1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Routing netlink socket interface: protocol independent part. 7 * 8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 * 15 * Fixes: 16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong. 17 */ 18 19 #include <linux/bitops.h> 20 #include <linux/errno.h> 21 #include <linux/module.h> 22 #include <linux/types.h> 23 #include <linux/socket.h> 24 #include <linux/kernel.h> 25 #include <linux/timer.h> 26 #include <linux/string.h> 27 #include <linux/sockios.h> 28 #include <linux/net.h> 29 #include <linux/fcntl.h> 30 #include <linux/mm.h> 31 #include <linux/slab.h> 32 #include <linux/interrupt.h> 33 #include <linux/capability.h> 34 #include <linux/skbuff.h> 35 #include <linux/init.h> 36 #include <linux/security.h> 37 #include <linux/mutex.h> 38 #include <linux/if_addr.h> 39 #include <linux/if_bridge.h> 40 #include <linux/if_vlan.h> 41 #include <linux/pci.h> 42 #include <linux/etherdevice.h> 43 #include <linux/bpf.h> 44 45 #include <linux/uaccess.h> 46 47 #include <linux/inet.h> 48 #include <linux/netdevice.h> 49 #include <net/switchdev.h> 50 #include <net/ip.h> 51 #include <net/protocol.h> 52 #include <net/arp.h> 53 #include <net/route.h> 54 #include <net/udp.h> 55 #include <net/tcp.h> 56 #include <net/sock.h> 57 #include <net/pkt_sched.h> 58 #include <net/fib_rules.h> 59 #include <net/rtnetlink.h> 60 #include <net/net_namespace.h> 61 62 #define RTNL_MAX_TYPE 48 63 #define RTNL_SLAVE_MAX_TYPE 36 64 65 struct rtnl_link { 66 rtnl_doit_func doit; 67 rtnl_dumpit_func dumpit; 68 struct module *owner; 69 unsigned int flags; 70 struct rcu_head rcu; 71 }; 72 73 static DEFINE_MUTEX(rtnl_mutex); 74 75 void rtnl_lock(void) 76 { 77 mutex_lock(&rtnl_mutex); 78 } 79 EXPORT_SYMBOL(rtnl_lock); 80 81 int rtnl_lock_killable(void) 82 { 83 return mutex_lock_killable(&rtnl_mutex); 84 } 85 EXPORT_SYMBOL(rtnl_lock_killable); 86 87 static struct sk_buff *defer_kfree_skb_list; 88 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail) 89 { 90 if (head && tail) { 91 tail->next = defer_kfree_skb_list; 92 defer_kfree_skb_list = head; 93 } 94 } 95 EXPORT_SYMBOL(rtnl_kfree_skbs); 96 97 void __rtnl_unlock(void) 98 { 99 struct sk_buff *head = defer_kfree_skb_list; 100 101 defer_kfree_skb_list = NULL; 102 103 mutex_unlock(&rtnl_mutex); 104 105 while (head) { 106 struct sk_buff *next = head->next; 107 108 kfree_skb(head); 109 cond_resched(); 110 head = next; 111 } 112 } 113 114 void rtnl_unlock(void) 115 { 116 /* This fellow will unlock it for us. */ 117 netdev_run_todo(); 118 } 119 EXPORT_SYMBOL(rtnl_unlock); 120 121 int rtnl_trylock(void) 122 { 123 return mutex_trylock(&rtnl_mutex); 124 } 125 EXPORT_SYMBOL(rtnl_trylock); 126 127 int rtnl_is_locked(void) 128 { 129 return mutex_is_locked(&rtnl_mutex); 130 } 131 EXPORT_SYMBOL(rtnl_is_locked); 132 133 #ifdef CONFIG_PROVE_LOCKING 134 bool lockdep_rtnl_is_held(void) 135 { 136 return lockdep_is_held(&rtnl_mutex); 137 } 138 EXPORT_SYMBOL(lockdep_rtnl_is_held); 139 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 140 141 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 142 143 static inline int rtm_msgindex(int msgtype) 144 { 145 int msgindex = msgtype - RTM_BASE; 146 147 /* 148 * msgindex < 0 implies someone tried to register a netlink 149 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 150 * the message type has not been added to linux/rtnetlink.h 151 */ 152 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 153 154 return msgindex; 155 } 156 157 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype) 158 { 159 struct rtnl_link **tab; 160 161 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers)) 162 protocol = PF_UNSPEC; 163 164 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]); 165 if (!tab) 166 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]); 167 168 return tab[msgtype]; 169 } 170 171 static int rtnl_register_internal(struct module *owner, 172 int protocol, int msgtype, 173 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 174 unsigned int flags) 175 { 176 struct rtnl_link *link, *old; 177 struct rtnl_link __rcu **tab; 178 int msgindex; 179 int ret = -ENOBUFS; 180 181 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 182 msgindex = rtm_msgindex(msgtype); 183 184 rtnl_lock(); 185 tab = rtnl_msg_handlers[protocol]; 186 if (tab == NULL) { 187 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL); 188 if (!tab) 189 goto unlock; 190 191 /* ensures we see the 0 stores */ 192 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab); 193 } 194 195 old = rtnl_dereference(tab[msgindex]); 196 if (old) { 197 link = kmemdup(old, sizeof(*old), GFP_KERNEL); 198 if (!link) 199 goto unlock; 200 } else { 201 link = kzalloc(sizeof(*link), GFP_KERNEL); 202 if (!link) 203 goto unlock; 204 } 205 206 WARN_ON(link->owner && link->owner != owner); 207 link->owner = owner; 208 209 WARN_ON(doit && link->doit && link->doit != doit); 210 if (doit) 211 link->doit = doit; 212 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit); 213 if (dumpit) 214 link->dumpit = dumpit; 215 216 link->flags |= flags; 217 218 /* publish protocol:msgtype */ 219 rcu_assign_pointer(tab[msgindex], link); 220 ret = 0; 221 if (old) 222 kfree_rcu(old, rcu); 223 unlock: 224 rtnl_unlock(); 225 return ret; 226 } 227 228 /** 229 * rtnl_register_module - Register a rtnetlink message type 230 * 231 * @owner: module registering the hook (THIS_MODULE) 232 * @protocol: Protocol family or PF_UNSPEC 233 * @msgtype: rtnetlink message type 234 * @doit: Function pointer called for each request message 235 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 236 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions 237 * 238 * Like rtnl_register, but for use by removable modules. 239 */ 240 int rtnl_register_module(struct module *owner, 241 int protocol, int msgtype, 242 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 243 unsigned int flags) 244 { 245 return rtnl_register_internal(owner, protocol, msgtype, 246 doit, dumpit, flags); 247 } 248 EXPORT_SYMBOL_GPL(rtnl_register_module); 249 250 /** 251 * rtnl_register - Register a rtnetlink message type 252 * @protocol: Protocol family or PF_UNSPEC 253 * @msgtype: rtnetlink message type 254 * @doit: Function pointer called for each request message 255 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 256 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions 257 * 258 * Registers the specified function pointers (at least one of them has 259 * to be non-NULL) to be called whenever a request message for the 260 * specified protocol family and message type is received. 261 * 262 * The special protocol family PF_UNSPEC may be used to define fallback 263 * function pointers for the case when no entry for the specific protocol 264 * family exists. 265 */ 266 void rtnl_register(int protocol, int msgtype, 267 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 268 unsigned int flags) 269 { 270 int err; 271 272 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit, 273 flags); 274 if (err) 275 pr_err("Unable to register rtnetlink message handler, " 276 "protocol = %d, message type = %d\n", protocol, msgtype); 277 } 278 279 /** 280 * rtnl_unregister - Unregister a rtnetlink message type 281 * @protocol: Protocol family or PF_UNSPEC 282 * @msgtype: rtnetlink message type 283 * 284 * Returns 0 on success or a negative error code. 285 */ 286 int rtnl_unregister(int protocol, int msgtype) 287 { 288 struct rtnl_link **tab, *link; 289 int msgindex; 290 291 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 292 msgindex = rtm_msgindex(msgtype); 293 294 rtnl_lock(); 295 tab = rtnl_dereference(rtnl_msg_handlers[protocol]); 296 if (!tab) { 297 rtnl_unlock(); 298 return -ENOENT; 299 } 300 301 link = tab[msgindex]; 302 rcu_assign_pointer(tab[msgindex], NULL); 303 rtnl_unlock(); 304 305 kfree_rcu(link, rcu); 306 307 return 0; 308 } 309 EXPORT_SYMBOL_GPL(rtnl_unregister); 310 311 /** 312 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 313 * @protocol : Protocol family or PF_UNSPEC 314 * 315 * Identical to calling rtnl_unregster() for all registered message types 316 * of a certain protocol family. 317 */ 318 void rtnl_unregister_all(int protocol) 319 { 320 struct rtnl_link **tab, *link; 321 int msgindex; 322 323 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 324 325 rtnl_lock(); 326 tab = rtnl_msg_handlers[protocol]; 327 if (!tab) { 328 rtnl_unlock(); 329 return; 330 } 331 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL); 332 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) { 333 link = tab[msgindex]; 334 if (!link) 335 continue; 336 337 rcu_assign_pointer(tab[msgindex], NULL); 338 kfree_rcu(link, rcu); 339 } 340 rtnl_unlock(); 341 342 synchronize_net(); 343 344 kfree(tab); 345 } 346 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 347 348 static LIST_HEAD(link_ops); 349 350 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 351 { 352 const struct rtnl_link_ops *ops; 353 354 list_for_each_entry(ops, &link_ops, list) { 355 if (!strcmp(ops->kind, kind)) 356 return ops; 357 } 358 return NULL; 359 } 360 361 /** 362 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 363 * @ops: struct rtnl_link_ops * to register 364 * 365 * The caller must hold the rtnl_mutex. This function should be used 366 * by drivers that create devices during module initialization. It 367 * must be called before registering the devices. 368 * 369 * Returns 0 on success or a negative error code. 370 */ 371 int __rtnl_link_register(struct rtnl_link_ops *ops) 372 { 373 if (rtnl_link_ops_get(ops->kind)) 374 return -EEXIST; 375 376 /* The check for setup is here because if ops 377 * does not have that filled up, it is not possible 378 * to use the ops for creating device. So do not 379 * fill up dellink as well. That disables rtnl_dellink. 380 */ 381 if (ops->setup && !ops->dellink) 382 ops->dellink = unregister_netdevice_queue; 383 384 list_add_tail(&ops->list, &link_ops); 385 return 0; 386 } 387 EXPORT_SYMBOL_GPL(__rtnl_link_register); 388 389 /** 390 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 391 * @ops: struct rtnl_link_ops * to register 392 * 393 * Returns 0 on success or a negative error code. 394 */ 395 int rtnl_link_register(struct rtnl_link_ops *ops) 396 { 397 int err; 398 399 /* Sanity-check max sizes to avoid stack buffer overflow. */ 400 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE || 401 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)) 402 return -EINVAL; 403 404 rtnl_lock(); 405 err = __rtnl_link_register(ops); 406 rtnl_unlock(); 407 return err; 408 } 409 EXPORT_SYMBOL_GPL(rtnl_link_register); 410 411 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 412 { 413 struct net_device *dev; 414 LIST_HEAD(list_kill); 415 416 for_each_netdev(net, dev) { 417 if (dev->rtnl_link_ops == ops) 418 ops->dellink(dev, &list_kill); 419 } 420 unregister_netdevice_many(&list_kill); 421 } 422 423 /** 424 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 425 * @ops: struct rtnl_link_ops * to unregister 426 * 427 * The caller must hold the rtnl_mutex and guarantee net_namespace_list 428 * integrity (hold pernet_ops_rwsem for writing to close the race 429 * with setup_net() and cleanup_net()). 430 */ 431 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 432 { 433 struct net *net; 434 435 for_each_net(net) { 436 __rtnl_kill_links(net, ops); 437 } 438 list_del(&ops->list); 439 } 440 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 441 442 /* Return with the rtnl_lock held when there are no network 443 * devices unregistering in any network namespace. 444 */ 445 static void rtnl_lock_unregistering_all(void) 446 { 447 struct net *net; 448 bool unregistering; 449 DEFINE_WAIT_FUNC(wait, woken_wake_function); 450 451 add_wait_queue(&netdev_unregistering_wq, &wait); 452 for (;;) { 453 unregistering = false; 454 rtnl_lock(); 455 /* We held write locked pernet_ops_rwsem, and parallel 456 * setup_net() and cleanup_net() are not possible. 457 */ 458 for_each_net(net) { 459 if (net->dev_unreg_count > 0) { 460 unregistering = true; 461 break; 462 } 463 } 464 if (!unregistering) 465 break; 466 __rtnl_unlock(); 467 468 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); 469 } 470 remove_wait_queue(&netdev_unregistering_wq, &wait); 471 } 472 473 /** 474 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 475 * @ops: struct rtnl_link_ops * to unregister 476 */ 477 void rtnl_link_unregister(struct rtnl_link_ops *ops) 478 { 479 /* Close the race with setup_net() and cleanup_net() */ 480 down_write(&pernet_ops_rwsem); 481 rtnl_lock_unregistering_all(); 482 __rtnl_link_unregister(ops); 483 rtnl_unlock(); 484 up_write(&pernet_ops_rwsem); 485 } 486 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 487 488 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev) 489 { 490 struct net_device *master_dev; 491 const struct rtnl_link_ops *ops; 492 size_t size = 0; 493 494 rcu_read_lock(); 495 496 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev); 497 if (!master_dev) 498 goto out; 499 500 ops = master_dev->rtnl_link_ops; 501 if (!ops || !ops->get_slave_size) 502 goto out; 503 /* IFLA_INFO_SLAVE_DATA + nested data */ 504 size = nla_total_size(sizeof(struct nlattr)) + 505 ops->get_slave_size(master_dev, dev); 506 507 out: 508 rcu_read_unlock(); 509 return size; 510 } 511 512 static size_t rtnl_link_get_size(const struct net_device *dev) 513 { 514 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 515 size_t size; 516 517 if (!ops) 518 return 0; 519 520 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 521 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 522 523 if (ops->get_size) 524 /* IFLA_INFO_DATA + nested data */ 525 size += nla_total_size(sizeof(struct nlattr)) + 526 ops->get_size(dev); 527 528 if (ops->get_xstats_size) 529 /* IFLA_INFO_XSTATS */ 530 size += nla_total_size(ops->get_xstats_size(dev)); 531 532 size += rtnl_link_get_slave_info_data_size(dev); 533 534 return size; 535 } 536 537 static LIST_HEAD(rtnl_af_ops); 538 539 static const struct rtnl_af_ops *rtnl_af_lookup(const int family) 540 { 541 const struct rtnl_af_ops *ops; 542 543 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) { 544 if (ops->family == family) 545 return ops; 546 } 547 548 return NULL; 549 } 550 551 /** 552 * rtnl_af_register - Register rtnl_af_ops with rtnetlink. 553 * @ops: struct rtnl_af_ops * to register 554 * 555 * Returns 0 on success or a negative error code. 556 */ 557 void rtnl_af_register(struct rtnl_af_ops *ops) 558 { 559 rtnl_lock(); 560 list_add_tail_rcu(&ops->list, &rtnl_af_ops); 561 rtnl_unlock(); 562 } 563 EXPORT_SYMBOL_GPL(rtnl_af_register); 564 565 /** 566 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 567 * @ops: struct rtnl_af_ops * to unregister 568 */ 569 void rtnl_af_unregister(struct rtnl_af_ops *ops) 570 { 571 rtnl_lock(); 572 list_del_rcu(&ops->list); 573 rtnl_unlock(); 574 575 synchronize_rcu(); 576 } 577 EXPORT_SYMBOL_GPL(rtnl_af_unregister); 578 579 static size_t rtnl_link_get_af_size(const struct net_device *dev, 580 u32 ext_filter_mask) 581 { 582 struct rtnl_af_ops *af_ops; 583 size_t size; 584 585 /* IFLA_AF_SPEC */ 586 size = nla_total_size(sizeof(struct nlattr)); 587 588 rcu_read_lock(); 589 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 590 if (af_ops->get_link_af_size) { 591 /* AF_* + nested data */ 592 size += nla_total_size(sizeof(struct nlattr)) + 593 af_ops->get_link_af_size(dev, ext_filter_mask); 594 } 595 } 596 rcu_read_unlock(); 597 598 return size; 599 } 600 601 static bool rtnl_have_link_slave_info(const struct net_device *dev) 602 { 603 struct net_device *master_dev; 604 bool ret = false; 605 606 rcu_read_lock(); 607 608 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev); 609 if (master_dev && master_dev->rtnl_link_ops) 610 ret = true; 611 rcu_read_unlock(); 612 return ret; 613 } 614 615 static int rtnl_link_slave_info_fill(struct sk_buff *skb, 616 const struct net_device *dev) 617 { 618 struct net_device *master_dev; 619 const struct rtnl_link_ops *ops; 620 struct nlattr *slave_data; 621 int err; 622 623 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 624 if (!master_dev) 625 return 0; 626 ops = master_dev->rtnl_link_ops; 627 if (!ops) 628 return 0; 629 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0) 630 return -EMSGSIZE; 631 if (ops->fill_slave_info) { 632 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA); 633 if (!slave_data) 634 return -EMSGSIZE; 635 err = ops->fill_slave_info(skb, master_dev, dev); 636 if (err < 0) 637 goto err_cancel_slave_data; 638 nla_nest_end(skb, slave_data); 639 } 640 return 0; 641 642 err_cancel_slave_data: 643 nla_nest_cancel(skb, slave_data); 644 return err; 645 } 646 647 static int rtnl_link_info_fill(struct sk_buff *skb, 648 const struct net_device *dev) 649 { 650 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 651 struct nlattr *data; 652 int err; 653 654 if (!ops) 655 return 0; 656 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 657 return -EMSGSIZE; 658 if (ops->fill_xstats) { 659 err = ops->fill_xstats(skb, dev); 660 if (err < 0) 661 return err; 662 } 663 if (ops->fill_info) { 664 data = nla_nest_start(skb, IFLA_INFO_DATA); 665 if (data == NULL) 666 return -EMSGSIZE; 667 err = ops->fill_info(skb, dev); 668 if (err < 0) 669 goto err_cancel_data; 670 nla_nest_end(skb, data); 671 } 672 return 0; 673 674 err_cancel_data: 675 nla_nest_cancel(skb, data); 676 return err; 677 } 678 679 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 680 { 681 struct nlattr *linkinfo; 682 int err = -EMSGSIZE; 683 684 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 685 if (linkinfo == NULL) 686 goto out; 687 688 err = rtnl_link_info_fill(skb, dev); 689 if (err < 0) 690 goto err_cancel_link; 691 692 err = rtnl_link_slave_info_fill(skb, dev); 693 if (err < 0) 694 goto err_cancel_link; 695 696 nla_nest_end(skb, linkinfo); 697 return 0; 698 699 err_cancel_link: 700 nla_nest_cancel(skb, linkinfo); 701 out: 702 return err; 703 } 704 705 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo) 706 { 707 struct sock *rtnl = net->rtnl; 708 int err = 0; 709 710 NETLINK_CB(skb).dst_group = group; 711 if (echo) 712 refcount_inc(&skb->users); 713 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 714 if (echo) 715 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 716 return err; 717 } 718 719 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 720 { 721 struct sock *rtnl = net->rtnl; 722 723 return nlmsg_unicast(rtnl, skb, pid); 724 } 725 EXPORT_SYMBOL(rtnl_unicast); 726 727 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 728 struct nlmsghdr *nlh, gfp_t flags) 729 { 730 struct sock *rtnl = net->rtnl; 731 int report = 0; 732 733 if (nlh) 734 report = nlmsg_report(nlh); 735 736 nlmsg_notify(rtnl, skb, pid, group, report, flags); 737 } 738 EXPORT_SYMBOL(rtnl_notify); 739 740 void rtnl_set_sk_err(struct net *net, u32 group, int error) 741 { 742 struct sock *rtnl = net->rtnl; 743 744 netlink_set_err(rtnl, 0, group, error); 745 } 746 EXPORT_SYMBOL(rtnl_set_sk_err); 747 748 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 749 { 750 struct nlattr *mx; 751 int i, valid = 0; 752 753 mx = nla_nest_start(skb, RTA_METRICS); 754 if (mx == NULL) 755 return -ENOBUFS; 756 757 for (i = 0; i < RTAX_MAX; i++) { 758 if (metrics[i]) { 759 if (i == RTAX_CC_ALGO - 1) { 760 char tmp[TCP_CA_NAME_MAX], *name; 761 762 name = tcp_ca_get_name_by_key(metrics[i], tmp); 763 if (!name) 764 continue; 765 if (nla_put_string(skb, i + 1, name)) 766 goto nla_put_failure; 767 } else if (i == RTAX_FEATURES - 1) { 768 u32 user_features = metrics[i] & RTAX_FEATURE_MASK; 769 770 if (!user_features) 771 continue; 772 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK); 773 if (nla_put_u32(skb, i + 1, user_features)) 774 goto nla_put_failure; 775 } else { 776 if (nla_put_u32(skb, i + 1, metrics[i])) 777 goto nla_put_failure; 778 } 779 valid++; 780 } 781 } 782 783 if (!valid) { 784 nla_nest_cancel(skb, mx); 785 return 0; 786 } 787 788 return nla_nest_end(skb, mx); 789 790 nla_put_failure: 791 nla_nest_cancel(skb, mx); 792 return -EMSGSIZE; 793 } 794 EXPORT_SYMBOL(rtnetlink_put_metrics); 795 796 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 797 long expires, u32 error) 798 { 799 struct rta_cacheinfo ci = { 800 .rta_error = error, 801 .rta_id = id, 802 }; 803 804 if (dst) { 805 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse); 806 ci.rta_used = dst->__use; 807 ci.rta_clntref = atomic_read(&dst->__refcnt); 808 } 809 if (expires) { 810 unsigned long clock; 811 812 clock = jiffies_to_clock_t(abs(expires)); 813 clock = min_t(unsigned long, clock, INT_MAX); 814 ci.rta_expires = (expires > 0) ? clock : -clock; 815 } 816 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 817 } 818 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 819 820 static void set_operstate(struct net_device *dev, unsigned char transition) 821 { 822 unsigned char operstate = dev->operstate; 823 824 switch (transition) { 825 case IF_OPER_UP: 826 if ((operstate == IF_OPER_DORMANT || 827 operstate == IF_OPER_UNKNOWN) && 828 !netif_dormant(dev)) 829 operstate = IF_OPER_UP; 830 break; 831 832 case IF_OPER_DORMANT: 833 if (operstate == IF_OPER_UP || 834 operstate == IF_OPER_UNKNOWN) 835 operstate = IF_OPER_DORMANT; 836 break; 837 } 838 839 if (dev->operstate != operstate) { 840 write_lock_bh(&dev_base_lock); 841 dev->operstate = operstate; 842 write_unlock_bh(&dev_base_lock); 843 netdev_state_change(dev); 844 } 845 } 846 847 static unsigned int rtnl_dev_get_flags(const struct net_device *dev) 848 { 849 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) | 850 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI)); 851 } 852 853 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 854 const struct ifinfomsg *ifm) 855 { 856 unsigned int flags = ifm->ifi_flags; 857 858 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 859 if (ifm->ifi_change) 860 flags = (flags & ifm->ifi_change) | 861 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change); 862 863 return flags; 864 } 865 866 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 867 const struct rtnl_link_stats64 *b) 868 { 869 a->rx_packets = b->rx_packets; 870 a->tx_packets = b->tx_packets; 871 a->rx_bytes = b->rx_bytes; 872 a->tx_bytes = b->tx_bytes; 873 a->rx_errors = b->rx_errors; 874 a->tx_errors = b->tx_errors; 875 a->rx_dropped = b->rx_dropped; 876 a->tx_dropped = b->tx_dropped; 877 878 a->multicast = b->multicast; 879 a->collisions = b->collisions; 880 881 a->rx_length_errors = b->rx_length_errors; 882 a->rx_over_errors = b->rx_over_errors; 883 a->rx_crc_errors = b->rx_crc_errors; 884 a->rx_frame_errors = b->rx_frame_errors; 885 a->rx_fifo_errors = b->rx_fifo_errors; 886 a->rx_missed_errors = b->rx_missed_errors; 887 888 a->tx_aborted_errors = b->tx_aborted_errors; 889 a->tx_carrier_errors = b->tx_carrier_errors; 890 a->tx_fifo_errors = b->tx_fifo_errors; 891 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 892 a->tx_window_errors = b->tx_window_errors; 893 894 a->rx_compressed = b->rx_compressed; 895 a->tx_compressed = b->tx_compressed; 896 897 a->rx_nohandler = b->rx_nohandler; 898 } 899 900 /* All VF info */ 901 static inline int rtnl_vfinfo_size(const struct net_device *dev, 902 u32 ext_filter_mask) 903 { 904 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) { 905 int num_vfs = dev_num_vf(dev->dev.parent); 906 size_t size = nla_total_size(0); 907 size += num_vfs * 908 (nla_total_size(0) + 909 nla_total_size(sizeof(struct ifla_vf_mac)) + 910 nla_total_size(sizeof(struct ifla_vf_vlan)) + 911 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */ 912 nla_total_size(MAX_VLAN_LIST_LEN * 913 sizeof(struct ifla_vf_vlan_info)) + 914 nla_total_size(sizeof(struct ifla_vf_spoofchk)) + 915 nla_total_size(sizeof(struct ifla_vf_tx_rate)) + 916 nla_total_size(sizeof(struct ifla_vf_rate)) + 917 nla_total_size(sizeof(struct ifla_vf_link_state)) + 918 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) + 919 nla_total_size(0) + /* nest IFLA_VF_STATS */ 920 /* IFLA_VF_STATS_RX_PACKETS */ 921 nla_total_size_64bit(sizeof(__u64)) + 922 /* IFLA_VF_STATS_TX_PACKETS */ 923 nla_total_size_64bit(sizeof(__u64)) + 924 /* IFLA_VF_STATS_RX_BYTES */ 925 nla_total_size_64bit(sizeof(__u64)) + 926 /* IFLA_VF_STATS_TX_BYTES */ 927 nla_total_size_64bit(sizeof(__u64)) + 928 /* IFLA_VF_STATS_BROADCAST */ 929 nla_total_size_64bit(sizeof(__u64)) + 930 /* IFLA_VF_STATS_MULTICAST */ 931 nla_total_size_64bit(sizeof(__u64)) + 932 /* IFLA_VF_STATS_RX_DROPPED */ 933 nla_total_size_64bit(sizeof(__u64)) + 934 /* IFLA_VF_STATS_TX_DROPPED */ 935 nla_total_size_64bit(sizeof(__u64)) + 936 nla_total_size(sizeof(struct ifla_vf_trust))); 937 return size; 938 } else 939 return 0; 940 } 941 942 static size_t rtnl_port_size(const struct net_device *dev, 943 u32 ext_filter_mask) 944 { 945 size_t port_size = nla_total_size(4) /* PORT_VF */ 946 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 947 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 948 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 949 + nla_total_size(1) /* PROT_VDP_REQUEST */ 950 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 951 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 952 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 953 + port_size; 954 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 955 + port_size; 956 957 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 958 !(ext_filter_mask & RTEXT_FILTER_VF)) 959 return 0; 960 if (dev_num_vf(dev->dev.parent)) 961 return port_self_size + vf_ports_size + 962 vf_port_size * dev_num_vf(dev->dev.parent); 963 else 964 return port_self_size; 965 } 966 967 static size_t rtnl_xdp_size(void) 968 { 969 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */ 970 nla_total_size(1) + /* XDP_ATTACHED */ 971 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */ 972 nla_total_size(4); /* XDP_<mode>_PROG_ID */ 973 974 return xdp_size; 975 } 976 977 static noinline size_t if_nlmsg_size(const struct net_device *dev, 978 u32 ext_filter_mask) 979 { 980 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 981 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 982 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 983 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 984 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap)) 985 + nla_total_size(sizeof(struct rtnl_link_stats)) 986 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64)) 987 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 988 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 989 + nla_total_size(4) /* IFLA_TXQLEN */ 990 + nla_total_size(4) /* IFLA_WEIGHT */ 991 + nla_total_size(4) /* IFLA_MTU */ 992 + nla_total_size(4) /* IFLA_LINK */ 993 + nla_total_size(4) /* IFLA_MASTER */ 994 + nla_total_size(1) /* IFLA_CARRIER */ 995 + nla_total_size(4) /* IFLA_PROMISCUITY */ 996 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 997 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 998 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */ 999 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */ 1000 + nla_total_size(1) /* IFLA_OPERSTATE */ 1001 + nla_total_size(1) /* IFLA_LINKMODE */ 1002 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 1003 + nla_total_size(4) /* IFLA_LINK_NETNSID */ 1004 + nla_total_size(4) /* IFLA_GROUP */ 1005 + nla_total_size(ext_filter_mask 1006 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */ 1007 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */ 1008 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 1009 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */ 1010 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */ 1011 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */ 1012 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */ 1013 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */ 1014 + rtnl_xdp_size() /* IFLA_XDP */ 1015 + nla_total_size(4) /* IFLA_EVENT */ 1016 + nla_total_size(4) /* IFLA_NEW_NETNSID */ 1017 + nla_total_size(4) /* IFLA_NEW_IFINDEX */ 1018 + nla_total_size(1) /* IFLA_PROTO_DOWN */ 1019 + nla_total_size(4) /* IFLA_IF_NETNSID */ 1020 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */ 1021 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */ 1022 + nla_total_size(4) /* IFLA_MIN_MTU */ 1023 + nla_total_size(4) /* IFLA_MAX_MTU */ 1024 + 0; 1025 } 1026 1027 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 1028 { 1029 struct nlattr *vf_ports; 1030 struct nlattr *vf_port; 1031 int vf; 1032 int err; 1033 1034 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS); 1035 if (!vf_ports) 1036 return -EMSGSIZE; 1037 1038 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 1039 vf_port = nla_nest_start(skb, IFLA_VF_PORT); 1040 if (!vf_port) 1041 goto nla_put_failure; 1042 if (nla_put_u32(skb, IFLA_PORT_VF, vf)) 1043 goto nla_put_failure; 1044 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 1045 if (err == -EMSGSIZE) 1046 goto nla_put_failure; 1047 if (err) { 1048 nla_nest_cancel(skb, vf_port); 1049 continue; 1050 } 1051 nla_nest_end(skb, vf_port); 1052 } 1053 1054 nla_nest_end(skb, vf_ports); 1055 1056 return 0; 1057 1058 nla_put_failure: 1059 nla_nest_cancel(skb, vf_ports); 1060 return -EMSGSIZE; 1061 } 1062 1063 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 1064 { 1065 struct nlattr *port_self; 1066 int err; 1067 1068 port_self = nla_nest_start(skb, IFLA_PORT_SELF); 1069 if (!port_self) 1070 return -EMSGSIZE; 1071 1072 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 1073 if (err) { 1074 nla_nest_cancel(skb, port_self); 1075 return (err == -EMSGSIZE) ? err : 0; 1076 } 1077 1078 nla_nest_end(skb, port_self); 1079 1080 return 0; 1081 } 1082 1083 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev, 1084 u32 ext_filter_mask) 1085 { 1086 int err; 1087 1088 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 1089 !(ext_filter_mask & RTEXT_FILTER_VF)) 1090 return 0; 1091 1092 err = rtnl_port_self_fill(skb, dev); 1093 if (err) 1094 return err; 1095 1096 if (dev_num_vf(dev->dev.parent)) { 1097 err = rtnl_vf_ports_fill(skb, dev); 1098 if (err) 1099 return err; 1100 } 1101 1102 return 0; 1103 } 1104 1105 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev) 1106 { 1107 int err; 1108 struct netdev_phys_item_id ppid; 1109 1110 err = dev_get_phys_port_id(dev, &ppid); 1111 if (err) { 1112 if (err == -EOPNOTSUPP) 1113 return 0; 1114 return err; 1115 } 1116 1117 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id)) 1118 return -EMSGSIZE; 1119 1120 return 0; 1121 } 1122 1123 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev) 1124 { 1125 char name[IFNAMSIZ]; 1126 int err; 1127 1128 err = dev_get_phys_port_name(dev, name, sizeof(name)); 1129 if (err) { 1130 if (err == -EOPNOTSUPP) 1131 return 0; 1132 return err; 1133 } 1134 1135 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name)) 1136 return -EMSGSIZE; 1137 1138 return 0; 1139 } 1140 1141 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev) 1142 { 1143 int err; 1144 struct switchdev_attr attr = { 1145 .orig_dev = dev, 1146 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, 1147 .flags = SWITCHDEV_F_NO_RECURSE, 1148 }; 1149 1150 err = switchdev_port_attr_get(dev, &attr); 1151 if (err) { 1152 if (err == -EOPNOTSUPP) 1153 return 0; 1154 return err; 1155 } 1156 1157 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len, 1158 attr.u.ppid.id)) 1159 return -EMSGSIZE; 1160 1161 return 0; 1162 } 1163 1164 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb, 1165 struct net_device *dev) 1166 { 1167 struct rtnl_link_stats64 *sp; 1168 struct nlattr *attr; 1169 1170 attr = nla_reserve_64bit(skb, IFLA_STATS64, 1171 sizeof(struct rtnl_link_stats64), IFLA_PAD); 1172 if (!attr) 1173 return -EMSGSIZE; 1174 1175 sp = nla_data(attr); 1176 dev_get_stats(dev, sp); 1177 1178 attr = nla_reserve(skb, IFLA_STATS, 1179 sizeof(struct rtnl_link_stats)); 1180 if (!attr) 1181 return -EMSGSIZE; 1182 1183 copy_rtnl_link_stats(nla_data(attr), sp); 1184 1185 return 0; 1186 } 1187 1188 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb, 1189 struct net_device *dev, 1190 int vfs_num, 1191 struct nlattr *vfinfo) 1192 { 1193 struct ifla_vf_rss_query_en vf_rss_query_en; 1194 struct nlattr *vf, *vfstats, *vfvlanlist; 1195 struct ifla_vf_link_state vf_linkstate; 1196 struct ifla_vf_vlan_info vf_vlan_info; 1197 struct ifla_vf_spoofchk vf_spoofchk; 1198 struct ifla_vf_tx_rate vf_tx_rate; 1199 struct ifla_vf_stats vf_stats; 1200 struct ifla_vf_trust vf_trust; 1201 struct ifla_vf_vlan vf_vlan; 1202 struct ifla_vf_rate vf_rate; 1203 struct ifla_vf_mac vf_mac; 1204 struct ifla_vf_info ivi; 1205 1206 memset(&ivi, 0, sizeof(ivi)); 1207 1208 /* Not all SR-IOV capable drivers support the 1209 * spoofcheck and "RSS query enable" query. Preset to 1210 * -1 so the user space tool can detect that the driver 1211 * didn't report anything. 1212 */ 1213 ivi.spoofchk = -1; 1214 ivi.rss_query_en = -1; 1215 ivi.trusted = -1; 1216 /* The default value for VF link state is "auto" 1217 * IFLA_VF_LINK_STATE_AUTO which equals zero 1218 */ 1219 ivi.linkstate = 0; 1220 /* VLAN Protocol by default is 802.1Q */ 1221 ivi.vlan_proto = htons(ETH_P_8021Q); 1222 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi)) 1223 return 0; 1224 1225 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info)); 1226 1227 vf_mac.vf = 1228 vf_vlan.vf = 1229 vf_vlan_info.vf = 1230 vf_rate.vf = 1231 vf_tx_rate.vf = 1232 vf_spoofchk.vf = 1233 vf_linkstate.vf = 1234 vf_rss_query_en.vf = 1235 vf_trust.vf = ivi.vf; 1236 1237 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 1238 vf_vlan.vlan = ivi.vlan; 1239 vf_vlan.qos = ivi.qos; 1240 vf_vlan_info.vlan = ivi.vlan; 1241 vf_vlan_info.qos = ivi.qos; 1242 vf_vlan_info.vlan_proto = ivi.vlan_proto; 1243 vf_tx_rate.rate = ivi.max_tx_rate; 1244 vf_rate.min_tx_rate = ivi.min_tx_rate; 1245 vf_rate.max_tx_rate = ivi.max_tx_rate; 1246 vf_spoofchk.setting = ivi.spoofchk; 1247 vf_linkstate.link_state = ivi.linkstate; 1248 vf_rss_query_en.setting = ivi.rss_query_en; 1249 vf_trust.setting = ivi.trusted; 1250 vf = nla_nest_start(skb, IFLA_VF_INFO); 1251 if (!vf) 1252 goto nla_put_vfinfo_failure; 1253 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) || 1254 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) || 1255 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate), 1256 &vf_rate) || 1257 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), 1258 &vf_tx_rate) || 1259 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk), 1260 &vf_spoofchk) || 1261 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate), 1262 &vf_linkstate) || 1263 nla_put(skb, IFLA_VF_RSS_QUERY_EN, 1264 sizeof(vf_rss_query_en), 1265 &vf_rss_query_en) || 1266 nla_put(skb, IFLA_VF_TRUST, 1267 sizeof(vf_trust), &vf_trust)) 1268 goto nla_put_vf_failure; 1269 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST); 1270 if (!vfvlanlist) 1271 goto nla_put_vf_failure; 1272 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info), 1273 &vf_vlan_info)) { 1274 nla_nest_cancel(skb, vfvlanlist); 1275 goto nla_put_vf_failure; 1276 } 1277 nla_nest_end(skb, vfvlanlist); 1278 memset(&vf_stats, 0, sizeof(vf_stats)); 1279 if (dev->netdev_ops->ndo_get_vf_stats) 1280 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num, 1281 &vf_stats); 1282 vfstats = nla_nest_start(skb, IFLA_VF_STATS); 1283 if (!vfstats) 1284 goto nla_put_vf_failure; 1285 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS, 1286 vf_stats.rx_packets, IFLA_VF_STATS_PAD) || 1287 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS, 1288 vf_stats.tx_packets, IFLA_VF_STATS_PAD) || 1289 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES, 1290 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) || 1291 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES, 1292 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) || 1293 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST, 1294 vf_stats.broadcast, IFLA_VF_STATS_PAD) || 1295 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST, 1296 vf_stats.multicast, IFLA_VF_STATS_PAD) || 1297 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED, 1298 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) || 1299 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED, 1300 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) { 1301 nla_nest_cancel(skb, vfstats); 1302 goto nla_put_vf_failure; 1303 } 1304 nla_nest_end(skb, vfstats); 1305 nla_nest_end(skb, vf); 1306 return 0; 1307 1308 nla_put_vf_failure: 1309 nla_nest_cancel(skb, vf); 1310 nla_put_vfinfo_failure: 1311 nla_nest_cancel(skb, vfinfo); 1312 return -EMSGSIZE; 1313 } 1314 1315 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb, 1316 struct net_device *dev, 1317 u32 ext_filter_mask) 1318 { 1319 struct nlattr *vfinfo; 1320 int i, num_vfs; 1321 1322 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0)) 1323 return 0; 1324 1325 num_vfs = dev_num_vf(dev->dev.parent); 1326 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs)) 1327 return -EMSGSIZE; 1328 1329 if (!dev->netdev_ops->ndo_get_vf_config) 1330 return 0; 1331 1332 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1333 if (!vfinfo) 1334 return -EMSGSIZE; 1335 1336 for (i = 0; i < num_vfs; i++) { 1337 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1338 return -EMSGSIZE; 1339 } 1340 1341 nla_nest_end(skb, vfinfo); 1342 return 0; 1343 } 1344 1345 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev) 1346 { 1347 struct rtnl_link_ifmap map; 1348 1349 memset(&map, 0, sizeof(map)); 1350 map.mem_start = dev->mem_start; 1351 map.mem_end = dev->mem_end; 1352 map.base_addr = dev->base_addr; 1353 map.irq = dev->irq; 1354 map.dma = dev->dma; 1355 map.port = dev->if_port; 1356 1357 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD)) 1358 return -EMSGSIZE; 1359 1360 return 0; 1361 } 1362 1363 static u32 rtnl_xdp_prog_skb(struct net_device *dev) 1364 { 1365 const struct bpf_prog *generic_xdp_prog; 1366 1367 ASSERT_RTNL(); 1368 1369 generic_xdp_prog = rtnl_dereference(dev->xdp_prog); 1370 if (!generic_xdp_prog) 1371 return 0; 1372 return generic_xdp_prog->aux->id; 1373 } 1374 1375 static u32 rtnl_xdp_prog_drv(struct net_device *dev) 1376 { 1377 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG); 1378 } 1379 1380 static u32 rtnl_xdp_prog_hw(struct net_device *dev) 1381 { 1382 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, 1383 XDP_QUERY_PROG_HW); 1384 } 1385 1386 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev, 1387 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr, 1388 u32 (*get_prog_id)(struct net_device *dev)) 1389 { 1390 u32 curr_id; 1391 int err; 1392 1393 curr_id = get_prog_id(dev); 1394 if (!curr_id) 1395 return 0; 1396 1397 *prog_id = curr_id; 1398 err = nla_put_u32(skb, attr, curr_id); 1399 if (err) 1400 return err; 1401 1402 if (*mode != XDP_ATTACHED_NONE) 1403 *mode = XDP_ATTACHED_MULTI; 1404 else 1405 *mode = tgt_mode; 1406 1407 return 0; 1408 } 1409 1410 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1411 { 1412 struct nlattr *xdp; 1413 u32 prog_id; 1414 int err; 1415 u8 mode; 1416 1417 xdp = nla_nest_start(skb, IFLA_XDP); 1418 if (!xdp) 1419 return -EMSGSIZE; 1420 1421 prog_id = 0; 1422 mode = XDP_ATTACHED_NONE; 1423 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB, 1424 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb); 1425 if (err) 1426 goto err_cancel; 1427 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV, 1428 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv); 1429 if (err) 1430 goto err_cancel; 1431 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW, 1432 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw); 1433 if (err) 1434 goto err_cancel; 1435 1436 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode); 1437 if (err) 1438 goto err_cancel; 1439 1440 if (prog_id && mode != XDP_ATTACHED_MULTI) { 1441 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id); 1442 if (err) 1443 goto err_cancel; 1444 } 1445 1446 nla_nest_end(skb, xdp); 1447 return 0; 1448 1449 err_cancel: 1450 nla_nest_cancel(skb, xdp); 1451 return err; 1452 } 1453 1454 static u32 rtnl_get_event(unsigned long event) 1455 { 1456 u32 rtnl_event_type = IFLA_EVENT_NONE; 1457 1458 switch (event) { 1459 case NETDEV_REBOOT: 1460 rtnl_event_type = IFLA_EVENT_REBOOT; 1461 break; 1462 case NETDEV_FEAT_CHANGE: 1463 rtnl_event_type = IFLA_EVENT_FEATURES; 1464 break; 1465 case NETDEV_BONDING_FAILOVER: 1466 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER; 1467 break; 1468 case NETDEV_NOTIFY_PEERS: 1469 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS; 1470 break; 1471 case NETDEV_RESEND_IGMP: 1472 rtnl_event_type = IFLA_EVENT_IGMP_RESEND; 1473 break; 1474 case NETDEV_CHANGEINFODATA: 1475 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS; 1476 break; 1477 default: 1478 break; 1479 } 1480 1481 return rtnl_event_type; 1482 } 1483 1484 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev) 1485 { 1486 const struct net_device *upper_dev; 1487 int ret = 0; 1488 1489 rcu_read_lock(); 1490 1491 upper_dev = netdev_master_upper_dev_get_rcu(dev); 1492 if (upper_dev) 1493 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex); 1494 1495 rcu_read_unlock(); 1496 return ret; 1497 } 1498 1499 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev) 1500 { 1501 int ifindex = dev_get_iflink(dev); 1502 1503 if (dev->ifindex == ifindex) 1504 return 0; 1505 1506 return nla_put_u32(skb, IFLA_LINK, ifindex); 1507 } 1508 1509 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb, 1510 struct net_device *dev) 1511 { 1512 char buf[IFALIASZ]; 1513 int ret; 1514 1515 ret = dev_get_alias(dev, buf, sizeof(buf)); 1516 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0; 1517 } 1518 1519 static int rtnl_fill_link_netnsid(struct sk_buff *skb, 1520 const struct net_device *dev, 1521 struct net *src_net) 1522 { 1523 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) { 1524 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1525 1526 if (!net_eq(dev_net(dev), link_net)) { 1527 int id = peernet2id_alloc(src_net, link_net); 1528 1529 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1530 return -EMSGSIZE; 1531 } 1532 } 1533 1534 return 0; 1535 } 1536 1537 static int rtnl_fill_link_af(struct sk_buff *skb, 1538 const struct net_device *dev, 1539 u32 ext_filter_mask) 1540 { 1541 const struct rtnl_af_ops *af_ops; 1542 struct nlattr *af_spec; 1543 1544 af_spec = nla_nest_start(skb, IFLA_AF_SPEC); 1545 if (!af_spec) 1546 return -EMSGSIZE; 1547 1548 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 1549 struct nlattr *af; 1550 int err; 1551 1552 if (!af_ops->fill_link_af) 1553 continue; 1554 1555 af = nla_nest_start(skb, af_ops->family); 1556 if (!af) 1557 return -EMSGSIZE; 1558 1559 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1560 /* 1561 * Caller may return ENODATA to indicate that there 1562 * was no data to be dumped. This is not an error, it 1563 * means we should trim the attribute header and 1564 * continue. 1565 */ 1566 if (err == -ENODATA) 1567 nla_nest_cancel(skb, af); 1568 else if (err < 0) 1569 return -EMSGSIZE; 1570 1571 nla_nest_end(skb, af); 1572 } 1573 1574 nla_nest_end(skb, af_spec); 1575 return 0; 1576 } 1577 1578 static int rtnl_fill_ifinfo(struct sk_buff *skb, 1579 struct net_device *dev, struct net *src_net, 1580 int type, u32 pid, u32 seq, u32 change, 1581 unsigned int flags, u32 ext_filter_mask, 1582 u32 event, int *new_nsid, int new_ifindex, 1583 int tgt_netnsid) 1584 { 1585 struct ifinfomsg *ifm; 1586 struct nlmsghdr *nlh; 1587 1588 ASSERT_RTNL(); 1589 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1590 if (nlh == NULL) 1591 return -EMSGSIZE; 1592 1593 ifm = nlmsg_data(nlh); 1594 ifm->ifi_family = AF_UNSPEC; 1595 ifm->__ifi_pad = 0; 1596 ifm->ifi_type = dev->type; 1597 ifm->ifi_index = dev->ifindex; 1598 ifm->ifi_flags = dev_get_flags(dev); 1599 ifm->ifi_change = change; 1600 1601 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid)) 1602 goto nla_put_failure; 1603 1604 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1605 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1606 nla_put_u8(skb, IFLA_OPERSTATE, 1607 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1608 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1609 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1610 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) || 1611 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) || 1612 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1613 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1614 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1615 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1616 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1617 #ifdef CONFIG_RPS 1618 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1619 #endif 1620 nla_put_iflink(skb, dev) || 1621 put_master_ifindex(skb, dev) || 1622 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1623 (dev->qdisc && 1624 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1625 nla_put_ifalias(skb, dev) || 1626 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1627 atomic_read(&dev->carrier_up_count) + 1628 atomic_read(&dev->carrier_down_count)) || 1629 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) || 1630 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT, 1631 atomic_read(&dev->carrier_up_count)) || 1632 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT, 1633 atomic_read(&dev->carrier_down_count))) 1634 goto nla_put_failure; 1635 1636 if (event != IFLA_EVENT_NONE) { 1637 if (nla_put_u32(skb, IFLA_EVENT, event)) 1638 goto nla_put_failure; 1639 } 1640 1641 if (rtnl_fill_link_ifmap(skb, dev)) 1642 goto nla_put_failure; 1643 1644 if (dev->addr_len) { 1645 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1646 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1647 goto nla_put_failure; 1648 } 1649 1650 if (rtnl_phys_port_id_fill(skb, dev)) 1651 goto nla_put_failure; 1652 1653 if (rtnl_phys_port_name_fill(skb, dev)) 1654 goto nla_put_failure; 1655 1656 if (rtnl_phys_switch_id_fill(skb, dev)) 1657 goto nla_put_failure; 1658 1659 if (rtnl_fill_stats(skb, dev)) 1660 goto nla_put_failure; 1661 1662 if (rtnl_fill_vf(skb, dev, ext_filter_mask)) 1663 goto nla_put_failure; 1664 1665 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1666 goto nla_put_failure; 1667 1668 if (rtnl_xdp_fill(skb, dev)) 1669 goto nla_put_failure; 1670 1671 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1672 if (rtnl_link_fill(skb, dev) < 0) 1673 goto nla_put_failure; 1674 } 1675 1676 if (rtnl_fill_link_netnsid(skb, dev, src_net)) 1677 goto nla_put_failure; 1678 1679 if (new_nsid && 1680 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0) 1681 goto nla_put_failure; 1682 if (new_ifindex && 1683 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0) 1684 goto nla_put_failure; 1685 1686 1687 rcu_read_lock(); 1688 if (rtnl_fill_link_af(skb, dev, ext_filter_mask)) 1689 goto nla_put_failure_rcu; 1690 rcu_read_unlock(); 1691 1692 nlmsg_end(skb, nlh); 1693 return 0; 1694 1695 nla_put_failure_rcu: 1696 rcu_read_unlock(); 1697 nla_put_failure: 1698 nlmsg_cancel(skb, nlh); 1699 return -EMSGSIZE; 1700 } 1701 1702 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1703 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1704 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1705 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1706 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1707 [IFLA_MTU] = { .type = NLA_U32 }, 1708 [IFLA_LINK] = { .type = NLA_U32 }, 1709 [IFLA_MASTER] = { .type = NLA_U32 }, 1710 [IFLA_CARRIER] = { .type = NLA_U8 }, 1711 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1712 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1713 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1714 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1715 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1716 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1717 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1718 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to 1719 * allow 0-length string (needed to remove an alias). 1720 */ 1721 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 }, 1722 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1723 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1724 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1725 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1726 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1727 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1728 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1729 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1730 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 }, 1731 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 }, 1732 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1733 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1734 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1735 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1736 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1737 [IFLA_XDP] = { .type = NLA_NESTED }, 1738 [IFLA_EVENT] = { .type = NLA_U32 }, 1739 [IFLA_GROUP] = { .type = NLA_U32 }, 1740 [IFLA_IF_NETNSID] = { .type = NLA_S32 }, 1741 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 }, 1742 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 }, 1743 [IFLA_MIN_MTU] = { .type = NLA_U32 }, 1744 [IFLA_MAX_MTU] = { .type = NLA_U32 }, 1745 }; 1746 1747 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1748 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1749 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1750 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1751 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1752 }; 1753 1754 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1755 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1756 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1757 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1758 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1759 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1760 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1761 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1762 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1763 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1764 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1765 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1766 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1767 }; 1768 1769 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1770 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1771 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1772 .len = PORT_PROFILE_MAX }, 1773 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1774 .len = PORT_UUID_MAX }, 1775 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1776 .len = PORT_UUID_MAX }, 1777 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1778 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1779 1780 /* Unused, but we need to keep it here since user space could 1781 * fill it. It's also broken with regard to NLA_BINARY use in 1782 * combination with structs. 1783 */ 1784 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1785 .len = sizeof(struct ifla_port_vsi) }, 1786 }; 1787 1788 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1789 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1790 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1791 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1792 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 }, 1793 }; 1794 1795 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1796 { 1797 const struct rtnl_link_ops *ops = NULL; 1798 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1799 1800 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, 1801 ifla_info_policy, NULL) < 0) 1802 return NULL; 1803 1804 if (linfo[IFLA_INFO_KIND]) { 1805 char kind[MODULE_NAME_LEN]; 1806 1807 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1808 ops = rtnl_link_ops_get(kind); 1809 } 1810 1811 return ops; 1812 } 1813 1814 static bool link_master_filtered(struct net_device *dev, int master_idx) 1815 { 1816 struct net_device *master; 1817 1818 if (!master_idx) 1819 return false; 1820 1821 master = netdev_master_upper_dev_get(dev); 1822 if (!master || master->ifindex != master_idx) 1823 return true; 1824 1825 return false; 1826 } 1827 1828 static bool link_kind_filtered(const struct net_device *dev, 1829 const struct rtnl_link_ops *kind_ops) 1830 { 1831 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1832 return true; 1833 1834 return false; 1835 } 1836 1837 static bool link_dump_filtered(struct net_device *dev, 1838 int master_idx, 1839 const struct rtnl_link_ops *kind_ops) 1840 { 1841 if (link_master_filtered(dev, master_idx) || 1842 link_kind_filtered(dev, kind_ops)) 1843 return true; 1844 1845 return false; 1846 } 1847 1848 static struct net *get_target_net(struct sock *sk, int netnsid) 1849 { 1850 struct net *net; 1851 1852 net = get_net_ns_by_id(sock_net(sk), netnsid); 1853 if (!net) 1854 return ERR_PTR(-EINVAL); 1855 1856 /* For now, the caller is required to have CAP_NET_ADMIN in 1857 * the user namespace owning the target net ns. 1858 */ 1859 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) { 1860 put_net(net); 1861 return ERR_PTR(-EACCES); 1862 } 1863 return net; 1864 } 1865 1866 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1867 { 1868 struct net *net = sock_net(skb->sk); 1869 struct net *tgt_net = net; 1870 int h, s_h; 1871 int idx = 0, s_idx; 1872 struct net_device *dev; 1873 struct hlist_head *head; 1874 struct nlattr *tb[IFLA_MAX+1]; 1875 u32 ext_filter_mask = 0; 1876 const struct rtnl_link_ops *kind_ops = NULL; 1877 unsigned int flags = NLM_F_MULTI; 1878 int master_idx = 0; 1879 int netnsid = -1; 1880 int err; 1881 int hdrlen; 1882 1883 s_h = cb->args[0]; 1884 s_idx = cb->args[1]; 1885 1886 /* A hack to preserve kernel<->userspace interface. 1887 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1888 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1889 * what iproute2 < v3.9.0 used. 1890 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1891 * attribute, its netlink message is shorter than struct ifinfomsg. 1892 */ 1893 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1894 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1895 1896 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, 1897 ifla_policy, NULL) >= 0) { 1898 if (tb[IFLA_IF_NETNSID]) { 1899 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]); 1900 tgt_net = get_target_net(skb->sk, netnsid); 1901 if (IS_ERR(tgt_net)) { 1902 tgt_net = net; 1903 netnsid = -1; 1904 } 1905 } 1906 1907 if (tb[IFLA_EXT_MASK]) 1908 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1909 1910 if (tb[IFLA_MASTER]) 1911 master_idx = nla_get_u32(tb[IFLA_MASTER]); 1912 1913 if (tb[IFLA_LINKINFO]) 1914 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]); 1915 1916 if (master_idx || kind_ops) 1917 flags |= NLM_F_DUMP_FILTERED; 1918 } 1919 1920 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1921 idx = 0; 1922 head = &tgt_net->dev_index_head[h]; 1923 hlist_for_each_entry(dev, head, index_hlist) { 1924 if (link_dump_filtered(dev, master_idx, kind_ops)) 1925 goto cont; 1926 if (idx < s_idx) 1927 goto cont; 1928 err = rtnl_fill_ifinfo(skb, dev, net, 1929 RTM_NEWLINK, 1930 NETLINK_CB(cb->skb).portid, 1931 cb->nlh->nlmsg_seq, 0, 1932 flags, 1933 ext_filter_mask, 0, NULL, 0, 1934 netnsid); 1935 1936 if (err < 0) { 1937 if (likely(skb->len)) 1938 goto out; 1939 1940 goto out_err; 1941 } 1942 cont: 1943 idx++; 1944 } 1945 } 1946 out: 1947 err = skb->len; 1948 out_err: 1949 cb->args[1] = idx; 1950 cb->args[0] = h; 1951 cb->seq = net->dev_base_seq; 1952 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1953 if (netnsid >= 0) 1954 put_net(tgt_net); 1955 1956 return err; 1957 } 1958 1959 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len, 1960 struct netlink_ext_ack *exterr) 1961 { 1962 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr); 1963 } 1964 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1965 1966 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1967 { 1968 struct net *net; 1969 /* Examine the link attributes and figure out which 1970 * network namespace we are talking about. 1971 */ 1972 if (tb[IFLA_NET_NS_PID]) 1973 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1974 else if (tb[IFLA_NET_NS_FD]) 1975 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1976 else 1977 net = get_net(src_net); 1978 return net; 1979 } 1980 EXPORT_SYMBOL(rtnl_link_get_net); 1981 1982 /* Figure out which network namespace we are talking about by 1983 * examining the link attributes in the following order: 1984 * 1985 * 1. IFLA_NET_NS_PID 1986 * 2. IFLA_NET_NS_FD 1987 * 3. IFLA_IF_NETNSID 1988 */ 1989 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net, 1990 struct nlattr *tb[]) 1991 { 1992 struct net *net; 1993 1994 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) 1995 return rtnl_link_get_net(src_net, tb); 1996 1997 if (!tb[IFLA_IF_NETNSID]) 1998 return get_net(src_net); 1999 2000 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID])); 2001 if (!net) 2002 return ERR_PTR(-EINVAL); 2003 2004 return net; 2005 } 2006 2007 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb, 2008 struct net *src_net, 2009 struct nlattr *tb[], int cap) 2010 { 2011 struct net *net; 2012 2013 net = rtnl_link_get_net_by_nlattr(src_net, tb); 2014 if (IS_ERR(net)) 2015 return net; 2016 2017 if (!netlink_ns_capable(skb, net->user_ns, cap)) { 2018 put_net(net); 2019 return ERR_PTR(-EPERM); 2020 } 2021 2022 return net; 2023 } 2024 2025 /* Verify that rtnetlink requests do not pass additional properties 2026 * potentially referring to different network namespaces. 2027 */ 2028 static int rtnl_ensure_unique_netns(struct nlattr *tb[], 2029 struct netlink_ext_ack *extack, 2030 bool netns_id_only) 2031 { 2032 2033 if (netns_id_only) { 2034 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD]) 2035 return 0; 2036 2037 NL_SET_ERR_MSG(extack, "specified netns attribute not supported"); 2038 return -EOPNOTSUPP; 2039 } 2040 2041 if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])) 2042 goto invalid_attr; 2043 2044 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD])) 2045 goto invalid_attr; 2046 2047 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID])) 2048 goto invalid_attr; 2049 2050 return 0; 2051 2052 invalid_attr: 2053 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified"); 2054 return -EINVAL; 2055 } 2056 2057 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 2058 { 2059 if (dev) { 2060 if (tb[IFLA_ADDRESS] && 2061 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 2062 return -EINVAL; 2063 2064 if (tb[IFLA_BROADCAST] && 2065 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 2066 return -EINVAL; 2067 } 2068 2069 if (tb[IFLA_AF_SPEC]) { 2070 struct nlattr *af; 2071 int rem, err; 2072 2073 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2074 const struct rtnl_af_ops *af_ops; 2075 2076 rcu_read_lock(); 2077 af_ops = rtnl_af_lookup(nla_type(af)); 2078 if (!af_ops) { 2079 rcu_read_unlock(); 2080 return -EAFNOSUPPORT; 2081 } 2082 2083 if (!af_ops->set_link_af) { 2084 rcu_read_unlock(); 2085 return -EOPNOTSUPP; 2086 } 2087 2088 if (af_ops->validate_link_af) { 2089 err = af_ops->validate_link_af(dev, af); 2090 if (err < 0) { 2091 rcu_read_unlock(); 2092 return err; 2093 } 2094 } 2095 2096 rcu_read_unlock(); 2097 } 2098 } 2099 2100 return 0; 2101 } 2102 2103 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 2104 int guid_type) 2105 { 2106 const struct net_device_ops *ops = dev->netdev_ops; 2107 2108 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 2109 } 2110 2111 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 2112 { 2113 if (dev->type != ARPHRD_INFINIBAND) 2114 return -EOPNOTSUPP; 2115 2116 return handle_infiniband_guid(dev, ivt, guid_type); 2117 } 2118 2119 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 2120 { 2121 const struct net_device_ops *ops = dev->netdev_ops; 2122 int err = -EINVAL; 2123 2124 if (tb[IFLA_VF_MAC]) { 2125 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 2126 2127 err = -EOPNOTSUPP; 2128 if (ops->ndo_set_vf_mac) 2129 err = ops->ndo_set_vf_mac(dev, ivm->vf, 2130 ivm->mac); 2131 if (err < 0) 2132 return err; 2133 } 2134 2135 if (tb[IFLA_VF_VLAN]) { 2136 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 2137 2138 err = -EOPNOTSUPP; 2139 if (ops->ndo_set_vf_vlan) 2140 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 2141 ivv->qos, 2142 htons(ETH_P_8021Q)); 2143 if (err < 0) 2144 return err; 2145 } 2146 2147 if (tb[IFLA_VF_VLAN_LIST]) { 2148 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 2149 struct nlattr *attr; 2150 int rem, len = 0; 2151 2152 err = -EOPNOTSUPP; 2153 if (!ops->ndo_set_vf_vlan) 2154 return err; 2155 2156 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 2157 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 2158 nla_len(attr) < NLA_HDRLEN) { 2159 return -EINVAL; 2160 } 2161 if (len >= MAX_VLAN_LIST_LEN) 2162 return -EOPNOTSUPP; 2163 ivvl[len] = nla_data(attr); 2164 2165 len++; 2166 } 2167 if (len == 0) 2168 return -EINVAL; 2169 2170 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 2171 ivvl[0]->qos, ivvl[0]->vlan_proto); 2172 if (err < 0) 2173 return err; 2174 } 2175 2176 if (tb[IFLA_VF_TX_RATE]) { 2177 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 2178 struct ifla_vf_info ivf; 2179 2180 err = -EOPNOTSUPP; 2181 if (ops->ndo_get_vf_config) 2182 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 2183 if (err < 0) 2184 return err; 2185 2186 err = -EOPNOTSUPP; 2187 if (ops->ndo_set_vf_rate) 2188 err = ops->ndo_set_vf_rate(dev, ivt->vf, 2189 ivf.min_tx_rate, 2190 ivt->rate); 2191 if (err < 0) 2192 return err; 2193 } 2194 2195 if (tb[IFLA_VF_RATE]) { 2196 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 2197 2198 err = -EOPNOTSUPP; 2199 if (ops->ndo_set_vf_rate) 2200 err = ops->ndo_set_vf_rate(dev, ivt->vf, 2201 ivt->min_tx_rate, 2202 ivt->max_tx_rate); 2203 if (err < 0) 2204 return err; 2205 } 2206 2207 if (tb[IFLA_VF_SPOOFCHK]) { 2208 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 2209 2210 err = -EOPNOTSUPP; 2211 if (ops->ndo_set_vf_spoofchk) 2212 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 2213 ivs->setting); 2214 if (err < 0) 2215 return err; 2216 } 2217 2218 if (tb[IFLA_VF_LINK_STATE]) { 2219 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 2220 2221 err = -EOPNOTSUPP; 2222 if (ops->ndo_set_vf_link_state) 2223 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 2224 ivl->link_state); 2225 if (err < 0) 2226 return err; 2227 } 2228 2229 if (tb[IFLA_VF_RSS_QUERY_EN]) { 2230 struct ifla_vf_rss_query_en *ivrssq_en; 2231 2232 err = -EOPNOTSUPP; 2233 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 2234 if (ops->ndo_set_vf_rss_query_en) 2235 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 2236 ivrssq_en->setting); 2237 if (err < 0) 2238 return err; 2239 } 2240 2241 if (tb[IFLA_VF_TRUST]) { 2242 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 2243 2244 err = -EOPNOTSUPP; 2245 if (ops->ndo_set_vf_trust) 2246 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 2247 if (err < 0) 2248 return err; 2249 } 2250 2251 if (tb[IFLA_VF_IB_NODE_GUID]) { 2252 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 2253 2254 if (!ops->ndo_set_vf_guid) 2255 return -EOPNOTSUPP; 2256 2257 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 2258 } 2259 2260 if (tb[IFLA_VF_IB_PORT_GUID]) { 2261 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 2262 2263 if (!ops->ndo_set_vf_guid) 2264 return -EOPNOTSUPP; 2265 2266 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 2267 } 2268 2269 return err; 2270 } 2271 2272 static int do_set_master(struct net_device *dev, int ifindex, 2273 struct netlink_ext_ack *extack) 2274 { 2275 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 2276 const struct net_device_ops *ops; 2277 int err; 2278 2279 if (upper_dev) { 2280 if (upper_dev->ifindex == ifindex) 2281 return 0; 2282 ops = upper_dev->netdev_ops; 2283 if (ops->ndo_del_slave) { 2284 err = ops->ndo_del_slave(upper_dev, dev); 2285 if (err) 2286 return err; 2287 } else { 2288 return -EOPNOTSUPP; 2289 } 2290 } 2291 2292 if (ifindex) { 2293 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 2294 if (!upper_dev) 2295 return -EINVAL; 2296 ops = upper_dev->netdev_ops; 2297 if (ops->ndo_add_slave) { 2298 err = ops->ndo_add_slave(upper_dev, dev, extack); 2299 if (err) 2300 return err; 2301 } else { 2302 return -EOPNOTSUPP; 2303 } 2304 } 2305 return 0; 2306 } 2307 2308 #define DO_SETLINK_MODIFIED 0x01 2309 /* notify flag means notify + modified. */ 2310 #define DO_SETLINK_NOTIFY 0x03 2311 static int do_setlink(const struct sk_buff *skb, 2312 struct net_device *dev, struct ifinfomsg *ifm, 2313 struct netlink_ext_ack *extack, 2314 struct nlattr **tb, char *ifname, int status) 2315 { 2316 const struct net_device_ops *ops = dev->netdev_ops; 2317 int err; 2318 2319 err = validate_linkmsg(dev, tb); 2320 if (err < 0) 2321 return err; 2322 2323 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) { 2324 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev), 2325 tb, CAP_NET_ADMIN); 2326 if (IS_ERR(net)) { 2327 err = PTR_ERR(net); 2328 goto errout; 2329 } 2330 2331 err = dev_change_net_namespace(dev, net, ifname); 2332 put_net(net); 2333 if (err) 2334 goto errout; 2335 status |= DO_SETLINK_MODIFIED; 2336 } 2337 2338 if (tb[IFLA_MAP]) { 2339 struct rtnl_link_ifmap *u_map; 2340 struct ifmap k_map; 2341 2342 if (!ops->ndo_set_config) { 2343 err = -EOPNOTSUPP; 2344 goto errout; 2345 } 2346 2347 if (!netif_device_present(dev)) { 2348 err = -ENODEV; 2349 goto errout; 2350 } 2351 2352 u_map = nla_data(tb[IFLA_MAP]); 2353 k_map.mem_start = (unsigned long) u_map->mem_start; 2354 k_map.mem_end = (unsigned long) u_map->mem_end; 2355 k_map.base_addr = (unsigned short) u_map->base_addr; 2356 k_map.irq = (unsigned char) u_map->irq; 2357 k_map.dma = (unsigned char) u_map->dma; 2358 k_map.port = (unsigned char) u_map->port; 2359 2360 err = ops->ndo_set_config(dev, &k_map); 2361 if (err < 0) 2362 goto errout; 2363 2364 status |= DO_SETLINK_NOTIFY; 2365 } 2366 2367 if (tb[IFLA_ADDRESS]) { 2368 struct sockaddr *sa; 2369 int len; 2370 2371 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len, 2372 sizeof(*sa)); 2373 sa = kmalloc(len, GFP_KERNEL); 2374 if (!sa) { 2375 err = -ENOMEM; 2376 goto errout; 2377 } 2378 sa->sa_family = dev->type; 2379 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 2380 dev->addr_len); 2381 err = dev_set_mac_address(dev, sa); 2382 kfree(sa); 2383 if (err) 2384 goto errout; 2385 status |= DO_SETLINK_MODIFIED; 2386 } 2387 2388 if (tb[IFLA_MTU]) { 2389 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack); 2390 if (err < 0) 2391 goto errout; 2392 status |= DO_SETLINK_MODIFIED; 2393 } 2394 2395 if (tb[IFLA_GROUP]) { 2396 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2397 status |= DO_SETLINK_NOTIFY; 2398 } 2399 2400 /* 2401 * Interface selected by interface index but interface 2402 * name provided implies that a name change has been 2403 * requested. 2404 */ 2405 if (ifm->ifi_index > 0 && ifname[0]) { 2406 err = dev_change_name(dev, ifname); 2407 if (err < 0) 2408 goto errout; 2409 status |= DO_SETLINK_MODIFIED; 2410 } 2411 2412 if (tb[IFLA_IFALIAS]) { 2413 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2414 nla_len(tb[IFLA_IFALIAS])); 2415 if (err < 0) 2416 goto errout; 2417 status |= DO_SETLINK_NOTIFY; 2418 } 2419 2420 if (tb[IFLA_BROADCAST]) { 2421 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2422 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2423 } 2424 2425 if (ifm->ifi_flags || ifm->ifi_change) { 2426 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2427 if (err < 0) 2428 goto errout; 2429 } 2430 2431 if (tb[IFLA_MASTER]) { 2432 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack); 2433 if (err) 2434 goto errout; 2435 status |= DO_SETLINK_MODIFIED; 2436 } 2437 2438 if (tb[IFLA_CARRIER]) { 2439 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2440 if (err) 2441 goto errout; 2442 status |= DO_SETLINK_MODIFIED; 2443 } 2444 2445 if (tb[IFLA_TXQLEN]) { 2446 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]); 2447 2448 err = dev_change_tx_queue_len(dev, value); 2449 if (err) 2450 goto errout; 2451 status |= DO_SETLINK_MODIFIED; 2452 } 2453 2454 if (tb[IFLA_GSO_MAX_SIZE]) { 2455 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]); 2456 2457 if (max_size > GSO_MAX_SIZE) { 2458 err = -EINVAL; 2459 goto errout; 2460 } 2461 2462 if (dev->gso_max_size ^ max_size) { 2463 netif_set_gso_max_size(dev, max_size); 2464 status |= DO_SETLINK_MODIFIED; 2465 } 2466 } 2467 2468 if (tb[IFLA_GSO_MAX_SEGS]) { 2469 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]); 2470 2471 if (max_segs > GSO_MAX_SEGS) { 2472 err = -EINVAL; 2473 goto errout; 2474 } 2475 2476 if (dev->gso_max_segs ^ max_segs) { 2477 dev->gso_max_segs = max_segs; 2478 status |= DO_SETLINK_MODIFIED; 2479 } 2480 } 2481 2482 if (tb[IFLA_OPERSTATE]) 2483 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2484 2485 if (tb[IFLA_LINKMODE]) { 2486 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2487 2488 write_lock_bh(&dev_base_lock); 2489 if (dev->link_mode ^ value) 2490 status |= DO_SETLINK_NOTIFY; 2491 dev->link_mode = value; 2492 write_unlock_bh(&dev_base_lock); 2493 } 2494 2495 if (tb[IFLA_VFINFO_LIST]) { 2496 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2497 struct nlattr *attr; 2498 int rem; 2499 2500 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2501 if (nla_type(attr) != IFLA_VF_INFO || 2502 nla_len(attr) < NLA_HDRLEN) { 2503 err = -EINVAL; 2504 goto errout; 2505 } 2506 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr, 2507 ifla_vf_policy, NULL); 2508 if (err < 0) 2509 goto errout; 2510 err = do_setvfinfo(dev, vfinfo); 2511 if (err < 0) 2512 goto errout; 2513 status |= DO_SETLINK_NOTIFY; 2514 } 2515 } 2516 err = 0; 2517 2518 if (tb[IFLA_VF_PORTS]) { 2519 struct nlattr *port[IFLA_PORT_MAX+1]; 2520 struct nlattr *attr; 2521 int vf; 2522 int rem; 2523 2524 err = -EOPNOTSUPP; 2525 if (!ops->ndo_set_vf_port) 2526 goto errout; 2527 2528 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2529 if (nla_type(attr) != IFLA_VF_PORT || 2530 nla_len(attr) < NLA_HDRLEN) { 2531 err = -EINVAL; 2532 goto errout; 2533 } 2534 err = nla_parse_nested(port, IFLA_PORT_MAX, attr, 2535 ifla_port_policy, NULL); 2536 if (err < 0) 2537 goto errout; 2538 if (!port[IFLA_PORT_VF]) { 2539 err = -EOPNOTSUPP; 2540 goto errout; 2541 } 2542 vf = nla_get_u32(port[IFLA_PORT_VF]); 2543 err = ops->ndo_set_vf_port(dev, vf, port); 2544 if (err < 0) 2545 goto errout; 2546 status |= DO_SETLINK_NOTIFY; 2547 } 2548 } 2549 err = 0; 2550 2551 if (tb[IFLA_PORT_SELF]) { 2552 struct nlattr *port[IFLA_PORT_MAX+1]; 2553 2554 err = nla_parse_nested(port, IFLA_PORT_MAX, 2555 tb[IFLA_PORT_SELF], ifla_port_policy, 2556 NULL); 2557 if (err < 0) 2558 goto errout; 2559 2560 err = -EOPNOTSUPP; 2561 if (ops->ndo_set_vf_port) 2562 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2563 if (err < 0) 2564 goto errout; 2565 status |= DO_SETLINK_NOTIFY; 2566 } 2567 2568 if (tb[IFLA_AF_SPEC]) { 2569 struct nlattr *af; 2570 int rem; 2571 2572 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2573 const struct rtnl_af_ops *af_ops; 2574 2575 rcu_read_lock(); 2576 2577 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af)))); 2578 2579 err = af_ops->set_link_af(dev, af); 2580 if (err < 0) { 2581 rcu_read_unlock(); 2582 goto errout; 2583 } 2584 2585 rcu_read_unlock(); 2586 status |= DO_SETLINK_NOTIFY; 2587 } 2588 } 2589 err = 0; 2590 2591 if (tb[IFLA_PROTO_DOWN]) { 2592 err = dev_change_proto_down(dev, 2593 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2594 if (err) 2595 goto errout; 2596 status |= DO_SETLINK_NOTIFY; 2597 } 2598 2599 if (tb[IFLA_XDP]) { 2600 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2601 u32 xdp_flags = 0; 2602 2603 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP], 2604 ifla_xdp_policy, NULL); 2605 if (err < 0) 2606 goto errout; 2607 2608 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) { 2609 err = -EINVAL; 2610 goto errout; 2611 } 2612 2613 if (xdp[IFLA_XDP_FLAGS]) { 2614 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2615 if (xdp_flags & ~XDP_FLAGS_MASK) { 2616 err = -EINVAL; 2617 goto errout; 2618 } 2619 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) { 2620 err = -EINVAL; 2621 goto errout; 2622 } 2623 } 2624 2625 if (xdp[IFLA_XDP_FD]) { 2626 err = dev_change_xdp_fd(dev, extack, 2627 nla_get_s32(xdp[IFLA_XDP_FD]), 2628 xdp_flags); 2629 if (err) 2630 goto errout; 2631 status |= DO_SETLINK_NOTIFY; 2632 } 2633 } 2634 2635 errout: 2636 if (status & DO_SETLINK_MODIFIED) { 2637 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY) 2638 netdev_state_change(dev); 2639 2640 if (err < 0) 2641 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n", 2642 dev->name); 2643 } 2644 2645 return err; 2646 } 2647 2648 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2649 struct netlink_ext_ack *extack) 2650 { 2651 struct net *net = sock_net(skb->sk); 2652 struct ifinfomsg *ifm; 2653 struct net_device *dev; 2654 int err; 2655 struct nlattr *tb[IFLA_MAX+1]; 2656 char ifname[IFNAMSIZ]; 2657 2658 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, 2659 extack); 2660 if (err < 0) 2661 goto errout; 2662 2663 err = rtnl_ensure_unique_netns(tb, extack, false); 2664 if (err < 0) 2665 goto errout; 2666 2667 if (tb[IFLA_IFNAME]) 2668 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2669 else 2670 ifname[0] = '\0'; 2671 2672 err = -EINVAL; 2673 ifm = nlmsg_data(nlh); 2674 if (ifm->ifi_index > 0) 2675 dev = __dev_get_by_index(net, ifm->ifi_index); 2676 else if (tb[IFLA_IFNAME]) 2677 dev = __dev_get_by_name(net, ifname); 2678 else 2679 goto errout; 2680 2681 if (dev == NULL) { 2682 err = -ENODEV; 2683 goto errout; 2684 } 2685 2686 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0); 2687 errout: 2688 return err; 2689 } 2690 2691 static int rtnl_group_dellink(const struct net *net, int group) 2692 { 2693 struct net_device *dev, *aux; 2694 LIST_HEAD(list_kill); 2695 bool found = false; 2696 2697 if (!group) 2698 return -EPERM; 2699 2700 for_each_netdev(net, dev) { 2701 if (dev->group == group) { 2702 const struct rtnl_link_ops *ops; 2703 2704 found = true; 2705 ops = dev->rtnl_link_ops; 2706 if (!ops || !ops->dellink) 2707 return -EOPNOTSUPP; 2708 } 2709 } 2710 2711 if (!found) 2712 return -ENODEV; 2713 2714 for_each_netdev_safe(net, dev, aux) { 2715 if (dev->group == group) { 2716 const struct rtnl_link_ops *ops; 2717 2718 ops = dev->rtnl_link_ops; 2719 ops->dellink(dev, &list_kill); 2720 } 2721 } 2722 unregister_netdevice_many(&list_kill); 2723 2724 return 0; 2725 } 2726 2727 int rtnl_delete_link(struct net_device *dev) 2728 { 2729 const struct rtnl_link_ops *ops; 2730 LIST_HEAD(list_kill); 2731 2732 ops = dev->rtnl_link_ops; 2733 if (!ops || !ops->dellink) 2734 return -EOPNOTSUPP; 2735 2736 ops->dellink(dev, &list_kill); 2737 unregister_netdevice_many(&list_kill); 2738 2739 return 0; 2740 } 2741 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2742 2743 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 2744 struct netlink_ext_ack *extack) 2745 { 2746 struct net *net = sock_net(skb->sk); 2747 struct net *tgt_net = net; 2748 struct net_device *dev = NULL; 2749 struct ifinfomsg *ifm; 2750 char ifname[IFNAMSIZ]; 2751 struct nlattr *tb[IFLA_MAX+1]; 2752 int err; 2753 int netnsid = -1; 2754 2755 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2756 if (err < 0) 2757 return err; 2758 2759 err = rtnl_ensure_unique_netns(tb, extack, true); 2760 if (err < 0) 2761 return err; 2762 2763 if (tb[IFLA_IFNAME]) 2764 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2765 2766 if (tb[IFLA_IF_NETNSID]) { 2767 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]); 2768 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid); 2769 if (IS_ERR(tgt_net)) 2770 return PTR_ERR(tgt_net); 2771 } 2772 2773 err = -EINVAL; 2774 ifm = nlmsg_data(nlh); 2775 if (ifm->ifi_index > 0) 2776 dev = __dev_get_by_index(tgt_net, ifm->ifi_index); 2777 else if (tb[IFLA_IFNAME]) 2778 dev = __dev_get_by_name(tgt_net, ifname); 2779 else if (tb[IFLA_GROUP]) 2780 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP])); 2781 else 2782 goto out; 2783 2784 if (!dev) { 2785 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0) 2786 err = -ENODEV; 2787 2788 goto out; 2789 } 2790 2791 err = rtnl_delete_link(dev); 2792 2793 out: 2794 if (netnsid >= 0) 2795 put_net(tgt_net); 2796 2797 return err; 2798 } 2799 2800 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2801 { 2802 unsigned int old_flags; 2803 int err; 2804 2805 old_flags = dev->flags; 2806 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2807 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2808 if (err < 0) 2809 return err; 2810 } 2811 2812 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) { 2813 __dev_notify_flags(dev, old_flags, 0U); 2814 } else { 2815 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2816 __dev_notify_flags(dev, old_flags, ~0U); 2817 } 2818 return 0; 2819 } 2820 EXPORT_SYMBOL(rtnl_configure_link); 2821 2822 struct net_device *rtnl_create_link(struct net *net, 2823 const char *ifname, unsigned char name_assign_type, 2824 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 2825 { 2826 struct net_device *dev; 2827 unsigned int num_tx_queues = 1; 2828 unsigned int num_rx_queues = 1; 2829 2830 if (tb[IFLA_NUM_TX_QUEUES]) 2831 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2832 else if (ops->get_num_tx_queues) 2833 num_tx_queues = ops->get_num_tx_queues(); 2834 2835 if (tb[IFLA_NUM_RX_QUEUES]) 2836 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2837 else if (ops->get_num_rx_queues) 2838 num_rx_queues = ops->get_num_rx_queues(); 2839 2840 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2841 ops->setup, num_tx_queues, num_rx_queues); 2842 if (!dev) 2843 return ERR_PTR(-ENOMEM); 2844 2845 dev_net_set(dev, net); 2846 dev->rtnl_link_ops = ops; 2847 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2848 2849 if (tb[IFLA_MTU]) 2850 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2851 if (tb[IFLA_ADDRESS]) { 2852 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2853 nla_len(tb[IFLA_ADDRESS])); 2854 dev->addr_assign_type = NET_ADDR_SET; 2855 } 2856 if (tb[IFLA_BROADCAST]) 2857 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2858 nla_len(tb[IFLA_BROADCAST])); 2859 if (tb[IFLA_TXQLEN]) 2860 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2861 if (tb[IFLA_OPERSTATE]) 2862 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2863 if (tb[IFLA_LINKMODE]) 2864 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2865 if (tb[IFLA_GROUP]) 2866 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2867 if (tb[IFLA_GSO_MAX_SIZE]) 2868 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE])); 2869 if (tb[IFLA_GSO_MAX_SEGS]) 2870 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]); 2871 2872 return dev; 2873 } 2874 EXPORT_SYMBOL(rtnl_create_link); 2875 2876 static int rtnl_group_changelink(const struct sk_buff *skb, 2877 struct net *net, int group, 2878 struct ifinfomsg *ifm, 2879 struct netlink_ext_ack *extack, 2880 struct nlattr **tb) 2881 { 2882 struct net_device *dev, *aux; 2883 int err; 2884 2885 for_each_netdev_safe(net, dev, aux) { 2886 if (dev->group == group) { 2887 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0); 2888 if (err < 0) 2889 return err; 2890 } 2891 } 2892 2893 return 0; 2894 } 2895 2896 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2897 struct netlink_ext_ack *extack) 2898 { 2899 struct net *net = sock_net(skb->sk); 2900 const struct rtnl_link_ops *ops; 2901 const struct rtnl_link_ops *m_ops = NULL; 2902 struct net_device *dev; 2903 struct net_device *master_dev = NULL; 2904 struct ifinfomsg *ifm; 2905 char kind[MODULE_NAME_LEN]; 2906 char ifname[IFNAMSIZ]; 2907 struct nlattr *tb[IFLA_MAX+1]; 2908 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 2909 unsigned char name_assign_type = NET_NAME_USER; 2910 int err; 2911 2912 #ifdef CONFIG_MODULES 2913 replay: 2914 #endif 2915 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2916 if (err < 0) 2917 return err; 2918 2919 err = rtnl_ensure_unique_netns(tb, extack, false); 2920 if (err < 0) 2921 return err; 2922 2923 if (tb[IFLA_IFNAME]) 2924 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2925 else 2926 ifname[0] = '\0'; 2927 2928 ifm = nlmsg_data(nlh); 2929 if (ifm->ifi_index > 0) 2930 dev = __dev_get_by_index(net, ifm->ifi_index); 2931 else { 2932 if (ifname[0]) 2933 dev = __dev_get_by_name(net, ifname); 2934 else 2935 dev = NULL; 2936 } 2937 2938 if (dev) { 2939 master_dev = netdev_master_upper_dev_get(dev); 2940 if (master_dev) 2941 m_ops = master_dev->rtnl_link_ops; 2942 } 2943 2944 err = validate_linkmsg(dev, tb); 2945 if (err < 0) 2946 return err; 2947 2948 if (tb[IFLA_LINKINFO]) { 2949 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 2950 tb[IFLA_LINKINFO], ifla_info_policy, 2951 NULL); 2952 if (err < 0) 2953 return err; 2954 } else 2955 memset(linkinfo, 0, sizeof(linkinfo)); 2956 2957 if (linkinfo[IFLA_INFO_KIND]) { 2958 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 2959 ops = rtnl_link_ops_get(kind); 2960 } else { 2961 kind[0] = '\0'; 2962 ops = NULL; 2963 } 2964 2965 if (1) { 2966 struct nlattr *attr[RTNL_MAX_TYPE + 1]; 2967 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1]; 2968 struct nlattr **data = NULL; 2969 struct nlattr **slave_data = NULL; 2970 struct net *dest_net, *link_net = NULL; 2971 2972 if (ops) { 2973 if (ops->maxtype > RTNL_MAX_TYPE) 2974 return -EINVAL; 2975 2976 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 2977 err = nla_parse_nested(attr, ops->maxtype, 2978 linkinfo[IFLA_INFO_DATA], 2979 ops->policy, NULL); 2980 if (err < 0) 2981 return err; 2982 data = attr; 2983 } 2984 if (ops->validate) { 2985 err = ops->validate(tb, data, extack); 2986 if (err < 0) 2987 return err; 2988 } 2989 } 2990 2991 if (m_ops) { 2992 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE) 2993 return -EINVAL; 2994 2995 if (m_ops->slave_maxtype && 2996 linkinfo[IFLA_INFO_SLAVE_DATA]) { 2997 err = nla_parse_nested(slave_attr, 2998 m_ops->slave_maxtype, 2999 linkinfo[IFLA_INFO_SLAVE_DATA], 3000 m_ops->slave_policy, 3001 NULL); 3002 if (err < 0) 3003 return err; 3004 slave_data = slave_attr; 3005 } 3006 } 3007 3008 if (dev) { 3009 int status = 0; 3010 3011 if (nlh->nlmsg_flags & NLM_F_EXCL) 3012 return -EEXIST; 3013 if (nlh->nlmsg_flags & NLM_F_REPLACE) 3014 return -EOPNOTSUPP; 3015 3016 if (linkinfo[IFLA_INFO_DATA]) { 3017 if (!ops || ops != dev->rtnl_link_ops || 3018 !ops->changelink) 3019 return -EOPNOTSUPP; 3020 3021 err = ops->changelink(dev, tb, data, extack); 3022 if (err < 0) 3023 return err; 3024 status |= DO_SETLINK_NOTIFY; 3025 } 3026 3027 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 3028 if (!m_ops || !m_ops->slave_changelink) 3029 return -EOPNOTSUPP; 3030 3031 err = m_ops->slave_changelink(master_dev, dev, 3032 tb, slave_data, 3033 extack); 3034 if (err < 0) 3035 return err; 3036 status |= DO_SETLINK_NOTIFY; 3037 } 3038 3039 return do_setlink(skb, dev, ifm, extack, tb, ifname, 3040 status); 3041 } 3042 3043 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 3044 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 3045 return rtnl_group_changelink(skb, net, 3046 nla_get_u32(tb[IFLA_GROUP]), 3047 ifm, extack, tb); 3048 return -ENODEV; 3049 } 3050 3051 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 3052 return -EOPNOTSUPP; 3053 3054 if (!ops) { 3055 #ifdef CONFIG_MODULES 3056 if (kind[0]) { 3057 __rtnl_unlock(); 3058 request_module("rtnl-link-%s", kind); 3059 rtnl_lock(); 3060 ops = rtnl_link_ops_get(kind); 3061 if (ops) 3062 goto replay; 3063 } 3064 #endif 3065 return -EOPNOTSUPP; 3066 } 3067 3068 if (!ops->setup) 3069 return -EOPNOTSUPP; 3070 3071 if (!ifname[0]) { 3072 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 3073 name_assign_type = NET_NAME_ENUM; 3074 } 3075 3076 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN); 3077 if (IS_ERR(dest_net)) 3078 return PTR_ERR(dest_net); 3079 3080 if (tb[IFLA_LINK_NETNSID]) { 3081 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 3082 3083 link_net = get_net_ns_by_id(dest_net, id); 3084 if (!link_net) { 3085 err = -EINVAL; 3086 goto out; 3087 } 3088 err = -EPERM; 3089 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 3090 goto out; 3091 } 3092 3093 dev = rtnl_create_link(link_net ? : dest_net, ifname, 3094 name_assign_type, ops, tb); 3095 if (IS_ERR(dev)) { 3096 err = PTR_ERR(dev); 3097 goto out; 3098 } 3099 3100 dev->ifindex = ifm->ifi_index; 3101 3102 if (ops->newlink) { 3103 err = ops->newlink(link_net ? : net, dev, tb, data, 3104 extack); 3105 /* Drivers should call free_netdev() in ->destructor 3106 * and unregister it on failure after registration 3107 * so that device could be finally freed in rtnl_unlock. 3108 */ 3109 if (err < 0) { 3110 /* If device is not registered at all, free it now */ 3111 if (dev->reg_state == NETREG_UNINITIALIZED) 3112 free_netdev(dev); 3113 goto out; 3114 } 3115 } else { 3116 err = register_netdevice(dev); 3117 if (err < 0) { 3118 free_netdev(dev); 3119 goto out; 3120 } 3121 } 3122 err = rtnl_configure_link(dev, ifm); 3123 if (err < 0) 3124 goto out_unregister; 3125 if (link_net) { 3126 err = dev_change_net_namespace(dev, dest_net, ifname); 3127 if (err < 0) 3128 goto out_unregister; 3129 } 3130 if (tb[IFLA_MASTER]) { 3131 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), 3132 extack); 3133 if (err) 3134 goto out_unregister; 3135 } 3136 out: 3137 if (link_net) 3138 put_net(link_net); 3139 put_net(dest_net); 3140 return err; 3141 out_unregister: 3142 if (ops->newlink) { 3143 LIST_HEAD(list_kill); 3144 3145 ops->dellink(dev, &list_kill); 3146 unregister_netdevice_many(&list_kill); 3147 } else { 3148 unregister_netdevice(dev); 3149 } 3150 goto out; 3151 } 3152 } 3153 3154 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3155 struct netlink_ext_ack *extack) 3156 { 3157 struct net *net = sock_net(skb->sk); 3158 struct net *tgt_net = net; 3159 struct ifinfomsg *ifm; 3160 char ifname[IFNAMSIZ]; 3161 struct nlattr *tb[IFLA_MAX+1]; 3162 struct net_device *dev = NULL; 3163 struct sk_buff *nskb; 3164 int netnsid = -1; 3165 int err; 3166 u32 ext_filter_mask = 0; 3167 3168 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 3169 if (err < 0) 3170 return err; 3171 3172 err = rtnl_ensure_unique_netns(tb, extack, true); 3173 if (err < 0) 3174 return err; 3175 3176 if (tb[IFLA_IF_NETNSID]) { 3177 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]); 3178 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid); 3179 if (IS_ERR(tgt_net)) 3180 return PTR_ERR(tgt_net); 3181 } 3182 3183 if (tb[IFLA_IFNAME]) 3184 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 3185 3186 if (tb[IFLA_EXT_MASK]) 3187 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 3188 3189 err = -EINVAL; 3190 ifm = nlmsg_data(nlh); 3191 if (ifm->ifi_index > 0) 3192 dev = __dev_get_by_index(tgt_net, ifm->ifi_index); 3193 else if (tb[IFLA_IFNAME]) 3194 dev = __dev_get_by_name(tgt_net, ifname); 3195 else 3196 goto out; 3197 3198 err = -ENODEV; 3199 if (dev == NULL) 3200 goto out; 3201 3202 err = -ENOBUFS; 3203 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 3204 if (nskb == NULL) 3205 goto out; 3206 3207 err = rtnl_fill_ifinfo(nskb, dev, net, 3208 RTM_NEWLINK, NETLINK_CB(skb).portid, 3209 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 3210 0, NULL, 0, netnsid); 3211 if (err < 0) { 3212 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 3213 WARN_ON(err == -EMSGSIZE); 3214 kfree_skb(nskb); 3215 } else 3216 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 3217 out: 3218 if (netnsid >= 0) 3219 put_net(tgt_net); 3220 3221 return err; 3222 } 3223 3224 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 3225 { 3226 struct net *net = sock_net(skb->sk); 3227 struct net_device *dev; 3228 struct nlattr *tb[IFLA_MAX+1]; 3229 u32 ext_filter_mask = 0; 3230 u16 min_ifinfo_dump_size = 0; 3231 int hdrlen; 3232 3233 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 3234 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 3235 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 3236 3237 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) { 3238 if (tb[IFLA_EXT_MASK]) 3239 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 3240 } 3241 3242 if (!ext_filter_mask) 3243 return NLMSG_GOODSIZE; 3244 /* 3245 * traverse the list of net devices and compute the minimum 3246 * buffer size based upon the filter mask. 3247 */ 3248 rcu_read_lock(); 3249 for_each_netdev_rcu(net, dev) { 3250 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 3251 if_nlmsg_size(dev, 3252 ext_filter_mask)); 3253 } 3254 rcu_read_unlock(); 3255 3256 return nlmsg_total_size(min_ifinfo_dump_size); 3257 } 3258 3259 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 3260 { 3261 int idx; 3262 int s_idx = cb->family; 3263 3264 if (s_idx == 0) 3265 s_idx = 1; 3266 3267 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 3268 struct rtnl_link **tab; 3269 int type = cb->nlh->nlmsg_type-RTM_BASE; 3270 struct rtnl_link *link; 3271 rtnl_dumpit_func dumpit; 3272 3273 if (idx < s_idx || idx == PF_PACKET) 3274 continue; 3275 3276 if (type < 0 || type >= RTM_NR_MSGTYPES) 3277 continue; 3278 3279 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]); 3280 if (!tab) 3281 continue; 3282 3283 link = tab[type]; 3284 if (!link) 3285 continue; 3286 3287 dumpit = link->dumpit; 3288 if (!dumpit) 3289 continue; 3290 3291 if (idx > s_idx) { 3292 memset(&cb->args[0], 0, sizeof(cb->args)); 3293 cb->prev_seq = 0; 3294 cb->seq = 0; 3295 } 3296 if (dumpit(skb, cb)) 3297 break; 3298 } 3299 cb->family = idx; 3300 3301 return skb->len; 3302 } 3303 3304 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 3305 unsigned int change, 3306 u32 event, gfp_t flags, int *new_nsid, 3307 int new_ifindex) 3308 { 3309 struct net *net = dev_net(dev); 3310 struct sk_buff *skb; 3311 int err = -ENOBUFS; 3312 size_t if_info_size; 3313 3314 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 3315 if (skb == NULL) 3316 goto errout; 3317 3318 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev), 3319 type, 0, 0, change, 0, 0, event, 3320 new_nsid, new_ifindex, -1); 3321 if (err < 0) { 3322 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 3323 WARN_ON(err == -EMSGSIZE); 3324 kfree_skb(skb); 3325 goto errout; 3326 } 3327 return skb; 3328 errout: 3329 if (err < 0) 3330 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3331 return NULL; 3332 } 3333 3334 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 3335 { 3336 struct net *net = dev_net(dev); 3337 3338 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 3339 } 3340 3341 static void rtmsg_ifinfo_event(int type, struct net_device *dev, 3342 unsigned int change, u32 event, 3343 gfp_t flags, int *new_nsid, int new_ifindex) 3344 { 3345 struct sk_buff *skb; 3346 3347 if (dev->reg_state != NETREG_REGISTERED) 3348 return; 3349 3350 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid, 3351 new_ifindex); 3352 if (skb) 3353 rtmsg_ifinfo_send(skb, dev, flags); 3354 } 3355 3356 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 3357 gfp_t flags) 3358 { 3359 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, 3360 NULL, 0); 3361 } 3362 3363 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change, 3364 gfp_t flags, int *new_nsid, int new_ifindex) 3365 { 3366 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, 3367 new_nsid, new_ifindex); 3368 } 3369 3370 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 3371 struct net_device *dev, 3372 u8 *addr, u16 vid, u32 pid, u32 seq, 3373 int type, unsigned int flags, 3374 int nlflags, u16 ndm_state) 3375 { 3376 struct nlmsghdr *nlh; 3377 struct ndmsg *ndm; 3378 3379 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 3380 if (!nlh) 3381 return -EMSGSIZE; 3382 3383 ndm = nlmsg_data(nlh); 3384 ndm->ndm_family = AF_BRIDGE; 3385 ndm->ndm_pad1 = 0; 3386 ndm->ndm_pad2 = 0; 3387 ndm->ndm_flags = flags; 3388 ndm->ndm_type = 0; 3389 ndm->ndm_ifindex = dev->ifindex; 3390 ndm->ndm_state = ndm_state; 3391 3392 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 3393 goto nla_put_failure; 3394 if (vid) 3395 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 3396 goto nla_put_failure; 3397 3398 nlmsg_end(skb, nlh); 3399 return 0; 3400 3401 nla_put_failure: 3402 nlmsg_cancel(skb, nlh); 3403 return -EMSGSIZE; 3404 } 3405 3406 static inline size_t rtnl_fdb_nlmsg_size(void) 3407 { 3408 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 3409 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 3410 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 3411 0; 3412 } 3413 3414 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 3415 u16 ndm_state) 3416 { 3417 struct net *net = dev_net(dev); 3418 struct sk_buff *skb; 3419 int err = -ENOBUFS; 3420 3421 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 3422 if (!skb) 3423 goto errout; 3424 3425 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 3426 0, 0, type, NTF_SELF, 0, ndm_state); 3427 if (err < 0) { 3428 kfree_skb(skb); 3429 goto errout; 3430 } 3431 3432 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 3433 return; 3434 errout: 3435 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 3436 } 3437 3438 /** 3439 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 3440 */ 3441 int ndo_dflt_fdb_add(struct ndmsg *ndm, 3442 struct nlattr *tb[], 3443 struct net_device *dev, 3444 const unsigned char *addr, u16 vid, 3445 u16 flags) 3446 { 3447 int err = -EINVAL; 3448 3449 /* If aging addresses are supported device will need to 3450 * implement its own handler for this. 3451 */ 3452 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 3453 pr_info("%s: FDB only supports static addresses\n", dev->name); 3454 return err; 3455 } 3456 3457 if (vid) { 3458 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 3459 return err; 3460 } 3461 3462 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3463 err = dev_uc_add_excl(dev, addr); 3464 else if (is_multicast_ether_addr(addr)) 3465 err = dev_mc_add_excl(dev, addr); 3466 3467 /* Only return duplicate errors if NLM_F_EXCL is set */ 3468 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 3469 err = 0; 3470 3471 return err; 3472 } 3473 EXPORT_SYMBOL(ndo_dflt_fdb_add); 3474 3475 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid, 3476 struct netlink_ext_ack *extack) 3477 { 3478 u16 vid = 0; 3479 3480 if (vlan_attr) { 3481 if (nla_len(vlan_attr) != sizeof(u16)) { 3482 NL_SET_ERR_MSG(extack, "invalid vlan attribute size"); 3483 return -EINVAL; 3484 } 3485 3486 vid = nla_get_u16(vlan_attr); 3487 3488 if (!vid || vid >= VLAN_VID_MASK) { 3489 NL_SET_ERR_MSG(extack, "invalid vlan id"); 3490 return -EINVAL; 3491 } 3492 } 3493 *p_vid = vid; 3494 return 0; 3495 } 3496 3497 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, 3498 struct netlink_ext_ack *extack) 3499 { 3500 struct net *net = sock_net(skb->sk); 3501 struct ndmsg *ndm; 3502 struct nlattr *tb[NDA_MAX+1]; 3503 struct net_device *dev; 3504 u8 *addr; 3505 u16 vid; 3506 int err; 3507 3508 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3509 if (err < 0) 3510 return err; 3511 3512 ndm = nlmsg_data(nlh); 3513 if (ndm->ndm_ifindex == 0) { 3514 NL_SET_ERR_MSG(extack, "invalid ifindex"); 3515 return -EINVAL; 3516 } 3517 3518 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3519 if (dev == NULL) { 3520 NL_SET_ERR_MSG(extack, "unknown ifindex"); 3521 return -ENODEV; 3522 } 3523 3524 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3525 NL_SET_ERR_MSG(extack, "invalid address"); 3526 return -EINVAL; 3527 } 3528 3529 addr = nla_data(tb[NDA_LLADDR]); 3530 3531 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack); 3532 if (err) 3533 return err; 3534 3535 err = -EOPNOTSUPP; 3536 3537 /* Support fdb on master device the net/bridge default case */ 3538 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3539 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3540 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3541 const struct net_device_ops *ops = br_dev->netdev_ops; 3542 3543 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3544 nlh->nlmsg_flags); 3545 if (err) 3546 goto out; 3547 else 3548 ndm->ndm_flags &= ~NTF_MASTER; 3549 } 3550 3551 /* Embedded bridge, macvlan, and any other device support */ 3552 if ((ndm->ndm_flags & NTF_SELF)) { 3553 if (dev->netdev_ops->ndo_fdb_add) 3554 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3555 vid, 3556 nlh->nlmsg_flags); 3557 else 3558 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3559 nlh->nlmsg_flags); 3560 3561 if (!err) { 3562 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3563 ndm->ndm_state); 3564 ndm->ndm_flags &= ~NTF_SELF; 3565 } 3566 } 3567 out: 3568 return err; 3569 } 3570 3571 /** 3572 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3573 */ 3574 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3575 struct nlattr *tb[], 3576 struct net_device *dev, 3577 const unsigned char *addr, u16 vid) 3578 { 3579 int err = -EINVAL; 3580 3581 /* If aging addresses are supported device will need to 3582 * implement its own handler for this. 3583 */ 3584 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3585 pr_info("%s: FDB only supports static addresses\n", dev->name); 3586 return err; 3587 } 3588 3589 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3590 err = dev_uc_del(dev, addr); 3591 else if (is_multicast_ether_addr(addr)) 3592 err = dev_mc_del(dev, addr); 3593 3594 return err; 3595 } 3596 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3597 3598 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, 3599 struct netlink_ext_ack *extack) 3600 { 3601 struct net *net = sock_net(skb->sk); 3602 struct ndmsg *ndm; 3603 struct nlattr *tb[NDA_MAX+1]; 3604 struct net_device *dev; 3605 int err = -EINVAL; 3606 __u8 *addr; 3607 u16 vid; 3608 3609 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3610 return -EPERM; 3611 3612 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3613 if (err < 0) 3614 return err; 3615 3616 ndm = nlmsg_data(nlh); 3617 if (ndm->ndm_ifindex == 0) { 3618 NL_SET_ERR_MSG(extack, "invalid ifindex"); 3619 return -EINVAL; 3620 } 3621 3622 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3623 if (dev == NULL) { 3624 NL_SET_ERR_MSG(extack, "unknown ifindex"); 3625 return -ENODEV; 3626 } 3627 3628 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3629 NL_SET_ERR_MSG(extack, "invalid address"); 3630 return -EINVAL; 3631 } 3632 3633 addr = nla_data(tb[NDA_LLADDR]); 3634 3635 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack); 3636 if (err) 3637 return err; 3638 3639 err = -EOPNOTSUPP; 3640 3641 /* Support fdb on master device the net/bridge default case */ 3642 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3643 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3644 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3645 const struct net_device_ops *ops = br_dev->netdev_ops; 3646 3647 if (ops->ndo_fdb_del) 3648 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3649 3650 if (err) 3651 goto out; 3652 else 3653 ndm->ndm_flags &= ~NTF_MASTER; 3654 } 3655 3656 /* Embedded bridge, macvlan, and any other device support */ 3657 if (ndm->ndm_flags & NTF_SELF) { 3658 if (dev->netdev_ops->ndo_fdb_del) 3659 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3660 vid); 3661 else 3662 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3663 3664 if (!err) { 3665 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3666 ndm->ndm_state); 3667 ndm->ndm_flags &= ~NTF_SELF; 3668 } 3669 } 3670 out: 3671 return err; 3672 } 3673 3674 static int nlmsg_populate_fdb(struct sk_buff *skb, 3675 struct netlink_callback *cb, 3676 struct net_device *dev, 3677 int *idx, 3678 struct netdev_hw_addr_list *list) 3679 { 3680 struct netdev_hw_addr *ha; 3681 int err; 3682 u32 portid, seq; 3683 3684 portid = NETLINK_CB(cb->skb).portid; 3685 seq = cb->nlh->nlmsg_seq; 3686 3687 list_for_each_entry(ha, &list->list, list) { 3688 if (*idx < cb->args[2]) 3689 goto skip; 3690 3691 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3692 portid, seq, 3693 RTM_NEWNEIGH, NTF_SELF, 3694 NLM_F_MULTI, NUD_PERMANENT); 3695 if (err < 0) 3696 return err; 3697 skip: 3698 *idx += 1; 3699 } 3700 return 0; 3701 } 3702 3703 /** 3704 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3705 * @nlh: netlink message header 3706 * @dev: netdevice 3707 * 3708 * Default netdevice operation to dump the existing unicast address list. 3709 * Returns number of addresses from list put in skb. 3710 */ 3711 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3712 struct netlink_callback *cb, 3713 struct net_device *dev, 3714 struct net_device *filter_dev, 3715 int *idx) 3716 { 3717 int err; 3718 3719 netif_addr_lock_bh(dev); 3720 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3721 if (err) 3722 goto out; 3723 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3724 out: 3725 netif_addr_unlock_bh(dev); 3726 return err; 3727 } 3728 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3729 3730 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 3731 { 3732 struct net_device *dev; 3733 struct nlattr *tb[IFLA_MAX+1]; 3734 struct net_device *br_dev = NULL; 3735 const struct net_device_ops *ops = NULL; 3736 const struct net_device_ops *cops = NULL; 3737 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 3738 struct net *net = sock_net(skb->sk); 3739 struct hlist_head *head; 3740 int brport_idx = 0; 3741 int br_idx = 0; 3742 int h, s_h; 3743 int idx = 0, s_idx; 3744 int err = 0; 3745 int fidx = 0; 3746 3747 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, 3748 IFLA_MAX, ifla_policy, NULL); 3749 if (err < 0) { 3750 return -EINVAL; 3751 } else if (err == 0) { 3752 if (tb[IFLA_MASTER]) 3753 br_idx = nla_get_u32(tb[IFLA_MASTER]); 3754 } 3755 3756 brport_idx = ifm->ifi_index; 3757 3758 if (br_idx) { 3759 br_dev = __dev_get_by_index(net, br_idx); 3760 if (!br_dev) 3761 return -ENODEV; 3762 3763 ops = br_dev->netdev_ops; 3764 } 3765 3766 s_h = cb->args[0]; 3767 s_idx = cb->args[1]; 3768 3769 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3770 idx = 0; 3771 head = &net->dev_index_head[h]; 3772 hlist_for_each_entry(dev, head, index_hlist) { 3773 3774 if (brport_idx && (dev->ifindex != brport_idx)) 3775 continue; 3776 3777 if (!br_idx) { /* user did not specify a specific bridge */ 3778 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3779 br_dev = netdev_master_upper_dev_get(dev); 3780 cops = br_dev->netdev_ops; 3781 } 3782 } else { 3783 if (dev != br_dev && 3784 !(dev->priv_flags & IFF_BRIDGE_PORT)) 3785 continue; 3786 3787 if (br_dev != netdev_master_upper_dev_get(dev) && 3788 !(dev->priv_flags & IFF_EBRIDGE)) 3789 continue; 3790 cops = ops; 3791 } 3792 3793 if (idx < s_idx) 3794 goto cont; 3795 3796 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3797 if (cops && cops->ndo_fdb_dump) { 3798 err = cops->ndo_fdb_dump(skb, cb, 3799 br_dev, dev, 3800 &fidx); 3801 if (err == -EMSGSIZE) 3802 goto out; 3803 } 3804 } 3805 3806 if (dev->netdev_ops->ndo_fdb_dump) 3807 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 3808 dev, NULL, 3809 &fidx); 3810 else 3811 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 3812 &fidx); 3813 if (err == -EMSGSIZE) 3814 goto out; 3815 3816 cops = NULL; 3817 3818 /* reset fdb offset to 0 for rest of the interfaces */ 3819 cb->args[2] = 0; 3820 fidx = 0; 3821 cont: 3822 idx++; 3823 } 3824 } 3825 3826 out: 3827 cb->args[0] = h; 3828 cb->args[1] = idx; 3829 cb->args[2] = fidx; 3830 3831 return skb->len; 3832 } 3833 3834 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 3835 unsigned int attrnum, unsigned int flag) 3836 { 3837 if (mask & flag) 3838 return nla_put_u8(skb, attrnum, !!(flags & flag)); 3839 return 0; 3840 } 3841 3842 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 3843 struct net_device *dev, u16 mode, 3844 u32 flags, u32 mask, int nlflags, 3845 u32 filter_mask, 3846 int (*vlan_fill)(struct sk_buff *skb, 3847 struct net_device *dev, 3848 u32 filter_mask)) 3849 { 3850 struct nlmsghdr *nlh; 3851 struct ifinfomsg *ifm; 3852 struct nlattr *br_afspec; 3853 struct nlattr *protinfo; 3854 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 3855 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3856 int err = 0; 3857 3858 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 3859 if (nlh == NULL) 3860 return -EMSGSIZE; 3861 3862 ifm = nlmsg_data(nlh); 3863 ifm->ifi_family = AF_BRIDGE; 3864 ifm->__ifi_pad = 0; 3865 ifm->ifi_type = dev->type; 3866 ifm->ifi_index = dev->ifindex; 3867 ifm->ifi_flags = dev_get_flags(dev); 3868 ifm->ifi_change = 0; 3869 3870 3871 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 3872 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 3873 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 3874 (br_dev && 3875 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 3876 (dev->addr_len && 3877 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 3878 (dev->ifindex != dev_get_iflink(dev) && 3879 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 3880 goto nla_put_failure; 3881 3882 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 3883 if (!br_afspec) 3884 goto nla_put_failure; 3885 3886 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 3887 nla_nest_cancel(skb, br_afspec); 3888 goto nla_put_failure; 3889 } 3890 3891 if (mode != BRIDGE_MODE_UNDEF) { 3892 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 3893 nla_nest_cancel(skb, br_afspec); 3894 goto nla_put_failure; 3895 } 3896 } 3897 if (vlan_fill) { 3898 err = vlan_fill(skb, dev, filter_mask); 3899 if (err) { 3900 nla_nest_cancel(skb, br_afspec); 3901 goto nla_put_failure; 3902 } 3903 } 3904 nla_nest_end(skb, br_afspec); 3905 3906 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED); 3907 if (!protinfo) 3908 goto nla_put_failure; 3909 3910 if (brport_nla_put_flag(skb, flags, mask, 3911 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 3912 brport_nla_put_flag(skb, flags, mask, 3913 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 3914 brport_nla_put_flag(skb, flags, mask, 3915 IFLA_BRPORT_FAST_LEAVE, 3916 BR_MULTICAST_FAST_LEAVE) || 3917 brport_nla_put_flag(skb, flags, mask, 3918 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 3919 brport_nla_put_flag(skb, flags, mask, 3920 IFLA_BRPORT_LEARNING, BR_LEARNING) || 3921 brport_nla_put_flag(skb, flags, mask, 3922 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 3923 brport_nla_put_flag(skb, flags, mask, 3924 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 3925 brport_nla_put_flag(skb, flags, mask, 3926 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 3927 nla_nest_cancel(skb, protinfo); 3928 goto nla_put_failure; 3929 } 3930 3931 nla_nest_end(skb, protinfo); 3932 3933 nlmsg_end(skb, nlh); 3934 return 0; 3935 nla_put_failure: 3936 nlmsg_cancel(skb, nlh); 3937 return err ? err : -EMSGSIZE; 3938 } 3939 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 3940 3941 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 3942 { 3943 struct net *net = sock_net(skb->sk); 3944 struct net_device *dev; 3945 int idx = 0; 3946 u32 portid = NETLINK_CB(cb->skb).portid; 3947 u32 seq = cb->nlh->nlmsg_seq; 3948 u32 filter_mask = 0; 3949 int err; 3950 3951 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 3952 struct nlattr *extfilt; 3953 3954 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 3955 IFLA_EXT_MASK); 3956 if (extfilt) { 3957 if (nla_len(extfilt) < sizeof(filter_mask)) 3958 return -EINVAL; 3959 3960 filter_mask = nla_get_u32(extfilt); 3961 } 3962 } 3963 3964 rcu_read_lock(); 3965 for_each_netdev_rcu(net, dev) { 3966 const struct net_device_ops *ops = dev->netdev_ops; 3967 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3968 3969 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 3970 if (idx >= cb->args[0]) { 3971 err = br_dev->netdev_ops->ndo_bridge_getlink( 3972 skb, portid, seq, dev, 3973 filter_mask, NLM_F_MULTI); 3974 if (err < 0 && err != -EOPNOTSUPP) { 3975 if (likely(skb->len)) 3976 break; 3977 3978 goto out_err; 3979 } 3980 } 3981 idx++; 3982 } 3983 3984 if (ops->ndo_bridge_getlink) { 3985 if (idx >= cb->args[0]) { 3986 err = ops->ndo_bridge_getlink(skb, portid, 3987 seq, dev, 3988 filter_mask, 3989 NLM_F_MULTI); 3990 if (err < 0 && err != -EOPNOTSUPP) { 3991 if (likely(skb->len)) 3992 break; 3993 3994 goto out_err; 3995 } 3996 } 3997 idx++; 3998 } 3999 } 4000 err = skb->len; 4001 out_err: 4002 rcu_read_unlock(); 4003 cb->args[0] = idx; 4004 4005 return err; 4006 } 4007 4008 static inline size_t bridge_nlmsg_size(void) 4009 { 4010 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 4011 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 4012 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 4013 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 4014 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 4015 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 4016 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 4017 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 4018 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 4019 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 4020 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 4021 } 4022 4023 static int rtnl_bridge_notify(struct net_device *dev) 4024 { 4025 struct net *net = dev_net(dev); 4026 struct sk_buff *skb; 4027 int err = -EOPNOTSUPP; 4028 4029 if (!dev->netdev_ops->ndo_bridge_getlink) 4030 return 0; 4031 4032 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 4033 if (!skb) { 4034 err = -ENOMEM; 4035 goto errout; 4036 } 4037 4038 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 4039 if (err < 0) 4040 goto errout; 4041 4042 if (!skb->len) 4043 goto errout; 4044 4045 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 4046 return 0; 4047 errout: 4048 WARN_ON(err == -EMSGSIZE); 4049 kfree_skb(skb); 4050 if (err) 4051 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 4052 return err; 4053 } 4054 4055 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 4056 struct netlink_ext_ack *extack) 4057 { 4058 struct net *net = sock_net(skb->sk); 4059 struct ifinfomsg *ifm; 4060 struct net_device *dev; 4061 struct nlattr *br_spec, *attr = NULL; 4062 int rem, err = -EOPNOTSUPP; 4063 u16 flags = 0; 4064 bool have_flags = false; 4065 4066 if (nlmsg_len(nlh) < sizeof(*ifm)) 4067 return -EINVAL; 4068 4069 ifm = nlmsg_data(nlh); 4070 if (ifm->ifi_family != AF_BRIDGE) 4071 return -EPFNOSUPPORT; 4072 4073 dev = __dev_get_by_index(net, ifm->ifi_index); 4074 if (!dev) { 4075 NL_SET_ERR_MSG(extack, "unknown ifindex"); 4076 return -ENODEV; 4077 } 4078 4079 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 4080 if (br_spec) { 4081 nla_for_each_nested(attr, br_spec, rem) { 4082 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 4083 if (nla_len(attr) < sizeof(flags)) 4084 return -EINVAL; 4085 4086 have_flags = true; 4087 flags = nla_get_u16(attr); 4088 break; 4089 } 4090 } 4091 } 4092 4093 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 4094 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4095 4096 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 4097 err = -EOPNOTSUPP; 4098 goto out; 4099 } 4100 4101 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags); 4102 if (err) 4103 goto out; 4104 4105 flags &= ~BRIDGE_FLAGS_MASTER; 4106 } 4107 4108 if ((flags & BRIDGE_FLAGS_SELF)) { 4109 if (!dev->netdev_ops->ndo_bridge_setlink) 4110 err = -EOPNOTSUPP; 4111 else 4112 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 4113 flags); 4114 if (!err) { 4115 flags &= ~BRIDGE_FLAGS_SELF; 4116 4117 /* Generate event to notify upper layer of bridge 4118 * change 4119 */ 4120 err = rtnl_bridge_notify(dev); 4121 } 4122 } 4123 4124 if (have_flags) 4125 memcpy(nla_data(attr), &flags, sizeof(flags)); 4126 out: 4127 return err; 4128 } 4129 4130 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 4131 struct netlink_ext_ack *extack) 4132 { 4133 struct net *net = sock_net(skb->sk); 4134 struct ifinfomsg *ifm; 4135 struct net_device *dev; 4136 struct nlattr *br_spec, *attr = NULL; 4137 int rem, err = -EOPNOTSUPP; 4138 u16 flags = 0; 4139 bool have_flags = false; 4140 4141 if (nlmsg_len(nlh) < sizeof(*ifm)) 4142 return -EINVAL; 4143 4144 ifm = nlmsg_data(nlh); 4145 if (ifm->ifi_family != AF_BRIDGE) 4146 return -EPFNOSUPPORT; 4147 4148 dev = __dev_get_by_index(net, ifm->ifi_index); 4149 if (!dev) { 4150 NL_SET_ERR_MSG(extack, "unknown ifindex"); 4151 return -ENODEV; 4152 } 4153 4154 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 4155 if (br_spec) { 4156 nla_for_each_nested(attr, br_spec, rem) { 4157 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 4158 if (nla_len(attr) < sizeof(flags)) 4159 return -EINVAL; 4160 4161 have_flags = true; 4162 flags = nla_get_u16(attr); 4163 break; 4164 } 4165 } 4166 } 4167 4168 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 4169 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4170 4171 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 4172 err = -EOPNOTSUPP; 4173 goto out; 4174 } 4175 4176 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 4177 if (err) 4178 goto out; 4179 4180 flags &= ~BRIDGE_FLAGS_MASTER; 4181 } 4182 4183 if ((flags & BRIDGE_FLAGS_SELF)) { 4184 if (!dev->netdev_ops->ndo_bridge_dellink) 4185 err = -EOPNOTSUPP; 4186 else 4187 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 4188 flags); 4189 4190 if (!err) { 4191 flags &= ~BRIDGE_FLAGS_SELF; 4192 4193 /* Generate event to notify upper layer of bridge 4194 * change 4195 */ 4196 err = rtnl_bridge_notify(dev); 4197 } 4198 } 4199 4200 if (have_flags) 4201 memcpy(nla_data(attr), &flags, sizeof(flags)); 4202 out: 4203 return err; 4204 } 4205 4206 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 4207 { 4208 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 4209 (!idxattr || idxattr == attrid); 4210 } 4211 4212 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 4213 static int rtnl_get_offload_stats_attr_size(int attr_id) 4214 { 4215 switch (attr_id) { 4216 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 4217 return sizeof(struct rtnl_link_stats64); 4218 } 4219 4220 return 0; 4221 } 4222 4223 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 4224 int *prividx) 4225 { 4226 struct nlattr *attr = NULL; 4227 int attr_id, size; 4228 void *attr_data; 4229 int err; 4230 4231 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 4232 dev->netdev_ops->ndo_get_offload_stats)) 4233 return -ENODATA; 4234 4235 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 4236 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 4237 if (attr_id < *prividx) 4238 continue; 4239 4240 size = rtnl_get_offload_stats_attr_size(attr_id); 4241 if (!size) 4242 continue; 4243 4244 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 4245 continue; 4246 4247 attr = nla_reserve_64bit(skb, attr_id, size, 4248 IFLA_OFFLOAD_XSTATS_UNSPEC); 4249 if (!attr) 4250 goto nla_put_failure; 4251 4252 attr_data = nla_data(attr); 4253 memset(attr_data, 0, size); 4254 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 4255 attr_data); 4256 if (err) 4257 goto get_offload_stats_failure; 4258 } 4259 4260 if (!attr) 4261 return -ENODATA; 4262 4263 *prividx = 0; 4264 return 0; 4265 4266 nla_put_failure: 4267 err = -EMSGSIZE; 4268 get_offload_stats_failure: 4269 *prividx = attr_id; 4270 return err; 4271 } 4272 4273 static int rtnl_get_offload_stats_size(const struct net_device *dev) 4274 { 4275 int nla_size = 0; 4276 int attr_id; 4277 int size; 4278 4279 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 4280 dev->netdev_ops->ndo_get_offload_stats)) 4281 return 0; 4282 4283 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 4284 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 4285 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 4286 continue; 4287 size = rtnl_get_offload_stats_attr_size(attr_id); 4288 nla_size += nla_total_size_64bit(size); 4289 } 4290 4291 if (nla_size != 0) 4292 nla_size += nla_total_size(0); 4293 4294 return nla_size; 4295 } 4296 4297 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 4298 int type, u32 pid, u32 seq, u32 change, 4299 unsigned int flags, unsigned int filter_mask, 4300 int *idxattr, int *prividx) 4301 { 4302 struct if_stats_msg *ifsm; 4303 struct nlmsghdr *nlh; 4304 struct nlattr *attr; 4305 int s_prividx = *prividx; 4306 int err; 4307 4308 ASSERT_RTNL(); 4309 4310 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 4311 if (!nlh) 4312 return -EMSGSIZE; 4313 4314 ifsm = nlmsg_data(nlh); 4315 ifsm->family = PF_UNSPEC; 4316 ifsm->pad1 = 0; 4317 ifsm->pad2 = 0; 4318 ifsm->ifindex = dev->ifindex; 4319 ifsm->filter_mask = filter_mask; 4320 4321 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 4322 struct rtnl_link_stats64 *sp; 4323 4324 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 4325 sizeof(struct rtnl_link_stats64), 4326 IFLA_STATS_UNSPEC); 4327 if (!attr) 4328 goto nla_put_failure; 4329 4330 sp = nla_data(attr); 4331 dev_get_stats(dev, sp); 4332 } 4333 4334 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 4335 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 4336 4337 if (ops && ops->fill_linkxstats) { 4338 *idxattr = IFLA_STATS_LINK_XSTATS; 4339 attr = nla_nest_start(skb, 4340 IFLA_STATS_LINK_XSTATS); 4341 if (!attr) 4342 goto nla_put_failure; 4343 4344 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 4345 nla_nest_end(skb, attr); 4346 if (err) 4347 goto nla_put_failure; 4348 *idxattr = 0; 4349 } 4350 } 4351 4352 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 4353 *idxattr)) { 4354 const struct rtnl_link_ops *ops = NULL; 4355 const struct net_device *master; 4356 4357 master = netdev_master_upper_dev_get(dev); 4358 if (master) 4359 ops = master->rtnl_link_ops; 4360 if (ops && ops->fill_linkxstats) { 4361 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 4362 attr = nla_nest_start(skb, 4363 IFLA_STATS_LINK_XSTATS_SLAVE); 4364 if (!attr) 4365 goto nla_put_failure; 4366 4367 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 4368 nla_nest_end(skb, attr); 4369 if (err) 4370 goto nla_put_failure; 4371 *idxattr = 0; 4372 } 4373 } 4374 4375 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 4376 *idxattr)) { 4377 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 4378 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS); 4379 if (!attr) 4380 goto nla_put_failure; 4381 4382 err = rtnl_get_offload_stats(skb, dev, prividx); 4383 if (err == -ENODATA) 4384 nla_nest_cancel(skb, attr); 4385 else 4386 nla_nest_end(skb, attr); 4387 4388 if (err && err != -ENODATA) 4389 goto nla_put_failure; 4390 *idxattr = 0; 4391 } 4392 4393 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 4394 struct rtnl_af_ops *af_ops; 4395 4396 *idxattr = IFLA_STATS_AF_SPEC; 4397 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC); 4398 if (!attr) 4399 goto nla_put_failure; 4400 4401 rcu_read_lock(); 4402 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 4403 if (af_ops->fill_stats_af) { 4404 struct nlattr *af; 4405 int err; 4406 4407 af = nla_nest_start(skb, af_ops->family); 4408 if (!af) { 4409 rcu_read_unlock(); 4410 goto nla_put_failure; 4411 } 4412 err = af_ops->fill_stats_af(skb, dev); 4413 4414 if (err == -ENODATA) { 4415 nla_nest_cancel(skb, af); 4416 } else if (err < 0) { 4417 rcu_read_unlock(); 4418 goto nla_put_failure; 4419 } 4420 4421 nla_nest_end(skb, af); 4422 } 4423 } 4424 rcu_read_unlock(); 4425 4426 nla_nest_end(skb, attr); 4427 4428 *idxattr = 0; 4429 } 4430 4431 nlmsg_end(skb, nlh); 4432 4433 return 0; 4434 4435 nla_put_failure: 4436 /* not a multi message or no progress mean a real error */ 4437 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 4438 nlmsg_cancel(skb, nlh); 4439 else 4440 nlmsg_end(skb, nlh); 4441 4442 return -EMSGSIZE; 4443 } 4444 4445 static size_t if_nlmsg_stats_size(const struct net_device *dev, 4446 u32 filter_mask) 4447 { 4448 size_t size = 0; 4449 4450 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 4451 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 4452 4453 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 4454 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 4455 int attr = IFLA_STATS_LINK_XSTATS; 4456 4457 if (ops && ops->get_linkxstats_size) { 4458 size += nla_total_size(ops->get_linkxstats_size(dev, 4459 attr)); 4460 /* for IFLA_STATS_LINK_XSTATS */ 4461 size += nla_total_size(0); 4462 } 4463 } 4464 4465 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 4466 struct net_device *_dev = (struct net_device *)dev; 4467 const struct rtnl_link_ops *ops = NULL; 4468 const struct net_device *master; 4469 4470 /* netdev_master_upper_dev_get can't take const */ 4471 master = netdev_master_upper_dev_get(_dev); 4472 if (master) 4473 ops = master->rtnl_link_ops; 4474 if (ops && ops->get_linkxstats_size) { 4475 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 4476 4477 size += nla_total_size(ops->get_linkxstats_size(dev, 4478 attr)); 4479 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 4480 size += nla_total_size(0); 4481 } 4482 } 4483 4484 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 4485 size += rtnl_get_offload_stats_size(dev); 4486 4487 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 4488 struct rtnl_af_ops *af_ops; 4489 4490 /* for IFLA_STATS_AF_SPEC */ 4491 size += nla_total_size(0); 4492 4493 rcu_read_lock(); 4494 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 4495 if (af_ops->get_stats_af_size) { 4496 size += nla_total_size( 4497 af_ops->get_stats_af_size(dev)); 4498 4499 /* for AF_* */ 4500 size += nla_total_size(0); 4501 } 4502 } 4503 rcu_read_unlock(); 4504 } 4505 4506 return size; 4507 } 4508 4509 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh, 4510 struct netlink_ext_ack *extack) 4511 { 4512 struct net *net = sock_net(skb->sk); 4513 struct net_device *dev = NULL; 4514 int idxattr = 0, prividx = 0; 4515 struct if_stats_msg *ifsm; 4516 struct sk_buff *nskb; 4517 u32 filter_mask; 4518 int err; 4519 4520 if (nlmsg_len(nlh) < sizeof(*ifsm)) 4521 return -EINVAL; 4522 4523 ifsm = nlmsg_data(nlh); 4524 if (ifsm->ifindex > 0) 4525 dev = __dev_get_by_index(net, ifsm->ifindex); 4526 else 4527 return -EINVAL; 4528 4529 if (!dev) 4530 return -ENODEV; 4531 4532 filter_mask = ifsm->filter_mask; 4533 if (!filter_mask) 4534 return -EINVAL; 4535 4536 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 4537 if (!nskb) 4538 return -ENOBUFS; 4539 4540 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 4541 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 4542 0, filter_mask, &idxattr, &prividx); 4543 if (err < 0) { 4544 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 4545 WARN_ON(err == -EMSGSIZE); 4546 kfree_skb(nskb); 4547 } else { 4548 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 4549 } 4550 4551 return err; 4552 } 4553 4554 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 4555 { 4556 int h, s_h, err, s_idx, s_idxattr, s_prividx; 4557 struct net *net = sock_net(skb->sk); 4558 unsigned int flags = NLM_F_MULTI; 4559 struct if_stats_msg *ifsm; 4560 struct hlist_head *head; 4561 struct net_device *dev; 4562 u32 filter_mask = 0; 4563 int idx = 0; 4564 4565 s_h = cb->args[0]; 4566 s_idx = cb->args[1]; 4567 s_idxattr = cb->args[2]; 4568 s_prividx = cb->args[3]; 4569 4570 cb->seq = net->dev_base_seq; 4571 4572 if (nlmsg_len(cb->nlh) < sizeof(*ifsm)) 4573 return -EINVAL; 4574 4575 ifsm = nlmsg_data(cb->nlh); 4576 filter_mask = ifsm->filter_mask; 4577 if (!filter_mask) 4578 return -EINVAL; 4579 4580 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4581 idx = 0; 4582 head = &net->dev_index_head[h]; 4583 hlist_for_each_entry(dev, head, index_hlist) { 4584 if (idx < s_idx) 4585 goto cont; 4586 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 4587 NETLINK_CB(cb->skb).portid, 4588 cb->nlh->nlmsg_seq, 0, 4589 flags, filter_mask, 4590 &s_idxattr, &s_prividx); 4591 /* If we ran out of room on the first message, 4592 * we're in trouble 4593 */ 4594 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 4595 4596 if (err < 0) 4597 goto out; 4598 s_prividx = 0; 4599 s_idxattr = 0; 4600 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4601 cont: 4602 idx++; 4603 } 4604 } 4605 out: 4606 cb->args[3] = s_prividx; 4607 cb->args[2] = s_idxattr; 4608 cb->args[1] = idx; 4609 cb->args[0] = h; 4610 4611 return skb->len; 4612 } 4613 4614 /* Process one rtnetlink message. */ 4615 4616 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, 4617 struct netlink_ext_ack *extack) 4618 { 4619 struct net *net = sock_net(skb->sk); 4620 struct rtnl_link *link; 4621 struct module *owner; 4622 int err = -EOPNOTSUPP; 4623 rtnl_doit_func doit; 4624 unsigned int flags; 4625 int kind; 4626 int family; 4627 int type; 4628 4629 type = nlh->nlmsg_type; 4630 if (type > RTM_MAX) 4631 return -EOPNOTSUPP; 4632 4633 type -= RTM_BASE; 4634 4635 /* All the messages must have at least 1 byte length */ 4636 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 4637 return 0; 4638 4639 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 4640 kind = type&3; 4641 4642 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 4643 return -EPERM; 4644 4645 rcu_read_lock(); 4646 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 4647 struct sock *rtnl; 4648 rtnl_dumpit_func dumpit; 4649 u16 min_dump_alloc = 0; 4650 4651 link = rtnl_get_link(family, type); 4652 if (!link || !link->dumpit) { 4653 family = PF_UNSPEC; 4654 link = rtnl_get_link(family, type); 4655 if (!link || !link->dumpit) 4656 goto err_unlock; 4657 } 4658 owner = link->owner; 4659 dumpit = link->dumpit; 4660 4661 if (type == RTM_GETLINK - RTM_BASE) 4662 min_dump_alloc = rtnl_calcit(skb, nlh); 4663 4664 err = 0; 4665 /* need to do this before rcu_read_unlock() */ 4666 if (!try_module_get(owner)) 4667 err = -EPROTONOSUPPORT; 4668 4669 rcu_read_unlock(); 4670 4671 rtnl = net->rtnl; 4672 if (err == 0) { 4673 struct netlink_dump_control c = { 4674 .dump = dumpit, 4675 .min_dump_alloc = min_dump_alloc, 4676 .module = owner, 4677 }; 4678 err = netlink_dump_start(rtnl, skb, nlh, &c); 4679 /* netlink_dump_start() will keep a reference on 4680 * module if dump is still in progress. 4681 */ 4682 module_put(owner); 4683 } 4684 return err; 4685 } 4686 4687 link = rtnl_get_link(family, type); 4688 if (!link || !link->doit) { 4689 family = PF_UNSPEC; 4690 link = rtnl_get_link(PF_UNSPEC, type); 4691 if (!link || !link->doit) 4692 goto out_unlock; 4693 } 4694 4695 owner = link->owner; 4696 if (!try_module_get(owner)) { 4697 err = -EPROTONOSUPPORT; 4698 goto out_unlock; 4699 } 4700 4701 flags = link->flags; 4702 if (flags & RTNL_FLAG_DOIT_UNLOCKED) { 4703 doit = link->doit; 4704 rcu_read_unlock(); 4705 if (doit) 4706 err = doit(skb, nlh, extack); 4707 module_put(owner); 4708 return err; 4709 } 4710 rcu_read_unlock(); 4711 4712 rtnl_lock(); 4713 link = rtnl_get_link(family, type); 4714 if (link && link->doit) 4715 err = link->doit(skb, nlh, extack); 4716 rtnl_unlock(); 4717 4718 module_put(owner); 4719 4720 return err; 4721 4722 out_unlock: 4723 rcu_read_unlock(); 4724 return err; 4725 4726 err_unlock: 4727 rcu_read_unlock(); 4728 return -EOPNOTSUPP; 4729 } 4730 4731 static void rtnetlink_rcv(struct sk_buff *skb) 4732 { 4733 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 4734 } 4735 4736 static int rtnetlink_bind(struct net *net, int group) 4737 { 4738 switch (group) { 4739 case RTNLGRP_IPV4_MROUTE_R: 4740 case RTNLGRP_IPV6_MROUTE_R: 4741 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 4742 return -EPERM; 4743 break; 4744 } 4745 return 0; 4746 } 4747 4748 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 4749 { 4750 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4751 4752 switch (event) { 4753 case NETDEV_REBOOT: 4754 case NETDEV_CHANGEMTU: 4755 case NETDEV_CHANGEADDR: 4756 case NETDEV_CHANGENAME: 4757 case NETDEV_FEAT_CHANGE: 4758 case NETDEV_BONDING_FAILOVER: 4759 case NETDEV_POST_TYPE_CHANGE: 4760 case NETDEV_NOTIFY_PEERS: 4761 case NETDEV_CHANGEUPPER: 4762 case NETDEV_RESEND_IGMP: 4763 case NETDEV_CHANGEINFODATA: 4764 case NETDEV_CHANGELOWERSTATE: 4765 case NETDEV_CHANGE_TX_QUEUE_LEN: 4766 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event), 4767 GFP_KERNEL, NULL, 0); 4768 break; 4769 default: 4770 break; 4771 } 4772 return NOTIFY_DONE; 4773 } 4774 4775 static struct notifier_block rtnetlink_dev_notifier = { 4776 .notifier_call = rtnetlink_event, 4777 }; 4778 4779 4780 static int __net_init rtnetlink_net_init(struct net *net) 4781 { 4782 struct sock *sk; 4783 struct netlink_kernel_cfg cfg = { 4784 .groups = RTNLGRP_MAX, 4785 .input = rtnetlink_rcv, 4786 .cb_mutex = &rtnl_mutex, 4787 .flags = NL_CFG_F_NONROOT_RECV, 4788 .bind = rtnetlink_bind, 4789 }; 4790 4791 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 4792 if (!sk) 4793 return -ENOMEM; 4794 net->rtnl = sk; 4795 return 0; 4796 } 4797 4798 static void __net_exit rtnetlink_net_exit(struct net *net) 4799 { 4800 netlink_kernel_release(net->rtnl); 4801 net->rtnl = NULL; 4802 } 4803 4804 static struct pernet_operations rtnetlink_net_ops = { 4805 .init = rtnetlink_net_init, 4806 .exit = rtnetlink_net_exit, 4807 }; 4808 4809 void __init rtnetlink_init(void) 4810 { 4811 if (register_pernet_subsys(&rtnetlink_net_ops)) 4812 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 4813 4814 register_netdevice_notifier(&rtnetlink_dev_notifier); 4815 4816 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 4817 rtnl_dump_ifinfo, 0); 4818 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0); 4819 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0); 4820 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0); 4821 4822 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0); 4823 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0); 4824 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0); 4825 4826 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0); 4827 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0); 4828 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0); 4829 4830 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0); 4831 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0); 4832 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0); 4833 4834 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 4835 0); 4836 } 4837