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 #define MACVLAN_F_PASSTHRU 1 43 #define MACVLAN_F_ADDRCHANGE 2 44 45 struct macvlan_port { 46 struct net_device *dev; 47 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 48 struct list_head vlans; 49 struct sk_buff_head bc_queue; 50 struct work_struct bc_work; 51 u32 flags; 52 int count; 53 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 54 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 55 unsigned char perm_addr[ETH_ALEN]; 56 }; 57 58 struct macvlan_source_entry { 59 struct hlist_node hlist; 60 struct macvlan_dev *vlan; 61 unsigned char addr[6+2] __aligned(sizeof(u16)); 62 struct rcu_head rcu; 63 }; 64 65 struct macvlan_skb_cb { 66 const struct macvlan_dev *src; 67 }; 68 69 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 70 71 static void macvlan_port_destroy(struct net_device *dev); 72 73 static inline bool macvlan_passthru(const struct macvlan_port *port) 74 { 75 return port->flags & MACVLAN_F_PASSTHRU; 76 } 77 78 static inline void macvlan_set_passthru(struct macvlan_port *port) 79 { 80 port->flags |= MACVLAN_F_PASSTHRU; 81 } 82 83 static inline bool macvlan_addr_change(const struct macvlan_port *port) 84 { 85 return port->flags & MACVLAN_F_ADDRCHANGE; 86 } 87 88 static inline void macvlan_set_addr_change(struct macvlan_port *port) 89 { 90 port->flags |= MACVLAN_F_ADDRCHANGE; 91 } 92 93 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 94 { 95 port->flags &= ~MACVLAN_F_ADDRCHANGE; 96 } 97 98 /* Hash Ethernet address */ 99 static u32 macvlan_eth_hash(const unsigned char *addr) 100 { 101 u64 value = get_unaligned((u64 *)addr); 102 103 /* only want 6 bytes */ 104 #ifdef __BIG_ENDIAN 105 value >>= 16; 106 #else 107 value <<= 16; 108 #endif 109 return hash_64(value, MACVLAN_HASH_BITS); 110 } 111 112 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 113 { 114 return rcu_dereference(dev->rx_handler_data); 115 } 116 117 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 118 { 119 return rtnl_dereference(dev->rx_handler_data); 120 } 121 122 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT) 123 124 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 125 const unsigned char *addr) 126 { 127 struct macvlan_dev *vlan; 128 u32 idx = macvlan_eth_hash(addr); 129 130 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) { 131 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 132 return vlan; 133 } 134 return NULL; 135 } 136 137 static struct macvlan_source_entry *macvlan_hash_lookup_source( 138 const struct macvlan_dev *vlan, 139 const unsigned char *addr) 140 { 141 struct macvlan_source_entry *entry; 142 u32 idx = macvlan_eth_hash(addr); 143 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 144 145 hlist_for_each_entry_rcu(entry, h, hlist) { 146 if (ether_addr_equal_64bits(entry->addr, addr) && 147 entry->vlan == vlan) 148 return entry; 149 } 150 return NULL; 151 } 152 153 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 154 const unsigned char *addr) 155 { 156 struct macvlan_port *port = vlan->port; 157 struct macvlan_source_entry *entry; 158 struct hlist_head *h; 159 160 entry = macvlan_hash_lookup_source(vlan, addr); 161 if (entry) 162 return 0; 163 164 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 165 if (!entry) 166 return -ENOMEM; 167 168 ether_addr_copy(entry->addr, addr); 169 entry->vlan = vlan; 170 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 171 hlist_add_head_rcu(&entry->hlist, h); 172 vlan->macaddr_count++; 173 174 return 0; 175 } 176 177 static void macvlan_hash_add(struct macvlan_dev *vlan) 178 { 179 struct macvlan_port *port = vlan->port; 180 const unsigned char *addr = vlan->dev->dev_addr; 181 u32 idx = macvlan_eth_hash(addr); 182 183 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 184 } 185 186 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 187 { 188 hlist_del_rcu(&entry->hlist); 189 kfree_rcu(entry, rcu); 190 } 191 192 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 193 { 194 hlist_del_rcu(&vlan->hlist); 195 if (sync) 196 synchronize_rcu(); 197 } 198 199 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 200 const unsigned char *addr) 201 { 202 macvlan_hash_del(vlan, true); 203 /* Now that we are unhashed it is safe to change the device 204 * address without confusing packet delivery. 205 */ 206 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 207 macvlan_hash_add(vlan); 208 } 209 210 static bool macvlan_addr_busy(const struct macvlan_port *port, 211 const unsigned char *addr) 212 { 213 /* Test to see if the specified address is 214 * currently in use by the underlying device or 215 * another macvlan. 216 */ 217 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 218 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 219 return true; 220 221 if (macvlan_hash_lookup(port, addr)) 222 return true; 223 224 return false; 225 } 226 227 228 static int macvlan_broadcast_one(struct sk_buff *skb, 229 const struct macvlan_dev *vlan, 230 const struct ethhdr *eth, bool local) 231 { 232 struct net_device *dev = vlan->dev; 233 234 if (local) 235 return __dev_forward_skb(dev, skb); 236 237 skb->dev = dev; 238 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 239 skb->pkt_type = PACKET_BROADCAST; 240 else 241 skb->pkt_type = PACKET_MULTICAST; 242 243 return 0; 244 } 245 246 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 247 { 248 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 249 } 250 251 252 static unsigned int mc_hash(const struct macvlan_dev *vlan, 253 const unsigned char *addr) 254 { 255 u32 val = __get_unaligned_cpu32(addr + 2); 256 257 val ^= macvlan_hash_mix(vlan); 258 return hash_32(val, MACVLAN_MC_FILTER_BITS); 259 } 260 261 static void macvlan_broadcast(struct sk_buff *skb, 262 const struct macvlan_port *port, 263 struct net_device *src, 264 enum macvlan_mode mode) 265 { 266 const struct ethhdr *eth = eth_hdr(skb); 267 const struct macvlan_dev *vlan; 268 struct sk_buff *nskb; 269 unsigned int i; 270 int err; 271 unsigned int hash; 272 273 if (skb->protocol == htons(ETH_P_PAUSE)) 274 return; 275 276 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 277 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 278 if (vlan->dev == src || !(vlan->mode & mode)) 279 continue; 280 281 hash = mc_hash(vlan, eth->h_dest); 282 if (!test_bit(hash, vlan->mc_filter)) 283 continue; 284 285 err = NET_RX_DROP; 286 nskb = skb_clone(skb, GFP_ATOMIC); 287 if (likely(nskb)) 288 err = macvlan_broadcast_one( 289 nskb, vlan, eth, 290 mode == MACVLAN_MODE_BRIDGE) ?: 291 netif_rx_ni(nskb); 292 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 293 err == NET_RX_SUCCESS, true); 294 } 295 } 296 } 297 298 static void macvlan_process_broadcast(struct work_struct *w) 299 { 300 struct macvlan_port *port = container_of(w, struct macvlan_port, 301 bc_work); 302 struct sk_buff *skb; 303 struct sk_buff_head list; 304 305 __skb_queue_head_init(&list); 306 307 spin_lock_bh(&port->bc_queue.lock); 308 skb_queue_splice_tail_init(&port->bc_queue, &list); 309 spin_unlock_bh(&port->bc_queue.lock); 310 311 while ((skb = __skb_dequeue(&list))) { 312 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 313 314 rcu_read_lock(); 315 316 if (!src) 317 /* frame comes from an external address */ 318 macvlan_broadcast(skb, port, NULL, 319 MACVLAN_MODE_PRIVATE | 320 MACVLAN_MODE_VEPA | 321 MACVLAN_MODE_PASSTHRU| 322 MACVLAN_MODE_BRIDGE); 323 else if (src->mode == MACVLAN_MODE_VEPA) 324 /* flood to everyone except source */ 325 macvlan_broadcast(skb, port, src->dev, 326 MACVLAN_MODE_VEPA | 327 MACVLAN_MODE_BRIDGE); 328 else 329 /* 330 * flood only to VEPA ports, bridge ports 331 * already saw the frame on the way out. 332 */ 333 macvlan_broadcast(skb, port, src->dev, 334 MACVLAN_MODE_VEPA); 335 336 rcu_read_unlock(); 337 338 if (src) 339 dev_put(src->dev); 340 kfree_skb(skb); 341 } 342 } 343 344 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 345 const struct macvlan_dev *src, 346 struct sk_buff *skb) 347 { 348 struct sk_buff *nskb; 349 int err = -ENOMEM; 350 351 nskb = skb_clone(skb, GFP_ATOMIC); 352 if (!nskb) 353 goto err; 354 355 MACVLAN_SKB_CB(nskb)->src = src; 356 357 spin_lock(&port->bc_queue.lock); 358 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) { 359 if (src) 360 dev_hold(src->dev); 361 __skb_queue_tail(&port->bc_queue, nskb); 362 err = 0; 363 } 364 spin_unlock(&port->bc_queue.lock); 365 366 if (err) 367 goto free_nskb; 368 369 schedule_work(&port->bc_work); 370 return; 371 372 free_nskb: 373 kfree_skb(nskb); 374 err: 375 atomic_long_inc(&skb->dev->rx_dropped); 376 } 377 378 static void macvlan_flush_sources(struct macvlan_port *port, 379 struct macvlan_dev *vlan) 380 { 381 int i; 382 383 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 384 struct hlist_node *h, *n; 385 386 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) { 387 struct macvlan_source_entry *entry; 388 389 entry = hlist_entry(h, struct macvlan_source_entry, 390 hlist); 391 if (entry->vlan == vlan) 392 macvlan_hash_del_source(entry); 393 } 394 } 395 vlan->macaddr_count = 0; 396 } 397 398 static void macvlan_forward_source_one(struct sk_buff *skb, 399 struct macvlan_dev *vlan) 400 { 401 struct sk_buff *nskb; 402 struct net_device *dev; 403 int len; 404 int ret; 405 406 dev = vlan->dev; 407 if (unlikely(!(dev->flags & IFF_UP))) 408 return; 409 410 nskb = skb_clone(skb, GFP_ATOMIC); 411 if (!nskb) 412 return; 413 414 len = nskb->len + ETH_HLEN; 415 nskb->dev = dev; 416 417 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr)) 418 nskb->pkt_type = PACKET_HOST; 419 420 ret = netif_rx(nskb); 421 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 422 } 423 424 static void macvlan_forward_source(struct sk_buff *skb, 425 struct macvlan_port *port, 426 const unsigned char *addr) 427 { 428 struct macvlan_source_entry *entry; 429 u32 idx = macvlan_eth_hash(addr); 430 struct hlist_head *h = &port->vlan_source_hash[idx]; 431 432 hlist_for_each_entry_rcu(entry, h, hlist) { 433 if (ether_addr_equal_64bits(entry->addr, addr)) 434 macvlan_forward_source_one(skb, entry->vlan); 435 } 436 } 437 438 /* called under rcu_read_lock() from netif_receive_skb */ 439 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 440 { 441 struct macvlan_port *port; 442 struct sk_buff *skb = *pskb; 443 const struct ethhdr *eth = eth_hdr(skb); 444 const struct macvlan_dev *vlan; 445 const struct macvlan_dev *src; 446 struct net_device *dev; 447 unsigned int len = 0; 448 int ret; 449 rx_handler_result_t handle_res; 450 451 port = macvlan_port_get_rcu(skb->dev); 452 if (is_multicast_ether_addr(eth->h_dest)) { 453 unsigned int hash; 454 455 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 456 if (!skb) 457 return RX_HANDLER_CONSUMED; 458 *pskb = skb; 459 eth = eth_hdr(skb); 460 macvlan_forward_source(skb, port, eth->h_source); 461 src = macvlan_hash_lookup(port, eth->h_source); 462 if (src && src->mode != MACVLAN_MODE_VEPA && 463 src->mode != MACVLAN_MODE_BRIDGE) { 464 /* forward to original port. */ 465 vlan = src; 466 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 467 netif_rx(skb); 468 handle_res = RX_HANDLER_CONSUMED; 469 goto out; 470 } 471 472 hash = mc_hash(NULL, eth->h_dest); 473 if (test_bit(hash, port->mc_filter)) 474 macvlan_broadcast_enqueue(port, src, skb); 475 476 return RX_HANDLER_PASS; 477 } 478 479 macvlan_forward_source(skb, port, eth->h_source); 480 if (macvlan_passthru(port)) 481 vlan = list_first_or_null_rcu(&port->vlans, 482 struct macvlan_dev, list); 483 else 484 vlan = macvlan_hash_lookup(port, eth->h_dest); 485 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 486 return RX_HANDLER_PASS; 487 488 dev = vlan->dev; 489 if (unlikely(!(dev->flags & IFF_UP))) { 490 kfree_skb(skb); 491 return RX_HANDLER_CONSUMED; 492 } 493 len = skb->len + ETH_HLEN; 494 skb = skb_share_check(skb, GFP_ATOMIC); 495 if (!skb) { 496 ret = NET_RX_DROP; 497 handle_res = RX_HANDLER_CONSUMED; 498 goto out; 499 } 500 501 *pskb = skb; 502 skb->dev = dev; 503 skb->pkt_type = PACKET_HOST; 504 505 ret = NET_RX_SUCCESS; 506 handle_res = RX_HANDLER_ANOTHER; 507 out: 508 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 509 return handle_res; 510 } 511 512 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 513 { 514 const struct macvlan_dev *vlan = netdev_priv(dev); 515 const struct macvlan_port *port = vlan->port; 516 const struct macvlan_dev *dest; 517 518 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 519 const struct ethhdr *eth = (void *)skb->data; 520 521 /* send to other bridge ports directly */ 522 if (is_multicast_ether_addr(eth->h_dest)) { 523 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 524 goto xmit_world; 525 } 526 527 dest = macvlan_hash_lookup(port, eth->h_dest); 528 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 529 /* send to lowerdev first for its network taps */ 530 dev_forward_skb(vlan->lowerdev, skb); 531 532 return NET_XMIT_SUCCESS; 533 } 534 } 535 536 xmit_world: 537 skb->dev = vlan->lowerdev; 538 return dev_queue_xmit(skb); 539 } 540 541 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 542 { 543 #ifdef CONFIG_NET_POLL_CONTROLLER 544 if (vlan->netpoll) 545 netpoll_send_skb(vlan->netpoll, skb); 546 #else 547 BUG(); 548 #endif 549 return NETDEV_TX_OK; 550 } 551 552 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 553 struct net_device *dev) 554 { 555 unsigned int len = skb->len; 556 int ret; 557 struct macvlan_dev *vlan = netdev_priv(dev); 558 559 if (unlikely(netpoll_tx_running(dev))) 560 return macvlan_netpoll_send_skb(vlan, skb); 561 562 if (vlan->fwd_priv) { 563 skb->dev = vlan->lowerdev; 564 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv); 565 } else { 566 ret = macvlan_queue_xmit(skb, dev); 567 } 568 569 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 570 struct vlan_pcpu_stats *pcpu_stats; 571 572 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 573 u64_stats_update_begin(&pcpu_stats->syncp); 574 pcpu_stats->tx_packets++; 575 pcpu_stats->tx_bytes += len; 576 u64_stats_update_end(&pcpu_stats->syncp); 577 } else { 578 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 579 } 580 return ret; 581 } 582 583 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 584 unsigned short type, const void *daddr, 585 const void *saddr, unsigned len) 586 { 587 const struct macvlan_dev *vlan = netdev_priv(dev); 588 struct net_device *lowerdev = vlan->lowerdev; 589 590 return dev_hard_header(skb, lowerdev, type, daddr, 591 saddr ? : dev->dev_addr, len); 592 } 593 594 static const struct header_ops macvlan_hard_header_ops = { 595 .create = macvlan_hard_header, 596 .parse = eth_header_parse, 597 .cache = eth_header_cache, 598 .cache_update = eth_header_cache_update, 599 }; 600 601 static struct rtnl_link_ops macvlan_link_ops; 602 603 static int macvlan_open(struct net_device *dev) 604 { 605 struct macvlan_dev *vlan = netdev_priv(dev); 606 struct net_device *lowerdev = vlan->lowerdev; 607 int err; 608 609 if (macvlan_passthru(vlan->port)) { 610 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 611 err = dev_set_promiscuity(lowerdev, 1); 612 if (err < 0) 613 goto out; 614 } 615 goto hash_add; 616 } 617 618 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD && 619 dev->rtnl_link_ops == &macvlan_link_ops) { 620 vlan->fwd_priv = 621 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 622 623 /* If we get a NULL pointer back, or if we get an error 624 * then we should just fall through to the non accelerated path 625 */ 626 if (IS_ERR_OR_NULL(vlan->fwd_priv)) { 627 vlan->fwd_priv = NULL; 628 } else 629 return 0; 630 } 631 632 err = -EBUSY; 633 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 634 goto out; 635 636 err = dev_uc_add(lowerdev, dev->dev_addr); 637 if (err < 0) 638 goto out; 639 if (dev->flags & IFF_ALLMULTI) { 640 err = dev_set_allmulti(lowerdev, 1); 641 if (err < 0) 642 goto del_unicast; 643 } 644 645 if (dev->flags & IFF_PROMISC) { 646 err = dev_set_promiscuity(lowerdev, 1); 647 if (err < 0) 648 goto clear_multi; 649 } 650 651 hash_add: 652 macvlan_hash_add(vlan); 653 return 0; 654 655 clear_multi: 656 if (dev->flags & IFF_ALLMULTI) 657 dev_set_allmulti(lowerdev, -1); 658 del_unicast: 659 dev_uc_del(lowerdev, dev->dev_addr); 660 out: 661 if (vlan->fwd_priv) { 662 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 663 vlan->fwd_priv); 664 vlan->fwd_priv = NULL; 665 } 666 return err; 667 } 668 669 static int macvlan_stop(struct net_device *dev) 670 { 671 struct macvlan_dev *vlan = netdev_priv(dev); 672 struct net_device *lowerdev = vlan->lowerdev; 673 674 if (vlan->fwd_priv) { 675 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 676 vlan->fwd_priv); 677 vlan->fwd_priv = NULL; 678 return 0; 679 } 680 681 dev_uc_unsync(lowerdev, dev); 682 dev_mc_unsync(lowerdev, dev); 683 684 if (macvlan_passthru(vlan->port)) { 685 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 686 dev_set_promiscuity(lowerdev, -1); 687 goto hash_del; 688 } 689 690 if (dev->flags & IFF_ALLMULTI) 691 dev_set_allmulti(lowerdev, -1); 692 693 if (dev->flags & IFF_PROMISC) 694 dev_set_promiscuity(lowerdev, -1); 695 696 dev_uc_del(lowerdev, dev->dev_addr); 697 698 hash_del: 699 macvlan_hash_del(vlan, !dev->dismantle); 700 return 0; 701 } 702 703 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 704 { 705 struct macvlan_dev *vlan = netdev_priv(dev); 706 struct net_device *lowerdev = vlan->lowerdev; 707 struct macvlan_port *port = vlan->port; 708 int err; 709 710 if (!(dev->flags & IFF_UP)) { 711 /* Just copy in the new address */ 712 ether_addr_copy(dev->dev_addr, addr); 713 } else { 714 /* Rehash and update the device filters */ 715 if (macvlan_addr_busy(vlan->port, addr)) 716 return -EBUSY; 717 718 if (!macvlan_passthru(port)) { 719 err = dev_uc_add(lowerdev, addr); 720 if (err) 721 return err; 722 723 dev_uc_del(lowerdev, dev->dev_addr); 724 } 725 726 macvlan_hash_change_addr(vlan, addr); 727 } 728 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 729 /* Since addr_change isn't set, we are here due to lower 730 * device change. Save the lower-dev address so we can 731 * restore it later. 732 */ 733 ether_addr_copy(vlan->port->perm_addr, 734 lowerdev->dev_addr); 735 } 736 macvlan_clear_addr_change(port); 737 return 0; 738 } 739 740 static int macvlan_set_mac_address(struct net_device *dev, void *p) 741 { 742 struct macvlan_dev *vlan = netdev_priv(dev); 743 struct sockaddr *addr = p; 744 745 if (!is_valid_ether_addr(addr->sa_data)) 746 return -EADDRNOTAVAIL; 747 748 /* If the addresses are the same, this is a no-op */ 749 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 750 return 0; 751 752 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 753 macvlan_set_addr_change(vlan->port); 754 return dev_set_mac_address(vlan->lowerdev, addr); 755 } 756 757 return macvlan_sync_address(dev, addr->sa_data); 758 } 759 760 static void macvlan_change_rx_flags(struct net_device *dev, int change) 761 { 762 struct macvlan_dev *vlan = netdev_priv(dev); 763 struct net_device *lowerdev = vlan->lowerdev; 764 765 if (dev->flags & IFF_UP) { 766 if (change & IFF_ALLMULTI) 767 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 768 if (change & IFF_PROMISC) 769 dev_set_promiscuity(lowerdev, 770 dev->flags & IFF_PROMISC ? 1 : -1); 771 772 } 773 } 774 775 static void macvlan_compute_filter(unsigned long *mc_filter, 776 struct net_device *dev, 777 struct macvlan_dev *vlan) 778 { 779 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 780 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 781 } else { 782 struct netdev_hw_addr *ha; 783 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 784 785 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 786 netdev_for_each_mc_addr(ha, dev) { 787 __set_bit(mc_hash(vlan, ha->addr), filter); 788 } 789 790 __set_bit(mc_hash(vlan, dev->broadcast), filter); 791 792 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 793 } 794 } 795 796 static void macvlan_set_mac_lists(struct net_device *dev) 797 { 798 struct macvlan_dev *vlan = netdev_priv(dev); 799 800 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 801 802 dev_uc_sync(vlan->lowerdev, dev); 803 dev_mc_sync(vlan->lowerdev, dev); 804 805 /* This is slightly inaccurate as we're including the subscription 806 * list of vlan->lowerdev too. 807 * 808 * Bug alert: This only works if everyone has the same broadcast 809 * address as lowerdev. As soon as someone changes theirs this 810 * will break. 811 * 812 * However, this is already broken as when you change your broadcast 813 * address we don't get called. 814 * 815 * The solution is to maintain a list of broadcast addresses like 816 * we do for uc/mc, if you care. 817 */ 818 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 819 } 820 821 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 822 { 823 struct macvlan_dev *vlan = netdev_priv(dev); 824 825 if (vlan->lowerdev->mtu < new_mtu) 826 return -EINVAL; 827 dev->mtu = new_mtu; 828 return 0; 829 } 830 831 /* 832 * macvlan network devices have devices nesting below it and are a special 833 * "super class" of normal network devices; split their locks off into a 834 * separate class since they always nest. 835 */ 836 static struct lock_class_key macvlan_netdev_addr_lock_key; 837 838 #define ALWAYS_ON_OFFLOADS \ 839 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 840 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 841 842 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 843 844 #define MACVLAN_FEATURES \ 845 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 846 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 847 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 848 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 849 850 #define MACVLAN_STATE_MASK \ 851 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 852 853 static int macvlan_get_nest_level(struct net_device *dev) 854 { 855 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level; 856 } 857 858 static void macvlan_set_lockdep_class(struct net_device *dev) 859 { 860 netdev_lockdep_set_classes(dev); 861 lockdep_set_class_and_subclass(&dev->addr_list_lock, 862 &macvlan_netdev_addr_lock_key, 863 macvlan_get_nest_level(dev)); 864 } 865 866 static int macvlan_init(struct net_device *dev) 867 { 868 struct macvlan_dev *vlan = netdev_priv(dev); 869 const struct net_device *lowerdev = vlan->lowerdev; 870 struct macvlan_port *port = vlan->port; 871 872 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 873 (lowerdev->state & MACVLAN_STATE_MASK); 874 dev->features = lowerdev->features & MACVLAN_FEATURES; 875 dev->features |= ALWAYS_ON_FEATURES; 876 dev->hw_features |= NETIF_F_LRO; 877 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 878 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 879 dev->hw_enc_features |= dev->features; 880 dev->gso_max_size = lowerdev->gso_max_size; 881 dev->gso_max_segs = lowerdev->gso_max_segs; 882 dev->hard_header_len = lowerdev->hard_header_len; 883 884 macvlan_set_lockdep_class(dev); 885 886 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 887 if (!vlan->pcpu_stats) 888 return -ENOMEM; 889 890 port->count += 1; 891 892 return 0; 893 } 894 895 static void macvlan_uninit(struct net_device *dev) 896 { 897 struct macvlan_dev *vlan = netdev_priv(dev); 898 struct macvlan_port *port = vlan->port; 899 900 free_percpu(vlan->pcpu_stats); 901 902 macvlan_flush_sources(port, vlan); 903 port->count -= 1; 904 if (!port->count) 905 macvlan_port_destroy(port->dev); 906 } 907 908 static void macvlan_dev_get_stats64(struct net_device *dev, 909 struct rtnl_link_stats64 *stats) 910 { 911 struct macvlan_dev *vlan = netdev_priv(dev); 912 913 if (vlan->pcpu_stats) { 914 struct vlan_pcpu_stats *p; 915 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 916 u32 rx_errors = 0, tx_dropped = 0; 917 unsigned int start; 918 int i; 919 920 for_each_possible_cpu(i) { 921 p = per_cpu_ptr(vlan->pcpu_stats, i); 922 do { 923 start = u64_stats_fetch_begin_irq(&p->syncp); 924 rx_packets = p->rx_packets; 925 rx_bytes = p->rx_bytes; 926 rx_multicast = p->rx_multicast; 927 tx_packets = p->tx_packets; 928 tx_bytes = p->tx_bytes; 929 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 930 931 stats->rx_packets += rx_packets; 932 stats->rx_bytes += rx_bytes; 933 stats->multicast += rx_multicast; 934 stats->tx_packets += tx_packets; 935 stats->tx_bytes += tx_bytes; 936 /* rx_errors & tx_dropped are u32, updated 937 * without syncp protection. 938 */ 939 rx_errors += p->rx_errors; 940 tx_dropped += p->tx_dropped; 941 } 942 stats->rx_errors = rx_errors; 943 stats->rx_dropped = rx_errors; 944 stats->tx_dropped = tx_dropped; 945 } 946 } 947 948 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 949 __be16 proto, u16 vid) 950 { 951 struct macvlan_dev *vlan = netdev_priv(dev); 952 struct net_device *lowerdev = vlan->lowerdev; 953 954 return vlan_vid_add(lowerdev, proto, vid); 955 } 956 957 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 958 __be16 proto, u16 vid) 959 { 960 struct macvlan_dev *vlan = netdev_priv(dev); 961 struct net_device *lowerdev = vlan->lowerdev; 962 963 vlan_vid_del(lowerdev, proto, vid); 964 return 0; 965 } 966 967 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 968 struct net_device *dev, 969 const unsigned char *addr, u16 vid, 970 u16 flags) 971 { 972 struct macvlan_dev *vlan = netdev_priv(dev); 973 int err = -EINVAL; 974 975 /* Support unicast filter only on passthru devices. 976 * Multicast filter should be allowed on all devices. 977 */ 978 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 979 return -EOPNOTSUPP; 980 981 if (flags & NLM_F_REPLACE) 982 return -EOPNOTSUPP; 983 984 if (is_unicast_ether_addr(addr)) 985 err = dev_uc_add_excl(dev, addr); 986 else if (is_multicast_ether_addr(addr)) 987 err = dev_mc_add_excl(dev, addr); 988 989 return err; 990 } 991 992 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 993 struct net_device *dev, 994 const unsigned char *addr, u16 vid) 995 { 996 struct macvlan_dev *vlan = netdev_priv(dev); 997 int err = -EINVAL; 998 999 /* Support unicast filter only on passthru devices. 1000 * Multicast filter should be allowed on all devices. 1001 */ 1002 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1003 return -EOPNOTSUPP; 1004 1005 if (is_unicast_ether_addr(addr)) 1006 err = dev_uc_del(dev, addr); 1007 else if (is_multicast_ether_addr(addr)) 1008 err = dev_mc_del(dev, addr); 1009 1010 return err; 1011 } 1012 1013 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1014 struct ethtool_drvinfo *drvinfo) 1015 { 1016 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1017 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1018 } 1019 1020 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1021 struct ethtool_link_ksettings *cmd) 1022 { 1023 const struct macvlan_dev *vlan = netdev_priv(dev); 1024 1025 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1026 } 1027 1028 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1029 netdev_features_t features) 1030 { 1031 struct macvlan_dev *vlan = netdev_priv(dev); 1032 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1033 netdev_features_t mask; 1034 1035 features |= NETIF_F_ALL_FOR_ALL; 1036 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1037 mask = features; 1038 1039 lowerdev_features &= (features | ~NETIF_F_LRO); 1040 features = netdev_increment_features(lowerdev_features, features, mask); 1041 features |= ALWAYS_ON_FEATURES; 1042 features &= ~NETIF_F_NETNS_LOCAL; 1043 1044 return features; 1045 } 1046 1047 #ifdef CONFIG_NET_POLL_CONTROLLER 1048 static void macvlan_dev_poll_controller(struct net_device *dev) 1049 { 1050 return; 1051 } 1052 1053 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1054 { 1055 struct macvlan_dev *vlan = netdev_priv(dev); 1056 struct net_device *real_dev = vlan->lowerdev; 1057 struct netpoll *netpoll; 1058 int err = 0; 1059 1060 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1061 err = -ENOMEM; 1062 if (!netpoll) 1063 goto out; 1064 1065 err = __netpoll_setup(netpoll, real_dev); 1066 if (err) { 1067 kfree(netpoll); 1068 goto out; 1069 } 1070 1071 vlan->netpoll = netpoll; 1072 1073 out: 1074 return err; 1075 } 1076 1077 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1078 { 1079 struct macvlan_dev *vlan = netdev_priv(dev); 1080 struct netpoll *netpoll = vlan->netpoll; 1081 1082 if (!netpoll) 1083 return; 1084 1085 vlan->netpoll = NULL; 1086 1087 __netpoll_free_async(netpoll); 1088 } 1089 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1090 1091 static int macvlan_dev_get_iflink(const struct net_device *dev) 1092 { 1093 struct macvlan_dev *vlan = netdev_priv(dev); 1094 1095 return vlan->lowerdev->ifindex; 1096 } 1097 1098 static const struct ethtool_ops macvlan_ethtool_ops = { 1099 .get_link = ethtool_op_get_link, 1100 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1101 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1102 }; 1103 1104 static const struct net_device_ops macvlan_netdev_ops = { 1105 .ndo_init = macvlan_init, 1106 .ndo_uninit = macvlan_uninit, 1107 .ndo_open = macvlan_open, 1108 .ndo_stop = macvlan_stop, 1109 .ndo_start_xmit = macvlan_start_xmit, 1110 .ndo_change_mtu = macvlan_change_mtu, 1111 .ndo_fix_features = macvlan_fix_features, 1112 .ndo_change_rx_flags = macvlan_change_rx_flags, 1113 .ndo_set_mac_address = macvlan_set_mac_address, 1114 .ndo_set_rx_mode = macvlan_set_mac_lists, 1115 .ndo_get_stats64 = macvlan_dev_get_stats64, 1116 .ndo_validate_addr = eth_validate_addr, 1117 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1118 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1119 .ndo_fdb_add = macvlan_fdb_add, 1120 .ndo_fdb_del = macvlan_fdb_del, 1121 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1122 .ndo_get_lock_subclass = macvlan_get_nest_level, 1123 #ifdef CONFIG_NET_POLL_CONTROLLER 1124 .ndo_poll_controller = macvlan_dev_poll_controller, 1125 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1126 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1127 #endif 1128 .ndo_get_iflink = macvlan_dev_get_iflink, 1129 .ndo_features_check = passthru_features_check, 1130 }; 1131 1132 void macvlan_common_setup(struct net_device *dev) 1133 { 1134 ether_setup(dev); 1135 1136 dev->min_mtu = 0; 1137 dev->max_mtu = ETH_MAX_MTU; 1138 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1139 netif_keep_dst(dev); 1140 dev->priv_flags |= IFF_UNICAST_FLT; 1141 dev->netdev_ops = &macvlan_netdev_ops; 1142 dev->needs_free_netdev = true; 1143 dev->header_ops = &macvlan_hard_header_ops; 1144 dev->ethtool_ops = &macvlan_ethtool_ops; 1145 } 1146 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1147 1148 static void macvlan_setup(struct net_device *dev) 1149 { 1150 macvlan_common_setup(dev); 1151 dev->priv_flags |= IFF_NO_QUEUE; 1152 } 1153 1154 static int macvlan_port_create(struct net_device *dev) 1155 { 1156 struct macvlan_port *port; 1157 unsigned int i; 1158 int err; 1159 1160 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1161 return -EINVAL; 1162 1163 if (netdev_is_rx_handler_busy(dev)) 1164 return -EBUSY; 1165 1166 port = kzalloc(sizeof(*port), GFP_KERNEL); 1167 if (port == NULL) 1168 return -ENOMEM; 1169 1170 port->dev = dev; 1171 ether_addr_copy(port->perm_addr, dev->dev_addr); 1172 INIT_LIST_HEAD(&port->vlans); 1173 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1174 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1175 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1176 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1177 1178 skb_queue_head_init(&port->bc_queue); 1179 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1180 1181 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1182 if (err) 1183 kfree(port); 1184 else 1185 dev->priv_flags |= IFF_MACVLAN_PORT; 1186 return err; 1187 } 1188 1189 static void macvlan_port_destroy(struct net_device *dev) 1190 { 1191 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1192 struct sk_buff *skb; 1193 1194 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1195 netdev_rx_handler_unregister(dev); 1196 1197 /* After this point, no packet can schedule bc_work anymore, 1198 * but we need to cancel it and purge left skbs if any. 1199 */ 1200 cancel_work_sync(&port->bc_work); 1201 1202 while ((skb = __skb_dequeue(&port->bc_queue))) { 1203 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1204 1205 if (src) 1206 dev_put(src->dev); 1207 1208 kfree_skb(skb); 1209 } 1210 1211 /* If the lower device address has been changed by passthru 1212 * macvlan, put it back. 1213 */ 1214 if (macvlan_passthru(port) && 1215 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1216 struct sockaddr sa; 1217 1218 sa.sa_family = port->dev->type; 1219 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1220 dev_set_mac_address(port->dev, &sa); 1221 } 1222 1223 kfree(port); 1224 } 1225 1226 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1227 struct netlink_ext_ack *extack) 1228 { 1229 if (tb[IFLA_ADDRESS]) { 1230 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1231 return -EINVAL; 1232 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1233 return -EADDRNOTAVAIL; 1234 } 1235 1236 if (!data) 1237 return 0; 1238 1239 if (data[IFLA_MACVLAN_FLAGS] && 1240 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1241 return -EINVAL; 1242 1243 if (data[IFLA_MACVLAN_MODE]) { 1244 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1245 case MACVLAN_MODE_PRIVATE: 1246 case MACVLAN_MODE_VEPA: 1247 case MACVLAN_MODE_BRIDGE: 1248 case MACVLAN_MODE_PASSTHRU: 1249 case MACVLAN_MODE_SOURCE: 1250 break; 1251 default: 1252 return -EINVAL; 1253 } 1254 } 1255 1256 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1257 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1258 case MACVLAN_MACADDR_ADD: 1259 case MACVLAN_MACADDR_DEL: 1260 case MACVLAN_MACADDR_FLUSH: 1261 case MACVLAN_MACADDR_SET: 1262 break; 1263 default: 1264 return -EINVAL; 1265 } 1266 } 1267 1268 if (data[IFLA_MACVLAN_MACADDR]) { 1269 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1270 return -EINVAL; 1271 1272 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1273 return -EADDRNOTAVAIL; 1274 } 1275 1276 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1277 return -EINVAL; 1278 1279 return 0; 1280 } 1281 1282 /** 1283 * reconfigure list of remote source mac address 1284 * (only for macvlan devices in source mode) 1285 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1286 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1287 */ 1288 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1289 struct nlattr *data[]) 1290 { 1291 char *addr = NULL; 1292 int ret, rem, len; 1293 struct nlattr *nla, *head; 1294 struct macvlan_source_entry *entry; 1295 1296 if (data[IFLA_MACVLAN_MACADDR]) 1297 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1298 1299 if (mode == MACVLAN_MACADDR_ADD) { 1300 if (!addr) 1301 return -EINVAL; 1302 1303 return macvlan_hash_add_source(vlan, addr); 1304 1305 } else if (mode == MACVLAN_MACADDR_DEL) { 1306 if (!addr) 1307 return -EINVAL; 1308 1309 entry = macvlan_hash_lookup_source(vlan, addr); 1310 if (entry) { 1311 macvlan_hash_del_source(entry); 1312 vlan->macaddr_count--; 1313 } 1314 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1315 macvlan_flush_sources(vlan->port, vlan); 1316 } else if (mode == MACVLAN_MACADDR_SET) { 1317 macvlan_flush_sources(vlan->port, vlan); 1318 1319 if (addr) { 1320 ret = macvlan_hash_add_source(vlan, addr); 1321 if (ret) 1322 return ret; 1323 } 1324 1325 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1326 return 0; 1327 1328 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1329 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1330 1331 nla_for_each_attr(nla, head, len, rem) { 1332 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1333 nla_len(nla) != ETH_ALEN) 1334 continue; 1335 1336 addr = nla_data(nla); 1337 ret = macvlan_hash_add_source(vlan, addr); 1338 if (ret) 1339 return ret; 1340 } 1341 } else { 1342 return -EINVAL; 1343 } 1344 1345 return 0; 1346 } 1347 1348 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1349 struct nlattr *tb[], struct nlattr *data[], 1350 struct netlink_ext_ack *extack) 1351 { 1352 struct macvlan_dev *vlan = netdev_priv(dev); 1353 struct macvlan_port *port; 1354 struct net_device *lowerdev; 1355 int err; 1356 int macmode; 1357 bool create = false; 1358 1359 if (!tb[IFLA_LINK]) 1360 return -EINVAL; 1361 1362 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1363 if (lowerdev == NULL) 1364 return -ENODEV; 1365 1366 /* When creating macvlans or macvtaps on top of other macvlans - use 1367 * the real device as the lowerdev. 1368 */ 1369 if (netif_is_macvlan(lowerdev)) 1370 lowerdev = macvlan_dev_real_dev(lowerdev); 1371 1372 if (!tb[IFLA_MTU]) 1373 dev->mtu = lowerdev->mtu; 1374 else if (dev->mtu > lowerdev->mtu) 1375 return -EINVAL; 1376 1377 /* MTU range: 68 - lowerdev->max_mtu */ 1378 dev->min_mtu = ETH_MIN_MTU; 1379 dev->max_mtu = lowerdev->max_mtu; 1380 1381 if (!tb[IFLA_ADDRESS]) 1382 eth_hw_addr_random(dev); 1383 1384 if (!macvlan_port_exists(lowerdev)) { 1385 err = macvlan_port_create(lowerdev); 1386 if (err < 0) 1387 return err; 1388 create = true; 1389 } 1390 port = macvlan_port_get_rtnl(lowerdev); 1391 1392 /* Only 1 macvlan device can be created in passthru mode */ 1393 if (macvlan_passthru(port)) { 1394 /* The macvlan port must be not created this time, 1395 * still goto destroy_macvlan_port for readability. 1396 */ 1397 err = -EINVAL; 1398 goto destroy_macvlan_port; 1399 } 1400 1401 vlan->lowerdev = lowerdev; 1402 vlan->dev = dev; 1403 vlan->port = port; 1404 vlan->set_features = MACVLAN_FEATURES; 1405 vlan->nest_level = dev_get_nest_level(lowerdev) + 1; 1406 1407 vlan->mode = MACVLAN_MODE_VEPA; 1408 if (data && data[IFLA_MACVLAN_MODE]) 1409 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1410 1411 if (data && data[IFLA_MACVLAN_FLAGS]) 1412 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1413 1414 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1415 if (port->count) { 1416 err = -EINVAL; 1417 goto destroy_macvlan_port; 1418 } 1419 macvlan_set_passthru(port); 1420 eth_hw_addr_inherit(dev, lowerdev); 1421 } 1422 1423 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1424 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1425 err = -EINVAL; 1426 goto destroy_macvlan_port; 1427 } 1428 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1429 err = macvlan_changelink_sources(vlan, macmode, data); 1430 if (err) 1431 goto destroy_macvlan_port; 1432 } 1433 1434 err = register_netdevice(dev); 1435 if (err < 0) 1436 goto destroy_macvlan_port; 1437 1438 dev->priv_flags |= IFF_MACVLAN; 1439 err = netdev_upper_dev_link(lowerdev, dev, extack); 1440 if (err) 1441 goto unregister_netdev; 1442 1443 list_add_tail_rcu(&vlan->list, &port->vlans); 1444 netif_stacked_transfer_operstate(lowerdev, dev); 1445 linkwatch_fire_event(dev); 1446 1447 return 0; 1448 1449 unregister_netdev: 1450 unregister_netdevice(dev); 1451 destroy_macvlan_port: 1452 if (create) 1453 macvlan_port_destroy(port->dev); 1454 return err; 1455 } 1456 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1457 1458 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1459 struct nlattr *tb[], struct nlattr *data[], 1460 struct netlink_ext_ack *extack) 1461 { 1462 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1463 } 1464 1465 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1466 { 1467 struct macvlan_dev *vlan = netdev_priv(dev); 1468 1469 if (vlan->mode == MACVLAN_MODE_SOURCE) 1470 macvlan_flush_sources(vlan->port, vlan); 1471 list_del_rcu(&vlan->list); 1472 unregister_netdevice_queue(dev, head); 1473 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1474 } 1475 EXPORT_SYMBOL_GPL(macvlan_dellink); 1476 1477 static int macvlan_changelink(struct net_device *dev, 1478 struct nlattr *tb[], struct nlattr *data[], 1479 struct netlink_ext_ack *extack) 1480 { 1481 struct macvlan_dev *vlan = netdev_priv(dev); 1482 enum macvlan_mode mode; 1483 bool set_mode = false; 1484 enum macvlan_macaddr_mode macmode; 1485 int ret; 1486 1487 /* Validate mode, but don't set yet: setting flags may fail. */ 1488 if (data && data[IFLA_MACVLAN_MODE]) { 1489 set_mode = true; 1490 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1491 /* Passthrough mode can't be set or cleared dynamically */ 1492 if ((mode == MACVLAN_MODE_PASSTHRU) != 1493 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1494 return -EINVAL; 1495 if (vlan->mode == MACVLAN_MODE_SOURCE && 1496 vlan->mode != mode) 1497 macvlan_flush_sources(vlan->port, vlan); 1498 } 1499 1500 if (data && data[IFLA_MACVLAN_FLAGS]) { 1501 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1502 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1503 if (macvlan_passthru(vlan->port) && promisc) { 1504 int err; 1505 1506 if (flags & MACVLAN_FLAG_NOPROMISC) 1507 err = dev_set_promiscuity(vlan->lowerdev, -1); 1508 else 1509 err = dev_set_promiscuity(vlan->lowerdev, 1); 1510 if (err < 0) 1511 return err; 1512 } 1513 vlan->flags = flags; 1514 } 1515 if (set_mode) 1516 vlan->mode = mode; 1517 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1518 if (vlan->mode != MACVLAN_MODE_SOURCE) 1519 return -EINVAL; 1520 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1521 ret = macvlan_changelink_sources(vlan, macmode, data); 1522 if (ret) 1523 return ret; 1524 } 1525 return 0; 1526 } 1527 1528 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1529 { 1530 if (vlan->macaddr_count == 0) 1531 return 0; 1532 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1533 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1534 } 1535 1536 static size_t macvlan_get_size(const struct net_device *dev) 1537 { 1538 struct macvlan_dev *vlan = netdev_priv(dev); 1539 1540 return (0 1541 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1542 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1543 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1544 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1545 ); 1546 } 1547 1548 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1549 const struct macvlan_dev *vlan, 1550 const int i) 1551 { 1552 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1553 struct macvlan_source_entry *entry; 1554 1555 hlist_for_each_entry_rcu(entry, h, hlist) { 1556 if (entry->vlan != vlan) 1557 continue; 1558 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1559 return 1; 1560 } 1561 return 0; 1562 } 1563 1564 static int macvlan_fill_info(struct sk_buff *skb, 1565 const struct net_device *dev) 1566 { 1567 struct macvlan_dev *vlan = netdev_priv(dev); 1568 int i; 1569 struct nlattr *nest; 1570 1571 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1572 goto nla_put_failure; 1573 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1574 goto nla_put_failure; 1575 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1576 goto nla_put_failure; 1577 if (vlan->macaddr_count > 0) { 1578 nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA); 1579 if (nest == NULL) 1580 goto nla_put_failure; 1581 1582 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1583 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1584 goto nla_put_failure; 1585 } 1586 nla_nest_end(skb, nest); 1587 } 1588 return 0; 1589 1590 nla_put_failure: 1591 return -EMSGSIZE; 1592 } 1593 1594 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1595 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1596 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1597 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1598 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1599 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1600 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1601 }; 1602 1603 int macvlan_link_register(struct rtnl_link_ops *ops) 1604 { 1605 /* common fields */ 1606 ops->validate = macvlan_validate; 1607 ops->maxtype = IFLA_MACVLAN_MAX; 1608 ops->policy = macvlan_policy; 1609 ops->changelink = macvlan_changelink; 1610 ops->get_size = macvlan_get_size; 1611 ops->fill_info = macvlan_fill_info; 1612 1613 return rtnl_link_register(ops); 1614 }; 1615 EXPORT_SYMBOL_GPL(macvlan_link_register); 1616 1617 static struct net *macvlan_get_link_net(const struct net_device *dev) 1618 { 1619 return dev_net(macvlan_dev_real_dev(dev)); 1620 } 1621 1622 static struct rtnl_link_ops macvlan_link_ops = { 1623 .kind = "macvlan", 1624 .setup = macvlan_setup, 1625 .newlink = macvlan_newlink, 1626 .dellink = macvlan_dellink, 1627 .get_link_net = macvlan_get_link_net, 1628 .priv_size = sizeof(struct macvlan_dev), 1629 }; 1630 1631 static int macvlan_device_event(struct notifier_block *unused, 1632 unsigned long event, void *ptr) 1633 { 1634 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1635 struct macvlan_dev *vlan, *next; 1636 struct macvlan_port *port; 1637 LIST_HEAD(list_kill); 1638 1639 if (!macvlan_port_exists(dev)) 1640 return NOTIFY_DONE; 1641 1642 port = macvlan_port_get_rtnl(dev); 1643 1644 switch (event) { 1645 case NETDEV_UP: 1646 case NETDEV_CHANGE: 1647 list_for_each_entry(vlan, &port->vlans, list) 1648 netif_stacked_transfer_operstate(vlan->lowerdev, 1649 vlan->dev); 1650 break; 1651 case NETDEV_FEAT_CHANGE: 1652 list_for_each_entry(vlan, &port->vlans, list) { 1653 vlan->dev->gso_max_size = dev->gso_max_size; 1654 vlan->dev->gso_max_segs = dev->gso_max_segs; 1655 netdev_update_features(vlan->dev); 1656 } 1657 break; 1658 case NETDEV_CHANGEMTU: 1659 list_for_each_entry(vlan, &port->vlans, list) { 1660 if (vlan->dev->mtu <= dev->mtu) 1661 continue; 1662 dev_set_mtu(vlan->dev, dev->mtu); 1663 } 1664 break; 1665 case NETDEV_CHANGEADDR: 1666 if (!macvlan_passthru(port)) 1667 return NOTIFY_DONE; 1668 1669 vlan = list_first_entry_or_null(&port->vlans, 1670 struct macvlan_dev, 1671 list); 1672 1673 if (macvlan_sync_address(vlan->dev, dev->dev_addr)) 1674 return NOTIFY_BAD; 1675 1676 break; 1677 case NETDEV_UNREGISTER: 1678 /* twiddle thumbs on netns device moves */ 1679 if (dev->reg_state != NETREG_UNREGISTERING) 1680 break; 1681 1682 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1683 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1684 unregister_netdevice_many(&list_kill); 1685 break; 1686 case NETDEV_PRE_TYPE_CHANGE: 1687 /* Forbid underlaying device to change its type. */ 1688 return NOTIFY_BAD; 1689 1690 case NETDEV_NOTIFY_PEERS: 1691 case NETDEV_BONDING_FAILOVER: 1692 case NETDEV_RESEND_IGMP: 1693 /* Propagate to all vlans */ 1694 list_for_each_entry(vlan, &port->vlans, list) 1695 call_netdevice_notifiers(event, vlan->dev); 1696 } 1697 return NOTIFY_DONE; 1698 } 1699 1700 static struct notifier_block macvlan_notifier_block __read_mostly = { 1701 .notifier_call = macvlan_device_event, 1702 }; 1703 1704 static int __init macvlan_init_module(void) 1705 { 1706 int err; 1707 1708 register_netdevice_notifier(&macvlan_notifier_block); 1709 1710 err = macvlan_link_register(&macvlan_link_ops); 1711 if (err < 0) 1712 goto err1; 1713 return 0; 1714 err1: 1715 unregister_netdevice_notifier(&macvlan_notifier_block); 1716 return err; 1717 } 1718 1719 static void __exit macvlan_cleanup_module(void) 1720 { 1721 rtnl_link_unregister(&macvlan_link_ops); 1722 unregister_netdevice_notifier(&macvlan_notifier_block); 1723 } 1724 1725 module_init(macvlan_init_module); 1726 module_exit(macvlan_cleanup_module); 1727 1728 MODULE_LICENSE("GPL"); 1729 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1730 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1731 MODULE_ALIAS_RTNL_LINK("macvlan"); 1732