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 if (vlan->netpoll) 546 netpoll_send_skb(vlan->netpoll, skb); 547 #else 548 BUG(); 549 #endif 550 return NETDEV_TX_OK; 551 } 552 553 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 554 struct net_device *dev) 555 { 556 struct macvlan_dev *vlan = netdev_priv(dev); 557 unsigned int len = skb->len; 558 int ret; 559 560 if (unlikely(netpoll_tx_running(dev))) 561 return macvlan_netpoll_send_skb(vlan, skb); 562 563 ret = macvlan_queue_xmit(skb, dev); 564 565 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 566 struct vlan_pcpu_stats *pcpu_stats; 567 568 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 569 u64_stats_update_begin(&pcpu_stats->syncp); 570 pcpu_stats->tx_packets++; 571 pcpu_stats->tx_bytes += len; 572 u64_stats_update_end(&pcpu_stats->syncp); 573 } else { 574 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 575 } 576 return ret; 577 } 578 579 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 580 unsigned short type, const void *daddr, 581 const void *saddr, unsigned len) 582 { 583 const struct macvlan_dev *vlan = netdev_priv(dev); 584 struct net_device *lowerdev = vlan->lowerdev; 585 586 return dev_hard_header(skb, lowerdev, type, daddr, 587 saddr ? : dev->dev_addr, len); 588 } 589 590 static const struct header_ops macvlan_hard_header_ops = { 591 .create = macvlan_hard_header, 592 .parse = eth_header_parse, 593 .cache = eth_header_cache, 594 .cache_update = eth_header_cache_update, 595 }; 596 597 static int macvlan_open(struct net_device *dev) 598 { 599 struct macvlan_dev *vlan = netdev_priv(dev); 600 struct net_device *lowerdev = vlan->lowerdev; 601 int err; 602 603 if (macvlan_passthru(vlan->port)) { 604 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 605 err = dev_set_promiscuity(lowerdev, 1); 606 if (err < 0) 607 goto out; 608 } 609 goto hash_add; 610 } 611 612 err = -EADDRINUSE; 613 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 614 goto out; 615 616 /* Attempt to populate accel_priv which is used to offload the L2 617 * forwarding requests for unicast packets. 618 */ 619 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 620 vlan->accel_priv = 621 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 622 623 /* If earlier attempt to offload failed, or accel_priv is not 624 * populated we must add the unicast address to the lower device. 625 */ 626 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 627 vlan->accel_priv = NULL; 628 err = dev_uc_add(lowerdev, dev->dev_addr); 629 if (err < 0) 630 goto out; 631 } 632 633 if (dev->flags & IFF_ALLMULTI) { 634 err = dev_set_allmulti(lowerdev, 1); 635 if (err < 0) 636 goto del_unicast; 637 } 638 639 if (dev->flags & IFF_PROMISC) { 640 err = dev_set_promiscuity(lowerdev, 1); 641 if (err < 0) 642 goto clear_multi; 643 } 644 645 hash_add: 646 macvlan_hash_add(vlan); 647 return 0; 648 649 clear_multi: 650 if (dev->flags & IFF_ALLMULTI) 651 dev_set_allmulti(lowerdev, -1); 652 del_unicast: 653 if (vlan->accel_priv) { 654 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 655 vlan->accel_priv); 656 vlan->accel_priv = NULL; 657 } else { 658 dev_uc_del(lowerdev, dev->dev_addr); 659 } 660 out: 661 return err; 662 } 663 664 static int macvlan_stop(struct net_device *dev) 665 { 666 struct macvlan_dev *vlan = netdev_priv(dev); 667 struct net_device *lowerdev = vlan->lowerdev; 668 669 if (vlan->accel_priv) { 670 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 671 vlan->accel_priv); 672 vlan->accel_priv = NULL; 673 } 674 675 dev_uc_unsync(lowerdev, dev); 676 dev_mc_unsync(lowerdev, dev); 677 678 if (macvlan_passthru(vlan->port)) { 679 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 680 dev_set_promiscuity(lowerdev, -1); 681 goto hash_del; 682 } 683 684 if (dev->flags & IFF_ALLMULTI) 685 dev_set_allmulti(lowerdev, -1); 686 687 if (dev->flags & IFF_PROMISC) 688 dev_set_promiscuity(lowerdev, -1); 689 690 dev_uc_del(lowerdev, dev->dev_addr); 691 692 hash_del: 693 macvlan_hash_del(vlan, !dev->dismantle); 694 return 0; 695 } 696 697 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 698 { 699 struct macvlan_dev *vlan = netdev_priv(dev); 700 struct net_device *lowerdev = vlan->lowerdev; 701 struct macvlan_port *port = vlan->port; 702 int err; 703 704 if (!(dev->flags & IFF_UP)) { 705 /* Just copy in the new address */ 706 ether_addr_copy(dev->dev_addr, addr); 707 } else { 708 /* Rehash and update the device filters */ 709 if (macvlan_addr_busy(vlan->port, addr)) 710 return -EADDRINUSE; 711 712 if (!macvlan_passthru(port)) { 713 err = dev_uc_add(lowerdev, addr); 714 if (err) 715 return err; 716 717 dev_uc_del(lowerdev, dev->dev_addr); 718 } 719 720 macvlan_hash_change_addr(vlan, addr); 721 } 722 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 723 /* Since addr_change isn't set, we are here due to lower 724 * device change. Save the lower-dev address so we can 725 * restore it later. 726 */ 727 ether_addr_copy(vlan->port->perm_addr, 728 lowerdev->dev_addr); 729 } 730 macvlan_clear_addr_change(port); 731 return 0; 732 } 733 734 static int macvlan_set_mac_address(struct net_device *dev, void *p) 735 { 736 struct macvlan_dev *vlan = netdev_priv(dev); 737 struct sockaddr *addr = p; 738 739 if (!is_valid_ether_addr(addr->sa_data)) 740 return -EADDRNOTAVAIL; 741 742 /* If the addresses are the same, this is a no-op */ 743 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 744 return 0; 745 746 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 747 macvlan_set_addr_change(vlan->port); 748 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 749 } 750 751 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 752 return -EADDRINUSE; 753 754 return macvlan_sync_address(dev, addr->sa_data); 755 } 756 757 static void macvlan_change_rx_flags(struct net_device *dev, int change) 758 { 759 struct macvlan_dev *vlan = netdev_priv(dev); 760 struct net_device *lowerdev = vlan->lowerdev; 761 762 if (dev->flags & IFF_UP) { 763 if (change & IFF_ALLMULTI) 764 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 765 if (change & IFF_PROMISC) 766 dev_set_promiscuity(lowerdev, 767 dev->flags & IFF_PROMISC ? 1 : -1); 768 769 } 770 } 771 772 static void macvlan_compute_filter(unsigned long *mc_filter, 773 struct net_device *dev, 774 struct macvlan_dev *vlan) 775 { 776 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 777 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 778 } else { 779 struct netdev_hw_addr *ha; 780 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 781 782 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 783 netdev_for_each_mc_addr(ha, dev) { 784 __set_bit(mc_hash(vlan, ha->addr), filter); 785 } 786 787 __set_bit(mc_hash(vlan, dev->broadcast), filter); 788 789 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 790 } 791 } 792 793 static void macvlan_set_mac_lists(struct net_device *dev) 794 { 795 struct macvlan_dev *vlan = netdev_priv(dev); 796 797 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 798 799 dev_uc_sync(vlan->lowerdev, dev); 800 dev_mc_sync(vlan->lowerdev, dev); 801 802 /* This is slightly inaccurate as we're including the subscription 803 * list of vlan->lowerdev too. 804 * 805 * Bug alert: This only works if everyone has the same broadcast 806 * address as lowerdev. As soon as someone changes theirs this 807 * will break. 808 * 809 * However, this is already broken as when you change your broadcast 810 * address we don't get called. 811 * 812 * The solution is to maintain a list of broadcast addresses like 813 * we do for uc/mc, if you care. 814 */ 815 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 816 } 817 818 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 819 { 820 struct macvlan_dev *vlan = netdev_priv(dev); 821 822 if (vlan->lowerdev->mtu < new_mtu) 823 return -EINVAL; 824 dev->mtu = new_mtu; 825 return 0; 826 } 827 828 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 829 { 830 struct net_device *real_dev = macvlan_dev_real_dev(dev); 831 const struct net_device_ops *ops = real_dev->netdev_ops; 832 struct ifreq ifrr; 833 int err = -EOPNOTSUPP; 834 835 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ); 836 ifrr.ifr_ifru = ifr->ifr_ifru; 837 838 switch (cmd) { 839 case SIOCSHWTSTAMP: 840 if (!net_eq(dev_net(dev), &init_net)) 841 break; 842 /* fall through */ 843 case SIOCGHWTSTAMP: 844 if (netif_device_present(real_dev) && ops->ndo_do_ioctl) 845 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd); 846 break; 847 } 848 849 if (!err) 850 ifr->ifr_ifru = ifrr.ifr_ifru; 851 852 return err; 853 } 854 855 /* 856 * macvlan network devices have devices nesting below it and are a special 857 * "super class" of normal network devices; split their locks off into a 858 * separate class since they always nest. 859 */ 860 #define ALWAYS_ON_OFFLOADS \ 861 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 862 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 863 864 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 865 866 #define MACVLAN_FEATURES \ 867 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 868 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 869 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 870 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 871 872 #define MACVLAN_STATE_MASK \ 873 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 874 875 static int macvlan_init(struct net_device *dev) 876 { 877 struct macvlan_dev *vlan = netdev_priv(dev); 878 const struct net_device *lowerdev = vlan->lowerdev; 879 struct macvlan_port *port = vlan->port; 880 881 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 882 (lowerdev->state & MACVLAN_STATE_MASK); 883 dev->features = lowerdev->features & MACVLAN_FEATURES; 884 dev->features |= ALWAYS_ON_FEATURES; 885 dev->hw_features |= NETIF_F_LRO; 886 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 887 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 888 dev->hw_enc_features |= dev->features; 889 dev->gso_max_size = lowerdev->gso_max_size; 890 dev->gso_max_segs = lowerdev->gso_max_segs; 891 dev->hard_header_len = lowerdev->hard_header_len; 892 893 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 894 if (!vlan->pcpu_stats) 895 return -ENOMEM; 896 897 port->count += 1; 898 899 return 0; 900 } 901 902 static void macvlan_uninit(struct net_device *dev) 903 { 904 struct macvlan_dev *vlan = netdev_priv(dev); 905 struct macvlan_port *port = vlan->port; 906 907 free_percpu(vlan->pcpu_stats); 908 909 macvlan_flush_sources(port, vlan); 910 port->count -= 1; 911 if (!port->count) 912 macvlan_port_destroy(port->dev); 913 } 914 915 static void macvlan_dev_get_stats64(struct net_device *dev, 916 struct rtnl_link_stats64 *stats) 917 { 918 struct macvlan_dev *vlan = netdev_priv(dev); 919 920 if (vlan->pcpu_stats) { 921 struct vlan_pcpu_stats *p; 922 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 923 u32 rx_errors = 0, tx_dropped = 0; 924 unsigned int start; 925 int i; 926 927 for_each_possible_cpu(i) { 928 p = per_cpu_ptr(vlan->pcpu_stats, i); 929 do { 930 start = u64_stats_fetch_begin_irq(&p->syncp); 931 rx_packets = p->rx_packets; 932 rx_bytes = p->rx_bytes; 933 rx_multicast = p->rx_multicast; 934 tx_packets = p->tx_packets; 935 tx_bytes = p->tx_bytes; 936 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 937 938 stats->rx_packets += rx_packets; 939 stats->rx_bytes += rx_bytes; 940 stats->multicast += rx_multicast; 941 stats->tx_packets += tx_packets; 942 stats->tx_bytes += tx_bytes; 943 /* rx_errors & tx_dropped are u32, updated 944 * without syncp protection. 945 */ 946 rx_errors += p->rx_errors; 947 tx_dropped += p->tx_dropped; 948 } 949 stats->rx_errors = rx_errors; 950 stats->rx_dropped = rx_errors; 951 stats->tx_dropped = tx_dropped; 952 } 953 } 954 955 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 956 __be16 proto, u16 vid) 957 { 958 struct macvlan_dev *vlan = netdev_priv(dev); 959 struct net_device *lowerdev = vlan->lowerdev; 960 961 return vlan_vid_add(lowerdev, proto, vid); 962 } 963 964 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 965 __be16 proto, u16 vid) 966 { 967 struct macvlan_dev *vlan = netdev_priv(dev); 968 struct net_device *lowerdev = vlan->lowerdev; 969 970 vlan_vid_del(lowerdev, proto, vid); 971 return 0; 972 } 973 974 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 975 struct net_device *dev, 976 const unsigned char *addr, u16 vid, 977 u16 flags, 978 struct netlink_ext_ack *extack) 979 { 980 struct macvlan_dev *vlan = netdev_priv(dev); 981 int err = -EINVAL; 982 983 /* Support unicast filter only on passthru devices. 984 * Multicast filter should be allowed on all devices. 985 */ 986 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 987 return -EOPNOTSUPP; 988 989 if (flags & NLM_F_REPLACE) 990 return -EOPNOTSUPP; 991 992 if (is_unicast_ether_addr(addr)) 993 err = dev_uc_add_excl(dev, addr); 994 else if (is_multicast_ether_addr(addr)) 995 err = dev_mc_add_excl(dev, addr); 996 997 return err; 998 } 999 1000 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1001 struct net_device *dev, 1002 const unsigned char *addr, u16 vid) 1003 { 1004 struct macvlan_dev *vlan = netdev_priv(dev); 1005 int err = -EINVAL; 1006 1007 /* Support unicast filter only on passthru devices. 1008 * Multicast filter should be allowed on all devices. 1009 */ 1010 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1011 return -EOPNOTSUPP; 1012 1013 if (is_unicast_ether_addr(addr)) 1014 err = dev_uc_del(dev, addr); 1015 else if (is_multicast_ether_addr(addr)) 1016 err = dev_mc_del(dev, addr); 1017 1018 return err; 1019 } 1020 1021 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1022 struct ethtool_drvinfo *drvinfo) 1023 { 1024 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1025 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1026 } 1027 1028 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1029 struct ethtool_link_ksettings *cmd) 1030 { 1031 const struct macvlan_dev *vlan = netdev_priv(dev); 1032 1033 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1034 } 1035 1036 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1037 struct ethtool_ts_info *info) 1038 { 1039 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1040 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1041 struct phy_device *phydev = real_dev->phydev; 1042 1043 if (phy_has_tsinfo(phydev)) { 1044 return phy_ts_info(phydev, info); 1045 } else if (ops->get_ts_info) { 1046 return ops->get_ts_info(real_dev, info); 1047 } else { 1048 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1049 SOF_TIMESTAMPING_SOFTWARE; 1050 info->phc_index = -1; 1051 } 1052 1053 return 0; 1054 } 1055 1056 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1057 netdev_features_t features) 1058 { 1059 struct macvlan_dev *vlan = netdev_priv(dev); 1060 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1061 netdev_features_t mask; 1062 1063 features |= NETIF_F_ALL_FOR_ALL; 1064 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1065 mask = features; 1066 1067 lowerdev_features &= (features | ~NETIF_F_LRO); 1068 features = netdev_increment_features(lowerdev_features, features, mask); 1069 features |= ALWAYS_ON_FEATURES; 1070 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1071 1072 return features; 1073 } 1074 1075 #ifdef CONFIG_NET_POLL_CONTROLLER 1076 static void macvlan_dev_poll_controller(struct net_device *dev) 1077 { 1078 return; 1079 } 1080 1081 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1082 { 1083 struct macvlan_dev *vlan = netdev_priv(dev); 1084 struct net_device *real_dev = vlan->lowerdev; 1085 struct netpoll *netpoll; 1086 int err = 0; 1087 1088 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1089 err = -ENOMEM; 1090 if (!netpoll) 1091 goto out; 1092 1093 err = __netpoll_setup(netpoll, real_dev); 1094 if (err) { 1095 kfree(netpoll); 1096 goto out; 1097 } 1098 1099 vlan->netpoll = netpoll; 1100 1101 out: 1102 return err; 1103 } 1104 1105 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1106 { 1107 struct macvlan_dev *vlan = netdev_priv(dev); 1108 struct netpoll *netpoll = vlan->netpoll; 1109 1110 if (!netpoll) 1111 return; 1112 1113 vlan->netpoll = NULL; 1114 1115 __netpoll_free(netpoll); 1116 } 1117 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1118 1119 static int macvlan_dev_get_iflink(const struct net_device *dev) 1120 { 1121 struct macvlan_dev *vlan = netdev_priv(dev); 1122 1123 return vlan->lowerdev->ifindex; 1124 } 1125 1126 static const struct ethtool_ops macvlan_ethtool_ops = { 1127 .get_link = ethtool_op_get_link, 1128 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1129 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1130 .get_ts_info = macvlan_ethtool_get_ts_info, 1131 }; 1132 1133 static const struct net_device_ops macvlan_netdev_ops = { 1134 .ndo_init = macvlan_init, 1135 .ndo_uninit = macvlan_uninit, 1136 .ndo_open = macvlan_open, 1137 .ndo_stop = macvlan_stop, 1138 .ndo_start_xmit = macvlan_start_xmit, 1139 .ndo_change_mtu = macvlan_change_mtu, 1140 .ndo_do_ioctl = macvlan_do_ioctl, 1141 .ndo_fix_features = macvlan_fix_features, 1142 .ndo_change_rx_flags = macvlan_change_rx_flags, 1143 .ndo_set_mac_address = macvlan_set_mac_address, 1144 .ndo_set_rx_mode = macvlan_set_mac_lists, 1145 .ndo_get_stats64 = macvlan_dev_get_stats64, 1146 .ndo_validate_addr = eth_validate_addr, 1147 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1148 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1149 .ndo_fdb_add = macvlan_fdb_add, 1150 .ndo_fdb_del = macvlan_fdb_del, 1151 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1152 #ifdef CONFIG_NET_POLL_CONTROLLER 1153 .ndo_poll_controller = macvlan_dev_poll_controller, 1154 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1155 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1156 #endif 1157 .ndo_get_iflink = macvlan_dev_get_iflink, 1158 .ndo_features_check = passthru_features_check, 1159 .ndo_change_proto_down = dev_change_proto_down_generic, 1160 }; 1161 1162 void macvlan_common_setup(struct net_device *dev) 1163 { 1164 ether_setup(dev); 1165 1166 dev->min_mtu = 0; 1167 dev->max_mtu = ETH_MAX_MTU; 1168 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1169 netif_keep_dst(dev); 1170 dev->priv_flags |= IFF_UNICAST_FLT; 1171 dev->netdev_ops = &macvlan_netdev_ops; 1172 dev->needs_free_netdev = true; 1173 dev->header_ops = &macvlan_hard_header_ops; 1174 dev->ethtool_ops = &macvlan_ethtool_ops; 1175 } 1176 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1177 1178 static void macvlan_setup(struct net_device *dev) 1179 { 1180 macvlan_common_setup(dev); 1181 dev->priv_flags |= IFF_NO_QUEUE; 1182 } 1183 1184 static int macvlan_port_create(struct net_device *dev) 1185 { 1186 struct macvlan_port *port; 1187 unsigned int i; 1188 int err; 1189 1190 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1191 return -EINVAL; 1192 1193 if (netdev_is_rx_handler_busy(dev)) 1194 return -EBUSY; 1195 1196 port = kzalloc(sizeof(*port), GFP_KERNEL); 1197 if (port == NULL) 1198 return -ENOMEM; 1199 1200 port->dev = dev; 1201 ether_addr_copy(port->perm_addr, dev->dev_addr); 1202 INIT_LIST_HEAD(&port->vlans); 1203 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1204 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1205 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1206 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1207 1208 skb_queue_head_init(&port->bc_queue); 1209 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1210 1211 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1212 if (err) 1213 kfree(port); 1214 else 1215 dev->priv_flags |= IFF_MACVLAN_PORT; 1216 return err; 1217 } 1218 1219 static void macvlan_port_destroy(struct net_device *dev) 1220 { 1221 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1222 struct sk_buff *skb; 1223 1224 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1225 netdev_rx_handler_unregister(dev); 1226 1227 /* After this point, no packet can schedule bc_work anymore, 1228 * but we need to cancel it and purge left skbs if any. 1229 */ 1230 cancel_work_sync(&port->bc_work); 1231 1232 while ((skb = __skb_dequeue(&port->bc_queue))) { 1233 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1234 1235 if (src) 1236 dev_put(src->dev); 1237 1238 kfree_skb(skb); 1239 } 1240 1241 /* If the lower device address has been changed by passthru 1242 * macvlan, put it back. 1243 */ 1244 if (macvlan_passthru(port) && 1245 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1246 struct sockaddr sa; 1247 1248 sa.sa_family = port->dev->type; 1249 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1250 dev_set_mac_address(port->dev, &sa, NULL); 1251 } 1252 1253 kfree(port); 1254 } 1255 1256 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1257 struct netlink_ext_ack *extack) 1258 { 1259 if (tb[IFLA_ADDRESS]) { 1260 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1261 return -EINVAL; 1262 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1263 return -EADDRNOTAVAIL; 1264 } 1265 1266 if (!data) 1267 return 0; 1268 1269 if (data[IFLA_MACVLAN_FLAGS] && 1270 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1271 return -EINVAL; 1272 1273 if (data[IFLA_MACVLAN_MODE]) { 1274 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1275 case MACVLAN_MODE_PRIVATE: 1276 case MACVLAN_MODE_VEPA: 1277 case MACVLAN_MODE_BRIDGE: 1278 case MACVLAN_MODE_PASSTHRU: 1279 case MACVLAN_MODE_SOURCE: 1280 break; 1281 default: 1282 return -EINVAL; 1283 } 1284 } 1285 1286 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1287 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1288 case MACVLAN_MACADDR_ADD: 1289 case MACVLAN_MACADDR_DEL: 1290 case MACVLAN_MACADDR_FLUSH: 1291 case MACVLAN_MACADDR_SET: 1292 break; 1293 default: 1294 return -EINVAL; 1295 } 1296 } 1297 1298 if (data[IFLA_MACVLAN_MACADDR]) { 1299 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1300 return -EINVAL; 1301 1302 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1303 return -EADDRNOTAVAIL; 1304 } 1305 1306 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1307 return -EINVAL; 1308 1309 return 0; 1310 } 1311 1312 /** 1313 * reconfigure list of remote source mac address 1314 * (only for macvlan devices in source mode) 1315 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1316 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1317 */ 1318 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1319 struct nlattr *data[]) 1320 { 1321 char *addr = NULL; 1322 int ret, rem, len; 1323 struct nlattr *nla, *head; 1324 struct macvlan_source_entry *entry; 1325 1326 if (data[IFLA_MACVLAN_MACADDR]) 1327 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1328 1329 if (mode == MACVLAN_MACADDR_ADD) { 1330 if (!addr) 1331 return -EINVAL; 1332 1333 return macvlan_hash_add_source(vlan, addr); 1334 1335 } else if (mode == MACVLAN_MACADDR_DEL) { 1336 if (!addr) 1337 return -EINVAL; 1338 1339 entry = macvlan_hash_lookup_source(vlan, addr); 1340 if (entry) { 1341 macvlan_hash_del_source(entry); 1342 vlan->macaddr_count--; 1343 } 1344 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1345 macvlan_flush_sources(vlan->port, vlan); 1346 } else if (mode == MACVLAN_MACADDR_SET) { 1347 macvlan_flush_sources(vlan->port, vlan); 1348 1349 if (addr) { 1350 ret = macvlan_hash_add_source(vlan, addr); 1351 if (ret) 1352 return ret; 1353 } 1354 1355 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1356 return 0; 1357 1358 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1359 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1360 1361 nla_for_each_attr(nla, head, len, rem) { 1362 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1363 nla_len(nla) != ETH_ALEN) 1364 continue; 1365 1366 addr = nla_data(nla); 1367 ret = macvlan_hash_add_source(vlan, addr); 1368 if (ret) 1369 return ret; 1370 } 1371 } else { 1372 return -EINVAL; 1373 } 1374 1375 return 0; 1376 } 1377 1378 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1379 struct nlattr *tb[], struct nlattr *data[], 1380 struct netlink_ext_ack *extack) 1381 { 1382 struct macvlan_dev *vlan = netdev_priv(dev); 1383 struct macvlan_port *port; 1384 struct net_device *lowerdev; 1385 int err; 1386 int macmode; 1387 bool create = false; 1388 1389 if (!tb[IFLA_LINK]) 1390 return -EINVAL; 1391 1392 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1393 if (lowerdev == NULL) 1394 return -ENODEV; 1395 1396 /* When creating macvlans or macvtaps on top of other macvlans - use 1397 * the real device as the lowerdev. 1398 */ 1399 if (netif_is_macvlan(lowerdev)) 1400 lowerdev = macvlan_dev_real_dev(lowerdev); 1401 1402 if (!tb[IFLA_MTU]) 1403 dev->mtu = lowerdev->mtu; 1404 else if (dev->mtu > lowerdev->mtu) 1405 return -EINVAL; 1406 1407 /* MTU range: 68 - lowerdev->max_mtu */ 1408 dev->min_mtu = ETH_MIN_MTU; 1409 dev->max_mtu = lowerdev->max_mtu; 1410 1411 if (!tb[IFLA_ADDRESS]) 1412 eth_hw_addr_random(dev); 1413 1414 if (!netif_is_macvlan_port(lowerdev)) { 1415 err = macvlan_port_create(lowerdev); 1416 if (err < 0) 1417 return err; 1418 create = true; 1419 } 1420 port = macvlan_port_get_rtnl(lowerdev); 1421 1422 /* Only 1 macvlan device can be created in passthru mode */ 1423 if (macvlan_passthru(port)) { 1424 /* The macvlan port must be not created this time, 1425 * still goto destroy_macvlan_port for readability. 1426 */ 1427 err = -EINVAL; 1428 goto destroy_macvlan_port; 1429 } 1430 1431 vlan->lowerdev = lowerdev; 1432 vlan->dev = dev; 1433 vlan->port = port; 1434 vlan->set_features = MACVLAN_FEATURES; 1435 1436 vlan->mode = MACVLAN_MODE_VEPA; 1437 if (data && data[IFLA_MACVLAN_MODE]) 1438 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1439 1440 if (data && data[IFLA_MACVLAN_FLAGS]) 1441 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1442 1443 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1444 if (port->count) { 1445 err = -EINVAL; 1446 goto destroy_macvlan_port; 1447 } 1448 macvlan_set_passthru(port); 1449 eth_hw_addr_inherit(dev, lowerdev); 1450 } 1451 1452 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1453 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1454 err = -EINVAL; 1455 goto destroy_macvlan_port; 1456 } 1457 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1458 err = macvlan_changelink_sources(vlan, macmode, data); 1459 if (err) 1460 goto destroy_macvlan_port; 1461 } 1462 1463 err = register_netdevice(dev); 1464 if (err < 0) 1465 goto destroy_macvlan_port; 1466 1467 dev->priv_flags |= IFF_MACVLAN; 1468 err = netdev_upper_dev_link(lowerdev, dev, extack); 1469 if (err) 1470 goto unregister_netdev; 1471 1472 list_add_tail_rcu(&vlan->list, &port->vlans); 1473 netif_stacked_transfer_operstate(lowerdev, dev); 1474 linkwatch_fire_event(dev); 1475 1476 return 0; 1477 1478 unregister_netdev: 1479 /* macvlan_uninit would free the macvlan port */ 1480 unregister_netdevice(dev); 1481 return err; 1482 destroy_macvlan_port: 1483 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1484 * so we destroy the macvlan port only when it's valid. 1485 */ 1486 if (create && macvlan_port_get_rtnl(lowerdev)) 1487 macvlan_port_destroy(port->dev); 1488 return err; 1489 } 1490 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1491 1492 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1493 struct nlattr *tb[], struct nlattr *data[], 1494 struct netlink_ext_ack *extack) 1495 { 1496 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1497 } 1498 1499 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1500 { 1501 struct macvlan_dev *vlan = netdev_priv(dev); 1502 1503 if (vlan->mode == MACVLAN_MODE_SOURCE) 1504 macvlan_flush_sources(vlan->port, vlan); 1505 list_del_rcu(&vlan->list); 1506 unregister_netdevice_queue(dev, head); 1507 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1508 } 1509 EXPORT_SYMBOL_GPL(macvlan_dellink); 1510 1511 static int macvlan_changelink(struct net_device *dev, 1512 struct nlattr *tb[], struct nlattr *data[], 1513 struct netlink_ext_ack *extack) 1514 { 1515 struct macvlan_dev *vlan = netdev_priv(dev); 1516 enum macvlan_mode mode; 1517 bool set_mode = false; 1518 enum macvlan_macaddr_mode macmode; 1519 int ret; 1520 1521 /* Validate mode, but don't set yet: setting flags may fail. */ 1522 if (data && data[IFLA_MACVLAN_MODE]) { 1523 set_mode = true; 1524 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1525 /* Passthrough mode can't be set or cleared dynamically */ 1526 if ((mode == MACVLAN_MODE_PASSTHRU) != 1527 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1528 return -EINVAL; 1529 if (vlan->mode == MACVLAN_MODE_SOURCE && 1530 vlan->mode != mode) 1531 macvlan_flush_sources(vlan->port, vlan); 1532 } 1533 1534 if (data && data[IFLA_MACVLAN_FLAGS]) { 1535 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1536 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1537 if (macvlan_passthru(vlan->port) && promisc) { 1538 int err; 1539 1540 if (flags & MACVLAN_FLAG_NOPROMISC) 1541 err = dev_set_promiscuity(vlan->lowerdev, -1); 1542 else 1543 err = dev_set_promiscuity(vlan->lowerdev, 1); 1544 if (err < 0) 1545 return err; 1546 } 1547 vlan->flags = flags; 1548 } 1549 if (set_mode) 1550 vlan->mode = mode; 1551 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1552 if (vlan->mode != MACVLAN_MODE_SOURCE) 1553 return -EINVAL; 1554 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1555 ret = macvlan_changelink_sources(vlan, macmode, data); 1556 if (ret) 1557 return ret; 1558 } 1559 return 0; 1560 } 1561 1562 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1563 { 1564 if (vlan->macaddr_count == 0) 1565 return 0; 1566 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1567 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1568 } 1569 1570 static size_t macvlan_get_size(const struct net_device *dev) 1571 { 1572 struct macvlan_dev *vlan = netdev_priv(dev); 1573 1574 return (0 1575 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1576 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1577 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1578 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1579 ); 1580 } 1581 1582 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1583 const struct macvlan_dev *vlan, 1584 const int i) 1585 { 1586 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1587 struct macvlan_source_entry *entry; 1588 1589 hlist_for_each_entry_rcu(entry, h, hlist) { 1590 if (entry->vlan != vlan) 1591 continue; 1592 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1593 return 1; 1594 } 1595 return 0; 1596 } 1597 1598 static int macvlan_fill_info(struct sk_buff *skb, 1599 const struct net_device *dev) 1600 { 1601 struct macvlan_dev *vlan = netdev_priv(dev); 1602 int i; 1603 struct nlattr *nest; 1604 1605 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1606 goto nla_put_failure; 1607 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1608 goto nla_put_failure; 1609 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1610 goto nla_put_failure; 1611 if (vlan->macaddr_count > 0) { 1612 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1613 if (nest == NULL) 1614 goto nla_put_failure; 1615 1616 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1617 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1618 goto nla_put_failure; 1619 } 1620 nla_nest_end(skb, nest); 1621 } 1622 return 0; 1623 1624 nla_put_failure: 1625 return -EMSGSIZE; 1626 } 1627 1628 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1629 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1630 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1631 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1632 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1633 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1634 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1635 }; 1636 1637 int macvlan_link_register(struct rtnl_link_ops *ops) 1638 { 1639 /* common fields */ 1640 ops->validate = macvlan_validate; 1641 ops->maxtype = IFLA_MACVLAN_MAX; 1642 ops->policy = macvlan_policy; 1643 ops->changelink = macvlan_changelink; 1644 ops->get_size = macvlan_get_size; 1645 ops->fill_info = macvlan_fill_info; 1646 1647 return rtnl_link_register(ops); 1648 }; 1649 EXPORT_SYMBOL_GPL(macvlan_link_register); 1650 1651 static struct net *macvlan_get_link_net(const struct net_device *dev) 1652 { 1653 return dev_net(macvlan_dev_real_dev(dev)); 1654 } 1655 1656 static struct rtnl_link_ops macvlan_link_ops = { 1657 .kind = "macvlan", 1658 .setup = macvlan_setup, 1659 .newlink = macvlan_newlink, 1660 .dellink = macvlan_dellink, 1661 .get_link_net = macvlan_get_link_net, 1662 .priv_size = sizeof(struct macvlan_dev), 1663 }; 1664 1665 static int macvlan_device_event(struct notifier_block *unused, 1666 unsigned long event, void *ptr) 1667 { 1668 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1669 struct macvlan_dev *vlan, *next; 1670 struct macvlan_port *port; 1671 LIST_HEAD(list_kill); 1672 1673 if (!netif_is_macvlan_port(dev)) 1674 return NOTIFY_DONE; 1675 1676 port = macvlan_port_get_rtnl(dev); 1677 1678 switch (event) { 1679 case NETDEV_UP: 1680 case NETDEV_DOWN: 1681 case NETDEV_CHANGE: 1682 list_for_each_entry(vlan, &port->vlans, list) 1683 netif_stacked_transfer_operstate(vlan->lowerdev, 1684 vlan->dev); 1685 break; 1686 case NETDEV_FEAT_CHANGE: 1687 list_for_each_entry(vlan, &port->vlans, list) { 1688 vlan->dev->gso_max_size = dev->gso_max_size; 1689 vlan->dev->gso_max_segs = dev->gso_max_segs; 1690 netdev_update_features(vlan->dev); 1691 } 1692 break; 1693 case NETDEV_CHANGEMTU: 1694 list_for_each_entry(vlan, &port->vlans, list) { 1695 if (vlan->dev->mtu <= dev->mtu) 1696 continue; 1697 dev_set_mtu(vlan->dev, dev->mtu); 1698 } 1699 break; 1700 case NETDEV_CHANGEADDR: 1701 if (!macvlan_passthru(port)) 1702 return NOTIFY_DONE; 1703 1704 vlan = list_first_entry_or_null(&port->vlans, 1705 struct macvlan_dev, 1706 list); 1707 1708 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1709 return NOTIFY_BAD; 1710 1711 break; 1712 case NETDEV_UNREGISTER: 1713 /* twiddle thumbs on netns device moves */ 1714 if (dev->reg_state != NETREG_UNREGISTERING) 1715 break; 1716 1717 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1718 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1719 unregister_netdevice_many(&list_kill); 1720 break; 1721 case NETDEV_PRE_TYPE_CHANGE: 1722 /* Forbid underlaying device to change its type. */ 1723 return NOTIFY_BAD; 1724 1725 case NETDEV_NOTIFY_PEERS: 1726 case NETDEV_BONDING_FAILOVER: 1727 case NETDEV_RESEND_IGMP: 1728 /* Propagate to all vlans */ 1729 list_for_each_entry(vlan, &port->vlans, list) 1730 call_netdevice_notifiers(event, vlan->dev); 1731 } 1732 return NOTIFY_DONE; 1733 } 1734 1735 static struct notifier_block macvlan_notifier_block __read_mostly = { 1736 .notifier_call = macvlan_device_event, 1737 }; 1738 1739 static int __init macvlan_init_module(void) 1740 { 1741 int err; 1742 1743 register_netdevice_notifier(&macvlan_notifier_block); 1744 1745 err = macvlan_link_register(&macvlan_link_ops); 1746 if (err < 0) 1747 goto err1; 1748 return 0; 1749 err1: 1750 unregister_netdevice_notifier(&macvlan_notifier_block); 1751 return err; 1752 } 1753 1754 static void __exit macvlan_cleanup_module(void) 1755 { 1756 rtnl_link_unregister(&macvlan_link_ops); 1757 unregister_netdevice_notifier(&macvlan_notifier_block); 1758 } 1759 1760 module_init(macvlan_init_module); 1761 module_exit(macvlan_cleanup_module); 1762 1763 MODULE_LICENSE("GPL"); 1764 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1765 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1766 MODULE_ALIAS_RTNL_LINK("macvlan"); 1767