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 nla_total_size(4); /* XDP_FLAGS */ 904 905 return xdp_size; 906 } 907 908 static noinline size_t if_nlmsg_size(const struct net_device *dev, 909 u32 ext_filter_mask) 910 { 911 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 912 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 913 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 914 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 915 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap)) 916 + nla_total_size(sizeof(struct rtnl_link_stats)) 917 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64)) 918 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 919 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 920 + nla_total_size(4) /* IFLA_TXQLEN */ 921 + nla_total_size(4) /* IFLA_WEIGHT */ 922 + nla_total_size(4) /* IFLA_MTU */ 923 + nla_total_size(4) /* IFLA_LINK */ 924 + nla_total_size(4) /* IFLA_MASTER */ 925 + nla_total_size(1) /* IFLA_CARRIER */ 926 + nla_total_size(4) /* IFLA_PROMISCUITY */ 927 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 928 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 929 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */ 930 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */ 931 + nla_total_size(1) /* IFLA_OPERSTATE */ 932 + nla_total_size(1) /* IFLA_LINKMODE */ 933 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 934 + nla_total_size(4) /* IFLA_LINK_NETNSID */ 935 + nla_total_size(ext_filter_mask 936 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */ 937 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */ 938 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 939 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */ 940 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */ 941 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */ 942 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */ 943 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */ 944 + rtnl_xdp_size() /* IFLA_XDP */ 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 int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1251 { 1252 struct nlattr *xdp; 1253 u32 xdp_flags = 0; 1254 u8 val = 0; 1255 int err; 1256 1257 xdp = nla_nest_start(skb, IFLA_XDP); 1258 if (!xdp) 1259 return -EMSGSIZE; 1260 if (rcu_access_pointer(dev->xdp_prog)) { 1261 xdp_flags = XDP_FLAGS_SKB_MODE; 1262 val = 1; 1263 } else if (dev->netdev_ops->ndo_xdp) { 1264 struct netdev_xdp xdp_op = {}; 1265 1266 xdp_op.command = XDP_QUERY_PROG; 1267 err = dev->netdev_ops->ndo_xdp(dev, &xdp_op); 1268 if (err) 1269 goto err_cancel; 1270 val = xdp_op.prog_attached; 1271 } 1272 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, val); 1273 if (err) 1274 goto err_cancel; 1275 1276 if (xdp_flags) { 1277 err = nla_put_u32(skb, IFLA_XDP_FLAGS, xdp_flags); 1278 if (err) 1279 goto err_cancel; 1280 } 1281 nla_nest_end(skb, xdp); 1282 return 0; 1283 1284 err_cancel: 1285 nla_nest_cancel(skb, xdp); 1286 return err; 1287 } 1288 1289 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 1290 int type, u32 pid, u32 seq, u32 change, 1291 unsigned int flags, u32 ext_filter_mask) 1292 { 1293 struct ifinfomsg *ifm; 1294 struct nlmsghdr *nlh; 1295 struct nlattr *af_spec; 1296 struct rtnl_af_ops *af_ops; 1297 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1298 1299 ASSERT_RTNL(); 1300 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1301 if (nlh == NULL) 1302 return -EMSGSIZE; 1303 1304 ifm = nlmsg_data(nlh); 1305 ifm->ifi_family = AF_UNSPEC; 1306 ifm->__ifi_pad = 0; 1307 ifm->ifi_type = dev->type; 1308 ifm->ifi_index = dev->ifindex; 1309 ifm->ifi_flags = dev_get_flags(dev); 1310 ifm->ifi_change = change; 1311 1312 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1313 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1314 nla_put_u8(skb, IFLA_OPERSTATE, 1315 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1316 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1317 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1318 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1319 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1320 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1321 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1322 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1323 #ifdef CONFIG_RPS 1324 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1325 #endif 1326 (dev->ifindex != dev_get_iflink(dev) && 1327 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) || 1328 (upper_dev && 1329 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) || 1330 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1331 (dev->qdisc && 1332 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1333 (dev->ifalias && 1334 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) || 1335 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1336 atomic_read(&dev->carrier_changes)) || 1337 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down)) 1338 goto nla_put_failure; 1339 1340 if (rtnl_fill_link_ifmap(skb, dev)) 1341 goto nla_put_failure; 1342 1343 if (dev->addr_len) { 1344 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1345 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1346 goto nla_put_failure; 1347 } 1348 1349 if (rtnl_phys_port_id_fill(skb, dev)) 1350 goto nla_put_failure; 1351 1352 if (rtnl_phys_port_name_fill(skb, dev)) 1353 goto nla_put_failure; 1354 1355 if (rtnl_phys_switch_id_fill(skb, dev)) 1356 goto nla_put_failure; 1357 1358 if (rtnl_fill_stats(skb, dev)) 1359 goto nla_put_failure; 1360 1361 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) && 1362 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent))) 1363 goto nla_put_failure; 1364 1365 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent && 1366 ext_filter_mask & RTEXT_FILTER_VF) { 1367 int i; 1368 struct nlattr *vfinfo; 1369 int num_vfs = dev_num_vf(dev->dev.parent); 1370 1371 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1372 if (!vfinfo) 1373 goto nla_put_failure; 1374 for (i = 0; i < num_vfs; i++) { 1375 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1376 goto nla_put_failure; 1377 } 1378 1379 nla_nest_end(skb, vfinfo); 1380 } 1381 1382 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1383 goto nla_put_failure; 1384 1385 if (rtnl_xdp_fill(skb, dev)) 1386 goto nla_put_failure; 1387 1388 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1389 if (rtnl_link_fill(skb, dev) < 0) 1390 goto nla_put_failure; 1391 } 1392 1393 if (dev->rtnl_link_ops && 1394 dev->rtnl_link_ops->get_link_net) { 1395 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1396 1397 if (!net_eq(dev_net(dev), link_net)) { 1398 int id = peernet2id_alloc(dev_net(dev), link_net); 1399 1400 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1401 goto nla_put_failure; 1402 } 1403 } 1404 1405 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC))) 1406 goto nla_put_failure; 1407 1408 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 1409 if (af_ops->fill_link_af) { 1410 struct nlattr *af; 1411 int err; 1412 1413 if (!(af = nla_nest_start(skb, af_ops->family))) 1414 goto nla_put_failure; 1415 1416 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1417 1418 /* 1419 * Caller may return ENODATA to indicate that there 1420 * was no data to be dumped. This is not an error, it 1421 * means we should trim the attribute header and 1422 * continue. 1423 */ 1424 if (err == -ENODATA) 1425 nla_nest_cancel(skb, af); 1426 else if (err < 0) 1427 goto nla_put_failure; 1428 1429 nla_nest_end(skb, af); 1430 } 1431 } 1432 1433 nla_nest_end(skb, af_spec); 1434 1435 nlmsg_end(skb, nlh); 1436 return 0; 1437 1438 nla_put_failure: 1439 nlmsg_cancel(skb, nlh); 1440 return -EMSGSIZE; 1441 } 1442 1443 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1444 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1445 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1446 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1447 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1448 [IFLA_MTU] = { .type = NLA_U32 }, 1449 [IFLA_LINK] = { .type = NLA_U32 }, 1450 [IFLA_MASTER] = { .type = NLA_U32 }, 1451 [IFLA_CARRIER] = { .type = NLA_U8 }, 1452 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1453 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1454 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1455 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1456 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1457 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1458 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1459 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 1460 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1461 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1462 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1463 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1464 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1465 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1466 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1467 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1468 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1469 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1470 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1471 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1472 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1473 [IFLA_XDP] = { .type = NLA_NESTED }, 1474 }; 1475 1476 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1477 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1478 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1479 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1480 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1481 }; 1482 1483 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1484 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1485 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1486 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1487 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1488 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1489 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1490 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1491 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1492 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1493 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1494 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1495 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1496 }; 1497 1498 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1499 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1500 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1501 .len = PORT_PROFILE_MAX }, 1502 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1503 .len = PORT_UUID_MAX }, 1504 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1505 .len = PORT_UUID_MAX }, 1506 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1507 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1508 1509 /* Unused, but we need to keep it here since user space could 1510 * fill it. It's also broken with regard to NLA_BINARY use in 1511 * combination with structs. 1512 */ 1513 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1514 .len = sizeof(struct ifla_port_vsi) }, 1515 }; 1516 1517 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1518 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1519 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1520 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1521 }; 1522 1523 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1524 { 1525 const struct rtnl_link_ops *ops = NULL; 1526 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1527 1528 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, 1529 ifla_info_policy, NULL) < 0) 1530 return NULL; 1531 1532 if (linfo[IFLA_INFO_KIND]) { 1533 char kind[MODULE_NAME_LEN]; 1534 1535 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1536 ops = rtnl_link_ops_get(kind); 1537 } 1538 1539 return ops; 1540 } 1541 1542 static bool link_master_filtered(struct net_device *dev, int master_idx) 1543 { 1544 struct net_device *master; 1545 1546 if (!master_idx) 1547 return false; 1548 1549 master = netdev_master_upper_dev_get(dev); 1550 if (!master || master->ifindex != master_idx) 1551 return true; 1552 1553 return false; 1554 } 1555 1556 static bool link_kind_filtered(const struct net_device *dev, 1557 const struct rtnl_link_ops *kind_ops) 1558 { 1559 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1560 return true; 1561 1562 return false; 1563 } 1564 1565 static bool link_dump_filtered(struct net_device *dev, 1566 int master_idx, 1567 const struct rtnl_link_ops *kind_ops) 1568 { 1569 if (link_master_filtered(dev, master_idx) || 1570 link_kind_filtered(dev, kind_ops)) 1571 return true; 1572 1573 return false; 1574 } 1575 1576 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1577 { 1578 struct net *net = sock_net(skb->sk); 1579 int h, s_h; 1580 int idx = 0, s_idx; 1581 struct net_device *dev; 1582 struct hlist_head *head; 1583 struct nlattr *tb[IFLA_MAX+1]; 1584 u32 ext_filter_mask = 0; 1585 const struct rtnl_link_ops *kind_ops = NULL; 1586 unsigned int flags = NLM_F_MULTI; 1587 int master_idx = 0; 1588 int err; 1589 int hdrlen; 1590 1591 s_h = cb->args[0]; 1592 s_idx = cb->args[1]; 1593 1594 cb->seq = net->dev_base_seq; 1595 1596 /* A hack to preserve kernel<->userspace interface. 1597 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1598 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1599 * what iproute2 < v3.9.0 used. 1600 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1601 * attribute, its netlink message is shorter than struct ifinfomsg. 1602 */ 1603 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1604 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1605 1606 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, 1607 ifla_policy, NULL) >= 0) { 1608 if (tb[IFLA_EXT_MASK]) 1609 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1610 1611 if (tb[IFLA_MASTER]) 1612 master_idx = nla_get_u32(tb[IFLA_MASTER]); 1613 1614 if (tb[IFLA_LINKINFO]) 1615 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]); 1616 1617 if (master_idx || kind_ops) 1618 flags |= NLM_F_DUMP_FILTERED; 1619 } 1620 1621 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1622 idx = 0; 1623 head = &net->dev_index_head[h]; 1624 hlist_for_each_entry(dev, head, index_hlist) { 1625 if (link_dump_filtered(dev, master_idx, kind_ops)) 1626 goto cont; 1627 if (idx < s_idx) 1628 goto cont; 1629 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 1630 NETLINK_CB(cb->skb).portid, 1631 cb->nlh->nlmsg_seq, 0, 1632 flags, 1633 ext_filter_mask); 1634 /* If we ran out of room on the first message, 1635 * we're in trouble 1636 */ 1637 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 1638 1639 if (err < 0) 1640 goto out; 1641 1642 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1643 cont: 1644 idx++; 1645 } 1646 } 1647 out: 1648 cb->args[1] = idx; 1649 cb->args[0] = h; 1650 1651 return skb->len; 1652 } 1653 1654 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len, 1655 struct netlink_ext_ack *exterr) 1656 { 1657 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr); 1658 } 1659 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1660 1661 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1662 { 1663 struct net *net; 1664 /* Examine the link attributes and figure out which 1665 * network namespace we are talking about. 1666 */ 1667 if (tb[IFLA_NET_NS_PID]) 1668 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1669 else if (tb[IFLA_NET_NS_FD]) 1670 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1671 else 1672 net = get_net(src_net); 1673 return net; 1674 } 1675 EXPORT_SYMBOL(rtnl_link_get_net); 1676 1677 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1678 { 1679 if (dev) { 1680 if (tb[IFLA_ADDRESS] && 1681 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1682 return -EINVAL; 1683 1684 if (tb[IFLA_BROADCAST] && 1685 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1686 return -EINVAL; 1687 } 1688 1689 if (tb[IFLA_AF_SPEC]) { 1690 struct nlattr *af; 1691 int rem, err; 1692 1693 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1694 const struct rtnl_af_ops *af_ops; 1695 1696 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1697 return -EAFNOSUPPORT; 1698 1699 if (!af_ops->set_link_af) 1700 return -EOPNOTSUPP; 1701 1702 if (af_ops->validate_link_af) { 1703 err = af_ops->validate_link_af(dev, af); 1704 if (err < 0) 1705 return err; 1706 } 1707 } 1708 } 1709 1710 return 0; 1711 } 1712 1713 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 1714 int guid_type) 1715 { 1716 const struct net_device_ops *ops = dev->netdev_ops; 1717 1718 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 1719 } 1720 1721 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 1722 { 1723 if (dev->type != ARPHRD_INFINIBAND) 1724 return -EOPNOTSUPP; 1725 1726 return handle_infiniband_guid(dev, ivt, guid_type); 1727 } 1728 1729 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 1730 { 1731 const struct net_device_ops *ops = dev->netdev_ops; 1732 int err = -EINVAL; 1733 1734 if (tb[IFLA_VF_MAC]) { 1735 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 1736 1737 err = -EOPNOTSUPP; 1738 if (ops->ndo_set_vf_mac) 1739 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1740 ivm->mac); 1741 if (err < 0) 1742 return err; 1743 } 1744 1745 if (tb[IFLA_VF_VLAN]) { 1746 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 1747 1748 err = -EOPNOTSUPP; 1749 if (ops->ndo_set_vf_vlan) 1750 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 1751 ivv->qos, 1752 htons(ETH_P_8021Q)); 1753 if (err < 0) 1754 return err; 1755 } 1756 1757 if (tb[IFLA_VF_VLAN_LIST]) { 1758 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 1759 struct nlattr *attr; 1760 int rem, len = 0; 1761 1762 err = -EOPNOTSUPP; 1763 if (!ops->ndo_set_vf_vlan) 1764 return err; 1765 1766 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 1767 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 1768 nla_len(attr) < NLA_HDRLEN) { 1769 return -EINVAL; 1770 } 1771 if (len >= MAX_VLAN_LIST_LEN) 1772 return -EOPNOTSUPP; 1773 ivvl[len] = nla_data(attr); 1774 1775 len++; 1776 } 1777 if (len == 0) 1778 return -EINVAL; 1779 1780 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 1781 ivvl[0]->qos, ivvl[0]->vlan_proto); 1782 if (err < 0) 1783 return err; 1784 } 1785 1786 if (tb[IFLA_VF_TX_RATE]) { 1787 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 1788 struct ifla_vf_info ivf; 1789 1790 err = -EOPNOTSUPP; 1791 if (ops->ndo_get_vf_config) 1792 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 1793 if (err < 0) 1794 return err; 1795 1796 err = -EOPNOTSUPP; 1797 if (ops->ndo_set_vf_rate) 1798 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1799 ivf.min_tx_rate, 1800 ivt->rate); 1801 if (err < 0) 1802 return err; 1803 } 1804 1805 if (tb[IFLA_VF_RATE]) { 1806 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 1807 1808 err = -EOPNOTSUPP; 1809 if (ops->ndo_set_vf_rate) 1810 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1811 ivt->min_tx_rate, 1812 ivt->max_tx_rate); 1813 if (err < 0) 1814 return err; 1815 } 1816 1817 if (tb[IFLA_VF_SPOOFCHK]) { 1818 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 1819 1820 err = -EOPNOTSUPP; 1821 if (ops->ndo_set_vf_spoofchk) 1822 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 1823 ivs->setting); 1824 if (err < 0) 1825 return err; 1826 } 1827 1828 if (tb[IFLA_VF_LINK_STATE]) { 1829 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 1830 1831 err = -EOPNOTSUPP; 1832 if (ops->ndo_set_vf_link_state) 1833 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 1834 ivl->link_state); 1835 if (err < 0) 1836 return err; 1837 } 1838 1839 if (tb[IFLA_VF_RSS_QUERY_EN]) { 1840 struct ifla_vf_rss_query_en *ivrssq_en; 1841 1842 err = -EOPNOTSUPP; 1843 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 1844 if (ops->ndo_set_vf_rss_query_en) 1845 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 1846 ivrssq_en->setting); 1847 if (err < 0) 1848 return err; 1849 } 1850 1851 if (tb[IFLA_VF_TRUST]) { 1852 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 1853 1854 err = -EOPNOTSUPP; 1855 if (ops->ndo_set_vf_trust) 1856 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 1857 if (err < 0) 1858 return err; 1859 } 1860 1861 if (tb[IFLA_VF_IB_NODE_GUID]) { 1862 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 1863 1864 if (!ops->ndo_set_vf_guid) 1865 return -EOPNOTSUPP; 1866 1867 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 1868 } 1869 1870 if (tb[IFLA_VF_IB_PORT_GUID]) { 1871 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 1872 1873 if (!ops->ndo_set_vf_guid) 1874 return -EOPNOTSUPP; 1875 1876 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 1877 } 1878 1879 return err; 1880 } 1881 1882 static int do_set_master(struct net_device *dev, int ifindex) 1883 { 1884 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1885 const struct net_device_ops *ops; 1886 int err; 1887 1888 if (upper_dev) { 1889 if (upper_dev->ifindex == ifindex) 1890 return 0; 1891 ops = upper_dev->netdev_ops; 1892 if (ops->ndo_del_slave) { 1893 err = ops->ndo_del_slave(upper_dev, dev); 1894 if (err) 1895 return err; 1896 } else { 1897 return -EOPNOTSUPP; 1898 } 1899 } 1900 1901 if (ifindex) { 1902 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 1903 if (!upper_dev) 1904 return -EINVAL; 1905 ops = upper_dev->netdev_ops; 1906 if (ops->ndo_add_slave) { 1907 err = ops->ndo_add_slave(upper_dev, dev); 1908 if (err) 1909 return err; 1910 } else { 1911 return -EOPNOTSUPP; 1912 } 1913 } 1914 return 0; 1915 } 1916 1917 #define DO_SETLINK_MODIFIED 0x01 1918 /* notify flag means notify + modified. */ 1919 #define DO_SETLINK_NOTIFY 0x03 1920 static int do_setlink(const struct sk_buff *skb, 1921 struct net_device *dev, struct ifinfomsg *ifm, 1922 struct netlink_ext_ack *extack, 1923 struct nlattr **tb, char *ifname, int status) 1924 { 1925 const struct net_device_ops *ops = dev->netdev_ops; 1926 int err; 1927 1928 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) { 1929 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1930 if (IS_ERR(net)) { 1931 err = PTR_ERR(net); 1932 goto errout; 1933 } 1934 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { 1935 put_net(net); 1936 err = -EPERM; 1937 goto errout; 1938 } 1939 err = dev_change_net_namespace(dev, net, ifname); 1940 put_net(net); 1941 if (err) 1942 goto errout; 1943 status |= DO_SETLINK_MODIFIED; 1944 } 1945 1946 if (tb[IFLA_MAP]) { 1947 struct rtnl_link_ifmap *u_map; 1948 struct ifmap k_map; 1949 1950 if (!ops->ndo_set_config) { 1951 err = -EOPNOTSUPP; 1952 goto errout; 1953 } 1954 1955 if (!netif_device_present(dev)) { 1956 err = -ENODEV; 1957 goto errout; 1958 } 1959 1960 u_map = nla_data(tb[IFLA_MAP]); 1961 k_map.mem_start = (unsigned long) u_map->mem_start; 1962 k_map.mem_end = (unsigned long) u_map->mem_end; 1963 k_map.base_addr = (unsigned short) u_map->base_addr; 1964 k_map.irq = (unsigned char) u_map->irq; 1965 k_map.dma = (unsigned char) u_map->dma; 1966 k_map.port = (unsigned char) u_map->port; 1967 1968 err = ops->ndo_set_config(dev, &k_map); 1969 if (err < 0) 1970 goto errout; 1971 1972 status |= DO_SETLINK_NOTIFY; 1973 } 1974 1975 if (tb[IFLA_ADDRESS]) { 1976 struct sockaddr *sa; 1977 int len; 1978 1979 len = sizeof(sa_family_t) + dev->addr_len; 1980 sa = kmalloc(len, GFP_KERNEL); 1981 if (!sa) { 1982 err = -ENOMEM; 1983 goto errout; 1984 } 1985 sa->sa_family = dev->type; 1986 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 1987 dev->addr_len); 1988 err = dev_set_mac_address(dev, sa); 1989 kfree(sa); 1990 if (err) 1991 goto errout; 1992 status |= DO_SETLINK_MODIFIED; 1993 } 1994 1995 if (tb[IFLA_MTU]) { 1996 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1997 if (err < 0) 1998 goto errout; 1999 status |= DO_SETLINK_MODIFIED; 2000 } 2001 2002 if (tb[IFLA_GROUP]) { 2003 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2004 status |= DO_SETLINK_NOTIFY; 2005 } 2006 2007 /* 2008 * Interface selected by interface index but interface 2009 * name provided implies that a name change has been 2010 * requested. 2011 */ 2012 if (ifm->ifi_index > 0 && ifname[0]) { 2013 err = dev_change_name(dev, ifname); 2014 if (err < 0) 2015 goto errout; 2016 status |= DO_SETLINK_MODIFIED; 2017 } 2018 2019 if (tb[IFLA_IFALIAS]) { 2020 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2021 nla_len(tb[IFLA_IFALIAS])); 2022 if (err < 0) 2023 goto errout; 2024 status |= DO_SETLINK_NOTIFY; 2025 } 2026 2027 if (tb[IFLA_BROADCAST]) { 2028 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2029 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2030 } 2031 2032 if (ifm->ifi_flags || ifm->ifi_change) { 2033 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2034 if (err < 0) 2035 goto errout; 2036 } 2037 2038 if (tb[IFLA_MASTER]) { 2039 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2040 if (err) 2041 goto errout; 2042 status |= DO_SETLINK_MODIFIED; 2043 } 2044 2045 if (tb[IFLA_CARRIER]) { 2046 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2047 if (err) 2048 goto errout; 2049 status |= DO_SETLINK_MODIFIED; 2050 } 2051 2052 if (tb[IFLA_TXQLEN]) { 2053 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]); 2054 unsigned long orig_len = dev->tx_queue_len; 2055 2056 if (dev->tx_queue_len ^ value) { 2057 dev->tx_queue_len = value; 2058 err = call_netdevice_notifiers( 2059 NETDEV_CHANGE_TX_QUEUE_LEN, dev); 2060 err = notifier_to_errno(err); 2061 if (err) { 2062 dev->tx_queue_len = orig_len; 2063 goto errout; 2064 } 2065 status |= DO_SETLINK_NOTIFY; 2066 } 2067 } 2068 2069 if (tb[IFLA_OPERSTATE]) 2070 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2071 2072 if (tb[IFLA_LINKMODE]) { 2073 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2074 2075 write_lock_bh(&dev_base_lock); 2076 if (dev->link_mode ^ value) 2077 status |= DO_SETLINK_NOTIFY; 2078 dev->link_mode = value; 2079 write_unlock_bh(&dev_base_lock); 2080 } 2081 2082 if (tb[IFLA_VFINFO_LIST]) { 2083 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2084 struct nlattr *attr; 2085 int rem; 2086 2087 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2088 if (nla_type(attr) != IFLA_VF_INFO || 2089 nla_len(attr) < NLA_HDRLEN) { 2090 err = -EINVAL; 2091 goto errout; 2092 } 2093 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr, 2094 ifla_vf_policy, NULL); 2095 if (err < 0) 2096 goto errout; 2097 err = do_setvfinfo(dev, vfinfo); 2098 if (err < 0) 2099 goto errout; 2100 status |= DO_SETLINK_NOTIFY; 2101 } 2102 } 2103 err = 0; 2104 2105 if (tb[IFLA_VF_PORTS]) { 2106 struct nlattr *port[IFLA_PORT_MAX+1]; 2107 struct nlattr *attr; 2108 int vf; 2109 int rem; 2110 2111 err = -EOPNOTSUPP; 2112 if (!ops->ndo_set_vf_port) 2113 goto errout; 2114 2115 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2116 if (nla_type(attr) != IFLA_VF_PORT || 2117 nla_len(attr) < NLA_HDRLEN) { 2118 err = -EINVAL; 2119 goto errout; 2120 } 2121 err = nla_parse_nested(port, IFLA_PORT_MAX, attr, 2122 ifla_port_policy, NULL); 2123 if (err < 0) 2124 goto errout; 2125 if (!port[IFLA_PORT_VF]) { 2126 err = -EOPNOTSUPP; 2127 goto errout; 2128 } 2129 vf = nla_get_u32(port[IFLA_PORT_VF]); 2130 err = ops->ndo_set_vf_port(dev, vf, port); 2131 if (err < 0) 2132 goto errout; 2133 status |= DO_SETLINK_NOTIFY; 2134 } 2135 } 2136 err = 0; 2137 2138 if (tb[IFLA_PORT_SELF]) { 2139 struct nlattr *port[IFLA_PORT_MAX+1]; 2140 2141 err = nla_parse_nested(port, IFLA_PORT_MAX, 2142 tb[IFLA_PORT_SELF], ifla_port_policy, 2143 NULL); 2144 if (err < 0) 2145 goto errout; 2146 2147 err = -EOPNOTSUPP; 2148 if (ops->ndo_set_vf_port) 2149 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2150 if (err < 0) 2151 goto errout; 2152 status |= DO_SETLINK_NOTIFY; 2153 } 2154 2155 if (tb[IFLA_AF_SPEC]) { 2156 struct nlattr *af; 2157 int rem; 2158 2159 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2160 const struct rtnl_af_ops *af_ops; 2161 2162 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 2163 BUG(); 2164 2165 err = af_ops->set_link_af(dev, af); 2166 if (err < 0) 2167 goto errout; 2168 2169 status |= DO_SETLINK_NOTIFY; 2170 } 2171 } 2172 err = 0; 2173 2174 if (tb[IFLA_PROTO_DOWN]) { 2175 err = dev_change_proto_down(dev, 2176 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2177 if (err) 2178 goto errout; 2179 status |= DO_SETLINK_NOTIFY; 2180 } 2181 2182 if (tb[IFLA_XDP]) { 2183 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2184 u32 xdp_flags = 0; 2185 2186 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP], 2187 ifla_xdp_policy, NULL); 2188 if (err < 0) 2189 goto errout; 2190 2191 if (xdp[IFLA_XDP_ATTACHED]) { 2192 err = -EINVAL; 2193 goto errout; 2194 } 2195 2196 if (xdp[IFLA_XDP_FLAGS]) { 2197 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2198 if (xdp_flags & ~XDP_FLAGS_MASK) { 2199 err = -EINVAL; 2200 goto errout; 2201 } 2202 if ((xdp_flags & XDP_FLAGS_SKB_MODE) && 2203 (xdp_flags & XDP_FLAGS_DRV_MODE)) { 2204 err = -EINVAL; 2205 goto errout; 2206 } 2207 } 2208 2209 if (xdp[IFLA_XDP_FD]) { 2210 err = dev_change_xdp_fd(dev, extack, 2211 nla_get_s32(xdp[IFLA_XDP_FD]), 2212 xdp_flags); 2213 if (err) 2214 goto errout; 2215 status |= DO_SETLINK_NOTIFY; 2216 } 2217 } 2218 2219 errout: 2220 if (status & DO_SETLINK_MODIFIED) { 2221 if (status & DO_SETLINK_NOTIFY) 2222 netdev_state_change(dev); 2223 2224 if (err < 0) 2225 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", 2226 dev->name); 2227 } 2228 2229 return err; 2230 } 2231 2232 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2233 struct netlink_ext_ack *extack) 2234 { 2235 struct net *net = sock_net(skb->sk); 2236 struct ifinfomsg *ifm; 2237 struct net_device *dev; 2238 int err; 2239 struct nlattr *tb[IFLA_MAX+1]; 2240 char ifname[IFNAMSIZ]; 2241 2242 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, 2243 extack); 2244 if (err < 0) 2245 goto errout; 2246 2247 if (tb[IFLA_IFNAME]) 2248 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2249 else 2250 ifname[0] = '\0'; 2251 2252 err = -EINVAL; 2253 ifm = nlmsg_data(nlh); 2254 if (ifm->ifi_index > 0) 2255 dev = __dev_get_by_index(net, ifm->ifi_index); 2256 else if (tb[IFLA_IFNAME]) 2257 dev = __dev_get_by_name(net, ifname); 2258 else 2259 goto errout; 2260 2261 if (dev == NULL) { 2262 err = -ENODEV; 2263 goto errout; 2264 } 2265 2266 err = validate_linkmsg(dev, tb); 2267 if (err < 0) 2268 goto errout; 2269 2270 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0); 2271 errout: 2272 return err; 2273 } 2274 2275 static int rtnl_group_dellink(const struct net *net, int group) 2276 { 2277 struct net_device *dev, *aux; 2278 LIST_HEAD(list_kill); 2279 bool found = false; 2280 2281 if (!group) 2282 return -EPERM; 2283 2284 for_each_netdev(net, dev) { 2285 if (dev->group == group) { 2286 const struct rtnl_link_ops *ops; 2287 2288 found = true; 2289 ops = dev->rtnl_link_ops; 2290 if (!ops || !ops->dellink) 2291 return -EOPNOTSUPP; 2292 } 2293 } 2294 2295 if (!found) 2296 return -ENODEV; 2297 2298 for_each_netdev_safe(net, dev, aux) { 2299 if (dev->group == group) { 2300 const struct rtnl_link_ops *ops; 2301 2302 ops = dev->rtnl_link_ops; 2303 ops->dellink(dev, &list_kill); 2304 } 2305 } 2306 unregister_netdevice_many(&list_kill); 2307 2308 return 0; 2309 } 2310 2311 int rtnl_delete_link(struct net_device *dev) 2312 { 2313 const struct rtnl_link_ops *ops; 2314 LIST_HEAD(list_kill); 2315 2316 ops = dev->rtnl_link_ops; 2317 if (!ops || !ops->dellink) 2318 return -EOPNOTSUPP; 2319 2320 ops->dellink(dev, &list_kill); 2321 unregister_netdevice_many(&list_kill); 2322 2323 return 0; 2324 } 2325 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2326 2327 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 2328 struct netlink_ext_ack *extack) 2329 { 2330 struct net *net = sock_net(skb->sk); 2331 struct net_device *dev; 2332 struct ifinfomsg *ifm; 2333 char ifname[IFNAMSIZ]; 2334 struct nlattr *tb[IFLA_MAX+1]; 2335 int err; 2336 2337 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2338 if (err < 0) 2339 return err; 2340 2341 if (tb[IFLA_IFNAME]) 2342 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2343 2344 ifm = nlmsg_data(nlh); 2345 if (ifm->ifi_index > 0) 2346 dev = __dev_get_by_index(net, ifm->ifi_index); 2347 else if (tb[IFLA_IFNAME]) 2348 dev = __dev_get_by_name(net, ifname); 2349 else if (tb[IFLA_GROUP]) 2350 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP])); 2351 else 2352 return -EINVAL; 2353 2354 if (!dev) 2355 return -ENODEV; 2356 2357 return rtnl_delete_link(dev); 2358 } 2359 2360 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2361 { 2362 unsigned int old_flags; 2363 int err; 2364 2365 old_flags = dev->flags; 2366 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2367 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2368 if (err < 0) 2369 return err; 2370 } 2371 2372 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2373 2374 __dev_notify_flags(dev, old_flags, ~0U); 2375 return 0; 2376 } 2377 EXPORT_SYMBOL(rtnl_configure_link); 2378 2379 struct net_device *rtnl_create_link(struct net *net, 2380 const char *ifname, unsigned char name_assign_type, 2381 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 2382 { 2383 struct net_device *dev; 2384 unsigned int num_tx_queues = 1; 2385 unsigned int num_rx_queues = 1; 2386 2387 if (tb[IFLA_NUM_TX_QUEUES]) 2388 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2389 else if (ops->get_num_tx_queues) 2390 num_tx_queues = ops->get_num_tx_queues(); 2391 2392 if (tb[IFLA_NUM_RX_QUEUES]) 2393 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2394 else if (ops->get_num_rx_queues) 2395 num_rx_queues = ops->get_num_rx_queues(); 2396 2397 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2398 ops->setup, num_tx_queues, num_rx_queues); 2399 if (!dev) 2400 return ERR_PTR(-ENOMEM); 2401 2402 dev_net_set(dev, net); 2403 dev->rtnl_link_ops = ops; 2404 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2405 2406 if (tb[IFLA_MTU]) 2407 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2408 if (tb[IFLA_ADDRESS]) { 2409 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2410 nla_len(tb[IFLA_ADDRESS])); 2411 dev->addr_assign_type = NET_ADDR_SET; 2412 } 2413 if (tb[IFLA_BROADCAST]) 2414 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2415 nla_len(tb[IFLA_BROADCAST])); 2416 if (tb[IFLA_TXQLEN]) 2417 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2418 if (tb[IFLA_OPERSTATE]) 2419 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2420 if (tb[IFLA_LINKMODE]) 2421 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2422 if (tb[IFLA_GROUP]) 2423 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2424 2425 return dev; 2426 } 2427 EXPORT_SYMBOL(rtnl_create_link); 2428 2429 static int rtnl_group_changelink(const struct sk_buff *skb, 2430 struct net *net, int group, 2431 struct ifinfomsg *ifm, 2432 struct netlink_ext_ack *extack, 2433 struct nlattr **tb) 2434 { 2435 struct net_device *dev, *aux; 2436 int err; 2437 2438 for_each_netdev_safe(net, dev, aux) { 2439 if (dev->group == group) { 2440 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0); 2441 if (err < 0) 2442 return err; 2443 } 2444 } 2445 2446 return 0; 2447 } 2448 2449 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2450 struct netlink_ext_ack *extack) 2451 { 2452 struct net *net = sock_net(skb->sk); 2453 const struct rtnl_link_ops *ops; 2454 const struct rtnl_link_ops *m_ops = NULL; 2455 struct net_device *dev; 2456 struct net_device *master_dev = NULL; 2457 struct ifinfomsg *ifm; 2458 char kind[MODULE_NAME_LEN]; 2459 char ifname[IFNAMSIZ]; 2460 struct nlattr *tb[IFLA_MAX+1]; 2461 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 2462 unsigned char name_assign_type = NET_NAME_USER; 2463 int err; 2464 2465 #ifdef CONFIG_MODULES 2466 replay: 2467 #endif 2468 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2469 if (err < 0) 2470 return err; 2471 2472 if (tb[IFLA_IFNAME]) 2473 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2474 else 2475 ifname[0] = '\0'; 2476 2477 ifm = nlmsg_data(nlh); 2478 if (ifm->ifi_index > 0) 2479 dev = __dev_get_by_index(net, ifm->ifi_index); 2480 else { 2481 if (ifname[0]) 2482 dev = __dev_get_by_name(net, ifname); 2483 else 2484 dev = NULL; 2485 } 2486 2487 if (dev) { 2488 master_dev = netdev_master_upper_dev_get(dev); 2489 if (master_dev) 2490 m_ops = master_dev->rtnl_link_ops; 2491 } 2492 2493 err = validate_linkmsg(dev, tb); 2494 if (err < 0) 2495 return err; 2496 2497 if (tb[IFLA_LINKINFO]) { 2498 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 2499 tb[IFLA_LINKINFO], ifla_info_policy, 2500 NULL); 2501 if (err < 0) 2502 return err; 2503 } else 2504 memset(linkinfo, 0, sizeof(linkinfo)); 2505 2506 if (linkinfo[IFLA_INFO_KIND]) { 2507 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 2508 ops = rtnl_link_ops_get(kind); 2509 } else { 2510 kind[0] = '\0'; 2511 ops = NULL; 2512 } 2513 2514 if (1) { 2515 struct nlattr *attr[ops ? ops->maxtype + 1 : 1]; 2516 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1]; 2517 struct nlattr **data = NULL; 2518 struct nlattr **slave_data = NULL; 2519 struct net *dest_net, *link_net = NULL; 2520 2521 if (ops) { 2522 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 2523 err = nla_parse_nested(attr, ops->maxtype, 2524 linkinfo[IFLA_INFO_DATA], 2525 ops->policy, NULL); 2526 if (err < 0) 2527 return err; 2528 data = attr; 2529 } 2530 if (ops->validate) { 2531 err = ops->validate(tb, data); 2532 if (err < 0) 2533 return err; 2534 } 2535 } 2536 2537 if (m_ops) { 2538 if (m_ops->slave_maxtype && 2539 linkinfo[IFLA_INFO_SLAVE_DATA]) { 2540 err = nla_parse_nested(slave_attr, 2541 m_ops->slave_maxtype, 2542 linkinfo[IFLA_INFO_SLAVE_DATA], 2543 m_ops->slave_policy, 2544 NULL); 2545 if (err < 0) 2546 return err; 2547 slave_data = slave_attr; 2548 } 2549 if (m_ops->slave_validate) { 2550 err = m_ops->slave_validate(tb, slave_data); 2551 if (err < 0) 2552 return err; 2553 } 2554 } 2555 2556 if (dev) { 2557 int status = 0; 2558 2559 if (nlh->nlmsg_flags & NLM_F_EXCL) 2560 return -EEXIST; 2561 if (nlh->nlmsg_flags & NLM_F_REPLACE) 2562 return -EOPNOTSUPP; 2563 2564 if (linkinfo[IFLA_INFO_DATA]) { 2565 if (!ops || ops != dev->rtnl_link_ops || 2566 !ops->changelink) 2567 return -EOPNOTSUPP; 2568 2569 err = ops->changelink(dev, tb, data); 2570 if (err < 0) 2571 return err; 2572 status |= DO_SETLINK_NOTIFY; 2573 } 2574 2575 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 2576 if (!m_ops || !m_ops->slave_changelink) 2577 return -EOPNOTSUPP; 2578 2579 err = m_ops->slave_changelink(master_dev, dev, 2580 tb, slave_data); 2581 if (err < 0) 2582 return err; 2583 status |= DO_SETLINK_NOTIFY; 2584 } 2585 2586 return do_setlink(skb, dev, ifm, extack, tb, ifname, 2587 status); 2588 } 2589 2590 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 2591 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 2592 return rtnl_group_changelink(skb, net, 2593 nla_get_u32(tb[IFLA_GROUP]), 2594 ifm, extack, tb); 2595 return -ENODEV; 2596 } 2597 2598 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 2599 return -EOPNOTSUPP; 2600 2601 if (!ops) { 2602 #ifdef CONFIG_MODULES 2603 if (kind[0]) { 2604 __rtnl_unlock(); 2605 request_module("rtnl-link-%s", kind); 2606 rtnl_lock(); 2607 ops = rtnl_link_ops_get(kind); 2608 if (ops) 2609 goto replay; 2610 } 2611 #endif 2612 return -EOPNOTSUPP; 2613 } 2614 2615 if (!ops->setup) 2616 return -EOPNOTSUPP; 2617 2618 if (!ifname[0]) { 2619 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 2620 name_assign_type = NET_NAME_ENUM; 2621 } 2622 2623 dest_net = rtnl_link_get_net(net, tb); 2624 if (IS_ERR(dest_net)) 2625 return PTR_ERR(dest_net); 2626 2627 err = -EPERM; 2628 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN)) 2629 goto out; 2630 2631 if (tb[IFLA_LINK_NETNSID]) { 2632 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 2633 2634 link_net = get_net_ns_by_id(dest_net, id); 2635 if (!link_net) { 2636 err = -EINVAL; 2637 goto out; 2638 } 2639 err = -EPERM; 2640 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 2641 goto out; 2642 } 2643 2644 dev = rtnl_create_link(link_net ? : dest_net, ifname, 2645 name_assign_type, ops, tb); 2646 if (IS_ERR(dev)) { 2647 err = PTR_ERR(dev); 2648 goto out; 2649 } 2650 2651 dev->ifindex = ifm->ifi_index; 2652 2653 if (ops->newlink) { 2654 err = ops->newlink(link_net ? : net, dev, tb, data); 2655 /* Drivers should call free_netdev() in ->destructor 2656 * and unregister it on failure after registration 2657 * so that device could be finally freed in rtnl_unlock. 2658 */ 2659 if (err < 0) { 2660 /* If device is not registered at all, free it now */ 2661 if (dev->reg_state == NETREG_UNINITIALIZED) 2662 free_netdev(dev); 2663 goto out; 2664 } 2665 } else { 2666 err = register_netdevice(dev); 2667 if (err < 0) { 2668 free_netdev(dev); 2669 goto out; 2670 } 2671 } 2672 err = rtnl_configure_link(dev, ifm); 2673 if (err < 0) 2674 goto out_unregister; 2675 if (link_net) { 2676 err = dev_change_net_namespace(dev, dest_net, ifname); 2677 if (err < 0) 2678 goto out_unregister; 2679 } 2680 if (tb[IFLA_MASTER]) { 2681 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2682 if (err) 2683 goto out_unregister; 2684 } 2685 out: 2686 if (link_net) 2687 put_net(link_net); 2688 put_net(dest_net); 2689 return err; 2690 out_unregister: 2691 if (ops->newlink) { 2692 LIST_HEAD(list_kill); 2693 2694 ops->dellink(dev, &list_kill); 2695 unregister_netdevice_many(&list_kill); 2696 } else { 2697 unregister_netdevice(dev); 2698 } 2699 goto out; 2700 } 2701 } 2702 2703 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2704 struct netlink_ext_ack *extack) 2705 { 2706 struct net *net = sock_net(skb->sk); 2707 struct ifinfomsg *ifm; 2708 char ifname[IFNAMSIZ]; 2709 struct nlattr *tb[IFLA_MAX+1]; 2710 struct net_device *dev = NULL; 2711 struct sk_buff *nskb; 2712 int err; 2713 u32 ext_filter_mask = 0; 2714 2715 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2716 if (err < 0) 2717 return err; 2718 2719 if (tb[IFLA_IFNAME]) 2720 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2721 2722 if (tb[IFLA_EXT_MASK]) 2723 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2724 2725 ifm = nlmsg_data(nlh); 2726 if (ifm->ifi_index > 0) 2727 dev = __dev_get_by_index(net, ifm->ifi_index); 2728 else if (tb[IFLA_IFNAME]) 2729 dev = __dev_get_by_name(net, ifname); 2730 else 2731 return -EINVAL; 2732 2733 if (dev == NULL) 2734 return -ENODEV; 2735 2736 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 2737 if (nskb == NULL) 2738 return -ENOBUFS; 2739 2740 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid, 2741 nlh->nlmsg_seq, 0, 0, ext_filter_mask); 2742 if (err < 0) { 2743 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 2744 WARN_ON(err == -EMSGSIZE); 2745 kfree_skb(nskb); 2746 } else 2747 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 2748 2749 return err; 2750 } 2751 2752 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 2753 { 2754 struct net *net = sock_net(skb->sk); 2755 struct net_device *dev; 2756 struct nlattr *tb[IFLA_MAX+1]; 2757 u32 ext_filter_mask = 0; 2758 u16 min_ifinfo_dump_size = 0; 2759 int hdrlen; 2760 2761 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 2762 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 2763 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 2764 2765 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) { 2766 if (tb[IFLA_EXT_MASK]) 2767 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2768 } 2769 2770 if (!ext_filter_mask) 2771 return NLMSG_GOODSIZE; 2772 /* 2773 * traverse the list of net devices and compute the minimum 2774 * buffer size based upon the filter mask. 2775 */ 2776 list_for_each_entry(dev, &net->dev_base_head, dev_list) { 2777 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 2778 if_nlmsg_size(dev, 2779 ext_filter_mask)); 2780 } 2781 2782 return nlmsg_total_size(min_ifinfo_dump_size); 2783 } 2784 2785 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 2786 { 2787 int idx; 2788 int s_idx = cb->family; 2789 2790 if (s_idx == 0) 2791 s_idx = 1; 2792 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 2793 int type = cb->nlh->nlmsg_type-RTM_BASE; 2794 if (idx < s_idx || idx == PF_PACKET) 2795 continue; 2796 if (rtnl_msg_handlers[idx] == NULL || 2797 rtnl_msg_handlers[idx][type].dumpit == NULL) 2798 continue; 2799 if (idx > s_idx) { 2800 memset(&cb->args[0], 0, sizeof(cb->args)); 2801 cb->prev_seq = 0; 2802 cb->seq = 0; 2803 } 2804 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 2805 break; 2806 } 2807 cb->family = idx; 2808 2809 return skb->len; 2810 } 2811 2812 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 2813 unsigned int change, gfp_t flags) 2814 { 2815 struct net *net = dev_net(dev); 2816 struct sk_buff *skb; 2817 int err = -ENOBUFS; 2818 size_t if_info_size; 2819 2820 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 2821 if (skb == NULL) 2822 goto errout; 2823 2824 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0); 2825 if (err < 0) { 2826 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 2827 WARN_ON(err == -EMSGSIZE); 2828 kfree_skb(skb); 2829 goto errout; 2830 } 2831 return skb; 2832 errout: 2833 if (err < 0) 2834 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2835 return NULL; 2836 } 2837 2838 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 2839 { 2840 struct net *net = dev_net(dev); 2841 2842 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 2843 } 2844 2845 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 2846 gfp_t flags) 2847 { 2848 struct sk_buff *skb; 2849 2850 if (dev->reg_state != NETREG_REGISTERED) 2851 return; 2852 2853 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags); 2854 if (skb) 2855 rtmsg_ifinfo_send(skb, dev, flags); 2856 } 2857 EXPORT_SYMBOL(rtmsg_ifinfo); 2858 2859 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 2860 struct net_device *dev, 2861 u8 *addr, u16 vid, u32 pid, u32 seq, 2862 int type, unsigned int flags, 2863 int nlflags, u16 ndm_state) 2864 { 2865 struct nlmsghdr *nlh; 2866 struct ndmsg *ndm; 2867 2868 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 2869 if (!nlh) 2870 return -EMSGSIZE; 2871 2872 ndm = nlmsg_data(nlh); 2873 ndm->ndm_family = AF_BRIDGE; 2874 ndm->ndm_pad1 = 0; 2875 ndm->ndm_pad2 = 0; 2876 ndm->ndm_flags = flags; 2877 ndm->ndm_type = 0; 2878 ndm->ndm_ifindex = dev->ifindex; 2879 ndm->ndm_state = ndm_state; 2880 2881 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 2882 goto nla_put_failure; 2883 if (vid) 2884 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 2885 goto nla_put_failure; 2886 2887 nlmsg_end(skb, nlh); 2888 return 0; 2889 2890 nla_put_failure: 2891 nlmsg_cancel(skb, nlh); 2892 return -EMSGSIZE; 2893 } 2894 2895 static inline size_t rtnl_fdb_nlmsg_size(void) 2896 { 2897 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 2898 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 2899 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 2900 0; 2901 } 2902 2903 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 2904 u16 ndm_state) 2905 { 2906 struct net *net = dev_net(dev); 2907 struct sk_buff *skb; 2908 int err = -ENOBUFS; 2909 2910 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 2911 if (!skb) 2912 goto errout; 2913 2914 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 2915 0, 0, type, NTF_SELF, 0, ndm_state); 2916 if (err < 0) { 2917 kfree_skb(skb); 2918 goto errout; 2919 } 2920 2921 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 2922 return; 2923 errout: 2924 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 2925 } 2926 2927 /** 2928 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 2929 */ 2930 int ndo_dflt_fdb_add(struct ndmsg *ndm, 2931 struct nlattr *tb[], 2932 struct net_device *dev, 2933 const unsigned char *addr, u16 vid, 2934 u16 flags) 2935 { 2936 int err = -EINVAL; 2937 2938 /* If aging addresses are supported device will need to 2939 * implement its own handler for this. 2940 */ 2941 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 2942 pr_info("%s: FDB only supports static addresses\n", dev->name); 2943 return err; 2944 } 2945 2946 if (vid) { 2947 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 2948 return err; 2949 } 2950 2951 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 2952 err = dev_uc_add_excl(dev, addr); 2953 else if (is_multicast_ether_addr(addr)) 2954 err = dev_mc_add_excl(dev, addr); 2955 2956 /* Only return duplicate errors if NLM_F_EXCL is set */ 2957 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 2958 err = 0; 2959 2960 return err; 2961 } 2962 EXPORT_SYMBOL(ndo_dflt_fdb_add); 2963 2964 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid) 2965 { 2966 u16 vid = 0; 2967 2968 if (vlan_attr) { 2969 if (nla_len(vlan_attr) != sizeof(u16)) { 2970 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n"); 2971 return -EINVAL; 2972 } 2973 2974 vid = nla_get_u16(vlan_attr); 2975 2976 if (!vid || vid >= VLAN_VID_MASK) { 2977 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n", 2978 vid); 2979 return -EINVAL; 2980 } 2981 } 2982 *p_vid = vid; 2983 return 0; 2984 } 2985 2986 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, 2987 struct netlink_ext_ack *extack) 2988 { 2989 struct net *net = sock_net(skb->sk); 2990 struct ndmsg *ndm; 2991 struct nlattr *tb[NDA_MAX+1]; 2992 struct net_device *dev; 2993 u8 *addr; 2994 u16 vid; 2995 int err; 2996 2997 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 2998 if (err < 0) 2999 return err; 3000 3001 ndm = nlmsg_data(nlh); 3002 if (ndm->ndm_ifindex == 0) { 3003 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n"); 3004 return -EINVAL; 3005 } 3006 3007 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3008 if (dev == NULL) { 3009 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n"); 3010 return -ENODEV; 3011 } 3012 3013 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3014 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n"); 3015 return -EINVAL; 3016 } 3017 3018 addr = nla_data(tb[NDA_LLADDR]); 3019 3020 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3021 if (err) 3022 return err; 3023 3024 err = -EOPNOTSUPP; 3025 3026 /* Support fdb on master device the net/bridge default case */ 3027 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3028 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3029 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3030 const struct net_device_ops *ops = br_dev->netdev_ops; 3031 3032 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3033 nlh->nlmsg_flags); 3034 if (err) 3035 goto out; 3036 else 3037 ndm->ndm_flags &= ~NTF_MASTER; 3038 } 3039 3040 /* Embedded bridge, macvlan, and any other device support */ 3041 if ((ndm->ndm_flags & NTF_SELF)) { 3042 if (dev->netdev_ops->ndo_fdb_add) 3043 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3044 vid, 3045 nlh->nlmsg_flags); 3046 else 3047 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3048 nlh->nlmsg_flags); 3049 3050 if (!err) { 3051 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3052 ndm->ndm_state); 3053 ndm->ndm_flags &= ~NTF_SELF; 3054 } 3055 } 3056 out: 3057 return err; 3058 } 3059 3060 /** 3061 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3062 */ 3063 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3064 struct nlattr *tb[], 3065 struct net_device *dev, 3066 const unsigned char *addr, u16 vid) 3067 { 3068 int err = -EINVAL; 3069 3070 /* If aging addresses are supported device will need to 3071 * implement its own handler for this. 3072 */ 3073 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3074 pr_info("%s: FDB only supports static addresses\n", dev->name); 3075 return err; 3076 } 3077 3078 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3079 err = dev_uc_del(dev, addr); 3080 else if (is_multicast_ether_addr(addr)) 3081 err = dev_mc_del(dev, addr); 3082 3083 return err; 3084 } 3085 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3086 3087 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, 3088 struct netlink_ext_ack *extack) 3089 { 3090 struct net *net = sock_net(skb->sk); 3091 struct ndmsg *ndm; 3092 struct nlattr *tb[NDA_MAX+1]; 3093 struct net_device *dev; 3094 int err = -EINVAL; 3095 __u8 *addr; 3096 u16 vid; 3097 3098 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3099 return -EPERM; 3100 3101 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3102 if (err < 0) 3103 return err; 3104 3105 ndm = nlmsg_data(nlh); 3106 if (ndm->ndm_ifindex == 0) { 3107 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n"); 3108 return -EINVAL; 3109 } 3110 3111 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3112 if (dev == NULL) { 3113 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n"); 3114 return -ENODEV; 3115 } 3116 3117 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3118 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n"); 3119 return -EINVAL; 3120 } 3121 3122 addr = nla_data(tb[NDA_LLADDR]); 3123 3124 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3125 if (err) 3126 return err; 3127 3128 err = -EOPNOTSUPP; 3129 3130 /* Support fdb on master device the net/bridge default case */ 3131 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3132 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3133 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3134 const struct net_device_ops *ops = br_dev->netdev_ops; 3135 3136 if (ops->ndo_fdb_del) 3137 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3138 3139 if (err) 3140 goto out; 3141 else 3142 ndm->ndm_flags &= ~NTF_MASTER; 3143 } 3144 3145 /* Embedded bridge, macvlan, and any other device support */ 3146 if (ndm->ndm_flags & NTF_SELF) { 3147 if (dev->netdev_ops->ndo_fdb_del) 3148 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3149 vid); 3150 else 3151 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3152 3153 if (!err) { 3154 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3155 ndm->ndm_state); 3156 ndm->ndm_flags &= ~NTF_SELF; 3157 } 3158 } 3159 out: 3160 return err; 3161 } 3162 3163 static int nlmsg_populate_fdb(struct sk_buff *skb, 3164 struct netlink_callback *cb, 3165 struct net_device *dev, 3166 int *idx, 3167 struct netdev_hw_addr_list *list) 3168 { 3169 struct netdev_hw_addr *ha; 3170 int err; 3171 u32 portid, seq; 3172 3173 portid = NETLINK_CB(cb->skb).portid; 3174 seq = cb->nlh->nlmsg_seq; 3175 3176 list_for_each_entry(ha, &list->list, list) { 3177 if (*idx < cb->args[2]) 3178 goto skip; 3179 3180 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3181 portid, seq, 3182 RTM_NEWNEIGH, NTF_SELF, 3183 NLM_F_MULTI, NUD_PERMANENT); 3184 if (err < 0) 3185 return err; 3186 skip: 3187 *idx += 1; 3188 } 3189 return 0; 3190 } 3191 3192 /** 3193 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3194 * @nlh: netlink message header 3195 * @dev: netdevice 3196 * 3197 * Default netdevice operation to dump the existing unicast address list. 3198 * Returns number of addresses from list put in skb. 3199 */ 3200 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3201 struct netlink_callback *cb, 3202 struct net_device *dev, 3203 struct net_device *filter_dev, 3204 int *idx) 3205 { 3206 int err; 3207 3208 netif_addr_lock_bh(dev); 3209 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3210 if (err) 3211 goto out; 3212 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3213 out: 3214 netif_addr_unlock_bh(dev); 3215 return err; 3216 } 3217 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3218 3219 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 3220 { 3221 struct net_device *dev; 3222 struct nlattr *tb[IFLA_MAX+1]; 3223 struct net_device *br_dev = NULL; 3224 const struct net_device_ops *ops = NULL; 3225 const struct net_device_ops *cops = NULL; 3226 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 3227 struct net *net = sock_net(skb->sk); 3228 struct hlist_head *head; 3229 int brport_idx = 0; 3230 int br_idx = 0; 3231 int h, s_h; 3232 int idx = 0, s_idx; 3233 int err = 0; 3234 int fidx = 0; 3235 3236 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, 3237 IFLA_MAX, ifla_policy, NULL) == 0) { 3238 if (tb[IFLA_MASTER]) 3239 br_idx = nla_get_u32(tb[IFLA_MASTER]); 3240 } 3241 3242 brport_idx = ifm->ifi_index; 3243 3244 if (br_idx) { 3245 br_dev = __dev_get_by_index(net, br_idx); 3246 if (!br_dev) 3247 return -ENODEV; 3248 3249 ops = br_dev->netdev_ops; 3250 } 3251 3252 s_h = cb->args[0]; 3253 s_idx = cb->args[1]; 3254 3255 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3256 idx = 0; 3257 head = &net->dev_index_head[h]; 3258 hlist_for_each_entry(dev, head, index_hlist) { 3259 3260 if (brport_idx && (dev->ifindex != brport_idx)) 3261 continue; 3262 3263 if (!br_idx) { /* user did not specify a specific bridge */ 3264 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3265 br_dev = netdev_master_upper_dev_get(dev); 3266 cops = br_dev->netdev_ops; 3267 } 3268 } else { 3269 if (dev != br_dev && 3270 !(dev->priv_flags & IFF_BRIDGE_PORT)) 3271 continue; 3272 3273 if (br_dev != netdev_master_upper_dev_get(dev) && 3274 !(dev->priv_flags & IFF_EBRIDGE)) 3275 continue; 3276 cops = ops; 3277 } 3278 3279 if (idx < s_idx) 3280 goto cont; 3281 3282 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3283 if (cops && cops->ndo_fdb_dump) { 3284 err = cops->ndo_fdb_dump(skb, cb, 3285 br_dev, dev, 3286 &fidx); 3287 if (err == -EMSGSIZE) 3288 goto out; 3289 } 3290 } 3291 3292 if (dev->netdev_ops->ndo_fdb_dump) 3293 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 3294 dev, NULL, 3295 &fidx); 3296 else 3297 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 3298 &fidx); 3299 if (err == -EMSGSIZE) 3300 goto out; 3301 3302 cops = NULL; 3303 3304 /* reset fdb offset to 0 for rest of the interfaces */ 3305 cb->args[2] = 0; 3306 fidx = 0; 3307 cont: 3308 idx++; 3309 } 3310 } 3311 3312 out: 3313 cb->args[0] = h; 3314 cb->args[1] = idx; 3315 cb->args[2] = fidx; 3316 3317 return skb->len; 3318 } 3319 3320 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 3321 unsigned int attrnum, unsigned int flag) 3322 { 3323 if (mask & flag) 3324 return nla_put_u8(skb, attrnum, !!(flags & flag)); 3325 return 0; 3326 } 3327 3328 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 3329 struct net_device *dev, u16 mode, 3330 u32 flags, u32 mask, int nlflags, 3331 u32 filter_mask, 3332 int (*vlan_fill)(struct sk_buff *skb, 3333 struct net_device *dev, 3334 u32 filter_mask)) 3335 { 3336 struct nlmsghdr *nlh; 3337 struct ifinfomsg *ifm; 3338 struct nlattr *br_afspec; 3339 struct nlattr *protinfo; 3340 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 3341 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3342 int err = 0; 3343 3344 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 3345 if (nlh == NULL) 3346 return -EMSGSIZE; 3347 3348 ifm = nlmsg_data(nlh); 3349 ifm->ifi_family = AF_BRIDGE; 3350 ifm->__ifi_pad = 0; 3351 ifm->ifi_type = dev->type; 3352 ifm->ifi_index = dev->ifindex; 3353 ifm->ifi_flags = dev_get_flags(dev); 3354 ifm->ifi_change = 0; 3355 3356 3357 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 3358 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 3359 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 3360 (br_dev && 3361 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 3362 (dev->addr_len && 3363 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 3364 (dev->ifindex != dev_get_iflink(dev) && 3365 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 3366 goto nla_put_failure; 3367 3368 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 3369 if (!br_afspec) 3370 goto nla_put_failure; 3371 3372 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 3373 nla_nest_cancel(skb, br_afspec); 3374 goto nla_put_failure; 3375 } 3376 3377 if (mode != BRIDGE_MODE_UNDEF) { 3378 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 3379 nla_nest_cancel(skb, br_afspec); 3380 goto nla_put_failure; 3381 } 3382 } 3383 if (vlan_fill) { 3384 err = vlan_fill(skb, dev, filter_mask); 3385 if (err) { 3386 nla_nest_cancel(skb, br_afspec); 3387 goto nla_put_failure; 3388 } 3389 } 3390 nla_nest_end(skb, br_afspec); 3391 3392 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED); 3393 if (!protinfo) 3394 goto nla_put_failure; 3395 3396 if (brport_nla_put_flag(skb, flags, mask, 3397 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 3398 brport_nla_put_flag(skb, flags, mask, 3399 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 3400 brport_nla_put_flag(skb, flags, mask, 3401 IFLA_BRPORT_FAST_LEAVE, 3402 BR_MULTICAST_FAST_LEAVE) || 3403 brport_nla_put_flag(skb, flags, mask, 3404 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 3405 brport_nla_put_flag(skb, flags, mask, 3406 IFLA_BRPORT_LEARNING, BR_LEARNING) || 3407 brport_nla_put_flag(skb, flags, mask, 3408 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 3409 brport_nla_put_flag(skb, flags, mask, 3410 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 3411 brport_nla_put_flag(skb, flags, mask, 3412 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 3413 nla_nest_cancel(skb, protinfo); 3414 goto nla_put_failure; 3415 } 3416 3417 nla_nest_end(skb, protinfo); 3418 3419 nlmsg_end(skb, nlh); 3420 return 0; 3421 nla_put_failure: 3422 nlmsg_cancel(skb, nlh); 3423 return err ? err : -EMSGSIZE; 3424 } 3425 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 3426 3427 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 3428 { 3429 struct net *net = sock_net(skb->sk); 3430 struct net_device *dev; 3431 int idx = 0; 3432 u32 portid = NETLINK_CB(cb->skb).portid; 3433 u32 seq = cb->nlh->nlmsg_seq; 3434 u32 filter_mask = 0; 3435 int err; 3436 3437 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 3438 struct nlattr *extfilt; 3439 3440 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 3441 IFLA_EXT_MASK); 3442 if (extfilt) { 3443 if (nla_len(extfilt) < sizeof(filter_mask)) 3444 return -EINVAL; 3445 3446 filter_mask = nla_get_u32(extfilt); 3447 } 3448 } 3449 3450 rcu_read_lock(); 3451 for_each_netdev_rcu(net, dev) { 3452 const struct net_device_ops *ops = dev->netdev_ops; 3453 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3454 3455 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 3456 if (idx >= cb->args[0]) { 3457 err = br_dev->netdev_ops->ndo_bridge_getlink( 3458 skb, portid, seq, dev, 3459 filter_mask, NLM_F_MULTI); 3460 if (err < 0 && err != -EOPNOTSUPP) 3461 break; 3462 } 3463 idx++; 3464 } 3465 3466 if (ops->ndo_bridge_getlink) { 3467 if (idx >= cb->args[0]) { 3468 err = ops->ndo_bridge_getlink(skb, portid, 3469 seq, dev, 3470 filter_mask, 3471 NLM_F_MULTI); 3472 if (err < 0 && err != -EOPNOTSUPP) 3473 break; 3474 } 3475 idx++; 3476 } 3477 } 3478 rcu_read_unlock(); 3479 cb->args[0] = idx; 3480 3481 return skb->len; 3482 } 3483 3484 static inline size_t bridge_nlmsg_size(void) 3485 { 3486 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 3487 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 3488 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 3489 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 3490 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 3491 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 3492 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 3493 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 3494 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 3495 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 3496 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 3497 } 3498 3499 static int rtnl_bridge_notify(struct net_device *dev) 3500 { 3501 struct net *net = dev_net(dev); 3502 struct sk_buff *skb; 3503 int err = -EOPNOTSUPP; 3504 3505 if (!dev->netdev_ops->ndo_bridge_getlink) 3506 return 0; 3507 3508 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 3509 if (!skb) { 3510 err = -ENOMEM; 3511 goto errout; 3512 } 3513 3514 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 3515 if (err < 0) 3516 goto errout; 3517 3518 if (!skb->len) 3519 goto errout; 3520 3521 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 3522 return 0; 3523 errout: 3524 WARN_ON(err == -EMSGSIZE); 3525 kfree_skb(skb); 3526 if (err) 3527 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3528 return err; 3529 } 3530 3531 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3532 struct netlink_ext_ack *extack) 3533 { 3534 struct net *net = sock_net(skb->sk); 3535 struct ifinfomsg *ifm; 3536 struct net_device *dev; 3537 struct nlattr *br_spec, *attr = NULL; 3538 int rem, err = -EOPNOTSUPP; 3539 u16 flags = 0; 3540 bool have_flags = false; 3541 3542 if (nlmsg_len(nlh) < sizeof(*ifm)) 3543 return -EINVAL; 3544 3545 ifm = nlmsg_data(nlh); 3546 if (ifm->ifi_family != AF_BRIDGE) 3547 return -EPFNOSUPPORT; 3548 3549 dev = __dev_get_by_index(net, ifm->ifi_index); 3550 if (!dev) { 3551 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3552 return -ENODEV; 3553 } 3554 3555 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3556 if (br_spec) { 3557 nla_for_each_nested(attr, br_spec, rem) { 3558 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3559 if (nla_len(attr) < sizeof(flags)) 3560 return -EINVAL; 3561 3562 have_flags = true; 3563 flags = nla_get_u16(attr); 3564 break; 3565 } 3566 } 3567 } 3568 3569 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3570 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3571 3572 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 3573 err = -EOPNOTSUPP; 3574 goto out; 3575 } 3576 3577 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags); 3578 if (err) 3579 goto out; 3580 3581 flags &= ~BRIDGE_FLAGS_MASTER; 3582 } 3583 3584 if ((flags & BRIDGE_FLAGS_SELF)) { 3585 if (!dev->netdev_ops->ndo_bridge_setlink) 3586 err = -EOPNOTSUPP; 3587 else 3588 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 3589 flags); 3590 if (!err) { 3591 flags &= ~BRIDGE_FLAGS_SELF; 3592 3593 /* Generate event to notify upper layer of bridge 3594 * change 3595 */ 3596 err = rtnl_bridge_notify(dev); 3597 } 3598 } 3599 3600 if (have_flags) 3601 memcpy(nla_data(attr), &flags, sizeof(flags)); 3602 out: 3603 return err; 3604 } 3605 3606 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 3607 struct netlink_ext_ack *extack) 3608 { 3609 struct net *net = sock_net(skb->sk); 3610 struct ifinfomsg *ifm; 3611 struct net_device *dev; 3612 struct nlattr *br_spec, *attr = NULL; 3613 int rem, err = -EOPNOTSUPP; 3614 u16 flags = 0; 3615 bool have_flags = false; 3616 3617 if (nlmsg_len(nlh) < sizeof(*ifm)) 3618 return -EINVAL; 3619 3620 ifm = nlmsg_data(nlh); 3621 if (ifm->ifi_family != AF_BRIDGE) 3622 return -EPFNOSUPPORT; 3623 3624 dev = __dev_get_by_index(net, ifm->ifi_index); 3625 if (!dev) { 3626 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3627 return -ENODEV; 3628 } 3629 3630 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3631 if (br_spec) { 3632 nla_for_each_nested(attr, br_spec, rem) { 3633 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3634 if (nla_len(attr) < sizeof(flags)) 3635 return -EINVAL; 3636 3637 have_flags = true; 3638 flags = nla_get_u16(attr); 3639 break; 3640 } 3641 } 3642 } 3643 3644 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3645 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3646 3647 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 3648 err = -EOPNOTSUPP; 3649 goto out; 3650 } 3651 3652 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 3653 if (err) 3654 goto out; 3655 3656 flags &= ~BRIDGE_FLAGS_MASTER; 3657 } 3658 3659 if ((flags & BRIDGE_FLAGS_SELF)) { 3660 if (!dev->netdev_ops->ndo_bridge_dellink) 3661 err = -EOPNOTSUPP; 3662 else 3663 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 3664 flags); 3665 3666 if (!err) { 3667 flags &= ~BRIDGE_FLAGS_SELF; 3668 3669 /* Generate event to notify upper layer of bridge 3670 * change 3671 */ 3672 err = rtnl_bridge_notify(dev); 3673 } 3674 } 3675 3676 if (have_flags) 3677 memcpy(nla_data(attr), &flags, sizeof(flags)); 3678 out: 3679 return err; 3680 } 3681 3682 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 3683 { 3684 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 3685 (!idxattr || idxattr == attrid); 3686 } 3687 3688 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 3689 static int rtnl_get_offload_stats_attr_size(int attr_id) 3690 { 3691 switch (attr_id) { 3692 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 3693 return sizeof(struct rtnl_link_stats64); 3694 } 3695 3696 return 0; 3697 } 3698 3699 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 3700 int *prividx) 3701 { 3702 struct nlattr *attr = NULL; 3703 int attr_id, size; 3704 void *attr_data; 3705 int err; 3706 3707 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3708 dev->netdev_ops->ndo_get_offload_stats)) 3709 return -ENODATA; 3710 3711 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3712 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3713 if (attr_id < *prividx) 3714 continue; 3715 3716 size = rtnl_get_offload_stats_attr_size(attr_id); 3717 if (!size) 3718 continue; 3719 3720 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3721 continue; 3722 3723 attr = nla_reserve_64bit(skb, attr_id, size, 3724 IFLA_OFFLOAD_XSTATS_UNSPEC); 3725 if (!attr) 3726 goto nla_put_failure; 3727 3728 attr_data = nla_data(attr); 3729 memset(attr_data, 0, size); 3730 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 3731 attr_data); 3732 if (err) 3733 goto get_offload_stats_failure; 3734 } 3735 3736 if (!attr) 3737 return -ENODATA; 3738 3739 *prividx = 0; 3740 return 0; 3741 3742 nla_put_failure: 3743 err = -EMSGSIZE; 3744 get_offload_stats_failure: 3745 *prividx = attr_id; 3746 return err; 3747 } 3748 3749 static int rtnl_get_offload_stats_size(const struct net_device *dev) 3750 { 3751 int nla_size = 0; 3752 int attr_id; 3753 int size; 3754 3755 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3756 dev->netdev_ops->ndo_get_offload_stats)) 3757 return 0; 3758 3759 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3760 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3761 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3762 continue; 3763 size = rtnl_get_offload_stats_attr_size(attr_id); 3764 nla_size += nla_total_size_64bit(size); 3765 } 3766 3767 if (nla_size != 0) 3768 nla_size += nla_total_size(0); 3769 3770 return nla_size; 3771 } 3772 3773 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 3774 int type, u32 pid, u32 seq, u32 change, 3775 unsigned int flags, unsigned int filter_mask, 3776 int *idxattr, int *prividx) 3777 { 3778 struct if_stats_msg *ifsm; 3779 struct nlmsghdr *nlh; 3780 struct nlattr *attr; 3781 int s_prividx = *prividx; 3782 int err; 3783 3784 ASSERT_RTNL(); 3785 3786 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 3787 if (!nlh) 3788 return -EMSGSIZE; 3789 3790 ifsm = nlmsg_data(nlh); 3791 ifsm->ifindex = dev->ifindex; 3792 ifsm->filter_mask = filter_mask; 3793 3794 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 3795 struct rtnl_link_stats64 *sp; 3796 3797 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 3798 sizeof(struct rtnl_link_stats64), 3799 IFLA_STATS_UNSPEC); 3800 if (!attr) 3801 goto nla_put_failure; 3802 3803 sp = nla_data(attr); 3804 dev_get_stats(dev, sp); 3805 } 3806 3807 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 3808 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3809 3810 if (ops && ops->fill_linkxstats) { 3811 *idxattr = IFLA_STATS_LINK_XSTATS; 3812 attr = nla_nest_start(skb, 3813 IFLA_STATS_LINK_XSTATS); 3814 if (!attr) 3815 goto nla_put_failure; 3816 3817 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3818 nla_nest_end(skb, attr); 3819 if (err) 3820 goto nla_put_failure; 3821 *idxattr = 0; 3822 } 3823 } 3824 3825 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 3826 *idxattr)) { 3827 const struct rtnl_link_ops *ops = NULL; 3828 const struct net_device *master; 3829 3830 master = netdev_master_upper_dev_get(dev); 3831 if (master) 3832 ops = master->rtnl_link_ops; 3833 if (ops && ops->fill_linkxstats) { 3834 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 3835 attr = nla_nest_start(skb, 3836 IFLA_STATS_LINK_XSTATS_SLAVE); 3837 if (!attr) 3838 goto nla_put_failure; 3839 3840 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3841 nla_nest_end(skb, attr); 3842 if (err) 3843 goto nla_put_failure; 3844 *idxattr = 0; 3845 } 3846 } 3847 3848 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 3849 *idxattr)) { 3850 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 3851 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS); 3852 if (!attr) 3853 goto nla_put_failure; 3854 3855 err = rtnl_get_offload_stats(skb, dev, prividx); 3856 if (err == -ENODATA) 3857 nla_nest_cancel(skb, attr); 3858 else 3859 nla_nest_end(skb, attr); 3860 3861 if (err && err != -ENODATA) 3862 goto nla_put_failure; 3863 *idxattr = 0; 3864 } 3865 3866 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 3867 struct rtnl_af_ops *af_ops; 3868 3869 *idxattr = IFLA_STATS_AF_SPEC; 3870 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC); 3871 if (!attr) 3872 goto nla_put_failure; 3873 3874 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3875 if (af_ops->fill_stats_af) { 3876 struct nlattr *af; 3877 int err; 3878 3879 af = nla_nest_start(skb, af_ops->family); 3880 if (!af) 3881 goto nla_put_failure; 3882 3883 err = af_ops->fill_stats_af(skb, dev); 3884 3885 if (err == -ENODATA) 3886 nla_nest_cancel(skb, af); 3887 else if (err < 0) 3888 goto nla_put_failure; 3889 3890 nla_nest_end(skb, af); 3891 } 3892 } 3893 3894 nla_nest_end(skb, attr); 3895 3896 *idxattr = 0; 3897 } 3898 3899 nlmsg_end(skb, nlh); 3900 3901 return 0; 3902 3903 nla_put_failure: 3904 /* not a multi message or no progress mean a real error */ 3905 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 3906 nlmsg_cancel(skb, nlh); 3907 else 3908 nlmsg_end(skb, nlh); 3909 3910 return -EMSGSIZE; 3911 } 3912 3913 static size_t if_nlmsg_stats_size(const struct net_device *dev, 3914 u32 filter_mask) 3915 { 3916 size_t size = 0; 3917 3918 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 3919 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 3920 3921 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 3922 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3923 int attr = IFLA_STATS_LINK_XSTATS; 3924 3925 if (ops && ops->get_linkxstats_size) { 3926 size += nla_total_size(ops->get_linkxstats_size(dev, 3927 attr)); 3928 /* for IFLA_STATS_LINK_XSTATS */ 3929 size += nla_total_size(0); 3930 } 3931 } 3932 3933 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 3934 struct net_device *_dev = (struct net_device *)dev; 3935 const struct rtnl_link_ops *ops = NULL; 3936 const struct net_device *master; 3937 3938 /* netdev_master_upper_dev_get can't take const */ 3939 master = netdev_master_upper_dev_get(_dev); 3940 if (master) 3941 ops = master->rtnl_link_ops; 3942 if (ops && ops->get_linkxstats_size) { 3943 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 3944 3945 size += nla_total_size(ops->get_linkxstats_size(dev, 3946 attr)); 3947 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 3948 size += nla_total_size(0); 3949 } 3950 } 3951 3952 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 3953 size += rtnl_get_offload_stats_size(dev); 3954 3955 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 3956 struct rtnl_af_ops *af_ops; 3957 3958 /* for IFLA_STATS_AF_SPEC */ 3959 size += nla_total_size(0); 3960 3961 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3962 if (af_ops->get_stats_af_size) { 3963 size += nla_total_size( 3964 af_ops->get_stats_af_size(dev)); 3965 3966 /* for AF_* */ 3967 size += nla_total_size(0); 3968 } 3969 } 3970 } 3971 3972 return size; 3973 } 3974 3975 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh, 3976 struct netlink_ext_ack *extack) 3977 { 3978 struct net *net = sock_net(skb->sk); 3979 struct net_device *dev = NULL; 3980 int idxattr = 0, prividx = 0; 3981 struct if_stats_msg *ifsm; 3982 struct sk_buff *nskb; 3983 u32 filter_mask; 3984 int err; 3985 3986 if (nlmsg_len(nlh) < sizeof(*ifsm)) 3987 return -EINVAL; 3988 3989 ifsm = nlmsg_data(nlh); 3990 if (ifsm->ifindex > 0) 3991 dev = __dev_get_by_index(net, ifsm->ifindex); 3992 else 3993 return -EINVAL; 3994 3995 if (!dev) 3996 return -ENODEV; 3997 3998 filter_mask = ifsm->filter_mask; 3999 if (!filter_mask) 4000 return -EINVAL; 4001 4002 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 4003 if (!nskb) 4004 return -ENOBUFS; 4005 4006 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 4007 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 4008 0, filter_mask, &idxattr, &prividx); 4009 if (err < 0) { 4010 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 4011 WARN_ON(err == -EMSGSIZE); 4012 kfree_skb(nskb); 4013 } else { 4014 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 4015 } 4016 4017 return err; 4018 } 4019 4020 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 4021 { 4022 int h, s_h, err, s_idx, s_idxattr, s_prividx; 4023 struct net *net = sock_net(skb->sk); 4024 unsigned int flags = NLM_F_MULTI; 4025 struct if_stats_msg *ifsm; 4026 struct hlist_head *head; 4027 struct net_device *dev; 4028 u32 filter_mask = 0; 4029 int idx = 0; 4030 4031 s_h = cb->args[0]; 4032 s_idx = cb->args[1]; 4033 s_idxattr = cb->args[2]; 4034 s_prividx = cb->args[3]; 4035 4036 cb->seq = net->dev_base_seq; 4037 4038 if (nlmsg_len(cb->nlh) < sizeof(*ifsm)) 4039 return -EINVAL; 4040 4041 ifsm = nlmsg_data(cb->nlh); 4042 filter_mask = ifsm->filter_mask; 4043 if (!filter_mask) 4044 return -EINVAL; 4045 4046 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4047 idx = 0; 4048 head = &net->dev_index_head[h]; 4049 hlist_for_each_entry(dev, head, index_hlist) { 4050 if (idx < s_idx) 4051 goto cont; 4052 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 4053 NETLINK_CB(cb->skb).portid, 4054 cb->nlh->nlmsg_seq, 0, 4055 flags, filter_mask, 4056 &s_idxattr, &s_prividx); 4057 /* If we ran out of room on the first message, 4058 * we're in trouble 4059 */ 4060 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 4061 4062 if (err < 0) 4063 goto out; 4064 s_prividx = 0; 4065 s_idxattr = 0; 4066 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4067 cont: 4068 idx++; 4069 } 4070 } 4071 out: 4072 cb->args[3] = s_prividx; 4073 cb->args[2] = s_idxattr; 4074 cb->args[1] = idx; 4075 cb->args[0] = h; 4076 4077 return skb->len; 4078 } 4079 4080 /* Process one rtnetlink message. */ 4081 4082 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, 4083 struct netlink_ext_ack *extack) 4084 { 4085 struct net *net = sock_net(skb->sk); 4086 rtnl_doit_func doit; 4087 int kind; 4088 int family; 4089 int type; 4090 int err; 4091 4092 type = nlh->nlmsg_type; 4093 if (type > RTM_MAX) 4094 return -EOPNOTSUPP; 4095 4096 type -= RTM_BASE; 4097 4098 /* All the messages must have at least 1 byte length */ 4099 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 4100 return 0; 4101 4102 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 4103 kind = type&3; 4104 4105 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 4106 return -EPERM; 4107 4108 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 4109 struct sock *rtnl; 4110 rtnl_dumpit_func dumpit; 4111 rtnl_calcit_func calcit; 4112 u16 min_dump_alloc = 0; 4113 4114 dumpit = rtnl_get_dumpit(family, type); 4115 if (dumpit == NULL) 4116 return -EOPNOTSUPP; 4117 calcit = rtnl_get_calcit(family, type); 4118 if (calcit) 4119 min_dump_alloc = calcit(skb, nlh); 4120 4121 __rtnl_unlock(); 4122 rtnl = net->rtnl; 4123 { 4124 struct netlink_dump_control c = { 4125 .dump = dumpit, 4126 .min_dump_alloc = min_dump_alloc, 4127 }; 4128 err = netlink_dump_start(rtnl, skb, nlh, &c); 4129 } 4130 rtnl_lock(); 4131 return err; 4132 } 4133 4134 doit = rtnl_get_doit(family, type); 4135 if (doit == NULL) 4136 return -EOPNOTSUPP; 4137 4138 return doit(skb, nlh, extack); 4139 } 4140 4141 static void rtnetlink_rcv(struct sk_buff *skb) 4142 { 4143 rtnl_lock(); 4144 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 4145 rtnl_unlock(); 4146 } 4147 4148 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 4149 { 4150 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4151 4152 switch (event) { 4153 case NETDEV_REBOOT: 4154 case NETDEV_CHANGENAME: 4155 case NETDEV_FEAT_CHANGE: 4156 case NETDEV_BONDING_FAILOVER: 4157 case NETDEV_NOTIFY_PEERS: 4158 case NETDEV_RESEND_IGMP: 4159 case NETDEV_CHANGEINFODATA: 4160 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL); 4161 break; 4162 default: 4163 break; 4164 } 4165 return NOTIFY_DONE; 4166 } 4167 4168 static struct notifier_block rtnetlink_dev_notifier = { 4169 .notifier_call = rtnetlink_event, 4170 }; 4171 4172 4173 static int __net_init rtnetlink_net_init(struct net *net) 4174 { 4175 struct sock *sk; 4176 struct netlink_kernel_cfg cfg = { 4177 .groups = RTNLGRP_MAX, 4178 .input = rtnetlink_rcv, 4179 .cb_mutex = &rtnl_mutex, 4180 .flags = NL_CFG_F_NONROOT_RECV, 4181 }; 4182 4183 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 4184 if (!sk) 4185 return -ENOMEM; 4186 net->rtnl = sk; 4187 return 0; 4188 } 4189 4190 static void __net_exit rtnetlink_net_exit(struct net *net) 4191 { 4192 netlink_kernel_release(net->rtnl); 4193 net->rtnl = NULL; 4194 } 4195 4196 static struct pernet_operations rtnetlink_net_ops = { 4197 .init = rtnetlink_net_init, 4198 .exit = rtnetlink_net_exit, 4199 }; 4200 4201 void __init rtnetlink_init(void) 4202 { 4203 if (register_pernet_subsys(&rtnetlink_net_ops)) 4204 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 4205 4206 register_netdevice_notifier(&rtnetlink_dev_notifier); 4207 4208 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 4209 rtnl_dump_ifinfo, rtnl_calcit); 4210 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL); 4211 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL); 4212 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL); 4213 4214 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL); 4215 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL); 4216 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL); 4217 4218 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL); 4219 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL); 4220 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL); 4221 4222 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL); 4223 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL); 4224 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL); 4225 4226 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 4227 NULL); 4228 } 4229