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