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                     rx_bytes;
  31         u64                     tx_packets;
  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         strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
  70         strlcpy(info->version, DRV_VERSION, sizeof(info->version));
  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 const struct ethtool_ops veth_ethtool_ops = {
 102         .get_settings           = veth_get_settings,
 103         .get_drvinfo            = veth_get_drvinfo,
 104         .get_link               = ethtool_op_get_link,
 105         .get_strings            = veth_get_strings,
 106         .get_sset_count         = veth_get_sset_count,
 107         .get_ethtool_stats      = veth_get_ethtool_stats,
 108 };
 109
 110 /*
 111  * xmit
 112  */
 113
 114 static netdev_tx_t veth_xmit(struct sk_buff *skb, struct net_device *dev)
 115 {
 116         struct net_device *rcv = NULL;
 117         struct veth_priv *priv, *rcv_priv;
 118         struct veth_net_stats *stats, *rcv_stats;
 119         int length;
 120
 121         priv = netdev_priv(dev);
 122         rcv = priv->peer;
 123         rcv_priv = netdev_priv(rcv);
 124
 125         stats = this_cpu_ptr(priv->stats);
 126         rcv_stats = this_cpu_ptr(rcv_priv->stats);
 127
 128         /* don't change ip_summed == CHECKSUM_PARTIAL, as that
 129            will cause bad checksum on forwarded packets */
 130         if (skb->ip_summed == CHECKSUM_NONE &&
 131             rcv->features & NETIF_F_RXCSUM)
 132                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 133
 134         length = skb->len;
 135         if (dev_forward_skb(rcv, skb) != NET_RX_SUCCESS)
 136                 goto rx_drop;
 137
 138         u64_stats_update_begin(&stats->syncp);
 139         stats->tx_bytes += length;
 140         stats->tx_packets++;
 141         u64_stats_update_end(&stats->syncp);
 142
 143         u64_stats_update_begin(&rcv_stats->syncp);
 144         rcv_stats->rx_bytes += length;
 145         rcv_stats->rx_packets++;
 146         u64_stats_update_end(&rcv_stats->syncp);
 147
 148         return NETDEV_TX_OK;
 149
 150 rx_drop:
 151         u64_stats_update_begin(&rcv_stats->syncp);
 152         rcv_stats->rx_dropped++;
 153         u64_stats_update_end(&rcv_stats->syncp);
 154         return NETDEV_TX_OK;
 155 }
 156
 157 /*
 158  * general routines
 159  */
 160
 161 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
 162                                                   struct rtnl_link_stats64 *tot)
 163 {
 164         struct veth_priv *priv = netdev_priv(dev);
 165         int cpu;
 166
 167         for_each_possible_cpu(cpu) {
 168                 struct veth_net_stats *stats = per_cpu_ptr(priv->stats, cpu);
 169                 u64 rx_packets, rx_bytes, rx_dropped;
 170                 u64 tx_packets, tx_bytes;
 171                 unsigned int start;
 172
 173                 do {
 174                         start = u64_stats_fetch_begin_bh(&stats->syncp);
 175                         rx_packets = stats->rx_packets;
 176                         tx_packets = stats->tx_packets;
 177                         rx_bytes = stats->rx_bytes;
 178                         tx_bytes = stats->tx_bytes;
 179                         rx_dropped = stats->rx_dropped;
 180                 } while (u64_stats_fetch_retry_bh(&stats->syncp, start));
 181                 tot->rx_packets += rx_packets;
 182                 tot->tx_packets += tx_packets;
 183                 tot->rx_bytes   += rx_bytes;
 184                 tot->tx_bytes   += tx_bytes;
 185                 tot->rx_dropped += rx_dropped;
 186         }
 187
 188         return tot;
 189 }
 190
 191 static int veth_open(struct net_device *dev)
 192 {
 193         struct veth_priv *priv;
 194
 195         priv = netdev_priv(dev);
 196         if (priv->peer == NULL)
 197                 return -ENOTCONN;
 198
 199         if (priv->peer->flags & IFF_UP) {
 200                 netif_carrier_on(dev);
 201                 netif_carrier_on(priv->peer);
 202         }
 203         return 0;
 204 }
 205
 206 static int veth_close(struct net_device *dev)
 207 {
 208         struct veth_priv *priv = netdev_priv(dev);
 209
 210         netif_carrier_off(dev);
 211         netif_carrier_off(priv->peer);
 212
 213         return 0;
 214 }
 215
 216 static int is_valid_veth_mtu(int new_mtu)
 217 {
 218         return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
 219 }
 220
 221 static int veth_change_mtu(struct net_device *dev, int new_mtu)
 222 {
 223         if (!is_valid_veth_mtu(new_mtu))
 224                 return -EINVAL;
 225         dev->mtu = new_mtu;
 226         return 0;
 227 }
 228
 229 static int veth_dev_init(struct net_device *dev)
 230 {
 231         struct veth_net_stats __percpu *stats;
 232         struct veth_priv *priv;
 233
 234         stats = alloc_percpu(struct veth_net_stats);
 235         if (stats == NULL)
 236                 return -ENOMEM;
 237
 238         priv = netdev_priv(dev);
 239         priv->stats = stats;
 240         return 0;
 241 }
 242
 243 static void veth_dev_free(struct net_device *dev)
 244 {
 245         struct veth_priv *priv;
 246
 247         priv = netdev_priv(dev);
 248         free_percpu(priv->stats);
 249         free_netdev(dev);
 250 }
 251
 252 static const struct net_device_ops veth_netdev_ops = {
 253         .ndo_init            = veth_dev_init,
 254         .ndo_open            = veth_open,
 255         .ndo_stop            = veth_close,
 256         .ndo_start_xmit      = veth_xmit,
 257         .ndo_change_mtu      = veth_change_mtu,
 258         .ndo_get_stats64     = veth_get_stats64,
 259         .ndo_set_mac_address = eth_mac_addr,
 260 };
 261
 262 static void veth_setup(struct net_device *dev)
 263 {
 264         ether_setup(dev);
 265
 266         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
 267
 268         dev->netdev_ops = &veth_netdev_ops;
 269         dev->ethtool_ops = &veth_ethtool_ops;
 270         dev->features |= NETIF_F_LLTX;
 271         dev->destructor = veth_dev_free;
 272
 273         dev->hw_features = NETIF_F_HW_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
 274 }
 275
 276 /*
 277  * netlink interface
 278  */
 279
 280 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 281 {
 282         if (tb[IFLA_ADDRESS]) {
 283                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 284                         return -EINVAL;
 285                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 286                         return -EADDRNOTAVAIL;
 287         }
 288         if (tb[IFLA_MTU]) {
 289                 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
 290                         return -EINVAL;
 291         }
 292         return 0;
 293 }
 294
 295 static struct rtnl_link_ops veth_link_ops;
 296
 297 static int veth_newlink(struct net *src_net, struct net_device *dev,
 298                          struct nlattr *tb[], struct nlattr *data[])
 299 {
 300         int err;
 301         struct net_device *peer;
 302         struct veth_priv *priv;
 303         char ifname[IFNAMSIZ];
 304         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 305         struct ifinfomsg *ifmp;
 306         struct net *net;
 307
 308         /*
 309          * create and register peer first
 310          */
 311         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 312                 struct nlattr *nla_peer;
 313
 314                 nla_peer = data[VETH_INFO_PEER];
 315                 ifmp = nla_data(nla_peer);
 316                 err = nla_parse(peer_tb, IFLA_MAX,
 317                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 318                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 319                                 ifla_policy);
 320                 if (err < 0)
 321                         return err;
 322
 323                 err = veth_validate(peer_tb, NULL);
 324                 if (err < 0)
 325                         return err;
 326
 327                 tbp = peer_tb;
 328         } else {
 329                 ifmp = NULL;
 330                 tbp = tb;
 331         }
 332
 333         if (tbp[IFLA_IFNAME])
 334                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 335         else
 336                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 337
 338         net = rtnl_link_get_net(src_net, tbp);
 339         if (IS_ERR(net))
 340                 return PTR_ERR(net);
 341
 342         peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
 343         if (IS_ERR(peer)) {
 344                 put_net(net);
 345                 return PTR_ERR(peer);
 346         }
 347
 348         if (tbp[IFLA_ADDRESS] == NULL)
 349                 random_ether_addr(peer->dev_addr);
 350
 351         err = register_netdevice(peer);
 352         put_net(net);
 353         net = NULL;
 354         if (err < 0)
 355                 goto err_register_peer;
 356
 357         netif_carrier_off(peer);
 358
 359         err = rtnl_configure_link(peer, ifmp);
 360         if (err < 0)
 361                 goto err_configure_peer;
 362
 363         /*
 364          * register dev last
 365          *
 366          * note, that since we've registered new device the dev's name
 367          * should be re-allocated
 368          */
 369
 370         if (tb[IFLA_ADDRESS] == NULL)
 371                 random_ether_addr(dev->dev_addr);
 372
 373         if (tb[IFLA_IFNAME])
 374                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 375         else
 376                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 377
 378         if (strchr(dev->name, '%')) {
 379                 err = dev_alloc_name(dev, dev->name);
 380                 if (err < 0)
 381                         goto err_alloc_name;
 382         }
 383
 384         err = register_netdevice(dev);
 385         if (err < 0)
 386                 goto err_register_dev;
 387
 388         netif_carrier_off(dev);
 389
 390         /*
 391          * tie the deviced together
 392          */
 393
 394         priv = netdev_priv(dev);
 395         priv->peer = peer;
 396
 397         priv = netdev_priv(peer);
 398         priv->peer = dev;
 399         return 0;
 400
 401 err_register_dev:
 402         /* nothing to do */
 403 err_alloc_name:
 404 err_configure_peer:
 405         unregister_netdevice(peer);
 406         return err;
 407
 408 err_register_peer:
 409         free_netdev(peer);
 410         return err;
 411 }
 412
 413 static void veth_dellink(struct net_device *dev, struct list_head *head)
 414 {
 415         struct veth_priv *priv;
 416         struct net_device *peer;
 417
 418         priv = netdev_priv(dev);
 419         peer = priv->peer;
 420
 421         unregister_netdevice_queue(dev, head);
 422         unregister_netdevice_queue(peer, head);
 423 }
 424
 425 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1] = {
 426         [VETH_INFO_PEER]        = { .len = sizeof(struct ifinfomsg) },
 427 };
 428
 429 static struct rtnl_link_ops veth_link_ops = {
 430         .kind           = DRV_NAME,
 431         .priv_size      = sizeof(struct veth_priv),
 432         .setup          = veth_setup,
 433         .validate       = veth_validate,
 434         .newlink        = veth_newlink,
 435         .dellink        = veth_dellink,
 436         .policy         = veth_policy,
 437         .maxtype        = VETH_INFO_MAX,
 438 };
 439
 440 /*
 441  * init/fini
 442  */
 443
 444 static __init int veth_init(void)
 445 {
 446         return rtnl_link_register(&veth_link_ops);
 447 }
 448
 449 static __exit void veth_exit(void)
 450 {
 451         rtnl_link_unregister(&veth_link_ops);
 452 }
 453
 454 module_init(veth_init);
 455 module_exit(veth_exit);
 456
 457 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 458 MODULE_LICENSE("GPL v2");
 459 MODULE_ALIAS_RTNL_LINK(DRV_NAME);