linux/drivers/net/veth.c
<<
>>
Prefs
   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
  26struct 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
  35struct 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
  45static struct {
  46        const char string[ETH_GSTRING_LEN];
  47} ethtool_stats_keys[] = {
  48        { "peer_ifindex" },
  49};
  50
  51static 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
  66static 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
  73static 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
  82static 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
  92static 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
 101static 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
 109static 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
 118static u32 veth_get_tx_csum(struct net_device *dev)
 119{
 120        return (dev->features & NETIF_F_NO_CSUM) != 0;
 121}
 122
 123static 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
 132static 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
 151static 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, 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->tstamp.tv64 = 0;
 175        skb->pkt_type = PACKET_HOST;
 176        skb->protocol = eth_type_trans(skb, rcv);
 177        if (dev->features & NETIF_F_NO_CSUM)
 178                skb->ip_summed = rcv_priv->ip_summed;
 179
 180        skb->mark = 0;
 181        secpath_reset(skb);
 182        nf_reset(skb);
 183
 184        length = skb->len;
 185
 186        stats->tx_bytes += length;
 187        stats->tx_packets++;
 188
 189        rcv_stats->rx_bytes += length;
 190        rcv_stats->rx_packets++;
 191
 192        netif_rx(skb);
 193        return NETDEV_TX_OK;
 194
 195tx_drop:
 196        kfree_skb(skb);
 197        stats->tx_dropped++;
 198        return NETDEV_TX_OK;
 199
 200rx_drop:
 201        kfree_skb(skb);
 202        rcv_stats->rx_dropped++;
 203        return NETDEV_TX_OK;
 204}
 205
 206/*
 207 * general routines
 208 */
 209
 210static struct net_device_stats *veth_get_stats(struct net_device *dev)
 211{
 212        struct veth_priv *priv;
 213        int cpu;
 214        struct veth_net_stats *stats, total = {0};
 215
 216        priv = netdev_priv(dev);
 217
 218        for_each_possible_cpu(cpu) {
 219                stats = per_cpu_ptr(priv->stats, cpu);
 220
 221                total.rx_packets += stats->rx_packets;
 222                total.tx_packets += stats->tx_packets;
 223                total.rx_bytes   += stats->rx_bytes;
 224                total.tx_bytes   += stats->tx_bytes;
 225                total.tx_dropped += stats->tx_dropped;
 226                total.rx_dropped += stats->rx_dropped;
 227        }
 228        dev->stats.rx_packets = total.rx_packets;
 229        dev->stats.tx_packets = total.tx_packets;
 230        dev->stats.rx_bytes   = total.rx_bytes;
 231        dev->stats.tx_bytes   = total.tx_bytes;
 232        dev->stats.tx_dropped = total.tx_dropped;
 233        dev->stats.rx_dropped = total.rx_dropped;
 234
 235        return &dev->stats;
 236}
 237
 238static 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
 253static 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        return 0;
 261}
 262
 263static int is_valid_veth_mtu(int new_mtu)
 264{
 265        return (new_mtu >= MIN_MTU && new_mtu <= MAX_MTU);
 266}
 267
 268static int veth_change_mtu(struct net_device *dev, int new_mtu)
 269{
 270        if (!is_valid_veth_mtu(new_mtu))
 271                return -EINVAL;
 272        dev->mtu = new_mtu;
 273        return 0;
 274}
 275
 276static int veth_dev_init(struct net_device *dev)
 277{
 278        struct veth_net_stats *stats;
 279        struct veth_priv *priv;
 280
 281        stats = alloc_percpu(struct veth_net_stats);
 282        if (stats == NULL)
 283                return -ENOMEM;
 284
 285        priv = netdev_priv(dev);
 286        priv->stats = stats;
 287        return 0;
 288}
 289
 290static void veth_dev_free(struct net_device *dev)
 291{
 292        struct veth_priv *priv;
 293
 294        priv = netdev_priv(dev);
 295        free_percpu(priv->stats);
 296        free_netdev(dev);
 297}
 298
 299static const struct net_device_ops veth_netdev_ops = {
 300        .ndo_init            = veth_dev_init,
 301        .ndo_open            = veth_open,
 302        .ndo_stop            = veth_close,
 303        .ndo_start_xmit      = veth_xmit,
 304        .ndo_change_mtu      = veth_change_mtu,
 305        .ndo_get_stats       = veth_get_stats,
 306        .ndo_set_mac_address = eth_mac_addr,
 307};
 308
 309static void veth_setup(struct net_device *dev)
 310{
 311        ether_setup(dev);
 312
 313        dev->netdev_ops = &veth_netdev_ops;
 314        dev->ethtool_ops = &veth_ethtool_ops;
 315        dev->features |= NETIF_F_LLTX;
 316        dev->destructor = veth_dev_free;
 317}
 318
 319/*
 320 * netlink interface
 321 */
 322
 323static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 324{
 325        if (tb[IFLA_ADDRESS]) {
 326                if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 327                        return -EINVAL;
 328                if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 329                        return -EADDRNOTAVAIL;
 330        }
 331        if (tb[IFLA_MTU]) {
 332                if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
 333                        return -EINVAL;
 334        }
 335        return 0;
 336}
 337
 338static struct rtnl_link_ops veth_link_ops;
 339
 340static int veth_newlink(struct net_device *dev,
 341                         struct nlattr *tb[], struct nlattr *data[])
 342{
 343        int err;
 344        struct net_device *peer;
 345        struct veth_priv *priv;
 346        char ifname[IFNAMSIZ];
 347        struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 348
 349        /*
 350         * create and register peer first
 351         *
 352         * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
 353         * skip it since no info from it is useful yet
 354         */
 355
 356        if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 357                struct nlattr *nla_peer;
 358
 359                nla_peer = data[VETH_INFO_PEER];
 360                err = nla_parse(peer_tb, IFLA_MAX,
 361                                nla_data(nla_peer) + sizeof(struct ifinfomsg),
 362                                nla_len(nla_peer) - sizeof(struct ifinfomsg),
 363                                ifla_policy);
 364                if (err < 0)
 365                        return err;
 366
 367                err = veth_validate(peer_tb, NULL);
 368                if (err < 0)
 369                        return err;
 370
 371                tbp = peer_tb;
 372        } else
 373                tbp = tb;
 374
 375        if (tbp[IFLA_IFNAME])
 376                nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 377        else
 378                snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 379
 380        peer = rtnl_create_link(dev_net(dev), ifname, &veth_link_ops, tbp);
 381        if (IS_ERR(peer))
 382                return PTR_ERR(peer);
 383
 384        if (tbp[IFLA_ADDRESS] == NULL)
 385                random_ether_addr(peer->dev_addr);
 386
 387        err = register_netdevice(peer);
 388        if (err < 0)
 389                goto err_register_peer;
 390
 391        netif_carrier_off(peer);
 392
 393        /*
 394         * register dev last
 395         *
 396         * note, that since we've registered new device the dev's name
 397         * should be re-allocated
 398         */
 399
 400        if (tb[IFLA_ADDRESS] == NULL)
 401                random_ether_addr(dev->dev_addr);
 402
 403        if (tb[IFLA_IFNAME])
 404                nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 405        else
 406                snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 407
 408        if (strchr(dev->name, '%')) {
 409                err = dev_alloc_name(dev, dev->name);
 410                if (err < 0)
 411                        goto err_alloc_name;
 412        }
 413
 414        err = register_netdevice(dev);
 415        if (err < 0)
 416                goto err_register_dev;
 417
 418        netif_carrier_off(dev);
 419
 420        /*
 421         * tie the deviced together
 422         */
 423
 424        priv = netdev_priv(dev);
 425        priv->peer = peer;
 426
 427        priv = netdev_priv(peer);
 428        priv->peer = dev;
 429        return 0;
 430
 431err_register_dev:
 432        /* nothing to do */
 433err_alloc_name:
 434        unregister_netdevice(peer);
 435        return err;
 436
 437err_register_peer:
 438        free_netdev(peer);
 439        return err;
 440}
 441
 442static void veth_dellink(struct net_device *dev)
 443{
 444        struct veth_priv *priv;
 445        struct net_device *peer;
 446
 447        priv = netdev_priv(dev);
 448        peer = priv->peer;
 449
 450        unregister_netdevice(dev);
 451        unregister_netdevice(peer);
 452}
 453
 454static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 455
 456static struct rtnl_link_ops veth_link_ops = {
 457        .kind           = DRV_NAME,
 458        .priv_size      = sizeof(struct veth_priv),
 459        .setup          = veth_setup,
 460        .validate       = veth_validate,
 461        .newlink        = veth_newlink,
 462        .dellink        = veth_dellink,
 463        .policy         = veth_policy,
 464        .maxtype        = VETH_INFO_MAX,
 465};
 466
 467/*
 468 * init/fini
 469 */
 470
 471static __init int veth_init(void)
 472{
 473        return rtnl_link_register(&veth_link_ops);
 474}
 475
 476static __exit void veth_exit(void)
 477{
 478        rtnl_link_unregister(&veth_link_ops);
 479}
 480
 481module_init(veth_init);
 482module_exit(veth_exit);
 483
 484MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 485MODULE_LICENSE("GPL v2");
 486MODULE_ALIAS_RTNL_LINK(DRV_NAME);
 487