1 /* 2 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net> 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public License as 6 * published by the Free Software Foundation; either version 2 of 7 * the License, or (at your option) any later version. 8 * 9 * The code this is based on carried the following copyright notice: 10 * --- 11 * (C) Copyright 2001-2006 12 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com 13 * Re-worked by Ben Greear <greearb@candelatech.com> 14 * --- 15 */ 16 #include <linux/kernel.h> 17 #include <linux/types.h> 18 #include <linux/module.h> 19 #include <linux/init.h> 20 #include <linux/errno.h> 21 #include <linux/slab.h> 22 #include <linux/string.h> 23 #include <linux/rculist.h> 24 #include <linux/notifier.h> 25 #include <linux/netdevice.h> 26 #include <linux/etherdevice.h> 27 #include <linux/net_tstamp.h> 28 #include <linux/ethtool.h> 29 #include <linux/if_arp.h> 30 #include <linux/if_vlan.h> 31 #include <linux/if_link.h> 32 #include <linux/if_macvlan.h> 33 #include <linux/hash.h> 34 #include <linux/workqueue.h> 35 #include <net/rtnetlink.h> 36 #include <net/xfrm.h> 37 #include <linux/netpoll.h> 38 #include <linux/phy.h> 39 40 #define MACVLAN_HASH_BITS 8 41 #define MACVLAN_HASH_SIZE (1<<MACVLAN_HASH_BITS) 42 #define MACVLAN_BC_QUEUE_LEN 1000 43 44 #define MACVLAN_F_PASSTHRU 1 45 #define MACVLAN_F_ADDRCHANGE 2 46 47 struct macvlan_port { 48 struct net_device *dev; 49 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE]; 50 struct list_head vlans; 51 struct sk_buff_head bc_queue; 52 struct work_struct bc_work; 53 u32 flags; 54 int count; 55 struct hlist_head vlan_source_hash[MACVLAN_HASH_SIZE]; 56 DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ); 57 unsigned char perm_addr[ETH_ALEN]; 58 }; 59 60 struct macvlan_source_entry { 61 struct hlist_node hlist; 62 struct macvlan_dev *vlan; 63 unsigned char addr[6+2] __aligned(sizeof(u16)); 64 struct rcu_head rcu; 65 }; 66 67 struct macvlan_skb_cb { 68 const struct macvlan_dev *src; 69 }; 70 71 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0])) 72 73 static void macvlan_port_destroy(struct net_device *dev); 74 75 static inline bool macvlan_passthru(const struct macvlan_port *port) 76 { 77 return port->flags & MACVLAN_F_PASSTHRU; 78 } 79 80 static inline void macvlan_set_passthru(struct macvlan_port *port) 81 { 82 port->flags |= MACVLAN_F_PASSTHRU; 83 } 84 85 static inline bool macvlan_addr_change(const struct macvlan_port *port) 86 { 87 return port->flags & MACVLAN_F_ADDRCHANGE; 88 } 89 90 static inline void macvlan_set_addr_change(struct macvlan_port *port) 91 { 92 port->flags |= MACVLAN_F_ADDRCHANGE; 93 } 94 95 static inline void macvlan_clear_addr_change(struct macvlan_port *port) 96 { 97 port->flags &= ~MACVLAN_F_ADDRCHANGE; 98 } 99 100 /* Hash Ethernet address */ 101 static u32 macvlan_eth_hash(const unsigned char *addr) 102 { 103 u64 value = get_unaligned((u64 *)addr); 104 105 /* only want 6 bytes */ 106 #ifdef __BIG_ENDIAN 107 value >>= 16; 108 #else 109 value <<= 16; 110 #endif 111 return hash_64(value, MACVLAN_HASH_BITS); 112 } 113 114 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev) 115 { 116 return rcu_dereference(dev->rx_handler_data); 117 } 118 119 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev) 120 { 121 return rtnl_dereference(dev->rx_handler_data); 122 } 123 124 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port, 125 const unsigned char *addr) 126 { 127 struct macvlan_dev *vlan; 128 u32 idx = macvlan_eth_hash(addr); 129 130 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) { 131 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr)) 132 return vlan; 133 } 134 return NULL; 135 } 136 137 static struct macvlan_source_entry *macvlan_hash_lookup_source( 138 const struct macvlan_dev *vlan, 139 const unsigned char *addr) 140 { 141 struct macvlan_source_entry *entry; 142 u32 idx = macvlan_eth_hash(addr); 143 struct hlist_head *h = &vlan->port->vlan_source_hash[idx]; 144 145 hlist_for_each_entry_rcu(entry, h, hlist) { 146 if (ether_addr_equal_64bits(entry->addr, addr) && 147 entry->vlan == vlan) 148 return entry; 149 } 150 return NULL; 151 } 152 153 static int macvlan_hash_add_source(struct macvlan_dev *vlan, 154 const unsigned char *addr) 155 { 156 struct macvlan_port *port = vlan->port; 157 struct macvlan_source_entry *entry; 158 struct hlist_head *h; 159 160 entry = macvlan_hash_lookup_source(vlan, addr); 161 if (entry) 162 return 0; 163 164 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 165 if (!entry) 166 return -ENOMEM; 167 168 ether_addr_copy(entry->addr, addr); 169 entry->vlan = vlan; 170 h = &port->vlan_source_hash[macvlan_eth_hash(addr)]; 171 hlist_add_head_rcu(&entry->hlist, h); 172 vlan->macaddr_count++; 173 174 return 0; 175 } 176 177 static void macvlan_hash_add(struct macvlan_dev *vlan) 178 { 179 struct macvlan_port *port = vlan->port; 180 const unsigned char *addr = vlan->dev->dev_addr; 181 u32 idx = macvlan_eth_hash(addr); 182 183 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]); 184 } 185 186 static void macvlan_hash_del_source(struct macvlan_source_entry *entry) 187 { 188 hlist_del_rcu(&entry->hlist); 189 kfree_rcu(entry, rcu); 190 } 191 192 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync) 193 { 194 hlist_del_rcu(&vlan->hlist); 195 if (sync) 196 synchronize_rcu(); 197 } 198 199 static void macvlan_hash_change_addr(struct macvlan_dev *vlan, 200 const unsigned char *addr) 201 { 202 macvlan_hash_del(vlan, true); 203 /* Now that we are unhashed it is safe to change the device 204 * address without confusing packet delivery. 205 */ 206 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN); 207 macvlan_hash_add(vlan); 208 } 209 210 static bool macvlan_addr_busy(const struct macvlan_port *port, 211 const unsigned char *addr) 212 { 213 /* Test to see if the specified address is 214 * currently in use by the underlying device or 215 * another macvlan. 216 */ 217 if (!macvlan_passthru(port) && !macvlan_addr_change(port) && 218 ether_addr_equal_64bits(port->dev->dev_addr, addr)) 219 return true; 220 221 if (macvlan_hash_lookup(port, addr)) 222 return true; 223 224 return false; 225 } 226 227 228 static int macvlan_broadcast_one(struct sk_buff *skb, 229 const struct macvlan_dev *vlan, 230 const struct ethhdr *eth, bool local) 231 { 232 struct net_device *dev = vlan->dev; 233 234 if (local) 235 return __dev_forward_skb(dev, skb); 236 237 skb->dev = dev; 238 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast)) 239 skb->pkt_type = PACKET_BROADCAST; 240 else 241 skb->pkt_type = PACKET_MULTICAST; 242 243 return 0; 244 } 245 246 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan) 247 { 248 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT); 249 } 250 251 252 static unsigned int mc_hash(const struct macvlan_dev *vlan, 253 const unsigned char *addr) 254 { 255 u32 val = __get_unaligned_cpu32(addr + 2); 256 257 val ^= macvlan_hash_mix(vlan); 258 return hash_32(val, MACVLAN_MC_FILTER_BITS); 259 } 260 261 static void macvlan_broadcast(struct sk_buff *skb, 262 const struct macvlan_port *port, 263 struct net_device *src, 264 enum macvlan_mode mode) 265 { 266 const struct ethhdr *eth = eth_hdr(skb); 267 const struct macvlan_dev *vlan; 268 struct sk_buff *nskb; 269 unsigned int i; 270 int err; 271 unsigned int hash; 272 273 if (skb->protocol == htons(ETH_P_PAUSE)) 274 return; 275 276 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 277 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) { 278 if (vlan->dev == src || !(vlan->mode & mode)) 279 continue; 280 281 hash = mc_hash(vlan, eth->h_dest); 282 if (!test_bit(hash, vlan->mc_filter)) 283 continue; 284 285 err = NET_RX_DROP; 286 nskb = skb_clone(skb, GFP_ATOMIC); 287 if (likely(nskb)) 288 err = macvlan_broadcast_one( 289 nskb, vlan, eth, 290 mode == MACVLAN_MODE_BRIDGE) ?: 291 netif_rx_ni(nskb); 292 macvlan_count_rx(vlan, skb->len + ETH_HLEN, 293 err == NET_RX_SUCCESS, true); 294 } 295 } 296 } 297 298 static void macvlan_process_broadcast(struct work_struct *w) 299 { 300 struct macvlan_port *port = container_of(w, struct macvlan_port, 301 bc_work); 302 struct sk_buff *skb; 303 struct sk_buff_head list; 304 305 __skb_queue_head_init(&list); 306 307 spin_lock_bh(&port->bc_queue.lock); 308 skb_queue_splice_tail_init(&port->bc_queue, &list); 309 spin_unlock_bh(&port->bc_queue.lock); 310 311 while ((skb = __skb_dequeue(&list))) { 312 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 313 314 rcu_read_lock(); 315 316 if (!src) 317 /* frame comes from an external address */ 318 macvlan_broadcast(skb, port, NULL, 319 MACVLAN_MODE_PRIVATE | 320 MACVLAN_MODE_VEPA | 321 MACVLAN_MODE_PASSTHRU| 322 MACVLAN_MODE_BRIDGE); 323 else if (src->mode == MACVLAN_MODE_VEPA) 324 /* flood to everyone except source */ 325 macvlan_broadcast(skb, port, src->dev, 326 MACVLAN_MODE_VEPA | 327 MACVLAN_MODE_BRIDGE); 328 else 329 /* 330 * flood only to VEPA ports, bridge ports 331 * already saw the frame on the way out. 332 */ 333 macvlan_broadcast(skb, port, src->dev, 334 MACVLAN_MODE_VEPA); 335 336 rcu_read_unlock(); 337 338 if (src) 339 dev_put(src->dev); 340 consume_skb(skb); 341 } 342 } 343 344 static void macvlan_broadcast_enqueue(struct macvlan_port *port, 345 const struct macvlan_dev *src, 346 struct sk_buff *skb) 347 { 348 struct sk_buff *nskb; 349 int err = -ENOMEM; 350 351 nskb = skb_clone(skb, GFP_ATOMIC); 352 if (!nskb) 353 goto err; 354 355 MACVLAN_SKB_CB(nskb)->src = src; 356 357 spin_lock(&port->bc_queue.lock); 358 if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) { 359 if (src) 360 dev_hold(src->dev); 361 __skb_queue_tail(&port->bc_queue, nskb); 362 err = 0; 363 } 364 spin_unlock(&port->bc_queue.lock); 365 366 if (err) 367 goto free_nskb; 368 369 schedule_work(&port->bc_work); 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 = (void *)skb->data; 520 521 /* send to other bridge ports directly */ 522 if (is_multicast_ether_addr(eth->h_dest)) { 523 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE); 524 goto xmit_world; 525 } 526 527 dest = macvlan_hash_lookup(port, eth->h_dest); 528 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) { 529 /* send to lowerdev first for its network taps */ 530 dev_forward_skb(vlan->lowerdev, skb); 531 532 return NET_XMIT_SUCCESS; 533 } 534 } 535 xmit_world: 536 skb->dev = vlan->lowerdev; 537 return dev_queue_xmit_accel(skb, 538 netdev_get_sb_channel(dev) ? dev : NULL); 539 } 540 541 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) 542 { 543 #ifdef CONFIG_NET_POLL_CONTROLLER 544 if (vlan->netpoll) 545 netpoll_send_skb(vlan->netpoll, skb); 546 #else 547 BUG(); 548 #endif 549 return NETDEV_TX_OK; 550 } 551 552 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, 553 struct net_device *dev) 554 { 555 struct macvlan_dev *vlan = netdev_priv(dev); 556 unsigned int len = skb->len; 557 int ret; 558 559 if (unlikely(netpoll_tx_running(dev))) 560 return macvlan_netpoll_send_skb(vlan, skb); 561 562 ret = macvlan_queue_xmit(skb, dev); 563 564 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 565 struct vlan_pcpu_stats *pcpu_stats; 566 567 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats); 568 u64_stats_update_begin(&pcpu_stats->syncp); 569 pcpu_stats->tx_packets++; 570 pcpu_stats->tx_bytes += len; 571 u64_stats_update_end(&pcpu_stats->syncp); 572 } else { 573 this_cpu_inc(vlan->pcpu_stats->tx_dropped); 574 } 575 return ret; 576 } 577 578 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 579 unsigned short type, const void *daddr, 580 const void *saddr, unsigned len) 581 { 582 const struct macvlan_dev *vlan = netdev_priv(dev); 583 struct net_device *lowerdev = vlan->lowerdev; 584 585 return dev_hard_header(skb, lowerdev, type, daddr, 586 saddr ? : dev->dev_addr, len); 587 } 588 589 static const struct header_ops macvlan_hard_header_ops = { 590 .create = macvlan_hard_header, 591 .parse = eth_header_parse, 592 .cache = eth_header_cache, 593 .cache_update = eth_header_cache_update, 594 }; 595 596 static int macvlan_open(struct net_device *dev) 597 { 598 struct macvlan_dev *vlan = netdev_priv(dev); 599 struct net_device *lowerdev = vlan->lowerdev; 600 int err; 601 602 if (macvlan_passthru(vlan->port)) { 603 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) { 604 err = dev_set_promiscuity(lowerdev, 1); 605 if (err < 0) 606 goto out; 607 } 608 goto hash_add; 609 } 610 611 err = -EADDRINUSE; 612 if (macvlan_addr_busy(vlan->port, dev->dev_addr)) 613 goto out; 614 615 /* Attempt to populate accel_priv which is used to offload the L2 616 * forwarding requests for unicast packets. 617 */ 618 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) 619 vlan->accel_priv = 620 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev); 621 622 /* If earlier attempt to offload failed, or accel_priv is not 623 * populated we must add the unicast address to the lower device. 624 */ 625 if (IS_ERR_OR_NULL(vlan->accel_priv)) { 626 vlan->accel_priv = NULL; 627 err = dev_uc_add(lowerdev, dev->dev_addr); 628 if (err < 0) 629 goto out; 630 } 631 632 if (dev->flags & IFF_ALLMULTI) { 633 err = dev_set_allmulti(lowerdev, 1); 634 if (err < 0) 635 goto del_unicast; 636 } 637 638 if (dev->flags & IFF_PROMISC) { 639 err = dev_set_promiscuity(lowerdev, 1); 640 if (err < 0) 641 goto clear_multi; 642 } 643 644 hash_add: 645 macvlan_hash_add(vlan); 646 return 0; 647 648 clear_multi: 649 if (dev->flags & IFF_ALLMULTI) 650 dev_set_allmulti(lowerdev, -1); 651 del_unicast: 652 if (vlan->accel_priv) { 653 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 654 vlan->accel_priv); 655 vlan->accel_priv = NULL; 656 } else { 657 dev_uc_del(lowerdev, dev->dev_addr); 658 } 659 out: 660 return err; 661 } 662 663 static int macvlan_stop(struct net_device *dev) 664 { 665 struct macvlan_dev *vlan = netdev_priv(dev); 666 struct net_device *lowerdev = vlan->lowerdev; 667 668 if (vlan->accel_priv) { 669 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev, 670 vlan->accel_priv); 671 vlan->accel_priv = NULL; 672 } 673 674 dev_uc_unsync(lowerdev, dev); 675 dev_mc_unsync(lowerdev, dev); 676 677 if (macvlan_passthru(vlan->port)) { 678 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) 679 dev_set_promiscuity(lowerdev, -1); 680 goto hash_del; 681 } 682 683 if (dev->flags & IFF_ALLMULTI) 684 dev_set_allmulti(lowerdev, -1); 685 686 if (dev->flags & IFF_PROMISC) 687 dev_set_promiscuity(lowerdev, -1); 688 689 dev_uc_del(lowerdev, dev->dev_addr); 690 691 hash_del: 692 macvlan_hash_del(vlan, !dev->dismantle); 693 return 0; 694 } 695 696 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr) 697 { 698 struct macvlan_dev *vlan = netdev_priv(dev); 699 struct net_device *lowerdev = vlan->lowerdev; 700 struct macvlan_port *port = vlan->port; 701 int err; 702 703 if (!(dev->flags & IFF_UP)) { 704 /* Just copy in the new address */ 705 ether_addr_copy(dev->dev_addr, addr); 706 } else { 707 /* Rehash and update the device filters */ 708 if (macvlan_addr_busy(vlan->port, addr)) 709 return -EADDRINUSE; 710 711 if (!macvlan_passthru(port)) { 712 err = dev_uc_add(lowerdev, addr); 713 if (err) 714 return err; 715 716 dev_uc_del(lowerdev, dev->dev_addr); 717 } 718 719 macvlan_hash_change_addr(vlan, addr); 720 } 721 if (macvlan_passthru(port) && !macvlan_addr_change(port)) { 722 /* Since addr_change isn't set, we are here due to lower 723 * device change. Save the lower-dev address so we can 724 * restore it later. 725 */ 726 ether_addr_copy(vlan->port->perm_addr, 727 lowerdev->dev_addr); 728 } 729 macvlan_clear_addr_change(port); 730 return 0; 731 } 732 733 static int macvlan_set_mac_address(struct net_device *dev, void *p) 734 { 735 struct macvlan_dev *vlan = netdev_priv(dev); 736 struct sockaddr *addr = p; 737 738 if (!is_valid_ether_addr(addr->sa_data)) 739 return -EADDRNOTAVAIL; 740 741 /* If the addresses are the same, this is a no-op */ 742 if (ether_addr_equal(dev->dev_addr, addr->sa_data)) 743 return 0; 744 745 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 746 macvlan_set_addr_change(vlan->port); 747 return dev_set_mac_address(vlan->lowerdev, addr, NULL); 748 } 749 750 if (macvlan_addr_busy(vlan->port, addr->sa_data)) 751 return -EADDRINUSE; 752 753 return macvlan_sync_address(dev, addr->sa_data); 754 } 755 756 static void macvlan_change_rx_flags(struct net_device *dev, int change) 757 { 758 struct macvlan_dev *vlan = netdev_priv(dev); 759 struct net_device *lowerdev = vlan->lowerdev; 760 761 if (dev->flags & IFF_UP) { 762 if (change & IFF_ALLMULTI) 763 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1); 764 if (change & IFF_PROMISC) 765 dev_set_promiscuity(lowerdev, 766 dev->flags & IFF_PROMISC ? 1 : -1); 767 768 } 769 } 770 771 static void macvlan_compute_filter(unsigned long *mc_filter, 772 struct net_device *dev, 773 struct macvlan_dev *vlan) 774 { 775 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 776 bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ); 777 } else { 778 struct netdev_hw_addr *ha; 779 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ); 780 781 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ); 782 netdev_for_each_mc_addr(ha, dev) { 783 __set_bit(mc_hash(vlan, ha->addr), filter); 784 } 785 786 __set_bit(mc_hash(vlan, dev->broadcast), filter); 787 788 bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ); 789 } 790 } 791 792 static void macvlan_set_mac_lists(struct net_device *dev) 793 { 794 struct macvlan_dev *vlan = netdev_priv(dev); 795 796 macvlan_compute_filter(vlan->mc_filter, dev, vlan); 797 798 dev_uc_sync(vlan->lowerdev, dev); 799 dev_mc_sync(vlan->lowerdev, dev); 800 801 /* This is slightly inaccurate as we're including the subscription 802 * list of vlan->lowerdev too. 803 * 804 * Bug alert: This only works if everyone has the same broadcast 805 * address as lowerdev. As soon as someone changes theirs this 806 * will break. 807 * 808 * However, this is already broken as when you change your broadcast 809 * address we don't get called. 810 * 811 * The solution is to maintain a list of broadcast addresses like 812 * we do for uc/mc, if you care. 813 */ 814 macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL); 815 } 816 817 static int macvlan_change_mtu(struct net_device *dev, int new_mtu) 818 { 819 struct macvlan_dev *vlan = netdev_priv(dev); 820 821 if (vlan->lowerdev->mtu < new_mtu) 822 return -EINVAL; 823 dev->mtu = new_mtu; 824 return 0; 825 } 826 827 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 828 { 829 struct net_device *real_dev = macvlan_dev_real_dev(dev); 830 const struct net_device_ops *ops = real_dev->netdev_ops; 831 struct ifreq ifrr; 832 int err = -EOPNOTSUPP; 833 834 strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ); 835 ifrr.ifr_ifru = ifr->ifr_ifru; 836 837 switch (cmd) { 838 case SIOCSHWTSTAMP: 839 case SIOCGHWTSTAMP: 840 if (netif_device_present(real_dev) && ops->ndo_do_ioctl) 841 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd); 842 break; 843 } 844 845 if (!err) 846 ifr->ifr_ifru = ifrr.ifr_ifru; 847 848 return err; 849 } 850 851 /* 852 * macvlan network devices have devices nesting below it and are a special 853 * "super class" of normal network devices; split their locks off into a 854 * separate class since they always nest. 855 */ 856 static struct lock_class_key macvlan_netdev_addr_lock_key; 857 858 #define ALWAYS_ON_OFFLOADS \ 859 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 860 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 861 862 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 863 864 #define MACVLAN_FEATURES \ 865 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 866 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 867 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 868 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 869 870 #define MACVLAN_STATE_MASK \ 871 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 872 873 static int macvlan_get_nest_level(struct net_device *dev) 874 { 875 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level; 876 } 877 878 static void macvlan_set_lockdep_class(struct net_device *dev) 879 { 880 netdev_lockdep_set_classes(dev); 881 lockdep_set_class_and_subclass(&dev->addr_list_lock, 882 &macvlan_netdev_addr_lock_key, 883 macvlan_get_nest_level(dev)); 884 } 885 886 static int macvlan_init(struct net_device *dev) 887 { 888 struct macvlan_dev *vlan = netdev_priv(dev); 889 const struct net_device *lowerdev = vlan->lowerdev; 890 struct macvlan_port *port = vlan->port; 891 892 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 893 (lowerdev->state & MACVLAN_STATE_MASK); 894 dev->features = lowerdev->features & MACVLAN_FEATURES; 895 dev->features |= ALWAYS_ON_FEATURES; 896 dev->hw_features |= NETIF_F_LRO; 897 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 898 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 899 dev->hw_enc_features |= dev->features; 900 dev->gso_max_size = lowerdev->gso_max_size; 901 dev->gso_max_segs = lowerdev->gso_max_segs; 902 dev->hard_header_len = lowerdev->hard_header_len; 903 904 macvlan_set_lockdep_class(dev); 905 906 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 907 if (!vlan->pcpu_stats) 908 return -ENOMEM; 909 910 port->count += 1; 911 912 return 0; 913 } 914 915 static void macvlan_uninit(struct net_device *dev) 916 { 917 struct macvlan_dev *vlan = netdev_priv(dev); 918 struct macvlan_port *port = vlan->port; 919 920 free_percpu(vlan->pcpu_stats); 921 922 macvlan_flush_sources(port, vlan); 923 port->count -= 1; 924 if (!port->count) 925 macvlan_port_destroy(port->dev); 926 } 927 928 static void macvlan_dev_get_stats64(struct net_device *dev, 929 struct rtnl_link_stats64 *stats) 930 { 931 struct macvlan_dev *vlan = netdev_priv(dev); 932 933 if (vlan->pcpu_stats) { 934 struct vlan_pcpu_stats *p; 935 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 936 u32 rx_errors = 0, tx_dropped = 0; 937 unsigned int start; 938 int i; 939 940 for_each_possible_cpu(i) { 941 p = per_cpu_ptr(vlan->pcpu_stats, i); 942 do { 943 start = u64_stats_fetch_begin_irq(&p->syncp); 944 rx_packets = p->rx_packets; 945 rx_bytes = p->rx_bytes; 946 rx_multicast = p->rx_multicast; 947 tx_packets = p->tx_packets; 948 tx_bytes = p->tx_bytes; 949 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 950 951 stats->rx_packets += rx_packets; 952 stats->rx_bytes += rx_bytes; 953 stats->multicast += rx_multicast; 954 stats->tx_packets += tx_packets; 955 stats->tx_bytes += tx_bytes; 956 /* rx_errors & tx_dropped are u32, updated 957 * without syncp protection. 958 */ 959 rx_errors += p->rx_errors; 960 tx_dropped += p->tx_dropped; 961 } 962 stats->rx_errors = rx_errors; 963 stats->rx_dropped = rx_errors; 964 stats->tx_dropped = tx_dropped; 965 } 966 } 967 968 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 969 __be16 proto, u16 vid) 970 { 971 struct macvlan_dev *vlan = netdev_priv(dev); 972 struct net_device *lowerdev = vlan->lowerdev; 973 974 return vlan_vid_add(lowerdev, proto, vid); 975 } 976 977 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 978 __be16 proto, u16 vid) 979 { 980 struct macvlan_dev *vlan = netdev_priv(dev); 981 struct net_device *lowerdev = vlan->lowerdev; 982 983 vlan_vid_del(lowerdev, proto, vid); 984 return 0; 985 } 986 987 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 988 struct net_device *dev, 989 const unsigned char *addr, u16 vid, 990 u16 flags, 991 struct netlink_ext_ack *extack) 992 { 993 struct macvlan_dev *vlan = netdev_priv(dev); 994 int err = -EINVAL; 995 996 /* Support unicast filter only on passthru devices. 997 * Multicast filter should be allowed on all devices. 998 */ 999 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1000 return -EOPNOTSUPP; 1001 1002 if (flags & NLM_F_REPLACE) 1003 return -EOPNOTSUPP; 1004 1005 if (is_unicast_ether_addr(addr)) 1006 err = dev_uc_add_excl(dev, addr); 1007 else if (is_multicast_ether_addr(addr)) 1008 err = dev_mc_add_excl(dev, addr); 1009 1010 return err; 1011 } 1012 1013 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1014 struct net_device *dev, 1015 const unsigned char *addr, u16 vid) 1016 { 1017 struct macvlan_dev *vlan = netdev_priv(dev); 1018 int err = -EINVAL; 1019 1020 /* Support unicast filter only on passthru devices. 1021 * Multicast filter should be allowed on all devices. 1022 */ 1023 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1024 return -EOPNOTSUPP; 1025 1026 if (is_unicast_ether_addr(addr)) 1027 err = dev_uc_del(dev, addr); 1028 else if (is_multicast_ether_addr(addr)) 1029 err = dev_mc_del(dev, addr); 1030 1031 return err; 1032 } 1033 1034 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1035 struct ethtool_drvinfo *drvinfo) 1036 { 1037 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1038 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1039 } 1040 1041 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1042 struct ethtool_link_ksettings *cmd) 1043 { 1044 const struct macvlan_dev *vlan = netdev_priv(dev); 1045 1046 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1047 } 1048 1049 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1050 struct ethtool_ts_info *info) 1051 { 1052 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1053 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1054 struct phy_device *phydev = real_dev->phydev; 1055 1056 if (phydev && phydev->drv && phydev->drv->ts_info) { 1057 return phydev->drv->ts_info(phydev, info); 1058 } else if (ops->get_ts_info) { 1059 return ops->get_ts_info(real_dev, info); 1060 } else { 1061 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1062 SOF_TIMESTAMPING_SOFTWARE; 1063 info->phc_index = -1; 1064 } 1065 1066 return 0; 1067 } 1068 1069 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1070 netdev_features_t features) 1071 { 1072 struct macvlan_dev *vlan = netdev_priv(dev); 1073 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1074 netdev_features_t mask; 1075 1076 features |= NETIF_F_ALL_FOR_ALL; 1077 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1078 mask = features; 1079 1080 lowerdev_features &= (features | ~NETIF_F_LRO); 1081 features = netdev_increment_features(lowerdev_features, features, mask); 1082 features |= ALWAYS_ON_FEATURES; 1083 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1084 1085 return features; 1086 } 1087 1088 #ifdef CONFIG_NET_POLL_CONTROLLER 1089 static void macvlan_dev_poll_controller(struct net_device *dev) 1090 { 1091 return; 1092 } 1093 1094 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1095 { 1096 struct macvlan_dev *vlan = netdev_priv(dev); 1097 struct net_device *real_dev = vlan->lowerdev; 1098 struct netpoll *netpoll; 1099 int err = 0; 1100 1101 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1102 err = -ENOMEM; 1103 if (!netpoll) 1104 goto out; 1105 1106 err = __netpoll_setup(netpoll, real_dev); 1107 if (err) { 1108 kfree(netpoll); 1109 goto out; 1110 } 1111 1112 vlan->netpoll = netpoll; 1113 1114 out: 1115 return err; 1116 } 1117 1118 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1119 { 1120 struct macvlan_dev *vlan = netdev_priv(dev); 1121 struct netpoll *netpoll = vlan->netpoll; 1122 1123 if (!netpoll) 1124 return; 1125 1126 vlan->netpoll = NULL; 1127 1128 __netpoll_free(netpoll); 1129 } 1130 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1131 1132 static int macvlan_dev_get_iflink(const struct net_device *dev) 1133 { 1134 struct macvlan_dev *vlan = netdev_priv(dev); 1135 1136 return vlan->lowerdev->ifindex; 1137 } 1138 1139 static const struct ethtool_ops macvlan_ethtool_ops = { 1140 .get_link = ethtool_op_get_link, 1141 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1142 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1143 .get_ts_info = macvlan_ethtool_get_ts_info, 1144 }; 1145 1146 static const struct net_device_ops macvlan_netdev_ops = { 1147 .ndo_init = macvlan_init, 1148 .ndo_uninit = macvlan_uninit, 1149 .ndo_open = macvlan_open, 1150 .ndo_stop = macvlan_stop, 1151 .ndo_start_xmit = macvlan_start_xmit, 1152 .ndo_change_mtu = macvlan_change_mtu, 1153 .ndo_do_ioctl = macvlan_do_ioctl, 1154 .ndo_fix_features = macvlan_fix_features, 1155 .ndo_change_rx_flags = macvlan_change_rx_flags, 1156 .ndo_set_mac_address = macvlan_set_mac_address, 1157 .ndo_set_rx_mode = macvlan_set_mac_lists, 1158 .ndo_get_stats64 = macvlan_dev_get_stats64, 1159 .ndo_validate_addr = eth_validate_addr, 1160 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1161 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1162 .ndo_fdb_add = macvlan_fdb_add, 1163 .ndo_fdb_del = macvlan_fdb_del, 1164 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1165 .ndo_get_lock_subclass = macvlan_get_nest_level, 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 vlan->nest_level = dev_get_nest_level(lowerdev) + 1; 1450 1451 vlan->mode = MACVLAN_MODE_VEPA; 1452 if (data && data[IFLA_MACVLAN_MODE]) 1453 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1454 1455 if (data && data[IFLA_MACVLAN_FLAGS]) 1456 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1457 1458 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1459 if (port->count) { 1460 err = -EINVAL; 1461 goto destroy_macvlan_port; 1462 } 1463 macvlan_set_passthru(port); 1464 eth_hw_addr_inherit(dev, lowerdev); 1465 } 1466 1467 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1468 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1469 err = -EINVAL; 1470 goto destroy_macvlan_port; 1471 } 1472 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1473 err = macvlan_changelink_sources(vlan, macmode, data); 1474 if (err) 1475 goto destroy_macvlan_port; 1476 } 1477 1478 err = register_netdevice(dev); 1479 if (err < 0) 1480 goto destroy_macvlan_port; 1481 1482 dev->priv_flags |= IFF_MACVLAN; 1483 err = netdev_upper_dev_link(lowerdev, dev, extack); 1484 if (err) 1485 goto unregister_netdev; 1486 1487 list_add_tail_rcu(&vlan->list, &port->vlans); 1488 netif_stacked_transfer_operstate(lowerdev, dev); 1489 linkwatch_fire_event(dev); 1490 1491 return 0; 1492 1493 unregister_netdev: 1494 /* macvlan_uninit would free the macvlan port */ 1495 unregister_netdevice(dev); 1496 return err; 1497 destroy_macvlan_port: 1498 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1499 * so we destroy the macvlan port only when it's valid. 1500 */ 1501 if (create && macvlan_port_get_rtnl(lowerdev)) 1502 macvlan_port_destroy(port->dev); 1503 return err; 1504 } 1505 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1506 1507 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1508 struct nlattr *tb[], struct nlattr *data[], 1509 struct netlink_ext_ack *extack) 1510 { 1511 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1512 } 1513 1514 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1515 { 1516 struct macvlan_dev *vlan = netdev_priv(dev); 1517 1518 if (vlan->mode == MACVLAN_MODE_SOURCE) 1519 macvlan_flush_sources(vlan->port, vlan); 1520 list_del_rcu(&vlan->list); 1521 unregister_netdevice_queue(dev, head); 1522 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1523 } 1524 EXPORT_SYMBOL_GPL(macvlan_dellink); 1525 1526 static int macvlan_changelink(struct net_device *dev, 1527 struct nlattr *tb[], struct nlattr *data[], 1528 struct netlink_ext_ack *extack) 1529 { 1530 struct macvlan_dev *vlan = netdev_priv(dev); 1531 enum macvlan_mode mode; 1532 bool set_mode = false; 1533 enum macvlan_macaddr_mode macmode; 1534 int ret; 1535 1536 /* Validate mode, but don't set yet: setting flags may fail. */ 1537 if (data && data[IFLA_MACVLAN_MODE]) { 1538 set_mode = true; 1539 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1540 /* Passthrough mode can't be set or cleared dynamically */ 1541 if ((mode == MACVLAN_MODE_PASSTHRU) != 1542 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1543 return -EINVAL; 1544 if (vlan->mode == MACVLAN_MODE_SOURCE && 1545 vlan->mode != mode) 1546 macvlan_flush_sources(vlan->port, vlan); 1547 } 1548 1549 if (data && data[IFLA_MACVLAN_FLAGS]) { 1550 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1551 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1552 if (macvlan_passthru(vlan->port) && promisc) { 1553 int err; 1554 1555 if (flags & MACVLAN_FLAG_NOPROMISC) 1556 err = dev_set_promiscuity(vlan->lowerdev, -1); 1557 else 1558 err = dev_set_promiscuity(vlan->lowerdev, 1); 1559 if (err < 0) 1560 return err; 1561 } 1562 vlan->flags = flags; 1563 } 1564 if (set_mode) 1565 vlan->mode = mode; 1566 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1567 if (vlan->mode != MACVLAN_MODE_SOURCE) 1568 return -EINVAL; 1569 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1570 ret = macvlan_changelink_sources(vlan, macmode, data); 1571 if (ret) 1572 return ret; 1573 } 1574 return 0; 1575 } 1576 1577 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1578 { 1579 if (vlan->macaddr_count == 0) 1580 return 0; 1581 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1582 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1583 } 1584 1585 static size_t macvlan_get_size(const struct net_device *dev) 1586 { 1587 struct macvlan_dev *vlan = netdev_priv(dev); 1588 1589 return (0 1590 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1591 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1592 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1593 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1594 ); 1595 } 1596 1597 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1598 const struct macvlan_dev *vlan, 1599 const int i) 1600 { 1601 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1602 struct macvlan_source_entry *entry; 1603 1604 hlist_for_each_entry_rcu(entry, h, hlist) { 1605 if (entry->vlan != vlan) 1606 continue; 1607 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1608 return 1; 1609 } 1610 return 0; 1611 } 1612 1613 static int macvlan_fill_info(struct sk_buff *skb, 1614 const struct net_device *dev) 1615 { 1616 struct macvlan_dev *vlan = netdev_priv(dev); 1617 int i; 1618 struct nlattr *nest; 1619 1620 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1621 goto nla_put_failure; 1622 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1623 goto nla_put_failure; 1624 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1625 goto nla_put_failure; 1626 if (vlan->macaddr_count > 0) { 1627 nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA); 1628 if (nest == NULL) 1629 goto nla_put_failure; 1630 1631 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1632 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1633 goto nla_put_failure; 1634 } 1635 nla_nest_end(skb, nest); 1636 } 1637 return 0; 1638 1639 nla_put_failure: 1640 return -EMSGSIZE; 1641 } 1642 1643 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1644 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1645 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1646 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1647 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1648 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1649 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1650 }; 1651 1652 int macvlan_link_register(struct rtnl_link_ops *ops) 1653 { 1654 /* common fields */ 1655 ops->validate = macvlan_validate; 1656 ops->maxtype = IFLA_MACVLAN_MAX; 1657 ops->policy = macvlan_policy; 1658 ops->changelink = macvlan_changelink; 1659 ops->get_size = macvlan_get_size; 1660 ops->fill_info = macvlan_fill_info; 1661 1662 return rtnl_link_register(ops); 1663 }; 1664 EXPORT_SYMBOL_GPL(macvlan_link_register); 1665 1666 static struct net *macvlan_get_link_net(const struct net_device *dev) 1667 { 1668 return dev_net(macvlan_dev_real_dev(dev)); 1669 } 1670 1671 static struct rtnl_link_ops macvlan_link_ops = { 1672 .kind = "macvlan", 1673 .setup = macvlan_setup, 1674 .newlink = macvlan_newlink, 1675 .dellink = macvlan_dellink, 1676 .get_link_net = macvlan_get_link_net, 1677 .priv_size = sizeof(struct macvlan_dev), 1678 }; 1679 1680 static int macvlan_device_event(struct notifier_block *unused, 1681 unsigned long event, void *ptr) 1682 { 1683 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1684 struct macvlan_dev *vlan, *next; 1685 struct macvlan_port *port; 1686 LIST_HEAD(list_kill); 1687 1688 if (!netif_is_macvlan_port(dev)) 1689 return NOTIFY_DONE; 1690 1691 port = macvlan_port_get_rtnl(dev); 1692 1693 switch (event) { 1694 case NETDEV_UP: 1695 case NETDEV_DOWN: 1696 case NETDEV_CHANGE: 1697 list_for_each_entry(vlan, &port->vlans, list) 1698 netif_stacked_transfer_operstate(vlan->lowerdev, 1699 vlan->dev); 1700 break; 1701 case NETDEV_FEAT_CHANGE: 1702 list_for_each_entry(vlan, &port->vlans, list) { 1703 vlan->dev->gso_max_size = dev->gso_max_size; 1704 vlan->dev->gso_max_segs = dev->gso_max_segs; 1705 netdev_update_features(vlan->dev); 1706 } 1707 break; 1708 case NETDEV_CHANGEMTU: 1709 list_for_each_entry(vlan, &port->vlans, list) { 1710 if (vlan->dev->mtu <= dev->mtu) 1711 continue; 1712 dev_set_mtu(vlan->dev, dev->mtu); 1713 } 1714 break; 1715 case NETDEV_CHANGEADDR: 1716 if (!macvlan_passthru(port)) 1717 return NOTIFY_DONE; 1718 1719 vlan = list_first_entry_or_null(&port->vlans, 1720 struct macvlan_dev, 1721 list); 1722 1723 if (macvlan_sync_address(vlan->dev, dev->dev_addr)) 1724 return NOTIFY_BAD; 1725 1726 break; 1727 case NETDEV_UNREGISTER: 1728 /* twiddle thumbs on netns device moves */ 1729 if (dev->reg_state != NETREG_UNREGISTERING) 1730 break; 1731 1732 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1733 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1734 unregister_netdevice_many(&list_kill); 1735 break; 1736 case NETDEV_PRE_TYPE_CHANGE: 1737 /* Forbid underlaying device to change its type. */ 1738 return NOTIFY_BAD; 1739 1740 case NETDEV_NOTIFY_PEERS: 1741 case NETDEV_BONDING_FAILOVER: 1742 case NETDEV_RESEND_IGMP: 1743 /* Propagate to all vlans */ 1744 list_for_each_entry(vlan, &port->vlans, list) 1745 call_netdevice_notifiers(event, vlan->dev); 1746 } 1747 return NOTIFY_DONE; 1748 } 1749 1750 static struct notifier_block macvlan_notifier_block __read_mostly = { 1751 .notifier_call = macvlan_device_event, 1752 }; 1753 1754 static int __init macvlan_init_module(void) 1755 { 1756 int err; 1757 1758 register_netdevice_notifier(&macvlan_notifier_block); 1759 1760 err = macvlan_link_register(&macvlan_link_ops); 1761 if (err < 0) 1762 goto err1; 1763 return 0; 1764 err1: 1765 unregister_netdevice_notifier(&macvlan_notifier_block); 1766 return err; 1767 } 1768 1769 static void __exit macvlan_cleanup_module(void) 1770 { 1771 rtnl_link_unregister(&macvlan_link_ops); 1772 unregister_netdevice_notifier(&macvlan_notifier_block); 1773 } 1774 1775 module_init(macvlan_init_module); 1776 module_exit(macvlan_cleanup_module); 1777 1778 MODULE_LICENSE("GPL"); 1779 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1780 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1781 MODULE_ALIAS_RTNL_LINK("macvlan"); 1782