Commit | Line | Data |
---|---|---|
e314dbdc PE |
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 <net/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, ðtool_stats_keys, sizeof(ethtool_stats_keys)); | |
78 | break; | |
79 | } | |
80 | } | |
81 | ||
82 | static int veth_get_stats_count(struct net_device *dev) | |
83 | { | |
84 | return ARRAY_SIZE(ethtool_stats_keys); | |
85 | } | |
86 | ||
87 | static void veth_get_ethtool_stats(struct net_device *dev, | |
88 | struct ethtool_stats *stats, u64 *data) | |
89 | { | |
90 | struct veth_priv *priv; | |
91 | ||
92 | priv = netdev_priv(dev); | |
93 | data[0] = priv->peer->ifindex; | |
94 | } | |
95 | ||
96 | static u32 veth_get_rx_csum(struct net_device *dev) | |
97 | { | |
98 | struct veth_priv *priv; | |
99 | ||
100 | priv = netdev_priv(dev); | |
101 | return priv->ip_summed == CHECKSUM_UNNECESSARY; | |
102 | } | |
103 | ||
104 | static int veth_set_rx_csum(struct net_device *dev, u32 data) | |
105 | { | |
106 | struct veth_priv *priv; | |
107 | ||
108 | priv = netdev_priv(dev); | |
109 | priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE; | |
110 | return 0; | |
111 | } | |
112 | ||
113 | static u32 veth_get_tx_csum(struct net_device *dev) | |
114 | { | |
115 | return (dev->features & NETIF_F_NO_CSUM) != 0; | |
116 | } | |
117 | ||
118 | static int veth_set_tx_csum(struct net_device *dev, u32 data) | |
119 | { | |
120 | if (data) | |
121 | dev->features |= NETIF_F_NO_CSUM; | |
122 | else | |
123 | dev->features &= ~NETIF_F_NO_CSUM; | |
124 | return 0; | |
125 | } | |
126 | ||
127 | static struct ethtool_ops veth_ethtool_ops = { | |
128 | .get_settings = veth_get_settings, | |
129 | .get_drvinfo = veth_get_drvinfo, | |
130 | .get_link = ethtool_op_get_link, | |
131 | .get_rx_csum = veth_get_rx_csum, | |
132 | .set_rx_csum = veth_set_rx_csum, | |
133 | .get_tx_csum = veth_get_tx_csum, | |
134 | .set_tx_csum = veth_set_tx_csum, | |
135 | .get_sg = ethtool_op_get_sg, | |
136 | .set_sg = ethtool_op_set_sg, | |
137 | .get_strings = veth_get_strings, | |
138 | .get_stats_count = veth_get_stats_count, | |
139 | .get_ethtool_stats = veth_get_ethtool_stats, | |
140 | }; | |
141 | ||
142 | /* | |
143 | * xmit | |
144 | */ | |
145 | ||
146 | static int veth_xmit(struct sk_buff *skb, struct net_device *dev) | |
147 | { | |
148 | struct net_device *rcv = NULL; | |
149 | struct veth_priv *priv, *rcv_priv; | |
150 | struct veth_net_stats *stats; | |
151 | int length, cpu; | |
152 | ||
153 | skb_orphan(skb); | |
154 | ||
155 | priv = netdev_priv(dev); | |
156 | rcv = priv->peer; | |
157 | rcv_priv = netdev_priv(rcv); | |
158 | ||
159 | cpu = smp_processor_id(); | |
160 | stats = per_cpu_ptr(priv->stats, cpu); | |
161 | ||
162 | if (!(rcv->flags & IFF_UP)) | |
163 | goto outf; | |
164 | ||
165 | skb->pkt_type = PACKET_HOST; | |
166 | skb->protocol = eth_type_trans(skb, rcv); | |
167 | if (dev->features & NETIF_F_NO_CSUM) | |
168 | skb->ip_summed = rcv_priv->ip_summed; | |
169 | ||
170 | dst_release(skb->dst); | |
171 | skb->dst = NULL; | |
172 | skb->mark = 0; | |
173 | secpath_reset(skb); | |
174 | nf_reset(skb); | |
175 | ||
176 | length = skb->len; | |
177 | ||
178 | stats->tx_bytes += length; | |
179 | stats->tx_packets++; | |
180 | ||
181 | stats = per_cpu_ptr(rcv_priv->stats, cpu); | |
182 | stats->rx_bytes += length; | |
183 | stats->rx_packets++; | |
184 | ||
185 | netif_rx(skb); | |
186 | return 0; | |
187 | ||
188 | outf: | |
189 | kfree_skb(skb); | |
190 | stats->tx_dropped++; | |
191 | return 0; | |
192 | } | |
193 | ||
194 | /* | |
195 | * general routines | |
196 | */ | |
197 | ||
198 | static struct net_device_stats *veth_get_stats(struct net_device *dev) | |
199 | { | |
200 | struct veth_priv *priv; | |
201 | struct net_device_stats *dev_stats; | |
202 | int cpu; | |
203 | struct veth_net_stats *stats; | |
204 | ||
205 | priv = netdev_priv(dev); | |
206 | dev_stats = &dev->stats; | |
207 | ||
208 | dev_stats->rx_packets = 0; | |
209 | dev_stats->tx_packets = 0; | |
210 | dev_stats->rx_bytes = 0; | |
211 | dev_stats->tx_bytes = 0; | |
212 | dev_stats->tx_dropped = 0; | |
213 | ||
214 | for_each_online_cpu(cpu) { | |
215 | stats = per_cpu_ptr(priv->stats, cpu); | |
216 | ||
217 | dev_stats->rx_packets += stats->rx_packets; | |
218 | dev_stats->tx_packets += stats->tx_packets; | |
219 | dev_stats->rx_bytes += stats->rx_bytes; | |
220 | dev_stats->tx_bytes += stats->tx_bytes; | |
221 | dev_stats->tx_dropped += stats->tx_dropped; | |
222 | } | |
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; | |
245 | ||
246 | if (netif_carrier_ok(dev)) { | |
247 | priv = netdev_priv(dev); | |
248 | netif_carrier_off(dev); | |
249 | netif_carrier_off(priv->peer); | |
250 | } | |
251 | return 0; | |
252 | } | |
253 | ||
254 | static int veth_dev_init(struct net_device *dev) | |
255 | { | |
256 | struct veth_net_stats *stats; | |
257 | struct veth_priv *priv; | |
258 | ||
259 | stats = alloc_percpu(struct veth_net_stats); | |
260 | if (stats == NULL) | |
261 | return -ENOMEM; | |
262 | ||
263 | priv = netdev_priv(dev); | |
264 | priv->stats = stats; | |
265 | return 0; | |
266 | } | |
267 | ||
268 | static void veth_dev_free(struct net_device *dev) | |
269 | { | |
270 | struct veth_priv *priv; | |
271 | ||
272 | priv = netdev_priv(dev); | |
273 | free_percpu(priv->stats); | |
274 | free_netdev(dev); | |
275 | } | |
276 | ||
277 | static void veth_setup(struct net_device *dev) | |
278 | { | |
279 | ether_setup(dev); | |
280 | ||
281 | dev->hard_start_xmit = veth_xmit; | |
282 | dev->get_stats = veth_get_stats; | |
283 | dev->open = veth_open; | |
284 | dev->stop = veth_close; | |
285 | dev->ethtool_ops = &veth_ethtool_ops; | |
286 | dev->features |= NETIF_F_LLTX; | |
287 | dev->init = veth_dev_init; | |
288 | dev->destructor = veth_dev_free; | |
289 | } | |
290 | ||
291 | /* | |
292 | * netlink interface | |
293 | */ | |
294 | ||
295 | static int veth_validate(struct nlattr *tb[], struct nlattr *data[]) | |
296 | { | |
297 | if (tb[IFLA_ADDRESS]) { | |
298 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
299 | return -EINVAL; | |
300 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
301 | return -EADDRNOTAVAIL; | |
302 | } | |
303 | return 0; | |
304 | } | |
305 | ||
306 | static struct rtnl_link_ops veth_link_ops; | |
307 | ||
308 | static int veth_newlink(struct net_device *dev, | |
309 | struct nlattr *tb[], struct nlattr *data[]) | |
310 | { | |
311 | int err; | |
312 | struct net_device *peer; | |
313 | struct veth_priv *priv; | |
314 | char ifname[IFNAMSIZ]; | |
315 | struct nlattr *peer_tb[IFLA_MAX + 1], **tbp; | |
316 | ||
317 | /* | |
318 | * create and register peer first | |
319 | * | |
320 | * struct ifinfomsg is at the head of VETH_INFO_PEER, but we | |
321 | * skip it since no info from it is useful yet | |
322 | */ | |
323 | ||
324 | if (data != NULL && data[VETH_INFO_PEER] != NULL) { | |
325 | struct nlattr *nla_peer; | |
326 | ||
327 | nla_peer = data[VETH_INFO_PEER]; | |
328 | err = nla_parse(peer_tb, IFLA_MAX, | |
329 | nla_data(nla_peer) + sizeof(struct ifinfomsg), | |
330 | nla_len(nla_peer) - sizeof(struct ifinfomsg), | |
331 | ifla_policy); | |
332 | if (err < 0) | |
333 | return err; | |
334 | ||
335 | err = veth_validate(peer_tb, NULL); | |
336 | if (err < 0) | |
337 | return err; | |
338 | ||
339 | tbp = peer_tb; | |
340 | } else | |
341 | tbp = tb; | |
342 | ||
343 | if (tbp[IFLA_IFNAME]) | |
344 | nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ); | |
345 | else | |
346 | snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d"); | |
347 | ||
348 | peer = rtnl_create_link(ifname, &veth_link_ops, tbp); | |
349 | if (IS_ERR(peer)) | |
350 | return PTR_ERR(peer); | |
351 | ||
352 | if (tbp[IFLA_ADDRESS] == NULL) | |
353 | random_ether_addr(peer->dev_addr); | |
354 | ||
355 | err = register_netdevice(peer); | |
356 | if (err < 0) | |
357 | goto err_register_peer; | |
358 | ||
359 | netif_carrier_off(peer); | |
360 | ||
361 | /* | |
362 | * register dev last | |
363 | * | |
364 | * note, that since we've registered new device the dev's name | |
365 | * should be re-allocated | |
366 | */ | |
367 | ||
368 | if (tb[IFLA_ADDRESS] == NULL) | |
369 | random_ether_addr(dev->dev_addr); | |
370 | ||
371 | if (tb[IFLA_IFNAME]) | |
372 | nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ); | |
373 | else | |
374 | snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d"); | |
375 | ||
376 | if (strchr(dev->name, '%')) { | |
377 | err = dev_alloc_name(dev, dev->name); | |
378 | if (err < 0) | |
379 | goto err_alloc_name; | |
380 | } | |
381 | ||
382 | err = register_netdevice(dev); | |
383 | if (err < 0) | |
384 | goto err_register_dev; | |
385 | ||
386 | netif_carrier_off(dev); | |
387 | ||
388 | /* | |
389 | * tie the deviced together | |
390 | */ | |
391 | ||
392 | priv = netdev_priv(dev); | |
393 | priv->dev = dev; | |
394 | priv->peer = peer; | |
395 | list_add(&priv->list, &veth_list); | |
396 | ||
397 | priv = netdev_priv(peer); | |
398 | priv->dev = peer; | |
399 | priv->peer = dev; | |
400 | INIT_LIST_HEAD(&priv->list); | |
401 | return 0; | |
402 | ||
403 | err_register_dev: | |
404 | /* nothing to do */ | |
405 | err_alloc_name: | |
406 | unregister_netdevice(peer); | |
407 | return err; | |
408 | ||
409 | err_register_peer: | |
410 | free_netdev(peer); | |
411 | return err; | |
412 | } | |
413 | ||
414 | static void veth_dellink(struct net_device *dev) | |
415 | { | |
416 | struct veth_priv *priv; | |
417 | struct net_device *peer; | |
418 | ||
419 | priv = netdev_priv(dev); | |
420 | peer = priv->peer; | |
421 | ||
422 | if (!list_empty(&priv->list)) | |
423 | list_del(&priv->list); | |
424 | ||
425 | priv = netdev_priv(peer); | |
426 | if (!list_empty(&priv->list)) | |
427 | list_del(&priv->list); | |
428 | ||
429 | unregister_netdevice(dev); | |
430 | unregister_netdevice(peer); | |
431 | } | |
432 | ||
433 | static const struct nla_policy veth_policy[VETH_INFO_MAX + 1]; | |
434 | ||
435 | static struct rtnl_link_ops veth_link_ops = { | |
436 | .kind = DRV_NAME, | |
437 | .priv_size = sizeof(struct veth_priv), | |
438 | .setup = veth_setup, | |
439 | .validate = veth_validate, | |
440 | .newlink = veth_newlink, | |
441 | .dellink = veth_dellink, | |
442 | .policy = veth_policy, | |
443 | .maxtype = VETH_INFO_MAX, | |
444 | }; | |
445 | ||
446 | /* | |
447 | * init/fini | |
448 | */ | |
449 | ||
450 | static __init int veth_init(void) | |
451 | { | |
452 | return rtnl_link_register(&veth_link_ops); | |
453 | } | |
454 | ||
455 | static __exit void veth_exit(void) | |
456 | { | |
457 | struct veth_priv *priv, *next; | |
458 | ||
459 | rtnl_lock(); | |
460 | /* | |
461 | * cannot trust __rtnl_link_unregister() to unregister all | |
462 | * devices, as each ->dellink call will remove two devices | |
463 | * from the list at once. | |
464 | */ | |
465 | list_for_each_entry_safe(priv, next, &veth_list, list) | |
466 | veth_dellink(priv->dev); | |
467 | ||
468 | __rtnl_link_unregister(&veth_link_ops); | |
469 | rtnl_unlock(); | |
470 | } | |
471 | ||
472 | module_init(veth_init); | |
473 | module_exit(veth_exit); | |
474 | ||
475 | MODULE_DESCRIPTION("Virtual Ethernet Tunnel"); | |
476 | MODULE_LICENSE("GPL v2"); | |
477 | MODULE_ALIAS_RTNL_LINK(DRV_NAME); |