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 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 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 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 276 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 277 if (vlan->dev == src || !(vlan->mode & mode)) 278 continue; 279 280 hash = mc_hash(vlan, eth->h_dest); 281 if (!test_bit(hash, vlan->mc_filter)) 282 continue; 283 284 err = NET_RX_DROP; 285 nskb = skb_clone(skb, GFP_ATOMIC); 286 if (likely(nskb)) 287 err = macvlan_broadcast_one( 288 nskb, vlan, eth, 289 mode == MACVLAN_MODE_BRIDGE) ?: 290 netif_rx_ni(nskb); 291 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 292 err == NET_RX_SUCCESS, true); 293 } 294 } 295 } 296 297 static void macvlan_process_broadcast(struct work_struct *w) 298 { 299 struct macvlan_port *port = container_of(w, struct macvlan_port, 300 bc_work); 301 struct sk_buff *skb; 302 struct sk_buff_head list; 303 304 __skb_queue_head_init(&list); 305 306 spin_lock_bh(&port->bc_queue.lock); 307 skb_queue_splice_tail_init(&port->bc_queue, &list); 308 spin_unlock_bh(&port->bc_queue.lock); 309 310 while ((skb = __skb_dequeue(&list))) { 311 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 312 313 rcu_read_lock(); 314 315 if (!src) 316 /* frame comes from an external address */ 317 macvlan_broadcast(skb, port, NULL, 318 MACVLAN_MODE_PRIVATE | 319 MACVLAN_MODE_VEPA | 320 MACVLAN_MODE_PASSTHRU| 321 MACVLAN_MODE_BRIDGE); 322 else if (src->mode == MACVLAN_MODE_VEPA) 323 /* flood to everyone except source */ 324 macvlan_broadcast(skb, port, src->dev, 325 MACVLAN_MODE_VEPA | 326 MACVLAN_MODE_BRIDGE); 327 else 328 /* 329 * flood only to VEPA ports, bridge ports 330 * already saw the frame on the way out. 331 */ 332 macvlan_broadcast(skb, port, src->dev, 333 MACVLAN_MODE_VEPA); 334 335 rcu_read_unlock(); 336 337 if (src) 338 dev_put(src->dev); 339 consume_skb(skb); 340 341 cond_resched(); 342 } 343 } 344 345 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 346 const struct macvlan_dev *src, 347 struct sk_buff *skb) 348 { 349 struct sk_buff *nskb; 350 int err = -ENOMEM; 351 352 nskb = skb_clone(skb, GFP_ATOMIC); 353 if (!nskb) 354 goto err; 355 356 MACVLAN_SKB_CB(nskb)->src = src; 357 358 spin_lock(&port->bc_queue.lock); 359 if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) { 360 if (src) 361 dev_hold(src->dev); 362 __skb_queue_tail(&port->bc_queue, nskb); 363 err = 0; 364 } 365 spin_unlock(&port->bc_queue.lock); 366 367 schedule_work(&port->bc_work); 368 369 if (err) 370 goto free_nskb; 371 372 return; 373 374 free_nskb: 375 kfree_skb(nskb); 376 err: 377 atomic_long_inc(&skb->dev->rx_dropped); 378 } 379 380 static void macvlan_flush_sources(struct macvlan_port *port, 381 struct macvlan_dev *vlan) 382 { 383 int i; 384 385 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 386 struct hlist_node *h, *n; 387 388 hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) { 389 struct macvlan_source_entry *entry; 390 391 entry = hlist_entry(h, struct macvlan_source_entry, 392 hlist); 393 if (entry->vlan == vlan) 394 macvlan_hash_del_source(entry); 395 } 396 } 397 vlan->macaddr_count = 0; 398 } 399 400 static void macvlan_forward_source_one(struct sk_buff *skb, 401 struct macvlan_dev *vlan) 402 { 403 struct sk_buff *nskb; 404 struct net_device *dev; 405 int len; 406 int ret; 407 408 dev = vlan->dev; 409 if (unlikely(!(dev->flags & IFF_UP))) 410 return; 411 412 nskb = skb_clone(skb, GFP_ATOMIC); 413 if (!nskb) 414 return; 415 416 len = nskb->len + ETH_HLEN; 417 nskb->dev = dev; 418 419 if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr)) 420 nskb->pkt_type = PACKET_HOST; 421 422 ret = netif_rx(nskb); 423 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 424 } 425 426 static void macvlan_forward_source(struct sk_buff *skb, 427 struct macvlan_port *port, 428 const unsigned char *addr) 429 { 430 struct macvlan_source_entry *entry; 431 u32 idx = macvlan_eth_hash(addr); 432 struct hlist_head *h = &port->vlan_source_hash[idx]; 433 434 hlist_for_each_entry_rcu(entry, h, hlist) { 435 if (ether_addr_equal_64bits(entry->addr, addr)) 436 macvlan_forward_source_one(skb, entry->vlan); 437 } 438 } 439 440 /* called under rcu_read_lock() from netif_receive_skb */ 441 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb) 442 { 443 struct macvlan_port *port; 444 struct sk_buff *skb = *pskb; 445 const struct ethhdr *eth = eth_hdr(skb); 446 const struct macvlan_dev *vlan; 447 const struct macvlan_dev *src; 448 struct net_device *dev; 449 unsigned int len = 0; 450 int ret; 451 rx_handler_result_t handle_res; 452 453 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */ 454 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 455 return RX_HANDLER_PASS; 456 457 port = macvlan_port_get_rcu(skb->dev); 458 if (is_multicast_ether_addr(eth->h_dest)) { 459 unsigned int hash; 460 461 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 462 if (!skb) 463 return RX_HANDLER_CONSUMED; 464 *pskb = skb; 465 eth = eth_hdr(skb); 466 macvlan_forward_source(skb, port, eth->h_source); 467 src = macvlan_hash_lookup(port, eth->h_source); 468 if (src && src->mode != MACVLAN_MODE_VEPA && 469 src->mode != MACVLAN_MODE_BRIDGE) { 470 /* forward to original port. */ 471 vlan = src; 472 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 473 netif_rx(skb); 474 handle_res = RX_HANDLER_CONSUMED; 475 goto out; 476 } 477 478 hash = mc_hash(NULL, eth->h_dest); 479 if (test_bit(hash, port->mc_filter)) 480 macvlan_broadcast_enqueue(port, src, skb); 481 482 return RX_HANDLER_PASS; 483 } 484 485 macvlan_forward_source(skb, port, eth->h_source); 486 if (macvlan_passthru(port)) 487 vlan = list_first_or_null_rcu(&port->vlans, 488 struct macvlan_dev, list); 489 else 490 vlan = macvlan_hash_lookup(port, eth->h_dest); 491 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 492 return RX_HANDLER_PASS; 493 494 dev = vlan->dev; 495 if (unlikely(!(dev->flags & IFF_UP))) { 496 kfree_skb(skb); 497 return RX_HANDLER_CONSUMED; 498 } 499 len = skb->len + ETH_HLEN; 500 skb = skb_share_check(skb, GFP_ATOMIC); 501 if (!skb) { 502 ret = NET_RX_DROP; 503 handle_res = RX_HANDLER_CONSUMED; 504 goto out; 505 } 506 507 *pskb = skb; 508 skb->dev = dev; 509 skb->pkt_type = PACKET_HOST; 510 511 ret = NET_RX_SUCCESS; 512 handle_res = RX_HANDLER_ANOTHER; 513 out: 514 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 515 return handle_res; 516 } 517 518 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 519 { 520 const struct macvlan_dev *vlan = netdev_priv(dev); 521 const struct macvlan_port *port = vlan->port; 522 const struct macvlan_dev *dest; 523 524 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 525 const struct ethhdr *eth = skb_eth_hdr(skb); 526 527 /* send to other bridge ports directly */ 528 if (is_multicast_ether_addr(eth->h_dest)) { 529 skb_reset_mac_header(skb); 530 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 531 goto xmit_world; 532 } 533 534 dest = macvlan_hash_lookup(port, eth->h_dest); 535 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 536 /* send to lowerdev first for its network taps */ 537 dev_forward_skb(vlan->lowerdev, skb); 538 539 return NET_XMIT_SUCCESS; 540 } 541 } 542 xmit_world: 543 skb->dev = vlan->lowerdev; 544 return dev_queue_xmit_accel(skb, 545 netdev_get_sb_channel(dev) ? dev : NULL); 546 } 547 548 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 549 { 550 #ifdef CONFIG_NET_POLL_CONTROLLER 551 return netpoll_send_skb(vlan->netpoll, skb); 552 #else 553 BUG(); 554 return NETDEV_TX_OK; 555 #endif 556 } 557 558 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 559 struct net_device *dev) 560 { 561 struct macvlan_dev *vlan = netdev_priv(dev); 562 unsigned int len = skb->len; 563 int ret; 564 565 if (unlikely(netpoll_tx_running(dev))) 566 return macvlan_netpoll_send_skb(vlan, skb); 567 568 ret = macvlan_queue_xmit(skb, dev); 569 570 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 571 struct vlan_pcpu_stats *pcpu_stats; 572 573 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 574 u64_stats_update_begin(&pcpu_stats->syncp); 575 pcpu_stats->tx_packets++; 576 pcpu_stats->tx_bytes += len; 577 u64_stats_update_end(&pcpu_stats->syncp); 578 } else { 579 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 580 } 581 return ret; 582 } 583 584 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 585 unsigned short type, const void *daddr, 586 const void *saddr, unsigned len) 587 { 588 const struct macvlan_dev *vlan = netdev_priv(dev); 589 struct net_device *lowerdev = vlan->lowerdev; 590 591 return dev_hard_header(skb, lowerdev, type, daddr, 592 saddr ? : dev->dev_addr, len); 593 } 594 595 static const struct header_ops macvlan_hard_header_ops = { 596 .create = macvlan_hard_header, 597 .parse = eth_header_parse, 598 .cache = eth_header_cache, 599 .cache_update = eth_header_cache_update, 600 }; 601 602 static int macvlan_open(struct net_device *dev) 603 { 604 struct macvlan_dev *vlan = netdev_priv(dev); 605 struct net_device *lowerdev = vlan->lowerdev; 606 int err; 607 608 if (macvlan_passthru(vlan->port)) { 609 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 610 err = dev_set_promiscuity(lowerdev, 1); 611 if (err < 0) 612 goto out; 613 } 614 goto hash_add; 615 } 616 617 err = -EADDRINUSE; 618 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 619 goto out; 620 621 /* Attempt to populate accel_priv which is used to offload the L2 622 * forwarding requests for unicast packets. 623 */ 624 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 625 vlan->accel_priv = 626 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 627 628 /* If earlier attempt to offload failed, or accel_priv is not 629 * populated we must add the unicast address to the lower device. 630 */ 631 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 632 vlan->accel_priv = NULL; 633 err = dev_uc_add(lowerdev, dev->dev_addr); 634 if (err < 0) 635 goto out; 636 } 637 638 if (dev->flags & IFF_ALLMULTI) { 639 err = dev_set_allmulti(lowerdev, 1); 640 if (err < 0) 641 goto del_unicast; 642 } 643 644 if (dev->flags & IFF_PROMISC) { 645 err = dev_set_promiscuity(lowerdev, 1); 646 if (err < 0) 647 goto clear_multi; 648 } 649 650 hash_add: 651 macvlan_hash_add(vlan); 652 return 0; 653 654 clear_multi: 655 if (dev->flags & IFF_ALLMULTI) 656 dev_set_allmulti(lowerdev, -1); 657 del_unicast: 658 if (vlan->accel_priv) { 659 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 660 vlan->accel_priv); 661 vlan->accel_priv = NULL; 662 } else { 663 dev_uc_del(lowerdev, dev->dev_addr); 664 } 665 out: 666 return err; 667 } 668 669 static int macvlan_stop(struct net_device *dev) 670 { 671 struct macvlan_dev *vlan = netdev_priv(dev); 672 struct net_device *lowerdev = vlan->lowerdev; 673 674 if (vlan->accel_priv) { 675 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 676 vlan->accel_priv); 677 vlan->accel_priv = NULL; 678 } 679 680 dev_uc_unsync(lowerdev, dev); 681 dev_mc_unsync(lowerdev, dev); 682 683 if (macvlan_passthru(vlan->port)) { 684 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 685 dev_set_promiscuity(lowerdev, -1); 686 goto hash_del; 687 } 688 689 if (dev->flags & IFF_ALLMULTI) 690 dev_set_allmulti(lowerdev, -1); 691 692 if (dev->flags & IFF_PROMISC) 693 dev_set_promiscuity(lowerdev, -1); 694 695 dev_uc_del(lowerdev, dev->dev_addr); 696 697 hash_del: 698 macvlan_hash_del(vlan, !dev->dismantle); 699 return 0; 700 } 701 702 static int macvlan_sync_address(struct net_device *dev, 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 ether_addr_copy(dev->dev_addr, 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_do_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_do_ioctl) 850 err = ops->ndo_do_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 const 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 dev->gso_max_size = lowerdev->gso_max_size; 904 dev->gso_max_segs = 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 return 0; 915 } 916 917 static void macvlan_uninit(struct net_device *dev) 918 { 919 struct macvlan_dev *vlan = netdev_priv(dev); 920 struct macvlan_port *port = vlan->port; 921 922 free_percpu(vlan->pcpu_stats); 923 924 macvlan_flush_sources(port, vlan); 925 port->count -= 1; 926 if (!port->count) 927 macvlan_port_destroy(port->dev); 928 } 929 930 static void macvlan_dev_get_stats64(struct net_device *dev, 931 struct rtnl_link_stats64 *stats) 932 { 933 struct macvlan_dev *vlan = netdev_priv(dev); 934 935 if (vlan->pcpu_stats) { 936 struct vlan_pcpu_stats *p; 937 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 938 u32 rx_errors = 0, tx_dropped = 0; 939 unsigned int start; 940 int i; 941 942 for_each_possible_cpu(i) { 943 p = per_cpu_ptr(vlan->pcpu_stats, i); 944 do { 945 start = u64_stats_fetch_begin_irq(&p->syncp); 946 rx_packets = p->rx_packets; 947 rx_bytes = p->rx_bytes; 948 rx_multicast = p->rx_multicast; 949 tx_packets = p->tx_packets; 950 tx_bytes = p->tx_bytes; 951 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 952 953 stats->rx_packets += rx_packets; 954 stats->rx_bytes += rx_bytes; 955 stats->multicast += rx_multicast; 956 stats->tx_packets += tx_packets; 957 stats->tx_bytes += tx_bytes; 958 /* rx_errors & tx_dropped are u32, updated 959 * without syncp protection. 960 */ 961 rx_errors += p->rx_errors; 962 tx_dropped += p->tx_dropped; 963 } 964 stats->rx_errors = rx_errors; 965 stats->rx_dropped = rx_errors; 966 stats->tx_dropped = tx_dropped; 967 } 968 } 969 970 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 971 __be16 proto, u16 vid) 972 { 973 struct macvlan_dev *vlan = netdev_priv(dev); 974 struct net_device *lowerdev = vlan->lowerdev; 975 976 return vlan_vid_add(lowerdev, proto, vid); 977 } 978 979 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 980 __be16 proto, u16 vid) 981 { 982 struct macvlan_dev *vlan = netdev_priv(dev); 983 struct net_device *lowerdev = vlan->lowerdev; 984 985 vlan_vid_del(lowerdev, proto, vid); 986 return 0; 987 } 988 989 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 990 struct net_device *dev, 991 const unsigned char *addr, u16 vid, 992 u16 flags, 993 struct netlink_ext_ack *extack) 994 { 995 struct macvlan_dev *vlan = netdev_priv(dev); 996 int err = -EINVAL; 997 998 /* Support unicast filter only on passthru devices. 999 * Multicast filter should be allowed on all devices. 1000 */ 1001 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1002 return -EOPNOTSUPP; 1003 1004 if (flags & NLM_F_REPLACE) 1005 return -EOPNOTSUPP; 1006 1007 if (is_unicast_ether_addr(addr)) 1008 err = dev_uc_add_excl(dev, addr); 1009 else if (is_multicast_ether_addr(addr)) 1010 err = dev_mc_add_excl(dev, addr); 1011 1012 return err; 1013 } 1014 1015 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1016 struct net_device *dev, 1017 const unsigned char *addr, u16 vid) 1018 { 1019 struct macvlan_dev *vlan = netdev_priv(dev); 1020 int err = -EINVAL; 1021 1022 /* Support unicast filter only on passthru devices. 1023 * Multicast filter should be allowed on all devices. 1024 */ 1025 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1026 return -EOPNOTSUPP; 1027 1028 if (is_unicast_ether_addr(addr)) 1029 err = dev_uc_del(dev, addr); 1030 else if (is_multicast_ether_addr(addr)) 1031 err = dev_mc_del(dev, addr); 1032 1033 return err; 1034 } 1035 1036 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1037 struct ethtool_drvinfo *drvinfo) 1038 { 1039 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1040 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1041 } 1042 1043 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1044 struct ethtool_link_ksettings *cmd) 1045 { 1046 const struct macvlan_dev *vlan = netdev_priv(dev); 1047 1048 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1049 } 1050 1051 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1052 struct ethtool_ts_info *info) 1053 { 1054 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1055 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1056 struct phy_device *phydev = real_dev->phydev; 1057 1058 if (phy_has_tsinfo(phydev)) { 1059 return phy_ts_info(phydev, info); 1060 } else if (ops->get_ts_info) { 1061 return ops->get_ts_info(real_dev, info); 1062 } else { 1063 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1064 SOF_TIMESTAMPING_SOFTWARE; 1065 info->phc_index = -1; 1066 } 1067 1068 return 0; 1069 } 1070 1071 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1072 netdev_features_t features) 1073 { 1074 struct macvlan_dev *vlan = netdev_priv(dev); 1075 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1076 netdev_features_t mask; 1077 1078 features |= NETIF_F_ALL_FOR_ALL; 1079 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1080 mask = features; 1081 1082 lowerdev_features &= (features | ~NETIF_F_LRO); 1083 features = netdev_increment_features(lowerdev_features, features, mask); 1084 features |= ALWAYS_ON_FEATURES; 1085 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1086 1087 return features; 1088 } 1089 1090 #ifdef CONFIG_NET_POLL_CONTROLLER 1091 static void macvlan_dev_poll_controller(struct net_device *dev) 1092 { 1093 return; 1094 } 1095 1096 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1097 { 1098 struct macvlan_dev *vlan = netdev_priv(dev); 1099 struct net_device *real_dev = vlan->lowerdev; 1100 struct netpoll *netpoll; 1101 int err; 1102 1103 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1104 err = -ENOMEM; 1105 if (!netpoll) 1106 goto out; 1107 1108 err = __netpoll_setup(netpoll, real_dev); 1109 if (err) { 1110 kfree(netpoll); 1111 goto out; 1112 } 1113 1114 vlan->netpoll = netpoll; 1115 1116 out: 1117 return err; 1118 } 1119 1120 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1121 { 1122 struct macvlan_dev *vlan = netdev_priv(dev); 1123 struct netpoll *netpoll = vlan->netpoll; 1124 1125 if (!netpoll) 1126 return; 1127 1128 vlan->netpoll = NULL; 1129 1130 __netpoll_free(netpoll); 1131 } 1132 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1133 1134 static int macvlan_dev_get_iflink(const struct net_device *dev) 1135 { 1136 struct macvlan_dev *vlan = netdev_priv(dev); 1137 1138 return vlan->lowerdev->ifindex; 1139 } 1140 1141 static const struct ethtool_ops macvlan_ethtool_ops = { 1142 .get_link = ethtool_op_get_link, 1143 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1144 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1145 .get_ts_info = macvlan_ethtool_get_ts_info, 1146 }; 1147 1148 static const struct net_device_ops macvlan_netdev_ops = { 1149 .ndo_init = macvlan_init, 1150 .ndo_uninit = macvlan_uninit, 1151 .ndo_open = macvlan_open, 1152 .ndo_stop = macvlan_stop, 1153 .ndo_start_xmit = macvlan_start_xmit, 1154 .ndo_change_mtu = macvlan_change_mtu, 1155 .ndo_do_ioctl = macvlan_do_ioctl, 1156 .ndo_fix_features = macvlan_fix_features, 1157 .ndo_change_rx_flags = macvlan_change_rx_flags, 1158 .ndo_set_mac_address = macvlan_set_mac_address, 1159 .ndo_set_rx_mode = macvlan_set_mac_lists, 1160 .ndo_get_stats64 = macvlan_dev_get_stats64, 1161 .ndo_validate_addr = eth_validate_addr, 1162 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1163 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1164 .ndo_fdb_add = macvlan_fdb_add, 1165 .ndo_fdb_del = macvlan_fdb_del, 1166 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1167 #ifdef CONFIG_NET_POLL_CONTROLLER 1168 .ndo_poll_controller = macvlan_dev_poll_controller, 1169 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1170 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1171 #endif 1172 .ndo_get_iflink = macvlan_dev_get_iflink, 1173 .ndo_features_check = passthru_features_check, 1174 .ndo_change_proto_down = dev_change_proto_down_generic, 1175 }; 1176 1177 void macvlan_common_setup(struct net_device *dev) 1178 { 1179 ether_setup(dev); 1180 1181 dev->min_mtu = 0; 1182 dev->max_mtu = ETH_MAX_MTU; 1183 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1184 netif_keep_dst(dev); 1185 dev->priv_flags |= IFF_UNICAST_FLT; 1186 dev->netdev_ops = &macvlan_netdev_ops; 1187 dev->needs_free_netdev = true; 1188 dev->header_ops = &macvlan_hard_header_ops; 1189 dev->ethtool_ops = &macvlan_ethtool_ops; 1190 } 1191 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1192 1193 static void macvlan_setup(struct net_device *dev) 1194 { 1195 macvlan_common_setup(dev); 1196 dev->priv_flags |= IFF_NO_QUEUE; 1197 } 1198 1199 static int macvlan_port_create(struct net_device *dev) 1200 { 1201 struct macvlan_port *port; 1202 unsigned int i; 1203 int err; 1204 1205 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1206 return -EINVAL; 1207 1208 if (netdev_is_rx_handler_busy(dev)) 1209 return -EBUSY; 1210 1211 port = kzalloc(sizeof(*port), GFP_KERNEL); 1212 if (port == NULL) 1213 return -ENOMEM; 1214 1215 port->dev = dev; 1216 ether_addr_copy(port->perm_addr, dev->dev_addr); 1217 INIT_LIST_HEAD(&port->vlans); 1218 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1219 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1220 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1221 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1222 1223 port->bc_queue_len_used = 0; 1224 skb_queue_head_init(&port->bc_queue); 1225 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1226 1227 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1228 if (err) 1229 kfree(port); 1230 else 1231 dev->priv_flags |= IFF_MACVLAN_PORT; 1232 return err; 1233 } 1234 1235 static void macvlan_port_destroy(struct net_device *dev) 1236 { 1237 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1238 struct sk_buff *skb; 1239 1240 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1241 netdev_rx_handler_unregister(dev); 1242 1243 /* After this point, no packet can schedule bc_work anymore, 1244 * but we need to cancel it and purge left skbs if any. 1245 */ 1246 cancel_work_sync(&port->bc_work); 1247 1248 while ((skb = __skb_dequeue(&port->bc_queue))) { 1249 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1250 1251 if (src) 1252 dev_put(src->dev); 1253 1254 kfree_skb(skb); 1255 } 1256 1257 /* If the lower device address has been changed by passthru 1258 * macvlan, put it back. 1259 */ 1260 if (macvlan_passthru(port) && 1261 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1262 struct sockaddr sa; 1263 1264 sa.sa_family = port->dev->type; 1265 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1266 dev_set_mac_address(port->dev, &sa, NULL); 1267 } 1268 1269 kfree(port); 1270 } 1271 1272 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1273 struct netlink_ext_ack *extack) 1274 { 1275 struct nlattr *nla, *head; 1276 int rem, len; 1277 1278 if (tb[IFLA_ADDRESS]) { 1279 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1280 return -EINVAL; 1281 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1282 return -EADDRNOTAVAIL; 1283 } 1284 1285 if (!data) 1286 return 0; 1287 1288 if (data[IFLA_MACVLAN_FLAGS] && 1289 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1290 return -EINVAL; 1291 1292 if (data[IFLA_MACVLAN_MODE]) { 1293 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1294 case MACVLAN_MODE_PRIVATE: 1295 case MACVLAN_MODE_VEPA: 1296 case MACVLAN_MODE_BRIDGE: 1297 case MACVLAN_MODE_PASSTHRU: 1298 case MACVLAN_MODE_SOURCE: 1299 break; 1300 default: 1301 return -EINVAL; 1302 } 1303 } 1304 1305 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1306 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1307 case MACVLAN_MACADDR_ADD: 1308 case MACVLAN_MACADDR_DEL: 1309 case MACVLAN_MACADDR_FLUSH: 1310 case MACVLAN_MACADDR_SET: 1311 break; 1312 default: 1313 return -EINVAL; 1314 } 1315 } 1316 1317 if (data[IFLA_MACVLAN_MACADDR]) { 1318 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1319 return -EINVAL; 1320 1321 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1322 return -EADDRNOTAVAIL; 1323 } 1324 1325 if (data[IFLA_MACVLAN_MACADDR_DATA]) { 1326 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1327 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1328 1329 nla_for_each_attr(nla, head, len, rem) { 1330 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1331 nla_len(nla) != ETH_ALEN) 1332 return -EINVAL; 1333 1334 if (!is_valid_ether_addr(nla_data(nla))) 1335 return -EADDRNOTAVAIL; 1336 } 1337 } 1338 1339 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1340 return -EINVAL; 1341 1342 return 0; 1343 } 1344 1345 /* 1346 * reconfigure list of remote source mac address 1347 * (only for macvlan devices in source mode) 1348 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1349 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1350 */ 1351 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1352 struct nlattr *data[]) 1353 { 1354 char *addr = NULL; 1355 int ret, rem, len; 1356 struct nlattr *nla, *head; 1357 struct macvlan_source_entry *entry; 1358 1359 if (data[IFLA_MACVLAN_MACADDR]) 1360 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1361 1362 if (mode == MACVLAN_MACADDR_ADD) { 1363 if (!addr) 1364 return -EINVAL; 1365 1366 return macvlan_hash_add_source(vlan, addr); 1367 1368 } else if (mode == MACVLAN_MACADDR_DEL) { 1369 if (!addr) 1370 return -EINVAL; 1371 1372 entry = macvlan_hash_lookup_source(vlan, addr); 1373 if (entry) { 1374 macvlan_hash_del_source(entry); 1375 vlan->macaddr_count--; 1376 } 1377 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1378 macvlan_flush_sources(vlan->port, vlan); 1379 } else if (mode == MACVLAN_MACADDR_SET) { 1380 macvlan_flush_sources(vlan->port, vlan); 1381 1382 if (addr) { 1383 ret = macvlan_hash_add_source(vlan, addr); 1384 if (ret) 1385 return ret; 1386 } 1387 1388 if (!data[IFLA_MACVLAN_MACADDR_DATA]) 1389 return 0; 1390 1391 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1392 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1393 1394 nla_for_each_attr(nla, head, len, rem) { 1395 addr = nla_data(nla); 1396 ret = macvlan_hash_add_source(vlan, addr); 1397 if (ret) 1398 return ret; 1399 } 1400 } else { 1401 return -EINVAL; 1402 } 1403 1404 return 0; 1405 } 1406 1407 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1408 struct nlattr *tb[], struct nlattr *data[], 1409 struct netlink_ext_ack *extack) 1410 { 1411 struct macvlan_dev *vlan = netdev_priv(dev); 1412 struct macvlan_port *port; 1413 struct net_device *lowerdev; 1414 int err; 1415 int macmode; 1416 bool create = false; 1417 1418 if (!tb[IFLA_LINK]) 1419 return -EINVAL; 1420 1421 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1422 if (lowerdev == NULL) 1423 return -ENODEV; 1424 1425 /* When creating macvlans or macvtaps on top of other macvlans - use 1426 * the real device as the lowerdev. 1427 */ 1428 if (netif_is_macvlan(lowerdev)) 1429 lowerdev = macvlan_dev_real_dev(lowerdev); 1430 1431 if (!tb[IFLA_MTU]) 1432 dev->mtu = lowerdev->mtu; 1433 else if (dev->mtu > lowerdev->mtu) 1434 return -EINVAL; 1435 1436 /* MTU range: 68 - lowerdev->max_mtu */ 1437 dev->min_mtu = ETH_MIN_MTU; 1438 dev->max_mtu = lowerdev->max_mtu; 1439 1440 if (!tb[IFLA_ADDRESS]) 1441 eth_hw_addr_random(dev); 1442 1443 if (!netif_is_macvlan_port(lowerdev)) { 1444 err = macvlan_port_create(lowerdev); 1445 if (err < 0) 1446 return err; 1447 create = true; 1448 } 1449 port = macvlan_port_get_rtnl(lowerdev); 1450 1451 /* Only 1 macvlan device can be created in passthru mode */ 1452 if (macvlan_passthru(port)) { 1453 /* The macvlan port must be not created this time, 1454 * still goto destroy_macvlan_port for readability. 1455 */ 1456 err = -EINVAL; 1457 goto destroy_macvlan_port; 1458 } 1459 1460 vlan->lowerdev = lowerdev; 1461 vlan->dev = dev; 1462 vlan->port = port; 1463 vlan->set_features = MACVLAN_FEATURES; 1464 1465 vlan->mode = MACVLAN_MODE_VEPA; 1466 if (data && data[IFLA_MACVLAN_MODE]) 1467 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1468 1469 if (data && data[IFLA_MACVLAN_FLAGS]) 1470 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1471 1472 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1473 if (port->count) { 1474 err = -EINVAL; 1475 goto destroy_macvlan_port; 1476 } 1477 macvlan_set_passthru(port); 1478 eth_hw_addr_inherit(dev, lowerdev); 1479 } 1480 1481 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1482 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1483 err = -EINVAL; 1484 goto destroy_macvlan_port; 1485 } 1486 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1487 err = macvlan_changelink_sources(vlan, macmode, data); 1488 if (err) 1489 goto destroy_macvlan_port; 1490 } 1491 1492 vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN; 1493 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) 1494 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1495 1496 err = register_netdevice(dev); 1497 if (err < 0) 1498 goto destroy_macvlan_port; 1499 1500 dev->priv_flags |= IFF_MACVLAN; 1501 err = netdev_upper_dev_link(lowerdev, dev, extack); 1502 if (err) 1503 goto unregister_netdev; 1504 1505 list_add_tail_rcu(&vlan->list, &port->vlans); 1506 update_port_bc_queue_len(vlan->port); 1507 netif_stacked_transfer_operstate(lowerdev, dev); 1508 linkwatch_fire_event(dev); 1509 1510 return 0; 1511 1512 unregister_netdev: 1513 /* macvlan_uninit would free the macvlan port */ 1514 unregister_netdevice(dev); 1515 return err; 1516 destroy_macvlan_port: 1517 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1518 * so we destroy the macvlan port only when it's valid. 1519 */ 1520 if (create && macvlan_port_get_rtnl(lowerdev)) 1521 macvlan_port_destroy(port->dev); 1522 return err; 1523 } 1524 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1525 1526 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1527 struct nlattr *tb[], struct nlattr *data[], 1528 struct netlink_ext_ack *extack) 1529 { 1530 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1531 } 1532 1533 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1534 { 1535 struct macvlan_dev *vlan = netdev_priv(dev); 1536 1537 if (vlan->mode == MACVLAN_MODE_SOURCE) 1538 macvlan_flush_sources(vlan->port, vlan); 1539 list_del_rcu(&vlan->list); 1540 update_port_bc_queue_len(vlan->port); 1541 unregister_netdevice_queue(dev, head); 1542 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1543 } 1544 EXPORT_SYMBOL_GPL(macvlan_dellink); 1545 1546 static int macvlan_changelink(struct net_device *dev, 1547 struct nlattr *tb[], struct nlattr *data[], 1548 struct netlink_ext_ack *extack) 1549 { 1550 struct macvlan_dev *vlan = netdev_priv(dev); 1551 enum macvlan_mode mode; 1552 bool set_mode = false; 1553 enum macvlan_macaddr_mode macmode; 1554 int ret; 1555 1556 /* Validate mode, but don't set yet: setting flags may fail. */ 1557 if (data && data[IFLA_MACVLAN_MODE]) { 1558 set_mode = true; 1559 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1560 /* Passthrough mode can't be set or cleared dynamically */ 1561 if ((mode == MACVLAN_MODE_PASSTHRU) != 1562 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1563 return -EINVAL; 1564 if (vlan->mode == MACVLAN_MODE_SOURCE && 1565 vlan->mode != mode) 1566 macvlan_flush_sources(vlan->port, vlan); 1567 } 1568 1569 if (data && data[IFLA_MACVLAN_FLAGS]) { 1570 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1571 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1572 if (macvlan_passthru(vlan->port) && promisc) { 1573 int err; 1574 1575 if (flags & MACVLAN_FLAG_NOPROMISC) 1576 err = dev_set_promiscuity(vlan->lowerdev, -1); 1577 else 1578 err = dev_set_promiscuity(vlan->lowerdev, 1); 1579 if (err < 0) 1580 return err; 1581 } 1582 vlan->flags = flags; 1583 } 1584 1585 if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) { 1586 vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]); 1587 update_port_bc_queue_len(vlan->port); 1588 } 1589 1590 if (set_mode) 1591 vlan->mode = mode; 1592 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1593 if (vlan->mode != MACVLAN_MODE_SOURCE) 1594 return -EINVAL; 1595 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1596 ret = macvlan_changelink_sources(vlan, macmode, data); 1597 if (ret) 1598 return ret; 1599 } 1600 return 0; 1601 } 1602 1603 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1604 { 1605 if (vlan->macaddr_count == 0) 1606 return 0; 1607 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1608 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1609 } 1610 1611 static size_t macvlan_get_size(const struct net_device *dev) 1612 { 1613 struct macvlan_dev *vlan = netdev_priv(dev); 1614 1615 return (0 1616 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1617 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1618 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1619 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1620 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */ 1621 + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */ 1622 ); 1623 } 1624 1625 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1626 const struct macvlan_dev *vlan, 1627 const int i) 1628 { 1629 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1630 struct macvlan_source_entry *entry; 1631 1632 hlist_for_each_entry_rcu(entry, h, hlist) { 1633 if (entry->vlan != vlan) 1634 continue; 1635 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1636 return 1; 1637 } 1638 return 0; 1639 } 1640 1641 static int macvlan_fill_info(struct sk_buff *skb, 1642 const struct net_device *dev) 1643 { 1644 struct macvlan_dev *vlan = netdev_priv(dev); 1645 struct macvlan_port *port = vlan->port; 1646 int i; 1647 struct nlattr *nest; 1648 1649 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1650 goto nla_put_failure; 1651 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1652 goto nla_put_failure; 1653 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1654 goto nla_put_failure; 1655 if (vlan->macaddr_count > 0) { 1656 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1657 if (nest == NULL) 1658 goto nla_put_failure; 1659 1660 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1661 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1662 goto nla_put_failure; 1663 } 1664 nla_nest_end(skb, nest); 1665 } 1666 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req)) 1667 goto nla_put_failure; 1668 if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used)) 1669 goto nla_put_failure; 1670 return 0; 1671 1672 nla_put_failure: 1673 return -EMSGSIZE; 1674 } 1675 1676 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1677 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1678 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1679 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1680 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1681 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1682 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1683 [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 }, 1684 [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT }, 1685 }; 1686 1687 int macvlan_link_register(struct rtnl_link_ops *ops) 1688 { 1689 /* common fields */ 1690 ops->validate = macvlan_validate; 1691 ops->maxtype = IFLA_MACVLAN_MAX; 1692 ops->policy = macvlan_policy; 1693 ops->changelink = macvlan_changelink; 1694 ops->get_size = macvlan_get_size; 1695 ops->fill_info = macvlan_fill_info; 1696 1697 return rtnl_link_register(ops); 1698 }; 1699 EXPORT_SYMBOL_GPL(macvlan_link_register); 1700 1701 static struct net *macvlan_get_link_net(const struct net_device *dev) 1702 { 1703 return dev_net(macvlan_dev_real_dev(dev)); 1704 } 1705 1706 static struct rtnl_link_ops macvlan_link_ops = { 1707 .kind = "macvlan", 1708 .setup = macvlan_setup, 1709 .newlink = macvlan_newlink, 1710 .dellink = macvlan_dellink, 1711 .get_link_net = macvlan_get_link_net, 1712 .priv_size = sizeof(struct macvlan_dev), 1713 }; 1714 1715 static void update_port_bc_queue_len(struct macvlan_port *port) 1716 { 1717 u32 max_bc_queue_len_req = 0; 1718 struct macvlan_dev *vlan; 1719 1720 list_for_each_entry(vlan, &port->vlans, list) { 1721 if (vlan->bc_queue_len_req > max_bc_queue_len_req) 1722 max_bc_queue_len_req = vlan->bc_queue_len_req; 1723 } 1724 port->bc_queue_len_used = max_bc_queue_len_req; 1725 } 1726 1727 static int macvlan_device_event(struct notifier_block *unused, 1728 unsigned long event, void *ptr) 1729 { 1730 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1731 struct macvlan_dev *vlan, *next; 1732 struct macvlan_port *port; 1733 LIST_HEAD(list_kill); 1734 1735 if (!netif_is_macvlan_port(dev)) 1736 return NOTIFY_DONE; 1737 1738 port = macvlan_port_get_rtnl(dev); 1739 1740 switch (event) { 1741 case NETDEV_UP: 1742 case NETDEV_DOWN: 1743 case NETDEV_CHANGE: 1744 list_for_each_entry(vlan, &port->vlans, list) 1745 netif_stacked_transfer_operstate(vlan->lowerdev, 1746 vlan->dev); 1747 break; 1748 case NETDEV_FEAT_CHANGE: 1749 list_for_each_entry(vlan, &port->vlans, list) { 1750 vlan->dev->gso_max_size = dev->gso_max_size; 1751 vlan->dev->gso_max_segs = dev->gso_max_segs; 1752 netdev_update_features(vlan->dev); 1753 } 1754 break; 1755 case NETDEV_CHANGEMTU: 1756 list_for_each_entry(vlan, &port->vlans, list) { 1757 if (vlan->dev->mtu <= dev->mtu) 1758 continue; 1759 dev_set_mtu(vlan->dev, dev->mtu); 1760 } 1761 break; 1762 case NETDEV_CHANGEADDR: 1763 if (!macvlan_passthru(port)) 1764 return NOTIFY_DONE; 1765 1766 vlan = list_first_entry_or_null(&port->vlans, 1767 struct macvlan_dev, 1768 list); 1769 1770 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1771 return NOTIFY_BAD; 1772 1773 break; 1774 case NETDEV_UNREGISTER: 1775 /* twiddle thumbs on netns device moves */ 1776 if (dev->reg_state != NETREG_UNREGISTERING) 1777 break; 1778 1779 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1780 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1781 unregister_netdevice_many(&list_kill); 1782 break; 1783 case NETDEV_PRE_TYPE_CHANGE: 1784 /* Forbid underlaying device to change its type. */ 1785 return NOTIFY_BAD; 1786 1787 case NETDEV_NOTIFY_PEERS: 1788 case NETDEV_BONDING_FAILOVER: 1789 case NETDEV_RESEND_IGMP: 1790 /* Propagate to all vlans */ 1791 list_for_each_entry(vlan, &port->vlans, list) 1792 call_netdevice_notifiers(event, vlan->dev); 1793 } 1794 return NOTIFY_DONE; 1795 } 1796 1797 static struct notifier_block macvlan_notifier_block __read_mostly = { 1798 .notifier_call = macvlan_device_event, 1799 }; 1800 1801 static int __init macvlan_init_module(void) 1802 { 1803 int err; 1804 1805 register_netdevice_notifier(&macvlan_notifier_block); 1806 1807 err = macvlan_link_register(&macvlan_link_ops); 1808 if (err < 0) 1809 goto err1; 1810 return 0; 1811 err1: 1812 unregister_netdevice_notifier(&macvlan_notifier_block); 1813 return err; 1814 } 1815 1816 static void __exit macvlan_cleanup_module(void) 1817 { 1818 rtnl_link_unregister(&macvlan_link_ops); 1819 unregister_netdevice_notifier(&macvlan_notifier_block); 1820 } 1821 1822 module_init(macvlan_init_module); 1823 module_exit(macvlan_cleanup_module); 1824 1825 MODULE_LICENSE("GPL"); 1826 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1827 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1828 MODULE_ALIAS_RTNL_LINK("macvlan"); 1829