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 if (!net_eq(dev_net(dev), &init_net)) 840 break; 841 /* fall through */ 842 case SIOCGHWTSTAMP: 843 if (netif_device_present(real_dev) && ops->ndo_do_ioctl) 844 err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd); 845 break; 846 } 847 848 if (!err) 849 ifr->ifr_ifru = ifrr.ifr_ifru; 850 851 return err; 852 } 853 854 /* 855 * macvlan network devices have devices nesting below it and are a special 856 * "super class" of normal network devices; split their locks off into a 857 * separate class since they always nest. 858 */ 859 static struct lock_class_key macvlan_netdev_addr_lock_key; 860 861 #define ALWAYS_ON_OFFLOADS \ 862 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \ 863 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL) 864 865 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX) 866 867 #define MACVLAN_FEATURES \ 868 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 869 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \ 870 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 871 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 872 873 #define MACVLAN_STATE_MASK \ 874 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 875 876 static int macvlan_get_nest_level(struct net_device *dev) 877 { 878 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level; 879 } 880 881 static void macvlan_set_lockdep_class(struct net_device *dev) 882 { 883 netdev_lockdep_set_classes(dev); 884 lockdep_set_class_and_subclass(&dev->addr_list_lock, 885 &macvlan_netdev_addr_lock_key, 886 macvlan_get_nest_level(dev)); 887 } 888 889 static int macvlan_init(struct net_device *dev) 890 { 891 struct macvlan_dev *vlan = netdev_priv(dev); 892 const struct net_device *lowerdev = vlan->lowerdev; 893 struct macvlan_port *port = vlan->port; 894 895 dev->state = (dev->state & ~MACVLAN_STATE_MASK) | 896 (lowerdev->state & MACVLAN_STATE_MASK); 897 dev->features = lowerdev->features & MACVLAN_FEATURES; 898 dev->features |= ALWAYS_ON_FEATURES; 899 dev->hw_features |= NETIF_F_LRO; 900 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES; 901 dev->vlan_features |= ALWAYS_ON_OFFLOADS; 902 dev->hw_enc_features |= dev->features; 903 dev->gso_max_size = lowerdev->gso_max_size; 904 dev->gso_max_segs = lowerdev->gso_max_segs; 905 dev->hard_header_len = lowerdev->hard_header_len; 906 907 macvlan_set_lockdep_class(dev); 908 909 vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats); 910 if (!vlan->pcpu_stats) 911 return -ENOMEM; 912 913 port->count += 1; 914 915 return 0; 916 } 917 918 static void macvlan_uninit(struct net_device *dev) 919 { 920 struct macvlan_dev *vlan = netdev_priv(dev); 921 struct macvlan_port *port = vlan->port; 922 923 free_percpu(vlan->pcpu_stats); 924 925 macvlan_flush_sources(port, vlan); 926 port->count -= 1; 927 if (!port->count) 928 macvlan_port_destroy(port->dev); 929 } 930 931 static void macvlan_dev_get_stats64(struct net_device *dev, 932 struct rtnl_link_stats64 *stats) 933 { 934 struct macvlan_dev *vlan = netdev_priv(dev); 935 936 if (vlan->pcpu_stats) { 937 struct vlan_pcpu_stats *p; 938 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes; 939 u32 rx_errors = 0, tx_dropped = 0; 940 unsigned int start; 941 int i; 942 943 for_each_possible_cpu(i) { 944 p = per_cpu_ptr(vlan->pcpu_stats, i); 945 do { 946 start = u64_stats_fetch_begin_irq(&p->syncp); 947 rx_packets = p->rx_packets; 948 rx_bytes = p->rx_bytes; 949 rx_multicast = p->rx_multicast; 950 tx_packets = p->tx_packets; 951 tx_bytes = p->tx_bytes; 952 } while (u64_stats_fetch_retry_irq(&p->syncp, start)); 953 954 stats->rx_packets += rx_packets; 955 stats->rx_bytes += rx_bytes; 956 stats->multicast += rx_multicast; 957 stats->tx_packets += tx_packets; 958 stats->tx_bytes += tx_bytes; 959 /* rx_errors & tx_dropped are u32, updated 960 * without syncp protection. 961 */ 962 rx_errors += p->rx_errors; 963 tx_dropped += p->tx_dropped; 964 } 965 stats->rx_errors = rx_errors; 966 stats->rx_dropped = rx_errors; 967 stats->tx_dropped = tx_dropped; 968 } 969 } 970 971 static int macvlan_vlan_rx_add_vid(struct net_device *dev, 972 __be16 proto, u16 vid) 973 { 974 struct macvlan_dev *vlan = netdev_priv(dev); 975 struct net_device *lowerdev = vlan->lowerdev; 976 977 return vlan_vid_add(lowerdev, proto, vid); 978 } 979 980 static int macvlan_vlan_rx_kill_vid(struct net_device *dev, 981 __be16 proto, u16 vid) 982 { 983 struct macvlan_dev *vlan = netdev_priv(dev); 984 struct net_device *lowerdev = vlan->lowerdev; 985 986 vlan_vid_del(lowerdev, proto, vid); 987 return 0; 988 } 989 990 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], 991 struct net_device *dev, 992 const unsigned char *addr, u16 vid, 993 u16 flags, 994 struct netlink_ext_ack *extack) 995 { 996 struct macvlan_dev *vlan = netdev_priv(dev); 997 int err = -EINVAL; 998 999 /* Support unicast filter only on passthru devices. 1000 * Multicast filter should be allowed on all devices. 1001 */ 1002 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1003 return -EOPNOTSUPP; 1004 1005 if (flags & NLM_F_REPLACE) 1006 return -EOPNOTSUPP; 1007 1008 if (is_unicast_ether_addr(addr)) 1009 err = dev_uc_add_excl(dev, addr); 1010 else if (is_multicast_ether_addr(addr)) 1011 err = dev_mc_add_excl(dev, addr); 1012 1013 return err; 1014 } 1015 1016 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], 1017 struct net_device *dev, 1018 const unsigned char *addr, u16 vid) 1019 { 1020 struct macvlan_dev *vlan = netdev_priv(dev); 1021 int err = -EINVAL; 1022 1023 /* Support unicast filter only on passthru devices. 1024 * Multicast filter should be allowed on all devices. 1025 */ 1026 if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr)) 1027 return -EOPNOTSUPP; 1028 1029 if (is_unicast_ether_addr(addr)) 1030 err = dev_uc_del(dev, addr); 1031 else if (is_multicast_ether_addr(addr)) 1032 err = dev_mc_del(dev, addr); 1033 1034 return err; 1035 } 1036 1037 static void macvlan_ethtool_get_drvinfo(struct net_device *dev, 1038 struct ethtool_drvinfo *drvinfo) 1039 { 1040 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver)); 1041 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version)); 1042 } 1043 1044 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev, 1045 struct ethtool_link_ksettings *cmd) 1046 { 1047 const struct macvlan_dev *vlan = netdev_priv(dev); 1048 1049 return __ethtool_get_link_ksettings(vlan->lowerdev, cmd); 1050 } 1051 1052 static int macvlan_ethtool_get_ts_info(struct net_device *dev, 1053 struct ethtool_ts_info *info) 1054 { 1055 struct net_device *real_dev = macvlan_dev_real_dev(dev); 1056 const struct ethtool_ops *ops = real_dev->ethtool_ops; 1057 struct phy_device *phydev = real_dev->phydev; 1058 1059 if (phydev && phydev->drv && phydev->drv->ts_info) { 1060 return phydev->drv->ts_info(phydev, info); 1061 } else if (ops->get_ts_info) { 1062 return ops->get_ts_info(real_dev, info); 1063 } else { 1064 info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE | 1065 SOF_TIMESTAMPING_SOFTWARE; 1066 info->phc_index = -1; 1067 } 1068 1069 return 0; 1070 } 1071 1072 static netdev_features_t macvlan_fix_features(struct net_device *dev, 1073 netdev_features_t features) 1074 { 1075 struct macvlan_dev *vlan = netdev_priv(dev); 1076 netdev_features_t lowerdev_features = vlan->lowerdev->features; 1077 netdev_features_t mask; 1078 1079 features |= NETIF_F_ALL_FOR_ALL; 1080 features &= (vlan->set_features | ~MACVLAN_FEATURES); 1081 mask = features; 1082 1083 lowerdev_features &= (features | ~NETIF_F_LRO); 1084 features = netdev_increment_features(lowerdev_features, features, mask); 1085 features |= ALWAYS_ON_FEATURES; 1086 features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES); 1087 1088 return features; 1089 } 1090 1091 #ifdef CONFIG_NET_POLL_CONTROLLER 1092 static void macvlan_dev_poll_controller(struct net_device *dev) 1093 { 1094 return; 1095 } 1096 1097 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo) 1098 { 1099 struct macvlan_dev *vlan = netdev_priv(dev); 1100 struct net_device *real_dev = vlan->lowerdev; 1101 struct netpoll *netpoll; 1102 int err = 0; 1103 1104 netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL); 1105 err = -ENOMEM; 1106 if (!netpoll) 1107 goto out; 1108 1109 err = __netpoll_setup(netpoll, real_dev); 1110 if (err) { 1111 kfree(netpoll); 1112 goto out; 1113 } 1114 1115 vlan->netpoll = netpoll; 1116 1117 out: 1118 return err; 1119 } 1120 1121 static void macvlan_dev_netpoll_cleanup(struct net_device *dev) 1122 { 1123 struct macvlan_dev *vlan = netdev_priv(dev); 1124 struct netpoll *netpoll = vlan->netpoll; 1125 1126 if (!netpoll) 1127 return; 1128 1129 vlan->netpoll = NULL; 1130 1131 __netpoll_free(netpoll); 1132 } 1133 #endif /* CONFIG_NET_POLL_CONTROLLER */ 1134 1135 static int macvlan_dev_get_iflink(const struct net_device *dev) 1136 { 1137 struct macvlan_dev *vlan = netdev_priv(dev); 1138 1139 return vlan->lowerdev->ifindex; 1140 } 1141 1142 static const struct ethtool_ops macvlan_ethtool_ops = { 1143 .get_link = ethtool_op_get_link, 1144 .get_link_ksettings = macvlan_ethtool_get_link_ksettings, 1145 .get_drvinfo = macvlan_ethtool_get_drvinfo, 1146 .get_ts_info = macvlan_ethtool_get_ts_info, 1147 }; 1148 1149 static const struct net_device_ops macvlan_netdev_ops = { 1150 .ndo_init = macvlan_init, 1151 .ndo_uninit = macvlan_uninit, 1152 .ndo_open = macvlan_open, 1153 .ndo_stop = macvlan_stop, 1154 .ndo_start_xmit = macvlan_start_xmit, 1155 .ndo_change_mtu = macvlan_change_mtu, 1156 .ndo_do_ioctl = macvlan_do_ioctl, 1157 .ndo_fix_features = macvlan_fix_features, 1158 .ndo_change_rx_flags = macvlan_change_rx_flags, 1159 .ndo_set_mac_address = macvlan_set_mac_address, 1160 .ndo_set_rx_mode = macvlan_set_mac_lists, 1161 .ndo_get_stats64 = macvlan_dev_get_stats64, 1162 .ndo_validate_addr = eth_validate_addr, 1163 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid, 1164 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid, 1165 .ndo_fdb_add = macvlan_fdb_add, 1166 .ndo_fdb_del = macvlan_fdb_del, 1167 .ndo_fdb_dump = ndo_dflt_fdb_dump, 1168 .ndo_get_lock_subclass = macvlan_get_nest_level, 1169 #ifdef CONFIG_NET_POLL_CONTROLLER 1170 .ndo_poll_controller = macvlan_dev_poll_controller, 1171 .ndo_netpoll_setup = macvlan_dev_netpoll_setup, 1172 .ndo_netpoll_cleanup = macvlan_dev_netpoll_cleanup, 1173 #endif 1174 .ndo_get_iflink = macvlan_dev_get_iflink, 1175 .ndo_features_check = passthru_features_check, 1176 .ndo_change_proto_down = dev_change_proto_down_generic, 1177 }; 1178 1179 void macvlan_common_setup(struct net_device *dev) 1180 { 1181 ether_setup(dev); 1182 1183 dev->min_mtu = 0; 1184 dev->max_mtu = ETH_MAX_MTU; 1185 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1186 netif_keep_dst(dev); 1187 dev->priv_flags |= IFF_UNICAST_FLT; 1188 dev->netdev_ops = &macvlan_netdev_ops; 1189 dev->needs_free_netdev = true; 1190 dev->header_ops = &macvlan_hard_header_ops; 1191 dev->ethtool_ops = &macvlan_ethtool_ops; 1192 } 1193 EXPORT_SYMBOL_GPL(macvlan_common_setup); 1194 1195 static void macvlan_setup(struct net_device *dev) 1196 { 1197 macvlan_common_setup(dev); 1198 dev->priv_flags |= IFF_NO_QUEUE; 1199 } 1200 1201 static int macvlan_port_create(struct net_device *dev) 1202 { 1203 struct macvlan_port *port; 1204 unsigned int i; 1205 int err; 1206 1207 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) 1208 return -EINVAL; 1209 1210 if (netdev_is_rx_handler_busy(dev)) 1211 return -EBUSY; 1212 1213 port = kzalloc(sizeof(*port), GFP_KERNEL); 1214 if (port == NULL) 1215 return -ENOMEM; 1216 1217 port->dev = dev; 1218 ether_addr_copy(port->perm_addr, dev->dev_addr); 1219 INIT_LIST_HEAD(&port->vlans); 1220 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1221 INIT_HLIST_HEAD(&port->vlan_hash[i]); 1222 for (i = 0; i < MACVLAN_HASH_SIZE; i++) 1223 INIT_HLIST_HEAD(&port->vlan_source_hash[i]); 1224 1225 skb_queue_head_init(&port->bc_queue); 1226 INIT_WORK(&port->bc_work, macvlan_process_broadcast); 1227 1228 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port); 1229 if (err) 1230 kfree(port); 1231 else 1232 dev->priv_flags |= IFF_MACVLAN_PORT; 1233 return err; 1234 } 1235 1236 static void macvlan_port_destroy(struct net_device *dev) 1237 { 1238 struct macvlan_port *port = macvlan_port_get_rtnl(dev); 1239 struct sk_buff *skb; 1240 1241 dev->priv_flags &= ~IFF_MACVLAN_PORT; 1242 netdev_rx_handler_unregister(dev); 1243 1244 /* After this point, no packet can schedule bc_work anymore, 1245 * but we need to cancel it and purge left skbs if any. 1246 */ 1247 cancel_work_sync(&port->bc_work); 1248 1249 while ((skb = __skb_dequeue(&port->bc_queue))) { 1250 const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src; 1251 1252 if (src) 1253 dev_put(src->dev); 1254 1255 kfree_skb(skb); 1256 } 1257 1258 /* If the lower device address has been changed by passthru 1259 * macvlan, put it back. 1260 */ 1261 if (macvlan_passthru(port) && 1262 !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) { 1263 struct sockaddr sa; 1264 1265 sa.sa_family = port->dev->type; 1266 memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len); 1267 dev_set_mac_address(port->dev, &sa, NULL); 1268 } 1269 1270 kfree(port); 1271 } 1272 1273 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[], 1274 struct netlink_ext_ack *extack) 1275 { 1276 if (tb[IFLA_ADDRESS]) { 1277 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1278 return -EINVAL; 1279 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1280 return -EADDRNOTAVAIL; 1281 } 1282 1283 if (!data) 1284 return 0; 1285 1286 if (data[IFLA_MACVLAN_FLAGS] && 1287 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC) 1288 return -EINVAL; 1289 1290 if (data[IFLA_MACVLAN_MODE]) { 1291 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) { 1292 case MACVLAN_MODE_PRIVATE: 1293 case MACVLAN_MODE_VEPA: 1294 case MACVLAN_MODE_BRIDGE: 1295 case MACVLAN_MODE_PASSTHRU: 1296 case MACVLAN_MODE_SOURCE: 1297 break; 1298 default: 1299 return -EINVAL; 1300 } 1301 } 1302 1303 if (data[IFLA_MACVLAN_MACADDR_MODE]) { 1304 switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) { 1305 case MACVLAN_MACADDR_ADD: 1306 case MACVLAN_MACADDR_DEL: 1307 case MACVLAN_MACADDR_FLUSH: 1308 case MACVLAN_MACADDR_SET: 1309 break; 1310 default: 1311 return -EINVAL; 1312 } 1313 } 1314 1315 if (data[IFLA_MACVLAN_MACADDR]) { 1316 if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN) 1317 return -EINVAL; 1318 1319 if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR]))) 1320 return -EADDRNOTAVAIL; 1321 } 1322 1323 if (data[IFLA_MACVLAN_MACADDR_COUNT]) 1324 return -EINVAL; 1325 1326 return 0; 1327 } 1328 1329 /** 1330 * reconfigure list of remote source mac address 1331 * (only for macvlan devices in source mode) 1332 * Note regarding alignment: all netlink data is aligned to 4 Byte, which 1333 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage. 1334 */ 1335 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode, 1336 struct nlattr *data[]) 1337 { 1338 char *addr = NULL; 1339 int ret, rem, len; 1340 struct nlattr *nla, *head; 1341 struct macvlan_source_entry *entry; 1342 1343 if (data[IFLA_MACVLAN_MACADDR]) 1344 addr = nla_data(data[IFLA_MACVLAN_MACADDR]); 1345 1346 if (mode == MACVLAN_MACADDR_ADD) { 1347 if (!addr) 1348 return -EINVAL; 1349 1350 return macvlan_hash_add_source(vlan, addr); 1351 1352 } else if (mode == MACVLAN_MACADDR_DEL) { 1353 if (!addr) 1354 return -EINVAL; 1355 1356 entry = macvlan_hash_lookup_source(vlan, addr); 1357 if (entry) { 1358 macvlan_hash_del_source(entry); 1359 vlan->macaddr_count--; 1360 } 1361 } else if (mode == MACVLAN_MACADDR_FLUSH) { 1362 macvlan_flush_sources(vlan->port, vlan); 1363 } else if (mode == MACVLAN_MACADDR_SET) { 1364 macvlan_flush_sources(vlan->port, vlan); 1365 1366 if (addr) { 1367 ret = macvlan_hash_add_source(vlan, addr); 1368 if (ret) 1369 return ret; 1370 } 1371 1372 if (!data || !data[IFLA_MACVLAN_MACADDR_DATA]) 1373 return 0; 1374 1375 head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]); 1376 len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]); 1377 1378 nla_for_each_attr(nla, head, len, rem) { 1379 if (nla_type(nla) != IFLA_MACVLAN_MACADDR || 1380 nla_len(nla) != ETH_ALEN) 1381 continue; 1382 1383 addr = nla_data(nla); 1384 ret = macvlan_hash_add_source(vlan, addr); 1385 if (ret) 1386 return ret; 1387 } 1388 } else { 1389 return -EINVAL; 1390 } 1391 1392 return 0; 1393 } 1394 1395 int macvlan_common_newlink(struct net *src_net, struct net_device *dev, 1396 struct nlattr *tb[], struct nlattr *data[], 1397 struct netlink_ext_ack *extack) 1398 { 1399 struct macvlan_dev *vlan = netdev_priv(dev); 1400 struct macvlan_port *port; 1401 struct net_device *lowerdev; 1402 int err; 1403 int macmode; 1404 bool create = false; 1405 1406 if (!tb[IFLA_LINK]) 1407 return -EINVAL; 1408 1409 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 1410 if (lowerdev == NULL) 1411 return -ENODEV; 1412 1413 /* When creating macvlans or macvtaps on top of other macvlans - use 1414 * the real device as the lowerdev. 1415 */ 1416 if (netif_is_macvlan(lowerdev)) 1417 lowerdev = macvlan_dev_real_dev(lowerdev); 1418 1419 if (!tb[IFLA_MTU]) 1420 dev->mtu = lowerdev->mtu; 1421 else if (dev->mtu > lowerdev->mtu) 1422 return -EINVAL; 1423 1424 /* MTU range: 68 - lowerdev->max_mtu */ 1425 dev->min_mtu = ETH_MIN_MTU; 1426 dev->max_mtu = lowerdev->max_mtu; 1427 1428 if (!tb[IFLA_ADDRESS]) 1429 eth_hw_addr_random(dev); 1430 1431 if (!netif_is_macvlan_port(lowerdev)) { 1432 err = macvlan_port_create(lowerdev); 1433 if (err < 0) 1434 return err; 1435 create = true; 1436 } 1437 port = macvlan_port_get_rtnl(lowerdev); 1438 1439 /* Only 1 macvlan device can be created in passthru mode */ 1440 if (macvlan_passthru(port)) { 1441 /* The macvlan port must be not created this time, 1442 * still goto destroy_macvlan_port for readability. 1443 */ 1444 err = -EINVAL; 1445 goto destroy_macvlan_port; 1446 } 1447 1448 vlan->lowerdev = lowerdev; 1449 vlan->dev = dev; 1450 vlan->port = port; 1451 vlan->set_features = MACVLAN_FEATURES; 1452 vlan->nest_level = dev_get_nest_level(lowerdev) + 1; 1453 1454 vlan->mode = MACVLAN_MODE_VEPA; 1455 if (data && data[IFLA_MACVLAN_MODE]) 1456 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1457 1458 if (data && data[IFLA_MACVLAN_FLAGS]) 1459 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1460 1461 if (vlan->mode == MACVLAN_MODE_PASSTHRU) { 1462 if (port->count) { 1463 err = -EINVAL; 1464 goto destroy_macvlan_port; 1465 } 1466 macvlan_set_passthru(port); 1467 eth_hw_addr_inherit(dev, lowerdev); 1468 } 1469 1470 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1471 if (vlan->mode != MACVLAN_MODE_SOURCE) { 1472 err = -EINVAL; 1473 goto destroy_macvlan_port; 1474 } 1475 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1476 err = macvlan_changelink_sources(vlan, macmode, data); 1477 if (err) 1478 goto destroy_macvlan_port; 1479 } 1480 1481 err = register_netdevice(dev); 1482 if (err < 0) 1483 goto destroy_macvlan_port; 1484 1485 dev->priv_flags |= IFF_MACVLAN; 1486 err = netdev_upper_dev_link(lowerdev, dev, extack); 1487 if (err) 1488 goto unregister_netdev; 1489 1490 list_add_tail_rcu(&vlan->list, &port->vlans); 1491 netif_stacked_transfer_operstate(lowerdev, dev); 1492 linkwatch_fire_event(dev); 1493 1494 return 0; 1495 1496 unregister_netdev: 1497 /* macvlan_uninit would free the macvlan port */ 1498 unregister_netdevice(dev); 1499 return err; 1500 destroy_macvlan_port: 1501 /* the macvlan port may be freed by macvlan_uninit when fail to register. 1502 * so we destroy the macvlan port only when it's valid. 1503 */ 1504 if (create && macvlan_port_get_rtnl(lowerdev)) 1505 macvlan_port_destroy(port->dev); 1506 return err; 1507 } 1508 EXPORT_SYMBOL_GPL(macvlan_common_newlink); 1509 1510 static int macvlan_newlink(struct net *src_net, struct net_device *dev, 1511 struct nlattr *tb[], struct nlattr *data[], 1512 struct netlink_ext_ack *extack) 1513 { 1514 return macvlan_common_newlink(src_net, dev, tb, data, extack); 1515 } 1516 1517 void macvlan_dellink(struct net_device *dev, struct list_head *head) 1518 { 1519 struct macvlan_dev *vlan = netdev_priv(dev); 1520 1521 if (vlan->mode == MACVLAN_MODE_SOURCE) 1522 macvlan_flush_sources(vlan->port, vlan); 1523 list_del_rcu(&vlan->list); 1524 unregister_netdevice_queue(dev, head); 1525 netdev_upper_dev_unlink(vlan->lowerdev, dev); 1526 } 1527 EXPORT_SYMBOL_GPL(macvlan_dellink); 1528 1529 static int macvlan_changelink(struct net_device *dev, 1530 struct nlattr *tb[], struct nlattr *data[], 1531 struct netlink_ext_ack *extack) 1532 { 1533 struct macvlan_dev *vlan = netdev_priv(dev); 1534 enum macvlan_mode mode; 1535 bool set_mode = false; 1536 enum macvlan_macaddr_mode macmode; 1537 int ret; 1538 1539 /* Validate mode, but don't set yet: setting flags may fail. */ 1540 if (data && data[IFLA_MACVLAN_MODE]) { 1541 set_mode = true; 1542 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]); 1543 /* Passthrough mode can't be set or cleared dynamically */ 1544 if ((mode == MACVLAN_MODE_PASSTHRU) != 1545 (vlan->mode == MACVLAN_MODE_PASSTHRU)) 1546 return -EINVAL; 1547 if (vlan->mode == MACVLAN_MODE_SOURCE && 1548 vlan->mode != mode) 1549 macvlan_flush_sources(vlan->port, vlan); 1550 } 1551 1552 if (data && data[IFLA_MACVLAN_FLAGS]) { 1553 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]); 1554 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC; 1555 if (macvlan_passthru(vlan->port) && promisc) { 1556 int err; 1557 1558 if (flags & MACVLAN_FLAG_NOPROMISC) 1559 err = dev_set_promiscuity(vlan->lowerdev, -1); 1560 else 1561 err = dev_set_promiscuity(vlan->lowerdev, 1); 1562 if (err < 0) 1563 return err; 1564 } 1565 vlan->flags = flags; 1566 } 1567 if (set_mode) 1568 vlan->mode = mode; 1569 if (data && data[IFLA_MACVLAN_MACADDR_MODE]) { 1570 if (vlan->mode != MACVLAN_MODE_SOURCE) 1571 return -EINVAL; 1572 macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]); 1573 ret = macvlan_changelink_sources(vlan, macmode, data); 1574 if (ret) 1575 return ret; 1576 } 1577 return 0; 1578 } 1579 1580 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan) 1581 { 1582 if (vlan->macaddr_count == 0) 1583 return 0; 1584 return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */ 1585 + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN); 1586 } 1587 1588 static size_t macvlan_get_size(const struct net_device *dev) 1589 { 1590 struct macvlan_dev *vlan = netdev_priv(dev); 1591 1592 return (0 1593 + nla_total_size(4) /* IFLA_MACVLAN_MODE */ 1594 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */ 1595 + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */ 1596 + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */ 1597 ); 1598 } 1599 1600 static int macvlan_fill_info_macaddr(struct sk_buff *skb, 1601 const struct macvlan_dev *vlan, 1602 const int i) 1603 { 1604 struct hlist_head *h = &vlan->port->vlan_source_hash[i]; 1605 struct macvlan_source_entry *entry; 1606 1607 hlist_for_each_entry_rcu(entry, h, hlist) { 1608 if (entry->vlan != vlan) 1609 continue; 1610 if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr)) 1611 return 1; 1612 } 1613 return 0; 1614 } 1615 1616 static int macvlan_fill_info(struct sk_buff *skb, 1617 const struct net_device *dev) 1618 { 1619 struct macvlan_dev *vlan = netdev_priv(dev); 1620 int i; 1621 struct nlattr *nest; 1622 1623 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode)) 1624 goto nla_put_failure; 1625 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags)) 1626 goto nla_put_failure; 1627 if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count)) 1628 goto nla_put_failure; 1629 if (vlan->macaddr_count > 0) { 1630 nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA); 1631 if (nest == NULL) 1632 goto nla_put_failure; 1633 1634 for (i = 0; i < MACVLAN_HASH_SIZE; i++) { 1635 if (macvlan_fill_info_macaddr(skb, vlan, i)) 1636 goto nla_put_failure; 1637 } 1638 nla_nest_end(skb, nest); 1639 } 1640 return 0; 1641 1642 nla_put_failure: 1643 return -EMSGSIZE; 1644 } 1645 1646 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = { 1647 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 }, 1648 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 }, 1649 [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 }, 1650 [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1651 [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED }, 1652 [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 }, 1653 }; 1654 1655 int macvlan_link_register(struct rtnl_link_ops *ops) 1656 { 1657 /* common fields */ 1658 ops->validate = macvlan_validate; 1659 ops->maxtype = IFLA_MACVLAN_MAX; 1660 ops->policy = macvlan_policy; 1661 ops->changelink = macvlan_changelink; 1662 ops->get_size = macvlan_get_size; 1663 ops->fill_info = macvlan_fill_info; 1664 1665 return rtnl_link_register(ops); 1666 }; 1667 EXPORT_SYMBOL_GPL(macvlan_link_register); 1668 1669 static struct net *macvlan_get_link_net(const struct net_device *dev) 1670 { 1671 return dev_net(macvlan_dev_real_dev(dev)); 1672 } 1673 1674 static struct rtnl_link_ops macvlan_link_ops = { 1675 .kind = "macvlan", 1676 .setup = macvlan_setup, 1677 .newlink = macvlan_newlink, 1678 .dellink = macvlan_dellink, 1679 .get_link_net = macvlan_get_link_net, 1680 .priv_size = sizeof(struct macvlan_dev), 1681 }; 1682 1683 static int macvlan_device_event(struct notifier_block *unused, 1684 unsigned long event, void *ptr) 1685 { 1686 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1687 struct macvlan_dev *vlan, *next; 1688 struct macvlan_port *port; 1689 LIST_HEAD(list_kill); 1690 1691 if (!netif_is_macvlan_port(dev)) 1692 return NOTIFY_DONE; 1693 1694 port = macvlan_port_get_rtnl(dev); 1695 1696 switch (event) { 1697 case NETDEV_UP: 1698 case NETDEV_DOWN: 1699 case NETDEV_CHANGE: 1700 list_for_each_entry(vlan, &port->vlans, list) 1701 netif_stacked_transfer_operstate(vlan->lowerdev, 1702 vlan->dev); 1703 break; 1704 case NETDEV_FEAT_CHANGE: 1705 list_for_each_entry(vlan, &port->vlans, list) { 1706 vlan->dev->gso_max_size = dev->gso_max_size; 1707 vlan->dev->gso_max_segs = dev->gso_max_segs; 1708 netdev_update_features(vlan->dev); 1709 } 1710 break; 1711 case NETDEV_CHANGEMTU: 1712 list_for_each_entry(vlan, &port->vlans, list) { 1713 if (vlan->dev->mtu <= dev->mtu) 1714 continue; 1715 dev_set_mtu(vlan->dev, dev->mtu); 1716 } 1717 break; 1718 case NETDEV_CHANGEADDR: 1719 if (!macvlan_passthru(port)) 1720 return NOTIFY_DONE; 1721 1722 vlan = list_first_entry_or_null(&port->vlans, 1723 struct macvlan_dev, 1724 list); 1725 1726 if (macvlan_sync_address(vlan->dev, dev->dev_addr)) 1727 return NOTIFY_BAD; 1728 1729 break; 1730 case NETDEV_UNREGISTER: 1731 /* twiddle thumbs on netns device moves */ 1732 if (dev->reg_state != NETREG_UNREGISTERING) 1733 break; 1734 1735 list_for_each_entry_safe(vlan, next, &port->vlans, list) 1736 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill); 1737 unregister_netdevice_many(&list_kill); 1738 break; 1739 case NETDEV_PRE_TYPE_CHANGE: 1740 /* Forbid underlaying device to change its type. */ 1741 return NOTIFY_BAD; 1742 1743 case NETDEV_NOTIFY_PEERS: 1744 case NETDEV_BONDING_FAILOVER: 1745 case NETDEV_RESEND_IGMP: 1746 /* Propagate to all vlans */ 1747 list_for_each_entry(vlan, &port->vlans, list) 1748 call_netdevice_notifiers(event, vlan->dev); 1749 } 1750 return NOTIFY_DONE; 1751 } 1752 1753 static struct notifier_block macvlan_notifier_block __read_mostly = { 1754 .notifier_call = macvlan_device_event, 1755 }; 1756 1757 static int __init macvlan_init_module(void) 1758 { 1759 int err; 1760 1761 register_netdevice_notifier(&macvlan_notifier_block); 1762 1763 err = macvlan_link_register(&macvlan_link_ops); 1764 if (err < 0) 1765 goto err1; 1766 return 0; 1767 err1: 1768 unregister_netdevice_notifier(&macvlan_notifier_block); 1769 return err; 1770 } 1771 1772 static void __exit macvlan_cleanup_module(void) 1773 { 1774 rtnl_link_unregister(&macvlan_link_ops); 1775 unregister_netdevice_notifier(&macvlan_notifier_block); 1776 } 1777 1778 module_init(macvlan_init_module); 1779 module_exit(macvlan_cleanup_module); 1780 1781 MODULE_LICENSE("GPL"); 1782 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>"); 1783 MODULE_DESCRIPTION("Driver for MAC address based VLANs"); 1784 MODULE_ALIAS_RTNL_LINK("macvlan"); 1785