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