1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Routing netlink socket interface: protocol independent part. 8 * 9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 10 * 11 * Fixes: 12 * Vitaly E. Lavrov RTA_OK arithmetics was wrong. 13 */ 14 15 #include <linux/bitops.h> 16 #include <linux/errno.h> 17 #include <linux/module.h> 18 #include <linux/types.h> 19 #include <linux/socket.h> 20 #include <linux/kernel.h> 21 #include <linux/timer.h> 22 #include <linux/string.h> 23 #include <linux/sockios.h> 24 #include <linux/net.h> 25 #include <linux/fcntl.h> 26 #include <linux/mm.h> 27 #include <linux/slab.h> 28 #include <linux/interrupt.h> 29 #include <linux/capability.h> 30 #include <linux/skbuff.h> 31 #include <linux/init.h> 32 #include <linux/security.h> 33 #include <linux/mutex.h> 34 #include <linux/if_addr.h> 35 #include <linux/if_bridge.h> 36 #include <linux/if_vlan.h> 37 #include <linux/pci.h> 38 #include <linux/etherdevice.h> 39 #include <linux/bpf.h> 40 41 #include <linux/uaccess.h> 42 43 #include <linux/inet.h> 44 #include <linux/netdevice.h> 45 #include <net/ip.h> 46 #include <net/protocol.h> 47 #include <net/arp.h> 48 #include <net/route.h> 49 #include <net/udp.h> 50 #include <net/tcp.h> 51 #include <net/sock.h> 52 #include <net/pkt_sched.h> 53 #include <net/fib_rules.h> 54 #include <net/rtnetlink.h> 55 #include <net/net_namespace.h> 56 57 #define RTNL_MAX_TYPE 50 58 #define RTNL_SLAVE_MAX_TYPE 36 59 60 struct rtnl_link { 61 rtnl_doit_func doit; 62 rtnl_dumpit_func dumpit; 63 struct module *owner; 64 unsigned int flags; 65 struct rcu_head rcu; 66 }; 67 68 static DEFINE_MUTEX(rtnl_mutex); 69 70 void rtnl_lock(void) 71 { 72 mutex_lock(&rtnl_mutex); 73 } 74 EXPORT_SYMBOL(rtnl_lock); 75 76 int rtnl_lock_killable(void) 77 { 78 return mutex_lock_killable(&rtnl_mutex); 79 } 80 EXPORT_SYMBOL(rtnl_lock_killable); 81 82 static struct sk_buff *defer_kfree_skb_list; 83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail) 84 { 85 if (head && tail) { 86 tail->next = defer_kfree_skb_list; 87 defer_kfree_skb_list = head; 88 } 89 } 90 EXPORT_SYMBOL(rtnl_kfree_skbs); 91 92 void __rtnl_unlock(void) 93 { 94 struct sk_buff *head = defer_kfree_skb_list; 95 96 defer_kfree_skb_list = NULL; 97 98 mutex_unlock(&rtnl_mutex); 99 100 while (head) { 101 struct sk_buff *next = head->next; 102 103 kfree_skb(head); 104 cond_resched(); 105 head = next; 106 } 107 } 108 109 void rtnl_unlock(void) 110 { 111 /* This fellow will unlock it for us. */ 112 netdev_run_todo(); 113 } 114 EXPORT_SYMBOL(rtnl_unlock); 115 116 int rtnl_trylock(void) 117 { 118 return mutex_trylock(&rtnl_mutex); 119 } 120 EXPORT_SYMBOL(rtnl_trylock); 121 122 int rtnl_is_locked(void) 123 { 124 return mutex_is_locked(&rtnl_mutex); 125 } 126 EXPORT_SYMBOL(rtnl_is_locked); 127 128 bool refcount_dec_and_rtnl_lock(refcount_t *r) 129 { 130 return refcount_dec_and_mutex_lock(r, &rtnl_mutex); 131 } 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock); 133 134 #ifdef CONFIG_PROVE_LOCKING 135 bool lockdep_rtnl_is_held(void) 136 { 137 return lockdep_is_held(&rtnl_mutex); 138 } 139 EXPORT_SYMBOL(lockdep_rtnl_is_held); 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 141 142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 143 144 static inline int rtm_msgindex(int msgtype) 145 { 146 int msgindex = msgtype - RTM_BASE; 147 148 /* 149 * msgindex < 0 implies someone tried to register a netlink 150 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 151 * the message type has not been added to linux/rtnetlink.h 152 */ 153 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 154 155 return msgindex; 156 } 157 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype) 159 { 160 struct rtnl_link **tab; 161 162 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers)) 163 protocol = PF_UNSPEC; 164 165 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]); 166 if (!tab) 167 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]); 168 169 return tab[msgtype]; 170 } 171 172 static int rtnl_register_internal(struct module *owner, 173 int protocol, int msgtype, 174 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 175 unsigned int flags) 176 { 177 struct rtnl_link *link, *old; 178 struct rtnl_link __rcu **tab; 179 int msgindex; 180 int ret = -ENOBUFS; 181 182 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 183 msgindex = rtm_msgindex(msgtype); 184 185 rtnl_lock(); 186 tab = rtnl_msg_handlers[protocol]; 187 if (tab == NULL) { 188 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL); 189 if (!tab) 190 goto unlock; 191 192 /* ensures we see the 0 stores */ 193 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab); 194 } 195 196 old = rtnl_dereference(tab[msgindex]); 197 if (old) { 198 link = kmemdup(old, sizeof(*old), GFP_KERNEL); 199 if (!link) 200 goto unlock; 201 } else { 202 link = kzalloc(sizeof(*link), GFP_KERNEL); 203 if (!link) 204 goto unlock; 205 } 206 207 WARN_ON(link->owner && link->owner != owner); 208 link->owner = owner; 209 210 WARN_ON(doit && link->doit && link->doit != doit); 211 if (doit) 212 link->doit = doit; 213 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit); 214 if (dumpit) 215 link->dumpit = dumpit; 216 217 link->flags |= flags; 218 219 /* publish protocol:msgtype */ 220 rcu_assign_pointer(tab[msgindex], link); 221 ret = 0; 222 if (old) 223 kfree_rcu(old, rcu); 224 unlock: 225 rtnl_unlock(); 226 return ret; 227 } 228 229 /** 230 * rtnl_register_module - Register a rtnetlink message type 231 * 232 * @owner: module registering the hook (THIS_MODULE) 233 * @protocol: Protocol family or PF_UNSPEC 234 * @msgtype: rtnetlink message type 235 * @doit: Function pointer called for each request message 236 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 237 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions 238 * 239 * Like rtnl_register, but for use by removable modules. 240 */ 241 int rtnl_register_module(struct module *owner, 242 int protocol, int msgtype, 243 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 244 unsigned int flags) 245 { 246 return rtnl_register_internal(owner, protocol, msgtype, 247 doit, dumpit, flags); 248 } 249 EXPORT_SYMBOL_GPL(rtnl_register_module); 250 251 /** 252 * rtnl_register - Register a rtnetlink message type 253 * @protocol: Protocol family or PF_UNSPEC 254 * @msgtype: rtnetlink message type 255 * @doit: Function pointer called for each request message 256 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 257 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions 258 * 259 * Registers the specified function pointers (at least one of them has 260 * to be non-NULL) to be called whenever a request message for the 261 * specified protocol family and message type is received. 262 * 263 * The special protocol family PF_UNSPEC may be used to define fallback 264 * function pointers for the case when no entry for the specific protocol 265 * family exists. 266 */ 267 void rtnl_register(int protocol, int msgtype, 268 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 269 unsigned int flags) 270 { 271 int err; 272 273 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit, 274 flags); 275 if (err) 276 pr_err("Unable to register rtnetlink message handler, " 277 "protocol = %d, message type = %d\n", protocol, msgtype); 278 } 279 280 /** 281 * rtnl_unregister - Unregister a rtnetlink message type 282 * @protocol: Protocol family or PF_UNSPEC 283 * @msgtype: rtnetlink message type 284 * 285 * Returns 0 on success or a negative error code. 286 */ 287 int rtnl_unregister(int protocol, int msgtype) 288 { 289 struct rtnl_link **tab, *link; 290 int msgindex; 291 292 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 293 msgindex = rtm_msgindex(msgtype); 294 295 rtnl_lock(); 296 tab = rtnl_dereference(rtnl_msg_handlers[protocol]); 297 if (!tab) { 298 rtnl_unlock(); 299 return -ENOENT; 300 } 301 302 link = tab[msgindex]; 303 rcu_assign_pointer(tab[msgindex], NULL); 304 rtnl_unlock(); 305 306 kfree_rcu(link, rcu); 307 308 return 0; 309 } 310 EXPORT_SYMBOL_GPL(rtnl_unregister); 311 312 /** 313 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 314 * @protocol : Protocol family or PF_UNSPEC 315 * 316 * Identical to calling rtnl_unregster() for all registered message types 317 * of a certain protocol family. 318 */ 319 void rtnl_unregister_all(int protocol) 320 { 321 struct rtnl_link **tab, *link; 322 int msgindex; 323 324 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 325 326 rtnl_lock(); 327 tab = rtnl_msg_handlers[protocol]; 328 if (!tab) { 329 rtnl_unlock(); 330 return; 331 } 332 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL); 333 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) { 334 link = tab[msgindex]; 335 if (!link) 336 continue; 337 338 rcu_assign_pointer(tab[msgindex], NULL); 339 kfree_rcu(link, rcu); 340 } 341 rtnl_unlock(); 342 343 synchronize_net(); 344 345 kfree(tab); 346 } 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 348 349 static LIST_HEAD(link_ops); 350 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 352 { 353 const struct rtnl_link_ops *ops; 354 355 list_for_each_entry(ops, &link_ops, list) { 356 if (!strcmp(ops->kind, kind)) 357 return ops; 358 } 359 return NULL; 360 } 361 362 /** 363 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 364 * @ops: struct rtnl_link_ops * to register 365 * 366 * The caller must hold the rtnl_mutex. This function should be used 367 * by drivers that create devices during module initialization. It 368 * must be called before registering the devices. 369 * 370 * Returns 0 on success or a negative error code. 371 */ 372 int __rtnl_link_register(struct rtnl_link_ops *ops) 373 { 374 if (rtnl_link_ops_get(ops->kind)) 375 return -EEXIST; 376 377 /* The check for setup is here because if ops 378 * does not have that filled up, it is not possible 379 * to use the ops for creating device. So do not 380 * fill up dellink as well. That disables rtnl_dellink. 381 */ 382 if (ops->setup && !ops->dellink) 383 ops->dellink = unregister_netdevice_queue; 384 385 list_add_tail(&ops->list, &link_ops); 386 return 0; 387 } 388 EXPORT_SYMBOL_GPL(__rtnl_link_register); 389 390 /** 391 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 392 * @ops: struct rtnl_link_ops * to register 393 * 394 * Returns 0 on success or a negative error code. 395 */ 396 int rtnl_link_register(struct rtnl_link_ops *ops) 397 { 398 int err; 399 400 /* Sanity-check max sizes to avoid stack buffer overflow. */ 401 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE || 402 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)) 403 return -EINVAL; 404 405 rtnl_lock(); 406 err = __rtnl_link_register(ops); 407 rtnl_unlock(); 408 return err; 409 } 410 EXPORT_SYMBOL_GPL(rtnl_link_register); 411 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 413 { 414 struct net_device *dev; 415 LIST_HEAD(list_kill); 416 417 for_each_netdev(net, dev) { 418 if (dev->rtnl_link_ops == ops) 419 ops->dellink(dev, &list_kill); 420 } 421 unregister_netdevice_many(&list_kill); 422 } 423 424 /** 425 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 426 * @ops: struct rtnl_link_ops * to unregister 427 * 428 * The caller must hold the rtnl_mutex and guarantee net_namespace_list 429 * integrity (hold pernet_ops_rwsem for writing to close the race 430 * with setup_net() and cleanup_net()). 431 */ 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 433 { 434 struct net *net; 435 436 for_each_net(net) { 437 __rtnl_kill_links(net, ops); 438 } 439 list_del(&ops->list); 440 } 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 442 443 /* Return with the rtnl_lock held when there are no network 444 * devices unregistering in any network namespace. 445 */ 446 static void rtnl_lock_unregistering_all(void) 447 { 448 struct net *net; 449 bool unregistering; 450 DEFINE_WAIT_FUNC(wait, woken_wake_function); 451 452 add_wait_queue(&netdev_unregistering_wq, &wait); 453 for (;;) { 454 unregistering = false; 455 rtnl_lock(); 456 /* We held write locked pernet_ops_rwsem, and parallel 457 * setup_net() and cleanup_net() are not possible. 458 */ 459 for_each_net(net) { 460 if (net->dev_unreg_count > 0) { 461 unregistering = true; 462 break; 463 } 464 } 465 if (!unregistering) 466 break; 467 __rtnl_unlock(); 468 469 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); 470 } 471 remove_wait_queue(&netdev_unregistering_wq, &wait); 472 } 473 474 /** 475 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 476 * @ops: struct rtnl_link_ops * to unregister 477 */ 478 void rtnl_link_unregister(struct rtnl_link_ops *ops) 479 { 480 /* Close the race with setup_net() and cleanup_net() */ 481 down_write(&pernet_ops_rwsem); 482 rtnl_lock_unregistering_all(); 483 __rtnl_link_unregister(ops); 484 rtnl_unlock(); 485 up_write(&pernet_ops_rwsem); 486 } 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 488 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev) 490 { 491 struct net_device *master_dev; 492 const struct rtnl_link_ops *ops; 493 size_t size = 0; 494 495 rcu_read_lock(); 496 497 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev); 498 if (!master_dev) 499 goto out; 500 501 ops = master_dev->rtnl_link_ops; 502 if (!ops || !ops->get_slave_size) 503 goto out; 504 /* IFLA_INFO_SLAVE_DATA + nested data */ 505 size = nla_total_size(sizeof(struct nlattr)) + 506 ops->get_slave_size(master_dev, dev); 507 508 out: 509 rcu_read_unlock(); 510 return size; 511 } 512 513 static size_t rtnl_link_get_size(const struct net_device *dev) 514 { 515 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 516 size_t size; 517 518 if (!ops) 519 return 0; 520 521 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 522 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 523 524 if (ops->get_size) 525 /* IFLA_INFO_DATA + nested data */ 526 size += nla_total_size(sizeof(struct nlattr)) + 527 ops->get_size(dev); 528 529 if (ops->get_xstats_size) 530 /* IFLA_INFO_XSTATS */ 531 size += nla_total_size(ops->get_xstats_size(dev)); 532 533 size += rtnl_link_get_slave_info_data_size(dev); 534 535 return size; 536 } 537 538 static LIST_HEAD(rtnl_af_ops); 539 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family) 541 { 542 const struct rtnl_af_ops *ops; 543 544 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) { 545 if (ops->family == family) 546 return ops; 547 } 548 549 return NULL; 550 } 551 552 /** 553 * rtnl_af_register - Register rtnl_af_ops with rtnetlink. 554 * @ops: struct rtnl_af_ops * to register 555 * 556 * Returns 0 on success or a negative error code. 557 */ 558 void rtnl_af_register(struct rtnl_af_ops *ops) 559 { 560 rtnl_lock(); 561 list_add_tail_rcu(&ops->list, &rtnl_af_ops); 562 rtnl_unlock(); 563 } 564 EXPORT_SYMBOL_GPL(rtnl_af_register); 565 566 /** 567 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 568 * @ops: struct rtnl_af_ops * to unregister 569 */ 570 void rtnl_af_unregister(struct rtnl_af_ops *ops) 571 { 572 rtnl_lock(); 573 list_del_rcu(&ops->list); 574 rtnl_unlock(); 575 576 synchronize_rcu(); 577 } 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister); 579 580 static size_t rtnl_link_get_af_size(const struct net_device *dev, 581 u32 ext_filter_mask) 582 { 583 struct rtnl_af_ops *af_ops; 584 size_t size; 585 586 /* IFLA_AF_SPEC */ 587 size = nla_total_size(sizeof(struct nlattr)); 588 589 rcu_read_lock(); 590 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 591 if (af_ops->get_link_af_size) { 592 /* AF_* + nested data */ 593 size += nla_total_size(sizeof(struct nlattr)) + 594 af_ops->get_link_af_size(dev, ext_filter_mask); 595 } 596 } 597 rcu_read_unlock(); 598 599 return size; 600 } 601 602 static bool rtnl_have_link_slave_info(const struct net_device *dev) 603 { 604 struct net_device *master_dev; 605 bool ret = false; 606 607 rcu_read_lock(); 608 609 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev); 610 if (master_dev && master_dev->rtnl_link_ops) 611 ret = true; 612 rcu_read_unlock(); 613 return ret; 614 } 615 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb, 617 const struct net_device *dev) 618 { 619 struct net_device *master_dev; 620 const struct rtnl_link_ops *ops; 621 struct nlattr *slave_data; 622 int err; 623 624 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 625 if (!master_dev) 626 return 0; 627 ops = master_dev->rtnl_link_ops; 628 if (!ops) 629 return 0; 630 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0) 631 return -EMSGSIZE; 632 if (ops->fill_slave_info) { 633 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA); 634 if (!slave_data) 635 return -EMSGSIZE; 636 err = ops->fill_slave_info(skb, master_dev, dev); 637 if (err < 0) 638 goto err_cancel_slave_data; 639 nla_nest_end(skb, slave_data); 640 } 641 return 0; 642 643 err_cancel_slave_data: 644 nla_nest_cancel(skb, slave_data); 645 return err; 646 } 647 648 static int rtnl_link_info_fill(struct sk_buff *skb, 649 const struct net_device *dev) 650 { 651 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 652 struct nlattr *data; 653 int err; 654 655 if (!ops) 656 return 0; 657 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 658 return -EMSGSIZE; 659 if (ops->fill_xstats) { 660 err = ops->fill_xstats(skb, dev); 661 if (err < 0) 662 return err; 663 } 664 if (ops->fill_info) { 665 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA); 666 if (data == NULL) 667 return -EMSGSIZE; 668 err = ops->fill_info(skb, dev); 669 if (err < 0) 670 goto err_cancel_data; 671 nla_nest_end(skb, data); 672 } 673 return 0; 674 675 err_cancel_data: 676 nla_nest_cancel(skb, data); 677 return err; 678 } 679 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 681 { 682 struct nlattr *linkinfo; 683 int err = -EMSGSIZE; 684 685 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO); 686 if (linkinfo == NULL) 687 goto out; 688 689 err = rtnl_link_info_fill(skb, dev); 690 if (err < 0) 691 goto err_cancel_link; 692 693 err = rtnl_link_slave_info_fill(skb, dev); 694 if (err < 0) 695 goto err_cancel_link; 696 697 nla_nest_end(skb, linkinfo); 698 return 0; 699 700 err_cancel_link: 701 nla_nest_cancel(skb, linkinfo); 702 out: 703 return err; 704 } 705 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo) 707 { 708 struct sock *rtnl = net->rtnl; 709 int err = 0; 710 711 NETLINK_CB(skb).dst_group = group; 712 if (echo) 713 refcount_inc(&skb->users); 714 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 715 if (echo) 716 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 717 return err; 718 } 719 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 721 { 722 struct sock *rtnl = net->rtnl; 723 724 return nlmsg_unicast(rtnl, skb, pid); 725 } 726 EXPORT_SYMBOL(rtnl_unicast); 727 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 729 struct nlmsghdr *nlh, gfp_t flags) 730 { 731 struct sock *rtnl = net->rtnl; 732 int report = 0; 733 734 if (nlh) 735 report = nlmsg_report(nlh); 736 737 nlmsg_notify(rtnl, skb, pid, group, report, flags); 738 } 739 EXPORT_SYMBOL(rtnl_notify); 740 741 void rtnl_set_sk_err(struct net *net, u32 group, int error) 742 { 743 struct sock *rtnl = net->rtnl; 744 745 netlink_set_err(rtnl, 0, group, error); 746 } 747 EXPORT_SYMBOL(rtnl_set_sk_err); 748 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 750 { 751 struct nlattr *mx; 752 int i, valid = 0; 753 754 /* nothing is dumped for dst_default_metrics, so just skip the loop */ 755 if (metrics == dst_default_metrics.metrics) 756 return 0; 757 758 mx = nla_nest_start_noflag(skb, RTA_METRICS); 759 if (mx == NULL) 760 return -ENOBUFS; 761 762 for (i = 0; i < RTAX_MAX; i++) { 763 if (metrics[i]) { 764 if (i == RTAX_CC_ALGO - 1) { 765 char tmp[TCP_CA_NAME_MAX], *name; 766 767 name = tcp_ca_get_name_by_key(metrics[i], tmp); 768 if (!name) 769 continue; 770 if (nla_put_string(skb, i + 1, name)) 771 goto nla_put_failure; 772 } else if (i == RTAX_FEATURES - 1) { 773 u32 user_features = metrics[i] & RTAX_FEATURE_MASK; 774 775 if (!user_features) 776 continue; 777 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK); 778 if (nla_put_u32(skb, i + 1, user_features)) 779 goto nla_put_failure; 780 } else { 781 if (nla_put_u32(skb, i + 1, metrics[i])) 782 goto nla_put_failure; 783 } 784 valid++; 785 } 786 } 787 788 if (!valid) { 789 nla_nest_cancel(skb, mx); 790 return 0; 791 } 792 793 return nla_nest_end(skb, mx); 794 795 nla_put_failure: 796 nla_nest_cancel(skb, mx); 797 return -EMSGSIZE; 798 } 799 EXPORT_SYMBOL(rtnetlink_put_metrics); 800 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 802 long expires, u32 error) 803 { 804 struct rta_cacheinfo ci = { 805 .rta_error = error, 806 .rta_id = id, 807 }; 808 809 if (dst) { 810 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse); 811 ci.rta_used = dst->__use; 812 ci.rta_clntref = atomic_read(&dst->__refcnt); 813 } 814 if (expires) { 815 unsigned long clock; 816 817 clock = jiffies_to_clock_t(abs(expires)); 818 clock = min_t(unsigned long, clock, INT_MAX); 819 ci.rta_expires = (expires > 0) ? clock : -clock; 820 } 821 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 822 } 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 824 825 static void set_operstate(struct net_device *dev, unsigned char transition) 826 { 827 unsigned char operstate = dev->operstate; 828 829 switch (transition) { 830 case IF_OPER_UP: 831 if ((operstate == IF_OPER_DORMANT || 832 operstate == IF_OPER_UNKNOWN) && 833 !netif_dormant(dev)) 834 operstate = IF_OPER_UP; 835 break; 836 837 case IF_OPER_DORMANT: 838 if (operstate == IF_OPER_UP || 839 operstate == IF_OPER_UNKNOWN) 840 operstate = IF_OPER_DORMANT; 841 break; 842 } 843 844 if (dev->operstate != operstate) { 845 write_lock_bh(&dev_base_lock); 846 dev->operstate = operstate; 847 write_unlock_bh(&dev_base_lock); 848 netdev_state_change(dev); 849 } 850 } 851 852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev) 853 { 854 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) | 855 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI)); 856 } 857 858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 859 const struct ifinfomsg *ifm) 860 { 861 unsigned int flags = ifm->ifi_flags; 862 863 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 864 if (ifm->ifi_change) 865 flags = (flags & ifm->ifi_change) | 866 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change); 867 868 return flags; 869 } 870 871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 872 const struct rtnl_link_stats64 *b) 873 { 874 a->rx_packets = b->rx_packets; 875 a->tx_packets = b->tx_packets; 876 a->rx_bytes = b->rx_bytes; 877 a->tx_bytes = b->tx_bytes; 878 a->rx_errors = b->rx_errors; 879 a->tx_errors = b->tx_errors; 880 a->rx_dropped = b->rx_dropped; 881 a->tx_dropped = b->tx_dropped; 882 883 a->multicast = b->multicast; 884 a->collisions = b->collisions; 885 886 a->rx_length_errors = b->rx_length_errors; 887 a->rx_over_errors = b->rx_over_errors; 888 a->rx_crc_errors = b->rx_crc_errors; 889 a->rx_frame_errors = b->rx_frame_errors; 890 a->rx_fifo_errors = b->rx_fifo_errors; 891 a->rx_missed_errors = b->rx_missed_errors; 892 893 a->tx_aborted_errors = b->tx_aborted_errors; 894 a->tx_carrier_errors = b->tx_carrier_errors; 895 a->tx_fifo_errors = b->tx_fifo_errors; 896 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 897 a->tx_window_errors = b->tx_window_errors; 898 899 a->rx_compressed = b->rx_compressed; 900 a->tx_compressed = b->tx_compressed; 901 902 a->rx_nohandler = b->rx_nohandler; 903 } 904 905 /* All VF info */ 906 static inline int rtnl_vfinfo_size(const struct net_device *dev, 907 u32 ext_filter_mask) 908 { 909 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) { 910 int num_vfs = dev_num_vf(dev->dev.parent); 911 size_t size = nla_total_size(0); 912 size += num_vfs * 913 (nla_total_size(0) + 914 nla_total_size(sizeof(struct ifla_vf_mac)) + 915 nla_total_size(sizeof(struct ifla_vf_broadcast)) + 916 nla_total_size(sizeof(struct ifla_vf_vlan)) + 917 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */ 918 nla_total_size(MAX_VLAN_LIST_LEN * 919 sizeof(struct ifla_vf_vlan_info)) + 920 nla_total_size(sizeof(struct ifla_vf_spoofchk)) + 921 nla_total_size(sizeof(struct ifla_vf_tx_rate)) + 922 nla_total_size(sizeof(struct ifla_vf_rate)) + 923 nla_total_size(sizeof(struct ifla_vf_link_state)) + 924 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) + 925 nla_total_size(0) + /* nest IFLA_VF_STATS */ 926 /* IFLA_VF_STATS_RX_PACKETS */ 927 nla_total_size_64bit(sizeof(__u64)) + 928 /* IFLA_VF_STATS_TX_PACKETS */ 929 nla_total_size_64bit(sizeof(__u64)) + 930 /* IFLA_VF_STATS_RX_BYTES */ 931 nla_total_size_64bit(sizeof(__u64)) + 932 /* IFLA_VF_STATS_TX_BYTES */ 933 nla_total_size_64bit(sizeof(__u64)) + 934 /* IFLA_VF_STATS_BROADCAST */ 935 nla_total_size_64bit(sizeof(__u64)) + 936 /* IFLA_VF_STATS_MULTICAST */ 937 nla_total_size_64bit(sizeof(__u64)) + 938 /* IFLA_VF_STATS_RX_DROPPED */ 939 nla_total_size_64bit(sizeof(__u64)) + 940 /* IFLA_VF_STATS_TX_DROPPED */ 941 nla_total_size_64bit(sizeof(__u64)) + 942 nla_total_size(sizeof(struct ifla_vf_trust))); 943 return size; 944 } else 945 return 0; 946 } 947 948 static size_t rtnl_port_size(const struct net_device *dev, 949 u32 ext_filter_mask) 950 { 951 size_t port_size = nla_total_size(4) /* PORT_VF */ 952 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 953 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 954 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 955 + nla_total_size(1) /* PROT_VDP_REQUEST */ 956 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 957 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 958 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 959 + port_size; 960 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 961 + port_size; 962 963 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 964 !(ext_filter_mask & RTEXT_FILTER_VF)) 965 return 0; 966 if (dev_num_vf(dev->dev.parent)) 967 return port_self_size + vf_ports_size + 968 vf_port_size * dev_num_vf(dev->dev.parent); 969 else 970 return port_self_size; 971 } 972 973 static size_t rtnl_xdp_size(void) 974 { 975 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */ 976 nla_total_size(1) + /* XDP_ATTACHED */ 977 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */ 978 nla_total_size(4); /* XDP_<mode>_PROG_ID */ 979 980 return xdp_size; 981 } 982 983 static noinline size_t if_nlmsg_size(const struct net_device *dev, 984 u32 ext_filter_mask) 985 { 986 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 987 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 988 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 989 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 990 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap)) 991 + nla_total_size(sizeof(struct rtnl_link_stats)) 992 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64)) 993 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 994 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 995 + nla_total_size(4) /* IFLA_TXQLEN */ 996 + nla_total_size(4) /* IFLA_WEIGHT */ 997 + nla_total_size(4) /* IFLA_MTU */ 998 + nla_total_size(4) /* IFLA_LINK */ 999 + nla_total_size(4) /* IFLA_MASTER */ 1000 + nla_total_size(1) /* IFLA_CARRIER */ 1001 + nla_total_size(4) /* IFLA_PROMISCUITY */ 1002 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 1003 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 1004 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */ 1005 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */ 1006 + nla_total_size(1) /* IFLA_OPERSTATE */ 1007 + nla_total_size(1) /* IFLA_LINKMODE */ 1008 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 1009 + nla_total_size(4) /* IFLA_LINK_NETNSID */ 1010 + nla_total_size(4) /* IFLA_GROUP */ 1011 + nla_total_size(ext_filter_mask 1012 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */ 1013 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */ 1014 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 1015 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */ 1016 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */ 1017 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */ 1018 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */ 1019 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */ 1020 + rtnl_xdp_size() /* IFLA_XDP */ 1021 + nla_total_size(4) /* IFLA_EVENT */ 1022 + nla_total_size(4) /* IFLA_NEW_NETNSID */ 1023 + nla_total_size(4) /* IFLA_NEW_IFINDEX */ 1024 + nla_total_size(1) /* IFLA_PROTO_DOWN */ 1025 + nla_total_size(4) /* IFLA_TARGET_NETNSID */ 1026 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */ 1027 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */ 1028 + nla_total_size(4) /* IFLA_MIN_MTU */ 1029 + nla_total_size(4) /* IFLA_MAX_MTU */ 1030 + 0; 1031 } 1032 1033 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 1034 { 1035 struct nlattr *vf_ports; 1036 struct nlattr *vf_port; 1037 int vf; 1038 int err; 1039 1040 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS); 1041 if (!vf_ports) 1042 return -EMSGSIZE; 1043 1044 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 1045 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT); 1046 if (!vf_port) 1047 goto nla_put_failure; 1048 if (nla_put_u32(skb, IFLA_PORT_VF, vf)) 1049 goto nla_put_failure; 1050 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 1051 if (err == -EMSGSIZE) 1052 goto nla_put_failure; 1053 if (err) { 1054 nla_nest_cancel(skb, vf_port); 1055 continue; 1056 } 1057 nla_nest_end(skb, vf_port); 1058 } 1059 1060 nla_nest_end(skb, vf_ports); 1061 1062 return 0; 1063 1064 nla_put_failure: 1065 nla_nest_cancel(skb, vf_ports); 1066 return -EMSGSIZE; 1067 } 1068 1069 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 1070 { 1071 struct nlattr *port_self; 1072 int err; 1073 1074 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF); 1075 if (!port_self) 1076 return -EMSGSIZE; 1077 1078 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 1079 if (err) { 1080 nla_nest_cancel(skb, port_self); 1081 return (err == -EMSGSIZE) ? err : 0; 1082 } 1083 1084 nla_nest_end(skb, port_self); 1085 1086 return 0; 1087 } 1088 1089 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev, 1090 u32 ext_filter_mask) 1091 { 1092 int err; 1093 1094 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 1095 !(ext_filter_mask & RTEXT_FILTER_VF)) 1096 return 0; 1097 1098 err = rtnl_port_self_fill(skb, dev); 1099 if (err) 1100 return err; 1101 1102 if (dev_num_vf(dev->dev.parent)) { 1103 err = rtnl_vf_ports_fill(skb, dev); 1104 if (err) 1105 return err; 1106 } 1107 1108 return 0; 1109 } 1110 1111 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev) 1112 { 1113 int err; 1114 struct netdev_phys_item_id ppid; 1115 1116 err = dev_get_phys_port_id(dev, &ppid); 1117 if (err) { 1118 if (err == -EOPNOTSUPP) 1119 return 0; 1120 return err; 1121 } 1122 1123 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id)) 1124 return -EMSGSIZE; 1125 1126 return 0; 1127 } 1128 1129 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev) 1130 { 1131 char name[IFNAMSIZ]; 1132 int err; 1133 1134 err = dev_get_phys_port_name(dev, name, sizeof(name)); 1135 if (err) { 1136 if (err == -EOPNOTSUPP) 1137 return 0; 1138 return err; 1139 } 1140 1141 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name)) 1142 return -EMSGSIZE; 1143 1144 return 0; 1145 } 1146 1147 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev) 1148 { 1149 struct netdev_phys_item_id ppid = { }; 1150 int err; 1151 1152 err = dev_get_port_parent_id(dev, &ppid, false); 1153 if (err) { 1154 if (err == -EOPNOTSUPP) 1155 return 0; 1156 return err; 1157 } 1158 1159 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id)) 1160 return -EMSGSIZE; 1161 1162 return 0; 1163 } 1164 1165 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb, 1166 struct net_device *dev) 1167 { 1168 struct rtnl_link_stats64 *sp; 1169 struct nlattr *attr; 1170 1171 attr = nla_reserve_64bit(skb, IFLA_STATS64, 1172 sizeof(struct rtnl_link_stats64), IFLA_PAD); 1173 if (!attr) 1174 return -EMSGSIZE; 1175 1176 sp = nla_data(attr); 1177 dev_get_stats(dev, sp); 1178 1179 attr = nla_reserve(skb, IFLA_STATS, 1180 sizeof(struct rtnl_link_stats)); 1181 if (!attr) 1182 return -EMSGSIZE; 1183 1184 copy_rtnl_link_stats(nla_data(attr), sp); 1185 1186 return 0; 1187 } 1188 1189 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb, 1190 struct net_device *dev, 1191 int vfs_num, 1192 struct nlattr *vfinfo) 1193 { 1194 struct ifla_vf_rss_query_en vf_rss_query_en; 1195 struct nlattr *vf, *vfstats, *vfvlanlist; 1196 struct ifla_vf_link_state vf_linkstate; 1197 struct ifla_vf_vlan_info vf_vlan_info; 1198 struct ifla_vf_spoofchk vf_spoofchk; 1199 struct ifla_vf_tx_rate vf_tx_rate; 1200 struct ifla_vf_stats vf_stats; 1201 struct ifla_vf_trust vf_trust; 1202 struct ifla_vf_vlan vf_vlan; 1203 struct ifla_vf_rate vf_rate; 1204 struct ifla_vf_mac vf_mac; 1205 struct ifla_vf_broadcast vf_broadcast; 1206 struct ifla_vf_info ivi; 1207 1208 memset(&ivi, 0, sizeof(ivi)); 1209 1210 /* Not all SR-IOV capable drivers support the 1211 * spoofcheck and "RSS query enable" query. Preset to 1212 * -1 so the user space tool can detect that the driver 1213 * didn't report anything. 1214 */ 1215 ivi.spoofchk = -1; 1216 ivi.rss_query_en = -1; 1217 ivi.trusted = -1; 1218 /* The default value for VF link state is "auto" 1219 * IFLA_VF_LINK_STATE_AUTO which equals zero 1220 */ 1221 ivi.linkstate = 0; 1222 /* VLAN Protocol by default is 802.1Q */ 1223 ivi.vlan_proto = htons(ETH_P_8021Q); 1224 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi)) 1225 return 0; 1226 1227 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info)); 1228 1229 vf_mac.vf = 1230 vf_vlan.vf = 1231 vf_vlan_info.vf = 1232 vf_rate.vf = 1233 vf_tx_rate.vf = 1234 vf_spoofchk.vf = 1235 vf_linkstate.vf = 1236 vf_rss_query_en.vf = 1237 vf_trust.vf = ivi.vf; 1238 1239 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 1240 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len); 1241 vf_vlan.vlan = ivi.vlan; 1242 vf_vlan.qos = ivi.qos; 1243 vf_vlan_info.vlan = ivi.vlan; 1244 vf_vlan_info.qos = ivi.qos; 1245 vf_vlan_info.vlan_proto = ivi.vlan_proto; 1246 vf_tx_rate.rate = ivi.max_tx_rate; 1247 vf_rate.min_tx_rate = ivi.min_tx_rate; 1248 vf_rate.max_tx_rate = ivi.max_tx_rate; 1249 vf_spoofchk.setting = ivi.spoofchk; 1250 vf_linkstate.link_state = ivi.linkstate; 1251 vf_rss_query_en.setting = ivi.rss_query_en; 1252 vf_trust.setting = ivi.trusted; 1253 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO); 1254 if (!vf) 1255 goto nla_put_vfinfo_failure; 1256 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) || 1257 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) || 1258 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) || 1259 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate), 1260 &vf_rate) || 1261 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), 1262 &vf_tx_rate) || 1263 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk), 1264 &vf_spoofchk) || 1265 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate), 1266 &vf_linkstate) || 1267 nla_put(skb, IFLA_VF_RSS_QUERY_EN, 1268 sizeof(vf_rss_query_en), 1269 &vf_rss_query_en) || 1270 nla_put(skb, IFLA_VF_TRUST, 1271 sizeof(vf_trust), &vf_trust)) 1272 goto nla_put_vf_failure; 1273 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST); 1274 if (!vfvlanlist) 1275 goto nla_put_vf_failure; 1276 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info), 1277 &vf_vlan_info)) { 1278 nla_nest_cancel(skb, vfvlanlist); 1279 goto nla_put_vf_failure; 1280 } 1281 nla_nest_end(skb, vfvlanlist); 1282 memset(&vf_stats, 0, sizeof(vf_stats)); 1283 if (dev->netdev_ops->ndo_get_vf_stats) 1284 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num, 1285 &vf_stats); 1286 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS); 1287 if (!vfstats) 1288 goto nla_put_vf_failure; 1289 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS, 1290 vf_stats.rx_packets, IFLA_VF_STATS_PAD) || 1291 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS, 1292 vf_stats.tx_packets, IFLA_VF_STATS_PAD) || 1293 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES, 1294 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) || 1295 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES, 1296 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) || 1297 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST, 1298 vf_stats.broadcast, IFLA_VF_STATS_PAD) || 1299 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST, 1300 vf_stats.multicast, IFLA_VF_STATS_PAD) || 1301 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED, 1302 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) || 1303 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED, 1304 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) { 1305 nla_nest_cancel(skb, vfstats); 1306 goto nla_put_vf_failure; 1307 } 1308 nla_nest_end(skb, vfstats); 1309 nla_nest_end(skb, vf); 1310 return 0; 1311 1312 nla_put_vf_failure: 1313 nla_nest_cancel(skb, vf); 1314 nla_put_vfinfo_failure: 1315 nla_nest_cancel(skb, vfinfo); 1316 return -EMSGSIZE; 1317 } 1318 1319 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb, 1320 struct net_device *dev, 1321 u32 ext_filter_mask) 1322 { 1323 struct nlattr *vfinfo; 1324 int i, num_vfs; 1325 1326 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0)) 1327 return 0; 1328 1329 num_vfs = dev_num_vf(dev->dev.parent); 1330 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs)) 1331 return -EMSGSIZE; 1332 1333 if (!dev->netdev_ops->ndo_get_vf_config) 1334 return 0; 1335 1336 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST); 1337 if (!vfinfo) 1338 return -EMSGSIZE; 1339 1340 for (i = 0; i < num_vfs; i++) { 1341 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1342 return -EMSGSIZE; 1343 } 1344 1345 nla_nest_end(skb, vfinfo); 1346 return 0; 1347 } 1348 1349 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev) 1350 { 1351 struct rtnl_link_ifmap map; 1352 1353 memset(&map, 0, sizeof(map)); 1354 map.mem_start = dev->mem_start; 1355 map.mem_end = dev->mem_end; 1356 map.base_addr = dev->base_addr; 1357 map.irq = dev->irq; 1358 map.dma = dev->dma; 1359 map.port = dev->if_port; 1360 1361 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD)) 1362 return -EMSGSIZE; 1363 1364 return 0; 1365 } 1366 1367 static u32 rtnl_xdp_prog_skb(struct net_device *dev) 1368 { 1369 const struct bpf_prog *generic_xdp_prog; 1370 1371 ASSERT_RTNL(); 1372 1373 generic_xdp_prog = rtnl_dereference(dev->xdp_prog); 1374 if (!generic_xdp_prog) 1375 return 0; 1376 return generic_xdp_prog->aux->id; 1377 } 1378 1379 static u32 rtnl_xdp_prog_drv(struct net_device *dev) 1380 { 1381 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG); 1382 } 1383 1384 static u32 rtnl_xdp_prog_hw(struct net_device *dev) 1385 { 1386 return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, 1387 XDP_QUERY_PROG_HW); 1388 } 1389 1390 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev, 1391 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr, 1392 u32 (*get_prog_id)(struct net_device *dev)) 1393 { 1394 u32 curr_id; 1395 int err; 1396 1397 curr_id = get_prog_id(dev); 1398 if (!curr_id) 1399 return 0; 1400 1401 *prog_id = curr_id; 1402 err = nla_put_u32(skb, attr, curr_id); 1403 if (err) 1404 return err; 1405 1406 if (*mode != XDP_ATTACHED_NONE) 1407 *mode = XDP_ATTACHED_MULTI; 1408 else 1409 *mode = tgt_mode; 1410 1411 return 0; 1412 } 1413 1414 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1415 { 1416 struct nlattr *xdp; 1417 u32 prog_id; 1418 int err; 1419 u8 mode; 1420 1421 xdp = nla_nest_start_noflag(skb, IFLA_XDP); 1422 if (!xdp) 1423 return -EMSGSIZE; 1424 1425 prog_id = 0; 1426 mode = XDP_ATTACHED_NONE; 1427 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB, 1428 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb); 1429 if (err) 1430 goto err_cancel; 1431 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV, 1432 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv); 1433 if (err) 1434 goto err_cancel; 1435 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW, 1436 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw); 1437 if (err) 1438 goto err_cancel; 1439 1440 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode); 1441 if (err) 1442 goto err_cancel; 1443 1444 if (prog_id && mode != XDP_ATTACHED_MULTI) { 1445 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id); 1446 if (err) 1447 goto err_cancel; 1448 } 1449 1450 nla_nest_end(skb, xdp); 1451 return 0; 1452 1453 err_cancel: 1454 nla_nest_cancel(skb, xdp); 1455 return err; 1456 } 1457 1458 static u32 rtnl_get_event(unsigned long event) 1459 { 1460 u32 rtnl_event_type = IFLA_EVENT_NONE; 1461 1462 switch (event) { 1463 case NETDEV_REBOOT: 1464 rtnl_event_type = IFLA_EVENT_REBOOT; 1465 break; 1466 case NETDEV_FEAT_CHANGE: 1467 rtnl_event_type = IFLA_EVENT_FEATURES; 1468 break; 1469 case NETDEV_BONDING_FAILOVER: 1470 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER; 1471 break; 1472 case NETDEV_NOTIFY_PEERS: 1473 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS; 1474 break; 1475 case NETDEV_RESEND_IGMP: 1476 rtnl_event_type = IFLA_EVENT_IGMP_RESEND; 1477 break; 1478 case NETDEV_CHANGEINFODATA: 1479 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS; 1480 break; 1481 default: 1482 break; 1483 } 1484 1485 return rtnl_event_type; 1486 } 1487 1488 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev) 1489 { 1490 const struct net_device *upper_dev; 1491 int ret = 0; 1492 1493 rcu_read_lock(); 1494 1495 upper_dev = netdev_master_upper_dev_get_rcu(dev); 1496 if (upper_dev) 1497 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex); 1498 1499 rcu_read_unlock(); 1500 return ret; 1501 } 1502 1503 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev, 1504 bool force) 1505 { 1506 int ifindex = dev_get_iflink(dev); 1507 1508 if (force || dev->ifindex != ifindex) 1509 return nla_put_u32(skb, IFLA_LINK, ifindex); 1510 1511 return 0; 1512 } 1513 1514 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb, 1515 struct net_device *dev) 1516 { 1517 char buf[IFALIASZ]; 1518 int ret; 1519 1520 ret = dev_get_alias(dev, buf, sizeof(buf)); 1521 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0; 1522 } 1523 1524 static int rtnl_fill_link_netnsid(struct sk_buff *skb, 1525 const struct net_device *dev, 1526 struct net *src_net, gfp_t gfp) 1527 { 1528 bool put_iflink = false; 1529 1530 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) { 1531 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1532 1533 if (!net_eq(dev_net(dev), link_net)) { 1534 int id = peernet2id_alloc(src_net, link_net, gfp); 1535 1536 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1537 return -EMSGSIZE; 1538 1539 put_iflink = true; 1540 } 1541 } 1542 1543 return nla_put_iflink(skb, dev, put_iflink); 1544 } 1545 1546 static int rtnl_fill_link_af(struct sk_buff *skb, 1547 const struct net_device *dev, 1548 u32 ext_filter_mask) 1549 { 1550 const struct rtnl_af_ops *af_ops; 1551 struct nlattr *af_spec; 1552 1553 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC); 1554 if (!af_spec) 1555 return -EMSGSIZE; 1556 1557 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 1558 struct nlattr *af; 1559 int err; 1560 1561 if (!af_ops->fill_link_af) 1562 continue; 1563 1564 af = nla_nest_start_noflag(skb, af_ops->family); 1565 if (!af) 1566 return -EMSGSIZE; 1567 1568 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1569 /* 1570 * Caller may return ENODATA to indicate that there 1571 * was no data to be dumped. This is not an error, it 1572 * means we should trim the attribute header and 1573 * continue. 1574 */ 1575 if (err == -ENODATA) 1576 nla_nest_cancel(skb, af); 1577 else if (err < 0) 1578 return -EMSGSIZE; 1579 1580 nla_nest_end(skb, af); 1581 } 1582 1583 nla_nest_end(skb, af_spec); 1584 return 0; 1585 } 1586 1587 static int rtnl_fill_ifinfo(struct sk_buff *skb, 1588 struct net_device *dev, struct net *src_net, 1589 int type, u32 pid, u32 seq, u32 change, 1590 unsigned int flags, u32 ext_filter_mask, 1591 u32 event, int *new_nsid, int new_ifindex, 1592 int tgt_netnsid, gfp_t gfp) 1593 { 1594 struct ifinfomsg *ifm; 1595 struct nlmsghdr *nlh; 1596 1597 ASSERT_RTNL(); 1598 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1599 if (nlh == NULL) 1600 return -EMSGSIZE; 1601 1602 ifm = nlmsg_data(nlh); 1603 ifm->ifi_family = AF_UNSPEC; 1604 ifm->__ifi_pad = 0; 1605 ifm->ifi_type = dev->type; 1606 ifm->ifi_index = dev->ifindex; 1607 ifm->ifi_flags = dev_get_flags(dev); 1608 ifm->ifi_change = change; 1609 1610 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid)) 1611 goto nla_put_failure; 1612 1613 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1614 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1615 nla_put_u8(skb, IFLA_OPERSTATE, 1616 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1617 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1618 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1619 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) || 1620 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) || 1621 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1622 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1623 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1624 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1625 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1626 #ifdef CONFIG_RPS 1627 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1628 #endif 1629 put_master_ifindex(skb, dev) || 1630 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1631 (dev->qdisc && 1632 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1633 nla_put_ifalias(skb, dev) || 1634 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1635 atomic_read(&dev->carrier_up_count) + 1636 atomic_read(&dev->carrier_down_count)) || 1637 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) || 1638 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT, 1639 atomic_read(&dev->carrier_up_count)) || 1640 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT, 1641 atomic_read(&dev->carrier_down_count))) 1642 goto nla_put_failure; 1643 1644 if (event != IFLA_EVENT_NONE) { 1645 if (nla_put_u32(skb, IFLA_EVENT, event)) 1646 goto nla_put_failure; 1647 } 1648 1649 if (rtnl_fill_link_ifmap(skb, dev)) 1650 goto nla_put_failure; 1651 1652 if (dev->addr_len) { 1653 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1654 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1655 goto nla_put_failure; 1656 } 1657 1658 if (rtnl_phys_port_id_fill(skb, dev)) 1659 goto nla_put_failure; 1660 1661 if (rtnl_phys_port_name_fill(skb, dev)) 1662 goto nla_put_failure; 1663 1664 if (rtnl_phys_switch_id_fill(skb, dev)) 1665 goto nla_put_failure; 1666 1667 if (rtnl_fill_stats(skb, dev)) 1668 goto nla_put_failure; 1669 1670 if (rtnl_fill_vf(skb, dev, ext_filter_mask)) 1671 goto nla_put_failure; 1672 1673 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1674 goto nla_put_failure; 1675 1676 if (rtnl_xdp_fill(skb, dev)) 1677 goto nla_put_failure; 1678 1679 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1680 if (rtnl_link_fill(skb, dev) < 0) 1681 goto nla_put_failure; 1682 } 1683 1684 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp)) 1685 goto nla_put_failure; 1686 1687 if (new_nsid && 1688 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0) 1689 goto nla_put_failure; 1690 if (new_ifindex && 1691 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0) 1692 goto nla_put_failure; 1693 1694 1695 rcu_read_lock(); 1696 if (rtnl_fill_link_af(skb, dev, ext_filter_mask)) 1697 goto nla_put_failure_rcu; 1698 rcu_read_unlock(); 1699 1700 nlmsg_end(skb, nlh); 1701 return 0; 1702 1703 nla_put_failure_rcu: 1704 rcu_read_unlock(); 1705 nla_put_failure: 1706 nlmsg_cancel(skb, nlh); 1707 return -EMSGSIZE; 1708 } 1709 1710 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1711 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1712 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1713 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1714 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1715 [IFLA_MTU] = { .type = NLA_U32 }, 1716 [IFLA_LINK] = { .type = NLA_U32 }, 1717 [IFLA_MASTER] = { .type = NLA_U32 }, 1718 [IFLA_CARRIER] = { .type = NLA_U8 }, 1719 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1720 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1721 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1722 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1723 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1724 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1725 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1726 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to 1727 * allow 0-length string (needed to remove an alias). 1728 */ 1729 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 }, 1730 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1731 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1732 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1733 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1734 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1735 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1736 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1737 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1738 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 }, 1739 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 }, 1740 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1741 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1742 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1743 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1744 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1745 [IFLA_XDP] = { .type = NLA_NESTED }, 1746 [IFLA_EVENT] = { .type = NLA_U32 }, 1747 [IFLA_GROUP] = { .type = NLA_U32 }, 1748 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 }, 1749 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 }, 1750 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 }, 1751 [IFLA_MIN_MTU] = { .type = NLA_U32 }, 1752 [IFLA_MAX_MTU] = { .type = NLA_U32 }, 1753 }; 1754 1755 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1756 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1757 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1758 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1759 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1760 }; 1761 1762 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1763 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1764 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT }, 1765 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1766 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1767 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1768 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1769 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1770 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1771 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1772 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1773 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1774 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1775 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1776 }; 1777 1778 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1779 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1780 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1781 .len = PORT_PROFILE_MAX }, 1782 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1783 .len = PORT_UUID_MAX }, 1784 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1785 .len = PORT_UUID_MAX }, 1786 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1787 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1788 1789 /* Unused, but we need to keep it here since user space could 1790 * fill it. It's also broken with regard to NLA_BINARY use in 1791 * combination with structs. 1792 */ 1793 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1794 .len = sizeof(struct ifla_port_vsi) }, 1795 }; 1796 1797 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1798 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1799 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1800 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1801 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 }, 1802 }; 1803 1804 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1805 { 1806 const struct rtnl_link_ops *ops = NULL; 1807 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1808 1809 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0) 1810 return NULL; 1811 1812 if (linfo[IFLA_INFO_KIND]) { 1813 char kind[MODULE_NAME_LEN]; 1814 1815 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1816 ops = rtnl_link_ops_get(kind); 1817 } 1818 1819 return ops; 1820 } 1821 1822 static bool link_master_filtered(struct net_device *dev, int master_idx) 1823 { 1824 struct net_device *master; 1825 1826 if (!master_idx) 1827 return false; 1828 1829 master = netdev_master_upper_dev_get(dev); 1830 if (!master || master->ifindex != master_idx) 1831 return true; 1832 1833 return false; 1834 } 1835 1836 static bool link_kind_filtered(const struct net_device *dev, 1837 const struct rtnl_link_ops *kind_ops) 1838 { 1839 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1840 return true; 1841 1842 return false; 1843 } 1844 1845 static bool link_dump_filtered(struct net_device *dev, 1846 int master_idx, 1847 const struct rtnl_link_ops *kind_ops) 1848 { 1849 if (link_master_filtered(dev, master_idx) || 1850 link_kind_filtered(dev, kind_ops)) 1851 return true; 1852 1853 return false; 1854 } 1855 1856 /** 1857 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged. 1858 * @sk: netlink socket 1859 * @netnsid: network namespace identifier 1860 * 1861 * Returns the network namespace identified by netnsid on success or an error 1862 * pointer on failure. 1863 */ 1864 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid) 1865 { 1866 struct net *net; 1867 1868 net = get_net_ns_by_id(sock_net(sk), netnsid); 1869 if (!net) 1870 return ERR_PTR(-EINVAL); 1871 1872 /* For now, the caller is required to have CAP_NET_ADMIN in 1873 * the user namespace owning the target net ns. 1874 */ 1875 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) { 1876 put_net(net); 1877 return ERR_PTR(-EACCES); 1878 } 1879 return net; 1880 } 1881 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable); 1882 1883 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh, 1884 bool strict_check, struct nlattr **tb, 1885 struct netlink_ext_ack *extack) 1886 { 1887 int hdrlen; 1888 1889 if (strict_check) { 1890 struct ifinfomsg *ifm; 1891 1892 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 1893 NL_SET_ERR_MSG(extack, "Invalid header for link dump"); 1894 return -EINVAL; 1895 } 1896 1897 ifm = nlmsg_data(nlh); 1898 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 1899 ifm->ifi_change) { 1900 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request"); 1901 return -EINVAL; 1902 } 1903 if (ifm->ifi_index) { 1904 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps"); 1905 return -EINVAL; 1906 } 1907 1908 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, 1909 IFLA_MAX, ifla_policy, 1910 extack); 1911 } 1912 1913 /* A hack to preserve kernel<->userspace interface. 1914 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1915 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1916 * what iproute2 < v3.9.0 used. 1917 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1918 * attribute, its netlink message is shorter than struct ifinfomsg. 1919 */ 1920 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 1921 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1922 1923 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, 1924 extack); 1925 } 1926 1927 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1928 { 1929 struct netlink_ext_ack *extack = cb->extack; 1930 const struct nlmsghdr *nlh = cb->nlh; 1931 struct net *net = sock_net(skb->sk); 1932 struct net *tgt_net = net; 1933 int h, s_h; 1934 int idx = 0, s_idx; 1935 struct net_device *dev; 1936 struct hlist_head *head; 1937 struct nlattr *tb[IFLA_MAX+1]; 1938 u32 ext_filter_mask = 0; 1939 const struct rtnl_link_ops *kind_ops = NULL; 1940 unsigned int flags = NLM_F_MULTI; 1941 int master_idx = 0; 1942 int netnsid = -1; 1943 int err, i; 1944 1945 s_h = cb->args[0]; 1946 s_idx = cb->args[1]; 1947 1948 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack); 1949 if (err < 0) { 1950 if (cb->strict_check) 1951 return err; 1952 1953 goto walk_entries; 1954 } 1955 1956 for (i = 0; i <= IFLA_MAX; ++i) { 1957 if (!tb[i]) 1958 continue; 1959 1960 /* new attributes should only be added with strict checking */ 1961 switch (i) { 1962 case IFLA_TARGET_NETNSID: 1963 netnsid = nla_get_s32(tb[i]); 1964 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid); 1965 if (IS_ERR(tgt_net)) { 1966 NL_SET_ERR_MSG(extack, "Invalid target network namespace id"); 1967 return PTR_ERR(tgt_net); 1968 } 1969 break; 1970 case IFLA_EXT_MASK: 1971 ext_filter_mask = nla_get_u32(tb[i]); 1972 break; 1973 case IFLA_MASTER: 1974 master_idx = nla_get_u32(tb[i]); 1975 break; 1976 case IFLA_LINKINFO: 1977 kind_ops = linkinfo_to_kind_ops(tb[i]); 1978 break; 1979 default: 1980 if (cb->strict_check) { 1981 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request"); 1982 return -EINVAL; 1983 } 1984 } 1985 } 1986 1987 if (master_idx || kind_ops) 1988 flags |= NLM_F_DUMP_FILTERED; 1989 1990 walk_entries: 1991 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1992 idx = 0; 1993 head = &tgt_net->dev_index_head[h]; 1994 hlist_for_each_entry(dev, head, index_hlist) { 1995 if (link_dump_filtered(dev, master_idx, kind_ops)) 1996 goto cont; 1997 if (idx < s_idx) 1998 goto cont; 1999 err = rtnl_fill_ifinfo(skb, dev, net, 2000 RTM_NEWLINK, 2001 NETLINK_CB(cb->skb).portid, 2002 nlh->nlmsg_seq, 0, flags, 2003 ext_filter_mask, 0, NULL, 0, 2004 netnsid, GFP_KERNEL); 2005 2006 if (err < 0) { 2007 if (likely(skb->len)) 2008 goto out; 2009 2010 goto out_err; 2011 } 2012 cont: 2013 idx++; 2014 } 2015 } 2016 out: 2017 err = skb->len; 2018 out_err: 2019 cb->args[1] = idx; 2020 cb->args[0] = h; 2021 cb->seq = net->dev_base_seq; 2022 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 2023 if (netnsid >= 0) 2024 put_net(tgt_net); 2025 2026 return err; 2027 } 2028 2029 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len, 2030 struct netlink_ext_ack *exterr) 2031 { 2032 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy, 2033 exterr); 2034 } 2035 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 2036 2037 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 2038 { 2039 struct net *net; 2040 /* Examine the link attributes and figure out which 2041 * network namespace we are talking about. 2042 */ 2043 if (tb[IFLA_NET_NS_PID]) 2044 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 2045 else if (tb[IFLA_NET_NS_FD]) 2046 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 2047 else 2048 net = get_net(src_net); 2049 return net; 2050 } 2051 EXPORT_SYMBOL(rtnl_link_get_net); 2052 2053 /* Figure out which network namespace we are talking about by 2054 * examining the link attributes in the following order: 2055 * 2056 * 1. IFLA_NET_NS_PID 2057 * 2. IFLA_NET_NS_FD 2058 * 3. IFLA_TARGET_NETNSID 2059 */ 2060 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net, 2061 struct nlattr *tb[]) 2062 { 2063 struct net *net; 2064 2065 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) 2066 return rtnl_link_get_net(src_net, tb); 2067 2068 if (!tb[IFLA_TARGET_NETNSID]) 2069 return get_net(src_net); 2070 2071 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID])); 2072 if (!net) 2073 return ERR_PTR(-EINVAL); 2074 2075 return net; 2076 } 2077 2078 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb, 2079 struct net *src_net, 2080 struct nlattr *tb[], int cap) 2081 { 2082 struct net *net; 2083 2084 net = rtnl_link_get_net_by_nlattr(src_net, tb); 2085 if (IS_ERR(net)) 2086 return net; 2087 2088 if (!netlink_ns_capable(skb, net->user_ns, cap)) { 2089 put_net(net); 2090 return ERR_PTR(-EPERM); 2091 } 2092 2093 return net; 2094 } 2095 2096 /* Verify that rtnetlink requests do not pass additional properties 2097 * potentially referring to different network namespaces. 2098 */ 2099 static int rtnl_ensure_unique_netns(struct nlattr *tb[], 2100 struct netlink_ext_ack *extack, 2101 bool netns_id_only) 2102 { 2103 2104 if (netns_id_only) { 2105 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD]) 2106 return 0; 2107 2108 NL_SET_ERR_MSG(extack, "specified netns attribute not supported"); 2109 return -EOPNOTSUPP; 2110 } 2111 2112 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])) 2113 goto invalid_attr; 2114 2115 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD])) 2116 goto invalid_attr; 2117 2118 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID])) 2119 goto invalid_attr; 2120 2121 return 0; 2122 2123 invalid_attr: 2124 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified"); 2125 return -EINVAL; 2126 } 2127 2128 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 2129 { 2130 if (dev) { 2131 if (tb[IFLA_ADDRESS] && 2132 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 2133 return -EINVAL; 2134 2135 if (tb[IFLA_BROADCAST] && 2136 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 2137 return -EINVAL; 2138 } 2139 2140 if (tb[IFLA_AF_SPEC]) { 2141 struct nlattr *af; 2142 int rem, err; 2143 2144 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2145 const struct rtnl_af_ops *af_ops; 2146 2147 rcu_read_lock(); 2148 af_ops = rtnl_af_lookup(nla_type(af)); 2149 if (!af_ops) { 2150 rcu_read_unlock(); 2151 return -EAFNOSUPPORT; 2152 } 2153 2154 if (!af_ops->set_link_af) { 2155 rcu_read_unlock(); 2156 return -EOPNOTSUPP; 2157 } 2158 2159 if (af_ops->validate_link_af) { 2160 err = af_ops->validate_link_af(dev, af); 2161 if (err < 0) { 2162 rcu_read_unlock(); 2163 return err; 2164 } 2165 } 2166 2167 rcu_read_unlock(); 2168 } 2169 } 2170 2171 return 0; 2172 } 2173 2174 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 2175 int guid_type) 2176 { 2177 const struct net_device_ops *ops = dev->netdev_ops; 2178 2179 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 2180 } 2181 2182 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 2183 { 2184 if (dev->type != ARPHRD_INFINIBAND) 2185 return -EOPNOTSUPP; 2186 2187 return handle_infiniband_guid(dev, ivt, guid_type); 2188 } 2189 2190 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 2191 { 2192 const struct net_device_ops *ops = dev->netdev_ops; 2193 int err = -EINVAL; 2194 2195 if (tb[IFLA_VF_MAC]) { 2196 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 2197 2198 if (ivm->vf >= INT_MAX) 2199 return -EINVAL; 2200 err = -EOPNOTSUPP; 2201 if (ops->ndo_set_vf_mac) 2202 err = ops->ndo_set_vf_mac(dev, ivm->vf, 2203 ivm->mac); 2204 if (err < 0) 2205 return err; 2206 } 2207 2208 if (tb[IFLA_VF_VLAN]) { 2209 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 2210 2211 if (ivv->vf >= INT_MAX) 2212 return -EINVAL; 2213 err = -EOPNOTSUPP; 2214 if (ops->ndo_set_vf_vlan) 2215 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 2216 ivv->qos, 2217 htons(ETH_P_8021Q)); 2218 if (err < 0) 2219 return err; 2220 } 2221 2222 if (tb[IFLA_VF_VLAN_LIST]) { 2223 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 2224 struct nlattr *attr; 2225 int rem, len = 0; 2226 2227 err = -EOPNOTSUPP; 2228 if (!ops->ndo_set_vf_vlan) 2229 return err; 2230 2231 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 2232 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 2233 nla_len(attr) < NLA_HDRLEN) { 2234 return -EINVAL; 2235 } 2236 if (len >= MAX_VLAN_LIST_LEN) 2237 return -EOPNOTSUPP; 2238 ivvl[len] = nla_data(attr); 2239 2240 len++; 2241 } 2242 if (len == 0) 2243 return -EINVAL; 2244 2245 if (ivvl[0]->vf >= INT_MAX) 2246 return -EINVAL; 2247 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 2248 ivvl[0]->qos, ivvl[0]->vlan_proto); 2249 if (err < 0) 2250 return err; 2251 } 2252 2253 if (tb[IFLA_VF_TX_RATE]) { 2254 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 2255 struct ifla_vf_info ivf; 2256 2257 if (ivt->vf >= INT_MAX) 2258 return -EINVAL; 2259 err = -EOPNOTSUPP; 2260 if (ops->ndo_get_vf_config) 2261 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 2262 if (err < 0) 2263 return err; 2264 2265 err = -EOPNOTSUPP; 2266 if (ops->ndo_set_vf_rate) 2267 err = ops->ndo_set_vf_rate(dev, ivt->vf, 2268 ivf.min_tx_rate, 2269 ivt->rate); 2270 if (err < 0) 2271 return err; 2272 } 2273 2274 if (tb[IFLA_VF_RATE]) { 2275 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 2276 2277 if (ivt->vf >= INT_MAX) 2278 return -EINVAL; 2279 err = -EOPNOTSUPP; 2280 if (ops->ndo_set_vf_rate) 2281 err = ops->ndo_set_vf_rate(dev, ivt->vf, 2282 ivt->min_tx_rate, 2283 ivt->max_tx_rate); 2284 if (err < 0) 2285 return err; 2286 } 2287 2288 if (tb[IFLA_VF_SPOOFCHK]) { 2289 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 2290 2291 if (ivs->vf >= INT_MAX) 2292 return -EINVAL; 2293 err = -EOPNOTSUPP; 2294 if (ops->ndo_set_vf_spoofchk) 2295 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 2296 ivs->setting); 2297 if (err < 0) 2298 return err; 2299 } 2300 2301 if (tb[IFLA_VF_LINK_STATE]) { 2302 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 2303 2304 if (ivl->vf >= INT_MAX) 2305 return -EINVAL; 2306 err = -EOPNOTSUPP; 2307 if (ops->ndo_set_vf_link_state) 2308 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 2309 ivl->link_state); 2310 if (err < 0) 2311 return err; 2312 } 2313 2314 if (tb[IFLA_VF_RSS_QUERY_EN]) { 2315 struct ifla_vf_rss_query_en *ivrssq_en; 2316 2317 err = -EOPNOTSUPP; 2318 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 2319 if (ivrssq_en->vf >= INT_MAX) 2320 return -EINVAL; 2321 if (ops->ndo_set_vf_rss_query_en) 2322 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 2323 ivrssq_en->setting); 2324 if (err < 0) 2325 return err; 2326 } 2327 2328 if (tb[IFLA_VF_TRUST]) { 2329 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 2330 2331 if (ivt->vf >= INT_MAX) 2332 return -EINVAL; 2333 err = -EOPNOTSUPP; 2334 if (ops->ndo_set_vf_trust) 2335 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 2336 if (err < 0) 2337 return err; 2338 } 2339 2340 if (tb[IFLA_VF_IB_NODE_GUID]) { 2341 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 2342 2343 if (ivt->vf >= INT_MAX) 2344 return -EINVAL; 2345 if (!ops->ndo_set_vf_guid) 2346 return -EOPNOTSUPP; 2347 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 2348 } 2349 2350 if (tb[IFLA_VF_IB_PORT_GUID]) { 2351 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 2352 2353 if (ivt->vf >= INT_MAX) 2354 return -EINVAL; 2355 if (!ops->ndo_set_vf_guid) 2356 return -EOPNOTSUPP; 2357 2358 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 2359 } 2360 2361 return err; 2362 } 2363 2364 static int do_set_master(struct net_device *dev, int ifindex, 2365 struct netlink_ext_ack *extack) 2366 { 2367 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 2368 const struct net_device_ops *ops; 2369 int err; 2370 2371 if (upper_dev) { 2372 if (upper_dev->ifindex == ifindex) 2373 return 0; 2374 ops = upper_dev->netdev_ops; 2375 if (ops->ndo_del_slave) { 2376 err = ops->ndo_del_slave(upper_dev, dev); 2377 if (err) 2378 return err; 2379 netdev_update_lockdep_key(dev); 2380 } else { 2381 return -EOPNOTSUPP; 2382 } 2383 } 2384 2385 if (ifindex) { 2386 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 2387 if (!upper_dev) 2388 return -EINVAL; 2389 ops = upper_dev->netdev_ops; 2390 if (ops->ndo_add_slave) { 2391 err = ops->ndo_add_slave(upper_dev, dev, extack); 2392 if (err) 2393 return err; 2394 } else { 2395 return -EOPNOTSUPP; 2396 } 2397 } 2398 return 0; 2399 } 2400 2401 #define DO_SETLINK_MODIFIED 0x01 2402 /* notify flag means notify + modified. */ 2403 #define DO_SETLINK_NOTIFY 0x03 2404 static int do_setlink(const struct sk_buff *skb, 2405 struct net_device *dev, struct ifinfomsg *ifm, 2406 struct netlink_ext_ack *extack, 2407 struct nlattr **tb, char *ifname, int status) 2408 { 2409 const struct net_device_ops *ops = dev->netdev_ops; 2410 int err; 2411 2412 err = validate_linkmsg(dev, tb); 2413 if (err < 0) 2414 return err; 2415 2416 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) { 2417 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev), 2418 tb, CAP_NET_ADMIN); 2419 if (IS_ERR(net)) { 2420 err = PTR_ERR(net); 2421 goto errout; 2422 } 2423 2424 err = dev_change_net_namespace(dev, net, ifname); 2425 put_net(net); 2426 if (err) 2427 goto errout; 2428 status |= DO_SETLINK_MODIFIED; 2429 } 2430 2431 if (tb[IFLA_MAP]) { 2432 struct rtnl_link_ifmap *u_map; 2433 struct ifmap k_map; 2434 2435 if (!ops->ndo_set_config) { 2436 err = -EOPNOTSUPP; 2437 goto errout; 2438 } 2439 2440 if (!netif_device_present(dev)) { 2441 err = -ENODEV; 2442 goto errout; 2443 } 2444 2445 u_map = nla_data(tb[IFLA_MAP]); 2446 k_map.mem_start = (unsigned long) u_map->mem_start; 2447 k_map.mem_end = (unsigned long) u_map->mem_end; 2448 k_map.base_addr = (unsigned short) u_map->base_addr; 2449 k_map.irq = (unsigned char) u_map->irq; 2450 k_map.dma = (unsigned char) u_map->dma; 2451 k_map.port = (unsigned char) u_map->port; 2452 2453 err = ops->ndo_set_config(dev, &k_map); 2454 if (err < 0) 2455 goto errout; 2456 2457 status |= DO_SETLINK_NOTIFY; 2458 } 2459 2460 if (tb[IFLA_ADDRESS]) { 2461 struct sockaddr *sa; 2462 int len; 2463 2464 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len, 2465 sizeof(*sa)); 2466 sa = kmalloc(len, GFP_KERNEL); 2467 if (!sa) { 2468 err = -ENOMEM; 2469 goto errout; 2470 } 2471 sa->sa_family = dev->type; 2472 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 2473 dev->addr_len); 2474 err = dev_set_mac_address(dev, sa, extack); 2475 kfree(sa); 2476 if (err) 2477 goto errout; 2478 status |= DO_SETLINK_MODIFIED; 2479 } 2480 2481 if (tb[IFLA_MTU]) { 2482 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack); 2483 if (err < 0) 2484 goto errout; 2485 status |= DO_SETLINK_MODIFIED; 2486 } 2487 2488 if (tb[IFLA_GROUP]) { 2489 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2490 status |= DO_SETLINK_NOTIFY; 2491 } 2492 2493 /* 2494 * Interface selected by interface index but interface 2495 * name provided implies that a name change has been 2496 * requested. 2497 */ 2498 if (ifm->ifi_index > 0 && ifname[0]) { 2499 err = dev_change_name(dev, ifname); 2500 if (err < 0) 2501 goto errout; 2502 status |= DO_SETLINK_MODIFIED; 2503 } 2504 2505 if (tb[IFLA_IFALIAS]) { 2506 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2507 nla_len(tb[IFLA_IFALIAS])); 2508 if (err < 0) 2509 goto errout; 2510 status |= DO_SETLINK_NOTIFY; 2511 } 2512 2513 if (tb[IFLA_BROADCAST]) { 2514 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2515 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2516 } 2517 2518 if (ifm->ifi_flags || ifm->ifi_change) { 2519 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm), 2520 extack); 2521 if (err < 0) 2522 goto errout; 2523 } 2524 2525 if (tb[IFLA_MASTER]) { 2526 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack); 2527 if (err) 2528 goto errout; 2529 status |= DO_SETLINK_MODIFIED; 2530 } 2531 2532 if (tb[IFLA_CARRIER]) { 2533 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2534 if (err) 2535 goto errout; 2536 status |= DO_SETLINK_MODIFIED; 2537 } 2538 2539 if (tb[IFLA_TXQLEN]) { 2540 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]); 2541 2542 err = dev_change_tx_queue_len(dev, value); 2543 if (err) 2544 goto errout; 2545 status |= DO_SETLINK_MODIFIED; 2546 } 2547 2548 if (tb[IFLA_GSO_MAX_SIZE]) { 2549 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]); 2550 2551 if (max_size > GSO_MAX_SIZE) { 2552 err = -EINVAL; 2553 goto errout; 2554 } 2555 2556 if (dev->gso_max_size ^ max_size) { 2557 netif_set_gso_max_size(dev, max_size); 2558 status |= DO_SETLINK_MODIFIED; 2559 } 2560 } 2561 2562 if (tb[IFLA_GSO_MAX_SEGS]) { 2563 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]); 2564 2565 if (max_segs > GSO_MAX_SEGS) { 2566 err = -EINVAL; 2567 goto errout; 2568 } 2569 2570 if (dev->gso_max_segs ^ max_segs) { 2571 dev->gso_max_segs = max_segs; 2572 status |= DO_SETLINK_MODIFIED; 2573 } 2574 } 2575 2576 if (tb[IFLA_OPERSTATE]) 2577 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2578 2579 if (tb[IFLA_LINKMODE]) { 2580 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2581 2582 write_lock_bh(&dev_base_lock); 2583 if (dev->link_mode ^ value) 2584 status |= DO_SETLINK_NOTIFY; 2585 dev->link_mode = value; 2586 write_unlock_bh(&dev_base_lock); 2587 } 2588 2589 if (tb[IFLA_VFINFO_LIST]) { 2590 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2591 struct nlattr *attr; 2592 int rem; 2593 2594 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2595 if (nla_type(attr) != IFLA_VF_INFO || 2596 nla_len(attr) < NLA_HDRLEN) { 2597 err = -EINVAL; 2598 goto errout; 2599 } 2600 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX, 2601 attr, 2602 ifla_vf_policy, 2603 NULL); 2604 if (err < 0) 2605 goto errout; 2606 err = do_setvfinfo(dev, vfinfo); 2607 if (err < 0) 2608 goto errout; 2609 status |= DO_SETLINK_NOTIFY; 2610 } 2611 } 2612 err = 0; 2613 2614 if (tb[IFLA_VF_PORTS]) { 2615 struct nlattr *port[IFLA_PORT_MAX+1]; 2616 struct nlattr *attr; 2617 int vf; 2618 int rem; 2619 2620 err = -EOPNOTSUPP; 2621 if (!ops->ndo_set_vf_port) 2622 goto errout; 2623 2624 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2625 if (nla_type(attr) != IFLA_VF_PORT || 2626 nla_len(attr) < NLA_HDRLEN) { 2627 err = -EINVAL; 2628 goto errout; 2629 } 2630 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX, 2631 attr, 2632 ifla_port_policy, 2633 NULL); 2634 if (err < 0) 2635 goto errout; 2636 if (!port[IFLA_PORT_VF]) { 2637 err = -EOPNOTSUPP; 2638 goto errout; 2639 } 2640 vf = nla_get_u32(port[IFLA_PORT_VF]); 2641 err = ops->ndo_set_vf_port(dev, vf, port); 2642 if (err < 0) 2643 goto errout; 2644 status |= DO_SETLINK_NOTIFY; 2645 } 2646 } 2647 err = 0; 2648 2649 if (tb[IFLA_PORT_SELF]) { 2650 struct nlattr *port[IFLA_PORT_MAX+1]; 2651 2652 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX, 2653 tb[IFLA_PORT_SELF], 2654 ifla_port_policy, NULL); 2655 if (err < 0) 2656 goto errout; 2657 2658 err = -EOPNOTSUPP; 2659 if (ops->ndo_set_vf_port) 2660 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2661 if (err < 0) 2662 goto errout; 2663 status |= DO_SETLINK_NOTIFY; 2664 } 2665 2666 if (tb[IFLA_AF_SPEC]) { 2667 struct nlattr *af; 2668 int rem; 2669 2670 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2671 const struct rtnl_af_ops *af_ops; 2672 2673 rcu_read_lock(); 2674 2675 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af)))); 2676 2677 err = af_ops->set_link_af(dev, af); 2678 if (err < 0) { 2679 rcu_read_unlock(); 2680 goto errout; 2681 } 2682 2683 rcu_read_unlock(); 2684 status |= DO_SETLINK_NOTIFY; 2685 } 2686 } 2687 err = 0; 2688 2689 if (tb[IFLA_PROTO_DOWN]) { 2690 err = dev_change_proto_down(dev, 2691 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2692 if (err) 2693 goto errout; 2694 status |= DO_SETLINK_NOTIFY; 2695 } 2696 2697 if (tb[IFLA_XDP]) { 2698 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2699 u32 xdp_flags = 0; 2700 2701 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX, 2702 tb[IFLA_XDP], 2703 ifla_xdp_policy, NULL); 2704 if (err < 0) 2705 goto errout; 2706 2707 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) { 2708 err = -EINVAL; 2709 goto errout; 2710 } 2711 2712 if (xdp[IFLA_XDP_FLAGS]) { 2713 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2714 if (xdp_flags & ~XDP_FLAGS_MASK) { 2715 err = -EINVAL; 2716 goto errout; 2717 } 2718 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) { 2719 err = -EINVAL; 2720 goto errout; 2721 } 2722 } 2723 2724 if (xdp[IFLA_XDP_FD]) { 2725 err = dev_change_xdp_fd(dev, extack, 2726 nla_get_s32(xdp[IFLA_XDP_FD]), 2727 xdp_flags); 2728 if (err) 2729 goto errout; 2730 status |= DO_SETLINK_NOTIFY; 2731 } 2732 } 2733 2734 errout: 2735 if (status & DO_SETLINK_MODIFIED) { 2736 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY) 2737 netdev_state_change(dev); 2738 2739 if (err < 0) 2740 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", 2741 dev->name); 2742 } 2743 2744 return err; 2745 } 2746 2747 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2748 struct netlink_ext_ack *extack) 2749 { 2750 struct net *net = sock_net(skb->sk); 2751 struct ifinfomsg *ifm; 2752 struct net_device *dev; 2753 int err; 2754 struct nlattr *tb[IFLA_MAX+1]; 2755 char ifname[IFNAMSIZ]; 2756 2757 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX, 2758 ifla_policy, extack); 2759 if (err < 0) 2760 goto errout; 2761 2762 err = rtnl_ensure_unique_netns(tb, extack, false); 2763 if (err < 0) 2764 goto errout; 2765 2766 if (tb[IFLA_IFNAME]) 2767 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2768 else 2769 ifname[0] = '\0'; 2770 2771 err = -EINVAL; 2772 ifm = nlmsg_data(nlh); 2773 if (ifm->ifi_index > 0) 2774 dev = __dev_get_by_index(net, ifm->ifi_index); 2775 else if (tb[IFLA_IFNAME]) 2776 dev = __dev_get_by_name(net, ifname); 2777 else 2778 goto errout; 2779 2780 if (dev == NULL) { 2781 err = -ENODEV; 2782 goto errout; 2783 } 2784 2785 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0); 2786 errout: 2787 return err; 2788 } 2789 2790 static int rtnl_group_dellink(const struct net *net, int group) 2791 { 2792 struct net_device *dev, *aux; 2793 LIST_HEAD(list_kill); 2794 bool found = false; 2795 2796 if (!group) 2797 return -EPERM; 2798 2799 for_each_netdev(net, dev) { 2800 if (dev->group == group) { 2801 const struct rtnl_link_ops *ops; 2802 2803 found = true; 2804 ops = dev->rtnl_link_ops; 2805 if (!ops || !ops->dellink) 2806 return -EOPNOTSUPP; 2807 } 2808 } 2809 2810 if (!found) 2811 return -ENODEV; 2812 2813 for_each_netdev_safe(net, dev, aux) { 2814 if (dev->group == group) { 2815 const struct rtnl_link_ops *ops; 2816 2817 ops = dev->rtnl_link_ops; 2818 ops->dellink(dev, &list_kill); 2819 } 2820 } 2821 unregister_netdevice_many(&list_kill); 2822 2823 return 0; 2824 } 2825 2826 int rtnl_delete_link(struct net_device *dev) 2827 { 2828 const struct rtnl_link_ops *ops; 2829 LIST_HEAD(list_kill); 2830 2831 ops = dev->rtnl_link_ops; 2832 if (!ops || !ops->dellink) 2833 return -EOPNOTSUPP; 2834 2835 ops->dellink(dev, &list_kill); 2836 unregister_netdevice_many(&list_kill); 2837 2838 return 0; 2839 } 2840 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2841 2842 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 2843 struct netlink_ext_ack *extack) 2844 { 2845 struct net *net = sock_net(skb->sk); 2846 struct net *tgt_net = net; 2847 struct net_device *dev = NULL; 2848 struct ifinfomsg *ifm; 2849 char ifname[IFNAMSIZ]; 2850 struct nlattr *tb[IFLA_MAX+1]; 2851 int err; 2852 int netnsid = -1; 2853 2854 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX, 2855 ifla_policy, extack); 2856 if (err < 0) 2857 return err; 2858 2859 err = rtnl_ensure_unique_netns(tb, extack, true); 2860 if (err < 0) 2861 return err; 2862 2863 if (tb[IFLA_IFNAME]) 2864 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2865 2866 if (tb[IFLA_TARGET_NETNSID]) { 2867 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]); 2868 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid); 2869 if (IS_ERR(tgt_net)) 2870 return PTR_ERR(tgt_net); 2871 } 2872 2873 err = -EINVAL; 2874 ifm = nlmsg_data(nlh); 2875 if (ifm->ifi_index > 0) 2876 dev = __dev_get_by_index(tgt_net, ifm->ifi_index); 2877 else if (tb[IFLA_IFNAME]) 2878 dev = __dev_get_by_name(tgt_net, ifname); 2879 else if (tb[IFLA_GROUP]) 2880 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP])); 2881 else 2882 goto out; 2883 2884 if (!dev) { 2885 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0) 2886 err = -ENODEV; 2887 2888 goto out; 2889 } 2890 2891 err = rtnl_delete_link(dev); 2892 2893 out: 2894 if (netnsid >= 0) 2895 put_net(tgt_net); 2896 2897 return err; 2898 } 2899 2900 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2901 { 2902 unsigned int old_flags; 2903 int err; 2904 2905 old_flags = dev->flags; 2906 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2907 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm), 2908 NULL); 2909 if (err < 0) 2910 return err; 2911 } 2912 2913 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) { 2914 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags)); 2915 } else { 2916 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2917 __dev_notify_flags(dev, old_flags, ~0U); 2918 } 2919 return 0; 2920 } 2921 EXPORT_SYMBOL(rtnl_configure_link); 2922 2923 struct net_device *rtnl_create_link(struct net *net, const char *ifname, 2924 unsigned char name_assign_type, 2925 const struct rtnl_link_ops *ops, 2926 struct nlattr *tb[], 2927 struct netlink_ext_ack *extack) 2928 { 2929 struct net_device *dev; 2930 unsigned int num_tx_queues = 1; 2931 unsigned int num_rx_queues = 1; 2932 2933 if (tb[IFLA_NUM_TX_QUEUES]) 2934 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2935 else if (ops->get_num_tx_queues) 2936 num_tx_queues = ops->get_num_tx_queues(); 2937 2938 if (tb[IFLA_NUM_RX_QUEUES]) 2939 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2940 else if (ops->get_num_rx_queues) 2941 num_rx_queues = ops->get_num_rx_queues(); 2942 2943 if (num_tx_queues < 1 || num_tx_queues > 4096) { 2944 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues"); 2945 return ERR_PTR(-EINVAL); 2946 } 2947 2948 if (num_rx_queues < 1 || num_rx_queues > 4096) { 2949 NL_SET_ERR_MSG(extack, "Invalid number of receive queues"); 2950 return ERR_PTR(-EINVAL); 2951 } 2952 2953 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2954 ops->setup, num_tx_queues, num_rx_queues); 2955 if (!dev) 2956 return ERR_PTR(-ENOMEM); 2957 2958 dev_net_set(dev, net); 2959 dev->rtnl_link_ops = ops; 2960 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2961 2962 if (tb[IFLA_MTU]) 2963 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2964 if (tb[IFLA_ADDRESS]) { 2965 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2966 nla_len(tb[IFLA_ADDRESS])); 2967 dev->addr_assign_type = NET_ADDR_SET; 2968 } 2969 if (tb[IFLA_BROADCAST]) 2970 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2971 nla_len(tb[IFLA_BROADCAST])); 2972 if (tb[IFLA_TXQLEN]) 2973 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2974 if (tb[IFLA_OPERSTATE]) 2975 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2976 if (tb[IFLA_LINKMODE]) 2977 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2978 if (tb[IFLA_GROUP]) 2979 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2980 if (tb[IFLA_GSO_MAX_SIZE]) 2981 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE])); 2982 if (tb[IFLA_GSO_MAX_SEGS]) 2983 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]); 2984 2985 return dev; 2986 } 2987 EXPORT_SYMBOL(rtnl_create_link); 2988 2989 static int rtnl_group_changelink(const struct sk_buff *skb, 2990 struct net *net, int group, 2991 struct ifinfomsg *ifm, 2992 struct netlink_ext_ack *extack, 2993 struct nlattr **tb) 2994 { 2995 struct net_device *dev, *aux; 2996 int err; 2997 2998 for_each_netdev_safe(net, dev, aux) { 2999 if (dev->group == group) { 3000 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0); 3001 if (err < 0) 3002 return err; 3003 } 3004 } 3005 3006 return 0; 3007 } 3008 3009 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3010 struct nlattr **attr, struct netlink_ext_ack *extack) 3011 { 3012 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1]; 3013 unsigned char name_assign_type = NET_NAME_USER; 3014 struct nlattr *linkinfo[IFLA_INFO_MAX + 1]; 3015 const struct rtnl_link_ops *m_ops = NULL; 3016 struct net_device *master_dev = NULL; 3017 struct net *net = sock_net(skb->sk); 3018 const struct rtnl_link_ops *ops; 3019 struct nlattr *tb[IFLA_MAX + 1]; 3020 struct net *dest_net, *link_net; 3021 struct nlattr **slave_data; 3022 char kind[MODULE_NAME_LEN]; 3023 struct net_device *dev; 3024 struct ifinfomsg *ifm; 3025 char ifname[IFNAMSIZ]; 3026 struct nlattr **data; 3027 int err; 3028 3029 #ifdef CONFIG_MODULES 3030 replay: 3031 #endif 3032 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX, 3033 ifla_policy, extack); 3034 if (err < 0) 3035 return err; 3036 3037 err = rtnl_ensure_unique_netns(tb, extack, false); 3038 if (err < 0) 3039 return err; 3040 3041 if (tb[IFLA_IFNAME]) 3042 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 3043 else 3044 ifname[0] = '\0'; 3045 3046 ifm = nlmsg_data(nlh); 3047 if (ifm->ifi_index > 0) 3048 dev = __dev_get_by_index(net, ifm->ifi_index); 3049 else { 3050 if (ifname[0]) 3051 dev = __dev_get_by_name(net, ifname); 3052 else 3053 dev = NULL; 3054 } 3055 3056 if (dev) { 3057 master_dev = netdev_master_upper_dev_get(dev); 3058 if (master_dev) 3059 m_ops = master_dev->rtnl_link_ops; 3060 } 3061 3062 err = validate_linkmsg(dev, tb); 3063 if (err < 0) 3064 return err; 3065 3066 if (tb[IFLA_LINKINFO]) { 3067 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX, 3068 tb[IFLA_LINKINFO], 3069 ifla_info_policy, NULL); 3070 if (err < 0) 3071 return err; 3072 } else 3073 memset(linkinfo, 0, sizeof(linkinfo)); 3074 3075 if (linkinfo[IFLA_INFO_KIND]) { 3076 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 3077 ops = rtnl_link_ops_get(kind); 3078 } else { 3079 kind[0] = '\0'; 3080 ops = NULL; 3081 } 3082 3083 data = NULL; 3084 if (ops) { 3085 if (ops->maxtype > RTNL_MAX_TYPE) 3086 return -EINVAL; 3087 3088 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 3089 err = nla_parse_nested_deprecated(attr, ops->maxtype, 3090 linkinfo[IFLA_INFO_DATA], 3091 ops->policy, extack); 3092 if (err < 0) 3093 return err; 3094 data = attr; 3095 } 3096 if (ops->validate) { 3097 err = ops->validate(tb, data, extack); 3098 if (err < 0) 3099 return err; 3100 } 3101 } 3102 3103 slave_data = NULL; 3104 if (m_ops) { 3105 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE) 3106 return -EINVAL; 3107 3108 if (m_ops->slave_maxtype && 3109 linkinfo[IFLA_INFO_SLAVE_DATA]) { 3110 err = nla_parse_nested_deprecated(slave_attr, 3111 m_ops->slave_maxtype, 3112 linkinfo[IFLA_INFO_SLAVE_DATA], 3113 m_ops->slave_policy, 3114 extack); 3115 if (err < 0) 3116 return err; 3117 slave_data = slave_attr; 3118 } 3119 } 3120 3121 if (dev) { 3122 int status = 0; 3123 3124 if (nlh->nlmsg_flags & NLM_F_EXCL) 3125 return -EEXIST; 3126 if (nlh->nlmsg_flags & NLM_F_REPLACE) 3127 return -EOPNOTSUPP; 3128 3129 if (linkinfo[IFLA_INFO_DATA]) { 3130 if (!ops || ops != dev->rtnl_link_ops || 3131 !ops->changelink) 3132 return -EOPNOTSUPP; 3133 3134 err = ops->changelink(dev, tb, data, extack); 3135 if (err < 0) 3136 return err; 3137 status |= DO_SETLINK_NOTIFY; 3138 } 3139 3140 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 3141 if (!m_ops || !m_ops->slave_changelink) 3142 return -EOPNOTSUPP; 3143 3144 err = m_ops->slave_changelink(master_dev, dev, tb, 3145 slave_data, extack); 3146 if (err < 0) 3147 return err; 3148 status |= DO_SETLINK_NOTIFY; 3149 } 3150 3151 return do_setlink(skb, dev, ifm, extack, tb, ifname, status); 3152 } 3153 3154 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 3155 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 3156 return rtnl_group_changelink(skb, net, 3157 nla_get_u32(tb[IFLA_GROUP]), 3158 ifm, extack, tb); 3159 return -ENODEV; 3160 } 3161 3162 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 3163 return -EOPNOTSUPP; 3164 3165 if (!ops) { 3166 #ifdef CONFIG_MODULES 3167 if (kind[0]) { 3168 __rtnl_unlock(); 3169 request_module("rtnl-link-%s", kind); 3170 rtnl_lock(); 3171 ops = rtnl_link_ops_get(kind); 3172 if (ops) 3173 goto replay; 3174 } 3175 #endif 3176 NL_SET_ERR_MSG(extack, "Unknown device type"); 3177 return -EOPNOTSUPP; 3178 } 3179 3180 if (!ops->setup) 3181 return -EOPNOTSUPP; 3182 3183 if (!ifname[0]) { 3184 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 3185 name_assign_type = NET_NAME_ENUM; 3186 } 3187 3188 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN); 3189 if (IS_ERR(dest_net)) 3190 return PTR_ERR(dest_net); 3191 3192 if (tb[IFLA_LINK_NETNSID]) { 3193 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 3194 3195 link_net = get_net_ns_by_id(dest_net, id); 3196 if (!link_net) { 3197 NL_SET_ERR_MSG(extack, "Unknown network namespace id"); 3198 err = -EINVAL; 3199 goto out; 3200 } 3201 err = -EPERM; 3202 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 3203 goto out; 3204 } else { 3205 link_net = NULL; 3206 } 3207 3208 dev = rtnl_create_link(link_net ? : dest_net, ifname, 3209 name_assign_type, ops, tb, extack); 3210 if (IS_ERR(dev)) { 3211 err = PTR_ERR(dev); 3212 goto out; 3213 } 3214 3215 dev->ifindex = ifm->ifi_index; 3216 3217 if (ops->newlink) { 3218 err = ops->newlink(link_net ? : net, dev, tb, data, extack); 3219 /* Drivers should call free_netdev() in ->destructor 3220 * and unregister it on failure after registration 3221 * so that device could be finally freed in rtnl_unlock. 3222 */ 3223 if (err < 0) { 3224 /* If device is not registered at all, free it now */ 3225 if (dev->reg_state == NETREG_UNINITIALIZED) 3226 free_netdev(dev); 3227 goto out; 3228 } 3229 } else { 3230 err = register_netdevice(dev); 3231 if (err < 0) { 3232 free_netdev(dev); 3233 goto out; 3234 } 3235 } 3236 err = rtnl_configure_link(dev, ifm); 3237 if (err < 0) 3238 goto out_unregister; 3239 if (link_net) { 3240 err = dev_change_net_namespace(dev, dest_net, ifname); 3241 if (err < 0) 3242 goto out_unregister; 3243 } 3244 if (tb[IFLA_MASTER]) { 3245 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack); 3246 if (err) 3247 goto out_unregister; 3248 } 3249 out: 3250 if (link_net) 3251 put_net(link_net); 3252 put_net(dest_net); 3253 return err; 3254 out_unregister: 3255 if (ops->newlink) { 3256 LIST_HEAD(list_kill); 3257 3258 ops->dellink(dev, &list_kill); 3259 unregister_netdevice_many(&list_kill); 3260 } else { 3261 unregister_netdevice(dev); 3262 } 3263 goto out; 3264 } 3265 3266 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3267 struct netlink_ext_ack *extack) 3268 { 3269 struct nlattr **attr; 3270 int ret; 3271 3272 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL); 3273 if (!attr) 3274 return -ENOMEM; 3275 3276 ret = __rtnl_newlink(skb, nlh, attr, extack); 3277 kfree(attr); 3278 return ret; 3279 } 3280 3281 static int rtnl_valid_getlink_req(struct sk_buff *skb, 3282 const struct nlmsghdr *nlh, 3283 struct nlattr **tb, 3284 struct netlink_ext_ack *extack) 3285 { 3286 struct ifinfomsg *ifm; 3287 int i, err; 3288 3289 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 3290 NL_SET_ERR_MSG(extack, "Invalid header for get link"); 3291 return -EINVAL; 3292 } 3293 3294 if (!netlink_strict_get_check(skb)) 3295 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX, 3296 ifla_policy, extack); 3297 3298 ifm = nlmsg_data(nlh); 3299 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 3300 ifm->ifi_change) { 3301 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request"); 3302 return -EINVAL; 3303 } 3304 3305 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX, 3306 ifla_policy, extack); 3307 if (err) 3308 return err; 3309 3310 for (i = 0; i <= IFLA_MAX; i++) { 3311 if (!tb[i]) 3312 continue; 3313 3314 switch (i) { 3315 case IFLA_IFNAME: 3316 case IFLA_EXT_MASK: 3317 case IFLA_TARGET_NETNSID: 3318 break; 3319 default: 3320 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request"); 3321 return -EINVAL; 3322 } 3323 } 3324 3325 return 0; 3326 } 3327 3328 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3329 struct netlink_ext_ack *extack) 3330 { 3331 struct net *net = sock_net(skb->sk); 3332 struct net *tgt_net = net; 3333 struct ifinfomsg *ifm; 3334 char ifname[IFNAMSIZ]; 3335 struct nlattr *tb[IFLA_MAX+1]; 3336 struct net_device *dev = NULL; 3337 struct sk_buff *nskb; 3338 int netnsid = -1; 3339 int err; 3340 u32 ext_filter_mask = 0; 3341 3342 err = rtnl_valid_getlink_req(skb, nlh, tb, extack); 3343 if (err < 0) 3344 return err; 3345 3346 err = rtnl_ensure_unique_netns(tb, extack, true); 3347 if (err < 0) 3348 return err; 3349 3350 if (tb[IFLA_TARGET_NETNSID]) { 3351 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]); 3352 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid); 3353 if (IS_ERR(tgt_net)) 3354 return PTR_ERR(tgt_net); 3355 } 3356 3357 if (tb[IFLA_IFNAME]) 3358 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 3359 3360 if (tb[IFLA_EXT_MASK]) 3361 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 3362 3363 err = -EINVAL; 3364 ifm = nlmsg_data(nlh); 3365 if (ifm->ifi_index > 0) 3366 dev = __dev_get_by_index(tgt_net, ifm->ifi_index); 3367 else if (tb[IFLA_IFNAME]) 3368 dev = __dev_get_by_name(tgt_net, ifname); 3369 else 3370 goto out; 3371 3372 err = -ENODEV; 3373 if (dev == NULL) 3374 goto out; 3375 3376 err = -ENOBUFS; 3377 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 3378 if (nskb == NULL) 3379 goto out; 3380 3381 err = rtnl_fill_ifinfo(nskb, dev, net, 3382 RTM_NEWLINK, NETLINK_CB(skb).portid, 3383 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 3384 0, NULL, 0, netnsid, GFP_KERNEL); 3385 if (err < 0) { 3386 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 3387 WARN_ON(err == -EMSGSIZE); 3388 kfree_skb(nskb); 3389 } else 3390 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 3391 out: 3392 if (netnsid >= 0) 3393 put_net(tgt_net); 3394 3395 return err; 3396 } 3397 3398 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 3399 { 3400 struct net *net = sock_net(skb->sk); 3401 struct net_device *dev; 3402 struct nlattr *tb[IFLA_MAX+1]; 3403 u32 ext_filter_mask = 0; 3404 u16 min_ifinfo_dump_size = 0; 3405 int hdrlen; 3406 3407 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 3408 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 3409 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 3410 3411 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) { 3412 if (tb[IFLA_EXT_MASK]) 3413 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 3414 } 3415 3416 if (!ext_filter_mask) 3417 return NLMSG_GOODSIZE; 3418 /* 3419 * traverse the list of net devices and compute the minimum 3420 * buffer size based upon the filter mask. 3421 */ 3422 rcu_read_lock(); 3423 for_each_netdev_rcu(net, dev) { 3424 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 3425 if_nlmsg_size(dev, 3426 ext_filter_mask)); 3427 } 3428 rcu_read_unlock(); 3429 3430 return nlmsg_total_size(min_ifinfo_dump_size); 3431 } 3432 3433 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 3434 { 3435 int idx; 3436 int s_idx = cb->family; 3437 int type = cb->nlh->nlmsg_type - RTM_BASE; 3438 int ret = 0; 3439 3440 if (s_idx == 0) 3441 s_idx = 1; 3442 3443 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 3444 struct rtnl_link **tab; 3445 struct rtnl_link *link; 3446 rtnl_dumpit_func dumpit; 3447 3448 if (idx < s_idx || idx == PF_PACKET) 3449 continue; 3450 3451 if (type < 0 || type >= RTM_NR_MSGTYPES) 3452 continue; 3453 3454 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]); 3455 if (!tab) 3456 continue; 3457 3458 link = tab[type]; 3459 if (!link) 3460 continue; 3461 3462 dumpit = link->dumpit; 3463 if (!dumpit) 3464 continue; 3465 3466 if (idx > s_idx) { 3467 memset(&cb->args[0], 0, sizeof(cb->args)); 3468 cb->prev_seq = 0; 3469 cb->seq = 0; 3470 } 3471 ret = dumpit(skb, cb); 3472 if (ret) 3473 break; 3474 } 3475 cb->family = idx; 3476 3477 return skb->len ? : ret; 3478 } 3479 3480 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 3481 unsigned int change, 3482 u32 event, gfp_t flags, int *new_nsid, 3483 int new_ifindex) 3484 { 3485 struct net *net = dev_net(dev); 3486 struct sk_buff *skb; 3487 int err = -ENOBUFS; 3488 size_t if_info_size; 3489 3490 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 3491 if (skb == NULL) 3492 goto errout; 3493 3494 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev), 3495 type, 0, 0, change, 0, 0, event, 3496 new_nsid, new_ifindex, -1, flags); 3497 if (err < 0) { 3498 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 3499 WARN_ON(err == -EMSGSIZE); 3500 kfree_skb(skb); 3501 goto errout; 3502 } 3503 return skb; 3504 errout: 3505 if (err < 0) 3506 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3507 return NULL; 3508 } 3509 3510 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 3511 { 3512 struct net *net = dev_net(dev); 3513 3514 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 3515 } 3516 3517 static void rtmsg_ifinfo_event(int type, struct net_device *dev, 3518 unsigned int change, u32 event, 3519 gfp_t flags, int *new_nsid, int new_ifindex) 3520 { 3521 struct sk_buff *skb; 3522 3523 if (dev->reg_state != NETREG_REGISTERED) 3524 return; 3525 3526 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid, 3527 new_ifindex); 3528 if (skb) 3529 rtmsg_ifinfo_send(skb, dev, flags); 3530 } 3531 3532 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 3533 gfp_t flags) 3534 { 3535 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, 3536 NULL, 0); 3537 } 3538 3539 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change, 3540 gfp_t flags, int *new_nsid, int new_ifindex) 3541 { 3542 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags, 3543 new_nsid, new_ifindex); 3544 } 3545 3546 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 3547 struct net_device *dev, 3548 u8 *addr, u16 vid, u32 pid, u32 seq, 3549 int type, unsigned int flags, 3550 int nlflags, u16 ndm_state) 3551 { 3552 struct nlmsghdr *nlh; 3553 struct ndmsg *ndm; 3554 3555 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 3556 if (!nlh) 3557 return -EMSGSIZE; 3558 3559 ndm = nlmsg_data(nlh); 3560 ndm->ndm_family = AF_BRIDGE; 3561 ndm->ndm_pad1 = 0; 3562 ndm->ndm_pad2 = 0; 3563 ndm->ndm_flags = flags; 3564 ndm->ndm_type = 0; 3565 ndm->ndm_ifindex = dev->ifindex; 3566 ndm->ndm_state = ndm_state; 3567 3568 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 3569 goto nla_put_failure; 3570 if (vid) 3571 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 3572 goto nla_put_failure; 3573 3574 nlmsg_end(skb, nlh); 3575 return 0; 3576 3577 nla_put_failure: 3578 nlmsg_cancel(skb, nlh); 3579 return -EMSGSIZE; 3580 } 3581 3582 static inline size_t rtnl_fdb_nlmsg_size(void) 3583 { 3584 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 3585 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 3586 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 3587 0; 3588 } 3589 3590 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 3591 u16 ndm_state) 3592 { 3593 struct net *net = dev_net(dev); 3594 struct sk_buff *skb; 3595 int err = -ENOBUFS; 3596 3597 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 3598 if (!skb) 3599 goto errout; 3600 3601 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 3602 0, 0, type, NTF_SELF, 0, ndm_state); 3603 if (err < 0) { 3604 kfree_skb(skb); 3605 goto errout; 3606 } 3607 3608 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 3609 return; 3610 errout: 3611 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 3612 } 3613 3614 /* 3615 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 3616 */ 3617 int ndo_dflt_fdb_add(struct ndmsg *ndm, 3618 struct nlattr *tb[], 3619 struct net_device *dev, 3620 const unsigned char *addr, u16 vid, 3621 u16 flags) 3622 { 3623 int err = -EINVAL; 3624 3625 /* If aging addresses are supported device will need to 3626 * implement its own handler for this. 3627 */ 3628 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 3629 pr_info("%s: FDB only supports static addresses\n", dev->name); 3630 return err; 3631 } 3632 3633 if (vid) { 3634 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 3635 return err; 3636 } 3637 3638 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3639 err = dev_uc_add_excl(dev, addr); 3640 else if (is_multicast_ether_addr(addr)) 3641 err = dev_mc_add_excl(dev, addr); 3642 3643 /* Only return duplicate errors if NLM_F_EXCL is set */ 3644 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 3645 err = 0; 3646 3647 return err; 3648 } 3649 EXPORT_SYMBOL(ndo_dflt_fdb_add); 3650 3651 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid, 3652 struct netlink_ext_ack *extack) 3653 { 3654 u16 vid = 0; 3655 3656 if (vlan_attr) { 3657 if (nla_len(vlan_attr) != sizeof(u16)) { 3658 NL_SET_ERR_MSG(extack, "invalid vlan attribute size"); 3659 return -EINVAL; 3660 } 3661 3662 vid = nla_get_u16(vlan_attr); 3663 3664 if (!vid || vid >= VLAN_VID_MASK) { 3665 NL_SET_ERR_MSG(extack, "invalid vlan id"); 3666 return -EINVAL; 3667 } 3668 } 3669 *p_vid = vid; 3670 return 0; 3671 } 3672 3673 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, 3674 struct netlink_ext_ack *extack) 3675 { 3676 struct net *net = sock_net(skb->sk); 3677 struct ndmsg *ndm; 3678 struct nlattr *tb[NDA_MAX+1]; 3679 struct net_device *dev; 3680 u8 *addr; 3681 u16 vid; 3682 int err; 3683 3684 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, 3685 extack); 3686 if (err < 0) 3687 return err; 3688 3689 ndm = nlmsg_data(nlh); 3690 if (ndm->ndm_ifindex == 0) { 3691 NL_SET_ERR_MSG(extack, "invalid ifindex"); 3692 return -EINVAL; 3693 } 3694 3695 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3696 if (dev == NULL) { 3697 NL_SET_ERR_MSG(extack, "unknown ifindex"); 3698 return -ENODEV; 3699 } 3700 3701 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3702 NL_SET_ERR_MSG(extack, "invalid address"); 3703 return -EINVAL; 3704 } 3705 3706 if (dev->type != ARPHRD_ETHER) { 3707 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices"); 3708 return -EINVAL; 3709 } 3710 3711 addr = nla_data(tb[NDA_LLADDR]); 3712 3713 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack); 3714 if (err) 3715 return err; 3716 3717 err = -EOPNOTSUPP; 3718 3719 /* Support fdb on master device the net/bridge default case */ 3720 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3721 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3722 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3723 const struct net_device_ops *ops = br_dev->netdev_ops; 3724 3725 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3726 nlh->nlmsg_flags, extack); 3727 if (err) 3728 goto out; 3729 else 3730 ndm->ndm_flags &= ~NTF_MASTER; 3731 } 3732 3733 /* Embedded bridge, macvlan, and any other device support */ 3734 if ((ndm->ndm_flags & NTF_SELF)) { 3735 if (dev->netdev_ops->ndo_fdb_add) 3736 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3737 vid, 3738 nlh->nlmsg_flags, 3739 extack); 3740 else 3741 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3742 nlh->nlmsg_flags); 3743 3744 if (!err) { 3745 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3746 ndm->ndm_state); 3747 ndm->ndm_flags &= ~NTF_SELF; 3748 } 3749 } 3750 out: 3751 return err; 3752 } 3753 3754 /* 3755 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3756 */ 3757 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3758 struct nlattr *tb[], 3759 struct net_device *dev, 3760 const unsigned char *addr, u16 vid) 3761 { 3762 int err = -EINVAL; 3763 3764 /* If aging addresses are supported device will need to 3765 * implement its own handler for this. 3766 */ 3767 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3768 pr_info("%s: FDB only supports static addresses\n", dev->name); 3769 return err; 3770 } 3771 3772 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3773 err = dev_uc_del(dev, addr); 3774 else if (is_multicast_ether_addr(addr)) 3775 err = dev_mc_del(dev, addr); 3776 3777 return err; 3778 } 3779 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3780 3781 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, 3782 struct netlink_ext_ack *extack) 3783 { 3784 struct net *net = sock_net(skb->sk); 3785 struct ndmsg *ndm; 3786 struct nlattr *tb[NDA_MAX+1]; 3787 struct net_device *dev; 3788 int err = -EINVAL; 3789 __u8 *addr; 3790 u16 vid; 3791 3792 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3793 return -EPERM; 3794 3795 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, 3796 extack); 3797 if (err < 0) 3798 return err; 3799 3800 ndm = nlmsg_data(nlh); 3801 if (ndm->ndm_ifindex == 0) { 3802 NL_SET_ERR_MSG(extack, "invalid ifindex"); 3803 return -EINVAL; 3804 } 3805 3806 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3807 if (dev == NULL) { 3808 NL_SET_ERR_MSG(extack, "unknown ifindex"); 3809 return -ENODEV; 3810 } 3811 3812 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3813 NL_SET_ERR_MSG(extack, "invalid address"); 3814 return -EINVAL; 3815 } 3816 3817 if (dev->type != ARPHRD_ETHER) { 3818 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices"); 3819 return -EINVAL; 3820 } 3821 3822 addr = nla_data(tb[NDA_LLADDR]); 3823 3824 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack); 3825 if (err) 3826 return err; 3827 3828 err = -EOPNOTSUPP; 3829 3830 /* Support fdb on master device the net/bridge default case */ 3831 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3832 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3833 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3834 const struct net_device_ops *ops = br_dev->netdev_ops; 3835 3836 if (ops->ndo_fdb_del) 3837 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3838 3839 if (err) 3840 goto out; 3841 else 3842 ndm->ndm_flags &= ~NTF_MASTER; 3843 } 3844 3845 /* Embedded bridge, macvlan, and any other device support */ 3846 if (ndm->ndm_flags & NTF_SELF) { 3847 if (dev->netdev_ops->ndo_fdb_del) 3848 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3849 vid); 3850 else 3851 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3852 3853 if (!err) { 3854 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3855 ndm->ndm_state); 3856 ndm->ndm_flags &= ~NTF_SELF; 3857 } 3858 } 3859 out: 3860 return err; 3861 } 3862 3863 static int nlmsg_populate_fdb(struct sk_buff *skb, 3864 struct netlink_callback *cb, 3865 struct net_device *dev, 3866 int *idx, 3867 struct netdev_hw_addr_list *list) 3868 { 3869 struct netdev_hw_addr *ha; 3870 int err; 3871 u32 portid, seq; 3872 3873 portid = NETLINK_CB(cb->skb).portid; 3874 seq = cb->nlh->nlmsg_seq; 3875 3876 list_for_each_entry(ha, &list->list, list) { 3877 if (*idx < cb->args[2]) 3878 goto skip; 3879 3880 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3881 portid, seq, 3882 RTM_NEWNEIGH, NTF_SELF, 3883 NLM_F_MULTI, NUD_PERMANENT); 3884 if (err < 0) 3885 return err; 3886 skip: 3887 *idx += 1; 3888 } 3889 return 0; 3890 } 3891 3892 /** 3893 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3894 * @skb: socket buffer to store message in 3895 * @cb: netlink callback 3896 * @dev: netdevice 3897 * @filter_dev: ignored 3898 * @idx: the number of FDB table entries dumped is added to *@idx 3899 * 3900 * Default netdevice operation to dump the existing unicast address list. 3901 * Returns number of addresses from list put in skb. 3902 */ 3903 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3904 struct netlink_callback *cb, 3905 struct net_device *dev, 3906 struct net_device *filter_dev, 3907 int *idx) 3908 { 3909 int err; 3910 3911 if (dev->type != ARPHRD_ETHER) 3912 return -EINVAL; 3913 3914 netif_addr_lock_bh(dev); 3915 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3916 if (err) 3917 goto out; 3918 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3919 out: 3920 netif_addr_unlock_bh(dev); 3921 return err; 3922 } 3923 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3924 3925 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh, 3926 int *br_idx, int *brport_idx, 3927 struct netlink_ext_ack *extack) 3928 { 3929 struct nlattr *tb[NDA_MAX + 1]; 3930 struct ndmsg *ndm; 3931 int err, i; 3932 3933 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) { 3934 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request"); 3935 return -EINVAL; 3936 } 3937 3938 ndm = nlmsg_data(nlh); 3939 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state || 3940 ndm->ndm_flags || ndm->ndm_type) { 3941 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request"); 3942 return -EINVAL; 3943 } 3944 3945 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb, 3946 NDA_MAX, NULL, extack); 3947 if (err < 0) 3948 return err; 3949 3950 *brport_idx = ndm->ndm_ifindex; 3951 for (i = 0; i <= NDA_MAX; ++i) { 3952 if (!tb[i]) 3953 continue; 3954 3955 switch (i) { 3956 case NDA_IFINDEX: 3957 if (nla_len(tb[i]) != sizeof(u32)) { 3958 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request"); 3959 return -EINVAL; 3960 } 3961 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]); 3962 break; 3963 case NDA_MASTER: 3964 if (nla_len(tb[i]) != sizeof(u32)) { 3965 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request"); 3966 return -EINVAL; 3967 } 3968 *br_idx = nla_get_u32(tb[NDA_MASTER]); 3969 break; 3970 default: 3971 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request"); 3972 return -EINVAL; 3973 } 3974 } 3975 3976 return 0; 3977 } 3978 3979 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh, 3980 int *br_idx, int *brport_idx, 3981 struct netlink_ext_ack *extack) 3982 { 3983 struct nlattr *tb[IFLA_MAX+1]; 3984 int err; 3985 3986 /* A hack to preserve kernel<->userspace interface. 3987 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0. 3988 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails. 3989 * So, check for ndmsg with an optional u32 attribute (not used here). 3990 * Fortunately these sizes don't conflict with the size of ifinfomsg 3991 * with an optional attribute. 3992 */ 3993 if (nlmsg_len(nlh) != sizeof(struct ndmsg) && 3994 (nlmsg_len(nlh) != sizeof(struct ndmsg) + 3995 nla_attr_size(sizeof(u32)))) { 3996 struct ifinfomsg *ifm; 3997 3998 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg), 3999 tb, IFLA_MAX, ifla_policy, 4000 extack); 4001 if (err < 0) { 4002 return -EINVAL; 4003 } else if (err == 0) { 4004 if (tb[IFLA_MASTER]) 4005 *br_idx = nla_get_u32(tb[IFLA_MASTER]); 4006 } 4007 4008 ifm = nlmsg_data(nlh); 4009 *brport_idx = ifm->ifi_index; 4010 } 4011 return 0; 4012 } 4013 4014 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 4015 { 4016 struct net_device *dev; 4017 struct net_device *br_dev = NULL; 4018 const struct net_device_ops *ops = NULL; 4019 const struct net_device_ops *cops = NULL; 4020 struct net *net = sock_net(skb->sk); 4021 struct hlist_head *head; 4022 int brport_idx = 0; 4023 int br_idx = 0; 4024 int h, s_h; 4025 int idx = 0, s_idx; 4026 int err = 0; 4027 int fidx = 0; 4028 4029 if (cb->strict_check) 4030 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx, 4031 cb->extack); 4032 else 4033 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx, 4034 cb->extack); 4035 if (err < 0) 4036 return err; 4037 4038 if (br_idx) { 4039 br_dev = __dev_get_by_index(net, br_idx); 4040 if (!br_dev) 4041 return -ENODEV; 4042 4043 ops = br_dev->netdev_ops; 4044 } 4045 4046 s_h = cb->args[0]; 4047 s_idx = cb->args[1]; 4048 4049 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4050 idx = 0; 4051 head = &net->dev_index_head[h]; 4052 hlist_for_each_entry(dev, head, index_hlist) { 4053 4054 if (brport_idx && (dev->ifindex != brport_idx)) 4055 continue; 4056 4057 if (!br_idx) { /* user did not specify a specific bridge */ 4058 if (dev->priv_flags & IFF_BRIDGE_PORT) { 4059 br_dev = netdev_master_upper_dev_get(dev); 4060 cops = br_dev->netdev_ops; 4061 } 4062 } else { 4063 if (dev != br_dev && 4064 !(dev->priv_flags & IFF_BRIDGE_PORT)) 4065 continue; 4066 4067 if (br_dev != netdev_master_upper_dev_get(dev) && 4068 !(dev->priv_flags & IFF_EBRIDGE)) 4069 continue; 4070 cops = ops; 4071 } 4072 4073 if (idx < s_idx) 4074 goto cont; 4075 4076 if (dev->priv_flags & IFF_BRIDGE_PORT) { 4077 if (cops && cops->ndo_fdb_dump) { 4078 err = cops->ndo_fdb_dump(skb, cb, 4079 br_dev, dev, 4080 &fidx); 4081 if (err == -EMSGSIZE) 4082 goto out; 4083 } 4084 } 4085 4086 if (dev->netdev_ops->ndo_fdb_dump) 4087 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 4088 dev, NULL, 4089 &fidx); 4090 else 4091 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 4092 &fidx); 4093 if (err == -EMSGSIZE) 4094 goto out; 4095 4096 cops = NULL; 4097 4098 /* reset fdb offset to 0 for rest of the interfaces */ 4099 cb->args[2] = 0; 4100 fidx = 0; 4101 cont: 4102 idx++; 4103 } 4104 } 4105 4106 out: 4107 cb->args[0] = h; 4108 cb->args[1] = idx; 4109 cb->args[2] = fidx; 4110 4111 return skb->len; 4112 } 4113 4114 static int valid_fdb_get_strict(const struct nlmsghdr *nlh, 4115 struct nlattr **tb, u8 *ndm_flags, 4116 int *br_idx, int *brport_idx, u8 **addr, 4117 u16 *vid, struct netlink_ext_ack *extack) 4118 { 4119 struct ndmsg *ndm; 4120 int err, i; 4121 4122 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) { 4123 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request"); 4124 return -EINVAL; 4125 } 4126 4127 ndm = nlmsg_data(nlh); 4128 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state || 4129 ndm->ndm_type) { 4130 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request"); 4131 return -EINVAL; 4132 } 4133 4134 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) { 4135 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request"); 4136 return -EINVAL; 4137 } 4138 4139 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb, 4140 NDA_MAX, nda_policy, extack); 4141 if (err < 0) 4142 return err; 4143 4144 *ndm_flags = ndm->ndm_flags; 4145 *brport_idx = ndm->ndm_ifindex; 4146 for (i = 0; i <= NDA_MAX; ++i) { 4147 if (!tb[i]) 4148 continue; 4149 4150 switch (i) { 4151 case NDA_MASTER: 4152 *br_idx = nla_get_u32(tb[i]); 4153 break; 4154 case NDA_LLADDR: 4155 if (nla_len(tb[i]) != ETH_ALEN) { 4156 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request"); 4157 return -EINVAL; 4158 } 4159 *addr = nla_data(tb[i]); 4160 break; 4161 case NDA_VLAN: 4162 err = fdb_vid_parse(tb[i], vid, extack); 4163 if (err) 4164 return err; 4165 break; 4166 case NDA_VNI: 4167 break; 4168 default: 4169 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request"); 4170 return -EINVAL; 4171 } 4172 } 4173 4174 return 0; 4175 } 4176 4177 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh, 4178 struct netlink_ext_ack *extack) 4179 { 4180 struct net_device *dev = NULL, *br_dev = NULL; 4181 const struct net_device_ops *ops = NULL; 4182 struct net *net = sock_net(in_skb->sk); 4183 struct nlattr *tb[NDA_MAX + 1]; 4184 struct sk_buff *skb; 4185 int brport_idx = 0; 4186 u8 ndm_flags = 0; 4187 int br_idx = 0; 4188 u8 *addr = NULL; 4189 u16 vid = 0; 4190 int err; 4191 4192 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx, 4193 &brport_idx, &addr, &vid, extack); 4194 if (err < 0) 4195 return err; 4196 4197 if (!addr) { 4198 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request"); 4199 return -EINVAL; 4200 } 4201 4202 if (brport_idx) { 4203 dev = __dev_get_by_index(net, brport_idx); 4204 if (!dev) { 4205 NL_SET_ERR_MSG(extack, "Unknown device ifindex"); 4206 return -ENODEV; 4207 } 4208 } 4209 4210 if (br_idx) { 4211 if (dev) { 4212 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive"); 4213 return -EINVAL; 4214 } 4215 4216 br_dev = __dev_get_by_index(net, br_idx); 4217 if (!br_dev) { 4218 NL_SET_ERR_MSG(extack, "Invalid master ifindex"); 4219 return -EINVAL; 4220 } 4221 ops = br_dev->netdev_ops; 4222 } 4223 4224 if (dev) { 4225 if (!ndm_flags || (ndm_flags & NTF_MASTER)) { 4226 if (!(dev->priv_flags & IFF_BRIDGE_PORT)) { 4227 NL_SET_ERR_MSG(extack, "Device is not a bridge port"); 4228 return -EINVAL; 4229 } 4230 br_dev = netdev_master_upper_dev_get(dev); 4231 if (!br_dev) { 4232 NL_SET_ERR_MSG(extack, "Master of device not found"); 4233 return -EINVAL; 4234 } 4235 ops = br_dev->netdev_ops; 4236 } else { 4237 if (!(ndm_flags & NTF_SELF)) { 4238 NL_SET_ERR_MSG(extack, "Missing NTF_SELF"); 4239 return -EINVAL; 4240 } 4241 ops = dev->netdev_ops; 4242 } 4243 } 4244 4245 if (!br_dev && !dev) { 4246 NL_SET_ERR_MSG(extack, "No device specified"); 4247 return -ENODEV; 4248 } 4249 4250 if (!ops || !ops->ndo_fdb_get) { 4251 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device"); 4252 return -EOPNOTSUPP; 4253 } 4254 4255 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); 4256 if (!skb) 4257 return -ENOBUFS; 4258 4259 if (br_dev) 4260 dev = br_dev; 4261 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid, 4262 NETLINK_CB(in_skb).portid, 4263 nlh->nlmsg_seq, extack); 4264 if (err) 4265 goto out; 4266 4267 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 4268 out: 4269 kfree_skb(skb); 4270 return err; 4271 } 4272 4273 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 4274 unsigned int attrnum, unsigned int flag) 4275 { 4276 if (mask & flag) 4277 return nla_put_u8(skb, attrnum, !!(flags & flag)); 4278 return 0; 4279 } 4280 4281 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 4282 struct net_device *dev, u16 mode, 4283 u32 flags, u32 mask, int nlflags, 4284 u32 filter_mask, 4285 int (*vlan_fill)(struct sk_buff *skb, 4286 struct net_device *dev, 4287 u32 filter_mask)) 4288 { 4289 struct nlmsghdr *nlh; 4290 struct ifinfomsg *ifm; 4291 struct nlattr *br_afspec; 4292 struct nlattr *protinfo; 4293 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 4294 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4295 int err = 0; 4296 4297 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 4298 if (nlh == NULL) 4299 return -EMSGSIZE; 4300 4301 ifm = nlmsg_data(nlh); 4302 ifm->ifi_family = AF_BRIDGE; 4303 ifm->__ifi_pad = 0; 4304 ifm->ifi_type = dev->type; 4305 ifm->ifi_index = dev->ifindex; 4306 ifm->ifi_flags = dev_get_flags(dev); 4307 ifm->ifi_change = 0; 4308 4309 4310 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 4311 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 4312 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 4313 (br_dev && 4314 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 4315 (dev->addr_len && 4316 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 4317 (dev->ifindex != dev_get_iflink(dev) && 4318 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 4319 goto nla_put_failure; 4320 4321 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC); 4322 if (!br_afspec) 4323 goto nla_put_failure; 4324 4325 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 4326 nla_nest_cancel(skb, br_afspec); 4327 goto nla_put_failure; 4328 } 4329 4330 if (mode != BRIDGE_MODE_UNDEF) { 4331 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 4332 nla_nest_cancel(skb, br_afspec); 4333 goto nla_put_failure; 4334 } 4335 } 4336 if (vlan_fill) { 4337 err = vlan_fill(skb, dev, filter_mask); 4338 if (err) { 4339 nla_nest_cancel(skb, br_afspec); 4340 goto nla_put_failure; 4341 } 4342 } 4343 nla_nest_end(skb, br_afspec); 4344 4345 protinfo = nla_nest_start(skb, IFLA_PROTINFO); 4346 if (!protinfo) 4347 goto nla_put_failure; 4348 4349 if (brport_nla_put_flag(skb, flags, mask, 4350 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 4351 brport_nla_put_flag(skb, flags, mask, 4352 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 4353 brport_nla_put_flag(skb, flags, mask, 4354 IFLA_BRPORT_FAST_LEAVE, 4355 BR_MULTICAST_FAST_LEAVE) || 4356 brport_nla_put_flag(skb, flags, mask, 4357 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 4358 brport_nla_put_flag(skb, flags, mask, 4359 IFLA_BRPORT_LEARNING, BR_LEARNING) || 4360 brport_nla_put_flag(skb, flags, mask, 4361 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 4362 brport_nla_put_flag(skb, flags, mask, 4363 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 4364 brport_nla_put_flag(skb, flags, mask, 4365 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 4366 nla_nest_cancel(skb, protinfo); 4367 goto nla_put_failure; 4368 } 4369 4370 nla_nest_end(skb, protinfo); 4371 4372 nlmsg_end(skb, nlh); 4373 return 0; 4374 nla_put_failure: 4375 nlmsg_cancel(skb, nlh); 4376 return err ? err : -EMSGSIZE; 4377 } 4378 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 4379 4380 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh, 4381 bool strict_check, u32 *filter_mask, 4382 struct netlink_ext_ack *extack) 4383 { 4384 struct nlattr *tb[IFLA_MAX+1]; 4385 int err, i; 4386 4387 if (strict_check) { 4388 struct ifinfomsg *ifm; 4389 4390 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 4391 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump"); 4392 return -EINVAL; 4393 } 4394 4395 ifm = nlmsg_data(nlh); 4396 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 4397 ifm->ifi_change || ifm->ifi_index) { 4398 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request"); 4399 return -EINVAL; 4400 } 4401 4402 err = nlmsg_parse_deprecated_strict(nlh, 4403 sizeof(struct ifinfomsg), 4404 tb, IFLA_MAX, ifla_policy, 4405 extack); 4406 } else { 4407 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg), 4408 tb, IFLA_MAX, ifla_policy, 4409 extack); 4410 } 4411 if (err < 0) 4412 return err; 4413 4414 /* new attributes should only be added with strict checking */ 4415 for (i = 0; i <= IFLA_MAX; ++i) { 4416 if (!tb[i]) 4417 continue; 4418 4419 switch (i) { 4420 case IFLA_EXT_MASK: 4421 *filter_mask = nla_get_u32(tb[i]); 4422 break; 4423 default: 4424 if (strict_check) { 4425 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request"); 4426 return -EINVAL; 4427 } 4428 } 4429 } 4430 4431 return 0; 4432 } 4433 4434 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 4435 { 4436 const struct nlmsghdr *nlh = cb->nlh; 4437 struct net *net = sock_net(skb->sk); 4438 struct net_device *dev; 4439 int idx = 0; 4440 u32 portid = NETLINK_CB(cb->skb).portid; 4441 u32 seq = nlh->nlmsg_seq; 4442 u32 filter_mask = 0; 4443 int err; 4444 4445 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask, 4446 cb->extack); 4447 if (err < 0 && cb->strict_check) 4448 return err; 4449 4450 rcu_read_lock(); 4451 for_each_netdev_rcu(net, dev) { 4452 const struct net_device_ops *ops = dev->netdev_ops; 4453 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4454 4455 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 4456 if (idx >= cb->args[0]) { 4457 err = br_dev->netdev_ops->ndo_bridge_getlink( 4458 skb, portid, seq, dev, 4459 filter_mask, NLM_F_MULTI); 4460 if (err < 0 && err != -EOPNOTSUPP) { 4461 if (likely(skb->len)) 4462 break; 4463 4464 goto out_err; 4465 } 4466 } 4467 idx++; 4468 } 4469 4470 if (ops->ndo_bridge_getlink) { 4471 if (idx >= cb->args[0]) { 4472 err = ops->ndo_bridge_getlink(skb, portid, 4473 seq, dev, 4474 filter_mask, 4475 NLM_F_MULTI); 4476 if (err < 0 && err != -EOPNOTSUPP) { 4477 if (likely(skb->len)) 4478 break; 4479 4480 goto out_err; 4481 } 4482 } 4483 idx++; 4484 } 4485 } 4486 err = skb->len; 4487 out_err: 4488 rcu_read_unlock(); 4489 cb->args[0] = idx; 4490 4491 return err; 4492 } 4493 4494 static inline size_t bridge_nlmsg_size(void) 4495 { 4496 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 4497 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 4498 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 4499 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 4500 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 4501 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 4502 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 4503 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 4504 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 4505 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 4506 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 4507 } 4508 4509 static int rtnl_bridge_notify(struct net_device *dev) 4510 { 4511 struct net *net = dev_net(dev); 4512 struct sk_buff *skb; 4513 int err = -EOPNOTSUPP; 4514 4515 if (!dev->netdev_ops->ndo_bridge_getlink) 4516 return 0; 4517 4518 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 4519 if (!skb) { 4520 err = -ENOMEM; 4521 goto errout; 4522 } 4523 4524 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 4525 if (err < 0) 4526 goto errout; 4527 4528 if (!skb->len) 4529 goto errout; 4530 4531 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 4532 return 0; 4533 errout: 4534 WARN_ON(err == -EMSGSIZE); 4535 kfree_skb(skb); 4536 if (err) 4537 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 4538 return err; 4539 } 4540 4541 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 4542 struct netlink_ext_ack *extack) 4543 { 4544 struct net *net = sock_net(skb->sk); 4545 struct ifinfomsg *ifm; 4546 struct net_device *dev; 4547 struct nlattr *br_spec, *attr = NULL; 4548 int rem, err = -EOPNOTSUPP; 4549 u16 flags = 0; 4550 bool have_flags = false; 4551 4552 if (nlmsg_len(nlh) < sizeof(*ifm)) 4553 return -EINVAL; 4554 4555 ifm = nlmsg_data(nlh); 4556 if (ifm->ifi_family != AF_BRIDGE) 4557 return -EPFNOSUPPORT; 4558 4559 dev = __dev_get_by_index(net, ifm->ifi_index); 4560 if (!dev) { 4561 NL_SET_ERR_MSG(extack, "unknown ifindex"); 4562 return -ENODEV; 4563 } 4564 4565 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 4566 if (br_spec) { 4567 nla_for_each_nested(attr, br_spec, rem) { 4568 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 4569 if (nla_len(attr) < sizeof(flags)) 4570 return -EINVAL; 4571 4572 have_flags = true; 4573 flags = nla_get_u16(attr); 4574 break; 4575 } 4576 } 4577 } 4578 4579 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 4580 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4581 4582 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 4583 err = -EOPNOTSUPP; 4584 goto out; 4585 } 4586 4587 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags, 4588 extack); 4589 if (err) 4590 goto out; 4591 4592 flags &= ~BRIDGE_FLAGS_MASTER; 4593 } 4594 4595 if ((flags & BRIDGE_FLAGS_SELF)) { 4596 if (!dev->netdev_ops->ndo_bridge_setlink) 4597 err = -EOPNOTSUPP; 4598 else 4599 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 4600 flags, 4601 extack); 4602 if (!err) { 4603 flags &= ~BRIDGE_FLAGS_SELF; 4604 4605 /* Generate event to notify upper layer of bridge 4606 * change 4607 */ 4608 err = rtnl_bridge_notify(dev); 4609 } 4610 } 4611 4612 if (have_flags) 4613 memcpy(nla_data(attr), &flags, sizeof(flags)); 4614 out: 4615 return err; 4616 } 4617 4618 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 4619 struct netlink_ext_ack *extack) 4620 { 4621 struct net *net = sock_net(skb->sk); 4622 struct ifinfomsg *ifm; 4623 struct net_device *dev; 4624 struct nlattr *br_spec, *attr = NULL; 4625 int rem, err = -EOPNOTSUPP; 4626 u16 flags = 0; 4627 bool have_flags = false; 4628 4629 if (nlmsg_len(nlh) < sizeof(*ifm)) 4630 return -EINVAL; 4631 4632 ifm = nlmsg_data(nlh); 4633 if (ifm->ifi_family != AF_BRIDGE) 4634 return -EPFNOSUPPORT; 4635 4636 dev = __dev_get_by_index(net, ifm->ifi_index); 4637 if (!dev) { 4638 NL_SET_ERR_MSG(extack, "unknown ifindex"); 4639 return -ENODEV; 4640 } 4641 4642 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 4643 if (br_spec) { 4644 nla_for_each_nested(attr, br_spec, rem) { 4645 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 4646 if (nla_len(attr) < sizeof(flags)) 4647 return -EINVAL; 4648 4649 have_flags = true; 4650 flags = nla_get_u16(attr); 4651 break; 4652 } 4653 } 4654 } 4655 4656 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 4657 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 4658 4659 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 4660 err = -EOPNOTSUPP; 4661 goto out; 4662 } 4663 4664 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 4665 if (err) 4666 goto out; 4667 4668 flags &= ~BRIDGE_FLAGS_MASTER; 4669 } 4670 4671 if ((flags & BRIDGE_FLAGS_SELF)) { 4672 if (!dev->netdev_ops->ndo_bridge_dellink) 4673 err = -EOPNOTSUPP; 4674 else 4675 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 4676 flags); 4677 4678 if (!err) { 4679 flags &= ~BRIDGE_FLAGS_SELF; 4680 4681 /* Generate event to notify upper layer of bridge 4682 * change 4683 */ 4684 err = rtnl_bridge_notify(dev); 4685 } 4686 } 4687 4688 if (have_flags) 4689 memcpy(nla_data(attr), &flags, sizeof(flags)); 4690 out: 4691 return err; 4692 } 4693 4694 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 4695 { 4696 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 4697 (!idxattr || idxattr == attrid); 4698 } 4699 4700 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 4701 static int rtnl_get_offload_stats_attr_size(int attr_id) 4702 { 4703 switch (attr_id) { 4704 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 4705 return sizeof(struct rtnl_link_stats64); 4706 } 4707 4708 return 0; 4709 } 4710 4711 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 4712 int *prividx) 4713 { 4714 struct nlattr *attr = NULL; 4715 int attr_id, size; 4716 void *attr_data; 4717 int err; 4718 4719 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 4720 dev->netdev_ops->ndo_get_offload_stats)) 4721 return -ENODATA; 4722 4723 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 4724 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 4725 if (attr_id < *prividx) 4726 continue; 4727 4728 size = rtnl_get_offload_stats_attr_size(attr_id); 4729 if (!size) 4730 continue; 4731 4732 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 4733 continue; 4734 4735 attr = nla_reserve_64bit(skb, attr_id, size, 4736 IFLA_OFFLOAD_XSTATS_UNSPEC); 4737 if (!attr) 4738 goto nla_put_failure; 4739 4740 attr_data = nla_data(attr); 4741 memset(attr_data, 0, size); 4742 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 4743 attr_data); 4744 if (err) 4745 goto get_offload_stats_failure; 4746 } 4747 4748 if (!attr) 4749 return -ENODATA; 4750 4751 *prividx = 0; 4752 return 0; 4753 4754 nla_put_failure: 4755 err = -EMSGSIZE; 4756 get_offload_stats_failure: 4757 *prividx = attr_id; 4758 return err; 4759 } 4760 4761 static int rtnl_get_offload_stats_size(const struct net_device *dev) 4762 { 4763 int nla_size = 0; 4764 int attr_id; 4765 int size; 4766 4767 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 4768 dev->netdev_ops->ndo_get_offload_stats)) 4769 return 0; 4770 4771 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 4772 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 4773 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 4774 continue; 4775 size = rtnl_get_offload_stats_attr_size(attr_id); 4776 nla_size += nla_total_size_64bit(size); 4777 } 4778 4779 if (nla_size != 0) 4780 nla_size += nla_total_size(0); 4781 4782 return nla_size; 4783 } 4784 4785 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 4786 int type, u32 pid, u32 seq, u32 change, 4787 unsigned int flags, unsigned int filter_mask, 4788 int *idxattr, int *prividx) 4789 { 4790 struct if_stats_msg *ifsm; 4791 struct nlmsghdr *nlh; 4792 struct nlattr *attr; 4793 int s_prividx = *prividx; 4794 int err; 4795 4796 ASSERT_RTNL(); 4797 4798 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 4799 if (!nlh) 4800 return -EMSGSIZE; 4801 4802 ifsm = nlmsg_data(nlh); 4803 ifsm->family = PF_UNSPEC; 4804 ifsm->pad1 = 0; 4805 ifsm->pad2 = 0; 4806 ifsm->ifindex = dev->ifindex; 4807 ifsm->filter_mask = filter_mask; 4808 4809 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 4810 struct rtnl_link_stats64 *sp; 4811 4812 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 4813 sizeof(struct rtnl_link_stats64), 4814 IFLA_STATS_UNSPEC); 4815 if (!attr) 4816 goto nla_put_failure; 4817 4818 sp = nla_data(attr); 4819 dev_get_stats(dev, sp); 4820 } 4821 4822 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 4823 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 4824 4825 if (ops && ops->fill_linkxstats) { 4826 *idxattr = IFLA_STATS_LINK_XSTATS; 4827 attr = nla_nest_start_noflag(skb, 4828 IFLA_STATS_LINK_XSTATS); 4829 if (!attr) 4830 goto nla_put_failure; 4831 4832 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 4833 nla_nest_end(skb, attr); 4834 if (err) 4835 goto nla_put_failure; 4836 *idxattr = 0; 4837 } 4838 } 4839 4840 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 4841 *idxattr)) { 4842 const struct rtnl_link_ops *ops = NULL; 4843 const struct net_device *master; 4844 4845 master = netdev_master_upper_dev_get(dev); 4846 if (master) 4847 ops = master->rtnl_link_ops; 4848 if (ops && ops->fill_linkxstats) { 4849 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 4850 attr = nla_nest_start_noflag(skb, 4851 IFLA_STATS_LINK_XSTATS_SLAVE); 4852 if (!attr) 4853 goto nla_put_failure; 4854 4855 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 4856 nla_nest_end(skb, attr); 4857 if (err) 4858 goto nla_put_failure; 4859 *idxattr = 0; 4860 } 4861 } 4862 4863 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 4864 *idxattr)) { 4865 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 4866 attr = nla_nest_start_noflag(skb, 4867 IFLA_STATS_LINK_OFFLOAD_XSTATS); 4868 if (!attr) 4869 goto nla_put_failure; 4870 4871 err = rtnl_get_offload_stats(skb, dev, prividx); 4872 if (err == -ENODATA) 4873 nla_nest_cancel(skb, attr); 4874 else 4875 nla_nest_end(skb, attr); 4876 4877 if (err && err != -ENODATA) 4878 goto nla_put_failure; 4879 *idxattr = 0; 4880 } 4881 4882 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 4883 struct rtnl_af_ops *af_ops; 4884 4885 *idxattr = IFLA_STATS_AF_SPEC; 4886 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC); 4887 if (!attr) 4888 goto nla_put_failure; 4889 4890 rcu_read_lock(); 4891 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 4892 if (af_ops->fill_stats_af) { 4893 struct nlattr *af; 4894 int err; 4895 4896 af = nla_nest_start_noflag(skb, 4897 af_ops->family); 4898 if (!af) { 4899 rcu_read_unlock(); 4900 goto nla_put_failure; 4901 } 4902 err = af_ops->fill_stats_af(skb, dev); 4903 4904 if (err == -ENODATA) { 4905 nla_nest_cancel(skb, af); 4906 } else if (err < 0) { 4907 rcu_read_unlock(); 4908 goto nla_put_failure; 4909 } 4910 4911 nla_nest_end(skb, af); 4912 } 4913 } 4914 rcu_read_unlock(); 4915 4916 nla_nest_end(skb, attr); 4917 4918 *idxattr = 0; 4919 } 4920 4921 nlmsg_end(skb, nlh); 4922 4923 return 0; 4924 4925 nla_put_failure: 4926 /* not a multi message or no progress mean a real error */ 4927 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 4928 nlmsg_cancel(skb, nlh); 4929 else 4930 nlmsg_end(skb, nlh); 4931 4932 return -EMSGSIZE; 4933 } 4934 4935 static size_t if_nlmsg_stats_size(const struct net_device *dev, 4936 u32 filter_mask) 4937 { 4938 size_t size = 0; 4939 4940 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 4941 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 4942 4943 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 4944 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 4945 int attr = IFLA_STATS_LINK_XSTATS; 4946 4947 if (ops && ops->get_linkxstats_size) { 4948 size += nla_total_size(ops->get_linkxstats_size(dev, 4949 attr)); 4950 /* for IFLA_STATS_LINK_XSTATS */ 4951 size += nla_total_size(0); 4952 } 4953 } 4954 4955 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 4956 struct net_device *_dev = (struct net_device *)dev; 4957 const struct rtnl_link_ops *ops = NULL; 4958 const struct net_device *master; 4959 4960 /* netdev_master_upper_dev_get can't take const */ 4961 master = netdev_master_upper_dev_get(_dev); 4962 if (master) 4963 ops = master->rtnl_link_ops; 4964 if (ops && ops->get_linkxstats_size) { 4965 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 4966 4967 size += nla_total_size(ops->get_linkxstats_size(dev, 4968 attr)); 4969 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 4970 size += nla_total_size(0); 4971 } 4972 } 4973 4974 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 4975 size += rtnl_get_offload_stats_size(dev); 4976 4977 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 4978 struct rtnl_af_ops *af_ops; 4979 4980 /* for IFLA_STATS_AF_SPEC */ 4981 size += nla_total_size(0); 4982 4983 rcu_read_lock(); 4984 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) { 4985 if (af_ops->get_stats_af_size) { 4986 size += nla_total_size( 4987 af_ops->get_stats_af_size(dev)); 4988 4989 /* for AF_* */ 4990 size += nla_total_size(0); 4991 } 4992 } 4993 rcu_read_unlock(); 4994 } 4995 4996 return size; 4997 } 4998 4999 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check, 5000 bool is_dump, struct netlink_ext_ack *extack) 5001 { 5002 struct if_stats_msg *ifsm; 5003 5004 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) { 5005 NL_SET_ERR_MSG(extack, "Invalid header for stats dump"); 5006 return -EINVAL; 5007 } 5008 5009 if (!strict_check) 5010 return 0; 5011 5012 ifsm = nlmsg_data(nlh); 5013 5014 /* only requests using strict checks can pass data to influence 5015 * the dump. The legacy exception is filter_mask. 5016 */ 5017 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) { 5018 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request"); 5019 return -EINVAL; 5020 } 5021 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) { 5022 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header"); 5023 return -EINVAL; 5024 } 5025 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) { 5026 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask"); 5027 return -EINVAL; 5028 } 5029 5030 return 0; 5031 } 5032 5033 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh, 5034 struct netlink_ext_ack *extack) 5035 { 5036 struct net *net = sock_net(skb->sk); 5037 struct net_device *dev = NULL; 5038 int idxattr = 0, prividx = 0; 5039 struct if_stats_msg *ifsm; 5040 struct sk_buff *nskb; 5041 u32 filter_mask; 5042 int err; 5043 5044 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb), 5045 false, extack); 5046 if (err) 5047 return err; 5048 5049 ifsm = nlmsg_data(nlh); 5050 if (ifsm->ifindex > 0) 5051 dev = __dev_get_by_index(net, ifsm->ifindex); 5052 else 5053 return -EINVAL; 5054 5055 if (!dev) 5056 return -ENODEV; 5057 5058 filter_mask = ifsm->filter_mask; 5059 if (!filter_mask) 5060 return -EINVAL; 5061 5062 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 5063 if (!nskb) 5064 return -ENOBUFS; 5065 5066 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 5067 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 5068 0, filter_mask, &idxattr, &prividx); 5069 if (err < 0) { 5070 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 5071 WARN_ON(err == -EMSGSIZE); 5072 kfree_skb(nskb); 5073 } else { 5074 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 5075 } 5076 5077 return err; 5078 } 5079 5080 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 5081 { 5082 struct netlink_ext_ack *extack = cb->extack; 5083 int h, s_h, err, s_idx, s_idxattr, s_prividx; 5084 struct net *net = sock_net(skb->sk); 5085 unsigned int flags = NLM_F_MULTI; 5086 struct if_stats_msg *ifsm; 5087 struct hlist_head *head; 5088 struct net_device *dev; 5089 u32 filter_mask = 0; 5090 int idx = 0; 5091 5092 s_h = cb->args[0]; 5093 s_idx = cb->args[1]; 5094 s_idxattr = cb->args[2]; 5095 s_prividx = cb->args[3]; 5096 5097 cb->seq = net->dev_base_seq; 5098 5099 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack); 5100 if (err) 5101 return err; 5102 5103 ifsm = nlmsg_data(cb->nlh); 5104 filter_mask = ifsm->filter_mask; 5105 if (!filter_mask) { 5106 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump"); 5107 return -EINVAL; 5108 } 5109 5110 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 5111 idx = 0; 5112 head = &net->dev_index_head[h]; 5113 hlist_for_each_entry(dev, head, index_hlist) { 5114 if (idx < s_idx) 5115 goto cont; 5116 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 5117 NETLINK_CB(cb->skb).portid, 5118 cb->nlh->nlmsg_seq, 0, 5119 flags, filter_mask, 5120 &s_idxattr, &s_prividx); 5121 /* If we ran out of room on the first message, 5122 * we're in trouble 5123 */ 5124 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 5125 5126 if (err < 0) 5127 goto out; 5128 s_prividx = 0; 5129 s_idxattr = 0; 5130 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 5131 cont: 5132 idx++; 5133 } 5134 } 5135 out: 5136 cb->args[3] = s_prividx; 5137 cb->args[2] = s_idxattr; 5138 cb->args[1] = idx; 5139 cb->args[0] = h; 5140 5141 return skb->len; 5142 } 5143 5144 /* Process one rtnetlink message. */ 5145 5146 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, 5147 struct netlink_ext_ack *extack) 5148 { 5149 struct net *net = sock_net(skb->sk); 5150 struct rtnl_link *link; 5151 struct module *owner; 5152 int err = -EOPNOTSUPP; 5153 rtnl_doit_func doit; 5154 unsigned int flags; 5155 int kind; 5156 int family; 5157 int type; 5158 5159 type = nlh->nlmsg_type; 5160 if (type > RTM_MAX) 5161 return -EOPNOTSUPP; 5162 5163 type -= RTM_BASE; 5164 5165 /* All the messages must have at least 1 byte length */ 5166 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 5167 return 0; 5168 5169 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 5170 kind = type&3; 5171 5172 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 5173 return -EPERM; 5174 5175 rcu_read_lock(); 5176 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 5177 struct sock *rtnl; 5178 rtnl_dumpit_func dumpit; 5179 u16 min_dump_alloc = 0; 5180 5181 link = rtnl_get_link(family, type); 5182 if (!link || !link->dumpit) { 5183 family = PF_UNSPEC; 5184 link = rtnl_get_link(family, type); 5185 if (!link || !link->dumpit) 5186 goto err_unlock; 5187 } 5188 owner = link->owner; 5189 dumpit = link->dumpit; 5190 5191 if (type == RTM_GETLINK - RTM_BASE) 5192 min_dump_alloc = rtnl_calcit(skb, nlh); 5193 5194 err = 0; 5195 /* need to do this before rcu_read_unlock() */ 5196 if (!try_module_get(owner)) 5197 err = -EPROTONOSUPPORT; 5198 5199 rcu_read_unlock(); 5200 5201 rtnl = net->rtnl; 5202 if (err == 0) { 5203 struct netlink_dump_control c = { 5204 .dump = dumpit, 5205 .min_dump_alloc = min_dump_alloc, 5206 .module = owner, 5207 }; 5208 err = netlink_dump_start(rtnl, skb, nlh, &c); 5209 /* netlink_dump_start() will keep a reference on 5210 * module if dump is still in progress. 5211 */ 5212 module_put(owner); 5213 } 5214 return err; 5215 } 5216 5217 link = rtnl_get_link(family, type); 5218 if (!link || !link->doit) { 5219 family = PF_UNSPEC; 5220 link = rtnl_get_link(PF_UNSPEC, type); 5221 if (!link || !link->doit) 5222 goto out_unlock; 5223 } 5224 5225 owner = link->owner; 5226 if (!try_module_get(owner)) { 5227 err = -EPROTONOSUPPORT; 5228 goto out_unlock; 5229 } 5230 5231 flags = link->flags; 5232 if (flags & RTNL_FLAG_DOIT_UNLOCKED) { 5233 doit = link->doit; 5234 rcu_read_unlock(); 5235 if (doit) 5236 err = doit(skb, nlh, extack); 5237 module_put(owner); 5238 return err; 5239 } 5240 rcu_read_unlock(); 5241 5242 rtnl_lock(); 5243 link = rtnl_get_link(family, type); 5244 if (link && link->doit) 5245 err = link->doit(skb, nlh, extack); 5246 rtnl_unlock(); 5247 5248 module_put(owner); 5249 5250 return err; 5251 5252 out_unlock: 5253 rcu_read_unlock(); 5254 return err; 5255 5256 err_unlock: 5257 rcu_read_unlock(); 5258 return -EOPNOTSUPP; 5259 } 5260 5261 static void rtnetlink_rcv(struct sk_buff *skb) 5262 { 5263 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 5264 } 5265 5266 static int rtnetlink_bind(struct net *net, int group) 5267 { 5268 switch (group) { 5269 case RTNLGRP_IPV4_MROUTE_R: 5270 case RTNLGRP_IPV6_MROUTE_R: 5271 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 5272 return -EPERM; 5273 break; 5274 } 5275 return 0; 5276 } 5277 5278 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 5279 { 5280 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 5281 5282 switch (event) { 5283 case NETDEV_REBOOT: 5284 case NETDEV_CHANGEMTU: 5285 case NETDEV_CHANGEADDR: 5286 case NETDEV_CHANGENAME: 5287 case NETDEV_FEAT_CHANGE: 5288 case NETDEV_BONDING_FAILOVER: 5289 case NETDEV_POST_TYPE_CHANGE: 5290 case NETDEV_NOTIFY_PEERS: 5291 case NETDEV_CHANGEUPPER: 5292 case NETDEV_RESEND_IGMP: 5293 case NETDEV_CHANGEINFODATA: 5294 case NETDEV_CHANGELOWERSTATE: 5295 case NETDEV_CHANGE_TX_QUEUE_LEN: 5296 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event), 5297 GFP_KERNEL, NULL, 0); 5298 break; 5299 default: 5300 break; 5301 } 5302 return NOTIFY_DONE; 5303 } 5304 5305 static struct notifier_block rtnetlink_dev_notifier = { 5306 .notifier_call = rtnetlink_event, 5307 }; 5308 5309 5310 static int __net_init rtnetlink_net_init(struct net *net) 5311 { 5312 struct sock *sk; 5313 struct netlink_kernel_cfg cfg = { 5314 .groups = RTNLGRP_MAX, 5315 .input = rtnetlink_rcv, 5316 .cb_mutex = &rtnl_mutex, 5317 .flags = NL_CFG_F_NONROOT_RECV, 5318 .bind = rtnetlink_bind, 5319 }; 5320 5321 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 5322 if (!sk) 5323 return -ENOMEM; 5324 net->rtnl = sk; 5325 return 0; 5326 } 5327 5328 static void __net_exit rtnetlink_net_exit(struct net *net) 5329 { 5330 netlink_kernel_release(net->rtnl); 5331 net->rtnl = NULL; 5332 } 5333 5334 static struct pernet_operations rtnetlink_net_ops = { 5335 .init = rtnetlink_net_init, 5336 .exit = rtnetlink_net_exit, 5337 }; 5338 5339 void __init rtnetlink_init(void) 5340 { 5341 if (register_pernet_subsys(&rtnetlink_net_ops)) 5342 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 5343 5344 register_netdevice_notifier(&rtnetlink_dev_notifier); 5345 5346 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 5347 rtnl_dump_ifinfo, 0); 5348 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0); 5349 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0); 5350 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0); 5351 5352 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0); 5353 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0); 5354 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0); 5355 5356 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0); 5357 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0); 5358 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0); 5359 5360 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0); 5361 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0); 5362 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0); 5363 5364 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 5365 0); 5366 } 5367