Loading...
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/slab.h>
13#include <linux/ethtool.h>
14#include <linux/etherdevice.h>
15#include <linux/u64_stats_sync.h>
16
17#include <net/rtnetlink.h>
18#include <net/dst.h>
19#include <net/xfrm.h>
20#include <linux/veth.h>
21#include <linux/module.h>
22
23#define DRV_NAME "veth"
24#define DRV_VERSION "1.0"
25
26#define MIN_MTU 68 /* Min L3 MTU */
27#define MAX_MTU 65535 /* Max L3 MTU (arbitrary) */
28
29struct pcpu_vstats {
30 u64 packets;
31 u64 bytes;
32 struct u64_stats_sync syncp;
33};
34
35struct veth_priv {
36 struct net_device __rcu *peer;
37 atomic64_t dropped;
38};
39
40/*
41 * ethtool interface
42 */
43
44static struct {
45 const char string[ETH_GSTRING_LEN];
46} ethtool_stats_keys[] = {
47 { "peer_ifindex" },
48};
49
50static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
51{
52 cmd->supported = 0;
53 cmd->advertising = 0;
54 ethtool_cmd_speed_set(cmd, SPEED_10000);
55 cmd->duplex = DUPLEX_FULL;
56 cmd->port = PORT_TP;
57 cmd->phy_address = 0;
58 cmd->transceiver = XCVR_INTERNAL;
59 cmd->autoneg = AUTONEG_DISABLE;
60 cmd->maxtxpkt = 0;
61 cmd->maxrxpkt = 0;
62 return 0;
63}
64
65static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
66{
67 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
68 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
69}
70
71static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
72{
73 switch(stringset) {
74 case ETH_SS_STATS:
75 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
76 break;
77 }
78}
79
80static int veth_get_sset_count(struct net_device *dev, int sset)
81{
82 switch (sset) {
83 case ETH_SS_STATS:
84 return ARRAY_SIZE(ethtool_stats_keys);
85 default:
86 return -EOPNOTSUPP;
87 }
88}
89
90static void veth_get_ethtool_stats(struct net_device *dev,
91 struct ethtool_stats *stats, u64 *data)
92{
93 struct veth_priv *priv = netdev_priv(dev);
94 struct net_device *peer = rtnl_dereference(priv->peer);
95
96 data[0] = peer ? peer->ifindex : 0;
97}
98
99static const struct ethtool_ops veth_ethtool_ops = {
100 .get_settings = veth_get_settings,
101 .get_drvinfo = veth_get_drvinfo,
102 .get_link = ethtool_op_get_link,
103 .get_strings = veth_get_strings,
104 .get_sset_count = veth_get_sset_count,
105 .get_ethtool_stats = veth_get_ethtool_stats,
106};
107
108static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
109{
110 struct veth_priv *priv = netdev_priv(dev);
111 struct net_device *rcv;
112 int length = skb->len;
113
114 rcu_read_lock();
115 rcv = rcu_dereference(priv->peer);
116 if (unlikely(!rcv)) {
117 kfree_skb(skb);
118 goto drop;
119 }
120 /* don't change ip_summed == CHECKSUM_PARTIAL, as that
121 * will cause bad checksum on forwarded packets
122 */
123 if (skb->ip_summed == CHECKSUM_NONE &&
124 rcv->features & NETIF_F_RXCSUM)
125 skb->ip_summed = CHECKSUM_UNNECESSARY;
126
127 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
128 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
129
130 u64_stats_update_begin(&stats->syncp);
131 stats->bytes += length;
132 stats->packets++;
133 u64_stats_update_end(&stats->syncp);
134 } else {
135drop:
136 atomic64_inc(&priv->dropped);
137 }
138 rcu_read_unlock();
139 return NETDEV_TX_OK;
140}
141
142/*
143 * general routines
144 */
145
146static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
147{
148 struct veth_priv *priv = netdev_priv(dev);
149 int cpu;
150
151 result->packets = 0;
152 result->bytes = 0;
153 for_each_possible_cpu(cpu) {
154 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
155 u64 packets, bytes;
156 unsigned int start;
157
158 do {
159 start = u64_stats_fetch_begin_irq(&stats->syncp);
160 packets = stats->packets;
161 bytes = stats->bytes;
162 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
163 result->packets += packets;
164 result->bytes += bytes;
165 }
166 return atomic64_read(&priv->dropped);
167}
168
169static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
170 struct rtnl_link_stats64 *tot)
171{
172 struct veth_priv *priv = netdev_priv(dev);
173 struct net_device *peer;
174 struct pcpu_vstats one;
175
176 tot->tx_dropped = veth_stats_one(&one, dev);
177 tot->tx_bytes = one.bytes;
178 tot->tx_packets = one.packets;
179
180 rcu_read_lock();
181 peer = rcu_dereference(priv->peer);
182 if (peer) {
183 tot->rx_dropped = veth_stats_one(&one, peer);
184 tot->rx_bytes = one.bytes;
185 tot->rx_packets = one.packets;
186 }
187 rcu_read_unlock();
188
189 return tot;
190}
191
192/* fake multicast ability */
193static void veth_set_multicast_list(struct net_device *dev)
194{
195}
196
197static int veth_open(struct net_device *dev)
198{
199 struct veth_priv *priv = netdev_priv(dev);
200 struct net_device *peer = rtnl_dereference(priv->peer);
201
202 if (!peer)
203 return -ENOTCONN;
204
205 if (peer->flags & IFF_UP) {
206 netif_carrier_on(dev);
207 netif_carrier_on(peer);
208 }
209 return 0;
210}
211
212static int veth_close(struct net_device *dev)
213{
214 struct veth_priv *priv = netdev_priv(dev);
215 struct net_device *peer = rtnl_dereference(priv->peer);
216
217 netif_carrier_off(dev);
218 if (peer)
219 netif_carrier_off(peer);
220
221 return 0;
222}
223
224static int is_valid_veth_mtu(int new_mtu)
225{
226 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
227}
228
229static int veth_change_mtu(struct net_device *dev, int new_mtu)
230{
231 if (!is_valid_veth_mtu(new_mtu))
232 return -EINVAL;
233 dev->mtu = new_mtu;
234 return 0;
235}
236
237static int veth_dev_init(struct net_device *dev)
238{
239 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
240 if (!dev->vstats)
241 return -ENOMEM;
242 return 0;
243}
244
245static void veth_dev_free(struct net_device *dev)
246{
247 free_percpu(dev->vstats);
248 free_netdev(dev);
249}
250
251static const struct net_device_ops veth_netdev_ops = {
252 .ndo_init = veth_dev_init,
253 .ndo_open = veth_open,
254 .ndo_stop = veth_close,
255 .ndo_start_xmit = veth_xmit,
256 .ndo_change_mtu = veth_change_mtu,
257 .ndo_get_stats64 = veth_get_stats64,
258 .ndo_set_rx_mode = veth_set_multicast_list,
259 .ndo_set_mac_address = eth_mac_addr,
260};
261
262#define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
263 NETIF_F_HW_CSUM | NETIF_F_RXCSUM | NETIF_F_HIGHDMA | \
264 NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL | \
265 NETIF_F_GSO_IPIP | NETIF_F_GSO_SIT | NETIF_F_UFO | \
266 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
267 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
268
269static void veth_setup(struct net_device *dev)
270{
271 ether_setup(dev);
272
273 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
274 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
275
276 dev->netdev_ops = &veth_netdev_ops;
277 dev->ethtool_ops = &veth_ethtool_ops;
278 dev->features |= NETIF_F_LLTX;
279 dev->features |= VETH_FEATURES;
280 dev->vlan_features = dev->features &
281 ~(NETIF_F_HW_VLAN_CTAG_TX |
282 NETIF_F_HW_VLAN_STAG_TX |
283 NETIF_F_HW_VLAN_CTAG_RX |
284 NETIF_F_HW_VLAN_STAG_RX);
285 dev->destructor = veth_dev_free;
286
287 dev->hw_features = VETH_FEATURES;
288 dev->hw_enc_features = VETH_FEATURES;
289}
290
291/*
292 * netlink interface
293 */
294
295static 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 if (tb[IFLA_MTU]) {
304 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
305 return -EINVAL;
306 }
307 return 0;
308}
309
310static struct rtnl_link_ops veth_link_ops;
311
312static int veth_newlink(struct net *src_net, struct net_device *dev,
313 struct nlattr *tb[], struct nlattr *data[])
314{
315 int err;
316 struct net_device *peer;
317 struct veth_priv *priv;
318 char ifname[IFNAMSIZ];
319 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
320 struct ifinfomsg *ifmp;
321 struct net *net;
322
323 /*
324 * create and register peer first
325 */
326 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
327 struct nlattr *nla_peer;
328
329 nla_peer = data[VETH_INFO_PEER];
330 ifmp = nla_data(nla_peer);
331 err = rtnl_nla_parse_ifla(peer_tb,
332 nla_data(nla_peer) + sizeof(struct ifinfomsg),
333 nla_len(nla_peer) - sizeof(struct ifinfomsg));
334 if (err < 0)
335 return err;
336
337 err = veth_validate(peer_tb, NULL);
338 if (err < 0)
339 return err;
340
341 tbp = peer_tb;
342 } else {
343 ifmp = NULL;
344 tbp = tb;
345 }
346
347 if (tbp[IFLA_IFNAME])
348 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
349 else
350 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
351
352 net = rtnl_link_get_net(src_net, tbp);
353 if (IS_ERR(net))
354 return PTR_ERR(net);
355
356 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp);
357 if (IS_ERR(peer)) {
358 put_net(net);
359 return PTR_ERR(peer);
360 }
361
362 if (tbp[IFLA_ADDRESS] == NULL)
363 eth_hw_addr_random(peer);
364
365 if (ifmp && (dev->ifindex != 0))
366 peer->ifindex = ifmp->ifi_index;
367
368 err = register_netdevice(peer);
369 put_net(net);
370 net = NULL;
371 if (err < 0)
372 goto err_register_peer;
373
374 netif_carrier_off(peer);
375
376 err = rtnl_configure_link(peer, ifmp);
377 if (err < 0)
378 goto err_configure_peer;
379
380 /*
381 * register dev last
382 *
383 * note, that since we've registered new device the dev's name
384 * should be re-allocated
385 */
386
387 if (tb[IFLA_ADDRESS] == NULL)
388 eth_hw_addr_random(dev);
389
390 if (tb[IFLA_IFNAME])
391 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
392 else
393 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
394
395 err = register_netdevice(dev);
396 if (err < 0)
397 goto err_register_dev;
398
399 netif_carrier_off(dev);
400
401 /*
402 * tie the deviced together
403 */
404
405 priv = netdev_priv(dev);
406 rcu_assign_pointer(priv->peer, peer);
407
408 priv = netdev_priv(peer);
409 rcu_assign_pointer(priv->peer, dev);
410 return 0;
411
412err_register_dev:
413 /* nothing to do */
414err_configure_peer:
415 unregister_netdevice(peer);
416 return err;
417
418err_register_peer:
419 free_netdev(peer);
420 return err;
421}
422
423static void veth_dellink(struct net_device *dev, struct list_head *head)
424{
425 struct veth_priv *priv;
426 struct net_device *peer;
427
428 priv = netdev_priv(dev);
429 peer = rtnl_dereference(priv->peer);
430
431 /* Note : dellink() is called from default_device_exit_batch(),
432 * before a rcu_synchronize() point. The devices are guaranteed
433 * not being freed before one RCU grace period.
434 */
435 RCU_INIT_POINTER(priv->peer, NULL);
436 unregister_netdevice_queue(dev, head);
437
438 if (peer) {
439 priv = netdev_priv(peer);
440 RCU_INIT_POINTER(priv->peer, NULL);
441 unregister_netdevice_queue(peer, head);
442 }
443}
444
445static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
446 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
447};
448
449static struct rtnl_link_ops veth_link_ops = {
450 .kind = DRV_NAME,
451 .priv_size = sizeof(struct veth_priv),
452 .setup = veth_setup,
453 .validate = veth_validate,
454 .newlink = veth_newlink,
455 .dellink = veth_dellink,
456 .policy = veth_policy,
457 .maxtype = VETH_INFO_MAX,
458};
459
460/*
461 * init/fini
462 */
463
464static __init int veth_init(void)
465{
466 return rtnl_link_register(&veth_link_ops);
467}
468
469static __exit void veth_exit(void)
470{
471 rtnl_link_unregister(&veth_link_ops);
472}
473
474module_init(veth_init);
475module_exit(veth_exit);
476
477MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
478MODULE_LICENSE("GPL v2");
479MODULE_ALIAS_RTNL_LINK(DRV_NAME);
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/slab.h>
13#include <linux/ethtool.h>
14#include <linux/etherdevice.h>
15#include <linux/u64_stats_sync.h>
16
17#include <net/rtnetlink.h>
18#include <net/dst.h>
19#include <net/xfrm.h>
20#include <linux/veth.h>
21#include <linux/module.h>
22
23#define DRV_NAME "veth"
24#define DRV_VERSION "1.0"
25
26struct pcpu_vstats {
27 u64 packets;
28 u64 bytes;
29 struct u64_stats_sync syncp;
30};
31
32struct veth_priv {
33 struct net_device __rcu *peer;
34 atomic64_t dropped;
35 unsigned requested_headroom;
36};
37
38/*
39 * ethtool interface
40 */
41
42static struct {
43 const char string[ETH_GSTRING_LEN];
44} ethtool_stats_keys[] = {
45 { "peer_ifindex" },
46};
47
48static int veth_get_link_ksettings(struct net_device *dev,
49 struct ethtool_link_ksettings *cmd)
50{
51 cmd->base.speed = SPEED_10000;
52 cmd->base.duplex = DUPLEX_FULL;
53 cmd->base.port = PORT_TP;
54 cmd->base.autoneg = AUTONEG_DISABLE;
55 return 0;
56}
57
58static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
59{
60 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
61 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
62}
63
64static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
65{
66 switch(stringset) {
67 case ETH_SS_STATS:
68 memcpy(buf, ðtool_stats_keys, sizeof(ethtool_stats_keys));
69 break;
70 }
71}
72
73static int veth_get_sset_count(struct net_device *dev, int sset)
74{
75 switch (sset) {
76 case ETH_SS_STATS:
77 return ARRAY_SIZE(ethtool_stats_keys);
78 default:
79 return -EOPNOTSUPP;
80 }
81}
82
83static void veth_get_ethtool_stats(struct net_device *dev,
84 struct ethtool_stats *stats, u64 *data)
85{
86 struct veth_priv *priv = netdev_priv(dev);
87 struct net_device *peer = rtnl_dereference(priv->peer);
88
89 data[0] = peer ? peer->ifindex : 0;
90}
91
92static const struct ethtool_ops veth_ethtool_ops = {
93 .get_drvinfo = veth_get_drvinfo,
94 .get_link = ethtool_op_get_link,
95 .get_strings = veth_get_strings,
96 .get_sset_count = veth_get_sset_count,
97 .get_ethtool_stats = veth_get_ethtool_stats,
98 .get_link_ksettings = veth_get_link_ksettings,
99};
100
101static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
102{
103 struct veth_priv *priv = netdev_priv(dev);
104 struct net_device *rcv;
105 int length = skb->len;
106
107 rcu_read_lock();
108 rcv = rcu_dereference(priv->peer);
109 if (unlikely(!rcv)) {
110 kfree_skb(skb);
111 goto drop;
112 }
113
114 if (likely(dev_forward_skb(rcv, skb) == NET_RX_SUCCESS)) {
115 struct pcpu_vstats *stats = this_cpu_ptr(dev->vstats);
116
117 u64_stats_update_begin(&stats->syncp);
118 stats->bytes += length;
119 stats->packets++;
120 u64_stats_update_end(&stats->syncp);
121 } else {
122drop:
123 atomic64_inc(&priv->dropped);
124 }
125 rcu_read_unlock();
126 return NETDEV_TX_OK;
127}
128
129/*
130 * general routines
131 */
132
133static u64 veth_stats_one(struct pcpu_vstats *result, struct net_device *dev)
134{
135 struct veth_priv *priv = netdev_priv(dev);
136 int cpu;
137
138 result->packets = 0;
139 result->bytes = 0;
140 for_each_possible_cpu(cpu) {
141 struct pcpu_vstats *stats = per_cpu_ptr(dev->vstats, cpu);
142 u64 packets, bytes;
143 unsigned int start;
144
145 do {
146 start = u64_stats_fetch_begin_irq(&stats->syncp);
147 packets = stats->packets;
148 bytes = stats->bytes;
149 } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
150 result->packets += packets;
151 result->bytes += bytes;
152 }
153 return atomic64_read(&priv->dropped);
154}
155
156static void veth_get_stats64(struct net_device *dev,
157 struct rtnl_link_stats64 *tot)
158{
159 struct veth_priv *priv = netdev_priv(dev);
160 struct net_device *peer;
161 struct pcpu_vstats one;
162
163 tot->tx_dropped = veth_stats_one(&one, dev);
164 tot->tx_bytes = one.bytes;
165 tot->tx_packets = one.packets;
166
167 rcu_read_lock();
168 peer = rcu_dereference(priv->peer);
169 if (peer) {
170 tot->rx_dropped = veth_stats_one(&one, peer);
171 tot->rx_bytes = one.bytes;
172 tot->rx_packets = one.packets;
173 }
174 rcu_read_unlock();
175}
176
177/* fake multicast ability */
178static void veth_set_multicast_list(struct net_device *dev)
179{
180}
181
182static int veth_open(struct net_device *dev)
183{
184 struct veth_priv *priv = netdev_priv(dev);
185 struct net_device *peer = rtnl_dereference(priv->peer);
186
187 if (!peer)
188 return -ENOTCONN;
189
190 if (peer->flags & IFF_UP) {
191 netif_carrier_on(dev);
192 netif_carrier_on(peer);
193 }
194 return 0;
195}
196
197static int veth_close(struct net_device *dev)
198{
199 struct veth_priv *priv = netdev_priv(dev);
200 struct net_device *peer = rtnl_dereference(priv->peer);
201
202 netif_carrier_off(dev);
203 if (peer)
204 netif_carrier_off(peer);
205
206 return 0;
207}
208
209static int is_valid_veth_mtu(int mtu)
210{
211 return mtu >= ETH_MIN_MTU && mtu <= ETH_MAX_MTU;
212}
213
214static int veth_dev_init(struct net_device *dev)
215{
216 dev->vstats = netdev_alloc_pcpu_stats(struct pcpu_vstats);
217 if (!dev->vstats)
218 return -ENOMEM;
219 return 0;
220}
221
222static void veth_dev_free(struct net_device *dev)
223{
224 free_percpu(dev->vstats);
225}
226
227#ifdef CONFIG_NET_POLL_CONTROLLER
228static void veth_poll_controller(struct net_device *dev)
229{
230 /* veth only receives frames when its peer sends one
231 * Since it's a synchronous operation, we are guaranteed
232 * never to have pending data when we poll for it so
233 * there is nothing to do here.
234 *
235 * We need this though so netpoll recognizes us as an interface that
236 * supports polling, which enables bridge devices in virt setups to
237 * still use netconsole
238 */
239}
240#endif /* CONFIG_NET_POLL_CONTROLLER */
241
242static int veth_get_iflink(const struct net_device *dev)
243{
244 struct veth_priv *priv = netdev_priv(dev);
245 struct net_device *peer;
246 int iflink;
247
248 rcu_read_lock();
249 peer = rcu_dereference(priv->peer);
250 iflink = peer ? peer->ifindex : 0;
251 rcu_read_unlock();
252
253 return iflink;
254}
255
256static void veth_set_rx_headroom(struct net_device *dev, int new_hr)
257{
258 struct veth_priv *peer_priv, *priv = netdev_priv(dev);
259 struct net_device *peer;
260
261 if (new_hr < 0)
262 new_hr = 0;
263
264 rcu_read_lock();
265 peer = rcu_dereference(priv->peer);
266 if (unlikely(!peer))
267 goto out;
268
269 peer_priv = netdev_priv(peer);
270 priv->requested_headroom = new_hr;
271 new_hr = max(priv->requested_headroom, peer_priv->requested_headroom);
272 dev->needed_headroom = new_hr;
273 peer->needed_headroom = new_hr;
274
275out:
276 rcu_read_unlock();
277}
278
279static const struct net_device_ops veth_netdev_ops = {
280 .ndo_init = veth_dev_init,
281 .ndo_open = veth_open,
282 .ndo_stop = veth_close,
283 .ndo_start_xmit = veth_xmit,
284 .ndo_get_stats64 = veth_get_stats64,
285 .ndo_set_rx_mode = veth_set_multicast_list,
286 .ndo_set_mac_address = eth_mac_addr,
287#ifdef CONFIG_NET_POLL_CONTROLLER
288 .ndo_poll_controller = veth_poll_controller,
289#endif
290 .ndo_get_iflink = veth_get_iflink,
291 .ndo_features_check = passthru_features_check,
292 .ndo_set_rx_headroom = veth_set_rx_headroom,
293};
294
295#define VETH_FEATURES (NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HW_CSUM | \
296 NETIF_F_RXCSUM | NETIF_F_SCTP_CRC | NETIF_F_HIGHDMA | \
297 NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL | \
298 NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX | \
299 NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_STAG_RX )
300
301static void veth_setup(struct net_device *dev)
302{
303 ether_setup(dev);
304
305 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
306 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
307 dev->priv_flags |= IFF_NO_QUEUE;
308 dev->priv_flags |= IFF_PHONY_HEADROOM;
309
310 dev->netdev_ops = &veth_netdev_ops;
311 dev->ethtool_ops = &veth_ethtool_ops;
312 dev->features |= NETIF_F_LLTX;
313 dev->features |= VETH_FEATURES;
314 dev->vlan_features = dev->features &
315 ~(NETIF_F_HW_VLAN_CTAG_TX |
316 NETIF_F_HW_VLAN_STAG_TX |
317 NETIF_F_HW_VLAN_CTAG_RX |
318 NETIF_F_HW_VLAN_STAG_RX);
319 dev->needs_free_netdev = true;
320 dev->priv_destructor = veth_dev_free;
321 dev->max_mtu = ETH_MAX_MTU;
322
323 dev->hw_features = VETH_FEATURES;
324 dev->hw_enc_features = VETH_FEATURES;
325 dev->mpls_features = NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
326}
327
328/*
329 * netlink interface
330 */
331
332static int veth_validate(struct nlattr *tb[], struct nlattr *data[],
333 struct netlink_ext_ack *extack)
334{
335 if (tb[IFLA_ADDRESS]) {
336 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
337 return -EINVAL;
338 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
339 return -EADDRNOTAVAIL;
340 }
341 if (tb[IFLA_MTU]) {
342 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
343 return -EINVAL;
344 }
345 return 0;
346}
347
348static struct rtnl_link_ops veth_link_ops;
349
350static int veth_newlink(struct net *src_net, struct net_device *dev,
351 struct nlattr *tb[], struct nlattr *data[],
352 struct netlink_ext_ack *extack)
353{
354 int err;
355 struct net_device *peer;
356 struct veth_priv *priv;
357 char ifname[IFNAMSIZ];
358 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
359 unsigned char name_assign_type;
360 struct ifinfomsg *ifmp;
361 struct net *net;
362
363 /*
364 * create and register peer first
365 */
366 if (data != NULL && data[VETH_INFO_PEER] != NULL) {
367 struct nlattr *nla_peer;
368
369 nla_peer = data[VETH_INFO_PEER];
370 ifmp = nla_data(nla_peer);
371 err = rtnl_nla_parse_ifla(peer_tb,
372 nla_data(nla_peer) + sizeof(struct ifinfomsg),
373 nla_len(nla_peer) - sizeof(struct ifinfomsg),
374 NULL);
375 if (err < 0)
376 return err;
377
378 err = veth_validate(peer_tb, NULL, extack);
379 if (err < 0)
380 return err;
381
382 tbp = peer_tb;
383 } else {
384 ifmp = NULL;
385 tbp = tb;
386 }
387
388 if (ifmp && tbp[IFLA_IFNAME]) {
389 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
390 name_assign_type = NET_NAME_USER;
391 } else {
392 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
393 name_assign_type = NET_NAME_ENUM;
394 }
395
396 net = rtnl_link_get_net(src_net, tbp);
397 if (IS_ERR(net))
398 return PTR_ERR(net);
399
400 peer = rtnl_create_link(net, ifname, name_assign_type,
401 &veth_link_ops, tbp);
402 if (IS_ERR(peer)) {
403 put_net(net);
404 return PTR_ERR(peer);
405 }
406
407 if (!ifmp || !tbp[IFLA_ADDRESS])
408 eth_hw_addr_random(peer);
409
410 if (ifmp && (dev->ifindex != 0))
411 peer->ifindex = ifmp->ifi_index;
412
413 peer->gso_max_size = dev->gso_max_size;
414 peer->gso_max_segs = dev->gso_max_segs;
415
416 err = register_netdevice(peer);
417 put_net(net);
418 net = NULL;
419 if (err < 0)
420 goto err_register_peer;
421
422 netif_carrier_off(peer);
423
424 err = rtnl_configure_link(peer, ifmp);
425 if (err < 0)
426 goto err_configure_peer;
427
428 /*
429 * register dev last
430 *
431 * note, that since we've registered new device the dev's name
432 * should be re-allocated
433 */
434
435 if (tb[IFLA_ADDRESS] == NULL)
436 eth_hw_addr_random(dev);
437
438 if (tb[IFLA_IFNAME])
439 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
440 else
441 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
442
443 err = register_netdevice(dev);
444 if (err < 0)
445 goto err_register_dev;
446
447 netif_carrier_off(dev);
448
449 /*
450 * tie the deviced together
451 */
452
453 priv = netdev_priv(dev);
454 rcu_assign_pointer(priv->peer, peer);
455
456 priv = netdev_priv(peer);
457 rcu_assign_pointer(priv->peer, dev);
458 return 0;
459
460err_register_dev:
461 /* nothing to do */
462err_configure_peer:
463 unregister_netdevice(peer);
464 return err;
465
466err_register_peer:
467 free_netdev(peer);
468 return err;
469}
470
471static void veth_dellink(struct net_device *dev, struct list_head *head)
472{
473 struct veth_priv *priv;
474 struct net_device *peer;
475
476 priv = netdev_priv(dev);
477 peer = rtnl_dereference(priv->peer);
478
479 /* Note : dellink() is called from default_device_exit_batch(),
480 * before a rcu_synchronize() point. The devices are guaranteed
481 * not being freed before one RCU grace period.
482 */
483 RCU_INIT_POINTER(priv->peer, NULL);
484 unregister_netdevice_queue(dev, head);
485
486 if (peer) {
487 priv = netdev_priv(peer);
488 RCU_INIT_POINTER(priv->peer, NULL);
489 unregister_netdevice_queue(peer, head);
490 }
491}
492
493static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
494 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) },
495};
496
497static struct net *veth_get_link_net(const struct net_device *dev)
498{
499 struct veth_priv *priv = netdev_priv(dev);
500 struct net_device *peer = rtnl_dereference(priv->peer);
501
502 return peer ? dev_net(peer) : dev_net(dev);
503}
504
505static struct rtnl_link_ops veth_link_ops = {
506 .kind = DRV_NAME,
507 .priv_size = sizeof(struct veth_priv),
508 .setup = veth_setup,
509 .validate = veth_validate,
510 .newlink = veth_newlink,
511 .dellink = veth_dellink,
512 .policy = veth_policy,
513 .maxtype = VETH_INFO_MAX,
514 .get_link_net = veth_get_link_net,
515};
516
517/*
518 * init/fini
519 */
520
521static __init int veth_init(void)
522{
523 return rtnl_link_register(&veth_link_ops);
524}
525
526static __exit void veth_exit(void)
527{
528 rtnl_link_unregister(&veth_link_ops);
529}
530
531module_init(veth_init);
532module_exit(veth_exit);
533
534MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
535MODULE_LICENSE("GPL v2");
536MODULE_ALIAS_RTNL_LINK(DRV_NAME);