1
2
3
4
5
6
7
8
9
10
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>
16
17#include <net/dst.h>
18#include <net/xfrm.h>
19#include <linux/veth.h>
20#include <linux/module.h>
21
22#define DRV_NAME "veth"
23#define DRV_VERSION "1.0"
24
25#define MIN_MTU 68
26#define MAX_MTU 65535
27
28struct pcpu_vstats {
29 u64 packets;
30 u64 bytes;
31 struct u64_stats_sync syncp;
32};
33
34struct veth_priv {
35 struct net_device __rcu *peer;
36 atomic64_t dropped;
37};
38
39
40
41
42
43static struct {
44 const char string[ETH_GSTRING_LEN];
45} ethtool_stats_keys[] = {
46 { "peer_ifindex" },
47};
48
49static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
50{
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;
62}
63
64static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
65{
66 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
67 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
68}
69
70static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
71{
72 switch(stringset) {
73 case ETH_SS_STATS:
74 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
75 break;
76 }
77}
78
79static int veth_get_sset_count(struct net_device *dev, int sset)
80{
81 switch (sset) {
82 case ETH_SS_STATS:
83 return ARRAY_SIZE(ethtool_stats_keys);
84 default:
85 return -EOPNOTSUPP;
86 }
87}
88
89static void veth_get_ethtool_stats(struct net_device *dev,
90 struct ethtool_stats *stats, u64 *data)
91{
92 struct veth_priv *priv = netdev_priv(dev);
93 struct net_device *peer = rtnl_dereference(priv->peer);
94
95 data[0] = peer ? peer->ifindex : 0;
96}
97
98static 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,
105};
106
107static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
108{
109 struct veth_priv *priv = netdev_priv(dev);
110 struct net_device *rcv;
111 int length = skb->len;
112
113 rcu_read_lock();
114 rcv = rcu_dereference(priv->peer);
115 if (unlikely(!rcv)) {
116 kfree_skb(skb);
117 goto drop;
118 }
119
120
121
122 if (skb->ip_summed == CHECKSUM_NONE &&
123 rcv->features & NETIF_F_RXCSUM)
124 skb->ip_summed = CHECKSUM_UNNECESSARY;
125
126 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
127 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
128
129 u64_stats_update_begin(&stats->syncp);
130 stats->bytes += length;
131 stats->packets++;
132 u64_stats_update_end(&stats->syncp);
133 } else {
134drop:
135 atomic64_inc(&priv->dropped);
136 }
137 rcu_read_unlock();
138 return NETDEV_TX_OK;
139}
140
141
142
143
144
145static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
146{
147 struct veth_priv *priv = netdev_priv(dev);
148 int cpu;
149
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;
156
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;
164 }
165 return atomic64_read(&priv->dropped);
166}
167
168static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
169 struct rtnl_link_stats64 *tot)
170{
171 struct veth_priv *priv = netdev_priv(dev);
172 struct net_device *peer;
173 struct pcpu_vstats one;
174
175 tot->tx_dropped = veth_stats_one(&one, dev);
176 tot->tx_bytes = one.bytes;
177 tot->tx_packets = one.packets;
178
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;
185 }
186 rcu_read_unlock();
187
188 return tot;
189}
190
191
192static void veth_set_multicast_list(struct net_device *dev)
193{
194}
195
196static int veth_open(struct net_device *dev)
197{
198 struct veth_priv *priv = netdev_priv(dev);
199 struct net_device *peer = rtnl_dereference(priv->peer);
200
201 if (!peer)
202 return -ENOTCONN;
203
204 if (peer->flags & IFF_UP) {
205 netif_carrier_on(dev);
206 netif_carrier_on(peer);
207 }
208 return 0;
209}
210
211static int veth_close(struct net_device *dev)
212{
213 struct veth_priv *priv = netdev_priv(dev);
214 struct net_device *peer = rtnl_dereference(priv->peer);
215
216 netif_carrier_off(dev);
217 if (peer)
218 netif_carrier_off(peer);
219
220 return 0;
221}
222
223static int is_valid_veth_mtu(int new_mtu)
224{
225 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
226}
227
228static int veth_change_mtu(struct net_device *dev, int new_mtu)
229{
230 if (!is_valid_veth_mtu(new_mtu))
231 return -EINVAL;
232 dev->mtu = new_mtu;
233 return 0;
234}
235
236static int veth_dev_init(struct net_device *dev)
237{
238 int i;
239
240 dev->vstats = alloc_percpu(struct pcpu_vstats);
241 if (!dev->vstats)
242 return -ENOMEM;
243
244 for_each_possible_cpu(i) {
245 struct pcpu_vstats *veth_stats;
246 veth_stats = per_cpu_ptr(dev->vstats, i);
247 u64_stats_init(&veth_stats->syncp);
248 }
249
250 return 0;
251}
252
253static void veth_dev_free(struct net_device *dev)
254{
255 free_percpu(dev->vstats);
256 free_netdev(dev);
257}
258
259static const struct net_device_ops veth_netdev_ops = {
260 .ndo_init = veth_dev_init,
261 .ndo_open = veth_open,
262 .ndo_stop = veth_close,
263 .ndo_start_xmit = veth_xmit,
264 .ndo_change_mtu = veth_change_mtu,
265 .ndo_get_stats64 = veth_get_stats64,
266 .ndo_set_rx_mode = veth_set_multicast_list,
267 .ndo_set_mac_address = eth_mac_addr,
268};
269
270#define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
271 NETIF_F_HW_CSUM | NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
272 NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL | \
273 NETIF_F_GSO_IPIP | NETIF_F_GSO_SIT | NETIF_F_UFO | \
274 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
275 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
276
277static void veth_setup(struct net_device *dev)
278{
279 ether_setup(dev);
280
281 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
282 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
283
284 dev->netdev_ops = &veth_netdev_ops;
285 dev->ethtool_ops = &veth_ethtool_ops;
286 dev->features |= NETIF_F_LLTX;
287 dev->features |= VETH_FEATURES;
288 dev->vlan_features = dev->features &
289 ~(NETIF_F_HW_VLAN_CTAG_TX |
290 NETIF_F_HW_VLAN_STAG_TX |
291 NETIF_F_HW_VLAN_CTAG_RX |
292 NETIF_F_HW_VLAN_STAG_RX);
293 dev->destructor = veth_dev_free;
294
295 dev->hw_features = VETH_FEATURES;
296 dev->hw_enc_features = VETH_FEATURES;
297}
298
299
300
301
302
303static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
304{
305 if (tb[IFLA_ADDRESS]) {
306 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
307 return -EINVAL;
308 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
309 return -EADDRNOTAVAIL;
310 }
311 if (tb[IFLA_MTU]) {
312 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
313 return -EINVAL;
314 }
315 return 0;
316}
317
318static struct rtnl_link_ops veth_link_ops;
319
320static int veth_newlink(struct net *src_net, struct net_device *dev,
321 struct nlattr *tb[], struct nlattr *data[])
322{
323 int err;
324 struct net_device *peer;
325 struct veth_priv *priv;
326 char ifname[IFNAMSIZ];
327 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
328 struct ifinfomsg *ifmp;
329 struct net *net;
330
331
332
333
334 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
335 struct nlattr *nla_peer;
336
337 nla_peer = data[VETH_INFO_PEER];
338 ifmp = nla_data(nla_peer);
339 err = nla_parse(peer_tb, IFLA_MAX,
340 nla_data(nla_peer) + sizeof(struct ifinfomsg),
341 nla_len(nla_peer) - sizeof(struct ifinfomsg),
342 ifla_policy);
343 if (err < 0)
344 return err;
345
346 err = veth_validate(peer_tb, NULL);
347 if (err < 0)
348 return err;
349
350 tbp = peer_tb;
351 } else {
352 ifmp = NULL;
353 tbp = tb;
354 }
355
356 if (tbp[IFLA_IFNAME])
357 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
358 else
359 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
360
361 net = rtnl_link_get_net(src_net, tbp);
362 if (IS_ERR(net))
363 return PTR_ERR(net);
364
365 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
366 if (IS_ERR(peer)) {
367 put_net(net);
368 return PTR_ERR(peer);
369 }
370
371 if (tbp[IFLA_ADDRESS] == NULL)
372 eth_hw_addr_random(peer);
373
374 if (ifmp && (dev->ifindex != 0))
375 peer->ifindex = ifmp->ifi_index;
376
377 err = register_netdevice(peer);
378 put_net(net);
379 net = NULL;
380 if (err < 0)
381 goto err_register_peer;
382
383 netif_carrier_off(peer);
384
385 err = rtnl_configure_link(peer, ifmp);
386 if (err < 0)
387 goto err_configure_peer;
388
389
390
391
392
393
394
395
396 if (tb[IFLA_ADDRESS] == NULL)
397 eth_hw_addr_random(dev);
398
399 if (tb[IFLA_IFNAME])
400 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
401 else
402 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
403
404 err = register_netdevice(dev);
405 if (err < 0)
406 goto err_register_dev;
407
408 netif_carrier_off(dev);
409
410
411
412
413
414 priv = netdev_priv(dev);
415 rcu_assign_pointer(priv->peer, peer);
416
417 priv = netdev_priv(peer);
418 rcu_assign_pointer(priv->peer, dev);
419 return 0;
420
421err_register_dev:
422
423err_configure_peer:
424 unregister_netdevice(peer);
425 return err;
426
427err_register_peer:
428 free_netdev(peer);
429 return err;
430}
431
432static void veth_dellink(struct net_device *dev, struct list_head *head)
433{
434 struct veth_priv *priv;
435 struct net_device *peer;
436
437 priv = netdev_priv(dev);
438 peer = rtnl_dereference(priv->peer);
439
440
441
442
443
444 RCU_INIT_POINTER(priv->peer, NULL);
445 unregister_netdevice_queue(dev, head);
446
447 if (peer) {
448 priv = netdev_priv(peer);
449 RCU_INIT_POINTER(priv->peer, NULL);
450 unregister_netdevice_queue(peer, head);
451 }
452}
453
454static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
455 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
456};
457
458static struct rtnl_link_ops veth_link_ops = {
459 .kind = DRV_NAME,
460 .priv_size = sizeof(struct veth_priv),
461 .setup = veth_setup,
462 .validate = veth_validate,
463 .newlink = veth_newlink,
464 .dellink = veth_dellink,
465 .policy = veth_policy,
466 .maxtype = VETH_INFO_MAX,
467};
468
469
470
471
472
473static __init int veth_init(void)
474{
475 return rtnl_link_register(&veth_link_ops);
476}
477
478static __exit void veth_exit(void)
479{
480 rtnl_link_unregister(&veth_link_ops);
481}
482
483module_init(veth_init);
484module_exit(veth_exit);
485
486MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
487MODULE_LICENSE("GPL v2");
488MODULE_ALIAS_RTNL_LINK(DRV_NAME);
489
