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_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 flags; 50 int count; 51 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 52 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 53 unsigned char perm_addr[ETH_ALEN]; 54 }; 55 56 struct macvlan_source_entry { 57 struct hlist_node hlist; 58 struct macvlan_dev *vlan; 59 unsigned char addr[6+2] __aligned(sizeof(u16)); 60 struct rcu_head rcu; 61 }; 62 63 struct macvlan_skb_cb { 64 const struct macvlan_dev *src; 65 }; 66 67 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 68 69 static void macvlan_port_destroy(struct net_device *dev); 70 71 static inline bool macvlan_passthru(const struct macvlan_port *port) 72 { 73 return port->flags & MACVLAN_F_PASSTHRU; 74 } 75 76 static inline void macvlan_set_passthru(struct macvlan_port *port) 77 { 78 port->flags |= MACVLAN_F_PASSTHRU; 79 } 80 81 static inline bool macvlan_addr_change(const struct macvlan_port *port) 82 { 83 return port->flags & MACVLAN_F_ADDRCHANGE; 84 } 85 86 static inline void macvlan_set_addr_change(struct macvlan_port *port) 87 { 88 port->flags |= MACVLAN_F_ADDRCHANGE; 89 } 90 91 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 92 { 93 port->flags &= ~MACVLAN_F_ADDRCHANGE; 94 } 95 96 /* Hash Ethernet address */ 97 static u32 macvlan_eth_hash(const unsigned char *addr) 98 { 99 u64 value = get_unaligned((u64 *)addr); 100 101 /* only want 6 bytes */ 102 #ifdef __BIG_ENDIAN 103 value >>= 16; 104 #else 105 value <<= 16; 106 #endif 107 return hash_64(value, MACVLAN_HASH_BITS); 108 } 109 110 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 111 { 112 return rcu_dereference(dev->rx_handler_data); 113 } 114 115 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 116 { 117 return rtnl_dereference(dev->rx_handler_data); 118 } 119 120 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 121 const unsigned char *addr) 122 { 123 struct macvlan_dev *vlan; 124 u32 idx = macvlan_eth_hash(addr); 125 126 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist, 127 lockdep_rtnl_is_held()) { 128 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 129 return vlan; 130 } 131 return NULL; 132 } 133 134 static struct macvlan_source_entry *macvlan_hash_lookup_source( 135 const struct macvlan_dev *vlan, 136 const unsigned char *addr) 137 { 138 struct macvlan_source_entry *entry; 139 u32 idx = macvlan_eth_hash(addr); 140 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 141 142 hlist_for_each_entry_rcu(entry, h, hlist) { 143 if (ether_addr_equal_64bits(entry->addr, addr) && 144 entry->vlan == vlan) 145 return entry; 146 } 147 return NULL; 148 } 149 150 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 151 const unsigned char *addr) 152 { 153 struct macvlan_port *port = vlan->port; 154 struct macvlan_source_entry *entry; 155 struct hlist_head *h; 156 157 entry = macvlan_hash_lookup_source(vlan, addr); 158 if (entry) 159 return 0; 160 161 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 162 if (!entry) 163 return -ENOMEM; 164 165 ether_addr_copy(entry->addr, addr); 166 entry->vlan = vlan; 167 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 168 hlist_add_head_rcu(&entry->hlist, h); 169 vlan->macaddr_count++; 170 171 return 0; 172 } 173 174 static void macvlan_hash_add(struct macvlan_dev *vlan) 175 { 176 struct macvlan_port *port = vlan->port; 177 const unsigned char *addr = vlan->dev->dev_addr; 178 u32 idx = macvlan_eth_hash(addr); 179 180 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 181 } 182 183 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 184 { 185 hlist_del_rcu(&entry->hlist); 186 kfree_rcu(entry, rcu); 187 } 188 189 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 190 { 191 hlist_del_rcu(&vlan->hlist); 192 if (sync) 193 synchronize_rcu(); 194 } 195 196 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 197 const unsigned char *addr) 198 { 199 macvlan_hash_del(vlan, true); 200 /* Now that we are unhashed it is safe to change the device 201 * address without confusing packet delivery. 202 */ 203 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 204 macvlan_hash_add(vlan); 205 } 206 207 static bool macvlan_addr_busy(const struct macvlan_port *port, 208 const unsigned char *addr) 209 { 210 /* Test to see if the specified address is 211 * currently in use by the underlying device or 212 * another macvlan. 213 */ 214 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 215 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 216 return true; 217 218 if (macvlan_hash_lookup(port, addr)) 219 return true; 220 221 return false; 222 } 223 224 225 static int macvlan_broadcast_one(struct sk_buff *skb, 226 const struct macvlan_dev *vlan, 227 const struct ethhdr *eth, bool local) 228 { 229 struct net_device *dev = vlan->dev; 230 231 if (local) 232 return __dev_forward_skb(dev, skb); 233 234 skb->dev = dev; 235 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 236 skb->pkt_type = PACKET_BROADCAST; 237 else 238 skb->pkt_type = PACKET_MULTICAST; 239 240 return 0; 241 } 242 243 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 244 { 245 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 246 } 247 248 249 static unsigned int mc_hash(const struct macvlan_dev *vlan, 250 const unsigned char *addr) 251 { 252 u32 val = __get_unaligned_cpu32(addr + 2); 253 254 val ^= macvlan_hash_mix(vlan); 255 return hash_32(val, MACVLAN_MC_FILTER_BITS); 256 } 257 258 static void macvlan_broadcast(struct sk_buff *skb, 259 const struct macvlan_port *port, 260 struct net_device *src, 261 enum macvlan_mode mode) 262 { 263 const struct ethhdr *eth = eth_hdr(skb); 264 const struct macvlan_dev *vlan; 265 struct sk_buff *nskb; 266 unsigned int i; 267 int err; 268 unsigned int hash; 269 270 if (skb->protocol == htons(ETH_P_PAUSE)) 271 return; 272 273 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 274 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 275 if (vlan->dev == src || !(vlan->mode & mode)) 276 continue; 277 278 hash = mc_hash(vlan, eth->h_dest); 279 if (!test_bit(hash, vlan->mc_filter)) 280 continue; 281 282 err = NET_RX_DROP; 283 nskb = skb_clone(skb, GFP_ATOMIC); 284 if (likely(nskb)) 285 err = macvlan_broadcast_one( 286 nskb, vlan, eth, 287 mode == MACVLAN_MODE_BRIDGE) ?: 288 netif_rx_ni(nskb); 289 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 290 err == NET_RX_SUCCESS, true); 291 } 292 } 293 } 294 295 static void macvlan_process_broadcast(struct work_struct *w) 296 { 297 struct macvlan_port *port = container_of(w, struct macvlan_port, 298 bc_work); 299 struct sk_buff *skb; 300 struct sk_buff_head list; 301 302 __skb_queue_head_init(&list); 303 304 spin_lock_bh(&port->bc_queue.lock); 305 skb_queue_splice_tail_init(&port->bc_queue, &list); 306 spin_unlock_bh(&port->bc_queue.lock); 307 308 while ((skb = __skb_dequeue(&list))) { 309 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 310 311 rcu_read_lock(); 312 313 if (!src) 314 /* frame comes from an external address */ 315 macvlan_broadcast(skb, port, NULL, 316 MACVLAN_MODE_PRIVATE | 317 MACVLAN_MODE_VEPA | 318 MACVLAN_MODE_PASSTHRU| 319 MACVLAN_MODE_BRIDGE); 320 else if (src->mode == MACVLAN_MODE_VEPA) 321 /* flood to everyone except source */ 322 macvlan_broadcast(skb, port, src->dev, 323 MACVLAN_MODE_VEPA | 324 MACVLAN_MODE_BRIDGE); 325 else 326 /* 327 * flood only to VEPA ports, bridge ports 328 * already saw the frame on the way out. 329 */ 330 macvlan_broadcast(skb, port, src->dev, 331 MACVLAN_MODE_VEPA); 332 333 rcu_read_unlock(); 334 335 if (src) 336 dev_put(src->dev); 337 consume_skb(skb); 338 339 cond_resched(); 340 } 341 } 342 343 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 344 const struct macvlan_dev *src, 345 struct sk_buff *skb) 346 { 347 struct sk_buff *nskb; 348 int err = -ENOMEM; 349 350 nskb = skb_clone(skb, GFP_ATOMIC); 351 if (!nskb) 352 goto err; 353 354 MACVLAN_SKB_CB(nskb)->src = src; 355 356 spin_lock(&port->bc_queue.lock); 357 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) { 358 if (src) 359 dev_hold(src->dev); 360 __skb_queue_tail(&port->bc_queue, nskb); 361 err = 0; 362 } 363 spin_unlock(&port->bc_queue.lock); 364 365 schedule_work(&port->bc_work); 366 367 if (err) 368 goto free_nskb; 369 370 return; 371 372 free_nskb: 373 kfree_skb(nskb); 374 err: 375 atomic_long_inc(&skb->dev->rx_dropped); 376 } 377 378 static void macvlan_flush_sources(struct macvlan_port *port, 379 struct macvlan_dev *vlan) 380 { 381 int i; 382 383 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 384 struct hlist_node *h, *n; 385 386 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) { 387 struct macvlan_source_entry *entry; 388 389 entry = hlist_entry(h, struct macvlan_source_entry, 390 hlist); 391 if (entry->vlan == vlan) 392 macvlan_hash_del_source(entry); 393 } 394 } 395 vlan->macaddr_count = 0; 396 } 397 398 static void macvlan_forward_source_one(struct sk_buff *skb, 399 struct macvlan_dev *vlan) 400 { 401 struct sk_buff *nskb; 402 struct net_device *dev; 403 int len; 404 int ret; 405 406 dev = vlan->dev; 407 if (unlikely(!(dev->flags & IFF_UP))) 408 return; 409 410 nskb = skb_clone(skb, GFP_ATOMIC); 411 if (!nskb) 412 return; 413 414 len = nskb->len + ETH_HLEN; 415 nskb->dev = dev; 416 417 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr)) 418 nskb->pkt_type = PACKET_HOST; 419 420 ret = netif_rx(nskb); 421 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 422 } 423 424 static void macvlan_forward_source(struct sk_buff *skb, 425 struct macvlan_port *port, 426 const unsigned char *addr) 427 { 428 struct macvlan_source_entry *entry; 429 u32 idx = macvlan_eth_hash(addr); 430 struct hlist_head *h = &port->vlan_source_hash[idx]; 431 432 hlist_for_each_entry_rcu(entry, h, hlist) { 433 if (ether_addr_equal_64bits(entry->addr, addr)) 434 macvlan_forward_source_one(skb, entry->vlan); 435 } 436 } 437 438 /* called under rcu_read_lock() from netif_receive_skb */ 439 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 440 { 441 struct macvlan_port *port; 442 struct sk_buff *skb = *pskb; 443 const struct ethhdr *eth = eth_hdr(skb); 444 const struct macvlan_dev *vlan; 445 const struct macvlan_dev *src; 446 struct net_device *dev; 447 unsigned int len = 0; 448 int ret; 449 rx_handler_result_t handle_res; 450 451 port = macvlan_port_get_rcu(skb->dev); 452 if (is_multicast_ether_addr(eth->h_dest)) { 453 unsigned int hash; 454 455 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 456 if (!skb) 457 return RX_HANDLER_CONSUMED; 458 *pskb = skb; 459 eth = eth_hdr(skb); 460 macvlan_forward_source(skb, port, eth->h_source); 461 src = macvlan_hash_lookup(port, eth->h_source); 462 if (src && src->mode != MACVLAN_MODE_VEPA && 463 src->mode != MACVLAN_MODE_BRIDGE) { 464 /* forward to original port. */ 465 vlan = src; 466 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 467 netif_rx(skb); 468 handle_res = RX_HANDLER_CONSUMED; 469 goto out; 470 } 471 472 hash = mc_hash(NULL, eth->h_dest); 473 if (test_bit(hash, port->mc_filter)) 474 macvlan_broadcast_enqueue(port, src, skb); 475 476 return RX_HANDLER_PASS; 477 } 478 479 macvlan_forward_source(skb, port, eth->h_source); 480 if (macvlan_passthru(port)) 481 vlan = list_first_or_null_rcu(&port->vlans, 482 struct macvlan_dev, list); 483 else 484 vlan = macvlan_hash_lookup(port, eth->h_dest); 485 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 486 return RX_HANDLER_PASS; 487 488 dev = vlan->dev; 489 if (unlikely(!(dev->flags & IFF_UP))) { 490 kfree_skb(skb); 491 return RX_HANDLER_CONSUMED; 492 } 493 len = skb->len + ETH_HLEN; 494 skb = skb_share_check(skb, GFP_ATOMIC); 495 if (!skb) { 496 ret = NET_RX_DROP; 497 handle_res = RX_HANDLER_CONSUMED; 498 goto out; 499 } 500 501 *pskb = skb; 502 skb->dev = dev; 503 skb->pkt_type = PACKET_HOST; 504 505 ret = NET_RX_SUCCESS; 506 handle_res = RX_HANDLER_ANOTHER; 507 out: 508 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 509 return handle_res; 510 } 511 512 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 513 { 514 const struct macvlan_dev *vlan = netdev_priv(dev); 515 const struct macvlan_port *port = vlan->port; 516 const struct macvlan_dev *dest; 517 518 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 519 const struct ethhdr *eth = skb_eth_hdr(skb); 520 521 /* send to other bridge ports directly */ 522 if (is_multicast_ether_addr(eth->h_dest)) { 523 skb_reset_mac_header(skb); 524 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 525 goto xmit_world; 526 } 527 528 dest = macvlan_hash_lookup(port, eth->h_dest); 529 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 530 /* send to lowerdev first for its network taps */ 531 dev_forward_skb(vlan->lowerdev, skb); 532 533 return NET_XMIT_SUCCESS; 534 } 535 } 536 xmit_world: 537 skb->dev = vlan->lowerdev; 538 return dev_queue_xmit_accel(skb, 539 netdev_get_sb_channel(dev) ? dev : NULL); 540 } 541 542 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 543 { 544 #ifdef CONFIG_NET_POLL_CONTROLLER 545 return netpoll_send_skb(vlan->netpoll, skb); 546 #else 547 BUG(); 548 return NETDEV_TX_OK; 549 #endif 550 } 551 552 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 553 struct net_device *dev) 554 { 555 struct macvlan_dev *vlan = netdev_priv(dev); 556 unsigned int len = skb->len; 557 int ret; 558 559 if (unlikely(netpoll_tx_running(dev))) 560 return macvlan_netpoll_send_skb(vlan, skb); 561 562 ret = macvlan_queue_xmit(skb, dev); 563 564 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 565 struct vlan_pcpu_stats *pcpu_stats; 566 567 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 568 u64_stats_update_begin(&pcpu_stats->syncp); 569 pcpu_stats->tx_packets++; 570 pcpu_stats->tx_bytes += len; 571 u64_stats_update_end(&pcpu_stats->syncp); 572 } else { 573 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 574 } 575 return ret; 576 } 577 578 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 579 unsigned short type, const void *daddr, 580 const void *saddr, unsigned len) 581 { 582 const struct macvlan_dev *vlan = netdev_priv(dev); 583 struct net_device *lowerdev = vlan->lowerdev; 584 585 return dev_hard_header(skb, lowerdev, type, daddr, 586 saddr ? : dev->dev_addr, len); 587 } 588 589 static const struct header_ops macvlan_hard_header_ops = { 590 .create = macvlan_hard_header, 591 .parse = eth_header_parse, 592 .cache = eth_header_cache, 593 .cache_update = eth_header_cache_update, 594 }; 595 596 static int macvlan_open(struct net_device *dev) 597 { 598 struct macvlan_dev *vlan = netdev_priv(dev); 599 struct net_device *lowerdev = vlan->lowerdev; 600 int err; 601 602 if (macvlan_passthru(vlan->port)) { 603 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 604 err = dev_set_promiscuity(lowerdev, 1); 605 if (err < 0) 606 goto out; 607 } 608 goto hash_add; 609 } 610 611 err = -EADDRINUSE; 612 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 613 goto out; 614 615 /* Attempt to populate accel_priv which is used to offload the L2 616 * forwarding requests for unicast packets. 617 */ 618 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 619 vlan->accel_priv = 620 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 621 622 /* If earlier attempt to offload failed, or accel_priv is not 623 * populated we must add the unicast address to the lower device. 624 */ 625 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 626 vlan->accel_priv = NULL; 627 err = dev_uc_add(lowerdev, dev->dev_addr); 628 if (err < 0) 629 goto out; 630 } 631 632 if (dev->flags & IFF_ALLMULTI) { 633 err = dev_set_allmulti(lowerdev, 1); 634 if (err < 0) 635 goto del_unicast; 636 } 637 638 if (dev->flags & IFF_PROMISC) { 639 err = dev_set_promiscuity(lowerdev, 1); 640 if (err < 0) 641 goto clear_multi; 642 } 643 644 hash_add: 645 macvlan_hash_add(vlan); 646 return 0; 647 648 clear_multi: 649 if (dev->flags & IFF_ALLMULTI) 650 dev_set_allmulti(lowerdev, -1); 651 del_unicast: 652 if (vlan->accel_priv) { 653 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 654 vlan->accel_priv); 655 vlan->accel_priv = NULL; 656 } else { 657 dev_uc_del(lowerdev, dev->dev_addr); 658 } 659 out: 660 return err; 661 } 662 663 static int macvlan_stop(struct net_device *dev) 664 { 665 struct macvlan_dev *vlan = netdev_priv(dev); 666 struct net_device *lowerdev = vlan->lowerdev; 667 668 if (vlan->accel_priv) { 669 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 670 vlan->accel_priv); 671 vlan->accel_priv = NULL; 672 } 673 674 dev_uc_unsync(lowerdev, dev); 675 dev_mc_unsync(lowerdev, dev); 676 677 if (macvlan_passthru(vlan->port)) { 678 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 679 dev_set_promiscuity(lowerdev, -1); 680 goto hash_del; 681 } 682 683 if (dev->flags & IFF_ALLMULTI) 684 dev_set_allmulti(lowerdev, -1); 685 686 if (dev->flags & IFF_PROMISC) 687 dev_set_promiscuity(lowerdev, -1); 688 689 dev_uc_del(lowerdev, dev->dev_addr); 690 691 hash_del: 692 macvlan_hash_del(vlan, !dev->dismantle); 693 return 0; 694 } 695 696 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 697 { 698 struct macvlan_dev *vlan = netdev_priv(dev); 699 struct net_device *lowerdev = vlan->lowerdev; 700 struct macvlan_port *port = vlan->port; 701 int err; 702 703 if (!(dev->flags & IFF_UP)) { 704 /* Just copy in the new address */ 705 ether_addr_copy(dev->dev_addr, addr); 706 } else { 707 /* Rehash and update the device filters */ 708 if (macvlan_addr_busy(vlan->port, addr)) 709 return -EADDRINUSE; 710 711 if (!macvlan_passthru(port)) { 712 err = dev_uc_add(lowerdev, addr); 713 if (err) 714 return err; 715 716 dev_uc_del(lowerdev, dev->dev_addr); 717 } 718 719 macvlan_hash_change_addr(vlan, addr); 720 } 721 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 722 /* Since addr_change isn't set, we are here due to lower 723 * device change. Save the lower-dev address so we can 724 * restore it later. 725 */ 726 ether_addr_copy(vlan->port->perm_addr, 727 lowerdev->dev_addr); 728 } 729 macvlan_clear_addr_change(port); 730 return 0; 731 } 732 733 static int macvlan_set_mac_address(struct net_device *dev, void *p) 734 { 735 struct macvlan_dev *vlan = netdev_priv(dev); 736 struct sockaddr *addr = p; 737 738 if (!is_valid_ether_addr(addr->sa_data)) 739 return -EADDRNOTAVAIL; 740 741 /* If the addresses are the same, this is a no-op */ 742 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 743 return 0; 744 745 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 746 macvlan_set_addr_change(vlan->port); 747 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 748 } 749 750 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 751 return -EADDRINUSE; 752 753 return macvlan_sync_address(dev, addr->sa_data); 754 } 755 756 static void macvlan_change_rx_flags(struct net_device *dev, int change) 757 { 758 struct macvlan_dev *vlan = netdev_priv(dev); 759 struct net_device *lowerdev = vlan->lowerdev; 760 761 if (dev->flags & IFF_UP) { 762 if (change & IFF_ALLMULTI) 763 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 764 if (change & IFF_PROMISC) 765 dev_set_promiscuity(lowerdev, 766 dev->flags & IFF_PROMISC ? 1 : -1); 767 768 } 769 } 770 771 static void macvlan_compute_filter(unsigned long *mc_filter, 772 struct net_device *dev, 773 struct macvlan_dev *vlan) 774 { 775 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 776 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 777 } else { 778 struct netdev_hw_addr *ha; 779 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 780 781 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 782 netdev_for_each_mc_addr(ha, dev) { 783 __set_bit(mc_hash(vlan, ha->addr), filter); 784 } 785 786 __set_bit(mc_hash(vlan, dev->broadcast), filter); 787 788 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 789 } 790 } 791 792 static void macvlan_set_mac_lists(struct net_device *dev) 793 { 794 struct macvlan_dev *vlan = netdev_priv(dev); 795 796 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 797 798 dev_uc_sync(vlan->lowerdev, dev); 799 dev_mc_sync(vlan->lowerdev, dev); 800 801 /* This is slightly inaccurate as we're including the subscription 802 * list of vlan->lowerdev too. 803 * 804 * Bug alert: This only works if everyone has the same broadcast 805 * address as lowerdev. As soon as someone changes theirs this 806 * will break. 807 * 808 * However, this is already broken as when you change your broadcast 809 * address we don't get called. 810 * 811 * The solution is to maintain a list of broadcast addresses like 812 * we do for uc/mc, if you care. 813 */ 814 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 815 } 816 817 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 818 { 819 struct macvlan_dev *vlan = netdev_priv(dev); 820 821 if (vlan->lowerdev->mtu < new_mtu) 822 return -EINVAL; 823 dev->mtu = new_mtu; 824 return 0; 825 } 826 827 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 828 { 829 struct net_device *real_dev = macvlan_dev_real_dev(dev); 830 const struct net_device_ops *ops = real_dev->netdev_ops; 831 struct ifreq ifrr; 832 int err = -EOPNOTSUPP; 833 834 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ); 835 ifrr.ifr_ifru = ifr->ifr_ifru; 836 837 switch (cmd) { 838 case SIOCSHWTSTAMP: 839 if (!net_eq(dev_net(dev), &init_net)) 840 break; 841 /* fall through */ 842 case SIOCGHWTSTAMP: 843 if (netif_device_present(real_dev) && ops->ndo_do_ioctl) 844 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd); 845 break; 846 } 847 848 if (!err) 849 ifr->ifr_ifru = ifrr.ifr_ifru; 850 851 return err; 852 } 853 854 /* 855 * macvlan network devices have devices nesting below it and are a special 856 * "super class" of normal network devices; split their locks off into a 857 * separate class since they always nest. 858 */ 859 #define ALWAYS_ON_OFFLOADS \ 860 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 861 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 862 863 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 864 865 #define MACVLAN_FEATURES \ 866 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 867 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 868 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 869 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 870 871 #define MACVLAN_STATE_MASK \ 872 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 873 874 static int macvlan_init(struct net_device *dev) 875 { 876 struct macvlan_dev *vlan = netdev_priv(dev); 877 const struct net_device *lowerdev = vlan->lowerdev; 878 struct macvlan_port *port = vlan->port; 879 880 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 881 (lowerdev->state & MACVLAN_STATE_MASK); 882 dev->features = lowerdev->features & MACVLAN_FEATURES; 883 dev->features |= ALWAYS_ON_FEATURES; 884 dev->hw_features |= NETIF_F_LRO; 885 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 886 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 887 dev->hw_enc_features |= dev->features; 888 dev->gso_max_size = lowerdev->gso_max_size; 889 dev->gso_max_segs = lowerdev->gso_max_segs; 890 dev->hard_header_len = lowerdev->hard_header_len; 891 892 netdev_lockdep_set_classes(dev); 893 894 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 895 if (!vlan->pcpu_stats) 896 return -ENOMEM; 897 898 port->count += 1; 899 900 return 0; 901 } 902 903 static void macvlan_uninit(struct net_device *dev) 904 { 905 struct macvlan_dev *vlan = netdev_priv(dev); 906 struct macvlan_port *port = vlan->port; 907 908 free_percpu(vlan->pcpu_stats); 909 910 macvlan_flush_sources(port, vlan); 911 port->count -= 1; 912 if (!port->count) 913 macvlan_port_destroy(port->dev); 914 } 915 916 static void macvlan_dev_get_stats64(struct net_device *dev, 917 struct rtnl_link_stats64 *stats) 918 { 919 struct macvlan_dev *vlan = netdev_priv(dev); 920 921 if (vlan->pcpu_stats) { 922 struct vlan_pcpu_stats *p; 923 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 924 u32 rx_errors = 0, tx_dropped = 0; 925 unsigned int start; 926 int i; 927 928 for_each_possible_cpu(i) { 929 p = per_cpu_ptr(vlan->pcpu_stats, i); 930 do { 931 start = u64_stats_fetch_begin_irq(&p->syncp); 932 rx_packets = p->rx_packets; 933 rx_bytes = p->rx_bytes; 934 rx_multicast = p->rx_multicast; 935 tx_packets = p->tx_packets; 936 tx_bytes = p->tx_bytes; 937 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 938 939 stats->rx_packets += rx_packets; 940 stats->rx_bytes += rx_bytes; 941 stats->multicast += rx_multicast; 942 stats->tx_packets += tx_packets; 943 stats->tx_bytes += tx_bytes; 944 /* rx_errors & tx_dropped are u32, updated 945 * without syncp protection. 946 */ 947 rx_errors += p->rx_errors; 948 tx_dropped += p->tx_dropped; 949 } 950 stats->rx_errors = rx_errors; 951 stats->rx_dropped = rx_errors; 952 stats->tx_dropped = tx_dropped; 953 } 954 } 955 956 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 957 __be16 proto, u16 vid) 958 { 959 struct macvlan_dev *vlan = netdev_priv(dev); 960 struct net_device *lowerdev = vlan->lowerdev; 961 962 return vlan_vid_add(lowerdev, proto, vid); 963 } 964 965 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 966 __be16 proto, u16 vid) 967 { 968 struct macvlan_dev *vlan = netdev_priv(dev); 969 struct net_device *lowerdev = vlan->lowerdev; 970 971 vlan_vid_del(lowerdev, proto, vid); 972 return 0; 973 } 974 975 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 976 struct net_device *dev, 977 const unsigned char *addr, u16 vid, 978 u16 flags, 979 struct netlink_ext_ack *extack) 980 { 981 struct macvlan_dev *vlan = netdev_priv(dev); 982 int err = -EINVAL; 983 984 /* Support unicast filter only on passthru devices. 985 * Multicast filter should be allowed on all devices. 986 */ 987 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 988 return -EOPNOTSUPP; 989 990 if (flags & NLM_F_REPLACE) 991 return -EOPNOTSUPP; 992 993 if (is_unicast_ether_addr(addr)) 994 err = dev_uc_add_excl(dev, addr); 995 else if (is_multicast_ether_addr(addr)) 996 err = dev_mc_add_excl(dev, addr); 997 998 return err; 999 } 1000 1001 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1002 struct net_device *dev, 1003 const unsigned char *addr, u16 vid) 1004 { 1005 struct macvlan_dev *vlan = netdev_priv(dev); 1006 int err = -EINVAL; 1007 1008 /* Support unicast filter only on passthru devices. 1009 * Multicast filter should be allowed on all devices. 1010 */ 1011 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1012 return -EOPNOTSUPP; 1013 1014 if (is_unicast_ether_addr(addr)) 1015 err = dev_uc_del(dev, addr); 1016 else if (is_multicast_ether_addr(addr)) 1017 err = dev_mc_del(dev, addr); 1018 1019 return err; 1020 } 1021 1022 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1023 struct ethtool_drvinfo *drvinfo) 1024 { 1025 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1026 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1027 } 1028 1029 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1030 struct ethtool_link_ksettings *cmd) 1031 { 1032 const struct macvlan_dev *vlan = netdev_priv(dev); 1033 1034 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1035 } 1036 1037 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1038 struct ethtool_ts_info *info) 1039 { 1040 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1041 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1042 struct phy_device *phydev = real_dev->phydev; 1043 1044 if (phy_has_tsinfo(phydev)) { 1045 return phy_ts_info(phydev, info); 1046 } else if (ops->get_ts_info) { 1047 return ops->get_ts_info(real_dev, info); 1048 } else { 1049 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1050 SOF_TIMESTAMPING_SOFTWARE; 1051 info->phc_index = -1; 1052 } 1053 1054 return 0; 1055 } 1056 1057 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1058 netdev_features_t features) 1059 { 1060 struct macvlan_dev *vlan = netdev_priv(dev); 1061 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1062 netdev_features_t mask; 1063 1064 features |= NETIF_F_ALL_FOR_ALL; 1065 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1066 mask = features; 1067 1068 lowerdev_features &= (features | ~NETIF_F_LRO); 1069 features = netdev_increment_features(lowerdev_features, features, mask); 1070 features |= ALWAYS_ON_FEATURES; 1071 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1072 1073 return features; 1074 } 1075 1076 #ifdef CONFIG_NET_POLL_CONTROLLER 1077 static void macvlan_dev_poll_controller(struct net_device *dev) 1078 { 1079 return; 1080 } 1081 1082 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1083 { 1084 struct macvlan_dev *vlan = netdev_priv(dev); 1085 struct net_device *real_dev = vlan->lowerdev; 1086 struct netpoll *netpoll; 1087 int err = 0; 1088 1089 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1090 err = -ENOMEM; 1091 if (!netpoll) 1092 goto out; 1093 1094 err = __netpoll_setup(netpoll, real_dev); 1095 if (err) { 1096 kfree(netpoll); 1097 goto out; 1098 } 1099 1100 vlan->netpoll = netpoll; 1101 1102 out: 1103 return err; 1104 } 1105 1106 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1107 { 1108 struct macvlan_dev *vlan = netdev_priv(dev); 1109 struct netpoll *netpoll = vlan->netpoll; 1110 1111 if (!netpoll) 1112 return; 1113 1114 vlan->netpoll = NULL; 1115 1116 __netpoll_free(netpoll); 1117 } 1118 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1119 1120 static int macvlan_dev_get_iflink(const struct net_device *dev) 1121 { 1122 struct macvlan_dev *vlan = netdev_priv(dev); 1123 1124 return vlan->lowerdev->ifindex; 1125 } 1126 1127 static const struct ethtool_ops macvlan_ethtool_ops = { 1128 .get_link = ethtool_op_get_link, 1129 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1130 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1131 .get_ts_info = macvlan_ethtool_get_ts_info, 1132 }; 1133 1134 static const struct net_device_ops macvlan_netdev_ops = { 1135 .ndo_init = macvlan_init, 1136 .ndo_uninit = macvlan_uninit, 1137 .ndo_open = macvlan_open, 1138 .ndo_stop = macvlan_stop, 1139 .ndo_start_xmit = macvlan_start_xmit, 1140 .ndo_change_mtu = macvlan_change_mtu, 1141 .ndo_do_ioctl = macvlan_do_ioctl, 1142 .ndo_fix_features = macvlan_fix_features, 1143 .ndo_change_rx_flags = macvlan_change_rx_flags, 1144 .ndo_set_mac_address = macvlan_set_mac_address, 1145 .ndo_set_rx_mode = macvlan_set_mac_lists, 1146 .ndo_get_stats64 = macvlan_dev_get_stats64, 1147 .ndo_validate_addr = eth_validate_addr, 1148 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1149 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1150 .ndo_fdb_add = macvlan_fdb_add, 1151 .ndo_fdb_del = macvlan_fdb_del, 1152 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1153 #ifdef CONFIG_NET_POLL_CONTROLLER 1154 .ndo_poll_controller = macvlan_dev_poll_controller, 1155 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1156 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1157 #endif 1158 .ndo_get_iflink = macvlan_dev_get_iflink, 1159 .ndo_features_check = passthru_features_check, 1160 .ndo_change_proto_down = dev_change_proto_down_generic, 1161 }; 1162 1163 void macvlan_common_setup(struct net_device *dev) 1164 { 1165 ether_setup(dev); 1166 1167 dev->min_mtu = 0; 1168 dev->max_mtu = ETH_MAX_MTU; 1169 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1170 netif_keep_dst(dev); 1171 dev->priv_flags |= IFF_UNICAST_FLT; 1172 dev->netdev_ops = &macvlan_netdev_ops; 1173 dev->needs_free_netdev = true; 1174 dev->header_ops = &macvlan_hard_header_ops; 1175 dev->ethtool_ops = &macvlan_ethtool_ops; 1176 } 1177 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1178 1179 static void macvlan_setup(struct net_device *dev) 1180 { 1181 macvlan_common_setup(dev); 1182 dev->priv_flags |= IFF_NO_QUEUE; 1183 } 1184 1185 static int macvlan_port_create(struct net_device *dev) 1186 { 1187 struct macvlan_port *port; 1188 unsigned int i; 1189 int err; 1190 1191 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1192 return -EINVAL; 1193 1194 if (netdev_is_rx_handler_busy(dev)) 1195 return -EBUSY; 1196 1197 port = kzalloc(sizeof(*port), GFP_KERNEL); 1198 if (port == NULL) 1199 return -ENOMEM; 1200 1201 port->dev = dev; 1202 ether_addr_copy(port->perm_addr, dev->dev_addr); 1203 INIT_LIST_HEAD(&port->vlans); 1204 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1205 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1206 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1207 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1208 1209 skb_queue_head_init(&port->bc_queue); 1210 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1211 1212 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1213 if (err) 1214 kfree(port); 1215 else 1216 dev->priv_flags |= IFF_MACVLAN_PORT; 1217 return err; 1218 } 1219 1220 static void macvlan_port_destroy(struct net_device *dev) 1221 { 1222 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1223 struct sk_buff *skb; 1224 1225 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1226 netdev_rx_handler_unregister(dev); 1227 1228 /* After this point, no packet can schedule bc_work anymore, 1229 * but we need to cancel it and purge left skbs if any. 1230 */ 1231 cancel_work_sync(&port->bc_work); 1232 1233 while ((skb = __skb_dequeue(&port->bc_queue))) { 1234 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1235 1236 if (src) 1237 dev_put(src->dev); 1238 1239 kfree_skb(skb); 1240 } 1241 1242 /* If the lower device address has been changed by passthru 1243 * macvlan, put it back. 1244 */ 1245 if (macvlan_passthru(port) && 1246 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1247 struct sockaddr sa; 1248 1249 sa.sa_family = port->dev->type; 1250 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1251 dev_set_mac_address(port->dev, &sa, NULL); 1252 } 1253 1254 kfree(port); 1255 } 1256 1257 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1258 struct netlink_ext_ack *extack) 1259 { 1260 if (tb[IFLA_ADDRESS]) { 1261 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1262 return -EINVAL; 1263 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1264 return -EADDRNOTAVAIL; 1265 } 1266 1267 if (!data) 1268 return 0; 1269 1270 if (data[IFLA_MACVLAN_FLAGS] && 1271 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1272 return -EINVAL; 1273 1274 if (data[IFLA_MACVLAN_MODE]) { 1275 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1276 case MACVLAN_MODE_PRIVATE: 1277 case MACVLAN_MODE_VEPA: 1278 case MACVLAN_MODE_BRIDGE: 1279 case MACVLAN_MODE_PASSTHRU: 1280 case MACVLAN_MODE_SOURCE: 1281 break; 1282 default: 1283 return -EINVAL; 1284 } 1285 } 1286 1287 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1288 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1289 case MACVLAN_MACADDR_ADD: 1290 case MACVLAN_MACADDR_DEL: 1291 case MACVLAN_MACADDR_FLUSH: 1292 case MACVLAN_MACADDR_SET: 1293 break; 1294 default: 1295 return -EINVAL; 1296 } 1297 } 1298 1299 if (data[IFLA_MACVLAN_MACADDR]) { 1300 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1301 return -EINVAL; 1302 1303 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1304 return -EADDRNOTAVAIL; 1305 } 1306 1307 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1308 return -EINVAL; 1309 1310 return 0; 1311 } 1312 1313 /** 1314 * reconfigure list of remote source mac address 1315 * (only for macvlan devices in source mode) 1316 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1317 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1318 */ 1319 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1320 struct nlattr *data[]) 1321 { 1322 char *addr = NULL; 1323 int ret, rem, len; 1324 struct nlattr *nla, *head; 1325 struct macvlan_source_entry *entry; 1326 1327 if (data[IFLA_MACVLAN_MACADDR]) 1328 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1329 1330 if (mode == MACVLAN_MACADDR_ADD) { 1331 if (!addr) 1332 return -EINVAL; 1333 1334 return macvlan_hash_add_source(vlan, addr); 1335 1336 } else if (mode == MACVLAN_MACADDR_DEL) { 1337 if (!addr) 1338 return -EINVAL; 1339 1340 entry = macvlan_hash_lookup_source(vlan, addr); 1341 if (entry) { 1342 macvlan_hash_del_source(entry); 1343 vlan->macaddr_count--; 1344 } 1345 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1346 macvlan_flush_sources(vlan->port, vlan); 1347 } else if (mode == MACVLAN_MACADDR_SET) { 1348 macvlan_flush_sources(vlan->port, vlan); 1349 1350 if (addr) { 1351 ret = macvlan_hash_add_source(vlan, addr); 1352 if (ret) 1353 return ret; 1354 } 1355 1356 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1357 return 0; 1358 1359 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1360 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1361 1362 nla_for_each_attr(nla, head, len, rem) { 1363 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1364 nla_len(nla) != ETH_ALEN) 1365 continue; 1366 1367 addr = nla_data(nla); 1368 ret = macvlan_hash_add_source(vlan, addr); 1369 if (ret) 1370 return ret; 1371 } 1372 } else { 1373 return -EINVAL; 1374 } 1375 1376 return 0; 1377 } 1378 1379 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1380 struct nlattr *tb[], struct nlattr *data[], 1381 struct netlink_ext_ack *extack) 1382 { 1383 struct macvlan_dev *vlan = netdev_priv(dev); 1384 struct macvlan_port *port; 1385 struct net_device *lowerdev; 1386 int err; 1387 int macmode; 1388 bool create = false; 1389 1390 if (!tb[IFLA_LINK]) 1391 return -EINVAL; 1392 1393 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1394 if (lowerdev == NULL) 1395 return -ENODEV; 1396 1397 /* When creating macvlans or macvtaps on top of other macvlans - use 1398 * the real device as the lowerdev. 1399 */ 1400 if (netif_is_macvlan(lowerdev)) 1401 lowerdev = macvlan_dev_real_dev(lowerdev); 1402 1403 if (!tb[IFLA_MTU]) 1404 dev->mtu = lowerdev->mtu; 1405 else if (dev->mtu > lowerdev->mtu) 1406 return -EINVAL; 1407 1408 /* MTU range: 68 - lowerdev->max_mtu */ 1409 dev->min_mtu = ETH_MIN_MTU; 1410 dev->max_mtu = lowerdev->max_mtu; 1411 1412 if (!tb[IFLA_ADDRESS]) 1413 eth_hw_addr_random(dev); 1414 1415 if (!netif_is_macvlan_port(lowerdev)) { 1416 err = macvlan_port_create(lowerdev); 1417 if (err < 0) 1418 return err; 1419 create = true; 1420 } 1421 port = macvlan_port_get_rtnl(lowerdev); 1422 1423 /* Only 1 macvlan device can be created in passthru mode */ 1424 if (macvlan_passthru(port)) { 1425 /* The macvlan port must be not created this time, 1426 * still goto destroy_macvlan_port for readability. 1427 */ 1428 err = -EINVAL; 1429 goto destroy_macvlan_port; 1430 } 1431 1432 vlan->lowerdev = lowerdev; 1433 vlan->dev = dev; 1434 vlan->port = port; 1435 vlan->set_features = MACVLAN_FEATURES; 1436 1437 vlan->mode = MACVLAN_MODE_VEPA; 1438 if (data && data[IFLA_MACVLAN_MODE]) 1439 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1440 1441 if (data && data[IFLA_MACVLAN_FLAGS]) 1442 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1443 1444 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1445 if (port->count) { 1446 err = -EINVAL; 1447 goto destroy_macvlan_port; 1448 } 1449 macvlan_set_passthru(port); 1450 eth_hw_addr_inherit(dev, lowerdev); 1451 } 1452 1453 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1454 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1455 err = -EINVAL; 1456 goto destroy_macvlan_port; 1457 } 1458 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1459 err = macvlan_changelink_sources(vlan, macmode, data); 1460 if (err) 1461 goto destroy_macvlan_port; 1462 } 1463 1464 err = register_netdevice(dev); 1465 if (err < 0) 1466 goto destroy_macvlan_port; 1467 1468 dev->priv_flags |= IFF_MACVLAN; 1469 err = netdev_upper_dev_link(lowerdev, dev, extack); 1470 if (err) 1471 goto unregister_netdev; 1472 1473 list_add_tail_rcu(&vlan->list, &port->vlans); 1474 netif_stacked_transfer_operstate(lowerdev, dev); 1475 linkwatch_fire_event(dev); 1476 1477 return 0; 1478 1479 unregister_netdev: 1480 /* macvlan_uninit would free the macvlan port */ 1481 unregister_netdevice(dev); 1482 return err; 1483 destroy_macvlan_port: 1484 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1485 * so we destroy the macvlan port only when it's valid. 1486 */ 1487 if (create && macvlan_port_get_rtnl(lowerdev)) 1488 macvlan_port_destroy(port->dev); 1489 return err; 1490 } 1491 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1492 1493 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1494 struct nlattr *tb[], struct nlattr *data[], 1495 struct netlink_ext_ack *extack) 1496 { 1497 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1498 } 1499 1500 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1501 { 1502 struct macvlan_dev *vlan = netdev_priv(dev); 1503 1504 if (vlan->mode == MACVLAN_MODE_SOURCE) 1505 macvlan_flush_sources(vlan->port, vlan); 1506 list_del_rcu(&vlan->list); 1507 unregister_netdevice_queue(dev, head); 1508 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1509 } 1510 EXPORT_SYMBOL_GPL(macvlan_dellink); 1511 1512 static int macvlan_changelink(struct net_device *dev, 1513 struct nlattr *tb[], struct nlattr *data[], 1514 struct netlink_ext_ack *extack) 1515 { 1516 struct macvlan_dev *vlan = netdev_priv(dev); 1517 enum macvlan_mode mode; 1518 bool set_mode = false; 1519 enum macvlan_macaddr_mode macmode; 1520 int ret; 1521 1522 /* Validate mode, but don't set yet: setting flags may fail. */ 1523 if (data && data[IFLA_MACVLAN_MODE]) { 1524 set_mode = true; 1525 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1526 /* Passthrough mode can't be set or cleared dynamically */ 1527 if ((mode == MACVLAN_MODE_PASSTHRU) != 1528 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1529 return -EINVAL; 1530 if (vlan->mode == MACVLAN_MODE_SOURCE && 1531 vlan->mode != mode) 1532 macvlan_flush_sources(vlan->port, vlan); 1533 } 1534 1535 if (data && data[IFLA_MACVLAN_FLAGS]) { 1536 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1537 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1538 if (macvlan_passthru(vlan->port) && promisc) { 1539 int err; 1540 1541 if (flags & MACVLAN_FLAG_NOPROMISC) 1542 err = dev_set_promiscuity(vlan->lowerdev, -1); 1543 else 1544 err = dev_set_promiscuity(vlan->lowerdev, 1); 1545 if (err < 0) 1546 return err; 1547 } 1548 vlan->flags = flags; 1549 } 1550 if (set_mode) 1551 vlan->mode = mode; 1552 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1553 if (vlan->mode != MACVLAN_MODE_SOURCE) 1554 return -EINVAL; 1555 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1556 ret = macvlan_changelink_sources(vlan, macmode, data); 1557 if (ret) 1558 return ret; 1559 } 1560 return 0; 1561 } 1562 1563 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1564 { 1565 if (vlan->macaddr_count == 0) 1566 return 0; 1567 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1568 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1569 } 1570 1571 static size_t macvlan_get_size(const struct net_device *dev) 1572 { 1573 struct macvlan_dev *vlan = netdev_priv(dev); 1574 1575 return (0 1576 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1577 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1578 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1579 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1580 ); 1581 } 1582 1583 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1584 const struct macvlan_dev *vlan, 1585 const int i) 1586 { 1587 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1588 struct macvlan_source_entry *entry; 1589 1590 hlist_for_each_entry_rcu(entry, h, hlist) { 1591 if (entry->vlan != vlan) 1592 continue; 1593 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1594 return 1; 1595 } 1596 return 0; 1597 } 1598 1599 static int macvlan_fill_info(struct sk_buff *skb, 1600 const struct net_device *dev) 1601 { 1602 struct macvlan_dev *vlan = netdev_priv(dev); 1603 int i; 1604 struct nlattr *nest; 1605 1606 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1607 goto nla_put_failure; 1608 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1609 goto nla_put_failure; 1610 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1611 goto nla_put_failure; 1612 if (vlan->macaddr_count > 0) { 1613 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1614 if (nest == NULL) 1615 goto nla_put_failure; 1616 1617 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1618 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1619 goto nla_put_failure; 1620 } 1621 nla_nest_end(skb, nest); 1622 } 1623 return 0; 1624 1625 nla_put_failure: 1626 return -EMSGSIZE; 1627 } 1628 1629 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1630 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1631 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1632 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1633 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1634 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1635 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1636 }; 1637 1638 int macvlan_link_register(struct rtnl_link_ops *ops) 1639 { 1640 /* common fields */ 1641 ops->validate = macvlan_validate; 1642 ops->maxtype = IFLA_MACVLAN_MAX; 1643 ops->policy = macvlan_policy; 1644 ops->changelink = macvlan_changelink; 1645 ops->get_size = macvlan_get_size; 1646 ops->fill_info = macvlan_fill_info; 1647 1648 return rtnl_link_register(ops); 1649 }; 1650 EXPORT_SYMBOL_GPL(macvlan_link_register); 1651 1652 static struct net *macvlan_get_link_net(const struct net_device *dev) 1653 { 1654 return dev_net(macvlan_dev_real_dev(dev)); 1655 } 1656 1657 static struct rtnl_link_ops macvlan_link_ops = { 1658 .kind = "macvlan", 1659 .setup = macvlan_setup, 1660 .newlink = macvlan_newlink, 1661 .dellink = macvlan_dellink, 1662 .get_link_net = macvlan_get_link_net, 1663 .priv_size = sizeof(struct macvlan_dev), 1664 }; 1665 1666 static int macvlan_device_event(struct notifier_block *unused, 1667 unsigned long event, void *ptr) 1668 { 1669 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1670 struct macvlan_dev *vlan, *next; 1671 struct macvlan_port *port; 1672 LIST_HEAD(list_kill); 1673 1674 if (!netif_is_macvlan_port(dev)) 1675 return NOTIFY_DONE; 1676 1677 port = macvlan_port_get_rtnl(dev); 1678 1679 switch (event) { 1680 case NETDEV_UP: 1681 case NETDEV_DOWN: 1682 case NETDEV_CHANGE: 1683 list_for_each_entry(vlan, &port->vlans, list) 1684 netif_stacked_transfer_operstate(vlan->lowerdev, 1685 vlan->dev); 1686 break; 1687 case NETDEV_FEAT_CHANGE: 1688 list_for_each_entry(vlan, &port->vlans, list) { 1689 vlan->dev->gso_max_size = dev->gso_max_size; 1690 vlan->dev->gso_max_segs = dev->gso_max_segs; 1691 netdev_update_features(vlan->dev); 1692 } 1693 break; 1694 case NETDEV_CHANGEMTU: 1695 list_for_each_entry(vlan, &port->vlans, list) { 1696 if (vlan->dev->mtu <= dev->mtu) 1697 continue; 1698 dev_set_mtu(vlan->dev, dev->mtu); 1699 } 1700 break; 1701 case NETDEV_CHANGEADDR: 1702 if (!macvlan_passthru(port)) 1703 return NOTIFY_DONE; 1704 1705 vlan = list_first_entry_or_null(&port->vlans, 1706 struct macvlan_dev, 1707 list); 1708 1709 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1710 return NOTIFY_BAD; 1711 1712 break; 1713 case NETDEV_UNREGISTER: 1714 /* twiddle thumbs on netns device moves */ 1715 if (dev->reg_state != NETREG_UNREGISTERING) 1716 break; 1717 1718 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1719 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1720 unregister_netdevice_many(&list_kill); 1721 break; 1722 case NETDEV_PRE_TYPE_CHANGE: 1723 /* Forbid underlaying device to change its type. */ 1724 return NOTIFY_BAD; 1725 1726 case NETDEV_NOTIFY_PEERS: 1727 case NETDEV_BONDING_FAILOVER: 1728 case NETDEV_RESEND_IGMP: 1729 /* Propagate to all vlans */ 1730 list_for_each_entry(vlan, &port->vlans, list) 1731 call_netdevice_notifiers(event, vlan->dev); 1732 } 1733 return NOTIFY_DONE; 1734 } 1735 1736 static struct notifier_block macvlan_notifier_block __read_mostly = { 1737 .notifier_call = macvlan_device_event, 1738 }; 1739 1740 static int __init macvlan_init_module(void) 1741 { 1742 int err; 1743 1744 register_netdevice_notifier(&macvlan_notifier_block); 1745 1746 err = macvlan_link_register(&macvlan_link_ops); 1747 if (err < 0) 1748 goto err1; 1749 return 0; 1750 err1: 1751 unregister_netdevice_notifier(&macvlan_notifier_block); 1752 return err; 1753 } 1754 1755 static void __exit macvlan_cleanup_module(void) 1756 { 1757 rtnl_link_unregister(&macvlan_link_ops); 1758 unregister_netdevice_notifier(&macvlan_notifier_block); 1759 } 1760 1761 module_init(macvlan_init_module); 1762 module_exit(macvlan_cleanup_module); 1763 1764 MODULE_LICENSE("GPL"); 1765 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1766 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1767 MODULE_ALIAS_RTNL_LINK("macvlan"); 1768