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