x86/xen: resume timer irqs early
[linux/fpc-iii.git] / drivers / net / veth.c
blob61c4044f644ec97f194f774b43f09ef05add049d
1 /*
2 * drivers/net/veth.c
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>
9 */
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/dst.h>
18 #include <net/xfrm.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) */
28 struct pcpu_vstats {
29 u64 packets;
30 u64 bytes;
31 struct u64_stats_sync syncp;
34 struct veth_priv {
35 struct net_device __rcu *peer;
36 atomic64_t dropped;
40 * ethtool interface
43 static struct {
44 const char string[ETH_GSTRING_LEN];
45 } ethtool_stats_keys[] = {
46 { "peer_ifindex" },
49 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
51 cmd->supported = 0;
52 cmd->advertising = 0;
53 ethtool_cmd_speed_set(cmd, SPEED_10000);
54 cmd->duplex = DUPLEX_FULL;
55 cmd->port = PORT_TP;
56 cmd->phy_address = 0;
57 cmd->transceiver = XCVR_INTERNAL;
58 cmd->autoneg = AUTONEG_DISABLE;
59 cmd->maxtxpkt = 0;
60 cmd->maxrxpkt = 0;
61 return 0;
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)
72 switch(stringset) {
73 case ETH_SS_STATS:
74 memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
75 break;
79 static int veth_get_sset_count(struct net_device *dev, int sset)
81 switch (sset) {
82 case ETH_SS_STATS:
83 return ARRAY_SIZE(ethtool_stats_keys);
84 default:
85 return -EOPNOTSUPP;
89 static void veth_get_ethtool_stats(struct net_device *dev,
90 struct ethtool_stats *stats, u64 *data)
92 struct veth_priv *priv = netdev_priv(dev);
93 struct net_device *peer = rtnl_dereference(priv->peer);
95 data[0] = peer ? peer->ifindex : 0;
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;
111 int length = skb->len;
113 rcu_read_lock();
114 rcv = rcu_dereference(priv->peer);
115 if (unlikely(!rcv)) {
116 kfree_skb(skb);
117 goto drop;
119 /* don't change ip_summed == CHECKSUM_PARTIAL, as that
120 * will cause bad checksum on forwarded packets
122 if (skb->ip_summed == CHECKSUM_NONE &&
123 rcv->features & NETIF_F_RXCSUM)
124 skb->ip_summed = CHECKSUM_UNNECESSARY;
126 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
127 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
129 u64_stats_update_begin(&stats->syncp);
130 stats->bytes += length;
131 stats->packets++;
132 u64_stats_update_end(&stats->syncp);
133 } else {
134 drop:
135 atomic64_inc(&priv->dropped);
137 rcu_read_unlock();
138 return NETDEV_TX_OK;
142 * general routines
145 static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
147 struct veth_priv *priv = netdev_priv(dev);
148 int cpu;
150 result->packets = 0;
151 result->bytes = 0;
152 for_each_possible_cpu(cpu) {
153 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
154 u64 packets, bytes;
155 unsigned int start;
157 do {
158 start = u64_stats_fetch_begin_bh(&stats->syncp);
159 packets = stats->packets;
160 bytes = stats->bytes;
161 } while (u64_stats_fetch_retry_bh(&stats->syncp, start));
162 result->packets += packets;
163 result->bytes += bytes;
165 return atomic64_read(&priv->dropped);
168 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
169 struct rtnl_link_stats64 *tot)
171 struct veth_priv *priv = netdev_priv(dev);
172 struct net_device *peer;
173 struct pcpu_vstats one;
175 tot->tx_dropped = veth_stats_one(&one, dev);
176 tot->tx_bytes = one.bytes;
177 tot->tx_packets = one.packets;
179 rcu_read_lock();
180 peer = rcu_dereference(priv->peer);
181 if (peer) {
182 tot->rx_dropped = veth_stats_one(&one, peer);
183 tot->rx_bytes = one.bytes;
184 tot->rx_packets = one.packets;
186 rcu_read_unlock();
188 return tot;
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);
196 if (!peer)
197 return -ENOTCONN;
199 if (peer->flags & IFF_UP) {
200 netif_carrier_on(dev);
201 netif_carrier_on(peer);
203 return 0;
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);
212 if (peer)
213 netif_carrier_off(peer);
215 return 0;
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))
226 return -EINVAL;
227 dev->mtu = new_mtu;
228 return 0;
231 static int veth_dev_init(struct net_device *dev)
233 dev->vstats = alloc_percpu(struct pcpu_vstats);
234 if (!dev->vstats)
235 return -ENOMEM;
237 return 0;
240 static void veth_dev_free(struct net_device *dev)
242 free_percpu(dev->vstats);
243 free_netdev(dev);
246 static const struct net_device_ops veth_netdev_ops = {
247 .ndo_init = veth_dev_init,
248 .ndo_open = veth_open,
249 .ndo_stop = veth_close,
250 .ndo_start_xmit = veth_xmit,
251 .ndo_change_mtu = veth_change_mtu,
252 .ndo_get_stats64 = veth_get_stats64,
253 .ndo_set_mac_address = eth_mac_addr,
256 #define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
257 NETIF_F_HW_CSUM | NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
258 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
259 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
261 static void veth_setup(struct net_device *dev)
263 ether_setup(dev);
265 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
266 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
268 dev->netdev_ops = &veth_netdev_ops;
269 dev->ethtool_ops = &veth_ethtool_ops;
270 dev->features |= NETIF_F_LLTX;
271 dev->features |= VETH_FEATURES;
272 dev->vlan_features = dev->features &
273 ~(NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX);
274 dev->destructor = veth_dev_free;
276 dev->hw_features = VETH_FEATURES;
280 * netlink interface
283 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
285 if (tb[IFLA_ADDRESS]) {
286 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
287 return -EINVAL;
288 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
289 return -EADDRNOTAVAIL;
291 if (tb[IFLA_MTU]) {
292 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
293 return -EINVAL;
295 return 0;
298 static struct rtnl_link_ops veth_link_ops;
300 static int veth_newlink(struct net *src_net, struct net_device *dev,
301 struct nlattr *tb[], struct nlattr *data[])
303 int err;
304 struct net_device *peer;
305 struct veth_priv *priv;
306 char ifname[IFNAMSIZ];
307 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
308 struct ifinfomsg *ifmp;
309 struct net *net;
312 * create and register peer first
314 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
315 struct nlattr *nla_peer;
317 nla_peer = data[VETH_INFO_PEER];
318 ifmp = nla_data(nla_peer);
319 err = nla_parse(peer_tb, IFLA_MAX,
320 nla_data(nla_peer) + sizeof(struct ifinfomsg),
321 nla_len(nla_peer) - sizeof(struct ifinfomsg),
322 ifla_policy);
323 if (err < 0)
324 return err;
326 err = veth_validate(peer_tb, NULL);
327 if (err < 0)
328 return err;
330 tbp = peer_tb;
331 } else {
332 ifmp = NULL;
333 tbp = tb;
336 if (tbp[IFLA_IFNAME])
337 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
338 else
339 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
341 net = rtnl_link_get_net(src_net, tbp);
342 if (IS_ERR(net))
343 return PTR_ERR(net);
345 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
346 if (IS_ERR(peer)) {
347 put_net(net);
348 return PTR_ERR(peer);
351 if (tbp[IFLA_ADDRESS] == NULL)
352 eth_hw_addr_random(peer);
354 if (ifmp && (dev->ifindex != 0))
355 peer->ifindex = ifmp->ifi_index;
357 err = register_netdevice(peer);
358 put_net(net);
359 net = NULL;
360 if (err < 0)
361 goto err_register_peer;
363 netif_carrier_off(peer);
365 err = rtnl_configure_link(peer, ifmp);
366 if (err < 0)
367 goto err_configure_peer;
370 * register dev last
372 * note, that since we've registered new device the dev's name
373 * should be re-allocated
376 if (tb[IFLA_ADDRESS] == NULL)
377 eth_hw_addr_random(dev);
379 if (tb[IFLA_IFNAME])
380 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
381 else
382 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
384 err = register_netdevice(dev);
385 if (err < 0)
386 goto err_register_dev;
388 netif_carrier_off(dev);
391 * tie the deviced together
394 priv = netdev_priv(dev);
395 rcu_assign_pointer(priv->peer, peer);
397 priv = netdev_priv(peer);
398 rcu_assign_pointer(priv->peer, dev);
399 return 0;
401 err_register_dev:
402 /* nothing to do */
403 err_configure_peer:
404 unregister_netdevice(peer);
405 return err;
407 err_register_peer:
408 free_netdev(peer);
409 return err;
412 static void veth_dellink(struct net_device *dev, struct list_head *head)
414 struct veth_priv *priv;
415 struct net_device *peer;
417 priv = netdev_priv(dev);
418 peer = rtnl_dereference(priv->peer);
420 /* Note : dellink() is called from default_device_exit_batch(),
421 * before a rcu_synchronize() point. The devices are guaranteed
422 * not being freed before one RCU grace period.
424 RCU_INIT_POINTER(priv->peer, NULL);
425 unregister_netdevice_queue(dev, head);
427 if (peer) {
428 priv = netdev_priv(peer);
429 RCU_INIT_POINTER(priv->peer, NULL);
430 unregister_netdevice_queue(peer, head);
434 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
435 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
438 static struct rtnl_link_ops veth_link_ops = {
439 .kind = DRV_NAME,
440 .priv_size = sizeof(struct veth_priv),
441 .setup = veth_setup,
442 .validate = veth_validate,
443 .newlink = veth_newlink,
444 .dellink = veth_dellink,
445 .policy = veth_policy,
446 .maxtype = VETH_INFO_MAX,
450 * init/fini
453 static __init int veth_init(void)
455 return rtnl_link_register(&veth_link_ops);
458 static __exit void veth_exit(void)
460 rtnl_link_unregister(&veth_link_ops);
463 module_init(veth_init);
464 module_exit(veth_exit);
466 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
467 MODULE_LICENSE("GPL v2");
468 MODULE_ALIAS_RTNL_LINK(DRV_NAME);