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/dst.h> 18 #include <net/xfrm.h> 19 #include <linux/veth.h> 20 #include <linux/module.h> 21 22 #define DRV_NAME "veth" 23 #define DRV_VERSION "1.0" 24 25 #define MIN_MTU 68 /* Min L3 MTU */ 26 #define MAX_MTU 65535 /* Max L3 MTU (arbitrary) */ 27 28 struct veth_net_stats { 29 u64 rx_packets; 30 u64 rx_bytes; 31 u64 tx_packets; 32 u64 tx_bytes; 33 u64 rx_dropped; 34 struct u64_stats_sync syncp; 35 }; 36 37 struct veth_priv { 38 struct net_device *peer; 39 struct veth_net_stats __percpu *stats; 40 }; 41 42 /* 43 * ethtool interface 44 */ 45 46 static struct { 47 const char string[ETH_GSTRING_LEN]; 48 } ethtool_stats_keys[] = { 49 { "peer_ifindex" }, 50 }; 51 52 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) 53 { 54 cmd->supported = 0; 55 cmd->advertising = 0; 56 ethtool_cmd_speed_set(cmd, SPEED_10000); 57 cmd->duplex = DUPLEX_FULL; 58 cmd->port = PORT_TP; 59 cmd->phy_address = 0; 60 cmd->transceiver = XCVR_INTERNAL; 61 cmd->autoneg = AUTONEG_DISABLE; 62 cmd->maxtxpkt = 0; 63 cmd->maxrxpkt = 0; 64 return 0; 65 } 66 67 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) 68 { 69 strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); 70 strlcpy(info->version, DRV_VERSION, sizeof(info->version)); 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 const struct ethtool_ops veth_ethtool_ops = { 102 .get_settings = veth_get_settings, 103 .get_drvinfo = veth_get_drvinfo, 104 .get_link = ethtool_op_get_link, 105 .get_strings = veth_get_strings, 106 .get_sset_count = veth_get_sset_count, 107 .get_ethtool_stats = veth_get_ethtool_stats, 108 }; 109 110 /* 111 * xmit 112 */ 113 114 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev) 115 { 116 struct net_device *rcv = NULL; 117 struct veth_priv *priv, *rcv_priv; 118 struct veth_net_stats *stats, *rcv_stats; 119 int length; 120 121 priv = netdev_priv(dev); 122 rcv = priv->peer; 123 rcv_priv = netdev_priv(rcv); 124 125 stats = this_cpu_ptr(priv->stats); 126 rcv_stats = this_cpu_ptr(rcv_priv->stats); 127 128 /* don't change ip_summed == CHECKSUM_PARTIAL, as that 129 will cause bad checksum on forwarded packets */ 130 if (skb->ip_summed == CHECKSUM_NONE && 131 rcv->features & NETIF_F_RXCSUM) 132 skb->ip_summed = CHECKSUM_UNNECESSARY; 133 134 length = skb->len; 135 if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS) 136 goto rx_drop; 137 138 u64_stats_update_begin(&stats->syncp); 139 stats->tx_bytes += length; 140 stats->tx_packets++; 141 u64_stats_update_end(&stats->syncp); 142 143 u64_stats_update_begin(&rcv_stats->syncp); 144 rcv_stats->rx_bytes += length; 145 rcv_stats->rx_packets++; 146 u64_stats_update_end(&rcv_stats->syncp); 147 148 return NETDEV_TX_OK; 149 150 rx_drop: 151 u64_stats_update_begin(&rcv_stats->syncp); 152 rcv_stats->rx_dropped++; 153 u64_stats_update_end(&rcv_stats->syncp); 154 return NETDEV_TX_OK; 155 } 156 157 /* 158 * general routines 159 */ 160 161 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev, 162 struct rtnl_link_stats64 *tot) 163 { 164 struct veth_priv *priv = netdev_priv(dev); 165 int cpu; 166 167 for_each_possible_cpu(cpu) { 168 struct veth_net_stats *stats = per_cpu_ptr(priv->stats, cpu); 169 u64 rx_packets, rx_bytes, rx_dropped; 170 u64 tx_packets, tx_bytes; 171 unsigned int start; 172 173 do { 174 start = u64_stats_fetch_begin_bh(&stats->syncp); 175 rx_packets = stats->rx_packets; 176 tx_packets = stats->tx_packets; 177 rx_bytes = stats->rx_bytes; 178 tx_bytes = stats->tx_bytes; 179 rx_dropped = stats->rx_dropped; 180 } while (u64_stats_fetch_retry_bh(&stats->syncp, start)); 181 tot->rx_packets += rx_packets; 182 tot->tx_packets += tx_packets; 183 tot->rx_bytes += rx_bytes; 184 tot->tx_bytes += tx_bytes; 185 tot->rx_dropped += rx_dropped; 186 } 187 188 return tot; 189 } 190 191 static int veth_open(struct net_device *dev) 192 { 193 struct veth_priv *priv; 194 195 priv = netdev_priv(dev); 196 if (priv->peer == NULL) 197 return -ENOTCONN; 198 199 if (priv->peer->flags & IFF_UP) { 200 netif_carrier_on(dev); 201 netif_carrier_on(priv->peer); 202 } 203 return 0; 204 } 205 206 static int veth_close(struct net_device *dev) 207 { 208 struct veth_priv *priv = netdev_priv(dev); 209 210 netif_carrier_off(dev); 211 netif_carrier_off(priv->peer); 212 213 return 0; 214 } 215 216 static int is_valid_veth_mtu(int new_mtu) 217 { 218 return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU; 219 } 220 221 static int veth_change_mtu(struct net_device *dev, int new_mtu) 222 { 223 if (!is_valid_veth_mtu(new_mtu)) 224 return -EINVAL; 225 dev->mtu = new_mtu; 226 return 0; 227 } 228 229 static int veth_dev_init(struct net_device *dev) 230 { 231 struct veth_net_stats __percpu *stats; 232 struct veth_priv *priv; 233 234 stats = alloc_percpu(struct veth_net_stats); 235 if (stats == NULL) 236 return -ENOMEM; 237 238 priv = netdev_priv(dev); 239 priv->stats = stats; 240 return 0; 241 } 242 243 static void veth_dev_free(struct net_device *dev) 244 { 245 struct veth_priv *priv; 246 247 priv = netdev_priv(dev); 248 free_percpu(priv->stats); 249 free_netdev(dev); 250 } 251 252 static const struct net_device_ops veth_netdev_ops = { 253 .ndo_init = veth_dev_init, 254 .ndo_open = veth_open, 255 .ndo_stop = veth_close, 256 .ndo_start_xmit = veth_xmit, 257 .ndo_change_mtu = veth_change_mtu, 258 .ndo_get_stats64 = veth_get_stats64, 259 .ndo_set_mac_address = eth_mac_addr, 260 }; 261 262 static void veth_setup(struct net_device *dev) 263 { 264 ether_setup(dev); 265 266 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 267 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 268 269 dev->netdev_ops = &veth_netdev_ops; 270 dev->ethtool_ops = &veth_ethtool_ops; 271 dev->features |= NETIF_F_LLTX; 272 dev->destructor = veth_dev_free; 273 274 dev->hw_features = NETIF_F_HW_CSUM | NETIF_F_SG | NETIF_F_RXCSUM; 275 } 276 277 /* 278 * netlink interface 279 */ 280 281 static int veth_validate(struct nlattr *tb[], struct nlattr *data[]) 282 { 283 if (tb[IFLA_ADDRESS]) { 284 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 285 return -EINVAL; 286 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 287 return -EADDRNOTAVAIL; 288 } 289 if (tb[IFLA_MTU]) { 290 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU]))) 291 return -EINVAL; 292 } 293 return 0; 294 } 295 296 static struct rtnl_link_ops veth_link_ops; 297 298 static int veth_newlink(struct net *src_net, struct net_device *dev, 299 struct nlattr *tb[], struct nlattr *data[]) 300 { 301 int err; 302 struct net_device *peer; 303 struct veth_priv *priv; 304 char ifname[IFNAMSIZ]; 305 struct nlattr *peer_tb[IFLA_MAX + 1], **tbp; 306 struct ifinfomsg *ifmp; 307 struct net *net; 308 309 /* 310 * create and register peer first 311 */ 312 if (data != NULL && data[VETH_INFO_PEER] != NULL) { 313 struct nlattr *nla_peer; 314 315 nla_peer = data[VETH_INFO_PEER]; 316 ifmp = nla_data(nla_peer); 317 err = nla_parse(peer_tb, IFLA_MAX, 318 nla_data(nla_peer) + sizeof(struct ifinfomsg), 319 nla_len(nla_peer) - sizeof(struct ifinfomsg), 320 ifla_policy); 321 if (err < 0) 322 return err; 323 324 err = veth_validate(peer_tb, NULL); 325 if (err < 0) 326 return err; 327 328 tbp = peer_tb; 329 } else { 330 ifmp = NULL; 331 tbp = tb; 332 } 333 334 if (tbp[IFLA_IFNAME]) 335 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ); 336 else 337 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d"); 338 339 net = rtnl_link_get_net(src_net, tbp); 340 if (IS_ERR(net)) 341 return PTR_ERR(net); 342 343 peer = rtnl_create_link(net, ifname, &veth_link_ops, tbp); 344 if (IS_ERR(peer)) { 345 put_net(net); 346 return PTR_ERR(peer); 347 } 348 349 if (tbp[IFLA_ADDRESS] == NULL) 350 eth_hw_addr_random(peer); 351 352 if (ifmp && (dev->ifindex != 0)) 353 peer->ifindex = ifmp->ifi_index; 354 355 err = register_netdevice(peer); 356 put_net(net); 357 net = NULL; 358 if (err < 0) 359 goto err_register_peer; 360 361 netif_carrier_off(peer); 362 363 err = rtnl_configure_link(peer, ifmp); 364 if (err < 0) 365 goto err_configure_peer; 366 367 /* 368 * register dev last 369 * 370 * note, that since we've registered new device the dev's name 371 * should be re-allocated 372 */ 373 374 if (tb[IFLA_ADDRESS] == NULL) 375 eth_hw_addr_random(dev); 376 377 if (tb[IFLA_IFNAME]) 378 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ); 379 else 380 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d"); 381 382 if (strchr(dev->name, '%')) { 383 err = dev_alloc_name(dev, dev->name); 384 if (err < 0) 385 goto err_alloc_name; 386 } 387 388 err = register_netdevice(dev); 389 if (err < 0) 390 goto err_register_dev; 391 392 netif_carrier_off(dev); 393 394 /* 395 * tie the deviced together 396 */ 397 398 priv = netdev_priv(dev); 399 priv->peer = peer; 400 401 priv = netdev_priv(peer); 402 priv->peer = dev; 403 return 0; 404 405 err_register_dev: 406 /* nothing to do */ 407 err_alloc_name: 408 err_configure_peer: 409 unregister_netdevice(peer); 410 return err; 411 412 err_register_peer: 413 free_netdev(peer); 414 return err; 415 } 416 417 static void veth_dellink(struct net_device *dev, struct list_head *head) 418 { 419 struct veth_priv *priv; 420 struct net_device *peer; 421 422 priv = netdev_priv(dev); 423 peer = priv->peer; 424 425 unregister_netdevice_queue(dev, head); 426 unregister_netdevice_queue(peer, head); 427 } 428 429 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = { 430 [VETH_INFO_PEER] = { .len = sizeof(struct ifinfomsg) }, 431 }; 432 433 static struct rtnl_link_ops veth_link_ops = { 434 .kind = DRV_NAME, 435 .priv_size = sizeof(struct veth_priv), 436 .setup = veth_setup, 437 .validate = veth_validate, 438 .newlink = veth_newlink, 439 .dellink = veth_dellink, 440 .policy = veth_policy, 441 .maxtype = VETH_INFO_MAX, 442 }; 443 444 /* 445 * init/fini 446 */ 447 448 static __init int veth_init(void) 449 { 450 return rtnl_link_register(&veth_link_ops); 451 } 452 453 static __exit void veth_exit(void) 454 { 455 rtnl_link_unregister(&veth_link_ops); 456 } 457 458 module_init(veth_init); 459 module_exit(veth_exit); 460 461 MODULE_DESCRIPTION("Virtual Ethernet Tunnel"); 462 MODULE_LICENSE("GPL v2"); 463 MODULE_ALIAS_RTNL_LINK(DRV_NAME); 464