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
  21 #define DRV_NAME        "veth"
  22 #define DRV_VERSION     "1.0"
  23
  24 #define MIN_MTU 68              /* Min L3 MTU */
  25 #define MAX_MTU 65535           /* Max L3 MTU (arbitrary) */
  26
  27 struct veth_net_stats {
  28         u64                     rx_packets;
  29         u64                     tx_packets;
  30         u64                     rx_bytes;
  31         u64                     tx_bytes;
  32         u64                     rx_dropped;
  33         u64                     tx_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 tx_drop:
 152         kfree_skb(skb);
 153         u64_stats_update_begin(&stats->syncp);
 154         stats->tx_dropped++;
 155         u64_stats_update_end(&stats->syncp);
 156         return NETDEV_TX_OK;
 157
 158 rx_drop:
 159         u64_stats_update_begin(&rcv_stats->syncp);
 160         rcv_stats->rx_dropped++;
 161         u64_stats_update_end(&rcv_stats->syncp);
 162         return NETDEV_TX_OK;
 163 }
 164
 165 /*
 166  * general routines
 167  */
 168
 169 static struct rtnl_link_stats64 *veth_get_stats64(struct net_device *dev,
 170                                                   struct rtnl_link_stats64 *tot)
 171 {
 172         struct veth_priv *priv = netdev_priv(dev);
 173         int cpu;
 174
 175         for_each_possible_cpu(cpu) {
 176                 struct veth_net_stats *stats = per_cpu_ptr(priv->stats, cpu);
 177                 u64 rx_packets, rx_bytes, rx_dropped;
 178                 u64 tx_packets, tx_bytes, tx_dropped;
 179                 unsigned int start;
 180
 181                 do {
 182                         start = u64_stats_fetch_begin_bh(&stats->syncp);
 183                         rx_packets = stats->rx_packets;
 184                         tx_packets = stats->tx_packets;
 185                         rx_bytes = stats->rx_bytes;
 186                         tx_bytes = stats->tx_bytes;
 187                         rx_dropped = stats->rx_dropped;
 188                         tx_dropped = stats->tx_dropped;
 189                 } while (u64_stats_fetch_retry_bh(&stats->syncp, start));
 190                 tot->rx_packets += rx_packets;
 191                 tot->tx_packets += tx_packets;
 192                 tot->rx_bytes   += rx_bytes;
 193                 tot->tx_bytes   += tx_bytes;
 194                 tot->rx_dropped += rx_dropped;
 195                 tot->tx_dropped += tx_dropped;
 196         }
 197
 198         return tot;
 199 }
 200
 201 static int veth_open(struct net_device *dev)
 202 {
 203         struct veth_priv *priv;
 204
 205         priv = netdev_priv(dev);
 206         if (priv->peer == NULL)
 207                 return -ENOTCONN;
 208
 209         if (priv->peer->flags & IFF_UP) {
 210                 netif_carrier_on(dev);
 211                 netif_carrier_on(priv->peer);
 212         }
 213         return 0;
 214 }
 215
 216 static int veth_close(struct net_device *dev)
 217 {
 218         struct veth_priv *priv = netdev_priv(dev);
 219
 220         netif_carrier_off(dev);
 221         netif_carrier_off(priv->peer);
 222
 223         return 0;
 224 }
 225
 226 static int is_valid_veth_mtu(int new_mtu)
 227 {
 228         return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
 229 }
 230
 231 static int veth_change_mtu(struct net_device *dev, int new_mtu)
 232 {
 233         if (!is_valid_veth_mtu(new_mtu))
 234                 return -EINVAL;
 235         dev->mtu = new_mtu;
 236         return 0;
 237 }
 238
 239 static int veth_dev_init(struct net_device *dev)
 240 {
 241         struct veth_net_stats __percpu *stats;
 242         struct veth_priv *priv;
 243
 244         stats = alloc_percpu(struct veth_net_stats);
 245         if (stats == NULL)
 246                 return -ENOMEM;
 247
 248         priv = netdev_priv(dev);
 249         priv->stats = stats;
 250         return 0;
 251 }
 252
 253 static void veth_dev_free(struct net_device *dev)
 254 {
 255         struct veth_priv *priv;
 256
 257         priv = netdev_priv(dev);
 258         free_percpu(priv->stats);
 259         free_netdev(dev);
 260 }
 261
 262 static const struct net_device_ops veth_netdev_ops = {
 263         .ndo_init            = veth_dev_init,
 264         .ndo_open            = veth_open,
 265         .ndo_stop            = veth_close,
 266         .ndo_start_xmit      = veth_xmit,
 267         .ndo_change_mtu      = veth_change_mtu,
 268         .ndo_get_stats64     = veth_get_stats64,
 269         .ndo_set_mac_address = eth_mac_addr,
 270 };
 271
 272 static void veth_setup(struct net_device *dev)
 273 {
 274         ether_setup(dev);
 275
 276         dev->netdev_ops = &veth_netdev_ops;
 277         dev->ethtool_ops = &veth_ethtool_ops;
 278         dev->features |= NETIF_F_LLTX;
 279         dev->destructor = veth_dev_free;
 280
 281         dev->hw_features = NETIF_F_NO_CSUM | NETIF_F_SG | NETIF_F_RXCSUM;
 282 }
 283
 284 /*
 285  * netlink interface
 286  */
 287
 288 static int veth_validate(struct nlattr *tb[], struct nlattr *data[])
 289 {
 290         if (tb[IFLA_ADDRESS]) {
 291                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 292                         return -EINVAL;
 293                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 294                         return -EADDRNOTAVAIL;
 295         }
 296         if (tb[IFLA_MTU]) {
 297                 if (!is_valid_veth_mtu(nla_get_u32(tb[IFLA_MTU])))
 298                         return -EINVAL;
 299         }
 300         return 0;
 301 }
 302
 303 static struct rtnl_link_ops veth_link_ops;
 304
 305 static int veth_newlink(struct net *src_net, struct net_device *dev,
 306                          struct nlattr *tb[], struct nlattr *data[])
 307 {
 308         int err;
 309         struct net_device *peer;
 310         struct veth_priv *priv;
 311         char ifname[IFNAMSIZ];
 312         struct nlattr *peer_tb[IFLA_MAX + 1], **tbp;
 313         struct ifinfomsg *ifmp;
 314         struct net *net;
 315
 316         /*
 317          * create and register peer first
 318          */
 319         if (data != NULL && data[VETH_INFO_PEER] != NULL) {
 320                 struct nlattr *nla_peer;
 321
 322                 nla_peer = data[VETH_INFO_PEER];
 323                 ifmp = nla_data(nla_peer);
 324                 err = nla_parse(peer_tb, IFLA_MAX,
 325                                 nla_data(nla_peer) + sizeof(struct ifinfomsg),
 326                                 nla_len(nla_peer) - sizeof(struct ifinfomsg),
 327                                 ifla_policy);
 328                 if (err < 0)
 329                         return err;
 330
 331                 err = veth_validate(peer_tb, NULL);
 332                 if (err < 0)
 333                         return err;
 334
 335                 tbp = peer_tb;
 336         } else {
 337                 ifmp = NULL;
 338                 tbp = tb;
 339         }
 340
 341         if (tbp[IFLA_IFNAME])
 342                 nla_strlcpy(ifname, tbp[IFLA_IFNAME], IFNAMSIZ);
 343         else
 344                 snprintf(ifname, IFNAMSIZ, DRV_NAME "%%d");
 345
 346         net = rtnl_link_get_net(src_net, tbp);
 347         if (IS_ERR(net))
 348                 return PTR_ERR(net);
 349
 350         peer = rtnl_create_link(src_net, net, ifname, &veth_link_ops, tbp);
 351         if (IS_ERR(peer)) {
 352                 put_net(net);
 353                 return PTR_ERR(peer);
 354         }
 355
 356         if (tbp[IFLA_ADDRESS] == NULL)
 357                 random_ether_addr(peer->dev_addr);
 358
 359         err = register_netdevice(peer);
 360         put_net(net);
 361         net = NULL;
 362         if (err < 0)
 363                 goto err_register_peer;
 364
 365         netif_carrier_off(peer);
 366
 367         err = rtnl_configure_link(peer, ifmp);
 368         if (err < 0)
 369                 goto err_configure_peer;
 370
 371         /*
 372          * register dev last
 373          *
 374          * note, that since we've registered new device the dev's name
 375          * should be re-allocated
 376          */
 377
 378         if (tb[IFLA_ADDRESS] == NULL)
 379                 random_ether_addr(dev->dev_addr);
 380
 381         if (tb[IFLA_IFNAME])
 382                 nla_strlcpy(dev->name, tb[IFLA_IFNAME], IFNAMSIZ);
 383         else
 384                 snprintf(dev->name, IFNAMSIZ, DRV_NAME "%%d");
 385
 386         if (strchr(dev->name, '%')) {
 387                 err = dev_alloc_name(dev, dev->name);
 388                 if (err < 0)
 389                         goto err_alloc_name;
 390         }
 391
 392         err = register_netdevice(dev);
 393         if (err < 0)
 394                 goto err_register_dev;
 395
 396         netif_carrier_off(dev);
 397
 398         /*
 399          * tie the deviced together
 400          */
 401
 402         priv = netdev_priv(dev);
 403         priv->peer = peer;
 404
 405         priv = netdev_priv(peer);
 406         priv->peer = dev;
 407         return 0;
 408
 409 err_register_dev:
 410         /* nothing to do */
 411 err_alloc_name:
 412 err_configure_peer:
 413         unregister_netdevice(peer);
 414         return err;
 415
 416 err_register_peer:
 417         free_netdev(peer);
 418         return err;
 419 }
 420
 421 static void veth_dellink(struct net_device *dev, struct list_head *head)
 422 {
 423         struct veth_priv *priv;
 424         struct net_device *peer;
 425
 426         priv = netdev_priv(dev);
 427         peer = priv->peer;
 428
 429         unregister_netdevice_queue(dev, head);
 430         unregister_netdevice_queue(peer, head);
 431 }
 432
 433 static const struct nla_policy veth_policy[VETH_INFO_MAX + 1];
 434
 435 static struct rtnl_link_ops veth_link_ops = {
 436         .kind           = DRV_NAME,
 437         .priv_size      = sizeof(struct veth_priv),
 438         .setup          = veth_setup,
 439         .validate       = veth_validate,
 440         .newlink        = veth_newlink,
 441         .dellink        = veth_dellink,
 442         .policy         = veth_policy,
 443         .maxtype        = VETH_INFO_MAX,
 444 };
 445
 446 /*
 447  * init/fini
 448  */
 449
 450 static __init int veth_init(void)
 451 {
 452         return rtnl_link_register(&veth_link_ops);
 453 }
 454
 455 static __exit void veth_exit(void)
 456 {
 457         rtnl_link_unregister(&veth_link_ops);
 458 }
 459
 460 module_init(veth_init);
 461 module_exit(veth_exit);
 462
 463 MODULE_DESCRIPTION("Virtual Ethernet Tunnel");
 464 MODULE_LICENSE("GPL v2");
 465 MODULE_ALIAS_RTNL_LINK(DRV_NAME);