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/slab.h>
  13 #include <linux/ethtool.h>
  14 #include <linux/etherdevice.h>
  15
  16 #include <net/dst.h>
  17 #include <net/xfrm.h>
  18 #include <linux/veth.h>
  19
  20 #define DRV_NAME        "veth"
  21 #define DRV_VERSION     "1.0"
  22
  23 #define MIN_MTU 68              /* Min L3 MTU */
  24 #define MAX_MTU 65535           /* Max L3 MTU (arbitrary) */
  25 #define MTU_PAD (ETH_HLEN + 4)  /* Max difference between L2 and L3 size MTU */
  26
  27 struct veth_net_stats {
  28         unsigned long   rx_packets;
  29         unsigned long   tx_packets;
  30         unsigned long   rx_bytes;
  31         unsigned long   tx_bytes;
  32         unsigned long   tx_dropped;
  33         unsigned long   rx_dropped;
  34 };
  35
  36 struct veth_priv {
  37         struct net_device *peer;
  38         struct veth_net_stats __percpu *stats;
  39         unsigned ip_summed;
  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         cmd->speed              = 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         strcpy(info->driver, DRV_NAME);
  70         strcpy(info->version, DRV_VERSION);
  71         strcpy(info->fw_version, "N/A");
  72 }
  73
  74 static void veth_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
  75 {
  76         switch(stringset) {
  77         case ETH_SS_STATS:
  78                 memcpy(buf, &ethtool_stats_keys, sizeof(ethtool_stats_keys));
  79                 break;
  80         }
  81 }
  82
  83 static int veth_get_sset_count(struct net_device *dev, int sset)
  84 {
  85         switch (sset) {
  86         case ETH_SS_STATS:
  87                 return ARRAY_SIZE(ethtool_stats_keys);
  88         default:
  89                 return -EOPNOTSUPP;
  90         }
  91 }
  92
  93 static void veth_get_ethtool_stats(struct net_device *dev,
  94                 struct ethtool_stats *stats, u64 *data)
  95 {
  96         struct veth_priv *priv;
  97
  98         priv = netdev_priv(dev);
  99         data[0] = priv->peer->ifindex;
 100 }
 101
 102 static u32 veth_get_rx_csum(struct net_device *dev)
 103 {
 104         struct veth_priv *priv;
 105
 106         priv = netdev_priv(dev);
 107         return priv->ip_summed == CHECKSUM_UNNECESSARY;
 108 }
 109
 110 static int veth_set_rx_csum(struct net_device *dev, u32 data)
 111 {
 112         struct veth_priv *priv;
 113
 114         priv = netdev_priv(dev);
 115         priv->ip_summed = data ? CHECKSUM_UNNECESSARY : CHECKSUM_NONE;
 116         return 0;
 117 }
 118
 119 static u32 veth_get_tx_csum(struct net_device *dev)
 120 {
 121         return (dev->features & NETIF_F_NO_CSUM) != 0;
 122 }
 123
 124 static int veth_set_tx_csum(struct net_device *dev, u32 data)
 125 {
 126         if (data)
 127                 dev->features |= NETIF_F_NO_CSUM;
 128         else
 129                 dev->features &= ~NETIF_F_NO_CSUM;
 130         return 0;
 131 }
 132
 133 static const struct ethtool_ops veth_ethtool_ops = {
 134         .get_settings           = veth_get_settings,
 135         .get_drvinfo            = veth_get_drvinfo,
 136         .get_link               = ethtool_op_get_link,
 137         .get_rx_csum            = veth_get_rx_csum,
 138         .set_rx_csum            = veth_set_rx_csum,
 139         .get_tx_csum            = veth_get_tx_csum,
 140         .set_tx_csum            = veth_set_tx_csum,
 141         .get_sg                 = ethtool_op_get_sg,
 142         .set_sg                 = ethtool_op_set_sg,
 143         .get_strings            = veth_get_strings,
 144         .get_sset_count         = veth_get_sset_count,
 145         .get_ethtool_stats      = veth_get_ethtool_stats,
 146 };
 147
 148 /*
 149  * xmit
 150  */
 151
 152 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
 153 {
 154         struct net_device *rcv = NULL;
 155         struct veth_priv *priv, *rcv_priv;
 156         struct veth_net_stats *stats, *rcv_stats;
 157         int length;
 158
 159         priv = netdev_priv(dev);
 160         rcv = priv->peer;
 161         rcv_priv = netdev_priv(rcv);
 162
 163         stats = this_cpu_ptr(priv->stats);
 164         rcv_stats = this_cpu_ptr(rcv_priv->stats);
 165
 166         if (!(rcv->flags & IFF_UP))
 167                 goto tx_drop;
 168
 169         if (dev->features & NETIF_F_NO_CSUM)
 170                 skb->ip_summed = rcv_priv->ip_summed;
 171
 172         length = skb->len + ETH_HLEN;
 173         if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
 174                 goto rx_drop;
 175
 176         stats->tx_bytes += length;
 177         stats->tx_packets++;
 178
 179         rcv_stats->rx_bytes += length;
 180         rcv_stats->rx_packets++;
 181
 182         return NETDEV_TX_OK;
 183
 184 tx_drop:
 185         kfree_skb(skb);
 186         stats->tx_dropped++;
 187         return NETDEV_TX_OK;
 188
 189 rx_drop:
 190         rcv_stats->rx_dropped++;
 191         return NETDEV_TX_OK;
 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         int cpu;
 202         struct veth_net_stats *stats, total = {0};
 203
 204         priv = netdev_priv(dev);
 205
 206         for_each_possible_cpu(cpu) {
 207                 stats = per_cpu_ptr(priv->stats, cpu);
 208
 209                 total.rx_packets += stats->rx_packets;
 210                 total.tx_packets += stats->tx_packets;
 211                 total.rx_bytes   += stats->rx_bytes;
 212                 total.tx_bytes   += stats->tx_bytes;
 213                 total.tx_dropped += stats->tx_dropped;
 214                 total.rx_dropped += stats->rx_dropped;
 215         }
 216         dev->stats.rx_packets = total.rx_packets;
 217         dev->stats.tx_packets = total.tx_packets;
 218         dev->stats.rx_bytes   = total.rx_bytes;
 219         dev->stats.tx_bytes   = total.tx_bytes;
 220         dev->stats.tx_dropped = total.tx_dropped;
 221         dev->stats.rx_dropped = total.rx_dropped;
 222
 223         return &dev->stats;
 224 }
 225
 226 static int veth_open(struct net_device *dev)
 227 {
 228         struct veth_priv *priv;
 229
 230         priv = netdev_priv(dev);
 231         if (priv->peer == NULL)
 232                 return -ENOTCONN;
 233
 234         if (priv->peer->flags & IFF_UP) {
 235                 netif_carrier_on(dev);
 236                 netif_carrier_on(priv->peer);
 237         }
 238         return 0;
 239 }
 240
 241 static int veth_close(struct net_device *dev)
 242 {
 243         struct veth_priv *priv = netdev_priv(dev);
 244
 245         netif_carrier_off(dev);
 246         netif_carrier_off(priv->peer);
 247
 248         return 0;
 249 }
 250
 251 static int is_valid_veth_mtu(int new_mtu)
 252 {
 253         return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
 254 }
 255
 256 static int veth_change_mtu(struct net_device *dev, int new_mtu)
 257 {
 258         if (!is_valid_veth_mtu(new_mtu))
 259                 return -EINVAL;
 260         dev->mtu = new_mtu;
 261         return 0;
 262 }
 263
 264 static int veth_dev_init(struct net_device *dev)
 265 {
 266         struct veth_net_stats __percpu *stats;
 267         struct veth_priv *priv;
 268
 269         stats = alloc_percpu(struct veth_net_stats);
 270         if (stats == NULL)
 271                 return -ENOMEM;
 272
 273         priv = netdev_priv(dev);
 274         priv->stats = stats;
 275         return 0;
 276 }
 277
 278 static void veth_dev_free(struct net_device *dev)
 279 {
 280         struct veth_priv *priv;
 281
 282         priv = netdev_priv(dev);
 283         free_percpu(priv->stats);
 284         free_netdev(dev);
 285 }
 286
 287 static const struct net_device_ops veth_netdev_ops = {
 288         .ndo_init            = veth_dev_init,
 289         .ndo_open            = veth_open,
 290         .ndo_stop            = veth_close,
 291         .ndo_start_xmit      = veth_xmit,
 292         .ndo_change_mtu      = veth_change_mtu,
 293         .ndo_get_stats       = veth_get_stats,
 294         .ndo_set_mac_address = eth_mac_addr,
 295 };
 296
 297 static void veth_setup(struct net_device *dev)
 298 {
 299         ether_setup(dev);
 300
 301         dev->netdev_ops = &veth_netdev_ops;
 302         dev->ethtool_ops = &veth_ethtool_ops;
 303         dev->features |= NETIF_F_LLTX;
 304         dev->destructor = veth_dev_free;
 305 }
 306
 307 /*
 308  * netlink interface
 309  */
 310
 311 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 312 {
 313         if (tb[IFLA_ADDRESS]) {
 314                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 315                         return -EINVAL;
 316                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 317                         return -EADDRNOTAVAIL;
 318         }
 319         if (tb[IFLA_MTU]) {
 320                 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
 321                         return -EINVAL;
 322         }
 323         return 0;
 324 }
 325
 326 static struct rtnl_link_ops veth_link_ops;
 327
 328 static int veth_newlink(struct net *src_net, struct net_device *dev,
 329                          struct nlattr *tb[], struct nlattr *data[])
 330 {
 331         int err;
 332         struct net_device *peer;
 333         struct veth_priv *priv;
 334         char ifname[IFNAMSIZ];
 335         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 336         struct ifinfomsg *ifmp;
 337         struct net *net;
 338
 339         /*
 340          * create and register peer first
 341          */
 342         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 343                 struct nlattr *nla_peer;
 344
 345                 nla_peer = data[VETH_INFO_PEER];
 346                 ifmp = nla_data(nla_peer);
 347                 err = nla_parse(peer_tb, IFLA_MAX,
 348                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 349                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 350                                 ifla_policy);
 351                 if (err < 0)
 352                         return err;
 353
 354                 err = veth_validate(peer_tb, NULL);
 355                 if (err < 0)
 356                         return err;
 357
 358                 tbp = peer_tb;
 359         } else {
 360                 ifmp = NULL;
 361                 tbp = tb;
 362         }
 363
 364         if (tbp[IFLA_IFNAME])
 365                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 366         else
 367                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 368
 369         net = rtnl_link_get_net(src_net, tbp);
 370         if (IS_ERR(net))
 371                 return PTR_ERR(net);
 372
 373         peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
 374         if (IS_ERR(peer)) {
 375                 put_net(net);
 376                 return PTR_ERR(peer);
 377         }
 378
 379         if (tbp[IFLA_ADDRESS] == NULL)
 380                 random_ether_addr(peer->dev_addr);
 381
 382         err = register_netdevice(peer);
 383         put_net(net);
 384         net = NULL;
 385         if (err < 0)
 386                 goto err_register_peer;
 387
 388         netif_carrier_off(peer);
 389
 390         err = rtnl_configure_link(peer, ifmp);
 391         if (err < 0)
 392                 goto err_configure_peer;
 393
 394         /*
 395          * register dev last
 396          *
 397          * note, that since we've registered new device the dev's name
 398          * should be re-allocated
 399          */
 400
 401         if (tb[IFLA_ADDRESS] == NULL)
 402                 random_ether_addr(dev->dev_addr);
 403
 404         if (tb[IFLA_IFNAME])
 405                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 406         else
 407                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 408
 409         if (strchr(dev->name, '%')) {
 410                 err = dev_alloc_name(dev, dev->name);
 411                 if (err < 0)
 412                         goto err_alloc_name;
 413         }
 414
 415         err = register_netdevice(dev);
 416         if (err < 0)
 417                 goto err_register_dev;
 418
 419         netif_carrier_off(dev);
 420
 421         /*
 422          * tie the deviced together
 423          */
 424
 425         priv = netdev_priv(dev);
 426         priv->peer = peer;
 427
 428         priv = netdev_priv(peer);
 429         priv->peer = dev;
 430         return 0;
 431
 432 err_register_dev:
 433         /* nothing to do */
 434 err_alloc_name:
 435 err_configure_peer:
 436         unregister_netdevice(peer);
 437         return err;
 438
 439 err_register_peer:
 440         free_netdev(peer);
 441         return err;
 442 }
 443
 444 static void veth_dellink(struct net_device *dev, struct list_head *head)
 445 {
 446         struct veth_priv *priv;
 447         struct net_device *peer;
 448
 449         priv = netdev_priv(dev);
 450         peer = priv->peer;
 451
 452         unregister_netdevice_queue(dev, head);
 453         unregister_netdevice_queue(peer, head);
 454 }
 455
 456 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 457
 458 static struct rtnl_link_ops veth_link_ops = {
 459         .kind           = DRV_NAME,
 460         .priv_size      = sizeof(struct veth_priv),
 461         .setup          = veth_setup,
 462         .validate       = veth_validate,
 463         .newlink        = veth_newlink,
 464         .dellink        = veth_dellink,
 465         .policy         = veth_policy,
 466         .maxtype        = VETH_INFO_MAX,
 467 };
 468
 469 /*
 470  * init/fini
 471  */
 472
 473 static __init int veth_init(void)
 474 {
 475         return rtnl_link_register(&veth_link_ops);
 476 }
 477
 478 static __exit void veth_exit(void)
 479 {
 480         rtnl_link_unregister(&veth_link_ops);
 481 }
 482
 483 module_init(veth_init);
 484 module_exit(veth_exit);
 485
 486 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 487 MODULE_LICENSE("GPL v2");
 488 MODULE_ALIAS_RTNL_LINK(DRV_NAME);