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