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