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 <linux/hash.h>
  33 #include <net/rtnetlink.h>
  34 #include <net/xfrm.h>
  35
  36 #define MACVLAN_HASH_SIZE       (1 << BITS_PER_BYTE)
  37
  38 struct macvlan_port {
  39         struct net_device       *dev;
  40         struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
  41         struct list_head        vlans;
  42         struct rcu_head         rcu;
  43         bool                    passthru;
  44         int                     count;
  45 };
  46
  47 static void macvlan_port_destroy(struct net_device *dev);
  48
  49 #define macvlan_port_get_rcu(dev) \
  50         ((struct macvlan_port *) rcu_dereference(dev->rx_handler_data))
  51 #define macvlan_port_get(dev) ((struct macvlan_port *) dev->rx_handler_data)
  52 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
  53
  54 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
  55                                                const unsigned char *addr)
  56 {
  57         struct macvlan_dev *vlan;
  58         struct hlist_node *n;
  59
  60         hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
  61                 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
  62                         return vlan;
  63         }
  64         return NULL;
  65 }
  66
  67 static void macvlan_hash_add(struct macvlan_dev *vlan)
  68 {
  69         struct macvlan_port *port = vlan->port;
  70         const unsigned char *addr = vlan->dev->dev_addr;
  71
  72         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
  73 }
  74
  75 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
  76 {
  77         hlist_del_rcu(&vlan->hlist);
  78         if (sync)
  79                 synchronize_rcu();
  80 }
  81
  82 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
  83                                         const unsigned char *addr)
  84 {
  85         macvlan_hash_del(vlan, true);
  86         /* Now that we are unhashed it is safe to change the device
  87          * address without confusing packet delivery.
  88          */
  89         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
  90         macvlan_hash_add(vlan);
  91 }
  92
  93 static int macvlan_addr_busy(const struct macvlan_port *port,
  94                                 const unsigned char *addr)
  95 {
  96         /* Test to see if the specified multicast address is
  97          * currently in use by the underlying device or
  98          * another macvlan.
  99          */
 100         if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
 101                 return 1;
 102
 103         if (macvlan_hash_lookup(port, addr))
 104                 return 1;
 105
 106         return 0;
 107 }
 108
 109
 110 static int macvlan_broadcast_one(struct sk_buff *skb,
 111                                  const struct macvlan_dev *vlan,
 112                                  const struct ethhdr *eth, bool local)
 113 {
 114         struct net_device *dev = vlan->dev;
 115         if (!skb)
 116                 return NET_RX_DROP;
 117
 118         if (local)
 119                 return vlan->forward(dev, skb);
 120
 121         skb->dev = dev;
 122         if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
 123                 skb->pkt_type = PACKET_BROADCAST;
 124         else
 125                 skb->pkt_type = PACKET_MULTICAST;
 126
 127         return vlan->receive(skb);
 128 }
 129
 130 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
 131 {
 132         return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
 133 }
 134
 135
 136 static unsigned int mc_hash(const struct macvlan_dev *vlan,
 137                             const unsigned char *addr)
 138 {
 139         u32 val = __get_unaligned_cpu32(addr + 2);
 140
 141         val ^= macvlan_hash_mix(vlan);
 142         return hash_32(val, MACVLAN_MC_FILTER_BITS);
 143 }
 144
 145 static void macvlan_broadcast(struct sk_buff *skb,
 146                               const struct macvlan_port *port,
 147                               struct net_device *src,
 148                               enum macvlan_mode mode)
 149 {
 150         const struct ethhdr *eth = eth_hdr(skb);
 151         const struct macvlan_dev *vlan;
 152         struct hlist_node *n;
 153         struct sk_buff *nskb;
 154         unsigned int i;
 155         int err;
 156         unsigned int hash;
 157
 158         if (skb->protocol == htons(ETH_P_PAUSE))
 159                 return;
 160
 161         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
 162                 hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
 163                         if (vlan->dev == src || !(vlan->mode & mode))
 164                                 continue;
 165
 166                         hash = mc_hash(vlan, eth->h_dest);
 167                         if (!test_bit(hash, vlan->mc_filter))
 168                                 continue;
 169                         nskb = skb_clone(skb, GFP_ATOMIC);
 170                         err = macvlan_broadcast_one(nskb, vlan, eth,
 171                                          mode == MACVLAN_MODE_BRIDGE);
 172                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
 173                                          err == NET_RX_SUCCESS, 1);
 174                 }
 175         }
 176 }
 177
 178 /* called under rcu_read_lock() from netif_receive_skb */
 179 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
 180 {
 181         struct macvlan_port *port;
 182         struct sk_buff *skb = *pskb;
 183         const struct ethhdr *eth = eth_hdr(skb);
 184         const struct macvlan_dev *vlan;
 185         const struct macvlan_dev *src;
 186         struct net_device *dev;
 187         unsigned int len = 0;
 188         int ret = NET_RX_DROP;
 189
 190         port = macvlan_port_get_rcu(skb->dev);
 191         if (is_multicast_ether_addr(eth->h_dest)) {
 192                 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
 193                 if (!skb)
 194                         return RX_HANDLER_CONSUMED;
 195                 eth = eth_hdr(skb);
 196                 src = macvlan_hash_lookup(port, eth->h_source);
 197                 if (!src)
 198                         /* frame comes from an external address */
 199                         macvlan_broadcast(skb, port, NULL,
 200                                           MACVLAN_MODE_PRIVATE |
 201                                           MACVLAN_MODE_VEPA    |
 202                                           MACVLAN_MODE_PASSTHRU|
 203                                           MACVLAN_MODE_BRIDGE);
 204                 else if (src->mode == MACVLAN_MODE_VEPA)
 205                         /* flood to everyone except source */
 206                         macvlan_broadcast(skb, port, src->dev,
 207                                           MACVLAN_MODE_VEPA |
 208                                           MACVLAN_MODE_BRIDGE);
 209                 else if (src->mode == MACVLAN_MODE_BRIDGE)
 210                         /*
 211                          * flood only to VEPA ports, bridge ports
 212                          * already saw the frame on the way out.
 213                          */
 214                         macvlan_broadcast(skb, port, src->dev,
 215                                           MACVLAN_MODE_VEPA);
 216                 else {
 217                         /* forward to original port. */
 218                         vlan = src;
 219                         ret = macvlan_broadcast_one(skb, vlan, eth, 0);
 220                         goto out;
 221                 }
 222
 223                 return RX_HANDLER_PASS;
 224         }
 225
 226         if (port->passthru)
 227                 vlan = list_first_entry(&port->vlans, struct macvlan_dev, list);
 228         else
 229                 vlan = macvlan_hash_lookup(port, eth->h_dest);
 230         if (vlan == NULL)
 231                 return RX_HANDLER_PASS;
 232
 233         dev = vlan->dev;
 234         if (unlikely(!(dev->flags & IFF_UP))) {
 235                 kfree_skb(skb);
 236                 return RX_HANDLER_CONSUMED;
 237         }
 238         len = skb->len + ETH_HLEN;
 239         skb = skb_share_check(skb, GFP_ATOMIC);
 240         if (!skb)
 241                 goto out;
 242
 243         skb->dev = dev;
 244         skb->pkt_type = PACKET_HOST;
 245
 246         ret = vlan->receive(skb);
 247
 248 out:
 249         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
 250         return RX_HANDLER_CONSUMED;
 251 }
 252
 253 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 254 {
 255         const struct macvlan_dev *vlan = netdev_priv(dev);
 256         const struct macvlan_port *port = vlan->port;
 257         const struct macvlan_dev *dest;
 258         __u8 ip_summed = skb->ip_summed;
 259
 260         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
 261                 const struct ethhdr *eth = (void *)skb->data;
 262                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 263
 264                 /* send to other bridge ports directly */
 265                 if (is_multicast_ether_addr(eth->h_dest)) {
 266                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
 267                         goto xmit_world;
 268                 }
 269
 270                 dest = macvlan_hash_lookup(port, eth->h_dest);
 271                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
 272                         /* send to lowerdev first for its network taps */
 273                         dev_forward_skb(vlan->lowerdev, skb);
 274
 275                         return NET_XMIT_SUCCESS;
 276                 }
 277         }
 278
 279 xmit_world:
 280         skb->ip_summed = ip_summed;
 281         skb->dev = vlan->lowerdev;
 282         return dev_queue_xmit(skb);
 283 }
 284
 285 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
 286                                struct net_device *dev)
 287 {
 288         unsigned int len = skb->len;
 289         int ret;
 290         const struct macvlan_dev *vlan = netdev_priv(dev);
 291
 292         ret = macvlan_queue_xmit(skb, dev);
 293         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
 294                 struct macvlan_pcpu_stats *pcpu_stats;
 295
 296                 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
 297                 u64_stats_update_begin(&pcpu_stats->syncp);
 298                 pcpu_stats->tx_packets++;
 299                 pcpu_stats->tx_bytes += len;
 300                 u64_stats_update_end(&pcpu_stats->syncp);
 301         } else {
 302                 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
 303         }
 304         return ret;
 305 }
 306 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
 307
 308 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 309                                unsigned short type, const void *daddr,
 310                                const void *saddr, unsigned len)
 311 {
 312         const struct macvlan_dev *vlan = netdev_priv(dev);
 313         struct net_device *lowerdev = vlan->lowerdev;
 314
 315         return dev_hard_header(skb, lowerdev, type, daddr,
 316                                saddr ? : dev->dev_addr, len);
 317 }
 318
 319 static const struct header_ops macvlan_hard_header_ops = {
 320         .create         = macvlan_hard_header,
 321         .rebuild        = eth_rebuild_header,
 322         .parse          = eth_header_parse,
 323         .cache          = eth_header_cache,
 324         .cache_update   = eth_header_cache_update,
 325 };
 326
 327 static int macvlan_open(struct net_device *dev)
 328 {
 329         struct macvlan_dev *vlan = netdev_priv(dev);
 330         struct net_device *lowerdev = vlan->lowerdev;
 331         int err;
 332
 333         if (vlan->port->passthru) {
 334                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
 335                         dev_set_promiscuity(lowerdev, 1);
 336                 goto hash_add;
 337         }
 338
 339         err = -EBUSY;
 340         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
 341                 goto out;
 342
 343         err = dev_uc_add(lowerdev, dev->dev_addr);
 344         if (err < 0)
 345                 goto out;
 346         if (dev->flags & IFF_ALLMULTI) {
 347                 err = dev_set_allmulti(lowerdev, 1);
 348                 if (err < 0)
 349                         goto del_unicast;
 350         }
 351
 352 hash_add:
 353         macvlan_hash_add(vlan);
 354         return 0;
 355
 356 del_unicast:
 357         dev_uc_del(lowerdev, dev->dev_addr);
 358 out:
 359         return err;
 360 }
 361
 362 static int macvlan_stop(struct net_device *dev)
 363 {
 364         struct macvlan_dev *vlan = netdev_priv(dev);
 365         struct net_device *lowerdev = vlan->lowerdev;
 366
 367         dev_uc_unsync(lowerdev, dev);
 368         dev_mc_unsync(lowerdev, dev);
 369
 370         if (vlan->port->passthru) {
 371                 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
 372                         dev_set_promiscuity(lowerdev, -1);
 373                 goto hash_del;
 374         }
 375
 376         if (dev->flags & IFF_ALLMULTI)
 377                 dev_set_allmulti(lowerdev, -1);
 378
 379         dev_uc_del(lowerdev, dev->dev_addr);
 380
 381 hash_del:
 382         macvlan_hash_del(vlan, !dev->dismantle);
 383         return 0;
 384 }
 385
 386 static int macvlan_set_mac_address(struct net_device *dev, void *p)
 387 {
 388         struct macvlan_dev *vlan = netdev_priv(dev);
 389         struct net_device *lowerdev = vlan->lowerdev;
 390         struct sockaddr *addr = p;
 391         int err;
 392
 393         if (!is_valid_ether_addr(addr->sa_data))
 394                 return -EADDRNOTAVAIL;
 395
 396         if (!(dev->flags & IFF_UP)) {
 397                 /* Just copy in the new address */
 398                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
 399         } else {
 400                 /* Rehash and update the device filters */
 401                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
 402                         return -EBUSY;
 403
 404                 err = dev_uc_add(lowerdev, addr->sa_data);
 405                 if (err)
 406                         return err;
 407
 408                 dev_uc_del(lowerdev, dev->dev_addr);
 409
 410                 macvlan_hash_change_addr(vlan, addr->sa_data);
 411         }
 412         return 0;
 413 }
 414
 415 static void macvlan_change_rx_flags(struct net_device *dev, int change)
 416 {
 417         struct macvlan_dev *vlan = netdev_priv(dev);
 418         struct net_device *lowerdev = vlan->lowerdev;
 419
 420         if (change & IFF_ALLMULTI)
 421                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
 422 }
 423
 424 static void macvlan_set_mac_lists(struct net_device *dev)
 425 {
 426         struct macvlan_dev *vlan = netdev_priv(dev);
 427
 428         if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
 429                 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
 430         } else {
 431                 struct netdev_hw_addr *ha;
 432                 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
 433
 434                 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
 435                 netdev_for_each_mc_addr(ha, dev) {
 436                         __set_bit(mc_hash(vlan, ha->addr), filter);
 437                 }
 438
 439                 __set_bit(mc_hash(vlan, dev->broadcast), filter);
 440
 441                 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
 442         }
 443         dev_uc_sync(vlan->lowerdev, dev);
 444         dev_mc_sync(vlan->lowerdev, dev);
 445 }
 446
 447 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 448 {
 449         struct macvlan_dev *vlan = netdev_priv(dev);
 450
 451         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
 452                 return -EINVAL;
 453         dev->mtu = new_mtu;
 454         return 0;
 455 }
 456
 457 /*
 458  * macvlan network devices have devices nesting below it and are a special
 459  * "super class" of normal network devices; split their locks off into a
 460  * separate class since they always nest.
 461  */
 462 static struct lock_class_key macvlan_netdev_xmit_lock_key;
 463 static struct lock_class_key macvlan_netdev_addr_lock_key;
 464
 465 #define MACVLAN_FEATURES \
 466         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
 467          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
 468          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
 469          NETIF_F_HW_VLAN_FILTER)
 470
 471 #define MACVLAN_STATE_MASK \
 472         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
 473
 474 static void macvlan_set_lockdep_class_one(struct net_device *dev,
 475                                           struct netdev_queue *txq,
 476                                           void *_unused)
 477 {
 478         lockdep_set_class(&txq->_xmit_lock,
 479                           &macvlan_netdev_xmit_lock_key);
 480 }
 481
 482 static void macvlan_set_lockdep_class(struct net_device *dev)
 483 {
 484         lockdep_set_class(&dev->addr_list_lock,
 485                           &macvlan_netdev_addr_lock_key);
 486         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
 487 }
 488
 489 static int macvlan_init(struct net_device *dev)
 490 {
 491         struct macvlan_dev *vlan = netdev_priv(dev);
 492         const struct net_device *lowerdev = vlan->lowerdev;
 493
 494         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
 495                                   (lowerdev->state & MACVLAN_STATE_MASK);
 496         dev->features           = lowerdev->features & MACVLAN_FEATURES;
 497         dev->features           |= NETIF_F_LLTX;
 498         dev->gso_max_size       = lowerdev->gso_max_size;
 499         dev->iflink             = lowerdev->ifindex;
 500         dev->hard_header_len    = lowerdev->hard_header_len;
 501
 502         macvlan_set_lockdep_class(dev);
 503
 504         vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
 505         if (!vlan->pcpu_stats)
 506                 return -ENOMEM;
 507
 508         return 0;
 509 }
 510
 511 static void macvlan_uninit(struct net_device *dev)
 512 {
 513         struct macvlan_dev *vlan = netdev_priv(dev);
 514         struct macvlan_port *port = vlan->port;
 515
 516         free_percpu(vlan->pcpu_stats);
 517
 518         port->count -= 1;
 519         if (!port->count)
 520                 macvlan_port_destroy(port->dev);
 521 }
 522
 523 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
 524                                                          struct rtnl_link_stats64 *stats)
 525 {
 526         struct macvlan_dev *vlan = netdev_priv(dev);
 527
 528         if (vlan->pcpu_stats) {
 529                 struct macvlan_pcpu_stats *p;
 530                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
 531                 u32 rx_errors = 0, tx_dropped = 0;
 532                 unsigned int start;
 533                 int i;
 534
 535                 for_each_possible_cpu(i) {
 536                         p = per_cpu_ptr(vlan->pcpu_stats, i);
 537                         do {
 538                                 start = u64_stats_fetch_begin_bh(&p->syncp);
 539                                 rx_packets      = p->rx_packets;
 540                                 rx_bytes        = p->rx_bytes;
 541                                 rx_multicast    = p->rx_multicast;
 542                                 tx_packets      = p->tx_packets;
 543                                 tx_bytes        = p->tx_bytes;
 544                         } while (u64_stats_fetch_retry_bh(&p->syncp, start));
 545
 546                         stats->rx_packets       += rx_packets;
 547                         stats->rx_bytes         += rx_bytes;
 548                         stats->multicast        += rx_multicast;
 549                         stats->tx_packets       += tx_packets;
 550                         stats->tx_bytes         += tx_bytes;
 551                         /* rx_errors & tx_dropped are u32, updated
 552                          * without syncp protection.
 553                          */
 554                         rx_errors       += p->rx_errors;
 555                         tx_dropped      += p->tx_dropped;
 556                 }
 557                 stats->rx_errors        = rx_errors;
 558                 stats->rx_dropped       = rx_errors;
 559                 stats->tx_dropped       = tx_dropped;
 560         }
 561         return stats;
 562 }
 563
 564 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
 565                                     unsigned short vid)
 566 {
 567         struct macvlan_dev *vlan = netdev_priv(dev);
 568         struct net_device *lowerdev = vlan->lowerdev;
 569
 570         return vlan_vid_add(lowerdev, vid);
 571 }
 572
 573 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
 574                                      unsigned short vid)
 575 {
 576         struct macvlan_dev *vlan = netdev_priv(dev);
 577         struct net_device *lowerdev = vlan->lowerdev;
 578
 579         vlan_vid_del(lowerdev, vid);
 580         return 0;
 581 }
 582
 583 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 584                            struct net_device *dev,
 585                            const unsigned char *addr,
 586                            u16 flags)
 587 {
 588         struct macvlan_dev *vlan = netdev_priv(dev);
 589         int err = -EINVAL;
 590
 591         if (!vlan->port->passthru)
 592                 return -EOPNOTSUPP;
 593
 594         if (is_unicast_ether_addr(addr))
 595                 err = dev_uc_add_excl(dev, addr);
 596         else if (is_multicast_ether_addr(addr))
 597                 err = dev_mc_add_excl(dev, addr);
 598
 599         return err;
 600 }
 601
 602 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
 603                            struct net_device *dev,
 604                            const unsigned char *addr)
 605 {
 606         struct macvlan_dev *vlan = netdev_priv(dev);
 607         int err = -EINVAL;
 608
 609         if (!vlan->port->passthru)
 610                 return -EOPNOTSUPP;
 611
 612         if (is_unicast_ether_addr(addr))
 613                 err = dev_uc_del(dev, addr);
 614         else if (is_multicast_ether_addr(addr))
 615                 err = dev_mc_del(dev, addr);
 616
 617         return err;
 618 }
 619
 620 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
 621                                         struct ethtool_drvinfo *drvinfo)
 622 {
 623         strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
 624         strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
 625 }
 626
 627 static int macvlan_ethtool_get_settings(struct net_device *dev,
 628                                         struct ethtool_cmd *cmd)
 629 {
 630         const struct macvlan_dev *vlan = netdev_priv(dev);
 631
 632         return __ethtool_get_settings(vlan->lowerdev, cmd);
 633 }
 634
 635 static const struct ethtool_ops macvlan_ethtool_ops = {
 636         .get_link               = ethtool_op_get_link,
 637         .get_settings           = macvlan_ethtool_get_settings,
 638         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
 639 };
 640
 641 static const struct net_device_ops macvlan_netdev_ops = {
 642         .ndo_init               = macvlan_init,
 643         .ndo_uninit             = macvlan_uninit,
 644         .ndo_open               = macvlan_open,
 645         .ndo_stop               = macvlan_stop,
 646         .ndo_start_xmit         = macvlan_start_xmit,
 647         .ndo_change_mtu         = macvlan_change_mtu,
 648         .ndo_change_rx_flags    = macvlan_change_rx_flags,
 649         .ndo_set_mac_address    = macvlan_set_mac_address,
 650         .ndo_set_rx_mode        = macvlan_set_mac_lists,
 651         .ndo_get_stats64        = macvlan_dev_get_stats64,
 652         .ndo_validate_addr      = eth_validate_addr,
 653         .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
 654         .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
 655         .ndo_fdb_add            = macvlan_fdb_add,
 656         .ndo_fdb_del            = macvlan_fdb_del,
 657         .ndo_fdb_dump           = ndo_dflt_fdb_dump,
 658 };
 659
 660 void macvlan_common_setup(struct net_device *dev)
 661 {
 662         ether_setup(dev);
 663
 664         dev->priv_flags        &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
 665         dev->netdev_ops         = &macvlan_netdev_ops;
 666         dev->destructor         = free_netdev;
 667         dev->header_ops         = &macvlan_hard_header_ops,
 668         dev->ethtool_ops        = &macvlan_ethtool_ops;
 669 }
 670 EXPORT_SYMBOL_GPL(macvlan_common_setup);
 671
 672 static void macvlan_setup(struct net_device *dev)
 673 {
 674         macvlan_common_setup(dev);
 675         dev->tx_queue_len       = 0;
 676 }
 677
 678 static int macvlan_port_create(struct net_device *dev)
 679 {
 680         struct macvlan_port *port;
 681         unsigned int i;
 682         int err;
 683
 684         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
 685                 return -EINVAL;
 686
 687         port = kzalloc(sizeof(*port), GFP_KERNEL);
 688         if (port == NULL)
 689                 return -ENOMEM;
 690
 691         port->passthru = false;
 692         port->dev = dev;
 693         INIT_LIST_HEAD(&port->vlans);
 694         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
 695                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
 696
 697         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
 698         if (err)
 699                 kfree(port);
 700         else
 701                 dev->priv_flags |= IFF_MACVLAN_PORT;
 702         return err;
 703 }
 704
 705 static void macvlan_port_destroy(struct net_device *dev)
 706 {
 707         struct macvlan_port *port = macvlan_port_get(dev);
 708
 709         dev->priv_flags &= ~IFF_MACVLAN_PORT;
 710         netdev_rx_handler_unregister(dev);
 711         kfree_rcu(port, rcu);
 712 }
 713
 714 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
 715 {
 716         if (tb[IFLA_ADDRESS]) {
 717                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
 718                         return -EINVAL;
 719                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
 720                         return -EADDRNOTAVAIL;
 721         }
 722
 723         if (data && data[IFLA_MACVLAN_MODE]) {
 724                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
 725                 case MACVLAN_MODE_PRIVATE:
 726                 case MACVLAN_MODE_VEPA:
 727                 case MACVLAN_MODE_BRIDGE:
 728                 case MACVLAN_MODE_PASSTHRU:
 729                         break;
 730                 default:
 731                         return -EINVAL;
 732                 }
 733         }
 734         return 0;
 735 }
 736
 737 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
 738                            struct nlattr *tb[], struct nlattr *data[],
 739                            int (*receive)(struct sk_buff *skb),
 740                            int (*forward)(struct net_device *dev,
 741                                           struct sk_buff *skb))
 742 {
 743         struct macvlan_dev *vlan = netdev_priv(dev);
 744         struct macvlan_port *port;
 745         struct net_device *lowerdev;
 746         int err;
 747
 748         if (!tb[IFLA_LINK])
 749                 return -EINVAL;
 750
 751         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
 752         if (lowerdev == NULL)
 753                 return -ENODEV;
 754
 755         /* When creating macvlans on top of other macvlans - use
 756          * the real device as the lowerdev.
 757          */
 758         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
 759                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
 760                 lowerdev = lowervlan->lowerdev;
 761         }
 762
 763         if (!tb[IFLA_MTU])
 764                 dev->mtu = lowerdev->mtu;
 765         else if (dev->mtu > lowerdev->mtu)
 766                 return -EINVAL;
 767
 768         if (!tb[IFLA_ADDRESS])
 769                 eth_hw_addr_random(dev);
 770
 771         if (!macvlan_port_exists(lowerdev)) {
 772                 err = macvlan_port_create(lowerdev);
 773                 if (err < 0)
 774                         return err;
 775         }
 776         port = macvlan_port_get(lowerdev);
 777
 778         /* Only 1 macvlan device can be created in passthru mode */
 779         if (port->passthru)
 780                 return -EINVAL;
 781
 782         vlan->lowerdev = lowerdev;
 783         vlan->dev      = dev;
 784         vlan->port     = port;
 785         vlan->receive  = receive;
 786         vlan->forward  = forward;
 787
 788         vlan->mode     = MACVLAN_MODE_VEPA;
 789         if (data && data[IFLA_MACVLAN_MODE])
 790                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 791
 792         if (data && data[IFLA_MACVLAN_FLAGS])
 793                 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
 794
 795         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
 796                 if (port->count)
 797                         return -EINVAL;
 798                 port->passthru = true;
 799                 memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
 800         }
 801
 802         err = netdev_upper_dev_link(lowerdev, dev);
 803         if (err)
 804                 goto destroy_port;
 805
 806         port->count += 1;
 807         err = register_netdevice(dev);
 808         if (err < 0)
 809                 goto upper_dev_unlink;
 810
 811         list_add_tail(&vlan->list, &port->vlans);
 812         netif_stacked_transfer_operstate(lowerdev, dev);
 813
 814         return 0;
 815
 816 upper_dev_unlink:
 817         netdev_upper_dev_unlink(lowerdev, dev);
 818 destroy_port:
 819         port->count -= 1;
 820         if (!port->count)
 821                 macvlan_port_destroy(lowerdev);
 822
 823         return err;
 824 }
 825 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
 826
 827 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
 828                            struct nlattr *tb[], struct nlattr *data[])
 829 {
 830         return macvlan_common_newlink(src_net, dev, tb, data,
 831                                       netif_rx,
 832                                       dev_forward_skb);
 833 }
 834
 835 void macvlan_dellink(struct net_device *dev, struct list_head *head)
 836 {
 837         struct macvlan_dev *vlan = netdev_priv(dev);
 838
 839         list_del(&vlan->list);
 840         unregister_netdevice_queue(dev, head);
 841         netdev_upper_dev_unlink(vlan->lowerdev, dev);
 842 }
 843 EXPORT_SYMBOL_GPL(macvlan_dellink);
 844
 845 static int macvlan_changelink(struct net_device *dev,
 846                 struct nlattr *tb[], struct nlattr *data[])
 847 {
 848         struct macvlan_dev *vlan = netdev_priv(dev);
 849         if (data && data[IFLA_MACVLAN_MODE])
 850                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
 851         if (data && data[IFLA_MACVLAN_FLAGS]) {
 852                 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
 853                 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
 854
 855                 if (promisc && (flags & MACVLAN_FLAG_NOPROMISC))
 856                         dev_set_promiscuity(vlan->lowerdev, -1);
 857                 else if (promisc && !(flags & MACVLAN_FLAG_NOPROMISC))
 858                         dev_set_promiscuity(vlan->lowerdev, 1);
 859                 vlan->flags = flags;
 860         }
 861         return 0;
 862 }
 863
 864 static size_t macvlan_get_size(const struct net_device *dev)
 865 {
 866         return (0
 867                 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
 868                 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
 869                 );
 870 }
 871
 872 static int macvlan_fill_info(struct sk_buff *skb,
 873                                 const struct net_device *dev)
 874 {
 875         struct macvlan_dev *vlan = netdev_priv(dev);
 876
 877         if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
 878                 goto nla_put_failure;
 879         if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
 880                 goto nla_put_failure;
 881         return 0;
 882
 883 nla_put_failure:
 884         return -EMSGSIZE;
 885 }
 886
 887 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
 888         [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
 889         [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
 890 };
 891
 892 int macvlan_link_register(struct rtnl_link_ops *ops)
 893 {
 894         /* common fields */
 895         ops->priv_size          = sizeof(struct macvlan_dev);
 896         ops->validate           = macvlan_validate;
 897         ops->maxtype            = IFLA_MACVLAN_MAX;
 898         ops->policy             = macvlan_policy;
 899         ops->changelink         = macvlan_changelink;
 900         ops->get_size           = macvlan_get_size;
 901         ops->fill_info          = macvlan_fill_info;
 902
 903         return rtnl_link_register(ops);
 904 };
 905 EXPORT_SYMBOL_GPL(macvlan_link_register);
 906
 907 static struct rtnl_link_ops macvlan_link_ops = {
 908         .kind           = "macvlan",
 909         .setup          = macvlan_setup,
 910         .newlink        = macvlan_newlink,
 911         .dellink        = macvlan_dellink,
 912 };
 913
 914 static int macvlan_device_event(struct notifier_block *unused,
 915                                 unsigned long event, void *ptr)
 916 {
 917         struct net_device *dev = ptr;
 918         struct macvlan_dev *vlan, *next;
 919         struct macvlan_port *port;
 920         LIST_HEAD(list_kill);
 921
 922         if (!macvlan_port_exists(dev))
 923                 return NOTIFY_DONE;
 924
 925         port = macvlan_port_get(dev);
 926
 927         switch (event) {
 928         case NETDEV_CHANGE:
 929                 list_for_each_entry(vlan, &port->vlans, list)
 930                         netif_stacked_transfer_operstate(vlan->lowerdev,
 931                                                          vlan->dev);
 932                 break;
 933         case NETDEV_FEAT_CHANGE:
 934                 list_for_each_entry(vlan, &port->vlans, list) {
 935                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
 936                         vlan->dev->gso_max_size = dev->gso_max_size;
 937                         netdev_features_change(vlan->dev);
 938                 }
 939                 break;
 940         case NETDEV_UNREGISTER:
 941                 /* twiddle thumbs on netns device moves */
 942                 if (dev->reg_state != NETREG_UNREGISTERING)
 943                         break;
 944
 945                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
 946                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
 947                 unregister_netdevice_many(&list_kill);
 948                 list_del(&list_kill);
 949                 break;
 950         case NETDEV_PRE_TYPE_CHANGE:
 951                 /* Forbid underlaying device to change its type. */
 952                 return NOTIFY_BAD;
 953         }
 954         return NOTIFY_DONE;
 955 }
 956
 957 static struct notifier_block macvlan_notifier_block __read_mostly = {
 958         .notifier_call  = macvlan_device_event,
 959 };
 960
 961 static int __init macvlan_init_module(void)
 962 {
 963         int err;
 964
 965         register_netdevice_notifier(&macvlan_notifier_block);
 966
 967         err = macvlan_link_register(&macvlan_link_ops);
 968         if (err < 0)
 969                 goto err1;
 970         return 0;
 971 err1:
 972         unregister_netdevice_notifier(&macvlan_notifier_block);
 973         return err;
 974 }
 975
 976 static void __exit macvlan_cleanup_module(void)
 977 {
 978         rtnl_link_unregister(&macvlan_link_ops);
 979         unregister_netdevice_notifier(&macvlan_notifier_block);
 980 }
 981
 982 module_init(macvlan_init_module);
 983 module_exit(macvlan_cleanup_module);
 984
 985 MODULE_LICENSE("GPL");
 986 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
 987 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
 988 MODULE_ALIAS_RTNL_LINK("macvlan");