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