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>
19 #include <linux/veth.h>
20 #include <linux/module.h>
22 #define DRV_NAME "veth"
23 #define DRV_VERSION "1.0"
25 #define MIN_MTU 68 /* Min L3 MTU */
26 #define MAX_MTU 65535 /* Max L3 MTU (arbitrary) */
31 struct u64_stats_sync syncp;
35 struct net_device *peer;
44 const char string[ETH_GSTRING_LEN];
45 } ethtool_stats_keys[] = {
49 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
53 ethtool_cmd_speed_set(cmd, SPEED_10000);
54 cmd->duplex = DUPLEX_FULL;
57 cmd->transceiver = XCVR_INTERNAL;
58 cmd->autoneg = AUTONEG_DISABLE;
64 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
66 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
67 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
70 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
74 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
79 static int veth_get_sset_count(struct net_device *dev, int sset)
83 return ARRAY_SIZE(ethtool_stats_keys);
89 static void veth_get_ethtool_stats(struct net_device *dev,
90 struct ethtool_stats *stats, u64 *data)
92 struct veth_priv *priv;
94 priv = netdev_priv(dev);
95 data[0] = priv->peer->ifindex;
98 static const struct ethtool_ops veth_ethtool_ops = {
99 .get_settings = veth_get_settings,
100 .get_drvinfo = veth_get_drvinfo,
101 .get_link = ethtool_op_get_link,
102 .get_strings = veth_get_strings,
103 .get_sset_count = veth_get_sset_count,
104 .get_ethtool_stats = veth_get_ethtool_stats,
107 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
109 struct veth_priv *priv = netdev_priv(dev);
110 struct net_device *rcv = priv->peer;
111 int length = skb->len;
113 /* don't change ip_summed == CHECKSUM_PARTIAL, as that
114 * will cause bad checksum on forwarded packets
116 if (skb->ip_summed == CHECKSUM_NONE &&
117 rcv->features & NETIF_F_RXCSUM)
118 skb->ip_summed = CHECKSUM_UNNECESSARY;
120 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
121 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
123 u64_stats_update_begin(&stats->syncp);
124 stats->bytes += length;
126 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_bh(&stats->syncp);
152 packets = stats->packets;
153 bytes = stats->bytes;
154 } while (u64_stats_fetch_retry_bh(&stats->syncp, start));
155 result->packets += packets;
156 result->bytes += bytes;
158 return atomic64_read(&priv->dropped);
161 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
162 struct rtnl_link_stats64 *tot)
164 struct veth_priv *priv = netdev_priv(dev);
165 struct pcpu_vstats one;
167 tot->tx_dropped = veth_stats_one(&one, dev);
168 tot->tx_bytes = one.bytes;
169 tot->tx_packets = one.packets;
171 tot->rx_dropped = veth_stats_one(&one, priv->peer);
172 tot->rx_bytes = one.bytes;
173 tot->rx_packets = one.packets;
178 static int veth_open(struct net_device *dev)
180 struct veth_priv *priv;
182 priv = netdev_priv(dev);
183 if (priv->peer == NULL)
186 if (priv->peer->flags & IFF_UP) {
187 netif_carrier_on(dev);
188 netif_carrier_on(priv->peer);
193 static int veth_close(struct net_device *dev)
195 struct veth_priv *priv = netdev_priv(dev);
197 netif_carrier_off(dev);
198 netif_carrier_off(priv->peer);
203 static int is_valid_veth_mtu(int new_mtu)
205 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
208 static int veth_change_mtu(struct net_device *dev, int new_mtu)
210 if (!is_valid_veth_mtu(new_mtu))
216 static int veth_dev_init(struct net_device *dev)
218 dev->vstats = alloc_percpu(struct pcpu_vstats);
225 static void veth_dev_free(struct net_device *dev)
227 free_percpu(dev->vstats);
231 static const struct net_device_ops veth_netdev_ops = {
232 .ndo_init = veth_dev_init,
233 .ndo_open = veth_open,
234 .ndo_stop = veth_close,
235 .ndo_start_xmit = veth_xmit,
236 .ndo_change_mtu = veth_change_mtu,
237 .ndo_get_stats64 = veth_get_stats64,
238 .ndo_set_mac_address = eth_mac_addr,
241 static void veth_setup(struct net_device *dev)
245 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
246 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
248 dev->netdev_ops = &veth_netdev_ops;
249 dev->ethtool_ops = &veth_ethtool_ops;
250 dev->features |= NETIF_F_LLTX;
251 dev->destructor = veth_dev_free;
253 dev->hw_features = NETIF_F_HW_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
260 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
262 if (tb[IFLA_ADDRESS]) {
263 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
265 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
266 return -EADDRNOTAVAIL;
269 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
275 static struct rtnl_link_ops veth_link_ops;
277 static int veth_newlink(struct net *src_net, struct net_device *dev,
278 struct nlattr *tb[], struct nlattr *data[])
281 struct net_device *peer;
282 struct veth_priv *priv;
283 char ifname[IFNAMSIZ];
284 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
285 struct ifinfomsg *ifmp;
289 * create and register peer first
291 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
292 struct nlattr *nla_peer;
294 nla_peer = data[VETH_INFO_PEER];
295 ifmp = nla_data(nla_peer);
296 err = nla_parse(peer_tb, IFLA_MAX,
297 nla_data(nla_peer) + sizeof(struct ifinfomsg),
298 nla_len(nla_peer) - sizeof(struct ifinfomsg),
303 err = veth_validate(peer_tb, NULL);
313 if (tbp[IFLA_IFNAME])
314 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
316 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
318 net = rtnl_link_get_net(src_net, tbp);
322 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
325 return PTR_ERR(peer);
328 if (tbp[IFLA_ADDRESS] == NULL)
329 eth_hw_addr_random(peer);
331 if (ifmp && (dev->ifindex != 0))
332 peer->ifindex = ifmp->ifi_index;
334 err = register_netdevice(peer);
338 goto err_register_peer;
340 netif_carrier_off(peer);
342 err = rtnl_configure_link(peer, ifmp);
344 goto err_configure_peer;
349 * note, that since we've registered new device the dev's name
350 * should be re-allocated
353 if (tb[IFLA_ADDRESS] == NULL)
354 eth_hw_addr_random(dev);
357 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
359 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
361 if (strchr(dev->name, '%')) {
362 err = dev_alloc_name(dev, dev->name);
367 err = register_netdevice(dev);
369 goto err_register_dev;
371 netif_carrier_off(dev);
374 * tie the deviced together
377 priv = netdev_priv(dev);
380 priv = netdev_priv(peer);
388 unregister_netdevice(peer);
396 static void veth_dellink(struct net_device *dev, struct list_head *head)
398 struct veth_priv *priv;
399 struct net_device *peer;
401 priv = netdev_priv(dev);
404 unregister_netdevice_queue(dev, head);
405 unregister_netdevice_queue(peer, head);
408 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
409 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
412 static struct rtnl_link_ops veth_link_ops = {
414 .priv_size = sizeof(struct veth_priv),
416 .validate = veth_validate,
417 .newlink = veth_newlink,
418 .dellink = veth_dellink,
419 .policy = veth_policy,
420 .maxtype = VETH_INFO_MAX,
427 static __init int veth_init(void)
429 return rtnl_link_register(&veth_link_ops);
432 static __exit void veth_exit(void)
434 rtnl_link_unregister(&veth_link_ops);
437 module_init(veth_init);
438 module_exit(veth_exit);
440 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
441 MODULE_LICENSE("GPL v2");
442 MODULE_ALIAS_RTNL_LINK(DRV_NAME);