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_vlan.h>
  30 #include <linux/if_link.h>
  31 #include <linux/if_macvlan.h>
  32 #include <net/rtnetlink.h>
  33 #include <net/xfrm.h>
  34
  35 #define MACVLAN_HASH_SIZE       (1 << BITS_PER_BYTE)
  36
  37 struct macvlan_port {
  38         struct net_device       *dev;
  39         struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
  40         struct list_head        vlans;
  41         struct rcu_head         rcu;
  42         bool                    passthru;
  43         int                     count;
  44 };
  45
  46 static void macvlan_port_destroy(struct net_device *dev);
  47
  48 #define macvlan_port_get_rcu(dev) \
  49         ((struct macvlan_port *) rcu_dereference(dev->rx_handler_data))
  50 #define macvlan_port_get(dev) ((struct macvlan_port *) dev->rx_handler_data)
  51 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
  52
  53 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
  54                                                const unsigned char *addr)
  55 {
  56         struct macvlan_dev *vlan;
  57         struct hlist_node *n;
  58
  59         hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
  60                 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
  61                         return vlan;
  62         }
  63         return NULL;
  64 }
  65
  66 static void macvlan_hash_add(struct macvlan_dev *vlan)
  67 {
  68         struct macvlan_port *port = vlan->port;
  69         const unsigned char *addr = vlan->dev->dev_addr;
  70
  71         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
  72 }
  73
  74 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
  75 {
  76         hlist_del_rcu(&vlan->hlist);
  77         if (sync)
  78                 synchronize_rcu();
  79 }
  80
  81 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
  82                                         const unsigned char *addr)
  83 {
  84         macvlan_hash_del(vlan, true);
  85         /* Now that we are unhashed it is safe to change the device
  86          * address without confusing packet delivery.
  87          */
  88         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
  89         macvlan_hash_add(vlan);
  90 }
  91
  92 static int macvlan_addr_busy(const struct macvlan_port *port,
  93                                 const unsigned char *addr)
  94 {
  95         /* Test to see if the specified multicast address is
  96          * currently in use by the underlying device or
  97          * another macvlan.
  98          */
  99         if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
 100                 return 1;
 101
 102         if (macvlan_hash_lookup(port, addr))
 103                 return 1;
 104
 105         return 0;
 106 }
 107
 108
 109 static int macvlan_broadcast_one(struct sk_buff *skb,
 110                                  const struct macvlan_dev *vlan,
 111                                  const struct ethhdr *eth, bool local)
 112 {
 113         struct net_device *dev = vlan->dev;
 114         if (!skb)
 115                 return NET_RX_DROP;
 116
 117         if (local)
 118                 return vlan->forward(dev, skb);
 119
 120         skb->dev = dev;
 121         if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
 122                 skb->pkt_type = PACKET_BROADCAST;
 123         else
 124                 skb->pkt_type = PACKET_MULTICAST;
 125
 126         return vlan->receive(skb);
 127 }
 128
 129 static void macvlan_broadcast(struct sk_buff *skb,
 130                               const struct macvlan_port *port,
 131                               struct net_device *src,
 132                               enum macvlan_mode mode)
 133 {
 134         const struct ethhdr *eth = eth_hdr(skb);
 135         const struct macvlan_dev *vlan;
 136         struct hlist_node *n;
 137         struct sk_buff *nskb;
 138         unsigned int i;
 139         int err;
 140
 141         if (skb->protocol == htons(ETH_P_PAUSE))
 142                 return;
 143
 144         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
 145                 hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
 146                         if (vlan->dev == src || !(vlan->mode & mode))
 147                                 continue;
 148
 149                         nskb = skb_clone(skb, GFP_ATOMIC);
 150                         err = macvlan_broadcast_one(nskb, vlan, eth,
 151                                          mode == MACVLAN_MODE_BRIDGE);
 152                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
 153                                          err == NET_RX_SUCCESS, 1);
 154                 }
 155         }
 156 }
 157
 158 /* called under rcu_read_lock() from netif_receive_skb */
 159 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
 160 {
 161         struct macvlan_port *port;
 162         struct sk_buff *skb = *pskb;
 163         const struct ethhdr *eth = eth_hdr(skb);
 164         const struct macvlan_dev *vlan;
 165         const struct macvlan_dev *src;
 166         struct net_device *dev;
 167         unsigned int len = 0;
 168         int ret = NET_RX_DROP;
 169
 170         port = macvlan_port_get_rcu(skb->dev);
 171         if (is_multicast_ether_addr(eth->h_dest)) {
 172                 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
 173                 if (!skb)
 174                         return RX_HANDLER_CONSUMED;
 175                 eth = eth_hdr(skb);
 176                 src = macvlan_hash_lookup(port, eth->h_source);
 177                 if (!src)
 178                         /* frame comes from an external address */
 179                         macvlan_broadcast(skb, port, NULL,
 180                                           MACVLAN_MODE_PRIVATE |
 181                                           MACVLAN_MODE_VEPA    |
 182                                           MACVLAN_MODE_PASSTHRU|
 183                                           MACVLAN_MODE_BRIDGE);
 184                 else if (src->mode == MACVLAN_MODE_VEPA)
 185                         /* flood to everyone except source */
 186                         macvlan_broadcast(skb, port, src->dev,
 187                                           MACVLAN_MODE_VEPA |
 188                                           MACVLAN_MODE_BRIDGE);
 189                 else if (src->mode == MACVLAN_MODE_BRIDGE)
 190                         /*
 191                          * flood only to VEPA ports, bridge ports
 192                          * already saw the frame on the way out.
 193                          */
 194                         macvlan_broadcast(skb, port, src->dev,
 195                                           MACVLAN_MODE_VEPA);
 196                 else {
 197                         /* forward to original port. */
 198                         vlan = src;
 199                         ret = macvlan_broadcast_one(skb, vlan, eth, 0);
 200                         goto out;
 201                 }
 202
 203                 return RX_HANDLER_PASS;
 204         }
 205
 206         if (port->passthru)
 207                 vlan = list_first_entry(&port->vlans, struct macvlan_dev, list);
 208         else
 209                 vlan = macvlan_hash_lookup(port, eth->h_dest);
 210         if (vlan == NULL)
 211                 return RX_HANDLER_PASS;
 212
 213         dev = vlan->dev;
 214         if (unlikely(!(dev->flags & IFF_UP))) {
 215                 kfree_skb(skb);
 216                 return RX_HANDLER_CONSUMED;
 217         }
 218         len = skb->len + ETH_HLEN;
 219         skb = skb_share_check(skb, GFP_ATOMIC);
 220         if (!skb)
 221                 goto out;
 222
 223         skb->dev = dev;
 224         skb->pkt_type = PACKET_HOST;
 225
 226         ret = vlan->receive(skb);
 227
 228 out:
 229         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
 230         return RX_HANDLER_CONSUMED;
 231 }
 232
 233 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 234 {
 235         const struct macvlan_dev *vlan = netdev_priv(dev);
 236         const struct macvlan_port *port = vlan->port;
 237         const struct macvlan_dev *dest;
 238         __u8 ip_summed = skb->ip_summed;
 239
 240         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
 241                 const struct ethhdr *eth = (void *)skb->data;
 242                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 243
 244                 /* send to other bridge ports directly */
 245                 if (is_multicast_ether_addr(eth->h_dest)) {
 246                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
 247                         goto xmit_world;
 248                 }
 249
 250                 dest = macvlan_hash_lookup(port, eth->h_dest);
 251                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
 252                         /* send to lowerdev first for its network taps */
 253                         dev_forward_skb(vlan->lowerdev, skb);
 254
 255                         return NET_XMIT_SUCCESS;
 256                 }
 257         }
 258
 259 xmit_world:
 260         skb->ip_summed = ip_summed;
 261         skb->dev = vlan->lowerdev;
 262         return dev_queue_xmit(skb);
 263 }
 264
 265 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
 266                                struct net_device *dev)
 267 {
 268         unsigned int len = skb->len;
 269         int ret;
 270         const struct macvlan_dev *vlan = netdev_priv(dev);
 271
 272         ret = macvlan_queue_xmit(skb, dev);
 273         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
 274                 struct macvlan_pcpu_stats *pcpu_stats;
 275
 276                 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
 277                 u64_stats_update_begin(&pcpu_stats->syncp);
 278                 pcpu_stats->tx_packets++;
 279                 pcpu_stats->tx_bytes += len;
 280                 u64_stats_update_end(&pcpu_stats->syncp);
 281         } else {
 282                 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
 283         }
 284         return ret;
 285 }
 286 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
 287
 288 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 289                                unsigned short type, const void *daddr,
 290                                const void *saddr, unsigned len)
 291 {
 292         const struct macvlan_dev *vlan = netdev_priv(dev);
 293         struct net_device *lowerdev = vlan->lowerdev;
 294
 295         return dev_hard_header(skb, lowerdev, type, daddr,
 296                                saddr ? : dev->dev_addr, len);
 297 }
 298
 299 static const struct header_ops macvlan_hard_header_ops = {
 300         .create         = macvlan_hard_header,
 301         .rebuild        = eth_rebuild_header,
 302         .parse          = eth_header_parse,
 303         .cache          = eth_header_cache,
 304         .cache_update   = eth_header_cache_update,
 305 };
 306
 307 static int macvlan_open(struct net_device *dev)
 308 {
 309         struct macvlan_dev *vlan = netdev_priv(dev);
 310         struct net_device *lowerdev = vlan->lowerdev;
 311         int err;
 312
 313         if (vlan->port->passthru) {
 314                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
 315                         dev_set_promiscuity(lowerdev, 1);
 316                 goto hash_add;
 317         }
 318
 319         err = -EBUSY;
 320         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
 321                 goto out;
 322
 323         err = dev_uc_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
 332 hash_add:
 333         macvlan_hash_add(vlan);
 334         return 0;
 335
 336 del_unicast:
 337         dev_uc_del(lowerdev, dev->dev_addr);
 338 out:
 339         return err;
 340 }
 341
 342 static int macvlan_stop(struct net_device *dev)
 343 {
 344         struct macvlan_dev *vlan = netdev_priv(dev);
 345         struct net_device *lowerdev = vlan->lowerdev;
 346
 347         dev_uc_unsync(lowerdev, dev);
 348         dev_mc_unsync(lowerdev, dev);
 349
 350         if (vlan->port->passthru) {
 351                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
 352                         dev_set_promiscuity(lowerdev, -1);
 353                 goto hash_del;
 354         }
 355
 356         if (dev->flags & IFF_ALLMULTI)
 357                 dev_set_allmulti(lowerdev, -1);
 358
 359         dev_uc_del(lowerdev, dev->dev_addr);
 360
 361 hash_del:
 362         macvlan_hash_del(vlan, !dev->dismantle);
 363         return 0;
 364 }
 365
 366 static int macvlan_set_mac_address(struct net_device *dev, void *p)
 367 {
 368         struct macvlan_dev *vlan = netdev_priv(dev);
 369         struct net_device *lowerdev = vlan->lowerdev;
 370         struct sockaddr *addr = p;
 371         int err;
 372
 373         if (!is_valid_ether_addr(addr->sa_data))
 374                 return -EADDRNOTAVAIL;
 375
 376         if (!(dev->flags & IFF_UP)) {
 377                 /* Just copy in the new address */
 378                 dev->addr_assign_type &= ~NET_ADDR_RANDOM;
 379                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
 380         } else {
 381                 /* Rehash and update the device filters */
 382                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
 383                         return -EBUSY;
 384
 385                 err = dev_uc_add(lowerdev, addr->sa_data);
 386                 if (err)
 387                         return err;
 388
 389                 dev_uc_del(lowerdev, dev->dev_addr);
 390
 391                 macvlan_hash_change_addr(vlan, addr->sa_data);
 392         }
 393         return 0;
 394 }
 395
 396 static void macvlan_change_rx_flags(struct net_device *dev, int change)
 397 {
 398         struct macvlan_dev *vlan = netdev_priv(dev);
 399         struct net_device *lowerdev = vlan->lowerdev;
 400
 401         if (change & IFF_ALLMULTI)
 402                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
 403 }
 404
 405 static void macvlan_set_mac_lists(struct net_device *dev)
 406 {
 407         struct macvlan_dev *vlan = netdev_priv(dev);
 408
 409         dev_uc_sync(vlan->lowerdev, dev);
 410         dev_mc_sync(vlan->lowerdev, dev);
 411 }
 412
 413 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 414 {
 415         struct macvlan_dev *vlan = netdev_priv(dev);
 416
 417         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
 418                 return -EINVAL;
 419         dev->mtu = new_mtu;
 420         return 0;
 421 }
 422
 423 /*
 424  * macvlan network devices have devices nesting below it and are a special
 425  * "super class" of normal network devices; split their locks off into a
 426  * separate class since they always nest.
 427  */
 428 static struct lock_class_key macvlan_netdev_xmit_lock_key;
 429 static struct lock_class_key macvlan_netdev_addr_lock_key;
 430
 431 #define MACVLAN_FEATURES \
 432         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
 433          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
 434          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
 435          NETIF_F_HW_VLAN_FILTER)
 436
 437 #define MACVLAN_STATE_MASK \
 438         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
 439
 440 static void macvlan_set_lockdep_class_one(struct net_device *dev,
 441                                           struct netdev_queue *txq,
 442                                           void *_unused)
 443 {
 444         lockdep_set_class(&txq->_xmit_lock,
 445                           &macvlan_netdev_xmit_lock_key);
 446 }
 447
 448 static void macvlan_set_lockdep_class(struct net_device *dev)
 449 {
 450         lockdep_set_class(&dev->addr_list_lock,
 451                           &macvlan_netdev_addr_lock_key);
 452         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
 453 }
 454
 455 static int macvlan_init(struct net_device *dev)
 456 {
 457         struct macvlan_dev *vlan = netdev_priv(dev);
 458         const struct net_device *lowerdev = vlan->lowerdev;
 459
 460         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
 461                                   (lowerdev->state & MACVLAN_STATE_MASK);
 462         dev->features           = lowerdev->features & MACVLAN_FEATURES;
 463         dev->features           |= NETIF_F_LLTX;
 464         dev->gso_max_size       = lowerdev->gso_max_size;
 465         dev->iflink             = lowerdev->ifindex;
 466         dev->hard_header_len    = lowerdev->hard_header_len;
 467
 468         macvlan_set_lockdep_class(dev);
 469
 470         vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
 471         if (!vlan->pcpu_stats)
 472                 return -ENOMEM;
 473
 474         return 0;
 475 }
 476
 477 static void macvlan_uninit(struct net_device *dev)
 478 {
 479         struct macvlan_dev *vlan = netdev_priv(dev);
 480         struct macvlan_port *port = vlan->port;
 481
 482         free_percpu(vlan->pcpu_stats);
 483
 484         port->count -= 1;
 485         if (!port->count)
 486                 macvlan_port_destroy(port->dev);
 487 }
 488
 489 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
 490                                                          struct rtnl_link_stats64 *stats)
 491 {
 492         struct macvlan_dev *vlan = netdev_priv(dev);
 493
 494         if (vlan->pcpu_stats) {
 495                 struct macvlan_pcpu_stats *p;
 496                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
 497                 u32 rx_errors = 0, tx_dropped = 0;
 498                 unsigned int start;
 499                 int i;
 500
 501                 for_each_possible_cpu(i) {
 502                         p = per_cpu_ptr(vlan->pcpu_stats, i);
 503                         do {
 504                                 start = u64_stats_fetch_begin_bh(&p->syncp);
 505                                 rx_packets      = p->rx_packets;
 506                                 rx_bytes        = p->rx_bytes;
 507                                 rx_multicast    = p->rx_multicast;
 508                                 tx_packets      = p->tx_packets;
 509                                 tx_bytes        = p->tx_bytes;
 510                         } while (u64_stats_fetch_retry_bh(&p->syncp, start));
 511
 512                         stats->rx_packets       += rx_packets;
 513                         stats->rx_bytes         += rx_bytes;
 514                         stats->multicast        += rx_multicast;
 515                         stats->tx_packets       += tx_packets;
 516                         stats->tx_bytes         += tx_bytes;
 517                         /* rx_errors & tx_dropped are u32, updated
 518                          * without syncp protection.
 519                          */
 520                         rx_errors       += p->rx_errors;
 521                         tx_dropped      += p->tx_dropped;
 522                 }
 523                 stats->rx_errors        = rx_errors;
 524                 stats->rx_dropped       = rx_errors;
 525                 stats->tx_dropped       = tx_dropped;
 526         }
 527         return stats;
 528 }
 529
 530 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
 531                                     unsigned short vid)
 532 {
 533         struct macvlan_dev *vlan = netdev_priv(dev);
 534         struct net_device *lowerdev = vlan->lowerdev;
 535
 536         return vlan_vid_add(lowerdev, vid);
 537 }
 538
 539 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
 540                                      unsigned short vid)
 541 {
 542         struct macvlan_dev *vlan = netdev_priv(dev);
 543         struct net_device *lowerdev = vlan->lowerdev;
 544
 545         vlan_vid_del(lowerdev, vid);
 546         return 0;
 547 }
 548
 549 static int macvlan_fdb_add(struct ndmsg *ndm,
 550                            struct net_device *dev,
 551                            unsigned char *addr,
 552                            u16 flags)
 553 {
 554         struct macvlan_dev *vlan = netdev_priv(dev);
 555         int err = -EINVAL;
 556
 557         if (!vlan->port->passthru)
 558                 return -EOPNOTSUPP;
 559
 560         if (is_unicast_ether_addr(addr))
 561                 err = dev_uc_add_excl(dev, addr);
 562         else if (is_multicast_ether_addr(addr))
 563                 err = dev_mc_add_excl(dev, addr);
 564
 565         return err;
 566 }
 567
 568 static int macvlan_fdb_del(struct ndmsg *ndm,
 569                            struct net_device *dev,
 570                            unsigned char *addr)
 571 {
 572         struct macvlan_dev *vlan = netdev_priv(dev);
 573         int err = -EINVAL;
 574
 575         if (!vlan->port->passthru)
 576                 return -EOPNOTSUPP;
 577
 578         if (is_unicast_ether_addr(addr))
 579                 err = dev_uc_del(dev, addr);
 580         else if (is_multicast_ether_addr(addr))
 581                 err = dev_mc_del(dev, addr);
 582
 583         return err;
 584 }
 585
 586 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
 587                                         struct ethtool_drvinfo *drvinfo)
 588 {
 589         snprintf(drvinfo->driver, 32, "macvlan");
 590         snprintf(drvinfo->version, 32, "0.1");
 591 }
 592
 593 static int macvlan_ethtool_get_settings(struct net_device *dev,
 594                                         struct ethtool_cmd *cmd)
 595 {
 596         const struct macvlan_dev *vlan = netdev_priv(dev);
 597
 598         return __ethtool_get_settings(vlan->lowerdev, cmd);
 599 }
 600
 601 static const struct ethtool_ops macvlan_ethtool_ops = {
 602         .get_link               = ethtool_op_get_link,
 603         .get_settings           = macvlan_ethtool_get_settings,
 604         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
 605 };
 606
 607 static const struct net_device_ops macvlan_netdev_ops = {
 608         .ndo_init               = macvlan_init,
 609         .ndo_uninit             = macvlan_uninit,
 610         .ndo_open               = macvlan_open,
 611         .ndo_stop               = macvlan_stop,
 612         .ndo_start_xmit         = macvlan_start_xmit,
 613         .ndo_change_mtu         = macvlan_change_mtu,
 614         .ndo_change_rx_flags    = macvlan_change_rx_flags,
 615         .ndo_set_mac_address    = macvlan_set_mac_address,
 616         .ndo_set_rx_mode        = macvlan_set_mac_lists,
 617         .ndo_get_stats64        = macvlan_dev_get_stats64,
 618         .ndo_validate_addr      = eth_validate_addr,
 619         .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
 620         .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
 621         .ndo_fdb_add            = macvlan_fdb_add,
 622         .ndo_fdb_del            = macvlan_fdb_del,
 623         .ndo_fdb_dump           = ndo_dflt_fdb_dump,
 624 };
 625
 626 void macvlan_common_setup(struct net_device *dev)
 627 {
 628         ether_setup(dev);
 629
 630         dev->priv_flags        &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
 631         dev->netdev_ops         = &macvlan_netdev_ops;
 632         dev->destructor         = free_netdev;
 633         dev->header_ops         = &macvlan_hard_header_ops,
 634         dev->ethtool_ops        = &macvlan_ethtool_ops;
 635 }
 636 EXPORT_SYMBOL_GPL(macvlan_common_setup);
 637
 638 static void macvlan_setup(struct net_device *dev)
 639 {
 640         macvlan_common_setup(dev);
 641         dev->tx_queue_len       = 0;
 642 }
 643
 644 static int macvlan_port_create(struct net_device *dev)
 645 {
 646         struct macvlan_port *port;
 647         unsigned int i;
 648         int err;
 649
 650         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
 651                 return -EINVAL;
 652
 653         port = kzalloc(sizeof(*port), GFP_KERNEL);
 654         if (port == NULL)
 655                 return -ENOMEM;
 656
 657         port->passthru = false;
 658         port->dev = dev;
 659         INIT_LIST_HEAD(&port->vlans);
 660         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
 661                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
 662
 663         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
 664         if (err)
 665                 kfree(port);
 666         else
 667                 dev->priv_flags |= IFF_MACVLAN_PORT;
 668         return err;
 669 }
 670
 671 static void macvlan_port_destroy(struct net_device *dev)
 672 {
 673         struct macvlan_port *port = macvlan_port_get(dev);
 674
 675         dev->priv_flags &= ~IFF_MACVLAN_PORT;
 676         netdev_rx_handler_unregister(dev);
 677         kfree_rcu(port, rcu);
 678 }
 679
 680 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
 681 {
 682         if (tb[IFLA_ADDRESS]) {
 683                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 684                         return -EINVAL;
 685                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 686                         return -EADDRNOTAVAIL;
 687         }
 688
 689         if (data && data[IFLA_MACVLAN_MODE]) {
 690                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
 691                 case MACVLAN_MODE_PRIVATE:
 692                 case MACVLAN_MODE_VEPA:
 693                 case MACVLAN_MODE_BRIDGE:
 694                 case MACVLAN_MODE_PASSTHRU:
 695                         break;
 696                 default:
 697                         return -EINVAL;
 698                 }
 699         }
 700         return 0;
 701 }
 702
 703 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
 704                            struct nlattr *tb[], struct nlattr *data[],
 705                            int (*receive)(struct sk_buff *skb),
 706                            int (*forward)(struct net_device *dev,
 707                                           struct sk_buff *skb))
 708 {
 709         struct macvlan_dev *vlan = netdev_priv(dev);
 710         struct macvlan_port *port;
 711         struct net_device *lowerdev;
 712         int err;
 713
 714         if (!tb[IFLA_LINK])
 715                 return -EINVAL;
 716
 717         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
 718         if (lowerdev == NULL)
 719                 return -ENODEV;
 720
 721         /* When creating macvlans on top of other macvlans - use
 722          * the real device as the lowerdev.
 723          */
 724         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
 725                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
 726                 lowerdev = lowervlan->lowerdev;
 727         }
 728
 729         if (!tb[IFLA_MTU])
 730                 dev->mtu = lowerdev->mtu;
 731         else if (dev->mtu > lowerdev->mtu)
 732                 return -EINVAL;
 733
 734         if (!tb[IFLA_ADDRESS])
 735                 eth_hw_addr_random(dev);
 736
 737         if (!macvlan_port_exists(lowerdev)) {
 738                 err = macvlan_port_create(lowerdev);
 739                 if (err < 0)
 740                         return err;
 741         }
 742         port = macvlan_port_get(lowerdev);
 743
 744         /* Only 1 macvlan device can be created in passthru mode */
 745         if (port->passthru)
 746                 return -EINVAL;
 747
 748         vlan->lowerdev = lowerdev;
 749         vlan->dev      = dev;
 750         vlan->port     = port;
 751         vlan->receive  = receive;
 752         vlan->forward  = forward;
 753
 754         vlan->mode     = MACVLAN_MODE_VEPA;
 755         if (data && data[IFLA_MACVLAN_MODE])
 756                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 757
 758         if (data && data[IFLA_MACVLAN_FLAGS])
 759                 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
 760
 761         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
 762                 if (port->count)
 763                         return -EINVAL;
 764                 port->passthru = true;
 765                 memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
 766         }
 767
 768         port->count += 1;
 769         err = register_netdevice(dev);
 770         if (err < 0)
 771                 goto destroy_port;
 772
 773         list_add_tail(&vlan->list, &port->vlans);
 774         netif_stacked_transfer_operstate(lowerdev, dev);
 775
 776         return 0;
 777
 778 destroy_port:
 779         port->count -= 1;
 780         if (!port->count)
 781                 macvlan_port_destroy(lowerdev);
 782
 783         return err;
 784 }
 785 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
 786
 787 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
 788                            struct nlattr *tb[], struct nlattr *data[])
 789 {
 790         return macvlan_common_newlink(src_net, dev, tb, data,
 791                                       netif_rx,
 792                                       dev_forward_skb);
 793 }
 794
 795 void macvlan_dellink(struct net_device *dev, struct list_head *head)
 796 {
 797         struct macvlan_dev *vlan = netdev_priv(dev);
 798
 799         list_del(&vlan->list);
 800         unregister_netdevice_queue(dev, head);
 801 }
 802 EXPORT_SYMBOL_GPL(macvlan_dellink);
 803
 804 static int macvlan_changelink(struct net_device *dev,
 805                 struct nlattr *tb[], struct nlattr *data[])
 806 {
 807         struct macvlan_dev *vlan = netdev_priv(dev);
 808         if (data && data[IFLA_MACVLAN_MODE])
 809                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 810         if (data && data[IFLA_MACVLAN_FLAGS]) {
 811                 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
 812                 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
 813
 814                 if (promisc && (flags & MACVLAN_FLAG_NOPROMISC))
 815                         dev_set_promiscuity(vlan->lowerdev, -1);
 816                 else if (promisc && !(flags & MACVLAN_FLAG_NOPROMISC))
 817                         dev_set_promiscuity(vlan->lowerdev, 1);
 818                 vlan->flags = flags;
 819         }
 820         return 0;
 821 }
 822
 823 static size_t macvlan_get_size(const struct net_device *dev)
 824 {
 825         return nla_total_size(4);
 826 }
 827
 828 static int macvlan_fill_info(struct sk_buff *skb,
 829                                 const struct net_device *dev)
 830 {
 831         struct macvlan_dev *vlan = netdev_priv(dev);
 832
 833         if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
 834                 goto nla_put_failure;
 835         if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
 836                 goto nla_put_failure;
 837         return 0;
 838
 839 nla_put_failure:
 840         return -EMSGSIZE;
 841 }
 842
 843 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
 844         [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
 845         [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
 846 };
 847
 848 int macvlan_link_register(struct rtnl_link_ops *ops)
 849 {
 850         /* common fields */
 851         ops->priv_size          = sizeof(struct macvlan_dev);
 852         ops->validate           = macvlan_validate;
 853         ops->maxtype            = IFLA_MACVLAN_MAX;
 854         ops->policy             = macvlan_policy;
 855         ops->changelink         = macvlan_changelink;
 856         ops->get_size           = macvlan_get_size;
 857         ops->fill_info          = macvlan_fill_info;
 858
 859         return rtnl_link_register(ops);
 860 };
 861 EXPORT_SYMBOL_GPL(macvlan_link_register);
 862
 863 static struct rtnl_link_ops macvlan_link_ops = {
 864         .kind           = "macvlan",
 865         .setup          = macvlan_setup,
 866         .newlink        = macvlan_newlink,
 867         .dellink        = macvlan_dellink,
 868 };
 869
 870 static int macvlan_device_event(struct notifier_block *unused,
 871                                 unsigned long event, void *ptr)
 872 {
 873         struct net_device *dev = ptr;
 874         struct macvlan_dev *vlan, *next;
 875         struct macvlan_port *port;
 876         LIST_HEAD(list_kill);
 877
 878         if (!macvlan_port_exists(dev))
 879                 return NOTIFY_DONE;
 880
 881         port = macvlan_port_get(dev);
 882
 883         switch (event) {
 884         case NETDEV_CHANGE:
 885                 list_for_each_entry(vlan, &port->vlans, list)
 886                         netif_stacked_transfer_operstate(vlan->lowerdev,
 887                                                          vlan->dev);
 888                 break;
 889         case NETDEV_FEAT_CHANGE:
 890                 list_for_each_entry(vlan, &port->vlans, list) {
 891                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
 892                         vlan->dev->gso_max_size = dev->gso_max_size;
 893                         netdev_features_change(vlan->dev);
 894                 }
 895                 break;
 896         case NETDEV_UNREGISTER:
 897                 /* twiddle thumbs on netns device moves */
 898                 if (dev->reg_state != NETREG_UNREGISTERING)
 899                         break;
 900
 901                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
 902                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
 903                 unregister_netdevice_many(&list_kill);
 904                 list_del(&list_kill);
 905                 break;
 906         case NETDEV_PRE_TYPE_CHANGE:
 907                 /* Forbid underlaying device to change its type. */
 908                 return NOTIFY_BAD;
 909         }
 910         return NOTIFY_DONE;
 911 }
 912
 913 static struct notifier_block macvlan_notifier_block __read_mostly = {
 914         .notifier_call  = macvlan_device_event,
 915 };
 916
 917 static int __init macvlan_init_module(void)
 918 {
 919         int err;
 920
 921         register_netdevice_notifier(&macvlan_notifier_block);
 922
 923         err = macvlan_link_register(&macvlan_link_ops);
 924         if (err < 0)
 925                 goto err1;
 926         return 0;
 927 err1:
 928         unregister_netdevice_notifier(&macvlan_notifier_block);
 929         return err;
 930 }
 931
 932 static void __exit macvlan_cleanup_module(void)
 933 {
 934         rtnl_link_unregister(&macvlan_link_ops);
 935         unregister_netdevice_notifier(&macvlan_notifier_block);
 936 }
 937
 938 module_init(macvlan_init_module);
 939 module_exit(macvlan_cleanup_module);
 940
 941 MODULE_LICENSE("GPL");
 942 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
 943 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
 944 MODULE_ALIAS_RTNL_LINK("macvlan");