1 /* 2 * INET 802.1Q VLAN 3 * Ethernet-type device handling. 4 * 5 * Authors: Ben Greear <greearb@candelatech.com> 6 * Please send support related email to: netdev@vger.kernel.org 7 * VLAN Home Page: http://www.candelatech.com/~greear/vlan.html 8 * 9 * Fixes: 10 * Fix for packet capture - Nick Eggleston <nick@dccinc.com>; 11 * Add HW acceleration hooks - David S. Miller <davem@redhat.com>; 12 * Correct all the locking - David S. Miller <davem@redhat.com>; 13 * Use hash table for VLAN groups - David S. Miller <davem@redhat.com> 14 * 15 * This program is free software; you can redistribute it and/or 16 * modify it under the terms of the GNU General Public License 17 * as published by the Free Software Foundation; either version 18 * 2 of the License, or (at your option) any later version. 19 */ 20 21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 22 23 #include <linux/capability.h> 24 #include <linux/module.h> 25 #include <linux/netdevice.h> 26 #include <linux/skbuff.h> 27 #include <linux/slab.h> 28 #include <linux/init.h> 29 #include <linux/rculist.h> 30 #include <net/p8022.h> 31 #include <net/arp.h> 32 #include <linux/rtnetlink.h> 33 #include <linux/notifier.h> 34 #include <net/rtnetlink.h> 35 #include <net/net_namespace.h> 36 #include <net/netns/generic.h> 37 #include <asm/uaccess.h> 38 39 #include <linux/if_vlan.h> 40 #include "vlan.h" 41 #include "vlanproc.h" 42 43 #define DRV_VERSION "1.8" 44 45 /* Global VLAN variables */ 46 47 int vlan_net_id __read_mostly; 48 49 const char vlan_fullname[] = "802.1Q VLAN Support"; 50 const char vlan_version[] = DRV_VERSION; 51 52 /* End of global variables definitions. */ 53 54 static int vlan_group_prealloc_vid(struct vlan_group *vg, u16 vlan_id) 55 { 56 struct net_device **array; 57 unsigned int size; 58 59 ASSERT_RTNL(); 60 61 array = vg->vlan_devices_arrays[vlan_id / VLAN_GROUP_ARRAY_PART_LEN]; 62 if (array != NULL) 63 return 0; 64 65 size = sizeof(struct net_device *) * VLAN_GROUP_ARRAY_PART_LEN; 66 array = kzalloc(size, GFP_KERNEL); 67 if (array == NULL) 68 return -ENOBUFS; 69 70 vg->vlan_devices_arrays[vlan_id / VLAN_GROUP_ARRAY_PART_LEN] = array; 71 return 0; 72 } 73 74 void unregister_vlan_dev(struct net_device *dev, struct list_head *head) 75 { 76 struct vlan_dev_priv *vlan = vlan_dev_priv(dev); 77 struct net_device *real_dev = vlan->real_dev; 78 struct vlan_info *vlan_info; 79 struct vlan_group *grp; 80 u16 vlan_id = vlan->vlan_id; 81 82 ASSERT_RTNL(); 83 84 vlan_info = rtnl_dereference(real_dev->vlan_info); 85 BUG_ON(!vlan_info); 86 87 grp = &vlan_info->grp; 88 89 /* Take it out of our own structures, but be sure to interlock with 90 * HW accelerating devices or SW vlan input packet processing if 91 * VLAN is not 0 (leave it there for 802.1p). 92 */ 93 if (vlan_id) 94 vlan_vid_del(real_dev, vlan_id); 95 96 grp->nr_vlan_devs--; 97 98 if (vlan->flags & VLAN_FLAG_GVRP) 99 vlan_gvrp_request_leave(dev); 100 101 vlan_group_set_device(grp, vlan_id, NULL); 102 /* Because unregister_netdevice_queue() makes sure at least one rcu 103 * grace period is respected before device freeing, 104 * we dont need to call synchronize_net() here. 105 */ 106 unregister_netdevice_queue(dev, head); 107 108 if (grp->nr_vlan_devs == 0) 109 vlan_gvrp_uninit_applicant(real_dev); 110 111 /* Get rid of the vlan's reference to real_dev */ 112 dev_put(real_dev); 113 } 114 115 int vlan_check_real_dev(struct net_device *real_dev, u16 vlan_id) 116 { 117 const char *name = real_dev->name; 118 const struct net_device_ops *ops = real_dev->netdev_ops; 119 120 if (real_dev->features & NETIF_F_VLAN_CHALLENGED) { 121 pr_info("VLANs not supported on %s\n", name); 122 return -EOPNOTSUPP; 123 } 124 125 if ((real_dev->features & NETIF_F_HW_VLAN_FILTER) && 126 (!ops->ndo_vlan_rx_add_vid || !ops->ndo_vlan_rx_kill_vid)) { 127 pr_info("Device %s has buggy VLAN hw accel\n", name); 128 return -EOPNOTSUPP; 129 } 130 131 if (vlan_find_dev(real_dev, vlan_id) != NULL) 132 return -EEXIST; 133 134 return 0; 135 } 136 137 int register_vlan_dev(struct net_device *dev) 138 { 139 struct vlan_dev_priv *vlan = vlan_dev_priv(dev); 140 struct net_device *real_dev = vlan->real_dev; 141 u16 vlan_id = vlan->vlan_id; 142 struct vlan_info *vlan_info; 143 struct vlan_group *grp; 144 int err; 145 146 err = vlan_vid_add(real_dev, vlan_id); 147 if (err) 148 return err; 149 150 vlan_info = rtnl_dereference(real_dev->vlan_info); 151 /* vlan_info should be there now. vlan_vid_add took care of it */ 152 BUG_ON(!vlan_info); 153 154 grp = &vlan_info->grp; 155 if (grp->nr_vlan_devs == 0) { 156 err = vlan_gvrp_init_applicant(real_dev); 157 if (err < 0) 158 goto out_vid_del; 159 } 160 161 err = vlan_group_prealloc_vid(grp, vlan_id); 162 if (err < 0) 163 goto out_uninit_applicant; 164 165 err = register_netdevice(dev); 166 if (err < 0) 167 goto out_uninit_applicant; 168 169 /* Account for reference in struct vlan_dev_priv */ 170 dev_hold(real_dev); 171 172 netif_stacked_transfer_operstate(real_dev, dev); 173 linkwatch_fire_event(dev); /* _MUST_ call rfc2863_policy() */ 174 175 /* So, got the sucker initialized, now lets place 176 * it into our local structure. 177 */ 178 vlan_group_set_device(grp, vlan_id, dev); 179 grp->nr_vlan_devs++; 180 181 return 0; 182 183 out_uninit_applicant: 184 if (grp->nr_vlan_devs == 0) 185 vlan_gvrp_uninit_applicant(real_dev); 186 out_vid_del: 187 vlan_vid_del(real_dev, vlan_id); 188 return err; 189 } 190 191 /* Attach a VLAN device to a mac address (ie Ethernet Card). 192 * Returns 0 if the device was created or a negative error code otherwise. 193 */ 194 static int register_vlan_device(struct net_device *real_dev, u16 vlan_id) 195 { 196 struct net_device *new_dev; 197 struct net *net = dev_net(real_dev); 198 struct vlan_net *vn = net_generic(net, vlan_net_id); 199 char name[IFNAMSIZ]; 200 int err; 201 202 if (vlan_id >= VLAN_VID_MASK) 203 return -ERANGE; 204 205 err = vlan_check_real_dev(real_dev, vlan_id); 206 if (err < 0) 207 return err; 208 209 /* Gotta set up the fields for the device. */ 210 switch (vn->name_type) { 211 case VLAN_NAME_TYPE_RAW_PLUS_VID: 212 /* name will look like: eth1.0005 */ 213 snprintf(name, IFNAMSIZ, "%s.%.4i", real_dev->name, vlan_id); 214 break; 215 case VLAN_NAME_TYPE_PLUS_VID_NO_PAD: 216 /* Put our vlan.VID in the name. 217 * Name will look like: vlan5 218 */ 219 snprintf(name, IFNAMSIZ, "vlan%i", vlan_id); 220 break; 221 case VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD: 222 /* Put our vlan.VID in the name. 223 * Name will look like: eth0.5 224 */ 225 snprintf(name, IFNAMSIZ, "%s.%i", real_dev->name, vlan_id); 226 break; 227 case VLAN_NAME_TYPE_PLUS_VID: 228 /* Put our vlan.VID in the name. 229 * Name will look like: vlan0005 230 */ 231 default: 232 snprintf(name, IFNAMSIZ, "vlan%.4i", vlan_id); 233 } 234 235 new_dev = alloc_netdev(sizeof(struct vlan_dev_priv), name, vlan_setup); 236 237 if (new_dev == NULL) 238 return -ENOBUFS; 239 240 dev_net_set(new_dev, net); 241 /* need 4 bytes for extra VLAN header info, 242 * hope the underlying device can handle it. 243 */ 244 new_dev->mtu = real_dev->mtu; 245 new_dev->priv_flags |= (real_dev->priv_flags & IFF_UNICAST_FLT); 246 247 vlan_dev_priv(new_dev)->vlan_id = vlan_id; 248 vlan_dev_priv(new_dev)->real_dev = real_dev; 249 vlan_dev_priv(new_dev)->dent = NULL; 250 vlan_dev_priv(new_dev)->flags = VLAN_FLAG_REORDER_HDR; 251 252 new_dev->rtnl_link_ops = &vlan_link_ops; 253 err = register_vlan_dev(new_dev); 254 if (err < 0) 255 goto out_free_newdev; 256 257 return 0; 258 259 out_free_newdev: 260 free_netdev(new_dev); 261 return err; 262 } 263 264 static void vlan_sync_address(struct net_device *dev, 265 struct net_device *vlandev) 266 { 267 struct vlan_dev_priv *vlan = vlan_dev_priv(vlandev); 268 269 /* May be called without an actual change */ 270 if (ether_addr_equal(vlan->real_dev_addr, dev->dev_addr)) 271 return; 272 273 /* vlan address was different from the old address and is equal to 274 * the new address */ 275 if (!ether_addr_equal(vlandev->dev_addr, vlan->real_dev_addr) && 276 ether_addr_equal(vlandev->dev_addr, dev->dev_addr)) 277 dev_uc_del(dev, vlandev->dev_addr); 278 279 /* vlan address was equal to the old address and is different from 280 * the new address */ 281 if (ether_addr_equal(vlandev->dev_addr, vlan->real_dev_addr) && 282 !ether_addr_equal(vlandev->dev_addr, dev->dev_addr)) 283 dev_uc_add(dev, vlandev->dev_addr); 284 285 memcpy(vlan->real_dev_addr, dev->dev_addr, ETH_ALEN); 286 } 287 288 static void vlan_transfer_features(struct net_device *dev, 289 struct net_device *vlandev) 290 { 291 vlandev->gso_max_size = dev->gso_max_size; 292 293 if (dev->features & NETIF_F_HW_VLAN_TX) 294 vlandev->hard_header_len = dev->hard_header_len; 295 else 296 vlandev->hard_header_len = dev->hard_header_len + VLAN_HLEN; 297 298 #if IS_ENABLED(CONFIG_FCOE) 299 vlandev->fcoe_ddp_xid = dev->fcoe_ddp_xid; 300 #endif 301 302 netdev_update_features(vlandev); 303 } 304 305 static void __vlan_device_event(struct net_device *dev, unsigned long event) 306 { 307 switch (event) { 308 case NETDEV_CHANGENAME: 309 vlan_proc_rem_dev(dev); 310 if (vlan_proc_add_dev(dev) < 0) 311 pr_warn("failed to change proc name for %s\n", 312 dev->name); 313 break; 314 case NETDEV_REGISTER: 315 if (vlan_proc_add_dev(dev) < 0) 316 pr_warn("failed to add proc entry for %s\n", dev->name); 317 break; 318 case NETDEV_UNREGISTER: 319 vlan_proc_rem_dev(dev); 320 break; 321 } 322 } 323 324 static int vlan_device_event(struct notifier_block *unused, unsigned long event, 325 void *ptr) 326 { 327 struct net_device *dev = ptr; 328 struct vlan_group *grp; 329 struct vlan_info *vlan_info; 330 int i, flgs; 331 struct net_device *vlandev; 332 struct vlan_dev_priv *vlan; 333 LIST_HEAD(list); 334 335 if (is_vlan_dev(dev)) 336 __vlan_device_event(dev, event); 337 338 if ((event == NETDEV_UP) && 339 (dev->features & NETIF_F_HW_VLAN_FILTER)) { 340 pr_info("adding VLAN 0 to HW filter on device %s\n", 341 dev->name); 342 vlan_vid_add(dev, 0); 343 } 344 345 vlan_info = rtnl_dereference(dev->vlan_info); 346 if (!vlan_info) 347 goto out; 348 grp = &vlan_info->grp; 349 350 /* It is OK that we do not hold the group lock right now, 351 * as we run under the RTNL lock. 352 */ 353 354 switch (event) { 355 case NETDEV_CHANGE: 356 /* Propagate real device state to vlan devices */ 357 for (i = 0; i < VLAN_N_VID; i++) { 358 vlandev = vlan_group_get_device(grp, i); 359 if (!vlandev) 360 continue; 361 362 netif_stacked_transfer_operstate(dev, vlandev); 363 } 364 break; 365 366 case NETDEV_CHANGEADDR: 367 /* Adjust unicast filters on underlying device */ 368 for (i = 0; i < VLAN_N_VID; i++) { 369 vlandev = vlan_group_get_device(grp, i); 370 if (!vlandev) 371 continue; 372 373 flgs = vlandev->flags; 374 if (!(flgs & IFF_UP)) 375 continue; 376 377 vlan_sync_address(dev, vlandev); 378 } 379 break; 380 381 case NETDEV_CHANGEMTU: 382 for (i = 0; i < VLAN_N_VID; i++) { 383 vlandev = vlan_group_get_device(grp, i); 384 if (!vlandev) 385 continue; 386 387 if (vlandev->mtu <= dev->mtu) 388 continue; 389 390 dev_set_mtu(vlandev, dev->mtu); 391 } 392 break; 393 394 case NETDEV_FEAT_CHANGE: 395 /* Propagate device features to underlying device */ 396 for (i = 0; i < VLAN_N_VID; i++) { 397 vlandev = vlan_group_get_device(grp, i); 398 if (!vlandev) 399 continue; 400 401 vlan_transfer_features(dev, vlandev); 402 } 403 404 break; 405 406 case NETDEV_DOWN: 407 if (dev->features & NETIF_F_HW_VLAN_FILTER) 408 vlan_vid_del(dev, 0); 409 410 /* Put all VLANs for this dev in the down state too. */ 411 for (i = 0; i < VLAN_N_VID; i++) { 412 vlandev = vlan_group_get_device(grp, i); 413 if (!vlandev) 414 continue; 415 416 flgs = vlandev->flags; 417 if (!(flgs & IFF_UP)) 418 continue; 419 420 vlan = vlan_dev_priv(vlandev); 421 if (!(vlan->flags & VLAN_FLAG_LOOSE_BINDING)) 422 dev_change_flags(vlandev, flgs & ~IFF_UP); 423 netif_stacked_transfer_operstate(dev, vlandev); 424 } 425 break; 426 427 case NETDEV_UP: 428 /* Put all VLANs for this dev in the up state too. */ 429 for (i = 0; i < VLAN_N_VID; i++) { 430 vlandev = vlan_group_get_device(grp, i); 431 if (!vlandev) 432 continue; 433 434 flgs = vlandev->flags; 435 if (flgs & IFF_UP) 436 continue; 437 438 vlan = vlan_dev_priv(vlandev); 439 if (!(vlan->flags & VLAN_FLAG_LOOSE_BINDING)) 440 dev_change_flags(vlandev, flgs | IFF_UP); 441 netif_stacked_transfer_operstate(dev, vlandev); 442 } 443 break; 444 445 case NETDEV_UNREGISTER: 446 /* twiddle thumbs on netns device moves */ 447 if (dev->reg_state != NETREG_UNREGISTERING) 448 break; 449 450 for (i = 0; i < VLAN_N_VID; i++) { 451 vlandev = vlan_group_get_device(grp, i); 452 if (!vlandev) 453 continue; 454 455 /* removal of last vid destroys vlan_info, abort 456 * afterwards */ 457 if (vlan_info->nr_vids == 1) 458 i = VLAN_N_VID; 459 460 unregister_vlan_dev(vlandev, &list); 461 } 462 unregister_netdevice_many(&list); 463 break; 464 465 case NETDEV_PRE_TYPE_CHANGE: 466 /* Forbid underlaying device to change its type. */ 467 if (vlan_uses_dev(dev)) 468 return NOTIFY_BAD; 469 break; 470 471 case NETDEV_NOTIFY_PEERS: 472 case NETDEV_BONDING_FAILOVER: 473 /* Propagate to vlan devices */ 474 for (i = 0; i < VLAN_N_VID; i++) { 475 vlandev = vlan_group_get_device(grp, i); 476 if (!vlandev) 477 continue; 478 479 call_netdevice_notifiers(event, vlandev); 480 } 481 break; 482 } 483 484 out: 485 return NOTIFY_DONE; 486 } 487 488 static struct notifier_block vlan_notifier_block __read_mostly = { 489 .notifier_call = vlan_device_event, 490 }; 491 492 /* 493 * VLAN IOCTL handler. 494 * o execute requested action or pass command to the device driver 495 * arg is really a struct vlan_ioctl_args __user *. 496 */ 497 static int vlan_ioctl_handler(struct net *net, void __user *arg) 498 { 499 int err; 500 struct vlan_ioctl_args args; 501 struct net_device *dev = NULL; 502 503 if (copy_from_user(&args, arg, sizeof(struct vlan_ioctl_args))) 504 return -EFAULT; 505 506 /* Null terminate this sucker, just in case. */ 507 args.device1[23] = 0; 508 args.u.device2[23] = 0; 509 510 rtnl_lock(); 511 512 switch (args.cmd) { 513 case SET_VLAN_INGRESS_PRIORITY_CMD: 514 case SET_VLAN_EGRESS_PRIORITY_CMD: 515 case SET_VLAN_FLAG_CMD: 516 case ADD_VLAN_CMD: 517 case DEL_VLAN_CMD: 518 case GET_VLAN_REALDEV_NAME_CMD: 519 case GET_VLAN_VID_CMD: 520 err = -ENODEV; 521 dev = __dev_get_by_name(net, args.device1); 522 if (!dev) 523 goto out; 524 525 err = -EINVAL; 526 if (args.cmd != ADD_VLAN_CMD && !is_vlan_dev(dev)) 527 goto out; 528 } 529 530 switch (args.cmd) { 531 case SET_VLAN_INGRESS_PRIORITY_CMD: 532 err = -EPERM; 533 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 534 break; 535 vlan_dev_set_ingress_priority(dev, 536 args.u.skb_priority, 537 args.vlan_qos); 538 err = 0; 539 break; 540 541 case SET_VLAN_EGRESS_PRIORITY_CMD: 542 err = -EPERM; 543 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 544 break; 545 err = vlan_dev_set_egress_priority(dev, 546 args.u.skb_priority, 547 args.vlan_qos); 548 break; 549 550 case SET_VLAN_FLAG_CMD: 551 err = -EPERM; 552 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 553 break; 554 err = vlan_dev_change_flags(dev, 555 args.vlan_qos ? args.u.flag : 0, 556 args.u.flag); 557 break; 558 559 case SET_VLAN_NAME_TYPE_CMD: 560 err = -EPERM; 561 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 562 break; 563 if ((args.u.name_type >= 0) && 564 (args.u.name_type < VLAN_NAME_TYPE_HIGHEST)) { 565 struct vlan_net *vn; 566 567 vn = net_generic(net, vlan_net_id); 568 vn->name_type = args.u.name_type; 569 err = 0; 570 } else { 571 err = -EINVAL; 572 } 573 break; 574 575 case ADD_VLAN_CMD: 576 err = -EPERM; 577 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 578 break; 579 err = register_vlan_device(dev, args.u.VID); 580 break; 581 582 case DEL_VLAN_CMD: 583 err = -EPERM; 584 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 585 break; 586 unregister_vlan_dev(dev, NULL); 587 err = 0; 588 break; 589 590 case GET_VLAN_REALDEV_NAME_CMD: 591 err = 0; 592 vlan_dev_get_realdev_name(dev, args.u.device2); 593 if (copy_to_user(arg, &args, 594 sizeof(struct vlan_ioctl_args))) 595 err = -EFAULT; 596 break; 597 598 case GET_VLAN_VID_CMD: 599 err = 0; 600 args.u.VID = vlan_dev_vlan_id(dev); 601 if (copy_to_user(arg, &args, 602 sizeof(struct vlan_ioctl_args))) 603 err = -EFAULT; 604 break; 605 606 default: 607 err = -EOPNOTSUPP; 608 break; 609 } 610 out: 611 rtnl_unlock(); 612 return err; 613 } 614 615 static int __net_init vlan_init_net(struct net *net) 616 { 617 struct vlan_net *vn = net_generic(net, vlan_net_id); 618 int err; 619 620 vn->name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD; 621 622 err = vlan_proc_init(net); 623 624 return err; 625 } 626 627 static void __net_exit vlan_exit_net(struct net *net) 628 { 629 vlan_proc_cleanup(net); 630 } 631 632 static struct pernet_operations vlan_net_ops = { 633 .init = vlan_init_net, 634 .exit = vlan_exit_net, 635 .id = &vlan_net_id, 636 .size = sizeof(struct vlan_net), 637 }; 638 639 static int __init vlan_proto_init(void) 640 { 641 int err; 642 643 pr_info("%s v%s\n", vlan_fullname, vlan_version); 644 645 err = register_pernet_subsys(&vlan_net_ops); 646 if (err < 0) 647 goto err0; 648 649 err = register_netdevice_notifier(&vlan_notifier_block); 650 if (err < 0) 651 goto err2; 652 653 err = vlan_gvrp_init(); 654 if (err < 0) 655 goto err3; 656 657 err = vlan_netlink_init(); 658 if (err < 0) 659 goto err4; 660 661 vlan_ioctl_set(vlan_ioctl_handler); 662 return 0; 663 664 err4: 665 vlan_gvrp_uninit(); 666 err3: 667 unregister_netdevice_notifier(&vlan_notifier_block); 668 err2: 669 unregister_pernet_subsys(&vlan_net_ops); 670 err0: 671 return err; 672 } 673 674 static void __exit vlan_cleanup_module(void) 675 { 676 vlan_ioctl_set(NULL); 677 vlan_netlink_fini(); 678 679 unregister_netdevice_notifier(&vlan_notifier_block); 680 681 unregister_pernet_subsys(&vlan_net_ops); 682 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 683 684 vlan_gvrp_uninit(); 685 } 686 687 module_init(vlan_proto_init); 688 module_exit(vlan_cleanup_module); 689 690 MODULE_LICENSE("GPL"); 691 MODULE_VERSION(DRV_VERSION); 692