xref: /linux/drivers/net/veth.c (revision 273b281fa22c293963ee3e6eec418f5dda2dbc83) 
1 /*
2  *  drivers/net/veth.c
3  *
4  *  Copyright (C) 2007 OpenVZ http://openvz.org, SWsoft Inc
5  *
6  * Author: Pavel Emelianov <xemul@openvz.org>
7  * Ethtool interface from: Eric W. Biederman <ebiederm@xmission.com>
8  *
9  */
10 
11 #include <linux/netdevice.h>
12 #include <linux/ethtool.h>
13 #include <linux/etherdevice.h>
14 
15 #include <net/dst.h>
16 #include <net/xfrm.h>
17 #include <linux/veth.h>
18 
19 #define DRV_NAME	"veth"
20 #define DRV_VERSION	"1.0"
21 
22 #define MIN_MTU 68		/* Min L3 MTU */
23 #define MAX_MTU 65535		/* Max L3 MTU (arbitrary) */
24 #define MTU_PAD (ETH_HLEN + 4)  /* Max difference between L2 and L3 size MTU */
25 
26 struct veth_net_stats {
27 	unsigned long	rx_packets;
28 	unsigned long	tx_packets;
29 	unsigned long	rx_bytes;
30 	unsigned long	tx_bytes;
31 	unsigned long	tx_dropped;
32 	unsigned long	rx_dropped;
33 };
34 
35 struct veth_priv {
36 	struct net_device *peer;
37 	struct veth_net_stats *stats;
38 	unsigned ip_summed;
39 };
40 
41 /*
42  * ethtool interface
43  */
44 
45 static struct {
46 	const char string[ETH_GSTRING_LEN];
47 } ethtool_stats_keys[] = {
48 	{ "peer_ifindex" },
49 };
50 
51 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
52 {
53 	cmd->supported		= 0;
54 	cmd->advertising	= 0;
55 	cmd->speed		= SPEED_10000;
56 	cmd->duplex		= DUPLEX_FULL;
57 	cmd->port		= PORT_TP;
58 	cmd->phy_address	= 0;
59 	cmd->transceiver	= XCVR_INTERNAL;
60 	cmd->autoneg		= AUTONEG_DISABLE;
61 	cmd->maxtxpkt		= 0;
62 	cmd->maxrxpkt		= 0;
63 	return 0;
64 }
65 
66 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
67 {
68 	strcpy(info->driver, DRV_NAME);
69 	strcpy(info->version, DRV_VERSION);
70 	strcpy(info->fw_version, "N/A");
71 }
72 
73 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
74 {
75 	switch(stringset) {
76 	case ETH_SS_STATS:
77 		memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
78 		break;
79 	}
80 }
81 
82 static int veth_get_sset_count(struct net_device *dev, int sset)
83 {
84 	switch (sset) {
85 	case ETH_SS_STATS:
86 		return ARRAY_SIZE(ethtool_stats_keys);
87 	default:
88 		return -EOPNOTSUPP;
89 	}
90 }
91 
92 static void veth_get_ethtool_stats(struct net_device *dev,
93 		struct ethtool_stats *stats, u64 *data)
94 {
95 	struct veth_priv *priv;
96 
97 	priv = netdev_priv(dev);
98 	data[0] = priv->peer->ifindex;
99 }
100 
101 static u32 veth_get_rx_csum(struct net_device *dev)
102 {
103 	struct veth_priv *priv;
104 
105 	priv = netdev_priv(dev);
106 	return priv->ip_summed == CHECKSUM_UNNECESSARY;
107 }
108 
109 static int veth_set_rx_csum(struct net_device *dev, u32 data)
110 {
111 	struct veth_priv *priv;
112 
113 	priv = netdev_priv(dev);
114 	priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
115 	return 0;
116 }
117 
118 static u32 veth_get_tx_csum(struct net_device *dev)
119 {
120 	return (dev->features & NETIF_F_NO_CSUM) != 0;
121 }
122 
123 static int veth_set_tx_csum(struct net_device *dev, u32 data)
124 {
125 	if (data)
126 		dev->features |= NETIF_F_NO_CSUM;
127 	else
128 		dev->features &= ~NETIF_F_NO_CSUM;
129 	return 0;
130 }
131 
132 static const struct ethtool_ops veth_ethtool_ops = {
133 	.get_settings		= veth_get_settings,
134 	.get_drvinfo		= veth_get_drvinfo,
135 	.get_link		= ethtool_op_get_link,
136 	.get_rx_csum		= veth_get_rx_csum,
137 	.set_rx_csum		= veth_set_rx_csum,
138 	.get_tx_csum		= veth_get_tx_csum,
139 	.set_tx_csum		= veth_set_tx_csum,
140 	.get_sg			= ethtool_op_get_sg,
141 	.set_sg			= ethtool_op_set_sg,
142 	.get_strings		= veth_get_strings,
143 	.get_sset_count		= veth_get_sset_count,
144 	.get_ethtool_stats	= veth_get_ethtool_stats,
145 };
146 
147 /*
148  * xmit
149  */
150 
151 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
152 {
153 	struct net_device *rcv = NULL;
154 	struct veth_priv *priv, *rcv_priv;
155 	struct veth_net_stats *stats, *rcv_stats;
156 	int length, cpu;
157 
158 	priv = netdev_priv(dev);
159 	rcv = priv->peer;
160 	rcv_priv = netdev_priv(rcv);
161 
162 	cpu = smp_processor_id();
163 	stats = per_cpu_ptr(priv->stats, cpu);
164 	rcv_stats = per_cpu_ptr(rcv_priv->stats, cpu);
165 
166 	if (!(rcv->flags & IFF_UP))
167 		goto tx_drop;
168 
169 	if (dev->features & NETIF_F_NO_CSUM)
170 		skb->ip_summed = rcv_priv->ip_summed;
171 
172 	length = skb->len + ETH_HLEN;
173 	if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
174 		goto rx_drop;
175 
176 	stats->tx_bytes += length;
177 	stats->tx_packets++;
178 
179 	rcv_stats->rx_bytes += length;
180 	rcv_stats->rx_packets++;
181 
182 	return NETDEV_TX_OK;
183 
184 tx_drop:
185 	kfree_skb(skb);
186 	stats->tx_dropped++;
187 	return NETDEV_TX_OK;
188 
189 rx_drop:
190 	kfree_skb(skb);
191 	rcv_stats->rx_dropped++;
192 	return NETDEV_TX_OK;
193 }
194 
195 /*
196  * general routines
197  */
198 
199 static struct net_device_stats *veth_get_stats(struct net_device *dev)
200 {
201 	struct veth_priv *priv;
202 	int cpu;
203 	struct veth_net_stats *stats, total = {0};
204 
205 	priv = netdev_priv(dev);
206 
207 	for_each_possible_cpu(cpu) {
208 		stats = per_cpu_ptr(priv->stats, cpu);
209 
210 		total.rx_packets += stats->rx_packets;
211 		total.tx_packets += stats->tx_packets;
212 		total.rx_bytes   += stats->rx_bytes;
213 		total.tx_bytes   += stats->tx_bytes;
214 		total.tx_dropped += stats->tx_dropped;
215 		total.rx_dropped += stats->rx_dropped;
216 	}
217 	dev->stats.rx_packets = total.rx_packets;
218 	dev->stats.tx_packets = total.tx_packets;
219 	dev->stats.rx_bytes   = total.rx_bytes;
220 	dev->stats.tx_bytes   = total.tx_bytes;
221 	dev->stats.tx_dropped = total.tx_dropped;
222 	dev->stats.rx_dropped = total.rx_dropped;
223 
224 	return &dev->stats;
225 }
226 
227 static int veth_open(struct net_device *dev)
228 {
229 	struct veth_priv *priv;
230 
231 	priv = netdev_priv(dev);
232 	if (priv->peer == NULL)
233 		return -ENOTCONN;
234 
235 	if (priv->peer->flags & IFF_UP) {
236 		netif_carrier_on(dev);
237 		netif_carrier_on(priv->peer);
238 	}
239 	return 0;
240 }
241 
242 static int veth_close(struct net_device *dev)
243 {
244 	struct veth_priv *priv = netdev_priv(dev);
245 
246 	netif_carrier_off(dev);
247 	netif_carrier_off(priv->peer);
248 
249 	return 0;
250 }
251 
252 static int is_valid_veth_mtu(int new_mtu)
253 {
254 	return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
255 }
256 
257 static int veth_change_mtu(struct net_device *dev, int new_mtu)
258 {
259 	if (!is_valid_veth_mtu(new_mtu))
260 		return -EINVAL;
261 	dev->mtu = new_mtu;
262 	return 0;
263 }
264 
265 static int veth_dev_init(struct net_device *dev)
266 {
267 	struct veth_net_stats *stats;
268 	struct veth_priv *priv;
269 
270 	stats = alloc_percpu(struct veth_net_stats);
271 	if (stats == NULL)
272 		return -ENOMEM;
273 
274 	priv = netdev_priv(dev);
275 	priv->stats = stats;
276 	return 0;
277 }
278 
279 static void veth_dev_free(struct net_device *dev)
280 {
281 	struct veth_priv *priv;
282 
283 	priv = netdev_priv(dev);
284 	free_percpu(priv->stats);
285 	free_netdev(dev);
286 }
287 
288 static const struct net_device_ops veth_netdev_ops = {
289 	.ndo_init            = veth_dev_init,
290 	.ndo_open            = veth_open,
291 	.ndo_stop            = veth_close,
292 	.ndo_start_xmit      = veth_xmit,
293 	.ndo_change_mtu      = veth_change_mtu,
294 	.ndo_get_stats       = veth_get_stats,
295 	.ndo_set_mac_address = eth_mac_addr,
296 };
297 
298 static void veth_setup(struct net_device *dev)
299 {
300 	ether_setup(dev);
301 
302 	dev->netdev_ops = &veth_netdev_ops;
303 	dev->ethtool_ops = &veth_ethtool_ops;
304 	dev->features |= NETIF_F_LLTX;
305 	dev->destructor = veth_dev_free;
306 }
307 
308 /*
309  * netlink interface
310  */
311 
312 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
313 {
314 	if (tb[IFLA_ADDRESS]) {
315 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
316 			return -EINVAL;
317 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
318 			return -EADDRNOTAVAIL;
319 	}
320 	if (tb[IFLA_MTU]) {
321 		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
322 			return -EINVAL;
323 	}
324 	return 0;
325 }
326 
327 static struct rtnl_link_ops veth_link_ops;
328 
329 static int veth_newlink(struct net *src_net, struct net_device *dev,
330 			 struct nlattr *tb[], struct nlattr *data[])
331 {
332 	int err;
333 	struct net_device *peer;
334 	struct veth_priv *priv;
335 	char ifname[IFNAMSIZ];
336 	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
337 	struct net *net;
338 
339 	/*
340 	 * create and register peer first
341 	 *
342 	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
343 	 * skip it since no info from it is useful yet
344 	 */
345 
346 	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
347 		struct nlattr *nla_peer;
348 
349 		nla_peer = data[VETH_INFO_PEER];
350 		err = nla_parse(peer_tb, IFLA_MAX,
351 				nla_data(nla_peer) + sizeof(struct ifinfomsg),
352 				nla_len(nla_peer) - sizeof(struct ifinfomsg),
353 				ifla_policy);
354 		if (err < 0)
355 			return err;
356 
357 		err = veth_validate(peer_tb, NULL);
358 		if (err < 0)
359 			return err;
360 
361 		tbp = peer_tb;
362 	} else
363 		tbp = tb;
364 
365 	if (tbp[IFLA_IFNAME])
366 		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
367 	else
368 		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
369 
370 	net = rtnl_link_get_net(src_net, tbp);
371 	if (IS_ERR(net))
372 		return PTR_ERR(net);
373 
374 	peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
375 	if (IS_ERR(peer)) {
376 		put_net(net);
377 		return PTR_ERR(peer);
378 	}
379 
380 	if (tbp[IFLA_ADDRESS] == NULL)
381 		random_ether_addr(peer->dev_addr);
382 
383 	err = register_netdevice(peer);
384 	put_net(net);
385 	net = NULL;
386 	if (err < 0)
387 		goto err_register_peer;
388 
389 	netif_carrier_off(peer);
390 
391 	/*
392 	 * register dev last
393 	 *
394 	 * note, that since we've registered new device the dev's name
395 	 * should be re-allocated
396 	 */
397 
398 	if (tb[IFLA_ADDRESS] == NULL)
399 		random_ether_addr(dev->dev_addr);
400 
401 	if (tb[IFLA_IFNAME])
402 		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
403 	else
404 		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
405 
406 	if (strchr(dev->name, '%')) {
407 		err = dev_alloc_name(dev, dev->name);
408 		if (err < 0)
409 			goto err_alloc_name;
410 	}
411 
412 	err = register_netdevice(dev);
413 	if (err < 0)
414 		goto err_register_dev;
415 
416 	netif_carrier_off(dev);
417 
418 	/*
419 	 * tie the deviced together
420 	 */
421 
422 	priv = netdev_priv(dev);
423 	priv->peer = peer;
424 
425 	priv = netdev_priv(peer);
426 	priv->peer = dev;
427 	return 0;
428 
429 err_register_dev:
430 	/* nothing to do */
431 err_alloc_name:
432 	unregister_netdevice(peer);
433 	return err;
434 
435 err_register_peer:
436 	free_netdev(peer);
437 	return err;
438 }
439 
440 static void veth_dellink(struct net_device *dev, struct list_head *head)
441 {
442 	struct veth_priv *priv;
443 	struct net_device *peer;
444 
445 	priv = netdev_priv(dev);
446 	peer = priv->peer;
447 
448 	unregister_netdevice_queue(dev, head);
449 	unregister_netdevice_queue(peer, head);
450 }
451 
452 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
453 
454 static struct rtnl_link_ops veth_link_ops = {
455 	.kind		= DRV_NAME,
456 	.priv_size	= sizeof(struct veth_priv),
457 	.setup		= veth_setup,
458 	.validate	= veth_validate,
459 	.newlink	= veth_newlink,
460 	.dellink	= veth_dellink,
461 	.policy		= veth_policy,
462 	.maxtype	= VETH_INFO_MAX,
463 };
464 
465 /*
466  * init/fini
467  */
468 
469 static __init int veth_init(void)
470 {
471 	return rtnl_link_register(&veth_link_ops);
472 }
473 
474 static __exit void veth_exit(void)
475 {
476 	rtnl_link_unregister(&veth_link_ops);
477 }
478 
479 module_init(veth_init);
480 module_exit(veth_exit);
481 
482 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
483 MODULE_LICENSE("GPL v2");
484 MODULE_ALIAS_RTNL_LINK(DRV_NAME);
485