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