drivers/net/ipvlan/ipvlan_core.c

   1 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
   2  *
   3  * This program is free software; you can redistribute it and/or
   4  * modify it under the terms of the GNU General Public License as
   5  * published by the Free Software Foundation; either version 2 of
   6  * the License, or (at your option) any later version.
   7  *
   8  */
   9
  10 #include "ipvlan.h"
  11
  12 static u32 ipvlan_jhash_secret __read_mostly;
  13
  14 void ipvlan_init_secret(void)
  15 {
  16         net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
  17 }
  18
  19 void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
  20                             unsigned int len, bool success, bool mcast)
  21 {
  22         if (likely(success)) {
  23                 struct ipvl_pcpu_stats *pcptr;
  24
  25                 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
  26                 u64_stats_update_begin(&pcptr->syncp);
  27                 pcptr->rx_pkts++;
  28                 pcptr->rx_bytes += len;
  29                 if (mcast)
  30                         pcptr->rx_mcast++;
  31                 u64_stats_update_end(&pcptr->syncp);
  32         } else {
  33                 this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
  34         }
  35 }
  36 EXPORT_SYMBOL_GPL(ipvlan_count_rx);
  37
  38 static u8 ipvlan_get_v6_hash(const void *iaddr)
  39 {
  40         const struct in6_addr *ip6_addr = iaddr;
  41
  42         return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
  43                IPVLAN_HASH_MASK;
  44 }
  45
  46 static u8 ipvlan_get_v4_hash(const void *iaddr)
  47 {
  48         const struct in_addr *ip4_addr = iaddr;
  49
  50         return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
  51                IPVLAN_HASH_MASK;
  52 }
  53
  54 static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
  55                                                const void *iaddr, bool is_v6)
  56 {
  57         struct ipvl_addr *addr;
  58         u8 hash;
  59
  60         hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
  61                ipvlan_get_v4_hash(iaddr);
  62         hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) {
  63                 if (is_v6 && addr->atype == IPVL_IPV6 &&
  64                     ipv6_addr_equal(&addr->ip6addr, iaddr))
  65                         return addr;
  66                 else if (!is_v6 && addr->atype == IPVL_IPV4 &&
  67                          addr->ip4addr.s_addr ==
  68                                 ((struct in_addr *)iaddr)->s_addr)
  69                         return addr;
  70         }
  71         return NULL;
  72 }
  73
  74 void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
  75 {
  76         struct ipvl_port *port = ipvlan->port;
  77         u8 hash;
  78
  79         hash = (addr->atype == IPVL_IPV6) ?
  80                ipvlan_get_v6_hash(&addr->ip6addr) :
  81                ipvlan_get_v4_hash(&addr->ip4addr);
  82         if (hlist_unhashed(&addr->hlnode))
  83                 hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
  84 }
  85
  86 void ipvlan_ht_addr_del(struct ipvl_addr *addr)
  87 {
  88         hlist_del_init_rcu(&addr->hlnode);
  89 }
  90
  91 struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
  92                                    const void *iaddr, bool is_v6)
  93 {
  94         struct ipvl_addr *addr;
  95
  96         list_for_each_entry(addr, &ipvlan->addrs, anode) {
  97                 if ((is_v6 && addr->atype == IPVL_IPV6 &&
  98                     ipv6_addr_equal(&addr->ip6addr, iaddr)) ||
  99                     (!is_v6 && addr->atype == IPVL_IPV4 &&
 100                     addr->ip4addr.s_addr == ((struct in_addr *)iaddr)->s_addr))
 101                         return addr;
 102         }
 103         return NULL;
 104 }
 105
 106 bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
 107 {
 108         struct ipvl_dev *ipvlan;
 109
 110         ASSERT_RTNL();
 111
 112         list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
 113                 if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
 114                         return true;
 115         }
 116         return false;
 117 }
 118
 119 static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
 120 {
 121         void *lyr3h = NULL;
 122
 123         switch (skb->protocol) {
 124         case htons(ETH_P_ARP): {
 125                 struct arphdr *arph;
 126
 127                 if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
 128                         return NULL;
 129
 130                 arph = arp_hdr(skb);
 131                 *type = IPVL_ARP;
 132                 lyr3h = arph;
 133                 break;
 134         }
 135         case htons(ETH_P_IP): {
 136                 u32 pktlen;
 137                 struct iphdr *ip4h;
 138
 139                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
 140                         return NULL;
 141
 142                 ip4h = ip_hdr(skb);
 143                 pktlen = ntohs(ip4h->tot_len);
 144                 if (ip4h->ihl < 5 || ip4h->version != 4)
 145                         return NULL;
 146                 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
 147                         return NULL;
 148
 149                 *type = IPVL_IPV4;
 150                 lyr3h = ip4h;
 151                 break;
 152         }
 153         case htons(ETH_P_IPV6): {
 154                 struct ipv6hdr *ip6h;
 155
 156                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
 157                         return NULL;
 158
 159                 ip6h = ipv6_hdr(skb);
 160                 if (ip6h->version != 6)
 161                         return NULL;
 162
 163                 *type = IPVL_IPV6;
 164                 lyr3h = ip6h;
 165                 /* Only Neighbour Solicitation pkts need different treatment */
 166                 if (ipv6_addr_any(&ip6h->saddr) &&
 167                     ip6h->nexthdr == NEXTHDR_ICMP) {
 168                         *type = IPVL_ICMPV6;
 169                         lyr3h = ip6h + 1;
 170                 }
 171                 break;
 172         }
 173         default:
 174                 return NULL;
 175         }
 176
 177         return lyr3h;
 178 }
 179
 180 unsigned int ipvlan_mac_hash(const unsigned char *addr)
 181 {
 182         u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
 183                                ipvlan_jhash_secret);
 184
 185         return hash & IPVLAN_MAC_FILTER_MASK;
 186 }
 187
 188 void ipvlan_process_multicast(struct work_struct *work)
 189 {
 190         struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
 191         struct ethhdr *ethh;
 192         struct ipvl_dev *ipvlan;
 193         struct sk_buff *skb, *nskb;
 194         struct sk_buff_head list;
 195         unsigned int len;
 196         unsigned int mac_hash;
 197         int ret;
 198         u8 pkt_type;
 199         bool tx_pkt;
 200
 201         __skb_queue_head_init(&list);
 202
 203         spin_lock_bh(&port->backlog.lock);
 204         skb_queue_splice_tail_init(&port->backlog, &list);
 205         spin_unlock_bh(&port->backlog.lock);
 206
 207         while ((skb = __skb_dequeue(&list)) != NULL) {
 208                 struct net_device *dev = skb->dev;
 209                 bool consumed = false;
 210
 211                 ethh = eth_hdr(skb);
 212                 tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
 213                 mac_hash = ipvlan_mac_hash(ethh->h_dest);
 214
 215                 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
 216                         pkt_type = PACKET_BROADCAST;
 217                 else
 218                         pkt_type = PACKET_MULTICAST;
 219
 220                 rcu_read_lock();
 221                 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
 222                         if (tx_pkt && (ipvlan->dev == skb->dev))
 223                                 continue;
 224                         if (!test_bit(mac_hash, ipvlan->mac_filters))
 225                                 continue;
 226                         if (!(ipvlan->dev->flags & IFF_UP))
 227                                 continue;
 228                         ret = NET_RX_DROP;
 229                         len = skb->len + ETH_HLEN;
 230                         nskb = skb_clone(skb, GFP_ATOMIC);
 231                         local_bh_disable();
 232                         if (nskb) {
 233                                 consumed = true;
 234                                 nskb->pkt_type = pkt_type;
 235                                 nskb->dev = ipvlan->dev;
 236                                 if (tx_pkt)
 237                                         ret = dev_forward_skb(ipvlan->dev, nskb);
 238                                 else
 239                                         ret = netif_rx(nskb);
 240                         }
 241                         ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
 242                         local_bh_enable();
 243                 }
 244                 rcu_read_unlock();
 245
 246                 if (tx_pkt) {
 247                         /* If the packet originated here, send it out. */
 248                         skb->dev = port->dev;
 249                         skb->pkt_type = pkt_type;
 250                         dev_queue_xmit(skb);
 251                 } else {
 252                         if (consumed)
 253                                 consume_skb(skb);
 254                         else
 255                                 kfree_skb(skb);
 256                 }
 257                 if (dev)
 258                         dev_put(dev);
 259         }
 260 }
 261
 262 static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
 263 {
 264         bool xnet = true;
 265
 266         if (dev)
 267                 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
 268
 269         skb_scrub_packet(skb, xnet);
 270         if (dev)
 271                 skb->dev = dev;
 272 }
 273
 274 static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
 275                             bool local)
 276 {
 277         struct ipvl_dev *ipvlan = addr->master;
 278         struct net_device *dev = ipvlan->dev;
 279         unsigned int len;
 280         rx_handler_result_t ret = RX_HANDLER_CONSUMED;
 281         bool success = false;
 282         struct sk_buff *skb = *pskb;
 283
 284         len = skb->len + ETH_HLEN;
 285         /* Only packets exchanged between two local slaves need to have
 286          * device-up check as well as skb-share check.
 287          */
 288         if (local) {
 289                 if (unlikely(!(dev->flags & IFF_UP))) {
 290                         kfree_skb(skb);
 291                         goto out;
 292                 }
 293
 294                 skb = skb_share_check(skb, GFP_ATOMIC);
 295                 if (!skb)
 296                         goto out;
 297
 298                 *pskb = skb;
 299         }
 300         ipvlan_skb_crossing_ns(skb, dev);
 301
 302         if (local) {
 303                 skb->pkt_type = PACKET_HOST;
 304                 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
 305                         success = true;
 306         } else {
 307                 ret = RX_HANDLER_ANOTHER;
 308                 success = true;
 309         }
 310
 311 out:
 312         ipvlan_count_rx(ipvlan, len, success, false);
 313         return ret;
 314 }
 315
 316 static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
 317                                             void *lyr3h, int addr_type,
 318                                             bool use_dest)
 319 {
 320         struct ipvl_addr *addr = NULL;
 321
 322         if (addr_type == IPVL_IPV6) {
 323                 struct ipv6hdr *ip6h;
 324                 struct in6_addr *i6addr;
 325
 326                 ip6h = (struct ipv6hdr *)lyr3h;
 327                 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
 328                 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 329         } else if (addr_type == IPVL_ICMPV6) {
 330                 struct nd_msg *ndmh;
 331                 struct in6_addr *i6addr;
 332
 333                 /* Make sure that the NeighborSolicitation ICMPv6 packets
 334                  * are handled to avoid DAD issue.
 335                  */
 336                 ndmh = (struct nd_msg *)lyr3h;
 337                 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
 338                         i6addr = &ndmh->target;
 339                         addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 340                 }
 341         } else if (addr_type == IPVL_IPV4) {
 342                 struct iphdr *ip4h;
 343                 __be32 *i4addr;
 344
 345                 ip4h = (struct iphdr *)lyr3h;
 346                 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
 347                 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
 348         } else if (addr_type == IPVL_ARP) {
 349                 struct arphdr *arph;
 350                 unsigned char *arp_ptr;
 351                 __be32 dip;
 352
 353                 arph = (struct arphdr *)lyr3h;
 354                 arp_ptr = (unsigned char *)(arph + 1);
 355                 if (use_dest)
 356                         arp_ptr += (2 * port->dev->addr_len) + 4;
 357                 else
 358                         arp_ptr += port->dev->addr_len;
 359
 360                 memcpy(&dip, arp_ptr, 4);
 361                 addr = ipvlan_ht_addr_lookup(port, &dip, false);
 362         }
 363
 364         return addr;
 365 }
 366
 367 static int ipvlan_process_v4_outbound(struct sk_buff *skb)
 368 {
 369         const struct iphdr *ip4h = ip_hdr(skb);
 370         struct net_device *dev = skb->dev;
 371         struct net *net = dev_net(dev);
 372         struct rtable *rt;
 373         int err, ret = NET_XMIT_DROP;
 374         struct flowi4 fl4 = {
 375                 .flowi4_oif = dev->ifindex,
 376                 .flowi4_tos = RT_TOS(ip4h->tos),
 377                 .flowi4_flags = FLOWI_FLAG_ANYSRC,
 378                 .daddr = ip4h->daddr,
 379                 .saddr = ip4h->saddr,
 380         };
 381
 382         rt = ip_route_output_flow(net, &fl4, NULL);
 383         if (IS_ERR(rt))
 384                 goto err;
 385
 386         if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 387                 ip_rt_put(rt);
 388                 goto err;
 389         }
 390         skb_dst_set(skb, &rt->dst);
 391         err = ip_local_out(net, skb->sk, skb);
 392         if (unlikely(net_xmit_eval(err)))
 393                 dev->stats.tx_errors++;
 394         else
 395                 ret = NET_XMIT_SUCCESS;
 396         goto out;
 397 err:
 398         dev->stats.tx_errors++;
 399         kfree_skb(skb);
 400 out:
 401         return ret;
 402 }
 403
 404 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
 405 {
 406         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 407         struct net_device *dev = skb->dev;
 408         struct net *net = dev_net(dev);
 409         struct dst_entry *dst;
 410         int err, ret = NET_XMIT_DROP;
 411         struct flowi6 fl6 = {
 412                 .flowi6_iif = dev->ifindex,
 413                 .daddr = ip6h->daddr,
 414                 .saddr = ip6h->saddr,
 415                 .flowi6_flags = FLOWI_FLAG_ANYSRC,
 416                 .flowlabel = ip6_flowinfo(ip6h),
 417                 .flowi6_mark = skb->mark,
 418                 .flowi6_proto = ip6h->nexthdr,
 419         };
 420
 421         dst = ip6_route_output(net, NULL, &fl6);
 422         if (dst->error) {
 423                 ret = dst->error;
 424                 dst_release(dst);
 425                 goto err;
 426         }
 427         skb_dst_set(skb, dst);
 428         err = ip6_local_out(net, skb->sk, skb);
 429         if (unlikely(net_xmit_eval(err)))
 430                 dev->stats.tx_errors++;
 431         else
 432                 ret = NET_XMIT_SUCCESS;
 433         goto out;
 434 err:
 435         dev->stats.tx_errors++;
 436         kfree_skb(skb);
 437 out:
 438         return ret;
 439 }
 440
 441 static int ipvlan_process_outbound(struct sk_buff *skb)
 442 {
 443         struct ethhdr *ethh = eth_hdr(skb);
 444         int ret = NET_XMIT_DROP;
 445
 446         /* In this mode we dont care about multicast and broadcast traffic */
 447         if (is_multicast_ether_addr(ethh->h_dest)) {
 448                 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
 449                                     ntohs(skb->protocol));
 450                 kfree_skb(skb);
 451                 goto out;
 452         }
 453
 454         /* The ipvlan is a pseudo-L2 device, so the packets that we receive
 455          * will have L2; which need to discarded and processed further
 456          * in the net-ns of the main-device.
 457          */
 458         if (skb_mac_header_was_set(skb)) {
 459                 skb_pull(skb, sizeof(*ethh));
 460                 skb->mac_header = (typeof(skb->mac_header))~0U;
 461                 skb_reset_network_header(skb);
 462         }
 463
 464         if (skb->protocol == htons(ETH_P_IPV6))
 465                 ret = ipvlan_process_v6_outbound(skb);
 466         else if (skb->protocol == htons(ETH_P_IP))
 467                 ret = ipvlan_process_v4_outbound(skb);
 468         else {
 469                 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
 470                                     ntohs(skb->protocol));
 471                 kfree_skb(skb);
 472         }
 473 out:
 474         return ret;
 475 }
 476
 477 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
 478                                      struct sk_buff *skb, bool tx_pkt)
 479 {
 480         if (skb->protocol == htons(ETH_P_PAUSE)) {
 481                 kfree_skb(skb);
 482                 return;
 483         }
 484
 485         /* Record that the deferred packet is from TX or RX path. By
 486          * looking at mac-addresses on packet will lead to erronus decisions.
 487          * (This would be true for a loopback-mode on master device or a
 488          * hair-pin mode of the switch.)
 489          */
 490         IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
 491
 492         spin_lock(&port->backlog.lock);
 493         if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
 494                 if (skb->dev)
 495                         dev_hold(skb->dev);
 496                 __skb_queue_tail(&port->backlog, skb);
 497                 spin_unlock(&port->backlog.lock);
 498                 schedule_work(&port->wq);
 499         } else {
 500                 spin_unlock(&port->backlog.lock);
 501                 atomic_long_inc(&skb->dev->rx_dropped);
 502                 kfree_skb(skb);
 503         }
 504 }
 505
 506 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
 507 {
 508         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 509         void *lyr3h;
 510         struct ipvl_addr *addr;
 511         int addr_type;
 512
 513         lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 514         if (!lyr3h)
 515                 goto out;
 516
 517         addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 518         if (addr)
 519                 return ipvlan_rcv_frame(addr, &skb, true);
 520
 521 out:
 522         ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
 523         return ipvlan_process_outbound(skb);
 524 }
 525
 526 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
 527 {
 528         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 529         struct ethhdr *eth = eth_hdr(skb);
 530         struct ipvl_addr *addr;
 531         void *lyr3h;
 532         int addr_type;
 533
 534         if (ether_addr_equal(eth->h_dest, eth->h_source)) {
 535                 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 536                 if (lyr3h) {
 537                         addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 538                         if (addr)
 539                                 return ipvlan_rcv_frame(addr, &skb, true);
 540                 }
 541                 skb = skb_share_check(skb, GFP_ATOMIC);
 542                 if (!skb)
 543                         return NET_XMIT_DROP;
 544
 545                 /* Packet definitely does not belong to any of the
 546                  * virtual devices, but the dest is local. So forward
 547                  * the skb for the main-dev. At the RX side we just return
 548                  * RX_PASS for it to be processed further on the stack.
 549                  */
 550                 return dev_forward_skb(ipvlan->phy_dev, skb);
 551
 552         } else if (is_multicast_ether_addr(eth->h_dest)) {
 553                 ipvlan_skb_crossing_ns(skb, NULL);
 554                 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
 555                 return NET_XMIT_SUCCESS;
 556         }
 557
 558         ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
 559         return dev_queue_xmit(skb);
 560 }
 561
 562 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 563 {
 564         struct ipvl_dev *ipvlan = netdev_priv(dev);
 565         struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
 566
 567         if (!port)
 568                 goto out;
 569
 570         if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
 571                 goto out;
 572
 573         switch(port->mode) {
 574         case IPVLAN_MODE_L2:
 575                 return ipvlan_xmit_mode_l2(skb, dev);
 576         case IPVLAN_MODE_L3:
 577         case IPVLAN_MODE_L3S:
 578                 return ipvlan_xmit_mode_l3(skb, dev);
 579         }
 580
 581         /* Should not reach here */
 582         WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
 583                           port->mode);
 584 out:
 585         kfree_skb(skb);
 586         return NET_XMIT_DROP;
 587 }
 588
 589 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
 590 {
 591         struct ethhdr *eth = eth_hdr(skb);
 592         struct ipvl_addr *addr;
 593         void *lyr3h;
 594         int addr_type;
 595
 596         if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
 597                 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 598                 if (!lyr3h)
 599                         return true;
 600
 601                 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
 602                 if (addr)
 603                         return false;
 604         }
 605
 606         return true;
 607 }
 608
 609 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
 610                                                  struct ipvl_port *port)
 611 {
 612         void *lyr3h;
 613         int addr_type;
 614         struct ipvl_addr *addr;
 615         struct sk_buff *skb = *pskb;
 616         rx_handler_result_t ret = RX_HANDLER_PASS;
 617
 618         lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 619         if (!lyr3h)
 620                 goto out;
 621
 622         addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 623         if (addr)
 624                 ret = ipvlan_rcv_frame(addr, pskb, false);
 625
 626 out:
 627         return ret;
 628 }
 629
 630 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
 631                                                  struct ipvl_port *port)
 632 {
 633         struct sk_buff *skb = *pskb;
 634         struct ethhdr *eth = eth_hdr(skb);
 635         rx_handler_result_t ret = RX_HANDLER_PASS;
 636         void *lyr3h;
 637         int addr_type;
 638
 639         if (is_multicast_ether_addr(eth->h_dest)) {
 640                 if (ipvlan_external_frame(skb, port)) {
 641                         struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
 642
 643                         /* External frames are queued for device local
 644                          * distribution, but a copy is given to master
 645                          * straight away to avoid sending duplicates later
 646                          * when work-queue processes this frame. This is
 647                          * achieved by returning RX_HANDLER_PASS.
 648                          */
 649                         if (nskb) {
 650                                 ipvlan_skb_crossing_ns(nskb, NULL);
 651                                 ipvlan_multicast_enqueue(port, nskb, false);
 652                         }
 653                 }
 654         } else {
 655                 struct ipvl_addr *addr;
 656
 657                 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 658                 if (!lyr3h)
 659                         return ret;
 660
 661                 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 662                 if (addr)
 663                         ret = ipvlan_rcv_frame(addr, pskb, false);
 664         }
 665
 666         return ret;
 667 }
 668
 669 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
 670 {
 671         struct sk_buff *skb = *pskb;
 672         struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
 673
 674         if (!port)
 675                 return RX_HANDLER_PASS;
 676
 677         switch (port->mode) {
 678         case IPVLAN_MODE_L2:
 679                 return ipvlan_handle_mode_l2(pskb, port);
 680         case IPVLAN_MODE_L3:
 681                 return ipvlan_handle_mode_l3(pskb, port);
 682         case IPVLAN_MODE_L3S:
 683                 return RX_HANDLER_PASS;
 684         }
 685
 686         /* Should not reach here */
 687         WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
 688                           port->mode);
 689         kfree_skb(skb);
 690         return RX_HANDLER_CONSUMED;
 691 }
 692
 693 static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
 694                                             struct net_device *dev)
 695 {
 696         struct ipvl_addr *addr = NULL;
 697         struct ipvl_port *port;
 698         void *lyr3h;
 699         int addr_type;
 700
 701         if (!dev || !netif_is_ipvlan_port(dev))
 702                 goto out;
 703
 704         port = ipvlan_port_get_rcu(dev);
 705         if (!port || port->mode != IPVLAN_MODE_L3S)
 706                 goto out;
 707
 708         lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 709         if (!lyr3h)
 710                 goto out;
 711
 712         addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 713 out:
 714         return addr;
 715 }
 716
 717 struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
 718                               u16 proto)
 719 {
 720         struct ipvl_addr *addr;
 721         struct net_device *sdev;
 722
 723         addr = ipvlan_skb_to_addr(skb, dev);
 724         if (!addr)
 725                 goto out;
 726
 727         sdev = addr->master->dev;
 728         switch (proto) {
 729         case AF_INET:
 730         {
 731                 int err;
 732                 struct iphdr *ip4h = ip_hdr(skb);
 733
 734                 err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
 735                                            ip4h->tos, sdev);
 736                 if (unlikely(err))
 737                         goto out;
 738                 break;
 739         }
 740         case AF_INET6:
 741         {
 742                 struct dst_entry *dst;
 743                 struct ipv6hdr *ip6h = ipv6_hdr(skb);
 744                 int flags = RT6_LOOKUP_F_HAS_SADDR;
 745                 struct flowi6 fl6 = {
 746                         .flowi6_iif   = sdev->ifindex,
 747                         .daddr        = ip6h->daddr,
 748                         .saddr        = ip6h->saddr,
 749                         .flowlabel    = ip6_flowinfo(ip6h),
 750                         .flowi6_mark  = skb->mark,
 751                         .flowi6_proto = ip6h->nexthdr,
 752                 };
 753
 754                 skb_dst_drop(skb);
 755                 dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
 756                 skb_dst_set(skb, dst);
 757                 break;
 758         }
 759         default:
 760                 break;
 761         }
 762
 763 out:
 764         return skb;
 765 }
 766
 767 unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
 768                              const struct nf_hook_state *state)
 769 {
 770         struct ipvl_addr *addr;
 771         unsigned int len;
 772
 773         addr = ipvlan_skb_to_addr(skb, skb->dev);
 774         if (!addr)
 775                 goto out;
 776
 777         skb->dev = addr->master->dev;
 778         len = skb->len + ETH_HLEN;
 779         ipvlan_count_rx(addr->master, len, true, false);
 780 out:
 781         return NF_ACCEPT;
 782 }