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 #include <linux/netpoll.h> 37 38 #define MACVLAN_HASH_BITS 8 39 #define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS) 40 #define MACVLAN_BC_QUEUE_LEN 1000 41 42 struct macvlan_port { 43 struct net_device *dev; 44 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 45 struct list_head vlans; 46 struct rcu_head rcu; 47 struct sk_buff_head bc_queue; 48 struct work_struct bc_work; 49 bool passthru; 50 int count; 51 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 52 }; 53 54 struct macvlan_source_entry { 55 struct hlist_node hlist; 56 struct macvlan_dev *vlan; 57 unsigned char addr[6+2] __aligned(sizeof(u16)); 58 struct rcu_head rcu; 59 }; 60 61 struct macvlan_skb_cb { 62 const struct macvlan_dev *src; 63 }; 64 65 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 66 67 static void macvlan_port_destroy(struct net_device *dev); 68 69 /* Hash Ethernet address */ 70 static u32 macvlan_eth_hash(const unsigned char *addr) 71 { 72 u64 value = get_unaligned((u64 *)addr); 73 74 /* only want 6 bytes */ 75 #ifdef __BIG_ENDIAN 76 value >>= 16; 77 #else 78 value <<= 16; 79 #endif 80 return hash_64(value, MACVLAN_HASH_BITS); 81 } 82 83 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 84 { 85 return rcu_dereference(dev->rx_handler_data); 86 } 87 88 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 89 { 90 return rtnl_dereference(dev->rx_handler_data); 91 } 92 93 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT) 94 95 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 96 const unsigned char *addr) 97 { 98 struct macvlan_dev *vlan; 99 u32 idx = macvlan_eth_hash(addr); 100 101 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) { 102 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 103 return vlan; 104 } 105 return NULL; 106 } 107 108 static struct macvlan_source_entry *macvlan_hash_lookup_source( 109 const struct macvlan_dev *vlan, 110 const unsigned char *addr) 111 { 112 struct macvlan_source_entry *entry; 113 u32 idx = macvlan_eth_hash(addr); 114 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 115 116 hlist_for_each_entry_rcu(entry, h, hlist) { 117 if (ether_addr_equal_64bits(entry->addr, addr) && 118 entry->vlan == vlan) 119 return entry; 120 } 121 return NULL; 122 } 123 124 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 125 const unsigned char *addr) 126 { 127 struct macvlan_port *port = vlan->port; 128 struct macvlan_source_entry *entry; 129 struct hlist_head *h; 130 131 entry = macvlan_hash_lookup_source(vlan, addr); 132 if (entry) 133 return 0; 134 135 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 136 if (!entry) 137 return -ENOMEM; 138 139 ether_addr_copy(entry->addr, addr); 140 entry->vlan = vlan; 141 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 142 hlist_add_head_rcu(&entry->hlist, h); 143 vlan->macaddr_count++; 144 145 return 0; 146 } 147 148 static void macvlan_hash_add(struct macvlan_dev *vlan) 149 { 150 struct macvlan_port *port = vlan->port; 151 const unsigned char *addr = vlan->dev->dev_addr; 152 u32 idx = macvlan_eth_hash(addr); 153 154 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 155 } 156 157 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 158 { 159 hlist_del_rcu(&entry->hlist); 160 kfree_rcu(entry, rcu); 161 } 162 163 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 164 { 165 hlist_del_rcu(&vlan->hlist); 166 if (sync) 167 synchronize_rcu(); 168 } 169 170 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 171 const unsigned char *addr) 172 { 173 macvlan_hash_del(vlan, true); 174 /* Now that we are unhashed it is safe to change the device 175 * address without confusing packet delivery. 176 */ 177 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 178 macvlan_hash_add(vlan); 179 } 180 181 static int macvlan_addr_busy(const struct macvlan_port *port, 182 const unsigned char *addr) 183 { 184 /* Test to see if the specified multicast address is 185 * currently in use by the underlying device or 186 * another macvlan. 187 */ 188 if (ether_addr_equal_64bits(port->dev->dev_addr, addr)) 189 return 1; 190 191 if (macvlan_hash_lookup(port, addr)) 192 return 1; 193 194 return 0; 195 } 196 197 198 static int macvlan_broadcast_one(struct sk_buff *skb, 199 const struct macvlan_dev *vlan, 200 const struct ethhdr *eth, bool local) 201 { 202 struct net_device *dev = vlan->dev; 203 204 if (local) 205 return __dev_forward_skb(dev, skb); 206 207 skb->dev = dev; 208 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 209 skb->pkt_type = PACKET_BROADCAST; 210 else 211 skb->pkt_type = PACKET_MULTICAST; 212 213 return 0; 214 } 215 216 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 217 { 218 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 219 } 220 221 222 static unsigned int mc_hash(const struct macvlan_dev *vlan, 223 const unsigned char *addr) 224 { 225 u32 val = __get_unaligned_cpu32(addr + 2); 226 227 val ^= macvlan_hash_mix(vlan); 228 return hash_32(val, MACVLAN_MC_FILTER_BITS); 229 } 230 231 static void macvlan_broadcast(struct sk_buff *skb, 232 const struct macvlan_port *port, 233 struct net_device *src, 234 enum macvlan_mode mode) 235 { 236 const struct ethhdr *eth = eth_hdr(skb); 237 const struct macvlan_dev *vlan; 238 struct sk_buff *nskb; 239 unsigned int i; 240 int err; 241 unsigned int hash; 242 243 if (skb->protocol == htons(ETH_P_PAUSE)) 244 return; 245 246 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 247 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 248 if (vlan->dev == src || !(vlan->mode & mode)) 249 continue; 250 251 hash = mc_hash(vlan, eth->h_dest); 252 if (!test_bit(hash, vlan->mc_filter)) 253 continue; 254 255 err = NET_RX_DROP; 256 nskb = skb_clone(skb, GFP_ATOMIC); 257 if (likely(nskb)) 258 err = macvlan_broadcast_one( 259 nskb, vlan, eth, 260 mode == MACVLAN_MODE_BRIDGE) ?: 261 netif_rx_ni(nskb); 262 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 263 err == NET_RX_SUCCESS, true); 264 } 265 } 266 } 267 268 static void macvlan_process_broadcast(struct work_struct *w) 269 { 270 struct macvlan_port *port = container_of(w, struct macvlan_port, 271 bc_work); 272 struct sk_buff *skb; 273 struct sk_buff_head list; 274 275 __skb_queue_head_init(&list); 276 277 spin_lock_bh(&port->bc_queue.lock); 278 skb_queue_splice_tail_init(&port->bc_queue, &list); 279 spin_unlock_bh(&port->bc_queue.lock); 280 281 while ((skb = __skb_dequeue(&list))) { 282 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 283 284 rcu_read_lock(); 285 286 if (!src) 287 /* frame comes from an external address */ 288 macvlan_broadcast(skb, port, NULL, 289 MACVLAN_MODE_PRIVATE | 290 MACVLAN_MODE_VEPA | 291 MACVLAN_MODE_PASSTHRU| 292 MACVLAN_MODE_BRIDGE); 293 else if (src->mode == MACVLAN_MODE_VEPA) 294 /* flood to everyone except source */ 295 macvlan_broadcast(skb, port, src->dev, 296 MACVLAN_MODE_VEPA | 297 MACVLAN_MODE_BRIDGE); 298 else 299 /* 300 * flood only to VEPA ports, bridge ports 301 * already saw the frame on the way out. 302 */ 303 macvlan_broadcast(skb, port, src->dev, 304 MACVLAN_MODE_VEPA); 305 306 rcu_read_unlock(); 307 308 kfree_skb(skb); 309 } 310 } 311 312 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 313 struct sk_buff *skb) 314 { 315 struct sk_buff *nskb; 316 int err = -ENOMEM; 317 318 nskb = skb_clone(skb, GFP_ATOMIC); 319 if (!nskb) 320 goto err; 321 322 spin_lock(&port->bc_queue.lock); 323 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) { 324 __skb_queue_tail(&port->bc_queue, nskb); 325 err = 0; 326 } 327 spin_unlock(&port->bc_queue.lock); 328 329 if (err) 330 goto free_nskb; 331 332 schedule_work(&port->bc_work); 333 return; 334 335 free_nskb: 336 kfree_skb(nskb); 337 err: 338 atomic_long_inc(&skb->dev->rx_dropped); 339 } 340 341 static void macvlan_flush_sources(struct macvlan_port *port, 342 struct macvlan_dev *vlan) 343 { 344 int i; 345 346 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 347 struct hlist_node *h, *n; 348 349 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) { 350 struct macvlan_source_entry *entry; 351 352 entry = hlist_entry(h, struct macvlan_source_entry, 353 hlist); 354 if (entry->vlan == vlan) 355 macvlan_hash_del_source(entry); 356 } 357 } 358 vlan->macaddr_count = 0; 359 } 360 361 static void macvlan_forward_source_one(struct sk_buff *skb, 362 struct macvlan_dev *vlan) 363 { 364 struct sk_buff *nskb; 365 struct net_device *dev; 366 int len; 367 int ret; 368 369 dev = vlan->dev; 370 if (unlikely(!(dev->flags & IFF_UP))) 371 return; 372 373 nskb = skb_clone(skb, GFP_ATOMIC); 374 if (!nskb) 375 return; 376 377 len = nskb->len + ETH_HLEN; 378 nskb->dev = dev; 379 nskb->pkt_type = PACKET_HOST; 380 381 ret = netif_rx(nskb); 382 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 383 } 384 385 static void macvlan_forward_source(struct sk_buff *skb, 386 struct macvlan_port *port, 387 const unsigned char *addr) 388 { 389 struct macvlan_source_entry *entry; 390 u32 idx = macvlan_eth_hash(addr); 391 struct hlist_head *h = &port->vlan_source_hash[idx]; 392 393 hlist_for_each_entry_rcu(entry, h, hlist) { 394 if (ether_addr_equal_64bits(entry->addr, addr)) 395 if (entry->vlan->dev->flags & IFF_UP) 396 macvlan_forward_source_one(skb, entry->vlan); 397 } 398 } 399 400 /* called under rcu_read_lock() from netif_receive_skb */ 401 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 402 { 403 struct macvlan_port *port; 404 struct sk_buff *skb = *pskb; 405 const struct ethhdr *eth = eth_hdr(skb); 406 const struct macvlan_dev *vlan; 407 const struct macvlan_dev *src; 408 struct net_device *dev; 409 unsigned int len = 0; 410 int ret; 411 rx_handler_result_t handle_res; 412 413 port = macvlan_port_get_rcu(skb->dev); 414 if (is_multicast_ether_addr(eth->h_dest)) { 415 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN); 416 if (!skb) 417 return RX_HANDLER_CONSUMED; 418 eth = eth_hdr(skb); 419 macvlan_forward_source(skb, port, eth->h_source); 420 src = macvlan_hash_lookup(port, eth->h_source); 421 if (src && src->mode != MACVLAN_MODE_VEPA && 422 src->mode != MACVLAN_MODE_BRIDGE) { 423 /* forward to original port. */ 424 vlan = src; 425 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 426 netif_rx(skb); 427 handle_res = RX_HANDLER_CONSUMED; 428 goto out; 429 } 430 431 MACVLAN_SKB_CB(skb)->src = src; 432 macvlan_broadcast_enqueue(port, skb); 433 434 return RX_HANDLER_PASS; 435 } 436 437 macvlan_forward_source(skb, port, eth->h_source); 438 if (port->passthru) 439 vlan = list_first_or_null_rcu(&port->vlans, 440 struct macvlan_dev, list); 441 else 442 vlan = macvlan_hash_lookup(port, eth->h_dest); 443 if (vlan == NULL) 444 return RX_HANDLER_PASS; 445 446 dev = vlan->dev; 447 if (unlikely(!(dev->flags & IFF_UP))) { 448 kfree_skb(skb); 449 return RX_HANDLER_CONSUMED; 450 } 451 len = skb->len + ETH_HLEN; 452 skb = skb_share_check(skb, GFP_ATOMIC); 453 if (!skb) { 454 ret = NET_RX_DROP; 455 handle_res = RX_HANDLER_CONSUMED; 456 goto out; 457 } 458 459 skb->dev = dev; 460 skb->pkt_type = PACKET_HOST; 461 462 ret = NET_RX_SUCCESS; 463 handle_res = RX_HANDLER_ANOTHER; 464 out: 465 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 466 return handle_res; 467 } 468 469 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 470 { 471 const struct macvlan_dev *vlan = netdev_priv(dev); 472 const struct macvlan_port *port = vlan->port; 473 const struct macvlan_dev *dest; 474 475 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 476 const struct ethhdr *eth = (void *)skb->data; 477 478 /* send to other bridge ports directly */ 479 if (is_multicast_ether_addr(eth->h_dest)) { 480 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 481 goto xmit_world; 482 } 483 484 dest = macvlan_hash_lookup(port, eth->h_dest); 485 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 486 /* send to lowerdev first for its network taps */ 487 dev_forward_skb(vlan->lowerdev, skb); 488 489 return NET_XMIT_SUCCESS; 490 } 491 } 492 493 xmit_world: 494 skb->dev = vlan->lowerdev; 495 return dev_queue_xmit(skb); 496 } 497 498 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 499 { 500 #ifdef CONFIG_NET_POLL_CONTROLLER 501 if (vlan->netpoll) 502 netpoll_send_skb(vlan->netpoll, skb); 503 #else 504 BUG(); 505 #endif 506 return NETDEV_TX_OK; 507 } 508 509 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 510 struct net_device *dev) 511 { 512 unsigned int len = skb->len; 513 int ret; 514 struct macvlan_dev *vlan = netdev_priv(dev); 515 516 if (unlikely(netpoll_tx_running(dev))) 517 return macvlan_netpoll_send_skb(vlan, skb); 518 519 if (vlan->fwd_priv) { 520 skb->dev = vlan->lowerdev; 521 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv); 522 } else { 523 ret = macvlan_queue_xmit(skb, dev); 524 } 525 526 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 527 struct vlan_pcpu_stats *pcpu_stats; 528 529 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 530 u64_stats_update_begin(&pcpu_stats->syncp); 531 pcpu_stats->tx_packets++; 532 pcpu_stats->tx_bytes += len; 533 u64_stats_update_end(&pcpu_stats->syncp); 534 } else { 535 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 536 } 537 return ret; 538 } 539 540 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 541 unsigned short type, const void *daddr, 542 const void *saddr, unsigned len) 543 { 544 const struct macvlan_dev *vlan = netdev_priv(dev); 545 struct net_device *lowerdev = vlan->lowerdev; 546 547 return dev_hard_header(skb, lowerdev, type, daddr, 548 saddr ? : dev->dev_addr, len); 549 } 550 551 static const struct header_ops macvlan_hard_header_ops = { 552 .create = macvlan_hard_header, 553 .parse = eth_header_parse, 554 .cache = eth_header_cache, 555 .cache_update = eth_header_cache_update, 556 }; 557 558 static struct rtnl_link_ops macvlan_link_ops; 559 560 static int macvlan_open(struct net_device *dev) 561 { 562 struct macvlan_dev *vlan = netdev_priv(dev); 563 struct net_device *lowerdev = vlan->lowerdev; 564 int err; 565 566 if (vlan->port->passthru) { 567 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 568 err = dev_set_promiscuity(lowerdev, 1); 569 if (err < 0) 570 goto out; 571 } 572 goto hash_add; 573 } 574 575 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD && 576 dev->rtnl_link_ops == &macvlan_link_ops) { 577 vlan->fwd_priv = 578 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 579 580 /* If we get a NULL pointer back, or if we get an error 581 * then we should just fall through to the non accelerated path 582 */ 583 if (IS_ERR_OR_NULL(vlan->fwd_priv)) { 584 vlan->fwd_priv = NULL; 585 } else 586 return 0; 587 } 588 589 err = -EBUSY; 590 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 591 goto out; 592 593 err = dev_uc_add(lowerdev, dev->dev_addr); 594 if (err < 0) 595 goto out; 596 if (dev->flags & IFF_ALLMULTI) { 597 err = dev_set_allmulti(lowerdev, 1); 598 if (err < 0) 599 goto del_unicast; 600 } 601 602 hash_add: 603 macvlan_hash_add(vlan); 604 return 0; 605 606 del_unicast: 607 dev_uc_del(lowerdev, dev->dev_addr); 608 out: 609 if (vlan->fwd_priv) { 610 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 611 vlan->fwd_priv); 612 vlan->fwd_priv = NULL; 613 } 614 return err; 615 } 616 617 static int macvlan_stop(struct net_device *dev) 618 { 619 struct macvlan_dev *vlan = netdev_priv(dev); 620 struct net_device *lowerdev = vlan->lowerdev; 621 622 if (vlan->fwd_priv) { 623 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 624 vlan->fwd_priv); 625 vlan->fwd_priv = NULL; 626 return 0; 627 } 628 629 dev_uc_unsync(lowerdev, dev); 630 dev_mc_unsync(lowerdev, dev); 631 632 if (vlan->port->passthru) { 633 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 634 dev_set_promiscuity(lowerdev, -1); 635 goto hash_del; 636 } 637 638 if (dev->flags & IFF_ALLMULTI) 639 dev_set_allmulti(lowerdev, -1); 640 641 dev_uc_del(lowerdev, dev->dev_addr); 642 643 hash_del: 644 macvlan_hash_del(vlan, !dev->dismantle); 645 return 0; 646 } 647 648 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 649 { 650 struct macvlan_dev *vlan = netdev_priv(dev); 651 struct net_device *lowerdev = vlan->lowerdev; 652 int err; 653 654 if (!(dev->flags & IFF_UP)) { 655 /* Just copy in the new address */ 656 ether_addr_copy(dev->dev_addr, addr); 657 } else { 658 /* Rehash and update the device filters */ 659 if (macvlan_addr_busy(vlan->port, addr)) 660 return -EBUSY; 661 662 if (!vlan->port->passthru) { 663 err = dev_uc_add(lowerdev, addr); 664 if (err) 665 return err; 666 667 dev_uc_del(lowerdev, dev->dev_addr); 668 } 669 670 macvlan_hash_change_addr(vlan, addr); 671 } 672 return 0; 673 } 674 675 static int macvlan_set_mac_address(struct net_device *dev, void *p) 676 { 677 struct macvlan_dev *vlan = netdev_priv(dev); 678 struct sockaddr *addr = p; 679 680 if (!is_valid_ether_addr(addr->sa_data)) 681 return -EADDRNOTAVAIL; 682 683 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 684 dev_set_mac_address(vlan->lowerdev, addr); 685 return 0; 686 } 687 688 return macvlan_sync_address(dev, addr->sa_data); 689 } 690 691 static void macvlan_change_rx_flags(struct net_device *dev, int change) 692 { 693 struct macvlan_dev *vlan = netdev_priv(dev); 694 struct net_device *lowerdev = vlan->lowerdev; 695 696 if (dev->flags & IFF_UP) { 697 if (change & IFF_ALLMULTI) 698 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 699 } 700 } 701 702 static void macvlan_set_mac_lists(struct net_device *dev) 703 { 704 struct macvlan_dev *vlan = netdev_priv(dev); 705 706 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 707 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ); 708 } else { 709 struct netdev_hw_addr *ha; 710 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 711 712 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 713 netdev_for_each_mc_addr(ha, dev) { 714 __set_bit(mc_hash(vlan, ha->addr), filter); 715 } 716 717 __set_bit(mc_hash(vlan, dev->broadcast), filter); 718 719 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ); 720 } 721 dev_uc_sync(vlan->lowerdev, dev); 722 dev_mc_sync(vlan->lowerdev, dev); 723 } 724 725 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 726 { 727 struct macvlan_dev *vlan = netdev_priv(dev); 728 729 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu) 730 return -EINVAL; 731 dev->mtu = new_mtu; 732 return 0; 733 } 734 735 /* 736 * macvlan network devices have devices nesting below it and are a special 737 * "super class" of normal network devices; split their locks off into a 738 * separate class since they always nest. 739 */ 740 static struct lock_class_key macvlan_netdev_xmit_lock_key; 741 static struct lock_class_key macvlan_netdev_addr_lock_key; 742 743 #define ALWAYS_ON_FEATURES \ 744 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX | \ 745 NETIF_F_GSO_ROBUST) 746 747 #define MACVLAN_FEATURES \ 748 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 749 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_LRO | \ 750 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 751 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 752 753 #define MACVLAN_STATE_MASK \ 754 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 755 756 static int macvlan_get_nest_level(struct net_device *dev) 757 { 758 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level; 759 } 760 761 static void macvlan_set_lockdep_class_one(struct net_device *dev, 762 struct netdev_queue *txq, 763 void *_unused) 764 { 765 lockdep_set_class(&txq->_xmit_lock, 766 &macvlan_netdev_xmit_lock_key); 767 } 768 769 static void macvlan_set_lockdep_class(struct net_device *dev) 770 { 771 lockdep_set_class_and_subclass(&dev->addr_list_lock, 772 &macvlan_netdev_addr_lock_key, 773 macvlan_get_nest_level(dev)); 774 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL); 775 } 776 777 static int macvlan_init(struct net_device *dev) 778 { 779 struct macvlan_dev *vlan = netdev_priv(dev); 780 const struct net_device *lowerdev = vlan->lowerdev; 781 782 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 783 (lowerdev->state & MACVLAN_STATE_MASK); 784 dev->features = lowerdev->features & MACVLAN_FEATURES; 785 dev->features |= ALWAYS_ON_FEATURES; 786 dev->hw_features |= NETIF_F_LRO; 787 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 788 dev->gso_max_size = lowerdev->gso_max_size; 789 dev->iflink = lowerdev->ifindex; 790 dev->hard_header_len = lowerdev->hard_header_len; 791 792 macvlan_set_lockdep_class(dev); 793 794 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 795 if (!vlan->pcpu_stats) 796 return -ENOMEM; 797 798 return 0; 799 } 800 801 static void macvlan_uninit(struct net_device *dev) 802 { 803 struct macvlan_dev *vlan = netdev_priv(dev); 804 struct macvlan_port *port = vlan->port; 805 806 free_percpu(vlan->pcpu_stats); 807 808 macvlan_flush_sources(port, vlan); 809 port->count -= 1; 810 if (!port->count) 811 macvlan_port_destroy(port->dev); 812 } 813 814 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev, 815 struct rtnl_link_stats64 *stats) 816 { 817 struct macvlan_dev *vlan = netdev_priv(dev); 818 819 if (vlan->pcpu_stats) { 820 struct vlan_pcpu_stats *p; 821 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 822 u32 rx_errors = 0, tx_dropped = 0; 823 unsigned int start; 824 int i; 825 826 for_each_possible_cpu(i) { 827 p = per_cpu_ptr(vlan->pcpu_stats, i); 828 do { 829 start = u64_stats_fetch_begin_irq(&p->syncp); 830 rx_packets = p->rx_packets; 831 rx_bytes = p->rx_bytes; 832 rx_multicast = p->rx_multicast; 833 tx_packets = p->tx_packets; 834 tx_bytes = p->tx_bytes; 835 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 836 837 stats->rx_packets += rx_packets; 838 stats->rx_bytes += rx_bytes; 839 stats->multicast += rx_multicast; 840 stats->tx_packets += tx_packets; 841 stats->tx_bytes += tx_bytes; 842 /* rx_errors & tx_dropped are u32, updated 843 * without syncp protection. 844 */ 845 rx_errors += p->rx_errors; 846 tx_dropped += p->tx_dropped; 847 } 848 stats->rx_errors = rx_errors; 849 stats->rx_dropped = rx_errors; 850 stats->tx_dropped = tx_dropped; 851 } 852 return stats; 853 } 854 855 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 856 __be16 proto, u16 vid) 857 { 858 struct macvlan_dev *vlan = netdev_priv(dev); 859 struct net_device *lowerdev = vlan->lowerdev; 860 861 return vlan_vid_add(lowerdev, proto, vid); 862 } 863 864 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 865 __be16 proto, u16 vid) 866 { 867 struct macvlan_dev *vlan = netdev_priv(dev); 868 struct net_device *lowerdev = vlan->lowerdev; 869 870 vlan_vid_del(lowerdev, proto, vid); 871 return 0; 872 } 873 874 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 875 struct net_device *dev, 876 const unsigned char *addr, u16 vid, 877 u16 flags) 878 { 879 struct macvlan_dev *vlan = netdev_priv(dev); 880 int err = -EINVAL; 881 882 /* Support unicast filter only on passthru devices. 883 * Multicast filter should be allowed on all devices. 884 */ 885 if (!vlan->port->passthru && is_unicast_ether_addr(addr)) 886 return -EOPNOTSUPP; 887 888 if (flags & NLM_F_REPLACE) 889 return -EOPNOTSUPP; 890 891 if (is_unicast_ether_addr(addr)) 892 err = dev_uc_add_excl(dev, addr); 893 else if (is_multicast_ether_addr(addr)) 894 err = dev_mc_add_excl(dev, addr); 895 896 return err; 897 } 898 899 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 900 struct net_device *dev, 901 const unsigned char *addr, u16 vid) 902 { 903 struct macvlan_dev *vlan = netdev_priv(dev); 904 int err = -EINVAL; 905 906 /* Support unicast filter only on passthru devices. 907 * Multicast filter should be allowed on all devices. 908 */ 909 if (!vlan->port->passthru && is_unicast_ether_addr(addr)) 910 return -EOPNOTSUPP; 911 912 if (is_unicast_ether_addr(addr)) 913 err = dev_uc_del(dev, addr); 914 else if (is_multicast_ether_addr(addr)) 915 err = dev_mc_del(dev, addr); 916 917 return err; 918 } 919 920 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 921 struct ethtool_drvinfo *drvinfo) 922 { 923 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 924 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 925 } 926 927 static int macvlan_ethtool_get_settings(struct net_device *dev, 928 struct ethtool_cmd *cmd) 929 { 930 const struct macvlan_dev *vlan = netdev_priv(dev); 931 932 return __ethtool_get_settings(vlan->lowerdev, cmd); 933 } 934 935 static netdev_features_t macvlan_fix_features(struct net_device *dev, 936 netdev_features_t features) 937 { 938 struct macvlan_dev *vlan = netdev_priv(dev); 939 netdev_features_t lowerdev_features = vlan->lowerdev->features; 940 netdev_features_t mask; 941 942 features |= NETIF_F_ALL_FOR_ALL; 943 features &= (vlan->set_features | ~MACVLAN_FEATURES); 944 mask = features; 945 946 lowerdev_features &= (features | ~NETIF_F_LRO); 947 features = netdev_increment_features(lowerdev_features, features, mask); 948 features |= ALWAYS_ON_FEATURES; 949 features &= ~NETIF_F_NETNS_LOCAL; 950 951 return features; 952 } 953 954 #ifdef CONFIG_NET_POLL_CONTROLLER 955 static void macvlan_dev_poll_controller(struct net_device *dev) 956 { 957 return; 958 } 959 960 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 961 { 962 struct macvlan_dev *vlan = netdev_priv(dev); 963 struct net_device *real_dev = vlan->lowerdev; 964 struct netpoll *netpoll; 965 int err = 0; 966 967 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 968 err = -ENOMEM; 969 if (!netpoll) 970 goto out; 971 972 err = __netpoll_setup(netpoll, real_dev); 973 if (err) { 974 kfree(netpoll); 975 goto out; 976 } 977 978 vlan->netpoll = netpoll; 979 980 out: 981 return err; 982 } 983 984 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 985 { 986 struct macvlan_dev *vlan = netdev_priv(dev); 987 struct netpoll *netpoll = vlan->netpoll; 988 989 if (!netpoll) 990 return; 991 992 vlan->netpoll = NULL; 993 994 __netpoll_free_async(netpoll); 995 } 996 #endif /* CONFIG_NET_POLL_CONTROLLER */ 997 998 static const struct ethtool_ops macvlan_ethtool_ops = { 999 .get_link = ethtool_op_get_link, 1000 .get_settings = macvlan_ethtool_get_settings, 1001 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1002 }; 1003 1004 static const struct net_device_ops macvlan_netdev_ops = { 1005 .ndo_init = macvlan_init, 1006 .ndo_uninit = macvlan_uninit, 1007 .ndo_open = macvlan_open, 1008 .ndo_stop = macvlan_stop, 1009 .ndo_start_xmit = macvlan_start_xmit, 1010 .ndo_change_mtu = macvlan_change_mtu, 1011 .ndo_fix_features = macvlan_fix_features, 1012 .ndo_change_rx_flags = macvlan_change_rx_flags, 1013 .ndo_set_mac_address = macvlan_set_mac_address, 1014 .ndo_set_rx_mode = macvlan_set_mac_lists, 1015 .ndo_get_stats64 = macvlan_dev_get_stats64, 1016 .ndo_validate_addr = eth_validate_addr, 1017 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1018 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1019 .ndo_fdb_add = macvlan_fdb_add, 1020 .ndo_fdb_del = macvlan_fdb_del, 1021 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1022 .ndo_get_lock_subclass = macvlan_get_nest_level, 1023 #ifdef CONFIG_NET_POLL_CONTROLLER 1024 .ndo_poll_controller = macvlan_dev_poll_controller, 1025 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1026 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1027 #endif 1028 }; 1029 1030 void macvlan_common_setup(struct net_device *dev) 1031 { 1032 ether_setup(dev); 1033 1034 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1035 netif_keep_dst(dev); 1036 dev->priv_flags |= IFF_UNICAST_FLT; 1037 dev->netdev_ops = &macvlan_netdev_ops; 1038 dev->destructor = free_netdev; 1039 dev->header_ops = &macvlan_hard_header_ops; 1040 dev->ethtool_ops = &macvlan_ethtool_ops; 1041 } 1042 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1043 1044 static void macvlan_setup(struct net_device *dev) 1045 { 1046 macvlan_common_setup(dev); 1047 dev->tx_queue_len = 0; 1048 } 1049 1050 static int macvlan_port_create(struct net_device *dev) 1051 { 1052 struct macvlan_port *port; 1053 unsigned int i; 1054 int err; 1055 1056 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1057 return -EINVAL; 1058 1059 if (netif_is_ipvlan_port(dev)) 1060 return -EBUSY; 1061 1062 port = kzalloc(sizeof(*port), GFP_KERNEL); 1063 if (port == NULL) 1064 return -ENOMEM; 1065 1066 port->passthru = false; 1067 port->dev = dev; 1068 INIT_LIST_HEAD(&port->vlans); 1069 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1070 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1071 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1072 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1073 1074 skb_queue_head_init(&port->bc_queue); 1075 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1076 1077 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1078 if (err) 1079 kfree(port); 1080 else 1081 dev->priv_flags |= IFF_MACVLAN_PORT; 1082 return err; 1083 } 1084 1085 static void macvlan_port_destroy(struct net_device *dev) 1086 { 1087 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1088 1089 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1090 netdev_rx_handler_unregister(dev); 1091 1092 /* After this point, no packet can schedule bc_work anymore, 1093 * but we need to cancel it and purge left skbs if any. 1094 */ 1095 cancel_work_sync(&port->bc_work); 1096 __skb_queue_purge(&port->bc_queue); 1097 1098 kfree_rcu(port, rcu); 1099 } 1100 1101 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[]) 1102 { 1103 if (tb[IFLA_ADDRESS]) { 1104 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1105 return -EINVAL; 1106 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1107 return -EADDRNOTAVAIL; 1108 } 1109 1110 if (data && data[IFLA_MACVLAN_FLAGS] && 1111 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1112 return -EINVAL; 1113 1114 if (data && data[IFLA_MACVLAN_MODE]) { 1115 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1116 case MACVLAN_MODE_PRIVATE: 1117 case MACVLAN_MODE_VEPA: 1118 case MACVLAN_MODE_BRIDGE: 1119 case MACVLAN_MODE_PASSTHRU: 1120 case MACVLAN_MODE_SOURCE: 1121 break; 1122 default: 1123 return -EINVAL; 1124 } 1125 } 1126 1127 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1128 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1129 case MACVLAN_MACADDR_ADD: 1130 case MACVLAN_MACADDR_DEL: 1131 case MACVLAN_MACADDR_FLUSH: 1132 case MACVLAN_MACADDR_SET: 1133 break; 1134 default: 1135 return -EINVAL; 1136 } 1137 } 1138 1139 if (data && data[IFLA_MACVLAN_MACADDR]) { 1140 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1141 return -EINVAL; 1142 1143 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1144 return -EADDRNOTAVAIL; 1145 } 1146 1147 if (data && data[IFLA_MACVLAN_MACADDR_COUNT]) 1148 return -EINVAL; 1149 1150 return 0; 1151 } 1152 1153 /** 1154 * reconfigure list of remote source mac address 1155 * (only for macvlan devices in source mode) 1156 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1157 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1158 */ 1159 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1160 struct nlattr *data[]) 1161 { 1162 char *addr = NULL; 1163 int ret, rem, len; 1164 struct nlattr *nla, *head; 1165 struct macvlan_source_entry *entry; 1166 1167 if (data[IFLA_MACVLAN_MACADDR]) 1168 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1169 1170 if (mode == MACVLAN_MACADDR_ADD) { 1171 if (!addr) 1172 return -EINVAL; 1173 1174 return macvlan_hash_add_source(vlan, addr); 1175 1176 } else if (mode == MACVLAN_MACADDR_DEL) { 1177 if (!addr) 1178 return -EINVAL; 1179 1180 entry = macvlan_hash_lookup_source(vlan, addr); 1181 if (entry) { 1182 macvlan_hash_del_source(entry); 1183 vlan->macaddr_count--; 1184 } 1185 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1186 macvlan_flush_sources(vlan->port, vlan); 1187 } else if (mode == MACVLAN_MACADDR_SET) { 1188 macvlan_flush_sources(vlan->port, vlan); 1189 1190 if (addr) { 1191 ret = macvlan_hash_add_source(vlan, addr); 1192 if (ret) 1193 return ret; 1194 } 1195 1196 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1197 return 0; 1198 1199 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1200 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1201 1202 nla_for_each_attr(nla, head, len, rem) { 1203 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1204 nla_len(nla) != ETH_ALEN) 1205 continue; 1206 1207 addr = nla_data(nla); 1208 ret = macvlan_hash_add_source(vlan, addr); 1209 if (ret) 1210 return ret; 1211 } 1212 } else { 1213 return -EINVAL; 1214 } 1215 1216 return 0; 1217 } 1218 1219 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1220 struct nlattr *tb[], struct nlattr *data[]) 1221 { 1222 struct macvlan_dev *vlan = netdev_priv(dev); 1223 struct macvlan_port *port; 1224 struct net_device *lowerdev; 1225 int err; 1226 int macmode; 1227 1228 if (!tb[IFLA_LINK]) 1229 return -EINVAL; 1230 1231 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1232 if (lowerdev == NULL) 1233 return -ENODEV; 1234 1235 /* When creating macvlans or macvtaps on top of other macvlans - use 1236 * the real device as the lowerdev. 1237 */ 1238 if (netif_is_macvlan(lowerdev)) 1239 lowerdev = macvlan_dev_real_dev(lowerdev); 1240 1241 if (!tb[IFLA_MTU]) 1242 dev->mtu = lowerdev->mtu; 1243 else if (dev->mtu > lowerdev->mtu) 1244 return -EINVAL; 1245 1246 if (!tb[IFLA_ADDRESS]) 1247 eth_hw_addr_random(dev); 1248 1249 if (!macvlan_port_exists(lowerdev)) { 1250 err = macvlan_port_create(lowerdev); 1251 if (err < 0) 1252 return err; 1253 } 1254 port = macvlan_port_get_rtnl(lowerdev); 1255 1256 /* Only 1 macvlan device can be created in passthru mode */ 1257 if (port->passthru) 1258 return -EINVAL; 1259 1260 vlan->lowerdev = lowerdev; 1261 vlan->dev = dev; 1262 vlan->port = port; 1263 vlan->set_features = MACVLAN_FEATURES; 1264 vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1; 1265 1266 vlan->mode = MACVLAN_MODE_VEPA; 1267 if (data && data[IFLA_MACVLAN_MODE]) 1268 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1269 1270 if (data && data[IFLA_MACVLAN_FLAGS]) 1271 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1272 1273 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1274 if (port->count) 1275 return -EINVAL; 1276 port->passthru = true; 1277 eth_hw_addr_inherit(dev, lowerdev); 1278 } 1279 1280 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1281 if (vlan->mode != MACVLAN_MODE_SOURCE) 1282 return -EINVAL; 1283 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1284 err = macvlan_changelink_sources(vlan, macmode, data); 1285 if (err) 1286 return err; 1287 } 1288 1289 port->count += 1; 1290 err = register_netdevice(dev); 1291 if (err < 0) 1292 goto destroy_port; 1293 1294 dev->priv_flags |= IFF_MACVLAN; 1295 err = netdev_upper_dev_link(lowerdev, dev); 1296 if (err) 1297 goto unregister_netdev; 1298 1299 list_add_tail_rcu(&vlan->list, &port->vlans); 1300 netif_stacked_transfer_operstate(lowerdev, dev); 1301 1302 return 0; 1303 1304 unregister_netdev: 1305 unregister_netdevice(dev); 1306 destroy_port: 1307 port->count -= 1; 1308 if (!port->count) 1309 macvlan_port_destroy(lowerdev); 1310 1311 return err; 1312 } 1313 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1314 1315 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1316 struct nlattr *tb[], struct nlattr *data[]) 1317 { 1318 return macvlan_common_newlink(src_net, dev, tb, data); 1319 } 1320 1321 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1322 { 1323 struct macvlan_dev *vlan = netdev_priv(dev); 1324 1325 if (vlan->mode == MACVLAN_MODE_SOURCE) 1326 macvlan_flush_sources(vlan->port, vlan); 1327 list_del_rcu(&vlan->list); 1328 unregister_netdevice_queue(dev, head); 1329 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1330 } 1331 EXPORT_SYMBOL_GPL(macvlan_dellink); 1332 1333 static int macvlan_changelink(struct net_device *dev, 1334 struct nlattr *tb[], struct nlattr *data[]) 1335 { 1336 struct macvlan_dev *vlan = netdev_priv(dev); 1337 enum macvlan_mode mode; 1338 bool set_mode = false; 1339 enum macvlan_macaddr_mode macmode; 1340 int ret; 1341 1342 /* Validate mode, but don't set yet: setting flags may fail. */ 1343 if (data && data[IFLA_MACVLAN_MODE]) { 1344 set_mode = true; 1345 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1346 /* Passthrough mode can't be set or cleared dynamically */ 1347 if ((mode == MACVLAN_MODE_PASSTHRU) != 1348 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1349 return -EINVAL; 1350 if (vlan->mode == MACVLAN_MODE_SOURCE && 1351 vlan->mode != mode) 1352 macvlan_flush_sources(vlan->port, vlan); 1353 } 1354 1355 if (data && data[IFLA_MACVLAN_FLAGS]) { 1356 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1357 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1358 if (vlan->port->passthru && promisc) { 1359 int err; 1360 1361 if (flags & MACVLAN_FLAG_NOPROMISC) 1362 err = dev_set_promiscuity(vlan->lowerdev, -1); 1363 else 1364 err = dev_set_promiscuity(vlan->lowerdev, 1); 1365 if (err < 0) 1366 return err; 1367 } 1368 vlan->flags = flags; 1369 } 1370 if (set_mode) 1371 vlan->mode = mode; 1372 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1373 if (vlan->mode != MACVLAN_MODE_SOURCE) 1374 return -EINVAL; 1375 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1376 ret = macvlan_changelink_sources(vlan, macmode, data); 1377 if (ret) 1378 return ret; 1379 } 1380 return 0; 1381 } 1382 1383 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1384 { 1385 if (vlan->macaddr_count == 0) 1386 return 0; 1387 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1388 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1389 } 1390 1391 static size_t macvlan_get_size(const struct net_device *dev) 1392 { 1393 struct macvlan_dev *vlan = netdev_priv(dev); 1394 1395 return (0 1396 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1397 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1398 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1399 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1400 ); 1401 } 1402 1403 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1404 const struct macvlan_dev *vlan, 1405 const int i) 1406 { 1407 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1408 struct macvlan_source_entry *entry; 1409 1410 hlist_for_each_entry_rcu(entry, h, hlist) { 1411 if (entry->vlan != vlan) 1412 continue; 1413 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1414 return 1; 1415 } 1416 return 0; 1417 } 1418 1419 static int macvlan_fill_info(struct sk_buff *skb, 1420 const struct net_device *dev) 1421 { 1422 struct macvlan_dev *vlan = netdev_priv(dev); 1423 int i; 1424 struct nlattr *nest; 1425 1426 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1427 goto nla_put_failure; 1428 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1429 goto nla_put_failure; 1430 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1431 goto nla_put_failure; 1432 if (vlan->macaddr_count > 0) { 1433 nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA); 1434 if (nest == NULL) 1435 goto nla_put_failure; 1436 1437 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1438 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1439 goto nla_put_failure; 1440 } 1441 nla_nest_end(skb, nest); 1442 } 1443 return 0; 1444 1445 nla_put_failure: 1446 return -EMSGSIZE; 1447 } 1448 1449 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1450 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1451 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1452 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1453 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1454 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1455 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1456 }; 1457 1458 int macvlan_link_register(struct rtnl_link_ops *ops) 1459 { 1460 /* common fields */ 1461 ops->priv_size = sizeof(struct macvlan_dev); 1462 ops->validate = macvlan_validate; 1463 ops->maxtype = IFLA_MACVLAN_MAX; 1464 ops->policy = macvlan_policy; 1465 ops->changelink = macvlan_changelink; 1466 ops->get_size = macvlan_get_size; 1467 ops->fill_info = macvlan_fill_info; 1468 1469 return rtnl_link_register(ops); 1470 }; 1471 EXPORT_SYMBOL_GPL(macvlan_link_register); 1472 1473 static struct net *macvlan_get_link_net(const struct net_device *dev) 1474 { 1475 return dev_net(macvlan_dev_real_dev(dev)); 1476 } 1477 1478 static struct rtnl_link_ops macvlan_link_ops = { 1479 .kind = "macvlan", 1480 .setup = macvlan_setup, 1481 .newlink = macvlan_newlink, 1482 .dellink = macvlan_dellink, 1483 .get_link_net = macvlan_get_link_net, 1484 }; 1485 1486 static int macvlan_device_event(struct notifier_block *unused, 1487 unsigned long event, void *ptr) 1488 { 1489 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1490 struct macvlan_dev *vlan, *next; 1491 struct macvlan_port *port; 1492 LIST_HEAD(list_kill); 1493 1494 if (!macvlan_port_exists(dev)) 1495 return NOTIFY_DONE; 1496 1497 port = macvlan_port_get_rtnl(dev); 1498 1499 switch (event) { 1500 case NETDEV_CHANGE: 1501 list_for_each_entry(vlan, &port->vlans, list) 1502 netif_stacked_transfer_operstate(vlan->lowerdev, 1503 vlan->dev); 1504 break; 1505 case NETDEV_FEAT_CHANGE: 1506 list_for_each_entry(vlan, &port->vlans, list) { 1507 vlan->dev->gso_max_size = dev->gso_max_size; 1508 netdev_update_features(vlan->dev); 1509 } 1510 break; 1511 case NETDEV_CHANGEMTU: 1512 list_for_each_entry(vlan, &port->vlans, list) { 1513 if (vlan->dev->mtu <= dev->mtu) 1514 continue; 1515 dev_set_mtu(vlan->dev, dev->mtu); 1516 } 1517 break; 1518 case NETDEV_CHANGEADDR: 1519 if (!port->passthru) 1520 return NOTIFY_DONE; 1521 1522 vlan = list_first_entry_or_null(&port->vlans, 1523 struct macvlan_dev, 1524 list); 1525 1526 if (macvlan_sync_address(vlan->dev, dev->dev_addr)) 1527 return NOTIFY_BAD; 1528 1529 break; 1530 case NETDEV_UNREGISTER: 1531 /* twiddle thumbs on netns device moves */ 1532 if (dev->reg_state != NETREG_UNREGISTERING) 1533 break; 1534 1535 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1536 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1537 unregister_netdevice_many(&list_kill); 1538 break; 1539 case NETDEV_PRE_TYPE_CHANGE: 1540 /* Forbid underlaying device to change its type. */ 1541 return NOTIFY_BAD; 1542 1543 case NETDEV_NOTIFY_PEERS: 1544 case NETDEV_BONDING_FAILOVER: 1545 case NETDEV_RESEND_IGMP: 1546 /* Propagate to all vlans */ 1547 list_for_each_entry(vlan, &port->vlans, list) 1548 call_netdevice_notifiers(event, vlan->dev); 1549 } 1550 return NOTIFY_DONE; 1551 } 1552 1553 static struct notifier_block macvlan_notifier_block __read_mostly = { 1554 .notifier_call = macvlan_device_event, 1555 }; 1556 1557 static int __init macvlan_init_module(void) 1558 { 1559 int err; 1560 1561 register_netdevice_notifier(&macvlan_notifier_block); 1562 1563 err = macvlan_link_register(&macvlan_link_ops); 1564 if (err < 0) 1565 goto err1; 1566 return 0; 1567 err1: 1568 unregister_netdevice_notifier(&macvlan_notifier_block); 1569 return err; 1570 } 1571 1572 static void __exit macvlan_cleanup_module(void) 1573 { 1574 rtnl_link_unregister(&macvlan_link_ops); 1575 unregister_netdevice_notifier(&macvlan_notifier_block); 1576 } 1577 1578 module_init(macvlan_init_module); 1579 module_exit(macvlan_cleanup_module); 1580 1581 MODULE_LICENSE("GPL"); 1582 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1583 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1584 MODULE_ALIAS_RTNL_LINK("macvlan"); 1585