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 <linux/workqueue.h> 34 #include <net/rtnetlink.h> 35 #include <net/xfrm.h> 36 37 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE) 38 39 struct macvlan_port { 40 struct net_device *dev; 41 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 42 struct list_head vlans; 43 struct rcu_head rcu; 44 struct sk_buff_head bc_queue; 45 struct work_struct bc_work; 46 bool passthru; 47 int count; 48 }; 49 50 struct macvlan_skb_cb { 51 const struct macvlan_dev *src; 52 }; 53 54 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 55 56 static void macvlan_port_destroy(struct net_device *dev); 57 58 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 59 { 60 return rcu_dereference(dev->rx_handler_data); 61 } 62 63 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 64 { 65 return rtnl_dereference(dev->rx_handler_data); 66 } 67 68 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT) 69 70 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 71 const unsigned char *addr) 72 { 73 struct macvlan_dev *vlan; 74 75 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) { 76 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 77 return vlan; 78 } 79 return NULL; 80 } 81 82 static void macvlan_hash_add(struct macvlan_dev *vlan) 83 { 84 struct macvlan_port *port = vlan->port; 85 const unsigned char *addr = vlan->dev->dev_addr; 86 87 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]); 88 } 89 90 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 91 { 92 hlist_del_rcu(&vlan->hlist); 93 if (sync) 94 synchronize_rcu(); 95 } 96 97 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 98 const unsigned char *addr) 99 { 100 macvlan_hash_del(vlan, true); 101 /* Now that we are unhashed it is safe to change the device 102 * address without confusing packet delivery. 103 */ 104 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 105 macvlan_hash_add(vlan); 106 } 107 108 static int macvlan_addr_busy(const struct macvlan_port *port, 109 const unsigned char *addr) 110 { 111 /* Test to see if the specified multicast address is 112 * currently in use by the underlying device or 113 * another macvlan. 114 */ 115 if (ether_addr_equal_64bits(port->dev->dev_addr, addr)) 116 return 1; 117 118 if (macvlan_hash_lookup(port, addr)) 119 return 1; 120 121 return 0; 122 } 123 124 125 static int macvlan_broadcast_one(struct sk_buff *skb, 126 const struct macvlan_dev *vlan, 127 const struct ethhdr *eth, bool local) 128 { 129 struct net_device *dev = vlan->dev; 130 131 if (local) 132 return __dev_forward_skb(dev, skb); 133 134 skb->dev = dev; 135 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 136 skb->pkt_type = PACKET_BROADCAST; 137 else 138 skb->pkt_type = PACKET_MULTICAST; 139 140 return 0; 141 } 142 143 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 144 { 145 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 146 } 147 148 149 static unsigned int mc_hash(const struct macvlan_dev *vlan, 150 const unsigned char *addr) 151 { 152 u32 val = __get_unaligned_cpu32(addr + 2); 153 154 val ^= macvlan_hash_mix(vlan); 155 return hash_32(val, MACVLAN_MC_FILTER_BITS); 156 } 157 158 static void macvlan_broadcast(struct sk_buff *skb, 159 const struct macvlan_port *port, 160 struct net_device *src, 161 enum macvlan_mode mode) 162 { 163 const struct ethhdr *eth = eth_hdr(skb); 164 const struct macvlan_dev *vlan; 165 struct sk_buff *nskb; 166 unsigned int i; 167 int err; 168 unsigned int hash; 169 170 if (skb->protocol == htons(ETH_P_PAUSE)) 171 return; 172 173 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 174 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 175 if (vlan->dev == src || !(vlan->mode & mode)) 176 continue; 177 178 hash = mc_hash(vlan, eth->h_dest); 179 if (!test_bit(hash, vlan->mc_filter)) 180 continue; 181 182 err = NET_RX_DROP; 183 nskb = skb_clone(skb, GFP_ATOMIC); 184 if (likely(nskb)) 185 err = macvlan_broadcast_one( 186 nskb, vlan, eth, 187 mode == MACVLAN_MODE_BRIDGE) ?: 188 netif_rx_ni(nskb); 189 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 190 err == NET_RX_SUCCESS, 1); 191 } 192 } 193 } 194 195 static void macvlan_process_broadcast(struct work_struct *w) 196 { 197 struct macvlan_port *port = container_of(w, struct macvlan_port, 198 bc_work); 199 struct sk_buff *skb; 200 struct sk_buff_head list; 201 202 skb_queue_head_init(&list); 203 204 spin_lock_bh(&port->bc_queue.lock); 205 skb_queue_splice_tail_init(&port->bc_queue, &list); 206 spin_unlock_bh(&port->bc_queue.lock); 207 208 while ((skb = __skb_dequeue(&list))) { 209 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 210 211 rcu_read_lock(); 212 213 if (!src) 214 /* frame comes from an external address */ 215 macvlan_broadcast(skb, port, NULL, 216 MACVLAN_MODE_PRIVATE | 217 MACVLAN_MODE_VEPA | 218 MACVLAN_MODE_PASSTHRU| 219 MACVLAN_MODE_BRIDGE); 220 else if (src->mode == MACVLAN_MODE_VEPA) 221 /* flood to everyone except source */ 222 macvlan_broadcast(skb, port, src->dev, 223 MACVLAN_MODE_VEPA | 224 MACVLAN_MODE_BRIDGE); 225 else 226 /* 227 * flood only to VEPA ports, bridge ports 228 * already saw the frame on the way out. 229 */ 230 macvlan_broadcast(skb, port, src->dev, 231 MACVLAN_MODE_VEPA); 232 233 rcu_read_unlock(); 234 235 kfree_skb(skb); 236 } 237 } 238 239 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 240 struct sk_buff *skb) 241 { 242 struct sk_buff *nskb; 243 int err = -ENOMEM; 244 245 nskb = skb_clone(skb, GFP_ATOMIC); 246 if (!nskb) 247 goto err; 248 249 spin_lock(&port->bc_queue.lock); 250 if (skb_queue_len(&port->bc_queue) < skb->dev->tx_queue_len) { 251 __skb_queue_tail(&port->bc_queue, nskb); 252 err = 0; 253 } 254 spin_unlock(&port->bc_queue.lock); 255 256 if (err) 257 goto free_nskb; 258 259 schedule_work(&port->bc_work); 260 return; 261 262 free_nskb: 263 kfree_skb(nskb); 264 err: 265 atomic_long_inc(&skb->dev->rx_dropped); 266 } 267 268 /* called under rcu_read_lock() from netif_receive_skb */ 269 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 270 { 271 struct macvlan_port *port; 272 struct sk_buff *skb = *pskb; 273 const struct ethhdr *eth = eth_hdr(skb); 274 const struct macvlan_dev *vlan; 275 const struct macvlan_dev *src; 276 struct net_device *dev; 277 unsigned int len = 0; 278 int ret = NET_RX_DROP; 279 280 port = macvlan_port_get_rcu(skb->dev); 281 if (is_multicast_ether_addr(eth->h_dest)) { 282 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN); 283 if (!skb) 284 return RX_HANDLER_CONSUMED; 285 eth = eth_hdr(skb); 286 src = macvlan_hash_lookup(port, eth->h_source); 287 if (src && src->mode != MACVLAN_MODE_VEPA && 288 src->mode != MACVLAN_MODE_BRIDGE) { 289 /* forward to original port. */ 290 vlan = src; 291 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 292 netif_rx(skb); 293 goto out; 294 } 295 296 MACVLAN_SKB_CB(skb)->src = src; 297 macvlan_broadcast_enqueue(port, skb); 298 299 return RX_HANDLER_PASS; 300 } 301 302 if (port->passthru) 303 vlan = list_first_or_null_rcu(&port->vlans, 304 struct macvlan_dev, list); 305 else 306 vlan = macvlan_hash_lookup(port, eth->h_dest); 307 if (vlan == NULL) 308 return RX_HANDLER_PASS; 309 310 dev = vlan->dev; 311 if (unlikely(!(dev->flags & IFF_UP))) { 312 kfree_skb(skb); 313 return RX_HANDLER_CONSUMED; 314 } 315 len = skb->len + ETH_HLEN; 316 skb = skb_share_check(skb, GFP_ATOMIC); 317 if (!skb) 318 goto out; 319 320 skb->dev = dev; 321 skb->pkt_type = PACKET_HOST; 322 323 ret = netif_rx(skb); 324 325 out: 326 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0); 327 return RX_HANDLER_CONSUMED; 328 } 329 330 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 331 { 332 const struct macvlan_dev *vlan = netdev_priv(dev); 333 const struct macvlan_port *port = vlan->port; 334 const struct macvlan_dev *dest; 335 336 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 337 const struct ethhdr *eth = (void *)skb->data; 338 339 /* send to other bridge ports directly */ 340 if (is_multicast_ether_addr(eth->h_dest)) { 341 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 342 goto xmit_world; 343 } 344 345 dest = macvlan_hash_lookup(port, eth->h_dest); 346 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 347 /* send to lowerdev first for its network taps */ 348 dev_forward_skb(vlan->lowerdev, skb); 349 350 return NET_XMIT_SUCCESS; 351 } 352 } 353 354 xmit_world: 355 skb->dev = vlan->lowerdev; 356 return dev_queue_xmit(skb); 357 } 358 359 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 360 struct net_device *dev) 361 { 362 unsigned int len = skb->len; 363 int ret; 364 const struct macvlan_dev *vlan = netdev_priv(dev); 365 366 if (vlan->fwd_priv) { 367 skb->dev = vlan->lowerdev; 368 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv); 369 } else { 370 ret = macvlan_queue_xmit(skb, dev); 371 } 372 373 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 374 struct vlan_pcpu_stats *pcpu_stats; 375 376 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 377 u64_stats_update_begin(&pcpu_stats->syncp); 378 pcpu_stats->tx_packets++; 379 pcpu_stats->tx_bytes += len; 380 u64_stats_update_end(&pcpu_stats->syncp); 381 } else { 382 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 383 } 384 return ret; 385 } 386 387 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 388 unsigned short type, const void *daddr, 389 const void *saddr, unsigned len) 390 { 391 const struct macvlan_dev *vlan = netdev_priv(dev); 392 struct net_device *lowerdev = vlan->lowerdev; 393 394 return dev_hard_header(skb, lowerdev, type, daddr, 395 saddr ? : dev->dev_addr, len); 396 } 397 398 static const struct header_ops macvlan_hard_header_ops = { 399 .create = macvlan_hard_header, 400 .rebuild = eth_rebuild_header, 401 .parse = eth_header_parse, 402 .cache = eth_header_cache, 403 .cache_update = eth_header_cache_update, 404 }; 405 406 static struct rtnl_link_ops macvlan_link_ops; 407 408 static int macvlan_open(struct net_device *dev) 409 { 410 struct macvlan_dev *vlan = netdev_priv(dev); 411 struct net_device *lowerdev = vlan->lowerdev; 412 int err; 413 414 if (vlan->port->passthru) { 415 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 416 err = dev_set_promiscuity(lowerdev, 1); 417 if (err < 0) 418 goto out; 419 } 420 goto hash_add; 421 } 422 423 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD && 424 dev->rtnl_link_ops == &macvlan_link_ops) { 425 vlan->fwd_priv = 426 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 427 428 /* If we get a NULL pointer back, or if we get an error 429 * then we should just fall through to the non accelerated path 430 */ 431 if (IS_ERR_OR_NULL(vlan->fwd_priv)) { 432 vlan->fwd_priv = NULL; 433 } else 434 return 0; 435 } 436 437 err = -EBUSY; 438 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 439 goto out; 440 441 err = dev_uc_add(lowerdev, dev->dev_addr); 442 if (err < 0) 443 goto out; 444 if (dev->flags & IFF_ALLMULTI) { 445 err = dev_set_allmulti(lowerdev, 1); 446 if (err < 0) 447 goto del_unicast; 448 } 449 450 hash_add: 451 macvlan_hash_add(vlan); 452 return 0; 453 454 del_unicast: 455 dev_uc_del(lowerdev, dev->dev_addr); 456 out: 457 if (vlan->fwd_priv) { 458 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 459 vlan->fwd_priv); 460 vlan->fwd_priv = NULL; 461 } 462 return err; 463 } 464 465 static int macvlan_stop(struct net_device *dev) 466 { 467 struct macvlan_dev *vlan = netdev_priv(dev); 468 struct net_device *lowerdev = vlan->lowerdev; 469 470 if (vlan->fwd_priv) { 471 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 472 vlan->fwd_priv); 473 vlan->fwd_priv = NULL; 474 return 0; 475 } 476 477 dev_uc_unsync(lowerdev, dev); 478 dev_mc_unsync(lowerdev, dev); 479 480 if (vlan->port->passthru) { 481 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 482 dev_set_promiscuity(lowerdev, -1); 483 goto hash_del; 484 } 485 486 if (dev->flags & IFF_ALLMULTI) 487 dev_set_allmulti(lowerdev, -1); 488 489 dev_uc_del(lowerdev, dev->dev_addr); 490 491 hash_del: 492 macvlan_hash_del(vlan, !dev->dismantle); 493 return 0; 494 } 495 496 static int macvlan_set_mac_address(struct net_device *dev, void *p) 497 { 498 struct macvlan_dev *vlan = netdev_priv(dev); 499 struct net_device *lowerdev = vlan->lowerdev; 500 struct sockaddr *addr = p; 501 int err; 502 503 if (!is_valid_ether_addr(addr->sa_data)) 504 return -EADDRNOTAVAIL; 505 506 if (!(dev->flags & IFF_UP)) { 507 /* Just copy in the new address */ 508 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN); 509 } else { 510 /* Rehash and update the device filters */ 511 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 512 return -EBUSY; 513 514 err = dev_uc_add(lowerdev, addr->sa_data); 515 if (err) 516 return err; 517 518 dev_uc_del(lowerdev, dev->dev_addr); 519 520 macvlan_hash_change_addr(vlan, addr->sa_data); 521 } 522 return 0; 523 } 524 525 static void macvlan_change_rx_flags(struct net_device *dev, int change) 526 { 527 struct macvlan_dev *vlan = netdev_priv(dev); 528 struct net_device *lowerdev = vlan->lowerdev; 529 530 if (change & IFF_ALLMULTI) 531 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 532 } 533 534 static void macvlan_set_mac_lists(struct net_device *dev) 535 { 536 struct macvlan_dev *vlan = netdev_priv(dev); 537 538 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 539 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ); 540 } else { 541 struct netdev_hw_addr *ha; 542 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 543 544 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 545 netdev_for_each_mc_addr(ha, dev) { 546 __set_bit(mc_hash(vlan, ha->addr), filter); 547 } 548 549 __set_bit(mc_hash(vlan, dev->broadcast), filter); 550 551 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ); 552 } 553 dev_uc_sync(vlan->lowerdev, dev); 554 dev_mc_sync(vlan->lowerdev, dev); 555 } 556 557 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 558 { 559 struct macvlan_dev *vlan = netdev_priv(dev); 560 561 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu) 562 return -EINVAL; 563 dev->mtu = new_mtu; 564 return 0; 565 } 566 567 /* 568 * macvlan network devices have devices nesting below it and are a special 569 * "super class" of normal network devices; split their locks off into a 570 * separate class since they always nest. 571 */ 572 static struct lock_class_key macvlan_netdev_xmit_lock_key; 573 static struct lock_class_key macvlan_netdev_addr_lock_key; 574 575 #define ALWAYS_ON_FEATURES \ 576 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX) 577 578 #define MACVLAN_FEATURES \ 579 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 580 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \ 581 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 582 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 583 584 #define MACVLAN_STATE_MASK \ 585 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 586 587 static void macvlan_set_lockdep_class_one(struct net_device *dev, 588 struct netdev_queue *txq, 589 void *_unused) 590 { 591 lockdep_set_class(&txq->_xmit_lock, 592 &macvlan_netdev_xmit_lock_key); 593 } 594 595 static void macvlan_set_lockdep_class(struct net_device *dev) 596 { 597 lockdep_set_class(&dev->addr_list_lock, 598 &macvlan_netdev_addr_lock_key); 599 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL); 600 } 601 602 static int macvlan_init(struct net_device *dev) 603 { 604 struct macvlan_dev *vlan = netdev_priv(dev); 605 const struct net_device *lowerdev = vlan->lowerdev; 606 607 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 608 (lowerdev->state & MACVLAN_STATE_MASK); 609 dev->features = lowerdev->features & MACVLAN_FEATURES; 610 dev->features |= ALWAYS_ON_FEATURES; 611 dev->gso_max_size = lowerdev->gso_max_size; 612 dev->iflink = lowerdev->ifindex; 613 dev->hard_header_len = lowerdev->hard_header_len; 614 615 macvlan_set_lockdep_class(dev); 616 617 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 618 if (!vlan->pcpu_stats) 619 return -ENOMEM; 620 621 return 0; 622 } 623 624 static void macvlan_uninit(struct net_device *dev) 625 { 626 struct macvlan_dev *vlan = netdev_priv(dev); 627 struct macvlan_port *port = vlan->port; 628 629 free_percpu(vlan->pcpu_stats); 630 631 port->count -= 1; 632 if (!port->count) 633 macvlan_port_destroy(port->dev); 634 } 635 636 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev, 637 struct rtnl_link_stats64 *stats) 638 { 639 struct macvlan_dev *vlan = netdev_priv(dev); 640 641 if (vlan->pcpu_stats) { 642 struct vlan_pcpu_stats *p; 643 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 644 u32 rx_errors = 0, tx_dropped = 0; 645 unsigned int start; 646 int i; 647 648 for_each_possible_cpu(i) { 649 p = per_cpu_ptr(vlan->pcpu_stats, i); 650 do { 651 start = u64_stats_fetch_begin_irq(&p->syncp); 652 rx_packets = p->rx_packets; 653 rx_bytes = p->rx_bytes; 654 rx_multicast = p->rx_multicast; 655 tx_packets = p->tx_packets; 656 tx_bytes = p->tx_bytes; 657 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 658 659 stats->rx_packets += rx_packets; 660 stats->rx_bytes += rx_bytes; 661 stats->multicast += rx_multicast; 662 stats->tx_packets += tx_packets; 663 stats->tx_bytes += tx_bytes; 664 /* rx_errors & tx_dropped are u32, updated 665 * without syncp protection. 666 */ 667 rx_errors += p->rx_errors; 668 tx_dropped += p->tx_dropped; 669 } 670 stats->rx_errors = rx_errors; 671 stats->rx_dropped = rx_errors; 672 stats->tx_dropped = tx_dropped; 673 } 674 return stats; 675 } 676 677 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 678 __be16 proto, u16 vid) 679 { 680 struct macvlan_dev *vlan = netdev_priv(dev); 681 struct net_device *lowerdev = vlan->lowerdev; 682 683 return vlan_vid_add(lowerdev, proto, vid); 684 } 685 686 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 687 __be16 proto, u16 vid) 688 { 689 struct macvlan_dev *vlan = netdev_priv(dev); 690 struct net_device *lowerdev = vlan->lowerdev; 691 692 vlan_vid_del(lowerdev, proto, vid); 693 return 0; 694 } 695 696 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 697 struct net_device *dev, 698 const unsigned char *addr, 699 u16 flags) 700 { 701 struct macvlan_dev *vlan = netdev_priv(dev); 702 int err = -EINVAL; 703 704 if (!vlan->port->passthru) 705 return -EOPNOTSUPP; 706 707 if (flags & NLM_F_REPLACE) 708 return -EOPNOTSUPP; 709 710 if (is_unicast_ether_addr(addr)) 711 err = dev_uc_add_excl(dev, addr); 712 else if (is_multicast_ether_addr(addr)) 713 err = dev_mc_add_excl(dev, addr); 714 715 return err; 716 } 717 718 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 719 struct net_device *dev, 720 const unsigned char *addr) 721 { 722 struct macvlan_dev *vlan = netdev_priv(dev); 723 int err = -EINVAL; 724 725 if (!vlan->port->passthru) 726 return -EOPNOTSUPP; 727 728 if (is_unicast_ether_addr(addr)) 729 err = dev_uc_del(dev, addr); 730 else if (is_multicast_ether_addr(addr)) 731 err = dev_mc_del(dev, addr); 732 733 return err; 734 } 735 736 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 737 struct ethtool_drvinfo *drvinfo) 738 { 739 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 740 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 741 } 742 743 static int macvlan_ethtool_get_settings(struct net_device *dev, 744 struct ethtool_cmd *cmd) 745 { 746 const struct macvlan_dev *vlan = netdev_priv(dev); 747 748 return __ethtool_get_settings(vlan->lowerdev, cmd); 749 } 750 751 static netdev_features_t macvlan_fix_features(struct net_device *dev, 752 netdev_features_t features) 753 { 754 struct macvlan_dev *vlan = netdev_priv(dev); 755 netdev_features_t mask; 756 757 features |= NETIF_F_ALL_FOR_ALL; 758 features &= (vlan->set_features | ~MACVLAN_FEATURES); 759 mask = features; 760 761 features = netdev_increment_features(vlan->lowerdev->features, 762 features, 763 mask); 764 features |= ALWAYS_ON_FEATURES; 765 766 return features; 767 } 768 769 static const struct ethtool_ops macvlan_ethtool_ops = { 770 .get_link = ethtool_op_get_link, 771 .get_settings = macvlan_ethtool_get_settings, 772 .get_drvinfo = macvlan_ethtool_get_drvinfo, 773 }; 774 775 static const struct net_device_ops macvlan_netdev_ops = { 776 .ndo_init = macvlan_init, 777 .ndo_uninit = macvlan_uninit, 778 .ndo_open = macvlan_open, 779 .ndo_stop = macvlan_stop, 780 .ndo_start_xmit = macvlan_start_xmit, 781 .ndo_change_mtu = macvlan_change_mtu, 782 .ndo_fix_features = macvlan_fix_features, 783 .ndo_change_rx_flags = macvlan_change_rx_flags, 784 .ndo_set_mac_address = macvlan_set_mac_address, 785 .ndo_set_rx_mode = macvlan_set_mac_lists, 786 .ndo_get_stats64 = macvlan_dev_get_stats64, 787 .ndo_validate_addr = eth_validate_addr, 788 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 789 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 790 .ndo_fdb_add = macvlan_fdb_add, 791 .ndo_fdb_del = macvlan_fdb_del, 792 .ndo_fdb_dump = ndo_dflt_fdb_dump, 793 }; 794 795 void macvlan_common_setup(struct net_device *dev) 796 { 797 ether_setup(dev); 798 799 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); 800 dev->priv_flags |= IFF_UNICAST_FLT; 801 dev->netdev_ops = &macvlan_netdev_ops; 802 dev->destructor = free_netdev; 803 dev->header_ops = &macvlan_hard_header_ops; 804 dev->ethtool_ops = &macvlan_ethtool_ops; 805 } 806 EXPORT_SYMBOL_GPL(macvlan_common_setup); 807 808 static void macvlan_setup(struct net_device *dev) 809 { 810 macvlan_common_setup(dev); 811 dev->tx_queue_len = 0; 812 } 813 814 static int macvlan_port_create(struct net_device *dev) 815 { 816 struct macvlan_port *port; 817 unsigned int i; 818 int err; 819 820 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 821 return -EINVAL; 822 823 port = kzalloc(sizeof(*port), GFP_KERNEL); 824 if (port == NULL) 825 return -ENOMEM; 826 827 port->passthru = false; 828 port->dev = dev; 829 INIT_LIST_HEAD(&port->vlans); 830 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 831 INIT_HLIST_HEAD(&port->vlan_hash[i]); 832 833 skb_queue_head_init(&port->bc_queue); 834 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 835 836 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 837 if (err) 838 kfree(port); 839 else 840 dev->priv_flags |= IFF_MACVLAN_PORT; 841 return err; 842 } 843 844 static void macvlan_port_destroy(struct net_device *dev) 845 { 846 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 847 848 cancel_work_sync(&port->bc_work); 849 dev->priv_flags &= ~IFF_MACVLAN_PORT; 850 netdev_rx_handler_unregister(dev); 851 kfree_rcu(port, rcu); 852 } 853 854 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[]) 855 { 856 if (tb[IFLA_ADDRESS]) { 857 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 858 return -EINVAL; 859 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 860 return -EADDRNOTAVAIL; 861 } 862 863 if (data && data[IFLA_MACVLAN_FLAGS] && 864 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 865 return -EINVAL; 866 867 if (data && data[IFLA_MACVLAN_MODE]) { 868 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 869 case MACVLAN_MODE_PRIVATE: 870 case MACVLAN_MODE_VEPA: 871 case MACVLAN_MODE_BRIDGE: 872 case MACVLAN_MODE_PASSTHRU: 873 break; 874 default: 875 return -EINVAL; 876 } 877 } 878 return 0; 879 } 880 881 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 882 struct nlattr *tb[], struct nlattr *data[]) 883 { 884 struct macvlan_dev *vlan = netdev_priv(dev); 885 struct macvlan_port *port; 886 struct net_device *lowerdev; 887 int err; 888 889 if (!tb[IFLA_LINK]) 890 return -EINVAL; 891 892 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 893 if (lowerdev == NULL) 894 return -ENODEV; 895 896 /* When creating macvlans or macvtaps on top of other macvlans - use 897 * the real device as the lowerdev. 898 */ 899 if (netif_is_macvlan(lowerdev)) 900 lowerdev = macvlan_dev_real_dev(lowerdev); 901 902 if (!tb[IFLA_MTU]) 903 dev->mtu = lowerdev->mtu; 904 else if (dev->mtu > lowerdev->mtu) 905 return -EINVAL; 906 907 if (!tb[IFLA_ADDRESS]) 908 eth_hw_addr_random(dev); 909 910 if (!macvlan_port_exists(lowerdev)) { 911 err = macvlan_port_create(lowerdev); 912 if (err < 0) 913 return err; 914 } 915 port = macvlan_port_get_rtnl(lowerdev); 916 917 /* Only 1 macvlan device can be created in passthru mode */ 918 if (port->passthru) 919 return -EINVAL; 920 921 vlan->lowerdev = lowerdev; 922 vlan->dev = dev; 923 vlan->port = port; 924 vlan->set_features = MACVLAN_FEATURES; 925 926 vlan->mode = MACVLAN_MODE_VEPA; 927 if (data && data[IFLA_MACVLAN_MODE]) 928 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 929 930 if (data && data[IFLA_MACVLAN_FLAGS]) 931 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 932 933 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 934 if (port->count) 935 return -EINVAL; 936 port->passthru = true; 937 eth_hw_addr_inherit(dev, lowerdev); 938 } 939 940 port->count += 1; 941 err = register_netdevice(dev); 942 if (err < 0) 943 goto destroy_port; 944 945 dev->priv_flags |= IFF_MACVLAN; 946 err = netdev_upper_dev_link(lowerdev, dev); 947 if (err) 948 goto unregister_netdev; 949 950 list_add_tail_rcu(&vlan->list, &port->vlans); 951 netif_stacked_transfer_operstate(lowerdev, dev); 952 953 return 0; 954 955 unregister_netdev: 956 unregister_netdevice(dev); 957 destroy_port: 958 port->count -= 1; 959 if (!port->count) 960 macvlan_port_destroy(lowerdev); 961 962 return err; 963 } 964 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 965 966 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 967 struct nlattr *tb[], struct nlattr *data[]) 968 { 969 return macvlan_common_newlink(src_net, dev, tb, data); 970 } 971 972 void macvlan_dellink(struct net_device *dev, struct list_head *head) 973 { 974 struct macvlan_dev *vlan = netdev_priv(dev); 975 976 list_del_rcu(&vlan->list); 977 unregister_netdevice_queue(dev, head); 978 netdev_upper_dev_unlink(vlan->lowerdev, dev); 979 } 980 EXPORT_SYMBOL_GPL(macvlan_dellink); 981 982 static int macvlan_changelink(struct net_device *dev, 983 struct nlattr *tb[], struct nlattr *data[]) 984 { 985 struct macvlan_dev *vlan = netdev_priv(dev); 986 enum macvlan_mode mode; 987 bool set_mode = false; 988 989 /* Validate mode, but don't set yet: setting flags may fail. */ 990 if (data && data[IFLA_MACVLAN_MODE]) { 991 set_mode = true; 992 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 993 /* Passthrough mode can't be set or cleared dynamically */ 994 if ((mode == MACVLAN_MODE_PASSTHRU) != 995 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 996 return -EINVAL; 997 } 998 999 if (data && data[IFLA_MACVLAN_FLAGS]) { 1000 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1001 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1002 if (vlan->port->passthru && promisc) { 1003 int err; 1004 1005 if (flags & MACVLAN_FLAG_NOPROMISC) 1006 err = dev_set_promiscuity(vlan->lowerdev, -1); 1007 else 1008 err = dev_set_promiscuity(vlan->lowerdev, 1); 1009 if (err < 0) 1010 return err; 1011 } 1012 vlan->flags = flags; 1013 } 1014 if (set_mode) 1015 vlan->mode = mode; 1016 return 0; 1017 } 1018 1019 static size_t macvlan_get_size(const struct net_device *dev) 1020 { 1021 return (0 1022 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1023 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1024 ); 1025 } 1026 1027 static int macvlan_fill_info(struct sk_buff *skb, 1028 const struct net_device *dev) 1029 { 1030 struct macvlan_dev *vlan = netdev_priv(dev); 1031 1032 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1033 goto nla_put_failure; 1034 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1035 goto nla_put_failure; 1036 return 0; 1037 1038 nla_put_failure: 1039 return -EMSGSIZE; 1040 } 1041 1042 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1043 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1044 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1045 }; 1046 1047 int macvlan_link_register(struct rtnl_link_ops *ops) 1048 { 1049 /* common fields */ 1050 ops->priv_size = sizeof(struct macvlan_dev); 1051 ops->validate = macvlan_validate; 1052 ops->maxtype = IFLA_MACVLAN_MAX; 1053 ops->policy = macvlan_policy; 1054 ops->changelink = macvlan_changelink; 1055 ops->get_size = macvlan_get_size; 1056 ops->fill_info = macvlan_fill_info; 1057 1058 return rtnl_link_register(ops); 1059 }; 1060 EXPORT_SYMBOL_GPL(macvlan_link_register); 1061 1062 static struct rtnl_link_ops macvlan_link_ops = { 1063 .kind = "macvlan", 1064 .setup = macvlan_setup, 1065 .newlink = macvlan_newlink, 1066 .dellink = macvlan_dellink, 1067 }; 1068 1069 static int macvlan_device_event(struct notifier_block *unused, 1070 unsigned long event, void *ptr) 1071 { 1072 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1073 struct macvlan_dev *vlan, *next; 1074 struct macvlan_port *port; 1075 LIST_HEAD(list_kill); 1076 1077 if (!macvlan_port_exists(dev)) 1078 return NOTIFY_DONE; 1079 1080 port = macvlan_port_get_rtnl(dev); 1081 1082 switch (event) { 1083 case NETDEV_CHANGE: 1084 list_for_each_entry(vlan, &port->vlans, list) 1085 netif_stacked_transfer_operstate(vlan->lowerdev, 1086 vlan->dev); 1087 break; 1088 case NETDEV_FEAT_CHANGE: 1089 list_for_each_entry(vlan, &port->vlans, list) { 1090 vlan->dev->gso_max_size = dev->gso_max_size; 1091 netdev_update_features(vlan->dev); 1092 } 1093 break; 1094 case NETDEV_UNREGISTER: 1095 /* twiddle thumbs on netns device moves */ 1096 if (dev->reg_state != NETREG_UNREGISTERING) 1097 break; 1098 1099 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1100 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1101 unregister_netdevice_many(&list_kill); 1102 list_del(&list_kill); 1103 break; 1104 case NETDEV_PRE_TYPE_CHANGE: 1105 /* Forbid underlaying device to change its type. */ 1106 return NOTIFY_BAD; 1107 } 1108 return NOTIFY_DONE; 1109 } 1110 1111 static struct notifier_block macvlan_notifier_block __read_mostly = { 1112 .notifier_call = macvlan_device_event, 1113 }; 1114 1115 static int __init macvlan_init_module(void) 1116 { 1117 int err; 1118 1119 register_netdevice_notifier(&macvlan_notifier_block); 1120 1121 err = macvlan_link_register(&macvlan_link_ops); 1122 if (err < 0) 1123 goto err1; 1124 return 0; 1125 err1: 1126 unregister_netdevice_notifier(&macvlan_notifier_block); 1127 return err; 1128 } 1129 1130 static void __exit macvlan_cleanup_module(void) 1131 { 1132 rtnl_link_unregister(&macvlan_link_ops); 1133 unregister_netdevice_notifier(&macvlan_notifier_block); 1134 } 1135 1136 module_init(macvlan_init_module); 1137 module_exit(macvlan_cleanup_module); 1138 1139 MODULE_LICENSE("GPL"); 1140 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1141 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1142 MODULE_ALIAS_RTNL_LINK("macvlan"); 1143