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 ipvl_port *port, 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, arp_hdr_len(port->dev))))
 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                         struct icmp6hdr *icmph;
 169
 170                         if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph))))
 171                                 return NULL;
 172
 173                         ip6h = ipv6_hdr(skb);
 174                         icmph = (struct icmp6hdr *)(ip6h + 1);
 175
 176                         if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
 177                                 /* Need to access the ipv6 address in body */
 178                                 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)
 179                                                 + sizeof(struct in6_addr))))
 180                                         return NULL;
 181
 182                                 ip6h = ipv6_hdr(skb);
 183                                 icmph = (struct icmp6hdr *)(ip6h + 1);
 184                         }
 185
 186                         *type = IPVL_ICMPV6;
 187                         lyr3h = icmph;
 188                 }
 189                 break;
 190         }
 191         default:
 192                 return NULL;
 193         }
 194
 195         return lyr3h;
 196 }
 197
 198 unsigned int ipvlan_mac_hash(const unsigned char *addr)
 199 {
 200         u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
 201                                ipvlan_jhash_secret);
 202
 203         return hash & IPVLAN_MAC_FILTER_MASK;
 204 }
 205
 206 void ipvlan_process_multicast(struct work_struct *work)
 207 {
 208         struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
 209         struct ethhdr *ethh;
 210         struct ipvl_dev *ipvlan;
 211         struct sk_buff *skb, *nskb;
 212         struct sk_buff_head list;
 213         unsigned int len;
 214         unsigned int mac_hash;
 215         int ret;
 216         u8 pkt_type;
 217         bool tx_pkt;
 218
 219         __skb_queue_head_init(&list);
 220
 221         spin_lock_bh(&port->backlog.lock);
 222         skb_queue_splice_tail_init(&port->backlog, &list);
 223         spin_unlock_bh(&port->backlog.lock);
 224
 225         while ((skb = __skb_dequeue(&list)) != NULL) {
 226                 struct net_device *dev = skb->dev;
 227                 bool consumed = false;
 228
 229                 ethh = eth_hdr(skb);
 230                 tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
 231                 mac_hash = ipvlan_mac_hash(ethh->h_dest);
 232
 233                 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
 234                         pkt_type = PACKET_BROADCAST;
 235                 else
 236                         pkt_type = PACKET_MULTICAST;
 237
 238                 rcu_read_lock();
 239                 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
 240                         if (tx_pkt && (ipvlan->dev == skb->dev))
 241                                 continue;
 242                         if (!test_bit(mac_hash, ipvlan->mac_filters))
 243                                 continue;
 244                         if (!(ipvlan->dev->flags & IFF_UP))
 245                                 continue;
 246                         ret = NET_RX_DROP;
 247                         len = skb->len + ETH_HLEN;
 248                         nskb = skb_clone(skb, GFP_ATOMIC);
 249                         local_bh_disable();
 250                         if (nskb) {
 251                                 consumed = true;
 252                                 nskb->pkt_type = pkt_type;
 253                                 nskb->dev = ipvlan->dev;
 254                                 if (tx_pkt)
 255                                         ret = dev_forward_skb(ipvlan->dev, nskb);
 256                                 else
 257                                         ret = netif_rx(nskb);
 258                         }
 259                         ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
 260                         local_bh_enable();
 261                 }
 262                 rcu_read_unlock();
 263
 264                 if (tx_pkt) {
 265                         /* If the packet originated here, send it out. */
 266                         skb->dev = port->dev;
 267                         skb->pkt_type = pkt_type;
 268                         dev_queue_xmit(skb);
 269                 } else {
 270                         if (consumed)
 271                                 consume_skb(skb);
 272                         else
 273                                 kfree_skb(skb);
 274                 }
 275                 if (dev)
 276                         dev_put(dev);
 277         }
 278 }
 279
 280 static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
 281 {
 282         bool xnet = true;
 283
 284         if (dev)
 285                 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
 286
 287         skb_scrub_packet(skb, xnet);
 288         if (dev)
 289                 skb->dev = dev;
 290 }
 291
 292 static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
 293                             bool local)
 294 {
 295         struct ipvl_dev *ipvlan = addr->master;
 296         struct net_device *dev = ipvlan->dev;
 297         unsigned int len;
 298         rx_handler_result_t ret = RX_HANDLER_CONSUMED;
 299         bool success = false;
 300         struct sk_buff *skb = *pskb;
 301
 302         len = skb->len + ETH_HLEN;
 303         /* Only packets exchanged between two local slaves need to have
 304          * device-up check as well as skb-share check.
 305          */
 306         if (local) {
 307                 if (unlikely(!(dev->flags & IFF_UP))) {
 308                         kfree_skb(skb);
 309                         goto out;
 310                 }
 311
 312                 skb = skb_share_check(skb, GFP_ATOMIC);
 313                 if (!skb)
 314                         goto out;
 315
 316                 *pskb = skb;
 317         }
 318         ipvlan_skb_crossing_ns(skb, dev);
 319
 320         if (local) {
 321                 skb->pkt_type = PACKET_HOST;
 322                 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
 323                         success = true;
 324         } else {
 325                 ret = RX_HANDLER_ANOTHER;
 326                 success = true;
 327         }
 328
 329 out:
 330         ipvlan_count_rx(ipvlan, len, success, false);
 331         return ret;
 332 }
 333
 334 static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
 335                                             void *lyr3h, int addr_type,
 336                                             bool use_dest)
 337 {
 338         struct ipvl_addr *addr = NULL;
 339
 340         if (addr_type == IPVL_IPV6) {
 341                 struct ipv6hdr *ip6h;
 342                 struct in6_addr *i6addr;
 343
 344                 ip6h = (struct ipv6hdr *)lyr3h;
 345                 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
 346                 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 347         } else if (addr_type == IPVL_ICMPV6) {
 348                 struct nd_msg *ndmh;
 349                 struct in6_addr *i6addr;
 350
 351                 /* Make sure that the NeighborSolicitation ICMPv6 packets
 352                  * are handled to avoid DAD issue.
 353                  */
 354                 ndmh = (struct nd_msg *)lyr3h;
 355                 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
 356                         i6addr = &ndmh->target;
 357                         addr = ipvlan_ht_addr_lookup(port, i6addr, true);
 358                 }
 359         } else if (addr_type == IPVL_IPV4) {
 360                 struct iphdr *ip4h;
 361                 __be32 *i4addr;
 362
 363                 ip4h = (struct iphdr *)lyr3h;
 364                 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
 365                 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
 366         } else if (addr_type == IPVL_ARP) {
 367                 struct arphdr *arph;
 368                 unsigned char *arp_ptr;
 369                 __be32 dip;
 370
 371                 arph = (struct arphdr *)lyr3h;
 372                 arp_ptr = (unsigned char *)(arph + 1);
 373                 if (use_dest)
 374                         arp_ptr += (2 * port->dev->addr_len) + 4;
 375                 else
 376                         arp_ptr += port->dev->addr_len;
 377
 378                 memcpy(&dip, arp_ptr, 4);
 379                 addr = ipvlan_ht_addr_lookup(port, &dip, false);
 380         }
 381
 382         return addr;
 383 }
 384
 385 static int ipvlan_process_v4_outbound(struct sk_buff *skb)
 386 {
 387         const struct iphdr *ip4h = ip_hdr(skb);
 388         struct net_device *dev = skb->dev;
 389         struct net *net = dev_net(dev);
 390         struct rtable *rt;
 391         int err, ret = NET_XMIT_DROP;
 392         struct flowi4 fl4 = {
 393                 .flowi4_oif = dev->ifindex,
 394                 .flowi4_tos = RT_TOS(ip4h->tos),
 395                 .flowi4_flags = FLOWI_FLAG_ANYSRC,
 396                 .flowi4_mark = skb->mark,
 397                 .daddr = ip4h->daddr,
 398                 .saddr = ip4h->saddr,
 399         };
 400
 401         rt = ip_route_output_flow(net, &fl4, NULL);
 402         if (IS_ERR(rt))
 403                 goto err;
 404
 405         if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 406                 ip_rt_put(rt);
 407                 goto err;
 408         }
 409         skb_dst_set(skb, &rt->dst);
 410         err = ip_local_out(net, skb->sk, skb);
 411         if (unlikely(net_xmit_eval(err)))
 412                 dev->stats.tx_errors++;
 413         else
 414                 ret = NET_XMIT_SUCCESS;
 415         goto out;
 416 err:
 417         dev->stats.tx_errors++;
 418         kfree_skb(skb);
 419 out:
 420         return ret;
 421 }
 422
 423 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
 424 {
 425         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 426         struct net_device *dev = skb->dev;
 427         struct net *net = dev_net(dev);
 428         struct dst_entry *dst;
 429         int err, ret = NET_XMIT_DROP;
 430         struct flowi6 fl6 = {
 431                 .flowi6_oif = dev->ifindex,
 432                 .daddr = ip6h->daddr,
 433                 .saddr = ip6h->saddr,
 434                 .flowi6_flags = FLOWI_FLAG_ANYSRC,
 435                 .flowlabel = ip6_flowinfo(ip6h),
 436                 .flowi6_mark = skb->mark,
 437                 .flowi6_proto = ip6h->nexthdr,
 438         };
 439
 440         dst = ip6_route_output(net, NULL, &fl6);
 441         if (dst->error) {
 442                 ret = dst->error;
 443                 dst_release(dst);
 444                 goto err;
 445         }
 446         skb_dst_set(skb, dst);
 447         err = ip6_local_out(net, skb->sk, skb);
 448         if (unlikely(net_xmit_eval(err)))
 449                 dev->stats.tx_errors++;
 450         else
 451                 ret = NET_XMIT_SUCCESS;
 452         goto out;
 453 err:
 454         dev->stats.tx_errors++;
 455         kfree_skb(skb);
 456 out:
 457         return ret;
 458 }
 459
 460 static int ipvlan_process_outbound(struct sk_buff *skb)
 461 {
 462         struct ethhdr *ethh = eth_hdr(skb);
 463         int ret = NET_XMIT_DROP;
 464
 465         /* In this mode we dont care about multicast and broadcast traffic */
 466         if (is_multicast_ether_addr(ethh->h_dest)) {
 467                 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
 468                                     ntohs(skb->protocol));
 469                 kfree_skb(skb);
 470                 goto out;
 471         }
 472
 473         /* The ipvlan is a pseudo-L2 device, so the packets that we receive
 474          * will have L2; which need to discarded and processed further
 475          * in the net-ns of the main-device.
 476          */
 477         if (skb_mac_header_was_set(skb)) {
 478                 skb_pull(skb, sizeof(*ethh));
 479                 skb->mac_header = (typeof(skb->mac_header))~0U;
 480                 skb_reset_network_header(skb);
 481         }
 482
 483         if (skb->protocol == htons(ETH_P_IPV6))
 484                 ret = ipvlan_process_v6_outbound(skb);
 485         else if (skb->protocol == htons(ETH_P_IP))
 486                 ret = ipvlan_process_v4_outbound(skb);
 487         else {
 488                 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
 489                                     ntohs(skb->protocol));
 490                 kfree_skb(skb);
 491         }
 492 out:
 493         return ret;
 494 }
 495
 496 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
 497                                      struct sk_buff *skb, bool tx_pkt)
 498 {
 499         if (skb->protocol == htons(ETH_P_PAUSE)) {
 500                 kfree_skb(skb);
 501                 return;
 502         }
 503
 504         /* Record that the deferred packet is from TX or RX path. By
 505          * looking at mac-addresses on packet will lead to erronus decisions.
 506          * (This would be true for a loopback-mode on master device or a
 507          * hair-pin mode of the switch.)
 508          */
 509         IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
 510
 511         spin_lock(&port->backlog.lock);
 512         if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
 513                 if (skb->dev)
 514                         dev_hold(skb->dev);
 515                 __skb_queue_tail(&port->backlog, skb);
 516                 spin_unlock(&port->backlog.lock);
 517                 schedule_work(&port->wq);
 518         } else {
 519                 spin_unlock(&port->backlog.lock);
 520                 atomic_long_inc(&skb->dev->rx_dropped);
 521                 kfree_skb(skb);
 522         }
 523 }
 524
 525 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
 526 {
 527         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 528         void *lyr3h;
 529         struct ipvl_addr *addr;
 530         int addr_type;
 531
 532         lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
 533         if (!lyr3h)
 534                 goto out;
 535
 536         if (!ipvlan_is_vepa(ipvlan->port)) {
 537                 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 538                 if (addr) {
 539                         if (ipvlan_is_private(ipvlan->port)) {
 540                                 consume_skb(skb);
 541                                 return NET_XMIT_DROP;
 542                         }
 543                         return ipvlan_rcv_frame(addr, &skb, true);
 544                 }
 545         }
 546 out:
 547         ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
 548         return ipvlan_process_outbound(skb);
 549 }
 550
 551 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
 552 {
 553         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 554         struct ethhdr *eth = eth_hdr(skb);
 555         struct ipvl_addr *addr;
 556         void *lyr3h;
 557         int addr_type;
 558
 559         if (!ipvlan_is_vepa(ipvlan->port) &&
 560             ether_addr_equal(eth->h_dest, eth->h_source)) {
 561                 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
 562                 if (lyr3h) {
 563                         addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
 564                         if (addr) {
 565                                 if (ipvlan_is_private(ipvlan->port)) {
 566                                         consume_skb(skb);
 567                                         return NET_XMIT_DROP;
 568                                 }
 569                                 return ipvlan_rcv_frame(addr, &skb, true);
 570                         }
 571                 }
 572                 skb = skb_share_check(skb, GFP_ATOMIC);
 573                 if (!skb)
 574                         return NET_XMIT_DROP;
 575
 576                 /* Packet definitely does not belong to any of the
 577                  * virtual devices, but the dest is local. So forward
 578                  * the skb for the main-dev. At the RX side we just return
 579                  * RX_PASS for it to be processed further on the stack.
 580                  */
 581                 return dev_forward_skb(ipvlan->phy_dev, skb);
 582
 583         } else if (is_multicast_ether_addr(eth->h_dest)) {
 584                 ipvlan_skb_crossing_ns(skb, NULL);
 585                 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
 586                 return NET_XMIT_SUCCESS;
 587         }
 588
 589         ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
 590         return dev_queue_xmit(skb);
 591 }
 592
 593 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
 594 {
 595         struct ipvl_dev *ipvlan = netdev_priv(dev);
 596         struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
 597
 598         if (!port)
 599                 goto out;
 600
 601         if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
 602                 goto out;
 603
 604         switch(port->mode) {
 605         case IPVLAN_MODE_L2:
 606                 return ipvlan_xmit_mode_l2(skb, dev);
 607         case IPVLAN_MODE_L3:
 608         case IPVLAN_MODE_L3S:
 609                 return ipvlan_xmit_mode_l3(skb, dev);
 610         }
 611
 612         /* Should not reach here */
 613         WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
 614                           port->mode);
 615 out:
 616         kfree_skb(skb);
 617         return NET_XMIT_DROP;
 618 }
 619
 620 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
 621 {
 622         struct ethhdr *eth = eth_hdr(skb);
 623         struct ipvl_addr *addr;
 624         void *lyr3h;
 625         int addr_type;
 626
 627         if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
 628                 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 629                 if (!lyr3h)
 630                         return true;
 631
 632                 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
 633                 if (addr)
 634                         return false;
 635         }
 636
 637         return true;
 638 }
 639
 640 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
 641                                                  struct ipvl_port *port)
 642 {
 643         void *lyr3h;
 644         int addr_type;
 645         struct ipvl_addr *addr;
 646         struct sk_buff *skb = *pskb;
 647         rx_handler_result_t ret = RX_HANDLER_PASS;
 648
 649         lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 650         if (!lyr3h)
 651                 goto out;
 652
 653         addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 654         if (addr)
 655                 ret = ipvlan_rcv_frame(addr, pskb, false);
 656
 657 out:
 658         return ret;
 659 }
 660
 661 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
 662                                                  struct ipvl_port *port)
 663 {
 664         struct sk_buff *skb = *pskb;
 665         struct ethhdr *eth = eth_hdr(skb);
 666         rx_handler_result_t ret = RX_HANDLER_PASS;
 667         void *lyr3h;
 668         int addr_type;
 669
 670         if (is_multicast_ether_addr(eth->h_dest)) {
 671                 if (ipvlan_external_frame(skb, port)) {
 672                         struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
 673
 674                         /* External frames are queued for device local
 675                          * distribution, but a copy is given to master
 676                          * straight away to avoid sending duplicates later
 677                          * when work-queue processes this frame. This is
 678                          * achieved by returning RX_HANDLER_PASS.
 679                          */
 680                         if (nskb) {
 681                                 ipvlan_skb_crossing_ns(nskb, NULL);
 682                                 ipvlan_multicast_enqueue(port, nskb, false);
 683                         }
 684                 }
 685         } else {
 686                 struct ipvl_addr *addr;
 687
 688                 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 689                 if (!lyr3h)
 690                         return ret;
 691
 692                 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 693                 if (addr)
 694                         ret = ipvlan_rcv_frame(addr, pskb, false);
 695         }
 696
 697         return ret;
 698 }
 699
 700 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
 701 {
 702         struct sk_buff *skb = *pskb;
 703         struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
 704
 705         if (!port)
 706                 return RX_HANDLER_PASS;
 707
 708         switch (port->mode) {
 709         case IPVLAN_MODE_L2:
 710                 return ipvlan_handle_mode_l2(pskb, port);
 711         case IPVLAN_MODE_L3:
 712                 return ipvlan_handle_mode_l3(pskb, port);
 713         case IPVLAN_MODE_L3S:
 714                 return RX_HANDLER_PASS;
 715         }
 716
 717         /* Should not reach here */
 718         WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
 719                           port->mode);
 720         kfree_skb(skb);
 721         return RX_HANDLER_CONSUMED;
 722 }
 723
 724 static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
 725                                             struct net_device *dev)
 726 {
 727         struct ipvl_addr *addr = NULL;
 728         struct ipvl_port *port;
 729         void *lyr3h;
 730         int addr_type;
 731
 732         if (!dev || !netif_is_ipvlan_port(dev))
 733                 goto out;
 734
 735         port = ipvlan_port_get_rcu(dev);
 736         if (!port || port->mode != IPVLAN_MODE_L3S)
 737                 goto out;
 738
 739         lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
 740         if (!lyr3h)
 741                 goto out;
 742
 743         addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
 744 out:
 745         return addr;
 746 }
 747
 748 struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
 749                               u16 proto)
 750 {
 751         struct ipvl_addr *addr;
 752         struct net_device *sdev;
 753
 754         addr = ipvlan_skb_to_addr(skb, dev);
 755         if (!addr)
 756                 goto out;
 757
 758         sdev = addr->master->dev;
 759         switch (proto) {
 760         case AF_INET:
 761         {
 762                 int err;
 763                 struct iphdr *ip4h = ip_hdr(skb);
 764
 765                 err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
 766                                            ip4h->tos, sdev);
 767                 if (unlikely(err))
 768                         goto out;
 769                 break;
 770         }
 771         case AF_INET6:
 772         {
 773                 struct dst_entry *dst;
 774                 struct ipv6hdr *ip6h = ipv6_hdr(skb);
 775                 int flags = RT6_LOOKUP_F_HAS_SADDR;
 776                 struct flowi6 fl6 = {
 777                         .flowi6_iif   = sdev->ifindex,
 778                         .daddr        = ip6h->daddr,
 779                         .saddr        = ip6h->saddr,
 780                         .flowlabel    = ip6_flowinfo(ip6h),
 781                         .flowi6_mark  = skb->mark,
 782                         .flowi6_proto = ip6h->nexthdr,
 783                 };
 784
 785                 skb_dst_drop(skb);
 786                 dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
 787                 skb_dst_set(skb, dst);
 788                 break;
 789         }
 790         default:
 791                 break;
 792         }
 793
 794 out:
 795         return skb;
 796 }
 797
 798 unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
 799                              const struct nf_hook_state *state)
 800 {
 801         struct ipvl_addr *addr;
 802         unsigned int len;
 803
 804         addr = ipvlan_skb_to_addr(skb, skb->dev);
 805         if (!addr)
 806                 goto out;
 807
 808         skb->dev = addr->master->dev;
 809         len = skb->len + ETH_HLEN;
 810         ipvlan_count_rx(addr->master, len, true, false);
 811 out:
 812         return NF_ACCEPT;
 813 }