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 #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                     tx_packets;
  31         u64                     rx_bytes;
  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         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 const struct ethtool_ops veth_ethtool_ops = {
 103         .get_settings           = veth_get_settings,
 104         .get_drvinfo            = veth_get_drvinfo,
 105         .get_link               = ethtool_op_get_link,
 106         .get_strings            = veth_get_strings,
 107         .get_sset_count         = veth_get_sset_count,
 108         .get_ethtool_stats      = veth_get_ethtool_stats,
 109 };
 110
 111 /*
 112  * xmit
 113  */
 114
 115 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
 116 {
 117         struct net_device *rcv = NULL;
 118         struct veth_priv *priv, *rcv_priv;
 119         struct veth_net_stats *stats, *rcv_stats;
 120         int length;
 121
 122         priv = netdev_priv(dev);
 123         rcv = priv->peer;
 124         rcv_priv = netdev_priv(rcv);
 125
 126         stats = this_cpu_ptr(priv->stats);
 127         rcv_stats = this_cpu_ptr(rcv_priv->stats);
 128
 129         /* don't change ip_summed == CHECKSUM_PARTIAL, as that
 130            will cause bad checksum on forwarded packets */
 131         if (skb->ip_summed == CHECKSUM_NONE &&
 132             rcv->features & NETIF_F_RXCSUM)
 133                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 134
 135         length = skb->len;
 136         if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
 137                 goto rx_drop;
 138
 139         u64_stats_update_begin(&stats->syncp);
 140         stats->tx_bytes += length;
 141         stats->tx_packets++;
 142         u64_stats_update_end(&stats->syncp);
 143
 144         u64_stats_update_begin(&rcv_stats->syncp);
 145         rcv_stats->rx_bytes += length;
 146         rcv_stats->rx_packets++;
 147         u64_stats_update_end(&rcv_stats->syncp);
 148
 149         return NETDEV_TX_OK;
 150
 151 rx_drop:
 152         u64_stats_update_begin(&rcv_stats->syncp);
 153         rcv_stats->rx_dropped++;
 154         u64_stats_update_end(&rcv_stats->syncp);
 155         return NETDEV_TX_OK;
 156 }
 157
 158 /*
 159  * general routines
 160  */
 161
 162 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
 163                                                   struct rtnl_link_stats64 *tot)
 164 {
 165         struct veth_priv *priv = netdev_priv(dev);
 166         int cpu;
 167
 168         for_each_possible_cpu(cpu) {
 169                 struct veth_net_stats *stats = per_cpu_ptr(priv->stats, cpu);
 170                 u64 rx_packets, rx_bytes, rx_dropped;
 171                 u64 tx_packets, tx_bytes;
 172                 unsigned int start;
 173
 174                 do {
 175                         start = u64_stats_fetch_begin_bh(&stats->syncp);
 176                         rx_packets = stats->rx_packets;
 177                         tx_packets = stats->tx_packets;
 178                         rx_bytes = stats->rx_bytes;
 179                         tx_bytes = stats->tx_bytes;
 180                         rx_dropped = stats->rx_dropped;
 181                 } while (u64_stats_fetch_retry_bh(&stats->syncp, start));
 182                 tot->rx_packets += rx_packets;
 183                 tot->tx_packets += tx_packets;
 184                 tot->rx_bytes   += rx_bytes;
 185                 tot->tx_bytes   += tx_bytes;
 186                 tot->rx_dropped += rx_dropped;
 187         }
 188
 189         return tot;
 190 }
 191
 192 static int veth_open(struct net_device *dev)
 193 {
 194         struct veth_priv *priv;
 195
 196         priv = netdev_priv(dev);
 197         if (priv->peer == NULL)
 198                 return -ENOTCONN;
 199
 200         if (priv->peer->flags & IFF_UP) {
 201                 netif_carrier_on(dev);
 202                 netif_carrier_on(priv->peer);
 203         }
 204         return 0;
 205 }
 206
 207 static int veth_close(struct net_device *dev)
 208 {
 209         struct veth_priv *priv = netdev_priv(dev);
 210
 211         netif_carrier_off(dev);
 212         netif_carrier_off(priv->peer);
 213
 214         return 0;
 215 }
 216
 217 static int is_valid_veth_mtu(int new_mtu)
 218 {
 219         return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
 220 }
 221
 222 static int veth_change_mtu(struct net_device *dev, int new_mtu)
 223 {
 224         if (!is_valid_veth_mtu(new_mtu))
 225                 return -EINVAL;
 226         dev->mtu = new_mtu;
 227         return 0;
 228 }
 229
 230 static int veth_dev_init(struct net_device *dev)
 231 {
 232         struct veth_net_stats __percpu *stats;
 233         struct veth_priv *priv;
 234
 235         stats = alloc_percpu(struct veth_net_stats);
 236         if (stats == NULL)
 237                 return -ENOMEM;
 238
 239         priv = netdev_priv(dev);
 240         priv->stats = stats;
 241         return 0;
 242 }
 243
 244 static void veth_dev_free(struct net_device *dev)
 245 {
 246         struct veth_priv *priv;
 247
 248         priv = netdev_priv(dev);
 249         free_percpu(priv->stats);
 250         free_netdev(dev);
 251 }
 252
 253 static const struct net_device_ops veth_netdev_ops = {
 254         .ndo_init            = veth_dev_init,
 255         .ndo_open            = veth_open,
 256         .ndo_stop            = veth_close,
 257         .ndo_start_xmit      = veth_xmit,
 258         .ndo_change_mtu      = veth_change_mtu,
 259         .ndo_get_stats64     = veth_get_stats64,
 260         .ndo_set_mac_address = eth_mac_addr,
 261 };
 262
 263 static void veth_setup(struct net_device *dev)
 264 {
 265         ether_setup(dev);
 266
 267         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
 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_NO_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(src_net, 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                 random_ether_addr(peer->dev_addr);
 351
 352         err = register_netdevice(peer);
 353         put_net(net);
 354         net = NULL;
 355         if (err < 0)
 356                 goto err_register_peer;
 357
 358         netif_carrier_off(peer);
 359
 360         err = rtnl_configure_link(peer, ifmp);
 361         if (err < 0)
 362                 goto err_configure_peer;
 363
 364         /*
 365          * register dev last
 366          *
 367          * note, that since we've registered new device the dev's name
 368          * should be re-allocated
 369          */
 370
 371         if (tb[IFLA_ADDRESS] == NULL)
 372                 random_ether_addr(dev->dev_addr);
 373
 374         if (tb[IFLA_IFNAME])
 375                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 376         else
 377                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 378
 379         if (strchr(dev->name, '%')) {
 380                 err = dev_alloc_name(dev, dev->name);
 381                 if (err < 0)
 382                         goto err_alloc_name;
 383         }
 384
 385         err = register_netdevice(dev);
 386         if (err < 0)
 387                 goto err_register_dev;
 388
 389         netif_carrier_off(dev);
 390
 391         /*
 392          * tie the deviced together
 393          */
 394
 395         priv = netdev_priv(dev);
 396         priv->peer = peer;
 397
 398         priv = netdev_priv(peer);
 399         priv->peer = dev;
 400         return 0;
 401
 402 err_register_dev:
 403         /* nothing to do */
 404 err_alloc_name:
 405 err_configure_peer:
 406         unregister_netdevice(peer);
 407         return err;
 408
 409 err_register_peer:
 410         free_netdev(peer);
 411         return err;
 412 }
 413
 414 static void veth_dellink(struct net_device *dev, struct list_head *head)
 415 {
 416         struct veth_priv *priv;
 417         struct net_device *peer;
 418
 419         priv = netdev_priv(dev);
 420         peer = priv->peer;
 421
 422         unregister_netdevice_queue(dev, head);
 423         unregister_netdevice_queue(peer, head);
 424 }
 425
 426 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
 427         [VETH_INFO_PEER]        = { .len = sizeof(struct ifinfomsg) },
 428 };
 429
 430 static struct rtnl_link_ops veth_link_ops = {
 431         .kind           = DRV_NAME,
 432         .priv_size      = sizeof(struct veth_priv),
 433         .setup          = veth_setup,
 434         .validate       = veth_validate,
 435         .newlink        = veth_newlink,
 436         .dellink        = veth_dellink,
 437         .policy         = veth_policy,
 438         .maxtype        = VETH_INFO_MAX,
 439 };
 440
 441 /*
 442  * init/fini
 443  */
 444
 445 static __init int veth_init(void)
 446 {
 447         return rtnl_link_register(&veth_link_ops);
 448 }
 449
 450 static __exit void veth_exit(void)
 451 {
 452         rtnl_link_unregister(&veth_link_ops);
 453 }
 454
 455 module_init(veth_init);
 456 module_exit(veth_exit);
 457
 458 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 459 MODULE_LICENSE("GPL v2");
 460 MODULE_ALIAS_RTNL_LINK(DRV_NAME);