1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com> 3 */ 4 5 #include <linux/ethtool.h> 6 7 #include "ipvlan.h" 8 9 static int ipvlan_set_port_mode(struct ipvl_port *port, u16 nval, 10 struct netlink_ext_ack *extack) 11 { 12 struct ipvl_dev *ipvlan; 13 unsigned int flags; 14 int err; 15 16 ASSERT_RTNL(); 17 if (port->mode != nval) { 18 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 19 flags = ipvlan->dev->flags; 20 if (nval == IPVLAN_MODE_L3 || nval == IPVLAN_MODE_L3S) { 21 err = dev_change_flags(ipvlan->dev, 22 flags | IFF_NOARP, 23 extack); 24 } else { 25 err = dev_change_flags(ipvlan->dev, 26 flags & ~IFF_NOARP, 27 extack); 28 } 29 if (unlikely(err)) 30 goto fail; 31 } 32 if (nval == IPVLAN_MODE_L3S) { 33 /* New mode is L3S */ 34 err = ipvlan_l3s_register(port); 35 if (err) 36 goto fail; 37 } else if (port->mode == IPVLAN_MODE_L3S) { 38 /* Old mode was L3S */ 39 ipvlan_l3s_unregister(port); 40 } 41 port->mode = nval; 42 } 43 return 0; 44 45 fail: 46 /* Undo the flags changes that have been done so far. */ 47 list_for_each_entry_continue_reverse(ipvlan, &port->ipvlans, pnode) { 48 flags = ipvlan->dev->flags; 49 if (port->mode == IPVLAN_MODE_L3 || 50 port->mode == IPVLAN_MODE_L3S) 51 dev_change_flags(ipvlan->dev, flags | IFF_NOARP, 52 NULL); 53 else 54 dev_change_flags(ipvlan->dev, flags & ~IFF_NOARP, 55 NULL); 56 } 57 58 return err; 59 } 60 61 static int ipvlan_port_create(struct net_device *dev) 62 { 63 struct ipvl_port *port; 64 int err, idx; 65 66 port = kzalloc(sizeof(struct ipvl_port), GFP_KERNEL); 67 if (!port) 68 return -ENOMEM; 69 70 write_pnet(&port->pnet, dev_net(dev)); 71 port->dev = dev; 72 port->mode = IPVLAN_MODE_L3; 73 INIT_LIST_HEAD(&port->ipvlans); 74 for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++) 75 INIT_HLIST_HEAD(&port->hlhead[idx]); 76 77 skb_queue_head_init(&port->backlog); 78 INIT_WORK(&port->wq, ipvlan_process_multicast); 79 ida_init(&port->ida); 80 port->dev_id_start = 1; 81 82 err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port); 83 if (err) 84 goto err; 85 86 netdev_hold(dev, &port->dev_tracker, GFP_KERNEL); 87 return 0; 88 89 err: 90 kfree(port); 91 return err; 92 } 93 94 static void ipvlan_port_destroy(struct net_device *dev) 95 { 96 struct ipvl_port *port = ipvlan_port_get_rtnl(dev); 97 struct sk_buff *skb; 98 99 netdev_put(dev, &port->dev_tracker); 100 if (port->mode == IPVLAN_MODE_L3S) 101 ipvlan_l3s_unregister(port); 102 netdev_rx_handler_unregister(dev); 103 cancel_work_sync(&port->wq); 104 while ((skb = __skb_dequeue(&port->backlog)) != NULL) { 105 dev_put(skb->dev); 106 kfree_skb(skb); 107 } 108 ida_destroy(&port->ida); 109 kfree(port); 110 } 111 112 #define IPVLAN_ALWAYS_ON_OFLOADS \ 113 (NETIF_F_SG | NETIF_F_HW_CSUM | \ 114 NETIF_F_GSO_ROBUST | NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL) 115 116 #define IPVLAN_ALWAYS_ON \ 117 (IPVLAN_ALWAYS_ON_OFLOADS | NETIF_F_VLAN_CHALLENGED) 118 119 #define IPVLAN_FEATURES \ 120 (NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \ 121 NETIF_F_GSO | NETIF_F_ALL_TSO | NETIF_F_GSO_ROBUST | \ 122 NETIF_F_GRO | NETIF_F_RXCSUM | \ 123 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER) 124 125 /* NETIF_F_GSO_ENCAP_ALL NETIF_F_GSO_SOFTWARE Newly added */ 126 127 #define IPVLAN_STATE_MASK \ 128 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT)) 129 130 static int ipvlan_init(struct net_device *dev) 131 { 132 struct ipvl_dev *ipvlan = netdev_priv(dev); 133 struct net_device *phy_dev = ipvlan->phy_dev; 134 struct ipvl_port *port; 135 int err; 136 137 dev->state = (dev->state & ~IPVLAN_STATE_MASK) | 138 (phy_dev->state & IPVLAN_STATE_MASK); 139 dev->features = phy_dev->features & IPVLAN_FEATURES; 140 dev->features |= IPVLAN_ALWAYS_ON; 141 dev->vlan_features = phy_dev->vlan_features & IPVLAN_FEATURES; 142 dev->vlan_features |= IPVLAN_ALWAYS_ON_OFLOADS; 143 dev->hw_enc_features |= dev->features; 144 dev->lltx = true; 145 netif_inherit_tso_max(dev, phy_dev); 146 dev->hard_header_len = phy_dev->hard_header_len; 147 148 netdev_lockdep_set_classes(dev); 149 150 ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats); 151 if (!ipvlan->pcpu_stats) 152 return -ENOMEM; 153 154 if (!netif_is_ipvlan_port(phy_dev)) { 155 err = ipvlan_port_create(phy_dev); 156 if (err < 0) { 157 free_percpu(ipvlan->pcpu_stats); 158 return err; 159 } 160 } 161 port = ipvlan_port_get_rtnl(phy_dev); 162 port->count += 1; 163 return 0; 164 } 165 166 static void ipvlan_uninit(struct net_device *dev) 167 { 168 struct ipvl_dev *ipvlan = netdev_priv(dev); 169 struct net_device *phy_dev = ipvlan->phy_dev; 170 struct ipvl_port *port; 171 172 free_percpu(ipvlan->pcpu_stats); 173 174 port = ipvlan_port_get_rtnl(phy_dev); 175 port->count -= 1; 176 if (!port->count) 177 ipvlan_port_destroy(port->dev); 178 } 179 180 static int ipvlan_open(struct net_device *dev) 181 { 182 struct ipvl_dev *ipvlan = netdev_priv(dev); 183 struct ipvl_addr *addr; 184 185 if (ipvlan->port->mode == IPVLAN_MODE_L3 || 186 ipvlan->port->mode == IPVLAN_MODE_L3S) 187 dev->flags |= IFF_NOARP; 188 else 189 dev->flags &= ~IFF_NOARP; 190 191 rcu_read_lock(); 192 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) 193 ipvlan_ht_addr_add(ipvlan, addr); 194 rcu_read_unlock(); 195 196 return 0; 197 } 198 199 static int ipvlan_stop(struct net_device *dev) 200 { 201 struct ipvl_dev *ipvlan = netdev_priv(dev); 202 struct net_device *phy_dev = ipvlan->phy_dev; 203 struct ipvl_addr *addr; 204 205 dev_uc_unsync(phy_dev, dev); 206 dev_mc_unsync(phy_dev, dev); 207 208 rcu_read_lock(); 209 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) 210 ipvlan_ht_addr_del(addr); 211 rcu_read_unlock(); 212 213 return 0; 214 } 215 216 static netdev_tx_t ipvlan_start_xmit(struct sk_buff *skb, 217 struct net_device *dev) 218 { 219 const struct ipvl_dev *ipvlan = netdev_priv(dev); 220 int skblen = skb->len; 221 int ret; 222 223 ret = ipvlan_queue_xmit(skb, dev); 224 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) { 225 struct ipvl_pcpu_stats *pcptr; 226 227 pcptr = this_cpu_ptr(ipvlan->pcpu_stats); 228 229 u64_stats_update_begin(&pcptr->syncp); 230 u64_stats_inc(&pcptr->tx_pkts); 231 u64_stats_add(&pcptr->tx_bytes, skblen); 232 u64_stats_update_end(&pcptr->syncp); 233 } else { 234 this_cpu_inc(ipvlan->pcpu_stats->tx_drps); 235 } 236 return ret; 237 } 238 239 static netdev_features_t ipvlan_fix_features(struct net_device *dev, 240 netdev_features_t features) 241 { 242 struct ipvl_dev *ipvlan = netdev_priv(dev); 243 244 features |= NETIF_F_ALL_FOR_ALL; 245 features &= (ipvlan->sfeatures | ~IPVLAN_FEATURES); 246 features = netdev_increment_features(ipvlan->phy_dev->features, 247 features, features); 248 features |= IPVLAN_ALWAYS_ON; 249 features &= (IPVLAN_FEATURES | IPVLAN_ALWAYS_ON); 250 251 return features; 252 } 253 254 static void ipvlan_change_rx_flags(struct net_device *dev, int change) 255 { 256 struct ipvl_dev *ipvlan = netdev_priv(dev); 257 struct net_device *phy_dev = ipvlan->phy_dev; 258 259 if (change & IFF_ALLMULTI) 260 dev_set_allmulti(phy_dev, dev->flags & IFF_ALLMULTI? 1 : -1); 261 } 262 263 static void ipvlan_set_multicast_mac_filter(struct net_device *dev) 264 { 265 struct ipvl_dev *ipvlan = netdev_priv(dev); 266 267 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) { 268 bitmap_fill(ipvlan->mac_filters, IPVLAN_MAC_FILTER_SIZE); 269 } else { 270 struct netdev_hw_addr *ha; 271 DECLARE_BITMAP(mc_filters, IPVLAN_MAC_FILTER_SIZE); 272 273 bitmap_zero(mc_filters, IPVLAN_MAC_FILTER_SIZE); 274 netdev_for_each_mc_addr(ha, dev) 275 __set_bit(ipvlan_mac_hash(ha->addr), mc_filters); 276 277 /* Turn-on broadcast bit irrespective of address family, 278 * since broadcast is deferred to a work-queue, hence no 279 * impact on fast-path processing. 280 */ 281 __set_bit(ipvlan_mac_hash(dev->broadcast), mc_filters); 282 283 bitmap_copy(ipvlan->mac_filters, mc_filters, 284 IPVLAN_MAC_FILTER_SIZE); 285 } 286 dev_uc_sync(ipvlan->phy_dev, dev); 287 dev_mc_sync(ipvlan->phy_dev, dev); 288 } 289 290 static void ipvlan_get_stats64(struct net_device *dev, 291 struct rtnl_link_stats64 *s) 292 { 293 struct ipvl_dev *ipvlan = netdev_priv(dev); 294 295 if (ipvlan->pcpu_stats) { 296 struct ipvl_pcpu_stats *pcptr; 297 u64 rx_pkts, rx_bytes, rx_mcast, tx_pkts, tx_bytes; 298 u32 rx_errs = 0, tx_drps = 0; 299 u32 strt; 300 int idx; 301 302 for_each_possible_cpu(idx) { 303 pcptr = per_cpu_ptr(ipvlan->pcpu_stats, idx); 304 do { 305 strt = u64_stats_fetch_begin(&pcptr->syncp); 306 rx_pkts = u64_stats_read(&pcptr->rx_pkts); 307 rx_bytes = u64_stats_read(&pcptr->rx_bytes); 308 rx_mcast = u64_stats_read(&pcptr->rx_mcast); 309 tx_pkts = u64_stats_read(&pcptr->tx_pkts); 310 tx_bytes = u64_stats_read(&pcptr->tx_bytes); 311 } while (u64_stats_fetch_retry(&pcptr->syncp, 312 strt)); 313 314 s->rx_packets += rx_pkts; 315 s->rx_bytes += rx_bytes; 316 s->multicast += rx_mcast; 317 s->tx_packets += tx_pkts; 318 s->tx_bytes += tx_bytes; 319 320 /* u32 values are updated without syncp protection. */ 321 rx_errs += READ_ONCE(pcptr->rx_errs); 322 tx_drps += READ_ONCE(pcptr->tx_drps); 323 } 324 s->rx_errors = rx_errs; 325 s->rx_dropped = rx_errs; 326 s->tx_dropped = tx_drps; 327 } 328 s->tx_errors = DEV_STATS_READ(dev, tx_errors); 329 } 330 331 static int ipvlan_vlan_rx_add_vid(struct net_device *dev, __be16 proto, u16 vid) 332 { 333 struct ipvl_dev *ipvlan = netdev_priv(dev); 334 struct net_device *phy_dev = ipvlan->phy_dev; 335 336 return vlan_vid_add(phy_dev, proto, vid); 337 } 338 339 static int ipvlan_vlan_rx_kill_vid(struct net_device *dev, __be16 proto, 340 u16 vid) 341 { 342 struct ipvl_dev *ipvlan = netdev_priv(dev); 343 struct net_device *phy_dev = ipvlan->phy_dev; 344 345 vlan_vid_del(phy_dev, proto, vid); 346 return 0; 347 } 348 349 static int ipvlan_get_iflink(const struct net_device *dev) 350 { 351 struct ipvl_dev *ipvlan = netdev_priv(dev); 352 353 return READ_ONCE(ipvlan->phy_dev->ifindex); 354 } 355 356 static const struct net_device_ops ipvlan_netdev_ops = { 357 .ndo_init = ipvlan_init, 358 .ndo_uninit = ipvlan_uninit, 359 .ndo_open = ipvlan_open, 360 .ndo_stop = ipvlan_stop, 361 .ndo_start_xmit = ipvlan_start_xmit, 362 .ndo_fix_features = ipvlan_fix_features, 363 .ndo_change_rx_flags = ipvlan_change_rx_flags, 364 .ndo_set_rx_mode = ipvlan_set_multicast_mac_filter, 365 .ndo_get_stats64 = ipvlan_get_stats64, 366 .ndo_vlan_rx_add_vid = ipvlan_vlan_rx_add_vid, 367 .ndo_vlan_rx_kill_vid = ipvlan_vlan_rx_kill_vid, 368 .ndo_get_iflink = ipvlan_get_iflink, 369 }; 370 371 static int ipvlan_hard_header(struct sk_buff *skb, struct net_device *dev, 372 unsigned short type, const void *daddr, 373 const void *saddr, unsigned len) 374 { 375 const struct ipvl_dev *ipvlan = netdev_priv(dev); 376 struct net_device *phy_dev = ipvlan->phy_dev; 377 378 /* TODO Probably use a different field than dev_addr so that the 379 * mac-address on the virtual device is portable and can be carried 380 * while the packets use the mac-addr on the physical device. 381 */ 382 return dev_hard_header(skb, phy_dev, type, daddr, 383 saddr ? : phy_dev->dev_addr, len); 384 } 385 386 static const struct header_ops ipvlan_header_ops = { 387 .create = ipvlan_hard_header, 388 .parse = eth_header_parse, 389 .cache = eth_header_cache, 390 .cache_update = eth_header_cache_update, 391 .parse_protocol = eth_header_parse_protocol, 392 }; 393 394 static void ipvlan_adjust_mtu(struct ipvl_dev *ipvlan, struct net_device *dev) 395 { 396 ipvlan->dev->mtu = dev->mtu; 397 } 398 399 static bool netif_is_ipvlan(const struct net_device *dev) 400 { 401 /* both ipvlan and ipvtap devices use the same netdev_ops */ 402 return dev->netdev_ops == &ipvlan_netdev_ops; 403 } 404 405 static int ipvlan_ethtool_get_link_ksettings(struct net_device *dev, 406 struct ethtool_link_ksettings *cmd) 407 { 408 const struct ipvl_dev *ipvlan = netdev_priv(dev); 409 410 return __ethtool_get_link_ksettings(ipvlan->phy_dev, cmd); 411 } 412 413 static void ipvlan_ethtool_get_drvinfo(struct net_device *dev, 414 struct ethtool_drvinfo *drvinfo) 415 { 416 strscpy(drvinfo->driver, IPVLAN_DRV, sizeof(drvinfo->driver)); 417 strscpy(drvinfo->version, IPV_DRV_VER, sizeof(drvinfo->version)); 418 } 419 420 static u32 ipvlan_ethtool_get_msglevel(struct net_device *dev) 421 { 422 const struct ipvl_dev *ipvlan = netdev_priv(dev); 423 424 return ipvlan->msg_enable; 425 } 426 427 static void ipvlan_ethtool_set_msglevel(struct net_device *dev, u32 value) 428 { 429 struct ipvl_dev *ipvlan = netdev_priv(dev); 430 431 ipvlan->msg_enable = value; 432 } 433 434 static const struct ethtool_ops ipvlan_ethtool_ops = { 435 .get_link = ethtool_op_get_link, 436 .get_link_ksettings = ipvlan_ethtool_get_link_ksettings, 437 .get_drvinfo = ipvlan_ethtool_get_drvinfo, 438 .get_msglevel = ipvlan_ethtool_get_msglevel, 439 .set_msglevel = ipvlan_ethtool_set_msglevel, 440 }; 441 442 static int ipvlan_nl_changelink(struct net_device *dev, 443 struct nlattr *tb[], struct nlattr *data[], 444 struct netlink_ext_ack *extack) 445 { 446 struct ipvl_dev *ipvlan = netdev_priv(dev); 447 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); 448 int err = 0; 449 450 if (!data) 451 return 0; 452 if (!ns_capable(dev_net(ipvlan->phy_dev)->user_ns, CAP_NET_ADMIN)) 453 return -EPERM; 454 455 if (data[IFLA_IPVLAN_MODE]) { 456 u16 nmode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 457 458 err = ipvlan_set_port_mode(port, nmode, extack); 459 } 460 461 if (!err && data[IFLA_IPVLAN_FLAGS]) { 462 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 463 464 if (flags & IPVLAN_F_PRIVATE) 465 ipvlan_mark_private(port); 466 else 467 ipvlan_clear_private(port); 468 469 if (flags & IPVLAN_F_VEPA) 470 ipvlan_mark_vepa(port); 471 else 472 ipvlan_clear_vepa(port); 473 } 474 475 return err; 476 } 477 478 static size_t ipvlan_nl_getsize(const struct net_device *dev) 479 { 480 return (0 481 + nla_total_size(2) /* IFLA_IPVLAN_MODE */ 482 + nla_total_size(2) /* IFLA_IPVLAN_FLAGS */ 483 ); 484 } 485 486 static int ipvlan_nl_validate(struct nlattr *tb[], struct nlattr *data[], 487 struct netlink_ext_ack *extack) 488 { 489 if (!data) 490 return 0; 491 492 if (data[IFLA_IPVLAN_MODE]) { 493 u16 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 494 495 if (mode >= IPVLAN_MODE_MAX) 496 return -EINVAL; 497 } 498 if (data[IFLA_IPVLAN_FLAGS]) { 499 u16 flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 500 501 /* Only two bits are used at this moment. */ 502 if (flags & ~(IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) 503 return -EINVAL; 504 /* Also both flags can't be active at the same time. */ 505 if ((flags & (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) == 506 (IPVLAN_F_PRIVATE | IPVLAN_F_VEPA)) 507 return -EINVAL; 508 } 509 510 return 0; 511 } 512 513 static int ipvlan_nl_fillinfo(struct sk_buff *skb, 514 const struct net_device *dev) 515 { 516 struct ipvl_dev *ipvlan = netdev_priv(dev); 517 struct ipvl_port *port = ipvlan_port_get_rtnl(ipvlan->phy_dev); 518 int ret = -EINVAL; 519 520 if (!port) 521 goto err; 522 523 ret = -EMSGSIZE; 524 if (nla_put_u16(skb, IFLA_IPVLAN_MODE, port->mode)) 525 goto err; 526 if (nla_put_u16(skb, IFLA_IPVLAN_FLAGS, port->flags)) 527 goto err; 528 529 return 0; 530 531 err: 532 return ret; 533 } 534 535 int ipvlan_link_new(struct net_device *dev, struct rtnl_newlink_params *params, 536 struct netlink_ext_ack *extack) 537 { 538 struct net *link_net = rtnl_newlink_link_net(params); 539 struct ipvl_dev *ipvlan = netdev_priv(dev); 540 struct nlattr **data = params->data; 541 struct nlattr **tb = params->tb; 542 struct ipvl_port *port; 543 struct net_device *phy_dev; 544 int err; 545 u16 mode = IPVLAN_MODE_L3; 546 547 if (!tb[IFLA_LINK]) 548 return -EINVAL; 549 550 phy_dev = __dev_get_by_index(link_net, nla_get_u32(tb[IFLA_LINK])); 551 if (!phy_dev) 552 return -ENODEV; 553 554 if (netif_is_ipvlan(phy_dev)) { 555 struct ipvl_dev *tmp = netdev_priv(phy_dev); 556 557 phy_dev = tmp->phy_dev; 558 if (!ns_capable(dev_net(phy_dev)->user_ns, CAP_NET_ADMIN)) 559 return -EPERM; 560 } else if (!netif_is_ipvlan_port(phy_dev)) { 561 /* Exit early if the underlying link is invalid or busy */ 562 if (phy_dev->type != ARPHRD_ETHER || 563 phy_dev->flags & IFF_LOOPBACK) { 564 netdev_err(phy_dev, 565 "Master is either lo or non-ether device\n"); 566 return -EINVAL; 567 } 568 569 if (netdev_is_rx_handler_busy(phy_dev)) { 570 netdev_err(phy_dev, "Device is already in use.\n"); 571 return -EBUSY; 572 } 573 } 574 575 ipvlan->phy_dev = phy_dev; 576 ipvlan->dev = dev; 577 ipvlan->sfeatures = IPVLAN_FEATURES; 578 if (!tb[IFLA_MTU]) 579 ipvlan_adjust_mtu(ipvlan, phy_dev); 580 INIT_LIST_HEAD(&ipvlan->addrs); 581 spin_lock_init(&ipvlan->addrs_lock); 582 583 /* TODO Probably put random address here to be presented to the 584 * world but keep using the physical-dev address for the outgoing 585 * packets. 586 */ 587 eth_hw_addr_set(dev, phy_dev->dev_addr); 588 589 dev->priv_flags |= IFF_NO_RX_HANDLER; 590 591 err = register_netdevice(dev); 592 if (err < 0) 593 return err; 594 595 /* ipvlan_init() would have created the port, if required */ 596 port = ipvlan_port_get_rtnl(phy_dev); 597 ipvlan->port = port; 598 599 /* If the port-id base is at the MAX value, then wrap it around and 600 * begin from 0x1 again. This may be due to a busy system where lots 601 * of slaves are getting created and deleted. 602 */ 603 if (port->dev_id_start == 0xFFFE) 604 port->dev_id_start = 0x1; 605 606 /* Since L2 address is shared among all IPvlan slaves including 607 * master, use unique 16 bit dev-ids to differentiate among them. 608 * Assign IDs between 0x1 and 0xFFFE (used by the master) to each 609 * slave link [see addrconf_ifid_eui48()]. 610 */ 611 err = ida_alloc_range(&port->ida, port->dev_id_start, 0xFFFD, 612 GFP_KERNEL); 613 if (err < 0) 614 err = ida_alloc_range(&port->ida, 0x1, port->dev_id_start - 1, 615 GFP_KERNEL); 616 if (err < 0) 617 goto unregister_netdev; 618 dev->dev_id = err; 619 620 /* Increment id-base to the next slot for the future assignment */ 621 port->dev_id_start = err + 1; 622 623 err = netdev_upper_dev_link(phy_dev, dev, extack); 624 if (err) 625 goto remove_ida; 626 627 /* Flags are per port and latest update overrides. User has 628 * to be consistent in setting it just like the mode attribute. 629 */ 630 if (data && data[IFLA_IPVLAN_FLAGS]) 631 port->flags = nla_get_u16(data[IFLA_IPVLAN_FLAGS]); 632 633 if (data && data[IFLA_IPVLAN_MODE]) 634 mode = nla_get_u16(data[IFLA_IPVLAN_MODE]); 635 636 err = ipvlan_set_port_mode(port, mode, extack); 637 if (err) 638 goto unlink_netdev; 639 640 list_add_tail_rcu(&ipvlan->pnode, &port->ipvlans); 641 netif_stacked_transfer_operstate(phy_dev, dev); 642 return 0; 643 644 unlink_netdev: 645 netdev_upper_dev_unlink(phy_dev, dev); 646 remove_ida: 647 ida_free(&port->ida, dev->dev_id); 648 unregister_netdev: 649 unregister_netdevice(dev); 650 return err; 651 } 652 EXPORT_SYMBOL_GPL(ipvlan_link_new); 653 654 void ipvlan_link_delete(struct net_device *dev, struct list_head *head) 655 { 656 struct ipvl_dev *ipvlan = netdev_priv(dev); 657 struct ipvl_addr *addr, *next; 658 659 spin_lock_bh(&ipvlan->addrs_lock); 660 list_for_each_entry_safe(addr, next, &ipvlan->addrs, anode) { 661 ipvlan_ht_addr_del(addr); 662 list_del_rcu(&addr->anode); 663 kfree_rcu(addr, rcu); 664 } 665 spin_unlock_bh(&ipvlan->addrs_lock); 666 667 ida_free(&ipvlan->port->ida, dev->dev_id); 668 list_del_rcu(&ipvlan->pnode); 669 unregister_netdevice_queue(dev, head); 670 netdev_upper_dev_unlink(ipvlan->phy_dev, dev); 671 } 672 EXPORT_SYMBOL_GPL(ipvlan_link_delete); 673 674 void ipvlan_link_setup(struct net_device *dev) 675 { 676 ether_setup(dev); 677 678 dev->max_mtu = ETH_MAX_MTU; 679 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); 680 dev->priv_flags |= IFF_UNICAST_FLT | IFF_NO_QUEUE; 681 dev->netdev_ops = &ipvlan_netdev_ops; 682 dev->needs_free_netdev = true; 683 dev->header_ops = &ipvlan_header_ops; 684 dev->ethtool_ops = &ipvlan_ethtool_ops; 685 } 686 EXPORT_SYMBOL_GPL(ipvlan_link_setup); 687 688 static const struct nla_policy ipvlan_nl_policy[IFLA_IPVLAN_MAX + 1] = 689 { 690 [IFLA_IPVLAN_MODE] = { .type = NLA_U16 }, 691 [IFLA_IPVLAN_FLAGS] = { .type = NLA_U16 }, 692 }; 693 694 static struct net *ipvlan_get_link_net(const struct net_device *dev) 695 { 696 struct ipvl_dev *ipvlan = netdev_priv(dev); 697 698 return dev_net(ipvlan->phy_dev); 699 } 700 701 static struct rtnl_link_ops ipvlan_link_ops = { 702 .kind = "ipvlan", 703 .priv_size = sizeof(struct ipvl_dev), 704 705 .setup = ipvlan_link_setup, 706 .newlink = ipvlan_link_new, 707 .dellink = ipvlan_link_delete, 708 .get_link_net = ipvlan_get_link_net, 709 }; 710 711 int ipvlan_link_register(struct rtnl_link_ops *ops) 712 { 713 ops->get_size = ipvlan_nl_getsize; 714 ops->policy = ipvlan_nl_policy; 715 ops->validate = ipvlan_nl_validate; 716 ops->fill_info = ipvlan_nl_fillinfo; 717 ops->changelink = ipvlan_nl_changelink; 718 ops->maxtype = IFLA_IPVLAN_MAX; 719 return rtnl_link_register(ops); 720 } 721 EXPORT_SYMBOL_GPL(ipvlan_link_register); 722 723 static int ipvlan_device_event(struct notifier_block *unused, 724 unsigned long event, void *ptr) 725 { 726 struct netlink_ext_ack *extack = netdev_notifier_info_to_extack(ptr); 727 struct netdev_notifier_pre_changeaddr_info *prechaddr_info; 728 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 729 struct ipvl_dev *ipvlan, *next; 730 struct ipvl_port *port; 731 LIST_HEAD(lst_kill); 732 int err; 733 734 if (!netif_is_ipvlan_port(dev)) 735 return NOTIFY_DONE; 736 737 port = ipvlan_port_get_rtnl(dev); 738 739 switch (event) { 740 case NETDEV_UP: 741 case NETDEV_DOWN: 742 case NETDEV_CHANGE: 743 list_for_each_entry(ipvlan, &port->ipvlans, pnode) 744 netif_stacked_transfer_operstate(ipvlan->phy_dev, 745 ipvlan->dev); 746 break; 747 748 case NETDEV_REGISTER: { 749 struct net *oldnet, *newnet = dev_net(dev); 750 751 oldnet = read_pnet(&port->pnet); 752 if (net_eq(newnet, oldnet)) 753 break; 754 755 write_pnet(&port->pnet, newnet); 756 757 if (port->mode == IPVLAN_MODE_L3S) 758 ipvlan_migrate_l3s_hook(oldnet, newnet); 759 break; 760 } 761 case NETDEV_UNREGISTER: 762 if (dev->reg_state != NETREG_UNREGISTERING) 763 break; 764 765 list_for_each_entry_safe(ipvlan, next, &port->ipvlans, pnode) 766 ipvlan->dev->rtnl_link_ops->dellink(ipvlan->dev, 767 &lst_kill); 768 unregister_netdevice_many(&lst_kill); 769 break; 770 771 case NETDEV_FEAT_CHANGE: 772 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 773 netif_inherit_tso_max(ipvlan->dev, dev); 774 netdev_update_features(ipvlan->dev); 775 } 776 break; 777 778 case NETDEV_CHANGEMTU: 779 list_for_each_entry(ipvlan, &port->ipvlans, pnode) 780 ipvlan_adjust_mtu(ipvlan, dev); 781 break; 782 783 case NETDEV_PRE_CHANGEADDR: 784 prechaddr_info = ptr; 785 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 786 err = dev_pre_changeaddr_notify(ipvlan->dev, 787 prechaddr_info->dev_addr, 788 extack); 789 if (err) 790 return notifier_from_errno(err); 791 } 792 break; 793 794 case NETDEV_CHANGEADDR: 795 list_for_each_entry(ipvlan, &port->ipvlans, pnode) { 796 eth_hw_addr_set(ipvlan->dev, dev->dev_addr); 797 call_netdevice_notifiers(NETDEV_CHANGEADDR, ipvlan->dev); 798 } 799 break; 800 801 case NETDEV_PRE_TYPE_CHANGE: 802 /* Forbid underlying device to change its type. */ 803 return NOTIFY_BAD; 804 805 case NETDEV_NOTIFY_PEERS: 806 case NETDEV_BONDING_FAILOVER: 807 case NETDEV_RESEND_IGMP: 808 list_for_each_entry(ipvlan, &port->ipvlans, pnode) 809 call_netdevice_notifiers(event, ipvlan->dev); 810 } 811 return NOTIFY_DONE; 812 } 813 814 /* the caller must held the addrs lock */ 815 static int ipvlan_add_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6) 816 { 817 struct ipvl_addr *addr; 818 819 addr = kzalloc(sizeof(struct ipvl_addr), GFP_ATOMIC); 820 if (!addr) 821 return -ENOMEM; 822 823 addr->master = ipvlan; 824 if (!is_v6) { 825 memcpy(&addr->ip4addr, iaddr, sizeof(struct in_addr)); 826 addr->atype = IPVL_IPV4; 827 #if IS_ENABLED(CONFIG_IPV6) 828 } else { 829 memcpy(&addr->ip6addr, iaddr, sizeof(struct in6_addr)); 830 addr->atype = IPVL_IPV6; 831 #endif 832 } 833 834 list_add_tail_rcu(&addr->anode, &ipvlan->addrs); 835 836 /* If the interface is not up, the address will be added to the hash 837 * list by ipvlan_open. 838 */ 839 if (netif_running(ipvlan->dev)) 840 ipvlan_ht_addr_add(ipvlan, addr); 841 842 return 0; 843 } 844 845 static void ipvlan_del_addr(struct ipvl_dev *ipvlan, void *iaddr, bool is_v6) 846 { 847 struct ipvl_addr *addr; 848 849 spin_lock_bh(&ipvlan->addrs_lock); 850 addr = ipvlan_find_addr(ipvlan, iaddr, is_v6); 851 if (!addr) { 852 spin_unlock_bh(&ipvlan->addrs_lock); 853 return; 854 } 855 856 ipvlan_ht_addr_del(addr); 857 list_del_rcu(&addr->anode); 858 spin_unlock_bh(&ipvlan->addrs_lock); 859 kfree_rcu(addr, rcu); 860 } 861 862 static bool ipvlan_is_valid_dev(const struct net_device *dev) 863 { 864 struct ipvl_dev *ipvlan = netdev_priv(dev); 865 866 if (!netif_is_ipvlan(dev)) 867 return false; 868 869 if (!ipvlan || !ipvlan->port) 870 return false; 871 872 return true; 873 } 874 875 #if IS_ENABLED(CONFIG_IPV6) 876 static int ipvlan_add_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) 877 { 878 int ret = -EINVAL; 879 880 spin_lock_bh(&ipvlan->addrs_lock); 881 if (ipvlan_addr_busy(ipvlan->port, ip6_addr, true)) 882 netif_err(ipvlan, ifup, ipvlan->dev, 883 "Failed to add IPv6=%pI6c addr for %s intf\n", 884 ip6_addr, ipvlan->dev->name); 885 else 886 ret = ipvlan_add_addr(ipvlan, ip6_addr, true); 887 spin_unlock_bh(&ipvlan->addrs_lock); 888 return ret; 889 } 890 891 static void ipvlan_del_addr6(struct ipvl_dev *ipvlan, struct in6_addr *ip6_addr) 892 { 893 return ipvlan_del_addr(ipvlan, ip6_addr, true); 894 } 895 896 static int ipvlan_addr6_event(struct notifier_block *unused, 897 unsigned long event, void *ptr) 898 { 899 struct inet6_ifaddr *if6 = (struct inet6_ifaddr *)ptr; 900 struct net_device *dev = (struct net_device *)if6->idev->dev; 901 struct ipvl_dev *ipvlan = netdev_priv(dev); 902 903 if (!ipvlan_is_valid_dev(dev)) 904 return NOTIFY_DONE; 905 906 switch (event) { 907 case NETDEV_UP: 908 if (ipvlan_add_addr6(ipvlan, &if6->addr)) 909 return NOTIFY_BAD; 910 break; 911 912 case NETDEV_DOWN: 913 ipvlan_del_addr6(ipvlan, &if6->addr); 914 break; 915 } 916 917 return NOTIFY_OK; 918 } 919 920 static int ipvlan_addr6_validator_event(struct notifier_block *unused, 921 unsigned long event, void *ptr) 922 { 923 struct in6_validator_info *i6vi = (struct in6_validator_info *)ptr; 924 struct net_device *dev = (struct net_device *)i6vi->i6vi_dev->dev; 925 struct ipvl_dev *ipvlan = netdev_priv(dev); 926 927 if (!ipvlan_is_valid_dev(dev)) 928 return NOTIFY_DONE; 929 930 switch (event) { 931 case NETDEV_UP: 932 if (ipvlan_addr_busy(ipvlan->port, &i6vi->i6vi_addr, true)) { 933 NL_SET_ERR_MSG(i6vi->extack, 934 "Address already assigned to an ipvlan device"); 935 return notifier_from_errno(-EADDRINUSE); 936 } 937 break; 938 } 939 940 return NOTIFY_OK; 941 } 942 #endif 943 944 static int ipvlan_add_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) 945 { 946 int ret = -EINVAL; 947 948 spin_lock_bh(&ipvlan->addrs_lock); 949 if (ipvlan_addr_busy(ipvlan->port, ip4_addr, false)) 950 netif_err(ipvlan, ifup, ipvlan->dev, 951 "Failed to add IPv4=%pI4 on %s intf.\n", 952 ip4_addr, ipvlan->dev->name); 953 else 954 ret = ipvlan_add_addr(ipvlan, ip4_addr, false); 955 spin_unlock_bh(&ipvlan->addrs_lock); 956 return ret; 957 } 958 959 static void ipvlan_del_addr4(struct ipvl_dev *ipvlan, struct in_addr *ip4_addr) 960 { 961 return ipvlan_del_addr(ipvlan, ip4_addr, false); 962 } 963 964 static int ipvlan_addr4_event(struct notifier_block *unused, 965 unsigned long event, void *ptr) 966 { 967 struct in_ifaddr *if4 = (struct in_ifaddr *)ptr; 968 struct net_device *dev = (struct net_device *)if4->ifa_dev->dev; 969 struct ipvl_dev *ipvlan = netdev_priv(dev); 970 struct in_addr ip4_addr; 971 972 if (!ipvlan_is_valid_dev(dev)) 973 return NOTIFY_DONE; 974 975 switch (event) { 976 case NETDEV_UP: 977 ip4_addr.s_addr = if4->ifa_address; 978 if (ipvlan_add_addr4(ipvlan, &ip4_addr)) 979 return NOTIFY_BAD; 980 break; 981 982 case NETDEV_DOWN: 983 ip4_addr.s_addr = if4->ifa_address; 984 ipvlan_del_addr4(ipvlan, &ip4_addr); 985 break; 986 } 987 988 return NOTIFY_OK; 989 } 990 991 static int ipvlan_addr4_validator_event(struct notifier_block *unused, 992 unsigned long event, void *ptr) 993 { 994 struct in_validator_info *ivi = (struct in_validator_info *)ptr; 995 struct net_device *dev = (struct net_device *)ivi->ivi_dev->dev; 996 struct ipvl_dev *ipvlan = netdev_priv(dev); 997 998 if (!ipvlan_is_valid_dev(dev)) 999 return NOTIFY_DONE; 1000 1001 switch (event) { 1002 case NETDEV_UP: 1003 if (ipvlan_addr_busy(ipvlan->port, &ivi->ivi_addr, false)) { 1004 NL_SET_ERR_MSG(ivi->extack, 1005 "Address already assigned to an ipvlan device"); 1006 return notifier_from_errno(-EADDRINUSE); 1007 } 1008 break; 1009 } 1010 1011 return NOTIFY_OK; 1012 } 1013 1014 static struct notifier_block ipvlan_addr4_notifier_block __read_mostly = { 1015 .notifier_call = ipvlan_addr4_event, 1016 }; 1017 1018 static struct notifier_block ipvlan_addr4_vtor_notifier_block __read_mostly = { 1019 .notifier_call = ipvlan_addr4_validator_event, 1020 }; 1021 1022 static struct notifier_block ipvlan_notifier_block __read_mostly = { 1023 .notifier_call = ipvlan_device_event, 1024 }; 1025 1026 #if IS_ENABLED(CONFIG_IPV6) 1027 static struct notifier_block ipvlan_addr6_notifier_block __read_mostly = { 1028 .notifier_call = ipvlan_addr6_event, 1029 }; 1030 1031 static struct notifier_block ipvlan_addr6_vtor_notifier_block __read_mostly = { 1032 .notifier_call = ipvlan_addr6_validator_event, 1033 }; 1034 #endif 1035 1036 static int __init ipvlan_init_module(void) 1037 { 1038 int err; 1039 1040 ipvlan_init_secret(); 1041 register_netdevice_notifier(&ipvlan_notifier_block); 1042 #if IS_ENABLED(CONFIG_IPV6) 1043 register_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1044 register_inet6addr_validator_notifier( 1045 &ipvlan_addr6_vtor_notifier_block); 1046 #endif 1047 register_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1048 register_inetaddr_validator_notifier(&ipvlan_addr4_vtor_notifier_block); 1049 1050 err = ipvlan_l3s_init(); 1051 if (err < 0) 1052 goto error; 1053 1054 err = ipvlan_link_register(&ipvlan_link_ops); 1055 if (err < 0) { 1056 ipvlan_l3s_cleanup(); 1057 goto error; 1058 } 1059 1060 return 0; 1061 error: 1062 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1063 unregister_inetaddr_validator_notifier( 1064 &ipvlan_addr4_vtor_notifier_block); 1065 #if IS_ENABLED(CONFIG_IPV6) 1066 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1067 unregister_inet6addr_validator_notifier( 1068 &ipvlan_addr6_vtor_notifier_block); 1069 #endif 1070 unregister_netdevice_notifier(&ipvlan_notifier_block); 1071 return err; 1072 } 1073 1074 static void __exit ipvlan_cleanup_module(void) 1075 { 1076 rtnl_link_unregister(&ipvlan_link_ops); 1077 ipvlan_l3s_cleanup(); 1078 unregister_netdevice_notifier(&ipvlan_notifier_block); 1079 unregister_inetaddr_notifier(&ipvlan_addr4_notifier_block); 1080 unregister_inetaddr_validator_notifier( 1081 &ipvlan_addr4_vtor_notifier_block); 1082 #if IS_ENABLED(CONFIG_IPV6) 1083 unregister_inet6addr_notifier(&ipvlan_addr6_notifier_block); 1084 unregister_inet6addr_validator_notifier( 1085 &ipvlan_addr6_vtor_notifier_block); 1086 #endif 1087 } 1088 1089 module_init(ipvlan_init_module); 1090 module_exit(ipvlan_cleanup_module); 1091 1092 MODULE_LICENSE("GPL"); 1093 MODULE_AUTHOR("Mahesh Bandewar <maheshb@google.com>"); 1094 MODULE_DESCRIPTION("Driver for L3 (IPv6/IPv4) based VLANs"); 1095 MODULE_ALIAS_RTNL_LINK("ipvlan"); 1096