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/netdev_lock.h> 32 #include <net/rtnetlink.h> 33 #include <net/xfrm.h> 34 #include <linux/netpoll.h> 35 #include <linux/phy.h> 36 37 #define MACVLAN_HASH_BITS 8 38 #define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS) 39 #define MACVLAN_DEFAULT_BC_QUEUE_LEN 1000 40 41 #define MACVLAN_F_PASSTHRU 1 42 #define MACVLAN_F_ADDRCHANGE 2 43 44 struct macvlan_port { 45 struct net_device *dev; 46 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 47 struct list_head vlans; 48 struct sk_buff_head bc_queue; 49 struct work_struct bc_work; 50 u32 bc_queue_len_used; 51 int bc_cutoff; 52 u32 flags; 53 int count; 54 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 55 DECLARE_BITMAP(bc_filter, MACVLAN_MC_FILTER_SZ); 56 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 57 unsigned char perm_addr[ETH_ALEN]; 58 }; 59 60 struct macvlan_source_entry { 61 struct hlist_node hlist; 62 struct macvlan_dev __rcu *vlan; 63 unsigned char addr[6+2] __aligned(sizeof(u16)); 64 struct rcu_head rcu; 65 }; 66 67 struct macvlan_skb_cb { 68 const struct macvlan_dev *src; 69 }; 70 71 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 72 73 static void macvlan_port_destroy(struct net_device *dev); 74 static void update_port_bc_queue_len(struct macvlan_port *port); 75 76 static inline bool macvlan_passthru(const struct macvlan_port *port) 77 { 78 return port->flags & MACVLAN_F_PASSTHRU; 79 } 80 81 static inline void macvlan_set_passthru(struct macvlan_port *port) 82 { 83 port->flags |= MACVLAN_F_PASSTHRU; 84 } 85 86 static inline bool macvlan_addr_change(const struct macvlan_port *port) 87 { 88 return port->flags & MACVLAN_F_ADDRCHANGE; 89 } 90 91 static inline void macvlan_set_addr_change(struct macvlan_port *port) 92 { 93 port->flags |= MACVLAN_F_ADDRCHANGE; 94 } 95 96 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 97 { 98 port->flags &= ~MACVLAN_F_ADDRCHANGE; 99 } 100 101 /* Hash Ethernet address */ 102 static u32 macvlan_eth_hash(const unsigned char *addr) 103 { 104 u64 value = get_unaligned((u64 *)addr); 105 106 /* only want 6 bytes */ 107 #ifdef __BIG_ENDIAN 108 value >>= 16; 109 #else 110 value <<= 16; 111 #endif 112 return hash_64(value, MACVLAN_HASH_BITS); 113 } 114 115 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 116 { 117 return rcu_dereference(dev->rx_handler_data); 118 } 119 120 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 121 { 122 return rtnl_dereference(dev->rx_handler_data); 123 } 124 125 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 126 const unsigned char *addr) 127 { 128 struct macvlan_dev *vlan; 129 u32 idx = macvlan_eth_hash(addr); 130 131 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist, 132 lockdep_rtnl_is_held()) { 133 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 134 return vlan; 135 } 136 return NULL; 137 } 138 139 static struct macvlan_source_entry *macvlan_hash_lookup_source( 140 const struct macvlan_dev *vlan, 141 const unsigned char *addr) 142 { 143 struct macvlan_source_entry *entry; 144 u32 idx = macvlan_eth_hash(addr); 145 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 146 147 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) { 148 if (ether_addr_equal_64bits(entry->addr, addr) && 149 rcu_access_pointer(entry->vlan) == vlan) 150 return entry; 151 } 152 return NULL; 153 } 154 155 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 156 const unsigned char *addr) 157 { 158 struct macvlan_port *port = vlan->port; 159 struct macvlan_source_entry *entry; 160 struct hlist_head *h; 161 162 entry = macvlan_hash_lookup_source(vlan, addr); 163 if (entry) 164 return 0; 165 166 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 167 if (!entry) 168 return -ENOMEM; 169 170 ether_addr_copy(entry->addr, addr); 171 RCU_INIT_POINTER(entry->vlan, vlan); 172 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 173 hlist_add_head_rcu(&entry->hlist, h); 174 vlan->macaddr_count++; 175 176 return 0; 177 } 178 179 static void macvlan_hash_add(struct macvlan_dev *vlan) 180 { 181 struct macvlan_port *port = vlan->port; 182 const unsigned char *addr = vlan->dev->dev_addr; 183 u32 idx = macvlan_eth_hash(addr); 184 185 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 186 } 187 188 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 189 { 190 RCU_INIT_POINTER(entry->vlan, NULL); 191 hlist_del_rcu(&entry->hlist); 192 kfree_rcu(entry, rcu); 193 } 194 195 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 196 { 197 hlist_del_rcu(&vlan->hlist); 198 if (sync) 199 synchronize_rcu(); 200 } 201 202 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 203 const unsigned char *addr) 204 { 205 macvlan_hash_del(vlan, true); 206 /* Now that we are unhashed it is safe to change the device 207 * address without confusing packet delivery. 208 */ 209 eth_hw_addr_set(vlan->dev, addr); 210 macvlan_hash_add(vlan); 211 } 212 213 static bool macvlan_addr_busy(const struct macvlan_port *port, 214 const unsigned char *addr) 215 { 216 /* Test to see if the specified address is 217 * currently in use by the underlying device or 218 * another macvlan. 219 */ 220 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 221 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 222 return true; 223 224 if (macvlan_hash_lookup(port, addr)) 225 return true; 226 227 return false; 228 } 229 230 231 static int macvlan_broadcast_one(struct sk_buff *skb, 232 const struct macvlan_dev *vlan, 233 const struct ethhdr *eth, bool local) 234 { 235 struct net_device *dev = vlan->dev; 236 237 if (local) 238 return __dev_forward_skb(dev, skb); 239 240 skb->dev = dev; 241 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 242 skb->pkt_type = PACKET_BROADCAST; 243 else 244 skb->pkt_type = PACKET_MULTICAST; 245 246 return 0; 247 } 248 249 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 250 { 251 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 252 } 253 254 255 static unsigned int mc_hash(const struct macvlan_dev *vlan, 256 const unsigned char *addr) 257 { 258 u32 val = get_unaligned((u32 *)(addr + 2)); 259 260 val ^= macvlan_hash_mix(vlan); 261 return hash_32(val, MACVLAN_MC_FILTER_BITS); 262 } 263 264 static void macvlan_broadcast(struct sk_buff *skb, 265 const struct macvlan_port *port, 266 struct net_device *src, 267 enum macvlan_mode mode) 268 { 269 const struct ethhdr *eth = eth_hdr(skb); 270 const struct macvlan_dev *vlan; 271 struct sk_buff *nskb; 272 unsigned int i; 273 int err; 274 unsigned int hash; 275 276 if (skb->protocol == htons(ETH_P_PAUSE)) 277 return; 278 279 hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) { 280 if (vlan->dev == src || !(vlan->mode & mode)) 281 continue; 282 283 hash = mc_hash(vlan, eth->h_dest); 284 if (!test_bit(hash, vlan->mc_filter)) 285 continue; 286 287 err = NET_RX_DROP; 288 nskb = skb_clone(skb, GFP_ATOMIC); 289 if (likely(nskb)) 290 err = macvlan_broadcast_one(nskb, vlan, eth, 291 mode == MACVLAN_MODE_BRIDGE) ?: 292 netif_rx(nskb); 293 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 294 err == NET_RX_SUCCESS, true); 295 } 296 } 297 298 static void macvlan_multicast_rx(const struct macvlan_port *port, 299 const struct macvlan_dev *src, 300 struct sk_buff *skb) 301 { 302 if (!src) 303 /* frame comes from an external address */ 304 macvlan_broadcast(skb, port, NULL, 305 MACVLAN_MODE_PRIVATE | 306 MACVLAN_MODE_VEPA | 307 MACVLAN_MODE_PASSTHRU| 308 MACVLAN_MODE_BRIDGE); 309 else if (src->mode == MACVLAN_MODE_VEPA) 310 /* flood to everyone except source */ 311 macvlan_broadcast(skb, port, src->dev, 312 MACVLAN_MODE_VEPA | 313 MACVLAN_MODE_BRIDGE); 314 else 315 /* 316 * flood only to VEPA ports, bridge ports 317 * already saw the frame on the way out. 318 */ 319 macvlan_broadcast(skb, port, src->dev, 320 MACVLAN_MODE_VEPA); 321 } 322 323 static void macvlan_process_broadcast(struct work_struct *w) 324 { 325 struct macvlan_port *port = container_of(w, struct macvlan_port, 326 bc_work); 327 struct sk_buff *skb; 328 struct sk_buff_head list; 329 330 __skb_queue_head_init(&list); 331 332 spin_lock_bh(&port->bc_queue.lock); 333 skb_queue_splice_tail_init(&port->bc_queue, &list); 334 spin_unlock_bh(&port->bc_queue.lock); 335 336 while ((skb = __skb_dequeue(&list))) { 337 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 338 339 rcu_read_lock(); 340 macvlan_multicast_rx(port, src, skb); 341 rcu_read_unlock(); 342 343 if (src) 344 dev_put(src->dev); 345 consume_skb(skb); 346 347 cond_resched(); 348 } 349 } 350 351 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 352 const struct macvlan_dev *src, 353 struct sk_buff *skb) 354 { 355 struct sk_buff *nskb; 356 int err = -ENOMEM; 357 358 nskb = skb_clone(skb, GFP_ATOMIC); 359 if (!nskb) 360 goto err; 361 362 MACVLAN_SKB_CB(nskb)->src = src; 363 364 spin_lock(&port->bc_queue.lock); 365 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) { 366 if (src) 367 dev_hold(src->dev); 368 __skb_queue_tail(&port->bc_queue, nskb); 369 err = 0; 370 } 371 spin_unlock(&port->bc_queue.lock); 372 373 queue_work(system_dfl_wq, &port->bc_work); 374 375 if (err) 376 goto free_nskb; 377 378 return; 379 380 free_nskb: 381 kfree_skb(nskb); 382 err: 383 dev_core_stats_rx_dropped_inc(skb->dev); 384 } 385 386 static void macvlan_flush_sources(struct macvlan_port *port, 387 struct macvlan_dev *vlan) 388 { 389 struct macvlan_source_entry *entry; 390 struct hlist_node *next; 391 int i; 392 393 hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist) 394 if (rcu_access_pointer(entry->vlan) == vlan) 395 macvlan_hash_del_source(entry); 396 397 vlan->macaddr_count = 0; 398 } 399 400 static void macvlan_forward_source_one(struct sk_buff *skb, 401 struct macvlan_dev *vlan) 402 { 403 struct sk_buff *nskb; 404 struct net_device *dev; 405 int len; 406 int ret; 407 408 dev = vlan->dev; 409 if (unlikely(!(dev->flags & IFF_UP))) 410 return; 411 412 nskb = skb_clone(skb, GFP_ATOMIC); 413 if (!nskb) 414 return; 415 416 len = nskb->len + ETH_HLEN; 417 nskb->dev = dev; 418 419 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr)) 420 nskb->pkt_type = PACKET_HOST; 421 422 ret = __netif_rx(nskb); 423 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 424 } 425 426 static bool macvlan_forward_source(struct sk_buff *skb, 427 struct macvlan_port *port, 428 const unsigned char *addr) 429 { 430 struct macvlan_source_entry *entry; 431 u32 idx = macvlan_eth_hash(addr); 432 struct hlist_head *h = &port->vlan_source_hash[idx]; 433 bool consume = false; 434 435 hlist_for_each_entry_rcu(entry, h, hlist) { 436 if (ether_addr_equal_64bits(entry->addr, addr)) { 437 struct macvlan_dev *vlan = rcu_dereference(entry->vlan); 438 439 if (!vlan) 440 continue; 441 442 if (vlan->flags & MACVLAN_FLAG_NODST) 443 consume = true; 444 macvlan_forward_source_one(skb, vlan); 445 } 446 } 447 448 return consume; 449 } 450 451 /* called under rcu_read_lock() from netif_receive_skb */ 452 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 453 { 454 struct macvlan_port *port; 455 struct sk_buff *skb = *pskb; 456 const struct ethhdr *eth = eth_hdr(skb); 457 const struct macvlan_dev *vlan; 458 const struct macvlan_dev *src; 459 struct net_device *dev; 460 unsigned int len = 0; 461 int ret; 462 rx_handler_result_t handle_res; 463 464 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */ 465 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 466 return RX_HANDLER_PASS; 467 468 port = macvlan_port_get_rcu(skb->dev); 469 if (is_multicast_ether_addr(eth->h_dest)) { 470 unsigned int hash; 471 472 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 473 if (!skb) 474 return RX_HANDLER_CONSUMED; 475 *pskb = skb; 476 eth = eth_hdr(skb); 477 if (macvlan_forward_source(skb, port, eth->h_source)) { 478 kfree_skb(skb); 479 return RX_HANDLER_CONSUMED; 480 } 481 src = macvlan_hash_lookup(port, eth->h_source); 482 if (src && src->mode != MACVLAN_MODE_VEPA && 483 src->mode != MACVLAN_MODE_BRIDGE) { 484 /* forward to original port. */ 485 vlan = src; 486 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 487 __netif_rx(skb); 488 handle_res = RX_HANDLER_CONSUMED; 489 goto out; 490 } 491 492 hash = mc_hash(NULL, eth->h_dest); 493 if (test_bit(hash, port->bc_filter)) 494 macvlan_broadcast_enqueue(port, src, skb); 495 else if (test_bit(hash, port->mc_filter)) 496 macvlan_multicast_rx(port, src, skb); 497 498 return RX_HANDLER_PASS; 499 } 500 501 if (macvlan_forward_source(skb, port, eth->h_source)) { 502 kfree_skb(skb); 503 return RX_HANDLER_CONSUMED; 504 } 505 if (macvlan_passthru(port)) 506 vlan = list_first_or_null_rcu(&port->vlans, 507 struct macvlan_dev, list); 508 else 509 vlan = macvlan_hash_lookup(port, eth->h_dest); 510 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 511 return RX_HANDLER_PASS; 512 513 dev = vlan->dev; 514 if (unlikely(!(dev->flags & IFF_UP))) { 515 kfree_skb(skb); 516 return RX_HANDLER_CONSUMED; 517 } 518 len = skb->len + ETH_HLEN; 519 skb = skb_share_check(skb, GFP_ATOMIC); 520 if (!skb) { 521 ret = NET_RX_DROP; 522 handle_res = RX_HANDLER_CONSUMED; 523 goto out; 524 } 525 526 *pskb = skb; 527 skb->dev = dev; 528 skb->pkt_type = PACKET_HOST; 529 530 ret = NET_RX_SUCCESS; 531 handle_res = RX_HANDLER_ANOTHER; 532 out: 533 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 534 return handle_res; 535 } 536 537 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 538 { 539 const struct macvlan_dev *vlan = netdev_priv(dev); 540 const struct macvlan_port *port = vlan->port; 541 const struct macvlan_dev *dest; 542 543 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 544 const struct ethhdr *eth = skb_eth_hdr(skb); 545 546 /* send to other bridge ports directly */ 547 if (is_multicast_ether_addr(eth->h_dest)) { 548 skb_reset_mac_header(skb); 549 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 550 goto xmit_world; 551 } 552 553 dest = macvlan_hash_lookup(port, eth->h_dest); 554 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 555 /* send to lowerdev first for its network taps */ 556 dev_forward_skb(vlan->lowerdev, skb); 557 558 return NET_XMIT_SUCCESS; 559 } 560 } 561 xmit_world: 562 skb->dev = vlan->lowerdev; 563 return dev_queue_xmit_accel(skb, 564 netdev_get_sb_channel(dev) ? dev : NULL); 565 } 566 567 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 568 { 569 #ifdef CONFIG_NET_POLL_CONTROLLER 570 return netpoll_send_skb(vlan->netpoll, skb); 571 #else 572 BUG(); 573 return NETDEV_TX_OK; 574 #endif 575 } 576 577 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 578 struct net_device *dev) 579 { 580 struct macvlan_dev *vlan = netdev_priv(dev); 581 unsigned int len = skb->len; 582 int ret; 583 584 if (unlikely(netpoll_tx_running(dev))) 585 return macvlan_netpoll_send_skb(vlan, skb); 586 587 ret = macvlan_queue_xmit(skb, dev); 588 589 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 590 struct vlan_pcpu_stats *pcpu_stats; 591 592 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 593 u64_stats_update_begin(&pcpu_stats->syncp); 594 u64_stats_inc(&pcpu_stats->tx_packets); 595 u64_stats_add(&pcpu_stats->tx_bytes, len); 596 u64_stats_update_end(&pcpu_stats->syncp); 597 } else { 598 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 599 } 600 return ret; 601 } 602 603 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 604 unsigned short type, const void *daddr, 605 const void *saddr, unsigned len) 606 { 607 const struct macvlan_dev *vlan = netdev_priv(dev); 608 struct net_device *lowerdev = vlan->lowerdev; 609 610 return dev_hard_header(skb, lowerdev, type, daddr, 611 saddr ? : dev->dev_addr, len); 612 } 613 614 static const struct header_ops macvlan_hard_header_ops = { 615 .create = macvlan_hard_header, 616 .parse = eth_header_parse, 617 .cache = eth_header_cache, 618 .cache_update = eth_header_cache_update, 619 .parse_protocol = eth_header_parse_protocol, 620 }; 621 622 static int macvlan_open(struct net_device *dev) 623 { 624 struct macvlan_dev *vlan = netdev_priv(dev); 625 struct net_device *lowerdev = vlan->lowerdev; 626 int err; 627 628 if (macvlan_passthru(vlan->port)) { 629 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 630 err = dev_set_promiscuity(lowerdev, 1); 631 if (err < 0) 632 goto out; 633 } 634 goto hash_add; 635 } 636 637 err = -EADDRINUSE; 638 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 639 goto out; 640 641 /* Attempt to populate accel_priv which is used to offload the L2 642 * forwarding requests for unicast packets. 643 */ 644 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 645 vlan->accel_priv = 646 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 647 648 /* If earlier attempt to offload failed, or accel_priv is not 649 * populated we must add the unicast address to the lower device. 650 */ 651 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 652 vlan->accel_priv = NULL; 653 err = dev_uc_add(lowerdev, dev->dev_addr); 654 if (err < 0) 655 goto out; 656 } 657 658 if (dev->flags & IFF_ALLMULTI) { 659 err = dev_set_allmulti(lowerdev, 1); 660 if (err < 0) 661 goto del_unicast; 662 } 663 664 if (dev->flags & IFF_PROMISC) { 665 err = dev_set_promiscuity(lowerdev, 1); 666 if (err < 0) 667 goto clear_multi; 668 } 669 670 hash_add: 671 macvlan_hash_add(vlan); 672 return 0; 673 674 clear_multi: 675 if (dev->flags & IFF_ALLMULTI) 676 dev_set_allmulti(lowerdev, -1); 677 del_unicast: 678 if (vlan->accel_priv) { 679 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 680 vlan->accel_priv); 681 vlan->accel_priv = NULL; 682 } else { 683 dev_uc_del(lowerdev, dev->dev_addr); 684 } 685 out: 686 return err; 687 } 688 689 static int macvlan_stop(struct net_device *dev) 690 { 691 struct macvlan_dev *vlan = netdev_priv(dev); 692 struct net_device *lowerdev = vlan->lowerdev; 693 694 if (vlan->accel_priv) { 695 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 696 vlan->accel_priv); 697 vlan->accel_priv = NULL; 698 } 699 700 dev_uc_unsync(lowerdev, dev); 701 dev_mc_unsync(lowerdev, dev); 702 703 if (macvlan_passthru(vlan->port)) { 704 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 705 dev_set_promiscuity(lowerdev, -1); 706 goto hash_del; 707 } 708 709 if (dev->flags & IFF_ALLMULTI) 710 dev_set_allmulti(lowerdev, -1); 711 712 if (dev->flags & IFF_PROMISC) 713 dev_set_promiscuity(lowerdev, -1); 714 715 dev_uc_del(lowerdev, dev->dev_addr); 716 717 hash_del: 718 macvlan_hash_del(vlan, !dev->dismantle); 719 return 0; 720 } 721 722 static int macvlan_sync_address(struct net_device *dev, 723 const unsigned char *addr) 724 { 725 struct macvlan_dev *vlan = netdev_priv(dev); 726 struct net_device *lowerdev = vlan->lowerdev; 727 struct macvlan_port *port = vlan->port; 728 int err; 729 730 if (!(dev->flags & IFF_UP)) { 731 /* Just copy in the new address */ 732 eth_hw_addr_set(dev, addr); 733 } else { 734 /* Rehash and update the device filters */ 735 if (macvlan_addr_busy(vlan->port, addr)) 736 return -EADDRINUSE; 737 738 if (!macvlan_passthru(port)) { 739 err = dev_uc_add(lowerdev, addr); 740 if (err) 741 return err; 742 743 dev_uc_del(lowerdev, dev->dev_addr); 744 } 745 746 macvlan_hash_change_addr(vlan, addr); 747 } 748 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 749 /* Since addr_change isn't set, we are here due to lower 750 * device change. Save the lower-dev address so we can 751 * restore it later. 752 */ 753 ether_addr_copy(vlan->port->perm_addr, 754 lowerdev->dev_addr); 755 } 756 macvlan_clear_addr_change(port); 757 return 0; 758 } 759 760 static int macvlan_set_mac_address(struct net_device *dev, void *p) 761 { 762 struct macvlan_dev *vlan = netdev_priv(dev); 763 struct sockaddr_storage *addr = p; 764 765 if (!is_valid_ether_addr(addr->__data)) 766 return -EADDRNOTAVAIL; 767 768 /* If the addresses are the same, this is a no-op */ 769 if (ether_addr_equal(dev->dev_addr, addr->__data)) 770 return 0; 771 772 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 773 macvlan_set_addr_change(vlan->port); 774 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 775 } 776 777 if (macvlan_addr_busy(vlan->port, addr->__data)) 778 return -EADDRINUSE; 779 780 return macvlan_sync_address(dev, addr->__data); 781 } 782 783 static void macvlan_change_rx_flags(struct net_device *dev, int change) 784 { 785 struct macvlan_dev *vlan = netdev_priv(dev); 786 struct net_device *lowerdev = vlan->lowerdev; 787 788 if (dev->flags & IFF_UP) { 789 if (change & IFF_ALLMULTI) 790 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 791 if (!macvlan_passthru(vlan->port) && change & IFF_PROMISC) 792 dev_set_promiscuity(lowerdev, 793 dev->flags & IFF_PROMISC ? 1 : -1); 794 795 } 796 } 797 798 static void macvlan_compute_filter(unsigned long *mc_filter, 799 struct net_device *dev, 800 struct macvlan_dev *vlan, int cutoff) 801 { 802 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 803 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 804 } else { 805 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 806 struct netdev_hw_addr *ha; 807 808 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 809 netdev_for_each_mc_addr(ha, dev) { 810 if (!vlan && ha->synced <= cutoff) 811 continue; 812 813 __set_bit(mc_hash(vlan, ha->addr), filter); 814 } 815 816 __set_bit(mc_hash(vlan, dev->broadcast), filter); 817 818 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 819 } 820 } 821 822 static void macvlan_recompute_bc_filter(struct macvlan_dev *vlan) 823 { 824 if (vlan->port->bc_cutoff < 0) { 825 bitmap_zero(vlan->port->bc_filter, MACVLAN_MC_FILTER_SZ); 826 return; 827 } 828 829 macvlan_compute_filter(vlan->port->bc_filter, vlan->lowerdev, NULL, 830 vlan->port->bc_cutoff); 831 } 832 833 static void macvlan_set_mac_lists(struct net_device *dev) 834 { 835 struct macvlan_dev *vlan = netdev_priv(dev); 836 837 macvlan_compute_filter(vlan->mc_filter, dev, vlan, 0); 838 839 dev_uc_sync(vlan->lowerdev, dev); 840 dev_mc_sync(vlan->lowerdev, dev); 841 842 /* This is slightly inaccurate as we're including the subscription 843 * list of vlan->lowerdev too. 844 * 845 * Bug alert: This only works if everyone has the same broadcast 846 * address as lowerdev. As soon as someone changes theirs this 847 * will break. 848 * 849 * However, this is already broken as when you change your broadcast 850 * address we don't get called. 851 * 852 * The solution is to maintain a list of broadcast addresses like 853 * we do for uc/mc, if you care. 854 */ 855 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL, 856 0); 857 macvlan_recompute_bc_filter(vlan); 858 } 859 860 static void update_port_bc_cutoff(struct macvlan_dev *vlan, int cutoff) 861 { 862 if (vlan->port->bc_cutoff == cutoff) 863 return; 864 865 vlan->port->bc_cutoff = cutoff; 866 macvlan_recompute_bc_filter(vlan); 867 } 868 869 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 870 { 871 struct macvlan_dev *vlan = netdev_priv(dev); 872 873 if (vlan->lowerdev->mtu < new_mtu) 874 return -EINVAL; 875 WRITE_ONCE(dev->mtu, new_mtu); 876 return 0; 877 } 878 879 static int macvlan_hwtstamp_get(struct net_device *dev, 880 struct kernel_hwtstamp_config *cfg) 881 { 882 struct net_device *real_dev = macvlan_dev_real_dev(dev); 883 884 return generic_hwtstamp_get_lower(real_dev, cfg); 885 } 886 887 static int macvlan_hwtstamp_set(struct net_device *dev, 888 struct kernel_hwtstamp_config *cfg, 889 struct netlink_ext_ack *extack) 890 { 891 struct net_device *real_dev = macvlan_dev_real_dev(dev); 892 893 if (!net_eq(dev_net(dev), &init_net)) 894 return -EOPNOTSUPP; 895 896 return generic_hwtstamp_set_lower(real_dev, cfg, extack); 897 } 898 899 /* 900 * macvlan network devices have devices nesting below it and are a special 901 * "super class" of normal network devices; split their locks off into a 902 * separate class since they always nest. 903 */ 904 static struct lock_class_key macvlan_netdev_addr_lock_key; 905 906 #define ALWAYS_ON_OFFLOADS \ 907 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 908 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 909 910 #define ALWAYS_ON_FEATURES ALWAYS_ON_OFFLOADS 911 912 #define MACVLAN_FEATURES \ 913 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 914 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 915 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 916 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 917 918 #define MACVLAN_STATE_MASK \ 919 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 920 921 static void macvlan_set_lockdep_class(struct net_device *dev) 922 { 923 netdev_lockdep_set_classes(dev); 924 lockdep_set_class(&dev->addr_list_lock, 925 &macvlan_netdev_addr_lock_key); 926 } 927 928 static int macvlan_init(struct net_device *dev) 929 { 930 struct macvlan_dev *vlan = netdev_priv(dev); 931 struct net_device *lowerdev = vlan->lowerdev; 932 struct macvlan_port *port = vlan->port; 933 934 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 935 (lowerdev->state & MACVLAN_STATE_MASK); 936 dev->features = lowerdev->features & MACVLAN_FEATURES; 937 dev->features |= ALWAYS_ON_FEATURES; 938 dev->hw_features |= NETIF_F_LRO; 939 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 940 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 941 dev->hw_enc_features |= dev->features; 942 dev->lltx = true; 943 netif_inherit_tso_max(dev, lowerdev); 944 dev->hard_header_len = lowerdev->hard_header_len; 945 macvlan_set_lockdep_class(dev); 946 947 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 948 if (!vlan->pcpu_stats) 949 return -ENOMEM; 950 951 port->count += 1; 952 953 /* Get macvlan's reference to lowerdev */ 954 netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL); 955 956 return 0; 957 } 958 959 static void macvlan_uninit(struct net_device *dev) 960 { 961 struct macvlan_dev *vlan = netdev_priv(dev); 962 struct macvlan_port *port = vlan->port; 963 964 free_percpu(vlan->pcpu_stats); 965 966 macvlan_flush_sources(port, vlan); 967 port->count -= 1; 968 if (!port->count) 969 macvlan_port_destroy(port->dev); 970 } 971 972 static void macvlan_dev_get_stats64(struct net_device *dev, 973 struct rtnl_link_stats64 *stats) 974 { 975 struct macvlan_dev *vlan = netdev_priv(dev); 976 977 if (vlan->pcpu_stats) { 978 struct vlan_pcpu_stats *p; 979 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 980 u32 rx_errors = 0, tx_dropped = 0; 981 unsigned int start; 982 int i; 983 984 for_each_possible_cpu(i) { 985 p = per_cpu_ptr(vlan->pcpu_stats, i); 986 do { 987 start = u64_stats_fetch_begin(&p->syncp); 988 rx_packets = u64_stats_read(&p->rx_packets); 989 rx_bytes = u64_stats_read(&p->rx_bytes); 990 rx_multicast = u64_stats_read(&p->rx_multicast); 991 tx_packets = u64_stats_read(&p->tx_packets); 992 tx_bytes = u64_stats_read(&p->tx_bytes); 993 } while (u64_stats_fetch_retry(&p->syncp, start)); 994 995 stats->rx_packets += rx_packets; 996 stats->rx_bytes += rx_bytes; 997 stats->multicast += rx_multicast; 998 stats->tx_packets += tx_packets; 999 stats->tx_bytes += tx_bytes; 1000 /* rx_errors & tx_dropped are u32, updated 1001 * without syncp protection. 1002 */ 1003 rx_errors += READ_ONCE(p->rx_errors); 1004 tx_dropped += READ_ONCE(p->tx_dropped); 1005 } 1006 stats->rx_errors = rx_errors; 1007 stats->rx_dropped = rx_errors; 1008 stats->tx_dropped = tx_dropped; 1009 } 1010 } 1011 1012 static int macvlan_vlan_rx_add_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 return vlan_vid_add(lowerdev, proto, vid); 1019 } 1020 1021 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 1022 __be16 proto, u16 vid) 1023 { 1024 struct macvlan_dev *vlan = netdev_priv(dev); 1025 struct net_device *lowerdev = vlan->lowerdev; 1026 1027 vlan_vid_del(lowerdev, proto, vid); 1028 return 0; 1029 } 1030 1031 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 1032 struct net_device *dev, 1033 const unsigned char *addr, u16 vid, 1034 u16 flags, bool *notified, 1035 struct netlink_ext_ack *extack) 1036 { 1037 struct macvlan_dev *vlan = netdev_priv(dev); 1038 int err = -EINVAL; 1039 1040 /* Support unicast filter only on passthru devices. 1041 * Multicast filter should be allowed on all devices. 1042 */ 1043 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1044 return -EOPNOTSUPP; 1045 1046 if (flags & NLM_F_REPLACE) 1047 return -EOPNOTSUPP; 1048 1049 if (is_unicast_ether_addr(addr)) 1050 err = dev_uc_add_excl(dev, addr); 1051 else if (is_multicast_ether_addr(addr)) 1052 err = dev_mc_add_excl(dev, addr); 1053 1054 return err; 1055 } 1056 1057 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1058 struct net_device *dev, 1059 const unsigned char *addr, u16 vid, bool *notified, 1060 struct netlink_ext_ack *extack) 1061 { 1062 struct macvlan_dev *vlan = netdev_priv(dev); 1063 int err = -EINVAL; 1064 1065 /* Support unicast filter only on passthru devices. 1066 * Multicast filter should be allowed on all devices. 1067 */ 1068 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1069 return -EOPNOTSUPP; 1070 1071 if (is_unicast_ether_addr(addr)) 1072 err = dev_uc_del(dev, addr); 1073 else if (is_multicast_ether_addr(addr)) 1074 err = dev_mc_del(dev, addr); 1075 1076 return err; 1077 } 1078 1079 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1080 struct ethtool_drvinfo *drvinfo) 1081 { 1082 strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1083 strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1084 } 1085 1086 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1087 struct ethtool_link_ksettings *cmd) 1088 { 1089 const struct macvlan_dev *vlan = netdev_priv(dev); 1090 1091 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1092 } 1093 1094 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1095 struct kernel_ethtool_ts_info *info) 1096 { 1097 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1098 1099 return ethtool_get_ts_info_by_layer(real_dev, info); 1100 } 1101 1102 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1103 netdev_features_t features) 1104 { 1105 struct macvlan_dev *vlan = netdev_priv(dev); 1106 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1107 netdev_features_t mask; 1108 1109 features |= NETIF_F_ALL_FOR_ALL; 1110 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1111 mask = features; 1112 1113 lowerdev_features &= (features | ~NETIF_F_LRO); 1114 features = netdev_increment_features(lowerdev_features, features, mask); 1115 features |= ALWAYS_ON_FEATURES; 1116 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1117 1118 return features; 1119 } 1120 1121 #ifdef CONFIG_NET_POLL_CONTROLLER 1122 static void macvlan_dev_poll_controller(struct net_device *dev) 1123 { 1124 return; 1125 } 1126 1127 static int macvlan_dev_netpoll_setup(struct net_device *dev) 1128 { 1129 struct macvlan_dev *vlan = netdev_priv(dev); 1130 struct net_device *real_dev = vlan->lowerdev; 1131 struct netpoll *netpoll; 1132 int err; 1133 1134 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1135 err = -ENOMEM; 1136 if (!netpoll) 1137 goto out; 1138 1139 err = __netpoll_setup(netpoll, real_dev); 1140 if (err) { 1141 kfree(netpoll); 1142 goto out; 1143 } 1144 1145 vlan->netpoll = netpoll; 1146 1147 out: 1148 return err; 1149 } 1150 1151 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1152 { 1153 struct macvlan_dev *vlan = netdev_priv(dev); 1154 struct netpoll *netpoll = vlan->netpoll; 1155 1156 if (!netpoll) 1157 return; 1158 1159 vlan->netpoll = NULL; 1160 1161 __netpoll_free(netpoll); 1162 } 1163 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1164 1165 static int macvlan_dev_get_iflink(const struct net_device *dev) 1166 { 1167 struct macvlan_dev *vlan = netdev_priv(dev); 1168 1169 return READ_ONCE(vlan->lowerdev->ifindex); 1170 } 1171 1172 static const struct ethtool_ops macvlan_ethtool_ops = { 1173 .get_link = ethtool_op_get_link, 1174 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1175 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1176 .get_ts_info = macvlan_ethtool_get_ts_info, 1177 }; 1178 1179 static const struct net_device_ops macvlan_netdev_ops = { 1180 .ndo_init = macvlan_init, 1181 .ndo_uninit = macvlan_uninit, 1182 .ndo_open = macvlan_open, 1183 .ndo_stop = macvlan_stop, 1184 .ndo_start_xmit = macvlan_start_xmit, 1185 .ndo_change_mtu = macvlan_change_mtu, 1186 .ndo_fix_features = macvlan_fix_features, 1187 .ndo_change_rx_flags = macvlan_change_rx_flags, 1188 .ndo_set_mac_address = macvlan_set_mac_address, 1189 .ndo_set_rx_mode = macvlan_set_mac_lists, 1190 .ndo_get_stats64 = macvlan_dev_get_stats64, 1191 .ndo_validate_addr = eth_validate_addr, 1192 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1193 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1194 .ndo_fdb_add = macvlan_fdb_add, 1195 .ndo_fdb_del = macvlan_fdb_del, 1196 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1197 #ifdef CONFIG_NET_POLL_CONTROLLER 1198 .ndo_poll_controller = macvlan_dev_poll_controller, 1199 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1200 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1201 #endif 1202 .ndo_get_iflink = macvlan_dev_get_iflink, 1203 .ndo_features_check = passthru_features_check, 1204 .ndo_hwtstamp_get = macvlan_hwtstamp_get, 1205 .ndo_hwtstamp_set = macvlan_hwtstamp_set, 1206 }; 1207 1208 static void macvlan_dev_free(struct net_device *dev) 1209 { 1210 struct macvlan_dev *vlan = netdev_priv(dev); 1211 1212 /* Get rid of the macvlan's reference to lowerdev */ 1213 netdev_put(vlan->lowerdev, &vlan->dev_tracker); 1214 } 1215 1216 void macvlan_common_setup(struct net_device *dev) 1217 { 1218 ether_setup(dev); 1219 1220 /* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */ 1221 dev->max_mtu = ETH_MAX_MTU; 1222 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1223 netif_keep_dst(dev); 1224 dev->priv_flags |= IFF_UNICAST_FLT; 1225 dev->change_proto_down = true; 1226 dev->netdev_ops = &macvlan_netdev_ops; 1227 dev->needs_free_netdev = true; 1228 dev->priv_destructor = macvlan_dev_free; 1229 dev->header_ops = &macvlan_hard_header_ops; 1230 dev->ethtool_ops = &macvlan_ethtool_ops; 1231 } 1232 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1233 1234 static void macvlan_setup(struct net_device *dev) 1235 { 1236 macvlan_common_setup(dev); 1237 dev->priv_flags |= IFF_NO_QUEUE; 1238 } 1239 1240 static int macvlan_port_create(struct net_device *dev) 1241 { 1242 struct macvlan_port *port; 1243 unsigned int i; 1244 int err; 1245 1246 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1247 return -EINVAL; 1248 1249 if (netdev_is_rx_handler_busy(dev)) 1250 return -EBUSY; 1251 1252 port = kzalloc(sizeof(*port), GFP_KERNEL); 1253 if (port == NULL) 1254 return -ENOMEM; 1255 1256 port->dev = dev; 1257 ether_addr_copy(port->perm_addr, dev->dev_addr); 1258 INIT_LIST_HEAD(&port->vlans); 1259 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1260 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1261 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1262 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1263 1264 port->bc_queue_len_used = 0; 1265 port->bc_cutoff = 1; 1266 skb_queue_head_init(&port->bc_queue); 1267 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1268 1269 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1270 if (err) 1271 kfree(port); 1272 else 1273 dev->priv_flags |= IFF_MACVLAN_PORT; 1274 return err; 1275 } 1276 1277 static void macvlan_port_destroy(struct net_device *dev) 1278 { 1279 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1280 struct sk_buff *skb; 1281 1282 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1283 netdev_rx_handler_unregister(dev); 1284 1285 /* After this point, no packet can schedule bc_work anymore, 1286 * but we need to cancel it and purge left skbs if any. 1287 */ 1288 cancel_work_sync(&port->bc_work); 1289 1290 while ((skb = __skb_dequeue(&port->bc_queue))) { 1291 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1292 1293 if (src) 1294 dev_put(src->dev); 1295 1296 kfree_skb(skb); 1297 } 1298 1299 /* If the lower device address has been changed by passthru 1300 * macvlan, put it back. 1301 */ 1302 if (macvlan_passthru(port) && 1303 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1304 struct sockaddr_storage ss; 1305 1306 ss.ss_family = port->dev->type; 1307 memcpy(&ss.__data, port->perm_addr, port->dev->addr_len); 1308 dev_set_mac_address(port->dev, &ss, NULL); 1309 } 1310 1311 kfree(port); 1312 } 1313 1314 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1315 struct netlink_ext_ack *extack) 1316 { 1317 struct nlattr *nla, *head; 1318 int rem, len; 1319 1320 if (tb[IFLA_ADDRESS]) { 1321 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1322 return -EINVAL; 1323 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1324 return -EADDRNOTAVAIL; 1325 } 1326 1327 if (!data) 1328 return 0; 1329 1330 if (data[IFLA_MACVLAN_FLAGS] && 1331 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC | 1332 MACVLAN_FLAG_NODST)) 1333 return -EINVAL; 1334 1335 if (data[IFLA_MACVLAN_MODE]) { 1336 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1337 case MACVLAN_MODE_PRIVATE: 1338 case MACVLAN_MODE_VEPA: 1339 case MACVLAN_MODE_BRIDGE: 1340 case MACVLAN_MODE_PASSTHRU: 1341 case MACVLAN_MODE_SOURCE: 1342 break; 1343 default: 1344 return -EINVAL; 1345 } 1346 } 1347 1348 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1349 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1350 case MACVLAN_MACADDR_ADD: 1351 case MACVLAN_MACADDR_DEL: 1352 case MACVLAN_MACADDR_FLUSH: 1353 case MACVLAN_MACADDR_SET: 1354 break; 1355 default: 1356 return -EINVAL; 1357 } 1358 } 1359 1360 if (data[IFLA_MACVLAN_MACADDR]) { 1361 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1362 return -EINVAL; 1363 1364 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1365 return -EADDRNOTAVAIL; 1366 } 1367 1368 if (data[IFLA_MACVLAN_MACADDR_DATA]) { 1369 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1370 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1371 1372 nla_for_each_attr(nla, head, len, rem) { 1373 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1374 nla_len(nla) != ETH_ALEN) 1375 return -EINVAL; 1376 1377 if (!is_valid_ether_addr(nla_data(nla))) 1378 return -EADDRNOTAVAIL; 1379 } 1380 } 1381 1382 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1383 return -EINVAL; 1384 1385 return 0; 1386 } 1387 1388 /* 1389 * reconfigure list of remote source mac address 1390 * (only for macvlan devices in source mode) 1391 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1392 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1393 */ 1394 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1395 struct nlattr *data[]) 1396 { 1397 char *addr = NULL; 1398 int ret, rem, len; 1399 struct nlattr *nla, *head; 1400 struct macvlan_source_entry *entry; 1401 1402 if (data[IFLA_MACVLAN_MACADDR]) 1403 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1404 1405 if (mode == MACVLAN_MACADDR_ADD) { 1406 if (!addr) 1407 return -EINVAL; 1408 1409 return macvlan_hash_add_source(vlan, addr); 1410 1411 } else if (mode == MACVLAN_MACADDR_DEL) { 1412 if (!addr) 1413 return -EINVAL; 1414 1415 entry = macvlan_hash_lookup_source(vlan, addr); 1416 if (entry) { 1417 macvlan_hash_del_source(entry); 1418 vlan->macaddr_count--; 1419 } 1420 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1421 macvlan_flush_sources(vlan->port, vlan); 1422 } else if (mode == MACVLAN_MACADDR_SET) { 1423 macvlan_flush_sources(vlan->port, vlan); 1424 1425 if (addr) { 1426 ret = macvlan_hash_add_source(vlan, addr); 1427 if (ret) 1428 return ret; 1429 } 1430 1431 if (!data[IFLA_MACVLAN_MACADDR_DATA]) 1432 return 0; 1433 1434 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1435 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1436 1437 nla_for_each_attr(nla, head, len, rem) { 1438 addr = nla_data(nla); 1439 ret = macvlan_hash_add_source(vlan, addr); 1440 if (ret) 1441 return ret; 1442 } 1443 } else { 1444 return -EINVAL; 1445 } 1446 1447 return 0; 1448 } 1449 1450 int macvlan_common_newlink(struct net_device *dev, 1451 struct rtnl_newlink_params *params, 1452 struct netlink_ext_ack *extack) 1453 { 1454 struct net *link_net = rtnl_newlink_link_net(params); 1455 struct macvlan_dev *vlan = netdev_priv(dev); 1456 struct nlattr **data = params->data; 1457 struct nlattr **tb = params->tb; 1458 struct net_device *lowerdev; 1459 struct macvlan_port *port; 1460 bool create = false; 1461 int macmode; 1462 int err; 1463 1464 if (!tb[IFLA_LINK]) 1465 return -EINVAL; 1466 1467 lowerdev = __dev_get_by_index(link_net, nla_get_u32(tb[IFLA_LINK])); 1468 if (lowerdev == NULL) 1469 return -ENODEV; 1470 1471 /* When creating macvlans or macvtaps on top of other macvlans - use 1472 * the real device as the lowerdev. 1473 */ 1474 if (netif_is_macvlan(lowerdev)) 1475 lowerdev = macvlan_dev_real_dev(lowerdev); 1476 1477 if (!tb[IFLA_MTU]) 1478 dev->mtu = lowerdev->mtu; 1479 else if (dev->mtu > lowerdev->mtu) 1480 return -EINVAL; 1481 1482 /* MTU range: 68 - lowerdev->max_mtu */ 1483 dev->min_mtu = ETH_MIN_MTU; 1484 dev->max_mtu = lowerdev->max_mtu; 1485 1486 if (!tb[IFLA_ADDRESS]) 1487 eth_hw_addr_random(dev); 1488 1489 if (!netif_is_macvlan_port(lowerdev)) { 1490 err = macvlan_port_create(lowerdev); 1491 if (err < 0) 1492 return err; 1493 create = true; 1494 } 1495 port = macvlan_port_get_rtnl(lowerdev); 1496 1497 /* Only 1 macvlan device can be created in passthru mode */ 1498 if (macvlan_passthru(port)) { 1499 /* The macvlan port must be not created this time, 1500 * still goto destroy_macvlan_port for readability. 1501 */ 1502 err = -EINVAL; 1503 goto destroy_macvlan_port; 1504 } 1505 1506 vlan->lowerdev = lowerdev; 1507 vlan->dev = dev; 1508 vlan->port = port; 1509 vlan->set_features = MACVLAN_FEATURES; 1510 1511 vlan->mode = MACVLAN_MODE_VEPA; 1512 if (data && data[IFLA_MACVLAN_MODE]) 1513 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1514 1515 if (data && data[IFLA_MACVLAN_FLAGS]) 1516 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1517 1518 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1519 if (port->count) { 1520 err = -EINVAL; 1521 goto destroy_macvlan_port; 1522 } 1523 macvlan_set_passthru(port); 1524 eth_hw_addr_inherit(dev, lowerdev); 1525 } 1526 1527 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1528 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1529 err = -EINVAL; 1530 goto destroy_macvlan_port; 1531 } 1532 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1533 err = macvlan_changelink_sources(vlan, macmode, data); 1534 if (err) 1535 goto destroy_macvlan_port; 1536 } 1537 1538 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN; 1539 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) 1540 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1541 1542 if (data && data[IFLA_MACVLAN_BC_CUTOFF]) 1543 update_port_bc_cutoff( 1544 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF])); 1545 1546 err = register_netdevice(dev); 1547 if (err < 0) 1548 goto destroy_macvlan_port; 1549 1550 dev->priv_flags |= IFF_MACVLAN; 1551 err = netdev_upper_dev_link(lowerdev, dev, extack); 1552 if (err) 1553 goto unregister_netdev; 1554 1555 list_add_tail_rcu(&vlan->list, &port->vlans); 1556 update_port_bc_queue_len(vlan->port); 1557 netif_stacked_transfer_operstate(lowerdev, dev); 1558 linkwatch_fire_event(dev); 1559 1560 return 0; 1561 1562 unregister_netdev: 1563 /* macvlan_uninit would free the macvlan port */ 1564 unregister_netdevice(dev); 1565 return err; 1566 destroy_macvlan_port: 1567 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1568 * so we destroy the macvlan port only when it's valid. 1569 */ 1570 if (create && macvlan_port_get_rtnl(lowerdev)) { 1571 macvlan_flush_sources(port, vlan); 1572 macvlan_port_destroy(port->dev); 1573 } 1574 return err; 1575 } 1576 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1577 1578 static int macvlan_newlink(struct net_device *dev, 1579 struct rtnl_newlink_params *params, 1580 struct netlink_ext_ack *extack) 1581 { 1582 return macvlan_common_newlink(dev, params, extack); 1583 } 1584 1585 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1586 { 1587 struct macvlan_dev *vlan = netdev_priv(dev); 1588 1589 if (vlan->mode == MACVLAN_MODE_SOURCE) 1590 macvlan_flush_sources(vlan->port, vlan); 1591 list_del_rcu(&vlan->list); 1592 update_port_bc_queue_len(vlan->port); 1593 unregister_netdevice_queue(dev, head); 1594 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1595 } 1596 EXPORT_SYMBOL_GPL(macvlan_dellink); 1597 1598 static int macvlan_changelink(struct net_device *dev, 1599 struct nlattr *tb[], struct nlattr *data[], 1600 struct netlink_ext_ack *extack) 1601 { 1602 struct macvlan_dev *vlan = netdev_priv(dev); 1603 enum macvlan_mode mode; 1604 bool set_mode = false; 1605 enum macvlan_macaddr_mode macmode; 1606 int ret; 1607 1608 /* Validate mode, but don't set yet: setting flags may fail. */ 1609 if (data && data[IFLA_MACVLAN_MODE]) { 1610 set_mode = true; 1611 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1612 /* Passthrough mode can't be set or cleared dynamically */ 1613 if ((mode == MACVLAN_MODE_PASSTHRU) != 1614 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1615 return -EINVAL; 1616 if (vlan->mode == MACVLAN_MODE_SOURCE && 1617 vlan->mode != mode) 1618 macvlan_flush_sources(vlan->port, vlan); 1619 } 1620 1621 if (data && data[IFLA_MACVLAN_FLAGS]) { 1622 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1623 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1624 if (macvlan_passthru(vlan->port) && promisc) { 1625 int err; 1626 1627 if (flags & MACVLAN_FLAG_NOPROMISC) 1628 err = dev_set_promiscuity(vlan->lowerdev, -1); 1629 else 1630 err = dev_set_promiscuity(vlan->lowerdev, 1); 1631 if (err < 0) 1632 return err; 1633 } 1634 vlan->flags = flags; 1635 } 1636 1637 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) { 1638 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1639 update_port_bc_queue_len(vlan->port); 1640 } 1641 1642 if (data && data[IFLA_MACVLAN_BC_CUTOFF]) 1643 update_port_bc_cutoff( 1644 vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF])); 1645 1646 if (set_mode) 1647 vlan->mode = mode; 1648 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1649 if (vlan->mode != MACVLAN_MODE_SOURCE) 1650 return -EINVAL; 1651 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1652 ret = macvlan_changelink_sources(vlan, macmode, data); 1653 if (ret) 1654 return ret; 1655 } 1656 return 0; 1657 } 1658 1659 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1660 { 1661 if (vlan->macaddr_count == 0) 1662 return 0; 1663 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1664 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1665 } 1666 1667 static size_t macvlan_get_size(const struct net_device *dev) 1668 { 1669 struct macvlan_dev *vlan = netdev_priv(dev); 1670 1671 return (0 1672 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1673 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1674 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1675 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1676 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */ 1677 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */ 1678 ); 1679 } 1680 1681 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1682 const struct macvlan_dev *vlan, 1683 const int i) 1684 { 1685 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1686 struct macvlan_source_entry *entry; 1687 1688 hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) { 1689 if (rcu_access_pointer(entry->vlan) != vlan) 1690 continue; 1691 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1692 return 1; 1693 } 1694 return 0; 1695 } 1696 1697 static int macvlan_fill_info(struct sk_buff *skb, 1698 const struct net_device *dev) 1699 { 1700 struct macvlan_dev *vlan = netdev_priv(dev); 1701 struct macvlan_port *port = vlan->port; 1702 int i; 1703 struct nlattr *nest; 1704 1705 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1706 goto nla_put_failure; 1707 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1708 goto nla_put_failure; 1709 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1710 goto nla_put_failure; 1711 if (vlan->macaddr_count > 0) { 1712 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1713 if (nest == NULL) 1714 goto nla_put_failure; 1715 1716 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1717 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1718 goto nla_put_failure; 1719 } 1720 nla_nest_end(skb, nest); 1721 } 1722 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req)) 1723 goto nla_put_failure; 1724 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used)) 1725 goto nla_put_failure; 1726 if (port->bc_cutoff != 1 && 1727 nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff)) 1728 goto nla_put_failure; 1729 return 0; 1730 1731 nla_put_failure: 1732 return -EMSGSIZE; 1733 } 1734 1735 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1736 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1737 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1738 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1739 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1740 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1741 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1742 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 }, 1743 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT }, 1744 [IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 }, 1745 }; 1746 1747 int macvlan_link_register(struct rtnl_link_ops *ops) 1748 { 1749 /* common fields */ 1750 ops->validate = macvlan_validate; 1751 ops->maxtype = IFLA_MACVLAN_MAX; 1752 ops->policy = macvlan_policy; 1753 ops->changelink = macvlan_changelink; 1754 ops->get_size = macvlan_get_size; 1755 ops->fill_info = macvlan_fill_info; 1756 1757 return rtnl_link_register(ops); 1758 }; 1759 EXPORT_SYMBOL_GPL(macvlan_link_register); 1760 1761 static struct net *macvlan_get_link_net(const struct net_device *dev) 1762 { 1763 return dev_net(macvlan_dev_real_dev(dev)); 1764 } 1765 1766 static struct rtnl_link_ops macvlan_link_ops = { 1767 .kind = "macvlan", 1768 .setup = macvlan_setup, 1769 .newlink = macvlan_newlink, 1770 .dellink = macvlan_dellink, 1771 .get_link_net = macvlan_get_link_net, 1772 .priv_size = sizeof(struct macvlan_dev), 1773 }; 1774 1775 static void update_port_bc_queue_len(struct macvlan_port *port) 1776 { 1777 u32 max_bc_queue_len_req = 0; 1778 struct macvlan_dev *vlan; 1779 1780 list_for_each_entry(vlan, &port->vlans, list) { 1781 if (vlan->bc_queue_len_req > max_bc_queue_len_req) 1782 max_bc_queue_len_req = vlan->bc_queue_len_req; 1783 } 1784 port->bc_queue_len_used = max_bc_queue_len_req; 1785 } 1786 1787 static int macvlan_device_event(struct notifier_block *unused, 1788 unsigned long event, void *ptr) 1789 { 1790 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1791 struct macvlan_dev *vlan, *next; 1792 struct macvlan_port *port; 1793 LIST_HEAD(list_kill); 1794 1795 if (!netif_is_macvlan_port(dev)) 1796 return NOTIFY_DONE; 1797 1798 port = macvlan_port_get_rtnl(dev); 1799 1800 switch (event) { 1801 case NETDEV_UP: 1802 case NETDEV_DOWN: 1803 case NETDEV_CHANGE: 1804 list_for_each_entry(vlan, &port->vlans, list) 1805 netif_stacked_transfer_operstate(vlan->lowerdev, 1806 vlan->dev); 1807 break; 1808 case NETDEV_FEAT_CHANGE: 1809 list_for_each_entry(vlan, &port->vlans, list) { 1810 netif_inherit_tso_max(vlan->dev, dev); 1811 netdev_update_features(vlan->dev); 1812 } 1813 break; 1814 case NETDEV_CHANGEMTU: 1815 list_for_each_entry(vlan, &port->vlans, list) { 1816 if (vlan->dev->mtu <= dev->mtu) 1817 continue; 1818 dev_set_mtu(vlan->dev, dev->mtu); 1819 } 1820 break; 1821 case NETDEV_CHANGEADDR: 1822 if (!macvlan_passthru(port)) 1823 return NOTIFY_DONE; 1824 1825 vlan = list_first_entry_or_null(&port->vlans, 1826 struct macvlan_dev, 1827 list); 1828 1829 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1830 return NOTIFY_BAD; 1831 1832 break; 1833 case NETDEV_UNREGISTER: 1834 /* twiddle thumbs on netns device moves */ 1835 if (dev->reg_state != NETREG_UNREGISTERING) 1836 break; 1837 1838 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1839 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1840 unregister_netdevice_many(&list_kill); 1841 break; 1842 case NETDEV_PRE_TYPE_CHANGE: 1843 /* Forbid underlying device to change its type. */ 1844 return NOTIFY_BAD; 1845 1846 case NETDEV_NOTIFY_PEERS: 1847 case NETDEV_BONDING_FAILOVER: 1848 case NETDEV_RESEND_IGMP: 1849 /* Propagate to all vlans */ 1850 list_for_each_entry(vlan, &port->vlans, list) 1851 call_netdevice_notifiers(event, vlan->dev); 1852 } 1853 return NOTIFY_DONE; 1854 } 1855 1856 static struct notifier_block macvlan_notifier_block __read_mostly = { 1857 .notifier_call = macvlan_device_event, 1858 }; 1859 1860 static int __init macvlan_init_module(void) 1861 { 1862 int err; 1863 1864 register_netdevice_notifier(&macvlan_notifier_block); 1865 1866 err = macvlan_link_register(&macvlan_link_ops); 1867 if (err < 0) 1868 goto err1; 1869 return 0; 1870 err1: 1871 unregister_netdevice_notifier(&macvlan_notifier_block); 1872 return err; 1873 } 1874 1875 static void __exit macvlan_cleanup_module(void) 1876 { 1877 rtnl_link_unregister(&macvlan_link_ops); 1878 unregister_netdevice_notifier(&macvlan_notifier_block); 1879 } 1880 1881 module_init(macvlan_init_module); 1882 module_exit(macvlan_cleanup_module); 1883 1884 MODULE_LICENSE("GPL"); 1885 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1886 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1887 MODULE_ALIAS_RTNL_LINK("macvlan"); 1888