drivers/net/veth.c

   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 const 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 netdev_tx_t 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;
 157
 158         priv = netdev_priv(dev);
 159         rcv = priv->peer;
 160         rcv_priv = netdev_priv(rcv);
 161
 162         stats = this_cpu_ptr(priv->stats);
 163         rcv_stats = this_cpu_ptr(rcv_priv->stats);
 164
 165         if (!(rcv->flags & IFF_UP))
 166                 goto tx_drop;
 167
 168         if (dev->features & NETIF_F_NO_CSUM)
 169                 skb->ip_summed = rcv_priv->ip_summed;
 170
 171         length = skb->len + ETH_HLEN;
 172         if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
 173                 goto rx_drop;
 174
 175         stats->tx_bytes += length;
 176         stats->tx_packets++;
 177
 178         rcv_stats->rx_bytes += length;
 179         rcv_stats->rx_packets++;
 180
 181         return NETDEV_TX_OK;
 182
 183 tx_drop:
 184         kfree_skb(skb);
 185         stats->tx_dropped++;
 186         return NETDEV_TX_OK;
 187
 188 rx_drop:
 189         rcv_stats->rx_dropped++;
 190         return NETDEV_TX_OK;
 191 }
 192
 193 /*
 194  * general routines
 195  */
 196
 197 static struct net_device_stats *veth_get_stats(struct net_device *dev)
 198 {
 199         struct veth_priv *priv;
 200         int cpu;
 201         struct veth_net_stats *stats, total = {0};
 202
 203         priv = netdev_priv(dev);
 204
 205         for_each_possible_cpu(cpu) {
 206                 stats = per_cpu_ptr(priv->stats, cpu);
 207
 208                 total.rx_packets += stats->rx_packets;
 209                 total.tx_packets += stats->tx_packets;
 210                 total.rx_bytes   += stats->rx_bytes;
 211                 total.tx_bytes   += stats->tx_bytes;
 212                 total.tx_dropped += stats->tx_dropped;
 213                 total.rx_dropped += stats->rx_dropped;
 214         }
 215         dev->stats.rx_packets = total.rx_packets;
 216         dev->stats.tx_packets = total.tx_packets;
 217         dev->stats.rx_bytes   = total.rx_bytes;
 218         dev->stats.tx_bytes   = total.tx_bytes;
 219         dev->stats.tx_dropped = total.tx_dropped;
 220         dev->stats.rx_dropped = total.rx_dropped;
 221
 222         return &dev->stats;
 223 }
 224
 225 static int veth_open(struct net_device *dev)
 226 {
 227         struct veth_priv *priv;
 228
 229         priv = netdev_priv(dev);
 230         if (priv->peer == NULL)
 231                 return -ENOTCONN;
 232
 233         if (priv->peer->flags & IFF_UP) {
 234                 netif_carrier_on(dev);
 235                 netif_carrier_on(priv->peer);
 236         }
 237         return 0;
 238 }
 239
 240 static int veth_close(struct net_device *dev)
 241 {
 242         struct veth_priv *priv = netdev_priv(dev);
 243
 244         netif_carrier_off(dev);
 245         netif_carrier_off(priv->peer);
 246
 247         return 0;
 248 }
 249
 250 static int is_valid_veth_mtu(int new_mtu)
 251 {
 252         return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
 253 }
 254
 255 static int veth_change_mtu(struct net_device *dev, int new_mtu)
 256 {
 257         if (!is_valid_veth_mtu(new_mtu))
 258                 return -EINVAL;
 259         dev->mtu = new_mtu;
 260         return 0;
 261 }
 262
 263 static int veth_dev_init(struct net_device *dev)
 264 {
 265         struct veth_net_stats *stats;
 266         struct veth_priv *priv;
 267
 268         stats = alloc_percpu(struct veth_net_stats);
 269         if (stats == NULL)
 270                 return -ENOMEM;
 271
 272         priv = netdev_priv(dev);
 273         priv->stats = stats;
 274         return 0;
 275 }
 276
 277 static void veth_dev_free(struct net_device *dev)
 278 {
 279         struct veth_priv *priv;
 280
 281         priv = netdev_priv(dev);
 282         free_percpu(priv->stats);
 283         free_netdev(dev);
 284 }
 285
 286 static const struct net_device_ops veth_netdev_ops = {
 287         .ndo_init            = veth_dev_init,
 288         .ndo_open            = veth_open,
 289         .ndo_stop            = veth_close,
 290         .ndo_start_xmit      = veth_xmit,
 291         .ndo_change_mtu      = veth_change_mtu,
 292         .ndo_get_stats       = veth_get_stats,
 293         .ndo_set_mac_address = eth_mac_addr,
 294 };
 295
 296 static void veth_setup(struct net_device *dev)
 297 {
 298         ether_setup(dev);
 299
 300         dev->netdev_ops = &veth_netdev_ops;
 301         dev->ethtool_ops = &veth_ethtool_ops;
 302         dev->features |= NETIF_F_LLTX;
 303         dev->destructor = veth_dev_free;
 304 }
 305
 306 /*
 307  * netlink interface
 308  */
 309
 310 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 311 {
 312         if (tb[IFLA_ADDRESS]) {
 313                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 314                         return -EINVAL;
 315                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 316                         return -EADDRNOTAVAIL;
 317         }
 318         if (tb[IFLA_MTU]) {
 319                 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
 320                         return -EINVAL;
 321         }
 322         return 0;
 323 }
 324
 325 static struct rtnl_link_ops veth_link_ops;
 326
 327 static int veth_newlink(struct net *src_net, struct net_device *dev,
 328                          struct nlattr *tb[], struct nlattr *data[])
 329 {
 330         int err;
 331         struct net_device *peer;
 332         struct veth_priv *priv;
 333         char ifname[IFNAMSIZ];
 334         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 335         struct net *net;
 336
 337         /*
 338          * create and register peer first
 339          *
 340          * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
 341          * skip it since no info from it is useful yet
 342          */
 343
 344         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 345                 struct nlattr *nla_peer;
 346
 347                 nla_peer = data[VETH_INFO_PEER];
 348                 err = nla_parse(peer_tb, IFLA_MAX,
 349                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 350                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 351                                 ifla_policy);
 352                 if (err < 0)
 353                         return err;
 354
 355                 err = veth_validate(peer_tb, NULL);
 356                 if (err < 0)
 357                         return err;
 358
 359                 tbp = peer_tb;
 360         } else
 361                 tbp = tb;
 362
 363         if (tbp[IFLA_IFNAME])
 364                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 365         else
 366                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 367
 368         net = rtnl_link_get_net(src_net, tbp);
 369         if (IS_ERR(net))
 370                 return PTR_ERR(net);
 371
 372         peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
 373         if (IS_ERR(peer)) {
 374                 put_net(net);
 375                 return PTR_ERR(peer);
 376         }
 377
 378         if (tbp[IFLA_ADDRESS] == NULL)
 379                 random_ether_addr(peer->dev_addr);
 380
 381         err = register_netdevice(peer);
 382         put_net(net);
 383         net = NULL;
 384         if (err < 0)
 385                 goto err_register_peer;
 386
 387         netif_carrier_off(peer);
 388
 389         /*
 390          * register dev last
 391          *
 392          * note, that since we've registered new device the dev's name
 393          * should be re-allocated
 394          */
 395
 396         if (tb[IFLA_ADDRESS] == NULL)
 397                 random_ether_addr(dev->dev_addr);
 398
 399         if (tb[IFLA_IFNAME])
 400                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 401         else
 402                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 403
 404         if (strchr(dev->name, '%')) {
 405                 err = dev_alloc_name(dev, dev->name);
 406                 if (err < 0)
 407                         goto err_alloc_name;
 408         }
 409
 410         err = register_netdevice(dev);
 411         if (err < 0)
 412                 goto err_register_dev;
 413
 414         netif_carrier_off(dev);
 415
 416         /*
 417          * tie the deviced together
 418          */
 419
 420         priv = netdev_priv(dev);
 421         priv->peer = peer;
 422
 423         priv = netdev_priv(peer);
 424         priv->peer = dev;
 425         return 0;
 426
 427 err_register_dev:
 428         /* nothing to do */
 429 err_alloc_name:
 430         unregister_netdevice(peer);
 431         return err;
 432
 433 err_register_peer:
 434         free_netdev(peer);
 435         return err;
 436 }
 437
 438 static void veth_dellink(struct net_device *dev, struct list_head *head)
 439 {
 440         struct veth_priv *priv;
 441         struct net_device *peer;
 442
 443         priv = netdev_priv(dev);
 444         peer = priv->peer;
 445
 446         unregister_netdevice_queue(dev, head);
 447         unregister_netdevice_queue(peer, head);
 448 }
 449
 450 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 451
 452 static struct rtnl_link_ops veth_link_ops = {
 453         .kind           = DRV_NAME,
 454         .priv_size      = sizeof(struct veth_priv),
 455         .setup          = veth_setup,
 456         .validate       = veth_validate,
 457         .newlink        = veth_newlink,
 458         .dellink        = veth_dellink,
 459         .policy         = veth_policy,
 460         .maxtype        = VETH_INFO_MAX,
 461 };
 462
 463 /*
 464  * init/fini
 465  */
 466
 467 static __init int veth_init(void)
 468 {
 469         return rtnl_link_register(&veth_link_ops);
 470 }
 471
 472 static __exit void veth_exit(void)
 473 {
 474         rtnl_link_unregister(&veth_link_ops);
 475 }
 476
 477 module_init(veth_init);
 478 module_exit(veth_exit);
 479
 480 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 481 MODULE_LICENSE("GPL v2");
 482 MODULE_ALIAS_RTNL_LINK(DRV_NAME);