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 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 50 { 51 return rcu_dereference(dev->rx_handler_data); 52 } 53 54 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 55 { 56 return rtnl_dereference(dev->rx_handler_data); 57 } 58 59 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT) 60 61 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 62 const unsigned char *addr) 63 { 64 struct macvlan_dev *vlan; 65 66 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) { 67 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 68 return vlan; 69 } 70 return NULL; 71 } 72 73 static void macvlan_hash_add(struct macvlan_dev *vlan) 74 { 75 struct macvlan_port *port = vlan->port; 76 const unsigned char *addr = vlan->dev->dev_addr; 77 78 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]); 79 } 80 81 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 82 { 83 hlist_del_rcu(&vlan->hlist); 84 if (sync) 85 synchronize_rcu(); 86 } 87 88 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 89 const unsigned char *addr) 90 { 91 macvlan_hash_del(vlan, true); 92 /* Now that we are unhashed it is safe to change the device 93 * address without confusing packet delivery. 94 */ 95 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 96 macvlan_hash_add(vlan); 97 } 98 99 static int macvlan_addr_busy(const struct macvlan_port *port, 100 const unsigned char *addr) 101 { 102 /* Test to see if the specified multicast address is 103 * currently in use by the underlying device or 104 * another macvlan. 105 */ 106 if (ether_addr_equal_64bits(port->dev->dev_addr, addr)) 107 return 1; 108 109 if (macvlan_hash_lookup(port, addr)) 110 return 1; 111 112 return 0; 113 } 114 115 116 static int macvlan_broadcast_one(struct sk_buff *skb, 117 const struct macvlan_dev *vlan, 118 const struct ethhdr *eth, bool local) 119 { 120 struct net_device *dev = vlan->dev; 121 if (!skb) 122 return NET_RX_DROP; 123 124 if (local) 125 return vlan->forward(dev, skb); 126 127 skb->dev = dev; 128 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 129 skb->pkt_type = PACKET_BROADCAST; 130 else 131 skb->pkt_type = PACKET_MULTICAST; 132 133 return vlan->receive(skb); 134 } 135 136 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 137 { 138 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 139 } 140 141 142 static unsigned int mc_hash(const struct macvlan_dev *vlan, 143 const unsigned char *addr) 144 { 145 u32 val = __get_unaligned_cpu32(addr + 2); 146 147 val ^= macvlan_hash_mix(vlan); 148 return hash_32(val, MACVLAN_MC_FILTER_BITS); 149 } 150 151 static void macvlan_broadcast(struct sk_buff *skb, 152 const struct macvlan_port *port, 153 struct net_device *src, 154 enum macvlan_mode mode) 155 { 156 const struct ethhdr *eth = eth_hdr(skb); 157 const struct macvlan_dev *vlan; 158 struct sk_buff *nskb; 159 unsigned int i; 160 int err; 161 unsigned int hash; 162 163 if (skb->protocol == htons(ETH_P_PAUSE)) 164 return; 165 166 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 167 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 168 if (vlan->dev == src || !(vlan->mode & mode)) 169 continue; 170 171 hash = mc_hash(vlan, eth->h_dest); 172 if (!test_bit(hash, vlan->mc_filter)) 173 continue; 174 nskb = skb_clone(skb, GFP_ATOMIC); 175 err = macvlan_broadcast_one(nskb, vlan, eth, 176 mode == MACVLAN_MODE_BRIDGE); 177 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 178 err == NET_RX_SUCCESS, 1); 179 } 180 } 181 } 182 183 /* called under rcu_read_lock() from netif_receive_skb */ 184 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 185 { 186 struct macvlan_port *port; 187 struct sk_buff *skb = *pskb; 188 const struct ethhdr *eth = eth_hdr(skb); 189 const struct macvlan_dev *vlan; 190 const struct macvlan_dev *src; 191 struct net_device *dev; 192 unsigned int len = 0; 193 int ret = NET_RX_DROP; 194 195 port = macvlan_port_get_rcu(skb->dev); 196 if (is_multicast_ether_addr(eth->h_dest)) { 197 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN); 198 if (!skb) 199 return RX_HANDLER_CONSUMED; 200 eth = eth_hdr(skb); 201 src = macvlan_hash_lookup(port, eth->h_source); 202 if (!src) 203 /* frame comes from an external address */ 204 macvlan_broadcast(skb, port, NULL, 205 MACVLAN_MODE_PRIVATE | 206 MACVLAN_MODE_VEPA | 207 MACVLAN_MODE_PASSTHRU| 208 MACVLAN_MODE_BRIDGE); 209 else if (src->mode == MACVLAN_MODE_VEPA) 210 /* flood to everyone except source */ 211 macvlan_broadcast(skb, port, src->dev, 212 MACVLAN_MODE_VEPA | 213 MACVLAN_MODE_BRIDGE); 214 else if (src->mode == MACVLAN_MODE_BRIDGE) 215 /* 216 * flood only to VEPA ports, bridge ports 217 * already saw the frame on the way out. 218 */ 219 macvlan_broadcast(skb, port, src->dev, 220 MACVLAN_MODE_VEPA); 221 else { 222 /* forward to original port. */ 223 vlan = src; 224 ret = macvlan_broadcast_one(skb, vlan, eth, 0); 225 goto out; 226 } 227 228 return RX_HANDLER_PASS; 229 } 230 231 if (port->passthru) 232 vlan = list_first_entry(&port->vlans, struct macvlan_dev, list); 233 else 234 vlan = macvlan_hash_lookup(port, eth->h_dest); 235 if (vlan == NULL) 236 return RX_HANDLER_PASS; 237 238 dev = vlan->dev; 239 if (unlikely(!(dev->flags & IFF_UP))) { 240 kfree_skb(skb); 241 return RX_HANDLER_CONSUMED; 242 } 243 len = skb->len + ETH_HLEN; 244 skb = skb_share_check(skb, GFP_ATOMIC); 245 if (!skb) 246 goto out; 247 248 skb->dev = dev; 249 skb->pkt_type = PACKET_HOST; 250 251 ret = vlan->receive(skb); 252 253 out: 254 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0); 255 return RX_HANDLER_CONSUMED; 256 } 257 258 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 259 { 260 const struct macvlan_dev *vlan = netdev_priv(dev); 261 const struct macvlan_port *port = vlan->port; 262 const struct macvlan_dev *dest; 263 __u8 ip_summed = skb->ip_summed; 264 265 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 266 const struct ethhdr *eth = (void *)skb->data; 267 skb->ip_summed = CHECKSUM_UNNECESSARY; 268 269 /* send to other bridge ports directly */ 270 if (is_multicast_ether_addr(eth->h_dest)) { 271 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 272 goto xmit_world; 273 } 274 275 dest = macvlan_hash_lookup(port, eth->h_dest); 276 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 277 /* send to lowerdev first for its network taps */ 278 dev_forward_skb(vlan->lowerdev, skb); 279 280 return NET_XMIT_SUCCESS; 281 } 282 } 283 284 xmit_world: 285 skb->ip_summed = ip_summed; 286 skb->dev = vlan->lowerdev; 287 return dev_queue_xmit(skb); 288 } 289 290 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 291 struct net_device *dev) 292 { 293 unsigned int len = skb->len; 294 int ret; 295 const struct macvlan_dev *vlan = netdev_priv(dev); 296 297 ret = macvlan_queue_xmit(skb, dev); 298 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 299 struct macvlan_pcpu_stats *pcpu_stats; 300 301 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 302 u64_stats_update_begin(&pcpu_stats->syncp); 303 pcpu_stats->tx_packets++; 304 pcpu_stats->tx_bytes += len; 305 u64_stats_update_end(&pcpu_stats->syncp); 306 } else { 307 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 308 } 309 return ret; 310 } 311 EXPORT_SYMBOL_GPL(macvlan_start_xmit); 312 313 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 314 unsigned short type, const void *daddr, 315 const void *saddr, unsigned len) 316 { 317 const struct macvlan_dev *vlan = netdev_priv(dev); 318 struct net_device *lowerdev = vlan->lowerdev; 319 320 return dev_hard_header(skb, lowerdev, type, daddr, 321 saddr ? : dev->dev_addr, len); 322 } 323 324 static const struct header_ops macvlan_hard_header_ops = { 325 .create = macvlan_hard_header, 326 .rebuild = eth_rebuild_header, 327 .parse = eth_header_parse, 328 .cache = eth_header_cache, 329 .cache_update = eth_header_cache_update, 330 }; 331 332 static int macvlan_open(struct net_device *dev) 333 { 334 struct macvlan_dev *vlan = netdev_priv(dev); 335 struct net_device *lowerdev = vlan->lowerdev; 336 int err; 337 338 if (vlan->port->passthru) { 339 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 340 dev_set_promiscuity(lowerdev, 1); 341 goto hash_add; 342 } 343 344 err = -EBUSY; 345 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 346 goto out; 347 348 err = dev_uc_add(lowerdev, dev->dev_addr); 349 if (err < 0) 350 goto out; 351 if (dev->flags & IFF_ALLMULTI) { 352 err = dev_set_allmulti(lowerdev, 1); 353 if (err < 0) 354 goto del_unicast; 355 } 356 357 hash_add: 358 macvlan_hash_add(vlan); 359 return 0; 360 361 del_unicast: 362 dev_uc_del(lowerdev, dev->dev_addr); 363 out: 364 return err; 365 } 366 367 static int macvlan_stop(struct net_device *dev) 368 { 369 struct macvlan_dev *vlan = netdev_priv(dev); 370 struct net_device *lowerdev = vlan->lowerdev; 371 372 dev_uc_unsync(lowerdev, dev); 373 dev_mc_unsync(lowerdev, dev); 374 375 if (vlan->port->passthru) { 376 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 377 dev_set_promiscuity(lowerdev, -1); 378 goto hash_del; 379 } 380 381 if (dev->flags & IFF_ALLMULTI) 382 dev_set_allmulti(lowerdev, -1); 383 384 dev_uc_del(lowerdev, dev->dev_addr); 385 386 hash_del: 387 macvlan_hash_del(vlan, !dev->dismantle); 388 return 0; 389 } 390 391 static int macvlan_set_mac_address(struct net_device *dev, void *p) 392 { 393 struct macvlan_dev *vlan = netdev_priv(dev); 394 struct net_device *lowerdev = vlan->lowerdev; 395 struct sockaddr *addr = p; 396 int err; 397 398 if (!is_valid_ether_addr(addr->sa_data)) 399 return -EADDRNOTAVAIL; 400 401 if (!(dev->flags & IFF_UP)) { 402 /* Just copy in the new address */ 403 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN); 404 } else { 405 /* Rehash and update the device filters */ 406 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 407 return -EBUSY; 408 409 err = dev_uc_add(lowerdev, addr->sa_data); 410 if (err) 411 return err; 412 413 dev_uc_del(lowerdev, dev->dev_addr); 414 415 macvlan_hash_change_addr(vlan, addr->sa_data); 416 } 417 return 0; 418 } 419 420 static void macvlan_change_rx_flags(struct net_device *dev, int change) 421 { 422 struct macvlan_dev *vlan = netdev_priv(dev); 423 struct net_device *lowerdev = vlan->lowerdev; 424 425 if (change & IFF_ALLMULTI) 426 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 427 } 428 429 static void macvlan_set_mac_lists(struct net_device *dev) 430 { 431 struct macvlan_dev *vlan = netdev_priv(dev); 432 433 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 434 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ); 435 } else { 436 struct netdev_hw_addr *ha; 437 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 438 439 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 440 netdev_for_each_mc_addr(ha, dev) { 441 __set_bit(mc_hash(vlan, ha->addr), filter); 442 } 443 444 __set_bit(mc_hash(vlan, dev->broadcast), filter); 445 446 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ); 447 } 448 dev_uc_sync(vlan->lowerdev, dev); 449 dev_mc_sync(vlan->lowerdev, dev); 450 } 451 452 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 453 { 454 struct macvlan_dev *vlan = netdev_priv(dev); 455 456 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu) 457 return -EINVAL; 458 dev->mtu = new_mtu; 459 return 0; 460 } 461 462 /* 463 * macvlan network devices have devices nesting below it and are a special 464 * "super class" of normal network devices; split their locks off into a 465 * separate class since they always nest. 466 */ 467 static struct lock_class_key macvlan_netdev_xmit_lock_key; 468 static struct lock_class_key macvlan_netdev_addr_lock_key; 469 470 #define MACVLAN_FEATURES \ 471 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 472 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \ 473 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 474 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 475 476 #define MACVLAN_STATE_MASK \ 477 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 478 479 static void macvlan_set_lockdep_class_one(struct net_device *dev, 480 struct netdev_queue *txq, 481 void *_unused) 482 { 483 lockdep_set_class(&txq->_xmit_lock, 484 &macvlan_netdev_xmit_lock_key); 485 } 486 487 static void macvlan_set_lockdep_class(struct net_device *dev) 488 { 489 lockdep_set_class(&dev->addr_list_lock, 490 &macvlan_netdev_addr_lock_key); 491 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL); 492 } 493 494 static int macvlan_init(struct net_device *dev) 495 { 496 struct macvlan_dev *vlan = netdev_priv(dev); 497 const struct net_device *lowerdev = vlan->lowerdev; 498 499 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 500 (lowerdev->state & MACVLAN_STATE_MASK); 501 dev->features = lowerdev->features & MACVLAN_FEATURES; 502 dev->features |= NETIF_F_LLTX; 503 dev->gso_max_size = lowerdev->gso_max_size; 504 dev->iflink = lowerdev->ifindex; 505 dev->hard_header_len = lowerdev->hard_header_len; 506 507 macvlan_set_lockdep_class(dev); 508 509 vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats); 510 if (!vlan->pcpu_stats) 511 return -ENOMEM; 512 513 return 0; 514 } 515 516 static void macvlan_uninit(struct net_device *dev) 517 { 518 struct macvlan_dev *vlan = netdev_priv(dev); 519 struct macvlan_port *port = vlan->port; 520 521 free_percpu(vlan->pcpu_stats); 522 523 port->count -= 1; 524 if (!port->count) 525 macvlan_port_destroy(port->dev); 526 } 527 528 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev, 529 struct rtnl_link_stats64 *stats) 530 { 531 struct macvlan_dev *vlan = netdev_priv(dev); 532 533 if (vlan->pcpu_stats) { 534 struct macvlan_pcpu_stats *p; 535 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 536 u32 rx_errors = 0, tx_dropped = 0; 537 unsigned int start; 538 int i; 539 540 for_each_possible_cpu(i) { 541 p = per_cpu_ptr(vlan->pcpu_stats, i); 542 do { 543 start = u64_stats_fetch_begin_bh(&p->syncp); 544 rx_packets = p->rx_packets; 545 rx_bytes = p->rx_bytes; 546 rx_multicast = p->rx_multicast; 547 tx_packets = p->tx_packets; 548 tx_bytes = p->tx_bytes; 549 } while (u64_stats_fetch_retry_bh(&p->syncp, start)); 550 551 stats->rx_packets += rx_packets; 552 stats->rx_bytes += rx_bytes; 553 stats->multicast += rx_multicast; 554 stats->tx_packets += tx_packets; 555 stats->tx_bytes += tx_bytes; 556 /* rx_errors & tx_dropped are u32, updated 557 * without syncp protection. 558 */ 559 rx_errors += p->rx_errors; 560 tx_dropped += p->tx_dropped; 561 } 562 stats->rx_errors = rx_errors; 563 stats->rx_dropped = rx_errors; 564 stats->tx_dropped = tx_dropped; 565 } 566 return stats; 567 } 568 569 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 570 __be16 proto, u16 vid) 571 { 572 struct macvlan_dev *vlan = netdev_priv(dev); 573 struct net_device *lowerdev = vlan->lowerdev; 574 575 return vlan_vid_add(lowerdev, proto, vid); 576 } 577 578 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 579 __be16 proto, u16 vid) 580 { 581 struct macvlan_dev *vlan = netdev_priv(dev); 582 struct net_device *lowerdev = vlan->lowerdev; 583 584 vlan_vid_del(lowerdev, proto, vid); 585 return 0; 586 } 587 588 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 589 struct net_device *dev, 590 const unsigned char *addr, 591 u16 flags) 592 { 593 struct macvlan_dev *vlan = netdev_priv(dev); 594 int err = -EINVAL; 595 596 if (!vlan->port->passthru) 597 return -EOPNOTSUPP; 598 599 if (is_unicast_ether_addr(addr)) 600 err = dev_uc_add_excl(dev, addr); 601 else if (is_multicast_ether_addr(addr)) 602 err = dev_mc_add_excl(dev, addr); 603 604 return err; 605 } 606 607 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 608 struct net_device *dev, 609 const unsigned char *addr) 610 { 611 struct macvlan_dev *vlan = netdev_priv(dev); 612 int err = -EINVAL; 613 614 if (!vlan->port->passthru) 615 return -EOPNOTSUPP; 616 617 if (is_unicast_ether_addr(addr)) 618 err = dev_uc_del(dev, addr); 619 else if (is_multicast_ether_addr(addr)) 620 err = dev_mc_del(dev, addr); 621 622 return err; 623 } 624 625 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 626 struct ethtool_drvinfo *drvinfo) 627 { 628 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 629 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 630 } 631 632 static int macvlan_ethtool_get_settings(struct net_device *dev, 633 struct ethtool_cmd *cmd) 634 { 635 const struct macvlan_dev *vlan = netdev_priv(dev); 636 637 return __ethtool_get_settings(vlan->lowerdev, cmd); 638 } 639 640 static const struct ethtool_ops macvlan_ethtool_ops = { 641 .get_link = ethtool_op_get_link, 642 .get_settings = macvlan_ethtool_get_settings, 643 .get_drvinfo = macvlan_ethtool_get_drvinfo, 644 }; 645 646 static const struct net_device_ops macvlan_netdev_ops = { 647 .ndo_init = macvlan_init, 648 .ndo_uninit = macvlan_uninit, 649 .ndo_open = macvlan_open, 650 .ndo_stop = macvlan_stop, 651 .ndo_start_xmit = macvlan_start_xmit, 652 .ndo_change_mtu = macvlan_change_mtu, 653 .ndo_change_rx_flags = macvlan_change_rx_flags, 654 .ndo_set_mac_address = macvlan_set_mac_address, 655 .ndo_set_rx_mode = macvlan_set_mac_lists, 656 .ndo_get_stats64 = macvlan_dev_get_stats64, 657 .ndo_validate_addr = eth_validate_addr, 658 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 659 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 660 .ndo_fdb_add = macvlan_fdb_add, 661 .ndo_fdb_del = macvlan_fdb_del, 662 .ndo_fdb_dump = ndo_dflt_fdb_dump, 663 }; 664 665 void macvlan_common_setup(struct net_device *dev) 666 { 667 ether_setup(dev); 668 669 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); 670 dev->priv_flags |= IFF_UNICAST_FLT; 671 dev->netdev_ops = &macvlan_netdev_ops; 672 dev->destructor = free_netdev; 673 dev->header_ops = &macvlan_hard_header_ops, 674 dev->ethtool_ops = &macvlan_ethtool_ops; 675 } 676 EXPORT_SYMBOL_GPL(macvlan_common_setup); 677 678 static void macvlan_setup(struct net_device *dev) 679 { 680 macvlan_common_setup(dev); 681 dev->tx_queue_len = 0; 682 } 683 684 static int macvlan_port_create(struct net_device *dev) 685 { 686 struct macvlan_port *port; 687 unsigned int i; 688 int err; 689 690 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 691 return -EINVAL; 692 693 port = kzalloc(sizeof(*port), GFP_KERNEL); 694 if (port == NULL) 695 return -ENOMEM; 696 697 port->passthru = false; 698 port->dev = dev; 699 INIT_LIST_HEAD(&port->vlans); 700 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 701 INIT_HLIST_HEAD(&port->vlan_hash[i]); 702 703 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 704 if (err) 705 kfree(port); 706 else 707 dev->priv_flags |= IFF_MACVLAN_PORT; 708 return err; 709 } 710 711 static void macvlan_port_destroy(struct net_device *dev) 712 { 713 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 714 715 dev->priv_flags &= ~IFF_MACVLAN_PORT; 716 netdev_rx_handler_unregister(dev); 717 kfree_rcu(port, rcu); 718 } 719 720 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[]) 721 { 722 if (tb[IFLA_ADDRESS]) { 723 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 724 return -EINVAL; 725 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 726 return -EADDRNOTAVAIL; 727 } 728 729 if (data && data[IFLA_MACVLAN_MODE]) { 730 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 731 case MACVLAN_MODE_PRIVATE: 732 case MACVLAN_MODE_VEPA: 733 case MACVLAN_MODE_BRIDGE: 734 case MACVLAN_MODE_PASSTHRU: 735 break; 736 default: 737 return -EINVAL; 738 } 739 } 740 return 0; 741 } 742 743 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 744 struct nlattr *tb[], struct nlattr *data[], 745 int (*receive)(struct sk_buff *skb), 746 int (*forward)(struct net_device *dev, 747 struct sk_buff *skb)) 748 { 749 struct macvlan_dev *vlan = netdev_priv(dev); 750 struct macvlan_port *port; 751 struct net_device *lowerdev; 752 int err; 753 754 if (!tb[IFLA_LINK]) 755 return -EINVAL; 756 757 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 758 if (lowerdev == NULL) 759 return -ENODEV; 760 761 /* When creating macvlans on top of other macvlans - use 762 * the real device as the lowerdev. 763 */ 764 if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) { 765 struct macvlan_dev *lowervlan = netdev_priv(lowerdev); 766 lowerdev = lowervlan->lowerdev; 767 } 768 769 if (!tb[IFLA_MTU]) 770 dev->mtu = lowerdev->mtu; 771 else if (dev->mtu > lowerdev->mtu) 772 return -EINVAL; 773 774 if (!tb[IFLA_ADDRESS]) 775 eth_hw_addr_random(dev); 776 777 if (!macvlan_port_exists(lowerdev)) { 778 err = macvlan_port_create(lowerdev); 779 if (err < 0) 780 return err; 781 } 782 port = macvlan_port_get_rtnl(lowerdev); 783 784 /* Only 1 macvlan device can be created in passthru mode */ 785 if (port->passthru) 786 return -EINVAL; 787 788 vlan->lowerdev = lowerdev; 789 vlan->dev = dev; 790 vlan->port = port; 791 vlan->receive = receive; 792 vlan->forward = forward; 793 794 vlan->mode = MACVLAN_MODE_VEPA; 795 if (data && data[IFLA_MACVLAN_MODE]) 796 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 797 798 if (data && data[IFLA_MACVLAN_FLAGS]) 799 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 800 801 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 802 if (port->count) 803 return -EINVAL; 804 port->passthru = true; 805 memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN); 806 } 807 808 err = netdev_upper_dev_link(lowerdev, dev); 809 if (err) 810 goto destroy_port; 811 812 port->count += 1; 813 err = register_netdevice(dev); 814 if (err < 0) 815 goto upper_dev_unlink; 816 817 list_add_tail(&vlan->list, &port->vlans); 818 netif_stacked_transfer_operstate(lowerdev, dev); 819 820 return 0; 821 822 upper_dev_unlink: 823 netdev_upper_dev_unlink(lowerdev, dev); 824 destroy_port: 825 port->count -= 1; 826 if (!port->count) 827 macvlan_port_destroy(lowerdev); 828 829 return err; 830 } 831 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 832 833 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 834 struct nlattr *tb[], struct nlattr *data[]) 835 { 836 return macvlan_common_newlink(src_net, dev, tb, data, 837 netif_rx, 838 dev_forward_skb); 839 } 840 841 void macvlan_dellink(struct net_device *dev, struct list_head *head) 842 { 843 struct macvlan_dev *vlan = netdev_priv(dev); 844 845 list_del(&vlan->list); 846 unregister_netdevice_queue(dev, head); 847 netdev_upper_dev_unlink(vlan->lowerdev, dev); 848 } 849 EXPORT_SYMBOL_GPL(macvlan_dellink); 850 851 static int macvlan_changelink(struct net_device *dev, 852 struct nlattr *tb[], struct nlattr *data[]) 853 { 854 struct macvlan_dev *vlan = netdev_priv(dev); 855 if (data && data[IFLA_MACVLAN_MODE]) 856 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 857 if (data && data[IFLA_MACVLAN_FLAGS]) { 858 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 859 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 860 861 if (promisc && (flags & MACVLAN_FLAG_NOPROMISC)) 862 dev_set_promiscuity(vlan->lowerdev, -1); 863 else if (promisc && !(flags & MACVLAN_FLAG_NOPROMISC)) 864 dev_set_promiscuity(vlan->lowerdev, 1); 865 vlan->flags = flags; 866 } 867 return 0; 868 } 869 870 static size_t macvlan_get_size(const struct net_device *dev) 871 { 872 return (0 873 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 874 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 875 ); 876 } 877 878 static int macvlan_fill_info(struct sk_buff *skb, 879 const struct net_device *dev) 880 { 881 struct macvlan_dev *vlan = netdev_priv(dev); 882 883 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 884 goto nla_put_failure; 885 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 886 goto nla_put_failure; 887 return 0; 888 889 nla_put_failure: 890 return -EMSGSIZE; 891 } 892 893 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 894 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 895 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 896 }; 897 898 int macvlan_link_register(struct rtnl_link_ops *ops) 899 { 900 /* common fields */ 901 ops->priv_size = sizeof(struct macvlan_dev); 902 ops->validate = macvlan_validate; 903 ops->maxtype = IFLA_MACVLAN_MAX; 904 ops->policy = macvlan_policy; 905 ops->changelink = macvlan_changelink; 906 ops->get_size = macvlan_get_size; 907 ops->fill_info = macvlan_fill_info; 908 909 return rtnl_link_register(ops); 910 }; 911 EXPORT_SYMBOL_GPL(macvlan_link_register); 912 913 static struct rtnl_link_ops macvlan_link_ops = { 914 .kind = "macvlan", 915 .setup = macvlan_setup, 916 .newlink = macvlan_newlink, 917 .dellink = macvlan_dellink, 918 }; 919 920 static int macvlan_device_event(struct notifier_block *unused, 921 unsigned long event, void *ptr) 922 { 923 struct net_device *dev = ptr; 924 struct macvlan_dev *vlan, *next; 925 struct macvlan_port *port; 926 LIST_HEAD(list_kill); 927 928 if (!macvlan_port_exists(dev)) 929 return NOTIFY_DONE; 930 931 port = macvlan_port_get_rtnl(dev); 932 933 switch (event) { 934 case NETDEV_CHANGE: 935 list_for_each_entry(vlan, &port->vlans, list) 936 netif_stacked_transfer_operstate(vlan->lowerdev, 937 vlan->dev); 938 break; 939 case NETDEV_FEAT_CHANGE: 940 list_for_each_entry(vlan, &port->vlans, list) { 941 vlan->dev->features = dev->features & MACVLAN_FEATURES; 942 vlan->dev->gso_max_size = dev->gso_max_size; 943 netdev_features_change(vlan->dev); 944 } 945 break; 946 case NETDEV_UNREGISTER: 947 /* twiddle thumbs on netns device moves */ 948 if (dev->reg_state != NETREG_UNREGISTERING) 949 break; 950 951 list_for_each_entry_safe(vlan, next, &port->vlans, list) 952 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 953 unregister_netdevice_many(&list_kill); 954 list_del(&list_kill); 955 break; 956 case NETDEV_PRE_TYPE_CHANGE: 957 /* Forbid underlaying device to change its type. */ 958 return NOTIFY_BAD; 959 } 960 return NOTIFY_DONE; 961 } 962 963 static struct notifier_block macvlan_notifier_block __read_mostly = { 964 .notifier_call = macvlan_device_event, 965 }; 966 967 static int __init macvlan_init_module(void) 968 { 969 int err; 970 971 register_netdevice_notifier(&macvlan_notifier_block); 972 973 err = macvlan_link_register(&macvlan_link_ops); 974 if (err < 0) 975 goto err1; 976 return 0; 977 err1: 978 unregister_netdevice_notifier(&macvlan_notifier_block); 979 return err; 980 } 981 982 static void __exit macvlan_cleanup_module(void) 983 { 984 rtnl_link_unregister(&macvlan_link_ops); 985 unregister_netdevice_notifier(&macvlan_notifier_block); 986 } 987 988 module_init(macvlan_init_module); 989 module_exit(macvlan_cleanup_module); 990 991 MODULE_LICENSE("GPL"); 992 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 993 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 994 MODULE_ALIAS_RTNL_LINK("macvlan"); 995