1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Routing netlink socket interface: protocol independent part. 7 * 8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 * 15 * Fixes: 16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong. 17 */ 18 19 #include <linux/errno.h> 20 #include <linux/module.h> 21 #include <linux/types.h> 22 #include <linux/socket.h> 23 #include <linux/kernel.h> 24 #include <linux/timer.h> 25 #include <linux/string.h> 26 #include <linux/sockios.h> 27 #include <linux/net.h> 28 #include <linux/fcntl.h> 29 #include <linux/mm.h> 30 #include <linux/slab.h> 31 #include <linux/interrupt.h> 32 #include <linux/capability.h> 33 #include <linux/skbuff.h> 34 #include <linux/init.h> 35 #include <linux/security.h> 36 #include <linux/mutex.h> 37 #include <linux/if_addr.h> 38 #include <linux/if_bridge.h> 39 #include <linux/if_vlan.h> 40 #include <linux/pci.h> 41 #include <linux/etherdevice.h> 42 43 #include <linux/uaccess.h> 44 45 #include <linux/inet.h> 46 #include <linux/netdevice.h> 47 #include <net/switchdev.h> 48 #include <net/ip.h> 49 #include <net/protocol.h> 50 #include <net/arp.h> 51 #include <net/route.h> 52 #include <net/udp.h> 53 #include <net/tcp.h> 54 #include <net/sock.h> 55 #include <net/pkt_sched.h> 56 #include <net/fib_rules.h> 57 #include <net/rtnetlink.h> 58 #include <net/net_namespace.h> 59 60 struct rtnl_link { 61 rtnl_doit_func doit; 62 rtnl_dumpit_func dumpit; 63 rtnl_calcit_func calcit; 64 }; 65 66 static DEFINE_MUTEX(rtnl_mutex); 67 68 void rtnl_lock(void) 69 { 70 mutex_lock(&rtnl_mutex); 71 } 72 EXPORT_SYMBOL(rtnl_lock); 73 74 static struct sk_buff *defer_kfree_skb_list; 75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail) 76 { 77 if (head && tail) { 78 tail->next = defer_kfree_skb_list; 79 defer_kfree_skb_list = head; 80 } 81 } 82 EXPORT_SYMBOL(rtnl_kfree_skbs); 83 84 void __rtnl_unlock(void) 85 { 86 struct sk_buff *head = defer_kfree_skb_list; 87 88 defer_kfree_skb_list = NULL; 89 90 mutex_unlock(&rtnl_mutex); 91 92 while (head) { 93 struct sk_buff *next = head->next; 94 95 kfree_skb(head); 96 cond_resched(); 97 head = next; 98 } 99 } 100 101 void rtnl_unlock(void) 102 { 103 /* This fellow will unlock it for us. */ 104 netdev_run_todo(); 105 } 106 EXPORT_SYMBOL(rtnl_unlock); 107 108 int rtnl_trylock(void) 109 { 110 return mutex_trylock(&rtnl_mutex); 111 } 112 EXPORT_SYMBOL(rtnl_trylock); 113 114 int rtnl_is_locked(void) 115 { 116 return mutex_is_locked(&rtnl_mutex); 117 } 118 EXPORT_SYMBOL(rtnl_is_locked); 119 120 #ifdef CONFIG_PROVE_LOCKING 121 bool lockdep_rtnl_is_held(void) 122 { 123 return lockdep_is_held(&rtnl_mutex); 124 } 125 EXPORT_SYMBOL(lockdep_rtnl_is_held); 126 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 127 128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 129 130 static inline int rtm_msgindex(int msgtype) 131 { 132 int msgindex = msgtype - RTM_BASE; 133 134 /* 135 * msgindex < 0 implies someone tried to register a netlink 136 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 137 * the message type has not been added to linux/rtnetlink.h 138 */ 139 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 140 141 return msgindex; 142 } 143 144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex) 145 { 146 struct rtnl_link *tab; 147 148 if (protocol <= RTNL_FAMILY_MAX) 149 tab = rtnl_msg_handlers[protocol]; 150 else 151 tab = NULL; 152 153 if (tab == NULL || tab[msgindex].doit == NULL) 154 tab = rtnl_msg_handlers[PF_UNSPEC]; 155 156 return tab[msgindex].doit; 157 } 158 159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex) 160 { 161 struct rtnl_link *tab; 162 163 if (protocol <= RTNL_FAMILY_MAX) 164 tab = rtnl_msg_handlers[protocol]; 165 else 166 tab = NULL; 167 168 if (tab == NULL || tab[msgindex].dumpit == NULL) 169 tab = rtnl_msg_handlers[PF_UNSPEC]; 170 171 return tab[msgindex].dumpit; 172 } 173 174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex) 175 { 176 struct rtnl_link *tab; 177 178 if (protocol <= RTNL_FAMILY_MAX) 179 tab = rtnl_msg_handlers[protocol]; 180 else 181 tab = NULL; 182 183 if (tab == NULL || tab[msgindex].calcit == NULL) 184 tab = rtnl_msg_handlers[PF_UNSPEC]; 185 186 return tab[msgindex].calcit; 187 } 188 189 /** 190 * __rtnl_register - Register a rtnetlink message type 191 * @protocol: Protocol family or PF_UNSPEC 192 * @msgtype: rtnetlink message type 193 * @doit: Function pointer called for each request message 194 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 195 * @calcit: Function pointer to calc size of dump message 196 * 197 * Registers the specified function pointers (at least one of them has 198 * to be non-NULL) to be called whenever a request message for the 199 * specified protocol family and message type is received. 200 * 201 * The special protocol family PF_UNSPEC may be used to define fallback 202 * function pointers for the case when no entry for the specific protocol 203 * family exists. 204 * 205 * Returns 0 on success or a negative error code. 206 */ 207 int __rtnl_register(int protocol, int msgtype, 208 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 209 rtnl_calcit_func calcit) 210 { 211 struct rtnl_link *tab; 212 int msgindex; 213 214 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 215 msgindex = rtm_msgindex(msgtype); 216 217 tab = rtnl_msg_handlers[protocol]; 218 if (tab == NULL) { 219 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL); 220 if (tab == NULL) 221 return -ENOBUFS; 222 223 rtnl_msg_handlers[protocol] = tab; 224 } 225 226 if (doit) 227 tab[msgindex].doit = doit; 228 229 if (dumpit) 230 tab[msgindex].dumpit = dumpit; 231 232 if (calcit) 233 tab[msgindex].calcit = calcit; 234 235 return 0; 236 } 237 EXPORT_SYMBOL_GPL(__rtnl_register); 238 239 /** 240 * rtnl_register - Register a rtnetlink message type 241 * 242 * Identical to __rtnl_register() but panics on failure. This is useful 243 * as failure of this function is very unlikely, it can only happen due 244 * to lack of memory when allocating the chain to store all message 245 * handlers for a protocol. Meant for use in init functions where lack 246 * of memory implies no sense in continuing. 247 */ 248 void rtnl_register(int protocol, int msgtype, 249 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 250 rtnl_calcit_func calcit) 251 { 252 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0) 253 panic("Unable to register rtnetlink message handler, " 254 "protocol = %d, message type = %d\n", 255 protocol, msgtype); 256 } 257 EXPORT_SYMBOL_GPL(rtnl_register); 258 259 /** 260 * rtnl_unregister - Unregister a rtnetlink message type 261 * @protocol: Protocol family or PF_UNSPEC 262 * @msgtype: rtnetlink message type 263 * 264 * Returns 0 on success or a negative error code. 265 */ 266 int rtnl_unregister(int protocol, int msgtype) 267 { 268 int msgindex; 269 270 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 271 msgindex = rtm_msgindex(msgtype); 272 273 if (rtnl_msg_handlers[protocol] == NULL) 274 return -ENOENT; 275 276 rtnl_msg_handlers[protocol][msgindex].doit = NULL; 277 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL; 278 rtnl_msg_handlers[protocol][msgindex].calcit = NULL; 279 280 return 0; 281 } 282 EXPORT_SYMBOL_GPL(rtnl_unregister); 283 284 /** 285 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 286 * @protocol : Protocol family or PF_UNSPEC 287 * 288 * Identical to calling rtnl_unregster() for all registered message types 289 * of a certain protocol family. 290 */ 291 void rtnl_unregister_all(int protocol) 292 { 293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 294 295 kfree(rtnl_msg_handlers[protocol]); 296 rtnl_msg_handlers[protocol] = NULL; 297 } 298 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 299 300 static LIST_HEAD(link_ops); 301 302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 303 { 304 const struct rtnl_link_ops *ops; 305 306 list_for_each_entry(ops, &link_ops, list) { 307 if (!strcmp(ops->kind, kind)) 308 return ops; 309 } 310 return NULL; 311 } 312 313 /** 314 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 315 * @ops: struct rtnl_link_ops * to register 316 * 317 * The caller must hold the rtnl_mutex. This function should be used 318 * by drivers that create devices during module initialization. It 319 * must be called before registering the devices. 320 * 321 * Returns 0 on success or a negative error code. 322 */ 323 int __rtnl_link_register(struct rtnl_link_ops *ops) 324 { 325 if (rtnl_link_ops_get(ops->kind)) 326 return -EEXIST; 327 328 /* The check for setup is here because if ops 329 * does not have that filled up, it is not possible 330 * to use the ops for creating device. So do not 331 * fill up dellink as well. That disables rtnl_dellink. 332 */ 333 if (ops->setup && !ops->dellink) 334 ops->dellink = unregister_netdevice_queue; 335 336 list_add_tail(&ops->list, &link_ops); 337 return 0; 338 } 339 EXPORT_SYMBOL_GPL(__rtnl_link_register); 340 341 /** 342 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 343 * @ops: struct rtnl_link_ops * to register 344 * 345 * Returns 0 on success or a negative error code. 346 */ 347 int rtnl_link_register(struct rtnl_link_ops *ops) 348 { 349 int err; 350 351 rtnl_lock(); 352 err = __rtnl_link_register(ops); 353 rtnl_unlock(); 354 return err; 355 } 356 EXPORT_SYMBOL_GPL(rtnl_link_register); 357 358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 359 { 360 struct net_device *dev; 361 LIST_HEAD(list_kill); 362 363 for_each_netdev(net, dev) { 364 if (dev->rtnl_link_ops == ops) 365 ops->dellink(dev, &list_kill); 366 } 367 unregister_netdevice_many(&list_kill); 368 } 369 370 /** 371 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 372 * @ops: struct rtnl_link_ops * to unregister 373 * 374 * The caller must hold the rtnl_mutex. 375 */ 376 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 377 { 378 struct net *net; 379 380 for_each_net(net) { 381 __rtnl_kill_links(net, ops); 382 } 383 list_del(&ops->list); 384 } 385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 386 387 /* Return with the rtnl_lock held when there are no network 388 * devices unregistering in any network namespace. 389 */ 390 static void rtnl_lock_unregistering_all(void) 391 { 392 struct net *net; 393 bool unregistering; 394 DEFINE_WAIT_FUNC(wait, woken_wake_function); 395 396 add_wait_queue(&netdev_unregistering_wq, &wait); 397 for (;;) { 398 unregistering = false; 399 rtnl_lock(); 400 for_each_net(net) { 401 if (net->dev_unreg_count > 0) { 402 unregistering = true; 403 break; 404 } 405 } 406 if (!unregistering) 407 break; 408 __rtnl_unlock(); 409 410 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); 411 } 412 remove_wait_queue(&netdev_unregistering_wq, &wait); 413 } 414 415 /** 416 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 417 * @ops: struct rtnl_link_ops * to unregister 418 */ 419 void rtnl_link_unregister(struct rtnl_link_ops *ops) 420 { 421 /* Close the race with cleanup_net() */ 422 mutex_lock(&net_mutex); 423 rtnl_lock_unregistering_all(); 424 __rtnl_link_unregister(ops); 425 rtnl_unlock(); 426 mutex_unlock(&net_mutex); 427 } 428 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 429 430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev) 431 { 432 struct net_device *master_dev; 433 const struct rtnl_link_ops *ops; 434 435 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 436 if (!master_dev) 437 return 0; 438 ops = master_dev->rtnl_link_ops; 439 if (!ops || !ops->get_slave_size) 440 return 0; 441 /* IFLA_INFO_SLAVE_DATA + nested data */ 442 return nla_total_size(sizeof(struct nlattr)) + 443 ops->get_slave_size(master_dev, dev); 444 } 445 446 static size_t rtnl_link_get_size(const struct net_device *dev) 447 { 448 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 449 size_t size; 450 451 if (!ops) 452 return 0; 453 454 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 455 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 456 457 if (ops->get_size) 458 /* IFLA_INFO_DATA + nested data */ 459 size += nla_total_size(sizeof(struct nlattr)) + 460 ops->get_size(dev); 461 462 if (ops->get_xstats_size) 463 /* IFLA_INFO_XSTATS */ 464 size += nla_total_size(ops->get_xstats_size(dev)); 465 466 size += rtnl_link_get_slave_info_data_size(dev); 467 468 return size; 469 } 470 471 static LIST_HEAD(rtnl_af_ops); 472 473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family) 474 { 475 const struct rtnl_af_ops *ops; 476 477 list_for_each_entry(ops, &rtnl_af_ops, list) { 478 if (ops->family == family) 479 return ops; 480 } 481 482 return NULL; 483 } 484 485 /** 486 * rtnl_af_register - Register rtnl_af_ops with rtnetlink. 487 * @ops: struct rtnl_af_ops * to register 488 * 489 * Returns 0 on success or a negative error code. 490 */ 491 void rtnl_af_register(struct rtnl_af_ops *ops) 492 { 493 rtnl_lock(); 494 list_add_tail(&ops->list, &rtnl_af_ops); 495 rtnl_unlock(); 496 } 497 EXPORT_SYMBOL_GPL(rtnl_af_register); 498 499 /** 500 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 501 * @ops: struct rtnl_af_ops * to unregister 502 * 503 * The caller must hold the rtnl_mutex. 504 */ 505 void __rtnl_af_unregister(struct rtnl_af_ops *ops) 506 { 507 list_del(&ops->list); 508 } 509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister); 510 511 /** 512 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 513 * @ops: struct rtnl_af_ops * to unregister 514 */ 515 void rtnl_af_unregister(struct rtnl_af_ops *ops) 516 { 517 rtnl_lock(); 518 __rtnl_af_unregister(ops); 519 rtnl_unlock(); 520 } 521 EXPORT_SYMBOL_GPL(rtnl_af_unregister); 522 523 static size_t rtnl_link_get_af_size(const struct net_device *dev, 524 u32 ext_filter_mask) 525 { 526 struct rtnl_af_ops *af_ops; 527 size_t size; 528 529 /* IFLA_AF_SPEC */ 530 size = nla_total_size(sizeof(struct nlattr)); 531 532 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 533 if (af_ops->get_link_af_size) { 534 /* AF_* + nested data */ 535 size += nla_total_size(sizeof(struct nlattr)) + 536 af_ops->get_link_af_size(dev, ext_filter_mask); 537 } 538 } 539 540 return size; 541 } 542 543 static bool rtnl_have_link_slave_info(const struct net_device *dev) 544 { 545 struct net_device *master_dev; 546 547 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 548 if (master_dev && master_dev->rtnl_link_ops) 549 return true; 550 return false; 551 } 552 553 static int rtnl_link_slave_info_fill(struct sk_buff *skb, 554 const struct net_device *dev) 555 { 556 struct net_device *master_dev; 557 const struct rtnl_link_ops *ops; 558 struct nlattr *slave_data; 559 int err; 560 561 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 562 if (!master_dev) 563 return 0; 564 ops = master_dev->rtnl_link_ops; 565 if (!ops) 566 return 0; 567 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0) 568 return -EMSGSIZE; 569 if (ops->fill_slave_info) { 570 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA); 571 if (!slave_data) 572 return -EMSGSIZE; 573 err = ops->fill_slave_info(skb, master_dev, dev); 574 if (err < 0) 575 goto err_cancel_slave_data; 576 nla_nest_end(skb, slave_data); 577 } 578 return 0; 579 580 err_cancel_slave_data: 581 nla_nest_cancel(skb, slave_data); 582 return err; 583 } 584 585 static int rtnl_link_info_fill(struct sk_buff *skb, 586 const struct net_device *dev) 587 { 588 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 589 struct nlattr *data; 590 int err; 591 592 if (!ops) 593 return 0; 594 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 595 return -EMSGSIZE; 596 if (ops->fill_xstats) { 597 err = ops->fill_xstats(skb, dev); 598 if (err < 0) 599 return err; 600 } 601 if (ops->fill_info) { 602 data = nla_nest_start(skb, IFLA_INFO_DATA); 603 if (data == NULL) 604 return -EMSGSIZE; 605 err = ops->fill_info(skb, dev); 606 if (err < 0) 607 goto err_cancel_data; 608 nla_nest_end(skb, data); 609 } 610 return 0; 611 612 err_cancel_data: 613 nla_nest_cancel(skb, data); 614 return err; 615 } 616 617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 618 { 619 struct nlattr *linkinfo; 620 int err = -EMSGSIZE; 621 622 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 623 if (linkinfo == NULL) 624 goto out; 625 626 err = rtnl_link_info_fill(skb, dev); 627 if (err < 0) 628 goto err_cancel_link; 629 630 err = rtnl_link_slave_info_fill(skb, dev); 631 if (err < 0) 632 goto err_cancel_link; 633 634 nla_nest_end(skb, linkinfo); 635 return 0; 636 637 err_cancel_link: 638 nla_nest_cancel(skb, linkinfo); 639 out: 640 return err; 641 } 642 643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo) 644 { 645 struct sock *rtnl = net->rtnl; 646 int err = 0; 647 648 NETLINK_CB(skb).dst_group = group; 649 if (echo) 650 atomic_inc(&skb->users); 651 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 652 if (echo) 653 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 654 return err; 655 } 656 657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 658 { 659 struct sock *rtnl = net->rtnl; 660 661 return nlmsg_unicast(rtnl, skb, pid); 662 } 663 EXPORT_SYMBOL(rtnl_unicast); 664 665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 666 struct nlmsghdr *nlh, gfp_t flags) 667 { 668 struct sock *rtnl = net->rtnl; 669 int report = 0; 670 671 if (nlh) 672 report = nlmsg_report(nlh); 673 674 nlmsg_notify(rtnl, skb, pid, group, report, flags); 675 } 676 EXPORT_SYMBOL(rtnl_notify); 677 678 void rtnl_set_sk_err(struct net *net, u32 group, int error) 679 { 680 struct sock *rtnl = net->rtnl; 681 682 netlink_set_err(rtnl, 0, group, error); 683 } 684 EXPORT_SYMBOL(rtnl_set_sk_err); 685 686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 687 { 688 struct nlattr *mx; 689 int i, valid = 0; 690 691 mx = nla_nest_start(skb, RTA_METRICS); 692 if (mx == NULL) 693 return -ENOBUFS; 694 695 for (i = 0; i < RTAX_MAX; i++) { 696 if (metrics[i]) { 697 if (i == RTAX_CC_ALGO - 1) { 698 char tmp[TCP_CA_NAME_MAX], *name; 699 700 name = tcp_ca_get_name_by_key(metrics[i], tmp); 701 if (!name) 702 continue; 703 if (nla_put_string(skb, i + 1, name)) 704 goto nla_put_failure; 705 } else if (i == RTAX_FEATURES - 1) { 706 u32 user_features = metrics[i] & RTAX_FEATURE_MASK; 707 708 if (!user_features) 709 continue; 710 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK); 711 if (nla_put_u32(skb, i + 1, user_features)) 712 goto nla_put_failure; 713 } else { 714 if (nla_put_u32(skb, i + 1, metrics[i])) 715 goto nla_put_failure; 716 } 717 valid++; 718 } 719 } 720 721 if (!valid) { 722 nla_nest_cancel(skb, mx); 723 return 0; 724 } 725 726 return nla_nest_end(skb, mx); 727 728 nla_put_failure: 729 nla_nest_cancel(skb, mx); 730 return -EMSGSIZE; 731 } 732 EXPORT_SYMBOL(rtnetlink_put_metrics); 733 734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 735 long expires, u32 error) 736 { 737 struct rta_cacheinfo ci = { 738 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse), 739 .rta_used = dst->__use, 740 .rta_clntref = atomic_read(&(dst->__refcnt)), 741 .rta_error = error, 742 .rta_id = id, 743 }; 744 745 if (expires) { 746 unsigned long clock; 747 748 clock = jiffies_to_clock_t(abs(expires)); 749 clock = min_t(unsigned long, clock, INT_MAX); 750 ci.rta_expires = (expires > 0) ? clock : -clock; 751 } 752 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 753 } 754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 755 756 static void set_operstate(struct net_device *dev, unsigned char transition) 757 { 758 unsigned char operstate = dev->operstate; 759 760 switch (transition) { 761 case IF_OPER_UP: 762 if ((operstate == IF_OPER_DORMANT || 763 operstate == IF_OPER_UNKNOWN) && 764 !netif_dormant(dev)) 765 operstate = IF_OPER_UP; 766 break; 767 768 case IF_OPER_DORMANT: 769 if (operstate == IF_OPER_UP || 770 operstate == IF_OPER_UNKNOWN) 771 operstate = IF_OPER_DORMANT; 772 break; 773 } 774 775 if (dev->operstate != operstate) { 776 write_lock_bh(&dev_base_lock); 777 dev->operstate = operstate; 778 write_unlock_bh(&dev_base_lock); 779 netdev_state_change(dev); 780 } 781 } 782 783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev) 784 { 785 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) | 786 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI)); 787 } 788 789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 790 const struct ifinfomsg *ifm) 791 { 792 unsigned int flags = ifm->ifi_flags; 793 794 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 795 if (ifm->ifi_change) 796 flags = (flags & ifm->ifi_change) | 797 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change); 798 799 return flags; 800 } 801 802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 803 const struct rtnl_link_stats64 *b) 804 { 805 a->rx_packets = b->rx_packets; 806 a->tx_packets = b->tx_packets; 807 a->rx_bytes = b->rx_bytes; 808 a->tx_bytes = b->tx_bytes; 809 a->rx_errors = b->rx_errors; 810 a->tx_errors = b->tx_errors; 811 a->rx_dropped = b->rx_dropped; 812 a->tx_dropped = b->tx_dropped; 813 814 a->multicast = b->multicast; 815 a->collisions = b->collisions; 816 817 a->rx_length_errors = b->rx_length_errors; 818 a->rx_over_errors = b->rx_over_errors; 819 a->rx_crc_errors = b->rx_crc_errors; 820 a->rx_frame_errors = b->rx_frame_errors; 821 a->rx_fifo_errors = b->rx_fifo_errors; 822 a->rx_missed_errors = b->rx_missed_errors; 823 824 a->tx_aborted_errors = b->tx_aborted_errors; 825 a->tx_carrier_errors = b->tx_carrier_errors; 826 a->tx_fifo_errors = b->tx_fifo_errors; 827 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 828 a->tx_window_errors = b->tx_window_errors; 829 830 a->rx_compressed = b->rx_compressed; 831 a->tx_compressed = b->tx_compressed; 832 833 a->rx_nohandler = b->rx_nohandler; 834 } 835 836 /* All VF info */ 837 static inline int rtnl_vfinfo_size(const struct net_device *dev, 838 u32 ext_filter_mask) 839 { 840 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) { 841 int num_vfs = dev_num_vf(dev->dev.parent); 842 size_t size = nla_total_size(0); 843 size += num_vfs * 844 (nla_total_size(0) + 845 nla_total_size(sizeof(struct ifla_vf_mac)) + 846 nla_total_size(sizeof(struct ifla_vf_vlan)) + 847 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */ 848 nla_total_size(MAX_VLAN_LIST_LEN * 849 sizeof(struct ifla_vf_vlan_info)) + 850 nla_total_size(sizeof(struct ifla_vf_spoofchk)) + 851 nla_total_size(sizeof(struct ifla_vf_tx_rate)) + 852 nla_total_size(sizeof(struct ifla_vf_rate)) + 853 nla_total_size(sizeof(struct ifla_vf_link_state)) + 854 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) + 855 nla_total_size(0) + /* nest IFLA_VF_STATS */ 856 /* IFLA_VF_STATS_RX_PACKETS */ 857 nla_total_size_64bit(sizeof(__u64)) + 858 /* IFLA_VF_STATS_TX_PACKETS */ 859 nla_total_size_64bit(sizeof(__u64)) + 860 /* IFLA_VF_STATS_RX_BYTES */ 861 nla_total_size_64bit(sizeof(__u64)) + 862 /* IFLA_VF_STATS_TX_BYTES */ 863 nla_total_size_64bit(sizeof(__u64)) + 864 /* IFLA_VF_STATS_BROADCAST */ 865 nla_total_size_64bit(sizeof(__u64)) + 866 /* IFLA_VF_STATS_MULTICAST */ 867 nla_total_size_64bit(sizeof(__u64)) + 868 nla_total_size(sizeof(struct ifla_vf_trust))); 869 return size; 870 } else 871 return 0; 872 } 873 874 static size_t rtnl_port_size(const struct net_device *dev, 875 u32 ext_filter_mask) 876 { 877 size_t port_size = nla_total_size(4) /* PORT_VF */ 878 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 879 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 880 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 881 + nla_total_size(1) /* PROT_VDP_REQUEST */ 882 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 883 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 884 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 885 + port_size; 886 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 887 + port_size; 888 889 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 890 !(ext_filter_mask & RTEXT_FILTER_VF)) 891 return 0; 892 if (dev_num_vf(dev->dev.parent)) 893 return port_self_size + vf_ports_size + 894 vf_port_size * dev_num_vf(dev->dev.parent); 895 else 896 return port_self_size; 897 } 898 899 static size_t rtnl_xdp_size(void) 900 { 901 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */ 902 nla_total_size(1); /* XDP_ATTACHED */ 903 904 return xdp_size; 905 } 906 907 static noinline size_t if_nlmsg_size(const struct net_device *dev, 908 u32 ext_filter_mask) 909 { 910 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 911 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 912 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 913 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 914 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap)) 915 + nla_total_size(sizeof(struct rtnl_link_stats)) 916 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64)) 917 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 918 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 919 + nla_total_size(4) /* IFLA_TXQLEN */ 920 + nla_total_size(4) /* IFLA_WEIGHT */ 921 + nla_total_size(4) /* IFLA_MTU */ 922 + nla_total_size(4) /* IFLA_LINK */ 923 + nla_total_size(4) /* IFLA_MASTER */ 924 + nla_total_size(1) /* IFLA_CARRIER */ 925 + nla_total_size(4) /* IFLA_PROMISCUITY */ 926 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 927 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 928 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */ 929 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */ 930 + nla_total_size(1) /* IFLA_OPERSTATE */ 931 + nla_total_size(1) /* IFLA_LINKMODE */ 932 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 933 + nla_total_size(4) /* IFLA_LINK_NETNSID */ 934 + nla_total_size(4) /* IFLA_GROUP */ 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 memset(&ivi, 0, sizeof(ivi)); 1129 1130 /* Not all SR-IOV capable drivers support the 1131 * spoofcheck and "RSS query enable" query. Preset to 1132 * -1 so the user space tool can detect that the driver 1133 * didn't report anything. 1134 */ 1135 ivi.spoofchk = -1; 1136 ivi.rss_query_en = -1; 1137 ivi.trusted = -1; 1138 /* The default value for VF link state is "auto" 1139 * IFLA_VF_LINK_STATE_AUTO which equals zero 1140 */ 1141 ivi.linkstate = 0; 1142 /* VLAN Protocol by default is 802.1Q */ 1143 ivi.vlan_proto = htons(ETH_P_8021Q); 1144 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi)) 1145 return 0; 1146 1147 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info)); 1148 1149 vf_mac.vf = 1150 vf_vlan.vf = 1151 vf_vlan_info.vf = 1152 vf_rate.vf = 1153 vf_tx_rate.vf = 1154 vf_spoofchk.vf = 1155 vf_linkstate.vf = 1156 vf_rss_query_en.vf = 1157 vf_trust.vf = ivi.vf; 1158 1159 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 1160 vf_vlan.vlan = ivi.vlan; 1161 vf_vlan.qos = ivi.qos; 1162 vf_vlan_info.vlan = ivi.vlan; 1163 vf_vlan_info.qos = ivi.qos; 1164 vf_vlan_info.vlan_proto = ivi.vlan_proto; 1165 vf_tx_rate.rate = ivi.max_tx_rate; 1166 vf_rate.min_tx_rate = ivi.min_tx_rate; 1167 vf_rate.max_tx_rate = ivi.max_tx_rate; 1168 vf_spoofchk.setting = ivi.spoofchk; 1169 vf_linkstate.link_state = ivi.linkstate; 1170 vf_rss_query_en.setting = ivi.rss_query_en; 1171 vf_trust.setting = ivi.trusted; 1172 vf = nla_nest_start(skb, IFLA_VF_INFO); 1173 if (!vf) 1174 goto nla_put_vfinfo_failure; 1175 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) || 1176 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) || 1177 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate), 1178 &vf_rate) || 1179 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), 1180 &vf_tx_rate) || 1181 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk), 1182 &vf_spoofchk) || 1183 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate), 1184 &vf_linkstate) || 1185 nla_put(skb, IFLA_VF_RSS_QUERY_EN, 1186 sizeof(vf_rss_query_en), 1187 &vf_rss_query_en) || 1188 nla_put(skb, IFLA_VF_TRUST, 1189 sizeof(vf_trust), &vf_trust)) 1190 goto nla_put_vf_failure; 1191 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST); 1192 if (!vfvlanlist) 1193 goto nla_put_vf_failure; 1194 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info), 1195 &vf_vlan_info)) { 1196 nla_nest_cancel(skb, vfvlanlist); 1197 goto nla_put_vf_failure; 1198 } 1199 nla_nest_end(skb, vfvlanlist); 1200 memset(&vf_stats, 0, sizeof(vf_stats)); 1201 if (dev->netdev_ops->ndo_get_vf_stats) 1202 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num, 1203 &vf_stats); 1204 vfstats = nla_nest_start(skb, IFLA_VF_STATS); 1205 if (!vfstats) 1206 goto nla_put_vf_failure; 1207 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS, 1208 vf_stats.rx_packets, IFLA_VF_STATS_PAD) || 1209 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS, 1210 vf_stats.tx_packets, IFLA_VF_STATS_PAD) || 1211 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES, 1212 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) || 1213 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES, 1214 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) || 1215 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST, 1216 vf_stats.broadcast, IFLA_VF_STATS_PAD) || 1217 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST, 1218 vf_stats.multicast, IFLA_VF_STATS_PAD)) { 1219 nla_nest_cancel(skb, vfstats); 1220 goto nla_put_vf_failure; 1221 } 1222 nla_nest_end(skb, vfstats); 1223 nla_nest_end(skb, vf); 1224 return 0; 1225 1226 nla_put_vf_failure: 1227 nla_nest_cancel(skb, vf); 1228 nla_put_vfinfo_failure: 1229 nla_nest_cancel(skb, vfinfo); 1230 return -EMSGSIZE; 1231 } 1232 1233 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev) 1234 { 1235 struct rtnl_link_ifmap map; 1236 1237 memset(&map, 0, sizeof(map)); 1238 map.mem_start = dev->mem_start; 1239 map.mem_end = dev->mem_end; 1240 map.base_addr = dev->base_addr; 1241 map.irq = dev->irq; 1242 map.dma = dev->dma; 1243 map.port = dev->if_port; 1244 1245 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD)) 1246 return -EMSGSIZE; 1247 1248 return 0; 1249 } 1250 1251 static u8 rtnl_xdp_attached_mode(struct net_device *dev) 1252 { 1253 const struct net_device_ops *ops = dev->netdev_ops; 1254 1255 ASSERT_RTNL(); 1256 1257 if (rcu_access_pointer(dev->xdp_prog)) 1258 return XDP_ATTACHED_SKB; 1259 if (ops->ndo_xdp && __dev_xdp_attached(dev, ops->ndo_xdp)) 1260 return XDP_ATTACHED_DRV; 1261 1262 return XDP_ATTACHED_NONE; 1263 } 1264 1265 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev) 1266 { 1267 struct nlattr *xdp; 1268 int err; 1269 1270 xdp = nla_nest_start(skb, IFLA_XDP); 1271 if (!xdp) 1272 return -EMSGSIZE; 1273 1274 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, 1275 rtnl_xdp_attached_mode(dev)); 1276 if (err) 1277 goto err_cancel; 1278 1279 nla_nest_end(skb, xdp); 1280 return 0; 1281 1282 err_cancel: 1283 nla_nest_cancel(skb, xdp); 1284 return err; 1285 } 1286 1287 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 1288 int type, u32 pid, u32 seq, u32 change, 1289 unsigned int flags, u32 ext_filter_mask) 1290 { 1291 struct ifinfomsg *ifm; 1292 struct nlmsghdr *nlh; 1293 struct nlattr *af_spec; 1294 struct rtnl_af_ops *af_ops; 1295 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1296 1297 ASSERT_RTNL(); 1298 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 1299 if (nlh == NULL) 1300 return -EMSGSIZE; 1301 1302 ifm = nlmsg_data(nlh); 1303 ifm->ifi_family = AF_UNSPEC; 1304 ifm->__ifi_pad = 0; 1305 ifm->ifi_type = dev->type; 1306 ifm->ifi_index = dev->ifindex; 1307 ifm->ifi_flags = dev_get_flags(dev); 1308 ifm->ifi_change = change; 1309 1310 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 1311 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 1312 nla_put_u8(skb, IFLA_OPERSTATE, 1313 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1314 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1315 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1316 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1317 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1318 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1319 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) || 1320 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) || 1321 #ifdef CONFIG_RPS 1322 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1323 #endif 1324 (dev->ifindex != dev_get_iflink(dev) && 1325 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) || 1326 (upper_dev && 1327 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) || 1328 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1329 (dev->qdisc && 1330 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1331 (dev->ifalias && 1332 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) || 1333 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1334 atomic_read(&dev->carrier_changes)) || 1335 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down)) 1336 goto nla_put_failure; 1337 1338 if (rtnl_fill_link_ifmap(skb, dev)) 1339 goto nla_put_failure; 1340 1341 if (dev->addr_len) { 1342 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1343 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1344 goto nla_put_failure; 1345 } 1346 1347 if (rtnl_phys_port_id_fill(skb, dev)) 1348 goto nla_put_failure; 1349 1350 if (rtnl_phys_port_name_fill(skb, dev)) 1351 goto nla_put_failure; 1352 1353 if (rtnl_phys_switch_id_fill(skb, dev)) 1354 goto nla_put_failure; 1355 1356 if (rtnl_fill_stats(skb, dev)) 1357 goto nla_put_failure; 1358 1359 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) && 1360 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent))) 1361 goto nla_put_failure; 1362 1363 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent && 1364 ext_filter_mask & RTEXT_FILTER_VF) { 1365 int i; 1366 struct nlattr *vfinfo; 1367 int num_vfs = dev_num_vf(dev->dev.parent); 1368 1369 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1370 if (!vfinfo) 1371 goto nla_put_failure; 1372 for (i = 0; i < num_vfs; i++) { 1373 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo)) 1374 goto nla_put_failure; 1375 } 1376 1377 nla_nest_end(skb, vfinfo); 1378 } 1379 1380 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1381 goto nla_put_failure; 1382 1383 if (rtnl_xdp_fill(skb, dev)) 1384 goto nla_put_failure; 1385 1386 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1387 if (rtnl_link_fill(skb, dev) < 0) 1388 goto nla_put_failure; 1389 } 1390 1391 if (dev->rtnl_link_ops && 1392 dev->rtnl_link_ops->get_link_net) { 1393 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev); 1394 1395 if (!net_eq(dev_net(dev), link_net)) { 1396 int id = peernet2id_alloc(dev_net(dev), link_net); 1397 1398 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id)) 1399 goto nla_put_failure; 1400 } 1401 } 1402 1403 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC))) 1404 goto nla_put_failure; 1405 1406 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 1407 if (af_ops->fill_link_af) { 1408 struct nlattr *af; 1409 int err; 1410 1411 if (!(af = nla_nest_start(skb, af_ops->family))) 1412 goto nla_put_failure; 1413 1414 err = af_ops->fill_link_af(skb, dev, ext_filter_mask); 1415 1416 /* 1417 * Caller may return ENODATA to indicate that there 1418 * was no data to be dumped. This is not an error, it 1419 * means we should trim the attribute header and 1420 * continue. 1421 */ 1422 if (err == -ENODATA) 1423 nla_nest_cancel(skb, af); 1424 else if (err < 0) 1425 goto nla_put_failure; 1426 1427 nla_nest_end(skb, af); 1428 } 1429 } 1430 1431 nla_nest_end(skb, af_spec); 1432 1433 nlmsg_end(skb, nlh); 1434 return 0; 1435 1436 nla_put_failure: 1437 nlmsg_cancel(skb, nlh); 1438 return -EMSGSIZE; 1439 } 1440 1441 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1442 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1443 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1444 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1445 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1446 [IFLA_MTU] = { .type = NLA_U32 }, 1447 [IFLA_LINK] = { .type = NLA_U32 }, 1448 [IFLA_MASTER] = { .type = NLA_U32 }, 1449 [IFLA_CARRIER] = { .type = NLA_U8 }, 1450 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1451 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1452 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1453 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1454 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1455 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1456 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1457 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 1458 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1459 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1460 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1461 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1462 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1463 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1464 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1465 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1466 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1467 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1468 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN }, 1469 [IFLA_LINK_NETNSID] = { .type = NLA_S32 }, 1470 [IFLA_PROTO_DOWN] = { .type = NLA_U8 }, 1471 [IFLA_XDP] = { .type = NLA_NESTED }, 1472 [IFLA_GROUP] = { .type = NLA_U32 }, 1473 }; 1474 1475 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1476 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1477 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1478 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1479 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1480 }; 1481 1482 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1483 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) }, 1484 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) }, 1485 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED }, 1486 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) }, 1487 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) }, 1488 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) }, 1489 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) }, 1490 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) }, 1491 [IFLA_VF_STATS] = { .type = NLA_NESTED }, 1492 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) }, 1493 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1494 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) }, 1495 }; 1496 1497 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1498 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1499 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1500 .len = PORT_PROFILE_MAX }, 1501 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1502 .len = PORT_UUID_MAX }, 1503 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1504 .len = PORT_UUID_MAX }, 1505 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1506 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1507 1508 /* Unused, but we need to keep it here since user space could 1509 * fill it. It's also broken with regard to NLA_BINARY use in 1510 * combination with structs. 1511 */ 1512 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1513 .len = sizeof(struct ifla_port_vsi) }, 1514 }; 1515 1516 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = { 1517 [IFLA_XDP_FD] = { .type = NLA_S32 }, 1518 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 }, 1519 [IFLA_XDP_FLAGS] = { .type = NLA_U32 }, 1520 }; 1521 1522 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla) 1523 { 1524 const struct rtnl_link_ops *ops = NULL; 1525 struct nlattr *linfo[IFLA_INFO_MAX + 1]; 1526 1527 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, 1528 ifla_info_policy, NULL) < 0) 1529 return NULL; 1530 1531 if (linfo[IFLA_INFO_KIND]) { 1532 char kind[MODULE_NAME_LEN]; 1533 1534 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind)); 1535 ops = rtnl_link_ops_get(kind); 1536 } 1537 1538 return ops; 1539 } 1540 1541 static bool link_master_filtered(struct net_device *dev, int master_idx) 1542 { 1543 struct net_device *master; 1544 1545 if (!master_idx) 1546 return false; 1547 1548 master = netdev_master_upper_dev_get(dev); 1549 if (!master || master->ifindex != master_idx) 1550 return true; 1551 1552 return false; 1553 } 1554 1555 static bool link_kind_filtered(const struct net_device *dev, 1556 const struct rtnl_link_ops *kind_ops) 1557 { 1558 if (kind_ops && dev->rtnl_link_ops != kind_ops) 1559 return true; 1560 1561 return false; 1562 } 1563 1564 static bool link_dump_filtered(struct net_device *dev, 1565 int master_idx, 1566 const struct rtnl_link_ops *kind_ops) 1567 { 1568 if (link_master_filtered(dev, master_idx) || 1569 link_kind_filtered(dev, kind_ops)) 1570 return true; 1571 1572 return false; 1573 } 1574 1575 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1576 { 1577 struct net *net = sock_net(skb->sk); 1578 int h, s_h; 1579 int idx = 0, s_idx; 1580 struct net_device *dev; 1581 struct hlist_head *head; 1582 struct nlattr *tb[IFLA_MAX+1]; 1583 u32 ext_filter_mask = 0; 1584 const struct rtnl_link_ops *kind_ops = NULL; 1585 unsigned int flags = NLM_F_MULTI; 1586 int master_idx = 0; 1587 int err; 1588 int hdrlen; 1589 1590 s_h = cb->args[0]; 1591 s_idx = cb->args[1]; 1592 1593 cb->seq = net->dev_base_seq; 1594 1595 /* A hack to preserve kernel<->userspace interface. 1596 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1597 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1598 * what iproute2 < v3.9.0 used. 1599 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1600 * attribute, its netlink message is shorter than struct ifinfomsg. 1601 */ 1602 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1603 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1604 1605 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, 1606 ifla_policy, NULL) >= 0) { 1607 if (tb[IFLA_EXT_MASK]) 1608 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1609 1610 if (tb[IFLA_MASTER]) 1611 master_idx = nla_get_u32(tb[IFLA_MASTER]); 1612 1613 if (tb[IFLA_LINKINFO]) 1614 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]); 1615 1616 if (master_idx || kind_ops) 1617 flags |= NLM_F_DUMP_FILTERED; 1618 } 1619 1620 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1621 idx = 0; 1622 head = &net->dev_index_head[h]; 1623 hlist_for_each_entry(dev, head, index_hlist) { 1624 if (link_dump_filtered(dev, master_idx, kind_ops)) 1625 goto cont; 1626 if (idx < s_idx) 1627 goto cont; 1628 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 1629 NETLINK_CB(cb->skb).portid, 1630 cb->nlh->nlmsg_seq, 0, 1631 flags, 1632 ext_filter_mask); 1633 1634 if (err < 0) { 1635 if (likely(skb->len)) 1636 goto out; 1637 1638 goto out_err; 1639 } 1640 1641 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1642 cont: 1643 idx++; 1644 } 1645 } 1646 out: 1647 err = skb->len; 1648 out_err: 1649 cb->args[1] = idx; 1650 cb->args[0] = h; 1651 1652 return err; 1653 } 1654 1655 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len, 1656 struct netlink_ext_ack *exterr) 1657 { 1658 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr); 1659 } 1660 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1661 1662 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1663 { 1664 struct net *net; 1665 /* Examine the link attributes and figure out which 1666 * network namespace we are talking about. 1667 */ 1668 if (tb[IFLA_NET_NS_PID]) 1669 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1670 else if (tb[IFLA_NET_NS_FD]) 1671 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1672 else 1673 net = get_net(src_net); 1674 return net; 1675 } 1676 EXPORT_SYMBOL(rtnl_link_get_net); 1677 1678 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1679 { 1680 if (dev) { 1681 if (tb[IFLA_ADDRESS] && 1682 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1683 return -EINVAL; 1684 1685 if (tb[IFLA_BROADCAST] && 1686 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1687 return -EINVAL; 1688 } 1689 1690 if (tb[IFLA_AF_SPEC]) { 1691 struct nlattr *af; 1692 int rem, err; 1693 1694 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1695 const struct rtnl_af_ops *af_ops; 1696 1697 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1698 return -EAFNOSUPPORT; 1699 1700 if (!af_ops->set_link_af) 1701 return -EOPNOTSUPP; 1702 1703 if (af_ops->validate_link_af) { 1704 err = af_ops->validate_link_af(dev, af); 1705 if (err < 0) 1706 return err; 1707 } 1708 } 1709 } 1710 1711 return 0; 1712 } 1713 1714 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt, 1715 int guid_type) 1716 { 1717 const struct net_device_ops *ops = dev->netdev_ops; 1718 1719 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type); 1720 } 1721 1722 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type) 1723 { 1724 if (dev->type != ARPHRD_INFINIBAND) 1725 return -EOPNOTSUPP; 1726 1727 return handle_infiniband_guid(dev, ivt, guid_type); 1728 } 1729 1730 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb) 1731 { 1732 const struct net_device_ops *ops = dev->netdev_ops; 1733 int err = -EINVAL; 1734 1735 if (tb[IFLA_VF_MAC]) { 1736 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]); 1737 1738 err = -EOPNOTSUPP; 1739 if (ops->ndo_set_vf_mac) 1740 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1741 ivm->mac); 1742 if (err < 0) 1743 return err; 1744 } 1745 1746 if (tb[IFLA_VF_VLAN]) { 1747 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]); 1748 1749 err = -EOPNOTSUPP; 1750 if (ops->ndo_set_vf_vlan) 1751 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan, 1752 ivv->qos, 1753 htons(ETH_P_8021Q)); 1754 if (err < 0) 1755 return err; 1756 } 1757 1758 if (tb[IFLA_VF_VLAN_LIST]) { 1759 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN]; 1760 struct nlattr *attr; 1761 int rem, len = 0; 1762 1763 err = -EOPNOTSUPP; 1764 if (!ops->ndo_set_vf_vlan) 1765 return err; 1766 1767 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) { 1768 if (nla_type(attr) != IFLA_VF_VLAN_INFO || 1769 nla_len(attr) < NLA_HDRLEN) { 1770 return -EINVAL; 1771 } 1772 if (len >= MAX_VLAN_LIST_LEN) 1773 return -EOPNOTSUPP; 1774 ivvl[len] = nla_data(attr); 1775 1776 len++; 1777 } 1778 if (len == 0) 1779 return -EINVAL; 1780 1781 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan, 1782 ivvl[0]->qos, ivvl[0]->vlan_proto); 1783 if (err < 0) 1784 return err; 1785 } 1786 1787 if (tb[IFLA_VF_TX_RATE]) { 1788 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]); 1789 struct ifla_vf_info ivf; 1790 1791 err = -EOPNOTSUPP; 1792 if (ops->ndo_get_vf_config) 1793 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf); 1794 if (err < 0) 1795 return err; 1796 1797 err = -EOPNOTSUPP; 1798 if (ops->ndo_set_vf_rate) 1799 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1800 ivf.min_tx_rate, 1801 ivt->rate); 1802 if (err < 0) 1803 return err; 1804 } 1805 1806 if (tb[IFLA_VF_RATE]) { 1807 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]); 1808 1809 err = -EOPNOTSUPP; 1810 if (ops->ndo_set_vf_rate) 1811 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1812 ivt->min_tx_rate, 1813 ivt->max_tx_rate); 1814 if (err < 0) 1815 return err; 1816 } 1817 1818 if (tb[IFLA_VF_SPOOFCHK]) { 1819 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]); 1820 1821 err = -EOPNOTSUPP; 1822 if (ops->ndo_set_vf_spoofchk) 1823 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 1824 ivs->setting); 1825 if (err < 0) 1826 return err; 1827 } 1828 1829 if (tb[IFLA_VF_LINK_STATE]) { 1830 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]); 1831 1832 err = -EOPNOTSUPP; 1833 if (ops->ndo_set_vf_link_state) 1834 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 1835 ivl->link_state); 1836 if (err < 0) 1837 return err; 1838 } 1839 1840 if (tb[IFLA_VF_RSS_QUERY_EN]) { 1841 struct ifla_vf_rss_query_en *ivrssq_en; 1842 1843 err = -EOPNOTSUPP; 1844 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]); 1845 if (ops->ndo_set_vf_rss_query_en) 1846 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf, 1847 ivrssq_en->setting); 1848 if (err < 0) 1849 return err; 1850 } 1851 1852 if (tb[IFLA_VF_TRUST]) { 1853 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]); 1854 1855 err = -EOPNOTSUPP; 1856 if (ops->ndo_set_vf_trust) 1857 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting); 1858 if (err < 0) 1859 return err; 1860 } 1861 1862 if (tb[IFLA_VF_IB_NODE_GUID]) { 1863 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]); 1864 1865 if (!ops->ndo_set_vf_guid) 1866 return -EOPNOTSUPP; 1867 1868 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID); 1869 } 1870 1871 if (tb[IFLA_VF_IB_PORT_GUID]) { 1872 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]); 1873 1874 if (!ops->ndo_set_vf_guid) 1875 return -EOPNOTSUPP; 1876 1877 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID); 1878 } 1879 1880 return err; 1881 } 1882 1883 static int do_set_master(struct net_device *dev, int ifindex) 1884 { 1885 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1886 const struct net_device_ops *ops; 1887 int err; 1888 1889 if (upper_dev) { 1890 if (upper_dev->ifindex == ifindex) 1891 return 0; 1892 ops = upper_dev->netdev_ops; 1893 if (ops->ndo_del_slave) { 1894 err = ops->ndo_del_slave(upper_dev, dev); 1895 if (err) 1896 return err; 1897 } else { 1898 return -EOPNOTSUPP; 1899 } 1900 } 1901 1902 if (ifindex) { 1903 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 1904 if (!upper_dev) 1905 return -EINVAL; 1906 ops = upper_dev->netdev_ops; 1907 if (ops->ndo_add_slave) { 1908 err = ops->ndo_add_slave(upper_dev, dev); 1909 if (err) 1910 return err; 1911 } else { 1912 return -EOPNOTSUPP; 1913 } 1914 } 1915 return 0; 1916 } 1917 1918 #define DO_SETLINK_MODIFIED 0x01 1919 /* notify flag means notify + modified. */ 1920 #define DO_SETLINK_NOTIFY 0x03 1921 static int do_setlink(const struct sk_buff *skb, 1922 struct net_device *dev, struct ifinfomsg *ifm, 1923 struct netlink_ext_ack *extack, 1924 struct nlattr **tb, char *ifname, int status) 1925 { 1926 const struct net_device_ops *ops = dev->netdev_ops; 1927 int err; 1928 1929 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) { 1930 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1931 if (IS_ERR(net)) { 1932 err = PTR_ERR(net); 1933 goto errout; 1934 } 1935 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { 1936 put_net(net); 1937 err = -EPERM; 1938 goto errout; 1939 } 1940 err = dev_change_net_namespace(dev, net, ifname); 1941 put_net(net); 1942 if (err) 1943 goto errout; 1944 status |= DO_SETLINK_MODIFIED; 1945 } 1946 1947 if (tb[IFLA_MAP]) { 1948 struct rtnl_link_ifmap *u_map; 1949 struct ifmap k_map; 1950 1951 if (!ops->ndo_set_config) { 1952 err = -EOPNOTSUPP; 1953 goto errout; 1954 } 1955 1956 if (!netif_device_present(dev)) { 1957 err = -ENODEV; 1958 goto errout; 1959 } 1960 1961 u_map = nla_data(tb[IFLA_MAP]); 1962 k_map.mem_start = (unsigned long) u_map->mem_start; 1963 k_map.mem_end = (unsigned long) u_map->mem_end; 1964 k_map.base_addr = (unsigned short) u_map->base_addr; 1965 k_map.irq = (unsigned char) u_map->irq; 1966 k_map.dma = (unsigned char) u_map->dma; 1967 k_map.port = (unsigned char) u_map->port; 1968 1969 err = ops->ndo_set_config(dev, &k_map); 1970 if (err < 0) 1971 goto errout; 1972 1973 status |= DO_SETLINK_NOTIFY; 1974 } 1975 1976 if (tb[IFLA_ADDRESS]) { 1977 struct sockaddr *sa; 1978 int len; 1979 1980 len = sizeof(sa_family_t) + dev->addr_len; 1981 sa = kmalloc(len, GFP_KERNEL); 1982 if (!sa) { 1983 err = -ENOMEM; 1984 goto errout; 1985 } 1986 sa->sa_family = dev->type; 1987 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 1988 dev->addr_len); 1989 err = dev_set_mac_address(dev, sa); 1990 kfree(sa); 1991 if (err) 1992 goto errout; 1993 status |= DO_SETLINK_MODIFIED; 1994 } 1995 1996 if (tb[IFLA_MTU]) { 1997 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1998 if (err < 0) 1999 goto errout; 2000 status |= DO_SETLINK_MODIFIED; 2001 } 2002 2003 if (tb[IFLA_GROUP]) { 2004 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2005 status |= DO_SETLINK_NOTIFY; 2006 } 2007 2008 /* 2009 * Interface selected by interface index but interface 2010 * name provided implies that a name change has been 2011 * requested. 2012 */ 2013 if (ifm->ifi_index > 0 && ifname[0]) { 2014 err = dev_change_name(dev, ifname); 2015 if (err < 0) 2016 goto errout; 2017 status |= DO_SETLINK_MODIFIED; 2018 } 2019 2020 if (tb[IFLA_IFALIAS]) { 2021 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 2022 nla_len(tb[IFLA_IFALIAS])); 2023 if (err < 0) 2024 goto errout; 2025 status |= DO_SETLINK_NOTIFY; 2026 } 2027 2028 if (tb[IFLA_BROADCAST]) { 2029 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 2030 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 2031 } 2032 2033 if (ifm->ifi_flags || ifm->ifi_change) { 2034 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2035 if (err < 0) 2036 goto errout; 2037 } 2038 2039 if (tb[IFLA_MASTER]) { 2040 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2041 if (err) 2042 goto errout; 2043 status |= DO_SETLINK_MODIFIED; 2044 } 2045 2046 if (tb[IFLA_CARRIER]) { 2047 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 2048 if (err) 2049 goto errout; 2050 status |= DO_SETLINK_MODIFIED; 2051 } 2052 2053 if (tb[IFLA_TXQLEN]) { 2054 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]); 2055 unsigned long orig_len = dev->tx_queue_len; 2056 2057 if (dev->tx_queue_len ^ value) { 2058 dev->tx_queue_len = value; 2059 err = call_netdevice_notifiers( 2060 NETDEV_CHANGE_TX_QUEUE_LEN, dev); 2061 err = notifier_to_errno(err); 2062 if (err) { 2063 dev->tx_queue_len = orig_len; 2064 goto errout; 2065 } 2066 status |= DO_SETLINK_NOTIFY; 2067 } 2068 } 2069 2070 if (tb[IFLA_OPERSTATE]) 2071 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2072 2073 if (tb[IFLA_LINKMODE]) { 2074 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 2075 2076 write_lock_bh(&dev_base_lock); 2077 if (dev->link_mode ^ value) 2078 status |= DO_SETLINK_NOTIFY; 2079 dev->link_mode = value; 2080 write_unlock_bh(&dev_base_lock); 2081 } 2082 2083 if (tb[IFLA_VFINFO_LIST]) { 2084 struct nlattr *vfinfo[IFLA_VF_MAX + 1]; 2085 struct nlattr *attr; 2086 int rem; 2087 2088 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 2089 if (nla_type(attr) != IFLA_VF_INFO || 2090 nla_len(attr) < NLA_HDRLEN) { 2091 err = -EINVAL; 2092 goto errout; 2093 } 2094 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr, 2095 ifla_vf_policy, NULL); 2096 if (err < 0) 2097 goto errout; 2098 err = do_setvfinfo(dev, vfinfo); 2099 if (err < 0) 2100 goto errout; 2101 status |= DO_SETLINK_NOTIFY; 2102 } 2103 } 2104 err = 0; 2105 2106 if (tb[IFLA_VF_PORTS]) { 2107 struct nlattr *port[IFLA_PORT_MAX+1]; 2108 struct nlattr *attr; 2109 int vf; 2110 int rem; 2111 2112 err = -EOPNOTSUPP; 2113 if (!ops->ndo_set_vf_port) 2114 goto errout; 2115 2116 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 2117 if (nla_type(attr) != IFLA_VF_PORT || 2118 nla_len(attr) < NLA_HDRLEN) { 2119 err = -EINVAL; 2120 goto errout; 2121 } 2122 err = nla_parse_nested(port, IFLA_PORT_MAX, attr, 2123 ifla_port_policy, NULL); 2124 if (err < 0) 2125 goto errout; 2126 if (!port[IFLA_PORT_VF]) { 2127 err = -EOPNOTSUPP; 2128 goto errout; 2129 } 2130 vf = nla_get_u32(port[IFLA_PORT_VF]); 2131 err = ops->ndo_set_vf_port(dev, vf, port); 2132 if (err < 0) 2133 goto errout; 2134 status |= DO_SETLINK_NOTIFY; 2135 } 2136 } 2137 err = 0; 2138 2139 if (tb[IFLA_PORT_SELF]) { 2140 struct nlattr *port[IFLA_PORT_MAX+1]; 2141 2142 err = nla_parse_nested(port, IFLA_PORT_MAX, 2143 tb[IFLA_PORT_SELF], ifla_port_policy, 2144 NULL); 2145 if (err < 0) 2146 goto errout; 2147 2148 err = -EOPNOTSUPP; 2149 if (ops->ndo_set_vf_port) 2150 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 2151 if (err < 0) 2152 goto errout; 2153 status |= DO_SETLINK_NOTIFY; 2154 } 2155 2156 if (tb[IFLA_AF_SPEC]) { 2157 struct nlattr *af; 2158 int rem; 2159 2160 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 2161 const struct rtnl_af_ops *af_ops; 2162 2163 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 2164 BUG(); 2165 2166 err = af_ops->set_link_af(dev, af); 2167 if (err < 0) 2168 goto errout; 2169 2170 status |= DO_SETLINK_NOTIFY; 2171 } 2172 } 2173 err = 0; 2174 2175 if (tb[IFLA_PROTO_DOWN]) { 2176 err = dev_change_proto_down(dev, 2177 nla_get_u8(tb[IFLA_PROTO_DOWN])); 2178 if (err) 2179 goto errout; 2180 status |= DO_SETLINK_NOTIFY; 2181 } 2182 2183 if (tb[IFLA_XDP]) { 2184 struct nlattr *xdp[IFLA_XDP_MAX + 1]; 2185 u32 xdp_flags = 0; 2186 2187 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP], 2188 ifla_xdp_policy, NULL); 2189 if (err < 0) 2190 goto errout; 2191 2192 if (xdp[IFLA_XDP_ATTACHED]) { 2193 err = -EINVAL; 2194 goto errout; 2195 } 2196 2197 if (xdp[IFLA_XDP_FLAGS]) { 2198 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]); 2199 if (xdp_flags & ~XDP_FLAGS_MASK) { 2200 err = -EINVAL; 2201 goto errout; 2202 } 2203 if ((xdp_flags & XDP_FLAGS_SKB_MODE) && 2204 (xdp_flags & XDP_FLAGS_DRV_MODE)) { 2205 err = -EINVAL; 2206 goto errout; 2207 } 2208 } 2209 2210 if (xdp[IFLA_XDP_FD]) { 2211 err = dev_change_xdp_fd(dev, extack, 2212 nla_get_s32(xdp[IFLA_XDP_FD]), 2213 xdp_flags); 2214 if (err) 2215 goto errout; 2216 status |= DO_SETLINK_NOTIFY; 2217 } 2218 } 2219 2220 errout: 2221 if (status & DO_SETLINK_MODIFIED) { 2222 if (status & DO_SETLINK_NOTIFY) 2223 netdev_state_change(dev); 2224 2225 if (err < 0) 2226 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", 2227 dev->name); 2228 } 2229 2230 return err; 2231 } 2232 2233 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2234 struct netlink_ext_ack *extack) 2235 { 2236 struct net *net = sock_net(skb->sk); 2237 struct ifinfomsg *ifm; 2238 struct net_device *dev; 2239 int err; 2240 struct nlattr *tb[IFLA_MAX+1]; 2241 char ifname[IFNAMSIZ]; 2242 2243 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, 2244 extack); 2245 if (err < 0) 2246 goto errout; 2247 2248 if (tb[IFLA_IFNAME]) 2249 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2250 else 2251 ifname[0] = '\0'; 2252 2253 err = -EINVAL; 2254 ifm = nlmsg_data(nlh); 2255 if (ifm->ifi_index > 0) 2256 dev = __dev_get_by_index(net, ifm->ifi_index); 2257 else if (tb[IFLA_IFNAME]) 2258 dev = __dev_get_by_name(net, ifname); 2259 else 2260 goto errout; 2261 2262 if (dev == NULL) { 2263 err = -ENODEV; 2264 goto errout; 2265 } 2266 2267 err = validate_linkmsg(dev, tb); 2268 if (err < 0) 2269 goto errout; 2270 2271 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0); 2272 errout: 2273 return err; 2274 } 2275 2276 static int rtnl_group_dellink(const struct net *net, int group) 2277 { 2278 struct net_device *dev, *aux; 2279 LIST_HEAD(list_kill); 2280 bool found = false; 2281 2282 if (!group) 2283 return -EPERM; 2284 2285 for_each_netdev(net, dev) { 2286 if (dev->group == group) { 2287 const struct rtnl_link_ops *ops; 2288 2289 found = true; 2290 ops = dev->rtnl_link_ops; 2291 if (!ops || !ops->dellink) 2292 return -EOPNOTSUPP; 2293 } 2294 } 2295 2296 if (!found) 2297 return -ENODEV; 2298 2299 for_each_netdev_safe(net, dev, aux) { 2300 if (dev->group == group) { 2301 const struct rtnl_link_ops *ops; 2302 2303 ops = dev->rtnl_link_ops; 2304 ops->dellink(dev, &list_kill); 2305 } 2306 } 2307 unregister_netdevice_many(&list_kill); 2308 2309 return 0; 2310 } 2311 2312 int rtnl_delete_link(struct net_device *dev) 2313 { 2314 const struct rtnl_link_ops *ops; 2315 LIST_HEAD(list_kill); 2316 2317 ops = dev->rtnl_link_ops; 2318 if (!ops || !ops->dellink) 2319 return -EOPNOTSUPP; 2320 2321 ops->dellink(dev, &list_kill); 2322 unregister_netdevice_many(&list_kill); 2323 2324 return 0; 2325 } 2326 EXPORT_SYMBOL_GPL(rtnl_delete_link); 2327 2328 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 2329 struct netlink_ext_ack *extack) 2330 { 2331 struct net *net = sock_net(skb->sk); 2332 struct net_device *dev; 2333 struct ifinfomsg *ifm; 2334 char ifname[IFNAMSIZ]; 2335 struct nlattr *tb[IFLA_MAX+1]; 2336 int err; 2337 2338 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2339 if (err < 0) 2340 return err; 2341 2342 if (tb[IFLA_IFNAME]) 2343 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2344 2345 ifm = nlmsg_data(nlh); 2346 if (ifm->ifi_index > 0) 2347 dev = __dev_get_by_index(net, ifm->ifi_index); 2348 else if (tb[IFLA_IFNAME]) 2349 dev = __dev_get_by_name(net, ifname); 2350 else if (tb[IFLA_GROUP]) 2351 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP])); 2352 else 2353 return -EINVAL; 2354 2355 if (!dev) 2356 return -ENODEV; 2357 2358 return rtnl_delete_link(dev); 2359 } 2360 2361 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 2362 { 2363 unsigned int old_flags; 2364 int err; 2365 2366 old_flags = dev->flags; 2367 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 2368 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 2369 if (err < 0) 2370 return err; 2371 } 2372 2373 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 2374 2375 __dev_notify_flags(dev, old_flags, ~0U); 2376 return 0; 2377 } 2378 EXPORT_SYMBOL(rtnl_configure_link); 2379 2380 struct net_device *rtnl_create_link(struct net *net, 2381 const char *ifname, unsigned char name_assign_type, 2382 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 2383 { 2384 struct net_device *dev; 2385 unsigned int num_tx_queues = 1; 2386 unsigned int num_rx_queues = 1; 2387 2388 if (tb[IFLA_NUM_TX_QUEUES]) 2389 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 2390 else if (ops->get_num_tx_queues) 2391 num_tx_queues = ops->get_num_tx_queues(); 2392 2393 if (tb[IFLA_NUM_RX_QUEUES]) 2394 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 2395 else if (ops->get_num_rx_queues) 2396 num_rx_queues = ops->get_num_rx_queues(); 2397 2398 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 2399 ops->setup, num_tx_queues, num_rx_queues); 2400 if (!dev) 2401 return ERR_PTR(-ENOMEM); 2402 2403 dev_net_set(dev, net); 2404 dev->rtnl_link_ops = ops; 2405 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 2406 2407 if (tb[IFLA_MTU]) 2408 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 2409 if (tb[IFLA_ADDRESS]) { 2410 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 2411 nla_len(tb[IFLA_ADDRESS])); 2412 dev->addr_assign_type = NET_ADDR_SET; 2413 } 2414 if (tb[IFLA_BROADCAST]) 2415 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 2416 nla_len(tb[IFLA_BROADCAST])); 2417 if (tb[IFLA_TXQLEN]) 2418 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 2419 if (tb[IFLA_OPERSTATE]) 2420 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 2421 if (tb[IFLA_LINKMODE]) 2422 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 2423 if (tb[IFLA_GROUP]) 2424 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 2425 2426 return dev; 2427 } 2428 EXPORT_SYMBOL(rtnl_create_link); 2429 2430 static int rtnl_group_changelink(const struct sk_buff *skb, 2431 struct net *net, int group, 2432 struct ifinfomsg *ifm, 2433 struct netlink_ext_ack *extack, 2434 struct nlattr **tb) 2435 { 2436 struct net_device *dev, *aux; 2437 int err; 2438 2439 for_each_netdev_safe(net, dev, aux) { 2440 if (dev->group == group) { 2441 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0); 2442 if (err < 0) 2443 return err; 2444 } 2445 } 2446 2447 return 0; 2448 } 2449 2450 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2451 struct netlink_ext_ack *extack) 2452 { 2453 struct net *net = sock_net(skb->sk); 2454 const struct rtnl_link_ops *ops; 2455 const struct rtnl_link_ops *m_ops = NULL; 2456 struct net_device *dev; 2457 struct net_device *master_dev = NULL; 2458 struct ifinfomsg *ifm; 2459 char kind[MODULE_NAME_LEN]; 2460 char ifname[IFNAMSIZ]; 2461 struct nlattr *tb[IFLA_MAX+1]; 2462 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 2463 unsigned char name_assign_type = NET_NAME_USER; 2464 int err; 2465 2466 #ifdef CONFIG_MODULES 2467 replay: 2468 #endif 2469 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2470 if (err < 0) 2471 return err; 2472 2473 if (tb[IFLA_IFNAME]) 2474 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2475 else 2476 ifname[0] = '\0'; 2477 2478 ifm = nlmsg_data(nlh); 2479 if (ifm->ifi_index > 0) 2480 dev = __dev_get_by_index(net, ifm->ifi_index); 2481 else { 2482 if (ifname[0]) 2483 dev = __dev_get_by_name(net, ifname); 2484 else 2485 dev = NULL; 2486 } 2487 2488 if (dev) { 2489 master_dev = netdev_master_upper_dev_get(dev); 2490 if (master_dev) 2491 m_ops = master_dev->rtnl_link_ops; 2492 } 2493 2494 err = validate_linkmsg(dev, tb); 2495 if (err < 0) 2496 return err; 2497 2498 if (tb[IFLA_LINKINFO]) { 2499 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 2500 tb[IFLA_LINKINFO], ifla_info_policy, 2501 NULL); 2502 if (err < 0) 2503 return err; 2504 } else 2505 memset(linkinfo, 0, sizeof(linkinfo)); 2506 2507 if (linkinfo[IFLA_INFO_KIND]) { 2508 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 2509 ops = rtnl_link_ops_get(kind); 2510 } else { 2511 kind[0] = '\0'; 2512 ops = NULL; 2513 } 2514 2515 if (1) { 2516 struct nlattr *attr[ops ? ops->maxtype + 1 : 1]; 2517 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1]; 2518 struct nlattr **data = NULL; 2519 struct nlattr **slave_data = NULL; 2520 struct net *dest_net, *link_net = NULL; 2521 2522 if (ops) { 2523 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 2524 err = nla_parse_nested(attr, ops->maxtype, 2525 linkinfo[IFLA_INFO_DATA], 2526 ops->policy, NULL); 2527 if (err < 0) 2528 return err; 2529 data = attr; 2530 } 2531 if (ops->validate) { 2532 err = ops->validate(tb, data); 2533 if (err < 0) 2534 return err; 2535 } 2536 } 2537 2538 if (m_ops) { 2539 if (m_ops->slave_maxtype && 2540 linkinfo[IFLA_INFO_SLAVE_DATA]) { 2541 err = nla_parse_nested(slave_attr, 2542 m_ops->slave_maxtype, 2543 linkinfo[IFLA_INFO_SLAVE_DATA], 2544 m_ops->slave_policy, 2545 NULL); 2546 if (err < 0) 2547 return err; 2548 slave_data = slave_attr; 2549 } 2550 if (m_ops->slave_validate) { 2551 err = m_ops->slave_validate(tb, slave_data); 2552 if (err < 0) 2553 return err; 2554 } 2555 } 2556 2557 if (dev) { 2558 int status = 0; 2559 2560 if (nlh->nlmsg_flags & NLM_F_EXCL) 2561 return -EEXIST; 2562 if (nlh->nlmsg_flags & NLM_F_REPLACE) 2563 return -EOPNOTSUPP; 2564 2565 if (linkinfo[IFLA_INFO_DATA]) { 2566 if (!ops || ops != dev->rtnl_link_ops || 2567 !ops->changelink) 2568 return -EOPNOTSUPP; 2569 2570 err = ops->changelink(dev, tb, data); 2571 if (err < 0) 2572 return err; 2573 status |= DO_SETLINK_NOTIFY; 2574 } 2575 2576 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 2577 if (!m_ops || !m_ops->slave_changelink) 2578 return -EOPNOTSUPP; 2579 2580 err = m_ops->slave_changelink(master_dev, dev, 2581 tb, slave_data); 2582 if (err < 0) 2583 return err; 2584 status |= DO_SETLINK_NOTIFY; 2585 } 2586 2587 return do_setlink(skb, dev, ifm, extack, tb, ifname, 2588 status); 2589 } 2590 2591 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 2592 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 2593 return rtnl_group_changelink(skb, net, 2594 nla_get_u32(tb[IFLA_GROUP]), 2595 ifm, extack, tb); 2596 return -ENODEV; 2597 } 2598 2599 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO]) 2600 return -EOPNOTSUPP; 2601 2602 if (!ops) { 2603 #ifdef CONFIG_MODULES 2604 if (kind[0]) { 2605 __rtnl_unlock(); 2606 request_module("rtnl-link-%s", kind); 2607 rtnl_lock(); 2608 ops = rtnl_link_ops_get(kind); 2609 if (ops) 2610 goto replay; 2611 } 2612 #endif 2613 return -EOPNOTSUPP; 2614 } 2615 2616 if (!ops->setup) 2617 return -EOPNOTSUPP; 2618 2619 if (!ifname[0]) { 2620 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 2621 name_assign_type = NET_NAME_ENUM; 2622 } 2623 2624 dest_net = rtnl_link_get_net(net, tb); 2625 if (IS_ERR(dest_net)) 2626 return PTR_ERR(dest_net); 2627 2628 err = -EPERM; 2629 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN)) 2630 goto out; 2631 2632 if (tb[IFLA_LINK_NETNSID]) { 2633 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]); 2634 2635 link_net = get_net_ns_by_id(dest_net, id); 2636 if (!link_net) { 2637 err = -EINVAL; 2638 goto out; 2639 } 2640 err = -EPERM; 2641 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN)) 2642 goto out; 2643 } 2644 2645 dev = rtnl_create_link(link_net ? : dest_net, ifname, 2646 name_assign_type, ops, tb); 2647 if (IS_ERR(dev)) { 2648 err = PTR_ERR(dev); 2649 goto out; 2650 } 2651 2652 dev->ifindex = ifm->ifi_index; 2653 2654 if (ops->newlink) { 2655 err = ops->newlink(link_net ? : net, dev, tb, data); 2656 /* Drivers should call free_netdev() in ->destructor 2657 * and unregister it on failure after registration 2658 * so that device could be finally freed in rtnl_unlock. 2659 */ 2660 if (err < 0) { 2661 /* If device is not registered at all, free it now */ 2662 if (dev->reg_state == NETREG_UNINITIALIZED) 2663 free_netdev(dev); 2664 goto out; 2665 } 2666 } else { 2667 err = register_netdevice(dev); 2668 if (err < 0) { 2669 free_netdev(dev); 2670 goto out; 2671 } 2672 } 2673 err = rtnl_configure_link(dev, ifm); 2674 if (err < 0) 2675 goto out_unregister; 2676 if (link_net) { 2677 err = dev_change_net_namespace(dev, dest_net, ifname); 2678 if (err < 0) 2679 goto out_unregister; 2680 } 2681 if (tb[IFLA_MASTER]) { 2682 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 2683 if (err) 2684 goto out_unregister; 2685 } 2686 out: 2687 if (link_net) 2688 put_net(link_net); 2689 put_net(dest_net); 2690 return err; 2691 out_unregister: 2692 if (ops->newlink) { 2693 LIST_HEAD(list_kill); 2694 2695 ops->dellink(dev, &list_kill); 2696 unregister_netdevice_many(&list_kill); 2697 } else { 2698 unregister_netdevice(dev); 2699 } 2700 goto out; 2701 } 2702 } 2703 2704 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh, 2705 struct netlink_ext_ack *extack) 2706 { 2707 struct net *net = sock_net(skb->sk); 2708 struct ifinfomsg *ifm; 2709 char ifname[IFNAMSIZ]; 2710 struct nlattr *tb[IFLA_MAX+1]; 2711 struct net_device *dev = NULL; 2712 struct sk_buff *nskb; 2713 int err; 2714 u32 ext_filter_mask = 0; 2715 2716 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack); 2717 if (err < 0) 2718 return err; 2719 2720 if (tb[IFLA_IFNAME]) 2721 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2722 2723 if (tb[IFLA_EXT_MASK]) 2724 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2725 2726 ifm = nlmsg_data(nlh); 2727 if (ifm->ifi_index > 0) 2728 dev = __dev_get_by_index(net, ifm->ifi_index); 2729 else if (tb[IFLA_IFNAME]) 2730 dev = __dev_get_by_name(net, ifname); 2731 else 2732 return -EINVAL; 2733 2734 if (dev == NULL) 2735 return -ENODEV; 2736 2737 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 2738 if (nskb == NULL) 2739 return -ENOBUFS; 2740 2741 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid, 2742 nlh->nlmsg_seq, 0, 0, ext_filter_mask); 2743 if (err < 0) { 2744 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 2745 WARN_ON(err == -EMSGSIZE); 2746 kfree_skb(nskb); 2747 } else 2748 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 2749 2750 return err; 2751 } 2752 2753 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 2754 { 2755 struct net *net = sock_net(skb->sk); 2756 struct net_device *dev; 2757 struct nlattr *tb[IFLA_MAX+1]; 2758 u32 ext_filter_mask = 0; 2759 u16 min_ifinfo_dump_size = 0; 2760 int hdrlen; 2761 2762 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 2763 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 2764 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 2765 2766 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) { 2767 if (tb[IFLA_EXT_MASK]) 2768 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2769 } 2770 2771 if (!ext_filter_mask) 2772 return NLMSG_GOODSIZE; 2773 /* 2774 * traverse the list of net devices and compute the minimum 2775 * buffer size based upon the filter mask. 2776 */ 2777 list_for_each_entry(dev, &net->dev_base_head, dev_list) { 2778 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 2779 if_nlmsg_size(dev, 2780 ext_filter_mask)); 2781 } 2782 2783 return nlmsg_total_size(min_ifinfo_dump_size); 2784 } 2785 2786 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 2787 { 2788 int idx; 2789 int s_idx = cb->family; 2790 2791 if (s_idx == 0) 2792 s_idx = 1; 2793 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 2794 int type = cb->nlh->nlmsg_type-RTM_BASE; 2795 if (idx < s_idx || idx == PF_PACKET) 2796 continue; 2797 if (rtnl_msg_handlers[idx] == NULL || 2798 rtnl_msg_handlers[idx][type].dumpit == NULL) 2799 continue; 2800 if (idx > s_idx) { 2801 memset(&cb->args[0], 0, sizeof(cb->args)); 2802 cb->prev_seq = 0; 2803 cb->seq = 0; 2804 } 2805 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 2806 break; 2807 } 2808 cb->family = idx; 2809 2810 return skb->len; 2811 } 2812 2813 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev, 2814 unsigned int change, gfp_t flags) 2815 { 2816 struct net *net = dev_net(dev); 2817 struct sk_buff *skb; 2818 int err = -ENOBUFS; 2819 size_t if_info_size; 2820 2821 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 2822 if (skb == NULL) 2823 goto errout; 2824 2825 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0); 2826 if (err < 0) { 2827 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 2828 WARN_ON(err == -EMSGSIZE); 2829 kfree_skb(skb); 2830 goto errout; 2831 } 2832 return skb; 2833 errout: 2834 if (err < 0) 2835 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2836 return NULL; 2837 } 2838 2839 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags) 2840 { 2841 struct net *net = dev_net(dev); 2842 2843 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 2844 } 2845 2846 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 2847 gfp_t flags) 2848 { 2849 struct sk_buff *skb; 2850 2851 if (dev->reg_state != NETREG_REGISTERED) 2852 return; 2853 2854 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags); 2855 if (skb) 2856 rtmsg_ifinfo_send(skb, dev, flags); 2857 } 2858 EXPORT_SYMBOL(rtmsg_ifinfo); 2859 2860 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 2861 struct net_device *dev, 2862 u8 *addr, u16 vid, u32 pid, u32 seq, 2863 int type, unsigned int flags, 2864 int nlflags, u16 ndm_state) 2865 { 2866 struct nlmsghdr *nlh; 2867 struct ndmsg *ndm; 2868 2869 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 2870 if (!nlh) 2871 return -EMSGSIZE; 2872 2873 ndm = nlmsg_data(nlh); 2874 ndm->ndm_family = AF_BRIDGE; 2875 ndm->ndm_pad1 = 0; 2876 ndm->ndm_pad2 = 0; 2877 ndm->ndm_flags = flags; 2878 ndm->ndm_type = 0; 2879 ndm->ndm_ifindex = dev->ifindex; 2880 ndm->ndm_state = ndm_state; 2881 2882 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 2883 goto nla_put_failure; 2884 if (vid) 2885 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid)) 2886 goto nla_put_failure; 2887 2888 nlmsg_end(skb, nlh); 2889 return 0; 2890 2891 nla_put_failure: 2892 nlmsg_cancel(skb, nlh); 2893 return -EMSGSIZE; 2894 } 2895 2896 static inline size_t rtnl_fdb_nlmsg_size(void) 2897 { 2898 return NLMSG_ALIGN(sizeof(struct ndmsg)) + 2899 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */ 2900 nla_total_size(sizeof(u16)) + /* NDA_VLAN */ 2901 0; 2902 } 2903 2904 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type, 2905 u16 ndm_state) 2906 { 2907 struct net *net = dev_net(dev); 2908 struct sk_buff *skb; 2909 int err = -ENOBUFS; 2910 2911 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 2912 if (!skb) 2913 goto errout; 2914 2915 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid, 2916 0, 0, type, NTF_SELF, 0, ndm_state); 2917 if (err < 0) { 2918 kfree_skb(skb); 2919 goto errout; 2920 } 2921 2922 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 2923 return; 2924 errout: 2925 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 2926 } 2927 2928 /** 2929 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 2930 */ 2931 int ndo_dflt_fdb_add(struct ndmsg *ndm, 2932 struct nlattr *tb[], 2933 struct net_device *dev, 2934 const unsigned char *addr, u16 vid, 2935 u16 flags) 2936 { 2937 int err = -EINVAL; 2938 2939 /* If aging addresses are supported device will need to 2940 * implement its own handler for this. 2941 */ 2942 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 2943 pr_info("%s: FDB only supports static addresses\n", dev->name); 2944 return err; 2945 } 2946 2947 if (vid) { 2948 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name); 2949 return err; 2950 } 2951 2952 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 2953 err = dev_uc_add_excl(dev, addr); 2954 else if (is_multicast_ether_addr(addr)) 2955 err = dev_mc_add_excl(dev, addr); 2956 2957 /* Only return duplicate errors if NLM_F_EXCL is set */ 2958 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 2959 err = 0; 2960 2961 return err; 2962 } 2963 EXPORT_SYMBOL(ndo_dflt_fdb_add); 2964 2965 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid) 2966 { 2967 u16 vid = 0; 2968 2969 if (vlan_attr) { 2970 if (nla_len(vlan_attr) != sizeof(u16)) { 2971 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n"); 2972 return -EINVAL; 2973 } 2974 2975 vid = nla_get_u16(vlan_attr); 2976 2977 if (!vid || vid >= VLAN_VID_MASK) { 2978 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n", 2979 vid); 2980 return -EINVAL; 2981 } 2982 } 2983 *p_vid = vid; 2984 return 0; 2985 } 2986 2987 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh, 2988 struct netlink_ext_ack *extack) 2989 { 2990 struct net *net = sock_net(skb->sk); 2991 struct ndmsg *ndm; 2992 struct nlattr *tb[NDA_MAX+1]; 2993 struct net_device *dev; 2994 u8 *addr; 2995 u16 vid; 2996 int err; 2997 2998 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 2999 if (err < 0) 3000 return err; 3001 3002 ndm = nlmsg_data(nlh); 3003 if (ndm->ndm_ifindex == 0) { 3004 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n"); 3005 return -EINVAL; 3006 } 3007 3008 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3009 if (dev == NULL) { 3010 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n"); 3011 return -ENODEV; 3012 } 3013 3014 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3015 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n"); 3016 return -EINVAL; 3017 } 3018 3019 addr = nla_data(tb[NDA_LLADDR]); 3020 3021 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3022 if (err) 3023 return err; 3024 3025 err = -EOPNOTSUPP; 3026 3027 /* Support fdb on master device the net/bridge default case */ 3028 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3029 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3030 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3031 const struct net_device_ops *ops = br_dev->netdev_ops; 3032 3033 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 3034 nlh->nlmsg_flags); 3035 if (err) 3036 goto out; 3037 else 3038 ndm->ndm_flags &= ~NTF_MASTER; 3039 } 3040 3041 /* Embedded bridge, macvlan, and any other device support */ 3042 if ((ndm->ndm_flags & NTF_SELF)) { 3043 if (dev->netdev_ops->ndo_fdb_add) 3044 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 3045 vid, 3046 nlh->nlmsg_flags); 3047 else 3048 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 3049 nlh->nlmsg_flags); 3050 3051 if (!err) { 3052 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH, 3053 ndm->ndm_state); 3054 ndm->ndm_flags &= ~NTF_SELF; 3055 } 3056 } 3057 out: 3058 return err; 3059 } 3060 3061 /** 3062 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 3063 */ 3064 int ndo_dflt_fdb_del(struct ndmsg *ndm, 3065 struct nlattr *tb[], 3066 struct net_device *dev, 3067 const unsigned char *addr, u16 vid) 3068 { 3069 int err = -EINVAL; 3070 3071 /* If aging addresses are supported device will need to 3072 * implement its own handler for this. 3073 */ 3074 if (!(ndm->ndm_state & NUD_PERMANENT)) { 3075 pr_info("%s: FDB only supports static addresses\n", dev->name); 3076 return err; 3077 } 3078 3079 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 3080 err = dev_uc_del(dev, addr); 3081 else if (is_multicast_ether_addr(addr)) 3082 err = dev_mc_del(dev, addr); 3083 3084 return err; 3085 } 3086 EXPORT_SYMBOL(ndo_dflt_fdb_del); 3087 3088 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh, 3089 struct netlink_ext_ack *extack) 3090 { 3091 struct net *net = sock_net(skb->sk); 3092 struct ndmsg *ndm; 3093 struct nlattr *tb[NDA_MAX+1]; 3094 struct net_device *dev; 3095 int err = -EINVAL; 3096 __u8 *addr; 3097 u16 vid; 3098 3099 if (!netlink_capable(skb, CAP_NET_ADMIN)) 3100 return -EPERM; 3101 3102 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack); 3103 if (err < 0) 3104 return err; 3105 3106 ndm = nlmsg_data(nlh); 3107 if (ndm->ndm_ifindex == 0) { 3108 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n"); 3109 return -EINVAL; 3110 } 3111 3112 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 3113 if (dev == NULL) { 3114 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n"); 3115 return -ENODEV; 3116 } 3117 3118 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 3119 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n"); 3120 return -EINVAL; 3121 } 3122 3123 addr = nla_data(tb[NDA_LLADDR]); 3124 3125 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 3126 if (err) 3127 return err; 3128 3129 err = -EOPNOTSUPP; 3130 3131 /* Support fdb on master device the net/bridge default case */ 3132 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 3133 (dev->priv_flags & IFF_BRIDGE_PORT)) { 3134 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3135 const struct net_device_ops *ops = br_dev->netdev_ops; 3136 3137 if (ops->ndo_fdb_del) 3138 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 3139 3140 if (err) 3141 goto out; 3142 else 3143 ndm->ndm_flags &= ~NTF_MASTER; 3144 } 3145 3146 /* Embedded bridge, macvlan, and any other device support */ 3147 if (ndm->ndm_flags & NTF_SELF) { 3148 if (dev->netdev_ops->ndo_fdb_del) 3149 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 3150 vid); 3151 else 3152 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 3153 3154 if (!err) { 3155 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH, 3156 ndm->ndm_state); 3157 ndm->ndm_flags &= ~NTF_SELF; 3158 } 3159 } 3160 out: 3161 return err; 3162 } 3163 3164 static int nlmsg_populate_fdb(struct sk_buff *skb, 3165 struct netlink_callback *cb, 3166 struct net_device *dev, 3167 int *idx, 3168 struct netdev_hw_addr_list *list) 3169 { 3170 struct netdev_hw_addr *ha; 3171 int err; 3172 u32 portid, seq; 3173 3174 portid = NETLINK_CB(cb->skb).portid; 3175 seq = cb->nlh->nlmsg_seq; 3176 3177 list_for_each_entry(ha, &list->list, list) { 3178 if (*idx < cb->args[2]) 3179 goto skip; 3180 3181 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0, 3182 portid, seq, 3183 RTM_NEWNEIGH, NTF_SELF, 3184 NLM_F_MULTI, NUD_PERMANENT); 3185 if (err < 0) 3186 return err; 3187 skip: 3188 *idx += 1; 3189 } 3190 return 0; 3191 } 3192 3193 /** 3194 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 3195 * @nlh: netlink message header 3196 * @dev: netdevice 3197 * 3198 * Default netdevice operation to dump the existing unicast address list. 3199 * Returns number of addresses from list put in skb. 3200 */ 3201 int ndo_dflt_fdb_dump(struct sk_buff *skb, 3202 struct netlink_callback *cb, 3203 struct net_device *dev, 3204 struct net_device *filter_dev, 3205 int *idx) 3206 { 3207 int err; 3208 3209 netif_addr_lock_bh(dev); 3210 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc); 3211 if (err) 3212 goto out; 3213 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc); 3214 out: 3215 netif_addr_unlock_bh(dev); 3216 return err; 3217 } 3218 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 3219 3220 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 3221 { 3222 struct net_device *dev; 3223 struct nlattr *tb[IFLA_MAX+1]; 3224 struct net_device *br_dev = NULL; 3225 const struct net_device_ops *ops = NULL; 3226 const struct net_device_ops *cops = NULL; 3227 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 3228 struct net *net = sock_net(skb->sk); 3229 struct hlist_head *head; 3230 int brport_idx = 0; 3231 int br_idx = 0; 3232 int h, s_h; 3233 int idx = 0, s_idx; 3234 int err = 0; 3235 int fidx = 0; 3236 3237 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, 3238 IFLA_MAX, ifla_policy, NULL); 3239 if (err < 0) { 3240 return -EINVAL; 3241 } else if (err == 0) { 3242 if (tb[IFLA_MASTER]) 3243 br_idx = nla_get_u32(tb[IFLA_MASTER]); 3244 } 3245 3246 brport_idx = ifm->ifi_index; 3247 3248 if (br_idx) { 3249 br_dev = __dev_get_by_index(net, br_idx); 3250 if (!br_dev) 3251 return -ENODEV; 3252 3253 ops = br_dev->netdev_ops; 3254 } 3255 3256 s_h = cb->args[0]; 3257 s_idx = cb->args[1]; 3258 3259 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3260 idx = 0; 3261 head = &net->dev_index_head[h]; 3262 hlist_for_each_entry(dev, head, index_hlist) { 3263 3264 if (brport_idx && (dev->ifindex != brport_idx)) 3265 continue; 3266 3267 if (!br_idx) { /* user did not specify a specific bridge */ 3268 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3269 br_dev = netdev_master_upper_dev_get(dev); 3270 cops = br_dev->netdev_ops; 3271 } 3272 } else { 3273 if (dev != br_dev && 3274 !(dev->priv_flags & IFF_BRIDGE_PORT)) 3275 continue; 3276 3277 if (br_dev != netdev_master_upper_dev_get(dev) && 3278 !(dev->priv_flags & IFF_EBRIDGE)) 3279 continue; 3280 cops = ops; 3281 } 3282 3283 if (idx < s_idx) 3284 goto cont; 3285 3286 if (dev->priv_flags & IFF_BRIDGE_PORT) { 3287 if (cops && cops->ndo_fdb_dump) { 3288 err = cops->ndo_fdb_dump(skb, cb, 3289 br_dev, dev, 3290 &fidx); 3291 if (err == -EMSGSIZE) 3292 goto out; 3293 } 3294 } 3295 3296 if (dev->netdev_ops->ndo_fdb_dump) 3297 err = dev->netdev_ops->ndo_fdb_dump(skb, cb, 3298 dev, NULL, 3299 &fidx); 3300 else 3301 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL, 3302 &fidx); 3303 if (err == -EMSGSIZE) 3304 goto out; 3305 3306 cops = NULL; 3307 3308 /* reset fdb offset to 0 for rest of the interfaces */ 3309 cb->args[2] = 0; 3310 fidx = 0; 3311 cont: 3312 idx++; 3313 } 3314 } 3315 3316 out: 3317 cb->args[0] = h; 3318 cb->args[1] = idx; 3319 cb->args[2] = fidx; 3320 3321 return skb->len; 3322 } 3323 3324 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask, 3325 unsigned int attrnum, unsigned int flag) 3326 { 3327 if (mask & flag) 3328 return nla_put_u8(skb, attrnum, !!(flags & flag)); 3329 return 0; 3330 } 3331 3332 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 3333 struct net_device *dev, u16 mode, 3334 u32 flags, u32 mask, int nlflags, 3335 u32 filter_mask, 3336 int (*vlan_fill)(struct sk_buff *skb, 3337 struct net_device *dev, 3338 u32 filter_mask)) 3339 { 3340 struct nlmsghdr *nlh; 3341 struct ifinfomsg *ifm; 3342 struct nlattr *br_afspec; 3343 struct nlattr *protinfo; 3344 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 3345 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3346 int err = 0; 3347 3348 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags); 3349 if (nlh == NULL) 3350 return -EMSGSIZE; 3351 3352 ifm = nlmsg_data(nlh); 3353 ifm->ifi_family = AF_BRIDGE; 3354 ifm->__ifi_pad = 0; 3355 ifm->ifi_type = dev->type; 3356 ifm->ifi_index = dev->ifindex; 3357 ifm->ifi_flags = dev_get_flags(dev); 3358 ifm->ifi_change = 0; 3359 3360 3361 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 3362 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 3363 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 3364 (br_dev && 3365 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 3366 (dev->addr_len && 3367 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 3368 (dev->ifindex != dev_get_iflink(dev) && 3369 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev)))) 3370 goto nla_put_failure; 3371 3372 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 3373 if (!br_afspec) 3374 goto nla_put_failure; 3375 3376 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) { 3377 nla_nest_cancel(skb, br_afspec); 3378 goto nla_put_failure; 3379 } 3380 3381 if (mode != BRIDGE_MODE_UNDEF) { 3382 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 3383 nla_nest_cancel(skb, br_afspec); 3384 goto nla_put_failure; 3385 } 3386 } 3387 if (vlan_fill) { 3388 err = vlan_fill(skb, dev, filter_mask); 3389 if (err) { 3390 nla_nest_cancel(skb, br_afspec); 3391 goto nla_put_failure; 3392 } 3393 } 3394 nla_nest_end(skb, br_afspec); 3395 3396 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED); 3397 if (!protinfo) 3398 goto nla_put_failure; 3399 3400 if (brport_nla_put_flag(skb, flags, mask, 3401 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) || 3402 brport_nla_put_flag(skb, flags, mask, 3403 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) || 3404 brport_nla_put_flag(skb, flags, mask, 3405 IFLA_BRPORT_FAST_LEAVE, 3406 BR_MULTICAST_FAST_LEAVE) || 3407 brport_nla_put_flag(skb, flags, mask, 3408 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) || 3409 brport_nla_put_flag(skb, flags, mask, 3410 IFLA_BRPORT_LEARNING, BR_LEARNING) || 3411 brport_nla_put_flag(skb, flags, mask, 3412 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) || 3413 brport_nla_put_flag(skb, flags, mask, 3414 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) || 3415 brport_nla_put_flag(skb, flags, mask, 3416 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) { 3417 nla_nest_cancel(skb, protinfo); 3418 goto nla_put_failure; 3419 } 3420 3421 nla_nest_end(skb, protinfo); 3422 3423 nlmsg_end(skb, nlh); 3424 return 0; 3425 nla_put_failure: 3426 nlmsg_cancel(skb, nlh); 3427 return err ? err : -EMSGSIZE; 3428 } 3429 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink); 3430 3431 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 3432 { 3433 struct net *net = sock_net(skb->sk); 3434 struct net_device *dev; 3435 int idx = 0; 3436 u32 portid = NETLINK_CB(cb->skb).portid; 3437 u32 seq = cb->nlh->nlmsg_seq; 3438 u32 filter_mask = 0; 3439 int err; 3440 3441 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 3442 struct nlattr *extfilt; 3443 3444 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 3445 IFLA_EXT_MASK); 3446 if (extfilt) { 3447 if (nla_len(extfilt) < sizeof(filter_mask)) 3448 return -EINVAL; 3449 3450 filter_mask = nla_get_u32(extfilt); 3451 } 3452 } 3453 3454 rcu_read_lock(); 3455 for_each_netdev_rcu(net, dev) { 3456 const struct net_device_ops *ops = dev->netdev_ops; 3457 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3458 3459 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 3460 if (idx >= cb->args[0]) { 3461 err = br_dev->netdev_ops->ndo_bridge_getlink( 3462 skb, portid, seq, dev, 3463 filter_mask, NLM_F_MULTI); 3464 if (err < 0 && err != -EOPNOTSUPP) { 3465 if (likely(skb->len)) 3466 break; 3467 3468 goto out_err; 3469 } 3470 } 3471 idx++; 3472 } 3473 3474 if (ops->ndo_bridge_getlink) { 3475 if (idx >= cb->args[0]) { 3476 err = ops->ndo_bridge_getlink(skb, portid, 3477 seq, dev, 3478 filter_mask, 3479 NLM_F_MULTI); 3480 if (err < 0 && err != -EOPNOTSUPP) { 3481 if (likely(skb->len)) 3482 break; 3483 3484 goto out_err; 3485 } 3486 } 3487 idx++; 3488 } 3489 } 3490 err = skb->len; 3491 out_err: 3492 rcu_read_unlock(); 3493 cb->args[0] = idx; 3494 3495 return err; 3496 } 3497 3498 static inline size_t bridge_nlmsg_size(void) 3499 { 3500 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 3501 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 3502 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 3503 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 3504 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 3505 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 3506 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 3507 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 3508 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 3509 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 3510 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 3511 } 3512 3513 static int rtnl_bridge_notify(struct net_device *dev) 3514 { 3515 struct net *net = dev_net(dev); 3516 struct sk_buff *skb; 3517 int err = -EOPNOTSUPP; 3518 3519 if (!dev->netdev_ops->ndo_bridge_getlink) 3520 return 0; 3521 3522 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 3523 if (!skb) { 3524 err = -ENOMEM; 3525 goto errout; 3526 } 3527 3528 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0); 3529 if (err < 0) 3530 goto errout; 3531 3532 if (!skb->len) 3533 goto errout; 3534 3535 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 3536 return 0; 3537 errout: 3538 WARN_ON(err == -EMSGSIZE); 3539 kfree_skb(skb); 3540 if (err) 3541 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 3542 return err; 3543 } 3544 3545 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, 3546 struct netlink_ext_ack *extack) 3547 { 3548 struct net *net = sock_net(skb->sk); 3549 struct ifinfomsg *ifm; 3550 struct net_device *dev; 3551 struct nlattr *br_spec, *attr = NULL; 3552 int rem, err = -EOPNOTSUPP; 3553 u16 flags = 0; 3554 bool have_flags = false; 3555 3556 if (nlmsg_len(nlh) < sizeof(*ifm)) 3557 return -EINVAL; 3558 3559 ifm = nlmsg_data(nlh); 3560 if (ifm->ifi_family != AF_BRIDGE) 3561 return -EPFNOSUPPORT; 3562 3563 dev = __dev_get_by_index(net, ifm->ifi_index); 3564 if (!dev) { 3565 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3566 return -ENODEV; 3567 } 3568 3569 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3570 if (br_spec) { 3571 nla_for_each_nested(attr, br_spec, rem) { 3572 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3573 if (nla_len(attr) < sizeof(flags)) 3574 return -EINVAL; 3575 3576 have_flags = true; 3577 flags = nla_get_u16(attr); 3578 break; 3579 } 3580 } 3581 } 3582 3583 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3584 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3585 3586 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 3587 err = -EOPNOTSUPP; 3588 goto out; 3589 } 3590 3591 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags); 3592 if (err) 3593 goto out; 3594 3595 flags &= ~BRIDGE_FLAGS_MASTER; 3596 } 3597 3598 if ((flags & BRIDGE_FLAGS_SELF)) { 3599 if (!dev->netdev_ops->ndo_bridge_setlink) 3600 err = -EOPNOTSUPP; 3601 else 3602 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh, 3603 flags); 3604 if (!err) { 3605 flags &= ~BRIDGE_FLAGS_SELF; 3606 3607 /* Generate event to notify upper layer of bridge 3608 * change 3609 */ 3610 err = rtnl_bridge_notify(dev); 3611 } 3612 } 3613 3614 if (have_flags) 3615 memcpy(nla_data(attr), &flags, sizeof(flags)); 3616 out: 3617 return err; 3618 } 3619 3620 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, 3621 struct netlink_ext_ack *extack) 3622 { 3623 struct net *net = sock_net(skb->sk); 3624 struct ifinfomsg *ifm; 3625 struct net_device *dev; 3626 struct nlattr *br_spec, *attr = NULL; 3627 int rem, err = -EOPNOTSUPP; 3628 u16 flags = 0; 3629 bool have_flags = false; 3630 3631 if (nlmsg_len(nlh) < sizeof(*ifm)) 3632 return -EINVAL; 3633 3634 ifm = nlmsg_data(nlh); 3635 if (ifm->ifi_family != AF_BRIDGE) 3636 return -EPFNOSUPPORT; 3637 3638 dev = __dev_get_by_index(net, ifm->ifi_index); 3639 if (!dev) { 3640 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 3641 return -ENODEV; 3642 } 3643 3644 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 3645 if (br_spec) { 3646 nla_for_each_nested(attr, br_spec, rem) { 3647 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 3648 if (nla_len(attr) < sizeof(flags)) 3649 return -EINVAL; 3650 3651 have_flags = true; 3652 flags = nla_get_u16(attr); 3653 break; 3654 } 3655 } 3656 } 3657 3658 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 3659 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 3660 3661 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 3662 err = -EOPNOTSUPP; 3663 goto out; 3664 } 3665 3666 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags); 3667 if (err) 3668 goto out; 3669 3670 flags &= ~BRIDGE_FLAGS_MASTER; 3671 } 3672 3673 if ((flags & BRIDGE_FLAGS_SELF)) { 3674 if (!dev->netdev_ops->ndo_bridge_dellink) 3675 err = -EOPNOTSUPP; 3676 else 3677 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh, 3678 flags); 3679 3680 if (!err) { 3681 flags &= ~BRIDGE_FLAGS_SELF; 3682 3683 /* Generate event to notify upper layer of bridge 3684 * change 3685 */ 3686 err = rtnl_bridge_notify(dev); 3687 } 3688 } 3689 3690 if (have_flags) 3691 memcpy(nla_data(attr), &flags, sizeof(flags)); 3692 out: 3693 return err; 3694 } 3695 3696 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr) 3697 { 3698 return (mask & IFLA_STATS_FILTER_BIT(attrid)) && 3699 (!idxattr || idxattr == attrid); 3700 } 3701 3702 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1) 3703 static int rtnl_get_offload_stats_attr_size(int attr_id) 3704 { 3705 switch (attr_id) { 3706 case IFLA_OFFLOAD_XSTATS_CPU_HIT: 3707 return sizeof(struct rtnl_link_stats64); 3708 } 3709 3710 return 0; 3711 } 3712 3713 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev, 3714 int *prividx) 3715 { 3716 struct nlattr *attr = NULL; 3717 int attr_id, size; 3718 void *attr_data; 3719 int err; 3720 3721 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3722 dev->netdev_ops->ndo_get_offload_stats)) 3723 return -ENODATA; 3724 3725 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3726 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3727 if (attr_id < *prividx) 3728 continue; 3729 3730 size = rtnl_get_offload_stats_attr_size(attr_id); 3731 if (!size) 3732 continue; 3733 3734 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3735 continue; 3736 3737 attr = nla_reserve_64bit(skb, attr_id, size, 3738 IFLA_OFFLOAD_XSTATS_UNSPEC); 3739 if (!attr) 3740 goto nla_put_failure; 3741 3742 attr_data = nla_data(attr); 3743 memset(attr_data, 0, size); 3744 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, 3745 attr_data); 3746 if (err) 3747 goto get_offload_stats_failure; 3748 } 3749 3750 if (!attr) 3751 return -ENODATA; 3752 3753 *prividx = 0; 3754 return 0; 3755 3756 nla_put_failure: 3757 err = -EMSGSIZE; 3758 get_offload_stats_failure: 3759 *prividx = attr_id; 3760 return err; 3761 } 3762 3763 static int rtnl_get_offload_stats_size(const struct net_device *dev) 3764 { 3765 int nla_size = 0; 3766 int attr_id; 3767 int size; 3768 3769 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats && 3770 dev->netdev_ops->ndo_get_offload_stats)) 3771 return 0; 3772 3773 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST; 3774 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) { 3775 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id)) 3776 continue; 3777 size = rtnl_get_offload_stats_attr_size(attr_id); 3778 nla_size += nla_total_size_64bit(size); 3779 } 3780 3781 if (nla_size != 0) 3782 nla_size += nla_total_size(0); 3783 3784 return nla_size; 3785 } 3786 3787 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev, 3788 int type, u32 pid, u32 seq, u32 change, 3789 unsigned int flags, unsigned int filter_mask, 3790 int *idxattr, int *prividx) 3791 { 3792 struct if_stats_msg *ifsm; 3793 struct nlmsghdr *nlh; 3794 struct nlattr *attr; 3795 int s_prividx = *prividx; 3796 int err; 3797 3798 ASSERT_RTNL(); 3799 3800 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags); 3801 if (!nlh) 3802 return -EMSGSIZE; 3803 3804 ifsm = nlmsg_data(nlh); 3805 ifsm->ifindex = dev->ifindex; 3806 ifsm->filter_mask = filter_mask; 3807 3808 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) { 3809 struct rtnl_link_stats64 *sp; 3810 3811 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64, 3812 sizeof(struct rtnl_link_stats64), 3813 IFLA_STATS_UNSPEC); 3814 if (!attr) 3815 goto nla_put_failure; 3816 3817 sp = nla_data(attr); 3818 dev_get_stats(dev, sp); 3819 } 3820 3821 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) { 3822 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3823 3824 if (ops && ops->fill_linkxstats) { 3825 *idxattr = IFLA_STATS_LINK_XSTATS; 3826 attr = nla_nest_start(skb, 3827 IFLA_STATS_LINK_XSTATS); 3828 if (!attr) 3829 goto nla_put_failure; 3830 3831 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3832 nla_nest_end(skb, attr); 3833 if (err) 3834 goto nla_put_failure; 3835 *idxattr = 0; 3836 } 3837 } 3838 3839 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 3840 *idxattr)) { 3841 const struct rtnl_link_ops *ops = NULL; 3842 const struct net_device *master; 3843 3844 master = netdev_master_upper_dev_get(dev); 3845 if (master) 3846 ops = master->rtnl_link_ops; 3847 if (ops && ops->fill_linkxstats) { 3848 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE; 3849 attr = nla_nest_start(skb, 3850 IFLA_STATS_LINK_XSTATS_SLAVE); 3851 if (!attr) 3852 goto nla_put_failure; 3853 3854 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr); 3855 nla_nest_end(skb, attr); 3856 if (err) 3857 goto nla_put_failure; 3858 *idxattr = 0; 3859 } 3860 } 3861 3862 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 3863 *idxattr)) { 3864 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS; 3865 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS); 3866 if (!attr) 3867 goto nla_put_failure; 3868 3869 err = rtnl_get_offload_stats(skb, dev, prividx); 3870 if (err == -ENODATA) 3871 nla_nest_cancel(skb, attr); 3872 else 3873 nla_nest_end(skb, attr); 3874 3875 if (err && err != -ENODATA) 3876 goto nla_put_failure; 3877 *idxattr = 0; 3878 } 3879 3880 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) { 3881 struct rtnl_af_ops *af_ops; 3882 3883 *idxattr = IFLA_STATS_AF_SPEC; 3884 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC); 3885 if (!attr) 3886 goto nla_put_failure; 3887 3888 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3889 if (af_ops->fill_stats_af) { 3890 struct nlattr *af; 3891 int err; 3892 3893 af = nla_nest_start(skb, af_ops->family); 3894 if (!af) 3895 goto nla_put_failure; 3896 3897 err = af_ops->fill_stats_af(skb, dev); 3898 3899 if (err == -ENODATA) 3900 nla_nest_cancel(skb, af); 3901 else if (err < 0) 3902 goto nla_put_failure; 3903 3904 nla_nest_end(skb, af); 3905 } 3906 } 3907 3908 nla_nest_end(skb, attr); 3909 3910 *idxattr = 0; 3911 } 3912 3913 nlmsg_end(skb, nlh); 3914 3915 return 0; 3916 3917 nla_put_failure: 3918 /* not a multi message or no progress mean a real error */ 3919 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx) 3920 nlmsg_cancel(skb, nlh); 3921 else 3922 nlmsg_end(skb, nlh); 3923 3924 return -EMSGSIZE; 3925 } 3926 3927 static size_t if_nlmsg_stats_size(const struct net_device *dev, 3928 u32 filter_mask) 3929 { 3930 size_t size = 0; 3931 3932 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0)) 3933 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64)); 3934 3935 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) { 3936 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 3937 int attr = IFLA_STATS_LINK_XSTATS; 3938 3939 if (ops && ops->get_linkxstats_size) { 3940 size += nla_total_size(ops->get_linkxstats_size(dev, 3941 attr)); 3942 /* for IFLA_STATS_LINK_XSTATS */ 3943 size += nla_total_size(0); 3944 } 3945 } 3946 3947 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) { 3948 struct net_device *_dev = (struct net_device *)dev; 3949 const struct rtnl_link_ops *ops = NULL; 3950 const struct net_device *master; 3951 3952 /* netdev_master_upper_dev_get can't take const */ 3953 master = netdev_master_upper_dev_get(_dev); 3954 if (master) 3955 ops = master->rtnl_link_ops; 3956 if (ops && ops->get_linkxstats_size) { 3957 int attr = IFLA_STATS_LINK_XSTATS_SLAVE; 3958 3959 size += nla_total_size(ops->get_linkxstats_size(dev, 3960 attr)); 3961 /* for IFLA_STATS_LINK_XSTATS_SLAVE */ 3962 size += nla_total_size(0); 3963 } 3964 } 3965 3966 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) 3967 size += rtnl_get_offload_stats_size(dev); 3968 3969 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) { 3970 struct rtnl_af_ops *af_ops; 3971 3972 /* for IFLA_STATS_AF_SPEC */ 3973 size += nla_total_size(0); 3974 3975 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 3976 if (af_ops->get_stats_af_size) { 3977 size += nla_total_size( 3978 af_ops->get_stats_af_size(dev)); 3979 3980 /* for AF_* */ 3981 size += nla_total_size(0); 3982 } 3983 } 3984 } 3985 3986 return size; 3987 } 3988 3989 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh, 3990 struct netlink_ext_ack *extack) 3991 { 3992 struct net *net = sock_net(skb->sk); 3993 struct net_device *dev = NULL; 3994 int idxattr = 0, prividx = 0; 3995 struct if_stats_msg *ifsm; 3996 struct sk_buff *nskb; 3997 u32 filter_mask; 3998 int err; 3999 4000 if (nlmsg_len(nlh) < sizeof(*ifsm)) 4001 return -EINVAL; 4002 4003 ifsm = nlmsg_data(nlh); 4004 if (ifsm->ifindex > 0) 4005 dev = __dev_get_by_index(net, ifsm->ifindex); 4006 else 4007 return -EINVAL; 4008 4009 if (!dev) 4010 return -ENODEV; 4011 4012 filter_mask = ifsm->filter_mask; 4013 if (!filter_mask) 4014 return -EINVAL; 4015 4016 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL); 4017 if (!nskb) 4018 return -ENOBUFS; 4019 4020 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS, 4021 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0, 4022 0, filter_mask, &idxattr, &prividx); 4023 if (err < 0) { 4024 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */ 4025 WARN_ON(err == -EMSGSIZE); 4026 kfree_skb(nskb); 4027 } else { 4028 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 4029 } 4030 4031 return err; 4032 } 4033 4034 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb) 4035 { 4036 int h, s_h, err, s_idx, s_idxattr, s_prividx; 4037 struct net *net = sock_net(skb->sk); 4038 unsigned int flags = NLM_F_MULTI; 4039 struct if_stats_msg *ifsm; 4040 struct hlist_head *head; 4041 struct net_device *dev; 4042 u32 filter_mask = 0; 4043 int idx = 0; 4044 4045 s_h = cb->args[0]; 4046 s_idx = cb->args[1]; 4047 s_idxattr = cb->args[2]; 4048 s_prividx = cb->args[3]; 4049 4050 cb->seq = net->dev_base_seq; 4051 4052 if (nlmsg_len(cb->nlh) < sizeof(*ifsm)) 4053 return -EINVAL; 4054 4055 ifsm = nlmsg_data(cb->nlh); 4056 filter_mask = ifsm->filter_mask; 4057 if (!filter_mask) 4058 return -EINVAL; 4059 4060 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4061 idx = 0; 4062 head = &net->dev_index_head[h]; 4063 hlist_for_each_entry(dev, head, index_hlist) { 4064 if (idx < s_idx) 4065 goto cont; 4066 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 4067 NETLINK_CB(cb->skb).portid, 4068 cb->nlh->nlmsg_seq, 0, 4069 flags, filter_mask, 4070 &s_idxattr, &s_prividx); 4071 /* If we ran out of room on the first message, 4072 * we're in trouble 4073 */ 4074 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 4075 4076 if (err < 0) 4077 goto out; 4078 s_prividx = 0; 4079 s_idxattr = 0; 4080 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 4081 cont: 4082 idx++; 4083 } 4084 } 4085 out: 4086 cb->args[3] = s_prividx; 4087 cb->args[2] = s_idxattr; 4088 cb->args[1] = idx; 4089 cb->args[0] = h; 4090 4091 return skb->len; 4092 } 4093 4094 /* Process one rtnetlink message. */ 4095 4096 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, 4097 struct netlink_ext_ack *extack) 4098 { 4099 struct net *net = sock_net(skb->sk); 4100 rtnl_doit_func doit; 4101 int kind; 4102 int family; 4103 int type; 4104 int err; 4105 4106 type = nlh->nlmsg_type; 4107 if (type > RTM_MAX) 4108 return -EOPNOTSUPP; 4109 4110 type -= RTM_BASE; 4111 4112 /* All the messages must have at least 1 byte length */ 4113 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 4114 return 0; 4115 4116 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 4117 kind = type&3; 4118 4119 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 4120 return -EPERM; 4121 4122 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 4123 struct sock *rtnl; 4124 rtnl_dumpit_func dumpit; 4125 rtnl_calcit_func calcit; 4126 u16 min_dump_alloc = 0; 4127 4128 dumpit = rtnl_get_dumpit(family, type); 4129 if (dumpit == NULL) 4130 return -EOPNOTSUPP; 4131 calcit = rtnl_get_calcit(family, type); 4132 if (calcit) 4133 min_dump_alloc = calcit(skb, nlh); 4134 4135 __rtnl_unlock(); 4136 rtnl = net->rtnl; 4137 { 4138 struct netlink_dump_control c = { 4139 .dump = dumpit, 4140 .min_dump_alloc = min_dump_alloc, 4141 }; 4142 err = netlink_dump_start(rtnl, skb, nlh, &c); 4143 } 4144 rtnl_lock(); 4145 return err; 4146 } 4147 4148 doit = rtnl_get_doit(family, type); 4149 if (doit == NULL) 4150 return -EOPNOTSUPP; 4151 4152 return doit(skb, nlh, extack); 4153 } 4154 4155 static void rtnetlink_rcv(struct sk_buff *skb) 4156 { 4157 rtnl_lock(); 4158 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 4159 rtnl_unlock(); 4160 } 4161 4162 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 4163 { 4164 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 4165 4166 switch (event) { 4167 case NETDEV_REBOOT: 4168 case NETDEV_CHANGENAME: 4169 case NETDEV_FEAT_CHANGE: 4170 case NETDEV_BONDING_FAILOVER: 4171 case NETDEV_NOTIFY_PEERS: 4172 case NETDEV_RESEND_IGMP: 4173 case NETDEV_CHANGEINFODATA: 4174 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL); 4175 break; 4176 default: 4177 break; 4178 } 4179 return NOTIFY_DONE; 4180 } 4181 4182 static struct notifier_block rtnetlink_dev_notifier = { 4183 .notifier_call = rtnetlink_event, 4184 }; 4185 4186 4187 static int __net_init rtnetlink_net_init(struct net *net) 4188 { 4189 struct sock *sk; 4190 struct netlink_kernel_cfg cfg = { 4191 .groups = RTNLGRP_MAX, 4192 .input = rtnetlink_rcv, 4193 .cb_mutex = &rtnl_mutex, 4194 .flags = NL_CFG_F_NONROOT_RECV, 4195 }; 4196 4197 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 4198 if (!sk) 4199 return -ENOMEM; 4200 net->rtnl = sk; 4201 return 0; 4202 } 4203 4204 static void __net_exit rtnetlink_net_exit(struct net *net) 4205 { 4206 netlink_kernel_release(net->rtnl); 4207 net->rtnl = NULL; 4208 } 4209 4210 static struct pernet_operations rtnetlink_net_ops = { 4211 .init = rtnetlink_net_init, 4212 .exit = rtnetlink_net_exit, 4213 }; 4214 4215 void __init rtnetlink_init(void) 4216 { 4217 if (register_pernet_subsys(&rtnetlink_net_ops)) 4218 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 4219 4220 register_netdevice_notifier(&rtnetlink_dev_notifier); 4221 4222 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 4223 rtnl_dump_ifinfo, rtnl_calcit); 4224 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL); 4225 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL); 4226 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL); 4227 4228 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL); 4229 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL); 4230 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL); 4231 4232 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL); 4233 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL); 4234 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL); 4235 4236 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL); 4237 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL); 4238 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL); 4239 4240 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump, 4241 NULL); 4242 } 4243