl_link_ops veth_link_ops = {
506 .priv_size = sizeof(struct veth_priv),
508 .validate = veth_validate,
509 .newlink = veth_newlink,
510 .dellink = veth_dellink,
511 .policy = veth_policy,
512 .maxtype = VETH_INFO_MAX,
513 .get_link_net = veth_get_link_net,
520 static __init int veth_init(void)
522 return rtnl_link_register(&veth_link_ops);
525 static __exit void veth_exit(void)
527 rtnl_link_unregister(&veth_link_ops);
530 module_init(veth_init);
531 module_exit(veth_exit);
533 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
534 MODULE_LICENSE("GPL v2");
535 MODULE_ALIAS_RTNL_LINK(DRV_NAME);