drivers/net/macvlan.c

   1 /*
   2  * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public License as
   6  * published by the Free Software Foundation; either version 2 of
   7  * the License, or (at your option) any later version.
   8  *
   9  * The code this is based on carried the following copyright notice:
  10  * ---
  11  * (C) Copyright 2001-2006
  12  * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
  13  * Re-worked by Ben Greear <greearb@candelatech.com>
  14  * ---
  15  */
  16 #include <linux/kernel.h>
  17 #include <linux/types.h>
  18 #include <linux/module.h>
  19 #include <linux/init.h>
  20 #include <linux/errno.h>
  21 #include <linux/slab.h>
  22 #include <linux/string.h>
  23 #include <linux/rculist.h>
  24 #include <linux/notifier.h>
  25 #include <linux/netdevice.h>
  26 #include <linux/etherdevice.h>
  27 #include <linux/ethtool.h>
  28 #include <linux/if_arp.h>
  29 #include <linux/if_link.h>
  30 #include <linux/if_macvlan.h>
  31 #include <net/rtnetlink.h>
  32 #include <net/xfrm.h>
  33
  34 #define MACVLAN_HASH_SIZE       (1 << BITS_PER_BYTE)
  35
  36 struct macvlan_port {
  37         struct net_device       *dev;
  38         struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
  39         struct list_head        vlans;
  40 };
  41
  42 /**
  43  *      struct macvlan_rx_stats - MACVLAN percpu rx stats
  44  *      @rx_packets: number of received packets
  45  *      @rx_bytes: number of received bytes
  46  *      @multicast: number of received multicast packets
  47  *      @rx_errors: number of errors
  48  */
  49 struct macvlan_rx_stats {
  50         unsigned long rx_packets;
  51         unsigned long rx_bytes;
  52         unsigned long multicast;
  53         unsigned long rx_errors;
  54 };
  55
  56 struct macvlan_dev {
  57         struct net_device       *dev;
  58         struct list_head        list;
  59         struct hlist_node       hlist;
  60         struct macvlan_port     *port;
  61         struct net_device       *lowerdev;
  62         struct macvlan_rx_stats *rx_stats;
  63         enum macvlan_mode       mode;
  64 };
  65
  66
  67 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
  68                                                const unsigned char *addr)
  69 {
  70         struct macvlan_dev *vlan;
  71         struct hlist_node *n;
  72
  73         hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
  74                 if (!compare_ether_addr_64bits(vlan->dev->dev_addr, addr))
  75                         return vlan;
  76         }
  77         return NULL;
  78 }
  79
  80 static void macvlan_hash_add(struct macvlan_dev *vlan)
  81 {
  82         struct macvlan_port *port = vlan->port;
  83         const unsigned char *addr = vlan->dev->dev_addr;
  84
  85         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
  86 }
  87
  88 static void macvlan_hash_del(struct macvlan_dev *vlan)
  89 {
  90         hlist_del_rcu(&vlan->hlist);
  91         synchronize_rcu();
  92 }
  93
  94 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
  95                                         const unsigned char *addr)
  96 {
  97         macvlan_hash_del(vlan);
  98         /* Now that we are unhashed it is safe to change the device
  99          * address without confusing packet delivery.
 100          */
 101         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
 102         macvlan_hash_add(vlan);
 103 }
 104
 105 static int macvlan_addr_busy(const struct macvlan_port *port,
 106                                 const unsigned char *addr)
 107 {
 108         /* Test to see if the specified multicast address is
 109          * currently in use by the underlying device or
 110          * another macvlan.
 111          */
 112         if (!compare_ether_addr_64bits(port->dev->dev_addr, addr))
 113                 return 1;
 114
 115         if (macvlan_hash_lookup(port, addr))
 116                 return 1;
 117
 118         return 0;
 119 }
 120
 121 static inline void macvlan_count_rx(const struct macvlan_dev *vlan,
 122                                     unsigned int len, bool success,
 123                                     bool multicast)
 124 {
 125         struct macvlan_rx_stats *rx_stats;
 126
 127         rx_stats = per_cpu_ptr(vlan->rx_stats, smp_processor_id());
 128         if (likely(success)) {
 129                 rx_stats->rx_packets++;;
 130                 rx_stats->rx_bytes += len;
 131                 if (multicast)
 132                         rx_stats->multicast++;
 133         } else {
 134                 rx_stats->rx_errors++;
 135         }
 136 }
 137
 138 static int macvlan_broadcast_one(struct sk_buff *skb, struct net_device *dev,
 139                                  const struct ethhdr *eth, bool local)
 140 {
 141         if (!skb)
 142                 return NET_RX_DROP;
 143
 144         if (local)
 145                 return dev_forward_skb(dev, skb);
 146
 147         skb->dev = dev;
 148         if (!compare_ether_addr_64bits(eth->h_dest,
 149                                        dev->broadcast))
 150                 skb->pkt_type = PACKET_BROADCAST;
 151         else
 152                 skb->pkt_type = PACKET_MULTICAST;
 153
 154         return netif_rx(skb);
 155 }
 156
 157 static void macvlan_broadcast(struct sk_buff *skb,
 158                               const struct macvlan_port *port,
 159                               struct net_device *src,
 160                               enum macvlan_mode mode)
 161 {
 162         const struct ethhdr *eth = eth_hdr(skb);
 163         const struct macvlan_dev *vlan;
 164         struct hlist_node *n;
 165         struct sk_buff *nskb;
 166         unsigned int i;
 167         int err;
 168
 169         if (skb->protocol == htons(ETH_P_PAUSE))
 170                 return;
 171
 172         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
 173                 hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
 174                         if (vlan->dev == src || !(vlan->mode & mode))
 175                                 continue;
 176
 177                         nskb = skb_clone(skb, GFP_ATOMIC);
 178                         err = macvlan_broadcast_one(nskb, vlan->dev, eth,
 179                                          mode == MACVLAN_MODE_BRIDGE);
 180                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
 181                                          err == NET_RX_SUCCESS, 1);
 182                 }
 183         }
 184 }
 185
 186 /* called under rcu_read_lock() from netif_receive_skb */
 187 static struct sk_buff *macvlan_handle_frame(struct sk_buff *skb)
 188 {
 189         const struct ethhdr *eth = eth_hdr(skb);
 190         const struct macvlan_port *port;
 191         const struct macvlan_dev *vlan;
 192         const struct macvlan_dev *src;
 193         struct net_device *dev;
 194         unsigned int len;
 195
 196         port = rcu_dereference(skb->dev->macvlan_port);
 197         if (port == NULL)
 198                 return skb;
 199
 200         if (is_multicast_ether_addr(eth->h_dest)) {
 201                 src = macvlan_hash_lookup(port, eth->h_source);
 202                 if (!src)
 203                         /* frame comes from an external address */
 204                         macvlan_broadcast(skb, port, NULL,
 205                                           MACVLAN_MODE_PRIVATE |
 206                                           MACVLAN_MODE_VEPA    |
 207                                           MACVLAN_MODE_BRIDGE);
 208                 else if (src->mode == MACVLAN_MODE_VEPA)
 209                         /* flood to everyone except source */
 210                         macvlan_broadcast(skb, port, src->dev,
 211                                           MACVLAN_MODE_VEPA |
 212                                           MACVLAN_MODE_BRIDGE);
 213                 else if (src->mode == MACVLAN_MODE_BRIDGE)
 214                         /*
 215                          * flood only to VEPA ports, bridge ports
 216                          * already saw the frame on the way out.
 217                          */
 218                         macvlan_broadcast(skb, port, src->dev,
 219                                           MACVLAN_MODE_VEPA);
 220                 return skb;
 221         }
 222
 223         vlan = macvlan_hash_lookup(port, eth->h_dest);
 224         if (vlan == NULL)
 225                 return skb;
 226
 227         dev = vlan->dev;
 228         if (unlikely(!(dev->flags & IFF_UP))) {
 229                 kfree_skb(skb);
 230                 return NULL;
 231         }
 232         len = skb->len + ETH_HLEN;
 233         skb = skb_share_check(skb, GFP_ATOMIC);
 234         macvlan_count_rx(vlan, len, skb != NULL, 0);
 235         if (!skb)
 236                 return NULL;
 237
 238         skb->dev = dev;
 239         skb->pkt_type = PACKET_HOST;
 240
 241         netif_rx(skb);
 242         return NULL;
 243 }
 244
 245 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 246 {
 247         const struct macvlan_dev *vlan = netdev_priv(dev);
 248         const struct macvlan_port *port = vlan->port;
 249         const struct macvlan_dev *dest;
 250
 251         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
 252                 const struct ethhdr *eth = (void *)skb->data;
 253
 254                 /* send to other bridge ports directly */
 255                 if (is_multicast_ether_addr(eth->h_dest)) {
 256                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
 257                         goto xmit_world;
 258                 }
 259
 260                 dest = macvlan_hash_lookup(port, eth->h_dest);
 261                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
 262                         unsigned int length = skb->len + ETH_HLEN;
 263                         int ret = dev_forward_skb(dest->dev, skb);
 264                         macvlan_count_rx(dest, length,
 265                                          ret == NET_RX_SUCCESS, 0);
 266
 267                         return NET_XMIT_SUCCESS;
 268                 }
 269         }
 270
 271 xmit_world:
 272         skb->dev = vlan->lowerdev;
 273         return dev_queue_xmit(skb);
 274 }
 275
 276 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
 277                                       struct net_device *dev)
 278 {
 279         int i = skb_get_queue_mapping(skb);
 280         struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
 281         unsigned int len = skb->len;
 282         int ret;
 283
 284         ret = macvlan_queue_xmit(skb, dev);
 285         if (likely(ret == NET_XMIT_SUCCESS)) {
 286                 txq->tx_packets++;
 287                 txq->tx_bytes += len;
 288         } else
 289                 txq->tx_dropped++;
 290
 291         return ret;
 292 }
 293
 294 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 295                                unsigned short type, const void *daddr,
 296                                const void *saddr, unsigned len)
 297 {
 298         const struct macvlan_dev *vlan = netdev_priv(dev);
 299         struct net_device *lowerdev = vlan->lowerdev;
 300
 301         return dev_hard_header(skb, lowerdev, type, daddr,
 302                                saddr ? : dev->dev_addr, len);
 303 }
 304
 305 static const struct header_ops macvlan_hard_header_ops = {
 306         .create         = macvlan_hard_header,
 307         .rebuild        = eth_rebuild_header,
 308         .parse          = eth_header_parse,
 309         .cache          = eth_header_cache,
 310         .cache_update   = eth_header_cache_update,
 311 };
 312
 313 static int macvlan_open(struct net_device *dev)
 314 {
 315         struct macvlan_dev *vlan = netdev_priv(dev);
 316         struct net_device *lowerdev = vlan->lowerdev;
 317         int err;
 318
 319         err = -EBUSY;
 320         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
 321                 goto out;
 322
 323         err = dev_unicast_add(lowerdev, dev->dev_addr);
 324         if (err < 0)
 325                 goto out;
 326         if (dev->flags & IFF_ALLMULTI) {
 327                 err = dev_set_allmulti(lowerdev, 1);
 328                 if (err < 0)
 329                         goto del_unicast;
 330         }
 331         macvlan_hash_add(vlan);
 332         return 0;
 333
 334 del_unicast:
 335         dev_unicast_delete(lowerdev, dev->dev_addr);
 336 out:
 337         return err;
 338 }
 339
 340 static int macvlan_stop(struct net_device *dev)
 341 {
 342         struct macvlan_dev *vlan = netdev_priv(dev);
 343         struct net_device *lowerdev = vlan->lowerdev;
 344
 345         dev_mc_unsync(lowerdev, dev);
 346         if (dev->flags & IFF_ALLMULTI)
 347                 dev_set_allmulti(lowerdev, -1);
 348
 349         dev_unicast_delete(lowerdev, dev->dev_addr);
 350
 351         macvlan_hash_del(vlan);
 352         return 0;
 353 }
 354
 355 static int macvlan_set_mac_address(struct net_device *dev, void *p)
 356 {
 357         struct macvlan_dev *vlan = netdev_priv(dev);
 358         struct net_device *lowerdev = vlan->lowerdev;
 359         struct sockaddr *addr = p;
 360         int err;
 361
 362         if (!is_valid_ether_addr(addr->sa_data))
 363                 return -EADDRNOTAVAIL;
 364
 365         if (!(dev->flags & IFF_UP)) {
 366                 /* Just copy in the new address */
 367                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
 368         } else {
 369                 /* Rehash and update the device filters */
 370                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
 371                         return -EBUSY;
 372
 373                 err = dev_unicast_add(lowerdev, addr->sa_data);
 374                 if (err)
 375                         return err;
 376
 377                 dev_unicast_delete(lowerdev, dev->dev_addr);
 378
 379                 macvlan_hash_change_addr(vlan, addr->sa_data);
 380         }
 381         return 0;
 382 }
 383
 384 static void macvlan_change_rx_flags(struct net_device *dev, int change)
 385 {
 386         struct macvlan_dev *vlan = netdev_priv(dev);
 387         struct net_device *lowerdev = vlan->lowerdev;
 388
 389         if (change & IFF_ALLMULTI)
 390                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
 391 }
 392
 393 static void macvlan_set_multicast_list(struct net_device *dev)
 394 {
 395         struct macvlan_dev *vlan = netdev_priv(dev);
 396
 397         dev_mc_sync(vlan->lowerdev, dev);
 398 }
 399
 400 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 401 {
 402         struct macvlan_dev *vlan = netdev_priv(dev);
 403
 404         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
 405                 return -EINVAL;
 406         dev->mtu = new_mtu;
 407         return 0;
 408 }
 409
 410 /*
 411  * macvlan network devices have devices nesting below it and are a special
 412  * "super class" of normal network devices; split their locks off into a
 413  * separate class since they always nest.
 414  */
 415 static struct lock_class_key macvlan_netdev_xmit_lock_key;
 416 static struct lock_class_key macvlan_netdev_addr_lock_key;
 417
 418 #define MACVLAN_FEATURES \
 419         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
 420          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
 421          NETIF_F_TSO_ECN | NETIF_F_TSO6)
 422
 423 #define MACVLAN_STATE_MASK \
 424         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
 425
 426 static void macvlan_set_lockdep_class_one(struct net_device *dev,
 427                                           struct netdev_queue *txq,
 428                                           void *_unused)
 429 {
 430         lockdep_set_class(&txq->_xmit_lock,
 431                           &macvlan_netdev_xmit_lock_key);
 432 }
 433
 434 static void macvlan_set_lockdep_class(struct net_device *dev)
 435 {
 436         lockdep_set_class(&dev->addr_list_lock,
 437                           &macvlan_netdev_addr_lock_key);
 438         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
 439 }
 440
 441 static int macvlan_init(struct net_device *dev)
 442 {
 443         struct macvlan_dev *vlan = netdev_priv(dev);
 444         const struct net_device *lowerdev = vlan->lowerdev;
 445
 446         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
 447                                   (lowerdev->state & MACVLAN_STATE_MASK);
 448         dev->features           = lowerdev->features & MACVLAN_FEATURES;
 449         dev->gso_max_size       = lowerdev->gso_max_size;
 450         dev->iflink             = lowerdev->ifindex;
 451         dev->hard_header_len    = lowerdev->hard_header_len;
 452
 453         macvlan_set_lockdep_class(dev);
 454
 455         vlan->rx_stats = alloc_percpu(struct macvlan_rx_stats);
 456         if (!vlan->rx_stats)
 457                 return -ENOMEM;
 458
 459         return 0;
 460 }
 461
 462 static void macvlan_uninit(struct net_device *dev)
 463 {
 464         struct macvlan_dev *vlan = netdev_priv(dev);
 465
 466         free_percpu(vlan->rx_stats);
 467 }
 468
 469 static struct net_device_stats *macvlan_dev_get_stats(struct net_device *dev)
 470 {
 471         struct net_device_stats *stats = &dev->stats;
 472         struct macvlan_dev *vlan = netdev_priv(dev);
 473
 474         dev_txq_stats_fold(dev, stats);
 475
 476         if (vlan->rx_stats) {
 477                 struct macvlan_rx_stats *p, rx = {0};
 478                 int i;
 479
 480                 for_each_possible_cpu(i) {
 481                         p = per_cpu_ptr(vlan->rx_stats, i);
 482                         rx.rx_packets += p->rx_packets;
 483                         rx.rx_bytes   += p->rx_bytes;
 484                         rx.rx_errors  += p->rx_errors;
 485                         rx.multicast  += p->multicast;
 486                 }
 487                 stats->rx_packets = rx.rx_packets;
 488                 stats->rx_bytes   = rx.rx_bytes;
 489                 stats->rx_errors  = rx.rx_errors;
 490                 stats->rx_dropped = rx.rx_errors;
 491                 stats->multicast  = rx.multicast;
 492         }
 493         return stats;
 494 }
 495
 496 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
 497                                         struct ethtool_drvinfo *drvinfo)
 498 {
 499         snprintf(drvinfo->driver, 32, "macvlan");
 500         snprintf(drvinfo->version, 32, "0.1");
 501 }
 502
 503 static u32 macvlan_ethtool_get_rx_csum(struct net_device *dev)
 504 {
 505         const struct macvlan_dev *vlan = netdev_priv(dev);
 506         return dev_ethtool_get_rx_csum(vlan->lowerdev);
 507 }
 508
 509 static int macvlan_ethtool_get_settings(struct net_device *dev,
 510                                         struct ethtool_cmd *cmd)
 511 {
 512         const struct macvlan_dev *vlan = netdev_priv(dev);
 513         return dev_ethtool_get_settings(vlan->lowerdev, cmd);
 514 }
 515
 516 static u32 macvlan_ethtool_get_flags(struct net_device *dev)
 517 {
 518         const struct macvlan_dev *vlan = netdev_priv(dev);
 519         return dev_ethtool_get_flags(vlan->lowerdev);
 520 }
 521
 522 static const struct ethtool_ops macvlan_ethtool_ops = {
 523         .get_link               = ethtool_op_get_link,
 524         .get_settings           = macvlan_ethtool_get_settings,
 525         .get_rx_csum            = macvlan_ethtool_get_rx_csum,
 526         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
 527         .get_flags              = macvlan_ethtool_get_flags,
 528 };
 529
 530 static const struct net_device_ops macvlan_netdev_ops = {
 531         .ndo_init               = macvlan_init,
 532         .ndo_uninit             = macvlan_uninit,
 533         .ndo_open               = macvlan_open,
 534         .ndo_stop               = macvlan_stop,
 535         .ndo_start_xmit         = macvlan_start_xmit,
 536         .ndo_change_mtu         = macvlan_change_mtu,
 537         .ndo_change_rx_flags    = macvlan_change_rx_flags,
 538         .ndo_set_mac_address    = macvlan_set_mac_address,
 539         .ndo_set_multicast_list = macvlan_set_multicast_list,
 540         .ndo_get_stats          = macvlan_dev_get_stats,
 541         .ndo_validate_addr      = eth_validate_addr,
 542 };
 543
 544 static void macvlan_setup(struct net_device *dev)
 545 {
 546         ether_setup(dev);
 547
 548         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
 549         dev->netdev_ops         = &macvlan_netdev_ops;
 550         dev->destructor         = free_netdev;
 551         dev->header_ops         = &macvlan_hard_header_ops,
 552         dev->ethtool_ops        = &macvlan_ethtool_ops;
 553         dev->tx_queue_len       = 0;
 554 }
 555
 556 static int macvlan_port_create(struct net_device *dev)
 557 {
 558         struct macvlan_port *port;
 559         unsigned int i;
 560
 561         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
 562                 return -EINVAL;
 563
 564         port = kzalloc(sizeof(*port), GFP_KERNEL);
 565         if (port == NULL)
 566                 return -ENOMEM;
 567
 568         port->dev = dev;
 569         INIT_LIST_HEAD(&port->vlans);
 570         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
 571                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
 572         rcu_assign_pointer(dev->macvlan_port, port);
 573         return 0;
 574 }
 575
 576 static void macvlan_port_destroy(struct net_device *dev)
 577 {
 578         struct macvlan_port *port = dev->macvlan_port;
 579
 580         rcu_assign_pointer(dev->macvlan_port, NULL);
 581         synchronize_rcu();
 582         kfree(port);
 583 }
 584
 585 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
 586 {
 587         if (tb[IFLA_ADDRESS]) {
 588                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 589                         return -EINVAL;
 590                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 591                         return -EADDRNOTAVAIL;
 592         }
 593
 594         if (data && data[IFLA_MACVLAN_MODE]) {
 595                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
 596                 case MACVLAN_MODE_PRIVATE:
 597                 case MACVLAN_MODE_VEPA:
 598                 case MACVLAN_MODE_BRIDGE:
 599                         break;
 600                 default:
 601                         return -EINVAL;
 602                 }
 603         }
 604         return 0;
 605 }
 606
 607 static int macvlan_get_tx_queues(struct net *net,
 608                                  struct nlattr *tb[],
 609                                  unsigned int *num_tx_queues,
 610                                  unsigned int *real_num_tx_queues)
 611 {
 612         struct net_device *real_dev;
 613
 614         if (!tb[IFLA_LINK])
 615                 return -EINVAL;
 616
 617         real_dev = __dev_get_by_index(net, nla_get_u32(tb[IFLA_LINK]));
 618         if (!real_dev)
 619                 return -ENODEV;
 620
 621         *num_tx_queues      = real_dev->num_tx_queues;
 622         *real_num_tx_queues = real_dev->real_num_tx_queues;
 623         return 0;
 624 }
 625
 626 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
 627                            struct nlattr *tb[], struct nlattr *data[])
 628 {
 629         struct macvlan_dev *vlan = netdev_priv(dev);
 630         struct macvlan_port *port;
 631         struct net_device *lowerdev;
 632         int err;
 633
 634         if (!tb[IFLA_LINK])
 635                 return -EINVAL;
 636
 637         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
 638         if (lowerdev == NULL)
 639                 return -ENODEV;
 640
 641         /* When creating macvlans on top of other macvlans - use
 642          * the real device as the lowerdev.
 643          */
 644         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
 645                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
 646                 lowerdev = lowervlan->lowerdev;
 647         }
 648
 649         if (!tb[IFLA_MTU])
 650                 dev->mtu = lowerdev->mtu;
 651         else if (dev->mtu > lowerdev->mtu)
 652                 return -EINVAL;
 653
 654         if (!tb[IFLA_ADDRESS])
 655                 random_ether_addr(dev->dev_addr);
 656
 657         if (lowerdev->macvlan_port == NULL) {
 658                 err = macvlan_port_create(lowerdev);
 659                 if (err < 0)
 660                         return err;
 661         }
 662         port = lowerdev->macvlan_port;
 663
 664         vlan->lowerdev = lowerdev;
 665         vlan->dev      = dev;
 666         vlan->port     = port;
 667
 668         vlan->mode     = MACVLAN_MODE_VEPA;
 669         if (data && data[IFLA_MACVLAN_MODE])
 670                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 671
 672         err = register_netdevice(dev);
 673         if (err < 0)
 674                 return err;
 675
 676         list_add_tail(&vlan->list, &port->vlans);
 677         netif_stacked_transfer_operstate(lowerdev, dev);
 678         return 0;
 679 }
 680
 681 static void macvlan_dellink(struct net_device *dev, struct list_head *head)
 682 {
 683         struct macvlan_dev *vlan = netdev_priv(dev);
 684         struct macvlan_port *port = vlan->port;
 685
 686         list_del(&vlan->list);
 687         unregister_netdevice_queue(dev, head);
 688
 689         if (list_empty(&port->vlans))
 690                 macvlan_port_destroy(port->dev);
 691 }
 692
 693 static int macvlan_changelink(struct net_device *dev,
 694                 struct nlattr *tb[], struct nlattr *data[])
 695 {
 696         struct macvlan_dev *vlan = netdev_priv(dev);
 697         if (data && data[IFLA_MACVLAN_MODE])
 698                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 699         return 0;
 700 }
 701
 702 static size_t macvlan_get_size(const struct net_device *dev)
 703 {
 704         return nla_total_size(4);
 705 }
 706
 707 static int macvlan_fill_info(struct sk_buff *skb,
 708                                 const struct net_device *dev)
 709 {
 710         struct macvlan_dev *vlan = netdev_priv(dev);
 711
 712         NLA_PUT_U32(skb, IFLA_MACVLAN_MODE, vlan->mode);
 713         return 0;
 714
 715 nla_put_failure:
 716         return -EMSGSIZE;
 717 }
 718
 719 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
 720         [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
 721 };
 722
 723 static struct rtnl_link_ops macvlan_link_ops __read_mostly = {
 724         .kind           = "macvlan",
 725         .priv_size      = sizeof(struct macvlan_dev),
 726         .get_tx_queues  = macvlan_get_tx_queues,
 727         .setup          = macvlan_setup,
 728         .validate       = macvlan_validate,
 729         .newlink        = macvlan_newlink,
 730         .dellink        = macvlan_dellink,
 731         .maxtype        = IFLA_MACVLAN_MAX,
 732         .policy         = macvlan_policy,
 733         .changelink     = macvlan_changelink,
 734         .get_size       = macvlan_get_size,
 735         .fill_info      = macvlan_fill_info,
 736 };
 737
 738 static int macvlan_device_event(struct notifier_block *unused,
 739                                 unsigned long event, void *ptr)
 740 {
 741         struct net_device *dev = ptr;
 742         struct macvlan_dev *vlan, *next;
 743         struct macvlan_port *port;
 744
 745         port = dev->macvlan_port;
 746         if (port == NULL)
 747                 return NOTIFY_DONE;
 748
 749         switch (event) {
 750         case NETDEV_CHANGE:
 751                 list_for_each_entry(vlan, &port->vlans, list)
 752                         netif_stacked_transfer_operstate(vlan->lowerdev,
 753                                                          vlan->dev);
 754                 break;
 755         case NETDEV_FEAT_CHANGE:
 756                 list_for_each_entry(vlan, &port->vlans, list) {
 757                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
 758                         vlan->dev->gso_max_size = dev->gso_max_size;
 759                         netdev_features_change(vlan->dev);
 760                 }
 761                 break;
 762         case NETDEV_UNREGISTER:
 763                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
 764                         macvlan_dellink(vlan->dev, NULL);
 765                 break;
 766         }
 767         return NOTIFY_DONE;
 768 }
 769
 770 static struct notifier_block macvlan_notifier_block __read_mostly = {
 771         .notifier_call  = macvlan_device_event,
 772 };
 773
 774 static int __init macvlan_init_module(void)
 775 {
 776         int err;
 777
 778         register_netdevice_notifier(&macvlan_notifier_block);
 779         macvlan_handle_frame_hook = macvlan_handle_frame;
 780
 781         err = rtnl_link_register(&macvlan_link_ops);
 782         if (err < 0)
 783                 goto err1;
 784         return 0;
 785 err1:
 786         macvlan_handle_frame_hook = NULL;
 787         unregister_netdevice_notifier(&macvlan_notifier_block);
 788         return err;
 789 }
 790
 791 static void __exit macvlan_cleanup_module(void)
 792 {
 793         rtnl_link_unregister(&macvlan_link_ops);
 794         macvlan_handle_frame_hook = NULL;
 795         unregister_netdevice_notifier(&macvlan_notifier_block);
 796 }
 797
 798 module_init(macvlan_init_module);
 799 module_exit(macvlan_cleanup_module);
 800
 801 MODULE_LICENSE("GPL");
 802 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
 803 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
 804 MODULE_ALIAS_RTNL_LINK("macvlan");