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;
38 unsigned requested_headroom;
46 const char string[ETH_GSTRING_LEN];
47 } ethtool_stats_keys[] = {
51 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
55 ethtool_cmd_speed_set(cmd, SPEED_10000);
56 cmd->duplex = DUPLEX_FULL;
59 cmd->transceiver = XCVR_INTERNAL;
60 cmd->autoneg = AUTONEG_DISABLE;
66 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
68 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
69 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
72 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
76 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
81 static int veth_get_sset_count(struct net_device *dev, int sset)
85 return ARRAY_SIZE(ethtool_stats_keys);
91 static void veth_get_ethtool_stats(struct net_device *dev,
92 struct ethtool_stats *stats, u64 *data)
94 struct veth_priv *priv = netdev_priv(dev);
95 struct net_device *peer = rtnl_dereference(priv->peer);
97 data[0] = peer ? peer->ifindex : 0;
100 static const struct ethtool_ops veth_ethtool_ops = {
101 .get_settings = veth_get_settings,
102 .get_drvinfo = veth_get_drvinfo,
103 .get_link = ethtool_op_get_link,
104 .get_strings = veth_get_strings,
105 .get_sset_count = veth_get_sset_count,
106 .get_ethtool_stats = veth_get_ethtool_stats,
109 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
111 struct veth_priv *priv = netdev_priv(dev);
112 struct net_device *rcv;
113 int length = skb->len;
116 rcv = rcu_dereference(priv->peer);
117 if (unlikely(!rcv)) {
122 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
123 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
125 u64_stats_update_begin(&stats->syncp);
126 stats->bytes += length;
128 u64_stats_update_end(&stats->syncp);
131 atomic64_inc(&priv->dropped);
141 static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
143 struct veth_priv *priv = netdev_priv(dev);
148 for_each_possible_cpu(cpu) {
149 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
154 start = u64_stats_fetch_begin_irq(&stats->syncp);
155 packets = stats->packets;
156 bytes = stats->bytes;
157 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
158 result->packets += packets;
159 result->bytes += bytes;
161 return atomic64_read(&priv->dropped);
164 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
165 struct rtnl_link_stats64 *tot)
167 struct veth_priv *priv = netdev_priv(dev);
168 struct net_device *peer;
169 struct pcpu_vstats one;
171 tot->tx_dropped = veth_stats_one(&one, dev);
172 tot->tx_bytes = one.bytes;
173 tot->tx_packets = one.packets;
176 peer = rcu_dereference(priv->peer);
178 tot->rx_dropped = veth_stats_one(&one, peer);
179 tot->rx_bytes = one.bytes;
180 tot->rx_packets = one.packets;
187 /* fake multicast ability */
188 static void veth_set_multicast_list(struct net_device *dev)
192 static int veth_open(struct net_device *dev)
194 struct veth_priv *priv = netdev_priv(dev);
195 struct net_device *peer = rtnl_dereference(priv->peer);
200 if (peer->flags & IFF_UP) {
201 netif_carrier_on(dev);
202 netif_carrier_on(peer);
207 static int veth_close(struct net_device *dev)
209 struct veth_priv *priv = netdev_priv(dev);
210 struct net_device *peer = rtnl_dereference(priv->peer);
212 netif_carrier_off(dev);
214 netif_carrier_off(peer);
219 static int is_valid_veth_mtu(int new_mtu)
221 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
224 static int veth_change_mtu(struct net_device *dev, int new_mtu)
226 if (!is_valid_veth_mtu(new_mtu))
232 static int veth_dev_init(struct net_device *dev)
234 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
240 static void veth_dev_free(struct net_device *dev)
242 free_percpu(dev->vstats);
246 #ifdef CONFIG_NET_POLL_CONTROLLER
247 static void veth_poll_controller(struct net_device *dev)
249 /* veth only receives frames when its peer sends one
250 * Since it's a synchronous operation, we are guaranteed
251 * never to have pending data when we poll for it so
252 * there is nothing to do here.
254 * We need this though so netpoll recognizes us as an interface that
255 * supports polling, which enables bridge devices in virt setups to
256 * still use netconsole
259 #endif /* CONFIG_NET_POLL_CONTROLLER */
261 static int veth_get_iflink(const struct net_device *dev)
263 struct veth_priv *priv = netdev_priv(dev);
264 struct net_device *peer;
268 peer = rcu_dereference(priv->peer);
269 iflink = peer ? peer->ifindex : 0;
275 static void veth_set_rx_headroom(struct net_device *dev, int new_hr)
277 struct veth_priv *peer_priv, *priv = netdev_priv(dev);
278 struct net_device *peer;
284 peer = rcu_dereference(priv->peer);
288 peer_priv = netdev_priv(peer);
289 priv->requested_headroom = new_hr;
290 new_hr = max(priv->requested_headroom, peer_priv->requested_headroom);
291 dev->needed_headroom = new_hr;
292 peer->needed_headroom = new_hr;
298 static const struct net_device_ops veth_netdev_ops = {
299 .ndo_init = veth_dev_init,
300 .ndo_open = veth_open,
301 .ndo_stop = veth_close,
302 .ndo_start_xmit = veth_xmit,
303 .ndo_change_mtu = veth_change_mtu,
304 .ndo_get_stats64 = veth_get_stats64,
305 .ndo_set_rx_mode = veth_set_multicast_list,
306 .ndo_set_mac_address = eth_mac_addr,
307 #ifdef CONFIG_NET_POLL_CONTROLLER
308 .ndo_poll_controller = veth_poll_controller,
310 .ndo_get_iflink = veth_get_iflink,
311 .ndo_features_check = passthru_features_check,
312 .ndo_set_rx_headroom = veth_set_rx_headroom,
315 #define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HW_CSUM | \
316 NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
317 NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL | \
318 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
319 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
321 static void veth_setup(struct net_device *dev)
325 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
326 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
327 dev->priv_flags |= IFF_NO_QUEUE;
328 dev->priv_flags |= IFF_PHONY_HEADROOM;
330 dev->netdev_ops = &veth_netdev_ops;
331 dev->ethtool_ops = &veth_ethtool_ops;
332 dev->features |= NETIF_F_LLTX;
333 dev->features |= VETH_FEATURES;
334 dev->vlan_features = dev->features &
335 ~(NETIF_F_HW_VLAN_CTAG_TX |
336 NETIF_F_HW_VLAN_STAG_TX |
337 NETIF_F_HW_VLAN_CTAG_RX |
338 NETIF_F_HW_VLAN_STAG_RX);
339 dev->destructor = veth_dev_free;
341 dev->hw_features = VETH_FEATURES;
342 dev->hw_enc_features = VETH_FEATURES;
349 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
351 if (tb[IFLA_ADDRESS]) {
352 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
354 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
355 return -EADDRNOTAVAIL;
358 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
364 static struct rtnl_link_ops veth_link_ops;
366 static int veth_newlink(struct net *src_net, struct net_device *dev,
367 struct nlattr *tb[], struct nlattr *data[])
370 struct net_device *peer;
371 struct veth_priv *priv;
372 char ifname[IFNAMSIZ];
373 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
374 unsigned char name_assign_type;
375 struct ifinfomsg *ifmp;
379 * create and register peer first
381 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
382 struct nlattr *nla_peer;
384 nla_peer = data[VETH_INFO_PEER];
385 ifmp = nla_data(nla_peer);
386 err = rtnl_nla_parse_ifla(peer_tb,
387 nla_data(nla_peer) + sizeof(struct ifinfomsg),
388 nla_len(nla_peer) - sizeof(struct ifinfomsg));
392 err = veth_validate(peer_tb, NULL);
402 if (tbp[IFLA_IFNAME]) {
403 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
404 name_assign_type = NET_NAME_USER;
406 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
407 name_assign_type = NET_NAME_ENUM;
410 net = rtnl_link_get_net(src_net, tbp);
414 peer = rtnl_create_link(net, ifname, name_assign_type,
415 &veth_link_ops, tbp);
418 return PTR_ERR(peer);
421 if (tbp[IFLA_ADDRESS] == NULL)
422 eth_hw_addr_random(peer);
424 if (ifmp && (dev->ifindex != 0))
425 peer->ifindex = ifmp->ifi_index;
427 err = register_netdevice(peer);
431 goto err_register_peer;
433 netif_carrier_off(peer);
435 err = rtnl_configure_link(peer, ifmp);
437 goto err_configure_peer;
442 * note, that since we've registered new device the dev's name
443 * should be re-allocated
446 if (tb[IFLA_ADDRESS] == NULL)
447 eth_hw_addr_random(dev);
450 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
452 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
454 err = register_netdevice(dev);
456 goto err_register_dev;
458 netif_carrier_off(dev);
461 * tie the deviced together
464 priv = netdev_priv(dev);
465 rcu_assign_pointer(priv->peer, peer);
467 priv = netdev_priv(peer);
468 rcu_assign_pointer(priv->peer, dev);
474 unregister_netdevice(peer);
482 static void veth_dellink(struct net_device *dev, struct list_head *head)
484 struct veth_priv *priv;
485 struct net_device *peer;
487 priv = netdev_priv(dev);
488 peer = rtnl_dereference(priv->peer);
490 /* Note : dellink() is called from default_device_exit_batch(),
491 * before a rcu_synchronize() point. The devices are guaranteed
492 * not being freed before one RCU grace period.
494 RCU_INIT_POINTER(priv->peer, NULL);
495 unregister_netdevice_queue(dev, head);
498 priv = netdev_priv(peer);
499 RCU_INIT_POINTER(priv->peer, NULL);
500 unregister_netdevice_queue(peer, head);
504 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
505 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
508 static struct net *veth_get_link_net(const struct net_device *dev)
510 struct veth_priv *priv = netdev_priv(dev);
511 struct net_device *peer = rtnl_dereference(priv->peer);
513 return peer ? dev_net(peer) : dev_net(dev);
516 static struct rtnl_link_ops veth_link_ops = {
518 .priv_size = sizeof(struct veth_priv),
520 .validate = veth_validate,
521 .newlink = veth_newlink,
522 .dellink = veth_dellink,
523 .policy = veth_policy,
524 .maxtype = VETH_INFO_MAX,
525 .get_link_net = veth_get_link_net,
532 static __init int veth_init(void)
534 return rtnl_link_register(&veth_link_ops);
537 static __exit void veth_exit(void)
539 rtnl_link_unregister(&veth_link_ops);
542 module_init(veth_init);
543 module_exit(veth_exit);
545 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
546 MODULE_LICENSE("GPL v2");
547 MODULE_ALIAS_RTNL_LINK(DRV_NAME);