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