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 struct veth_net_stats {
  23         unsigned long   rx_packets;
  24         unsigned long   tx_packets;
  25         unsigned long   rx_bytes;
  26         unsigned long   tx_bytes;
  27         unsigned long   tx_dropped;
  28 };
  29
  30 struct veth_priv {
  31         struct net_device *peer;
  32         struct veth_net_stats *stats;
  33         unsigned ip_summed;
  34 };
  35
  36 /*
  37  * ethtool interface
  38  */
  39
  40 static struct {
  41         const char string[ETH_GSTRING_LEN];
  42 } ethtool_stats_keys[] = {
  43         { "peer_ifindex" },
  44 };
  45
  46 static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
  47 {
  48         cmd->supported          = 0;
  49         cmd->advertising        = 0;
  50         cmd->speed              = SPEED_10000;
  51         cmd->duplex             = DUPLEX_FULL;
  52         cmd->port               = PORT_TP;
  53         cmd->phy_address        = 0;
  54         cmd->transceiver        = XCVR_INTERNAL;
  55         cmd->autoneg            = AUTONEG_DISABLE;
  56         cmd->maxtxpkt           = 0;
  57         cmd->maxrxpkt           = 0;
  58         return 0;
  59 }
  60
  61 static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
  62 {
  63         strcpy(info->driver, DRV_NAME);
  64         strcpy(info->version, DRV_VERSION);
  65         strcpy(info->fw_version, "N/A");
  66 }
  67
  68 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
  69 {
  70         switch(stringset) {
  71         case ETH_SS_STATS:
  72                 memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
  73                 break;
  74         }
  75 }
  76
  77 static int veth_get_sset_count(struct net_device *dev, int sset)
  78 {
  79         switch (sset) {
  80         case ETH_SS_STATS:
  81                 return ARRAY_SIZE(ethtool_stats_keys);
  82         default:
  83                 return -EOPNOTSUPP;
  84         }
  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_sset_count         = veth_get_sset_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_dev_init(struct net_device *dev)
 243 {
 244         struct veth_net_stats *stats;
 245         struct veth_priv *priv;
 246
 247         stats = alloc_percpu(struct veth_net_stats);
 248         if (stats == NULL)
 249                 return -ENOMEM;
 250
 251         priv = netdev_priv(dev);
 252         priv->stats = stats;
 253         return 0;
 254 }
 255
 256 static void veth_dev_free(struct net_device *dev)
 257 {
 258         struct veth_priv *priv;
 259
 260         priv = netdev_priv(dev);
 261         free_percpu(priv->stats);
 262         free_netdev(dev);
 263 }
 264
 265 static const struct net_device_ops veth_netdev_ops = {
 266         .ndo_init            = veth_dev_init,
 267         .ndo_open            = veth_open,
 268         .ndo_start_xmit      = veth_xmit,
 269         .ndo_get_stats       = veth_get_stats,
 270         .ndo_set_mac_address = eth_mac_addr,
 271 };
 272
 273 static void veth_setup(struct net_device *dev)
 274 {
 275         ether_setup(dev);
 276
 277         dev->netdev_ops = &veth_netdev_ops;
 278         dev->ethtool_ops = &veth_ethtool_ops;
 279         dev->features |= NETIF_F_LLTX;
 280         dev->destructor = veth_dev_free;
 281 }
 282
 283 static void veth_change_state(struct net_device *dev)
 284 {
 285         struct net_device *peer;
 286         struct veth_priv *priv;
 287
 288         priv = netdev_priv(dev);
 289         peer = priv->peer;
 290
 291         if (netif_carrier_ok(peer)) {
 292                 if (!netif_carrier_ok(dev))
 293                         netif_carrier_on(dev);
 294         } else {
 295                 if (netif_carrier_ok(dev))
 296                         netif_carrier_off(dev);
 297         }
 298 }
 299
 300 static int veth_device_event(struct notifier_block *unused,
 301                              unsigned long event, void *ptr)
 302 {
 303         struct net_device *dev = ptr;
 304
 305         if (dev->netdev_ops->ndo_open != veth_open)
 306                 goto out;
 307
 308         switch (event) {
 309         case NETDEV_CHANGE:
 310                 veth_change_state(dev);
 311                 break;
 312         }
 313 out:
 314         return NOTIFY_DONE;
 315 }
 316
 317 static struct notifier_block veth_notifier_block __read_mostly = {
 318         .notifier_call  = veth_device_event,
 319 };
 320
 321 /*
 322  * netlink interface
 323  */
 324
 325 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 326 {
 327         if (tb[IFLA_ADDRESS]) {
 328                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 329                         return -EINVAL;
 330                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 331                         return -EADDRNOTAVAIL;
 332         }
 333         return 0;
 334 }
 335
 336 static struct rtnl_link_ops veth_link_ops;
 337
 338 static int veth_newlink(struct net_device *dev,
 339                          struct nlattr *tb[], struct nlattr *data[])
 340 {
 341         int err;
 342         struct net_device *peer;
 343         struct veth_priv *priv;
 344         char ifname[IFNAMSIZ];
 345         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 346
 347         /*
 348          * create and register peer first
 349          *
 350          * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
 351          * skip it since no info from it is useful yet
 352          */
 353
 354         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 355                 struct nlattr *nla_peer;
 356
 357                 nla_peer = data[VETH_INFO_PEER];
 358                 err = nla_parse(peer_tb, IFLA_MAX,
 359                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 360                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 361                                 ifla_policy);
 362                 if (err < 0)
 363                         return err;
 364
 365                 err = veth_validate(peer_tb, NULL);
 366                 if (err < 0)
 367                         return err;
 368
 369                 tbp = peer_tb;
 370         } else
 371                 tbp = tb;
 372
 373         if (tbp[IFLA_IFNAME])
 374                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 375         else
 376                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 377
 378         peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
 379         if (IS_ERR(peer))
 380                 return PTR_ERR(peer);
 381
 382         if (tbp[IFLA_ADDRESS] == NULL)
 383                 random_ether_addr(peer->dev_addr);
 384
 385         err = register_netdevice(peer);
 386         if (err < 0)
 387                 goto err_register_peer;
 388
 389         netif_carrier_off(peer);
 390
 391         /*
 392          * register dev last
 393          *
 394          * note, that since we've registered new device the dev's name
 395          * should be re-allocated
 396          */
 397
 398         if (tb[IFLA_ADDRESS] == NULL)
 399                 random_ether_addr(dev->dev_addr);
 400
 401         if (tb[IFLA_IFNAME])
 402                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 403         else
 404                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 405
 406         if (strchr(dev->name, '%')) {
 407                 err = dev_alloc_name(dev, dev->name);
 408                 if (err < 0)
 409                         goto err_alloc_name;
 410         }
 411
 412         err = register_netdevice(dev);
 413         if (err < 0)
 414                 goto err_register_dev;
 415
 416         netif_carrier_off(dev);
 417
 418         /*
 419          * tie the deviced together
 420          */
 421
 422         priv = netdev_priv(dev);
 423         priv->peer = peer;
 424
 425         priv = netdev_priv(peer);
 426         priv->peer = dev;
 427         return 0;
 428
 429 err_register_dev:
 430         /* nothing to do */
 431 err_alloc_name:
 432         unregister_netdevice(peer);
 433         return err;
 434
 435 err_register_peer:
 436         free_netdev(peer);
 437         return err;
 438 }
 439
 440 static void veth_dellink(struct net_device *dev)
 441 {
 442         struct veth_priv *priv;
 443         struct net_device *peer;
 444
 445         priv = netdev_priv(dev);
 446         peer = priv->peer;
 447
 448         unregister_netdevice(dev);
 449         unregister_netdevice(peer);
 450 }
 451
 452 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 453
 454 static struct rtnl_link_ops veth_link_ops = {
 455         .kind           = DRV_NAME,
 456         .priv_size      = sizeof(struct veth_priv),
 457         .setup          = veth_setup,
 458         .validate       = veth_validate,
 459         .newlink        = veth_newlink,
 460         .dellink        = veth_dellink,
 461         .policy         = veth_policy,
 462         .maxtype        = VETH_INFO_MAX,
 463 };
 464
 465 /*
 466  * init/fini
 467  */
 468
 469 static __init int veth_init(void)
 470 {
 471         register_netdevice_notifier(&veth_notifier_block);
 472         return rtnl_link_register(&veth_link_ops);
 473 }
 474
 475 static __exit void veth_exit(void)
 476 {
 477         rtnl_link_unregister(&veth_link_ops);
 478         unregister_netdevice_notifier(&veth_notifier_block);
 479 }
 480
 481 module_init(veth_init);
 482 module_exit(veth_exit);
 483
 484 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 485 MODULE_LICENSE("GPL v2");
 486 MODULE_ALIAS_RTNL_LINK(DRV_NAME);