xref: /linux/drivers/net/veth.c (revision 8fa5723aa7e053d498336b48448b292fc2e0458b)
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/list.h>
12 #include <linux/netdevice.h>
13 #include <linux/ethtool.h>
14 #include <linux/etherdevice.h>
15 
16 #include <net/dst.h>
17 #include <net/xfrm.h>
18 #include <linux/veth.h>
19 
20 #define DRV_NAME	"veth"
21 #define DRV_VERSION	"1.0"
22 
23 struct veth_net_stats {
24 	unsigned long	rx_packets;
25 	unsigned long	tx_packets;
26 	unsigned long	rx_bytes;
27 	unsigned long	tx_bytes;
28 	unsigned long	tx_dropped;
29 };
30 
31 struct veth_priv {
32 	struct net_device *peer;
33 	struct net_device *dev;
34 	struct list_head list;
35 	struct veth_net_stats *stats;
36 	unsigned ip_summed;
37 };
38 
39 static LIST_HEAD(veth_list);
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 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 int 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;
156 	int length, cpu;
157 
158 	skb_orphan(skb);
159 
160 	priv = netdev_priv(dev);
161 	rcv = priv->peer;
162 	rcv_priv = netdev_priv(rcv);
163 
164 	cpu = smp_processor_id();
165 	stats = per_cpu_ptr(priv->stats, cpu);
166 
167 	if (!(rcv->flags & IFF_UP))
168 		goto outf;
169 
170 	skb->pkt_type = PACKET_HOST;
171 	skb->protocol = eth_type_trans(skb, rcv);
172 	if (dev->features & NETIF_F_NO_CSUM)
173 		skb->ip_summed = rcv_priv->ip_summed;
174 
175 	dst_release(skb->dst);
176 	skb->dst = NULL;
177 	skb->mark = 0;
178 	secpath_reset(skb);
179 	nf_reset(skb);
180 
181 	length = skb->len;
182 
183 	stats->tx_bytes += length;
184 	stats->tx_packets++;
185 
186 	stats = per_cpu_ptr(rcv_priv->stats, cpu);
187 	stats->rx_bytes += length;
188 	stats->rx_packets++;
189 
190 	netif_rx(skb);
191 	return 0;
192 
193 outf:
194 	kfree_skb(skb);
195 	stats->tx_dropped++;
196 	return 0;
197 }
198 
199 /*
200  * general routines
201  */
202 
203 static struct net_device_stats *veth_get_stats(struct net_device *dev)
204 {
205 	struct veth_priv *priv;
206 	struct net_device_stats *dev_stats;
207 	int cpu;
208 	struct veth_net_stats *stats;
209 
210 	priv = netdev_priv(dev);
211 	dev_stats = &dev->stats;
212 
213 	dev_stats->rx_packets = 0;
214 	dev_stats->tx_packets = 0;
215 	dev_stats->rx_bytes = 0;
216 	dev_stats->tx_bytes = 0;
217 	dev_stats->tx_dropped = 0;
218 
219 	for_each_online_cpu(cpu) {
220 		stats = per_cpu_ptr(priv->stats, cpu);
221 
222 		dev_stats->rx_packets += stats->rx_packets;
223 		dev_stats->tx_packets += stats->tx_packets;
224 		dev_stats->rx_bytes += stats->rx_bytes;
225 		dev_stats->tx_bytes += stats->tx_bytes;
226 		dev_stats->tx_dropped += stats->tx_dropped;
227 	}
228 
229 	return dev_stats;
230 }
231 
232 static int veth_open(struct net_device *dev)
233 {
234 	struct veth_priv *priv;
235 
236 	priv = netdev_priv(dev);
237 	if (priv->peer == NULL)
238 		return -ENOTCONN;
239 
240 	if (priv->peer->flags & IFF_UP) {
241 		netif_carrier_on(dev);
242 		netif_carrier_on(priv->peer);
243 	}
244 	return 0;
245 }
246 
247 static int veth_dev_init(struct net_device *dev)
248 {
249 	struct veth_net_stats *stats;
250 	struct veth_priv *priv;
251 
252 	stats = alloc_percpu(struct veth_net_stats);
253 	if (stats == NULL)
254 		return -ENOMEM;
255 
256 	priv = netdev_priv(dev);
257 	priv->stats = stats;
258 	return 0;
259 }
260 
261 static void veth_dev_free(struct net_device *dev)
262 {
263 	struct veth_priv *priv;
264 
265 	priv = netdev_priv(dev);
266 	free_percpu(priv->stats);
267 	free_netdev(dev);
268 }
269 
270 static void veth_setup(struct net_device *dev)
271 {
272 	ether_setup(dev);
273 
274 	dev->hard_start_xmit = veth_xmit;
275 	dev->get_stats = veth_get_stats;
276 	dev->open = veth_open;
277 	dev->ethtool_ops = &veth_ethtool_ops;
278 	dev->features |= NETIF_F_LLTX;
279 	dev->init = veth_dev_init;
280 	dev->destructor = veth_dev_free;
281 }
282 
283 static void veth_change_state(struct net_device *dev)
284 {
285 	struct net_device *peer;
286 	struct veth_priv *priv;
287 
288 	priv = netdev_priv(dev);
289 	peer = priv->peer;
290 
291 	if (netif_carrier_ok(peer)) {
292 		if (!netif_carrier_ok(dev))
293 			netif_carrier_on(dev);
294 	} else {
295 		if (netif_carrier_ok(dev))
296 			netif_carrier_off(dev);
297 	}
298 }
299 
300 static int veth_device_event(struct notifier_block *unused,
301 			     unsigned long event, void *ptr)
302 {
303 	struct net_device *dev = ptr;
304 
305 	if (dev->open != veth_open)
306 		goto out;
307 
308 	switch (event) {
309 	case NETDEV_CHANGE:
310 		veth_change_state(dev);
311 		break;
312 	}
313 out:
314 	return NOTIFY_DONE;
315 }
316 
317 static struct notifier_block veth_notifier_block __read_mostly = {
318 	.notifier_call	= veth_device_event,
319 };
320 
321 /*
322  * netlink interface
323  */
324 
325 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
326 {
327 	if (tb[IFLA_ADDRESS]) {
328 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
329 			return -EINVAL;
330 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
331 			return -EADDRNOTAVAIL;
332 	}
333 	return 0;
334 }
335 
336 static struct rtnl_link_ops veth_link_ops;
337 
338 static int veth_newlink(struct net_device *dev,
339 			 struct nlattr *tb[], struct nlattr *data[])
340 {
341 	int err;
342 	struct net_device *peer;
343 	struct veth_priv *priv;
344 	char ifname[IFNAMSIZ];
345 	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
346 
347 	/*
348 	 * create and register peer first
349 	 *
350 	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
351 	 * skip it since no info from it is useful yet
352 	 */
353 
354 	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
355 		struct nlattr *nla_peer;
356 
357 		nla_peer = data[VETH_INFO_PEER];
358 		err = nla_parse(peer_tb, IFLA_MAX,
359 				nla_data(nla_peer) + sizeof(struct ifinfomsg),
360 				nla_len(nla_peer) - sizeof(struct ifinfomsg),
361 				ifla_policy);
362 		if (err < 0)
363 			return err;
364 
365 		err = veth_validate(peer_tb, NULL);
366 		if (err < 0)
367 			return err;
368 
369 		tbp = peer_tb;
370 	} else
371 		tbp = tb;
372 
373 	if (tbp[IFLA_IFNAME])
374 		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
375 	else
376 		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
377 
378 	peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
379 	if (IS_ERR(peer))
380 		return PTR_ERR(peer);
381 
382 	if (tbp[IFLA_ADDRESS] == NULL)
383 		random_ether_addr(peer->dev_addr);
384 
385 	err = register_netdevice(peer);
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->dev = dev;
424 	priv->peer = peer;
425 	list_add(&priv->list, &veth_list);
426 
427 	priv = netdev_priv(peer);
428 	priv->dev = peer;
429 	priv->peer = dev;
430 	INIT_LIST_HEAD(&priv->list);
431 	return 0;
432 
433 err_register_dev:
434 	/* nothing to do */
435 err_alloc_name:
436 	unregister_netdevice(peer);
437 	return err;
438 
439 err_register_peer:
440 	free_netdev(peer);
441 	return err;
442 }
443 
444 static void veth_dellink(struct net_device *dev)
445 {
446 	struct veth_priv *priv;
447 	struct net_device *peer;
448 
449 	priv = netdev_priv(dev);
450 	peer = priv->peer;
451 
452 	if (!list_empty(&priv->list))
453 		list_del(&priv->list);
454 
455 	priv = netdev_priv(peer);
456 	if (!list_empty(&priv->list))
457 		list_del(&priv->list);
458 
459 	unregister_netdevice(dev);
460 	unregister_netdevice(peer);
461 }
462 
463 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
464 
465 static struct rtnl_link_ops veth_link_ops = {
466 	.kind		= DRV_NAME,
467 	.priv_size	= sizeof(struct veth_priv),
468 	.setup		= veth_setup,
469 	.validate	= veth_validate,
470 	.newlink	= veth_newlink,
471 	.dellink	= veth_dellink,
472 	.policy		= veth_policy,
473 	.maxtype	= VETH_INFO_MAX,
474 };
475 
476 /*
477  * init/fini
478  */
479 
480 static __init int veth_init(void)
481 {
482 	register_netdevice_notifier(&veth_notifier_block);
483 	return rtnl_link_register(&veth_link_ops);
484 }
485 
486 static __exit void veth_exit(void)
487 {
488 	rtnl_link_unregister(&veth_link_ops);
489 	unregister_netdevice_notifier(&veth_notifier_block);
490 }
491 
492 module_init(veth_init);
493 module_exit(veth_exit);
494 
495 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
496 MODULE_LICENSE("GPL v2");
497 MODULE_ALIAS_RTNL_LINK(DRV_NAME);
498