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 unsigned int ipvlan_netid __read_mostly; 13 14 struct ipvlan_netns { 15 unsigned int ipvl_nf_hook_refcnt; 16 }; 17 18 static const struct nf_hook_ops ipvl_nfops[] = { 19 { 20 .hook = ipvlan_nf_input, 21 .pf = NFPROTO_IPV4, 22 .hooknum = NF_INET_LOCAL_IN, 23 .priority = INT_MAX, 24 }, 25 #if IS_ENABLED(CONFIG_IPV6) 26 { 27 .hook = ipvlan_nf_input, 28 .pf = NFPROTO_IPV6, 29 .hooknum = NF_INET_LOCAL_IN, 30 .priority = INT_MAX, 31 }, 32 #endif 33 }; 34 35 static const struct l3mdev_ops ipvl_l3mdev_ops = { 36 .l3mdev_l3_rcv = ipvlan_l3_rcv, 37 }; 38 39 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev) 40 { 41 ipvlan->dev->mtu = dev->mtu; 42 } 43 44 static int ipvlan_register_nf_hook(struct net *net) 45 { 46 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid); 47 int err = 0; 48 49 if (!vnet->ipvl_nf_hook_refcnt) { 50 err = nf_register_net_hooks(net, ipvl_nfops, 51 ARRAY_SIZE(ipvl_nfops)); 52 if (!err) 53 vnet->ipvl_nf_hook_refcnt = 1; 54 } else { 55 vnet->ipvl_nf_hook_refcnt++; 56 } 57 58 return err; 59 } 60 61 static void ipvlan_unregister_nf_hook(struct net *net) 62 { 63 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid); 64 65 if (WARN_ON(!vnet->ipvl_nf_hook_refcnt)) 66 return; 67 68 vnet->ipvl_nf_hook_refcnt--; 69 if (!vnet->ipvl_nf_hook_refcnt) 70 nf_unregister_net_hooks(net, ipvl_nfops, 71 ARRAY_SIZE(ipvl_nfops)); 72 } 73 74 static int ipvlan_set_port_mode(struct ipvl_port *port, u16 nval) 75 { 76 struct ipvl_dev *ipvlan; 77 struct net_device *mdev = port->dev; 78 int err = 0; 79 80 ASSERT_RTNL(); 81 if (port->mode != nval) { 82 if (nval == IPVLAN_MODE_L3S) { 83 /* New mode is L3S */ 84 err = ipvlan_register_nf_hook(read_pnet(&port->pnet)); 85 if (!err) { 86 mdev->l3mdev_ops = &ipvl_l3mdev_ops; 87 mdev->priv_flags |= IFF_L3MDEV_MASTER; 88 } else 89 return err; 90 } else if (port->mode == IPVLAN_MODE_L3S) { 91 /* Old mode was L3S */ 92 mdev->priv_flags &= ~IFF_L3MDEV_MASTER; 93 ipvlan_unregister_nf_hook(read_pnet(&port->pnet)); 94 mdev->l3mdev_ops = NULL; 95 } 96 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 97 if (nval == IPVLAN_MODE_L3 || nval == IPVLAN_MODE_L3S) 98 ipvlan->dev->flags |= IFF_NOARP; 99 else 100 ipvlan->dev->flags &= ~IFF_NOARP; 101 } 102 port->mode = nval; 103 } 104 return err; 105 } 106 107 static int ipvlan_port_create(struct net_device *dev) 108 { 109 struct ipvl_port *port; 110 int err, idx; 111 112 port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL); 113 if (!port) 114 return -ENOMEM; 115 116 write_pnet(&port->pnet, dev_net(dev)); 117 port->dev = dev; 118 port->mode = IPVLAN_MODE_L3; 119 INIT_LIST_HEAD(&port->ipvlans); 120 for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++) 121 INIT_HLIST_HEAD(&port->hlhead[idx]); 122 123 skb_queue_head_init(&port->backlog); 124 INIT_WORK(&port->wq, ipvlan_process_multicast); 125 ida_init(&port->ida); 126 port->dev_id_start = 1; 127 128 err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port); 129 if (err) 130 goto err; 131 132 return 0; 133 134 err: 135 kfree(port); 136 return err; 137 } 138 139 static void ipvlan_port_destroy(struct net_device *dev) 140 { 141 struct ipvl_port *port = ipvlan_port_get_rtnl(dev); 142 struct sk_buff *skb; 143 144 if (port->mode == IPVLAN_MODE_L3S) { 145 dev->priv_flags &= ~IFF_L3MDEV_MASTER; 146 ipvlan_unregister_nf_hook(dev_net(dev)); 147 dev->l3mdev_ops = NULL; 148 } 149 netdev_rx_handler_unregister(dev); 150 cancel_work_sync(&port->wq); 151 while ((skb = __skb_dequeue(&port->backlog)) != NULL) { 152 if (skb->dev) 153 dev_put(skb->dev); 154 kfree_skb(skb); 155 } 156 ida_destroy(&port->ida); 157 kfree(port); 158 } 159 160 #define IPVLAN_FEATURES \ 161 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 162 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_GSO_ROBUST | \ 163 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \ 164 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 165 166 #define IPVLAN_STATE_MASK \ 167 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 168 169 static int ipvlan_init(struct net_device *dev) 170 { 171 struct ipvl_dev *ipvlan = netdev_priv(dev); 172 struct net_device *phy_dev = ipvlan->phy_dev; 173 struct ipvl_port *port; 174 int err; 175 176 dev->state = (dev->state & ~IPVLAN_STATE_MASK) | 177 (phy_dev->state & IPVLAN_STATE_MASK); 178 dev->features = phy_dev->features & IPVLAN_FEATURES; 179 dev->features |= NETIF_F_LLTX | NETIF_F_VLAN_CHALLENGED; 180 dev->gso_max_size = phy_dev->gso_max_size; 181 dev->gso_max_segs = phy_dev->gso_max_segs; 182 dev->hard_header_len = phy_dev->hard_header_len; 183 184 netdev_lockdep_set_classes(dev); 185 186 ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats); 187 if (!ipvlan->pcpu_stats) 188 return -ENOMEM; 189 190 if (!netif_is_ipvlan_port(phy_dev)) { 191 err = ipvlan_port_create(phy_dev); 192 if (err < 0) { 193 free_percpu(ipvlan->pcpu_stats); 194 return err; 195 } 196 } 197 port = ipvlan_port_get_rtnl(phy_dev); 198 port->count += 1; 199 return 0; 200 } 201 202 static void ipvlan_uninit(struct net_device *dev) 203 { 204 struct ipvl_dev *ipvlan = netdev_priv(dev); 205 struct net_device *phy_dev = ipvlan->phy_dev; 206 struct ipvl_port *port; 207 208 free_percpu(ipvlan->pcpu_stats); 209 210 port = ipvlan_port_get_rtnl(phy_dev); 211 port->count -= 1; 212 if (!port->count) 213 ipvlan_port_destroy(port->dev); 214 } 215 216 static int ipvlan_open(struct net_device *dev) 217 { 218 struct ipvl_dev *ipvlan = netdev_priv(dev); 219 struct net_device *phy_dev = ipvlan->phy_dev; 220 struct ipvl_addr *addr; 221 222 if (ipvlan->port->mode == IPVLAN_MODE_L3 || 223 ipvlan->port->mode == IPVLAN_MODE_L3S) 224 dev->flags |= IFF_NOARP; 225 else 226 dev->flags &= ~IFF_NOARP; 227 228 rcu_read_lock(); 229 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) 230 ipvlan_ht_addr_add(ipvlan, addr); 231 rcu_read_unlock(); 232 233 return dev_uc_add(phy_dev, phy_dev->dev_addr); 234 } 235 236 static int ipvlan_stop(struct net_device *dev) 237 { 238 struct ipvl_dev *ipvlan = netdev_priv(dev); 239 struct net_device *phy_dev = ipvlan->phy_dev; 240 struct ipvl_addr *addr; 241 242 dev_uc_unsync(phy_dev, dev); 243 dev_mc_unsync(phy_dev, dev); 244 245 dev_uc_del(phy_dev, phy_dev->dev_addr); 246 247 rcu_read_lock(); 248 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) 249 ipvlan_ht_addr_del(addr); 250 rcu_read_unlock(); 251 252 return 0; 253 } 254 255 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb, 256 struct net_device *dev) 257 { 258 const struct ipvl_dev *ipvlan = netdev_priv(dev); 259 int skblen = skb->len; 260 int ret; 261 262 ret = ipvlan_queue_xmit(skb, dev); 263 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 264 struct ipvl_pcpu_stats *pcptr; 265 266 pcptr = this_cpu_ptr(ipvlan->pcpu_stats); 267 268 u64_stats_update_begin(&pcptr->syncp); 269 pcptr->tx_pkts++; 270 pcptr->tx_bytes += skblen; 271 u64_stats_update_end(&pcptr->syncp); 272 } else { 273 this_cpu_inc(ipvlan->pcpu_stats->tx_drps); 274 } 275 return ret; 276 } 277 278 static netdev_features_t ipvlan_fix_features(struct net_device *dev, 279 netdev_features_t features) 280 { 281 struct ipvl_dev *ipvlan = netdev_priv(dev); 282 283 return features & (ipvlan->sfeatures | ~IPVLAN_FEATURES); 284 } 285 286 static void ipvlan_change_rx_flags(struct net_device *dev, int change) 287 { 288 struct ipvl_dev *ipvlan = netdev_priv(dev); 289 struct net_device *phy_dev = ipvlan->phy_dev; 290 291 if (change & IFF_ALLMULTI) 292 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1); 293 } 294 295 static void ipvlan_set_multicast_mac_filter(struct net_device *dev) 296 { 297 struct ipvl_dev *ipvlan = netdev_priv(dev); 298 299 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 300 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE); 301 } else { 302 struct netdev_hw_addr *ha; 303 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE); 304 305 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE); 306 netdev_for_each_mc_addr(ha, dev) 307 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters); 308 309 /* Turn-on broadcast bit irrespective of address family, 310 * since broadcast is deferred to a work-queue, hence no 311 * impact on fast-path processing. 312 */ 313 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters); 314 315 bitmap_copy(ipvlan->mac_filters, mc_filters, 316 IPVLAN_MAC_FILTER_SIZE); 317 } 318 dev_uc_sync(ipvlan->phy_dev, dev); 319 dev_mc_sync(ipvlan->phy_dev, dev); 320 } 321 322 static void ipvlan_get_stats64(struct net_device *dev, 323 struct rtnl_link_stats64 *s) 324 { 325 struct ipvl_dev *ipvlan = netdev_priv(dev); 326 327 if (ipvlan->pcpu_stats) { 328 struct ipvl_pcpu_stats *pcptr; 329 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes; 330 u32 rx_errs = 0, tx_drps = 0; 331 u32 strt; 332 int idx; 333 334 for_each_possible_cpu(idx) { 335 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx); 336 do { 337 strt= u64_stats_fetch_begin_irq(&pcptr->syncp); 338 rx_pkts = pcptr->rx_pkts; 339 rx_bytes = pcptr->rx_bytes; 340 rx_mcast = pcptr->rx_mcast; 341 tx_pkts = pcptr->tx_pkts; 342 tx_bytes = pcptr->tx_bytes; 343 } while (u64_stats_fetch_retry_irq(&pcptr->syncp, 344 strt)); 345 346 s->rx_packets += rx_pkts; 347 s->rx_bytes += rx_bytes; 348 s->multicast += rx_mcast; 349 s->tx_packets += tx_pkts; 350 s->tx_bytes += tx_bytes; 351 352 /* u32 values are updated without syncp protection. */ 353 rx_errs += pcptr->rx_errs; 354 tx_drps += pcptr->tx_drps; 355 } 356 s->rx_errors = rx_errs; 357 s->rx_dropped = rx_errs; 358 s->tx_dropped = tx_drps; 359 } 360 } 361 362 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid) 363 { 364 struct ipvl_dev *ipvlan = netdev_priv(dev); 365 struct net_device *phy_dev = ipvlan->phy_dev; 366 367 return vlan_vid_add(phy_dev, proto, vid); 368 } 369 370 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto, 371 u16 vid) 372 { 373 struct ipvl_dev *ipvlan = netdev_priv(dev); 374 struct net_device *phy_dev = ipvlan->phy_dev; 375 376 vlan_vid_del(phy_dev, proto, vid); 377 return 0; 378 } 379 380 static int ipvlan_get_iflink(const struct net_device *dev) 381 { 382 struct ipvl_dev *ipvlan = netdev_priv(dev); 383 384 return ipvlan->phy_dev->ifindex; 385 } 386 387 static const struct net_device_ops ipvlan_netdev_ops = { 388 .ndo_init = ipvlan_init, 389 .ndo_uninit = ipvlan_uninit, 390 .ndo_open = ipvlan_open, 391 .ndo_stop = ipvlan_stop, 392 .ndo_start_xmit = ipvlan_start_xmit, 393 .ndo_fix_features = ipvlan_fix_features, 394 .ndo_change_rx_flags = ipvlan_change_rx_flags, 395 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter, 396 .ndo_get_stats64 = ipvlan_get_stats64, 397 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid, 398 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid, 399 .ndo_get_iflink = ipvlan_get_iflink, 400 }; 401 402 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 403 unsigned short type, const void *daddr, 404 const void *saddr, unsigned len) 405 { 406 const struct ipvl_dev *ipvlan = netdev_priv(dev); 407 struct net_device *phy_dev = ipvlan->phy_dev; 408 409 /* TODO Probably use a different field than dev_addr so that the 410 * mac-address on the virtual device is portable and can be carried 411 * while the packets use the mac-addr on the physical device. 412 */ 413 return dev_hard_header(skb, phy_dev, type, daddr, 414 saddr ? : phy_dev->dev_addr, len); 415 } 416 417 static const struct header_ops ipvlan_header_ops = { 418 .create = ipvlan_hard_header, 419 .parse = eth_header_parse, 420 .cache = eth_header_cache, 421 .cache_update = eth_header_cache_update, 422 }; 423 424 static bool netif_is_ipvlan(const struct net_device *dev) 425 { 426 /* both ipvlan and ipvtap devices use the same netdev_ops */ 427 return dev->netdev_ops == &ipvlan_netdev_ops; 428 } 429 430 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev, 431 struct ethtool_link_ksettings *cmd) 432 { 433 const struct ipvl_dev *ipvlan = netdev_priv(dev); 434 435 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd); 436 } 437 438 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev, 439 struct ethtool_drvinfo *drvinfo) 440 { 441 strlcpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver)); 442 strlcpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version)); 443 } 444 445 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev) 446 { 447 const struct ipvl_dev *ipvlan = netdev_priv(dev); 448 449 return ipvlan->msg_enable; 450 } 451 452 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value) 453 { 454 struct ipvl_dev *ipvlan = netdev_priv(dev); 455 456 ipvlan->msg_enable = value; 457 } 458 459 static const struct ethtool_ops ipvlan_ethtool_ops = { 460 .get_link = ethtool_op_get_link, 461 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings, 462 .get_drvinfo = ipvlan_ethtool_get_drvinfo, 463 .get_msglevel = ipvlan_ethtool_get_msglevel, 464 .set_msglevel = ipvlan_ethtool_set_msglevel, 465 }; 466 467 static int ipvlan_nl_changelink(struct net_device *dev, 468 struct nlattr *tb[], struct nlattr *data[], 469 struct netlink_ext_ack *extack) 470 { 471 struct ipvl_dev *ipvlan = netdev_priv(dev); 472 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); 473 int err = 0; 474 475 if (!data) 476 return 0; 477 478 if (data[IFLA_IPVLAN_MODE]) { 479 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 480 481 err = ipvlan_set_port_mode(port, nmode); 482 } 483 484 if (!err && data[IFLA_IPVLAN_FLAGS]) { 485 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 486 487 if (flags & IPVLAN_F_PRIVATE) 488 ipvlan_mark_private(port); 489 else 490 ipvlan_clear_private(port); 491 492 if (flags & IPVLAN_F_VEPA) 493 ipvlan_mark_vepa(port); 494 else 495 ipvlan_clear_vepa(port); 496 } 497 498 return err; 499 } 500 501 static size_t ipvlan_nl_getsize(const struct net_device *dev) 502 { 503 return (0 504 + nla_total_size(2) /* IFLA_IPVLAN_MODE */ 505 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */ 506 ); 507 } 508 509 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[], 510 struct netlink_ext_ack *extack) 511 { 512 if (!data) 513 return 0; 514 515 if (data[IFLA_IPVLAN_MODE]) { 516 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 517 518 if (mode < IPVLAN_MODE_L2 || mode >= IPVLAN_MODE_MAX) 519 return -EINVAL; 520 } 521 if (data[IFLA_IPVLAN_FLAGS]) { 522 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 523 524 /* Only two bits are used at this moment. */ 525 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) 526 return -EINVAL; 527 /* Also both flags can't be active at the same time. */ 528 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) == 529 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) 530 return -EINVAL; 531 } 532 533 return 0; 534 } 535 536 static int ipvlan_nl_fillinfo(struct sk_buff *skb, 537 const struct net_device *dev) 538 { 539 struct ipvl_dev *ipvlan = netdev_priv(dev); 540 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); 541 int ret = -EINVAL; 542 543 if (!port) 544 goto err; 545 546 ret = -EMSGSIZE; 547 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode)) 548 goto err; 549 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags)) 550 goto err; 551 552 return 0; 553 554 err: 555 return ret; 556 } 557 558 int ipvlan_link_new(struct net *src_net, struct net_device *dev, 559 struct nlattr *tb[], struct nlattr *data[], 560 struct netlink_ext_ack *extack) 561 { 562 struct ipvl_dev *ipvlan = netdev_priv(dev); 563 struct ipvl_port *port; 564 struct net_device *phy_dev; 565 int err; 566 u16 mode = IPVLAN_MODE_L3; 567 568 if (!tb[IFLA_LINK]) 569 return -EINVAL; 570 571 phy_dev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK])); 572 if (!phy_dev) 573 return -ENODEV; 574 575 if (netif_is_ipvlan(phy_dev)) { 576 struct ipvl_dev *tmp = netdev_priv(phy_dev); 577 578 phy_dev = tmp->phy_dev; 579 } else if (!netif_is_ipvlan_port(phy_dev)) { 580 /* Exit early if the underlying link is invalid or busy */ 581 if (phy_dev->type != ARPHRD_ETHER || 582 phy_dev->flags & IFF_LOOPBACK) { 583 netdev_err(phy_dev, 584 "Master is either lo or non-ether device\n"); 585 return -EINVAL; 586 } 587 588 if (netdev_is_rx_handler_busy(phy_dev)) { 589 netdev_err(phy_dev, "Device is already in use.\n"); 590 return -EBUSY; 591 } 592 } 593 594 ipvlan->phy_dev = phy_dev; 595 ipvlan->dev = dev; 596 ipvlan->sfeatures = IPVLAN_FEATURES; 597 if (!tb[IFLA_MTU]) 598 ipvlan_adjust_mtu(ipvlan, phy_dev); 599 INIT_LIST_HEAD(&ipvlan->addrs); 600 spin_lock_init(&ipvlan->addrs_lock); 601 602 /* TODO Probably put random address here to be presented to the 603 * world but keep using the physical-dev address for the outgoing 604 * packets. 605 */ 606 memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN); 607 608 dev->priv_flags |= IFF_NO_RX_HANDLER; 609 610 err = register_netdevice(dev); 611 if (err < 0) 612 return err; 613 614 /* ipvlan_init() would have created the port, if required */ 615 port = ipvlan_port_get_rtnl(phy_dev); 616 ipvlan->port = port; 617 618 /* If the port-id base is at the MAX value, then wrap it around and 619 * begin from 0x1 again. This may be due to a busy system where lots 620 * of slaves are getting created and deleted. 621 */ 622 if (port->dev_id_start == 0xFFFE) 623 port->dev_id_start = 0x1; 624 625 /* Since L2 address is shared among all IPvlan slaves including 626 * master, use unique 16 bit dev-ids to diffentiate among them. 627 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each 628 * slave link [see addrconf_ifid_eui48()]. 629 */ 630 err = ida_simple_get(&port->ida, port->dev_id_start, 0xFFFE, 631 GFP_KERNEL); 632 if (err < 0) 633 err = ida_simple_get(&port->ida, 0x1, port->dev_id_start, 634 GFP_KERNEL); 635 if (err < 0) 636 goto unregister_netdev; 637 dev->dev_id = err; 638 639 /* Increment id-base to the next slot for the future assignment */ 640 port->dev_id_start = err + 1; 641 642 err = netdev_upper_dev_link(phy_dev, dev, extack); 643 if (err) 644 goto remove_ida; 645 646 /* Flags are per port and latest update overrides. User has 647 * to be consistent in setting it just like the mode attribute. 648 */ 649 if (data && data[IFLA_IPVLAN_FLAGS]) 650 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 651 652 if (data && data[IFLA_IPVLAN_MODE]) 653 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 654 655 err = ipvlan_set_port_mode(port, mode); 656 if (err) 657 goto unlink_netdev; 658 659 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans); 660 netif_stacked_transfer_operstate(phy_dev, dev); 661 return 0; 662 663 unlink_netdev: 664 netdev_upper_dev_unlink(phy_dev, dev); 665 remove_ida: 666 ida_simple_remove(&port->ida, dev->dev_id); 667 unregister_netdev: 668 unregister_netdevice(dev); 669 return err; 670 } 671 EXPORT_SYMBOL_GPL(ipvlan_link_new); 672 673 void ipvlan_link_delete(struct net_device *dev, struct list_head *head) 674 { 675 struct ipvl_dev *ipvlan = netdev_priv(dev); 676 struct ipvl_addr *addr, *next; 677 678 spin_lock_bh(&ipvlan->addrs_lock); 679 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) { 680 ipvlan_ht_addr_del(addr); 681 list_del_rcu(&addr->anode); 682 kfree_rcu(addr, rcu); 683 } 684 spin_unlock_bh(&ipvlan->addrs_lock); 685 686 ida_simple_remove(&ipvlan->port->ida, dev->dev_id); 687 list_del_rcu(&ipvlan->pnode); 688 unregister_netdevice_queue(dev, head); 689 netdev_upper_dev_unlink(ipvlan->phy_dev, dev); 690 } 691 EXPORT_SYMBOL_GPL(ipvlan_link_delete); 692 693 void ipvlan_link_setup(struct net_device *dev) 694 { 695 ether_setup(dev); 696 697 dev->max_mtu = ETH_MAX_MTU; 698 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); 699 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE; 700 dev->netdev_ops = &ipvlan_netdev_ops; 701 dev->needs_free_netdev = true; 702 dev->header_ops = &ipvlan_header_ops; 703 dev->ethtool_ops = &ipvlan_ethtool_ops; 704 } 705 EXPORT_SYMBOL_GPL(ipvlan_link_setup); 706 707 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] = 708 { 709 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 }, 710 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 }, 711 }; 712 713 static struct rtnl_link_ops ipvlan_link_ops = { 714 .kind = "ipvlan", 715 .priv_size = sizeof(struct ipvl_dev), 716 717 .setup = ipvlan_link_setup, 718 .newlink = ipvlan_link_new, 719 .dellink = ipvlan_link_delete, 720 }; 721 722 int ipvlan_link_register(struct rtnl_link_ops *ops) 723 { 724 ops->get_size = ipvlan_nl_getsize; 725 ops->policy = ipvlan_nl_policy; 726 ops->validate = ipvlan_nl_validate; 727 ops->fill_info = ipvlan_nl_fillinfo; 728 ops->changelink = ipvlan_nl_changelink; 729 ops->maxtype = IFLA_IPVLAN_MAX; 730 return rtnl_link_register(ops); 731 } 732 EXPORT_SYMBOL_GPL(ipvlan_link_register); 733 734 static int ipvlan_device_event(struct notifier_block *unused, 735 unsigned long event, void *ptr) 736 { 737 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 738 struct ipvl_dev *ipvlan, *next; 739 struct ipvl_port *port; 740 LIST_HEAD(lst_kill); 741 742 if (!netif_is_ipvlan_port(dev)) 743 return NOTIFY_DONE; 744 745 port = ipvlan_port_get_rtnl(dev); 746 747 switch (event) { 748 case NETDEV_CHANGE: 749 list_for_each_entry(ipvlan, &port->ipvlans, pnode) 750 netif_stacked_transfer_operstate(ipvlan->phy_dev, 751 ipvlan->dev); 752 break; 753 754 case NETDEV_REGISTER: { 755 struct net *oldnet, *newnet = dev_net(dev); 756 struct ipvlan_netns *old_vnet; 757 758 oldnet = read_pnet(&port->pnet); 759 if (net_eq(newnet, oldnet)) 760 break; 761 762 write_pnet(&port->pnet, newnet); 763 764 old_vnet = net_generic(oldnet, ipvlan_netid); 765 if (!old_vnet->ipvl_nf_hook_refcnt) 766 break; 767 768 ipvlan_register_nf_hook(newnet); 769 ipvlan_unregister_nf_hook(oldnet); 770 break; 771 } 772 case NETDEV_UNREGISTER: 773 if (dev->reg_state != NETREG_UNREGISTERING) 774 break; 775 776 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode) 777 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev, 778 &lst_kill); 779 unregister_netdevice_many(&lst_kill); 780 break; 781 782 case NETDEV_FEAT_CHANGE: 783 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 784 ipvlan->dev->features = dev->features & IPVLAN_FEATURES; 785 ipvlan->dev->gso_max_size = dev->gso_max_size; 786 ipvlan->dev->gso_max_segs = dev->gso_max_segs; 787 netdev_features_change(ipvlan->dev); 788 } 789 break; 790 791 case NETDEV_CHANGEMTU: 792 list_for_each_entry(ipvlan, &port->ipvlans, pnode) 793 ipvlan_adjust_mtu(ipvlan, dev); 794 break; 795 796 case NETDEV_CHANGEADDR: 797 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 798 ether_addr_copy(ipvlan->dev->dev_addr, dev->dev_addr); 799 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev); 800 } 801 break; 802 803 case NETDEV_PRE_TYPE_CHANGE: 804 /* Forbid underlying device to change its type. */ 805 return NOTIFY_BAD; 806 } 807 return NOTIFY_DONE; 808 } 809 810 /* the caller must held the addrs lock */ 811 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6) 812 { 813 struct ipvl_addr *addr; 814 815 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC); 816 if (!addr) 817 return -ENOMEM; 818 819 addr->master = ipvlan; 820 if (!is_v6) { 821 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr)); 822 addr->atype = IPVL_IPV4; 823 #if IS_ENABLED(CONFIG_IPV6) 824 } else { 825 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr)); 826 addr->atype = IPVL_IPV6; 827 #endif 828 } 829 830 list_add_tail_rcu(&addr->anode, &ipvlan->addrs); 831 832 /* If the interface is not up, the address will be added to the hash 833 * list by ipvlan_open. 834 */ 835 if (netif_running(ipvlan->dev)) 836 ipvlan_ht_addr_add(ipvlan, addr); 837 838 return 0; 839 } 840 841 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6) 842 { 843 struct ipvl_addr *addr; 844 845 spin_lock_bh(&ipvlan->addrs_lock); 846 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6); 847 if (!addr) { 848 spin_unlock_bh(&ipvlan->addrs_lock); 849 return; 850 } 851 852 ipvlan_ht_addr_del(addr); 853 list_del_rcu(&addr->anode); 854 spin_unlock_bh(&ipvlan->addrs_lock); 855 kfree_rcu(addr, rcu); 856 } 857 858 static bool ipvlan_is_valid_dev(const struct net_device *dev) 859 { 860 struct ipvl_dev *ipvlan = netdev_priv(dev); 861 862 if (!netif_is_ipvlan(dev)) 863 return false; 864 865 if (!ipvlan || !ipvlan->port) 866 return false; 867 868 return true; 869 } 870 871 #if IS_ENABLED(CONFIG_IPV6) 872 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) 873 { 874 int ret = -EINVAL; 875 876 spin_lock_bh(&ipvlan->addrs_lock); 877 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true)) 878 netif_err(ipvlan, ifup, ipvlan->dev, 879 "Failed to add IPv6=%pI6c addr for %s intf\n", 880 ip6_addr, ipvlan->dev->name); 881 else 882 ret = ipvlan_add_addr(ipvlan, ip6_addr, true); 883 spin_unlock_bh(&ipvlan->addrs_lock); 884 return ret; 885 } 886 887 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) 888 { 889 return ipvlan_del_addr(ipvlan, ip6_addr, true); 890 } 891 892 static int ipvlan_addr6_event(struct notifier_block *unused, 893 unsigned long event, void *ptr) 894 { 895 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr; 896 struct net_device *dev = (struct net_device *)if6->idev->dev; 897 struct ipvl_dev *ipvlan = netdev_priv(dev); 898 899 if (!ipvlan_is_valid_dev(dev)) 900 return NOTIFY_DONE; 901 902 switch (event) { 903 case NETDEV_UP: 904 if (ipvlan_add_addr6(ipvlan, &if6->addr)) 905 return NOTIFY_BAD; 906 break; 907 908 case NETDEV_DOWN: 909 ipvlan_del_addr6(ipvlan, &if6->addr); 910 break; 911 } 912 913 return NOTIFY_OK; 914 } 915 916 static int ipvlan_addr6_validator_event(struct notifier_block *unused, 917 unsigned long event, void *ptr) 918 { 919 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr; 920 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev; 921 struct ipvl_dev *ipvlan = netdev_priv(dev); 922 923 if (!ipvlan_is_valid_dev(dev)) 924 return NOTIFY_DONE; 925 926 switch (event) { 927 case NETDEV_UP: 928 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) { 929 NL_SET_ERR_MSG(i6vi->extack, 930 "Address already assigned to an ipvlan device"); 931 return notifier_from_errno(-EADDRINUSE); 932 } 933 break; 934 } 935 936 return NOTIFY_OK; 937 } 938 #endif 939 940 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) 941 { 942 int ret = -EINVAL; 943 944 spin_lock_bh(&ipvlan->addrs_lock); 945 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false)) 946 netif_err(ipvlan, ifup, ipvlan->dev, 947 "Failed to add IPv4=%pI4 on %s intf.\n", 948 ip4_addr, ipvlan->dev->name); 949 else 950 ret = ipvlan_add_addr(ipvlan, ip4_addr, false); 951 spin_unlock_bh(&ipvlan->addrs_lock); 952 return ret; 953 } 954 955 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) 956 { 957 return ipvlan_del_addr(ipvlan, ip4_addr, false); 958 } 959 960 static int ipvlan_addr4_event(struct notifier_block *unused, 961 unsigned long event, void *ptr) 962 { 963 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr; 964 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev; 965 struct ipvl_dev *ipvlan = netdev_priv(dev); 966 struct in_addr ip4_addr; 967 968 if (!ipvlan_is_valid_dev(dev)) 969 return NOTIFY_DONE; 970 971 switch (event) { 972 case NETDEV_UP: 973 ip4_addr.s_addr = if4->ifa_address; 974 if (ipvlan_add_addr4(ipvlan, &ip4_addr)) 975 return NOTIFY_BAD; 976 break; 977 978 case NETDEV_DOWN: 979 ip4_addr.s_addr = if4->ifa_address; 980 ipvlan_del_addr4(ipvlan, &ip4_addr); 981 break; 982 } 983 984 return NOTIFY_OK; 985 } 986 987 static int ipvlan_addr4_validator_event(struct notifier_block *unused, 988 unsigned long event, void *ptr) 989 { 990 struct in_validator_info *ivi = (struct in_validator_info *)ptr; 991 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev; 992 struct ipvl_dev *ipvlan = netdev_priv(dev); 993 994 if (!ipvlan_is_valid_dev(dev)) 995 return NOTIFY_DONE; 996 997 switch (event) { 998 case NETDEV_UP: 999 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) { 1000 NL_SET_ERR_MSG(ivi->extack, 1001 "Address already assigned to an ipvlan device"); 1002 return notifier_from_errno(-EADDRINUSE); 1003 } 1004 break; 1005 } 1006 1007 return NOTIFY_OK; 1008 } 1009 1010 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = { 1011 .notifier_call = ipvlan_addr4_event, 1012 }; 1013 1014 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = { 1015 .notifier_call = ipvlan_addr4_validator_event, 1016 }; 1017 1018 static struct notifier_block ipvlan_notifier_block __read_mostly = { 1019 .notifier_call = ipvlan_device_event, 1020 }; 1021 1022 #if IS_ENABLED(CONFIG_IPV6) 1023 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = { 1024 .notifier_call = ipvlan_addr6_event, 1025 }; 1026 1027 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = { 1028 .notifier_call = ipvlan_addr6_validator_event, 1029 }; 1030 #endif 1031 1032 static void ipvlan_ns_exit(struct net *net) 1033 { 1034 struct ipvlan_netns *vnet = net_generic(net, ipvlan_netid); 1035 1036 if (WARN_ON_ONCE(vnet->ipvl_nf_hook_refcnt)) { 1037 vnet->ipvl_nf_hook_refcnt = 0; 1038 nf_unregister_net_hooks(net, ipvl_nfops, 1039 ARRAY_SIZE(ipvl_nfops)); 1040 } 1041 } 1042 1043 static struct pernet_operations ipvlan_net_ops = { 1044 .id = &ipvlan_netid, 1045 .size = sizeof(struct ipvlan_netns), 1046 .exit = ipvlan_ns_exit, 1047 }; 1048 1049 static int __init ipvlan_init_module(void) 1050 { 1051 int err; 1052 1053 ipvlan_init_secret(); 1054 register_netdevice_notifier(&ipvlan_notifier_block); 1055 #if IS_ENABLED(CONFIG_IPV6) 1056 register_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1057 register_inet6addr_validator_notifier( 1058 &ipvlan_addr6_vtor_notifier_block); 1059 #endif 1060 register_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1061 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block); 1062 1063 err = register_pernet_subsys(&ipvlan_net_ops); 1064 if (err < 0) 1065 goto error; 1066 1067 err = ipvlan_link_register(&ipvlan_link_ops); 1068 if (err < 0) { 1069 unregister_pernet_subsys(&ipvlan_net_ops); 1070 goto error; 1071 } 1072 1073 return 0; 1074 error: 1075 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1076 unregister_inetaddr_validator_notifier( 1077 &ipvlan_addr4_vtor_notifier_block); 1078 #if IS_ENABLED(CONFIG_IPV6) 1079 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1080 unregister_inet6addr_validator_notifier( 1081 &ipvlan_addr6_vtor_notifier_block); 1082 #endif 1083 unregister_netdevice_notifier(&ipvlan_notifier_block); 1084 return err; 1085 } 1086 1087 static void __exit ipvlan_cleanup_module(void) 1088 { 1089 rtnl_link_unregister(&ipvlan_link_ops); 1090 unregister_pernet_subsys(&ipvlan_net_ops); 1091 unregister_netdevice_notifier(&ipvlan_notifier_block); 1092 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1093 unregister_inetaddr_validator_notifier( 1094 &ipvlan_addr4_vtor_notifier_block); 1095 #if IS_ENABLED(CONFIG_IPV6) 1096 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1097 unregister_inet6addr_validator_notifier( 1098 &ipvlan_addr6_vtor_notifier_block); 1099 #endif 1100 } 1101 1102 module_init(ipvlan_init_module); 1103 module_exit(ipvlan_cleanup_module); 1104 1105 MODULE_LICENSE("GPL"); 1106 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>"); 1107 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs"); 1108 MODULE_ALIAS_RTNL_LINK("ipvlan"); 1109