drivers/net/ipvlan/ipvlan_core.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
   3  */
   4
   5 #include "ipvlan.h"
   6
   7 static u32 ipvlan_jhash_secret __read_mostly;
   8
   9 void ipvlan_init_secret(void)
  10 {
  11         net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
  12 }
  13
  14 void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
  15                             unsigned int len, bool success, bool mcast)
  16 {
  17         if (likely(success)) {
  18                 struct ipvl_pcpu_stats *pcptr;
  19
  20                 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
  21                 u64_stats_update_begin(&pcptr->syncp);
  22                 pcptr->rx_pkts++;
  23                 pcptr->rx_bytes += len;
  24                 if (mcast)
  25                         pcptr->rx_mcast++;
  26                 u64_stats_update_end(&pcptr->syncp);
  27         } else {
  28                 this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
  29         }
  30 }
  31 EXPORT_SYMBOL_GPL(ipvlan_count_rx);
  32
  33 #if IS_ENABLED(CONFIG_IPV6)
  34 static u8 ipvlan_get_v6_hash(const void *iaddr)
  35 {
  36         const struct in6_addr *ip6_addr = iaddr;
  37
  38         return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
  39                IPVLAN_HASH_MASK;
  40 }
  41 #else
  42 static u8 ipvlan_get_v6_hash(const void *iaddr)
  43 {
  44         return 0;
  45 }
  46 #endif
  47
  48 static u8 ipvlan_get_v4_hash(const void *iaddr)
  49 {
  50         const struct in_addr *ip4_addr = iaddr;
  51
  52         return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
  53                IPVLAN_HASH_MASK;
  54 }
  55
  56 static bool addr_equal(bool is_v6, struct ipvl_addr *addr, const void *iaddr)
  57 {
  58         if (!is_v6 && addr->atype == IPVL_IPV4) {
  59                 struct in_addr *i4addr = (struct in_addr *)iaddr;
  60
  61                 return addr->ip4addr.s_addr == i4addr->s_addr;
  62 #if IS_ENABLED(CONFIG_IPV6)
  63         } else if (is_v6 && addr->atype == IPVL_IPV6) {
  64                 struct in6_addr *i6addr = (struct in6_addr *)iaddr;
  65
  66                 return ipv6_addr_equal(&addr->ip6addr, i6addr);
  67 #endif
  68         }
  69
  70         return false;
  71 }
  72
  73 static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
  74                                                const void *iaddr, bool is_v6)
  75 {
  76         struct ipvl_addr *addr;
  77         u8 hash;
  78
  79         hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
  80                ipvlan_get_v4_hash(iaddr);
  81         hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode)
  82                 if (addr_equal(is_v6, addr, iaddr))
  83                         return addr;
  84         return NULL;
  85 }
  86
  87 void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
  88 {
  89         struct ipvl_port *port = ipvlan->port;
  90         u8 hash;
  91
  92         hash = (addr->atype == IPVL_IPV6) ?
  93                ipvlan_get_v6_hash(&addr->ip6addr) :
  94                ipvlan_get_v4_hash(&addr->ip4addr);
  95         if (hlist_unhashed(&addr->hlnode))
  96                 hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
  97 }
  98
  99 void ipvlan_ht_addr_del(struct ipvl_addr *addr)
 100 {
 101         hlist_del_init_rcu(&addr->hlnode);
 102 }
 103
 104 struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
 105                                    const void *iaddr, bool is_v6)
 106 {
 107         struct ipvl_addr *addr, *ret = NULL;
 108
 109         rcu_read_lock();
 110         list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) {
 111                 if (addr_equal(is_v6, addr, iaddr)) {
 112                         ret = addr;
 113                         break;
 114                 }
 115         }
 116         rcu_read_unlock();
 117         return ret;
 118 }
 119
 120 bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
 121 {
 122         struct ipvl_dev *ipvlan;
 123         bool ret = false;
 124
 125         rcu_read_lock();
 126         list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
 127                 if (ipvlan_find_addr(ipvlan, iaddr, is_v6)) {
 128                         ret = true;
 129                         break;
 130                 }
 131         }
 132         rcu_read_unlock();
 133         return ret;
 134 }
 135
 136 void *ipvlan_get_L3_hdr(struct ipvl_port *port, struct sk_buff *skb, int *type)
 137 {
 138         void *lyr3h = NULL;
 139
 140         switch (skb->protocol) {
 141         case htons(ETH_P_ARP): {
 142                 struct arphdr *arph;
 143
 144                 if (unlikely(!pskb_may_pull(skb, arp_hdr_len(port->dev))))
 145                         return NULL;
 146
 147                 arph = arp_hdr(skb);
 148                 *type = IPVL_ARP;
 149                 lyr3h = arph;
 150                 break;
 151         }
 152         case htons(ETH_P_IP): {
 153                 u32 pktlen;
 154                 struct iphdr *ip4h;
 155
 156                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
 157                         return NULL;
 158
 159                 ip4h = ip_hdr(skb);
 160                 pktlen = ntohs(ip4h->tot_len);
 161                 if (ip4h->ihl < 5 || ip4h->version != 4)
 162                         return NULL;
 163                 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
 164                         return NULL;
 165
 166                 *type = IPVL_IPV4;
 167                 lyr3h = ip4h;
 168                 break;
 169         }
 170 #if IS_ENABLED(CONFIG_IPV6)
 171         case htons(ETH_P_IPV6): {
 172                 struct ipv6hdr *ip6h;
 173
 174                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
 175                         return NULL;
 176
 177                 ip6h = ipv6_hdr(skb);
 178                 if (ip6h->version != 6)
 179                         return NULL;
 180
 181                 *type = IPVL_IPV6;
 182                 lyr3h = ip6h;
 183                 /* Only Neighbour Solicitation pkts need different treatment */
 184                 if (ipv6_addr_any(&ip6h->saddr) &&
 185                     ip6h->nexthdr == NEXTHDR_ICMP) {
 186                         struct icmp6hdr *icmph;
 187
 188                         if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph))))
 189                                 return NULL;
 190
 191                         ip6h = ipv6_hdr(skb);
 192                         icmph = (struct icmp6hdr *)(ip6h + 1);
 193
 194                         if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
 195                                 /* Need to access the ipv6 address in body */
 196                                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)
 197                                                 + sizeof(struct in6_addr))))
 198                                         return NULL;
 199
 200                                 ip6h = ipv6_hdr(skb);
 201                                 icmph = (struct icmp6hdr *)(ip6h + 1);
 202                         }
 203
 204                         *type = IPVL_ICMPV6;
 205                         lyr3h = icmph;
 206                 }
 207                 break;
 208         }
 209 #endif
 210         default:
 211                 return NULL;
 212         }
 213
 214         return lyr3h;
 215 }
 216
 217 unsigned int ipvlan_mac_hash(const unsigned char *addr)
 218 {
 219         u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
 220                                ipvlan_jhash_secret);
 221
 222         return hash & IPVLAN_MAC_FILTER_MASK;
 223 }
 224
 225 void ipvlan_process_multicast(struct work_struct *work)
 226 {
 227         struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
 228         struct ethhdr *ethh;
 229         struct ipvl_dev *ipvlan;
 230         struct sk_buff *skb, *nskb;
 231         struct sk_buff_head list;
 232         unsigned int len;
 233         unsigned int mac_hash;
 234         int ret;
 235         u8 pkt_type;
 236         bool tx_pkt;
 237
 238         __skb_queue_head_init(&list);
 239
 240         spin_lock_bh(&port->backlog.lock);
 241         skb_queue_splice_tail_init(&port->backlog, &list);
 242         spin_unlock_bh(&port->backlog.lock);
 243
 244         while ((skb = __skb_dequeue(&list)) != NULL) {
 245                 struct net_device *dev = skb->dev;
 246                 bool consumed = false;
 247
 248                 ethh = eth_hdr(skb);
 249                 tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
 250                 mac_hash = ipvlan_mac_hash(ethh->h_dest);
 251
 252                 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
 253                         pkt_type = PACKET_BROADCAST;
 254                 else
 255                         pkt_type = PACKET_MULTICAST;
 256
 257                 rcu_read_lock();
 258                 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
 259                         if (tx_pkt && (ipvlan->dev == skb->dev))
 260                                 continue;
 261                         if (!test_bit(mac_hash, ipvlan->mac_filters))
 262                                 continue;
 263                         if (!(ipvlan->dev->flags & IFF_UP))
 264                                 continue;
 265                         ret = NET_RX_DROP;
 266                         len = skb->len + ETH_HLEN;
 267                         nskb = skb_clone(skb, GFP_ATOMIC);
 268                         local_bh_disable();
 269                         if (nskb) {
 270                                 consumed = true;
 271                                 nskb->pkt_type = pkt_type;
 272                                 nskb->dev = ipvlan->dev;
 273                                 if (tx_pkt)
 274                                         ret = dev_forward_skb(ipvlan->dev, nskb);
 275                                 else
 276                                         ret = netif_rx(nskb);
 277                         }
 278                         ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
 279                         local_bh_enable();
 280                 }
 281                 rcu_read_unlock();
 282
 283                 if (tx_pkt) {
 284                         /* If the packet originated here, send it out. */
 285                         skb->dev = port->dev;
 286                         skb->pkt_type = pkt_type;
 287                         dev_queue_xmit(skb);
 288                 } else {
 289                         if (consumed)
 290                                 consume_skb(skb);
 291                         else
 292                                 kfree_skb(skb);
 293                 }
 294                 if (dev)
 295                         dev_put(dev);
 296                 cond_resched();
 297         }
 298 }
 299
 300 static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
 301 {
 302         bool xnet = true;
 303
 304         if (dev)
 305                 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
 306
 307         skb_scrub_packet(skb, xnet);
 308         if (dev)
 309                 skb->dev = dev;
 310 }
 311
 312 static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
 313                             bool local)
 314 {
 315         struct ipvl_dev *ipvlan = addr->master;
 316         struct net_device *dev = ipvlan->dev;
 317         unsigned int len;
 318         rx_handler_result_t ret = RX_HANDLER_CONSUMED;
 319         bool success = false;
 320         struct sk_buff *skb = *pskb;
 321
 322         len = skb->len + ETH_HLEN;
 323         /* Only packets exchanged between two local slaves need to have
 324          * device-up check as well as skb-share check.
 325          */
 326         if (local) {
 327                 if (unlikely(!(dev->flags & IFF_UP))) {
 328                         kfree_skb(skb);
 329                         goto out;
 330                 }
 331
 332                 skb = skb_share_check(skb, GFP_ATOMIC);
 333                 if (!skb)
 334                         goto out;
 335
 336                 *pskb = skb;
 337         }
 338
 339         if (local) {
 340                 skb->pkt_type = PACKET_HOST;
 341                 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
 342                         success = true;
 343         } else {
 344                 skb->dev = dev;
 345                 ret = RX_HANDLER_ANOTHER;
 346                 success = true;
 347         }
 348
 349 out:
 350         ipvlan_count_rx(ipvlan, len, success, false);
 351         return ret;
 352 }
 353
 354 struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port, void *lyr3h,
 355                                      int addr_type, bool use_dest)
 356 {
 357         struct ipvl_addr *addr = NULL;
 358
 359         switch (addr_type) {
 360 #if IS_ENABLED(CONFIG_IPV6)
 361         case IPVL_IPV6: {
 362                 struct ipv6hdr *ip6h;
 363                 struct in6_addr *i6addr;
 364
 365                 ip6h = (struct ipv6hdr *)lyr3h;
 366                 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
 367                 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 368                 break;
 369         }
 370         case IPVL_ICMPV6: {
 371                 struct nd_msg *ndmh;
 372                 struct in6_addr *i6addr;
 373
 374                 /* Make sure that the NeighborSolicitation ICMPv6 packets
 375                  * are handled to avoid DAD issue.
 376                  */
 377                 ndmh = (struct nd_msg *)lyr3h;
 378                 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
 379                         i6addr = &ndmh->target;
 380                         addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 381                 }
 382                 break;
 383         }
 384 #endif
 385         case IPVL_IPV4: {
 386                 struct iphdr *ip4h;
 387                 __be32 *i4addr;
 388
 389                 ip4h = (struct iphdr *)lyr3h;
 390                 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
 391                 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
 392                 break;
 393         }
 394         case IPVL_ARP: {
 395                 struct arphdr *arph;
 396                 unsigned char *arp_ptr;
 397                 __be32 dip;
 398
 399                 arph = (struct arphdr *)lyr3h;
 400                 arp_ptr = (unsigned char *)(arph + 1);
 401                 if (use_dest)
 402                         arp_ptr += (2 * port->dev->addr_len) + 4;
 403                 else
 404                         arp_ptr += port->dev->addr_len;
 405
 406                 memcpy(&dip, arp_ptr, 4);
 407                 addr = ipvlan_ht_addr_lookup(port, &dip, false);
 408                 break;
 409         }
 410         }
 411
 412         return addr;
 413 }
 414
 415 static noinline_for_stack int ipvlan_process_v4_outbound(struct sk_buff *skb)
 416 {
 417         const struct iphdr *ip4h = ip_hdr(skb);
 418         struct net_device *dev = skb->dev;
 419         struct net *net = dev_net(dev);
 420         struct rtable *rt;
 421         int err, ret = NET_XMIT_DROP;
 422         struct flowi4 fl4 = {
 423                 .flowi4_oif = dev->ifindex,
 424                 .flowi4_tos = RT_TOS(ip4h->tos),
 425                 .flowi4_flags = FLOWI_FLAG_ANYSRC,
 426                 .flowi4_mark = skb->mark,
 427                 .daddr = ip4h->daddr,
 428                 .saddr = ip4h->saddr,
 429         };
 430
 431         rt = ip_route_output_flow(net, &fl4, NULL);
 432         if (IS_ERR(rt))
 433                 goto err;
 434
 435         if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 436                 ip_rt_put(rt);
 437                 goto err;
 438         }
 439         skb_dst_set(skb, &rt->dst);
 440
 441         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 442
 443         err = ip_local_out(net, skb->sk, skb);
 444         if (unlikely(net_xmit_eval(err)))
 445                 DEV_STATS_INC(dev, tx_errors);
 446         else
 447                 ret = NET_XMIT_SUCCESS;
 448         goto out;
 449 err:
 450         DEV_STATS_INC(dev, tx_errors);
 451         kfree_skb(skb);
 452 out:
 453         return ret;
 454 }
 455
 456 #if IS_ENABLED(CONFIG_IPV6)
 457
 458 static noinline_for_stack int
 459 ipvlan_route_v6_outbound(struct net_device *dev, struct sk_buff *skb)
 460 {
 461         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 462         struct flowi6 fl6 = {
 463                 .flowi6_oif = dev->ifindex,
 464                 .daddr = ip6h->daddr,
 465                 .saddr = ip6h->saddr,
 466                 .flowi6_flags = FLOWI_FLAG_ANYSRC,
 467                 .flowlabel = ip6_flowinfo(ip6h),
 468                 .flowi6_mark = skb->mark,
 469                 .flowi6_proto = ip6h->nexthdr,
 470         };
 471         struct dst_entry *dst;
 472         int err;
 473
 474         dst = ip6_route_output(dev_net(dev), NULL, &fl6);
 475         err = dst->error;
 476         if (err) {
 477                 dst_release(dst);
 478                 return err;
 479         }
 480         skb_dst_set(skb, dst);
 481         return 0;
 482 }
 483
 484 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
 485 {
 486         struct net_device *dev = skb->dev;
 487         int err, ret = NET_XMIT_DROP;
 488
 489         err = ipvlan_route_v6_outbound(dev, skb);
 490         if (unlikely(err)) {
 491                 DEV_STATS_INC(dev, tx_errors);
 492                 kfree_skb(skb);
 493                 return err;
 494         }
 495
 496         memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
 497
 498         err = ip6_local_out(dev_net(dev), skb->sk, skb);
 499         if (unlikely(net_xmit_eval(err)))
 500                 DEV_STATS_INC(dev, tx_errors);
 501         else
 502                 ret = NET_XMIT_SUCCESS;
 503         return ret;
 504 }
 505 #else
 506 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
 507 {
 508         return NET_XMIT_DROP;
 509 }
 510 #endif
 511
 512 static int ipvlan_process_outbound(struct sk_buff *skb)
 513 {
 514         int ret = NET_XMIT_DROP;
 515
 516         /* The ipvlan is a pseudo-L2 device, so the packets that we receive
 517          * will have L2; which need to discarded and processed further
 518          * in the net-ns of the main-device.
 519          */
 520         if (skb_mac_header_was_set(skb)) {
 521                 /* In this mode we dont care about
 522                  * multicast and broadcast traffic */
 523                 struct ethhdr *ethh = eth_hdr(skb);
 524
 525                 if (is_multicast_ether_addr(ethh->h_dest)) {
 526                         pr_debug_ratelimited(
 527                                 "Dropped {multi|broad}cast of type=[%x]\n",
 528                                 ntohs(skb->protocol));
 529                         kfree_skb(skb);
 530                         goto out;
 531                 }
 532
 533                 skb_pull(skb, sizeof(*ethh));
 534                 skb->mac_header = (typeof(skb->mac_header))~0U;
 535                 skb_reset_network_header(skb);
 536         }
 537
 538         if (skb->protocol == htons(ETH_P_IPV6))
 539                 ret = ipvlan_process_v6_outbound(skb);
 540         else if (skb->protocol == htons(ETH_P_IP))
 541                 ret = ipvlan_process_v4_outbound(skb);
 542         else {
 543                 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
 544                                     ntohs(skb->protocol));
 545                 kfree_skb(skb);
 546         }
 547 out:
 548         return ret;
 549 }
 550
 551 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
 552                                      struct sk_buff *skb, bool tx_pkt)
 553 {
 554         if (skb->protocol == htons(ETH_P_PAUSE)) {
 555                 kfree_skb(skb);
 556                 return;
 557         }
 558
 559         /* Record that the deferred packet is from TX or RX path. By
 560          * looking at mac-addresses on packet will lead to erronus decisions.
 561          * (This would be true for a loopback-mode on master device or a
 562          * hair-pin mode of the switch.)
 563          */
 564         IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
 565
 566         spin_lock(&port->backlog.lock);
 567         if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
 568                 if (skb->dev)
 569                         dev_hold(skb->dev);
 570                 __skb_queue_tail(&port->backlog, skb);
 571                 spin_unlock(&port->backlog.lock);
 572                 schedule_work(&port->wq);
 573         } else {
 574                 spin_unlock(&port->backlog.lock);
 575                 atomic_long_inc(&skb->dev->rx_dropped);
 576                 kfree_skb(skb);
 577         }
 578 }
 579
 580 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
 581 {
 582         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 583         void *lyr3h;
 584         struct ipvl_addr *addr;
 585         int addr_type;
 586
 587         lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
 588         if (!lyr3h)
 589                 goto out;
 590
 591         if (!ipvlan_is_vepa(ipvlan->port)) {
 592                 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 593                 if (addr) {
 594                         if (ipvlan_is_private(ipvlan->port)) {
 595                                 consume_skb(skb);
 596                                 return NET_XMIT_DROP;
 597                         }
 598                         ipvlan_rcv_frame(addr, &skb, true);
 599                         return NET_XMIT_SUCCESS;
 600                 }
 601         }
 602 out:
 603         ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
 604         return ipvlan_process_outbound(skb);
 605 }
 606
 607 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
 608 {
 609         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 610         struct ethhdr *eth = skb_eth_hdr(skb);
 611         struct ipvl_addr *addr;
 612         void *lyr3h;
 613         int addr_type;
 614
 615         if (!ipvlan_is_vepa(ipvlan->port) &&
 616             ether_addr_equal(eth->h_dest, eth->h_source)) {
 617                 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
 618                 if (lyr3h) {
 619                         addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 620                         if (addr) {
 621                                 if (ipvlan_is_private(ipvlan->port)) {
 622                                         consume_skb(skb);
 623                                         return NET_XMIT_DROP;
 624                                 }
 625                                 ipvlan_rcv_frame(addr, &skb, true);
 626                                 return NET_XMIT_SUCCESS;
 627                         }
 628                 }
 629                 skb = skb_share_check(skb, GFP_ATOMIC);
 630                 if (!skb)
 631                         return NET_XMIT_DROP;
 632
 633                 /* Packet definitely does not belong to any of the
 634                  * virtual devices, but the dest is local. So forward
 635                  * the skb for the main-dev. At the RX side we just return
 636                  * RX_PASS for it to be processed further on the stack.
 637                  */
 638                 dev_forward_skb(ipvlan->phy_dev, skb);
 639                 return NET_XMIT_SUCCESS;
 640
 641         } else if (is_multicast_ether_addr(eth->h_dest)) {
 642                 skb_reset_mac_header(skb);
 643                 ipvlan_skb_crossing_ns(skb, NULL);
 644                 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
 645                 return NET_XMIT_SUCCESS;
 646         }
 647
 648         skb->dev = ipvlan->phy_dev;
 649         return dev_queue_xmit(skb);
 650 }
 651
 652 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 653 {
 654         struct ipvl_dev *ipvlan = netdev_priv(dev);
 655         struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
 656
 657         if (!port)
 658                 goto out;
 659
 660         if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
 661                 goto out;
 662
 663         switch(port->mode) {
 664         case IPVLAN_MODE_L2:
 665                 return ipvlan_xmit_mode_l2(skb, dev);
 666         case IPVLAN_MODE_L3:
 667 #ifdef CONFIG_IPVLAN_L3S
 668         case IPVLAN_MODE_L3S:
 669 #endif
 670                 return ipvlan_xmit_mode_l3(skb, dev);
 671         }
 672
 673         /* Should not reach here */
 674         WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
 675                           port->mode);
 676 out:
 677         kfree_skb(skb);
 678         return NET_XMIT_DROP;
 679 }
 680
 681 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
 682 {
 683         struct ethhdr *eth = eth_hdr(skb);
 684         struct ipvl_addr *addr;
 685         void *lyr3h;
 686         int addr_type;
 687
 688         if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
 689                 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 690                 if (!lyr3h)
 691                         return true;
 692
 693                 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
 694                 if (addr)
 695                         return false;
 696         }
 697
 698         return true;
 699 }
 700
 701 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
 702                                                  struct ipvl_port *port)
 703 {
 704         void *lyr3h;
 705         int addr_type;
 706         struct ipvl_addr *addr;
 707         struct sk_buff *skb = *pskb;
 708         rx_handler_result_t ret = RX_HANDLER_PASS;
 709
 710         lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 711         if (!lyr3h)
 712                 goto out;
 713
 714         addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 715         if (addr)
 716                 ret = ipvlan_rcv_frame(addr, pskb, false);
 717
 718 out:
 719         return ret;
 720 }
 721
 722 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
 723                                                  struct ipvl_port *port)
 724 {
 725         struct sk_buff *skb = *pskb;
 726         struct ethhdr *eth = eth_hdr(skb);
 727         rx_handler_result_t ret = RX_HANDLER_PASS;
 728
 729         if (is_multicast_ether_addr(eth->h_dest)) {
 730                 if (ipvlan_external_frame(skb, port)) {
 731                         struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
 732
 733                         /* External frames are queued for device local
 734                          * distribution, but a copy is given to master
 735                          * straight away to avoid sending duplicates later
 736                          * when work-queue processes this frame. This is
 737                          * achieved by returning RX_HANDLER_PASS.
 738                          */
 739                         if (nskb) {
 740                                 ipvlan_skb_crossing_ns(nskb, NULL);
 741                                 ipvlan_multicast_enqueue(port, nskb, false);
 742                         }
 743                 }
 744         } else {
 745                 /* Perform like l3 mode for non-multicast packet */
 746                 ret = ipvlan_handle_mode_l3(pskb, port);
 747         }
 748
 749         return ret;
 750 }
 751
 752 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
 753 {
 754         struct sk_buff *skb = *pskb;
 755         struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
 756
 757         if (!port)
 758                 return RX_HANDLER_PASS;
 759
 760         switch (port->mode) {
 761         case IPVLAN_MODE_L2:
 762                 return ipvlan_handle_mode_l2(pskb, port);
 763         case IPVLAN_MODE_L3:
 764                 return ipvlan_handle_mode_l3(pskb, port);
 765 #ifdef CONFIG_IPVLAN_L3S
 766         case IPVLAN_MODE_L3S:
 767                 return RX_HANDLER_PASS;
 768 #endif
 769         }
 770
 771         /* Should not reach here */
 772         WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
 773                           port->mode);
 774         kfree_skb(skb);
 775         return RX_HANDLER_CONSUMED;
 776 }