1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net> 4 * 5 * The code this is based on carried the following copyright notice: 6 * --- 7 * (C) Copyright 2001-2006 8 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com 9 * Re-worked by Ben Greear <greearb@candelatech.com> 10 * --- 11 */ 12 #include <linux/kernel.h> 13 #include <linux/types.h> 14 #include <linux/module.h> 15 #include <linux/init.h> 16 #include <linux/errno.h> 17 #include <linux/slab.h> 18 #include <linux/string.h> 19 #include <linux/rculist.h> 20 #include <linux/notifier.h> 21 #include <linux/netdevice.h> 22 #include <linux/etherdevice.h> 23 #include <linux/net_tstamp.h> 24 #include <linux/ethtool.h> 25 #include <linux/if_arp.h> 26 #include <linux/if_vlan.h> 27 #include <linux/if_link.h> 28 #include <linux/if_macvlan.h> 29 #include <linux/hash.h> 30 #include <linux/workqueue.h> 31 #include <net/rtnetlink.h> 32 #include <net/xfrm.h> 33 #include <linux/netpoll.h> 34 #include <linux/phy.h> 35 36 #define MACVLAN_HASH_BITS 8 37 #define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS) 38 #define MACVLAN_DEFAULT_BC_QUEUE_LEN 1000 39 40 #define MACVLAN_F_PASSTHRU 1 41 #define MACVLAN_F_ADDRCHANGE 2 42 43 struct macvlan_port { 44 struct net_device *dev; 45 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 46 struct list_head vlans; 47 struct sk_buff_head bc_queue; 48 struct work_struct bc_work; 49 u32 bc_queue_len_used; 50 int bc_cutoff; 51 u32 flags; 52 int count; 53 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 54 DECLARE_BITMAP(bc_filter, MACVLAN_MC_FILTER_SZ); 55 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 56 unsigned char perm_addr[ETH_ALEN]; 57 }; 58 59 struct macvlan_source_entry { 60 struct hlist_node hlist; 61 struct macvlan_dev *vlan; 62 unsigned char addr[6+2] __aligned(sizeof(u16)); 63 struct rcu_head rcu; 64 }; 65 66 struct macvlan_skb_cb { 67 const struct macvlan_dev *src; 68 }; 69 70 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 71 72 static void macvlan_port_destroy(struct net_device *dev); 73 static void update_port_bc_queue_len(struct macvlan_port *port); 74 75 static inline bool macvlan_passthru(const struct macvlan_port *port) 76 { 77 return port->flags & MACVLAN_F_PASSTHRU; 78 } 79 80 static inline void macvlan_set_passthru(struct macvlan_port *port) 81 { 82 port->flags |= MACVLAN_F_PASSTHRU; 83 } 84 85 static inline bool macvlan_addr_change(const struct macvlan_port *port) 86 { 87 return port->flags & MACVLAN_F_ADDRCHANGE; 88 } 89 90 static inline void macvlan_set_addr_change(struct macvlan_port *port) 91 { 92 port->flags |= MACVLAN_F_ADDRCHANGE; 93 } 94 95 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 96 { 97 port->flags &= ~MACVLAN_F_ADDRCHANGE; 98 } 99 100 /* Hash Ethernet address */ 101 static u32 macvlan_eth_hash(const unsigned char *addr) 102 { 103 u64 value = get_unaligned((u64 *)addr); 104 105 /* only want 6 bytes */ 106 #ifdef __BIG_ENDIAN 107 value >>= 16; 108 #else 109 value <<= 16; 110 #endif 111 return hash_64(value, MACVLAN_HASH_BITS); 112 } 113 114 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 115 { 116 return rcu_dereference(dev->rx_handler_data); 117 } 118 119 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 120 { 121 return rtnl_dereference(dev->rx_handler_data); 122 } 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 lockdep_rtnl_is_held()) { 132 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 133 return vlan; 134 } 135 return NULL; 136 } 137 138 static struct macvlan_source_entry *macvlan_hash_lookup_source( 139 const struct macvlan_dev *vlan, 140 const unsigned char *addr) 141 { 142 struct macvlan_source_entry *entry; 143 u32 idx = macvlan_eth_hash(addr); 144 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 145 146 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) { 147 if (ether_addr_equal_64bits(entry->addr, addr) && 148 entry->vlan == vlan) 149 return entry; 150 } 151 return NULL; 152 } 153 154 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 155 const unsigned char *addr) 156 { 157 struct macvlan_port *port = vlan->port; 158 struct macvlan_source_entry *entry; 159 struct hlist_head *h; 160 161 entry = macvlan_hash_lookup_source(vlan, addr); 162 if (entry) 163 return 0; 164 165 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 166 if (!entry) 167 return -ENOMEM; 168 169 ether_addr_copy(entry->addr, addr); 170 entry->vlan = vlan; 171 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 172 hlist_add_head_rcu(&entry->hlist, h); 173 vlan->macaddr_count++; 174 175 return 0; 176 } 177 178 static void macvlan_hash_add(struct macvlan_dev *vlan) 179 { 180 struct macvlan_port *port = vlan->port; 181 const unsigned char *addr = vlan->dev->dev_addr; 182 u32 idx = macvlan_eth_hash(addr); 183 184 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 185 } 186 187 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 188 { 189 hlist_del_rcu(&entry->hlist); 190 kfree_rcu(entry, rcu); 191 } 192 193 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 194 { 195 hlist_del_rcu(&vlan->hlist); 196 if (sync) 197 synchronize_rcu(); 198 } 199 200 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 201 const unsigned char *addr) 202 { 203 macvlan_hash_del(vlan, true); 204 /* Now that we are unhashed it is safe to change the device 205 * address without confusing packet delivery. 206 */ 207 eth_hw_addr_set(vlan->dev, addr); 208 macvlan_hash_add(vlan); 209 } 210 211 static bool macvlan_addr_busy(const struct macvlan_port *port, 212 const unsigned char *addr) 213 { 214 /* Test to see if the specified address is 215 * currently in use by the underlying device or 216 * another macvlan. 217 */ 218 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 219 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 220 return true; 221 222 if (macvlan_hash_lookup(port, addr)) 223 return true; 224 225 return false; 226 } 227 228 229 static int macvlan_broadcast_one(struct sk_buff *skb, 230 const struct macvlan_dev *vlan, 231 const struct ethhdr *eth, bool local) 232 { 233 struct net_device *dev = vlan->dev; 234 235 if (local) 236 return __dev_forward_skb(dev, skb); 237 238 skb->dev = dev; 239 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 240 skb->pkt_type = PACKET_BROADCAST; 241 else 242 skb->pkt_type = PACKET_MULTICAST; 243 244 return 0; 245 } 246 247 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 248 { 249 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 250 } 251 252 253 static unsigned int mc_hash(const struct macvlan_dev *vlan, 254 const unsigned char *addr) 255 { 256 u32 val = __get_unaligned_cpu32(addr + 2); 257 258 val ^= macvlan_hash_mix(vlan); 259 return hash_32(val, MACVLAN_MC_FILTER_BITS); 260 } 261 262 static void macvlan_broadcast(struct sk_buff *skb, 263 const struct macvlan_port *port, 264 struct net_device *src, 265 enum macvlan_mode mode) 266 { 267 const struct ethhdr *eth = eth_hdr(skb); 268 const struct macvlan_dev *vlan; 269 struct sk_buff *nskb; 270 unsigned int i; 271 int err; 272 unsigned int hash; 273 274 if (skb->protocol == htons(ETH_P_PAUSE)) 275 return; 276 277 hash_for_each_rcu(port->vlan_hash, i, vlan, 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(nskb, vlan, eth, 289 mode == MACVLAN_MODE_BRIDGE) ?: 290 netif_rx(nskb); 291 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 292 err == NET_RX_SUCCESS, true); 293 } 294 } 295 296 static void macvlan_multicast_rx(const struct macvlan_port *port, 297 const struct macvlan_dev *src, 298 struct sk_buff *skb) 299 { 300 if (!src) 301 /* frame comes from an external address */ 302 macvlan_broadcast(skb, port, NULL, 303 MACVLAN_MODE_PRIVATE | 304 MACVLAN_MODE_VEPA | 305 MACVLAN_MODE_PASSTHRU| 306 MACVLAN_MODE_BRIDGE); 307 else if (src->mode == MACVLAN_MODE_VEPA) 308 /* flood to everyone except source */ 309 macvlan_broadcast(skb, port, src->dev, 310 MACVLAN_MODE_VEPA | 311 MACVLAN_MODE_BRIDGE); 312 else 313 /* 314 * flood only to VEPA ports, bridge ports 315 * already saw the frame on the way out. 316 */ 317 macvlan_broadcast(skb, port, src->dev, 318 MACVLAN_MODE_VEPA); 319 } 320 321 static void macvlan_process_broadcast(struct work_struct *w) 322 { 323 struct macvlan_port *port = container_of(w, struct macvlan_port, 324 bc_work); 325 struct sk_buff *skb; 326 struct sk_buff_head list; 327 328 __skb_queue_head_init(&list); 329 330 spin_lock_bh(&port->bc_queue.lock); 331 skb_queue_splice_tail_init(&port->bc_queue, &list); 332 spin_unlock_bh(&port->bc_queue.lock); 333 334 while ((skb = __skb_dequeue(&list))) { 335 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 336 337 rcu_read_lock(); 338 macvlan_multicast_rx(port, src, skb); 339 rcu_read_unlock(); 340 341 if (src) 342 dev_put(src->dev); 343 consume_skb(skb); 344 345 cond_resched(); 346 } 347 } 348 349 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 350 const struct macvlan_dev *src, 351 struct sk_buff *skb) 352 { 353 struct sk_buff *nskb; 354 int err = -ENOMEM; 355 356 nskb = skb_clone(skb, GFP_ATOMIC); 357 if (!nskb) 358 goto err; 359 360 MACVLAN_SKB_CB(nskb)->src = src; 361 362 spin_lock(&port->bc_queue.lock); 363 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) { 364 if (src) 365 dev_hold(src->dev); 366 __skb_queue_tail(&port->bc_queue, nskb); 367 err = 0; 368 } 369 spin_unlock(&port->bc_queue.lock); 370 371 queue_work(system_unbound_wq, &port->bc_work); 372 373 if (err) 374 goto free_nskb; 375 376 return; 377 378 free_nskb: 379 kfree_skb(nskb); 380 err: 381 dev_core_stats_rx_dropped_inc(skb->dev); 382 } 383 384 static void macvlan_flush_sources(struct macvlan_port *port, 385 struct macvlan_dev *vlan) 386 { 387 struct macvlan_source_entry *entry; 388 struct hlist_node *next; 389 int i; 390 391 hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist) 392 if (entry->vlan == vlan) 393 macvlan_hash_del_source(entry); 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 bool 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 bool consume = false; 432 433 hlist_for_each_entry_rcu(entry, h, hlist) { 434 if (ether_addr_equal_64bits(entry->addr, addr)) { 435 if (entry->vlan->flags & MACVLAN_FLAG_NODST) 436 consume = true; 437 macvlan_forward_source_one(skb, entry->vlan); 438 } 439 } 440 441 return consume; 442 } 443 444 /* called under rcu_read_lock() from netif_receive_skb */ 445 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 446 { 447 struct macvlan_port *port; 448 struct sk_buff *skb = *pskb; 449 const struct ethhdr *eth = eth_hdr(skb); 450 const struct macvlan_dev *vlan; 451 const struct macvlan_dev *src; 452 struct net_device *dev; 453 unsigned int len = 0; 454 int ret; 455 rx_handler_result_t handle_res; 456 457 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */ 458 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 459 return RX_HANDLER_PASS; 460 461 port = macvlan_port_get_rcu(skb->dev); 462 if (is_multicast_ether_addr(eth->h_dest)) { 463 unsigned int hash; 464 465 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 466 if (!skb) 467 return RX_HANDLER_CONSUMED; 468 *pskb = skb; 469 eth = eth_hdr(skb); 470 if (macvlan_forward_source(skb, port, eth->h_source)) { 471 kfree_skb(skb); 472 return RX_HANDLER_CONSUMED; 473 } 474 src = macvlan_hash_lookup(port, eth->h_source); 475 if (src && src->mode != MACVLAN_MODE_VEPA && 476 src->mode != MACVLAN_MODE_BRIDGE) { 477 /* forward to original port. */ 478 vlan = src; 479 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 480 __netif_rx(skb); 481 handle_res = RX_HANDLER_CONSUMED; 482 goto out; 483 } 484 485 hash = mc_hash(NULL, eth->h_dest); 486 if (test_bit(hash, port->bc_filter)) 487 macvlan_broadcast_enqueue(port, src, skb); 488 else if (test_bit(hash, port->mc_filter)) 489 macvlan_multicast_rx(port, src, skb); 490 491 return RX_HANDLER_PASS; 492 } 493 494 if (macvlan_forward_source(skb, port, eth->h_source)) { 495 kfree_skb(skb); 496 return RX_HANDLER_CONSUMED; 497 } 498 if (macvlan_passthru(port)) 499 vlan = list_first_or_null_rcu(&port->vlans, 500 struct macvlan_dev, list); 501 else 502 vlan = macvlan_hash_lookup(port, eth->h_dest); 503 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 504 return RX_HANDLER_PASS; 505 506 dev = vlan->dev; 507 if (unlikely(!(dev->flags & IFF_UP))) { 508 kfree_skb(skb); 509 return RX_HANDLER_CONSUMED; 510 } 511 len = skb->len + ETH_HLEN; 512 skb = skb_share_check(skb, GFP_ATOMIC); 513 if (!skb) { 514 ret = NET_RX_DROP; 515 handle_res = RX_HANDLER_CONSUMED; 516 goto out; 517 } 518 519 *pskb = skb; 520 skb->dev = dev; 521 skb->pkt_type = PACKET_HOST; 522 523 ret = NET_RX_SUCCESS; 524 handle_res = RX_HANDLER_ANOTHER; 525 out: 526 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 527 return handle_res; 528 } 529 530 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 531 { 532 const struct macvlan_dev *vlan = netdev_priv(dev); 533 const struct macvlan_port *port = vlan->port; 534 const struct macvlan_dev *dest; 535 536 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 537 const struct ethhdr *eth = skb_eth_hdr(skb); 538 539 /* send to other bridge ports directly */ 540 if (is_multicast_ether_addr(eth->h_dest)) { 541 skb_reset_mac_header(skb); 542 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 543 goto xmit_world; 544 } 545 546 dest = macvlan_hash_lookup(port, eth->h_dest); 547 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 548 /* send to lowerdev first for its network taps */ 549 dev_forward_skb(vlan->lowerdev, skb); 550 551 return NET_XMIT_SUCCESS; 552 } 553 } 554 xmit_world: 555 skb->dev = vlan->lowerdev; 556 return dev_queue_xmit_accel(skb, 557 netdev_get_sb_channel(dev) ? dev : NULL); 558 } 559 560 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 561 { 562 #ifdef CONFIG_NET_POLL_CONTROLLER 563 return netpoll_send_skb(vlan->netpoll, skb); 564 #else 565 BUG(); 566 return NETDEV_TX_OK; 567 #endif 568 } 569 570 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 571 struct net_device *dev) 572 { 573 struct macvlan_dev *vlan = netdev_priv(dev); 574 unsigned int len = skb->len; 575 int ret; 576 577 if (unlikely(netpoll_tx_running(dev))) 578 return macvlan_netpoll_send_skb(vlan, skb); 579 580 ret = macvlan_queue_xmit(skb, dev); 581 582 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 583 struct vlan_pcpu_stats *pcpu_stats; 584 585 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 586 u64_stats_update_begin(&pcpu_stats->syncp); 587 u64_stats_inc(&pcpu_stats->tx_packets); 588 u64_stats_add(&pcpu_stats->tx_bytes, len); 589 u64_stats_update_end(&pcpu_stats->syncp); 590 } else { 591 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 592 } 593 return ret; 594 } 595 596 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 597 unsigned short type, const void *daddr, 598 const void *saddr, unsigned len) 599 { 600 const struct macvlan_dev *vlan = netdev_priv(dev); 601 struct net_device *lowerdev = vlan->lowerdev; 602 603 return dev_hard_header(skb, lowerdev, type, daddr, 604 saddr ? : dev->dev_addr, len); 605 } 606 607 static const struct header_ops macvlan_hard_header_ops = { 608 .create = macvlan_hard_header, 609 .parse = eth_header_parse, 610 .cache = eth_header_cache, 611 .cache_update = eth_header_cache_update, 612 }; 613 614 static int macvlan_open(struct net_device *dev) 615 { 616 struct macvlan_dev *vlan = netdev_priv(dev); 617 struct net_device *lowerdev = vlan->lowerdev; 618 int err; 619 620 if (macvlan_passthru(vlan->port)) { 621 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 622 err = dev_set_promiscuity(lowerdev, 1); 623 if (err < 0) 624 goto out; 625 } 626 goto hash_add; 627 } 628 629 err = -EADDRINUSE; 630 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 631 goto out; 632 633 /* Attempt to populate accel_priv which is used to offload the L2 634 * forwarding requests for unicast packets. 635 */ 636 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 637 vlan->accel_priv = 638 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 639 640 /* If earlier attempt to offload failed, or accel_priv is not 641 * populated we must add the unicast address to the lower device. 642 */ 643 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 644 vlan->accel_priv = NULL; 645 err = dev_uc_add(lowerdev, dev->dev_addr); 646 if (err < 0) 647 goto out; 648 } 649 650 if (dev->flags & IFF_ALLMULTI) { 651 err = dev_set_allmulti(lowerdev, 1); 652 if (err < 0) 653 goto del_unicast; 654 } 655 656 if (dev->flags & IFF_PROMISC) { 657 err = dev_set_promiscuity(lowerdev, 1); 658 if (err < 0) 659 goto clear_multi; 660 } 661 662 hash_add: 663 macvlan_hash_add(vlan); 664 return 0; 665 666 clear_multi: 667 if (dev->flags & IFF_ALLMULTI) 668 dev_set_allmulti(lowerdev, -1); 669 del_unicast: 670 if (vlan->accel_priv) { 671 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 672 vlan->accel_priv); 673 vlan->accel_priv = NULL; 674 } else { 675 dev_uc_del(lowerdev, dev->dev_addr); 676 } 677 out: 678 return err; 679 } 680 681 static int macvlan_stop(struct net_device *dev) 682 { 683 struct macvlan_dev *vlan = netdev_priv(dev); 684 struct net_device *lowerdev = vlan->lowerdev; 685 686 if (vlan->accel_priv) { 687 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 688 vlan->accel_priv); 689 vlan->accel_priv = NULL; 690 } 691 692 dev_uc_unsync(lowerdev, dev); 693 dev_mc_unsync(lowerdev, dev); 694 695 if (macvlan_passthru(vlan->port)) { 696 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 697 dev_set_promiscuity(lowerdev, -1); 698 goto hash_del; 699 } 700 701 if (dev->flags & IFF_ALLMULTI) 702 dev_set_allmulti(lowerdev, -1); 703 704 if (dev->flags & IFF_PROMISC) 705 dev_set_promiscuity(lowerdev, -1); 706 707 dev_uc_del(lowerdev, dev->dev_addr); 708 709 hash_del: 710 macvlan_hash_del(vlan, !dev->dismantle); 711 return 0; 712 } 713 714 static int macvlan_sync_address(struct net_device *dev, 715 const unsigned char *addr) 716 { 717 struct macvlan_dev *vlan = netdev_priv(dev); 718 struct net_device *lowerdev = vlan->lowerdev; 719 struct macvlan_port *port = vlan->port; 720 int err; 721 722 if (!(dev->flags & IFF_UP)) { 723 /* Just copy in the new address */ 724 eth_hw_addr_set(dev, addr); 725 } else { 726 /* Rehash and update the device filters */ 727 if (macvlan_addr_busy(vlan->port, addr)) 728 return -EADDRINUSE; 729 730 if (!macvlan_passthru(port)) { 731 err = dev_uc_add(lowerdev, addr); 732 if (err) 733 return err; 734 735 dev_uc_del(lowerdev, dev->dev_addr); 736 } 737 738 macvlan_hash_change_addr(vlan, addr); 739 } 740 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 741 /* Since addr_change isn't set, we are here due to lower 742 * device change. Save the lower-dev address so we can 743 * restore it later. 744 */ 745 ether_addr_copy(vlan->port->perm_addr, 746 lowerdev->dev_addr); 747 } 748 macvlan_clear_addr_change(port); 749 return 0; 750 } 751 752 static int macvlan_set_mac_address(struct net_device *dev, void *p) 753 { 754 struct macvlan_dev *vlan = netdev_priv(dev); 755 struct sockaddr *addr = p; 756 757 if (!is_valid_ether_addr(addr->sa_data)) 758 return -EADDRNOTAVAIL; 759 760 /* If the addresses are the same, this is a no-op */ 761 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 762 return 0; 763 764 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 765 macvlan_set_addr_change(vlan->port); 766 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 767 } 768 769 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 770 return -EADDRINUSE; 771 772 return macvlan_sync_address(dev, addr->sa_data); 773 } 774 775 static void macvlan_change_rx_flags(struct net_device *dev, int change) 776 { 777 struct macvlan_dev *vlan = netdev_priv(dev); 778 struct net_device *lowerdev = vlan->lowerdev; 779 780 if (dev->flags & IFF_UP) { 781 if (change & IFF_ALLMULTI) 782 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 783 if (!macvlan_passthru(vlan->port) && change & IFF_PROMISC) 784 dev_set_promiscuity(lowerdev, 785 dev->flags & IFF_PROMISC ? 1 : -1); 786 787 } 788 } 789 790 static void macvlan_compute_filter(unsigned long *mc_filter, 791 struct net_device *dev, 792 struct macvlan_dev *vlan, int cutoff) 793 { 794 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 795 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 796 } else { 797 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 798 struct netdev_hw_addr *ha; 799 800 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 801 netdev_for_each_mc_addr(ha, dev) { 802 if (!vlan && ha->synced <= cutoff) 803 continue; 804 805 __set_bit(mc_hash(vlan, ha->addr), filter); 806 } 807 808 __set_bit(mc_hash(vlan, dev->broadcast), filter); 809 810 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 811 } 812 } 813 814 static void macvlan_recompute_bc_filter(struct macvlan_dev *vlan) 815 { 816 if (vlan->port->bc_cutoff < 0) { 817 bitmap_zero(vlan->port->bc_filter, MACVLAN_MC_FILTER_SZ); 818 return; 819 } 820 821 macvlan_compute_filter(vlan->port->bc_filter, vlan->lowerdev, NULL, 822 vlan->port->bc_cutoff); 823 } 824 825 static void macvlan_set_mac_lists(struct net_device *dev) 826 { 827 struct macvlan_dev *vlan = netdev_priv(dev); 828 829 macvlan_compute_filter(vlan->mc_filter, dev, vlan, 0); 830 831 dev_uc_sync(vlan->lowerdev, dev); 832 dev_mc_sync(vlan->lowerdev, dev); 833 834 /* This is slightly inaccurate as we're including the subscription 835 * list of vlan->lowerdev too. 836 * 837 * Bug alert: This only works if everyone has the same broadcast 838 * address as lowerdev. As soon as someone changes theirs this 839 * will break. 840 * 841 * However, this is already broken as when you change your broadcast 842 * address we don't get called. 843 * 844 * The solution is to maintain a list of broadcast addresses like 845 * we do for uc/mc, if you care. 846 */ 847 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL, 848 0); 849 macvlan_recompute_bc_filter(vlan); 850 } 851 852 static void update_port_bc_cutoff(struct macvlan_dev *vlan, int cutoff) 853 { 854 if (vlan->port->bc_cutoff == cutoff) 855 return; 856 857 vlan->port->bc_cutoff = cutoff; 858 macvlan_recompute_bc_filter(vlan); 859 } 860 861 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 862 { 863 struct macvlan_dev *vlan = netdev_priv(dev); 864 865 if (vlan->lowerdev->mtu < new_mtu) 866 return -EINVAL; 867 dev->mtu = new_mtu; 868 return 0; 869 } 870 871 static int macvlan_hwtstamp_get(struct net_device *dev, 872 struct kernel_hwtstamp_config *cfg) 873 { 874 struct net_device *real_dev = macvlan_dev_real_dev(dev); 875 876 return generic_hwtstamp_get_lower(real_dev, cfg); 877 } 878 879 static int macvlan_hwtstamp_set(struct net_device *dev, 880 struct kernel_hwtstamp_config *cfg, 881 struct netlink_ext_ack *extack) 882 { 883 struct net_device *real_dev = macvlan_dev_real_dev(dev); 884 885 if (!net_eq(dev_net(dev), &init_net)) 886 return -EOPNOTSUPP; 887 888 return generic_hwtstamp_set_lower(real_dev, cfg, extack); 889 } 890 891 /* 892 * macvlan network devices have devices nesting below it and are a special 893 * "super class" of normal network devices; split their locks off into a 894 * separate class since they always nest. 895 */ 896 static struct lock_class_key macvlan_netdev_addr_lock_key; 897 898 #define ALWAYS_ON_OFFLOADS \ 899 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 900 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 901 902 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 903 904 #define MACVLAN_FEATURES \ 905 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 906 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 907 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 908 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 909 910 #define MACVLAN_STATE_MASK \ 911 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 912 913 static void macvlan_set_lockdep_class(struct net_device *dev) 914 { 915 netdev_lockdep_set_classes(dev); 916 lockdep_set_class(&dev->addr_list_lock, 917 &macvlan_netdev_addr_lock_key); 918 } 919 920 static int macvlan_init(struct net_device *dev) 921 { 922 struct macvlan_dev *vlan = netdev_priv(dev); 923 struct net_device *lowerdev = vlan->lowerdev; 924 struct macvlan_port *port = vlan->port; 925 926 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 927 (lowerdev->state & MACVLAN_STATE_MASK); 928 dev->features = lowerdev->features & MACVLAN_FEATURES; 929 dev->features |= ALWAYS_ON_FEATURES; 930 dev->hw_features |= NETIF_F_LRO; 931 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 932 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 933 dev->hw_enc_features |= dev->features; 934 netif_inherit_tso_max(dev, lowerdev); 935 dev->hard_header_len = lowerdev->hard_header_len; 936 macvlan_set_lockdep_class(dev); 937 938 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 939 if (!vlan->pcpu_stats) 940 return -ENOMEM; 941 942 port->count += 1; 943 944 /* Get macvlan's reference to lowerdev */ 945 netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL); 946 947 return 0; 948 } 949 950 static void macvlan_uninit(struct net_device *dev) 951 { 952 struct macvlan_dev *vlan = netdev_priv(dev); 953 struct macvlan_port *port = vlan->port; 954 955 free_percpu(vlan->pcpu_stats); 956 957 macvlan_flush_sources(port, vlan); 958 port->count -= 1; 959 if (!port->count) 960 macvlan_port_destroy(port->dev); 961 } 962 963 static void macvlan_dev_get_stats64(struct net_device *dev, 964 struct rtnl_link_stats64 *stats) 965 { 966 struct macvlan_dev *vlan = netdev_priv(dev); 967 968 if (vlan->pcpu_stats) { 969 struct vlan_pcpu_stats *p; 970 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 971 u32 rx_errors = 0, tx_dropped = 0; 972 unsigned int start; 973 int i; 974 975 for_each_possible_cpu(i) { 976 p = per_cpu_ptr(vlan->pcpu_stats, i); 977 do { 978 start = u64_stats_fetch_begin(&p->syncp); 979 rx_packets = u64_stats_read(&p->rx_packets); 980 rx_bytes = u64_stats_read(&p->rx_bytes); 981 rx_multicast = u64_stats_read(&p->rx_multicast); 982 tx_packets = u64_stats_read(&p->tx_packets); 983 tx_bytes = u64_stats_read(&p->tx_bytes); 984 } while (u64_stats_fetch_retry(&p->syncp, start)); 985 986 stats->rx_packets += rx_packets; 987 stats->rx_bytes += rx_bytes; 988 stats->multicast += rx_multicast; 989 stats->tx_packets += tx_packets; 990 stats->tx_bytes += tx_bytes; 991 /* rx_errors & tx_dropped are u32, updated 992 * without syncp protection. 993 */ 994 rx_errors += READ_ONCE(p->rx_errors); 995 tx_dropped += READ_ONCE(p->tx_dropped); 996 } 997 stats->rx_errors = rx_errors; 998 stats->rx_dropped = rx_errors; 999 stats->tx_dropped = tx_dropped; 1000 } 1001 } 1002 1003 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 1004 __be16 proto, u16 vid) 1005 { 1006 struct macvlan_dev *vlan = netdev_priv(dev); 1007 struct net_device *lowerdev = vlan->lowerdev; 1008 1009 return vlan_vid_add(lowerdev, proto, vid); 1010 } 1011 1012 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 1013 __be16 proto, u16 vid) 1014 { 1015 struct macvlan_dev *vlan = netdev_priv(dev); 1016 struct net_device *lowerdev = vlan->lowerdev; 1017 1018 vlan_vid_del(lowerdev, proto, vid); 1019 return 0; 1020 } 1021 1022 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1023 struct net_device *dev, 1024 const unsigned char *addr, u16 vid, 1025 u16 flags, 1026 struct netlink_ext_ack *extack) 1027 { 1028 struct macvlan_dev *vlan = netdev_priv(dev); 1029 int err = -EINVAL; 1030 1031 /* Support unicast filter only on passthru devices. 1032 * Multicast filter should be allowed on all devices. 1033 */ 1034 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1035 return -EOPNOTSUPP; 1036 1037 if (flags & NLM_F_REPLACE) 1038 return -EOPNOTSUPP; 1039 1040 if (is_unicast_ether_addr(addr)) 1041 err = dev_uc_add_excl(dev, addr); 1042 else if (is_multicast_ether_addr(addr)) 1043 err = dev_mc_add_excl(dev, addr); 1044 1045 return err; 1046 } 1047 1048 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1049 struct net_device *dev, 1050 const unsigned char *addr, u16 vid, 1051 struct netlink_ext_ack *extack) 1052 { 1053 struct macvlan_dev *vlan = netdev_priv(dev); 1054 int err = -EINVAL; 1055 1056 /* Support unicast filter only on passthru devices. 1057 * Multicast filter should be allowed on all devices. 1058 */ 1059 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1060 return -EOPNOTSUPP; 1061 1062 if (is_unicast_ether_addr(addr)) 1063 err = dev_uc_del(dev, addr); 1064 else if (is_multicast_ether_addr(addr)) 1065 err = dev_mc_del(dev, addr); 1066 1067 return err; 1068 } 1069 1070 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1071 struct ethtool_drvinfo *drvinfo) 1072 { 1073 strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1074 strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1075 } 1076 1077 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1078 struct ethtool_link_ksettings *cmd) 1079 { 1080 const struct macvlan_dev *vlan = netdev_priv(dev); 1081 1082 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1083 } 1084 1085 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1086 struct ethtool_ts_info *info) 1087 { 1088 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1089 1090 return ethtool_get_ts_info_by_layer(real_dev, info); 1091 } 1092 1093 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1094 netdev_features_t features) 1095 { 1096 struct macvlan_dev *vlan = netdev_priv(dev); 1097 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1098 netdev_features_t mask; 1099 1100 features |= NETIF_F_ALL_FOR_ALL; 1101 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1102 mask = features; 1103 1104 lowerdev_features &= (features | ~NETIF_F_LRO); 1105 features = netdev_increment_features(lowerdev_features, features, mask); 1106 features |= ALWAYS_ON_FEATURES; 1107 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1108 1109 return features; 1110 } 1111 1112 #ifdef CONFIG_NET_POLL_CONTROLLER 1113 static void macvlan_dev_poll_controller(struct net_device *dev) 1114 { 1115 return; 1116 } 1117 1118 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1119 { 1120 struct macvlan_dev *vlan = netdev_priv(dev); 1121 struct net_device *real_dev = vlan->lowerdev; 1122 struct netpoll *netpoll; 1123 int err; 1124 1125 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1126 err = -ENOMEM; 1127 if (!netpoll) 1128 goto out; 1129 1130 err = __netpoll_setup(netpoll, real_dev); 1131 if (err) { 1132 kfree(netpoll); 1133 goto out; 1134 } 1135 1136 vlan->netpoll = netpoll; 1137 1138 out: 1139 return err; 1140 } 1141 1142 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1143 { 1144 struct macvlan_dev *vlan = netdev_priv(dev); 1145 struct netpoll *netpoll = vlan->netpoll; 1146 1147 if (!netpoll) 1148 return; 1149 1150 vlan->netpoll = NULL; 1151 1152 __netpoll_free(netpoll); 1153 } 1154 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1155 1156 static int macvlan_dev_get_iflink(const struct net_device *dev) 1157 { 1158 struct macvlan_dev *vlan = netdev_priv(dev); 1159 1160 return vlan->lowerdev->ifindex; 1161 } 1162 1163 static const struct ethtool_ops macvlan_ethtool_ops = { 1164 .get_link = ethtool_op_get_link, 1165 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1166 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1167 .get_ts_info = macvlan_ethtool_get_ts_info, 1168 }; 1169 1170 static const struct net_device_ops macvlan_netdev_ops = { 1171 .ndo_init = macvlan_init, 1172 .ndo_uninit = macvlan_uninit, 1173 .ndo_open = macvlan_open, 1174 .ndo_stop = macvlan_stop, 1175 .ndo_start_xmit = macvlan_start_xmit, 1176 .ndo_change_mtu = macvlan_change_mtu, 1177 .ndo_fix_features = macvlan_fix_features, 1178 .ndo_change_rx_flags = macvlan_change_rx_flags, 1179 .ndo_set_mac_address = macvlan_set_mac_address, 1180 .ndo_set_rx_mode = macvlan_set_mac_lists, 1181 .ndo_get_stats64 = macvlan_dev_get_stats64, 1182 .ndo_validate_addr = eth_validate_addr, 1183 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1184 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1185 .ndo_fdb_add = macvlan_fdb_add, 1186 .ndo_fdb_del = macvlan_fdb_del, 1187 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1188 #ifdef CONFIG_NET_POLL_CONTROLLER 1189 .ndo_poll_controller = macvlan_dev_poll_controller, 1190 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1191 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1192 #endif 1193 .ndo_get_iflink = macvlan_dev_get_iflink, 1194 .ndo_features_check = passthru_features_check, 1195 .ndo_hwtstamp_get = macvlan_hwtstamp_get, 1196 .ndo_hwtstamp_set = macvlan_hwtstamp_set, 1197 }; 1198 1199 static void macvlan_dev_free(struct net_device *dev) 1200 { 1201 struct macvlan_dev *vlan = netdev_priv(dev); 1202 1203 /* Get rid of the macvlan's reference to lowerdev */ 1204 netdev_put(vlan->lowerdev, &vlan->dev_tracker); 1205 } 1206 1207 void macvlan_common_setup(struct net_device *dev) 1208 { 1209 ether_setup(dev); 1210 1211 /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */ 1212 dev->max_mtu = ETH_MAX_MTU; 1213 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1214 netif_keep_dst(dev); 1215 dev->priv_flags |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN; 1216 dev->netdev_ops = &macvlan_netdev_ops; 1217 dev->needs_free_netdev = true; 1218 dev->priv_destructor = macvlan_dev_free; 1219 dev->header_ops = &macvlan_hard_header_ops; 1220 dev->ethtool_ops = &macvlan_ethtool_ops; 1221 } 1222 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1223 1224 static void macvlan_setup(struct net_device *dev) 1225 { 1226 macvlan_common_setup(dev); 1227 dev->priv_flags |= IFF_NO_QUEUE; 1228 } 1229 1230 static int macvlan_port_create(struct net_device *dev) 1231 { 1232 struct macvlan_port *port; 1233 unsigned int i; 1234 int err; 1235 1236 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1237 return -EINVAL; 1238 1239 if (netdev_is_rx_handler_busy(dev)) 1240 return -EBUSY; 1241 1242 port = kzalloc(sizeof(*port), GFP_KERNEL); 1243 if (port == NULL) 1244 return -ENOMEM; 1245 1246 port->dev = dev; 1247 ether_addr_copy(port->perm_addr, dev->dev_addr); 1248 INIT_LIST_HEAD(&port->vlans); 1249 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1250 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1251 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1252 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1253 1254 port->bc_queue_len_used = 0; 1255 port->bc_cutoff = 1; 1256 skb_queue_head_init(&port->bc_queue); 1257 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1258 1259 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1260 if (err) 1261 kfree(port); 1262 else 1263 dev->priv_flags |= IFF_MACVLAN_PORT; 1264 return err; 1265 } 1266 1267 static void macvlan_port_destroy(struct net_device *dev) 1268 { 1269 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1270 struct sk_buff *skb; 1271 1272 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1273 netdev_rx_handler_unregister(dev); 1274 1275 /* After this point, no packet can schedule bc_work anymore, 1276 * but we need to cancel it and purge left skbs if any. 1277 */ 1278 cancel_work_sync(&port->bc_work); 1279 1280 while ((skb = __skb_dequeue(&port->bc_queue))) { 1281 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1282 1283 if (src) 1284 dev_put(src->dev); 1285 1286 kfree_skb(skb); 1287 } 1288 1289 /* If the lower device address has been changed by passthru 1290 * macvlan, put it back. 1291 */ 1292 if (macvlan_passthru(port) && 1293 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1294 struct sockaddr sa; 1295 1296 sa.sa_family = port->dev->type; 1297 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1298 dev_set_mac_address(port->dev, &sa, NULL); 1299 } 1300 1301 kfree(port); 1302 } 1303 1304 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1305 struct netlink_ext_ack *extack) 1306 { 1307 struct nlattr *nla, *head; 1308 int rem, len; 1309 1310 if (tb[IFLA_ADDRESS]) { 1311 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1312 return -EINVAL; 1313 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1314 return -EADDRNOTAVAIL; 1315 } 1316 1317 if (!data) 1318 return 0; 1319 1320 if (data[IFLA_MACVLAN_FLAGS] && 1321 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC | 1322 MACVLAN_FLAG_NODST)) 1323 return -EINVAL; 1324 1325 if (data[IFLA_MACVLAN_MODE]) { 1326 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1327 case MACVLAN_MODE_PRIVATE: 1328 case MACVLAN_MODE_VEPA: 1329 case MACVLAN_MODE_BRIDGE: 1330 case MACVLAN_MODE_PASSTHRU: 1331 case MACVLAN_MODE_SOURCE: 1332 break; 1333 default: 1334 return -EINVAL; 1335 } 1336 } 1337 1338 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1339 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1340 case MACVLAN_MACADDR_ADD: 1341 case MACVLAN_MACADDR_DEL: 1342 case MACVLAN_MACADDR_FLUSH: 1343 case MACVLAN_MACADDR_SET: 1344 break; 1345 default: 1346 return -EINVAL; 1347 } 1348 } 1349 1350 if (data[IFLA_MACVLAN_MACADDR]) { 1351 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1352 return -EINVAL; 1353 1354 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1355 return -EADDRNOTAVAIL; 1356 } 1357 1358 if (data[IFLA_MACVLAN_MACADDR_DATA]) { 1359 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1360 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1361 1362 nla_for_each_attr(nla, head, len, rem) { 1363 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1364 nla_len(nla) != ETH_ALEN) 1365 return -EINVAL; 1366 1367 if (!is_valid_ether_addr(nla_data(nla))) 1368 return -EADDRNOTAVAIL; 1369 } 1370 } 1371 1372 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1373 return -EINVAL; 1374 1375 return 0; 1376 } 1377 1378 /* 1379 * reconfigure list of remote source mac address 1380 * (only for macvlan devices in source mode) 1381 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1382 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1383 */ 1384 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1385 struct nlattr *data[]) 1386 { 1387 char *addr = NULL; 1388 int ret, rem, len; 1389 struct nlattr *nla, *head; 1390 struct macvlan_source_entry *entry; 1391 1392 if (data[IFLA_MACVLAN_MACADDR]) 1393 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1394 1395 if (mode == MACVLAN_MACADDR_ADD) { 1396 if (!addr) 1397 return -EINVAL; 1398 1399 return macvlan_hash_add_source(vlan, addr); 1400 1401 } else if (mode == MACVLAN_MACADDR_DEL) { 1402 if (!addr) 1403 return -EINVAL; 1404 1405 entry = macvlan_hash_lookup_source(vlan, addr); 1406 if (entry) { 1407 macvlan_hash_del_source(entry); 1408 vlan->macaddr_count--; 1409 } 1410 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1411 macvlan_flush_sources(vlan->port, vlan); 1412 } else if (mode == MACVLAN_MACADDR_SET) { 1413 macvlan_flush_sources(vlan->port, vlan); 1414 1415 if (addr) { 1416 ret = macvlan_hash_add_source(vlan, addr); 1417 if (ret) 1418 return ret; 1419 } 1420 1421 if (!data[IFLA_MACVLAN_MACADDR_DATA]) 1422 return 0; 1423 1424 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1425 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1426 1427 nla_for_each_attr(nla, head, len, rem) { 1428 addr = nla_data(nla); 1429 ret = macvlan_hash_add_source(vlan, addr); 1430 if (ret) 1431 return ret; 1432 } 1433 } else { 1434 return -EINVAL; 1435 } 1436 1437 return 0; 1438 } 1439 1440 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1441 struct nlattr *tb[], struct nlattr *data[], 1442 struct netlink_ext_ack *extack) 1443 { 1444 struct macvlan_dev *vlan = netdev_priv(dev); 1445 struct macvlan_port *port; 1446 struct net_device *lowerdev; 1447 int err; 1448 int macmode; 1449 bool create = false; 1450 1451 if (!tb[IFLA_LINK]) 1452 return -EINVAL; 1453 1454 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1455 if (lowerdev == NULL) 1456 return -ENODEV; 1457 1458 /* When creating macvlans or macvtaps on top of other macvlans - use 1459 * the real device as the lowerdev. 1460 */ 1461 if (netif_is_macvlan(lowerdev)) 1462 lowerdev = macvlan_dev_real_dev(lowerdev); 1463 1464 if (!tb[IFLA_MTU]) 1465 dev->mtu = lowerdev->mtu; 1466 else if (dev->mtu > lowerdev->mtu) 1467 return -EINVAL; 1468 1469 /* MTU range: 68 - lowerdev->max_mtu */ 1470 dev->min_mtu = ETH_MIN_MTU; 1471 dev->max_mtu = lowerdev->max_mtu; 1472 1473 if (!tb[IFLA_ADDRESS]) 1474 eth_hw_addr_random(dev); 1475 1476 if (!netif_is_macvlan_port(lowerdev)) { 1477 err = macvlan_port_create(lowerdev); 1478 if (err < 0) 1479 return err; 1480 create = true; 1481 } 1482 port = macvlan_port_get_rtnl(lowerdev); 1483 1484 /* Only 1 macvlan device can be created in passthru mode */ 1485 if (macvlan_passthru(port)) { 1486 /* The macvlan port must be not created this time, 1487 * still goto destroy_macvlan_port for readability. 1488 */ 1489 err = -EINVAL; 1490 goto destroy_macvlan_port; 1491 } 1492 1493 vlan->lowerdev = lowerdev; 1494 vlan->dev = dev; 1495 vlan->port = port; 1496 vlan->set_features = MACVLAN_FEATURES; 1497 1498 vlan->mode = MACVLAN_MODE_VEPA; 1499 if (data && data[IFLA_MACVLAN_MODE]) 1500 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1501 1502 if (data && data[IFLA_MACVLAN_FLAGS]) 1503 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1504 1505 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1506 if (port->count) { 1507 err = -EINVAL; 1508 goto destroy_macvlan_port; 1509 } 1510 macvlan_set_passthru(port); 1511 eth_hw_addr_inherit(dev, lowerdev); 1512 } 1513 1514 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1515 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1516 err = -EINVAL; 1517 goto destroy_macvlan_port; 1518 } 1519 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1520 err = macvlan_changelink_sources(vlan, macmode, data); 1521 if (err) 1522 goto destroy_macvlan_port; 1523 } 1524 1525 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN; 1526 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) 1527 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1528 1529 if (data && data[IFLA_MACVLAN_BC_CUTOFF]) 1530 update_port_bc_cutoff( 1531 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF])); 1532 1533 err = register_netdevice(dev); 1534 if (err < 0) 1535 goto destroy_macvlan_port; 1536 1537 dev->priv_flags |= IFF_MACVLAN; 1538 err = netdev_upper_dev_link(lowerdev, dev, extack); 1539 if (err) 1540 goto unregister_netdev; 1541 1542 list_add_tail_rcu(&vlan->list, &port->vlans); 1543 update_port_bc_queue_len(vlan->port); 1544 netif_stacked_transfer_operstate(lowerdev, dev); 1545 linkwatch_fire_event(dev); 1546 1547 return 0; 1548 1549 unregister_netdev: 1550 /* macvlan_uninit would free the macvlan port */ 1551 unregister_netdevice(dev); 1552 return err; 1553 destroy_macvlan_port: 1554 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1555 * so we destroy the macvlan port only when it's valid. 1556 */ 1557 if (create && macvlan_port_get_rtnl(lowerdev)) { 1558 macvlan_flush_sources(port, vlan); 1559 macvlan_port_destroy(port->dev); 1560 } 1561 return err; 1562 } 1563 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1564 1565 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1566 struct nlattr *tb[], struct nlattr *data[], 1567 struct netlink_ext_ack *extack) 1568 { 1569 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1570 } 1571 1572 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1573 { 1574 struct macvlan_dev *vlan = netdev_priv(dev); 1575 1576 if (vlan->mode == MACVLAN_MODE_SOURCE) 1577 macvlan_flush_sources(vlan->port, vlan); 1578 list_del_rcu(&vlan->list); 1579 update_port_bc_queue_len(vlan->port); 1580 unregister_netdevice_queue(dev, head); 1581 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1582 } 1583 EXPORT_SYMBOL_GPL(macvlan_dellink); 1584 1585 static int macvlan_changelink(struct net_device *dev, 1586 struct nlattr *tb[], struct nlattr *data[], 1587 struct netlink_ext_ack *extack) 1588 { 1589 struct macvlan_dev *vlan = netdev_priv(dev); 1590 enum macvlan_mode mode; 1591 bool set_mode = false; 1592 enum macvlan_macaddr_mode macmode; 1593 int ret; 1594 1595 /* Validate mode, but don't set yet: setting flags may fail. */ 1596 if (data && data[IFLA_MACVLAN_MODE]) { 1597 set_mode = true; 1598 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1599 /* Passthrough mode can't be set or cleared dynamically */ 1600 if ((mode == MACVLAN_MODE_PASSTHRU) != 1601 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1602 return -EINVAL; 1603 if (vlan->mode == MACVLAN_MODE_SOURCE && 1604 vlan->mode != mode) 1605 macvlan_flush_sources(vlan->port, vlan); 1606 } 1607 1608 if (data && data[IFLA_MACVLAN_FLAGS]) { 1609 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1610 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1611 if (macvlan_passthru(vlan->port) && promisc) { 1612 int err; 1613 1614 if (flags & MACVLAN_FLAG_NOPROMISC) 1615 err = dev_set_promiscuity(vlan->lowerdev, -1); 1616 else 1617 err = dev_set_promiscuity(vlan->lowerdev, 1); 1618 if (err < 0) 1619 return err; 1620 } 1621 vlan->flags = flags; 1622 } 1623 1624 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) { 1625 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1626 update_port_bc_queue_len(vlan->port); 1627 } 1628 1629 if (data && data[IFLA_MACVLAN_BC_CUTOFF]) 1630 update_port_bc_cutoff( 1631 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF])); 1632 1633 if (set_mode) 1634 vlan->mode = mode; 1635 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1636 if (vlan->mode != MACVLAN_MODE_SOURCE) 1637 return -EINVAL; 1638 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1639 ret = macvlan_changelink_sources(vlan, macmode, data); 1640 if (ret) 1641 return ret; 1642 } 1643 return 0; 1644 } 1645 1646 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1647 { 1648 if (vlan->macaddr_count == 0) 1649 return 0; 1650 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1651 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1652 } 1653 1654 static size_t macvlan_get_size(const struct net_device *dev) 1655 { 1656 struct macvlan_dev *vlan = netdev_priv(dev); 1657 1658 return (0 1659 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1660 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1661 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1662 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1663 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */ 1664 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */ 1665 ); 1666 } 1667 1668 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1669 const struct macvlan_dev *vlan, 1670 const int i) 1671 { 1672 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1673 struct macvlan_source_entry *entry; 1674 1675 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) { 1676 if (entry->vlan != vlan) 1677 continue; 1678 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1679 return 1; 1680 } 1681 return 0; 1682 } 1683 1684 static int macvlan_fill_info(struct sk_buff *skb, 1685 const struct net_device *dev) 1686 { 1687 struct macvlan_dev *vlan = netdev_priv(dev); 1688 struct macvlan_port *port = vlan->port; 1689 int i; 1690 struct nlattr *nest; 1691 1692 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1693 goto nla_put_failure; 1694 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1695 goto nla_put_failure; 1696 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1697 goto nla_put_failure; 1698 if (vlan->macaddr_count > 0) { 1699 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1700 if (nest == NULL) 1701 goto nla_put_failure; 1702 1703 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1704 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1705 goto nla_put_failure; 1706 } 1707 nla_nest_end(skb, nest); 1708 } 1709 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req)) 1710 goto nla_put_failure; 1711 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used)) 1712 goto nla_put_failure; 1713 if (port->bc_cutoff != 1 && 1714 nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff)) 1715 goto nla_put_failure; 1716 return 0; 1717 1718 nla_put_failure: 1719 return -EMSGSIZE; 1720 } 1721 1722 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1723 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1724 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1725 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1726 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1727 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1728 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1729 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 }, 1730 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT }, 1731 [IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 }, 1732 }; 1733 1734 int macvlan_link_register(struct rtnl_link_ops *ops) 1735 { 1736 /* common fields */ 1737 ops->validate = macvlan_validate; 1738 ops->maxtype = IFLA_MACVLAN_MAX; 1739 ops->policy = macvlan_policy; 1740 ops->changelink = macvlan_changelink; 1741 ops->get_size = macvlan_get_size; 1742 ops->fill_info = macvlan_fill_info; 1743 1744 return rtnl_link_register(ops); 1745 }; 1746 EXPORT_SYMBOL_GPL(macvlan_link_register); 1747 1748 static struct net *macvlan_get_link_net(const struct net_device *dev) 1749 { 1750 return dev_net(macvlan_dev_real_dev(dev)); 1751 } 1752 1753 static struct rtnl_link_ops macvlan_link_ops = { 1754 .kind = "macvlan", 1755 .setup = macvlan_setup, 1756 .newlink = macvlan_newlink, 1757 .dellink = macvlan_dellink, 1758 .get_link_net = macvlan_get_link_net, 1759 .priv_size = sizeof(struct macvlan_dev), 1760 }; 1761 1762 static void update_port_bc_queue_len(struct macvlan_port *port) 1763 { 1764 u32 max_bc_queue_len_req = 0; 1765 struct macvlan_dev *vlan; 1766 1767 list_for_each_entry(vlan, &port->vlans, list) { 1768 if (vlan->bc_queue_len_req > max_bc_queue_len_req) 1769 max_bc_queue_len_req = vlan->bc_queue_len_req; 1770 } 1771 port->bc_queue_len_used = max_bc_queue_len_req; 1772 } 1773 1774 static int macvlan_device_event(struct notifier_block *unused, 1775 unsigned long event, void *ptr) 1776 { 1777 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1778 struct macvlan_dev *vlan, *next; 1779 struct macvlan_port *port; 1780 LIST_HEAD(list_kill); 1781 1782 if (!netif_is_macvlan_port(dev)) 1783 return NOTIFY_DONE; 1784 1785 port = macvlan_port_get_rtnl(dev); 1786 1787 switch (event) { 1788 case NETDEV_UP: 1789 case NETDEV_DOWN: 1790 case NETDEV_CHANGE: 1791 list_for_each_entry(vlan, &port->vlans, list) 1792 netif_stacked_transfer_operstate(vlan->lowerdev, 1793 vlan->dev); 1794 break; 1795 case NETDEV_FEAT_CHANGE: 1796 list_for_each_entry(vlan, &port->vlans, list) { 1797 netif_inherit_tso_max(vlan->dev, dev); 1798 netdev_update_features(vlan->dev); 1799 } 1800 break; 1801 case NETDEV_CHANGEMTU: 1802 list_for_each_entry(vlan, &port->vlans, list) { 1803 if (vlan->dev->mtu <= dev->mtu) 1804 continue; 1805 dev_set_mtu(vlan->dev, dev->mtu); 1806 } 1807 break; 1808 case NETDEV_CHANGEADDR: 1809 if (!macvlan_passthru(port)) 1810 return NOTIFY_DONE; 1811 1812 vlan = list_first_entry_or_null(&port->vlans, 1813 struct macvlan_dev, 1814 list); 1815 1816 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1817 return NOTIFY_BAD; 1818 1819 break; 1820 case NETDEV_UNREGISTER: 1821 /* twiddle thumbs on netns device moves */ 1822 if (dev->reg_state != NETREG_UNREGISTERING) 1823 break; 1824 1825 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1826 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1827 unregister_netdevice_many(&list_kill); 1828 break; 1829 case NETDEV_PRE_TYPE_CHANGE: 1830 /* Forbid underlying device to change its type. */ 1831 return NOTIFY_BAD; 1832 1833 case NETDEV_NOTIFY_PEERS: 1834 case NETDEV_BONDING_FAILOVER: 1835 case NETDEV_RESEND_IGMP: 1836 /* Propagate to all vlans */ 1837 list_for_each_entry(vlan, &port->vlans, list) 1838 call_netdevice_notifiers(event, vlan->dev); 1839 } 1840 return NOTIFY_DONE; 1841 } 1842 1843 static struct notifier_block macvlan_notifier_block __read_mostly = { 1844 .notifier_call = macvlan_device_event, 1845 }; 1846 1847 static int __init macvlan_init_module(void) 1848 { 1849 int err; 1850 1851 register_netdevice_notifier(&macvlan_notifier_block); 1852 1853 err = macvlan_link_register(&macvlan_link_ops); 1854 if (err < 0) 1855 goto err1; 1856 return 0; 1857 err1: 1858 unregister_netdevice_notifier(&macvlan_notifier_block); 1859 return err; 1860 } 1861 1862 static void __exit macvlan_cleanup_module(void) 1863 { 1864 rtnl_link_unregister(&macvlan_link_ops); 1865 unregister_netdevice_notifier(&macvlan_notifier_block); 1866 } 1867 1868 module_init(macvlan_init_module); 1869 module_exit(macvlan_cleanup_module); 1870 1871 MODULE_LICENSE("GPL"); 1872 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1873 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1874 MODULE_ALIAS_RTNL_LINK("macvlan"); 1875