xref: /linux/drivers/net/veth.c (revision 5bdef865eb358b6f3760e25e591ae115e9eeddef)
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 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, *rcv_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 	rcv_stats = per_cpu_ptr(rcv_priv->stats, cpu);
167 
168 	if (!(rcv->flags & IFF_UP))
169 		goto tx_drop;
170 
171 	if (skb->len > (rcv->mtu + MTU_PAD))
172 		goto rx_drop;
173 
174 	skb->pkt_type = PACKET_HOST;
175 	skb->protocol = eth_type_trans(skb, rcv);
176 	if (dev->features & NETIF_F_NO_CSUM)
177 		skb->ip_summed = rcv_priv->ip_summed;
178 
179 	skb->mark = 0;
180 	secpath_reset(skb);
181 	nf_reset(skb);
182 
183 	length = skb->len;
184 
185 	stats->tx_bytes += length;
186 	stats->tx_packets++;
187 
188 	rcv_stats->rx_bytes += length;
189 	rcv_stats->rx_packets++;
190 
191 	netif_rx(skb);
192 	return 0;
193 
194 tx_drop:
195 	kfree_skb(skb);
196 	stats->tx_dropped++;
197 	return 0;
198 
199 rx_drop:
200 	kfree_skb(skb);
201 	rcv_stats->rx_dropped++;
202 	return 0;
203 }
204 
205 /*
206  * general routines
207  */
208 
209 static struct net_device_stats *veth_get_stats(struct net_device *dev)
210 {
211 	struct veth_priv *priv;
212 	struct net_device_stats *dev_stats;
213 	int cpu;
214 	struct veth_net_stats *stats;
215 
216 	priv = netdev_priv(dev);
217 	dev_stats = &dev->stats;
218 
219 	dev_stats->rx_packets = 0;
220 	dev_stats->tx_packets = 0;
221 	dev_stats->rx_bytes = 0;
222 	dev_stats->tx_bytes = 0;
223 	dev_stats->tx_dropped = 0;
224 	dev_stats->rx_dropped = 0;
225 
226 	for_each_online_cpu(cpu) {
227 		stats = per_cpu_ptr(priv->stats, cpu);
228 
229 		dev_stats->rx_packets += stats->rx_packets;
230 		dev_stats->tx_packets += stats->tx_packets;
231 		dev_stats->rx_bytes += stats->rx_bytes;
232 		dev_stats->tx_bytes += stats->tx_bytes;
233 		dev_stats->tx_dropped += stats->tx_dropped;
234 		dev_stats->rx_dropped += stats->rx_dropped;
235 	}
236 
237 	return dev_stats;
238 }
239 
240 static int veth_open(struct net_device *dev)
241 {
242 	struct veth_priv *priv;
243 
244 	priv = netdev_priv(dev);
245 	if (priv->peer == NULL)
246 		return -ENOTCONN;
247 
248 	if (priv->peer->flags & IFF_UP) {
249 		netif_carrier_on(dev);
250 		netif_carrier_on(priv->peer);
251 	}
252 	return 0;
253 }
254 
255 static int veth_close(struct net_device *dev)
256 {
257 	struct veth_priv *priv = netdev_priv(dev);
258 
259 	netif_carrier_off(dev);
260 	netif_carrier_off(priv->peer);
261 
262 	return 0;
263 }
264 
265 static int is_valid_veth_mtu(int new_mtu)
266 {
267 	return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
268 }
269 
270 static int veth_change_mtu(struct net_device *dev, int new_mtu)
271 {
272 	if (!is_valid_veth_mtu(new_mtu))
273 		return -EINVAL;
274 	dev->mtu = new_mtu;
275 	return 0;
276 }
277 
278 static int veth_dev_init(struct net_device *dev)
279 {
280 	struct veth_net_stats *stats;
281 	struct veth_priv *priv;
282 
283 	stats = alloc_percpu(struct veth_net_stats);
284 	if (stats == NULL)
285 		return -ENOMEM;
286 
287 	priv = netdev_priv(dev);
288 	priv->stats = stats;
289 	return 0;
290 }
291 
292 static void veth_dev_free(struct net_device *dev)
293 {
294 	struct veth_priv *priv;
295 
296 	priv = netdev_priv(dev);
297 	free_percpu(priv->stats);
298 	free_netdev(dev);
299 }
300 
301 static const struct net_device_ops veth_netdev_ops = {
302 	.ndo_init            = veth_dev_init,
303 	.ndo_open            = veth_open,
304 	.ndo_stop            = veth_close,
305 	.ndo_start_xmit      = veth_xmit,
306 	.ndo_change_mtu      = veth_change_mtu,
307 	.ndo_get_stats       = veth_get_stats,
308 	.ndo_set_mac_address = eth_mac_addr,
309 };
310 
311 static void veth_setup(struct net_device *dev)
312 {
313 	ether_setup(dev);
314 
315 	dev->netdev_ops = &veth_netdev_ops;
316 	dev->ethtool_ops = &veth_ethtool_ops;
317 	dev->features |= NETIF_F_LLTX;
318 	dev->destructor = veth_dev_free;
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 	if (tb[IFLA_MTU]) {
334 		if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
335 			return -EINVAL;
336 	}
337 	return 0;
338 }
339 
340 static struct rtnl_link_ops veth_link_ops;
341 
342 static int veth_newlink(struct net_device *dev,
343 			 struct nlattr *tb[], struct nlattr *data[])
344 {
345 	int err;
346 	struct net_device *peer;
347 	struct veth_priv *priv;
348 	char ifname[IFNAMSIZ];
349 	struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
350 
351 	/*
352 	 * create and register peer first
353 	 *
354 	 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
355 	 * skip it since no info from it is useful yet
356 	 */
357 
358 	if (data != NULL && data[VETH_INFO_PEER] != NULL) {
359 		struct nlattr *nla_peer;
360 
361 		nla_peer = data[VETH_INFO_PEER];
362 		err = nla_parse(peer_tb, IFLA_MAX,
363 				nla_data(nla_peer) + sizeof(struct ifinfomsg),
364 				nla_len(nla_peer) - sizeof(struct ifinfomsg),
365 				ifla_policy);
366 		if (err < 0)
367 			return err;
368 
369 		err = veth_validate(peer_tb, NULL);
370 		if (err < 0)
371 			return err;
372 
373 		tbp = peer_tb;
374 	} else
375 		tbp = tb;
376 
377 	if (tbp[IFLA_IFNAME])
378 		nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
379 	else
380 		snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
381 
382 	peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
383 	if (IS_ERR(peer))
384 		return PTR_ERR(peer);
385 
386 	if (tbp[IFLA_ADDRESS] == NULL)
387 		random_ether_addr(peer->dev_addr);
388 
389 	err = register_netdevice(peer);
390 	if (err < 0)
391 		goto err_register_peer;
392 
393 	netif_carrier_off(peer);
394 
395 	/*
396 	 * register dev last
397 	 *
398 	 * note, that since we've registered new device the dev's name
399 	 * should be re-allocated
400 	 */
401 
402 	if (tb[IFLA_ADDRESS] == NULL)
403 		random_ether_addr(dev->dev_addr);
404 
405 	if (tb[IFLA_IFNAME])
406 		nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
407 	else
408 		snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
409 
410 	if (strchr(dev->name, '%')) {
411 		err = dev_alloc_name(dev, dev->name);
412 		if (err < 0)
413 			goto err_alloc_name;
414 	}
415 
416 	err = register_netdevice(dev);
417 	if (err < 0)
418 		goto err_register_dev;
419 
420 	netif_carrier_off(dev);
421 
422 	/*
423 	 * tie the deviced together
424 	 */
425 
426 	priv = netdev_priv(dev);
427 	priv->peer = peer;
428 
429 	priv = netdev_priv(peer);
430 	priv->peer = dev;
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 	unregister_netdevice(dev);
453 	unregister_netdevice(peer);
454 }
455 
456 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
457 
458 static 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  * init/fini
471  */
472 
473 static __init int veth_init(void)
474 {
475 	return rtnl_link_register(&veth_link_ops);
476 }
477 
478 static __exit void veth_exit(void)
479 {
480 	rtnl_link_unregister(&veth_link_ops);
481 }
482 
483 module_init(veth_init);
484 module_exit(veth_exit);
485 
486 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
487 MODULE_LICENSE("GPL v2");
488 MODULE_ALIAS_RTNL_LINK(DRV_NAME);
489