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, ðtool_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 = netdev_priv(dev); 212 struct net_device_stats *dev_stats = &dev->stats; 213 unsigned int cpu; 214 struct veth_net_stats *stats; 215 216 dev_stats->rx_packets = 0; 217 dev_stats->tx_packets = 0; 218 dev_stats->rx_bytes = 0; 219 dev_stats->tx_bytes = 0; 220 dev_stats->tx_dropped = 0; 221 dev_stats->rx_dropped = 0; 222 223 if (priv->stats) 224 for_each_online_cpu(cpu) { 225 stats = per_cpu_ptr(priv->stats, cpu); 226 227 dev_stats->rx_packets += stats->rx_packets; 228 dev_stats->tx_packets += stats->tx_packets; 229 dev_stats->rx_bytes += stats->rx_bytes; 230 dev_stats->tx_bytes += stats->tx_bytes; 231 dev_stats->tx_dropped += stats->tx_dropped; 232 dev_stats->rx_dropped += stats->rx_dropped; 233 } 234 235 return dev_stats; 236 } 237 238 static int veth_open(struct net_device *dev) 239 { 240 struct veth_priv *priv; 241 242 priv = netdev_priv(dev); 243 if (priv->peer == NULL) 244 return -ENOTCONN; 245 246 if (priv->peer->flags & IFF_UP) { 247 netif_carrier_on(dev); 248 netif_carrier_on(priv->peer); 249 } 250 return 0; 251 } 252 253 static int veth_close(struct net_device *dev) 254 { 255 struct veth_priv *priv = netdev_priv(dev); 256 257 netif_carrier_off(dev); 258 netif_carrier_off(priv->peer); 259 260 free_percpu(priv->stats); 261 priv->stats = NULL; 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 const struct net_device_ops veth_netdev_ops = { 293 .ndo_init = veth_dev_init, 294 .ndo_open = veth_open, 295 .ndo_stop = veth_close, 296 .ndo_start_xmit = veth_xmit, 297 .ndo_change_mtu = veth_change_mtu, 298 .ndo_get_stats = veth_get_stats, 299 .ndo_set_mac_address = eth_mac_addr, 300 }; 301 302 static void veth_setup(struct net_device *dev) 303 { 304 ether_setup(dev); 305 306 dev->netdev_ops = &veth_netdev_ops; 307 dev->ethtool_ops = &veth_ethtool_ops; 308 dev->features |= NETIF_F_LLTX; 309 dev->destructor = free_netdev; 310 } 311 312 /* 313 * netlink interface 314 */ 315 316 static int veth_validate(struct nlattr *tb[], struct nlattr *data[]) 317 { 318 if (tb[IFLA_ADDRESS]) { 319 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 320 return -EINVAL; 321 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 322 return -EADDRNOTAVAIL; 323 } 324 if (tb[IFLA_MTU]) { 325 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU]))) 326 return -EINVAL; 327 } 328 return 0; 329 } 330 331 static struct rtnl_link_ops veth_link_ops; 332 333 static int veth_newlink(struct net_device *dev, 334 struct nlattr *tb[], struct nlattr *data[]) 335 { 336 int err; 337 struct net_device *peer; 338 struct veth_priv *priv; 339 char ifname[IFNAMSIZ]; 340 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp; 341 342 /* 343 * create and register peer first 344 * 345 * struct ifinfomsg is at the head of VETH_INFO_PEER, but we 346 * skip it since no info from it is useful yet 347 */ 348 349 if (data != NULL && data[VETH_INFO_PEER] != NULL) { 350 struct nlattr *nla_peer; 351 352 nla_peer = data[VETH_INFO_PEER]; 353 err = nla_parse(peer_tb, IFLA_MAX, 354 nla_data(nla_peer) + sizeof(struct ifinfomsg), 355 nla_len(nla_peer) - sizeof(struct ifinfomsg), 356 ifla_policy); 357 if (err < 0) 358 return err; 359 360 err = veth_validate(peer_tb, NULL); 361 if (err < 0) 362 return err; 363 364 tbp = peer_tb; 365 } else 366 tbp = tb; 367 368 if (tbp[IFLA_IFNAME]) 369 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ); 370 else 371 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d"); 372 373 peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp); 374 if (IS_ERR(peer)) 375 return PTR_ERR(peer); 376 377 if (tbp[IFLA_ADDRESS] == NULL) 378 random_ether_addr(peer->dev_addr); 379 380 err = register_netdevice(peer); 381 if (err < 0) 382 goto err_register_peer; 383 384 netif_carrier_off(peer); 385 386 /* 387 * register dev last 388 * 389 * note, that since we've registered new device the dev's name 390 * should be re-allocated 391 */ 392 393 if (tb[IFLA_ADDRESS] == NULL) 394 random_ether_addr(dev->dev_addr); 395 396 if (tb[IFLA_IFNAME]) 397 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ); 398 else 399 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d"); 400 401 if (strchr(dev->name, '%')) { 402 err = dev_alloc_name(dev, dev->name); 403 if (err < 0) 404 goto err_alloc_name; 405 } 406 407 err = register_netdevice(dev); 408 if (err < 0) 409 goto err_register_dev; 410 411 netif_carrier_off(dev); 412 413 /* 414 * tie the deviced together 415 */ 416 417 priv = netdev_priv(dev); 418 priv->peer = peer; 419 420 priv = netdev_priv(peer); 421 priv->peer = dev; 422 return 0; 423 424 err_register_dev: 425 /* nothing to do */ 426 err_alloc_name: 427 unregister_netdevice(peer); 428 return err; 429 430 err_register_peer: 431 free_netdev(peer); 432 return err; 433 } 434 435 static void veth_dellink(struct net_device *dev) 436 { 437 struct veth_priv *priv; 438 struct net_device *peer; 439 440 priv = netdev_priv(dev); 441 peer = priv->peer; 442 443 unregister_netdevice(dev); 444 unregister_netdevice(peer); 445 } 446 447 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1]; 448 449 static 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 */ 463 464 static __init int veth_init(void) 465 { 466 return rtnl_link_register(&veth_link_ops); 467 } 468 469 static __exit void veth_exit(void) 470 { 471 rtnl_link_unregister(&veth_link_ops); 472 } 473 474 module_init(veth_init); 475 module_exit(veth_exit); 476 477 MODULE_DESCRIPTION("Virtual Ethernet Tunnel"); 478 MODULE_LICENSE("GPL v2"); 479 MODULE_ALIAS_RTNL_LINK(DRV_NAME); 480