drivers/net/ipvlan/ipvlan_main.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 void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev)
  13 {
  14         ipvlan->dev->mtu = dev->mtu - ipvlan->mtu_adj;
  15 }
  16
  17 void ipvlan_set_port_mode(struct ipvl_port *port, u32 nval)
  18 {
  19         struct ipvl_dev *ipvlan;
  20
  21         if (port->mode != nval) {
  22                 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
  23                         if (nval == IPVLAN_MODE_L3)
  24                                 ipvlan->dev->flags |= IFF_NOARP;
  25                         else
  26                                 ipvlan->dev->flags &= ~IFF_NOARP;
  27                 }
  28                 port->mode = nval;
  29         }
  30 }
  31
  32 static int ipvlan_port_create(struct net_device *dev)
  33 {
  34         struct ipvl_port *port;
  35         int err, idx;
  36
  37         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK) {
  38                 netdev_err(dev, "Master is either lo or non-ether device\n");
  39                 return -EINVAL;
  40         }
  41
  42         if (netif_is_macvlan_port(dev)) {
  43                 netdev_err(dev, "Master is a macvlan port.\n");
  44                 return -EBUSY;
  45         }
  46
  47         port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL);
  48         if (!port)
  49                 return -ENOMEM;
  50
  51         port->dev = dev;
  52         port->mode = IPVLAN_MODE_L3;
  53         INIT_LIST_HEAD(&port->ipvlans);
  54         for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++)
  55                 INIT_HLIST_HEAD(&port->hlhead[idx]);
  56
  57         err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port);
  58         if (err)
  59                 goto err;
  60
  61         dev->priv_flags |= IFF_IPVLAN_MASTER;
  62         return 0;
  63
  64 err:
  65         kfree_rcu(port, rcu);
  66         return err;
  67 }
  68
  69 static void ipvlan_port_destroy(struct net_device *dev)
  70 {
  71         struct ipvl_port *port = ipvlan_port_get_rtnl(dev);
  72
  73         dev->priv_flags &= ~IFF_IPVLAN_MASTER;
  74         netdev_rx_handler_unregister(dev);
  75         kfree_rcu(port, rcu);
  76 }
  77
  78 /* ipvlan network devices have devices nesting below it and are a special
  79  * "super class" of normal network devices; split their locks off into a
  80  * separate class since they always nest.
  81  */
  82 static struct lock_class_key ipvlan_netdev_xmit_lock_key;
  83 static struct lock_class_key ipvlan_netdev_addr_lock_key;
  84
  85 #define IPVLAN_FEATURES \
  86         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
  87          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
  88          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
  89          NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
  90
  91 #define IPVLAN_STATE_MASK \
  92         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
  93
  94 static void ipvlan_set_lockdep_class_one(struct net_device *dev,
  95                                          struct netdev_queue *txq,
  96                                          void *_unused)
  97 {
  98         lockdep_set_class(&txq->_xmit_lock, &ipvlan_netdev_xmit_lock_key);
  99 }
 100
 101 static void ipvlan_set_lockdep_class(struct net_device *dev)
 102 {
 103         lockdep_set_class(&dev->addr_list_lock, &ipvlan_netdev_addr_lock_key);
 104         netdev_for_each_tx_queue(dev, ipvlan_set_lockdep_class_one, NULL);
 105 }
 106
 107 static int ipvlan_init(struct net_device *dev)
 108 {
 109         struct ipvl_dev *ipvlan = netdev_priv(dev);
 110         const struct net_device *phy_dev = ipvlan->phy_dev;
 111
 112         dev->state = (dev->state & ~IPVLAN_STATE_MASK) |
 113                      (phy_dev->state & IPVLAN_STATE_MASK);
 114         dev->features = phy_dev->features & IPVLAN_FEATURES;
 115         dev->features |= NETIF_F_LLTX;
 116         dev->gso_max_size = phy_dev->gso_max_size;
 117         dev->iflink = phy_dev->ifindex;
 118         dev->hard_header_len = phy_dev->hard_header_len;
 119
 120         ipvlan_set_lockdep_class(dev);
 121
 122         ipvlan->pcpu_stats = alloc_percpu(struct ipvl_pcpu_stats);
 123         if (!ipvlan->pcpu_stats)
 124                 return -ENOMEM;
 125
 126         return 0;
 127 }
 128
 129 static void ipvlan_uninit(struct net_device *dev)
 130 {
 131         struct ipvl_dev *ipvlan = netdev_priv(dev);
 132         struct ipvl_port *port = ipvlan->port;
 133
 134         free_percpu(ipvlan->pcpu_stats);
 135
 136         port->count -= 1;
 137         if (!port->count)
 138                 ipvlan_port_destroy(port->dev);
 139 }
 140
 141 static int ipvlan_open(struct net_device *dev)
 142 {
 143         struct ipvl_dev *ipvlan = netdev_priv(dev);
 144         struct net_device *phy_dev = ipvlan->phy_dev;
 145         struct ipvl_addr *addr;
 146
 147         if (ipvlan->port->mode == IPVLAN_MODE_L3)
 148                 dev->flags |= IFF_NOARP;
 149         else
 150                 dev->flags &= ~IFF_NOARP;
 151
 152         if (ipvlan->ipv6cnt > 0 || ipvlan->ipv4cnt > 0) {
 153                 list_for_each_entry(addr, &ipvlan->addrs, anode)
 154                         ipvlan_ht_addr_add(ipvlan, addr);
 155         }
 156         return dev_uc_add(phy_dev, phy_dev->dev_addr);
 157 }
 158
 159 static int ipvlan_stop(struct net_device *dev)
 160 {
 161         struct ipvl_dev *ipvlan = netdev_priv(dev);
 162         struct net_device *phy_dev = ipvlan->phy_dev;
 163         struct ipvl_addr *addr;
 164
 165         dev_uc_unsync(phy_dev, dev);
 166         dev_mc_unsync(phy_dev, dev);
 167
 168         dev_uc_del(phy_dev, phy_dev->dev_addr);
 169
 170         if (ipvlan->ipv6cnt > 0 || ipvlan->ipv4cnt > 0) {
 171                 list_for_each_entry(addr, &ipvlan->addrs, anode)
 172                         ipvlan_ht_addr_del(addr, !dev->dismantle);
 173         }
 174         return 0;
 175 }
 176
 177 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb,
 178                                      struct net_device *dev)
 179 {
 180         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 181         int skblen = skb->len;
 182         int ret;
 183
 184         ret = ipvlan_queue_xmit(skb, dev);
 185         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
 186                 struct ipvl_pcpu_stats *pcptr;
 187
 188                 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
 189
 190                 u64_stats_update_begin(&pcptr->syncp);
 191                 pcptr->tx_pkts++;
 192                 pcptr->tx_bytes += skblen;
 193                 u64_stats_update_end(&pcptr->syncp);
 194         } else {
 195                 this_cpu_inc(ipvlan->pcpu_stats->tx_drps);
 196         }
 197         return ret;
 198 }
 199
 200 static netdev_features_t ipvlan_fix_features(struct net_device *dev,
 201                                              netdev_features_t features)
 202 {
 203         struct ipvl_dev *ipvlan = netdev_priv(dev);
 204
 205         return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES);
 206 }
 207
 208 static void ipvlan_change_rx_flags(struct net_device *dev, int change)
 209 {
 210         struct ipvl_dev *ipvlan = netdev_priv(dev);
 211         struct net_device *phy_dev = ipvlan->phy_dev;
 212
 213         if (change & IFF_ALLMULTI)
 214                 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1);
 215 }
 216
 217 static void ipvlan_set_broadcast_mac_filter(struct ipvl_dev *ipvlan, bool set)
 218 {
 219         struct net_device *dev = ipvlan->dev;
 220         unsigned int hashbit = ipvlan_mac_hash(dev->broadcast);
 221
 222         if (set && !test_bit(hashbit, ipvlan->mac_filters))
 223                 __set_bit(hashbit, ipvlan->mac_filters);
 224         else if (!set && test_bit(hashbit, ipvlan->mac_filters))
 225                 __clear_bit(hashbit, ipvlan->mac_filters);
 226 }
 227
 228 static void ipvlan_set_multicast_mac_filter(struct net_device *dev)
 229 {
 230         struct ipvl_dev *ipvlan = netdev_priv(dev);
 231
 232         if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
 233                 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE);
 234         } else {
 235                 struct netdev_hw_addr *ha;
 236                 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE);
 237
 238                 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE);
 239                 netdev_for_each_mc_addr(ha, dev)
 240                         __set_bit(ipvlan_mac_hash(ha->addr), mc_filters);
 241
 242                 bitmap_copy(ipvlan->mac_filters, mc_filters,
 243                             IPVLAN_MAC_FILTER_SIZE);
 244         }
 245         dev_uc_sync(ipvlan->phy_dev, dev);
 246         dev_mc_sync(ipvlan->phy_dev, dev);
 247 }
 248
 249 static struct rtnl_link_stats64 *ipvlan_get_stats64(struct net_device *dev,
 250                                                     struct rtnl_link_stats64 *s)
 251 {
 252         struct ipvl_dev *ipvlan = netdev_priv(dev);
 253
 254         if (ipvlan->pcpu_stats) {
 255                 struct ipvl_pcpu_stats *pcptr;
 256                 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes;
 257                 u32 rx_errs = 0, tx_drps = 0;
 258                 u32 strt;
 259                 int idx;
 260
 261                 for_each_possible_cpu(idx) {
 262                         pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx);
 263                         do {
 264                                 strt= u64_stats_fetch_begin_irq(&pcptr->syncp);
 265                                 rx_pkts = pcptr->rx_pkts;
 266                                 rx_bytes = pcptr->rx_bytes;
 267                                 rx_mcast = pcptr->rx_mcast;
 268                                 tx_pkts = pcptr->tx_pkts;
 269                                 tx_bytes = pcptr->tx_bytes;
 270                         } while (u64_stats_fetch_retry_irq(&pcptr->syncp,
 271                                                            strt));
 272
 273                         s->rx_packets += rx_pkts;
 274                         s->rx_bytes += rx_bytes;
 275                         s->multicast += rx_mcast;
 276                         s->tx_packets += tx_pkts;
 277                         s->tx_bytes += tx_bytes;
 278
 279                         /* u32 values are updated without syncp protection. */
 280                         rx_errs += pcptr->rx_errs;
 281                         tx_drps += pcptr->tx_drps;
 282                 }
 283                 s->rx_errors = rx_errs;
 284                 s->rx_dropped = rx_errs;
 285                 s->tx_dropped = tx_drps;
 286         }
 287         return s;
 288 }
 289
 290 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid)
 291 {
 292         struct ipvl_dev *ipvlan = netdev_priv(dev);
 293         struct net_device *phy_dev = ipvlan->phy_dev;
 294
 295         return vlan_vid_add(phy_dev, proto, vid);
 296 }
 297
 298 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto,
 299                                    u16 vid)
 300 {
 301         struct ipvl_dev *ipvlan = netdev_priv(dev);
 302         struct net_device *phy_dev = ipvlan->phy_dev;
 303
 304         vlan_vid_del(phy_dev, proto, vid);
 305         return 0;
 306 }
 307
 308 static const struct net_device_ops ipvlan_netdev_ops = {
 309         .ndo_init               = ipvlan_init,
 310         .ndo_uninit             = ipvlan_uninit,
 311         .ndo_open               = ipvlan_open,
 312         .ndo_stop               = ipvlan_stop,
 313         .ndo_start_xmit         = ipvlan_start_xmit,
 314         .ndo_fix_features       = ipvlan_fix_features,
 315         .ndo_change_rx_flags    = ipvlan_change_rx_flags,
 316         .ndo_set_rx_mode        = ipvlan_set_multicast_mac_filter,
 317         .ndo_get_stats64        = ipvlan_get_stats64,
 318         .ndo_vlan_rx_add_vid    = ipvlan_vlan_rx_add_vid,
 319         .ndo_vlan_rx_kill_vid   = ipvlan_vlan_rx_kill_vid,
 320 };
 321
 322 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
 323                               unsigned short type, const void *daddr,
 324                               const void *saddr, unsigned len)
 325 {
 326         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 327         struct net_device *phy_dev = ipvlan->phy_dev;
 328
 329         /* TODO Probably use a different field than dev_addr so that the
 330          * mac-address on the virtual device is portable and can be carried
 331          * while the packets use the mac-addr on the physical device.
 332          */
 333         return dev_hard_header(skb, phy_dev, type, daddr,
 334                                saddr ? : dev->dev_addr, len);
 335 }
 336
 337 static const struct header_ops ipvlan_header_ops = {
 338         .create         = ipvlan_hard_header,
 339         .parse          = eth_header_parse,
 340         .cache          = eth_header_cache,
 341         .cache_update   = eth_header_cache_update,
 342 };
 343
 344 static int ipvlan_ethtool_get_settings(struct net_device *dev,
 345                                        struct ethtool_cmd *cmd)
 346 {
 347         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 348
 349         return __ethtool_get_settings(ipvlan->phy_dev, cmd);
 350 }
 351
 352 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev,
 353                                        struct ethtool_drvinfo *drvinfo)
 354 {
 355         strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver));
 356         strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version));
 357 }
 358
 359 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev)
 360 {
 361         const struct ipvl_dev *ipvlan = netdev_priv(dev);
 362
 363         return ipvlan->msg_enable;
 364 }
 365
 366 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value)
 367 {
 368         struct ipvl_dev *ipvlan = netdev_priv(dev);
 369
 370         ipvlan->msg_enable = value;
 371 }
 372
 373 static const struct ethtool_ops ipvlan_ethtool_ops = {
 374         .get_link       = ethtool_op_get_link,
 375         .get_settings   = ipvlan_ethtool_get_settings,
 376         .get_drvinfo    = ipvlan_ethtool_get_drvinfo,
 377         .get_msglevel   = ipvlan_ethtool_get_msglevel,
 378         .set_msglevel   = ipvlan_ethtool_set_msglevel,
 379 };
 380
 381 static int ipvlan_nl_changelink(struct net_device *dev,
 382                                 struct nlattr *tb[], struct nlattr *data[])
 383 {
 384         struct ipvl_dev *ipvlan = netdev_priv(dev);
 385         struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
 386
 387         if (data && data[IFLA_IPVLAN_MODE]) {
 388                 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
 389
 390                 ipvlan_set_port_mode(port, nmode);
 391         }
 392         return 0;
 393 }
 394
 395 static size_t ipvlan_nl_getsize(const struct net_device *dev)
 396 {
 397         return (0
 398                 + nla_total_size(2) /* IFLA_IPVLAN_MODE */
 399                 );
 400 }
 401
 402 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[])
 403 {
 404         if (data && data[IFLA_IPVLAN_MODE]) {
 405                 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
 406
 407                 if (mode < IPVLAN_MODE_L2 || mode >= IPVLAN_MODE_MAX)
 408                         return -EINVAL;
 409         }
 410         return 0;
 411 }
 412
 413 static int ipvlan_nl_fillinfo(struct sk_buff *skb,
 414                               const struct net_device *dev)
 415 {
 416         struct ipvl_dev *ipvlan = netdev_priv(dev);
 417         struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev);
 418         int ret = -EINVAL;
 419
 420         if (!port)
 421                 goto err;
 422
 423         ret = -EMSGSIZE;
 424         if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode))
 425                 goto err;
 426
 427         return 0;
 428
 429 err:
 430         return ret;
 431 }
 432
 433 static int ipvlan_link_new(struct net *src_net, struct net_device *dev,
 434                            struct nlattr *tb[], struct nlattr *data[])
 435 {
 436         struct ipvl_dev *ipvlan = netdev_priv(dev);
 437         struct ipvl_port *port;
 438         struct net_device *phy_dev;
 439         int err;
 440
 441         if (!tb[IFLA_LINK])
 442                 return -EINVAL;
 443
 444         phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
 445         if (!phy_dev)
 446                 return -ENODEV;
 447
 448         if (netif_is_ipvlan(phy_dev)) {
 449                 struct ipvl_dev *tmp = netdev_priv(phy_dev);
 450
 451                 phy_dev = tmp->phy_dev;
 452         } else if (!netif_is_ipvlan_port(phy_dev)) {
 453                 err = ipvlan_port_create(phy_dev);
 454                 if (err < 0)
 455                         return err;
 456         }
 457
 458         port = ipvlan_port_get_rtnl(phy_dev);
 459         if (data && data[IFLA_IPVLAN_MODE])
 460                 port->mode = nla_get_u16(data[IFLA_IPVLAN_MODE]);
 461
 462         ipvlan->phy_dev = phy_dev;
 463         ipvlan->dev = dev;
 464         ipvlan->port = port;
 465         ipvlan->sfeatures = IPVLAN_FEATURES;
 466         INIT_LIST_HEAD(&ipvlan->addrs);
 467         ipvlan->ipv4cnt = 0;
 468         ipvlan->ipv6cnt = 0;
 469
 470         /* TODO Probably put random address here to be presented to the
 471          * world but keep using the physical-dev address for the outgoing
 472          * packets.
 473          */
 474         memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN);
 475
 476         dev->priv_flags |= IFF_IPVLAN_SLAVE;
 477
 478         port->count += 1;
 479         err = register_netdevice(dev);
 480         if (err < 0)
 481                 goto ipvlan_destroy_port;
 482
 483         err = netdev_upper_dev_link(phy_dev, dev);
 484         if (err)
 485                 goto ipvlan_destroy_port;
 486
 487         list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans);
 488         netif_stacked_transfer_operstate(phy_dev, dev);
 489         return 0;
 490
 491 ipvlan_destroy_port:
 492         port->count -= 1;
 493         if (!port->count)
 494                 ipvlan_port_destroy(phy_dev);
 495
 496         return err;
 497 }
 498
 499 static void ipvlan_link_delete(struct net_device *dev, struct list_head *head)
 500 {
 501         struct ipvl_dev *ipvlan = netdev_priv(dev);
 502         struct ipvl_addr *addr, *next;
 503
 504         if (ipvlan->ipv6cnt > 0 || ipvlan->ipv4cnt > 0) {
 505                 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) {
 506                         ipvlan_ht_addr_del(addr, !dev->dismantle);
 507                         list_del_rcu(&addr->anode);
 508                 }
 509         }
 510         list_del_rcu(&ipvlan->pnode);
 511         unregister_netdevice_queue(dev, head);
 512         netdev_upper_dev_unlink(ipvlan->phy_dev, dev);
 513 }
 514
 515 static void ipvlan_link_setup(struct net_device *dev)
 516 {
 517         ether_setup(dev);
 518
 519         dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
 520         dev->priv_flags |= IFF_UNICAST_FLT;
 521         dev->netdev_ops = &ipvlan_netdev_ops;
 522         dev->destructor = free_netdev;
 523         dev->header_ops = &ipvlan_header_ops;
 524         dev->ethtool_ops = &ipvlan_ethtool_ops;
 525         dev->tx_queue_len = 0;
 526 }
 527
 528 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] =
 529 {
 530         [IFLA_IPVLAN_MODE] = { .type = NLA_U16 },
 531 };
 532
 533 static struct rtnl_link_ops ipvlan_link_ops = {
 534         .kind           = "ipvlan",
 535         .priv_size      = sizeof(struct ipvl_dev),
 536
 537         .get_size       = ipvlan_nl_getsize,
 538         .policy         = ipvlan_nl_policy,
 539         .validate       = ipvlan_nl_validate,
 540         .fill_info      = ipvlan_nl_fillinfo,
 541         .changelink     = ipvlan_nl_changelink,
 542         .maxtype        = IFLA_IPVLAN_MAX,
 543
 544         .setup          = ipvlan_link_setup,
 545         .newlink        = ipvlan_link_new,
 546         .dellink        = ipvlan_link_delete,
 547 };
 548
 549 static int ipvlan_link_register(struct rtnl_link_ops *ops)
 550 {
 551         return rtnl_link_register(ops);
 552 }
 553
 554 static int ipvlan_device_event(struct notifier_block *unused,
 555                                unsigned long event, void *ptr)
 556 {
 557         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 558         struct ipvl_dev *ipvlan, *next;
 559         struct ipvl_port *port;
 560         LIST_HEAD(lst_kill);
 561
 562         if (!netif_is_ipvlan_port(dev))
 563                 return NOTIFY_DONE;
 564
 565         port = ipvlan_port_get_rtnl(dev);
 566
 567         switch (event) {
 568         case NETDEV_CHANGE:
 569                 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
 570                         netif_stacked_transfer_operstate(ipvlan->phy_dev,
 571                                                          ipvlan->dev);
 572                 break;
 573
 574         case NETDEV_UNREGISTER:
 575                 if (dev->reg_state != NETREG_UNREGISTERING)
 576                         break;
 577
 578                 list_for_each_entry_safe(ipvlan, next, &port->ipvlans,
 579                                          pnode)
 580                         ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev,
 581                                                             &lst_kill);
 582                 unregister_netdevice_many(&lst_kill);
 583                 break;
 584
 585         case NETDEV_FEAT_CHANGE:
 586                 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
 587                         ipvlan->dev->features = dev->features & IPVLAN_FEATURES;
 588                         ipvlan->dev->gso_max_size = dev->gso_max_size;
 589                         netdev_features_change(ipvlan->dev);
 590                 }
 591                 break;
 592
 593         case NETDEV_CHANGEMTU:
 594                 list_for_each_entry(ipvlan, &port->ipvlans, pnode)
 595                         ipvlan_adjust_mtu(ipvlan, dev);
 596                 break;
 597
 598         case NETDEV_PRE_TYPE_CHANGE:
 599                 /* Forbid underlying device to change its type. */
 600                 return NOTIFY_BAD;
 601         }
 602         return NOTIFY_DONE;
 603 }
 604
 605 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
 606 {
 607         struct ipvl_addr *addr;
 608
 609         if (ipvlan_addr_busy(ipvlan, ip6_addr, true)) {
 610                 netif_err(ipvlan, ifup, ipvlan->dev,
 611                           "Failed to add IPv6=%pI6c addr for %s intf\n",
 612                           ip6_addr, ipvlan->dev->name);
 613                 return -EINVAL;
 614         }
 615         addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC);
 616         if (!addr)
 617                 return -ENOMEM;
 618
 619         addr->master = ipvlan;
 620         memcpy(&addr->ip6addr, ip6_addr, sizeof(struct in6_addr));
 621         addr->atype = IPVL_IPV6;
 622         list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
 623         ipvlan->ipv6cnt++;
 624         ipvlan_ht_addr_add(ipvlan, addr);
 625
 626         return 0;
 627 }
 628
 629 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr)
 630 {
 631         struct ipvl_addr *addr;
 632
 633         addr = ipvlan_ht_addr_lookup(ipvlan->port, ip6_addr, true);
 634         if (!addr)
 635                 return;
 636
 637         ipvlan_ht_addr_del(addr, true);
 638         list_del_rcu(&addr->anode);
 639         ipvlan->ipv6cnt--;
 640         WARN_ON(ipvlan->ipv6cnt < 0);
 641         kfree_rcu(addr, rcu);
 642
 643         return;
 644 }
 645
 646 static int ipvlan_addr6_event(struct notifier_block *unused,
 647                               unsigned long event, void *ptr)
 648 {
 649         struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr;
 650         struct net_device *dev = (struct net_device *)if6->idev->dev;
 651         struct ipvl_dev *ipvlan = netdev_priv(dev);
 652
 653         if (!netif_is_ipvlan(dev))
 654                 return NOTIFY_DONE;
 655
 656         if (!ipvlan || !ipvlan->port)
 657                 return NOTIFY_DONE;
 658
 659         switch (event) {
 660         case NETDEV_UP:
 661                 if (ipvlan_add_addr6(ipvlan, &if6->addr))
 662                         return NOTIFY_BAD;
 663                 break;
 664
 665         case NETDEV_DOWN:
 666                 ipvlan_del_addr6(ipvlan, &if6->addr);
 667                 break;
 668         }
 669
 670         return NOTIFY_OK;
 671 }
 672
 673 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
 674 {
 675         struct ipvl_addr *addr;
 676
 677         if (ipvlan_addr_busy(ipvlan, ip4_addr, false)) {
 678                 netif_err(ipvlan, ifup, ipvlan->dev,
 679                           "Failed to add IPv4=%pI4 on %s intf.\n",
 680                           ip4_addr, ipvlan->dev->name);
 681                 return -EINVAL;
 682         }
 683         addr = kzalloc(sizeof(struct ipvl_addr), GFP_KERNEL);
 684         if (!addr)
 685                 return -ENOMEM;
 686
 687         addr->master = ipvlan;
 688         memcpy(&addr->ip4addr, ip4_addr, sizeof(struct in_addr));
 689         addr->atype = IPVL_IPV4;
 690         list_add_tail_rcu(&addr->anode, &ipvlan->addrs);
 691         ipvlan->ipv4cnt++;
 692         ipvlan_ht_addr_add(ipvlan, addr);
 693         ipvlan_set_broadcast_mac_filter(ipvlan, true);
 694
 695         return 0;
 696 }
 697
 698 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr)
 699 {
 700         struct ipvl_addr *addr;
 701
 702         addr = ipvlan_ht_addr_lookup(ipvlan->port, ip4_addr, false);
 703         if (!addr)
 704                 return;
 705
 706         ipvlan_ht_addr_del(addr, true);
 707         list_del_rcu(&addr->anode);
 708         ipvlan->ipv4cnt--;
 709         WARN_ON(ipvlan->ipv4cnt < 0);
 710         if (!ipvlan->ipv4cnt)
 711             ipvlan_set_broadcast_mac_filter(ipvlan, false);
 712         kfree_rcu(addr, rcu);
 713
 714         return;
 715 }
 716
 717 static int ipvlan_addr4_event(struct notifier_block *unused,
 718                               unsigned long event, void *ptr)
 719 {
 720         struct in_ifaddr *if4 = (struct in_ifaddr *)ptr;
 721         struct net_device *dev = (struct net_device *)if4->ifa_dev->dev;
 722         struct ipvl_dev *ipvlan = netdev_priv(dev);
 723         struct in_addr ip4_addr;
 724
 725         if (!netif_is_ipvlan(dev))
 726                 return NOTIFY_DONE;
 727
 728         if (!ipvlan || !ipvlan->port)
 729                 return NOTIFY_DONE;
 730
 731         switch (event) {
 732         case NETDEV_UP:
 733                 ip4_addr.s_addr = if4->ifa_address;
 734                 if (ipvlan_add_addr4(ipvlan, &ip4_addr))
 735                         return NOTIFY_BAD;
 736                 break;
 737
 738         case NETDEV_DOWN:
 739                 ip4_addr.s_addr = if4->ifa_address;
 740                 ipvlan_del_addr4(ipvlan, &ip4_addr);
 741                 break;
 742         }
 743
 744         return NOTIFY_OK;
 745 }
 746
 747 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = {
 748         .notifier_call = ipvlan_addr4_event,
 749 };
 750
 751 static struct notifier_block ipvlan_notifier_block __read_mostly = {
 752         .notifier_call = ipvlan_device_event,
 753 };
 754
 755 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = {
 756         .notifier_call = ipvlan_addr6_event,
 757 };
 758
 759 static int __init ipvlan_init_module(void)
 760 {
 761         int err;
 762
 763         ipvlan_init_secret();
 764         register_netdevice_notifier(&ipvlan_notifier_block);
 765         register_inet6addr_notifier(&ipvlan_addr6_notifier_block);
 766         register_inetaddr_notifier(&ipvlan_addr4_notifier_block);
 767
 768         err = ipvlan_link_register(&ipvlan_link_ops);
 769         if (err < 0)
 770                 goto error;
 771
 772         return 0;
 773 error:
 774         unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
 775         unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
 776         unregister_netdevice_notifier(&ipvlan_notifier_block);
 777         return err;
 778 }
 779
 780 static void __exit ipvlan_cleanup_module(void)
 781 {
 782         rtnl_link_unregister(&ipvlan_link_ops);
 783         unregister_netdevice_notifier(&ipvlan_notifier_block);
 784         unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block);
 785         unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block);
 786 }
 787
 788 module_init(ipvlan_init_module);
 789 module_exit(ipvlan_cleanup_module);
 790
 791 MODULE_LICENSE("GPL");
 792 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>");
 793 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs");
 794 MODULE_ALIAS_RTNL_LINK("ipvlan");