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/list.h>
  12 #include <linux/netdevice.h>
  13 #include <linux/ethtool.h>
  14 #include <linux/etherdevice.h>
  15
  16 #include <net/dst.h>
  17 #include <net/xfrm.h>
  18 #include <net/veth.h>
  19
  20 #define DRV_NAME        "veth"
  21 #define DRV_VERSION     "1.0"
  22
  23 struct veth_net_stats {
  24         unsigned long   rx_packets;
  25         unsigned long   tx_packets;
  26         unsigned long   rx_bytes;
  27         unsigned long   tx_bytes;
  28         unsigned long   tx_dropped;
  29 };
  30
  31 struct veth_priv {
  32         struct net_device *peer;
  33         struct net_device *dev;
  34         struct list_head list;
  35         struct veth_net_stats *stats;
  36         unsigned ip_summed;
  37 };
  38
  39 static LIST_HEAD(veth_list);
  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 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 int 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;
 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
 167         if (!(rcv->flags & IFF_UP))
 168                 goto outf;
 169
 170         skb->pkt_type = PACKET_HOST;
 171         skb->protocol = eth_type_trans(skb, rcv);
 172         if (dev->features & NETIF_F_NO_CSUM)
 173                 skb->ip_summed = rcv_priv->ip_summed;
 174
 175         dst_release(skb->dst);
 176         skb->dst = NULL;
 177         skb->mark = 0;
 178         secpath_reset(skb);
 179         nf_reset(skb);
 180
 181         length = skb->len;
 182
 183         stats->tx_bytes += length;
 184         stats->tx_packets++;
 185
 186         stats = per_cpu_ptr(rcv_priv->stats, cpu);
 187         stats->rx_bytes += length;
 188         stats->rx_packets++;
 189
 190         netif_rx(skb);
 191         return 0;
 192
 193 outf:
 194         kfree_skb(skb);
 195         stats->tx_dropped++;
 196         return 0;
 197 }
 198
 199 /*
 200  * general routines
 201  */
 202
 203 static struct net_device_stats *veth_get_stats(struct net_device *dev)
 204 {
 205         struct veth_priv *priv;
 206         struct net_device_stats *dev_stats;
 207         int cpu;
 208         struct veth_net_stats *stats;
 209
 210         priv = netdev_priv(dev);
 211         dev_stats = &dev->stats;
 212
 213         dev_stats->rx_packets = 0;
 214         dev_stats->tx_packets = 0;
 215         dev_stats->rx_bytes = 0;
 216         dev_stats->tx_bytes = 0;
 217         dev_stats->tx_dropped = 0;
 218
 219         for_each_online_cpu(cpu) {
 220                 stats = per_cpu_ptr(priv->stats, cpu);
 221
 222                 dev_stats->rx_packets += stats->rx_packets;
 223                 dev_stats->tx_packets += stats->tx_packets;
 224                 dev_stats->rx_bytes += stats->rx_bytes;
 225                 dev_stats->tx_bytes += stats->tx_bytes;
 226                 dev_stats->tx_dropped += stats->tx_dropped;
 227         }
 228
 229         return dev_stats;
 230 }
 231
 232 static int veth_open(struct net_device *dev)
 233 {
 234         struct veth_priv *priv;
 235
 236         priv = netdev_priv(dev);
 237         if (priv->peer == NULL)
 238                 return -ENOTCONN;
 239
 240         if (priv->peer->flags & IFF_UP) {
 241                 netif_carrier_on(dev);
 242                 netif_carrier_on(priv->peer);
 243         }
 244         return 0;
 245 }
 246
 247 static int veth_close(struct net_device *dev)
 248 {
 249         struct veth_priv *priv;
 250
 251         if (netif_carrier_ok(dev)) {
 252                 priv = netdev_priv(dev);
 253                 netif_carrier_off(dev);
 254                 netif_carrier_off(priv->peer);
 255         }
 256         return 0;
 257 }
 258
 259 static int veth_dev_init(struct net_device *dev)
 260 {
 261         struct veth_net_stats *stats;
 262         struct veth_priv *priv;
 263
 264         stats = alloc_percpu(struct veth_net_stats);
 265         if (stats == NULL)
 266                 return -ENOMEM;
 267
 268         priv = netdev_priv(dev);
 269         priv->stats = stats;
 270         return 0;
 271 }
 272
 273 static void veth_dev_free(struct net_device *dev)
 274 {
 275         struct veth_priv *priv;
 276
 277         priv = netdev_priv(dev);
 278         free_percpu(priv->stats);
 279         free_netdev(dev);
 280 }
 281
 282 static void veth_setup(struct net_device *dev)
 283 {
 284         ether_setup(dev);
 285
 286         dev->hard_start_xmit = veth_xmit;
 287         dev->get_stats = veth_get_stats;
 288         dev->open = veth_open;
 289         dev->stop = veth_close;
 290         dev->ethtool_ops = &veth_ethtool_ops;
 291         dev->features |= NETIF_F_LLTX;
 292         dev->init = veth_dev_init;
 293         dev->destructor = veth_dev_free;
 294 }
 295
 296 /*
 297  * netlink interface
 298  */
 299
 300 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 301 {
 302         if (tb[IFLA_ADDRESS]) {
 303                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 304                         return -EINVAL;
 305                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 306                         return -EADDRNOTAVAIL;
 307         }
 308         return 0;
 309 }
 310
 311 static struct rtnl_link_ops veth_link_ops;
 312
 313 static int veth_newlink(struct net_device *dev,
 314                          struct nlattr *tb[], struct nlattr *data[])
 315 {
 316         int err;
 317         struct net_device *peer;
 318         struct veth_priv *priv;
 319         char ifname[IFNAMSIZ];
 320         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 321
 322         /*
 323          * create and register peer first
 324          *
 325          * struct ifinfomsg is at the head of VETH_INFO_PEER, but we
 326          * skip it since no info from it is useful yet
 327          */
 328
 329         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 330                 struct nlattr *nla_peer;
 331
 332                 nla_peer = data[VETH_INFO_PEER];
 333                 err = nla_parse(peer_tb, IFLA_MAX,
 334                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 335                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 336                                 ifla_policy);
 337                 if (err < 0)
 338                         return err;
 339
 340                 err = veth_validate(peer_tb, NULL);
 341                 if (err < 0)
 342                         return err;
 343
 344                 tbp = peer_tb;
 345         } else
 346                 tbp = tb;
 347
 348         if (tbp[IFLA_IFNAME])
 349                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 350         else
 351                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 352
 353         peer = rtnl_create_link(dev->nd_net, ifname, &veth_link_ops, tbp);
 354         if (IS_ERR(peer))
 355                 return PTR_ERR(peer);
 356
 357         if (tbp[IFLA_ADDRESS] == NULL)
 358                 random_ether_addr(peer->dev_addr);
 359
 360         err = register_netdevice(peer);
 361         if (err < 0)
 362                 goto err_register_peer;
 363
 364         netif_carrier_off(peer);
 365
 366         /*
 367          * register dev last
 368          *
 369          * note, that since we've registered new device the dev's name
 370          * should be re-allocated
 371          */
 372
 373         if (tb[IFLA_ADDRESS] == NULL)
 374                 random_ether_addr(dev->dev_addr);
 375
 376         if (tb[IFLA_IFNAME])
 377                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 378         else
 379                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 380
 381         if (strchr(dev->name, '%')) {
 382                 err = dev_alloc_name(dev, dev->name);
 383                 if (err < 0)
 384                         goto err_alloc_name;
 385         }
 386
 387         err = register_netdevice(dev);
 388         if (err < 0)
 389                 goto err_register_dev;
 390
 391         netif_carrier_off(dev);
 392
 393         /*
 394          * tie the deviced together
 395          */
 396
 397         priv = netdev_priv(dev);
 398         priv->dev = dev;
 399         priv->peer = peer;
 400         list_add(&priv->list, &veth_list);
 401
 402         priv = netdev_priv(peer);
 403         priv->dev = peer;
 404         priv->peer = dev;
 405         INIT_LIST_HEAD(&priv->list);
 406         return 0;
 407
 408 err_register_dev:
 409         /* nothing to do */
 410 err_alloc_name:
 411         unregister_netdevice(peer);
 412         return err;
 413
 414 err_register_peer:
 415         free_netdev(peer);
 416         return err;
 417 }
 418
 419 static void veth_dellink(struct net_device *dev)
 420 {
 421         struct veth_priv *priv;
 422         struct net_device *peer;
 423
 424         priv = netdev_priv(dev);
 425         peer = priv->peer;
 426
 427         if (!list_empty(&priv->list))
 428                 list_del(&priv->list);
 429
 430         priv = netdev_priv(peer);
 431         if (!list_empty(&priv->list))
 432                 list_del(&priv->list);
 433
 434         unregister_netdevice(dev);
 435         unregister_netdevice(peer);
 436 }
 437
 438 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 439
 440 static struct rtnl_link_ops veth_link_ops = {
 441         .kind           = DRV_NAME,
 442         .priv_size      = sizeof(struct veth_priv),
 443         .setup          = veth_setup,
 444         .validate       = veth_validate,
 445         .newlink        = veth_newlink,
 446         .dellink        = veth_dellink,
 447         .policy         = veth_policy,
 448         .maxtype        = VETH_INFO_MAX,
 449 };
 450
 451 /*
 452  * init/fini
 453  */
 454
 455 static __init int veth_init(void)
 456 {
 457         return rtnl_link_register(&veth_link_ops);
 458 }
 459
 460 static __exit void veth_exit(void)
 461 {
 462         struct veth_priv *priv, *next;
 463
 464         rtnl_lock();
 465         /*
 466          * cannot trust __rtnl_link_unregister() to unregister all
 467          * devices, as each ->dellink call will remove two devices
 468          * from the list at once.
 469          */
 470         list_for_each_entry_safe(priv, next, &veth_list, list)
 471                 veth_dellink(priv->dev);
 472
 473         __rtnl_link_unregister(&veth_link_ops);
 474         rtnl_unlock();
 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);