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/errno.h> 20 #include <linux/module.h> 21 #include <linux/types.h> 22 #include <linux/socket.h> 23 #include <linux/kernel.h> 24 #include <linux/timer.h> 25 #include <linux/string.h> 26 #include <linux/sockios.h> 27 #include <linux/net.h> 28 #include <linux/fcntl.h> 29 #include <linux/mm.h> 30 #include <linux/slab.h> 31 #include <linux/interrupt.h> 32 #include <linux/capability.h> 33 #include <linux/skbuff.h> 34 #include <linux/init.h> 35 #include <linux/security.h> 36 #include <linux/mutex.h> 37 #include <linux/if_addr.h> 38 #include <linux/if_bridge.h> 39 #include <linux/if_vlan.h> 40 #include <linux/pci.h> 41 #include <linux/etherdevice.h> 42 43 #include <linux/uaccess.h> 44 45 #include <linux/inet.h> 46 #include <linux/netdevice.h> 47 #include <net/switchdev.h> 48 #include <net/ip.h> 49 #include <net/protocol.h> 50 #include <net/arp.h> 51 #include <net/route.h> 52 #include <net/udp.h> 53 #include <net/tcp.h> 54 #include <net/sock.h> 55 #include <net/pkt_sched.h> 56 #include <net/fib_rules.h> 57 #include <net/rtnetlink.h> 58 #include <net/net_namespace.h> 59 60 struct rtnl_link { 61 rtnl_doit_func doit; 62 rtnl_dumpit_func dumpit; 63 rtnl_calcit_func calcit; 64 }; 65 66 static DEFINE_MUTEX(rtnl_mutex); 67 68 void rtnl_lock(void) 69 { 70 mutex_lock(&rtnl_mutex); 71 } 72 EXPORT_SYMBOL(rtnl_lock); 73 74 static struct sk_buff *defer_kfree_skb_list; 75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail) 76 { 77 if (head && tail) { 78 tail->next = defer_kfree_skb_list; 79 defer_kfree_skb_list = head; 80 } 81 } 82 EXPORT_SYMBOL(rtnl_kfree_skbs); 83 84 void __rtnl_unlock(void) 85 { 86 struct sk_buff *head = defer_kfree_skb_list; 87 88 defer_kfree_skb_list = NULL; 89 90 mutex_unlock(&rtnl_mutex); 91 92 while (head) { 93 struct sk_buff *next = head->next; 94 95 kfree_skb(head); 96 cond_resched(); 97 head = next; 98 } 99 } 100 101 void rtnl_unlock(void) 102 { 103 /* This fellow will unlock it for us. */ 104 netdev_run_todo(); 105 } 106 EXPORT_SYMBOL(rtnl_unlock); 107 108 int rtnl_trylock(void) 109 { 110 return mutex_trylock(&rtnl_mutex); 111 } 112 EXPORT_SYMBOL(rtnl_trylock); 113 114 int rtnl_is_locked(void) 115 { 116 return mutex_is_locked(&rtnl_mutex); 117 } 118 EXPORT_SYMBOL(rtnl_is_locked); 119 120 #ifdef CONFIG_PROVE_LOCKING 121 bool lockdep_rtnl_is_held(void) 122 { 123 return lockdep_is_held(&rtnl_mutex); 124 } 125 EXPORT_SYMBOL(lockdep_rtnl_is_held); 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 127 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 129 130 static inline int rtm_msgindex(int msgtype) 131 { 132 int msgindex = msgtype - RTM_BASE; 133 134 /* 135 * msgindex < 0 implies someone tried to register a netlink 136 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 137 * the message type has not been added to linux/rtnetlink.h 138 */ 139 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 140 141 return msgindex; 142 } 143 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex) 145 { 146 struct rtnl_link *tab; 147 148 if (protocol <= RTNL_FAMILY_MAX) 149 tab = rtnl_msg_handlers[protocol]; 150 else 151 tab = NULL; 152 153 if (tab == NULL || tab[msgindex].doit == NULL) 154 tab = rtnl_msg_handlers[PF_UNSPEC]; 155 156 return tab[msgindex].doit; 157 } 158 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex) 160 { 161 struct rtnl_link *tab; 162 163 if (protocol <= RTNL_FAMILY_MAX) 164 tab = rtnl_msg_handlers[protocol]; 165 else 166 tab = NULL; 167 168 if (tab == NULL || tab[msgindex].dumpit == NULL) 169 tab = rtnl_msg_handlers[PF_UNSPEC]; 170 171 return tab[msgindex].dumpit; 172 } 173 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex) 175 { 176 struct rtnl_link *tab; 177 178 if (protocol <= RTNL_FAMILY_MAX) 179 tab = rtnl_msg_handlers[protocol]; 180 else 181 tab = NULL; 182 183 if (tab == NULL || tab[msgindex].calcit == NULL) 184 tab = rtnl_msg_handlers[PF_UNSPEC]; 185 186 return tab[msgindex].calcit; 187 } 188 189 /** 190 * __rtnl_register - Register a rtnetlink message type 191 * @protocol: Protocol family or PF_UNSPEC 192 * @msgtype: rtnetlink message type 193 * @doit: Function pointer called for each request message 194 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 195 * @calcit: Function pointer to calc size of dump message 196 * 197 * Registers the specified function pointers (at least one of them has 198 * to be non-NULL) to be called whenever a request message for the 199 * specified protocol family and message type is received. 200 * 201 * The special protocol family PF_UNSPEC may be used to define fallback 202 * function pointers for the case when no entry for the specific protocol 203 * family exists. 204 * 205 * Returns 0 on success or a negative error code. 206 */ 207 int __rtnl_register(int protocol, int msgtype, 208 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 209 rtnl_calcit_func calcit) 210 { 211 struct rtnl_link *tab; 212 int msgindex; 213 214 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 215 msgindex = rtm_msgindex(msgtype); 216 217 tab = rtnl_msg_handlers[protocol]; 218 if (tab == NULL) { 219 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL); 220 if (tab == NULL) 221 return -ENOBUFS; 222 223 rtnl_msg_handlers[protocol] = tab; 224 } 225 226 if (doit) 227 tab[msgindex].doit = doit; 228 229 if (dumpit) 230 tab[msgindex].dumpit = dumpit; 231 232 if (calcit) 233 tab[msgindex].calcit = calcit; 234 235 return 0; 236 } 237 EXPORT_SYMBOL_GPL(__rtnl_register); 238 239 /** 240 * rtnl_register - Register a rtnetlink message type 241 * 242 * Identical to __rtnl_register() but panics on failure. This is useful 243 * as failure of this function is very unlikely, it can only happen due 244 * to lack of memory when allocating the chain to store all message 245 * handlers for a protocol. Meant for use in init functions where lack 246 * of memory implies no sense in continuing. 247 */ 248 void rtnl_register(int protocol, int msgtype, 249 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 250 rtnl_calcit_func calcit) 251 { 252 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0) 253 panic("Unable to register rtnetlink message handler, " 254 "protocol = %d, message type = %d\n", 255 protocol, msgtype); 256 } 257 EXPORT_SYMBOL_GPL(rtnl_register); 258 259 /** 260 * rtnl_unregister - Unregister a rtnetlink message type 261 * @protocol: Protocol family or PF_UNSPEC 262 * @msgtype: rtnetlink message type 263 * 264 * Returns 0 on success or a negative error code. 265 */ 266 int rtnl_unregister(int protocol, int msgtype) 267 { 268 int msgindex; 269 270 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 271 msgindex = rtm_msgindex(msgtype); 272 273 if (rtnl_msg_handlers[protocol] == NULL) 274 return -ENOENT; 275 276 rtnl_msg_handlers[protocol][msgindex].doit = NULL; 277 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL; 278 rtnl_msg_handlers[protocol][msgindex].calcit = NULL; 279 280 return 0; 281 } 282 EXPORT_SYMBOL_GPL(rtnl_unregister); 283 284 /** 285 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 286 * @protocol : Protocol family or PF_UNSPEC 287 * 288 * Identical to calling rtnl_unregster() for all registered message types 289 * of a certain protocol family. 290 */ 291 void rtnl_unregister_all(int protocol) 292 { 293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 294 295 kfree(rtnl_msg_handlers[protocol]); 296 rtnl_msg_handlers[protocol] = NULL; 297 } 298 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 299 300 static LIST_HEAD(link_ops); 301 302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 303 { 304 const struct rtnl_link_ops *ops; 305 306 list_for_each_entry(ops, &link_ops, list) { 307 if (!strcmp(ops->kind, kind)) 308 return ops; 309 } 310 return NULL; 311 } 312 313 /** 314 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 315 * @ops: struct rtnl_link_ops * to register 316 * 317 * The caller must hold the rtnl_mutex. This function should be used 318 * by drivers that create devices during module initialization. It 319 * must be called before registering the devices. 320 * 321 * Returns 0 on success or a negative error code. 322 */ 323 int __rtnl_link_register(struct rtnl_link_ops *ops) 324 { 325 if (rtnl_link_ops_get(ops->kind)) 326 return -EEXIST; 327 328 /* The check for setup is here because if ops 329 * does not have that filled up, it is not possible 330 * to use the ops for creating device. So do not 331 * fill up dellink as well. That disables rtnl_dellink. 332 */ 333 if (ops->setup && !ops->dellink) 334 ops->dellink = unregister_netdevice_queue; 335 336 list_add_tail(&ops->list, &link_ops); 337 return 0; 338 } 339 EXPORT_SYMBOL_GPL(__rtnl_link_register); 340 341 /** 342 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 343 * @ops: struct rtnl_link_ops * to register 344 * 345 * Returns 0 on success or a negative error code. 346 */ 347 int rtnl_link_register(struct rtnl_link_ops *ops) 348 { 349 int err; 350 351 rtnl_lock(); 352 err = __rtnl_link_register(ops); 353 rtnl_unlock(); 354 return err; 355 } 356 EXPORT_SYMBOL_GPL(rtnl_link_register); 357 358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 359 { 360 struct net_device *dev; 361 LIST_HEAD(list_kill); 362 363 for_each_netdev(net, dev) { 364 if (dev->rtnl_link_ops == ops) 365 ops->dellink(dev, &list_kill); 366 } 367 unregister_netdevice_many(&list_kill); 368 } 369 370 /** 371 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 372 * @ops: struct rtnl_link_ops * to unregister 373 * 374 * The caller must hold the rtnl_mutex. 375 */ 376 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 377 { 378 struct net *net; 379 380 for_each_net(net) { 381 __rtnl_kill_links(net, ops); 382 } 383 list_del(&ops->list); 384 } 385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 386 387 /* Return with the rtnl_lock held when there are no network 388 * devices unregistering in any network namespace. 389 */ 390 static void rtnl_lock_unregistering_all(void) 391 { 392 struct net *net; 393 bool unregistering; 394 DEFINE_WAIT_FUNC(wait, woken_wake_function); 395 396 add_wait_queue(&netdev_unregistering_wq, &wait); 397 for (;;) { 398 unregistering = false; 399 rtnl_lock(); 400 for_each_net(net) { 401 if (net->dev_unreg_count > 0) { 402 unregistering = true; 403 break; 404 } 405 } 406 if (!unregistering) 407 break; 408 __rtnl_unlock(); 409 410 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); 411 } 412 remove_wait_queue(&netdev_unregistering_wq, &wait); 413 } 414 415 /** 416 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 417 * @ops: struct rtnl_link_ops * to unregister 418 */ 419 void rtnl_link_unregister(struct rtnl_link_ops *ops) 420 { 421 /* Close the race with cleanup_net() */ 422 mutex_lock(&net_mutex); 423 rtnl_lock_unregistering_all(); 424 __rtnl_link_unregister(ops); 425 rtnl_unlock(); 426 mutex_unlock(&net_mutex); 427 } 428 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 429 430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev) 431 { 432 struct net_device *master_dev; 433 const struct rtnl_link_ops *ops; 434 435 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 436 if (!master_dev) 437 return 0; 438 ops = master_dev->rtnl_link_ops; 439 if (!ops || !ops->get_slave_size) 440 return 0; 441 /* IFLA_INFO_SLAVE_DATA + nested data */ 442 return nla_total_size(sizeof(struct nlattr)) + 443 ops->get_slave_size(master_dev, dev); 444 } 445 446 static size_t rtnl_link_get_size(const struct net_device *dev) 447 { 448 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 449 size_t size; 450 451 if (!ops) 452 return 0; 453 454 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 455 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 456 457 if (ops->get_size) 458 /* IFLA_INFO_DATA + nested data */ 459 size += nla_total_size(sizeof(struct nlattr)) + 460 ops->get_size(dev); 461 462 if (ops->get_xstats_size) 463 /* IFLA_INFO_XSTATS */ 464 size += nla_total_size(ops->get_xstats_size(dev)); 465 466 size += rtnl_link_get_slave_info_data_size(dev); 467 468 return size; 469 } 470 471 static LIST_HEAD(rtnl_af_ops); 472 473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family) 474 { 475 const struct rtnl_af_ops *ops; 476 477 list_for_each_entry(ops, &rtnl_af_ops, list) { 478 if (ops->family == family) 479 return ops; 480 } 481 482 return NULL; 483 } 484 485 /** 486 * rtnl_af_register - Register rtnl_af_ops with rtnetlink. 487 * @ops: struct rtnl_af_ops * to register 488 * 489 * Returns 0 on success or a negative error code. 490 */ 491 void rtnl_af_register(struct rtnl_af_ops *ops) 492 { 493 rtnl_lock(); 494 list_add_tail(&ops->list, &rtnl_af_ops); 495 rtnl_unlock(); 496 } 497 EXPORT_SYMBOL_GPL(rtnl_af_register); 498 499 /** 500 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 501 * @ops: struct rtnl_af_ops * to unregister 502 * 503 * The caller must hold the rtnl_mutex. 504 */ 505 void __rtnl_af_unregister(struct rtnl_af_ops *ops) 506 { 507 list_del(&ops->list); 508 } 509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister); 510 511 /** 512 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 513 * @ops: struct rtnl_af_ops * to unregister 514 */ 515 void rtnl_af_unregister(struct rtnl_af_ops *ops) 516 { 517 rtnl_lock(); 518 __rtnl_af_unregister(ops); 519 rtnl_unlock(); 520 } 521 EXPORT_SYMBOL_GPL(rtnl_af_unregister); 522 523 static size_t rtnl_link_get_af_size(const struct net_device *dev, 524 u32 ext_filter_mask) 525 { 526 struct rtnl_af_ops *af_ops; 527 size_t size; 528 529 /* IFLA_AF_SPEC */ 530 size = nla_total_size(sizeof(struct nlattr)); 531 532 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 533 if (af_ops->get_link_af_size) { 534 /* AF_* + nested data */ 535 size += nla_total_size(sizeof(struct nlattr)) + 536 af_ops->get_link_af_size(dev, ext_filter_mask); 537 } 538 } 539 540 return size; 541 } 542 543 static bool rtnl_have_link_slave_info(const struct net_device *dev) 544 { 545 struct net_device *master_dev; 546 547 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 548 if (master_dev && master_dev->rtnl_link_ops) 549 return true; 550 return false; 551 } 552 553 static int rtnl_link_slave_info_fill(struct sk_buff *skb, 554 const struct net_device *dev) 555 { 556 struct net_device *master_dev; 557 const struct rtnl_link_ops *ops; 558 struct nlattr *slave_data; 559 int err; 560 561 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 562 if (!master_dev) 563 return 0; 564 ops = master_dev->rtnl_link_ops; 565 if (!ops) 566 return 0; 567 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0) 568 return -EMSGSIZE; 569 if (ops->fill_slave_info) { 570 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA); 571 if (!slave_data) 572 return -EMSGSIZE; 573 err = ops->fill_slave_info(skb, master_dev, dev); 574 if (err < 0) 575 goto err_cancel_slave_data; 576 nla_nest_end(skb, slave_data); 577 } 578 return 0; 579 580 err_cancel_slave_data: 581 nla_nest_cancel(skb, slave_data); 582 return err; 583 } 584 585 static int rtnl_link_info_fill(struct sk_buff *skb, 586 const struct net_device *dev) 587 { 588 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 589 struct nlattr *data; 590 int err; 591 592 if (!ops) 593 return 0; 594 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 595 return -EMSGSIZE; 596 if (ops->fill_xstats) { 597 err = ops->fill_xstats(skb, dev); 598 if (err < 0) 599 return err; 600 } 601 if (ops->fill_info) { 602 data = nla_nest_start(skb, IFLA_INFO_DATA); 603 if (data == NULL) 604 return -EMSGSIZE; 605 err = ops->fill_info(skb, dev); 606 if (err < 0) 607 goto err_cancel_data; 608 nla_nest_end(skb, data); 609 } 610 return 0; 611 612 err_cancel_data: 613 nla_nest_cancel(skb, data); 614 return err; 615 } 616 617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 618 { 619 struct nlattr *linkinfo; 620 int err = -EMSGSIZE; 621 622 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 623 if (linkinfo == NULL) 624 goto out; 625 626 err = rtnl_link_info_fill(skb, dev); 627 if (err < 0) 628 goto err_cancel_link; 629 630 err = rtnl_link_slave_info_fill(skb, dev); 631 if (err < 0) 632 goto err_cancel_link; 633 634 nla_nest_end(skb, linkinfo); 635 return 0; 636 637 err_cancel_link: 638 nla_nest_cancel(skb, linkinfo); 639 out: 640 return err; 641 } 642 643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo) 644 { 645 struct sock *rtnl = net->rtnl; 646 int err = 0; 647 648 NETLINK_CB(skb).dst_group = group; 649 if (echo) 650 atomic_inc(&skb->users); 651 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 652 if (echo) 653 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 654 return err; 655 } 656 657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 658 { 659 struct sock *rtnl = net->rtnl; 660 661 return nlmsg_unicast(rtnl, skb, pid); 662 } 663 EXPORT_SYMBOL(rtnl_unicast); 664 665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 666 struct nlmsghdr *nlh, gfp_t flags) 667 { 668 struct sock *rtnl = net->rtnl; 669 int report = 0; 670 671 if (nlh) 672 report = nlmsg_report(nlh); 673 674 nlmsg_notify(rtnl, skb, pid, group, report, flags); 675 } 676 EXPORT_SYMBOL(rtnl_notify); 677 678 void rtnl_set_sk_err(struct net *net, u32 group, int error) 679 { 680 struct sock *rtnl = net->rtnl; 681 682 netlink_set_err(rtnl, 0, group, error); 683 } 684 EXPORT_SYMBOL(rtnl_set_sk_err); 685 686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 687 { 688 struct nlattr *mx; 689 int i, valid = 0; 690 691 mx = nla_nest_start(skb, RTA_METRICS); 692 if (mx == NULL) 693 return -ENOBUFS; 694 695 for (i = 0; i < RTAX_MAX; i++) { 696 if (metrics[i]) { 697 if (i == RTAX_CC_ALGO - 1) { 698 char tmp[TCP_CA_NAME_MAX], *name; 699 700 name = tcp_ca_get_name_by_key(metrics[i], tmp); 701 if (!name) 702 continue; 703 if (nla_put_string(skb, i + 1, name)) 704 goto nla_put_failure; 705 } else if (i == RTAX_FEATURES - 1) { 706 u32 user_features = metrics[i] & RTAX_FEATURE_MASK; 707 708 if (!user_features) 709 continue; 710 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK); 711 if (nla_put_u32(skb, i + 1, user_features)) 712 goto nla_put_failure; 713 } else { 714 if (nla_put_u32(skb, i + 1, metrics[i])) 715 goto nla_put_failure; 716 } 717 valid++; 718 } 719 } 720 721 if (!valid) { 722 nla_nest_cancel(skb, mx); 723 return 0; 724 } 725 726 return nla_nest_end(skb, mx); 727 728 nla_put_failure: 729 nla_nest_cancel(skb, mx); 730 return -EMSGSIZE; 731 } 732 EXPORT_SYMBOL(rtnetlink_put_metrics); 733 734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 735 long expires, u32 error) 736 { 737 struct rta_cacheinfo ci = { 738 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse), 739 .rta_used = dst->__use, 740 .rta_clntref = atomic_read(&(dst->__refcnt)), 741 .rta_error = error, 742 .rta_id = id, 743 }; 744 745 if (expires) { 746 unsigned long clock; 747 748 clock = jiffies_to_clock_t(abs(expires)); 749 clock = min_t(unsigned long, clock, INT_MAX); 750 ci.rta_expires = (expires > 0) ? clock : -clock; 751 } 752 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 753 } 754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 755 756 static void set_operstate(struct net_device *dev, unsigned char transition) 757 { 758 unsigned char operstate = dev->operstate; 759 760 switch (transition) { 761 case IF_OPER_UP: 762 if ((operstate == IF_OPER_DORMANT || 763 operstate == IF_OPER_UNKNOWN) && 764 !netif_dormant(dev)) 765 operstate = IF_OPER_UP; 766 break; 767 768 case IF_OPER_DORMANT: 769 if (operstate == IF_OPER_UP || 770 operstate == IF_OPER_UNKNOWN) 771 operstate = IF_OPER_DORMANT; 772 break; 773 } 774 775 if (dev->operstate != operstate) { 776 write_lock_bh(&dev_base_lock); 777 dev->operstate = operstate; 778 write_unlock_bh(&dev_base_lock); 779 netdev_state_change(dev); 780 } 781 } 782 783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev) 784 { 785 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) | 786 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI)); 787 } 788 789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 790 const struct ifinfomsg *ifm) 791 { 792 unsigned int flags = ifm->ifi_flags; 793 794 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 795 if (ifm->ifi_change) 796 flags = (flags & ifm->ifi_change) | 797 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change); 798 799 return flags; 800 } 801 802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 803 const struct rtnl_link_stats64 *b) 804 { 805 a->rx_packets = b->rx_packets; 806 a->tx_packets = b->tx_packets; 807 a->rx_bytes = b->rx_bytes; 808 a->tx_bytes = b->tx_bytes; 809 a->rx_errors = b->rx_errors; 810 a->tx_errors = b->tx_errors; 811 a->rx_dropped = b->rx_dropped; 812 a->tx_dropped = b->tx_dropped; 813 814 a->multicast = b->multicast; 815 a->collisions = b->collisions; 816 817 a->rx_length_errors = b->rx_length_errors; 818 a->rx_over_errors = b->rx_over_errors; 819 a->rx_crc_errors = b->rx_crc_errors; 820 a->rx_frame_errors = b->rx_frame_errors; 821 a->rx_fifo_errors = b->rx_fifo_errors; 822 a->rx_missed_errors = b->rx_missed_errors; 823 824 a->tx_aborted_errors = b->tx_aborted_errors; 825 a->tx_carrier_errors = b->tx_carrier_errors; 826 a->tx_fifo_errors = b->tx_fifo_errors; 827 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 828 a->tx_window_errors = b->tx_window_errors; 829 830 a->rx_compressed = b->rx_compressed; 831 a->tx_compressed = b->tx_compressed; 832 833 a->rx_nohandler = b->rx_nohandler; 834 } 835 836 /* All VF info */ 837 static inline int rtnl_vfinfo_size(const struct net_device *dev, 838 u32 ext_filter_mask) 839 { 840 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) { 841 int num_vfs = dev_num_vf(dev->dev.parent); 842 size_t size = nla_total_size(0); 843 size += num_vfs * 844 (nla_total_size(0) + 845 nla_total_size(sizeof(struct ifla_vf_mac)) + 846 nla_total_size(sizeof(struct ifla_vf_vlan)) + 847 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */ 848 nla_total_size(MAX_VLAN_LIST_LEN * 849 sizeof(struct ifla_vf_vlan_info)) + 850 nla_total_size(sizeof(struct ifla_vf_spoofchk)) + 851 nla_total_size(sizeof(struct ifla_vf_tx_rate)) + 852 nla_total_size(sizeof(struct ifla_vf_rate)) + 853 nla_total_size(sizeof(struct ifla_vf_link_state)) + 854 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) + 855 nla_total_size(0) + /* nest IFLA_VF_STATS */ 856 /* IFLA_VF_STATS_RX_PACKETS */ 857 nla_total_size_64bit(sizeof(__u64)) + 858 /* IFLA_VF_STATS_TX_PACKETS */ 859 nla_total_size_64bit(sizeof(__u64)) + 860 /* IFLA_VF_STATS_RX_BYTES */ 861 nla_total_size_64bit(sizeof(__u64)) + 862 /* IFLA_VF_STATS_TX_BYTES */ 863 nla_total_size_64bit(sizeof(__u64)) + 864 /* IFLA_VF_STATS_BROADCAST */ 865 nla_total_size_64bit(sizeof(__u64)) + 866 /* IFLA_VF_STATS_MULTICAST */ 867 nla_total_size_64bit(sizeof(__u64)) + 868 nla_total_size(sizeof(struct ifla_vf_trust))); 869 return size; 870 } else 871 return 0; 872 } 873 874 static size_t rtnl_port_size(const struct net_device *dev, 875 u32 ext_filter_mask) 876 { 877 size_t port_size = nla_total_size(4) /* PORT_VF */ 878 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 879 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 880 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 881 + nla_total_size(1) /* PROT_VDP_REQUEST */ 882 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 883 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 884 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 885 + port_size; 886 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 887 + port_size; 888 889 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 890 !(ext_filter_mask & RTEXT_FILTER_VF)) 891 return 0; 892 if (dev_num_vf(dev->dev.parent)) 893 return port_self_size + vf_ports_size + 894 vf_port_size * dev_num_vf(dev->dev.parent); 895 else 896 return port_self_size; 897 } 898 899 static size_t rtnl_xdp_size(const struct net_device *dev) 900 { 901 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */ 902 nla_total_size(1); /* XDP_ATTACHED */ 903 904 if (!dev->netdev_ops->ndo_xdp) 905 return 0; 906 else 907 return xdp_size; 908 } 909 910 static noinline size_t if_nlmsg_size(const struct net_device *dev, 911 u32 ext_filter_mask) 912 { 913 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 914 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 915 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 916 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 917 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap)) 918 + nla_total_size(sizeof(struct rtnl_link_stats)) 919 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64)) 920 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 921 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 922 + nla_total_size(4) /* IFLA_TXQLEN */ 923 + nla_total_size(4) /* IFLA_WEIGHT */ 924 + nla_total_size(4) /* IFLA_MTU */ 925 + nla_total_size(4) /* IFLA_LINK */ 926 + nla_total_size(4) /* IFLA_MASTER */ 927 + nla_total_size(1) /* IFLA_CARRIER */ 928 + nla_total_size(4) /* IFLA_PROMISCUITY */ 929 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 930 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 931 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */ 932 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */ 933 + nla_total_size(1) /* IFLA_OPERSTATE */ 934 + nla_total_size(1) /* IFLA_LINKMODE */ 935 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 936 + nla_total_size(4) /* IFLA_LINK_NETNSID */ 937 + nla_total_size(ext_filter_mask 938 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */ 939 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */ 940 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 941 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */ 942 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */ 943 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */ 944 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */ 945 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */ 946 + rtnl_xdp_size(dev) /* IFLA_XDP */ 947 + nla_total_size(4) /* IFLA_EVENT */ 948 + nla_total_size(1); /* IFLA_PROTO_DOWN */ 949 950 } 951 952 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 953 { 954 struct nlattr *vf_ports; 955 struct nlattr *vf_port; 956 int vf; 957 int err; 958 959 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS); 960 if (!vf_ports) 961 return -EMSGSIZE; 962 963 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 964 vf_port = nla_nest_start(skb, IFLA_VF_PORT); 965 if (!vf_port) 966 goto nla_put_failure; 967 if (nla_put_u32(skb, IFLA_PORT_VF, vf)) 968 goto nla_put_failure; 969 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 970 if (err == -EMSGSIZE) 971 goto nla_put_failure; 972 if (err) { 973 nla_nest_cancel(skb, vf_port); 974 continue; 975 } 976 nla_nest_end(skb, vf_port); 977 } 978 979 nla_nest_end(skb, vf_ports); 980 981 return 0; 982 983 nla_put_failure: 984 nla_nest_cancel(skb, vf_ports); 985 return -EMSGSIZE; 986 } 987 988 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 989 { 990 struct nlattr *port_self; 991 int err; 992 993 port_self = nla_nest_start(skb, IFLA_PORT_SELF); 994 if (!port_self) 995 return -EMSGSIZE; 996 997 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 998 if (err) { 999 nla_nest_cancel(skb, port_self); 1000 return (err == -EMSGSIZE) ? err : 0; 1001 } 1002 1003 nla_nest_end(skb, port_self); 1004 1005 return 0; 1006 } 1007 1008 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev, 1009 u32 ext_filter_mask) 1010 { 1011 int err; 1012 1013 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 1014 !(ext_filter_mask & RTEXT_FILTER_VF)) 1015 return 0; 1016 1017 err = rtnl_port_self_fill(skb, dev); 1018 if (err) 1019 return err; 1020 1021 if (dev_num_vf(dev->dev.parent)) { 1022 err = rtnl_vf_ports_fill(skb, dev); 1023 if (err) 1024 return err; 1025 } 1026 1027 return 0; 1028 } 1029 1030 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev) 1031 { 1032 int err; 1033 struct netdev_phys_item_id ppid; 1034 1035 err = dev_get_phys_port_id(dev, &ppid); 1036 if (err) { 1037 if (err == -EOPNOTSUPP) 1038 return 0; 1039 return err; 1040 } 1041 1042 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id)) 1043 return -EMSGSIZE; 1044 1045 return 0; 1046 } 1047 1048 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev) 1049 { 1050 char name[IFNAMSIZ]; 1051 int err; 1052 1053 err = dev_get_phys_port_name(dev, name, sizeof(name)); 1054 if (err) { 1055 if (err == -EOPNOTSUPP) 1056 return 0; 1057 return err; 1058 } 1059 1060 if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name)) 1061 return -EMSGSIZE; 1062 1063 return 0; 1064 } 1065 1066 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev) 1067 { 1068 int err; 1069 struct switchdev_attr attr = { 1070 .orig_dev = dev, 1071 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, 1072 .flags = SWITCHDEV_F_NO_RECURSE, 1073 }; 1074 1075 err = switchdev_port_attr_get(dev, &attr); 1076 if (err) { 1077 if (err == -EOPNOTSUPP) 1078 return 0; 1079 return err; 1080 } 1081 1082 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len, 1083 attr.u.ppid.id)) 1084 return -EMSGSIZE; 1085 1086 return 0; 1087 } 1088 1089 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb, 1090 struct net_device *dev) 1091 { 1092 struct rtnl_link_stats64 *sp; 1093 struct nlattr *attr; 1094 1095 attr = nla_reserve_64bit(skb, IFLA_STATS64, 1096 sizeof(struct rtnl_link_stats64), IFLA_PAD); 1097 if (!attr) 1098 return -EMSGSIZE; 1099 1100 sp = nla_data(attr); 1101 dev_get_stats(dev, sp); 1102 1103 attr = nla_reserve(skb, IFLA_STATS, 1104 sizeof(struct rtnl_link_stats)); 1105 if (!attr) 1106 return -EMSGSIZE; 1107 1108 copy_rtnl_link_stats(nla_data(attr), sp); 1109 1110 return 0; 1111 } 1112 1113 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb, 1114 struct net_device *dev, 1115 int vfs_num, 1116 struct nlattr *vfinfo) 1117 { 1118 struct ifla_vf_rss_query_en vf_rss_query_en; 1119 struct nlattr *vf, *vfstats, *vfvlanlist; 1120 struct ifla_vf_link_state vf_linkstate; 1121 struct ifla_vf_vlan_info vf_vlan_info; 1122 struct ifla_vf_spoofchk vf_spoofchk; 1123 struct ifla_vf_tx_rate vf_tx_rate; 1124 struct ifla_vf_stats vf_stats; 1125 struct ifla_vf_trust vf_trust; 1126 struct ifla_vf_vlan vf_vlan; 1127 struct ifla_vf_rate vf_rate; 1128 struct ifla_vf_mac vf_mac; 1129 struct ifla_vf_info ivi; 1130 1131 /* Not all SR-IOV capable drivers support the 1132 * spoofcheck and "RSS query enable" query. Preset to 1133 * -1 so the user space tool can detect that the driver 1134 * didn't report anything. 1135 */ 1136 ivi.spoofchk = -1; 1137 ivi.rss_query_en = -1; 1138 ivi.trusted = -1; 1139 memset(ivi.mac, 0, sizeof(ivi.mac)); 1140 /* The default value for VF link state is "auto" 1141 * IFLA_VF_LINK_STATE_AUTO which equals zero 1142 */ 1143 ivi.linkstate = 0; 1144 /* VLAN Protocol by default is 802.1Q */ 1145 ivi.vlan_proto = htons(ETH_P_8021Q); 1146 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi)) 1147 return 0; 1148 1149 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info)); 1150 1151 vf_mac.vf = 1152 vf_vlan.vf = 1153 vf_vlan_info.vf = 1154 vf_rate.vf = 1155 vf_tx_rate.vf = 1156 vf_spoofchk.vf = 1157 vf_linkstate.vf = 1158 vf_rss_query_en.vf = 1159 vf_trust.vf = ivi.vf; 1160 1161 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 1162 vf_vlan.vlan = ivi.vlan; 1163 vf_vlan.qos = ivi.qos; 1164 vf_vlan_info.vlan = ivi.vlan; 1165 vf_vlan_info.qos = ivi.qos; 1166 vf_vlan_info.vlan_proto = ivi.vlan_proto; 1167 vf_tx_rate.rate = ivi.max_tx_rate; 1168 vf_rate.min_tx_rate = ivi.min_tx_rate; 1169 vf_rate.max_tx_rate = ivi.max_tx_rate; 1170 vf_spoofchk.setting = ivi.spoofchk; 1171 vf_linkstate.link_state = ivi.linkstate; 1172 vf_rss_query_en.setting = ivi.rss_query_en; 1173 vf_trust.setting = ivi.trusted; 1174 vf = nla_nest_start(skb, IFLA_VF_INFO); 1175 if (!vf) 1176 goto nla_put_vfinfo_failure; 1177 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) || 1178 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) || 1179 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate), 1180 &vf_rate) || 1181 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), 1182 &vf_tx_rate) || 1183 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk), 1184 &vf_spoofchk) || 1185 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate), 1186 &vf_linkstate) || 1187 nla_put(skb, IFLA_VF_RSS_QUERY_EN, 1188 sizeof(vf_rss_query_en), 1189 &vf_rss_query_en) || 1190 nla_put(skb, IFLA_VF_TRUST, 1191 sizeof(vf_trust), &vf_trust)) 1192 goto nla_put_vf_failure; 1193 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST); 1194 if (!vfvlanlist) 1195 goto nla_put_vf_failure; 1196 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info), 1197 &vf_vlan_info)) { 1198 nla_nest_cancel(skb, vfvlanlist); 1199 goto nla_put_vf_failure; 1200 } 1201 nla_nest_end(skb, vfvlanlist); 1202 memset(&vf_stats, 0, sizeof(vf_stats)); 1203 if (dev->netdev_ops->ndo_get_vf_stats) 1204 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num, 1205 &vf_stats); 1206 vfstats = nla_nest_start(skb, IFLA_VF_STATS); 1207 if (!vfstats) 1208 goto nla_put_vf_failure; 1209 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS, 1210 vf_stats.rx_packets, IFLA_VF_STATS_PAD) || 1211 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS, 1212 vf_stats.tx_packets, IFLA_VF_STATS_PAD) || 1213 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES, 1214 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) || 1215 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES, 1216 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) || 1217 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST, 1218 vf_stats.broadcast, IFLA_VF_STATS_PAD) || 1219 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST, 1220 vf_stats.multicast, IFLA_VF_STATS_PAD)) { 1221 nla_nest_cancel(skb, vfstats); 1222 goto nla_put_vf_failure; 1223 } 1224 nla_nest_end(skb, vfstats); 1225 nla_nest_end(skb, vf); 1226 return 0; 1227 1228 nla_put_vf_failure: 1229 nla_nest_cancel(skb, vf); 1230 nla_put_vfinfo_failure: 1231 nla_nest_cancel(skb, vfinfo); 1232 return -EMSGSIZE; 1233 } 1234 1235 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev) 1236 { 1237 struct rtnl_link_ifmap map; 1238 1239 memset(&map, 0, sizeof(map)); 1240 map.mem_start = dev->mem_start; 1241 map.mem_end = dev->mem_end; 1242 map.base_addr = dev->base_addr; 1243 map.irq = dev->irq; 1244 map.dma = dev->dma; 1245 map.port = dev->if_port; 1246 1247 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD)) 1248 return -EMSGSIZE; 1249 1250 return 0; 1251 } 1252 1253 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1254 { 1255 struct netdev_xdp xdp_op = {}; 1256 struct nlattr *xdp; 1257 int err; 1258 1259 if (!dev->netdev_ops->ndo_xdp) 1260 return 0; 1261 xdp = nla_nest_start(skb, IFLA_XDP); 1262 if (!xdp) 1263 return -EMSGSIZE; 1264 xdp_op.command = XDP_QUERY_PROG; 1265 err = dev->netdev_ops->ndo_xdp(dev, &xdp_op); 1266 if (err) 1267 goto err_cancel; 1268 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached); 1269 if (err) 1270 goto err_cancel; 1271 1272 nla_nest_end(skb, xdp); 1273 return 0; 1274 1275 err_cancel: 1276 nla_nest_cancel(skb, xdp); 1277 return err; 1278 } 1279 1280 static int rtnl_fill_link_event(struct sk_buff *skb, unsigned long event) 1281 { 1282 u32 rtnl_event; 1283 1284 switch (event) { 1285 case NETDEV_REBOOT: 1286 rtnl_event = IFLA_EVENT_REBOOT; 1287 break; 1288 case NETDEV_CHANGEMTU: 1289 rtnl_event = IFLA_EVENT_CHANGE_MTU; 1290 break; 1291 case NETDEV_CHANGEADDR: 1292 rtnl_event = IFLA_EVENT_CHANGE_ADDR; 1293 break; 1294 case NETDEV_CHANGENAME: 1295 rtnl_event = IFLA_EVENT_CHANGE_NAME; 1296 break; 1297 case NETDEV_FEAT_CHANGE: 1298 rtnl_event = IFLA_EVENT_FEAT_CHANGE; 1299 break; 1300 case NETDEV_BONDING_FAILOVER: 1301 rtnl_event = IFLA_EVENT_BONDING_FAILOVER; 1302 break; 1303 case NETDEV_POST_TYPE_CHANGE: 1304 rtnl_event = IFLA_EVENT_POST_TYPE_CHANGE; 1305 break; 1306 case NETDEV_NOTIFY_PEERS: 1307 rtnl_event = IFLA_EVENT_NOTIFY_PEERS; 1308 break; 1309 case NETDEV_CHANGEUPPER: 1310 rtnl_event = IFLA_EVENT_CHANGE_UPPER; 1311 break; 1312 case NETDEV_RESEND_IGMP: 1313 rtnl_event = IFLA_EVENT_RESEND_IGMP; 1314 break; 1315 case NETDEV_PRECHANGEMTU: 1316 rtnl_event = IFLA_EVENT_PRE_CHANGE_MTU; 1317 break; 1318 case NETDEV_CHANGEINFODATA: 1319 rtnl_event = IFLA_EVENT_CHANGE_INFO_DATA; 1320 break; 1321 case NETDEV_PRECHANGEUPPER: 1322 rtnl_event = IFLA_EVENT_PRE_CHANGE_UPPER; 1323 break; 1324 case NETDEV_CHANGELOWERSTATE: 1325 rtnl_event = IFLA_EVENT_CHANGE_LOWER_STATE; 1326 break; 1327 case NETDEV_UDP_TUNNEL_PUSH_INFO: 1328 rtnl_event = IFLA_EVENT_UDP_TUNNEL_PUSH_INFO; 1329 break; 1330 case NETDEV_CHANGE_TX_QUEUE_LEN: 1331 rtnl_event = IFLA_EVENT_CHANGE_TX_QUEUE_LEN; 1332 break; 1333 default: 1334 return 0; 1335 } 1336 1337 return nla_put_u32(skb, IFLA_EVENT, rtnl_event); 1338 } 1339 1340 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 1341 int type, u32 pid, u32 seq, u32 change, 1342 unsigned int flags, u32 ext_filter_mask, 1343 unsigned long event) 1344 { 1345 struct ifinfomsg *ifm; 1346 struct nlmsghdr *nlh; 1347 struct nlattr *af_spec; 1348 struct rtnl_af_ops *af_ops; 1349 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1350 1351 ASSERT_RTNL(); 1352 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1353 if (nlh == NULL) 1354 return -EMSGSIZE; 1355 1356 ifm = nlmsg_data(nlh); 1357 ifm->ifi_family = AF_UNSPEC; 1358 ifm->__ifi_pad = 0; 1359 ifm->ifi_type = dev->type; 1360 ifm->ifi_index = dev->ifindex; 1361 ifm->ifi_flags = dev_get_flags(dev); 1362 ifm->ifi_change = change; 1363 1364 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1365 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1366 nla_put_u8(skb, IFLA_OPERSTATE, 1367 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1368 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1369 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1370 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1371 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1372 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1373 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1374 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1375 #ifdef CONFIG_RPS 1376 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1377 #endif 1378 (dev->ifindex != dev_get_iflink(dev) && 1379 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) || 1380 (upper_dev && 1381 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) || 1382 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1383 (dev->qdisc && 1384 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1385 (dev->ifalias && 1386 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) || 1387 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1388 atomic_read(&dev->carrier_changes)) || 1389 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down)) 1390 goto nla_put_failure; 1391 1392 if (rtnl_fill_link_event(skb, event)) 1393 goto nla_put_failure; 1394 1395 if (rtnl_fill_link_ifmap(skb, dev)) 1396 goto nla_put_failure; 1397 1398 if (dev->addr_len) { 1399 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1400 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1401 goto nla_put_failure; 1402 } 1403 1404 if (rtnl_phys_port_id_fill(skb, dev)) 1405 goto nla_put_failure; 1406 1407 if (rtnl_phys_port_name_fill(skb, dev)) 1408 goto nla_put_failure; 1409 1410 if (rtnl_phys_switch_id_fill(skb, dev)) 1411 goto nla_put_failure; 1412 1413 if (rtnl_fill_stats(skb, dev)) 1414 goto nla_put_failure; 1415 1416 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) && 1417 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent))) 1418 goto nla_put_failure; 1419 1420 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent && 1421 ext_filter_mask & RTEXT_FILTER_VF) { 1422 int i; 1423 struct nlattr *vfinfo; 1424 int num_vfs = dev_num_vf(dev->dev.parent); 1425 1426 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1427 if (!vfinfo) 1428 goto nla_put_failure; 1429 for (i = 0; i < num_vfs; i++) { 1430 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1431 goto nla_put_failure; 1432 } 1433 1434 nla_nest_end(skb, vfinfo); 1435 } 1436 1437 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1438 goto nla_put_failure; 1439 1440 if (rtnl_xdp_fill(skb, dev)) 1441 goto nla_put_failure; 1442 1443 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1444 if (rtnl_link_fill(skb, dev) < 0) 1445 goto nla_put_failure; 1446 } 1447 1448 if (dev->rtnl_link_ops && 1449 dev->rtnl_link_ops->get_link_net) { 1450 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1451 1452 if (!net_eq(dev_net(dev), link_net)) { 1453 int id = peernet2id_alloc(dev_net(dev), link_net); 1454 1455 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1456 goto nla_put_failure; 1457 } 1458 } 1459 1460 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC))) 1461 goto nla_put_failure; 1462 1463 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 1464 if (af_ops->fill_link_af) { 1465 struct nlattr *af; 1466 int err; 1467 1468 if (!(af = nla_nest_start(skb, af_ops->family))) 1469 goto nla_put_failure; 1470 1471 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1472 1473 /* 1474 * Caller may return ENODATA to indicate that there 1475 * was no data to be dumped. This is not an error, it 1476 * means we should trim the attribute header and 1477 * continue. 1478 */ 1479 if (err == -ENODATA) 1480 nla_nest_cancel(skb, af); 1481 else if (err < 0) 1482 goto nla_put_failure; 1483 1484 nla_nest_end(skb, af); 1485 } 1486 } 1487 1488 nla_nest_end(skb, af_spec); 1489 1490 nlmsg_end(skb, nlh); 1491 return 0; 1492 1493 nla_put_failure: 1494 nlmsg_cancel(skb, nlh); 1495 return -EMSGSIZE; 1496 } 1497 1498 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1499 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1500 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1501 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1502 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1503 [IFLA_MTU] = { .type = NLA_U32 }, 1504 [IFLA_LINK] = { .type = NLA_U32 }, 1505 [IFLA_MASTER] = { .type = NLA_U32 }, 1506 [IFLA_CARRIER] = { .type = NLA_U8 }, 1507 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1508 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1509 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1510 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1511 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1512 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1513 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1514 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 1515 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1516 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1517 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1518 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1519 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1520 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1521 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1522 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1523 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1524 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1525 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1526 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1527 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1528 [IFLA_XDP] = { .type = NLA_NESTED }, 1529 [IFLA_EVENT] = { .type = NLA_U32 }, 1530 }; 1531 1532 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1533 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1534 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1535 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1536 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1537 }; 1538 1539 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1540 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1541 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1542 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1543 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1544 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1545 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1546 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1547 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1548 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1549 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1550 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1551 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1552 }; 1553 1554 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1555 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1556 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1557 .len = PORT_PROFILE_MAX }, 1558 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1559 .len = PORT_UUID_MAX }, 1560 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1561 .len = PORT_UUID_MAX }, 1562 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1563 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1564 1565 /* Unused, but we need to keep it here since user space could 1566 * fill it. It's also broken with regard to NLA_BINARY use in 1567 * combination with structs. 1568 */ 1569 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1570 .len = sizeof(struct ifla_port_vsi) }, 1571 }; 1572 1573 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1574 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1575 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1576 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1577 }; 1578 1579 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1580 { 1581 const struct rtnl_link_ops *ops = NULL; 1582 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1583 1584 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0) 1585 return NULL; 1586 1587 if (linfo[IFLA_INFO_KIND]) { 1588 char kind[MODULE_NAME_LEN]; 1589 1590 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1591 ops = rtnl_link_ops_get(kind); 1592 } 1593 1594 return ops; 1595 } 1596 1597 static bool link_master_filtered(struct net_device *dev, int master_idx) 1598 { 1599 struct net_device *master; 1600 1601 if (!master_idx) 1602 return false; 1603 1604 master = netdev_master_upper_dev_get(dev); 1605 if (!master || master->ifindex != master_idx) 1606 return true; 1607 1608 return false; 1609 } 1610 1611 static bool link_kind_filtered(const struct net_device *dev, 1612 const struct rtnl_link_ops *kind_ops) 1613 { 1614 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1615 return true; 1616 1617 return false; 1618 } 1619 1620 static bool link_dump_filtered(struct net_device *dev, 1621 int master_idx, 1622 const struct rtnl_link_ops *kind_ops) 1623 { 1624 if (link_master_filtered(dev, master_idx) || 1625 link_kind_filtered(dev, kind_ops)) 1626 return true; 1627 1628 return false; 1629 } 1630 1631 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1632 { 1633 struct net *net = sock_net(skb->sk); 1634 int h, s_h; 1635 int idx = 0, s_idx; 1636 struct net_device *dev; 1637 struct hlist_head *head; 1638 struct nlattr *tb[IFLA_MAX+1]; 1639 u32 ext_filter_mask = 0; 1640 const struct rtnl_link_ops *kind_ops = NULL; 1641 unsigned int flags = NLM_F_MULTI; 1642 int master_idx = 0; 1643 int err; 1644 int hdrlen; 1645 1646 s_h = cb->args[0]; 1647 s_idx = cb->args[1]; 1648 1649 cb->seq = net->dev_base_seq; 1650 1651 /* A hack to preserve kernel<->userspace interface. 1652 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1653 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1654 * what iproute2 < v3.9.0 used. 1655 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1656 * attribute, its netlink message is shorter than struct ifinfomsg. 1657 */ 1658 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1659 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1660 1661 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) { 1662 1663 if (tb[IFLA_EXT_MASK]) 1664 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1665 1666 if (tb[IFLA_MASTER]) 1667 master_idx = nla_get_u32(tb[IFLA_MASTER]); 1668 1669 if (tb[IFLA_LINKINFO]) 1670 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]); 1671 1672 if (master_idx || kind_ops) 1673 flags |= NLM_F_DUMP_FILTERED; 1674 } 1675 1676 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1677 idx = 0; 1678 head = &net->dev_index_head[h]; 1679 hlist_for_each_entry(dev, head, index_hlist) { 1680 if (link_dump_filtered(dev, master_idx, kind_ops)) 1681 goto cont; 1682 if (idx < s_idx) 1683 goto cont; 1684 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 1685 NETLINK_CB(cb->skb).portid, 1686 cb->nlh->nlmsg_seq, 0, 1687 flags, 1688 ext_filter_mask, 0); 1689 /* If we ran out of room on the first message, 1690 * we're in trouble 1691 */ 1692 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 1693 1694 if (err < 0) 1695 goto out; 1696 1697 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1698 cont: 1699 idx++; 1700 } 1701 } 1702 out: 1703 cb->args[1] = idx; 1704 cb->args[0] = h; 1705 1706 return skb->len; 1707 } 1708 1709 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len) 1710 { 1711 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy); 1712 } 1713 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1714 1715 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1716 { 1717 struct net *net; 1718 /* Examine the link attributes and figure out which 1719 * network namespace we are talking about. 1720 */ 1721 if (tb[IFLA_NET_NS_PID]) 1722 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1723 else if (tb[IFLA_NET_NS_FD]) 1724 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1725 else 1726 net = get_net(src_net); 1727 return net; 1728 } 1729 EXPORT_SYMBOL(rtnl_link_get_net); 1730 1731 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1732 { 1733 if (dev) { 1734 if (tb[IFLA_ADDRESS] && 1735 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1736 return -EINVAL; 1737 1738 if (tb[IFLA_BROADCAST] && 1739 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1740 return -EINVAL; 1741 } 1742 1743 if (tb[IFLA_AF_SPEC]) { 1744 struct nlattr *af; 1745 int rem, err; 1746 1747 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1748 const struct rtnl_af_ops *af_ops; 1749 1750 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1751 return -EAFNOSUPPORT; 1752 1753 if (!af_ops->set_link_af) 1754 return -EOPNOTSUPP; 1755 1756 if (af_ops->validate_link_af) { 1757 err = af_ops->validate_link_af(dev, af); 1758 if (err < 0) 1759 return err; 1760 } 1761 } 1762 } 1763 1764 return 0; 1765 } 1766 1767 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 1768 int guid_type) 1769 { 1770 const struct net_device_ops *ops = dev->netdev_ops; 1771 1772 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 1773 } 1774 1775 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 1776 { 1777 if (dev->type != ARPHRD_INFINIBAND) 1778 return -EOPNOTSUPP; 1779 1780 return handle_infiniband_guid(dev, ivt, guid_type); 1781 } 1782 1783 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 1784 { 1785 const struct net_device_ops *ops = dev->netdev_ops; 1786 int err = -EINVAL; 1787 1788 if (tb[IFLA_VF_MAC]) { 1789 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 1790 1791 err = -EOPNOTSUPP; 1792 if (ops->ndo_set_vf_mac) 1793 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1794 ivm->mac); 1795 if (err < 0) 1796 return err; 1797 } 1798 1799 if (tb[IFLA_VF_VLAN]) { 1800 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 1801 1802 err = -EOPNOTSUPP; 1803 if (ops->ndo_set_vf_vlan) 1804 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 1805 ivv->qos, 1806 htons(ETH_P_8021Q)); 1807 if (err < 0) 1808 return err; 1809 } 1810 1811 if (tb[IFLA_VF_VLAN_LIST]) { 1812 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 1813 struct nlattr *attr; 1814 int rem, len = 0; 1815 1816 err = -EOPNOTSUPP; 1817 if (!ops->ndo_set_vf_vlan) 1818 return err; 1819 1820 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 1821 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 1822 nla_len(attr) < NLA_HDRLEN) { 1823 return -EINVAL; 1824 } 1825 if (len >= MAX_VLAN_LIST_LEN) 1826 return -EOPNOTSUPP; 1827 ivvl[len] = nla_data(attr); 1828 1829 len++; 1830 } 1831 if (len == 0) 1832 return -EINVAL; 1833 1834 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 1835 ivvl[0]->qos, ivvl[0]->vlan_proto); 1836 if (err < 0) 1837 return err; 1838 } 1839 1840 if (tb[IFLA_VF_TX_RATE]) { 1841 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 1842 struct ifla_vf_info ivf; 1843 1844 err = -EOPNOTSUPP; 1845 if (ops->ndo_get_vf_config) 1846 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 1847 if (err < 0) 1848 return err; 1849 1850 err = -EOPNOTSUPP; 1851 if (ops->ndo_set_vf_rate) 1852 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1853 ivf.min_tx_rate, 1854 ivt->rate); 1855 if (err < 0) 1856 return err; 1857 } 1858 1859 if (tb[IFLA_VF_RATE]) { 1860 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 1861 1862 err = -EOPNOTSUPP; 1863 if (ops->ndo_set_vf_rate) 1864 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1865 ivt->min_tx_rate, 1866 ivt->max_tx_rate); 1867 if (err < 0) 1868 return err; 1869 } 1870 1871 if (tb[IFLA_VF_SPOOFCHK]) { 1872 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 1873 1874 err = -EOPNOTSUPP; 1875 if (ops->ndo_set_vf_spoofchk) 1876 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 1877 ivs->setting); 1878 if (err < 0) 1879 return err; 1880 } 1881 1882 if (tb[IFLA_VF_LINK_STATE]) { 1883 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 1884 1885 err = -EOPNOTSUPP; 1886 if (ops->ndo_set_vf_link_state) 1887 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 1888 ivl->link_state); 1889 if (err < 0) 1890 return err; 1891 } 1892 1893 if (tb[IFLA_VF_RSS_QUERY_EN]) { 1894 struct ifla_vf_rss_query_en *ivrssq_en; 1895 1896 err = -EOPNOTSUPP; 1897 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 1898 if (ops->ndo_set_vf_rss_query_en) 1899 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 1900 ivrssq_en->setting); 1901 if (err < 0) 1902 return err; 1903 } 1904 1905 if (tb[IFLA_VF_TRUST]) { 1906 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 1907 1908 err = -EOPNOTSUPP; 1909 if (ops->ndo_set_vf_trust) 1910 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 1911 if (err < 0) 1912 return err; 1913 } 1914 1915 if (tb[IFLA_VF_IB_NODE_GUID]) { 1916 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 1917 1918 if (!ops->ndo_set_vf_guid) 1919 return -EOPNOTSUPP; 1920 1921 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 1922 } 1923 1924 if (tb[IFLA_VF_IB_PORT_GUID]) { 1925 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 1926 1927 if (!ops->ndo_set_vf_guid) 1928 return -EOPNOTSUPP; 1929 1930 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 1931 } 1932 1933 return err; 1934 } 1935 1936 static int do_set_master(struct net_device *dev, int ifindex) 1937 { 1938 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1939 const struct net_device_ops *ops; 1940 int err; 1941 1942 if (upper_dev) { 1943 if (upper_dev->ifindex == ifindex) 1944 return 0; 1945 ops = upper_dev->netdev_ops; 1946 if (ops->ndo_del_slave) { 1947 err = ops->ndo_del_slave(upper_dev, dev); 1948 if (err) 1949 return err; 1950 } else { 1951 return -EOPNOTSUPP; 1952 } 1953 } 1954 1955 if (ifindex) { 1956 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 1957 if (!upper_dev) 1958 return -EINVAL; 1959 ops = upper_dev->netdev_ops; 1960 if (ops->ndo_add_slave) { 1961 err = ops->ndo_add_slave(upper_dev, dev); 1962 if (err) 1963 return err; 1964 } else { 1965 return -EOPNOTSUPP; 1966 } 1967 } 1968 return 0; 1969 } 1970 1971 #define DO_SETLINK_MODIFIED 0x01 1972 /* notify flag means notify + modified. */ 1973 #define DO_SETLINK_NOTIFY 0x03 1974 static int do_setlink(const struct sk_buff *skb, 1975 struct net_device *dev, struct ifinfomsg *ifm, 1976 struct nlattr **tb, char *ifname, int status) 1977 { 1978 const struct net_device_ops *ops = dev->netdev_ops; 1979 int err; 1980 1981 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) { 1982 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1983 if (IS_ERR(net)) { 1984 err = PTR_ERR(net); 1985 goto errout; 1986 } 1987 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { 1988 put_net(net); 1989 err = -EPERM; 1990 goto errout; 1991 } 1992 err = dev_change_net_namespace(dev, net, ifname); 1993 put_net(net); 1994 if (err) 1995 goto errout; 1996 status |= DO_SETLINK_MODIFIED; 1997 } 1998 1999 if (tb[IFLA_MAP]) { 2000 struct rtnl_link_ifmap *u_map; 2001 struct ifmap k_map; 2002 2003 if (!ops->ndo_set_config) { 2004 err = -EOPNOTSUPP; 2005 goto errout; 2006 } 2007 2008 if (!netif_device_present(dev)) { 2009 err = -ENODEV; 2010 goto errout; 2011 } 2012 2013 u_map = nla_data(tb[IFLA_MAP]); 2014 k_map.mem_start = (unsigned long) u_map->mem_start; 2015 k_map.mem_end = (unsigned long) u_map->mem_end; 2016 k_map.base_addr = (unsigned short) u_map->base_addr; 2017 k_map.irq = (unsigned char) u_map->irq; 2018 k_map.dma = (unsigned char) u_map->dma; 2019 k_map.port = (unsigned char) u_map->port; 2020 2021 err = ops->ndo_set_config(dev, &k_map); 2022 if (err < 0) 2023 goto errout; 2024 2025 status |= DO_SETLINK_NOTIFY; 2026 } 2027 2028 if (tb[IFLA_ADDRESS]) { 2029 struct sockaddr *sa; 2030 int len; 2031 2032 len = sizeof(sa_family_t) + dev->addr_len; 2033 sa = kmalloc(len, GFP_KERNEL); 2034 if (!sa) { 2035 err = -ENOMEM; 2036 goto errout; 2037 } 2038 sa->sa_family = dev->type; 2039 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 2040 dev->addr_len); 2041 err = dev_set_mac_address(dev, sa); 2042 kfree(sa); 2043 if (err) 2044 goto errout; 2045 status |= DO_SETLINK_MODIFIED; 2046 } 2047 2048 if (tb[IFLA_MTU]) { 2049 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2050 if (err < 0) 2051 goto errout; 2052 status |= DO_SETLINK_MODIFIED; 2053 } 2054 2055 if (tb[IFLA_GROUP]) { 2056 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2057 status |= DO_SETLINK_NOTIFY; 2058 } 2059 2060 /* 2061 * Interface selected by interface index but interface 2062 * name provided implies that a name change has been 2063 * requested. 2064 */ 2065 if (ifm->ifi_index > 0 && ifname[0]) { 2066 err = dev_change_name(dev, ifname); 2067 if (err < 0) 2068 goto errout; 2069 status |= DO_SETLINK_MODIFIED; 2070 } 2071 2072 if (tb[IFLA_IFALIAS]) { 2073 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2074 nla_len(tb[IFLA_IFALIAS])); 2075 if (err < 0) 2076 goto errout; 2077 status |= DO_SETLINK_NOTIFY; 2078 } 2079 2080 if (tb[IFLA_BROADCAST]) { 2081 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2082 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2083 } 2084 2085 if (ifm->ifi_flags || ifm->ifi_change) { 2086 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2087 if (err < 0) 2088 goto errout; 2089 } 2090 2091 if (tb[IFLA_MASTER]) { 2092 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2093 if (err) 2094 goto errout; 2095 status |= DO_SETLINK_MODIFIED; 2096 } 2097 2098 if (tb[IFLA_CARRIER]) { 2099 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2100 if (err) 2101 goto errout; 2102 status |= DO_SETLINK_MODIFIED; 2103 } 2104 2105 if (tb[IFLA_TXQLEN]) { 2106 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]); 2107 unsigned long orig_len = dev->tx_queue_len; 2108 2109 if (dev->tx_queue_len ^ value) { 2110 dev->tx_queue_len = value; 2111 err = call_netdevice_notifiers( 2112 NETDEV_CHANGE_TX_QUEUE_LEN, dev); 2113 err = notifier_to_errno(err); 2114 if (err) { 2115 dev->tx_queue_len = orig_len; 2116 goto errout; 2117 } 2118 status |= DO_SETLINK_NOTIFY; 2119 } 2120 } 2121 2122 if (tb[IFLA_OPERSTATE]) 2123 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2124 2125 if (tb[IFLA_LINKMODE]) { 2126 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2127 2128 write_lock_bh(&dev_base_lock); 2129 if (dev->link_mode ^ value) 2130 status |= DO_SETLINK_NOTIFY; 2131 dev->link_mode = value; 2132 write_unlock_bh(&dev_base_lock); 2133 } 2134 2135 if (tb[IFLA_VFINFO_LIST]) { 2136 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2137 struct nlattr *attr; 2138 int rem; 2139 2140 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2141 if (nla_type(attr) != IFLA_VF_INFO || 2142 nla_len(attr) < NLA_HDRLEN) { 2143 err = -EINVAL; 2144 goto errout; 2145 } 2146 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr, 2147 ifla_vf_policy); 2148 if (err < 0) 2149 goto errout; 2150 err = do_setvfinfo(dev, vfinfo); 2151 if (err < 0) 2152 goto errout; 2153 status |= DO_SETLINK_NOTIFY; 2154 } 2155 } 2156 err = 0; 2157 2158 if (tb[IFLA_VF_PORTS]) { 2159 struct nlattr *port[IFLA_PORT_MAX+1]; 2160 struct nlattr *attr; 2161 int vf; 2162 int rem; 2163 2164 err = -EOPNOTSUPP; 2165 if (!ops->ndo_set_vf_port) 2166 goto errout; 2167 2168 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2169 if (nla_type(attr) != IFLA_VF_PORT || 2170 nla_len(attr) < NLA_HDRLEN) { 2171 err = -EINVAL; 2172 goto errout; 2173 } 2174 err = nla_parse_nested(port, IFLA_PORT_MAX, attr, 2175 ifla_port_policy); 2176 if (err < 0) 2177 goto errout; 2178 if (!port[IFLA_PORT_VF]) { 2179 err = -EOPNOTSUPP; 2180 goto errout; 2181 } 2182 vf = nla_get_u32(port[IFLA_PORT_VF]); 2183 err = ops->ndo_set_vf_port(dev, vf, port); 2184 if (err < 0) 2185 goto errout; 2186 status |= DO_SETLINK_NOTIFY; 2187 } 2188 } 2189 err = 0; 2190 2191 if (tb[IFLA_PORT_SELF]) { 2192 struct nlattr *port[IFLA_PORT_MAX+1]; 2193 2194 err = nla_parse_nested(port, IFLA_PORT_MAX, 2195 tb[IFLA_PORT_SELF], ifla_port_policy); 2196 if (err < 0) 2197 goto errout; 2198 2199 err = -EOPNOTSUPP; 2200 if (ops->ndo_set_vf_port) 2201 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2202 if (err < 0) 2203 goto errout; 2204 status |= DO_SETLINK_NOTIFY; 2205 } 2206 2207 if (tb[IFLA_AF_SPEC]) { 2208 struct nlattr *af; 2209 int rem; 2210 2211 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2212 const struct rtnl_af_ops *af_ops; 2213 2214 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 2215 BUG(); 2216 2217 err = af_ops->set_link_af(dev, af); 2218 if (err < 0) 2219 goto errout; 2220 2221 status |= DO_SETLINK_NOTIFY; 2222 } 2223 } 2224 err = 0; 2225 2226 if (tb[IFLA_PROTO_DOWN]) { 2227 err = dev_change_proto_down(dev, 2228 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2229 if (err) 2230 goto errout; 2231 status |= DO_SETLINK_NOTIFY; 2232 } 2233 2234 if (tb[IFLA_XDP]) { 2235 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2236 u32 xdp_flags = 0; 2237 2238 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP], 2239 ifla_xdp_policy); 2240 if (err < 0) 2241 goto errout; 2242 2243 if (xdp[IFLA_XDP_ATTACHED]) { 2244 err = -EINVAL; 2245 goto errout; 2246 } 2247 2248 if (xdp[IFLA_XDP_FLAGS]) { 2249 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2250 if (xdp_flags & ~XDP_FLAGS_MASK) { 2251 err = -EINVAL; 2252 goto errout; 2253 } 2254 } 2255 2256 if (xdp[IFLA_XDP_FD]) { 2257 err = dev_change_xdp_fd(dev, 2258 nla_get_s32(xdp[IFLA_XDP_FD]), 2259 xdp_flags); 2260 if (err) 2261 goto errout; 2262 status |= DO_SETLINK_NOTIFY; 2263 } 2264 } 2265 2266 errout: 2267 if (status & DO_SETLINK_MODIFIED) { 2268 if (status & DO_SETLINK_NOTIFY) 2269 netdev_state_change(dev); 2270 2271 if (err < 0) 2272 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", 2273 dev->name); 2274 } 2275 2276 return err; 2277 } 2278 2279 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh) 2280 { 2281 struct net *net = sock_net(skb->sk); 2282 struct ifinfomsg *ifm; 2283 struct net_device *dev; 2284 int err; 2285 struct nlattr *tb[IFLA_MAX+1]; 2286 char ifname[IFNAMSIZ]; 2287 2288 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 2289 if (err < 0) 2290 goto errout; 2291 2292 if (tb[IFLA_IFNAME]) 2293 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2294 else 2295 ifname[0] = '\0'; 2296 2297 err = -EINVAL; 2298 ifm = nlmsg_data(nlh); 2299 if (ifm->ifi_index > 0) 2300 dev = __dev_get_by_index(net, ifm->ifi_index); 2301 else if (tb[IFLA_IFNAME]) 2302 dev = __dev_get_by_name(net, ifname); 2303 else 2304 goto errout; 2305 2306 if (dev == NULL) { 2307 err = -ENODEV; 2308 goto errout; 2309 } 2310 2311 err = validate_linkmsg(dev, tb); 2312 if (err < 0) 2313 goto errout; 2314 2315 err = do_setlink(skb, dev, ifm, tb, ifname, 0); 2316 errout: 2317 return err; 2318 } 2319 2320 static int rtnl_group_dellink(const struct net *net, int group) 2321 { 2322 struct net_device *dev, *aux; 2323 LIST_HEAD(list_kill); 2324 bool found = false; 2325 2326 if (!group) 2327 return -EPERM; 2328 2329 for_each_netdev(net, dev) { 2330 if (dev->group == group) { 2331 const struct rtnl_link_ops *ops; 2332 2333 found = true; 2334 ops = dev->rtnl_link_ops; 2335 if (!ops || !ops->dellink) 2336 return -EOPNOTSUPP; 2337 } 2338 } 2339 2340 if (!found) 2341 return -ENODEV; 2342 2343 for_each_netdev_safe(net, dev, aux) { 2344 if (dev->group == group) { 2345 const struct rtnl_link_ops *ops; 2346 2347 ops = dev->rtnl_link_ops; 2348 ops->dellink(dev, &list_kill); 2349 } 2350 } 2351 unregister_netdevice_many(&list_kill); 2352 2353 return 0; 2354 } 2355 2356 int rtnl_delete_link(struct net_device *dev) 2357 { 2358 const struct rtnl_link_ops *ops; 2359 LIST_HEAD(list_kill); 2360 2361 ops = dev->rtnl_link_ops; 2362 if (!ops || !ops->dellink) 2363 return -EOPNOTSUPP; 2364 2365 ops->dellink(dev, &list_kill); 2366 unregister_netdevice_many(&list_kill); 2367 2368 return 0; 2369 } 2370 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2371 2372 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh) 2373 { 2374 struct net *net = sock_net(skb->sk); 2375 struct net_device *dev; 2376 struct ifinfomsg *ifm; 2377 char ifname[IFNAMSIZ]; 2378 struct nlattr *tb[IFLA_MAX+1]; 2379 int err; 2380 2381 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 2382 if (err < 0) 2383 return err; 2384 2385 if (tb[IFLA_IFNAME]) 2386 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2387 2388 ifm = nlmsg_data(nlh); 2389 if (ifm->ifi_index > 0) 2390 dev = __dev_get_by_index(net, ifm->ifi_index); 2391 else if (tb[IFLA_IFNAME]) 2392 dev = __dev_get_by_name(net, ifname); 2393 else if (tb[IFLA_GROUP]) 2394 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP])); 2395 else 2396 return -EINVAL; 2397 2398 if (!dev) 2399 return -ENODEV; 2400 2401 return rtnl_delete_link(dev); 2402 } 2403 2404 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2405 { 2406 unsigned int old_flags; 2407 int err; 2408 2409 old_flags = dev->flags; 2410 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2411 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2412 if (err < 0) 2413 return err; 2414 } 2415 2416 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2417 2418 __dev_notify_flags(dev, old_flags, ~0U); 2419 return 0; 2420 } 2421 EXPORT_SYMBOL(rtnl_configure_link); 2422 2423 struct net_device *rtnl_create_link(struct net *net, 2424 const char *ifname, unsigned char name_assign_type, 2425 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 2426 { 2427 struct net_device *dev; 2428 unsigned int num_tx_queues = 1; 2429 unsigned int num_rx_queues = 1; 2430 2431 if (tb[IFLA_NUM_TX_QUEUES]) 2432 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2433 else if (ops->get_num_tx_queues) 2434 num_tx_queues = ops->get_num_tx_queues(); 2435 2436 if (tb[IFLA_NUM_RX_QUEUES]) 2437 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2438 else if (ops->get_num_rx_queues) 2439 num_rx_queues = ops->get_num_rx_queues(); 2440 2441 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2442 ops->setup, num_tx_queues, num_rx_queues); 2443 if (!dev) 2444 return ERR_PTR(-ENOMEM); 2445 2446 dev_net_set(dev, net); 2447 dev->rtnl_link_ops = ops; 2448 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2449 2450 if (tb[IFLA_MTU]) 2451 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2452 if (tb[IFLA_ADDRESS]) { 2453 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2454 nla_len(tb[IFLA_ADDRESS])); 2455 dev->addr_assign_type = NET_ADDR_SET; 2456 } 2457 if (tb[IFLA_BROADCAST]) 2458 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2459 nla_len(tb[IFLA_BROADCAST])); 2460 if (tb[IFLA_TXQLEN]) 2461 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2462 if (tb[IFLA_OPERSTATE]) 2463 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2464 if (tb[IFLA_LINKMODE]) 2465 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2466 if (tb[IFLA_GROUP]) 2467 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2468 2469 return dev; 2470 } 2471 EXPORT_SYMBOL(rtnl_create_link); 2472 2473 static int rtnl_group_changelink(const struct sk_buff *skb, 2474 struct net *net, int group, 2475 struct ifinfomsg *ifm, 2476 struct nlattr **tb) 2477 { 2478 struct net_device *dev, *aux; 2479 int err; 2480 2481 for_each_netdev_safe(net, dev, aux) { 2482 if (dev->group == group) { 2483 err = do_setlink(skb, dev, ifm, tb, NULL, 0); 2484 if (err < 0) 2485 return err; 2486 } 2487 } 2488 2489 return 0; 2490 } 2491 2492 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh) 2493 { 2494 struct net *net = sock_net(skb->sk); 2495 const struct rtnl_link_ops *ops; 2496 const struct rtnl_link_ops *m_ops = NULL; 2497 struct net_device *dev; 2498 struct net_device *master_dev = NULL; 2499 struct ifinfomsg *ifm; 2500 char kind[MODULE_NAME_LEN]; 2501 char ifname[IFNAMSIZ]; 2502 struct nlattr *tb[IFLA_MAX+1]; 2503 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 2504 unsigned char name_assign_type = NET_NAME_USER; 2505 int err; 2506 2507 #ifdef CONFIG_MODULES 2508 replay: 2509 #endif 2510 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 2511 if (err < 0) 2512 return err; 2513 2514 if (tb[IFLA_IFNAME]) 2515 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2516 else 2517 ifname[0] = '\0'; 2518 2519 ifm = nlmsg_data(nlh); 2520 if (ifm->ifi_index > 0) 2521 dev = __dev_get_by_index(net, ifm->ifi_index); 2522 else { 2523 if (ifname[0]) 2524 dev = __dev_get_by_name(net, ifname); 2525 else 2526 dev = NULL; 2527 } 2528 2529 if (dev) { 2530 master_dev = netdev_master_upper_dev_get(dev); 2531 if (master_dev) 2532 m_ops = master_dev->rtnl_link_ops; 2533 } 2534 2535 err = validate_linkmsg(dev, tb); 2536 if (err < 0) 2537 return err; 2538 2539 if (tb[IFLA_LINKINFO]) { 2540 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 2541 tb[IFLA_LINKINFO], ifla_info_policy); 2542 if (err < 0) 2543 return err; 2544 } else 2545 memset(linkinfo, 0, sizeof(linkinfo)); 2546 2547 if (linkinfo[IFLA_INFO_KIND]) { 2548 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 2549 ops = rtnl_link_ops_get(kind); 2550 } else { 2551 kind[0] = '\0'; 2552 ops = NULL; 2553 } 2554 2555 if (1) { 2556 struct nlattr *attr[ops ? ops->maxtype + 1 : 1]; 2557 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1]; 2558 struct nlattr **data = NULL; 2559 struct nlattr **slave_data = NULL; 2560 struct net *dest_net, *link_net = NULL; 2561 2562 if (ops) { 2563 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 2564 err = nla_parse_nested(attr, ops->maxtype, 2565 linkinfo[IFLA_INFO_DATA], 2566 ops->policy); 2567 if (err < 0) 2568 return err; 2569 data = attr; 2570 } 2571 if (ops->validate) { 2572 err = ops->validate(tb, data); 2573 if (err < 0) 2574 return err; 2575 } 2576 } 2577 2578 if (m_ops) { 2579 if (m_ops->slave_maxtype && 2580 linkinfo[IFLA_INFO_SLAVE_DATA]) { 2581 err = nla_parse_nested(slave_attr, 2582 m_ops->slave_maxtype, 2583 linkinfo[IFLA_INFO_SLAVE_DATA], 2584 m_ops->slave_policy); 2585 if (err < 0) 2586 return err; 2587 slave_data = slave_attr; 2588 } 2589 if (m_ops->slave_validate) { 2590 err = m_ops->slave_validate(tb, slave_data); 2591 if (err < 0) 2592 return err; 2593 } 2594 } 2595 2596 if (dev) { 2597 int status = 0; 2598 2599 if (nlh->nlmsg_flags & NLM_F_EXCL) 2600 return -EEXIST; 2601 if (nlh->nlmsg_flags & NLM_F_REPLACE) 2602 return -EOPNOTSUPP; 2603 2604 if (linkinfo[IFLA_INFO_DATA]) { 2605 if (!ops || ops != dev->rtnl_link_ops || 2606 !ops->changelink) 2607 return -EOPNOTSUPP; 2608 2609 err = ops->changelink(dev, tb, data); 2610 if (err < 0) 2611 return err; 2612 status |= DO_SETLINK_NOTIFY; 2613 } 2614 2615 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 2616 if (!m_ops || !m_ops->slave_changelink) 2617 return -EOPNOTSUPP; 2618 2619 err = m_ops->slave_changelink(master_dev, dev, 2620 tb, slave_data); 2621 if (err < 0) 2622 return err; 2623 status |= DO_SETLINK_NOTIFY; 2624 } 2625 2626 return do_setlink(skb, dev, ifm, tb, ifname, status); 2627 } 2628 2629 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 2630 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 2631 return rtnl_group_changelink(skb, net, 2632 nla_get_u32(tb[IFLA_GROUP]), 2633 ifm, tb); 2634 return -ENODEV; 2635 } 2636 2637 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 2638 return -EOPNOTSUPP; 2639 2640 if (!ops) { 2641 #ifdef CONFIG_MODULES 2642 if (kind[0]) { 2643 __rtnl_unlock(); 2644 request_module("rtnl-link-%s", kind); 2645 rtnl_lock(); 2646 ops = rtnl_link_ops_get(kind); 2647 if (ops) 2648 goto replay; 2649 } 2650 #endif 2651 return -EOPNOTSUPP; 2652 } 2653 2654 if (!ops->setup) 2655 return -EOPNOTSUPP; 2656 2657 if (!ifname[0]) { 2658 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 2659 name_assign_type = NET_NAME_ENUM; 2660 } 2661 2662 dest_net = rtnl_link_get_net(net, tb); 2663 if (IS_ERR(dest_net)) 2664 return PTR_ERR(dest_net); 2665 2666 err = -EPERM; 2667 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN)) 2668 goto out; 2669 2670 if (tb[IFLA_LINK_NETNSID]) { 2671 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 2672 2673 link_net = get_net_ns_by_id(dest_net, id); 2674 if (!link_net) { 2675 err = -EINVAL; 2676 goto out; 2677 } 2678 err = -EPERM; 2679 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 2680 goto out; 2681 } 2682 2683 dev = rtnl_create_link(link_net ? : dest_net, ifname, 2684 name_assign_type, ops, tb); 2685 if (IS_ERR(dev)) { 2686 err = PTR_ERR(dev); 2687 goto out; 2688 } 2689 2690 dev->ifindex = ifm->ifi_index; 2691 2692 if (ops->newlink) { 2693 err = ops->newlink(link_net ? : net, dev, tb, data); 2694 /* Drivers should call free_netdev() in ->destructor 2695 * and unregister it on failure after registration 2696 * so that device could be finally freed in rtnl_unlock. 2697 */ 2698 if (err < 0) { 2699 /* If device is not registered at all, free it now */ 2700 if (dev->reg_state == NETREG_UNINITIALIZED) 2701 free_netdev(dev); 2702 goto out; 2703 } 2704 } else { 2705 err = register_netdevice(dev); 2706 if (err < 0) { 2707 free_netdev(dev); 2708 goto out; 2709 } 2710 } 2711 err = rtnl_configure_link(dev, ifm); 2712 if (err < 0) 2713 goto out_unregister; 2714 if (link_net) { 2715 err = dev_change_net_namespace(dev, dest_net, ifname); 2716 if (err < 0) 2717 goto out_unregister; 2718 } 2719 if (tb[IFLA_MASTER]) { 2720 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2721 if (err) 2722 goto out_unregister; 2723 } 2724 out: 2725 if (link_net) 2726 put_net(link_net); 2727 put_net(dest_net); 2728 return err; 2729 out_unregister: 2730 if (ops->newlink) { 2731 LIST_HEAD(list_kill); 2732 2733 ops->dellink(dev, &list_kill); 2734 unregister_netdevice_many(&list_kill); 2735 } else { 2736 unregister_netdevice(dev); 2737 } 2738 goto out; 2739 } 2740 } 2741 2742 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh) 2743 { 2744 struct net *net = sock_net(skb->sk); 2745 struct ifinfomsg *ifm; 2746 char ifname[IFNAMSIZ]; 2747 struct nlattr *tb[IFLA_MAX+1]; 2748 struct net_device *dev = NULL; 2749 struct sk_buff *nskb; 2750 int err; 2751 u32 ext_filter_mask = 0; 2752 2753 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 2754 if (err < 0) 2755 return err; 2756 2757 if (tb[IFLA_IFNAME]) 2758 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2759 2760 if (tb[IFLA_EXT_MASK]) 2761 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2762 2763 ifm = nlmsg_data(nlh); 2764 if (ifm->ifi_index > 0) 2765 dev = __dev_get_by_index(net, ifm->ifi_index); 2766 else if (tb[IFLA_IFNAME]) 2767 dev = __dev_get_by_name(net, ifname); 2768 else 2769 return -EINVAL; 2770 2771 if (dev == NULL) 2772 return -ENODEV; 2773 2774 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 2775 if (nskb == NULL) 2776 return -ENOBUFS; 2777 2778 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid, 2779 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0); 2780 if (err < 0) { 2781 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 2782 WARN_ON(err == -EMSGSIZE); 2783 kfree_skb(nskb); 2784 } else 2785 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 2786 2787 return err; 2788 } 2789 2790 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 2791 { 2792 struct net *net = sock_net(skb->sk); 2793 struct net_device *dev; 2794 struct nlattr *tb[IFLA_MAX+1]; 2795 u32 ext_filter_mask = 0; 2796 u16 min_ifinfo_dump_size = 0; 2797 int hdrlen; 2798 2799 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 2800 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 2801 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 2802 2803 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) { 2804 if (tb[IFLA_EXT_MASK]) 2805 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2806 } 2807 2808 if (!ext_filter_mask) 2809 return NLMSG_GOODSIZE; 2810 /* 2811 * traverse the list of net devices and compute the minimum 2812 * buffer size based upon the filter mask. 2813 */ 2814 list_for_each_entry(dev, &net->dev_base_head, dev_list) { 2815 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 2816 if_nlmsg_size(dev, 2817 ext_filter_mask)); 2818 } 2819 2820 return nlmsg_total_size(min_ifinfo_dump_size); 2821 } 2822 2823 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 2824 { 2825 int idx; 2826 int s_idx = cb->family; 2827 2828 if (s_idx == 0) 2829 s_idx = 1; 2830 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 2831 int type = cb->nlh->nlmsg_type-RTM_BASE; 2832 if (idx < s_idx || idx == PF_PACKET) 2833 continue; 2834 if (rtnl_msg_handlers[idx] == NULL || 2835 rtnl_msg_handlers[idx][type].dumpit == NULL) 2836 continue; 2837 if (idx > s_idx) { 2838 memset(&cb->args[0], 0, sizeof(cb->args)); 2839 cb->prev_seq = 0; 2840 cb->seq = 0; 2841 } 2842 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 2843 break; 2844 } 2845 cb->family = idx; 2846 2847 return skb->len; 2848 } 2849 2850 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 2851 unsigned int change, 2852 unsigned long event, gfp_t flags) 2853 { 2854 struct net *net = dev_net(dev); 2855 struct sk_buff *skb; 2856 int err = -ENOBUFS; 2857 size_t if_info_size; 2858 2859 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 2860 if (skb == NULL) 2861 goto errout; 2862 2863 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event); 2864 if (err < 0) { 2865 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 2866 WARN_ON(err == -EMSGSIZE); 2867 kfree_skb(skb); 2868 goto errout; 2869 } 2870 return skb; 2871 errout: 2872 if (err < 0) 2873 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2874 return NULL; 2875 } 2876 2877 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 2878 { 2879 struct net *net = dev_net(dev); 2880 2881 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 2882 } 2883 2884 static void rtmsg_ifinfo_event(int type, struct net_device *dev, 2885 unsigned int change, unsigned long event, 2886 gfp_t flags) 2887 { 2888 struct sk_buff *skb; 2889 2890 if (dev->reg_state != NETREG_REGISTERED) 2891 return; 2892 2893 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags); 2894 if (skb) 2895 rtmsg_ifinfo_send(skb, dev, flags); 2896 } 2897 2898 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 2899 gfp_t flags) 2900 { 2901 rtmsg_ifinfo_event(type, dev, change, 0, flags); 2902 } 2903 EXPORT_SYMBOL(rtmsg_ifinfo); 2904 2905 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 2906 struct net_device *dev, 2907 u8 *addr, u16 vid, u32 pid, u32 seq, 2908 int type, unsigned int flags, 2909 int nlflags, u16 ndm_state) 2910 { 2911 struct nlmsghdr *nlh; 2912 struct ndmsg *ndm; 2913 2914 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 2915 if (!nlh) 2916 return -EMSGSIZE; 2917 2918 ndm = nlmsg_data(nlh); 2919 ndm->ndm_family = AF_BRIDGE; 2920 ndm->ndm_pad1 = 0; 2921 ndm->ndm_pad2 = 0; 2922 ndm->ndm_flags = flags; 2923 ndm->ndm_type = 0; 2924 ndm->ndm_ifindex = dev->ifindex; 2925 ndm->ndm_state = ndm_state; 2926 2927 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 2928 goto nla_put_failure; 2929 if (vid) 2930 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 2931 goto nla_put_failure; 2932 2933 nlmsg_end(skb, nlh); 2934 return 0; 2935 2936 nla_put_failure: 2937 nlmsg_cancel(skb, nlh); 2938 return -EMSGSIZE; 2939 } 2940 2941 static inline size_t rtnl_fdb_nlmsg_size(void) 2942 { 2943 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 2944 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 2945 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 2946 0; 2947 } 2948 2949 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 2950 u16 ndm_state) 2951 { 2952 struct net *net = dev_net(dev); 2953 struct sk_buff *skb; 2954 int err = -ENOBUFS; 2955 2956 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 2957 if (!skb) 2958 goto errout; 2959 2960 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 2961 0, 0, type, NTF_SELF, 0, ndm_state); 2962 if (err < 0) { 2963 kfree_skb(skb); 2964 goto errout; 2965 } 2966 2967 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 2968 return; 2969 errout: 2970 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 2971 } 2972 2973 /** 2974 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 2975 */ 2976 int ndo_dflt_fdb_add(struct ndmsg *ndm, 2977 struct nlattr *tb[], 2978 struct net_device *dev, 2979 const unsigned char *addr, u16 vid, 2980 u16 flags) 2981 { 2982 int err = -EINVAL; 2983 2984 /* If aging addresses are supported device will need to 2985 * implement its own handler for this. 2986 */ 2987 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 2988 pr_info("%s: FDB only supports static addresses\n", dev->name); 2989 return err; 2990 } 2991 2992 if (vid) { 2993 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 2994 return err; 2995 } 2996 2997 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 2998 err = dev_uc_add_excl(dev, addr); 2999 else if (is_multicast_ether_addr(addr)) 3000 err = dev_mc_add_excl(dev, addr); 3001 3002 /* Only return duplicate errors if NLM_F_EXCL is set */ 3003 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 3004 err = 0; 3005 3006 return err; 3007 } 3008 EXPORT_SYMBOL(ndo_dflt_fdb_add); 3009 3010 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid) 3011 { 3012 u16 vid = 0; 3013 3014 if (vlan_attr) { 3015 if (nla_len(vlan_attr) != sizeof(u16)) { 3016 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n"); 3017 return -EINVAL; 3018 } 3019 3020 vid = nla_get_u16(vlan_attr); 3021 3022 if (!vid || vid >= VLAN_VID_MASK) { 3023 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n", 3024 vid); 3025 return -EINVAL; 3026 } 3027 } 3028 *p_vid = vid; 3029 return 0; 3030 } 3031 3032 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh) 3033 { 3034 struct net *net = sock_net(skb->sk); 3035 struct ndmsg *ndm; 3036 struct nlattr *tb[NDA_MAX+1]; 3037 struct net_device *dev; 3038 u8 *addr; 3039 u16 vid; 3040 int err; 3041 3042 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL); 3043 if (err < 0) 3044 return err; 3045 3046 ndm = nlmsg_data(nlh); 3047 if (ndm->ndm_ifindex == 0) { 3048 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n"); 3049 return -EINVAL; 3050 } 3051 3052 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3053 if (dev == NULL) { 3054 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n"); 3055 return -ENODEV; 3056 } 3057 3058 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3059 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n"); 3060 return -EINVAL; 3061 } 3062 3063 addr = nla_data(tb[NDA_LLADDR]); 3064 3065 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3066 if (err) 3067 return err; 3068 3069 err = -EOPNOTSUPP; 3070 3071 /* Support fdb on master device the net/bridge default case */ 3072 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3073 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3074 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3075 const struct net_device_ops *ops = br_dev->netdev_ops; 3076 3077 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3078 nlh->nlmsg_flags); 3079 if (err) 3080 goto out; 3081 else 3082 ndm->ndm_flags &= ~NTF_MASTER; 3083 } 3084 3085 /* Embedded bridge, macvlan, and any other device support */ 3086 if ((ndm->ndm_flags & NTF_SELF)) { 3087 if (dev->netdev_ops->ndo_fdb_add) 3088 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3089 vid, 3090 nlh->nlmsg_flags); 3091 else 3092 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3093 nlh->nlmsg_flags); 3094 3095 if (!err) { 3096 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3097 ndm->ndm_state); 3098 ndm->ndm_flags &= ~NTF_SELF; 3099 } 3100 } 3101 out: 3102 return err; 3103 } 3104 3105 /** 3106 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3107 */ 3108 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3109 struct nlattr *tb[], 3110 struct net_device *dev, 3111 const unsigned char *addr, u16 vid) 3112 { 3113 int err = -EINVAL; 3114 3115 /* If aging addresses are supported device will need to 3116 * implement its own handler for this. 3117 */ 3118 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3119 pr_info("%s: FDB only supports static addresses\n", dev->name); 3120 return err; 3121 } 3122 3123 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3124 err = dev_uc_del(dev, addr); 3125 else if (is_multicast_ether_addr(addr)) 3126 err = dev_mc_del(dev, addr); 3127 3128 return err; 3129 } 3130 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3131 3132 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh) 3133 { 3134 struct net *net = sock_net(skb->sk); 3135 struct ndmsg *ndm; 3136 struct nlattr *tb[NDA_MAX+1]; 3137 struct net_device *dev; 3138 int err = -EINVAL; 3139 __u8 *addr; 3140 u16 vid; 3141 3142 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3143 return -EPERM; 3144 3145 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL); 3146 if (err < 0) 3147 return err; 3148 3149 ndm = nlmsg_data(nlh); 3150 if (ndm->ndm_ifindex == 0) { 3151 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n"); 3152 return -EINVAL; 3153 } 3154 3155 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3156 if (dev == NULL) { 3157 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n"); 3158 return -ENODEV; 3159 } 3160 3161 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3162 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n"); 3163 return -EINVAL; 3164 } 3165 3166 addr = nla_data(tb[NDA_LLADDR]); 3167 3168 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3169 if (err) 3170 return err; 3171 3172 err = -EOPNOTSUPP; 3173 3174 /* Support fdb on master device the net/bridge default case */ 3175 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3176 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3177 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3178 const struct net_device_ops *ops = br_dev->netdev_ops; 3179 3180 if (ops->ndo_fdb_del) 3181 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3182 3183 if (err) 3184 goto out; 3185 else 3186 ndm->ndm_flags &= ~NTF_MASTER; 3187 } 3188 3189 /* Embedded bridge, macvlan, and any other device support */ 3190 if (ndm->ndm_flags & NTF_SELF) { 3191 if (dev->netdev_ops->ndo_fdb_del) 3192 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3193 vid); 3194 else 3195 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3196 3197 if (!err) { 3198 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3199 ndm->ndm_state); 3200 ndm->ndm_flags &= ~NTF_SELF; 3201 } 3202 } 3203 out: 3204 return err; 3205 } 3206 3207 static int nlmsg_populate_fdb(struct sk_buff *skb, 3208 struct netlink_callback *cb, 3209 struct net_device *dev, 3210 int *idx, 3211 struct netdev_hw_addr_list *list) 3212 { 3213 struct netdev_hw_addr *ha; 3214 int err; 3215 u32 portid, seq; 3216 3217 portid = NETLINK_CB(cb->skb).portid; 3218 seq = cb->nlh->nlmsg_seq; 3219 3220 list_for_each_entry(ha, &list->list, list) { 3221 if (*idx < cb->args[2]) 3222 goto skip; 3223 3224 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3225 portid, seq, 3226 RTM_NEWNEIGH, NTF_SELF, 3227 NLM_F_MULTI, NUD_PERMANENT); 3228 if (err < 0) 3229 return err; 3230 skip: 3231 *idx += 1; 3232 } 3233 return 0; 3234 } 3235 3236 /** 3237 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3238 * @nlh: netlink message header 3239 * @dev: netdevice 3240 * 3241 * Default netdevice operation to dump the existing unicast address list. 3242 * Returns number of addresses from list put in skb. 3243 */ 3244 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3245 struct netlink_callback *cb, 3246 struct net_device *dev, 3247 struct net_device *filter_dev, 3248 int *idx) 3249 { 3250 int err; 3251 3252 netif_addr_lock_bh(dev); 3253 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3254 if (err) 3255 goto out; 3256 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3257 out: 3258 netif_addr_unlock_bh(dev); 3259 return err; 3260 } 3261 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3262 3263 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 3264 { 3265 struct net_device *dev; 3266 struct nlattr *tb[IFLA_MAX+1]; 3267 struct net_device *br_dev = NULL; 3268 const struct net_device_ops *ops = NULL; 3269 const struct net_device_ops *cops = NULL; 3270 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 3271 struct net *net = sock_net(skb->sk); 3272 struct hlist_head *head; 3273 int brport_idx = 0; 3274 int br_idx = 0; 3275 int h, s_h; 3276 int idx = 0, s_idx; 3277 int err = 0; 3278 int fidx = 0; 3279 3280 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX, 3281 ifla_policy) == 0) { 3282 if (tb[IFLA_MASTER]) 3283 br_idx = nla_get_u32(tb[IFLA_MASTER]); 3284 } 3285 3286 brport_idx = ifm->ifi_index; 3287 3288 if (br_idx) { 3289 br_dev = __dev_get_by_index(net, br_idx); 3290 if (!br_dev) 3291 return -ENODEV; 3292 3293 ops = br_dev->netdev_ops; 3294 } 3295 3296 s_h = cb->args[0]; 3297 s_idx = cb->args[1]; 3298 3299 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3300 idx = 0; 3301 head = &net->dev_index_head[h]; 3302 hlist_for_each_entry(dev, head, index_hlist) { 3303 3304 if (brport_idx && (dev->ifindex != brport_idx)) 3305 continue; 3306 3307 if (!br_idx) { /* user did not specify a specific bridge */ 3308 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3309 br_dev = netdev_master_upper_dev_get(dev); 3310 cops = br_dev->netdev_ops; 3311 } 3312 } else { 3313 if (dev != br_dev && 3314 !(dev->priv_flags & IFF_BRIDGE_PORT)) 3315 continue; 3316 3317 if (br_dev != netdev_master_upper_dev_get(dev) && 3318 !(dev->priv_flags & IFF_EBRIDGE)) 3319 continue; 3320 cops = ops; 3321 } 3322 3323 if (idx < s_idx) 3324 goto cont; 3325 3326 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3327 if (cops && cops->ndo_fdb_dump) { 3328 err = cops->ndo_fdb_dump(skb, cb, 3329 br_dev, dev, 3330 &fidx); 3331 if (err == -EMSGSIZE) 3332 goto out; 3333 } 3334 } 3335 3336 if (dev->netdev_ops->ndo_fdb_dump) 3337 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 3338 dev, NULL, 3339 &fidx); 3340 else 3341 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 3342 &fidx); 3343 if (err == -EMSGSIZE) 3344 goto out; 3345 3346 cops = NULL; 3347 3348 /* reset fdb offset to 0 for rest of the interfaces */ 3349 cb->args[2] = 0; 3350 fidx = 0; 3351 cont: 3352 idx++; 3353 } 3354 } 3355 3356 out: 3357 cb->args[0] = h; 3358 cb->args[1] = idx; 3359 cb->args[2] = fidx; 3360 3361 return skb->len; 3362 } 3363 3364 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 3365 unsigned int attrnum, unsigned int flag) 3366 { 3367 if (mask & flag) 3368 return nla_put_u8(skb, attrnum, !!(flags & flag)); 3369 return 0; 3370 } 3371 3372 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 3373 struct net_device *dev, u16 mode, 3374 u32 flags, u32 mask, int nlflags, 3375 u32 filter_mask, 3376 int (*vlan_fill)(struct sk_buff *skb, 3377 struct net_device *dev, 3378 u32 filter_mask)) 3379 { 3380 struct nlmsghdr *nlh; 3381 struct ifinfomsg *ifm; 3382 struct nlattr *br_afspec; 3383 struct nlattr *protinfo; 3384 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 3385 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3386 int err = 0; 3387 3388 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 3389 if (nlh == NULL) 3390 return -EMSGSIZE; 3391 3392 ifm = nlmsg_data(nlh); 3393 ifm->ifi_family = AF_BRIDGE; 3394 ifm->__ifi_pad = 0; 3395 ifm->ifi_type = dev->type; 3396 ifm->ifi_index = dev->ifindex; 3397 ifm->ifi_flags = dev_get_flags(dev); 3398 ifm->ifi_change = 0; 3399 3400 3401 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 3402 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 3403 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 3404 (br_dev && 3405 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 3406 (dev->addr_len && 3407 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 3408 (dev->ifindex != dev_get_iflink(dev) && 3409 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 3410 goto nla_put_failure; 3411 3412 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 3413 if (!br_afspec) 3414 goto nla_put_failure; 3415 3416 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 3417 nla_nest_cancel(skb, br_afspec); 3418 goto nla_put_failure; 3419 } 3420 3421 if (mode != BRIDGE_MODE_UNDEF) { 3422 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 3423 nla_nest_cancel(skb, br_afspec); 3424 goto nla_put_failure; 3425 } 3426 } 3427 if (vlan_fill) { 3428 err = vlan_fill(skb, dev, filter_mask); 3429 if (err) { 3430 nla_nest_cancel(skb, br_afspec); 3431 goto nla_put_failure; 3432 } 3433 } 3434 nla_nest_end(skb, br_afspec); 3435 3436 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED); 3437 if (!protinfo) 3438 goto nla_put_failure; 3439 3440 if (brport_nla_put_flag(skb, flags, mask, 3441 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 3442 brport_nla_put_flag(skb, flags, mask, 3443 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 3444 brport_nla_put_flag(skb, flags, mask, 3445 IFLA_BRPORT_FAST_LEAVE, 3446 BR_MULTICAST_FAST_LEAVE) || 3447 brport_nla_put_flag(skb, flags, mask, 3448 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 3449 brport_nla_put_flag(skb, flags, mask, 3450 IFLA_BRPORT_LEARNING, BR_LEARNING) || 3451 brport_nla_put_flag(skb, flags, mask, 3452 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 3453 brport_nla_put_flag(skb, flags, mask, 3454 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 3455 brport_nla_put_flag(skb, flags, mask, 3456 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 3457 nla_nest_cancel(skb, protinfo); 3458 goto nla_put_failure; 3459 } 3460 3461 nla_nest_end(skb, protinfo); 3462 3463 nlmsg_end(skb, nlh); 3464 return 0; 3465 nla_put_failure: 3466 nlmsg_cancel(skb, nlh); 3467 return err ? err : -EMSGSIZE; 3468 } 3469 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 3470 3471 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 3472 { 3473 struct net *net = sock_net(skb->sk); 3474 struct net_device *dev; 3475 int idx = 0; 3476 u32 portid = NETLINK_CB(cb->skb).portid; 3477 u32 seq = cb->nlh->nlmsg_seq; 3478 u32 filter_mask = 0; 3479 int err; 3480 3481 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 3482 struct nlattr *extfilt; 3483 3484 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 3485 IFLA_EXT_MASK); 3486 if (extfilt) { 3487 if (nla_len(extfilt) < sizeof(filter_mask)) 3488 return -EINVAL; 3489 3490 filter_mask = nla_get_u32(extfilt); 3491 } 3492 } 3493 3494 rcu_read_lock(); 3495 for_each_netdev_rcu(net, dev) { 3496 const struct net_device_ops *ops = dev->netdev_ops; 3497 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3498 3499 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 3500 if (idx >= cb->args[0]) { 3501 err = br_dev->netdev_ops->ndo_bridge_getlink( 3502 skb, portid, seq, dev, 3503 filter_mask, NLM_F_MULTI); 3504 if (err < 0 && err != -EOPNOTSUPP) 3505 break; 3506 } 3507 idx++; 3508 } 3509 3510 if (ops->ndo_bridge_getlink) { 3511 if (idx >= cb->args[0]) { 3512 err = ops->ndo_bridge_getlink(skb, portid, 3513 seq, dev, 3514 filter_mask, 3515 NLM_F_MULTI); 3516 if (err < 0 && err != -EOPNOTSUPP) 3517 break; 3518 } 3519 idx++; 3520 } 3521 } 3522 rcu_read_unlock(); 3523 cb->args[0] = idx; 3524 3525 return skb->len; 3526 } 3527 3528 static inline size_t bridge_nlmsg_size(void) 3529 { 3530 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 3531 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 3532 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 3533 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 3534 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 3535 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 3536 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 3537 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 3538 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 3539 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 3540 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 3541 } 3542 3543 static int rtnl_bridge_notify(struct net_device *dev) 3544 { 3545 struct net *net = dev_net(dev); 3546 struct sk_buff *skb; 3547 int err = -EOPNOTSUPP; 3548 3549 if (!dev->netdev_ops->ndo_bridge_getlink) 3550 return 0; 3551 3552 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 3553 if (!skb) { 3554 err = -ENOMEM; 3555 goto errout; 3556 } 3557 3558 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 3559 if (err < 0) 3560 goto errout; 3561 3562 if (!skb->len) 3563 goto errout; 3564 3565 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 3566 return 0; 3567 errout: 3568 WARN_ON(err == -EMSGSIZE); 3569 kfree_skb(skb); 3570 if (err) 3571 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3572 return err; 3573 } 3574 3575 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh) 3576 { 3577 struct net *net = sock_net(skb->sk); 3578 struct ifinfomsg *ifm; 3579 struct net_device *dev; 3580 struct nlattr *br_spec, *attr = NULL; 3581 int rem, err = -EOPNOTSUPP; 3582 u16 flags = 0; 3583 bool have_flags = false; 3584 3585 if (nlmsg_len(nlh) < sizeof(*ifm)) 3586 return -EINVAL; 3587 3588 ifm = nlmsg_data(nlh); 3589 if (ifm->ifi_family != AF_BRIDGE) 3590 return -EPFNOSUPPORT; 3591 3592 dev = __dev_get_by_index(net, ifm->ifi_index); 3593 if (!dev) { 3594 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3595 return -ENODEV; 3596 } 3597 3598 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3599 if (br_spec) { 3600 nla_for_each_nested(attr, br_spec, rem) { 3601 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3602 if (nla_len(attr) < sizeof(flags)) 3603 return -EINVAL; 3604 3605 have_flags = true; 3606 flags = nla_get_u16(attr); 3607 break; 3608 } 3609 } 3610 } 3611 3612 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3613 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3614 3615 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 3616 err = -EOPNOTSUPP; 3617 goto out; 3618 } 3619 3620 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags); 3621 if (err) 3622 goto out; 3623 3624 flags &= ~BRIDGE_FLAGS_MASTER; 3625 } 3626 3627 if ((flags & BRIDGE_FLAGS_SELF)) { 3628 if (!dev->netdev_ops->ndo_bridge_setlink) 3629 err = -EOPNOTSUPP; 3630 else 3631 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 3632 flags); 3633 if (!err) { 3634 flags &= ~BRIDGE_FLAGS_SELF; 3635 3636 /* Generate event to notify upper layer of bridge 3637 * change 3638 */ 3639 err = rtnl_bridge_notify(dev); 3640 } 3641 } 3642 3643 if (have_flags) 3644 memcpy(nla_data(attr), &flags, sizeof(flags)); 3645 out: 3646 return err; 3647 } 3648 3649 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh) 3650 { 3651 struct net *net = sock_net(skb->sk); 3652 struct ifinfomsg *ifm; 3653 struct net_device *dev; 3654 struct nlattr *br_spec, *attr = NULL; 3655 int rem, err = -EOPNOTSUPP; 3656 u16 flags = 0; 3657 bool have_flags = false; 3658 3659 if (nlmsg_len(nlh) < sizeof(*ifm)) 3660 return -EINVAL; 3661 3662 ifm = nlmsg_data(nlh); 3663 if (ifm->ifi_family != AF_BRIDGE) 3664 return -EPFNOSUPPORT; 3665 3666 dev = __dev_get_by_index(net, ifm->ifi_index); 3667 if (!dev) { 3668 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3669 return -ENODEV; 3670 } 3671 3672 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3673 if (br_spec) { 3674 nla_for_each_nested(attr, br_spec, rem) { 3675 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3676 if (nla_len(attr) < sizeof(flags)) 3677 return -EINVAL; 3678 3679 have_flags = true; 3680 flags = nla_get_u16(attr); 3681 break; 3682 } 3683 } 3684 } 3685 3686 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3687 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3688 3689 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 3690 err = -EOPNOTSUPP; 3691 goto out; 3692 } 3693 3694 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 3695 if (err) 3696 goto out; 3697 3698 flags &= ~BRIDGE_FLAGS_MASTER; 3699 } 3700 3701 if ((flags & BRIDGE_FLAGS_SELF)) { 3702 if (!dev->netdev_ops->ndo_bridge_dellink) 3703 err = -EOPNOTSUPP; 3704 else 3705 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 3706 flags); 3707 3708 if (!err) { 3709 flags &= ~BRIDGE_FLAGS_SELF; 3710 3711 /* Generate event to notify upper layer of bridge 3712 * change 3713 */ 3714 err = rtnl_bridge_notify(dev); 3715 } 3716 } 3717 3718 if (have_flags) 3719 memcpy(nla_data(attr), &flags, sizeof(flags)); 3720 out: 3721 return err; 3722 } 3723 3724 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 3725 { 3726 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 3727 (!idxattr || idxattr == attrid); 3728 } 3729 3730 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 3731 static int rtnl_get_offload_stats_attr_size(int attr_id) 3732 { 3733 switch (attr_id) { 3734 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 3735 return sizeof(struct rtnl_link_stats64); 3736 } 3737 3738 return 0; 3739 } 3740 3741 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 3742 int *prividx) 3743 { 3744 struct nlattr *attr = NULL; 3745 int attr_id, size; 3746 void *attr_data; 3747 int err; 3748 3749 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3750 dev->netdev_ops->ndo_get_offload_stats)) 3751 return -ENODATA; 3752 3753 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3754 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3755 if (attr_id < *prividx) 3756 continue; 3757 3758 size = rtnl_get_offload_stats_attr_size(attr_id); 3759 if (!size) 3760 continue; 3761 3762 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3763 continue; 3764 3765 attr = nla_reserve_64bit(skb, attr_id, size, 3766 IFLA_OFFLOAD_XSTATS_UNSPEC); 3767 if (!attr) 3768 goto nla_put_failure; 3769 3770 attr_data = nla_data(attr); 3771 memset(attr_data, 0, size); 3772 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 3773 attr_data); 3774 if (err) 3775 goto get_offload_stats_failure; 3776 } 3777 3778 if (!attr) 3779 return -ENODATA; 3780 3781 *prividx = 0; 3782 return 0; 3783 3784 nla_put_failure: 3785 err = -EMSGSIZE; 3786 get_offload_stats_failure: 3787 *prividx = attr_id; 3788 return err; 3789 } 3790 3791 static int rtnl_get_offload_stats_size(const struct net_device *dev) 3792 { 3793 int nla_size = 0; 3794 int attr_id; 3795 int size; 3796 3797 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3798 dev->netdev_ops->ndo_get_offload_stats)) 3799 return 0; 3800 3801 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3802 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3803 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3804 continue; 3805 size = rtnl_get_offload_stats_attr_size(attr_id); 3806 nla_size += nla_total_size_64bit(size); 3807 } 3808 3809 if (nla_size != 0) 3810 nla_size += nla_total_size(0); 3811 3812 return nla_size; 3813 } 3814 3815 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 3816 int type, u32 pid, u32 seq, u32 change, 3817 unsigned int flags, unsigned int filter_mask, 3818 int *idxattr, int *prividx) 3819 { 3820 struct if_stats_msg *ifsm; 3821 struct nlmsghdr *nlh; 3822 struct nlattr *attr; 3823 int s_prividx = *prividx; 3824 int err; 3825 3826 ASSERT_RTNL(); 3827 3828 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 3829 if (!nlh) 3830 return -EMSGSIZE; 3831 3832 ifsm = nlmsg_data(nlh); 3833 ifsm->ifindex = dev->ifindex; 3834 ifsm->filter_mask = filter_mask; 3835 3836 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 3837 struct rtnl_link_stats64 *sp; 3838 3839 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 3840 sizeof(struct rtnl_link_stats64), 3841 IFLA_STATS_UNSPEC); 3842 if (!attr) 3843 goto nla_put_failure; 3844 3845 sp = nla_data(attr); 3846 dev_get_stats(dev, sp); 3847 } 3848 3849 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 3850 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3851 3852 if (ops && ops->fill_linkxstats) { 3853 *idxattr = IFLA_STATS_LINK_XSTATS; 3854 attr = nla_nest_start(skb, 3855 IFLA_STATS_LINK_XSTATS); 3856 if (!attr) 3857 goto nla_put_failure; 3858 3859 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3860 nla_nest_end(skb, attr); 3861 if (err) 3862 goto nla_put_failure; 3863 *idxattr = 0; 3864 } 3865 } 3866 3867 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 3868 *idxattr)) { 3869 const struct rtnl_link_ops *ops = NULL; 3870 const struct net_device *master; 3871 3872 master = netdev_master_upper_dev_get(dev); 3873 if (master) 3874 ops = master->rtnl_link_ops; 3875 if (ops && ops->fill_linkxstats) { 3876 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 3877 attr = nla_nest_start(skb, 3878 IFLA_STATS_LINK_XSTATS_SLAVE); 3879 if (!attr) 3880 goto nla_put_failure; 3881 3882 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3883 nla_nest_end(skb, attr); 3884 if (err) 3885 goto nla_put_failure; 3886 *idxattr = 0; 3887 } 3888 } 3889 3890 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 3891 *idxattr)) { 3892 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 3893 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS); 3894 if (!attr) 3895 goto nla_put_failure; 3896 3897 err = rtnl_get_offload_stats(skb, dev, prividx); 3898 if (err == -ENODATA) 3899 nla_nest_cancel(skb, attr); 3900 else 3901 nla_nest_end(skb, attr); 3902 3903 if (err && err != -ENODATA) 3904 goto nla_put_failure; 3905 *idxattr = 0; 3906 } 3907 3908 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 3909 struct rtnl_af_ops *af_ops; 3910 3911 *idxattr = IFLA_STATS_AF_SPEC; 3912 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC); 3913 if (!attr) 3914 goto nla_put_failure; 3915 3916 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3917 if (af_ops->fill_stats_af) { 3918 struct nlattr *af; 3919 int err; 3920 3921 af = nla_nest_start(skb, af_ops->family); 3922 if (!af) 3923 goto nla_put_failure; 3924 3925 err = af_ops->fill_stats_af(skb, dev); 3926 3927 if (err == -ENODATA) 3928 nla_nest_cancel(skb, af); 3929 else if (err < 0) 3930 goto nla_put_failure; 3931 3932 nla_nest_end(skb, af); 3933 } 3934 } 3935 3936 nla_nest_end(skb, attr); 3937 3938 *idxattr = 0; 3939 } 3940 3941 nlmsg_end(skb, nlh); 3942 3943 return 0; 3944 3945 nla_put_failure: 3946 /* not a multi message or no progress mean a real error */ 3947 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 3948 nlmsg_cancel(skb, nlh); 3949 else 3950 nlmsg_end(skb, nlh); 3951 3952 return -EMSGSIZE; 3953 } 3954 3955 static size_t if_nlmsg_stats_size(const struct net_device *dev, 3956 u32 filter_mask) 3957 { 3958 size_t size = 0; 3959 3960 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 3961 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 3962 3963 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 3964 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3965 int attr = IFLA_STATS_LINK_XSTATS; 3966 3967 if (ops && ops->get_linkxstats_size) { 3968 size += nla_total_size(ops->get_linkxstats_size(dev, 3969 attr)); 3970 /* for IFLA_STATS_LINK_XSTATS */ 3971 size += nla_total_size(0); 3972 } 3973 } 3974 3975 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 3976 struct net_device *_dev = (struct net_device *)dev; 3977 const struct rtnl_link_ops *ops = NULL; 3978 const struct net_device *master; 3979 3980 /* netdev_master_upper_dev_get can't take const */ 3981 master = netdev_master_upper_dev_get(_dev); 3982 if (master) 3983 ops = master->rtnl_link_ops; 3984 if (ops && ops->get_linkxstats_size) { 3985 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 3986 3987 size += nla_total_size(ops->get_linkxstats_size(dev, 3988 attr)); 3989 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 3990 size += nla_total_size(0); 3991 } 3992 } 3993 3994 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 3995 size += rtnl_get_offload_stats_size(dev); 3996 3997 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 3998 struct rtnl_af_ops *af_ops; 3999 4000 /* for IFLA_STATS_AF_SPEC */ 4001 size += nla_total_size(0); 4002 4003 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 4004 if (af_ops->get_stats_af_size) { 4005 size += nla_total_size( 4006 af_ops->get_stats_af_size(dev)); 4007 4008 /* for AF_* */ 4009 size += nla_total_size(0); 4010 } 4011 } 4012 } 4013 4014 return size; 4015 } 4016 4017 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh) 4018 { 4019 struct net *net = sock_net(skb->sk); 4020 struct net_device *dev = NULL; 4021 int idxattr = 0, prividx = 0; 4022 struct if_stats_msg *ifsm; 4023 struct sk_buff *nskb; 4024 u32 filter_mask; 4025 int err; 4026 4027 if (nlmsg_len(nlh) < sizeof(*ifsm)) 4028 return -EINVAL; 4029 4030 ifsm = nlmsg_data(nlh); 4031 if (ifsm->ifindex > 0) 4032 dev = __dev_get_by_index(net, ifsm->ifindex); 4033 else 4034 return -EINVAL; 4035 4036 if (!dev) 4037 return -ENODEV; 4038 4039 filter_mask = ifsm->filter_mask; 4040 if (!filter_mask) 4041 return -EINVAL; 4042 4043 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 4044 if (!nskb) 4045 return -ENOBUFS; 4046 4047 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 4048 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 4049 0, filter_mask, &idxattr, &prividx); 4050 if (err < 0) { 4051 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 4052 WARN_ON(err == -EMSGSIZE); 4053 kfree_skb(nskb); 4054 } else { 4055 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 4056 } 4057 4058 return err; 4059 } 4060 4061 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 4062 { 4063 int h, s_h, err, s_idx, s_idxattr, s_prividx; 4064 struct net *net = sock_net(skb->sk); 4065 unsigned int flags = NLM_F_MULTI; 4066 struct if_stats_msg *ifsm; 4067 struct hlist_head *head; 4068 struct net_device *dev; 4069 u32 filter_mask = 0; 4070 int idx = 0; 4071 4072 s_h = cb->args[0]; 4073 s_idx = cb->args[1]; 4074 s_idxattr = cb->args[2]; 4075 s_prividx = cb->args[3]; 4076 4077 cb->seq = net->dev_base_seq; 4078 4079 if (nlmsg_len(cb->nlh) < sizeof(*ifsm)) 4080 return -EINVAL; 4081 4082 ifsm = nlmsg_data(cb->nlh); 4083 filter_mask = ifsm->filter_mask; 4084 if (!filter_mask) 4085 return -EINVAL; 4086 4087 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4088 idx = 0; 4089 head = &net->dev_index_head[h]; 4090 hlist_for_each_entry(dev, head, index_hlist) { 4091 if (idx < s_idx) 4092 goto cont; 4093 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 4094 NETLINK_CB(cb->skb).portid, 4095 cb->nlh->nlmsg_seq, 0, 4096 flags, filter_mask, 4097 &s_idxattr, &s_prividx); 4098 /* If we ran out of room on the first message, 4099 * we're in trouble 4100 */ 4101 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 4102 4103 if (err < 0) 4104 goto out; 4105 s_prividx = 0; 4106 s_idxattr = 0; 4107 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4108 cont: 4109 idx++; 4110 } 4111 } 4112 out: 4113 cb->args[3] = s_prividx; 4114 cb->args[2] = s_idxattr; 4115 cb->args[1] = idx; 4116 cb->args[0] = h; 4117 4118 return skb->len; 4119 } 4120 4121 /* Process one rtnetlink message. */ 4122 4123 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 4124 { 4125 struct net *net = sock_net(skb->sk); 4126 rtnl_doit_func doit; 4127 int kind; 4128 int family; 4129 int type; 4130 int err; 4131 4132 type = nlh->nlmsg_type; 4133 if (type > RTM_MAX) 4134 return -EOPNOTSUPP; 4135 4136 type -= RTM_BASE; 4137 4138 /* All the messages must have at least 1 byte length */ 4139 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 4140 return 0; 4141 4142 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 4143 kind = type&3; 4144 4145 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 4146 return -EPERM; 4147 4148 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 4149 struct sock *rtnl; 4150 rtnl_dumpit_func dumpit; 4151 rtnl_calcit_func calcit; 4152 u16 min_dump_alloc = 0; 4153 4154 dumpit = rtnl_get_dumpit(family, type); 4155 if (dumpit == NULL) 4156 return -EOPNOTSUPP; 4157 calcit = rtnl_get_calcit(family, type); 4158 if (calcit) 4159 min_dump_alloc = calcit(skb, nlh); 4160 4161 __rtnl_unlock(); 4162 rtnl = net->rtnl; 4163 { 4164 struct netlink_dump_control c = { 4165 .dump = dumpit, 4166 .min_dump_alloc = min_dump_alloc, 4167 }; 4168 err = netlink_dump_start(rtnl, skb, nlh, &c); 4169 } 4170 rtnl_lock(); 4171 return err; 4172 } 4173 4174 doit = rtnl_get_doit(family, type); 4175 if (doit == NULL) 4176 return -EOPNOTSUPP; 4177 4178 return doit(skb, nlh); 4179 } 4180 4181 static void rtnetlink_rcv(struct sk_buff *skb) 4182 { 4183 rtnl_lock(); 4184 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 4185 rtnl_unlock(); 4186 } 4187 4188 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 4189 { 4190 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4191 4192 switch (event) { 4193 case NETDEV_REBOOT: 4194 case NETDEV_CHANGEMTU: 4195 case NETDEV_CHANGEADDR: 4196 case NETDEV_CHANGENAME: 4197 case NETDEV_FEAT_CHANGE: 4198 case NETDEV_BONDING_FAILOVER: 4199 case NETDEV_POST_TYPE_CHANGE: 4200 case NETDEV_NOTIFY_PEERS: 4201 case NETDEV_CHANGEUPPER: 4202 case NETDEV_RESEND_IGMP: 4203 case NETDEV_PRECHANGEMTU: 4204 case NETDEV_CHANGEINFODATA: 4205 case NETDEV_PRECHANGEUPPER: 4206 case NETDEV_CHANGELOWERSTATE: 4207 case NETDEV_UDP_TUNNEL_PUSH_INFO: 4208 case NETDEV_CHANGE_TX_QUEUE_LEN: 4209 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, event, GFP_KERNEL); 4210 break; 4211 default: 4212 break; 4213 } 4214 return NOTIFY_DONE; 4215 } 4216 4217 static struct notifier_block rtnetlink_dev_notifier = { 4218 .notifier_call = rtnetlink_event, 4219 }; 4220 4221 4222 static int __net_init rtnetlink_net_init(struct net *net) 4223 { 4224 struct sock *sk; 4225 struct netlink_kernel_cfg cfg = { 4226 .groups = RTNLGRP_MAX, 4227 .input = rtnetlink_rcv, 4228 .cb_mutex = &rtnl_mutex, 4229 .flags = NL_CFG_F_NONROOT_RECV, 4230 }; 4231 4232 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 4233 if (!sk) 4234 return -ENOMEM; 4235 net->rtnl = sk; 4236 return 0; 4237 } 4238 4239 static void __net_exit rtnetlink_net_exit(struct net *net) 4240 { 4241 netlink_kernel_release(net->rtnl); 4242 net->rtnl = NULL; 4243 } 4244 4245 static struct pernet_operations rtnetlink_net_ops = { 4246 .init = rtnetlink_net_init, 4247 .exit = rtnetlink_net_exit, 4248 }; 4249 4250 void __init rtnetlink_init(void) 4251 { 4252 if (register_pernet_subsys(&rtnetlink_net_ops)) 4253 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 4254 4255 register_netdevice_notifier(&rtnetlink_dev_notifier); 4256 4257 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 4258 rtnl_dump_ifinfo, rtnl_calcit); 4259 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL); 4260 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL); 4261 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL); 4262 4263 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL); 4264 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL); 4265 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL); 4266 4267 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL); 4268 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL); 4269 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL); 4270 4271 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL); 4272 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL); 4273 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL); 4274 4275 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 4276 NULL); 4277 } 4278