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         struct rcu_head         rcu;
  41 };
  42
  43 #define macvlan_port_get_rcu(dev) \
  44         ((struct macvlan_port *) rcu_dereference(dev->rx_handler_data))
  45 #define macvlan_port_get(dev) ((struct macvlan_port *) dev->rx_handler_data)
  46 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
  47
  48 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
  49                                                const unsigned char *addr)
  50 {
  51         struct macvlan_dev *vlan;
  52         struct hlist_node *n;
  53
  54         hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
  55                 if (!compare_ether_addr_64bits(vlan->dev->dev_addr, addr))
  56                         return vlan;
  57         }
  58         return NULL;
  59 }
  60
  61 static void macvlan_hash_add(struct macvlan_dev *vlan)
  62 {
  63         struct macvlan_port *port = vlan->port;
  64         const unsigned char *addr = vlan->dev->dev_addr;
  65
  66         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
  67 }
  68
  69 static void macvlan_hash_del(struct macvlan_dev *vlan)
  70 {
  71         hlist_del_rcu(&vlan->hlist);
  72         synchronize_rcu();
  73 }
  74
  75 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
  76                                         const unsigned char *addr)
  77 {
  78         macvlan_hash_del(vlan);
  79         /* Now that we are unhashed it is safe to change the device
  80          * address without confusing packet delivery.
  81          */
  82         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
  83         macvlan_hash_add(vlan);
  84 }
  85
  86 static int macvlan_addr_busy(const struct macvlan_port *port,
  87                                 const unsigned char *addr)
  88 {
  89         /* Test to see if the specified multicast address is
  90          * currently in use by the underlying device or
  91          * another macvlan.
  92          */
  93         if (!compare_ether_addr_64bits(port->dev->dev_addr, addr))
  94                 return 1;
  95
  96         if (macvlan_hash_lookup(port, addr))
  97                 return 1;
  98
  99         return 0;
 100 }
 101
 102
 103 static int macvlan_broadcast_one(struct sk_buff *skb,
 104                                  const struct macvlan_dev *vlan,
 105                                  const struct ethhdr *eth, bool local)
 106 {
 107         struct net_device *dev = vlan->dev;
 108         if (!skb)
 109                 return NET_RX_DROP;
 110
 111         if (local)
 112                 return vlan->forward(dev, skb);
 113
 114         skb->dev = dev;
 115         if (!compare_ether_addr_64bits(eth->h_dest,
 116                                        dev->broadcast))
 117                 skb->pkt_type = PACKET_BROADCAST;
 118         else
 119                 skb->pkt_type = PACKET_MULTICAST;
 120
 121         return vlan->receive(skb);
 122 }
 123
 124 static void macvlan_broadcast(struct sk_buff *skb,
 125                               const struct macvlan_port *port,
 126                               struct net_device *src,
 127                               enum macvlan_mode mode)
 128 {
 129         const struct ethhdr *eth = eth_hdr(skb);
 130         const struct macvlan_dev *vlan;
 131         struct hlist_node *n;
 132         struct sk_buff *nskb;
 133         unsigned int i;
 134         int err;
 135
 136         if (skb->protocol == htons(ETH_P_PAUSE))
 137                 return;
 138
 139         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
 140                 hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
 141                         if (vlan->dev == src || !(vlan->mode & mode))
 142                                 continue;
 143
 144                         nskb = skb_clone(skb, GFP_ATOMIC);
 145                         err = macvlan_broadcast_one(nskb, vlan, eth,
 146                                          mode == MACVLAN_MODE_BRIDGE);
 147                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
 148                                          err == NET_RX_SUCCESS, 1);
 149                 }
 150         }
 151 }
 152
 153 /* called under rcu_read_lock() from netif_receive_skb */
 154 static struct sk_buff *macvlan_handle_frame(struct sk_buff *skb)
 155 {
 156         struct macvlan_port *port;
 157         const struct ethhdr *eth = eth_hdr(skb);
 158         const struct macvlan_dev *vlan;
 159         const struct macvlan_dev *src;
 160         struct net_device *dev;
 161         unsigned int len;
 162
 163         port = macvlan_port_get_rcu(skb->dev);
 164         if (is_multicast_ether_addr(eth->h_dest)) {
 165                 src = macvlan_hash_lookup(port, eth->h_source);
 166                 if (!src)
 167                         /* frame comes from an external address */
 168                         macvlan_broadcast(skb, port, NULL,
 169                                           MACVLAN_MODE_PRIVATE |
 170                                           MACVLAN_MODE_VEPA    |
 171                                           MACVLAN_MODE_BRIDGE);
 172                 else if (src->mode == MACVLAN_MODE_VEPA)
 173                         /* flood to everyone except source */
 174                         macvlan_broadcast(skb, port, src->dev,
 175                                           MACVLAN_MODE_VEPA |
 176                                           MACVLAN_MODE_BRIDGE);
 177                 else if (src->mode == MACVLAN_MODE_BRIDGE)
 178                         /*
 179                          * flood only to VEPA ports, bridge ports
 180                          * already saw the frame on the way out.
 181                          */
 182                         macvlan_broadcast(skb, port, src->dev,
 183                                           MACVLAN_MODE_VEPA);
 184                 return skb;
 185         }
 186
 187         vlan = macvlan_hash_lookup(port, eth->h_dest);
 188         if (vlan == NULL)
 189                 return skb;
 190
 191         dev = vlan->dev;
 192         if (unlikely(!(dev->flags & IFF_UP))) {
 193                 kfree_skb(skb);
 194                 return NULL;
 195         }
 196         len = skb->len + ETH_HLEN;
 197         skb = skb_share_check(skb, GFP_ATOMIC);
 198         macvlan_count_rx(vlan, len, skb != NULL, 0);
 199         if (!skb)
 200                 return NULL;
 201
 202         skb->dev = dev;
 203         skb->pkt_type = PACKET_HOST;
 204
 205         vlan->receive(skb);
 206         return NULL;
 207 }
 208
 209 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 210 {
 211         const struct macvlan_dev *vlan = netdev_priv(dev);
 212         const struct macvlan_port *port = vlan->port;
 213         const struct macvlan_dev *dest;
 214
 215         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
 216                 const struct ethhdr *eth = (void *)skb->data;
 217
 218                 /* send to other bridge ports directly */
 219                 if (is_multicast_ether_addr(eth->h_dest)) {
 220                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
 221                         goto xmit_world;
 222                 }
 223
 224                 dest = macvlan_hash_lookup(port, eth->h_dest);
 225                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
 226                         unsigned int length = skb->len + ETH_HLEN;
 227                         int ret = dest->forward(dest->dev, skb);
 228                         macvlan_count_rx(dest, length,
 229                                          ret == NET_RX_SUCCESS, 0);
 230
 231                         return NET_XMIT_SUCCESS;
 232                 }
 233         }
 234
 235 xmit_world:
 236         skb_set_dev(skb, vlan->lowerdev);
 237         return dev_queue_xmit(skb);
 238 }
 239
 240 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
 241                                struct net_device *dev)
 242 {
 243         int i = skb_get_queue_mapping(skb);
 244         struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
 245         unsigned int len = skb->len;
 246         int ret;
 247
 248         ret = macvlan_queue_xmit(skb, dev);
 249         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
 250                 txq->tx_packets++;
 251                 txq->tx_bytes += len;
 252         } else
 253                 txq->tx_dropped++;
 254
 255         return ret;
 256 }
 257 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
 258
 259 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 260                                unsigned short type, const void *daddr,
 261                                const void *saddr, unsigned len)
 262 {
 263         const struct macvlan_dev *vlan = netdev_priv(dev);
 264         struct net_device *lowerdev = vlan->lowerdev;
 265
 266         return dev_hard_header(skb, lowerdev, type, daddr,
 267                                saddr ? : dev->dev_addr, len);
 268 }
 269
 270 static const struct header_ops macvlan_hard_header_ops = {
 271         .create         = macvlan_hard_header,
 272         .rebuild        = eth_rebuild_header,
 273         .parse          = eth_header_parse,
 274         .cache          = eth_header_cache,
 275         .cache_update   = eth_header_cache_update,
 276 };
 277
 278 static int macvlan_open(struct net_device *dev)
 279 {
 280         struct macvlan_dev *vlan = netdev_priv(dev);
 281         struct net_device *lowerdev = vlan->lowerdev;
 282         int err;
 283
 284         err = -EBUSY;
 285         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
 286                 goto out;
 287
 288         err = dev_uc_add(lowerdev, dev->dev_addr);
 289         if (err < 0)
 290                 goto out;
 291         if (dev->flags & IFF_ALLMULTI) {
 292                 err = dev_set_allmulti(lowerdev, 1);
 293                 if (err < 0)
 294                         goto del_unicast;
 295         }
 296         macvlan_hash_add(vlan);
 297         return 0;
 298
 299 del_unicast:
 300         dev_uc_del(lowerdev, dev->dev_addr);
 301 out:
 302         return err;
 303 }
 304
 305 static int macvlan_stop(struct net_device *dev)
 306 {
 307         struct macvlan_dev *vlan = netdev_priv(dev);
 308         struct net_device *lowerdev = vlan->lowerdev;
 309
 310         dev_mc_unsync(lowerdev, dev);
 311         if (dev->flags & IFF_ALLMULTI)
 312                 dev_set_allmulti(lowerdev, -1);
 313
 314         dev_uc_del(lowerdev, dev->dev_addr);
 315
 316         macvlan_hash_del(vlan);
 317         return 0;
 318 }
 319
 320 static int macvlan_set_mac_address(struct net_device *dev, void *p)
 321 {
 322         struct macvlan_dev *vlan = netdev_priv(dev);
 323         struct net_device *lowerdev = vlan->lowerdev;
 324         struct sockaddr *addr = p;
 325         int err;
 326
 327         if (!is_valid_ether_addr(addr->sa_data))
 328                 return -EADDRNOTAVAIL;
 329
 330         if (!(dev->flags & IFF_UP)) {
 331                 /* Just copy in the new address */
 332                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
 333         } else {
 334                 /* Rehash and update the device filters */
 335                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
 336                         return -EBUSY;
 337
 338                 err = dev_uc_add(lowerdev, addr->sa_data);
 339                 if (err)
 340                         return err;
 341
 342                 dev_uc_del(lowerdev, dev->dev_addr);
 343
 344                 macvlan_hash_change_addr(vlan, addr->sa_data);
 345         }
 346         return 0;
 347 }
 348
 349 static void macvlan_change_rx_flags(struct net_device *dev, int change)
 350 {
 351         struct macvlan_dev *vlan = netdev_priv(dev);
 352         struct net_device *lowerdev = vlan->lowerdev;
 353
 354         if (change & IFF_ALLMULTI)
 355                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
 356 }
 357
 358 static void macvlan_set_multicast_list(struct net_device *dev)
 359 {
 360         struct macvlan_dev *vlan = netdev_priv(dev);
 361
 362         dev_mc_sync(vlan->lowerdev, dev);
 363 }
 364
 365 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 366 {
 367         struct macvlan_dev *vlan = netdev_priv(dev);
 368
 369         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
 370                 return -EINVAL;
 371         dev->mtu = new_mtu;
 372         return 0;
 373 }
 374
 375 /*
 376  * macvlan network devices have devices nesting below it and are a special
 377  * "super class" of normal network devices; split their locks off into a
 378  * separate class since they always nest.
 379  */
 380 static struct lock_class_key macvlan_netdev_xmit_lock_key;
 381 static struct lock_class_key macvlan_netdev_addr_lock_key;
 382
 383 #define MACVLAN_FEATURES \
 384         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
 385          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
 386          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO)
 387
 388 #define MACVLAN_STATE_MASK \
 389         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
 390
 391 static void macvlan_set_lockdep_class_one(struct net_device *dev,
 392                                           struct netdev_queue *txq,
 393                                           void *_unused)
 394 {
 395         lockdep_set_class(&txq->_xmit_lock,
 396                           &macvlan_netdev_xmit_lock_key);
 397 }
 398
 399 static void macvlan_set_lockdep_class(struct net_device *dev)
 400 {
 401         lockdep_set_class(&dev->addr_list_lock,
 402                           &macvlan_netdev_addr_lock_key);
 403         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
 404 }
 405
 406 static int macvlan_init(struct net_device *dev)
 407 {
 408         struct macvlan_dev *vlan = netdev_priv(dev);
 409         const struct net_device *lowerdev = vlan->lowerdev;
 410
 411         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
 412                                   (lowerdev->state & MACVLAN_STATE_MASK);
 413         dev->features           = lowerdev->features & MACVLAN_FEATURES;
 414         dev->gso_max_size       = lowerdev->gso_max_size;
 415         dev->iflink             = lowerdev->ifindex;
 416         dev->hard_header_len    = lowerdev->hard_header_len;
 417
 418         macvlan_set_lockdep_class(dev);
 419
 420         vlan->rx_stats = alloc_percpu(struct macvlan_rx_stats);
 421         if (!vlan->rx_stats)
 422                 return -ENOMEM;
 423
 424         return 0;
 425 }
 426
 427 static void macvlan_uninit(struct net_device *dev)
 428 {
 429         struct macvlan_dev *vlan = netdev_priv(dev);
 430
 431         free_percpu(vlan->rx_stats);
 432 }
 433
 434 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
 435                                                          struct rtnl_link_stats64 *stats)
 436 {
 437         struct macvlan_dev *vlan = netdev_priv(dev);
 438
 439         dev_txq_stats_fold(dev, stats);
 440
 441         if (vlan->rx_stats) {
 442                 struct macvlan_rx_stats *p, accum = {0};
 443                 u64 rx_packets, rx_bytes, rx_multicast;
 444                 unsigned int start;
 445                 int i;
 446
 447                 for_each_possible_cpu(i) {
 448                         p = per_cpu_ptr(vlan->rx_stats, i);
 449                         do {
 450                                 start = u64_stats_fetch_begin_bh(&p->syncp);
 451                                 rx_packets      = p->rx_packets;
 452                                 rx_bytes        = p->rx_bytes;
 453                                 rx_multicast    = p->rx_multicast;
 454                         } while (u64_stats_fetch_retry_bh(&p->syncp, start));
 455                         accum.rx_packets        += rx_packets;
 456                         accum.rx_bytes          += rx_bytes;
 457                         accum.rx_multicast      += rx_multicast;
 458                         /* rx_errors is an ulong, updated without syncp protection */
 459                         accum.rx_errors         += p->rx_errors;
 460                 }
 461                 stats->rx_packets = accum.rx_packets;
 462                 stats->rx_bytes   = accum.rx_bytes;
 463                 stats->rx_errors  = accum.rx_errors;
 464                 stats->rx_dropped = accum.rx_errors;
 465                 stats->multicast  = accum.rx_multicast;
 466         }
 467         return stats;
 468 }
 469
 470 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
 471                                         struct ethtool_drvinfo *drvinfo)
 472 {
 473         snprintf(drvinfo->driver, 32, "macvlan");
 474         snprintf(drvinfo->version, 32, "0.1");
 475 }
 476
 477 static u32 macvlan_ethtool_get_rx_csum(struct net_device *dev)
 478 {
 479         const struct macvlan_dev *vlan = netdev_priv(dev);
 480         return dev_ethtool_get_rx_csum(vlan->lowerdev);
 481 }
 482
 483 static int macvlan_ethtool_get_settings(struct net_device *dev,
 484                                         struct ethtool_cmd *cmd)
 485 {
 486         const struct macvlan_dev *vlan = netdev_priv(dev);
 487         return dev_ethtool_get_settings(vlan->lowerdev, cmd);
 488 }
 489
 490 static u32 macvlan_ethtool_get_flags(struct net_device *dev)
 491 {
 492         const struct macvlan_dev *vlan = netdev_priv(dev);
 493         return dev_ethtool_get_flags(vlan->lowerdev);
 494 }
 495
 496 static const struct ethtool_ops macvlan_ethtool_ops = {
 497         .get_link               = ethtool_op_get_link,
 498         .get_settings           = macvlan_ethtool_get_settings,
 499         .get_rx_csum            = macvlan_ethtool_get_rx_csum,
 500         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
 501         .get_flags              = macvlan_ethtool_get_flags,
 502 };
 503
 504 static const struct net_device_ops macvlan_netdev_ops = {
 505         .ndo_init               = macvlan_init,
 506         .ndo_uninit             = macvlan_uninit,
 507         .ndo_open               = macvlan_open,
 508         .ndo_stop               = macvlan_stop,
 509         .ndo_start_xmit         = macvlan_start_xmit,
 510         .ndo_change_mtu         = macvlan_change_mtu,
 511         .ndo_change_rx_flags    = macvlan_change_rx_flags,
 512         .ndo_set_mac_address    = macvlan_set_mac_address,
 513         .ndo_set_multicast_list = macvlan_set_multicast_list,
 514         .ndo_get_stats64        = macvlan_dev_get_stats64,
 515         .ndo_validate_addr      = eth_validate_addr,
 516 };
 517
 518 static void macvlan_setup(struct net_device *dev)
 519 {
 520         ether_setup(dev);
 521
 522         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
 523         dev->netdev_ops         = &macvlan_netdev_ops;
 524         dev->destructor         = free_netdev;
 525         dev->header_ops         = &macvlan_hard_header_ops,
 526         dev->ethtool_ops        = &macvlan_ethtool_ops;
 527         dev->tx_queue_len       = 0;
 528 }
 529
 530 static int macvlan_port_create(struct net_device *dev)
 531 {
 532         struct macvlan_port *port;
 533         unsigned int i;
 534         int err;
 535
 536         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
 537                 return -EINVAL;
 538
 539         port = kzalloc(sizeof(*port), GFP_KERNEL);
 540         if (port == NULL)
 541                 return -ENOMEM;
 542
 543         port->dev = dev;
 544         INIT_LIST_HEAD(&port->vlans);
 545         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
 546                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
 547
 548         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
 549         if (err)
 550                 kfree(port);
 551
 552         dev->priv_flags |= IFF_MACVLAN_PORT;
 553         return err;
 554 }
 555
 556 static void macvlan_port_rcu_free(struct rcu_head *head)
 557 {
 558         struct macvlan_port *port;
 559
 560         port = container_of(head, struct macvlan_port, rcu);
 561         kfree(port);
 562 }
 563
 564 static void macvlan_port_destroy(struct net_device *dev)
 565 {
 566         struct macvlan_port *port = macvlan_port_get(dev);
 567
 568         dev->priv_flags &= ~IFF_MACVLAN_PORT;
 569         netdev_rx_handler_unregister(dev);
 570         call_rcu(&port->rcu, macvlan_port_rcu_free);
 571 }
 572
 573 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
 574 {
 575         if (tb[IFLA_ADDRESS]) {
 576                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 577                         return -EINVAL;
 578                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 579                         return -EADDRNOTAVAIL;
 580         }
 581
 582         if (data && data[IFLA_MACVLAN_MODE]) {
 583                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
 584                 case MACVLAN_MODE_PRIVATE:
 585                 case MACVLAN_MODE_VEPA:
 586                 case MACVLAN_MODE_BRIDGE:
 587                         break;
 588                 default:
 589                         return -EINVAL;
 590                 }
 591         }
 592         return 0;
 593 }
 594
 595 static int macvlan_get_tx_queues(struct net *net,
 596                                  struct nlattr *tb[],
 597                                  unsigned int *num_tx_queues,
 598                                  unsigned int *real_num_tx_queues)
 599 {
 600         struct net_device *real_dev;
 601
 602         if (!tb[IFLA_LINK])
 603                 return -EINVAL;
 604
 605         real_dev = __dev_get_by_index(net, nla_get_u32(tb[IFLA_LINK]));
 606         if (!real_dev)
 607                 return -ENODEV;
 608
 609         *num_tx_queues      = real_dev->num_tx_queues;
 610         *real_num_tx_queues = real_dev->real_num_tx_queues;
 611         return 0;
 612 }
 613
 614 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
 615                            struct nlattr *tb[], struct nlattr *data[],
 616                            int (*receive)(struct sk_buff *skb),
 617                            int (*forward)(struct net_device *dev,
 618                                           struct sk_buff *skb))
 619 {
 620         struct macvlan_dev *vlan = netdev_priv(dev);
 621         struct macvlan_port *port;
 622         struct net_device *lowerdev;
 623         int err;
 624
 625         if (!tb[IFLA_LINK])
 626                 return -EINVAL;
 627
 628         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
 629         if (lowerdev == NULL)
 630                 return -ENODEV;
 631
 632         /* When creating macvlans on top of other macvlans - use
 633          * the real device as the lowerdev.
 634          */
 635         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
 636                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
 637                 lowerdev = lowervlan->lowerdev;
 638         }
 639
 640         if (!tb[IFLA_MTU])
 641                 dev->mtu = lowerdev->mtu;
 642         else if (dev->mtu > lowerdev->mtu)
 643                 return -EINVAL;
 644
 645         if (!tb[IFLA_ADDRESS])
 646                 random_ether_addr(dev->dev_addr);
 647
 648         if (!macvlan_port_exists(lowerdev)) {
 649                 err = macvlan_port_create(lowerdev);
 650                 if (err < 0)
 651                         return err;
 652         }
 653         port = macvlan_port_get(lowerdev);
 654
 655         vlan->lowerdev = lowerdev;
 656         vlan->dev      = dev;
 657         vlan->port     = port;
 658         vlan->receive  = receive;
 659         vlan->forward  = forward;
 660
 661         vlan->mode     = MACVLAN_MODE_VEPA;
 662         if (data && data[IFLA_MACVLAN_MODE])
 663                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 664
 665         err = register_netdevice(dev);
 666         if (err < 0)
 667                 goto destroy_port;
 668
 669         list_add_tail(&vlan->list, &port->vlans);
 670         netif_stacked_transfer_operstate(lowerdev, dev);
 671
 672         return 0;
 673
 674 destroy_port:
 675         if (list_empty(&port->vlans))
 676                 macvlan_port_destroy(lowerdev);
 677
 678         return err;
 679 }
 680 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
 681
 682 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
 683                            struct nlattr *tb[], struct nlattr *data[])
 684 {
 685         return macvlan_common_newlink(src_net, dev, tb, data,
 686                                       netif_rx,
 687                                       dev_forward_skb);
 688 }
 689
 690 void macvlan_dellink(struct net_device *dev, struct list_head *head)
 691 {
 692         struct macvlan_dev *vlan = netdev_priv(dev);
 693         struct macvlan_port *port = vlan->port;
 694
 695         list_del(&vlan->list);
 696         unregister_netdevice_queue(dev, head);
 697
 698         if (list_empty(&port->vlans))
 699                 macvlan_port_destroy(port->dev);
 700 }
 701 EXPORT_SYMBOL_GPL(macvlan_dellink);
 702
 703 static int macvlan_changelink(struct net_device *dev,
 704                 struct nlattr *tb[], struct nlattr *data[])
 705 {
 706         struct macvlan_dev *vlan = netdev_priv(dev);
 707         if (data && data[IFLA_MACVLAN_MODE])
 708                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 709         return 0;
 710 }
 711
 712 static size_t macvlan_get_size(const struct net_device *dev)
 713 {
 714         return nla_total_size(4);
 715 }
 716
 717 static int macvlan_fill_info(struct sk_buff *skb,
 718                                 const struct net_device *dev)
 719 {
 720         struct macvlan_dev *vlan = netdev_priv(dev);
 721
 722         NLA_PUT_U32(skb, IFLA_MACVLAN_MODE, vlan->mode);
 723         return 0;
 724
 725 nla_put_failure:
 726         return -EMSGSIZE;
 727 }
 728
 729 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
 730         [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
 731 };
 732
 733 int macvlan_link_register(struct rtnl_link_ops *ops)
 734 {
 735         /* common fields */
 736         ops->priv_size          = sizeof(struct macvlan_dev);
 737         ops->get_tx_queues      = macvlan_get_tx_queues;
 738         ops->setup              = macvlan_setup;
 739         ops->validate           = macvlan_validate;
 740         ops->maxtype            = IFLA_MACVLAN_MAX;
 741         ops->policy             = macvlan_policy;
 742         ops->changelink         = macvlan_changelink;
 743         ops->get_size           = macvlan_get_size;
 744         ops->fill_info          = macvlan_fill_info;
 745
 746         return rtnl_link_register(ops);
 747 };
 748 EXPORT_SYMBOL_GPL(macvlan_link_register);
 749
 750 static struct rtnl_link_ops macvlan_link_ops = {
 751         .kind           = "macvlan",
 752         .newlink        = macvlan_newlink,
 753         .dellink        = macvlan_dellink,
 754 };
 755
 756 static int macvlan_device_event(struct notifier_block *unused,
 757                                 unsigned long event, void *ptr)
 758 {
 759         struct net_device *dev = ptr;
 760         struct macvlan_dev *vlan, *next;
 761         struct macvlan_port *port;
 762
 763         if (!macvlan_port_exists(dev))
 764                 return NOTIFY_DONE;
 765
 766         port = macvlan_port_get(dev);
 767
 768         switch (event) {
 769         case NETDEV_CHANGE:
 770                 list_for_each_entry(vlan, &port->vlans, list)
 771                         netif_stacked_transfer_operstate(vlan->lowerdev,
 772                                                          vlan->dev);
 773                 break;
 774         case NETDEV_FEAT_CHANGE:
 775                 list_for_each_entry(vlan, &port->vlans, list) {
 776                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
 777                         vlan->dev->gso_max_size = dev->gso_max_size;
 778                         netdev_features_change(vlan->dev);
 779                 }
 780                 break;
 781         case NETDEV_UNREGISTER:
 782                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
 783                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, NULL);
 784                 break;
 785         case NETDEV_PRE_TYPE_CHANGE:
 786                 /* Forbid underlaying device to change its type. */
 787                 return NOTIFY_BAD;
 788         }
 789         return NOTIFY_DONE;
 790 }
 791
 792 static struct notifier_block macvlan_notifier_block __read_mostly = {
 793         .notifier_call  = macvlan_device_event,
 794 };
 795
 796 static int __init macvlan_init_module(void)
 797 {
 798         int err;
 799
 800         register_netdevice_notifier(&macvlan_notifier_block);
 801
 802         err = macvlan_link_register(&macvlan_link_ops);
 803         if (err < 0)
 804                 goto err1;
 805         return 0;
 806 err1:
 807         unregister_netdevice_notifier(&macvlan_notifier_block);
 808         return err;
 809 }
 810
 811 static void __exit macvlan_cleanup_module(void)
 812 {
 813         rtnl_link_unregister(&macvlan_link_ops);
 814         unregister_netdevice_notifier(&macvlan_notifier_block);
 815 }
 816
 817 module_init(macvlan_init_module);
 818 module_exit(macvlan_cleanup_module);
 819
 820 MODULE_LICENSE("GPL");
 821 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
 822 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
 823 MODULE_ALIAS_RTNL_LINK("macvlan");