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 noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb, 1215 struct net_device *dev, 1216 u32 ext_filter_mask) 1217 { 1218 struct nlattr *vfinfo; 1219 int i, num_vfs; 1220 1221 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0)) 1222 return 0; 1223 1224 num_vfs = dev_num_vf(dev->dev.parent); 1225 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs)) 1226 return -EMSGSIZE; 1227 1228 if (!dev->netdev_ops->ndo_get_vf_config) 1229 return 0; 1230 1231 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1232 if (!vfinfo) 1233 return -EMSGSIZE; 1234 1235 for (i = 0; i < num_vfs; i++) { 1236 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1237 return -EMSGSIZE; 1238 } 1239 1240 nla_nest_end(skb, vfinfo); 1241 return 0; 1242 } 1243 1244 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev) 1245 { 1246 struct rtnl_link_ifmap map; 1247 1248 memset(&map, 0, sizeof(map)); 1249 map.mem_start = dev->mem_start; 1250 map.mem_end = dev->mem_end; 1251 map.base_addr = dev->base_addr; 1252 map.irq = dev->irq; 1253 map.dma = dev->dma; 1254 map.port = dev->if_port; 1255 1256 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD)) 1257 return -EMSGSIZE; 1258 1259 return 0; 1260 } 1261 1262 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id) 1263 { 1264 const struct net_device_ops *ops = dev->netdev_ops; 1265 const struct bpf_prog *generic_xdp_prog; 1266 1267 ASSERT_RTNL(); 1268 1269 *prog_id = 0; 1270 generic_xdp_prog = rtnl_dereference(dev->xdp_prog); 1271 if (generic_xdp_prog) { 1272 *prog_id = generic_xdp_prog->aux->id; 1273 return XDP_ATTACHED_SKB; 1274 } 1275 if (!ops->ndo_xdp) 1276 return XDP_ATTACHED_NONE; 1277 1278 return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id); 1279 } 1280 1281 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1282 { 1283 struct nlattr *xdp; 1284 u32 prog_id; 1285 int err; 1286 1287 xdp = nla_nest_start(skb, IFLA_XDP); 1288 if (!xdp) 1289 return -EMSGSIZE; 1290 1291 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, 1292 rtnl_xdp_attached_mode(dev, &prog_id)); 1293 if (err) 1294 goto err_cancel; 1295 1296 if (prog_id) { 1297 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id); 1298 if (err) 1299 goto err_cancel; 1300 } 1301 1302 nla_nest_end(skb, xdp); 1303 return 0; 1304 1305 err_cancel: 1306 nla_nest_cancel(skb, xdp); 1307 return err; 1308 } 1309 1310 static u32 rtnl_get_event(unsigned long event) 1311 { 1312 u32 rtnl_event_type = IFLA_EVENT_NONE; 1313 1314 switch (event) { 1315 case NETDEV_REBOOT: 1316 rtnl_event_type = IFLA_EVENT_REBOOT; 1317 break; 1318 case NETDEV_FEAT_CHANGE: 1319 rtnl_event_type = IFLA_EVENT_FEATURES; 1320 break; 1321 case NETDEV_BONDING_FAILOVER: 1322 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER; 1323 break; 1324 case NETDEV_NOTIFY_PEERS: 1325 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS; 1326 break; 1327 case NETDEV_RESEND_IGMP: 1328 rtnl_event_type = IFLA_EVENT_IGMP_RESEND; 1329 break; 1330 case NETDEV_CHANGEINFODATA: 1331 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS; 1332 break; 1333 default: 1334 break; 1335 } 1336 1337 return rtnl_event_type; 1338 } 1339 1340 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev) 1341 { 1342 const struct net_device *upper_dev; 1343 int ret = 0; 1344 1345 rcu_read_lock(); 1346 1347 upper_dev = netdev_master_upper_dev_get_rcu(dev); 1348 if (upper_dev) 1349 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex); 1350 1351 rcu_read_unlock(); 1352 return ret; 1353 } 1354 1355 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev) 1356 { 1357 int ifindex = dev_get_iflink(dev); 1358 1359 if (dev->ifindex == ifindex) 1360 return 0; 1361 1362 return nla_put_u32(skb, IFLA_LINK, ifindex); 1363 } 1364 1365 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 1366 int type, u32 pid, u32 seq, u32 change, 1367 unsigned int flags, u32 ext_filter_mask, 1368 u32 event) 1369 { 1370 struct ifinfomsg *ifm; 1371 struct nlmsghdr *nlh; 1372 struct nlattr *af_spec; 1373 struct rtnl_af_ops *af_ops; 1374 1375 ASSERT_RTNL(); 1376 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1377 if (nlh == NULL) 1378 return -EMSGSIZE; 1379 1380 ifm = nlmsg_data(nlh); 1381 ifm->ifi_family = AF_UNSPEC; 1382 ifm->__ifi_pad = 0; 1383 ifm->ifi_type = dev->type; 1384 ifm->ifi_index = dev->ifindex; 1385 ifm->ifi_flags = dev_get_flags(dev); 1386 ifm->ifi_change = change; 1387 1388 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1389 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1390 nla_put_u8(skb, IFLA_OPERSTATE, 1391 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1392 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1393 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1394 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1395 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1396 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1397 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1398 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1399 #ifdef CONFIG_RPS 1400 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1401 #endif 1402 nla_put_iflink(skb, dev) || 1403 put_master_ifindex(skb, dev) || 1404 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1405 (dev->qdisc && 1406 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1407 (dev->ifalias && 1408 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) || 1409 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1410 atomic_read(&dev->carrier_changes)) || 1411 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down)) 1412 goto nla_put_failure; 1413 1414 if (event != IFLA_EVENT_NONE) { 1415 if (nla_put_u32(skb, IFLA_EVENT, event)) 1416 goto nla_put_failure; 1417 } 1418 1419 if (rtnl_fill_link_ifmap(skb, dev)) 1420 goto nla_put_failure; 1421 1422 if (dev->addr_len) { 1423 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1424 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1425 goto nla_put_failure; 1426 } 1427 1428 if (rtnl_phys_port_id_fill(skb, dev)) 1429 goto nla_put_failure; 1430 1431 if (rtnl_phys_port_name_fill(skb, dev)) 1432 goto nla_put_failure; 1433 1434 if (rtnl_phys_switch_id_fill(skb, dev)) 1435 goto nla_put_failure; 1436 1437 if (rtnl_fill_stats(skb, dev)) 1438 goto nla_put_failure; 1439 1440 if (rtnl_fill_vf(skb, dev, ext_filter_mask)) 1441 goto nla_put_failure; 1442 1443 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1444 goto nla_put_failure; 1445 1446 if (rtnl_xdp_fill(skb, dev)) 1447 goto nla_put_failure; 1448 1449 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1450 if (rtnl_link_fill(skb, dev) < 0) 1451 goto nla_put_failure; 1452 } 1453 1454 if (dev->rtnl_link_ops && 1455 dev->rtnl_link_ops->get_link_net) { 1456 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1457 1458 if (!net_eq(dev_net(dev), link_net)) { 1459 int id = peernet2id_alloc(dev_net(dev), link_net); 1460 1461 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1462 goto nla_put_failure; 1463 } 1464 } 1465 1466 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC))) 1467 goto nla_put_failure; 1468 1469 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 1470 if (af_ops->fill_link_af) { 1471 struct nlattr *af; 1472 int err; 1473 1474 if (!(af = nla_nest_start(skb, af_ops->family))) 1475 goto nla_put_failure; 1476 1477 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1478 1479 /* 1480 * Caller may return ENODATA to indicate that there 1481 * was no data to be dumped. This is not an error, it 1482 * means we should trim the attribute header and 1483 * continue. 1484 */ 1485 if (err == -ENODATA) 1486 nla_nest_cancel(skb, af); 1487 else if (err < 0) 1488 goto nla_put_failure; 1489 1490 nla_nest_end(skb, af); 1491 } 1492 } 1493 1494 nla_nest_end(skb, af_spec); 1495 1496 nlmsg_end(skb, nlh); 1497 return 0; 1498 1499 nla_put_failure: 1500 nlmsg_cancel(skb, nlh); 1501 return -EMSGSIZE; 1502 } 1503 1504 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1505 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1506 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1507 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1508 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1509 [IFLA_MTU] = { .type = NLA_U32 }, 1510 [IFLA_LINK] = { .type = NLA_U32 }, 1511 [IFLA_MASTER] = { .type = NLA_U32 }, 1512 [IFLA_CARRIER] = { .type = NLA_U8 }, 1513 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1514 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1515 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1516 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1517 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1518 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1519 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1520 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 1521 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1522 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1523 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1524 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1525 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1526 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1527 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1528 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1529 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1530 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1531 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1532 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1533 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1534 [IFLA_XDP] = { .type = NLA_NESTED }, 1535 [IFLA_EVENT] = { .type = NLA_U32 }, 1536 [IFLA_GROUP] = { .type = NLA_U32 }, 1537 }; 1538 1539 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1540 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1541 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1542 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1543 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1544 }; 1545 1546 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1547 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1548 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1549 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1550 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1551 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1552 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1553 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1554 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1555 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1556 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1557 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1558 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1559 }; 1560 1561 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1562 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1563 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1564 .len = PORT_PROFILE_MAX }, 1565 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1566 .len = PORT_UUID_MAX }, 1567 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1568 .len = PORT_UUID_MAX }, 1569 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1570 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1571 1572 /* Unused, but we need to keep it here since user space could 1573 * fill it. It's also broken with regard to NLA_BINARY use in 1574 * combination with structs. 1575 */ 1576 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1577 .len = sizeof(struct ifla_port_vsi) }, 1578 }; 1579 1580 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1581 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1582 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1583 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1584 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 }, 1585 }; 1586 1587 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1588 { 1589 const struct rtnl_link_ops *ops = NULL; 1590 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1591 1592 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, 1593 ifla_info_policy, NULL) < 0) 1594 return NULL; 1595 1596 if (linfo[IFLA_INFO_KIND]) { 1597 char kind[MODULE_NAME_LEN]; 1598 1599 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1600 ops = rtnl_link_ops_get(kind); 1601 } 1602 1603 return ops; 1604 } 1605 1606 static bool link_master_filtered(struct net_device *dev, int master_idx) 1607 { 1608 struct net_device *master; 1609 1610 if (!master_idx) 1611 return false; 1612 1613 master = netdev_master_upper_dev_get(dev); 1614 if (!master || master->ifindex != master_idx) 1615 return true; 1616 1617 return false; 1618 } 1619 1620 static bool link_kind_filtered(const struct net_device *dev, 1621 const struct rtnl_link_ops *kind_ops) 1622 { 1623 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1624 return true; 1625 1626 return false; 1627 } 1628 1629 static bool link_dump_filtered(struct net_device *dev, 1630 int master_idx, 1631 const struct rtnl_link_ops *kind_ops) 1632 { 1633 if (link_master_filtered(dev, master_idx) || 1634 link_kind_filtered(dev, kind_ops)) 1635 return true; 1636 1637 return false; 1638 } 1639 1640 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1641 { 1642 struct net *net = sock_net(skb->sk); 1643 int h, s_h; 1644 int idx = 0, s_idx; 1645 struct net_device *dev; 1646 struct hlist_head *head; 1647 struct nlattr *tb[IFLA_MAX+1]; 1648 u32 ext_filter_mask = 0; 1649 const struct rtnl_link_ops *kind_ops = NULL; 1650 unsigned int flags = NLM_F_MULTI; 1651 int master_idx = 0; 1652 int err; 1653 int hdrlen; 1654 1655 s_h = cb->args[0]; 1656 s_idx = cb->args[1]; 1657 1658 /* A hack to preserve kernel<->userspace interface. 1659 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1660 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1661 * what iproute2 < v3.9.0 used. 1662 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1663 * attribute, its netlink message is shorter than struct ifinfomsg. 1664 */ 1665 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1666 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1667 1668 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, 1669 ifla_policy, NULL) >= 0) { 1670 if (tb[IFLA_EXT_MASK]) 1671 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1672 1673 if (tb[IFLA_MASTER]) 1674 master_idx = nla_get_u32(tb[IFLA_MASTER]); 1675 1676 if (tb[IFLA_LINKINFO]) 1677 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]); 1678 1679 if (master_idx || kind_ops) 1680 flags |= NLM_F_DUMP_FILTERED; 1681 } 1682 1683 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1684 idx = 0; 1685 head = &net->dev_index_head[h]; 1686 hlist_for_each_entry(dev, head, index_hlist) { 1687 if (link_dump_filtered(dev, master_idx, kind_ops)) 1688 goto cont; 1689 if (idx < s_idx) 1690 goto cont; 1691 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 1692 NETLINK_CB(cb->skb).portid, 1693 cb->nlh->nlmsg_seq, 0, 1694 flags, 1695 ext_filter_mask, 0); 1696 1697 if (err < 0) { 1698 if (likely(skb->len)) 1699 goto out; 1700 1701 goto out_err; 1702 } 1703 cont: 1704 idx++; 1705 } 1706 } 1707 out: 1708 err = skb->len; 1709 out_err: 1710 cb->args[1] = idx; 1711 cb->args[0] = h; 1712 cb->seq = net->dev_base_seq; 1713 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1714 1715 return err; 1716 } 1717 1718 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len, 1719 struct netlink_ext_ack *exterr) 1720 { 1721 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr); 1722 } 1723 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1724 1725 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1726 { 1727 struct net *net; 1728 /* Examine the link attributes and figure out which 1729 * network namespace we are talking about. 1730 */ 1731 if (tb[IFLA_NET_NS_PID]) 1732 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1733 else if (tb[IFLA_NET_NS_FD]) 1734 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1735 else 1736 net = get_net(src_net); 1737 return net; 1738 } 1739 EXPORT_SYMBOL(rtnl_link_get_net); 1740 1741 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1742 { 1743 if (dev) { 1744 if (tb[IFLA_ADDRESS] && 1745 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1746 return -EINVAL; 1747 1748 if (tb[IFLA_BROADCAST] && 1749 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1750 return -EINVAL; 1751 } 1752 1753 if (tb[IFLA_AF_SPEC]) { 1754 struct nlattr *af; 1755 int rem, err; 1756 1757 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1758 const struct rtnl_af_ops *af_ops; 1759 1760 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1761 return -EAFNOSUPPORT; 1762 1763 if (!af_ops->set_link_af) 1764 return -EOPNOTSUPP; 1765 1766 if (af_ops->validate_link_af) { 1767 err = af_ops->validate_link_af(dev, af); 1768 if (err < 0) 1769 return err; 1770 } 1771 } 1772 } 1773 1774 return 0; 1775 } 1776 1777 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 1778 int guid_type) 1779 { 1780 const struct net_device_ops *ops = dev->netdev_ops; 1781 1782 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 1783 } 1784 1785 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 1786 { 1787 if (dev->type != ARPHRD_INFINIBAND) 1788 return -EOPNOTSUPP; 1789 1790 return handle_infiniband_guid(dev, ivt, guid_type); 1791 } 1792 1793 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 1794 { 1795 const struct net_device_ops *ops = dev->netdev_ops; 1796 int err = -EINVAL; 1797 1798 if (tb[IFLA_VF_MAC]) { 1799 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 1800 1801 err = -EOPNOTSUPP; 1802 if (ops->ndo_set_vf_mac) 1803 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1804 ivm->mac); 1805 if (err < 0) 1806 return err; 1807 } 1808 1809 if (tb[IFLA_VF_VLAN]) { 1810 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 1811 1812 err = -EOPNOTSUPP; 1813 if (ops->ndo_set_vf_vlan) 1814 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 1815 ivv->qos, 1816 htons(ETH_P_8021Q)); 1817 if (err < 0) 1818 return err; 1819 } 1820 1821 if (tb[IFLA_VF_VLAN_LIST]) { 1822 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 1823 struct nlattr *attr; 1824 int rem, len = 0; 1825 1826 err = -EOPNOTSUPP; 1827 if (!ops->ndo_set_vf_vlan) 1828 return err; 1829 1830 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 1831 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 1832 nla_len(attr) < NLA_HDRLEN) { 1833 return -EINVAL; 1834 } 1835 if (len >= MAX_VLAN_LIST_LEN) 1836 return -EOPNOTSUPP; 1837 ivvl[len] = nla_data(attr); 1838 1839 len++; 1840 } 1841 if (len == 0) 1842 return -EINVAL; 1843 1844 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 1845 ivvl[0]->qos, ivvl[0]->vlan_proto); 1846 if (err < 0) 1847 return err; 1848 } 1849 1850 if (tb[IFLA_VF_TX_RATE]) { 1851 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 1852 struct ifla_vf_info ivf; 1853 1854 err = -EOPNOTSUPP; 1855 if (ops->ndo_get_vf_config) 1856 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 1857 if (err < 0) 1858 return err; 1859 1860 err = -EOPNOTSUPP; 1861 if (ops->ndo_set_vf_rate) 1862 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1863 ivf.min_tx_rate, 1864 ivt->rate); 1865 if (err < 0) 1866 return err; 1867 } 1868 1869 if (tb[IFLA_VF_RATE]) { 1870 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 1871 1872 err = -EOPNOTSUPP; 1873 if (ops->ndo_set_vf_rate) 1874 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1875 ivt->min_tx_rate, 1876 ivt->max_tx_rate); 1877 if (err < 0) 1878 return err; 1879 } 1880 1881 if (tb[IFLA_VF_SPOOFCHK]) { 1882 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 1883 1884 err = -EOPNOTSUPP; 1885 if (ops->ndo_set_vf_spoofchk) 1886 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 1887 ivs->setting); 1888 if (err < 0) 1889 return err; 1890 } 1891 1892 if (tb[IFLA_VF_LINK_STATE]) { 1893 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 1894 1895 err = -EOPNOTSUPP; 1896 if (ops->ndo_set_vf_link_state) 1897 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 1898 ivl->link_state); 1899 if (err < 0) 1900 return err; 1901 } 1902 1903 if (tb[IFLA_VF_RSS_QUERY_EN]) { 1904 struct ifla_vf_rss_query_en *ivrssq_en; 1905 1906 err = -EOPNOTSUPP; 1907 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 1908 if (ops->ndo_set_vf_rss_query_en) 1909 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 1910 ivrssq_en->setting); 1911 if (err < 0) 1912 return err; 1913 } 1914 1915 if (tb[IFLA_VF_TRUST]) { 1916 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 1917 1918 err = -EOPNOTSUPP; 1919 if (ops->ndo_set_vf_trust) 1920 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 1921 if (err < 0) 1922 return err; 1923 } 1924 1925 if (tb[IFLA_VF_IB_NODE_GUID]) { 1926 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 1927 1928 if (!ops->ndo_set_vf_guid) 1929 return -EOPNOTSUPP; 1930 1931 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 1932 } 1933 1934 if (tb[IFLA_VF_IB_PORT_GUID]) { 1935 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 1936 1937 if (!ops->ndo_set_vf_guid) 1938 return -EOPNOTSUPP; 1939 1940 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 1941 } 1942 1943 return err; 1944 } 1945 1946 static int do_set_master(struct net_device *dev, int ifindex) 1947 { 1948 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1949 const struct net_device_ops *ops; 1950 int err; 1951 1952 if (upper_dev) { 1953 if (upper_dev->ifindex == ifindex) 1954 return 0; 1955 ops = upper_dev->netdev_ops; 1956 if (ops->ndo_del_slave) { 1957 err = ops->ndo_del_slave(upper_dev, dev); 1958 if (err) 1959 return err; 1960 } else { 1961 return -EOPNOTSUPP; 1962 } 1963 } 1964 1965 if (ifindex) { 1966 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 1967 if (!upper_dev) 1968 return -EINVAL; 1969 ops = upper_dev->netdev_ops; 1970 if (ops->ndo_add_slave) { 1971 err = ops->ndo_add_slave(upper_dev, dev); 1972 if (err) 1973 return err; 1974 } else { 1975 return -EOPNOTSUPP; 1976 } 1977 } 1978 return 0; 1979 } 1980 1981 #define DO_SETLINK_MODIFIED 0x01 1982 /* notify flag means notify + modified. */ 1983 #define DO_SETLINK_NOTIFY 0x03 1984 static int do_setlink(const struct sk_buff *skb, 1985 struct net_device *dev, struct ifinfomsg *ifm, 1986 struct netlink_ext_ack *extack, 1987 struct nlattr **tb, char *ifname, int status) 1988 { 1989 const struct net_device_ops *ops = dev->netdev_ops; 1990 int err; 1991 1992 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) { 1993 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1994 if (IS_ERR(net)) { 1995 err = PTR_ERR(net); 1996 goto errout; 1997 } 1998 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { 1999 put_net(net); 2000 err = -EPERM; 2001 goto errout; 2002 } 2003 err = dev_change_net_namespace(dev, net, ifname); 2004 put_net(net); 2005 if (err) 2006 goto errout; 2007 status |= DO_SETLINK_MODIFIED; 2008 } 2009 2010 if (tb[IFLA_MAP]) { 2011 struct rtnl_link_ifmap *u_map; 2012 struct ifmap k_map; 2013 2014 if (!ops->ndo_set_config) { 2015 err = -EOPNOTSUPP; 2016 goto errout; 2017 } 2018 2019 if (!netif_device_present(dev)) { 2020 err = -ENODEV; 2021 goto errout; 2022 } 2023 2024 u_map = nla_data(tb[IFLA_MAP]); 2025 k_map.mem_start = (unsigned long) u_map->mem_start; 2026 k_map.mem_end = (unsigned long) u_map->mem_end; 2027 k_map.base_addr = (unsigned short) u_map->base_addr; 2028 k_map.irq = (unsigned char) u_map->irq; 2029 k_map.dma = (unsigned char) u_map->dma; 2030 k_map.port = (unsigned char) u_map->port; 2031 2032 err = ops->ndo_set_config(dev, &k_map); 2033 if (err < 0) 2034 goto errout; 2035 2036 status |= DO_SETLINK_NOTIFY; 2037 } 2038 2039 if (tb[IFLA_ADDRESS]) { 2040 struct sockaddr *sa; 2041 int len; 2042 2043 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len, 2044 sizeof(*sa)); 2045 sa = kmalloc(len, GFP_KERNEL); 2046 if (!sa) { 2047 err = -ENOMEM; 2048 goto errout; 2049 } 2050 sa->sa_family = dev->type; 2051 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 2052 dev->addr_len); 2053 err = dev_set_mac_address(dev, sa); 2054 kfree(sa); 2055 if (err) 2056 goto errout; 2057 status |= DO_SETLINK_MODIFIED; 2058 } 2059 2060 if (tb[IFLA_MTU]) { 2061 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2062 if (err < 0) 2063 goto errout; 2064 status |= DO_SETLINK_MODIFIED; 2065 } 2066 2067 if (tb[IFLA_GROUP]) { 2068 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2069 status |= DO_SETLINK_NOTIFY; 2070 } 2071 2072 /* 2073 * Interface selected by interface index but interface 2074 * name provided implies that a name change has been 2075 * requested. 2076 */ 2077 if (ifm->ifi_index > 0 && ifname[0]) { 2078 err = dev_change_name(dev, ifname); 2079 if (err < 0) 2080 goto errout; 2081 status |= DO_SETLINK_MODIFIED; 2082 } 2083 2084 if (tb[IFLA_IFALIAS]) { 2085 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2086 nla_len(tb[IFLA_IFALIAS])); 2087 if (err < 0) 2088 goto errout; 2089 status |= DO_SETLINK_NOTIFY; 2090 } 2091 2092 if (tb[IFLA_BROADCAST]) { 2093 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2094 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2095 } 2096 2097 if (ifm->ifi_flags || ifm->ifi_change) { 2098 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2099 if (err < 0) 2100 goto errout; 2101 } 2102 2103 if (tb[IFLA_MASTER]) { 2104 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2105 if (err) 2106 goto errout; 2107 status |= DO_SETLINK_MODIFIED; 2108 } 2109 2110 if (tb[IFLA_CARRIER]) { 2111 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2112 if (err) 2113 goto errout; 2114 status |= DO_SETLINK_MODIFIED; 2115 } 2116 2117 if (tb[IFLA_TXQLEN]) { 2118 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]); 2119 unsigned int orig_len = dev->tx_queue_len; 2120 2121 if (dev->tx_queue_len ^ value) { 2122 dev->tx_queue_len = value; 2123 err = call_netdevice_notifiers( 2124 NETDEV_CHANGE_TX_QUEUE_LEN, dev); 2125 err = notifier_to_errno(err); 2126 if (err) { 2127 dev->tx_queue_len = orig_len; 2128 goto errout; 2129 } 2130 status |= DO_SETLINK_NOTIFY; 2131 } 2132 } 2133 2134 if (tb[IFLA_OPERSTATE]) 2135 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2136 2137 if (tb[IFLA_LINKMODE]) { 2138 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2139 2140 write_lock_bh(&dev_base_lock); 2141 if (dev->link_mode ^ value) 2142 status |= DO_SETLINK_NOTIFY; 2143 dev->link_mode = value; 2144 write_unlock_bh(&dev_base_lock); 2145 } 2146 2147 if (tb[IFLA_VFINFO_LIST]) { 2148 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2149 struct nlattr *attr; 2150 int rem; 2151 2152 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2153 if (nla_type(attr) != IFLA_VF_INFO || 2154 nla_len(attr) < NLA_HDRLEN) { 2155 err = -EINVAL; 2156 goto errout; 2157 } 2158 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr, 2159 ifla_vf_policy, NULL); 2160 if (err < 0) 2161 goto errout; 2162 err = do_setvfinfo(dev, vfinfo); 2163 if (err < 0) 2164 goto errout; 2165 status |= DO_SETLINK_NOTIFY; 2166 } 2167 } 2168 err = 0; 2169 2170 if (tb[IFLA_VF_PORTS]) { 2171 struct nlattr *port[IFLA_PORT_MAX+1]; 2172 struct nlattr *attr; 2173 int vf; 2174 int rem; 2175 2176 err = -EOPNOTSUPP; 2177 if (!ops->ndo_set_vf_port) 2178 goto errout; 2179 2180 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2181 if (nla_type(attr) != IFLA_VF_PORT || 2182 nla_len(attr) < NLA_HDRLEN) { 2183 err = -EINVAL; 2184 goto errout; 2185 } 2186 err = nla_parse_nested(port, IFLA_PORT_MAX, attr, 2187 ifla_port_policy, NULL); 2188 if (err < 0) 2189 goto errout; 2190 if (!port[IFLA_PORT_VF]) { 2191 err = -EOPNOTSUPP; 2192 goto errout; 2193 } 2194 vf = nla_get_u32(port[IFLA_PORT_VF]); 2195 err = ops->ndo_set_vf_port(dev, vf, port); 2196 if (err < 0) 2197 goto errout; 2198 status |= DO_SETLINK_NOTIFY; 2199 } 2200 } 2201 err = 0; 2202 2203 if (tb[IFLA_PORT_SELF]) { 2204 struct nlattr *port[IFLA_PORT_MAX+1]; 2205 2206 err = nla_parse_nested(port, IFLA_PORT_MAX, 2207 tb[IFLA_PORT_SELF], ifla_port_policy, 2208 NULL); 2209 if (err < 0) 2210 goto errout; 2211 2212 err = -EOPNOTSUPP; 2213 if (ops->ndo_set_vf_port) 2214 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2215 if (err < 0) 2216 goto errout; 2217 status |= DO_SETLINK_NOTIFY; 2218 } 2219 2220 if (tb[IFLA_AF_SPEC]) { 2221 struct nlattr *af; 2222 int rem; 2223 2224 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2225 const struct rtnl_af_ops *af_ops; 2226 2227 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 2228 BUG(); 2229 2230 err = af_ops->set_link_af(dev, af); 2231 if (err < 0) 2232 goto errout; 2233 2234 status |= DO_SETLINK_NOTIFY; 2235 } 2236 } 2237 err = 0; 2238 2239 if (tb[IFLA_PROTO_DOWN]) { 2240 err = dev_change_proto_down(dev, 2241 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2242 if (err) 2243 goto errout; 2244 status |= DO_SETLINK_NOTIFY; 2245 } 2246 2247 if (tb[IFLA_XDP]) { 2248 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2249 u32 xdp_flags = 0; 2250 2251 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP], 2252 ifla_xdp_policy, NULL); 2253 if (err < 0) 2254 goto errout; 2255 2256 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) { 2257 err = -EINVAL; 2258 goto errout; 2259 } 2260 2261 if (xdp[IFLA_XDP_FLAGS]) { 2262 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2263 if (xdp_flags & ~XDP_FLAGS_MASK) { 2264 err = -EINVAL; 2265 goto errout; 2266 } 2267 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) { 2268 err = -EINVAL; 2269 goto errout; 2270 } 2271 } 2272 2273 if (xdp[IFLA_XDP_FD]) { 2274 err = dev_change_xdp_fd(dev, extack, 2275 nla_get_s32(xdp[IFLA_XDP_FD]), 2276 xdp_flags); 2277 if (err) 2278 goto errout; 2279 status |= DO_SETLINK_NOTIFY; 2280 } 2281 } 2282 2283 errout: 2284 if (status & DO_SETLINK_MODIFIED) { 2285 if (status & DO_SETLINK_NOTIFY) 2286 netdev_state_change(dev); 2287 2288 if (err < 0) 2289 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", 2290 dev->name); 2291 } 2292 2293 return err; 2294 } 2295 2296 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2297 struct netlink_ext_ack *extack) 2298 { 2299 struct net *net = sock_net(skb->sk); 2300 struct ifinfomsg *ifm; 2301 struct net_device *dev; 2302 int err; 2303 struct nlattr *tb[IFLA_MAX+1]; 2304 char ifname[IFNAMSIZ]; 2305 2306 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, 2307 extack); 2308 if (err < 0) 2309 goto errout; 2310 2311 if (tb[IFLA_IFNAME]) 2312 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2313 else 2314 ifname[0] = '\0'; 2315 2316 err = -EINVAL; 2317 ifm = nlmsg_data(nlh); 2318 if (ifm->ifi_index > 0) 2319 dev = __dev_get_by_index(net, ifm->ifi_index); 2320 else if (tb[IFLA_IFNAME]) 2321 dev = __dev_get_by_name(net, ifname); 2322 else 2323 goto errout; 2324 2325 if (dev == NULL) { 2326 err = -ENODEV; 2327 goto errout; 2328 } 2329 2330 err = validate_linkmsg(dev, tb); 2331 if (err < 0) 2332 goto errout; 2333 2334 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0); 2335 errout: 2336 return err; 2337 } 2338 2339 static int rtnl_group_dellink(const struct net *net, int group) 2340 { 2341 struct net_device *dev, *aux; 2342 LIST_HEAD(list_kill); 2343 bool found = false; 2344 2345 if (!group) 2346 return -EPERM; 2347 2348 for_each_netdev(net, dev) { 2349 if (dev->group == group) { 2350 const struct rtnl_link_ops *ops; 2351 2352 found = true; 2353 ops = dev->rtnl_link_ops; 2354 if (!ops || !ops->dellink) 2355 return -EOPNOTSUPP; 2356 } 2357 } 2358 2359 if (!found) 2360 return -ENODEV; 2361 2362 for_each_netdev_safe(net, dev, aux) { 2363 if (dev->group == group) { 2364 const struct rtnl_link_ops *ops; 2365 2366 ops = dev->rtnl_link_ops; 2367 ops->dellink(dev, &list_kill); 2368 } 2369 } 2370 unregister_netdevice_many(&list_kill); 2371 2372 return 0; 2373 } 2374 2375 int rtnl_delete_link(struct net_device *dev) 2376 { 2377 const struct rtnl_link_ops *ops; 2378 LIST_HEAD(list_kill); 2379 2380 ops = dev->rtnl_link_ops; 2381 if (!ops || !ops->dellink) 2382 return -EOPNOTSUPP; 2383 2384 ops->dellink(dev, &list_kill); 2385 unregister_netdevice_many(&list_kill); 2386 2387 return 0; 2388 } 2389 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2390 2391 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 2392 struct netlink_ext_ack *extack) 2393 { 2394 struct net *net = sock_net(skb->sk); 2395 struct net_device *dev; 2396 struct ifinfomsg *ifm; 2397 char ifname[IFNAMSIZ]; 2398 struct nlattr *tb[IFLA_MAX+1]; 2399 int err; 2400 2401 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2402 if (err < 0) 2403 return err; 2404 2405 if (tb[IFLA_IFNAME]) 2406 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2407 2408 ifm = nlmsg_data(nlh); 2409 if (ifm->ifi_index > 0) 2410 dev = __dev_get_by_index(net, ifm->ifi_index); 2411 else if (tb[IFLA_IFNAME]) 2412 dev = __dev_get_by_name(net, ifname); 2413 else if (tb[IFLA_GROUP]) 2414 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP])); 2415 else 2416 return -EINVAL; 2417 2418 if (!dev) 2419 return -ENODEV; 2420 2421 return rtnl_delete_link(dev); 2422 } 2423 2424 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2425 { 2426 unsigned int old_flags; 2427 int err; 2428 2429 old_flags = dev->flags; 2430 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2431 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2432 if (err < 0) 2433 return err; 2434 } 2435 2436 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2437 2438 __dev_notify_flags(dev, old_flags, ~0U); 2439 return 0; 2440 } 2441 EXPORT_SYMBOL(rtnl_configure_link); 2442 2443 struct net_device *rtnl_create_link(struct net *net, 2444 const char *ifname, unsigned char name_assign_type, 2445 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 2446 { 2447 struct net_device *dev; 2448 unsigned int num_tx_queues = 1; 2449 unsigned int num_rx_queues = 1; 2450 2451 if (tb[IFLA_NUM_TX_QUEUES]) 2452 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2453 else if (ops->get_num_tx_queues) 2454 num_tx_queues = ops->get_num_tx_queues(); 2455 2456 if (tb[IFLA_NUM_RX_QUEUES]) 2457 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2458 else if (ops->get_num_rx_queues) 2459 num_rx_queues = ops->get_num_rx_queues(); 2460 2461 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2462 ops->setup, num_tx_queues, num_rx_queues); 2463 if (!dev) 2464 return ERR_PTR(-ENOMEM); 2465 2466 dev_net_set(dev, net); 2467 dev->rtnl_link_ops = ops; 2468 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2469 2470 if (tb[IFLA_MTU]) 2471 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2472 if (tb[IFLA_ADDRESS]) { 2473 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2474 nla_len(tb[IFLA_ADDRESS])); 2475 dev->addr_assign_type = NET_ADDR_SET; 2476 } 2477 if (tb[IFLA_BROADCAST]) 2478 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2479 nla_len(tb[IFLA_BROADCAST])); 2480 if (tb[IFLA_TXQLEN]) 2481 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2482 if (tb[IFLA_OPERSTATE]) 2483 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2484 if (tb[IFLA_LINKMODE]) 2485 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2486 if (tb[IFLA_GROUP]) 2487 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2488 2489 return dev; 2490 } 2491 EXPORT_SYMBOL(rtnl_create_link); 2492 2493 static int rtnl_group_changelink(const struct sk_buff *skb, 2494 struct net *net, int group, 2495 struct ifinfomsg *ifm, 2496 struct netlink_ext_ack *extack, 2497 struct nlattr **tb) 2498 { 2499 struct net_device *dev, *aux; 2500 int err; 2501 2502 for_each_netdev_safe(net, dev, aux) { 2503 if (dev->group == group) { 2504 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0); 2505 if (err < 0) 2506 return err; 2507 } 2508 } 2509 2510 return 0; 2511 } 2512 2513 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2514 struct netlink_ext_ack *extack) 2515 { 2516 struct net *net = sock_net(skb->sk); 2517 const struct rtnl_link_ops *ops; 2518 const struct rtnl_link_ops *m_ops = NULL; 2519 struct net_device *dev; 2520 struct net_device *master_dev = NULL; 2521 struct ifinfomsg *ifm; 2522 char kind[MODULE_NAME_LEN]; 2523 char ifname[IFNAMSIZ]; 2524 struct nlattr *tb[IFLA_MAX+1]; 2525 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 2526 unsigned char name_assign_type = NET_NAME_USER; 2527 int err; 2528 2529 #ifdef CONFIG_MODULES 2530 replay: 2531 #endif 2532 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2533 if (err < 0) 2534 return err; 2535 2536 if (tb[IFLA_IFNAME]) 2537 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2538 else 2539 ifname[0] = '\0'; 2540 2541 ifm = nlmsg_data(nlh); 2542 if (ifm->ifi_index > 0) 2543 dev = __dev_get_by_index(net, ifm->ifi_index); 2544 else { 2545 if (ifname[0]) 2546 dev = __dev_get_by_name(net, ifname); 2547 else 2548 dev = NULL; 2549 } 2550 2551 if (dev) { 2552 master_dev = netdev_master_upper_dev_get(dev); 2553 if (master_dev) 2554 m_ops = master_dev->rtnl_link_ops; 2555 } 2556 2557 err = validate_linkmsg(dev, tb); 2558 if (err < 0) 2559 return err; 2560 2561 if (tb[IFLA_LINKINFO]) { 2562 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 2563 tb[IFLA_LINKINFO], ifla_info_policy, 2564 NULL); 2565 if (err < 0) 2566 return err; 2567 } else 2568 memset(linkinfo, 0, sizeof(linkinfo)); 2569 2570 if (linkinfo[IFLA_INFO_KIND]) { 2571 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 2572 ops = rtnl_link_ops_get(kind); 2573 } else { 2574 kind[0] = '\0'; 2575 ops = NULL; 2576 } 2577 2578 if (1) { 2579 struct nlattr *attr[ops ? ops->maxtype + 1 : 1]; 2580 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1]; 2581 struct nlattr **data = NULL; 2582 struct nlattr **slave_data = NULL; 2583 struct net *dest_net, *link_net = NULL; 2584 2585 if (ops) { 2586 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 2587 err = nla_parse_nested(attr, ops->maxtype, 2588 linkinfo[IFLA_INFO_DATA], 2589 ops->policy, NULL); 2590 if (err < 0) 2591 return err; 2592 data = attr; 2593 } 2594 if (ops->validate) { 2595 err = ops->validate(tb, data, extack); 2596 if (err < 0) 2597 return err; 2598 } 2599 } 2600 2601 if (m_ops) { 2602 if (m_ops->slave_maxtype && 2603 linkinfo[IFLA_INFO_SLAVE_DATA]) { 2604 err = nla_parse_nested(slave_attr, 2605 m_ops->slave_maxtype, 2606 linkinfo[IFLA_INFO_SLAVE_DATA], 2607 m_ops->slave_policy, 2608 NULL); 2609 if (err < 0) 2610 return err; 2611 slave_data = slave_attr; 2612 } 2613 if (m_ops->slave_validate) { 2614 err = m_ops->slave_validate(tb, slave_data, 2615 extack); 2616 if (err < 0) 2617 return err; 2618 } 2619 } 2620 2621 if (dev) { 2622 int status = 0; 2623 2624 if (nlh->nlmsg_flags & NLM_F_EXCL) 2625 return -EEXIST; 2626 if (nlh->nlmsg_flags & NLM_F_REPLACE) 2627 return -EOPNOTSUPP; 2628 2629 if (linkinfo[IFLA_INFO_DATA]) { 2630 if (!ops || ops != dev->rtnl_link_ops || 2631 !ops->changelink) 2632 return -EOPNOTSUPP; 2633 2634 err = ops->changelink(dev, tb, data, extack); 2635 if (err < 0) 2636 return err; 2637 status |= DO_SETLINK_NOTIFY; 2638 } 2639 2640 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 2641 if (!m_ops || !m_ops->slave_changelink) 2642 return -EOPNOTSUPP; 2643 2644 err = m_ops->slave_changelink(master_dev, dev, 2645 tb, slave_data, 2646 extack); 2647 if (err < 0) 2648 return err; 2649 status |= DO_SETLINK_NOTIFY; 2650 } 2651 2652 return do_setlink(skb, dev, ifm, extack, tb, ifname, 2653 status); 2654 } 2655 2656 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 2657 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 2658 return rtnl_group_changelink(skb, net, 2659 nla_get_u32(tb[IFLA_GROUP]), 2660 ifm, extack, tb); 2661 return -ENODEV; 2662 } 2663 2664 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 2665 return -EOPNOTSUPP; 2666 2667 if (!ops) { 2668 #ifdef CONFIG_MODULES 2669 if (kind[0]) { 2670 __rtnl_unlock(); 2671 request_module("rtnl-link-%s", kind); 2672 rtnl_lock(); 2673 ops = rtnl_link_ops_get(kind); 2674 if (ops) 2675 goto replay; 2676 } 2677 #endif 2678 return -EOPNOTSUPP; 2679 } 2680 2681 if (!ops->setup) 2682 return -EOPNOTSUPP; 2683 2684 if (!ifname[0]) { 2685 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 2686 name_assign_type = NET_NAME_ENUM; 2687 } 2688 2689 dest_net = rtnl_link_get_net(net, tb); 2690 if (IS_ERR(dest_net)) 2691 return PTR_ERR(dest_net); 2692 2693 err = -EPERM; 2694 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN)) 2695 goto out; 2696 2697 if (tb[IFLA_LINK_NETNSID]) { 2698 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 2699 2700 link_net = get_net_ns_by_id(dest_net, id); 2701 if (!link_net) { 2702 err = -EINVAL; 2703 goto out; 2704 } 2705 err = -EPERM; 2706 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 2707 goto out; 2708 } 2709 2710 dev = rtnl_create_link(link_net ? : dest_net, ifname, 2711 name_assign_type, ops, tb); 2712 if (IS_ERR(dev)) { 2713 err = PTR_ERR(dev); 2714 goto out; 2715 } 2716 2717 dev->ifindex = ifm->ifi_index; 2718 2719 if (ops->newlink) { 2720 err = ops->newlink(link_net ? : net, dev, tb, data, 2721 extack); 2722 /* Drivers should call free_netdev() in ->destructor 2723 * and unregister it on failure after registration 2724 * so that device could be finally freed in rtnl_unlock. 2725 */ 2726 if (err < 0) { 2727 /* If device is not registered at all, free it now */ 2728 if (dev->reg_state == NETREG_UNINITIALIZED) 2729 free_netdev(dev); 2730 goto out; 2731 } 2732 } else { 2733 err = register_netdevice(dev); 2734 if (err < 0) { 2735 free_netdev(dev); 2736 goto out; 2737 } 2738 } 2739 err = rtnl_configure_link(dev, ifm); 2740 if (err < 0) 2741 goto out_unregister; 2742 if (link_net) { 2743 err = dev_change_net_namespace(dev, dest_net, ifname); 2744 if (err < 0) 2745 goto out_unregister; 2746 } 2747 if (tb[IFLA_MASTER]) { 2748 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2749 if (err) 2750 goto out_unregister; 2751 } 2752 out: 2753 if (link_net) 2754 put_net(link_net); 2755 put_net(dest_net); 2756 return err; 2757 out_unregister: 2758 if (ops->newlink) { 2759 LIST_HEAD(list_kill); 2760 2761 ops->dellink(dev, &list_kill); 2762 unregister_netdevice_many(&list_kill); 2763 } else { 2764 unregister_netdevice(dev); 2765 } 2766 goto out; 2767 } 2768 } 2769 2770 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2771 struct netlink_ext_ack *extack) 2772 { 2773 struct net *net = sock_net(skb->sk); 2774 struct ifinfomsg *ifm; 2775 char ifname[IFNAMSIZ]; 2776 struct nlattr *tb[IFLA_MAX+1]; 2777 struct net_device *dev = NULL; 2778 struct sk_buff *nskb; 2779 int err; 2780 u32 ext_filter_mask = 0; 2781 2782 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2783 if (err < 0) 2784 return err; 2785 2786 if (tb[IFLA_IFNAME]) 2787 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2788 2789 if (tb[IFLA_EXT_MASK]) 2790 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2791 2792 ifm = nlmsg_data(nlh); 2793 if (ifm->ifi_index > 0) 2794 dev = __dev_get_by_index(net, ifm->ifi_index); 2795 else if (tb[IFLA_IFNAME]) 2796 dev = __dev_get_by_name(net, ifname); 2797 else 2798 return -EINVAL; 2799 2800 if (dev == NULL) 2801 return -ENODEV; 2802 2803 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 2804 if (nskb == NULL) 2805 return -ENOBUFS; 2806 2807 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid, 2808 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0); 2809 if (err < 0) { 2810 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 2811 WARN_ON(err == -EMSGSIZE); 2812 kfree_skb(nskb); 2813 } else 2814 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 2815 2816 return err; 2817 } 2818 2819 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 2820 { 2821 struct net *net = sock_net(skb->sk); 2822 struct net_device *dev; 2823 struct nlattr *tb[IFLA_MAX+1]; 2824 u32 ext_filter_mask = 0; 2825 u16 min_ifinfo_dump_size = 0; 2826 int hdrlen; 2827 2828 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 2829 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 2830 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 2831 2832 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) { 2833 if (tb[IFLA_EXT_MASK]) 2834 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2835 } 2836 2837 if (!ext_filter_mask) 2838 return NLMSG_GOODSIZE; 2839 /* 2840 * traverse the list of net devices and compute the minimum 2841 * buffer size based upon the filter mask. 2842 */ 2843 rcu_read_lock(); 2844 for_each_netdev_rcu(net, dev) { 2845 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 2846 if_nlmsg_size(dev, 2847 ext_filter_mask)); 2848 } 2849 rcu_read_unlock(); 2850 2851 return nlmsg_total_size(min_ifinfo_dump_size); 2852 } 2853 2854 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 2855 { 2856 int idx; 2857 int s_idx = cb->family; 2858 2859 if (s_idx == 0) 2860 s_idx = 1; 2861 2862 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 2863 int type = cb->nlh->nlmsg_type-RTM_BASE; 2864 struct rtnl_link *handlers; 2865 rtnl_dumpit_func dumpit; 2866 2867 if (idx < s_idx || idx == PF_PACKET) 2868 continue; 2869 2870 handlers = rtnl_dereference(rtnl_msg_handlers[idx]); 2871 if (!handlers) 2872 continue; 2873 2874 dumpit = READ_ONCE(handlers[type].dumpit); 2875 if (!dumpit) 2876 continue; 2877 2878 if (idx > s_idx) { 2879 memset(&cb->args[0], 0, sizeof(cb->args)); 2880 cb->prev_seq = 0; 2881 cb->seq = 0; 2882 } 2883 if (dumpit(skb, cb)) 2884 break; 2885 } 2886 cb->family = idx; 2887 2888 return skb->len; 2889 } 2890 2891 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 2892 unsigned int change, 2893 u32 event, gfp_t flags) 2894 { 2895 struct net *net = dev_net(dev); 2896 struct sk_buff *skb; 2897 int err = -ENOBUFS; 2898 size_t if_info_size; 2899 2900 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 2901 if (skb == NULL) 2902 goto errout; 2903 2904 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event); 2905 if (err < 0) { 2906 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 2907 WARN_ON(err == -EMSGSIZE); 2908 kfree_skb(skb); 2909 goto errout; 2910 } 2911 return skb; 2912 errout: 2913 if (err < 0) 2914 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2915 return NULL; 2916 } 2917 2918 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 2919 { 2920 struct net *net = dev_net(dev); 2921 2922 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 2923 } 2924 2925 static void rtmsg_ifinfo_event(int type, struct net_device *dev, 2926 unsigned int change, u32 event, 2927 gfp_t flags) 2928 { 2929 struct sk_buff *skb; 2930 2931 if (dev->reg_state != NETREG_REGISTERED) 2932 return; 2933 2934 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags); 2935 if (skb) 2936 rtmsg_ifinfo_send(skb, dev, flags); 2937 } 2938 2939 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 2940 gfp_t flags) 2941 { 2942 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags); 2943 } 2944 EXPORT_SYMBOL(rtmsg_ifinfo); 2945 2946 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 2947 struct net_device *dev, 2948 u8 *addr, u16 vid, u32 pid, u32 seq, 2949 int type, unsigned int flags, 2950 int nlflags, u16 ndm_state) 2951 { 2952 struct nlmsghdr *nlh; 2953 struct ndmsg *ndm; 2954 2955 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 2956 if (!nlh) 2957 return -EMSGSIZE; 2958 2959 ndm = nlmsg_data(nlh); 2960 ndm->ndm_family = AF_BRIDGE; 2961 ndm->ndm_pad1 = 0; 2962 ndm->ndm_pad2 = 0; 2963 ndm->ndm_flags = flags; 2964 ndm->ndm_type = 0; 2965 ndm->ndm_ifindex = dev->ifindex; 2966 ndm->ndm_state = ndm_state; 2967 2968 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 2969 goto nla_put_failure; 2970 if (vid) 2971 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 2972 goto nla_put_failure; 2973 2974 nlmsg_end(skb, nlh); 2975 return 0; 2976 2977 nla_put_failure: 2978 nlmsg_cancel(skb, nlh); 2979 return -EMSGSIZE; 2980 } 2981 2982 static inline size_t rtnl_fdb_nlmsg_size(void) 2983 { 2984 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 2985 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 2986 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 2987 0; 2988 } 2989 2990 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 2991 u16 ndm_state) 2992 { 2993 struct net *net = dev_net(dev); 2994 struct sk_buff *skb; 2995 int err = -ENOBUFS; 2996 2997 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 2998 if (!skb) 2999 goto errout; 3000 3001 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 3002 0, 0, type, NTF_SELF, 0, ndm_state); 3003 if (err < 0) { 3004 kfree_skb(skb); 3005 goto errout; 3006 } 3007 3008 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 3009 return; 3010 errout: 3011 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 3012 } 3013 3014 /** 3015 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 3016 */ 3017 int ndo_dflt_fdb_add(struct ndmsg *ndm, 3018 struct nlattr *tb[], 3019 struct net_device *dev, 3020 const unsigned char *addr, u16 vid, 3021 u16 flags) 3022 { 3023 int err = -EINVAL; 3024 3025 /* If aging addresses are supported device will need to 3026 * implement its own handler for this. 3027 */ 3028 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 3029 pr_info("%s: FDB only supports static addresses\n", dev->name); 3030 return err; 3031 } 3032 3033 if (vid) { 3034 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 3035 return err; 3036 } 3037 3038 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3039 err = dev_uc_add_excl(dev, addr); 3040 else if (is_multicast_ether_addr(addr)) 3041 err = dev_mc_add_excl(dev, addr); 3042 3043 /* Only return duplicate errors if NLM_F_EXCL is set */ 3044 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 3045 err = 0; 3046 3047 return err; 3048 } 3049 EXPORT_SYMBOL(ndo_dflt_fdb_add); 3050 3051 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid) 3052 { 3053 u16 vid = 0; 3054 3055 if (vlan_attr) { 3056 if (nla_len(vlan_attr) != sizeof(u16)) { 3057 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n"); 3058 return -EINVAL; 3059 } 3060 3061 vid = nla_get_u16(vlan_attr); 3062 3063 if (!vid || vid >= VLAN_VID_MASK) { 3064 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n", 3065 vid); 3066 return -EINVAL; 3067 } 3068 } 3069 *p_vid = vid; 3070 return 0; 3071 } 3072 3073 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, 3074 struct netlink_ext_ack *extack) 3075 { 3076 struct net *net = sock_net(skb->sk); 3077 struct ndmsg *ndm; 3078 struct nlattr *tb[NDA_MAX+1]; 3079 struct net_device *dev; 3080 u8 *addr; 3081 u16 vid; 3082 int err; 3083 3084 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3085 if (err < 0) 3086 return err; 3087 3088 ndm = nlmsg_data(nlh); 3089 if (ndm->ndm_ifindex == 0) { 3090 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n"); 3091 return -EINVAL; 3092 } 3093 3094 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3095 if (dev == NULL) { 3096 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n"); 3097 return -ENODEV; 3098 } 3099 3100 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3101 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n"); 3102 return -EINVAL; 3103 } 3104 3105 addr = nla_data(tb[NDA_LLADDR]); 3106 3107 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3108 if (err) 3109 return err; 3110 3111 err = -EOPNOTSUPP; 3112 3113 /* Support fdb on master device the net/bridge default case */ 3114 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3115 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3116 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3117 const struct net_device_ops *ops = br_dev->netdev_ops; 3118 3119 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3120 nlh->nlmsg_flags); 3121 if (err) 3122 goto out; 3123 else 3124 ndm->ndm_flags &= ~NTF_MASTER; 3125 } 3126 3127 /* Embedded bridge, macvlan, and any other device support */ 3128 if ((ndm->ndm_flags & NTF_SELF)) { 3129 if (dev->netdev_ops->ndo_fdb_add) 3130 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3131 vid, 3132 nlh->nlmsg_flags); 3133 else 3134 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3135 nlh->nlmsg_flags); 3136 3137 if (!err) { 3138 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3139 ndm->ndm_state); 3140 ndm->ndm_flags &= ~NTF_SELF; 3141 } 3142 } 3143 out: 3144 return err; 3145 } 3146 3147 /** 3148 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3149 */ 3150 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3151 struct nlattr *tb[], 3152 struct net_device *dev, 3153 const unsigned char *addr, u16 vid) 3154 { 3155 int err = -EINVAL; 3156 3157 /* If aging addresses are supported device will need to 3158 * implement its own handler for this. 3159 */ 3160 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3161 pr_info("%s: FDB only supports static addresses\n", dev->name); 3162 return err; 3163 } 3164 3165 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3166 err = dev_uc_del(dev, addr); 3167 else if (is_multicast_ether_addr(addr)) 3168 err = dev_mc_del(dev, addr); 3169 3170 return err; 3171 } 3172 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3173 3174 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, 3175 struct netlink_ext_ack *extack) 3176 { 3177 struct net *net = sock_net(skb->sk); 3178 struct ndmsg *ndm; 3179 struct nlattr *tb[NDA_MAX+1]; 3180 struct net_device *dev; 3181 int err = -EINVAL; 3182 __u8 *addr; 3183 u16 vid; 3184 3185 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3186 return -EPERM; 3187 3188 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3189 if (err < 0) 3190 return err; 3191 3192 ndm = nlmsg_data(nlh); 3193 if (ndm->ndm_ifindex == 0) { 3194 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n"); 3195 return -EINVAL; 3196 } 3197 3198 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3199 if (dev == NULL) { 3200 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n"); 3201 return -ENODEV; 3202 } 3203 3204 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3205 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n"); 3206 return -EINVAL; 3207 } 3208 3209 addr = nla_data(tb[NDA_LLADDR]); 3210 3211 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3212 if (err) 3213 return err; 3214 3215 err = -EOPNOTSUPP; 3216 3217 /* Support fdb on master device the net/bridge default case */ 3218 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3219 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3220 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3221 const struct net_device_ops *ops = br_dev->netdev_ops; 3222 3223 if (ops->ndo_fdb_del) 3224 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3225 3226 if (err) 3227 goto out; 3228 else 3229 ndm->ndm_flags &= ~NTF_MASTER; 3230 } 3231 3232 /* Embedded bridge, macvlan, and any other device support */ 3233 if (ndm->ndm_flags & NTF_SELF) { 3234 if (dev->netdev_ops->ndo_fdb_del) 3235 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3236 vid); 3237 else 3238 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3239 3240 if (!err) { 3241 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3242 ndm->ndm_state); 3243 ndm->ndm_flags &= ~NTF_SELF; 3244 } 3245 } 3246 out: 3247 return err; 3248 } 3249 3250 static int nlmsg_populate_fdb(struct sk_buff *skb, 3251 struct netlink_callback *cb, 3252 struct net_device *dev, 3253 int *idx, 3254 struct netdev_hw_addr_list *list) 3255 { 3256 struct netdev_hw_addr *ha; 3257 int err; 3258 u32 portid, seq; 3259 3260 portid = NETLINK_CB(cb->skb).portid; 3261 seq = cb->nlh->nlmsg_seq; 3262 3263 list_for_each_entry(ha, &list->list, list) { 3264 if (*idx < cb->args[2]) 3265 goto skip; 3266 3267 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3268 portid, seq, 3269 RTM_NEWNEIGH, NTF_SELF, 3270 NLM_F_MULTI, NUD_PERMANENT); 3271 if (err < 0) 3272 return err; 3273 skip: 3274 *idx += 1; 3275 } 3276 return 0; 3277 } 3278 3279 /** 3280 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3281 * @nlh: netlink message header 3282 * @dev: netdevice 3283 * 3284 * Default netdevice operation to dump the existing unicast address list. 3285 * Returns number of addresses from list put in skb. 3286 */ 3287 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3288 struct netlink_callback *cb, 3289 struct net_device *dev, 3290 struct net_device *filter_dev, 3291 int *idx) 3292 { 3293 int err; 3294 3295 netif_addr_lock_bh(dev); 3296 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3297 if (err) 3298 goto out; 3299 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3300 out: 3301 netif_addr_unlock_bh(dev); 3302 return err; 3303 } 3304 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3305 3306 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 3307 { 3308 struct net_device *dev; 3309 struct nlattr *tb[IFLA_MAX+1]; 3310 struct net_device *br_dev = NULL; 3311 const struct net_device_ops *ops = NULL; 3312 const struct net_device_ops *cops = NULL; 3313 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 3314 struct net *net = sock_net(skb->sk); 3315 struct hlist_head *head; 3316 int brport_idx = 0; 3317 int br_idx = 0; 3318 int h, s_h; 3319 int idx = 0, s_idx; 3320 int err = 0; 3321 int fidx = 0; 3322 3323 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, 3324 IFLA_MAX, ifla_policy, NULL); 3325 if (err < 0) { 3326 return -EINVAL; 3327 } else if (err == 0) { 3328 if (tb[IFLA_MASTER]) 3329 br_idx = nla_get_u32(tb[IFLA_MASTER]); 3330 } 3331 3332 brport_idx = ifm->ifi_index; 3333 3334 if (br_idx) { 3335 br_dev = __dev_get_by_index(net, br_idx); 3336 if (!br_dev) 3337 return -ENODEV; 3338 3339 ops = br_dev->netdev_ops; 3340 } 3341 3342 s_h = cb->args[0]; 3343 s_idx = cb->args[1]; 3344 3345 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3346 idx = 0; 3347 head = &net->dev_index_head[h]; 3348 hlist_for_each_entry(dev, head, index_hlist) { 3349 3350 if (brport_idx && (dev->ifindex != brport_idx)) 3351 continue; 3352 3353 if (!br_idx) { /* user did not specify a specific bridge */ 3354 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3355 br_dev = netdev_master_upper_dev_get(dev); 3356 cops = br_dev->netdev_ops; 3357 } 3358 } else { 3359 if (dev != br_dev && 3360 !(dev->priv_flags & IFF_BRIDGE_PORT)) 3361 continue; 3362 3363 if (br_dev != netdev_master_upper_dev_get(dev) && 3364 !(dev->priv_flags & IFF_EBRIDGE)) 3365 continue; 3366 cops = ops; 3367 } 3368 3369 if (idx < s_idx) 3370 goto cont; 3371 3372 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3373 if (cops && cops->ndo_fdb_dump) { 3374 err = cops->ndo_fdb_dump(skb, cb, 3375 br_dev, dev, 3376 &fidx); 3377 if (err == -EMSGSIZE) 3378 goto out; 3379 } 3380 } 3381 3382 if (dev->netdev_ops->ndo_fdb_dump) 3383 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 3384 dev, NULL, 3385 &fidx); 3386 else 3387 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 3388 &fidx); 3389 if (err == -EMSGSIZE) 3390 goto out; 3391 3392 cops = NULL; 3393 3394 /* reset fdb offset to 0 for rest of the interfaces */ 3395 cb->args[2] = 0; 3396 fidx = 0; 3397 cont: 3398 idx++; 3399 } 3400 } 3401 3402 out: 3403 cb->args[0] = h; 3404 cb->args[1] = idx; 3405 cb->args[2] = fidx; 3406 3407 return skb->len; 3408 } 3409 3410 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 3411 unsigned int attrnum, unsigned int flag) 3412 { 3413 if (mask & flag) 3414 return nla_put_u8(skb, attrnum, !!(flags & flag)); 3415 return 0; 3416 } 3417 3418 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 3419 struct net_device *dev, u16 mode, 3420 u32 flags, u32 mask, int nlflags, 3421 u32 filter_mask, 3422 int (*vlan_fill)(struct sk_buff *skb, 3423 struct net_device *dev, 3424 u32 filter_mask)) 3425 { 3426 struct nlmsghdr *nlh; 3427 struct ifinfomsg *ifm; 3428 struct nlattr *br_afspec; 3429 struct nlattr *protinfo; 3430 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 3431 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3432 int err = 0; 3433 3434 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 3435 if (nlh == NULL) 3436 return -EMSGSIZE; 3437 3438 ifm = nlmsg_data(nlh); 3439 ifm->ifi_family = AF_BRIDGE; 3440 ifm->__ifi_pad = 0; 3441 ifm->ifi_type = dev->type; 3442 ifm->ifi_index = dev->ifindex; 3443 ifm->ifi_flags = dev_get_flags(dev); 3444 ifm->ifi_change = 0; 3445 3446 3447 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 3448 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 3449 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 3450 (br_dev && 3451 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 3452 (dev->addr_len && 3453 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 3454 (dev->ifindex != dev_get_iflink(dev) && 3455 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 3456 goto nla_put_failure; 3457 3458 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 3459 if (!br_afspec) 3460 goto nla_put_failure; 3461 3462 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 3463 nla_nest_cancel(skb, br_afspec); 3464 goto nla_put_failure; 3465 } 3466 3467 if (mode != BRIDGE_MODE_UNDEF) { 3468 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 3469 nla_nest_cancel(skb, br_afspec); 3470 goto nla_put_failure; 3471 } 3472 } 3473 if (vlan_fill) { 3474 err = vlan_fill(skb, dev, filter_mask); 3475 if (err) { 3476 nla_nest_cancel(skb, br_afspec); 3477 goto nla_put_failure; 3478 } 3479 } 3480 nla_nest_end(skb, br_afspec); 3481 3482 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED); 3483 if (!protinfo) 3484 goto nla_put_failure; 3485 3486 if (brport_nla_put_flag(skb, flags, mask, 3487 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 3488 brport_nla_put_flag(skb, flags, mask, 3489 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 3490 brport_nla_put_flag(skb, flags, mask, 3491 IFLA_BRPORT_FAST_LEAVE, 3492 BR_MULTICAST_FAST_LEAVE) || 3493 brport_nla_put_flag(skb, flags, mask, 3494 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 3495 brport_nla_put_flag(skb, flags, mask, 3496 IFLA_BRPORT_LEARNING, BR_LEARNING) || 3497 brport_nla_put_flag(skb, flags, mask, 3498 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 3499 brport_nla_put_flag(skb, flags, mask, 3500 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 3501 brport_nla_put_flag(skb, flags, mask, 3502 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 3503 nla_nest_cancel(skb, protinfo); 3504 goto nla_put_failure; 3505 } 3506 3507 nla_nest_end(skb, protinfo); 3508 3509 nlmsg_end(skb, nlh); 3510 return 0; 3511 nla_put_failure: 3512 nlmsg_cancel(skb, nlh); 3513 return err ? err : -EMSGSIZE; 3514 } 3515 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 3516 3517 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 3518 { 3519 struct net *net = sock_net(skb->sk); 3520 struct net_device *dev; 3521 int idx = 0; 3522 u32 portid = NETLINK_CB(cb->skb).portid; 3523 u32 seq = cb->nlh->nlmsg_seq; 3524 u32 filter_mask = 0; 3525 int err; 3526 3527 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 3528 struct nlattr *extfilt; 3529 3530 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 3531 IFLA_EXT_MASK); 3532 if (extfilt) { 3533 if (nla_len(extfilt) < sizeof(filter_mask)) 3534 return -EINVAL; 3535 3536 filter_mask = nla_get_u32(extfilt); 3537 } 3538 } 3539 3540 rcu_read_lock(); 3541 for_each_netdev_rcu(net, dev) { 3542 const struct net_device_ops *ops = dev->netdev_ops; 3543 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3544 3545 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 3546 if (idx >= cb->args[0]) { 3547 err = br_dev->netdev_ops->ndo_bridge_getlink( 3548 skb, portid, seq, dev, 3549 filter_mask, NLM_F_MULTI); 3550 if (err < 0 && err != -EOPNOTSUPP) { 3551 if (likely(skb->len)) 3552 break; 3553 3554 goto out_err; 3555 } 3556 } 3557 idx++; 3558 } 3559 3560 if (ops->ndo_bridge_getlink) { 3561 if (idx >= cb->args[0]) { 3562 err = ops->ndo_bridge_getlink(skb, portid, 3563 seq, dev, 3564 filter_mask, 3565 NLM_F_MULTI); 3566 if (err < 0 && err != -EOPNOTSUPP) { 3567 if (likely(skb->len)) 3568 break; 3569 3570 goto out_err; 3571 } 3572 } 3573 idx++; 3574 } 3575 } 3576 err = skb->len; 3577 out_err: 3578 rcu_read_unlock(); 3579 cb->args[0] = idx; 3580 3581 return err; 3582 } 3583 3584 static inline size_t bridge_nlmsg_size(void) 3585 { 3586 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 3587 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 3588 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 3589 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 3590 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 3591 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 3592 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 3593 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 3594 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 3595 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 3596 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 3597 } 3598 3599 static int rtnl_bridge_notify(struct net_device *dev) 3600 { 3601 struct net *net = dev_net(dev); 3602 struct sk_buff *skb; 3603 int err = -EOPNOTSUPP; 3604 3605 if (!dev->netdev_ops->ndo_bridge_getlink) 3606 return 0; 3607 3608 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 3609 if (!skb) { 3610 err = -ENOMEM; 3611 goto errout; 3612 } 3613 3614 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 3615 if (err < 0) 3616 goto errout; 3617 3618 if (!skb->len) 3619 goto errout; 3620 3621 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 3622 return 0; 3623 errout: 3624 WARN_ON(err == -EMSGSIZE); 3625 kfree_skb(skb); 3626 if (err) 3627 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3628 return err; 3629 } 3630 3631 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3632 struct netlink_ext_ack *extack) 3633 { 3634 struct net *net = sock_net(skb->sk); 3635 struct ifinfomsg *ifm; 3636 struct net_device *dev; 3637 struct nlattr *br_spec, *attr = NULL; 3638 int rem, err = -EOPNOTSUPP; 3639 u16 flags = 0; 3640 bool have_flags = false; 3641 3642 if (nlmsg_len(nlh) < sizeof(*ifm)) 3643 return -EINVAL; 3644 3645 ifm = nlmsg_data(nlh); 3646 if (ifm->ifi_family != AF_BRIDGE) 3647 return -EPFNOSUPPORT; 3648 3649 dev = __dev_get_by_index(net, ifm->ifi_index); 3650 if (!dev) { 3651 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3652 return -ENODEV; 3653 } 3654 3655 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3656 if (br_spec) { 3657 nla_for_each_nested(attr, br_spec, rem) { 3658 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3659 if (nla_len(attr) < sizeof(flags)) 3660 return -EINVAL; 3661 3662 have_flags = true; 3663 flags = nla_get_u16(attr); 3664 break; 3665 } 3666 } 3667 } 3668 3669 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3670 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3671 3672 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 3673 err = -EOPNOTSUPP; 3674 goto out; 3675 } 3676 3677 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags); 3678 if (err) 3679 goto out; 3680 3681 flags &= ~BRIDGE_FLAGS_MASTER; 3682 } 3683 3684 if ((flags & BRIDGE_FLAGS_SELF)) { 3685 if (!dev->netdev_ops->ndo_bridge_setlink) 3686 err = -EOPNOTSUPP; 3687 else 3688 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 3689 flags); 3690 if (!err) { 3691 flags &= ~BRIDGE_FLAGS_SELF; 3692 3693 /* Generate event to notify upper layer of bridge 3694 * change 3695 */ 3696 err = rtnl_bridge_notify(dev); 3697 } 3698 } 3699 3700 if (have_flags) 3701 memcpy(nla_data(attr), &flags, sizeof(flags)); 3702 out: 3703 return err; 3704 } 3705 3706 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 3707 struct netlink_ext_ack *extack) 3708 { 3709 struct net *net = sock_net(skb->sk); 3710 struct ifinfomsg *ifm; 3711 struct net_device *dev; 3712 struct nlattr *br_spec, *attr = NULL; 3713 int rem, err = -EOPNOTSUPP; 3714 u16 flags = 0; 3715 bool have_flags = false; 3716 3717 if (nlmsg_len(nlh) < sizeof(*ifm)) 3718 return -EINVAL; 3719 3720 ifm = nlmsg_data(nlh); 3721 if (ifm->ifi_family != AF_BRIDGE) 3722 return -EPFNOSUPPORT; 3723 3724 dev = __dev_get_by_index(net, ifm->ifi_index); 3725 if (!dev) { 3726 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3727 return -ENODEV; 3728 } 3729 3730 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3731 if (br_spec) { 3732 nla_for_each_nested(attr, br_spec, rem) { 3733 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3734 if (nla_len(attr) < sizeof(flags)) 3735 return -EINVAL; 3736 3737 have_flags = true; 3738 flags = nla_get_u16(attr); 3739 break; 3740 } 3741 } 3742 } 3743 3744 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3745 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3746 3747 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 3748 err = -EOPNOTSUPP; 3749 goto out; 3750 } 3751 3752 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 3753 if (err) 3754 goto out; 3755 3756 flags &= ~BRIDGE_FLAGS_MASTER; 3757 } 3758 3759 if ((flags & BRIDGE_FLAGS_SELF)) { 3760 if (!dev->netdev_ops->ndo_bridge_dellink) 3761 err = -EOPNOTSUPP; 3762 else 3763 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 3764 flags); 3765 3766 if (!err) { 3767 flags &= ~BRIDGE_FLAGS_SELF; 3768 3769 /* Generate event to notify upper layer of bridge 3770 * change 3771 */ 3772 err = rtnl_bridge_notify(dev); 3773 } 3774 } 3775 3776 if (have_flags) 3777 memcpy(nla_data(attr), &flags, sizeof(flags)); 3778 out: 3779 return err; 3780 } 3781 3782 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 3783 { 3784 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 3785 (!idxattr || idxattr == attrid); 3786 } 3787 3788 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 3789 static int rtnl_get_offload_stats_attr_size(int attr_id) 3790 { 3791 switch (attr_id) { 3792 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 3793 return sizeof(struct rtnl_link_stats64); 3794 } 3795 3796 return 0; 3797 } 3798 3799 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 3800 int *prividx) 3801 { 3802 struct nlattr *attr = NULL; 3803 int attr_id, size; 3804 void *attr_data; 3805 int err; 3806 3807 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3808 dev->netdev_ops->ndo_get_offload_stats)) 3809 return -ENODATA; 3810 3811 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3812 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3813 if (attr_id < *prividx) 3814 continue; 3815 3816 size = rtnl_get_offload_stats_attr_size(attr_id); 3817 if (!size) 3818 continue; 3819 3820 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3821 continue; 3822 3823 attr = nla_reserve_64bit(skb, attr_id, size, 3824 IFLA_OFFLOAD_XSTATS_UNSPEC); 3825 if (!attr) 3826 goto nla_put_failure; 3827 3828 attr_data = nla_data(attr); 3829 memset(attr_data, 0, size); 3830 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 3831 attr_data); 3832 if (err) 3833 goto get_offload_stats_failure; 3834 } 3835 3836 if (!attr) 3837 return -ENODATA; 3838 3839 *prividx = 0; 3840 return 0; 3841 3842 nla_put_failure: 3843 err = -EMSGSIZE; 3844 get_offload_stats_failure: 3845 *prividx = attr_id; 3846 return err; 3847 } 3848 3849 static int rtnl_get_offload_stats_size(const struct net_device *dev) 3850 { 3851 int nla_size = 0; 3852 int attr_id; 3853 int size; 3854 3855 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3856 dev->netdev_ops->ndo_get_offload_stats)) 3857 return 0; 3858 3859 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3860 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3861 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3862 continue; 3863 size = rtnl_get_offload_stats_attr_size(attr_id); 3864 nla_size += nla_total_size_64bit(size); 3865 } 3866 3867 if (nla_size != 0) 3868 nla_size += nla_total_size(0); 3869 3870 return nla_size; 3871 } 3872 3873 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 3874 int type, u32 pid, u32 seq, u32 change, 3875 unsigned int flags, unsigned int filter_mask, 3876 int *idxattr, int *prividx) 3877 { 3878 struct if_stats_msg *ifsm; 3879 struct nlmsghdr *nlh; 3880 struct nlattr *attr; 3881 int s_prividx = *prividx; 3882 int err; 3883 3884 ASSERT_RTNL(); 3885 3886 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 3887 if (!nlh) 3888 return -EMSGSIZE; 3889 3890 ifsm = nlmsg_data(nlh); 3891 ifsm->ifindex = dev->ifindex; 3892 ifsm->filter_mask = filter_mask; 3893 3894 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 3895 struct rtnl_link_stats64 *sp; 3896 3897 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 3898 sizeof(struct rtnl_link_stats64), 3899 IFLA_STATS_UNSPEC); 3900 if (!attr) 3901 goto nla_put_failure; 3902 3903 sp = nla_data(attr); 3904 dev_get_stats(dev, sp); 3905 } 3906 3907 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 3908 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3909 3910 if (ops && ops->fill_linkxstats) { 3911 *idxattr = IFLA_STATS_LINK_XSTATS; 3912 attr = nla_nest_start(skb, 3913 IFLA_STATS_LINK_XSTATS); 3914 if (!attr) 3915 goto nla_put_failure; 3916 3917 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3918 nla_nest_end(skb, attr); 3919 if (err) 3920 goto nla_put_failure; 3921 *idxattr = 0; 3922 } 3923 } 3924 3925 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 3926 *idxattr)) { 3927 const struct rtnl_link_ops *ops = NULL; 3928 const struct net_device *master; 3929 3930 master = netdev_master_upper_dev_get(dev); 3931 if (master) 3932 ops = master->rtnl_link_ops; 3933 if (ops && ops->fill_linkxstats) { 3934 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 3935 attr = nla_nest_start(skb, 3936 IFLA_STATS_LINK_XSTATS_SLAVE); 3937 if (!attr) 3938 goto nla_put_failure; 3939 3940 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3941 nla_nest_end(skb, attr); 3942 if (err) 3943 goto nla_put_failure; 3944 *idxattr = 0; 3945 } 3946 } 3947 3948 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 3949 *idxattr)) { 3950 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 3951 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS); 3952 if (!attr) 3953 goto nla_put_failure; 3954 3955 err = rtnl_get_offload_stats(skb, dev, prividx); 3956 if (err == -ENODATA) 3957 nla_nest_cancel(skb, attr); 3958 else 3959 nla_nest_end(skb, attr); 3960 3961 if (err && err != -ENODATA) 3962 goto nla_put_failure; 3963 *idxattr = 0; 3964 } 3965 3966 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 3967 struct rtnl_af_ops *af_ops; 3968 3969 *idxattr = IFLA_STATS_AF_SPEC; 3970 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC); 3971 if (!attr) 3972 goto nla_put_failure; 3973 3974 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3975 if (af_ops->fill_stats_af) { 3976 struct nlattr *af; 3977 int err; 3978 3979 af = nla_nest_start(skb, af_ops->family); 3980 if (!af) 3981 goto nla_put_failure; 3982 3983 err = af_ops->fill_stats_af(skb, dev); 3984 3985 if (err == -ENODATA) 3986 nla_nest_cancel(skb, af); 3987 else if (err < 0) 3988 goto nla_put_failure; 3989 3990 nla_nest_end(skb, af); 3991 } 3992 } 3993 3994 nla_nest_end(skb, attr); 3995 3996 *idxattr = 0; 3997 } 3998 3999 nlmsg_end(skb, nlh); 4000 4001 return 0; 4002 4003 nla_put_failure: 4004 /* not a multi message or no progress mean a real error */ 4005 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 4006 nlmsg_cancel(skb, nlh); 4007 else 4008 nlmsg_end(skb, nlh); 4009 4010 return -EMSGSIZE; 4011 } 4012 4013 static size_t if_nlmsg_stats_size(const struct net_device *dev, 4014 u32 filter_mask) 4015 { 4016 size_t size = 0; 4017 4018 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 4019 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 4020 4021 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 4022 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 4023 int attr = IFLA_STATS_LINK_XSTATS; 4024 4025 if (ops && ops->get_linkxstats_size) { 4026 size += nla_total_size(ops->get_linkxstats_size(dev, 4027 attr)); 4028 /* for IFLA_STATS_LINK_XSTATS */ 4029 size += nla_total_size(0); 4030 } 4031 } 4032 4033 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 4034 struct net_device *_dev = (struct net_device *)dev; 4035 const struct rtnl_link_ops *ops = NULL; 4036 const struct net_device *master; 4037 4038 /* netdev_master_upper_dev_get can't take const */ 4039 master = netdev_master_upper_dev_get(_dev); 4040 if (master) 4041 ops = master->rtnl_link_ops; 4042 if (ops && ops->get_linkxstats_size) { 4043 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 4044 4045 size += nla_total_size(ops->get_linkxstats_size(dev, 4046 attr)); 4047 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 4048 size += nla_total_size(0); 4049 } 4050 } 4051 4052 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 4053 size += rtnl_get_offload_stats_size(dev); 4054 4055 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 4056 struct rtnl_af_ops *af_ops; 4057 4058 /* for IFLA_STATS_AF_SPEC */ 4059 size += nla_total_size(0); 4060 4061 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 4062 if (af_ops->get_stats_af_size) { 4063 size += nla_total_size( 4064 af_ops->get_stats_af_size(dev)); 4065 4066 /* for AF_* */ 4067 size += nla_total_size(0); 4068 } 4069 } 4070 } 4071 4072 return size; 4073 } 4074 4075 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh, 4076 struct netlink_ext_ack *extack) 4077 { 4078 struct net *net = sock_net(skb->sk); 4079 struct net_device *dev = NULL; 4080 int idxattr = 0, prividx = 0; 4081 struct if_stats_msg *ifsm; 4082 struct sk_buff *nskb; 4083 u32 filter_mask; 4084 int err; 4085 4086 if (nlmsg_len(nlh) < sizeof(*ifsm)) 4087 return -EINVAL; 4088 4089 ifsm = nlmsg_data(nlh); 4090 if (ifsm->ifindex > 0) 4091 dev = __dev_get_by_index(net, ifsm->ifindex); 4092 else 4093 return -EINVAL; 4094 4095 if (!dev) 4096 return -ENODEV; 4097 4098 filter_mask = ifsm->filter_mask; 4099 if (!filter_mask) 4100 return -EINVAL; 4101 4102 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 4103 if (!nskb) 4104 return -ENOBUFS; 4105 4106 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 4107 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 4108 0, filter_mask, &idxattr, &prividx); 4109 if (err < 0) { 4110 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 4111 WARN_ON(err == -EMSGSIZE); 4112 kfree_skb(nskb); 4113 } else { 4114 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 4115 } 4116 4117 return err; 4118 } 4119 4120 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 4121 { 4122 int h, s_h, err, s_idx, s_idxattr, s_prividx; 4123 struct net *net = sock_net(skb->sk); 4124 unsigned int flags = NLM_F_MULTI; 4125 struct if_stats_msg *ifsm; 4126 struct hlist_head *head; 4127 struct net_device *dev; 4128 u32 filter_mask = 0; 4129 int idx = 0; 4130 4131 s_h = cb->args[0]; 4132 s_idx = cb->args[1]; 4133 s_idxattr = cb->args[2]; 4134 s_prividx = cb->args[3]; 4135 4136 cb->seq = net->dev_base_seq; 4137 4138 if (nlmsg_len(cb->nlh) < sizeof(*ifsm)) 4139 return -EINVAL; 4140 4141 ifsm = nlmsg_data(cb->nlh); 4142 filter_mask = ifsm->filter_mask; 4143 if (!filter_mask) 4144 return -EINVAL; 4145 4146 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4147 idx = 0; 4148 head = &net->dev_index_head[h]; 4149 hlist_for_each_entry(dev, head, index_hlist) { 4150 if (idx < s_idx) 4151 goto cont; 4152 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 4153 NETLINK_CB(cb->skb).portid, 4154 cb->nlh->nlmsg_seq, 0, 4155 flags, filter_mask, 4156 &s_idxattr, &s_prividx); 4157 /* If we ran out of room on the first message, 4158 * we're in trouble 4159 */ 4160 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 4161 4162 if (err < 0) 4163 goto out; 4164 s_prividx = 0; 4165 s_idxattr = 0; 4166 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4167 cont: 4168 idx++; 4169 } 4170 } 4171 out: 4172 cb->args[3] = s_prividx; 4173 cb->args[2] = s_idxattr; 4174 cb->args[1] = idx; 4175 cb->args[0] = h; 4176 4177 return skb->len; 4178 } 4179 4180 /* Process one rtnetlink message. */ 4181 4182 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, 4183 struct netlink_ext_ack *extack) 4184 { 4185 struct net *net = sock_net(skb->sk); 4186 struct rtnl_link *handlers; 4187 int err = -EOPNOTSUPP; 4188 rtnl_doit_func doit; 4189 unsigned int flags; 4190 int kind; 4191 int family; 4192 int type; 4193 4194 type = nlh->nlmsg_type; 4195 if (type > RTM_MAX) 4196 return -EOPNOTSUPP; 4197 4198 type -= RTM_BASE; 4199 4200 /* All the messages must have at least 1 byte length */ 4201 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 4202 return 0; 4203 4204 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 4205 kind = type&3; 4206 4207 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 4208 return -EPERM; 4209 4210 if (family >= ARRAY_SIZE(rtnl_msg_handlers)) 4211 family = PF_UNSPEC; 4212 4213 rcu_read_lock(); 4214 handlers = rcu_dereference(rtnl_msg_handlers[family]); 4215 if (!handlers) { 4216 family = PF_UNSPEC; 4217 handlers = rcu_dereference(rtnl_msg_handlers[family]); 4218 } 4219 4220 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 4221 struct sock *rtnl; 4222 rtnl_dumpit_func dumpit; 4223 u16 min_dump_alloc = 0; 4224 4225 dumpit = READ_ONCE(handlers[type].dumpit); 4226 if (!dumpit) { 4227 family = PF_UNSPEC; 4228 handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]); 4229 if (!handlers) 4230 goto err_unlock; 4231 4232 dumpit = READ_ONCE(handlers[type].dumpit); 4233 if (!dumpit) 4234 goto err_unlock; 4235 } 4236 4237 refcount_inc(&rtnl_msg_handlers_ref[family]); 4238 4239 if (type == RTM_GETLINK - RTM_BASE) 4240 min_dump_alloc = rtnl_calcit(skb, nlh); 4241 4242 rcu_read_unlock(); 4243 4244 rtnl = net->rtnl; 4245 { 4246 struct netlink_dump_control c = { 4247 .dump = dumpit, 4248 .min_dump_alloc = min_dump_alloc, 4249 }; 4250 err = netlink_dump_start(rtnl, skb, nlh, &c); 4251 } 4252 refcount_dec(&rtnl_msg_handlers_ref[family]); 4253 return err; 4254 } 4255 4256 doit = READ_ONCE(handlers[type].doit); 4257 if (!doit) { 4258 family = PF_UNSPEC; 4259 handlers = rcu_dereference(rtnl_msg_handlers[family]); 4260 } 4261 4262 flags = READ_ONCE(handlers[type].flags); 4263 if (flags & RTNL_FLAG_DOIT_UNLOCKED) { 4264 refcount_inc(&rtnl_msg_handlers_ref[family]); 4265 doit = READ_ONCE(handlers[type].doit); 4266 rcu_read_unlock(); 4267 if (doit) 4268 err = doit(skb, nlh, extack); 4269 refcount_dec(&rtnl_msg_handlers_ref[family]); 4270 return err; 4271 } 4272 4273 rcu_read_unlock(); 4274 4275 rtnl_lock(); 4276 handlers = rtnl_dereference(rtnl_msg_handlers[family]); 4277 if (handlers) { 4278 doit = READ_ONCE(handlers[type].doit); 4279 if (doit) 4280 err = doit(skb, nlh, extack); 4281 } 4282 rtnl_unlock(); 4283 return err; 4284 4285 err_unlock: 4286 rcu_read_unlock(); 4287 return -EOPNOTSUPP; 4288 } 4289 4290 static void rtnetlink_rcv(struct sk_buff *skb) 4291 { 4292 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 4293 } 4294 4295 static int rtnetlink_bind(struct net *net, int group) 4296 { 4297 switch (group) { 4298 case RTNLGRP_IPV4_MROUTE_R: 4299 case RTNLGRP_IPV6_MROUTE_R: 4300 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 4301 return -EPERM; 4302 break; 4303 } 4304 return 0; 4305 } 4306 4307 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 4308 { 4309 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4310 4311 switch (event) { 4312 case NETDEV_REBOOT: 4313 case NETDEV_CHANGEADDR: 4314 case NETDEV_CHANGENAME: 4315 case NETDEV_FEAT_CHANGE: 4316 case NETDEV_BONDING_FAILOVER: 4317 case NETDEV_NOTIFY_PEERS: 4318 case NETDEV_RESEND_IGMP: 4319 case NETDEV_CHANGEINFODATA: 4320 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event), 4321 GFP_KERNEL); 4322 break; 4323 default: 4324 break; 4325 } 4326 return NOTIFY_DONE; 4327 } 4328 4329 static struct notifier_block rtnetlink_dev_notifier = { 4330 .notifier_call = rtnetlink_event, 4331 }; 4332 4333 4334 static int __net_init rtnetlink_net_init(struct net *net) 4335 { 4336 struct sock *sk; 4337 struct netlink_kernel_cfg cfg = { 4338 .groups = RTNLGRP_MAX, 4339 .input = rtnetlink_rcv, 4340 .cb_mutex = &rtnl_mutex, 4341 .flags = NL_CFG_F_NONROOT_RECV, 4342 .bind = rtnetlink_bind, 4343 }; 4344 4345 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 4346 if (!sk) 4347 return -ENOMEM; 4348 net->rtnl = sk; 4349 return 0; 4350 } 4351 4352 static void __net_exit rtnetlink_net_exit(struct net *net) 4353 { 4354 netlink_kernel_release(net->rtnl); 4355 net->rtnl = NULL; 4356 } 4357 4358 static struct pernet_operations rtnetlink_net_ops = { 4359 .init = rtnetlink_net_init, 4360 .exit = rtnetlink_net_exit, 4361 }; 4362 4363 void __init rtnetlink_init(void) 4364 { 4365 int i; 4366 4367 for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++) 4368 refcount_set(&rtnl_msg_handlers_ref[i], 1); 4369 4370 if (register_pernet_subsys(&rtnetlink_net_ops)) 4371 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 4372 4373 register_netdevice_notifier(&rtnetlink_dev_notifier); 4374 4375 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 4376 rtnl_dump_ifinfo, 0); 4377 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0); 4378 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0); 4379 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0); 4380 4381 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0); 4382 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0); 4383 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0); 4384 4385 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0); 4386 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0); 4387 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0); 4388 4389 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0); 4390 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0); 4391 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0); 4392 4393 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 4394 0); 4395 } 4396