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