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