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 /* Packets from dev_loopback_xmit() do not have L2 header, bail out */ 452 if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) 453 return RX_HANDLER_PASS; 454 455 port = macvlan_port_get_rcu(skb->dev); 456 if (is_multicast_ether_addr(eth->h_dest)) { 457 unsigned int hash; 458 459 skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN); 460 if (!skb) 461 return RX_HANDLER_CONSUMED; 462 *pskb = skb; 463 eth = eth_hdr(skb); 464 macvlan_forward_source(skb, port, eth->h_source); 465 src = macvlan_hash_lookup(port, eth->h_source); 466 if (src && src->mode != MACVLAN_MODE_VEPA && 467 src->mode != MACVLAN_MODE_BRIDGE) { 468 /* forward to original port. */ 469 vlan = src; 470 ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?: 471 netif_rx(skb); 472 handle_res = RX_HANDLER_CONSUMED; 473 goto out; 474 } 475 476 hash = mc_hash(NULL, eth->h_dest); 477 if (test_bit(hash, port->mc_filter)) 478 macvlan_broadcast_enqueue(port, src, skb); 479 480 return RX_HANDLER_PASS; 481 } 482 483 macvlan_forward_source(skb, port, eth->h_source); 484 if (macvlan_passthru(port)) 485 vlan = list_first_or_null_rcu(&port->vlans, 486 struct macvlan_dev, list); 487 else 488 vlan = macvlan_hash_lookup(port, eth->h_dest); 489 if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE) 490 return RX_HANDLER_PASS; 491 492 dev = vlan->dev; 493 if (unlikely(!(dev->flags & IFF_UP))) { 494 kfree_skb(skb); 495 return RX_HANDLER_CONSUMED; 496 } 497 len = skb->len + ETH_HLEN; 498 skb = skb_share_check(skb, GFP_ATOMIC); 499 if (!skb) { 500 ret = NET_RX_DROP; 501 handle_res = RX_HANDLER_CONSUMED; 502 goto out; 503 } 504 505 *pskb = skb; 506 skb->dev = dev; 507 skb->pkt_type = PACKET_HOST; 508 509 ret = NET_RX_SUCCESS; 510 handle_res = RX_HANDLER_ANOTHER; 511 out: 512 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false); 513 return handle_res; 514 } 515 516 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev) 517 { 518 const struct macvlan_dev *vlan = netdev_priv(dev); 519 const struct macvlan_port *port = vlan->port; 520 const struct macvlan_dev *dest; 521 522 if (vlan->mode == MACVLAN_MODE_BRIDGE) { 523 const struct ethhdr *eth = skb_eth_hdr(skb); 524 525 /* send to other bridge ports directly */ 526 if (is_multicast_ether_addr(eth->h_dest)) { 527 skb_reset_mac_header(skb); 528 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 529 goto xmit_world; 530 } 531 532 dest = macvlan_hash_lookup(port, eth->h_dest); 533 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 534 /* send to lowerdev first for its network taps */ 535 dev_forward_skb(vlan->lowerdev, skb); 536 537 return NET_XMIT_SUCCESS; 538 } 539 } 540 xmit_world: 541 skb->dev = vlan->lowerdev; 542 return dev_queue_xmit_accel(skb, 543 netdev_get_sb_channel(dev) ? dev : NULL); 544 } 545 546 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 547 { 548 #ifdef CONFIG_NET_POLL_CONTROLLER 549 return netpoll_send_skb(vlan->netpoll, skb); 550 #else 551 BUG(); 552 return NETDEV_TX_OK; 553 #endif 554 } 555 556 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 557 struct net_device *dev) 558 { 559 struct macvlan_dev *vlan = netdev_priv(dev); 560 unsigned int len = skb->len; 561 int ret; 562 563 if (unlikely(netpoll_tx_running(dev))) 564 return macvlan_netpoll_send_skb(vlan, skb); 565 566 ret = macvlan_queue_xmit(skb, dev); 567 568 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 569 struct vlan_pcpu_stats *pcpu_stats; 570 571 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 572 u64_stats_update_begin(&pcpu_stats->syncp); 573 pcpu_stats->tx_packets++; 574 pcpu_stats->tx_bytes += len; 575 u64_stats_update_end(&pcpu_stats->syncp); 576 } else { 577 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 578 } 579 return ret; 580 } 581 582 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 583 unsigned short type, const void *daddr, 584 const void *saddr, unsigned len) 585 { 586 const struct macvlan_dev *vlan = netdev_priv(dev); 587 struct net_device *lowerdev = vlan->lowerdev; 588 589 return dev_hard_header(skb, lowerdev, type, daddr, 590 saddr ? : dev->dev_addr, len); 591 } 592 593 static const struct header_ops macvlan_hard_header_ops = { 594 .create = macvlan_hard_header, 595 .parse = eth_header_parse, 596 .cache = eth_header_cache, 597 .cache_update = eth_header_cache_update, 598 }; 599 600 static int macvlan_open(struct net_device *dev) 601 { 602 struct macvlan_dev *vlan = netdev_priv(dev); 603 struct net_device *lowerdev = vlan->lowerdev; 604 int err; 605 606 if (macvlan_passthru(vlan->port)) { 607 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 608 err = dev_set_promiscuity(lowerdev, 1); 609 if (err < 0) 610 goto out; 611 } 612 goto hash_add; 613 } 614 615 err = -EADDRINUSE; 616 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 617 goto out; 618 619 /* Attempt to populate accel_priv which is used to offload the L2 620 * forwarding requests for unicast packets. 621 */ 622 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 623 vlan->accel_priv = 624 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 625 626 /* If earlier attempt to offload failed, or accel_priv is not 627 * populated we must add the unicast address to the lower device. 628 */ 629 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 630 vlan->accel_priv = NULL; 631 err = dev_uc_add(lowerdev, dev->dev_addr); 632 if (err < 0) 633 goto out; 634 } 635 636 if (dev->flags & IFF_ALLMULTI) { 637 err = dev_set_allmulti(lowerdev, 1); 638 if (err < 0) 639 goto del_unicast; 640 } 641 642 if (dev->flags & IFF_PROMISC) { 643 err = dev_set_promiscuity(lowerdev, 1); 644 if (err < 0) 645 goto clear_multi; 646 } 647 648 hash_add: 649 macvlan_hash_add(vlan); 650 return 0; 651 652 clear_multi: 653 if (dev->flags & IFF_ALLMULTI) 654 dev_set_allmulti(lowerdev, -1); 655 del_unicast: 656 if (vlan->accel_priv) { 657 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 658 vlan->accel_priv); 659 vlan->accel_priv = NULL; 660 } else { 661 dev_uc_del(lowerdev, dev->dev_addr); 662 } 663 out: 664 return err; 665 } 666 667 static int macvlan_stop(struct net_device *dev) 668 { 669 struct macvlan_dev *vlan = netdev_priv(dev); 670 struct net_device *lowerdev = vlan->lowerdev; 671 672 if (vlan->accel_priv) { 673 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 674 vlan->accel_priv); 675 vlan->accel_priv = NULL; 676 } 677 678 dev_uc_unsync(lowerdev, dev); 679 dev_mc_unsync(lowerdev, dev); 680 681 if (macvlan_passthru(vlan->port)) { 682 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 683 dev_set_promiscuity(lowerdev, -1); 684 goto hash_del; 685 } 686 687 if (dev->flags & IFF_ALLMULTI) 688 dev_set_allmulti(lowerdev, -1); 689 690 if (dev->flags & IFF_PROMISC) 691 dev_set_promiscuity(lowerdev, -1); 692 693 dev_uc_del(lowerdev, dev->dev_addr); 694 695 hash_del: 696 macvlan_hash_del(vlan, !dev->dismantle); 697 return 0; 698 } 699 700 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 701 { 702 struct macvlan_dev *vlan = netdev_priv(dev); 703 struct net_device *lowerdev = vlan->lowerdev; 704 struct macvlan_port *port = vlan->port; 705 int err; 706 707 if (!(dev->flags & IFF_UP)) { 708 /* Just copy in the new address */ 709 ether_addr_copy(dev->dev_addr, addr); 710 } else { 711 /* Rehash and update the device filters */ 712 if (macvlan_addr_busy(vlan->port, addr)) 713 return -EADDRINUSE; 714 715 if (!macvlan_passthru(port)) { 716 err = dev_uc_add(lowerdev, addr); 717 if (err) 718 return err; 719 720 dev_uc_del(lowerdev, dev->dev_addr); 721 } 722 723 macvlan_hash_change_addr(vlan, addr); 724 } 725 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 726 /* Since addr_change isn't set, we are here due to lower 727 * device change. Save the lower-dev address so we can 728 * restore it later. 729 */ 730 ether_addr_copy(vlan->port->perm_addr, 731 lowerdev->dev_addr); 732 } 733 macvlan_clear_addr_change(port); 734 return 0; 735 } 736 737 static int macvlan_set_mac_address(struct net_device *dev, void *p) 738 { 739 struct macvlan_dev *vlan = netdev_priv(dev); 740 struct sockaddr *addr = p; 741 742 if (!is_valid_ether_addr(addr->sa_data)) 743 return -EADDRNOTAVAIL; 744 745 /* If the addresses are the same, this is a no-op */ 746 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 747 return 0; 748 749 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 750 macvlan_set_addr_change(vlan->port); 751 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 752 } 753 754 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 755 return -EADDRINUSE; 756 757 return macvlan_sync_address(dev, addr->sa_data); 758 } 759 760 static void macvlan_change_rx_flags(struct net_device *dev, int change) 761 { 762 struct macvlan_dev *vlan = netdev_priv(dev); 763 struct net_device *lowerdev = vlan->lowerdev; 764 765 if (dev->flags & IFF_UP) { 766 if (change & IFF_ALLMULTI) 767 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 768 if (change & IFF_PROMISC) 769 dev_set_promiscuity(lowerdev, 770 dev->flags & IFF_PROMISC ? 1 : -1); 771 772 } 773 } 774 775 static void macvlan_compute_filter(unsigned long *mc_filter, 776 struct net_device *dev, 777 struct macvlan_dev *vlan) 778 { 779 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 780 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 781 } else { 782 struct netdev_hw_addr *ha; 783 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 784 785 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 786 netdev_for_each_mc_addr(ha, dev) { 787 __set_bit(mc_hash(vlan, ha->addr), filter); 788 } 789 790 __set_bit(mc_hash(vlan, dev->broadcast), filter); 791 792 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 793 } 794 } 795 796 static void macvlan_set_mac_lists(struct net_device *dev) 797 { 798 struct macvlan_dev *vlan = netdev_priv(dev); 799 800 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 801 802 dev_uc_sync(vlan->lowerdev, dev); 803 dev_mc_sync(vlan->lowerdev, dev); 804 805 /* This is slightly inaccurate as we're including the subscription 806 * list of vlan->lowerdev too. 807 * 808 * Bug alert: This only works if everyone has the same broadcast 809 * address as lowerdev. As soon as someone changes theirs this 810 * will break. 811 * 812 * However, this is already broken as when you change your broadcast 813 * address we don't get called. 814 * 815 * The solution is to maintain a list of broadcast addresses like 816 * we do for uc/mc, if you care. 817 */ 818 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 819 } 820 821 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 822 { 823 struct macvlan_dev *vlan = netdev_priv(dev); 824 825 if (vlan->lowerdev->mtu < new_mtu) 826 return -EINVAL; 827 dev->mtu = new_mtu; 828 return 0; 829 } 830 831 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 832 { 833 struct net_device *real_dev = macvlan_dev_real_dev(dev); 834 const struct net_device_ops *ops = real_dev->netdev_ops; 835 struct ifreq ifrr; 836 int err = -EOPNOTSUPP; 837 838 strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ); 839 ifrr.ifr_ifru = ifr->ifr_ifru; 840 841 switch (cmd) { 842 case SIOCSHWTSTAMP: 843 if (!net_eq(dev_net(dev), &init_net)) 844 break; 845 /* fall through */ 846 case SIOCGHWTSTAMP: 847 if (netif_device_present(real_dev) && ops->ndo_do_ioctl) 848 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd); 849 break; 850 } 851 852 if (!err) 853 ifr->ifr_ifru = ifrr.ifr_ifru; 854 855 return err; 856 } 857 858 /* 859 * macvlan network devices have devices nesting below it and are a special 860 * "super class" of normal network devices; split their locks off into a 861 * separate class since they always nest. 862 */ 863 static struct lock_class_key macvlan_netdev_addr_lock_key; 864 865 #define ALWAYS_ON_OFFLOADS \ 866 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 867 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 868 869 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 870 871 #define MACVLAN_FEATURES \ 872 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 873 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 874 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 875 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 876 877 #define MACVLAN_STATE_MASK \ 878 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 879 880 static void macvlan_set_lockdep_class(struct net_device *dev) 881 { 882 netdev_lockdep_set_classes(dev); 883 lockdep_set_class_and_subclass(&dev->addr_list_lock, 884 &macvlan_netdev_addr_lock_key, 885 dev->lower_level); 886 } 887 888 static int macvlan_init(struct net_device *dev) 889 { 890 struct macvlan_dev *vlan = netdev_priv(dev); 891 const struct net_device *lowerdev = vlan->lowerdev; 892 struct macvlan_port *port = vlan->port; 893 894 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 895 (lowerdev->state & MACVLAN_STATE_MASK); 896 dev->features = lowerdev->features & MACVLAN_FEATURES; 897 dev->features |= ALWAYS_ON_FEATURES; 898 dev->hw_features |= NETIF_F_LRO; 899 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 900 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 901 dev->hw_enc_features |= dev->features; 902 dev->gso_max_size = lowerdev->gso_max_size; 903 dev->gso_max_segs = lowerdev->gso_max_segs; 904 dev->hard_header_len = lowerdev->hard_header_len; 905 macvlan_set_lockdep_class(dev); 906 907 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 908 if (!vlan->pcpu_stats) 909 return -ENOMEM; 910 911 port->count += 1; 912 913 return 0; 914 } 915 916 static void macvlan_uninit(struct net_device *dev) 917 { 918 struct macvlan_dev *vlan = netdev_priv(dev); 919 struct macvlan_port *port = vlan->port; 920 921 free_percpu(vlan->pcpu_stats); 922 923 macvlan_flush_sources(port, vlan); 924 port->count -= 1; 925 if (!port->count) 926 macvlan_port_destroy(port->dev); 927 } 928 929 static void macvlan_dev_get_stats64(struct net_device *dev, 930 struct rtnl_link_stats64 *stats) 931 { 932 struct macvlan_dev *vlan = netdev_priv(dev); 933 934 if (vlan->pcpu_stats) { 935 struct vlan_pcpu_stats *p; 936 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 937 u32 rx_errors = 0, tx_dropped = 0; 938 unsigned int start; 939 int i; 940 941 for_each_possible_cpu(i) { 942 p = per_cpu_ptr(vlan->pcpu_stats, i); 943 do { 944 start = u64_stats_fetch_begin_irq(&p->syncp); 945 rx_packets = p->rx_packets; 946 rx_bytes = p->rx_bytes; 947 rx_multicast = p->rx_multicast; 948 tx_packets = p->tx_packets; 949 tx_bytes = p->tx_bytes; 950 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 951 952 stats->rx_packets += rx_packets; 953 stats->rx_bytes += rx_bytes; 954 stats->multicast += rx_multicast; 955 stats->tx_packets += tx_packets; 956 stats->tx_bytes += tx_bytes; 957 /* rx_errors & tx_dropped are u32, updated 958 * without syncp protection. 959 */ 960 rx_errors += p->rx_errors; 961 tx_dropped += p->tx_dropped; 962 } 963 stats->rx_errors = rx_errors; 964 stats->rx_dropped = rx_errors; 965 stats->tx_dropped = tx_dropped; 966 } 967 } 968 969 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 970 __be16 proto, u16 vid) 971 { 972 struct macvlan_dev *vlan = netdev_priv(dev); 973 struct net_device *lowerdev = vlan->lowerdev; 974 975 return vlan_vid_add(lowerdev, proto, vid); 976 } 977 978 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 979 __be16 proto, u16 vid) 980 { 981 struct macvlan_dev *vlan = netdev_priv(dev); 982 struct net_device *lowerdev = vlan->lowerdev; 983 984 vlan_vid_del(lowerdev, proto, vid); 985 return 0; 986 } 987 988 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 989 struct net_device *dev, 990 const unsigned char *addr, u16 vid, 991 u16 flags, 992 struct netlink_ext_ack *extack) 993 { 994 struct macvlan_dev *vlan = netdev_priv(dev); 995 int err = -EINVAL; 996 997 /* Support unicast filter only on passthru devices. 998 * Multicast filter should be allowed on all devices. 999 */ 1000 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1001 return -EOPNOTSUPP; 1002 1003 if (flags & NLM_F_REPLACE) 1004 return -EOPNOTSUPP; 1005 1006 if (is_unicast_ether_addr(addr)) 1007 err = dev_uc_add_excl(dev, addr); 1008 else if (is_multicast_ether_addr(addr)) 1009 err = dev_mc_add_excl(dev, addr); 1010 1011 return err; 1012 } 1013 1014 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1015 struct net_device *dev, 1016 const unsigned char *addr, u16 vid) 1017 { 1018 struct macvlan_dev *vlan = netdev_priv(dev); 1019 int err = -EINVAL; 1020 1021 /* Support unicast filter only on passthru devices. 1022 * Multicast filter should be allowed on all devices. 1023 */ 1024 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1025 return -EOPNOTSUPP; 1026 1027 if (is_unicast_ether_addr(addr)) 1028 err = dev_uc_del(dev, addr); 1029 else if (is_multicast_ether_addr(addr)) 1030 err = dev_mc_del(dev, addr); 1031 1032 return err; 1033 } 1034 1035 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1036 struct ethtool_drvinfo *drvinfo) 1037 { 1038 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1039 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1040 } 1041 1042 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1043 struct ethtool_link_ksettings *cmd) 1044 { 1045 const struct macvlan_dev *vlan = netdev_priv(dev); 1046 1047 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1048 } 1049 1050 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1051 struct ethtool_ts_info *info) 1052 { 1053 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1054 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1055 struct phy_device *phydev = real_dev->phydev; 1056 1057 if (phy_has_tsinfo(phydev)) { 1058 return phy_ts_info(phydev, info); 1059 } else if (ops->get_ts_info) { 1060 return ops->get_ts_info(real_dev, info); 1061 } else { 1062 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1063 SOF_TIMESTAMPING_SOFTWARE; 1064 info->phc_index = -1; 1065 } 1066 1067 return 0; 1068 } 1069 1070 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1071 netdev_features_t features) 1072 { 1073 struct macvlan_dev *vlan = netdev_priv(dev); 1074 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1075 netdev_features_t mask; 1076 1077 features |= NETIF_F_ALL_FOR_ALL; 1078 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1079 mask = features; 1080 1081 lowerdev_features &= (features | ~NETIF_F_LRO); 1082 features = netdev_increment_features(lowerdev_features, features, mask); 1083 features |= ALWAYS_ON_FEATURES; 1084 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1085 1086 return features; 1087 } 1088 1089 #ifdef CONFIG_NET_POLL_CONTROLLER 1090 static void macvlan_dev_poll_controller(struct net_device *dev) 1091 { 1092 return; 1093 } 1094 1095 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1096 { 1097 struct macvlan_dev *vlan = netdev_priv(dev); 1098 struct net_device *real_dev = vlan->lowerdev; 1099 struct netpoll *netpoll; 1100 int err = 0; 1101 1102 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1103 err = -ENOMEM; 1104 if (!netpoll) 1105 goto out; 1106 1107 err = __netpoll_setup(netpoll, real_dev); 1108 if (err) { 1109 kfree(netpoll); 1110 goto out; 1111 } 1112 1113 vlan->netpoll = netpoll; 1114 1115 out: 1116 return err; 1117 } 1118 1119 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1120 { 1121 struct macvlan_dev *vlan = netdev_priv(dev); 1122 struct netpoll *netpoll = vlan->netpoll; 1123 1124 if (!netpoll) 1125 return; 1126 1127 vlan->netpoll = NULL; 1128 1129 __netpoll_free(netpoll); 1130 } 1131 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1132 1133 static int macvlan_dev_get_iflink(const struct net_device *dev) 1134 { 1135 struct macvlan_dev *vlan = netdev_priv(dev); 1136 1137 return vlan->lowerdev->ifindex; 1138 } 1139 1140 static const struct ethtool_ops macvlan_ethtool_ops = { 1141 .get_link = ethtool_op_get_link, 1142 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1143 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1144 .get_ts_info = macvlan_ethtool_get_ts_info, 1145 }; 1146 1147 static const struct net_device_ops macvlan_netdev_ops = { 1148 .ndo_init = macvlan_init, 1149 .ndo_uninit = macvlan_uninit, 1150 .ndo_open = macvlan_open, 1151 .ndo_stop = macvlan_stop, 1152 .ndo_start_xmit = macvlan_start_xmit, 1153 .ndo_change_mtu = macvlan_change_mtu, 1154 .ndo_do_ioctl = macvlan_do_ioctl, 1155 .ndo_fix_features = macvlan_fix_features, 1156 .ndo_change_rx_flags = macvlan_change_rx_flags, 1157 .ndo_set_mac_address = macvlan_set_mac_address, 1158 .ndo_set_rx_mode = macvlan_set_mac_lists, 1159 .ndo_get_stats64 = macvlan_dev_get_stats64, 1160 .ndo_validate_addr = eth_validate_addr, 1161 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1162 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1163 .ndo_fdb_add = macvlan_fdb_add, 1164 .ndo_fdb_del = macvlan_fdb_del, 1165 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1166 #ifdef CONFIG_NET_POLL_CONTROLLER 1167 .ndo_poll_controller = macvlan_dev_poll_controller, 1168 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1169 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1170 #endif 1171 .ndo_get_iflink = macvlan_dev_get_iflink, 1172 .ndo_features_check = passthru_features_check, 1173 .ndo_change_proto_down = dev_change_proto_down_generic, 1174 }; 1175 1176 void macvlan_common_setup(struct net_device *dev) 1177 { 1178 ether_setup(dev); 1179 1180 dev->min_mtu = 0; 1181 dev->max_mtu = ETH_MAX_MTU; 1182 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1183 netif_keep_dst(dev); 1184 dev->priv_flags |= IFF_UNICAST_FLT; 1185 dev->netdev_ops = &macvlan_netdev_ops; 1186 dev->needs_free_netdev = true; 1187 dev->header_ops = &macvlan_hard_header_ops; 1188 dev->ethtool_ops = &macvlan_ethtool_ops; 1189 } 1190 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1191 1192 static void macvlan_setup(struct net_device *dev) 1193 { 1194 macvlan_common_setup(dev); 1195 dev->priv_flags |= IFF_NO_QUEUE; 1196 } 1197 1198 static int macvlan_port_create(struct net_device *dev) 1199 { 1200 struct macvlan_port *port; 1201 unsigned int i; 1202 int err; 1203 1204 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1205 return -EINVAL; 1206 1207 if (netdev_is_rx_handler_busy(dev)) 1208 return -EBUSY; 1209 1210 port = kzalloc(sizeof(*port), GFP_KERNEL); 1211 if (port == NULL) 1212 return -ENOMEM; 1213 1214 port->dev = dev; 1215 ether_addr_copy(port->perm_addr, dev->dev_addr); 1216 INIT_LIST_HEAD(&port->vlans); 1217 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1218 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1219 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1220 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1221 1222 skb_queue_head_init(&port->bc_queue); 1223 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1224 1225 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1226 if (err) 1227 kfree(port); 1228 else 1229 dev->priv_flags |= IFF_MACVLAN_PORT; 1230 return err; 1231 } 1232 1233 static void macvlan_port_destroy(struct net_device *dev) 1234 { 1235 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1236 struct sk_buff *skb; 1237 1238 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1239 netdev_rx_handler_unregister(dev); 1240 1241 /* After this point, no packet can schedule bc_work anymore, 1242 * but we need to cancel it and purge left skbs if any. 1243 */ 1244 cancel_work_sync(&port->bc_work); 1245 1246 while ((skb = __skb_dequeue(&port->bc_queue))) { 1247 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1248 1249 if (src) 1250 dev_put(src->dev); 1251 1252 kfree_skb(skb); 1253 } 1254 1255 /* If the lower device address has been changed by passthru 1256 * macvlan, put it back. 1257 */ 1258 if (macvlan_passthru(port) && 1259 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1260 struct sockaddr sa; 1261 1262 sa.sa_family = port->dev->type; 1263 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1264 dev_set_mac_address(port->dev, &sa, NULL); 1265 } 1266 1267 kfree(port); 1268 } 1269 1270 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1271 struct netlink_ext_ack *extack) 1272 { 1273 if (tb[IFLA_ADDRESS]) { 1274 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1275 return -EINVAL; 1276 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1277 return -EADDRNOTAVAIL; 1278 } 1279 1280 if (!data) 1281 return 0; 1282 1283 if (data[IFLA_MACVLAN_FLAGS] && 1284 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1285 return -EINVAL; 1286 1287 if (data[IFLA_MACVLAN_MODE]) { 1288 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1289 case MACVLAN_MODE_PRIVATE: 1290 case MACVLAN_MODE_VEPA: 1291 case MACVLAN_MODE_BRIDGE: 1292 case MACVLAN_MODE_PASSTHRU: 1293 case MACVLAN_MODE_SOURCE: 1294 break; 1295 default: 1296 return -EINVAL; 1297 } 1298 } 1299 1300 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1301 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1302 case MACVLAN_MACADDR_ADD: 1303 case MACVLAN_MACADDR_DEL: 1304 case MACVLAN_MACADDR_FLUSH: 1305 case MACVLAN_MACADDR_SET: 1306 break; 1307 default: 1308 return -EINVAL; 1309 } 1310 } 1311 1312 if (data[IFLA_MACVLAN_MACADDR]) { 1313 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1314 return -EINVAL; 1315 1316 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1317 return -EADDRNOTAVAIL; 1318 } 1319 1320 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1321 return -EINVAL; 1322 1323 return 0; 1324 } 1325 1326 /** 1327 * reconfigure list of remote source mac address 1328 * (only for macvlan devices in source mode) 1329 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1330 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1331 */ 1332 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1333 struct nlattr *data[]) 1334 { 1335 char *addr = NULL; 1336 int ret, rem, len; 1337 struct nlattr *nla, *head; 1338 struct macvlan_source_entry *entry; 1339 1340 if (data[IFLA_MACVLAN_MACADDR]) 1341 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1342 1343 if (mode == MACVLAN_MACADDR_ADD) { 1344 if (!addr) 1345 return -EINVAL; 1346 1347 return macvlan_hash_add_source(vlan, addr); 1348 1349 } else if (mode == MACVLAN_MACADDR_DEL) { 1350 if (!addr) 1351 return -EINVAL; 1352 1353 entry = macvlan_hash_lookup_source(vlan, addr); 1354 if (entry) { 1355 macvlan_hash_del_source(entry); 1356 vlan->macaddr_count--; 1357 } 1358 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1359 macvlan_flush_sources(vlan->port, vlan); 1360 } else if (mode == MACVLAN_MACADDR_SET) { 1361 macvlan_flush_sources(vlan->port, vlan); 1362 1363 if (addr) { 1364 ret = macvlan_hash_add_source(vlan, addr); 1365 if (ret) 1366 return ret; 1367 } 1368 1369 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1370 return 0; 1371 1372 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1373 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1374 1375 nla_for_each_attr(nla, head, len, rem) { 1376 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1377 nla_len(nla) != ETH_ALEN) 1378 continue; 1379 1380 addr = nla_data(nla); 1381 ret = macvlan_hash_add_source(vlan, addr); 1382 if (ret) 1383 return ret; 1384 } 1385 } else { 1386 return -EINVAL; 1387 } 1388 1389 return 0; 1390 } 1391 1392 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1393 struct nlattr *tb[], struct nlattr *data[], 1394 struct netlink_ext_ack *extack) 1395 { 1396 struct macvlan_dev *vlan = netdev_priv(dev); 1397 struct macvlan_port *port; 1398 struct net_device *lowerdev; 1399 int err; 1400 int macmode; 1401 bool create = false; 1402 1403 if (!tb[IFLA_LINK]) 1404 return -EINVAL; 1405 1406 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1407 if (lowerdev == NULL) 1408 return -ENODEV; 1409 1410 /* When creating macvlans or macvtaps on top of other macvlans - use 1411 * the real device as the lowerdev. 1412 */ 1413 if (netif_is_macvlan(lowerdev)) 1414 lowerdev = macvlan_dev_real_dev(lowerdev); 1415 1416 if (!tb[IFLA_MTU]) 1417 dev->mtu = lowerdev->mtu; 1418 else if (dev->mtu > lowerdev->mtu) 1419 return -EINVAL; 1420 1421 /* MTU range: 68 - lowerdev->max_mtu */ 1422 dev->min_mtu = ETH_MIN_MTU; 1423 dev->max_mtu = lowerdev->max_mtu; 1424 1425 if (!tb[IFLA_ADDRESS]) 1426 eth_hw_addr_random(dev); 1427 1428 if (!netif_is_macvlan_port(lowerdev)) { 1429 err = macvlan_port_create(lowerdev); 1430 if (err < 0) 1431 return err; 1432 create = true; 1433 } 1434 port = macvlan_port_get_rtnl(lowerdev); 1435 1436 /* Only 1 macvlan device can be created in passthru mode */ 1437 if (macvlan_passthru(port)) { 1438 /* The macvlan port must be not created this time, 1439 * still goto destroy_macvlan_port for readability. 1440 */ 1441 err = -EINVAL; 1442 goto destroy_macvlan_port; 1443 } 1444 1445 vlan->lowerdev = lowerdev; 1446 vlan->dev = dev; 1447 vlan->port = port; 1448 vlan->set_features = MACVLAN_FEATURES; 1449 1450 vlan->mode = MACVLAN_MODE_VEPA; 1451 if (data && data[IFLA_MACVLAN_MODE]) 1452 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1453 1454 if (data && data[IFLA_MACVLAN_FLAGS]) 1455 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1456 1457 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1458 if (port->count) { 1459 err = -EINVAL; 1460 goto destroy_macvlan_port; 1461 } 1462 macvlan_set_passthru(port); 1463 eth_hw_addr_inherit(dev, lowerdev); 1464 } 1465 1466 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1467 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1468 err = -EINVAL; 1469 goto destroy_macvlan_port; 1470 } 1471 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1472 err = macvlan_changelink_sources(vlan, macmode, data); 1473 if (err) 1474 goto destroy_macvlan_port; 1475 } 1476 1477 err = register_netdevice(dev); 1478 if (err < 0) 1479 goto destroy_macvlan_port; 1480 1481 dev->priv_flags |= IFF_MACVLAN; 1482 err = netdev_upper_dev_link(lowerdev, dev, extack); 1483 if (err) 1484 goto unregister_netdev; 1485 1486 list_add_tail_rcu(&vlan->list, &port->vlans); 1487 netif_stacked_transfer_operstate(lowerdev, dev); 1488 linkwatch_fire_event(dev); 1489 1490 return 0; 1491 1492 unregister_netdev: 1493 /* macvlan_uninit would free the macvlan port */ 1494 unregister_netdevice(dev); 1495 return err; 1496 destroy_macvlan_port: 1497 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1498 * so we destroy the macvlan port only when it's valid. 1499 */ 1500 if (create && macvlan_port_get_rtnl(lowerdev)) 1501 macvlan_port_destroy(port->dev); 1502 return err; 1503 } 1504 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1505 1506 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1507 struct nlattr *tb[], struct nlattr *data[], 1508 struct netlink_ext_ack *extack) 1509 { 1510 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1511 } 1512 1513 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1514 { 1515 struct macvlan_dev *vlan = netdev_priv(dev); 1516 1517 if (vlan->mode == MACVLAN_MODE_SOURCE) 1518 macvlan_flush_sources(vlan->port, vlan); 1519 list_del_rcu(&vlan->list); 1520 unregister_netdevice_queue(dev, head); 1521 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1522 } 1523 EXPORT_SYMBOL_GPL(macvlan_dellink); 1524 1525 static int macvlan_changelink(struct net_device *dev, 1526 struct nlattr *tb[], struct nlattr *data[], 1527 struct netlink_ext_ack *extack) 1528 { 1529 struct macvlan_dev *vlan = netdev_priv(dev); 1530 enum macvlan_mode mode; 1531 bool set_mode = false; 1532 enum macvlan_macaddr_mode macmode; 1533 int ret; 1534 1535 /* Validate mode, but don't set yet: setting flags may fail. */ 1536 if (data && data[IFLA_MACVLAN_MODE]) { 1537 set_mode = true; 1538 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1539 /* Passthrough mode can't be set or cleared dynamically */ 1540 if ((mode == MACVLAN_MODE_PASSTHRU) != 1541 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1542 return -EINVAL; 1543 if (vlan->mode == MACVLAN_MODE_SOURCE && 1544 vlan->mode != mode) 1545 macvlan_flush_sources(vlan->port, vlan); 1546 } 1547 1548 if (data && data[IFLA_MACVLAN_FLAGS]) { 1549 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1550 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1551 if (macvlan_passthru(vlan->port) && promisc) { 1552 int err; 1553 1554 if (flags & MACVLAN_FLAG_NOPROMISC) 1555 err = dev_set_promiscuity(vlan->lowerdev, -1); 1556 else 1557 err = dev_set_promiscuity(vlan->lowerdev, 1); 1558 if (err < 0) 1559 return err; 1560 } 1561 vlan->flags = flags; 1562 } 1563 if (set_mode) 1564 vlan->mode = mode; 1565 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1566 if (vlan->mode != MACVLAN_MODE_SOURCE) 1567 return -EINVAL; 1568 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1569 ret = macvlan_changelink_sources(vlan, macmode, data); 1570 if (ret) 1571 return ret; 1572 } 1573 return 0; 1574 } 1575 1576 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1577 { 1578 if (vlan->macaddr_count == 0) 1579 return 0; 1580 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1581 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1582 } 1583 1584 static size_t macvlan_get_size(const struct net_device *dev) 1585 { 1586 struct macvlan_dev *vlan = netdev_priv(dev); 1587 1588 return (0 1589 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1590 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1591 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1592 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1593 ); 1594 } 1595 1596 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1597 const struct macvlan_dev *vlan, 1598 const int i) 1599 { 1600 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1601 struct macvlan_source_entry *entry; 1602 1603 hlist_for_each_entry_rcu(entry, h, hlist) { 1604 if (entry->vlan != vlan) 1605 continue; 1606 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1607 return 1; 1608 } 1609 return 0; 1610 } 1611 1612 static int macvlan_fill_info(struct sk_buff *skb, 1613 const struct net_device *dev) 1614 { 1615 struct macvlan_dev *vlan = netdev_priv(dev); 1616 int i; 1617 struct nlattr *nest; 1618 1619 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1620 goto nla_put_failure; 1621 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1622 goto nla_put_failure; 1623 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1624 goto nla_put_failure; 1625 if (vlan->macaddr_count > 0) { 1626 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1627 if (nest == NULL) 1628 goto nla_put_failure; 1629 1630 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1631 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1632 goto nla_put_failure; 1633 } 1634 nla_nest_end(skb, nest); 1635 } 1636 return 0; 1637 1638 nla_put_failure: 1639 return -EMSGSIZE; 1640 } 1641 1642 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1643 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1644 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1645 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1646 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1647 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1648 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1649 }; 1650 1651 int macvlan_link_register(struct rtnl_link_ops *ops) 1652 { 1653 /* common fields */ 1654 ops->validate = macvlan_validate; 1655 ops->maxtype = IFLA_MACVLAN_MAX; 1656 ops->policy = macvlan_policy; 1657 ops->changelink = macvlan_changelink; 1658 ops->get_size = macvlan_get_size; 1659 ops->fill_info = macvlan_fill_info; 1660 1661 return rtnl_link_register(ops); 1662 }; 1663 EXPORT_SYMBOL_GPL(macvlan_link_register); 1664 1665 static struct net *macvlan_get_link_net(const struct net_device *dev) 1666 { 1667 return dev_net(macvlan_dev_real_dev(dev)); 1668 } 1669 1670 static struct rtnl_link_ops macvlan_link_ops = { 1671 .kind = "macvlan", 1672 .setup = macvlan_setup, 1673 .newlink = macvlan_newlink, 1674 .dellink = macvlan_dellink, 1675 .get_link_net = macvlan_get_link_net, 1676 .priv_size = sizeof(struct macvlan_dev), 1677 }; 1678 1679 static int macvlan_device_event(struct notifier_block *unused, 1680 unsigned long event, void *ptr) 1681 { 1682 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1683 struct macvlan_dev *vlan, *next; 1684 struct macvlan_port *port; 1685 LIST_HEAD(list_kill); 1686 1687 if (!netif_is_macvlan_port(dev)) 1688 return NOTIFY_DONE; 1689 1690 port = macvlan_port_get_rtnl(dev); 1691 1692 switch (event) { 1693 case NETDEV_UP: 1694 case NETDEV_DOWN: 1695 case NETDEV_CHANGE: 1696 list_for_each_entry(vlan, &port->vlans, list) 1697 netif_stacked_transfer_operstate(vlan->lowerdev, 1698 vlan->dev); 1699 break; 1700 case NETDEV_FEAT_CHANGE: 1701 list_for_each_entry(vlan, &port->vlans, list) { 1702 vlan->dev->gso_max_size = dev->gso_max_size; 1703 vlan->dev->gso_max_segs = dev->gso_max_segs; 1704 netdev_update_features(vlan->dev); 1705 } 1706 break; 1707 case NETDEV_CHANGEMTU: 1708 list_for_each_entry(vlan, &port->vlans, list) { 1709 if (vlan->dev->mtu <= dev->mtu) 1710 continue; 1711 dev_set_mtu(vlan->dev, dev->mtu); 1712 } 1713 break; 1714 case NETDEV_CHANGEADDR: 1715 if (!macvlan_passthru(port)) 1716 return NOTIFY_DONE; 1717 1718 vlan = list_first_entry_or_null(&port->vlans, 1719 struct macvlan_dev, 1720 list); 1721 1722 if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr)) 1723 return NOTIFY_BAD; 1724 1725 break; 1726 case NETDEV_UNREGISTER: 1727 /* twiddle thumbs on netns device moves */ 1728 if (dev->reg_state != NETREG_UNREGISTERING) 1729 break; 1730 1731 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1732 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1733 unregister_netdevice_many(&list_kill); 1734 break; 1735 case NETDEV_PRE_TYPE_CHANGE: 1736 /* Forbid underlaying device to change its type. */ 1737 return NOTIFY_BAD; 1738 1739 case NETDEV_NOTIFY_PEERS: 1740 case NETDEV_BONDING_FAILOVER: 1741 case NETDEV_RESEND_IGMP: 1742 /* Propagate to all vlans */ 1743 list_for_each_entry(vlan, &port->vlans, list) 1744 call_netdevice_notifiers(event, vlan->dev); 1745 } 1746 return NOTIFY_DONE; 1747 } 1748 1749 static struct notifier_block macvlan_notifier_block __read_mostly = { 1750 .notifier_call = macvlan_device_event, 1751 }; 1752 1753 static int __init macvlan_init_module(void) 1754 { 1755 int err; 1756 1757 register_netdevice_notifier(&macvlan_notifier_block); 1758 1759 err = macvlan_link_register(&macvlan_link_ops); 1760 if (err < 0) 1761 goto err1; 1762 return 0; 1763 err1: 1764 unregister_netdevice_notifier(&macvlan_notifier_block); 1765 return err; 1766 } 1767 1768 static void __exit macvlan_cleanup_module(void) 1769 { 1770 rtnl_link_unregister(&macvlan_link_ops); 1771 unregister_netdevice_notifier(&macvlan_notifier_block); 1772 } 1773 1774 module_init(macvlan_init_module); 1775 module_exit(macvlan_cleanup_module); 1776 1777 MODULE_LICENSE("GPL"); 1778 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1779 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1780 MODULE_ALIAS_RTNL_LINK("macvlan"); 1781