linux/drivers/net/macvlan.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
   4 *
   5 * The code this is based on carried the following copyright notice:
   6 * ---
   7 * (C) Copyright 2001-2006
   8 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
   9 * Re-worked by Ben Greear <greearb@candelatech.com>
  10 * ---
  11 */
  12#include <linux/kernel.h>
  13#include <linux/types.h>
  14#include <linux/module.h>
  15#include <linux/init.h>
  16#include <linux/errno.h>
  17#include <linux/slab.h>
  18#include <linux/string.h>
  19#include <linux/rculist.h>
  20#include <linux/notifier.h>
  21#include <linux/netdevice.h>
  22#include <linux/etherdevice.h>
  23#include <linux/net_tstamp.h>
  24#include <linux/ethtool.h>
  25#include <linux/if_arp.h>
  26#include <linux/if_vlan.h>
  27#include <linux/if_link.h>
  28#include <linux/if_macvlan.h>
  29#include <linux/hash.h>
  30#include <linux/workqueue.h>
  31#include <net/rtnetlink.h>
  32#include <net/xfrm.h>
  33#include <linux/netpoll.h>
  34#include <linux/phy.h>
  35
  36#define MACVLAN_HASH_BITS       8
  37#define MACVLAN_HASH_SIZE       (1<<MACVLAN_HASH_BITS)
  38#define MACVLAN_DEFAULT_BC_QUEUE_LEN    1000
  39
  40#define MACVLAN_F_PASSTHRU      1
  41#define MACVLAN_F_ADDRCHANGE    2
  42
  43struct macvlan_port {
  44        struct net_device       *dev;
  45        struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
  46        struct list_head        vlans;
  47        struct sk_buff_head     bc_queue;
  48        struct work_struct      bc_work;
  49        u32                     bc_queue_len_used;
  50        u32                     flags;
  51        int                     count;
  52        struct hlist_head       vlan_source_hash[MACVLAN_HASH_SIZE];
  53        DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
  54        unsigned char           perm_addr[ETH_ALEN];
  55};
  56
  57struct macvlan_source_entry {
  58        struct hlist_node       hlist;
  59        struct macvlan_dev      *vlan;
  60        unsigned char           addr[6+2] __aligned(sizeof(u16));
  61        struct rcu_head         rcu;
  62};
  63
  64struct macvlan_skb_cb {
  65        const struct macvlan_dev *src;
  66};
  67
  68#define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
  69
  70static void macvlan_port_destroy(struct net_device *dev);
  71static void update_port_bc_queue_len(struct macvlan_port *port);
  72
  73static inline bool macvlan_passthru(const struct macvlan_port *port)
  74{
  75        return port->flags & MACVLAN_F_PASSTHRU;
  76}
  77
  78static inline void macvlan_set_passthru(struct macvlan_port *port)
  79{
  80        port->flags |= MACVLAN_F_PASSTHRU;
  81}
  82
  83static inline bool macvlan_addr_change(const struct macvlan_port *port)
  84{
  85        return port->flags & MACVLAN_F_ADDRCHANGE;
  86}
  87
  88static inline void macvlan_set_addr_change(struct macvlan_port *port)
  89{
  90        port->flags |= MACVLAN_F_ADDRCHANGE;
  91}
  92
  93static inline void macvlan_clear_addr_change(struct macvlan_port *port)
  94{
  95        port->flags &= ~MACVLAN_F_ADDRCHANGE;
  96}
  97
  98/* Hash Ethernet address */
  99static u32 macvlan_eth_hash(const unsigned char *addr)
 100{
 101        u64 value = get_unaligned((u64 *)addr);
 102
 103        /* only want 6 bytes */
 104#ifdef __BIG_ENDIAN
 105        value >>= 16;
 106#else
 107        value <<= 16;
 108#endif
 109        return hash_64(value, MACVLAN_HASH_BITS);
 110}
 111
 112static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
 113{
 114        return rcu_dereference(dev->rx_handler_data);
 115}
 116
 117static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
 118{
 119        return rtnl_dereference(dev->rx_handler_data);
 120}
 121
 122static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
 123                                               const unsigned char *addr)
 124{
 125        struct macvlan_dev *vlan;
 126        u32 idx = macvlan_eth_hash(addr);
 127
 128        hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
 129                                 lockdep_rtnl_is_held()) {
 130                if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
 131                        return vlan;
 132        }
 133        return NULL;
 134}
 135
 136static struct macvlan_source_entry *macvlan_hash_lookup_source(
 137        const struct macvlan_dev *vlan,
 138        const unsigned char *addr)
 139{
 140        struct macvlan_source_entry *entry;
 141        u32 idx = macvlan_eth_hash(addr);
 142        struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
 143
 144        hlist_for_each_entry_rcu(entry, h, hlist) {
 145                if (ether_addr_equal_64bits(entry->addr, addr) &&
 146                    entry->vlan == vlan)
 147                        return entry;
 148        }
 149        return NULL;
 150}
 151
 152static int macvlan_hash_add_source(struct macvlan_dev *vlan,
 153                                   const unsigned char *addr)
 154{
 155        struct macvlan_port *port = vlan->port;
 156        struct macvlan_source_entry *entry;
 157        struct hlist_head *h;
 158
 159        entry = macvlan_hash_lookup_source(vlan, addr);
 160        if (entry)
 161                return 0;
 162
 163        entry = kmalloc(sizeof(*entry), GFP_KERNEL);
 164        if (!entry)
 165                return -ENOMEM;
 166
 167        ether_addr_copy(entry->addr, addr);
 168        entry->vlan = vlan;
 169        h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
 170        hlist_add_head_rcu(&entry->hlist, h);
 171        vlan->macaddr_count++;
 172
 173        return 0;
 174}
 175
 176static void macvlan_hash_add(struct macvlan_dev *vlan)
 177{
 178        struct macvlan_port *port = vlan->port;
 179        const unsigned char *addr = vlan->dev->dev_addr;
 180        u32 idx = macvlan_eth_hash(addr);
 181
 182        hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
 183}
 184
 185static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
 186{
 187        hlist_del_rcu(&entry->hlist);
 188        kfree_rcu(entry, rcu);
 189}
 190
 191static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
 192{
 193        hlist_del_rcu(&vlan->hlist);
 194        if (sync)
 195                synchronize_rcu();
 196}
 197
 198static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
 199                                        const unsigned char *addr)
 200{
 201        macvlan_hash_del(vlan, true);
 202        /* Now that we are unhashed it is safe to change the device
 203         * address without confusing packet delivery.
 204         */
 205        memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
 206        macvlan_hash_add(vlan);
 207}
 208
 209static bool macvlan_addr_busy(const struct macvlan_port *port,
 210                              const unsigned char *addr)
 211{
 212        /* Test to see if the specified address is
 213         * currently in use by the underlying device or
 214         * another macvlan.
 215         */
 216        if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
 217            ether_addr_equal_64bits(port->dev->dev_addr, addr))
 218                return true;
 219
 220        if (macvlan_hash_lookup(port, addr))
 221                return true;
 222
 223        return false;
 224}
 225
 226
 227static int macvlan_broadcast_one(struct sk_buff *skb,
 228                                 const struct macvlan_dev *vlan,
 229                                 const struct ethhdr *eth, bool local)
 230{
 231        struct net_device *dev = vlan->dev;
 232
 233        if (local)
 234                return __dev_forward_skb(dev, skb);
 235
 236        skb->dev = dev;
 237        if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
 238                skb->pkt_type = PACKET_BROADCAST;
 239        else
 240                skb->pkt_type = PACKET_MULTICAST;
 241
 242        return 0;
 243}
 244
 245static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
 246{
 247        return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
 248}
 249
 250
 251static unsigned int mc_hash(const struct macvlan_dev *vlan,
 252                            const unsigned char *addr)
 253{
 254        u32 val = __get_unaligned_cpu32(addr + 2);
 255
 256        val ^= macvlan_hash_mix(vlan);
 257        return hash_32(val, MACVLAN_MC_FILTER_BITS);
 258}
 259
 260static void macvlan_broadcast(struct sk_buff *skb,
 261                              const struct macvlan_port *port,
 262                              struct net_device *src,
 263                              enum macvlan_mode mode)
 264{
 265        const struct ethhdr *eth = eth_hdr(skb);
 266        const struct macvlan_dev *vlan;
 267        struct sk_buff *nskb;
 268        unsigned int i;
 269        int err;
 270        unsigned int hash;
 271
 272        if (skb->protocol == htons(ETH_P_PAUSE))
 273                return;
 274
 275        hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) {
 276                if (vlan->dev == src || !(vlan->mode & mode))
 277                        continue;
 278
 279                hash = mc_hash(vlan, eth->h_dest);
 280                if (!test_bit(hash, vlan->mc_filter))
 281                        continue;
 282
 283                err = NET_RX_DROP;
 284                nskb = skb_clone(skb, GFP_ATOMIC);
 285                if (likely(nskb))
 286                        err = macvlan_broadcast_one(nskb, vlan, eth,
 287                                        mode == MACVLAN_MODE_BRIDGE) ?:
 288                              netif_rx_ni(nskb);
 289                macvlan_count_rx(vlan, skb->len + ETH_HLEN,
 290                                 err == NET_RX_SUCCESS, true);
 291        }
 292}
 293
 294static void macvlan_process_broadcast(struct work_struct *w)
 295{
 296        struct macvlan_port *port = container_of(w, struct macvlan_port,
 297                                                 bc_work);
 298        struct sk_buff *skb;
 299        struct sk_buff_head list;
 300
 301        __skb_queue_head_init(&list);
 302
 303        spin_lock_bh(&port->bc_queue.lock);
 304        skb_queue_splice_tail_init(&port->bc_queue, &list);
 305        spin_unlock_bh(&port->bc_queue.lock);
 306
 307        while ((skb = __skb_dequeue(&list))) {
 308                const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
 309
 310                rcu_read_lock();
 311
 312                if (!src)
 313                        /* frame comes from an external address */
 314                        macvlan_broadcast(skb, port, NULL,
 315                                          MACVLAN_MODE_PRIVATE |
 316                                          MACVLAN_MODE_VEPA    |
 317                                          MACVLAN_MODE_PASSTHRU|
 318                                          MACVLAN_MODE_BRIDGE);
 319                else if (src->mode == MACVLAN_MODE_VEPA)
 320                        /* flood to everyone except source */
 321                        macvlan_broadcast(skb, port, src->dev,
 322                                          MACVLAN_MODE_VEPA |
 323                                          MACVLAN_MODE_BRIDGE);
 324                else
 325                        /*
 326                         * flood only to VEPA ports, bridge ports
 327                         * already saw the frame on the way out.
 328                         */
 329                        macvlan_broadcast(skb, port, src->dev,
 330                                          MACVLAN_MODE_VEPA);
 331
 332                rcu_read_unlock();
 333
 334                if (src)
 335                        dev_put(src->dev);
 336                consume_skb(skb);
 337
 338                cond_resched();
 339        }
 340}
 341
 342static void macvlan_broadcast_enqueue(struct macvlan_port *port,
 343                                      const struct macvlan_dev *src,
 344                                      struct sk_buff *skb)
 345{
 346        struct sk_buff *nskb;
 347        int err = -ENOMEM;
 348
 349        nskb = skb_clone(skb, GFP_ATOMIC);
 350        if (!nskb)
 351                goto err;
 352
 353        MACVLAN_SKB_CB(nskb)->src = src;
 354
 355        spin_lock(&port->bc_queue.lock);
 356        if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) {
 357                if (src)
 358                        dev_hold(src->dev);
 359                __skb_queue_tail(&port->bc_queue, nskb);
 360                err = 0;
 361        }
 362        spin_unlock(&port->bc_queue.lock);
 363
 364        schedule_work(&port->bc_work);
 365
 366        if (err)
 367                goto free_nskb;
 368
 369        return;
 370
 371free_nskb:
 372        kfree_skb(nskb);
 373err:
 374        atomic_long_inc(&skb->dev->rx_dropped);
 375}
 376
 377static void macvlan_flush_sources(struct macvlan_port *port,
 378                                  struct macvlan_dev *vlan)
 379{
 380        struct macvlan_source_entry *entry;
 381        struct hlist_node *next;
 382        int i;
 383
 384        hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
 385                if (entry->vlan == vlan)
 386                        macvlan_hash_del_source(entry);
 387
 388        vlan->macaddr_count = 0;
 389}
 390
 391static void macvlan_forward_source_one(struct sk_buff *skb,
 392                                       struct macvlan_dev *vlan)
 393{
 394        struct sk_buff *nskb;
 395        struct net_device *dev;
 396        int len;
 397        int ret;
 398
 399        dev = vlan->dev;
 400        if (unlikely(!(dev->flags & IFF_UP)))
 401                return;
 402
 403        nskb = skb_clone(skb, GFP_ATOMIC);
 404        if (!nskb)
 405                return;
 406
 407        len = nskb->len + ETH_HLEN;
 408        nskb->dev = dev;
 409
 410        if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
 411                nskb->pkt_type = PACKET_HOST;
 412
 413        ret = netif_rx(nskb);
 414        macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
 415}
 416
 417static bool macvlan_forward_source(struct sk_buff *skb,
 418                                   struct macvlan_port *port,
 419                                   const unsigned char *addr)
 420{
 421        struct macvlan_source_entry *entry;
 422        u32 idx = macvlan_eth_hash(addr);
 423        struct hlist_head *h = &port->vlan_source_hash[idx];
 424        bool consume = false;
 425
 426        hlist_for_each_entry_rcu(entry, h, hlist) {
 427                if (ether_addr_equal_64bits(entry->addr, addr)) {
 428                        if (entry->vlan->flags & MACVLAN_FLAG_NODST)
 429                                consume = true;
 430                        macvlan_forward_source_one(skb, entry->vlan);
 431                }
 432        }
 433
 434        return consume;
 435}
 436
 437/* called under rcu_read_lock() from netif_receive_skb */
 438static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
 439{
 440        struct macvlan_port *port;
 441        struct sk_buff *skb = *pskb;
 442        const struct ethhdr *eth = eth_hdr(skb);
 443        const struct macvlan_dev *vlan;
 444        const struct macvlan_dev *src;
 445        struct net_device *dev;
 446        unsigned int len = 0;
 447        int ret;
 448        rx_handler_result_t handle_res;
 449
 450        /* Packets from dev_loopback_xmit() do not have L2 header, bail out */
 451        if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
 452                return RX_HANDLER_PASS;
 453
 454        port = macvlan_port_get_rcu(skb->dev);
 455        if (is_multicast_ether_addr(eth->h_dest)) {
 456                unsigned int hash;
 457
 458                skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
 459                if (!skb)
 460                        return RX_HANDLER_CONSUMED;
 461                *pskb = skb;
 462                eth = eth_hdr(skb);
 463                if (macvlan_forward_source(skb, port, eth->h_source))
 464                        return RX_HANDLER_CONSUMED;
 465                src = macvlan_hash_lookup(port, eth->h_source);
 466                if (src && src->mode != MACVLAN_MODE_VEPA &&
 467                    src->mode != MACVLAN_MODE_BRIDGE) {
 468                        /* forward to original port. */
 469                        vlan = src;
 470                        ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
 471                              netif_rx(skb);
 472                        handle_res = RX_HANDLER_CONSUMED;
 473                        goto out;
 474                }
 475
 476                hash = mc_hash(NULL, eth->h_dest);
 477                if (test_bit(hash, port->mc_filter))
 478                        macvlan_broadcast_enqueue(port, src, skb);
 479
 480                return RX_HANDLER_PASS;
 481        }
 482
 483        if (macvlan_forward_source(skb, port, eth->h_source))
 484                return RX_HANDLER_CONSUMED;
 485        if (macvlan_passthru(port))
 486                vlan = list_first_or_null_rcu(&port->vlans,
 487                                              struct macvlan_dev, list);
 488        else
 489                vlan = macvlan_hash_lookup(port, eth->h_dest);
 490        if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
 491                return RX_HANDLER_PASS;
 492
 493        dev = vlan->dev;
 494        if (unlikely(!(dev->flags & IFF_UP))) {
 495                kfree_skb(skb);
 496                return RX_HANDLER_CONSUMED;
 497        }
 498        len = skb->len + ETH_HLEN;
 499        skb = skb_share_check(skb, GFP_ATOMIC);
 500        if (!skb) {
 501                ret = NET_RX_DROP;
 502                handle_res = RX_HANDLER_CONSUMED;
 503                goto out;
 504        }
 505
 506        *pskb = skb;
 507        skb->dev = dev;
 508        skb->pkt_type = PACKET_HOST;
 509
 510        ret = NET_RX_SUCCESS;
 511        handle_res = RX_HANDLER_ANOTHER;
 512out:
 513        macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
 514        return handle_res;
 515}
 516
 517static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 518{
 519        const struct macvlan_dev *vlan = netdev_priv(dev);
 520        const struct macvlan_port *port = vlan->port;
 521        const struct macvlan_dev *dest;
 522
 523        if (vlan->mode == MACVLAN_MODE_BRIDGE) {
 524                const struct ethhdr *eth = skb_eth_hdr(skb);
 525
 526                /* send to other bridge ports directly */
 527                if (is_multicast_ether_addr(eth->h_dest)) {
 528                        skb_reset_mac_header(skb);
 529                        macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
 530                        goto xmit_world;
 531                }
 532
 533                dest = macvlan_hash_lookup(port, eth->h_dest);
 534                if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
 535                        /* send to lowerdev first for its network taps */
 536                        dev_forward_skb(vlan->lowerdev, skb);
 537
 538                        return NET_XMIT_SUCCESS;
 539                }
 540        }
 541xmit_world:
 542        skb->dev = vlan->lowerdev;
 543        return dev_queue_xmit_accel(skb,
 544                                    netdev_get_sb_channel(dev) ? dev : NULL);
 545}
 546
 547static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
 548{
 549#ifdef CONFIG_NET_POLL_CONTROLLER
 550        return netpoll_send_skb(vlan->netpoll, skb);
 551#else
 552        BUG();
 553        return NETDEV_TX_OK;
 554#endif
 555}
 556
 557static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
 558                                      struct net_device *dev)
 559{
 560        struct macvlan_dev *vlan = netdev_priv(dev);
 561        unsigned int len = skb->len;
 562        int ret;
 563
 564        if (unlikely(netpoll_tx_running(dev)))
 565                return macvlan_netpoll_send_skb(vlan, skb);
 566
 567        ret = macvlan_queue_xmit(skb, dev);
 568
 569        if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
 570                struct vlan_pcpu_stats *pcpu_stats;
 571
 572                pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
 573                u64_stats_update_begin(&pcpu_stats->syncp);
 574                pcpu_stats->tx_packets++;
 575                pcpu_stats->tx_bytes += len;
 576                u64_stats_update_end(&pcpu_stats->syncp);
 577        } else {
 578                this_cpu_inc(vlan->pcpu_stats->tx_dropped);
 579        }
 580        return ret;
 581}
 582
 583static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 584                               unsigned short type, const void *daddr,
 585                               const void *saddr, unsigned len)
 586{
 587        const struct macvlan_dev *vlan = netdev_priv(dev);
 588        struct net_device *lowerdev = vlan->lowerdev;
 589
 590        return dev_hard_header(skb, lowerdev, type, daddr,
 591                               saddr ? : dev->dev_addr, len);
 592}
 593
 594static const struct header_ops macvlan_hard_header_ops = {
 595        .create         = macvlan_hard_header,
 596        .parse          = eth_header_parse,
 597        .cache          = eth_header_cache,
 598        .cache_update   = eth_header_cache_update,
 599};
 600
 601static int macvlan_open(struct net_device *dev)
 602{
 603        struct macvlan_dev *vlan = netdev_priv(dev);
 604        struct net_device *lowerdev = vlan->lowerdev;
 605        int err;
 606
 607        if (macvlan_passthru(vlan->port)) {
 608                if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
 609                        err = dev_set_promiscuity(lowerdev, 1);
 610                        if (err < 0)
 611                                goto out;
 612                }
 613                goto hash_add;
 614        }
 615
 616        err = -EADDRINUSE;
 617        if (macvlan_addr_busy(vlan->port, dev->dev_addr))
 618                goto out;
 619
 620        /* Attempt to populate accel_priv which is used to offload the L2
 621         * forwarding requests for unicast packets.
 622         */
 623        if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
 624                vlan->accel_priv =
 625                      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
 626
 627        /* If earlier attempt to offload failed, or accel_priv is not
 628         * populated we must add the unicast address to the lower device.
 629         */
 630        if (IS_ERR_OR_NULL(vlan->accel_priv)) {
 631                vlan->accel_priv = NULL;
 632                err = dev_uc_add(lowerdev, dev->dev_addr);
 633                if (err < 0)
 634                        goto out;
 635        }
 636
 637        if (dev->flags & IFF_ALLMULTI) {
 638                err = dev_set_allmulti(lowerdev, 1);
 639                if (err < 0)
 640                        goto del_unicast;
 641        }
 642
 643        if (dev->flags & IFF_PROMISC) {
 644                err = dev_set_promiscuity(lowerdev, 1);
 645                if (err < 0)
 646                        goto clear_multi;
 647        }
 648
 649hash_add:
 650        macvlan_hash_add(vlan);
 651        return 0;
 652
 653clear_multi:
 654        if (dev->flags & IFF_ALLMULTI)
 655                dev_set_allmulti(lowerdev, -1);
 656del_unicast:
 657        if (vlan->accel_priv) {
 658                lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
 659                                                           vlan->accel_priv);
 660                vlan->accel_priv = NULL;
 661        } else {
 662                dev_uc_del(lowerdev, dev->dev_addr);
 663        }
 664out:
 665        return err;
 666}
 667
 668static int macvlan_stop(struct net_device *dev)
 669{
 670        struct macvlan_dev *vlan = netdev_priv(dev);
 671        struct net_device *lowerdev = vlan->lowerdev;
 672
 673        if (vlan->accel_priv) {
 674                lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
 675                                                           vlan->accel_priv);
 676                vlan->accel_priv = NULL;
 677        }
 678
 679        dev_uc_unsync(lowerdev, dev);
 680        dev_mc_unsync(lowerdev, dev);
 681
 682        if (macvlan_passthru(vlan->port)) {
 683                if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
 684                        dev_set_promiscuity(lowerdev, -1);
 685                goto hash_del;
 686        }
 687
 688        if (dev->flags & IFF_ALLMULTI)
 689                dev_set_allmulti(lowerdev, -1);
 690
 691        if (dev->flags & IFF_PROMISC)
 692                dev_set_promiscuity(lowerdev, -1);
 693
 694        dev_uc_del(lowerdev, dev->dev_addr);
 695
 696hash_del:
 697        macvlan_hash_del(vlan, !dev->dismantle);
 698        return 0;
 699}
 700
 701static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
 702{
 703        struct macvlan_dev *vlan = netdev_priv(dev);
 704        struct net_device *lowerdev = vlan->lowerdev;
 705        struct macvlan_port *port = vlan->port;
 706        int err;
 707
 708        if (!(dev->flags & IFF_UP)) {
 709                /* Just copy in the new address */
 710                ether_addr_copy(dev->dev_addr, addr);
 711        } else {
 712                /* Rehash and update the device filters */
 713                if (macvlan_addr_busy(vlan->port, addr))
 714                        return -EADDRINUSE;
 715
 716                if (!macvlan_passthru(port)) {
 717                        err = dev_uc_add(lowerdev, addr);
 718                        if (err)
 719                                return err;
 720
 721                        dev_uc_del(lowerdev, dev->dev_addr);
 722                }
 723
 724                macvlan_hash_change_addr(vlan, addr);
 725        }
 726        if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
 727                /* Since addr_change isn't set, we are here due to lower
 728                 * device change.  Save the lower-dev address so we can
 729                 * restore it later.
 730                 */
 731                ether_addr_copy(vlan->port->perm_addr,
 732                                lowerdev->dev_addr);
 733        }
 734        macvlan_clear_addr_change(port);
 735        return 0;
 736}
 737
 738static int macvlan_set_mac_address(struct net_device *dev, void *p)
 739{
 740        struct macvlan_dev *vlan = netdev_priv(dev);
 741        struct sockaddr *addr = p;
 742
 743        if (!is_valid_ether_addr(addr->sa_data))
 744                return -EADDRNOTAVAIL;
 745
 746        /* If the addresses are the same, this is a no-op */
 747        if (ether_addr_equal(dev->dev_addr, addr->sa_data))
 748                return 0;
 749
 750        if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
 751                macvlan_set_addr_change(vlan->port);
 752                return dev_set_mac_address(vlan->lowerdev, addr, NULL);
 753        }
 754
 755        if (macvlan_addr_busy(vlan->port, addr->sa_data))
 756                return -EADDRINUSE;
 757
 758        return macvlan_sync_address(dev, addr->sa_data);
 759}
 760
 761static void macvlan_change_rx_flags(struct net_device *dev, int change)
 762{
 763        struct macvlan_dev *vlan = netdev_priv(dev);
 764        struct net_device *lowerdev = vlan->lowerdev;
 765
 766        if (dev->flags & IFF_UP) {
 767                if (change & IFF_ALLMULTI)
 768                        dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
 769                if (change & IFF_PROMISC)
 770                        dev_set_promiscuity(lowerdev,
 771                                            dev->flags & IFF_PROMISC ? 1 : -1);
 772
 773        }
 774}
 775
 776static void macvlan_compute_filter(unsigned long *mc_filter,
 777                                   struct net_device *dev,
 778                                   struct macvlan_dev *vlan)
 779{
 780        if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
 781                bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
 782        } else {
 783                struct netdev_hw_addr *ha;
 784                DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
 785
 786                bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
 787                netdev_for_each_mc_addr(ha, dev) {
 788                        __set_bit(mc_hash(vlan, ha->addr), filter);
 789                }
 790
 791                __set_bit(mc_hash(vlan, dev->broadcast), filter);
 792
 793                bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
 794        }
 795}
 796
 797static void macvlan_set_mac_lists(struct net_device *dev)
 798{
 799        struct macvlan_dev *vlan = netdev_priv(dev);
 800
 801        macvlan_compute_filter(vlan->mc_filter, dev, vlan);
 802
 803        dev_uc_sync(vlan->lowerdev, dev);
 804        dev_mc_sync(vlan->lowerdev, dev);
 805
 806        /* This is slightly inaccurate as we're including the subscription
 807         * list of vlan->lowerdev too.
 808         *
 809         * Bug alert: This only works if everyone has the same broadcast
 810         * address as lowerdev.  As soon as someone changes theirs this
 811         * will break.
 812         *
 813         * However, this is already broken as when you change your broadcast
 814         * address we don't get called.
 815         *
 816         * The solution is to maintain a list of broadcast addresses like
 817         * we do for uc/mc, if you care.
 818         */
 819        macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
 820}
 821
 822static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 823{
 824        struct macvlan_dev *vlan = netdev_priv(dev);
 825
 826        if (vlan->lowerdev->mtu < new_mtu)
 827                return -EINVAL;
 828        dev->mtu = new_mtu;
 829        return 0;
 830}
 831
 832static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 833{
 834        struct net_device *real_dev = macvlan_dev_real_dev(dev);
 835        const struct net_device_ops *ops = real_dev->netdev_ops;
 836        struct ifreq ifrr;
 837        int err = -EOPNOTSUPP;
 838
 839        strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
 840        ifrr.ifr_ifru = ifr->ifr_ifru;
 841
 842        switch (cmd) {
 843        case SIOCSHWTSTAMP:
 844                if (!net_eq(dev_net(dev), &init_net))
 845                        break;
 846                fallthrough;
 847        case SIOCGHWTSTAMP:
 848                if (netif_device_present(real_dev) && ops->ndo_do_ioctl)
 849                        err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd);
 850                break;
 851        }
 852
 853        if (!err)
 854                ifr->ifr_ifru = ifrr.ifr_ifru;
 855
 856        return err;
 857}
 858
 859/*
 860 * macvlan network devices have devices nesting below it and are a special
 861 * "super class" of normal network devices; split their locks off into a
 862 * separate class since they always nest.
 863 */
 864static struct lock_class_key macvlan_netdev_addr_lock_key;
 865
 866#define ALWAYS_ON_OFFLOADS \
 867        (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
 868         NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
 869
 870#define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
 871
 872#define MACVLAN_FEATURES \
 873        (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
 874         NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
 875         NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
 876         NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
 877
 878#define MACVLAN_STATE_MASK \
 879        ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
 880
 881static void macvlan_set_lockdep_class(struct net_device *dev)
 882{
 883        netdev_lockdep_set_classes(dev);
 884        lockdep_set_class(&dev->addr_list_lock,
 885                          &macvlan_netdev_addr_lock_key);
 886}
 887
 888static int macvlan_init(struct net_device *dev)
 889{
 890        struct macvlan_dev *vlan = netdev_priv(dev);
 891        const struct net_device *lowerdev = vlan->lowerdev;
 892        struct macvlan_port *port = vlan->port;
 893
 894        dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
 895                                  (lowerdev->state & MACVLAN_STATE_MASK);
 896        dev->features           = lowerdev->features & MACVLAN_FEATURES;
 897        dev->features           |= ALWAYS_ON_FEATURES;
 898        dev->hw_features        |= NETIF_F_LRO;
 899        dev->vlan_features      = lowerdev->vlan_features & MACVLAN_FEATURES;
 900        dev->vlan_features      |= ALWAYS_ON_OFFLOADS;
 901        dev->hw_enc_features    |= dev->features;
 902        dev->gso_max_size       = lowerdev->gso_max_size;
 903        dev->gso_max_segs       = lowerdev->gso_max_segs;
 904        dev->hard_header_len    = lowerdev->hard_header_len;
 905        macvlan_set_lockdep_class(dev);
 906
 907        vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
 908        if (!vlan->pcpu_stats)
 909                return -ENOMEM;
 910
 911        port->count += 1;
 912
 913        return 0;
 914}
 915
 916static void macvlan_uninit(struct net_device *dev)
 917{
 918        struct macvlan_dev *vlan = netdev_priv(dev);
 919        struct macvlan_port *port = vlan->port;
 920
 921        free_percpu(vlan->pcpu_stats);
 922
 923        macvlan_flush_sources(port, vlan);
 924        port->count -= 1;
 925        if (!port->count)
 926                macvlan_port_destroy(port->dev);
 927}
 928
 929static void macvlan_dev_get_stats64(struct net_device *dev,
 930                                    struct rtnl_link_stats64 *stats)
 931{
 932        struct macvlan_dev *vlan = netdev_priv(dev);
 933
 934        if (vlan->pcpu_stats) {
 935                struct vlan_pcpu_stats *p;
 936                u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
 937                u32 rx_errors = 0, tx_dropped = 0;
 938                unsigned int start;
 939                int i;
 940
 941                for_each_possible_cpu(i) {
 942                        p = per_cpu_ptr(vlan->pcpu_stats, i);
 943                        do {
 944                                start = u64_stats_fetch_begin_irq(&p->syncp);
 945                                rx_packets      = p->rx_packets;
 946                                rx_bytes        = p->rx_bytes;
 947                                rx_multicast    = p->rx_multicast;
 948                                tx_packets      = p->tx_packets;
 949                                tx_bytes        = p->tx_bytes;
 950                        } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 951
 952                        stats->rx_packets       += rx_packets;
 953                        stats->rx_bytes         += rx_bytes;
 954                        stats->multicast        += rx_multicast;
 955                        stats->tx_packets       += tx_packets;
 956                        stats->tx_bytes         += tx_bytes;
 957                        /* rx_errors & tx_dropped are u32, updated
 958                         * without syncp protection.
 959                         */
 960                        rx_errors       += p->rx_errors;
 961                        tx_dropped      += p->tx_dropped;
 962                }
 963                stats->rx_errors        = rx_errors;
 964                stats->rx_dropped       = rx_errors;
 965                stats->tx_dropped       = tx_dropped;
 966        }
 967}
 968
 969static int macvlan_vlan_rx_add_vid(struct net_device *dev,
 970                                   __be16 proto, u16 vid)
 971{
 972        struct macvlan_dev *vlan = netdev_priv(dev);
 973        struct net_device *lowerdev = vlan->lowerdev;
 974
 975        return vlan_vid_add(lowerdev, proto, vid);
 976}
 977
 978static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
 979                                    __be16 proto, u16 vid)
 980{
 981        struct macvlan_dev *vlan = netdev_priv(dev);
 982        struct net_device *lowerdev = vlan->lowerdev;
 983
 984        vlan_vid_del(lowerdev, proto, vid);
 985        return 0;
 986}
 987
 988static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 989                           struct net_device *dev,
 990                           const unsigned char *addr, u16 vid,
 991                           u16 flags,
 992                           struct netlink_ext_ack *extack)
 993{
 994        struct macvlan_dev *vlan = netdev_priv(dev);
 995        int err = -EINVAL;
 996
 997        /* Support unicast filter only on passthru devices.
 998         * Multicast filter should be allowed on all devices.
 999         */
1000        if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1001                return -EOPNOTSUPP;
1002
1003        if (flags & NLM_F_REPLACE)
1004                return -EOPNOTSUPP;
1005
1006        if (is_unicast_ether_addr(addr))
1007                err = dev_uc_add_excl(dev, addr);
1008        else if (is_multicast_ether_addr(addr))
1009                err = dev_mc_add_excl(dev, addr);
1010
1011        return err;
1012}
1013
1014static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1015                           struct net_device *dev,
1016                           const unsigned char *addr, u16 vid)
1017{
1018        struct macvlan_dev *vlan = netdev_priv(dev);
1019        int err = -EINVAL;
1020
1021        /* Support unicast filter only on passthru devices.
1022         * Multicast filter should be allowed on all devices.
1023         */
1024        if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1025                return -EOPNOTSUPP;
1026
1027        if (is_unicast_ether_addr(addr))
1028                err = dev_uc_del(dev, addr);
1029        else if (is_multicast_ether_addr(addr))
1030                err = dev_mc_del(dev, addr);
1031
1032        return err;
1033}
1034
1035static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1036                                        struct ethtool_drvinfo *drvinfo)
1037{
1038        strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1039        strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1040}
1041
1042static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1043                                              struct ethtool_link_ksettings *cmd)
1044{
1045        const struct macvlan_dev *vlan = netdev_priv(dev);
1046
1047        return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1048}
1049
1050static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1051                                       struct ethtool_ts_info *info)
1052{
1053        struct net_device *real_dev = macvlan_dev_real_dev(dev);
1054        const struct ethtool_ops *ops = real_dev->ethtool_ops;
1055        struct phy_device *phydev = real_dev->phydev;
1056
1057        if (phy_has_tsinfo(phydev)) {
1058                return phy_ts_info(phydev, info);
1059        } else if (ops->get_ts_info) {
1060                return ops->get_ts_info(real_dev, info);
1061        } else {
1062                info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1063                        SOF_TIMESTAMPING_SOFTWARE;
1064                info->phc_index = -1;
1065        }
1066
1067        return 0;
1068}
1069
1070static netdev_features_t macvlan_fix_features(struct net_device *dev,
1071                                              netdev_features_t features)
1072{
1073        struct macvlan_dev *vlan = netdev_priv(dev);
1074        netdev_features_t lowerdev_features = vlan->lowerdev->features;
1075        netdev_features_t mask;
1076
1077        features |= NETIF_F_ALL_FOR_ALL;
1078        features &= (vlan->set_features | ~MACVLAN_FEATURES);
1079        mask = features;
1080
1081        lowerdev_features &= (features | ~NETIF_F_LRO);
1082        features = netdev_increment_features(lowerdev_features, features, mask);
1083        features |= ALWAYS_ON_FEATURES;
1084        features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1085
1086        return features;
1087}
1088
1089#ifdef CONFIG_NET_POLL_CONTROLLER
1090static void macvlan_dev_poll_controller(struct net_device *dev)
1091{
1092        return;
1093}
1094
1095static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1096{
1097        struct macvlan_dev *vlan = netdev_priv(dev);
1098        struct net_device *real_dev = vlan->lowerdev;
1099        struct netpoll *netpoll;
1100        int err;
1101
1102        netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1103        err = -ENOMEM;
1104        if (!netpoll)
1105                goto out;
1106
1107        err = __netpoll_setup(netpoll, real_dev);
1108        if (err) {
1109                kfree(netpoll);
1110                goto out;
1111        }
1112
1113        vlan->netpoll = netpoll;
1114
1115out:
1116        return err;
1117}
1118
1119static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1120{
1121        struct macvlan_dev *vlan = netdev_priv(dev);
1122        struct netpoll *netpoll = vlan->netpoll;
1123
1124        if (!netpoll)
1125                return;
1126
1127        vlan->netpoll = NULL;
1128
1129        __netpoll_free(netpoll);
1130}
1131#endif  /* CONFIG_NET_POLL_CONTROLLER */
1132
1133static int macvlan_dev_get_iflink(const struct net_device *dev)
1134{
1135        struct macvlan_dev *vlan = netdev_priv(dev);
1136
1137        return vlan->lowerdev->ifindex;
1138}
1139
1140static const struct ethtool_ops macvlan_ethtool_ops = {
1141        .get_link               = ethtool_op_get_link,
1142        .get_link_ksettings     = macvlan_ethtool_get_link_ksettings,
1143        .get_drvinfo            = macvlan_ethtool_get_drvinfo,
1144        .get_ts_info            = macvlan_ethtool_get_ts_info,
1145};
1146
1147static const struct net_device_ops macvlan_netdev_ops = {
1148        .ndo_init               = macvlan_init,
1149        .ndo_uninit             = macvlan_uninit,
1150        .ndo_open               = macvlan_open,
1151        .ndo_stop               = macvlan_stop,
1152        .ndo_start_xmit         = macvlan_start_xmit,
1153        .ndo_change_mtu         = macvlan_change_mtu,
1154        .ndo_do_ioctl           = macvlan_do_ioctl,
1155        .ndo_fix_features       = macvlan_fix_features,
1156        .ndo_change_rx_flags    = macvlan_change_rx_flags,
1157        .ndo_set_mac_address    = macvlan_set_mac_address,
1158        .ndo_set_rx_mode        = macvlan_set_mac_lists,
1159        .ndo_get_stats64        = macvlan_dev_get_stats64,
1160        .ndo_validate_addr      = eth_validate_addr,
1161        .ndo_vlan_rx_add_vid    = macvlan_vlan_rx_add_vid,
1162        .ndo_vlan_rx_kill_vid   = macvlan_vlan_rx_kill_vid,
1163        .ndo_fdb_add            = macvlan_fdb_add,
1164        .ndo_fdb_del            = macvlan_fdb_del,
1165        .ndo_fdb_dump           = ndo_dflt_fdb_dump,
1166#ifdef CONFIG_NET_POLL_CONTROLLER
1167        .ndo_poll_controller    = macvlan_dev_poll_controller,
1168        .ndo_netpoll_setup      = macvlan_dev_netpoll_setup,
1169        .ndo_netpoll_cleanup    = macvlan_dev_netpoll_cleanup,
1170#endif
1171        .ndo_get_iflink         = macvlan_dev_get_iflink,
1172        .ndo_features_check     = passthru_features_check,
1173        .ndo_change_proto_down  = dev_change_proto_down_generic,
1174};
1175
1176void 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}
1190EXPORT_SYMBOL_GPL(macvlan_common_setup);
1191
1192static void macvlan_setup(struct net_device *dev)
1193{
1194        macvlan_common_setup(dev);
1195        dev->priv_flags |= IFF_NO_QUEUE;
1196}
1197
1198static 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        port->bc_queue_len_used = 0;
1223        skb_queue_head_init(&port->bc_queue);
1224        INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1225
1226        err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1227        if (err)
1228                kfree(port);
1229        else
1230                dev->priv_flags |= IFF_MACVLAN_PORT;
1231        return err;
1232}
1233
1234static void macvlan_port_destroy(struct net_device *dev)
1235{
1236        struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1237        struct sk_buff *skb;
1238
1239        dev->priv_flags &= ~IFF_MACVLAN_PORT;
1240        netdev_rx_handler_unregister(dev);
1241
1242        /* After this point, no packet can schedule bc_work anymore,
1243         * but we need to cancel it and purge left skbs if any.
1244         */
1245        cancel_work_sync(&port->bc_work);
1246
1247        while ((skb = __skb_dequeue(&port->bc_queue))) {
1248                const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1249
1250                if (src)
1251                        dev_put(src->dev);
1252
1253                kfree_skb(skb);
1254        }
1255
1256        /* If the lower device address has been changed by passthru
1257         * macvlan, put it back.
1258         */
1259        if (macvlan_passthru(port) &&
1260            !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1261                struct sockaddr sa;
1262
1263                sa.sa_family = port->dev->type;
1264                memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1265                dev_set_mac_address(port->dev, &sa, NULL);
1266        }
1267
1268        kfree(port);
1269}
1270
1271static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1272                            struct netlink_ext_ack *extack)
1273{
1274        struct nlattr *nla, *head;
1275        int rem, len;
1276
1277        if (tb[IFLA_ADDRESS]) {
1278                if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1279                        return -EINVAL;
1280                if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1281                        return -EADDRNOTAVAIL;
1282        }
1283
1284        if (!data)
1285                return 0;
1286
1287        if (data[IFLA_MACVLAN_FLAGS] &&
1288            nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1289                                                      MACVLAN_FLAG_NODST))
1290                return -EINVAL;
1291
1292        if (data[IFLA_MACVLAN_MODE]) {
1293                switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1294                case MACVLAN_MODE_PRIVATE:
1295                case MACVLAN_MODE_VEPA:
1296                case MACVLAN_MODE_BRIDGE:
1297                case MACVLAN_MODE_PASSTHRU:
1298                case MACVLAN_MODE_SOURCE:
1299                        break;
1300                default:
1301                        return -EINVAL;
1302                }
1303        }
1304
1305        if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1306                switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1307                case MACVLAN_MACADDR_ADD:
1308                case MACVLAN_MACADDR_DEL:
1309                case MACVLAN_MACADDR_FLUSH:
1310                case MACVLAN_MACADDR_SET:
1311                        break;
1312                default:
1313                        return -EINVAL;
1314                }
1315        }
1316
1317        if (data[IFLA_MACVLAN_MACADDR]) {
1318                if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1319                        return -EINVAL;
1320
1321                if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1322                        return -EADDRNOTAVAIL;
1323        }
1324
1325        if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1326                head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1327                len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1328
1329                nla_for_each_attr(nla, head, len, rem) {
1330                        if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1331                            nla_len(nla) != ETH_ALEN)
1332                                return -EINVAL;
1333
1334                        if (!is_valid_ether_addr(nla_data(nla)))
1335                                return -EADDRNOTAVAIL;
1336                }
1337        }
1338
1339        if (data[IFLA_MACVLAN_MACADDR_COUNT])
1340                return -EINVAL;
1341
1342        return 0;
1343}
1344
1345/*
1346 * reconfigure list of remote source mac address
1347 * (only for macvlan devices in source mode)
1348 * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1349 * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1350 */
1351static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1352                                      struct nlattr *data[])
1353{
1354        char *addr = NULL;
1355        int ret, rem, len;
1356        struct nlattr *nla, *head;
1357        struct macvlan_source_entry *entry;
1358
1359        if (data[IFLA_MACVLAN_MACADDR])
1360                addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1361
1362        if (mode == MACVLAN_MACADDR_ADD) {
1363                if (!addr)
1364                        return -EINVAL;
1365
1366                return macvlan_hash_add_source(vlan, addr);
1367
1368        } else if (mode == MACVLAN_MACADDR_DEL) {
1369                if (!addr)
1370                        return -EINVAL;
1371
1372                entry = macvlan_hash_lookup_source(vlan, addr);
1373                if (entry) {
1374                        macvlan_hash_del_source(entry);
1375                        vlan->macaddr_count--;
1376                }
1377        } else if (mode == MACVLAN_MACADDR_FLUSH) {
1378                macvlan_flush_sources(vlan->port, vlan);
1379        } else if (mode == MACVLAN_MACADDR_SET) {
1380                macvlan_flush_sources(vlan->port, vlan);
1381
1382                if (addr) {
1383                        ret = macvlan_hash_add_source(vlan, addr);
1384                        if (ret)
1385                                return ret;
1386                }
1387
1388                if (!data[IFLA_MACVLAN_MACADDR_DATA])
1389                        return 0;
1390
1391                head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1392                len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1393
1394                nla_for_each_attr(nla, head, len, rem) {
1395                        addr = nla_data(nla);
1396                        ret = macvlan_hash_add_source(vlan, addr);
1397                        if (ret)
1398                                return ret;
1399                }
1400        } else {
1401                return -EINVAL;
1402        }
1403
1404        return 0;
1405}
1406
1407int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1408                           struct nlattr *tb[], struct nlattr *data[],
1409                           struct netlink_ext_ack *extack)
1410{
1411        struct macvlan_dev *vlan = netdev_priv(dev);
1412        struct macvlan_port *port;
1413        struct net_device *lowerdev;
1414        int err;
1415        int macmode;
1416        bool create = false;
1417
1418        if (!tb[IFLA_LINK])
1419                return -EINVAL;
1420
1421        lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1422        if (lowerdev == NULL)
1423                return -ENODEV;
1424
1425        /* When creating macvlans or macvtaps on top of other macvlans - use
1426         * the real device as the lowerdev.
1427         */
1428        if (netif_is_macvlan(lowerdev))
1429                lowerdev = macvlan_dev_real_dev(lowerdev);
1430
1431        if (!tb[IFLA_MTU])
1432                dev->mtu = lowerdev->mtu;
1433        else if (dev->mtu > lowerdev->mtu)
1434                return -EINVAL;
1435
1436        /* MTU range: 68 - lowerdev->max_mtu */
1437        dev->min_mtu = ETH_MIN_MTU;
1438        dev->max_mtu = lowerdev->max_mtu;
1439
1440        if (!tb[IFLA_ADDRESS])
1441                eth_hw_addr_random(dev);
1442
1443        if (!netif_is_macvlan_port(lowerdev)) {
1444                err = macvlan_port_create(lowerdev);
1445                if (err < 0)
1446                        return err;
1447                create = true;
1448        }
1449        port = macvlan_port_get_rtnl(lowerdev);
1450
1451        /* Only 1 macvlan device can be created in passthru mode */
1452        if (macvlan_passthru(port)) {
1453                /* The macvlan port must be not created this time,
1454                 * still goto destroy_macvlan_port for readability.
1455                 */
1456                err = -EINVAL;
1457                goto destroy_macvlan_port;
1458        }
1459
1460        vlan->lowerdev = lowerdev;
1461        vlan->dev      = dev;
1462        vlan->port     = port;
1463        vlan->set_features = MACVLAN_FEATURES;
1464
1465        vlan->mode     = MACVLAN_MODE_VEPA;
1466        if (data && data[IFLA_MACVLAN_MODE])
1467                vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1468
1469        if (data && data[IFLA_MACVLAN_FLAGS])
1470                vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1471
1472        if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1473                if (port->count) {
1474                        err = -EINVAL;
1475                        goto destroy_macvlan_port;
1476                }
1477                macvlan_set_passthru(port);
1478                eth_hw_addr_inherit(dev, lowerdev);
1479        }
1480
1481        if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1482                if (vlan->mode != MACVLAN_MODE_SOURCE) {
1483                        err = -EINVAL;
1484                        goto destroy_macvlan_port;
1485                }
1486                macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1487                err = macvlan_changelink_sources(vlan, macmode, data);
1488                if (err)
1489                        goto destroy_macvlan_port;
1490        }
1491
1492        vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1493        if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1494                vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1495
1496        err = register_netdevice(dev);
1497        if (err < 0)
1498                goto destroy_macvlan_port;
1499
1500        dev->priv_flags |= IFF_MACVLAN;
1501        err = netdev_upper_dev_link(lowerdev, dev, extack);
1502        if (err)
1503                goto unregister_netdev;
1504
1505        list_add_tail_rcu(&vlan->list, &port->vlans);
1506        update_port_bc_queue_len(vlan->port);
1507        netif_stacked_transfer_operstate(lowerdev, dev);
1508        linkwatch_fire_event(dev);
1509
1510        return 0;
1511
1512unregister_netdev:
1513        /* macvlan_uninit would free the macvlan port */
1514        unregister_netdevice(dev);
1515        return err;
1516destroy_macvlan_port:
1517        /* the macvlan port may be freed by macvlan_uninit when fail to register.
1518         * so we destroy the macvlan port only when it's valid.
1519         */
1520        if (create && macvlan_port_get_rtnl(lowerdev))
1521                macvlan_port_destroy(port->dev);
1522        return err;
1523}
1524EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1525
1526static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1527                           struct nlattr *tb[], struct nlattr *data[],
1528                           struct netlink_ext_ack *extack)
1529{
1530        return macvlan_common_newlink(src_net, dev, tb, data, extack);
1531}
1532
1533void macvlan_dellink(struct net_device *dev, struct list_head *head)
1534{
1535        struct macvlan_dev *vlan = netdev_priv(dev);
1536
1537        if (vlan->mode == MACVLAN_MODE_SOURCE)
1538                macvlan_flush_sources(vlan->port, vlan);
1539        list_del_rcu(&vlan->list);
1540        update_port_bc_queue_len(vlan->port);
1541        unregister_netdevice_queue(dev, head);
1542        netdev_upper_dev_unlink(vlan->lowerdev, dev);
1543}
1544EXPORT_SYMBOL_GPL(macvlan_dellink);
1545
1546static int macvlan_changelink(struct net_device *dev,
1547                              struct nlattr *tb[], struct nlattr *data[],
1548                              struct netlink_ext_ack *extack)
1549{
1550        struct macvlan_dev *vlan = netdev_priv(dev);
1551        enum macvlan_mode mode;
1552        bool set_mode = false;
1553        enum macvlan_macaddr_mode macmode;
1554        int ret;
1555
1556        /* Validate mode, but don't set yet: setting flags may fail. */
1557        if (data && data[IFLA_MACVLAN_MODE]) {
1558                set_mode = true;
1559                mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1560                /* Passthrough mode can't be set or cleared dynamically */
1561                if ((mode == MACVLAN_MODE_PASSTHRU) !=
1562                    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1563                        return -EINVAL;
1564                if (vlan->mode == MACVLAN_MODE_SOURCE &&
1565                    vlan->mode != mode)
1566                        macvlan_flush_sources(vlan->port, vlan);
1567        }
1568
1569        if (data && data[IFLA_MACVLAN_FLAGS]) {
1570                __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1571                bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1572                if (macvlan_passthru(vlan->port) && promisc) {
1573                        int err;
1574
1575                        if (flags & MACVLAN_FLAG_NOPROMISC)
1576                                err = dev_set_promiscuity(vlan->lowerdev, -1);
1577                        else
1578                                err = dev_set_promiscuity(vlan->lowerdev, 1);
1579                        if (err < 0)
1580                                return err;
1581                }
1582                vlan->flags = flags;
1583        }
1584
1585        if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1586                vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1587                update_port_bc_queue_len(vlan->port);
1588        }
1589
1590        if (set_mode)
1591                vlan->mode = mode;
1592        if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1593                if (vlan->mode != MACVLAN_MODE_SOURCE)
1594                        return -EINVAL;
1595                macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1596                ret = macvlan_changelink_sources(vlan, macmode, data);
1597                if (ret)
1598                        return ret;
1599        }
1600        return 0;
1601}
1602
1603static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1604{
1605        if (vlan->macaddr_count == 0)
1606                return 0;
1607        return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1608                + vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1609}
1610
1611static size_t macvlan_get_size(const struct net_device *dev)
1612{
1613        struct macvlan_dev *vlan = netdev_priv(dev);
1614
1615        return (0
1616                + nla_total_size(4) /* IFLA_MACVLAN_MODE */
1617                + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1618                + nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1619                + macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1620                + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1621                + nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1622                );
1623}
1624
1625static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1626                                     const struct macvlan_dev *vlan,
1627                                     const int i)
1628{
1629        struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1630        struct macvlan_source_entry *entry;
1631
1632        hlist_for_each_entry_rcu(entry, h, hlist) {
1633                if (entry->vlan != vlan)
1634                        continue;
1635                if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1636                        return 1;
1637        }
1638        return 0;
1639}
1640
1641static int macvlan_fill_info(struct sk_buff *skb,
1642                                const struct net_device *dev)
1643{
1644        struct macvlan_dev *vlan = netdev_priv(dev);
1645        struct macvlan_port *port = vlan->port;
1646        int i;
1647        struct nlattr *nest;
1648
1649        if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1650                goto nla_put_failure;
1651        if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1652                goto nla_put_failure;
1653        if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1654                goto nla_put_failure;
1655        if (vlan->macaddr_count > 0) {
1656                nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1657                if (nest == NULL)
1658                        goto nla_put_failure;
1659
1660                for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1661                        if (macvlan_fill_info_macaddr(skb, vlan, i))
1662                                goto nla_put_failure;
1663                }
1664                nla_nest_end(skb, nest);
1665        }
1666        if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1667                goto nla_put_failure;
1668        if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1669                goto nla_put_failure;
1670        return 0;
1671
1672nla_put_failure:
1673        return -EMSGSIZE;
1674}
1675
1676static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1677        [IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1678        [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1679        [IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1680        [IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1681        [IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1682        [IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1683        [IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1684        [IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1685};
1686
1687int macvlan_link_register(struct rtnl_link_ops *ops)
1688{
1689        /* common fields */
1690        ops->validate           = macvlan_validate;
1691        ops->maxtype            = IFLA_MACVLAN_MAX;
1692        ops->policy             = macvlan_policy;
1693        ops->changelink         = macvlan_changelink;
1694        ops->get_size           = macvlan_get_size;
1695        ops->fill_info          = macvlan_fill_info;
1696
1697        return rtnl_link_register(ops);
1698};
1699EXPORT_SYMBOL_GPL(macvlan_link_register);
1700
1701static struct net *macvlan_get_link_net(const struct net_device *dev)
1702{
1703        return dev_net(macvlan_dev_real_dev(dev));
1704}
1705
1706static struct rtnl_link_ops macvlan_link_ops = {
1707        .kind           = "macvlan",
1708        .setup          = macvlan_setup,
1709        .newlink        = macvlan_newlink,
1710        .dellink        = macvlan_dellink,
1711        .get_link_net   = macvlan_get_link_net,
1712        .priv_size      = sizeof(struct macvlan_dev),
1713};
1714
1715static void update_port_bc_queue_len(struct macvlan_port *port)
1716{
1717        u32 max_bc_queue_len_req = 0;
1718        struct macvlan_dev *vlan;
1719
1720        list_for_each_entry(vlan, &port->vlans, list) {
1721                if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1722                        max_bc_queue_len_req = vlan->bc_queue_len_req;
1723        }
1724        port->bc_queue_len_used = max_bc_queue_len_req;
1725}
1726
1727static int macvlan_device_event(struct notifier_block *unused,
1728                                unsigned long event, void *ptr)
1729{
1730        struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1731        struct macvlan_dev *vlan, *next;
1732        struct macvlan_port *port;
1733        LIST_HEAD(list_kill);
1734
1735        if (!netif_is_macvlan_port(dev))
1736                return NOTIFY_DONE;
1737
1738        port = macvlan_port_get_rtnl(dev);
1739
1740        switch (event) {
1741        case NETDEV_UP:
1742        case NETDEV_DOWN:
1743        case NETDEV_CHANGE:
1744                list_for_each_entry(vlan, &port->vlans, list)
1745                        netif_stacked_transfer_operstate(vlan->lowerdev,
1746                                                         vlan->dev);
1747                break;
1748        case NETDEV_FEAT_CHANGE:
1749                list_for_each_entry(vlan, &port->vlans, list) {
1750                        vlan->dev->gso_max_size = dev->gso_max_size;
1751                        vlan->dev->gso_max_segs = dev->gso_max_segs;
1752                        netdev_update_features(vlan->dev);
1753                }
1754                break;
1755        case NETDEV_CHANGEMTU:
1756                list_for_each_entry(vlan, &port->vlans, list) {
1757                        if (vlan->dev->mtu <= dev->mtu)
1758                                continue;
1759                        dev_set_mtu(vlan->dev, dev->mtu);
1760                }
1761                break;
1762        case NETDEV_CHANGEADDR:
1763                if (!macvlan_passthru(port))
1764                        return NOTIFY_DONE;
1765
1766                vlan = list_first_entry_or_null(&port->vlans,
1767                                                struct macvlan_dev,
1768                                                list);
1769
1770                if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1771                        return NOTIFY_BAD;
1772
1773                break;
1774        case NETDEV_UNREGISTER:
1775                /* twiddle thumbs on netns device moves */
1776                if (dev->reg_state != NETREG_UNREGISTERING)
1777                        break;
1778
1779                list_for_each_entry_safe(vlan, next, &port->vlans, list)
1780                        vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1781                unregister_netdevice_many(&list_kill);
1782                break;
1783        case NETDEV_PRE_TYPE_CHANGE:
1784                /* Forbid underlying device to change its type. */
1785                return NOTIFY_BAD;
1786
1787        case NETDEV_NOTIFY_PEERS:
1788        case NETDEV_BONDING_FAILOVER:
1789        case NETDEV_RESEND_IGMP:
1790                /* Propagate to all vlans */
1791                list_for_each_entry(vlan, &port->vlans, list)
1792                        call_netdevice_notifiers(event, vlan->dev);
1793        }
1794        return NOTIFY_DONE;
1795}
1796
1797static struct notifier_block macvlan_notifier_block __read_mostly = {
1798        .notifier_call  = macvlan_device_event,
1799};
1800
1801static int __init macvlan_init_module(void)
1802{
1803        int err;
1804
1805        register_netdevice_notifier(&macvlan_notifier_block);
1806
1807        err = macvlan_link_register(&macvlan_link_ops);
1808        if (err < 0)
1809                goto err1;
1810        return 0;
1811err1:
1812        unregister_netdevice_notifier(&macvlan_notifier_block);
1813        return err;
1814}
1815
1816static void __exit macvlan_cleanup_module(void)
1817{
1818        rtnl_link_unregister(&macvlan_link_ops);
1819        unregister_netdevice_notifier(&macvlan_notifier_block);
1820}
1821
1822module_init(macvlan_init_module);
1823module_exit(macvlan_cleanup_module);
1824
1825MODULE_LICENSE("GPL");
1826MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1827MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1828MODULE_ALIAS_RTNL_LINK("macvlan");
1829