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 u32 flags; 51 int count; 52 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 53 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 54 unsigned char perm_addr[ETH_ALEN]; 55 }; 56 57 struct macvlan_source_entry { 58 struct hlist_node hlist; 59 struct macvlan_dev *vlan; 60 unsigned char addr[6+2] __aligned(sizeof(u16)); 61 struct rcu_head rcu; 62 }; 63 64 struct macvlan_skb_cb { 65 const struct macvlan_dev *src; 66 }; 67 68 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 69 70 static void macvlan_port_destroy(struct net_device *dev); 71 static void update_port_bc_queue_len(struct macvlan_port *port); 72 73 static inline bool macvlan_passthru(const struct macvlan_port *port) 74 { 75 return port->flags & MACVLAN_F_PASSTHRU; 76 } 77 78 static inline void macvlan_set_passthru(struct macvlan_port *port) 79 { 80 port->flags |= MACVLAN_F_PASSTHRU; 81 } 82 83 static inline bool macvlan_addr_change(const struct macvlan_port *port) 84 { 85 return port->flags & MACVLAN_F_ADDRCHANGE; 86 } 87 88 static inline void macvlan_set_addr_change(struct macvlan_port *port) 89 { 90 port->flags |= MACVLAN_F_ADDRCHANGE; 91 } 92 93 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 94 { 95 port->flags &= ~MACVLAN_F_ADDRCHANGE; 96 } 97 98 /* Hash Ethernet address */ 99 static u32 macvlan_eth_hash(const unsigned char *addr) 100 { 101 u64 value = get_unaligned((u64 *)addr); 102 103 /* only want 6 bytes */ 104 #ifdef __BIG_ENDIAN 105 value >>= 16; 106 #else 107 value <<= 16; 108 #endif 109 return hash_64(value, MACVLAN_HASH_BITS); 110 } 111 112 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 113 { 114 return rcu_dereference(dev->rx_handler_data); 115 } 116 117 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 118 { 119 return rtnl_dereference(dev->rx_handler_data); 120 } 121 122 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 123 const unsigned char *addr) 124 { 125 struct macvlan_dev *vlan; 126 u32 idx = macvlan_eth_hash(addr); 127 128 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist, 129 lockdep_rtnl_is_held()) { 130 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 131 return vlan; 132 } 133 return NULL; 134 } 135 136 static struct macvlan_source_entry *macvlan_hash_lookup_source( 137 const struct macvlan_dev *vlan, 138 const unsigned char *addr) 139 { 140 struct macvlan_source_entry *entry; 141 u32 idx = macvlan_eth_hash(addr); 142 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 143 144 hlist_for_each_entry_rcu(entry, h, hlist) { 145 if (ether_addr_equal_64bits(entry->addr, addr) && 146 entry->vlan == vlan) 147 return entry; 148 } 149 return NULL; 150 } 151 152 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 153 const unsigned char *addr) 154 { 155 struct macvlan_port *port = vlan->port; 156 struct macvlan_source_entry *entry; 157 struct hlist_head *h; 158 159 entry = macvlan_hash_lookup_source(vlan, addr); 160 if (entry) 161 return 0; 162 163 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 164 if (!entry) 165 return -ENOMEM; 166 167 ether_addr_copy(entry->addr, addr); 168 entry->vlan = vlan; 169 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 170 hlist_add_head_rcu(&entry->hlist, h); 171 vlan->macaddr_count++; 172 173 return 0; 174 } 175 176 static void macvlan_hash_add(struct macvlan_dev *vlan) 177 { 178 struct macvlan_port *port = vlan->port; 179 const unsigned char *addr = vlan->dev->dev_addr; 180 u32 idx = macvlan_eth_hash(addr); 181 182 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 183 } 184 185 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 186 { 187 hlist_del_rcu(&entry->hlist); 188 kfree_rcu(entry, rcu); 189 } 190 191 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 192 { 193 hlist_del_rcu(&vlan->hlist); 194 if (sync) 195 synchronize_rcu(); 196 } 197 198 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 199 const unsigned char *addr) 200 { 201 macvlan_hash_del(vlan, true); 202 /* Now that we are unhashed it is safe to change the device 203 * address without confusing packet delivery. 204 */ 205 eth_hw_addr_set(vlan->dev, addr); 206 macvlan_hash_add(vlan); 207 } 208 209 static bool macvlan_addr_busy(const struct macvlan_port *port, 210 const unsigned char *addr) 211 { 212 /* Test to see if the specified address is 213 * currently in use by the underlying device or 214 * another macvlan. 215 */ 216 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 217 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 218 return true; 219 220 if (macvlan_hash_lookup(port, addr)) 221 return true; 222 223 return false; 224 } 225 226 227 static int macvlan_broadcast_one(struct sk_buff *skb, 228 const struct macvlan_dev *vlan, 229 const struct ethhdr *eth, bool local) 230 { 231 struct net_device *dev = vlan->dev; 232 233 if (local) 234 return __dev_forward_skb(dev, skb); 235 236 skb->dev = dev; 237 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 238 skb->pkt_type = PACKET_BROADCAST; 239 else 240 skb->pkt_type = PACKET_MULTICAST; 241 242 return 0; 243 } 244 245 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 246 { 247 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 248 } 249 250 251 static unsigned int mc_hash(const struct macvlan_dev *vlan, 252 const unsigned char *addr) 253 { 254 u32 val = __get_unaligned_cpu32(addr + 2); 255 256 val ^= macvlan_hash_mix(vlan); 257 return hash_32(val, MACVLAN_MC_FILTER_BITS); 258 } 259 260 static void macvlan_broadcast(struct sk_buff *skb, 261 const struct macvlan_port *port, 262 struct net_device *src, 263 enum macvlan_mode mode) 264 { 265 const struct ethhdr *eth = eth_hdr(skb); 266 const struct macvlan_dev *vlan; 267 struct sk_buff *nskb; 268 unsigned int i; 269 int err; 270 unsigned int hash; 271 272 if (skb->protocol == htons(ETH_P_PAUSE)) 273 return; 274 275 hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) { 276 if (vlan->dev == src || !(vlan->mode & mode)) 277 continue; 278 279 hash = mc_hash(vlan, eth->h_dest); 280 if (!test_bit(hash, vlan->mc_filter)) 281 continue; 282 283 err = NET_RX_DROP; 284 nskb = skb_clone(skb, GFP_ATOMIC); 285 if (likely(nskb)) 286 err = macvlan_broadcast_one(nskb, vlan, eth, 287 mode == MACVLAN_MODE_BRIDGE) ?: 288 netif_rx(nskb); 289 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 290 err == NET_RX_SUCCESS, true); 291 } 292 } 293 294 static void macvlan_process_broadcast(struct work_struct *w) 295 { 296 struct macvlan_port *port = container_of(w, struct macvlan_port, 297 bc_work); 298 struct sk_buff *skb; 299 struct sk_buff_head list; 300 301 __skb_queue_head_init(&list); 302 303 spin_lock_bh(&port->bc_queue.lock); 304 skb_queue_splice_tail_init(&port->bc_queue, &list); 305 spin_unlock_bh(&port->bc_queue.lock); 306 307 while ((skb = __skb_dequeue(&list))) { 308 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 309 310 rcu_read_lock(); 311 312 if (!src) 313 /* frame comes from an external address */ 314 macvlan_broadcast(skb, port, NULL, 315 MACVLAN_MODE_PRIVATE | 316 MACVLAN_MODE_VEPA | 317 MACVLAN_MODE_PASSTHRU| 318 MACVLAN_MODE_BRIDGE); 319 else if (src->mode == MACVLAN_MODE_VEPA) 320 /* flood to everyone except source */ 321 macvlan_broadcast(skb, port, src->dev, 322 MACVLAN_MODE_VEPA | 323 MACVLAN_MODE_BRIDGE); 324 else 325 /* 326 * flood only to VEPA ports, bridge ports 327 * already saw the frame on the way out. 328 */ 329 macvlan_broadcast(skb, port, src->dev, 330 MACVLAN_MODE_VEPA); 331 332 rcu_read_unlock(); 333 334 if (src) 335 dev_put(src->dev); 336 consume_skb(skb); 337 338 cond_resched(); 339 } 340 } 341 342 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 343 const struct macvlan_dev *src, 344 struct sk_buff *skb) 345 { 346 struct sk_buff *nskb; 347 int err = -ENOMEM; 348 349 nskb = skb_clone(skb, GFP_ATOMIC); 350 if (!nskb) 351 goto err; 352 353 MACVLAN_SKB_CB(nskb)->src = src; 354 355 spin_lock(&port->bc_queue.lock); 356 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) { 357 if (src) 358 dev_hold(src->dev); 359 __skb_queue_tail(&port->bc_queue, nskb); 360 err = 0; 361 } 362 spin_unlock(&port->bc_queue.lock); 363 364 schedule_work(&port->bc_work); 365 366 if (err) 367 goto free_nskb; 368 369 return; 370 371 free_nskb: 372 kfree_skb(nskb); 373 err: 374 dev_core_stats_rx_dropped_inc(skb->dev); 375 } 376 377 static void macvlan_flush_sources(struct macvlan_port *port, 378 struct macvlan_dev *vlan) 379 { 380 struct macvlan_source_entry *entry; 381 struct hlist_node *next; 382 int i; 383 384 hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist) 385 if (entry->vlan == vlan) 386 macvlan_hash_del_source(entry); 387 388 vlan->macaddr_count = 0; 389 } 390 391 static void macvlan_forward_source_one(struct sk_buff *skb, 392 struct macvlan_dev *vlan) 393 { 394 struct sk_buff *nskb; 395 struct net_device *dev; 396 int len; 397 int ret; 398 399 dev = vlan->dev; 400 if (unlikely(!(dev->flags & IFF_UP))) 401 return; 402 403 nskb = skb_clone(skb, GFP_ATOMIC); 404 if (!nskb) 405 return; 406 407 len = nskb->len + ETH_HLEN; 408 nskb->dev = dev; 409 410 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr)) 411 nskb->pkt_type = PACKET_HOST; 412 413 ret = __netif_rx(nskb); 414 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 415 } 416 417 static bool macvlan_forward_source(struct sk_buff *skb, 418 struct macvlan_port *port, 419 const unsigned char *addr) 420 { 421 struct macvlan_source_entry *entry; 422 u32 idx = macvlan_eth_hash(addr); 423 struct hlist_head *h = &port->vlan_source_hash[idx]; 424 bool consume = false; 425 426 hlist_for_each_entry_rcu(entry, h, hlist) { 427 if (ether_addr_equal_64bits(entry->addr, addr)) { 428 if (entry->vlan->flags & MACVLAN_FLAG_NODST) 429 consume = true; 430 macvlan_forward_source_one(skb, entry->vlan); 431 } 432 } 433 434 return consume; 435 } 436 437 /* called under rcu_read_lock() from netif_receive_skb */ 438 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 439 { 440 struct macvlan_port *port; 441 struct sk_buff *skb = *pskb; 442 const struct ethhdr *eth = eth_hdr(skb); 443 const struct macvlan_dev *vlan; 444 const struct macvlan_dev *src; 445 struct net_device *dev; 446 unsigned int len = 0; 447 int ret; 448 rx_handler_result_t handle_res; 449 450 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */ 451 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 452 return RX_HANDLER_PASS; 453 454 port = macvlan_port_get_rcu(skb->dev); 455 if (is_multicast_ether_addr(eth->h_dest)) { 456 unsigned int hash; 457 458 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 459 if (!skb) 460 return RX_HANDLER_CONSUMED; 461 *pskb = skb; 462 eth = eth_hdr(skb); 463 if (macvlan_forward_source(skb, port, eth->h_source)) 464 return RX_HANDLER_CONSUMED; 465 src = macvlan_hash_lookup(port, eth->h_source); 466 if (src && src->mode != MACVLAN_MODE_VEPA && 467 src->mode != MACVLAN_MODE_BRIDGE) { 468 /* forward to original port. */ 469 vlan = src; 470 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 471 __netif_rx(skb); 472 handle_res = RX_HANDLER_CONSUMED; 473 goto out; 474 } 475 476 hash = mc_hash(NULL, eth->h_dest); 477 if (test_bit(hash, port->mc_filter)) 478 macvlan_broadcast_enqueue(port, src, skb); 479 480 return RX_HANDLER_PASS; 481 } 482 483 if (macvlan_forward_source(skb, port, eth->h_source)) 484 return RX_HANDLER_CONSUMED; 485 if (macvlan_passthru(port)) 486 vlan = list_first_or_null_rcu(&port->vlans, 487 struct macvlan_dev, list); 488 else 489 vlan = macvlan_hash_lookup(port, eth->h_dest); 490 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 491 return RX_HANDLER_PASS; 492 493 dev = vlan->dev; 494 if (unlikely(!(dev->flags & IFF_UP))) { 495 kfree_skb(skb); 496 return RX_HANDLER_CONSUMED; 497 } 498 len = skb->len + ETH_HLEN; 499 skb = skb_share_check(skb, GFP_ATOMIC); 500 if (!skb) { 501 ret = NET_RX_DROP; 502 handle_res = RX_HANDLER_CONSUMED; 503 goto out; 504 } 505 506 *pskb = skb; 507 skb->dev = dev; 508 skb->pkt_type = PACKET_HOST; 509 510 ret = NET_RX_SUCCESS; 511 handle_res = RX_HANDLER_ANOTHER; 512 out: 513 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 514 return handle_res; 515 } 516 517 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 518 { 519 const struct macvlan_dev *vlan = netdev_priv(dev); 520 const struct macvlan_port *port = vlan->port; 521 const struct macvlan_dev *dest; 522 523 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 524 const struct ethhdr *eth = skb_eth_hdr(skb); 525 526 /* send to other bridge ports directly */ 527 if (is_multicast_ether_addr(eth->h_dest)) { 528 skb_reset_mac_header(skb); 529 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 530 goto xmit_world; 531 } 532 533 dest = macvlan_hash_lookup(port, eth->h_dest); 534 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 535 /* send to lowerdev first for its network taps */ 536 dev_forward_skb(vlan->lowerdev, skb); 537 538 return NET_XMIT_SUCCESS; 539 } 540 } 541 xmit_world: 542 skb->dev = vlan->lowerdev; 543 return dev_queue_xmit_accel(skb, 544 netdev_get_sb_channel(dev) ? dev : NULL); 545 } 546 547 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 548 { 549 #ifdef CONFIG_NET_POLL_CONTROLLER 550 return netpoll_send_skb(vlan->netpoll, skb); 551 #else 552 BUG(); 553 return NETDEV_TX_OK; 554 #endif 555 } 556 557 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 558 struct net_device *dev) 559 { 560 struct macvlan_dev *vlan = netdev_priv(dev); 561 unsigned int len = skb->len; 562 int ret; 563 564 if (unlikely(netpoll_tx_running(dev))) 565 return macvlan_netpoll_send_skb(vlan, skb); 566 567 ret = macvlan_queue_xmit(skb, dev); 568 569 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 570 struct vlan_pcpu_stats *pcpu_stats; 571 572 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 573 u64_stats_update_begin(&pcpu_stats->syncp); 574 pcpu_stats->tx_packets++; 575 pcpu_stats->tx_bytes += len; 576 u64_stats_update_end(&pcpu_stats->syncp); 577 } else { 578 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 579 } 580 return ret; 581 } 582 583 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 584 unsigned short type, const void *daddr, 585 const void *saddr, unsigned len) 586 { 587 const struct macvlan_dev *vlan = netdev_priv(dev); 588 struct net_device *lowerdev = vlan->lowerdev; 589 590 return dev_hard_header(skb, lowerdev, type, daddr, 591 saddr ? : dev->dev_addr, len); 592 } 593 594 static const struct header_ops macvlan_hard_header_ops = { 595 .create = macvlan_hard_header, 596 .parse = eth_header_parse, 597 .cache = eth_header_cache, 598 .cache_update = eth_header_cache_update, 599 }; 600 601 static int macvlan_open(struct net_device *dev) 602 { 603 struct macvlan_dev *vlan = netdev_priv(dev); 604 struct net_device *lowerdev = vlan->lowerdev; 605 int err; 606 607 if (macvlan_passthru(vlan->port)) { 608 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 609 err = dev_set_promiscuity(lowerdev, 1); 610 if (err < 0) 611 goto out; 612 } 613 goto hash_add; 614 } 615 616 err = -EADDRINUSE; 617 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 618 goto out; 619 620 /* Attempt to populate accel_priv which is used to offload the L2 621 * forwarding requests for unicast packets. 622 */ 623 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 624 vlan->accel_priv = 625 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 626 627 /* If earlier attempt to offload failed, or accel_priv is not 628 * populated we must add the unicast address to the lower device. 629 */ 630 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 631 vlan->accel_priv = NULL; 632 err = dev_uc_add(lowerdev, dev->dev_addr); 633 if (err < 0) 634 goto out; 635 } 636 637 if (dev->flags & IFF_ALLMULTI) { 638 err = dev_set_allmulti(lowerdev, 1); 639 if (err < 0) 640 goto del_unicast; 641 } 642 643 if (dev->flags & IFF_PROMISC) { 644 err = dev_set_promiscuity(lowerdev, 1); 645 if (err < 0) 646 goto clear_multi; 647 } 648 649 hash_add: 650 macvlan_hash_add(vlan); 651 return 0; 652 653 clear_multi: 654 if (dev->flags & IFF_ALLMULTI) 655 dev_set_allmulti(lowerdev, -1); 656 del_unicast: 657 if (vlan->accel_priv) { 658 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 659 vlan->accel_priv); 660 vlan->accel_priv = NULL; 661 } else { 662 dev_uc_del(lowerdev, dev->dev_addr); 663 } 664 out: 665 return err; 666 } 667 668 static int macvlan_stop(struct net_device *dev) 669 { 670 struct macvlan_dev *vlan = netdev_priv(dev); 671 struct net_device *lowerdev = vlan->lowerdev; 672 673 if (vlan->accel_priv) { 674 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 675 vlan->accel_priv); 676 vlan->accel_priv = NULL; 677 } 678 679 dev_uc_unsync(lowerdev, dev); 680 dev_mc_unsync(lowerdev, dev); 681 682 if (macvlan_passthru(vlan->port)) { 683 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 684 dev_set_promiscuity(lowerdev, -1); 685 goto hash_del; 686 } 687 688 if (dev->flags & IFF_ALLMULTI) 689 dev_set_allmulti(lowerdev, -1); 690 691 if (dev->flags & IFF_PROMISC) 692 dev_set_promiscuity(lowerdev, -1); 693 694 dev_uc_del(lowerdev, dev->dev_addr); 695 696 hash_del: 697 macvlan_hash_del(vlan, !dev->dismantle); 698 return 0; 699 } 700 701 static int macvlan_sync_address(struct net_device *dev, 702 const unsigned char *addr) 703 { 704 struct macvlan_dev *vlan = netdev_priv(dev); 705 struct net_device *lowerdev = vlan->lowerdev; 706 struct macvlan_port *port = vlan->port; 707 int err; 708 709 if (!(dev->flags & IFF_UP)) { 710 /* Just copy in the new address */ 711 eth_hw_addr_set(dev, addr); 712 } else { 713 /* Rehash and update the device filters */ 714 if (macvlan_addr_busy(vlan->port, addr)) 715 return -EADDRINUSE; 716 717 if (!macvlan_passthru(port)) { 718 err = dev_uc_add(lowerdev, addr); 719 if (err) 720 return err; 721 722 dev_uc_del(lowerdev, dev->dev_addr); 723 } 724 725 macvlan_hash_change_addr(vlan, addr); 726 } 727 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 728 /* Since addr_change isn't set, we are here due to lower 729 * device change. Save the lower-dev address so we can 730 * restore it later. 731 */ 732 ether_addr_copy(vlan->port->perm_addr, 733 lowerdev->dev_addr); 734 } 735 macvlan_clear_addr_change(port); 736 return 0; 737 } 738 739 static int macvlan_set_mac_address(struct net_device *dev, void *p) 740 { 741 struct macvlan_dev *vlan = netdev_priv(dev); 742 struct sockaddr *addr = p; 743 744 if (!is_valid_ether_addr(addr->sa_data)) 745 return -EADDRNOTAVAIL; 746 747 /* If the addresses are the same, this is a no-op */ 748 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 749 return 0; 750 751 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 752 macvlan_set_addr_change(vlan->port); 753 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 754 } 755 756 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 757 return -EADDRINUSE; 758 759 return macvlan_sync_address(dev, addr->sa_data); 760 } 761 762 static void macvlan_change_rx_flags(struct net_device *dev, int change) 763 { 764 struct macvlan_dev *vlan = netdev_priv(dev); 765 struct net_device *lowerdev = vlan->lowerdev; 766 767 if (dev->flags & IFF_UP) { 768 if (change & IFF_ALLMULTI) 769 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 770 if (change & IFF_PROMISC) 771 dev_set_promiscuity(lowerdev, 772 dev->flags & IFF_PROMISC ? 1 : -1); 773 774 } 775 } 776 777 static void macvlan_compute_filter(unsigned long *mc_filter, 778 struct net_device *dev, 779 struct macvlan_dev *vlan) 780 { 781 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 782 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 783 } else { 784 struct netdev_hw_addr *ha; 785 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 786 787 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 788 netdev_for_each_mc_addr(ha, dev) { 789 __set_bit(mc_hash(vlan, ha->addr), filter); 790 } 791 792 __set_bit(mc_hash(vlan, dev->broadcast), filter); 793 794 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 795 } 796 } 797 798 static void macvlan_set_mac_lists(struct net_device *dev) 799 { 800 struct macvlan_dev *vlan = netdev_priv(dev); 801 802 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 803 804 dev_uc_sync(vlan->lowerdev, dev); 805 dev_mc_sync(vlan->lowerdev, dev); 806 807 /* This is slightly inaccurate as we're including the subscription 808 * list of vlan->lowerdev too. 809 * 810 * Bug alert: This only works if everyone has the same broadcast 811 * address as lowerdev. As soon as someone changes theirs this 812 * will break. 813 * 814 * However, this is already broken as when you change your broadcast 815 * address we don't get called. 816 * 817 * The solution is to maintain a list of broadcast addresses like 818 * we do for uc/mc, if you care. 819 */ 820 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 821 } 822 823 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 824 { 825 struct macvlan_dev *vlan = netdev_priv(dev); 826 827 if (vlan->lowerdev->mtu < new_mtu) 828 return -EINVAL; 829 dev->mtu = new_mtu; 830 return 0; 831 } 832 833 static int macvlan_eth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 834 { 835 struct net_device *real_dev = macvlan_dev_real_dev(dev); 836 const struct net_device_ops *ops = real_dev->netdev_ops; 837 struct ifreq ifrr; 838 int err = -EOPNOTSUPP; 839 840 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ); 841 ifrr.ifr_ifru = ifr->ifr_ifru; 842 843 switch (cmd) { 844 case SIOCSHWTSTAMP: 845 if (!net_eq(dev_net(dev), &init_net)) 846 break; 847 fallthrough; 848 case SIOCGHWTSTAMP: 849 if (netif_device_present(real_dev) && ops->ndo_eth_ioctl) 850 err = ops->ndo_eth_ioctl(real_dev, &ifrr, cmd); 851 break; 852 } 853 854 if (!err) 855 ifr->ifr_ifru = ifrr.ifr_ifru; 856 857 return err; 858 } 859 860 /* 861 * macvlan network devices have devices nesting below it and are a special 862 * "super class" of normal network devices; split their locks off into a 863 * separate class since they always nest. 864 */ 865 static struct lock_class_key macvlan_netdev_addr_lock_key; 866 867 #define ALWAYS_ON_OFFLOADS \ 868 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 869 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 870 871 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 872 873 #define MACVLAN_FEATURES \ 874 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 875 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 876 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 877 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 878 879 #define MACVLAN_STATE_MASK \ 880 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 881 882 static void macvlan_set_lockdep_class(struct net_device *dev) 883 { 884 netdev_lockdep_set_classes(dev); 885 lockdep_set_class(&dev->addr_list_lock, 886 &macvlan_netdev_addr_lock_key); 887 } 888 889 static int macvlan_init(struct net_device *dev) 890 { 891 struct macvlan_dev *vlan = netdev_priv(dev); 892 struct net_device *lowerdev = vlan->lowerdev; 893 struct macvlan_port *port = vlan->port; 894 895 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 896 (lowerdev->state & MACVLAN_STATE_MASK); 897 dev->features = lowerdev->features & MACVLAN_FEATURES; 898 dev->features |= ALWAYS_ON_FEATURES; 899 dev->hw_features |= NETIF_F_LRO; 900 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 901 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 902 dev->hw_enc_features |= dev->features; 903 netif_set_gso_max_size(dev, lowerdev->gso_max_size); 904 netif_set_gso_max_segs(dev, lowerdev->gso_max_segs); 905 dev->hard_header_len = lowerdev->hard_header_len; 906 macvlan_set_lockdep_class(dev); 907 908 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 909 if (!vlan->pcpu_stats) 910 return -ENOMEM; 911 912 port->count += 1; 913 914 /* Get macvlan's reference to lowerdev */ 915 dev_hold_track(lowerdev, &vlan->dev_tracker, GFP_KERNEL); 916 917 return 0; 918 } 919 920 static void macvlan_uninit(struct net_device *dev) 921 { 922 struct macvlan_dev *vlan = netdev_priv(dev); 923 struct macvlan_port *port = vlan->port; 924 925 free_percpu(vlan->pcpu_stats); 926 927 macvlan_flush_sources(port, vlan); 928 port->count -= 1; 929 if (!port->count) 930 macvlan_port_destroy(port->dev); 931 } 932 933 static void macvlan_dev_get_stats64(struct net_device *dev, 934 struct rtnl_link_stats64 *stats) 935 { 936 struct macvlan_dev *vlan = netdev_priv(dev); 937 938 if (vlan->pcpu_stats) { 939 struct vlan_pcpu_stats *p; 940 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 941 u32 rx_errors = 0, tx_dropped = 0; 942 unsigned int start; 943 int i; 944 945 for_each_possible_cpu(i) { 946 p = per_cpu_ptr(vlan->pcpu_stats, i); 947 do { 948 start = u64_stats_fetch_begin_irq(&p->syncp); 949 rx_packets = p->rx_packets; 950 rx_bytes = p->rx_bytes; 951 rx_multicast = p->rx_multicast; 952 tx_packets = p->tx_packets; 953 tx_bytes = p->tx_bytes; 954 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 955 956 stats->rx_packets += rx_packets; 957 stats->rx_bytes += rx_bytes; 958 stats->multicast += rx_multicast; 959 stats->tx_packets += tx_packets; 960 stats->tx_bytes += tx_bytes; 961 /* rx_errors & tx_dropped are u32, updated 962 * without syncp protection. 963 */ 964 rx_errors += p->rx_errors; 965 tx_dropped += p->tx_dropped; 966 } 967 stats->rx_errors = rx_errors; 968 stats->rx_dropped = rx_errors; 969 stats->tx_dropped = tx_dropped; 970 } 971 } 972 973 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 974 __be16 proto, u16 vid) 975 { 976 struct macvlan_dev *vlan = netdev_priv(dev); 977 struct net_device *lowerdev = vlan->lowerdev; 978 979 return vlan_vid_add(lowerdev, proto, vid); 980 } 981 982 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 983 __be16 proto, u16 vid) 984 { 985 struct macvlan_dev *vlan = netdev_priv(dev); 986 struct net_device *lowerdev = vlan->lowerdev; 987 988 vlan_vid_del(lowerdev, proto, vid); 989 return 0; 990 } 991 992 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 993 struct net_device *dev, 994 const unsigned char *addr, u16 vid, 995 u16 flags, 996 struct netlink_ext_ack *extack) 997 { 998 struct macvlan_dev *vlan = netdev_priv(dev); 999 int err = -EINVAL; 1000 1001 /* Support unicast filter only on passthru devices. 1002 * Multicast filter should be allowed on all devices. 1003 */ 1004 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1005 return -EOPNOTSUPP; 1006 1007 if (flags & NLM_F_REPLACE) 1008 return -EOPNOTSUPP; 1009 1010 if (is_unicast_ether_addr(addr)) 1011 err = dev_uc_add_excl(dev, addr); 1012 else if (is_multicast_ether_addr(addr)) 1013 err = dev_mc_add_excl(dev, addr); 1014 1015 return err; 1016 } 1017 1018 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1019 struct net_device *dev, 1020 const unsigned char *addr, u16 vid) 1021 { 1022 struct macvlan_dev *vlan = netdev_priv(dev); 1023 int err = -EINVAL; 1024 1025 /* Support unicast filter only on passthru devices. 1026 * Multicast filter should be allowed on all devices. 1027 */ 1028 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1029 return -EOPNOTSUPP; 1030 1031 if (is_unicast_ether_addr(addr)) 1032 err = dev_uc_del(dev, addr); 1033 else if (is_multicast_ether_addr(addr)) 1034 err = dev_mc_del(dev, addr); 1035 1036 return err; 1037 } 1038 1039 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1040 struct ethtool_drvinfo *drvinfo) 1041 { 1042 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1043 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1044 } 1045 1046 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1047 struct ethtool_link_ksettings *cmd) 1048 { 1049 const struct macvlan_dev *vlan = netdev_priv(dev); 1050 1051 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1052 } 1053 1054 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1055 struct ethtool_ts_info *info) 1056 { 1057 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1058 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1059 struct phy_device *phydev = real_dev->phydev; 1060 1061 if (phy_has_tsinfo(phydev)) { 1062 return phy_ts_info(phydev, info); 1063 } else if (ops->get_ts_info) { 1064 return ops->get_ts_info(real_dev, info); 1065 } else { 1066 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1067 SOF_TIMESTAMPING_SOFTWARE; 1068 info->phc_index = -1; 1069 } 1070 1071 return 0; 1072 } 1073 1074 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1075 netdev_features_t features) 1076 { 1077 struct macvlan_dev *vlan = netdev_priv(dev); 1078 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1079 netdev_features_t mask; 1080 1081 features |= NETIF_F_ALL_FOR_ALL; 1082 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1083 mask = features; 1084 1085 lowerdev_features &= (features | ~NETIF_F_LRO); 1086 features = netdev_increment_features(lowerdev_features, features, mask); 1087 features |= ALWAYS_ON_FEATURES; 1088 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1089 1090 return features; 1091 } 1092 1093 #ifdef CONFIG_NET_POLL_CONTROLLER 1094 static void macvlan_dev_poll_controller(struct net_device *dev) 1095 { 1096 return; 1097 } 1098 1099 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1100 { 1101 struct macvlan_dev *vlan = netdev_priv(dev); 1102 struct net_device *real_dev = vlan->lowerdev; 1103 struct netpoll *netpoll; 1104 int err; 1105 1106 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1107 err = -ENOMEM; 1108 if (!netpoll) 1109 goto out; 1110 1111 err = __netpoll_setup(netpoll, real_dev); 1112 if (err) { 1113 kfree(netpoll); 1114 goto out; 1115 } 1116 1117 vlan->netpoll = netpoll; 1118 1119 out: 1120 return err; 1121 } 1122 1123 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1124 { 1125 struct macvlan_dev *vlan = netdev_priv(dev); 1126 struct netpoll *netpoll = vlan->netpoll; 1127 1128 if (!netpoll) 1129 return; 1130 1131 vlan->netpoll = NULL; 1132 1133 __netpoll_free(netpoll); 1134 } 1135 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1136 1137 static int macvlan_dev_get_iflink(const struct net_device *dev) 1138 { 1139 struct macvlan_dev *vlan = netdev_priv(dev); 1140 1141 return vlan->lowerdev->ifindex; 1142 } 1143 1144 static const struct ethtool_ops macvlan_ethtool_ops = { 1145 .get_link = ethtool_op_get_link, 1146 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1147 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1148 .get_ts_info = macvlan_ethtool_get_ts_info, 1149 }; 1150 1151 static const struct net_device_ops macvlan_netdev_ops = { 1152 .ndo_init = macvlan_init, 1153 .ndo_uninit = macvlan_uninit, 1154 .ndo_open = macvlan_open, 1155 .ndo_stop = macvlan_stop, 1156 .ndo_start_xmit = macvlan_start_xmit, 1157 .ndo_change_mtu = macvlan_change_mtu, 1158 .ndo_eth_ioctl = macvlan_eth_ioctl, 1159 .ndo_fix_features = macvlan_fix_features, 1160 .ndo_change_rx_flags = macvlan_change_rx_flags, 1161 .ndo_set_mac_address = macvlan_set_mac_address, 1162 .ndo_set_rx_mode = macvlan_set_mac_lists, 1163 .ndo_get_stats64 = macvlan_dev_get_stats64, 1164 .ndo_validate_addr = eth_validate_addr, 1165 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1166 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1167 .ndo_fdb_add = macvlan_fdb_add, 1168 .ndo_fdb_del = macvlan_fdb_del, 1169 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1170 #ifdef CONFIG_NET_POLL_CONTROLLER 1171 .ndo_poll_controller = macvlan_dev_poll_controller, 1172 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1173 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1174 #endif 1175 .ndo_get_iflink = macvlan_dev_get_iflink, 1176 .ndo_features_check = passthru_features_check, 1177 }; 1178 1179 static void macvlan_dev_free(struct net_device *dev) 1180 { 1181 struct macvlan_dev *vlan = netdev_priv(dev); 1182 1183 /* Get rid of the macvlan's reference to lowerdev */ 1184 dev_put_track(vlan->lowerdev, &vlan->dev_tracker); 1185 } 1186 1187 void macvlan_common_setup(struct net_device *dev) 1188 { 1189 ether_setup(dev); 1190 1191 dev->min_mtu = 0; 1192 dev->max_mtu = ETH_MAX_MTU; 1193 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1194 netif_keep_dst(dev); 1195 dev->priv_flags |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN; 1196 dev->netdev_ops = &macvlan_netdev_ops; 1197 dev->needs_free_netdev = true; 1198 dev->priv_destructor = macvlan_dev_free; 1199 dev->header_ops = &macvlan_hard_header_ops; 1200 dev->ethtool_ops = &macvlan_ethtool_ops; 1201 } 1202 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1203 1204 static void macvlan_setup(struct net_device *dev) 1205 { 1206 macvlan_common_setup(dev); 1207 dev->priv_flags |= IFF_NO_QUEUE; 1208 } 1209 1210 static int macvlan_port_create(struct net_device *dev) 1211 { 1212 struct macvlan_port *port; 1213 unsigned int i; 1214 int err; 1215 1216 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1217 return -EINVAL; 1218 1219 if (netdev_is_rx_handler_busy(dev)) 1220 return -EBUSY; 1221 1222 port = kzalloc(sizeof(*port), GFP_KERNEL); 1223 if (port == NULL) 1224 return -ENOMEM; 1225 1226 port->dev = dev; 1227 ether_addr_copy(port->perm_addr, dev->dev_addr); 1228 INIT_LIST_HEAD(&port->vlans); 1229 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1230 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1231 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1232 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1233 1234 port->bc_queue_len_used = 0; 1235 skb_queue_head_init(&port->bc_queue); 1236 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1237 1238 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1239 if (err) 1240 kfree(port); 1241 else 1242 dev->priv_flags |= IFF_MACVLAN_PORT; 1243 return err; 1244 } 1245 1246 static void macvlan_port_destroy(struct net_device *dev) 1247 { 1248 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1249 struct sk_buff *skb; 1250 1251 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1252 netdev_rx_handler_unregister(dev); 1253 1254 /* After this point, no packet can schedule bc_work anymore, 1255 * but we need to cancel it and purge left skbs if any. 1256 */ 1257 cancel_work_sync(&port->bc_work); 1258 1259 while ((skb = __skb_dequeue(&port->bc_queue))) { 1260 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1261 1262 if (src) 1263 dev_put(src->dev); 1264 1265 kfree_skb(skb); 1266 } 1267 1268 /* If the lower device address has been changed by passthru 1269 * macvlan, put it back. 1270 */ 1271 if (macvlan_passthru(port) && 1272 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1273 struct sockaddr sa; 1274 1275 sa.sa_family = port->dev->type; 1276 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1277 dev_set_mac_address(port->dev, &sa, NULL); 1278 } 1279 1280 kfree(port); 1281 } 1282 1283 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1284 struct netlink_ext_ack *extack) 1285 { 1286 struct nlattr *nla, *head; 1287 int rem, len; 1288 1289 if (tb[IFLA_ADDRESS]) { 1290 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1291 return -EINVAL; 1292 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1293 return -EADDRNOTAVAIL; 1294 } 1295 1296 if (!data) 1297 return 0; 1298 1299 if (data[IFLA_MACVLAN_FLAGS] && 1300 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC | 1301 MACVLAN_FLAG_NODST)) 1302 return -EINVAL; 1303 1304 if (data[IFLA_MACVLAN_MODE]) { 1305 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1306 case MACVLAN_MODE_PRIVATE: 1307 case MACVLAN_MODE_VEPA: 1308 case MACVLAN_MODE_BRIDGE: 1309 case MACVLAN_MODE_PASSTHRU: 1310 case MACVLAN_MODE_SOURCE: 1311 break; 1312 default: 1313 return -EINVAL; 1314 } 1315 } 1316 1317 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1318 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1319 case MACVLAN_MACADDR_ADD: 1320 case MACVLAN_MACADDR_DEL: 1321 case MACVLAN_MACADDR_FLUSH: 1322 case MACVLAN_MACADDR_SET: 1323 break; 1324 default: 1325 return -EINVAL; 1326 } 1327 } 1328 1329 if (data[IFLA_MACVLAN_MACADDR]) { 1330 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1331 return -EINVAL; 1332 1333 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1334 return -EADDRNOTAVAIL; 1335 } 1336 1337 if (data[IFLA_MACVLAN_MACADDR_DATA]) { 1338 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1339 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1340 1341 nla_for_each_attr(nla, head, len, rem) { 1342 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1343 nla_len(nla) != ETH_ALEN) 1344 return -EINVAL; 1345 1346 if (!is_valid_ether_addr(nla_data(nla))) 1347 return -EADDRNOTAVAIL; 1348 } 1349 } 1350 1351 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1352 return -EINVAL; 1353 1354 return 0; 1355 } 1356 1357 /* 1358 * reconfigure list of remote source mac address 1359 * (only for macvlan devices in source mode) 1360 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1361 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1362 */ 1363 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1364 struct nlattr *data[]) 1365 { 1366 char *addr = NULL; 1367 int ret, rem, len; 1368 struct nlattr *nla, *head; 1369 struct macvlan_source_entry *entry; 1370 1371 if (data[IFLA_MACVLAN_MACADDR]) 1372 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1373 1374 if (mode == MACVLAN_MACADDR_ADD) { 1375 if (!addr) 1376 return -EINVAL; 1377 1378 return macvlan_hash_add_source(vlan, addr); 1379 1380 } else if (mode == MACVLAN_MACADDR_DEL) { 1381 if (!addr) 1382 return -EINVAL; 1383 1384 entry = macvlan_hash_lookup_source(vlan, addr); 1385 if (entry) { 1386 macvlan_hash_del_source(entry); 1387 vlan->macaddr_count--; 1388 } 1389 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1390 macvlan_flush_sources(vlan->port, vlan); 1391 } else if (mode == MACVLAN_MACADDR_SET) { 1392 macvlan_flush_sources(vlan->port, vlan); 1393 1394 if (addr) { 1395 ret = macvlan_hash_add_source(vlan, addr); 1396 if (ret) 1397 return ret; 1398 } 1399 1400 if (!data[IFLA_MACVLAN_MACADDR_DATA]) 1401 return 0; 1402 1403 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1404 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1405 1406 nla_for_each_attr(nla, head, len, rem) { 1407 addr = nla_data(nla); 1408 ret = macvlan_hash_add_source(vlan, addr); 1409 if (ret) 1410 return ret; 1411 } 1412 } else { 1413 return -EINVAL; 1414 } 1415 1416 return 0; 1417 } 1418 1419 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1420 struct nlattr *tb[], struct nlattr *data[], 1421 struct netlink_ext_ack *extack) 1422 { 1423 struct macvlan_dev *vlan = netdev_priv(dev); 1424 struct macvlan_port *port; 1425 struct net_device *lowerdev; 1426 int err; 1427 int macmode; 1428 bool create = false; 1429 1430 if (!tb[IFLA_LINK]) 1431 return -EINVAL; 1432 1433 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1434 if (lowerdev == NULL) 1435 return -ENODEV; 1436 1437 /* When creating macvlans or macvtaps on top of other macvlans - use 1438 * the real device as the lowerdev. 1439 */ 1440 if (netif_is_macvlan(lowerdev)) 1441 lowerdev = macvlan_dev_real_dev(lowerdev); 1442 1443 if (!tb[IFLA_MTU]) 1444 dev->mtu = lowerdev->mtu; 1445 else if (dev->mtu > lowerdev->mtu) 1446 return -EINVAL; 1447 1448 /* MTU range: 68 - lowerdev->max_mtu */ 1449 dev->min_mtu = ETH_MIN_MTU; 1450 dev->max_mtu = lowerdev->max_mtu; 1451 1452 if (!tb[IFLA_ADDRESS]) 1453 eth_hw_addr_random(dev); 1454 1455 if (!netif_is_macvlan_port(lowerdev)) { 1456 err = macvlan_port_create(lowerdev); 1457 if (err < 0) 1458 return err; 1459 create = true; 1460 } 1461 port = macvlan_port_get_rtnl(lowerdev); 1462 1463 /* Only 1 macvlan device can be created in passthru mode */ 1464 if (macvlan_passthru(port)) { 1465 /* The macvlan port must be not created this time, 1466 * still goto destroy_macvlan_port for readability. 1467 */ 1468 err = -EINVAL; 1469 goto destroy_macvlan_port; 1470 } 1471 1472 vlan->lowerdev = lowerdev; 1473 vlan->dev = dev; 1474 vlan->port = port; 1475 vlan->set_features = MACVLAN_FEATURES; 1476 1477 vlan->mode = MACVLAN_MODE_VEPA; 1478 if (data && data[IFLA_MACVLAN_MODE]) 1479 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1480 1481 if (data && data[IFLA_MACVLAN_FLAGS]) 1482 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1483 1484 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1485 if (port->count) { 1486 err = -EINVAL; 1487 goto destroy_macvlan_port; 1488 } 1489 macvlan_set_passthru(port); 1490 eth_hw_addr_inherit(dev, lowerdev); 1491 } 1492 1493 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1494 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1495 err = -EINVAL; 1496 goto destroy_macvlan_port; 1497 } 1498 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1499 err = macvlan_changelink_sources(vlan, macmode, data); 1500 if (err) 1501 goto destroy_macvlan_port; 1502 } 1503 1504 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN; 1505 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) 1506 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1507 1508 err = register_netdevice(dev); 1509 if (err < 0) 1510 goto destroy_macvlan_port; 1511 1512 dev->priv_flags |= IFF_MACVLAN; 1513 err = netdev_upper_dev_link(lowerdev, dev, extack); 1514 if (err) 1515 goto unregister_netdev; 1516 1517 list_add_tail_rcu(&vlan->list, &port->vlans); 1518 update_port_bc_queue_len(vlan->port); 1519 netif_stacked_transfer_operstate(lowerdev, dev); 1520 linkwatch_fire_event(dev); 1521 1522 return 0; 1523 1524 unregister_netdev: 1525 /* macvlan_uninit would free the macvlan port */ 1526 unregister_netdevice(dev); 1527 return err; 1528 destroy_macvlan_port: 1529 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1530 * so we destroy the macvlan port only when it's valid. 1531 */ 1532 if (create && macvlan_port_get_rtnl(lowerdev)) 1533 macvlan_port_destroy(port->dev); 1534 return err; 1535 } 1536 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1537 1538 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1539 struct nlattr *tb[], struct nlattr *data[], 1540 struct netlink_ext_ack *extack) 1541 { 1542 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1543 } 1544 1545 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1546 { 1547 struct macvlan_dev *vlan = netdev_priv(dev); 1548 1549 if (vlan->mode == MACVLAN_MODE_SOURCE) 1550 macvlan_flush_sources(vlan->port, vlan); 1551 list_del_rcu(&vlan->list); 1552 update_port_bc_queue_len(vlan->port); 1553 unregister_netdevice_queue(dev, head); 1554 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1555 } 1556 EXPORT_SYMBOL_GPL(macvlan_dellink); 1557 1558 static int macvlan_changelink(struct net_device *dev, 1559 struct nlattr *tb[], struct nlattr *data[], 1560 struct netlink_ext_ack *extack) 1561 { 1562 struct macvlan_dev *vlan = netdev_priv(dev); 1563 enum macvlan_mode mode; 1564 bool set_mode = false; 1565 enum macvlan_macaddr_mode macmode; 1566 int ret; 1567 1568 /* Validate mode, but don't set yet: setting flags may fail. */ 1569 if (data && data[IFLA_MACVLAN_MODE]) { 1570 set_mode = true; 1571 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1572 /* Passthrough mode can't be set or cleared dynamically */ 1573 if ((mode == MACVLAN_MODE_PASSTHRU) != 1574 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1575 return -EINVAL; 1576 if (vlan->mode == MACVLAN_MODE_SOURCE && 1577 vlan->mode != mode) 1578 macvlan_flush_sources(vlan->port, vlan); 1579 } 1580 1581 if (data && data[IFLA_MACVLAN_FLAGS]) { 1582 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1583 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1584 if (macvlan_passthru(vlan->port) && promisc) { 1585 int err; 1586 1587 if (flags & MACVLAN_FLAG_NOPROMISC) 1588 err = dev_set_promiscuity(vlan->lowerdev, -1); 1589 else 1590 err = dev_set_promiscuity(vlan->lowerdev, 1); 1591 if (err < 0) 1592 return err; 1593 } 1594 vlan->flags = flags; 1595 } 1596 1597 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) { 1598 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1599 update_port_bc_queue_len(vlan->port); 1600 } 1601 1602 if (set_mode) 1603 vlan->mode = mode; 1604 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1605 if (vlan->mode != MACVLAN_MODE_SOURCE) 1606 return -EINVAL; 1607 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1608 ret = macvlan_changelink_sources(vlan, macmode, data); 1609 if (ret) 1610 return ret; 1611 } 1612 return 0; 1613 } 1614 1615 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1616 { 1617 if (vlan->macaddr_count == 0) 1618 return 0; 1619 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1620 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1621 } 1622 1623 static size_t macvlan_get_size(const struct net_device *dev) 1624 { 1625 struct macvlan_dev *vlan = netdev_priv(dev); 1626 1627 return (0 1628 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1629 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1630 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1631 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1632 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */ 1633 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */ 1634 ); 1635 } 1636 1637 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1638 const struct macvlan_dev *vlan, 1639 const int i) 1640 { 1641 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1642 struct macvlan_source_entry *entry; 1643 1644 hlist_for_each_entry_rcu(entry, h, hlist) { 1645 if (entry->vlan != vlan) 1646 continue; 1647 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1648 return 1; 1649 } 1650 return 0; 1651 } 1652 1653 static int macvlan_fill_info(struct sk_buff *skb, 1654 const struct net_device *dev) 1655 { 1656 struct macvlan_dev *vlan = netdev_priv(dev); 1657 struct macvlan_port *port = vlan->port; 1658 int i; 1659 struct nlattr *nest; 1660 1661 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1662 goto nla_put_failure; 1663 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1664 goto nla_put_failure; 1665 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1666 goto nla_put_failure; 1667 if (vlan->macaddr_count > 0) { 1668 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1669 if (nest == NULL) 1670 goto nla_put_failure; 1671 1672 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1673 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1674 goto nla_put_failure; 1675 } 1676 nla_nest_end(skb, nest); 1677 } 1678 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req)) 1679 goto nla_put_failure; 1680 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used)) 1681 goto nla_put_failure; 1682 return 0; 1683 1684 nla_put_failure: 1685 return -EMSGSIZE; 1686 } 1687 1688 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1689 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1690 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1691 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1692 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1693 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1694 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1695 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 }, 1696 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT }, 1697 }; 1698 1699 int macvlan_link_register(struct rtnl_link_ops *ops) 1700 { 1701 /* common fields */ 1702 ops->validate = macvlan_validate; 1703 ops->maxtype = IFLA_MACVLAN_MAX; 1704 ops->policy = macvlan_policy; 1705 ops->changelink = macvlan_changelink; 1706 ops->get_size = macvlan_get_size; 1707 ops->fill_info = macvlan_fill_info; 1708 1709 return rtnl_link_register(ops); 1710 }; 1711 EXPORT_SYMBOL_GPL(macvlan_link_register); 1712 1713 static struct net *macvlan_get_link_net(const struct net_device *dev) 1714 { 1715 return dev_net(macvlan_dev_real_dev(dev)); 1716 } 1717 1718 static struct rtnl_link_ops macvlan_link_ops = { 1719 .kind = "macvlan", 1720 .setup = macvlan_setup, 1721 .newlink = macvlan_newlink, 1722 .dellink = macvlan_dellink, 1723 .get_link_net = macvlan_get_link_net, 1724 .priv_size = sizeof(struct macvlan_dev), 1725 }; 1726 1727 static void update_port_bc_queue_len(struct macvlan_port *port) 1728 { 1729 u32 max_bc_queue_len_req = 0; 1730 struct macvlan_dev *vlan; 1731 1732 list_for_each_entry(vlan, &port->vlans, list) { 1733 if (vlan->bc_queue_len_req > max_bc_queue_len_req) 1734 max_bc_queue_len_req = vlan->bc_queue_len_req; 1735 } 1736 port->bc_queue_len_used = max_bc_queue_len_req; 1737 } 1738 1739 static int macvlan_device_event(struct notifier_block *unused, 1740 unsigned long event, void *ptr) 1741 { 1742 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1743 struct macvlan_dev *vlan, *next; 1744 struct macvlan_port *port; 1745 LIST_HEAD(list_kill); 1746 1747 if (!netif_is_macvlan_port(dev)) 1748 return NOTIFY_DONE; 1749 1750 port = macvlan_port_get_rtnl(dev); 1751 1752 switch (event) { 1753 case NETDEV_UP: 1754 case NETDEV_DOWN: 1755 case NETDEV_CHANGE: 1756 list_for_each_entry(vlan, &port->vlans, list) 1757 netif_stacked_transfer_operstate(vlan->lowerdev, 1758 vlan->dev); 1759 break; 1760 case NETDEV_FEAT_CHANGE: 1761 list_for_each_entry(vlan, &port->vlans, list) { 1762 netif_set_gso_max_size(vlan->dev, dev->gso_max_size); 1763 netif_set_gso_max_segs(vlan->dev, dev->gso_max_segs); 1764 netdev_update_features(vlan->dev); 1765 } 1766 break; 1767 case NETDEV_CHANGEMTU: 1768 list_for_each_entry(vlan, &port->vlans, list) { 1769 if (vlan->dev->mtu <= dev->mtu) 1770 continue; 1771 dev_set_mtu(vlan->dev, dev->mtu); 1772 } 1773 break; 1774 case NETDEV_CHANGEADDR: 1775 if (!macvlan_passthru(port)) 1776 return NOTIFY_DONE; 1777 1778 vlan = list_first_entry_or_null(&port->vlans, 1779 struct macvlan_dev, 1780 list); 1781 1782 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1783 return NOTIFY_BAD; 1784 1785 break; 1786 case NETDEV_UNREGISTER: 1787 /* twiddle thumbs on netns device moves */ 1788 if (dev->reg_state != NETREG_UNREGISTERING) 1789 break; 1790 1791 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1792 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1793 unregister_netdevice_many(&list_kill); 1794 break; 1795 case NETDEV_PRE_TYPE_CHANGE: 1796 /* Forbid underlying device to change its type. */ 1797 return NOTIFY_BAD; 1798 1799 case NETDEV_NOTIFY_PEERS: 1800 case NETDEV_BONDING_FAILOVER: 1801 case NETDEV_RESEND_IGMP: 1802 /* Propagate to all vlans */ 1803 list_for_each_entry(vlan, &port->vlans, list) 1804 call_netdevice_notifiers(event, vlan->dev); 1805 } 1806 return NOTIFY_DONE; 1807 } 1808 1809 static struct notifier_block macvlan_notifier_block __read_mostly = { 1810 .notifier_call = macvlan_device_event, 1811 }; 1812 1813 static int __init macvlan_init_module(void) 1814 { 1815 int err; 1816 1817 register_netdevice_notifier(&macvlan_notifier_block); 1818 1819 err = macvlan_link_register(&macvlan_link_ops); 1820 if (err < 0) 1821 goto err1; 1822 return 0; 1823 err1: 1824 unregister_netdevice_notifier(&macvlan_notifier_block); 1825 return err; 1826 } 1827 1828 static void __exit macvlan_cleanup_module(void) 1829 { 1830 rtnl_link_unregister(&macvlan_link_ops); 1831 unregister_netdevice_notifier(&macvlan_notifier_block); 1832 } 1833 1834 module_init(macvlan_init_module); 1835 module_exit(macvlan_cleanup_module); 1836 1837 MODULE_LICENSE("GPL"); 1838 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1839 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1840 MODULE_ALIAS_RTNL_LINK("macvlan"); 1841