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/ip.h> 48 #include <net/protocol.h> 49 #include <net/arp.h> 50 #include <net/route.h> 51 #include <net/udp.h> 52 #include <net/sock.h> 53 #include <net/pkt_sched.h> 54 #include <net/fib_rules.h> 55 #include <net/rtnetlink.h> 56 #include <net/net_namespace.h> 57 58 struct rtnl_link { 59 rtnl_doit_func doit; 60 rtnl_dumpit_func dumpit; 61 rtnl_calcit_func calcit; 62 }; 63 64 static DEFINE_MUTEX(rtnl_mutex); 65 66 void rtnl_lock(void) 67 { 68 mutex_lock(&rtnl_mutex); 69 } 70 EXPORT_SYMBOL(rtnl_lock); 71 72 void __rtnl_unlock(void) 73 { 74 mutex_unlock(&rtnl_mutex); 75 } 76 77 void rtnl_unlock(void) 78 { 79 /* This fellow will unlock it for us. */ 80 netdev_run_todo(); 81 } 82 EXPORT_SYMBOL(rtnl_unlock); 83 84 int rtnl_trylock(void) 85 { 86 return mutex_trylock(&rtnl_mutex); 87 } 88 EXPORT_SYMBOL(rtnl_trylock); 89 90 int rtnl_is_locked(void) 91 { 92 return mutex_is_locked(&rtnl_mutex); 93 } 94 EXPORT_SYMBOL(rtnl_is_locked); 95 96 #ifdef CONFIG_PROVE_LOCKING 97 int lockdep_rtnl_is_held(void) 98 { 99 return lockdep_is_held(&rtnl_mutex); 100 } 101 EXPORT_SYMBOL(lockdep_rtnl_is_held); 102 #endif /* #ifdef CONFIG_PROVE_LOCKING */ 103 104 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1]; 105 106 static inline int rtm_msgindex(int msgtype) 107 { 108 int msgindex = msgtype - RTM_BASE; 109 110 /* 111 * msgindex < 0 implies someone tried to register a netlink 112 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that 113 * the message type has not been added to linux/rtnetlink.h 114 */ 115 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES); 116 117 return msgindex; 118 } 119 120 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex) 121 { 122 struct rtnl_link *tab; 123 124 if (protocol <= RTNL_FAMILY_MAX) 125 tab = rtnl_msg_handlers[protocol]; 126 else 127 tab = NULL; 128 129 if (tab == NULL || tab[msgindex].doit == NULL) 130 tab = rtnl_msg_handlers[PF_UNSPEC]; 131 132 return tab[msgindex].doit; 133 } 134 135 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex) 136 { 137 struct rtnl_link *tab; 138 139 if (protocol <= RTNL_FAMILY_MAX) 140 tab = rtnl_msg_handlers[protocol]; 141 else 142 tab = NULL; 143 144 if (tab == NULL || tab[msgindex].dumpit == NULL) 145 tab = rtnl_msg_handlers[PF_UNSPEC]; 146 147 return tab[msgindex].dumpit; 148 } 149 150 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex) 151 { 152 struct rtnl_link *tab; 153 154 if (protocol <= RTNL_FAMILY_MAX) 155 tab = rtnl_msg_handlers[protocol]; 156 else 157 tab = NULL; 158 159 if (tab == NULL || tab[msgindex].calcit == NULL) 160 tab = rtnl_msg_handlers[PF_UNSPEC]; 161 162 return tab[msgindex].calcit; 163 } 164 165 /** 166 * __rtnl_register - Register a rtnetlink message type 167 * @protocol: Protocol family or PF_UNSPEC 168 * @msgtype: rtnetlink message type 169 * @doit: Function pointer called for each request message 170 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message 171 * @calcit: Function pointer to calc size of dump message 172 * 173 * Registers the specified function pointers (at least one of them has 174 * to be non-NULL) to be called whenever a request message for the 175 * specified protocol family and message type is received. 176 * 177 * The special protocol family PF_UNSPEC may be used to define fallback 178 * function pointers for the case when no entry for the specific protocol 179 * family exists. 180 * 181 * Returns 0 on success or a negative error code. 182 */ 183 int __rtnl_register(int protocol, int msgtype, 184 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 185 rtnl_calcit_func calcit) 186 { 187 struct rtnl_link *tab; 188 int msgindex; 189 190 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 191 msgindex = rtm_msgindex(msgtype); 192 193 tab = rtnl_msg_handlers[protocol]; 194 if (tab == NULL) { 195 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL); 196 if (tab == NULL) 197 return -ENOBUFS; 198 199 rtnl_msg_handlers[protocol] = tab; 200 } 201 202 if (doit) 203 tab[msgindex].doit = doit; 204 205 if (dumpit) 206 tab[msgindex].dumpit = dumpit; 207 208 if (calcit) 209 tab[msgindex].calcit = calcit; 210 211 return 0; 212 } 213 EXPORT_SYMBOL_GPL(__rtnl_register); 214 215 /** 216 * rtnl_register - Register a rtnetlink message type 217 * 218 * Identical to __rtnl_register() but panics on failure. This is useful 219 * as failure of this function is very unlikely, it can only happen due 220 * to lack of memory when allocating the chain to store all message 221 * handlers for a protocol. Meant for use in init functions where lack 222 * of memory implies no sense in continuing. 223 */ 224 void rtnl_register(int protocol, int msgtype, 225 rtnl_doit_func doit, rtnl_dumpit_func dumpit, 226 rtnl_calcit_func calcit) 227 { 228 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0) 229 panic("Unable to register rtnetlink message handler, " 230 "protocol = %d, message type = %d\n", 231 protocol, msgtype); 232 } 233 EXPORT_SYMBOL_GPL(rtnl_register); 234 235 /** 236 * rtnl_unregister - Unregister a rtnetlink message type 237 * @protocol: Protocol family or PF_UNSPEC 238 * @msgtype: rtnetlink message type 239 * 240 * Returns 0 on success or a negative error code. 241 */ 242 int rtnl_unregister(int protocol, int msgtype) 243 { 244 int msgindex; 245 246 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 247 msgindex = rtm_msgindex(msgtype); 248 249 if (rtnl_msg_handlers[protocol] == NULL) 250 return -ENOENT; 251 252 rtnl_msg_handlers[protocol][msgindex].doit = NULL; 253 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL; 254 255 return 0; 256 } 257 EXPORT_SYMBOL_GPL(rtnl_unregister); 258 259 /** 260 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol 261 * @protocol : Protocol family or PF_UNSPEC 262 * 263 * Identical to calling rtnl_unregster() for all registered message types 264 * of a certain protocol family. 265 */ 266 void rtnl_unregister_all(int protocol) 267 { 268 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX); 269 270 kfree(rtnl_msg_handlers[protocol]); 271 rtnl_msg_handlers[protocol] = NULL; 272 } 273 EXPORT_SYMBOL_GPL(rtnl_unregister_all); 274 275 static LIST_HEAD(link_ops); 276 277 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind) 278 { 279 const struct rtnl_link_ops *ops; 280 281 list_for_each_entry(ops, &link_ops, list) { 282 if (!strcmp(ops->kind, kind)) 283 return ops; 284 } 285 return NULL; 286 } 287 288 /** 289 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink. 290 * @ops: struct rtnl_link_ops * to register 291 * 292 * The caller must hold the rtnl_mutex. This function should be used 293 * by drivers that create devices during module initialization. It 294 * must be called before registering the devices. 295 * 296 * Returns 0 on success or a negative error code. 297 */ 298 int __rtnl_link_register(struct rtnl_link_ops *ops) 299 { 300 if (rtnl_link_ops_get(ops->kind)) 301 return -EEXIST; 302 303 /* The check for setup is here because if ops 304 * does not have that filled up, it is not possible 305 * to use the ops for creating device. So do not 306 * fill up dellink as well. That disables rtnl_dellink. 307 */ 308 if (ops->setup && !ops->dellink) 309 ops->dellink = unregister_netdevice_queue; 310 311 list_add_tail(&ops->list, &link_ops); 312 return 0; 313 } 314 EXPORT_SYMBOL_GPL(__rtnl_link_register); 315 316 /** 317 * rtnl_link_register - Register rtnl_link_ops with rtnetlink. 318 * @ops: struct rtnl_link_ops * to register 319 * 320 * Returns 0 on success or a negative error code. 321 */ 322 int rtnl_link_register(struct rtnl_link_ops *ops) 323 { 324 int err; 325 326 rtnl_lock(); 327 err = __rtnl_link_register(ops); 328 rtnl_unlock(); 329 return err; 330 } 331 EXPORT_SYMBOL_GPL(rtnl_link_register); 332 333 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops) 334 { 335 struct net_device *dev; 336 LIST_HEAD(list_kill); 337 338 for_each_netdev(net, dev) { 339 if (dev->rtnl_link_ops == ops) 340 ops->dellink(dev, &list_kill); 341 } 342 unregister_netdevice_many(&list_kill); 343 } 344 345 /** 346 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 347 * @ops: struct rtnl_link_ops * to unregister 348 * 349 * The caller must hold the rtnl_mutex. 350 */ 351 void __rtnl_link_unregister(struct rtnl_link_ops *ops) 352 { 353 struct net *net; 354 355 for_each_net(net) { 356 __rtnl_kill_links(net, ops); 357 } 358 list_del(&ops->list); 359 } 360 EXPORT_SYMBOL_GPL(__rtnl_link_unregister); 361 362 /* Return with the rtnl_lock held when there are no network 363 * devices unregistering in any network namespace. 364 */ 365 static void rtnl_lock_unregistering_all(void) 366 { 367 struct net *net; 368 bool unregistering; 369 DEFINE_WAIT(wait); 370 371 for (;;) { 372 prepare_to_wait(&netdev_unregistering_wq, &wait, 373 TASK_UNINTERRUPTIBLE); 374 unregistering = false; 375 rtnl_lock(); 376 for_each_net(net) { 377 if (net->dev_unreg_count > 0) { 378 unregistering = true; 379 break; 380 } 381 } 382 if (!unregistering) 383 break; 384 __rtnl_unlock(); 385 schedule(); 386 } 387 finish_wait(&netdev_unregistering_wq, &wait); 388 } 389 390 /** 391 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink. 392 * @ops: struct rtnl_link_ops * to unregister 393 */ 394 void rtnl_link_unregister(struct rtnl_link_ops *ops) 395 { 396 /* Close the race with cleanup_net() */ 397 mutex_lock(&net_mutex); 398 rtnl_lock_unregistering_all(); 399 __rtnl_link_unregister(ops); 400 rtnl_unlock(); 401 mutex_unlock(&net_mutex); 402 } 403 EXPORT_SYMBOL_GPL(rtnl_link_unregister); 404 405 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev) 406 { 407 struct net_device *master_dev; 408 const struct rtnl_link_ops *ops; 409 410 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 411 if (!master_dev) 412 return 0; 413 ops = master_dev->rtnl_link_ops; 414 if (!ops || !ops->get_slave_size) 415 return 0; 416 /* IFLA_INFO_SLAVE_DATA + nested data */ 417 return nla_total_size(sizeof(struct nlattr)) + 418 ops->get_slave_size(master_dev, dev); 419 } 420 421 static size_t rtnl_link_get_size(const struct net_device *dev) 422 { 423 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 424 size_t size; 425 426 if (!ops) 427 return 0; 428 429 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */ 430 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */ 431 432 if (ops->get_size) 433 /* IFLA_INFO_DATA + nested data */ 434 size += nla_total_size(sizeof(struct nlattr)) + 435 ops->get_size(dev); 436 437 if (ops->get_xstats_size) 438 /* IFLA_INFO_XSTATS */ 439 size += nla_total_size(ops->get_xstats_size(dev)); 440 441 size += rtnl_link_get_slave_info_data_size(dev); 442 443 return size; 444 } 445 446 static LIST_HEAD(rtnl_af_ops); 447 448 static const struct rtnl_af_ops *rtnl_af_lookup(const int family) 449 { 450 const struct rtnl_af_ops *ops; 451 452 list_for_each_entry(ops, &rtnl_af_ops, list) { 453 if (ops->family == family) 454 return ops; 455 } 456 457 return NULL; 458 } 459 460 /** 461 * rtnl_af_register - Register rtnl_af_ops with rtnetlink. 462 * @ops: struct rtnl_af_ops * to register 463 * 464 * Returns 0 on success or a negative error code. 465 */ 466 void rtnl_af_register(struct rtnl_af_ops *ops) 467 { 468 rtnl_lock(); 469 list_add_tail(&ops->list, &rtnl_af_ops); 470 rtnl_unlock(); 471 } 472 EXPORT_SYMBOL_GPL(rtnl_af_register); 473 474 /** 475 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 476 * @ops: struct rtnl_af_ops * to unregister 477 * 478 * The caller must hold the rtnl_mutex. 479 */ 480 void __rtnl_af_unregister(struct rtnl_af_ops *ops) 481 { 482 list_del(&ops->list); 483 } 484 EXPORT_SYMBOL_GPL(__rtnl_af_unregister); 485 486 /** 487 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink. 488 * @ops: struct rtnl_af_ops * to unregister 489 */ 490 void rtnl_af_unregister(struct rtnl_af_ops *ops) 491 { 492 rtnl_lock(); 493 __rtnl_af_unregister(ops); 494 rtnl_unlock(); 495 } 496 EXPORT_SYMBOL_GPL(rtnl_af_unregister); 497 498 static size_t rtnl_link_get_af_size(const struct net_device *dev) 499 { 500 struct rtnl_af_ops *af_ops; 501 size_t size; 502 503 /* IFLA_AF_SPEC */ 504 size = nla_total_size(sizeof(struct nlattr)); 505 506 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 507 if (af_ops->get_link_af_size) { 508 /* AF_* + nested data */ 509 size += nla_total_size(sizeof(struct nlattr)) + 510 af_ops->get_link_af_size(dev); 511 } 512 } 513 514 return size; 515 } 516 517 static bool rtnl_have_link_slave_info(const struct net_device *dev) 518 { 519 struct net_device *master_dev; 520 521 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 522 if (master_dev && master_dev->rtnl_link_ops) 523 return true; 524 return false; 525 } 526 527 static int rtnl_link_slave_info_fill(struct sk_buff *skb, 528 const struct net_device *dev) 529 { 530 struct net_device *master_dev; 531 const struct rtnl_link_ops *ops; 532 struct nlattr *slave_data; 533 int err; 534 535 master_dev = netdev_master_upper_dev_get((struct net_device *) dev); 536 if (!master_dev) 537 return 0; 538 ops = master_dev->rtnl_link_ops; 539 if (!ops) 540 return 0; 541 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0) 542 return -EMSGSIZE; 543 if (ops->fill_slave_info) { 544 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA); 545 if (!slave_data) 546 return -EMSGSIZE; 547 err = ops->fill_slave_info(skb, master_dev, dev); 548 if (err < 0) 549 goto err_cancel_slave_data; 550 nla_nest_end(skb, slave_data); 551 } 552 return 0; 553 554 err_cancel_slave_data: 555 nla_nest_cancel(skb, slave_data); 556 return err; 557 } 558 559 static int rtnl_link_info_fill(struct sk_buff *skb, 560 const struct net_device *dev) 561 { 562 const struct rtnl_link_ops *ops = dev->rtnl_link_ops; 563 struct nlattr *data; 564 int err; 565 566 if (!ops) 567 return 0; 568 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0) 569 return -EMSGSIZE; 570 if (ops->fill_xstats) { 571 err = ops->fill_xstats(skb, dev); 572 if (err < 0) 573 return err; 574 } 575 if (ops->fill_info) { 576 data = nla_nest_start(skb, IFLA_INFO_DATA); 577 if (data == NULL) 578 return -EMSGSIZE; 579 err = ops->fill_info(skb, dev); 580 if (err < 0) 581 goto err_cancel_data; 582 nla_nest_end(skb, data); 583 } 584 return 0; 585 586 err_cancel_data: 587 nla_nest_cancel(skb, data); 588 return err; 589 } 590 591 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev) 592 { 593 struct nlattr *linkinfo; 594 int err = -EMSGSIZE; 595 596 linkinfo = nla_nest_start(skb, IFLA_LINKINFO); 597 if (linkinfo == NULL) 598 goto out; 599 600 err = rtnl_link_info_fill(skb, dev); 601 if (err < 0) 602 goto err_cancel_link; 603 604 err = rtnl_link_slave_info_fill(skb, dev); 605 if (err < 0) 606 goto err_cancel_link; 607 608 nla_nest_end(skb, linkinfo); 609 return 0; 610 611 err_cancel_link: 612 nla_nest_cancel(skb, linkinfo); 613 out: 614 return err; 615 } 616 617 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo) 618 { 619 struct sock *rtnl = net->rtnl; 620 int err = 0; 621 622 NETLINK_CB(skb).dst_group = group; 623 if (echo) 624 atomic_inc(&skb->users); 625 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL); 626 if (echo) 627 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT); 628 return err; 629 } 630 631 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid) 632 { 633 struct sock *rtnl = net->rtnl; 634 635 return nlmsg_unicast(rtnl, skb, pid); 636 } 637 EXPORT_SYMBOL(rtnl_unicast); 638 639 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group, 640 struct nlmsghdr *nlh, gfp_t flags) 641 { 642 struct sock *rtnl = net->rtnl; 643 int report = 0; 644 645 if (nlh) 646 report = nlmsg_report(nlh); 647 648 nlmsg_notify(rtnl, skb, pid, group, report, flags); 649 } 650 EXPORT_SYMBOL(rtnl_notify); 651 652 void rtnl_set_sk_err(struct net *net, u32 group, int error) 653 { 654 struct sock *rtnl = net->rtnl; 655 656 netlink_set_err(rtnl, 0, group, error); 657 } 658 EXPORT_SYMBOL(rtnl_set_sk_err); 659 660 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics) 661 { 662 struct nlattr *mx; 663 int i, valid = 0; 664 665 mx = nla_nest_start(skb, RTA_METRICS); 666 if (mx == NULL) 667 return -ENOBUFS; 668 669 for (i = 0; i < RTAX_MAX; i++) { 670 if (metrics[i]) { 671 valid++; 672 if (nla_put_u32(skb, i+1, metrics[i])) 673 goto nla_put_failure; 674 } 675 } 676 677 if (!valid) { 678 nla_nest_cancel(skb, mx); 679 return 0; 680 } 681 682 return nla_nest_end(skb, mx); 683 684 nla_put_failure: 685 nla_nest_cancel(skb, mx); 686 return -EMSGSIZE; 687 } 688 EXPORT_SYMBOL(rtnetlink_put_metrics); 689 690 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id, 691 long expires, u32 error) 692 { 693 struct rta_cacheinfo ci = { 694 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse), 695 .rta_used = dst->__use, 696 .rta_clntref = atomic_read(&(dst->__refcnt)), 697 .rta_error = error, 698 .rta_id = id, 699 }; 700 701 if (expires) { 702 unsigned long clock; 703 704 clock = jiffies_to_clock_t(abs(expires)); 705 clock = min_t(unsigned long, clock, INT_MAX); 706 ci.rta_expires = (expires > 0) ? clock : -clock; 707 } 708 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci); 709 } 710 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo); 711 712 static void set_operstate(struct net_device *dev, unsigned char transition) 713 { 714 unsigned char operstate = dev->operstate; 715 716 switch (transition) { 717 case IF_OPER_UP: 718 if ((operstate == IF_OPER_DORMANT || 719 operstate == IF_OPER_UNKNOWN) && 720 !netif_dormant(dev)) 721 operstate = IF_OPER_UP; 722 break; 723 724 case IF_OPER_DORMANT: 725 if (operstate == IF_OPER_UP || 726 operstate == IF_OPER_UNKNOWN) 727 operstate = IF_OPER_DORMANT; 728 break; 729 } 730 731 if (dev->operstate != operstate) { 732 write_lock_bh(&dev_base_lock); 733 dev->operstate = operstate; 734 write_unlock_bh(&dev_base_lock); 735 netdev_state_change(dev); 736 } 737 } 738 739 static unsigned int rtnl_dev_get_flags(const struct net_device *dev) 740 { 741 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) | 742 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI)); 743 } 744 745 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev, 746 const struct ifinfomsg *ifm) 747 { 748 unsigned int flags = ifm->ifi_flags; 749 750 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */ 751 if (ifm->ifi_change) 752 flags = (flags & ifm->ifi_change) | 753 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change); 754 755 return flags; 756 } 757 758 static void copy_rtnl_link_stats(struct rtnl_link_stats *a, 759 const struct rtnl_link_stats64 *b) 760 { 761 a->rx_packets = b->rx_packets; 762 a->tx_packets = b->tx_packets; 763 a->rx_bytes = b->rx_bytes; 764 a->tx_bytes = b->tx_bytes; 765 a->rx_errors = b->rx_errors; 766 a->tx_errors = b->tx_errors; 767 a->rx_dropped = b->rx_dropped; 768 a->tx_dropped = b->tx_dropped; 769 770 a->multicast = b->multicast; 771 a->collisions = b->collisions; 772 773 a->rx_length_errors = b->rx_length_errors; 774 a->rx_over_errors = b->rx_over_errors; 775 a->rx_crc_errors = b->rx_crc_errors; 776 a->rx_frame_errors = b->rx_frame_errors; 777 a->rx_fifo_errors = b->rx_fifo_errors; 778 a->rx_missed_errors = b->rx_missed_errors; 779 780 a->tx_aborted_errors = b->tx_aborted_errors; 781 a->tx_carrier_errors = b->tx_carrier_errors; 782 a->tx_fifo_errors = b->tx_fifo_errors; 783 a->tx_heartbeat_errors = b->tx_heartbeat_errors; 784 a->tx_window_errors = b->tx_window_errors; 785 786 a->rx_compressed = b->rx_compressed; 787 a->tx_compressed = b->tx_compressed; 788 } 789 790 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b) 791 { 792 memcpy(v, b, sizeof(*b)); 793 } 794 795 /* All VF info */ 796 static inline int rtnl_vfinfo_size(const struct net_device *dev, 797 u32 ext_filter_mask) 798 { 799 if (dev->dev.parent && dev_is_pci(dev->dev.parent) && 800 (ext_filter_mask & RTEXT_FILTER_VF)) { 801 int num_vfs = dev_num_vf(dev->dev.parent); 802 size_t size = nla_total_size(sizeof(struct nlattr)); 803 size += nla_total_size(num_vfs * sizeof(struct nlattr)); 804 size += num_vfs * 805 (nla_total_size(sizeof(struct ifla_vf_mac)) + 806 nla_total_size(sizeof(struct ifla_vf_vlan)) + 807 nla_total_size(sizeof(struct ifla_vf_spoofchk)) + 808 nla_total_size(sizeof(struct ifla_vf_rate)) + 809 nla_total_size(sizeof(struct ifla_vf_link_state))); 810 return size; 811 } else 812 return 0; 813 } 814 815 static size_t rtnl_port_size(const struct net_device *dev, 816 u32 ext_filter_mask) 817 { 818 size_t port_size = nla_total_size(4) /* PORT_VF */ 819 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */ 820 + nla_total_size(sizeof(struct ifla_port_vsi)) 821 /* PORT_VSI_TYPE */ 822 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */ 823 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */ 824 + nla_total_size(1) /* PROT_VDP_REQUEST */ 825 + nla_total_size(2); /* PORT_VDP_RESPONSE */ 826 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr)); 827 size_t vf_port_size = nla_total_size(sizeof(struct nlattr)) 828 + port_size; 829 size_t port_self_size = nla_total_size(sizeof(struct nlattr)) 830 + port_size; 831 832 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 833 !(ext_filter_mask & RTEXT_FILTER_VF)) 834 return 0; 835 if (dev_num_vf(dev->dev.parent)) 836 return port_self_size + vf_ports_size + 837 vf_port_size * dev_num_vf(dev->dev.parent); 838 else 839 return port_self_size; 840 } 841 842 static noinline size_t if_nlmsg_size(const struct net_device *dev, 843 u32 ext_filter_mask) 844 { 845 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 846 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 847 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */ 848 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */ 849 + nla_total_size(sizeof(struct rtnl_link_ifmap)) 850 + nla_total_size(sizeof(struct rtnl_link_stats)) 851 + nla_total_size(sizeof(struct rtnl_link_stats64)) 852 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 853 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */ 854 + nla_total_size(4) /* IFLA_TXQLEN */ 855 + nla_total_size(4) /* IFLA_WEIGHT */ 856 + nla_total_size(4) /* IFLA_MTU */ 857 + nla_total_size(4) /* IFLA_LINK */ 858 + nla_total_size(4) /* IFLA_MASTER */ 859 + nla_total_size(1) /* IFLA_CARRIER */ 860 + nla_total_size(4) /* IFLA_PROMISCUITY */ 861 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */ 862 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */ 863 + nla_total_size(1) /* IFLA_OPERSTATE */ 864 + nla_total_size(1) /* IFLA_LINKMODE */ 865 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */ 866 + nla_total_size(ext_filter_mask 867 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */ 868 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */ 869 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */ 870 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */ 871 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */ 872 + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */ 873 } 874 875 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev) 876 { 877 struct nlattr *vf_ports; 878 struct nlattr *vf_port; 879 int vf; 880 int err; 881 882 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS); 883 if (!vf_ports) 884 return -EMSGSIZE; 885 886 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) { 887 vf_port = nla_nest_start(skb, IFLA_VF_PORT); 888 if (!vf_port) 889 goto nla_put_failure; 890 if (nla_put_u32(skb, IFLA_PORT_VF, vf)) 891 goto nla_put_failure; 892 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb); 893 if (err == -EMSGSIZE) 894 goto nla_put_failure; 895 if (err) { 896 nla_nest_cancel(skb, vf_port); 897 continue; 898 } 899 nla_nest_end(skb, vf_port); 900 } 901 902 nla_nest_end(skb, vf_ports); 903 904 return 0; 905 906 nla_put_failure: 907 nla_nest_cancel(skb, vf_ports); 908 return -EMSGSIZE; 909 } 910 911 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev) 912 { 913 struct nlattr *port_self; 914 int err; 915 916 port_self = nla_nest_start(skb, IFLA_PORT_SELF); 917 if (!port_self) 918 return -EMSGSIZE; 919 920 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb); 921 if (err) { 922 nla_nest_cancel(skb, port_self); 923 return (err == -EMSGSIZE) ? err : 0; 924 } 925 926 nla_nest_end(skb, port_self); 927 928 return 0; 929 } 930 931 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev, 932 u32 ext_filter_mask) 933 { 934 int err; 935 936 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent || 937 !(ext_filter_mask & RTEXT_FILTER_VF)) 938 return 0; 939 940 err = rtnl_port_self_fill(skb, dev); 941 if (err) 942 return err; 943 944 if (dev_num_vf(dev->dev.parent)) { 945 err = rtnl_vf_ports_fill(skb, dev); 946 if (err) 947 return err; 948 } 949 950 return 0; 951 } 952 953 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev) 954 { 955 int err; 956 struct netdev_phys_port_id ppid; 957 958 err = dev_get_phys_port_id(dev, &ppid); 959 if (err) { 960 if (err == -EOPNOTSUPP) 961 return 0; 962 return err; 963 } 964 965 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id)) 966 return -EMSGSIZE; 967 968 return 0; 969 } 970 971 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev, 972 int type, u32 pid, u32 seq, u32 change, 973 unsigned int flags, u32 ext_filter_mask) 974 { 975 struct ifinfomsg *ifm; 976 struct nlmsghdr *nlh; 977 struct rtnl_link_stats64 temp; 978 const struct rtnl_link_stats64 *stats; 979 struct nlattr *attr, *af_spec; 980 struct rtnl_af_ops *af_ops; 981 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 982 983 ASSERT_RTNL(); 984 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags); 985 if (nlh == NULL) 986 return -EMSGSIZE; 987 988 ifm = nlmsg_data(nlh); 989 ifm->ifi_family = AF_UNSPEC; 990 ifm->__ifi_pad = 0; 991 ifm->ifi_type = dev->type; 992 ifm->ifi_index = dev->ifindex; 993 ifm->ifi_flags = dev_get_flags(dev); 994 ifm->ifi_change = change; 995 996 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 997 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) || 998 nla_put_u8(skb, IFLA_OPERSTATE, 999 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) || 1000 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) || 1001 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 1002 nla_put_u32(skb, IFLA_GROUP, dev->group) || 1003 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) || 1004 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) || 1005 #ifdef CONFIG_RPS 1006 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) || 1007 #endif 1008 (dev->ifindex != dev->iflink && 1009 nla_put_u32(skb, IFLA_LINK, dev->iflink)) || 1010 (upper_dev && 1011 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) || 1012 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) || 1013 (dev->qdisc && 1014 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) || 1015 (dev->ifalias && 1016 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) || 1017 nla_put_u32(skb, IFLA_CARRIER_CHANGES, 1018 atomic_read(&dev->carrier_changes))) 1019 goto nla_put_failure; 1020 1021 if (1) { 1022 struct rtnl_link_ifmap map = { 1023 .mem_start = dev->mem_start, 1024 .mem_end = dev->mem_end, 1025 .base_addr = dev->base_addr, 1026 .irq = dev->irq, 1027 .dma = dev->dma, 1028 .port = dev->if_port, 1029 }; 1030 if (nla_put(skb, IFLA_MAP, sizeof(map), &map)) 1031 goto nla_put_failure; 1032 } 1033 1034 if (dev->addr_len) { 1035 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) || 1036 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast)) 1037 goto nla_put_failure; 1038 } 1039 1040 if (rtnl_phys_port_id_fill(skb, dev)) 1041 goto nla_put_failure; 1042 1043 attr = nla_reserve(skb, IFLA_STATS, 1044 sizeof(struct rtnl_link_stats)); 1045 if (attr == NULL) 1046 goto nla_put_failure; 1047 1048 stats = dev_get_stats(dev, &temp); 1049 copy_rtnl_link_stats(nla_data(attr), stats); 1050 1051 attr = nla_reserve(skb, IFLA_STATS64, 1052 sizeof(struct rtnl_link_stats64)); 1053 if (attr == NULL) 1054 goto nla_put_failure; 1055 copy_rtnl_link_stats64(nla_data(attr), stats); 1056 1057 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) && 1058 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent))) 1059 goto nla_put_failure; 1060 1061 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent 1062 && (ext_filter_mask & RTEXT_FILTER_VF)) { 1063 int i; 1064 1065 struct nlattr *vfinfo, *vf; 1066 int num_vfs = dev_num_vf(dev->dev.parent); 1067 1068 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST); 1069 if (!vfinfo) 1070 goto nla_put_failure; 1071 for (i = 0; i < num_vfs; i++) { 1072 struct ifla_vf_info ivi; 1073 struct ifla_vf_mac vf_mac; 1074 struct ifla_vf_vlan vf_vlan; 1075 struct ifla_vf_rate vf_rate; 1076 struct ifla_vf_tx_rate vf_tx_rate; 1077 struct ifla_vf_spoofchk vf_spoofchk; 1078 struct ifla_vf_link_state vf_linkstate; 1079 1080 /* 1081 * Not all SR-IOV capable drivers support the 1082 * spoofcheck query. Preset to -1 so the user 1083 * space tool can detect that the driver didn't 1084 * report anything. 1085 */ 1086 ivi.spoofchk = -1; 1087 memset(ivi.mac, 0, sizeof(ivi.mac)); 1088 /* The default value for VF link state is "auto" 1089 * IFLA_VF_LINK_STATE_AUTO which equals zero 1090 */ 1091 ivi.linkstate = 0; 1092 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi)) 1093 break; 1094 vf_mac.vf = 1095 vf_vlan.vf = 1096 vf_rate.vf = 1097 vf_tx_rate.vf = 1098 vf_spoofchk.vf = 1099 vf_linkstate.vf = ivi.vf; 1100 1101 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac)); 1102 vf_vlan.vlan = ivi.vlan; 1103 vf_vlan.qos = ivi.qos; 1104 vf_tx_rate.rate = ivi.max_tx_rate; 1105 vf_rate.min_tx_rate = ivi.min_tx_rate; 1106 vf_rate.max_tx_rate = ivi.max_tx_rate; 1107 vf_spoofchk.setting = ivi.spoofchk; 1108 vf_linkstate.link_state = ivi.linkstate; 1109 vf = nla_nest_start(skb, IFLA_VF_INFO); 1110 if (!vf) { 1111 nla_nest_cancel(skb, vfinfo); 1112 goto nla_put_failure; 1113 } 1114 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) || 1115 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) || 1116 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate), 1117 &vf_rate) || 1118 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), 1119 &vf_tx_rate) || 1120 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk), 1121 &vf_spoofchk) || 1122 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate), 1123 &vf_linkstate)) 1124 goto nla_put_failure; 1125 nla_nest_end(skb, vf); 1126 } 1127 nla_nest_end(skb, vfinfo); 1128 } 1129 1130 if (rtnl_port_fill(skb, dev, ext_filter_mask)) 1131 goto nla_put_failure; 1132 1133 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) { 1134 if (rtnl_link_fill(skb, dev) < 0) 1135 goto nla_put_failure; 1136 } 1137 1138 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC))) 1139 goto nla_put_failure; 1140 1141 list_for_each_entry(af_ops, &rtnl_af_ops, list) { 1142 if (af_ops->fill_link_af) { 1143 struct nlattr *af; 1144 int err; 1145 1146 if (!(af = nla_nest_start(skb, af_ops->family))) 1147 goto nla_put_failure; 1148 1149 err = af_ops->fill_link_af(skb, dev); 1150 1151 /* 1152 * Caller may return ENODATA to indicate that there 1153 * was no data to be dumped. This is not an error, it 1154 * means we should trim the attribute header and 1155 * continue. 1156 */ 1157 if (err == -ENODATA) 1158 nla_nest_cancel(skb, af); 1159 else if (err < 0) 1160 goto nla_put_failure; 1161 1162 nla_nest_end(skb, af); 1163 } 1164 } 1165 1166 nla_nest_end(skb, af_spec); 1167 1168 return nlmsg_end(skb, nlh); 1169 1170 nla_put_failure: 1171 nlmsg_cancel(skb, nlh); 1172 return -EMSGSIZE; 1173 } 1174 1175 static const struct nla_policy ifla_policy[IFLA_MAX+1] = { 1176 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 }, 1177 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1178 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN }, 1179 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) }, 1180 [IFLA_MTU] = { .type = NLA_U32 }, 1181 [IFLA_LINK] = { .type = NLA_U32 }, 1182 [IFLA_MASTER] = { .type = NLA_U32 }, 1183 [IFLA_CARRIER] = { .type = NLA_U8 }, 1184 [IFLA_TXQLEN] = { .type = NLA_U32 }, 1185 [IFLA_WEIGHT] = { .type = NLA_U32 }, 1186 [IFLA_OPERSTATE] = { .type = NLA_U8 }, 1187 [IFLA_LINKMODE] = { .type = NLA_U8 }, 1188 [IFLA_LINKINFO] = { .type = NLA_NESTED }, 1189 [IFLA_NET_NS_PID] = { .type = NLA_U32 }, 1190 [IFLA_NET_NS_FD] = { .type = NLA_U32 }, 1191 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 }, 1192 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED }, 1193 [IFLA_VF_PORTS] = { .type = NLA_NESTED }, 1194 [IFLA_PORT_SELF] = { .type = NLA_NESTED }, 1195 [IFLA_AF_SPEC] = { .type = NLA_NESTED }, 1196 [IFLA_EXT_MASK] = { .type = NLA_U32 }, 1197 [IFLA_PROMISCUITY] = { .type = NLA_U32 }, 1198 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 }, 1199 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 }, 1200 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN }, 1201 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */ 1202 }; 1203 1204 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = { 1205 [IFLA_INFO_KIND] = { .type = NLA_STRING }, 1206 [IFLA_INFO_DATA] = { .type = NLA_NESTED }, 1207 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING }, 1208 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED }, 1209 }; 1210 1211 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = { 1212 [IFLA_VF_INFO] = { .type = NLA_NESTED }, 1213 }; 1214 1215 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = { 1216 [IFLA_VF_MAC] = { .type = NLA_BINARY, 1217 .len = sizeof(struct ifla_vf_mac) }, 1218 [IFLA_VF_VLAN] = { .type = NLA_BINARY, 1219 .len = sizeof(struct ifla_vf_vlan) }, 1220 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY, 1221 .len = sizeof(struct ifla_vf_tx_rate) }, 1222 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY, 1223 .len = sizeof(struct ifla_vf_spoofchk) }, 1224 [IFLA_VF_RATE] = { .type = NLA_BINARY, 1225 .len = sizeof(struct ifla_vf_rate) }, 1226 [IFLA_VF_LINK_STATE] = { .type = NLA_BINARY, 1227 .len = sizeof(struct ifla_vf_link_state) }, 1228 }; 1229 1230 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = { 1231 [IFLA_PORT_VF] = { .type = NLA_U32 }, 1232 [IFLA_PORT_PROFILE] = { .type = NLA_STRING, 1233 .len = PORT_PROFILE_MAX }, 1234 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY, 1235 .len = sizeof(struct ifla_port_vsi)}, 1236 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY, 1237 .len = PORT_UUID_MAX }, 1238 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING, 1239 .len = PORT_UUID_MAX }, 1240 [IFLA_PORT_REQUEST] = { .type = NLA_U8, }, 1241 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, }, 1242 }; 1243 1244 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 1245 { 1246 struct net *net = sock_net(skb->sk); 1247 int h, s_h; 1248 int idx = 0, s_idx; 1249 struct net_device *dev; 1250 struct hlist_head *head; 1251 struct nlattr *tb[IFLA_MAX+1]; 1252 u32 ext_filter_mask = 0; 1253 int err; 1254 int hdrlen; 1255 1256 s_h = cb->args[0]; 1257 s_idx = cb->args[1]; 1258 1259 rcu_read_lock(); 1260 cb->seq = net->dev_base_seq; 1261 1262 /* A hack to preserve kernel<->userspace interface. 1263 * The correct header is ifinfomsg. It is consistent with rtnl_getlink. 1264 * However, before Linux v3.9 the code here assumed rtgenmsg and that's 1265 * what iproute2 < v3.9.0 used. 1266 * We can detect the old iproute2. Even including the IFLA_EXT_MASK 1267 * attribute, its netlink message is shorter than struct ifinfomsg. 1268 */ 1269 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ? 1270 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 1271 1272 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) { 1273 1274 if (tb[IFLA_EXT_MASK]) 1275 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 1276 } 1277 1278 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 1279 idx = 0; 1280 head = &net->dev_index_head[h]; 1281 hlist_for_each_entry_rcu(dev, head, index_hlist) { 1282 if (idx < s_idx) 1283 goto cont; 1284 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK, 1285 NETLINK_CB(cb->skb).portid, 1286 cb->nlh->nlmsg_seq, 0, 1287 NLM_F_MULTI, 1288 ext_filter_mask); 1289 /* If we ran out of room on the first message, 1290 * we're in trouble 1291 */ 1292 WARN_ON((err == -EMSGSIZE) && (skb->len == 0)); 1293 1294 if (err <= 0) 1295 goto out; 1296 1297 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1298 cont: 1299 idx++; 1300 } 1301 } 1302 out: 1303 rcu_read_unlock(); 1304 cb->args[1] = idx; 1305 cb->args[0] = h; 1306 1307 return skb->len; 1308 } 1309 1310 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len) 1311 { 1312 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy); 1313 } 1314 EXPORT_SYMBOL(rtnl_nla_parse_ifla); 1315 1316 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[]) 1317 { 1318 struct net *net; 1319 /* Examine the link attributes and figure out which 1320 * network namespace we are talking about. 1321 */ 1322 if (tb[IFLA_NET_NS_PID]) 1323 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID])); 1324 else if (tb[IFLA_NET_NS_FD]) 1325 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD])); 1326 else 1327 net = get_net(src_net); 1328 return net; 1329 } 1330 EXPORT_SYMBOL(rtnl_link_get_net); 1331 1332 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[]) 1333 { 1334 if (dev) { 1335 if (tb[IFLA_ADDRESS] && 1336 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len) 1337 return -EINVAL; 1338 1339 if (tb[IFLA_BROADCAST] && 1340 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len) 1341 return -EINVAL; 1342 } 1343 1344 if (tb[IFLA_AF_SPEC]) { 1345 struct nlattr *af; 1346 int rem, err; 1347 1348 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1349 const struct rtnl_af_ops *af_ops; 1350 1351 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1352 return -EAFNOSUPPORT; 1353 1354 if (!af_ops->set_link_af) 1355 return -EOPNOTSUPP; 1356 1357 if (af_ops->validate_link_af) { 1358 err = af_ops->validate_link_af(dev, af); 1359 if (err < 0) 1360 return err; 1361 } 1362 } 1363 } 1364 1365 return 0; 1366 } 1367 1368 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr) 1369 { 1370 int rem, err = -EINVAL; 1371 struct nlattr *vf; 1372 const struct net_device_ops *ops = dev->netdev_ops; 1373 1374 nla_for_each_nested(vf, attr, rem) { 1375 switch (nla_type(vf)) { 1376 case IFLA_VF_MAC: { 1377 struct ifla_vf_mac *ivm; 1378 ivm = nla_data(vf); 1379 err = -EOPNOTSUPP; 1380 if (ops->ndo_set_vf_mac) 1381 err = ops->ndo_set_vf_mac(dev, ivm->vf, 1382 ivm->mac); 1383 break; 1384 } 1385 case IFLA_VF_VLAN: { 1386 struct ifla_vf_vlan *ivv; 1387 ivv = nla_data(vf); 1388 err = -EOPNOTSUPP; 1389 if (ops->ndo_set_vf_vlan) 1390 err = ops->ndo_set_vf_vlan(dev, ivv->vf, 1391 ivv->vlan, 1392 ivv->qos); 1393 break; 1394 } 1395 case IFLA_VF_TX_RATE: { 1396 struct ifla_vf_tx_rate *ivt; 1397 struct ifla_vf_info ivf; 1398 ivt = nla_data(vf); 1399 err = -EOPNOTSUPP; 1400 if (ops->ndo_get_vf_config) 1401 err = ops->ndo_get_vf_config(dev, ivt->vf, 1402 &ivf); 1403 if (err) 1404 break; 1405 err = -EOPNOTSUPP; 1406 if (ops->ndo_set_vf_rate) 1407 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1408 ivf.min_tx_rate, 1409 ivt->rate); 1410 break; 1411 } 1412 case IFLA_VF_RATE: { 1413 struct ifla_vf_rate *ivt; 1414 ivt = nla_data(vf); 1415 err = -EOPNOTSUPP; 1416 if (ops->ndo_set_vf_rate) 1417 err = ops->ndo_set_vf_rate(dev, ivt->vf, 1418 ivt->min_tx_rate, 1419 ivt->max_tx_rate); 1420 break; 1421 } 1422 case IFLA_VF_SPOOFCHK: { 1423 struct ifla_vf_spoofchk *ivs; 1424 ivs = nla_data(vf); 1425 err = -EOPNOTSUPP; 1426 if (ops->ndo_set_vf_spoofchk) 1427 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf, 1428 ivs->setting); 1429 break; 1430 } 1431 case IFLA_VF_LINK_STATE: { 1432 struct ifla_vf_link_state *ivl; 1433 ivl = nla_data(vf); 1434 err = -EOPNOTSUPP; 1435 if (ops->ndo_set_vf_link_state) 1436 err = ops->ndo_set_vf_link_state(dev, ivl->vf, 1437 ivl->link_state); 1438 break; 1439 } 1440 default: 1441 err = -EINVAL; 1442 break; 1443 } 1444 if (err) 1445 break; 1446 } 1447 return err; 1448 } 1449 1450 static int do_set_master(struct net_device *dev, int ifindex) 1451 { 1452 struct net_device *upper_dev = netdev_master_upper_dev_get(dev); 1453 const struct net_device_ops *ops; 1454 int err; 1455 1456 if (upper_dev) { 1457 if (upper_dev->ifindex == ifindex) 1458 return 0; 1459 ops = upper_dev->netdev_ops; 1460 if (ops->ndo_del_slave) { 1461 err = ops->ndo_del_slave(upper_dev, dev); 1462 if (err) 1463 return err; 1464 } else { 1465 return -EOPNOTSUPP; 1466 } 1467 } 1468 1469 if (ifindex) { 1470 upper_dev = __dev_get_by_index(dev_net(dev), ifindex); 1471 if (!upper_dev) 1472 return -EINVAL; 1473 ops = upper_dev->netdev_ops; 1474 if (ops->ndo_add_slave) { 1475 err = ops->ndo_add_slave(upper_dev, dev); 1476 if (err) 1477 return err; 1478 } else { 1479 return -EOPNOTSUPP; 1480 } 1481 } 1482 return 0; 1483 } 1484 1485 #define DO_SETLINK_MODIFIED 0x01 1486 /* notify flag means notify + modified. */ 1487 #define DO_SETLINK_NOTIFY 0x03 1488 static int do_setlink(const struct sk_buff *skb, 1489 struct net_device *dev, struct ifinfomsg *ifm, 1490 struct nlattr **tb, char *ifname, int status) 1491 { 1492 const struct net_device_ops *ops = dev->netdev_ops; 1493 int err; 1494 1495 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) { 1496 struct net *net = rtnl_link_get_net(dev_net(dev), tb); 1497 if (IS_ERR(net)) { 1498 err = PTR_ERR(net); 1499 goto errout; 1500 } 1501 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { 1502 err = -EPERM; 1503 goto errout; 1504 } 1505 err = dev_change_net_namespace(dev, net, ifname); 1506 put_net(net); 1507 if (err) 1508 goto errout; 1509 status |= DO_SETLINK_MODIFIED; 1510 } 1511 1512 if (tb[IFLA_MAP]) { 1513 struct rtnl_link_ifmap *u_map; 1514 struct ifmap k_map; 1515 1516 if (!ops->ndo_set_config) { 1517 err = -EOPNOTSUPP; 1518 goto errout; 1519 } 1520 1521 if (!netif_device_present(dev)) { 1522 err = -ENODEV; 1523 goto errout; 1524 } 1525 1526 u_map = nla_data(tb[IFLA_MAP]); 1527 k_map.mem_start = (unsigned long) u_map->mem_start; 1528 k_map.mem_end = (unsigned long) u_map->mem_end; 1529 k_map.base_addr = (unsigned short) u_map->base_addr; 1530 k_map.irq = (unsigned char) u_map->irq; 1531 k_map.dma = (unsigned char) u_map->dma; 1532 k_map.port = (unsigned char) u_map->port; 1533 1534 err = ops->ndo_set_config(dev, &k_map); 1535 if (err < 0) 1536 goto errout; 1537 1538 status |= DO_SETLINK_NOTIFY; 1539 } 1540 1541 if (tb[IFLA_ADDRESS]) { 1542 struct sockaddr *sa; 1543 int len; 1544 1545 len = sizeof(sa_family_t) + dev->addr_len; 1546 sa = kmalloc(len, GFP_KERNEL); 1547 if (!sa) { 1548 err = -ENOMEM; 1549 goto errout; 1550 } 1551 sa->sa_family = dev->type; 1552 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]), 1553 dev->addr_len); 1554 err = dev_set_mac_address(dev, sa); 1555 kfree(sa); 1556 if (err) 1557 goto errout; 1558 status |= DO_SETLINK_MODIFIED; 1559 } 1560 1561 if (tb[IFLA_MTU]) { 1562 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1563 if (err < 0) 1564 goto errout; 1565 status |= DO_SETLINK_MODIFIED; 1566 } 1567 1568 if (tb[IFLA_GROUP]) { 1569 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 1570 status |= DO_SETLINK_NOTIFY; 1571 } 1572 1573 /* 1574 * Interface selected by interface index but interface 1575 * name provided implies that a name change has been 1576 * requested. 1577 */ 1578 if (ifm->ifi_index > 0 && ifname[0]) { 1579 err = dev_change_name(dev, ifname); 1580 if (err < 0) 1581 goto errout; 1582 status |= DO_SETLINK_MODIFIED; 1583 } 1584 1585 if (tb[IFLA_IFALIAS]) { 1586 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]), 1587 nla_len(tb[IFLA_IFALIAS])); 1588 if (err < 0) 1589 goto errout; 1590 status |= DO_SETLINK_NOTIFY; 1591 } 1592 1593 if (tb[IFLA_BROADCAST]) { 1594 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len); 1595 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev); 1596 } 1597 1598 if (ifm->ifi_flags || ifm->ifi_change) { 1599 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1600 if (err < 0) 1601 goto errout; 1602 } 1603 1604 if (tb[IFLA_MASTER]) { 1605 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER])); 1606 if (err) 1607 goto errout; 1608 status |= DO_SETLINK_MODIFIED; 1609 } 1610 1611 if (tb[IFLA_CARRIER]) { 1612 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER])); 1613 if (err) 1614 goto errout; 1615 status |= DO_SETLINK_MODIFIED; 1616 } 1617 1618 if (tb[IFLA_TXQLEN]) { 1619 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]); 1620 1621 if (dev->tx_queue_len ^ value) 1622 status |= DO_SETLINK_NOTIFY; 1623 1624 dev->tx_queue_len = value; 1625 } 1626 1627 if (tb[IFLA_OPERSTATE]) 1628 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1629 1630 if (tb[IFLA_LINKMODE]) { 1631 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]); 1632 1633 write_lock_bh(&dev_base_lock); 1634 if (dev->link_mode ^ value) 1635 status |= DO_SETLINK_NOTIFY; 1636 dev->link_mode = value; 1637 write_unlock_bh(&dev_base_lock); 1638 } 1639 1640 if (tb[IFLA_VFINFO_LIST]) { 1641 struct nlattr *attr; 1642 int rem; 1643 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) { 1644 if (nla_type(attr) != IFLA_VF_INFO) { 1645 err = -EINVAL; 1646 goto errout; 1647 } 1648 err = do_setvfinfo(dev, attr); 1649 if (err < 0) 1650 goto errout; 1651 status |= DO_SETLINK_NOTIFY; 1652 } 1653 } 1654 err = 0; 1655 1656 if (tb[IFLA_VF_PORTS]) { 1657 struct nlattr *port[IFLA_PORT_MAX+1]; 1658 struct nlattr *attr; 1659 int vf; 1660 int rem; 1661 1662 err = -EOPNOTSUPP; 1663 if (!ops->ndo_set_vf_port) 1664 goto errout; 1665 1666 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) { 1667 if (nla_type(attr) != IFLA_VF_PORT) 1668 continue; 1669 err = nla_parse_nested(port, IFLA_PORT_MAX, 1670 attr, ifla_port_policy); 1671 if (err < 0) 1672 goto errout; 1673 if (!port[IFLA_PORT_VF]) { 1674 err = -EOPNOTSUPP; 1675 goto errout; 1676 } 1677 vf = nla_get_u32(port[IFLA_PORT_VF]); 1678 err = ops->ndo_set_vf_port(dev, vf, port); 1679 if (err < 0) 1680 goto errout; 1681 status |= DO_SETLINK_NOTIFY; 1682 } 1683 } 1684 err = 0; 1685 1686 if (tb[IFLA_PORT_SELF]) { 1687 struct nlattr *port[IFLA_PORT_MAX+1]; 1688 1689 err = nla_parse_nested(port, IFLA_PORT_MAX, 1690 tb[IFLA_PORT_SELF], ifla_port_policy); 1691 if (err < 0) 1692 goto errout; 1693 1694 err = -EOPNOTSUPP; 1695 if (ops->ndo_set_vf_port) 1696 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port); 1697 if (err < 0) 1698 goto errout; 1699 status |= DO_SETLINK_NOTIFY; 1700 } 1701 1702 if (tb[IFLA_AF_SPEC]) { 1703 struct nlattr *af; 1704 int rem; 1705 1706 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) { 1707 const struct rtnl_af_ops *af_ops; 1708 1709 if (!(af_ops = rtnl_af_lookup(nla_type(af)))) 1710 BUG(); 1711 1712 err = af_ops->set_link_af(dev, af); 1713 if (err < 0) 1714 goto errout; 1715 1716 status |= DO_SETLINK_NOTIFY; 1717 } 1718 } 1719 err = 0; 1720 1721 errout: 1722 if (status & DO_SETLINK_MODIFIED) { 1723 if (status & DO_SETLINK_NOTIFY) 1724 netdev_state_change(dev); 1725 1726 if (err < 0) 1727 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", 1728 dev->name); 1729 } 1730 1731 return err; 1732 } 1733 1734 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh) 1735 { 1736 struct net *net = sock_net(skb->sk); 1737 struct ifinfomsg *ifm; 1738 struct net_device *dev; 1739 int err; 1740 struct nlattr *tb[IFLA_MAX+1]; 1741 char ifname[IFNAMSIZ]; 1742 1743 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1744 if (err < 0) 1745 goto errout; 1746 1747 if (tb[IFLA_IFNAME]) 1748 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1749 else 1750 ifname[0] = '\0'; 1751 1752 err = -EINVAL; 1753 ifm = nlmsg_data(nlh); 1754 if (ifm->ifi_index > 0) 1755 dev = __dev_get_by_index(net, ifm->ifi_index); 1756 else if (tb[IFLA_IFNAME]) 1757 dev = __dev_get_by_name(net, ifname); 1758 else 1759 goto errout; 1760 1761 if (dev == NULL) { 1762 err = -ENODEV; 1763 goto errout; 1764 } 1765 1766 err = validate_linkmsg(dev, tb); 1767 if (err < 0) 1768 goto errout; 1769 1770 err = do_setlink(skb, dev, ifm, tb, ifname, 0); 1771 errout: 1772 return err; 1773 } 1774 1775 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh) 1776 { 1777 struct net *net = sock_net(skb->sk); 1778 const struct rtnl_link_ops *ops; 1779 struct net_device *dev; 1780 struct ifinfomsg *ifm; 1781 char ifname[IFNAMSIZ]; 1782 struct nlattr *tb[IFLA_MAX+1]; 1783 int err; 1784 LIST_HEAD(list_kill); 1785 1786 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1787 if (err < 0) 1788 return err; 1789 1790 if (tb[IFLA_IFNAME]) 1791 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1792 1793 ifm = nlmsg_data(nlh); 1794 if (ifm->ifi_index > 0) 1795 dev = __dev_get_by_index(net, ifm->ifi_index); 1796 else if (tb[IFLA_IFNAME]) 1797 dev = __dev_get_by_name(net, ifname); 1798 else 1799 return -EINVAL; 1800 1801 if (!dev) 1802 return -ENODEV; 1803 1804 ops = dev->rtnl_link_ops; 1805 if (!ops || !ops->dellink) 1806 return -EOPNOTSUPP; 1807 1808 ops->dellink(dev, &list_kill); 1809 unregister_netdevice_many(&list_kill); 1810 return 0; 1811 } 1812 1813 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm) 1814 { 1815 unsigned int old_flags; 1816 int err; 1817 1818 old_flags = dev->flags; 1819 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) { 1820 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm)); 1821 if (err < 0) 1822 return err; 1823 } 1824 1825 dev->rtnl_link_state = RTNL_LINK_INITIALIZED; 1826 1827 __dev_notify_flags(dev, old_flags, ~0U); 1828 return 0; 1829 } 1830 EXPORT_SYMBOL(rtnl_configure_link); 1831 1832 struct net_device *rtnl_create_link(struct net *net, 1833 char *ifname, unsigned char name_assign_type, 1834 const struct rtnl_link_ops *ops, struct nlattr *tb[]) 1835 { 1836 int err; 1837 struct net_device *dev; 1838 unsigned int num_tx_queues = 1; 1839 unsigned int num_rx_queues = 1; 1840 1841 if (tb[IFLA_NUM_TX_QUEUES]) 1842 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]); 1843 else if (ops->get_num_tx_queues) 1844 num_tx_queues = ops->get_num_tx_queues(); 1845 1846 if (tb[IFLA_NUM_RX_QUEUES]) 1847 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]); 1848 else if (ops->get_num_rx_queues) 1849 num_rx_queues = ops->get_num_rx_queues(); 1850 1851 err = -ENOMEM; 1852 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type, 1853 ops->setup, num_tx_queues, num_rx_queues); 1854 if (!dev) 1855 goto err; 1856 1857 dev_net_set(dev, net); 1858 dev->rtnl_link_ops = ops; 1859 dev->rtnl_link_state = RTNL_LINK_INITIALIZING; 1860 1861 if (tb[IFLA_MTU]) 1862 dev->mtu = nla_get_u32(tb[IFLA_MTU]); 1863 if (tb[IFLA_ADDRESS]) { 1864 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]), 1865 nla_len(tb[IFLA_ADDRESS])); 1866 dev->addr_assign_type = NET_ADDR_SET; 1867 } 1868 if (tb[IFLA_BROADCAST]) 1869 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]), 1870 nla_len(tb[IFLA_BROADCAST])); 1871 if (tb[IFLA_TXQLEN]) 1872 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1873 if (tb[IFLA_OPERSTATE]) 1874 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1875 if (tb[IFLA_LINKMODE]) 1876 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]); 1877 if (tb[IFLA_GROUP]) 1878 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP])); 1879 1880 return dev; 1881 1882 err: 1883 return ERR_PTR(err); 1884 } 1885 EXPORT_SYMBOL(rtnl_create_link); 1886 1887 static int rtnl_group_changelink(const struct sk_buff *skb, 1888 struct net *net, int group, 1889 struct ifinfomsg *ifm, 1890 struct nlattr **tb) 1891 { 1892 struct net_device *dev; 1893 int err; 1894 1895 for_each_netdev(net, dev) { 1896 if (dev->group == group) { 1897 err = do_setlink(skb, dev, ifm, tb, NULL, 0); 1898 if (err < 0) 1899 return err; 1900 } 1901 } 1902 1903 return 0; 1904 } 1905 1906 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh) 1907 { 1908 struct net *net = sock_net(skb->sk); 1909 const struct rtnl_link_ops *ops; 1910 const struct rtnl_link_ops *m_ops = NULL; 1911 struct net_device *dev; 1912 struct net_device *master_dev = NULL; 1913 struct ifinfomsg *ifm; 1914 char kind[MODULE_NAME_LEN]; 1915 char ifname[IFNAMSIZ]; 1916 struct nlattr *tb[IFLA_MAX+1]; 1917 struct nlattr *linkinfo[IFLA_INFO_MAX+1]; 1918 unsigned char name_assign_type = NET_NAME_USER; 1919 int err; 1920 1921 #ifdef CONFIG_MODULES 1922 replay: 1923 #endif 1924 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 1925 if (err < 0) 1926 return err; 1927 1928 if (tb[IFLA_IFNAME]) 1929 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 1930 else 1931 ifname[0] = '\0'; 1932 1933 ifm = nlmsg_data(nlh); 1934 if (ifm->ifi_index > 0) 1935 dev = __dev_get_by_index(net, ifm->ifi_index); 1936 else { 1937 if (ifname[0]) 1938 dev = __dev_get_by_name(net, ifname); 1939 else 1940 dev = NULL; 1941 } 1942 1943 if (dev) { 1944 master_dev = netdev_master_upper_dev_get(dev); 1945 if (master_dev) 1946 m_ops = master_dev->rtnl_link_ops; 1947 } 1948 1949 err = validate_linkmsg(dev, tb); 1950 if (err < 0) 1951 return err; 1952 1953 if (tb[IFLA_LINKINFO]) { 1954 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX, 1955 tb[IFLA_LINKINFO], ifla_info_policy); 1956 if (err < 0) 1957 return err; 1958 } else 1959 memset(linkinfo, 0, sizeof(linkinfo)); 1960 1961 if (linkinfo[IFLA_INFO_KIND]) { 1962 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind)); 1963 ops = rtnl_link_ops_get(kind); 1964 } else { 1965 kind[0] = '\0'; 1966 ops = NULL; 1967 } 1968 1969 if (1) { 1970 struct nlattr *attr[ops ? ops->maxtype + 1 : 0]; 1971 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0]; 1972 struct nlattr **data = NULL; 1973 struct nlattr **slave_data = NULL; 1974 struct net *dest_net; 1975 1976 if (ops) { 1977 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) { 1978 err = nla_parse_nested(attr, ops->maxtype, 1979 linkinfo[IFLA_INFO_DATA], 1980 ops->policy); 1981 if (err < 0) 1982 return err; 1983 data = attr; 1984 } 1985 if (ops->validate) { 1986 err = ops->validate(tb, data); 1987 if (err < 0) 1988 return err; 1989 } 1990 } 1991 1992 if (m_ops) { 1993 if (m_ops->slave_maxtype && 1994 linkinfo[IFLA_INFO_SLAVE_DATA]) { 1995 err = nla_parse_nested(slave_attr, 1996 m_ops->slave_maxtype, 1997 linkinfo[IFLA_INFO_SLAVE_DATA], 1998 m_ops->slave_policy); 1999 if (err < 0) 2000 return err; 2001 slave_data = slave_attr; 2002 } 2003 if (m_ops->slave_validate) { 2004 err = m_ops->slave_validate(tb, slave_data); 2005 if (err < 0) 2006 return err; 2007 } 2008 } 2009 2010 if (dev) { 2011 int status = 0; 2012 2013 if (nlh->nlmsg_flags & NLM_F_EXCL) 2014 return -EEXIST; 2015 if (nlh->nlmsg_flags & NLM_F_REPLACE) 2016 return -EOPNOTSUPP; 2017 2018 if (linkinfo[IFLA_INFO_DATA]) { 2019 if (!ops || ops != dev->rtnl_link_ops || 2020 !ops->changelink) 2021 return -EOPNOTSUPP; 2022 2023 err = ops->changelink(dev, tb, data); 2024 if (err < 0) 2025 return err; 2026 status |= DO_SETLINK_NOTIFY; 2027 } 2028 2029 if (linkinfo[IFLA_INFO_SLAVE_DATA]) { 2030 if (!m_ops || !m_ops->slave_changelink) 2031 return -EOPNOTSUPP; 2032 2033 err = m_ops->slave_changelink(master_dev, dev, 2034 tb, slave_data); 2035 if (err < 0) 2036 return err; 2037 status |= DO_SETLINK_NOTIFY; 2038 } 2039 2040 return do_setlink(skb, dev, ifm, tb, ifname, status); 2041 } 2042 2043 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) { 2044 if (ifm->ifi_index == 0 && tb[IFLA_GROUP]) 2045 return rtnl_group_changelink(skb, net, 2046 nla_get_u32(tb[IFLA_GROUP]), 2047 ifm, tb); 2048 return -ENODEV; 2049 } 2050 2051 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO]) 2052 return -EOPNOTSUPP; 2053 2054 if (!ops) { 2055 #ifdef CONFIG_MODULES 2056 if (kind[0]) { 2057 __rtnl_unlock(); 2058 request_module("rtnl-link-%s", kind); 2059 rtnl_lock(); 2060 ops = rtnl_link_ops_get(kind); 2061 if (ops) 2062 goto replay; 2063 } 2064 #endif 2065 return -EOPNOTSUPP; 2066 } 2067 2068 if (!ops->setup) 2069 return -EOPNOTSUPP; 2070 2071 if (!ifname[0]) { 2072 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind); 2073 name_assign_type = NET_NAME_ENUM; 2074 } 2075 2076 dest_net = rtnl_link_get_net(net, tb); 2077 if (IS_ERR(dest_net)) 2078 return PTR_ERR(dest_net); 2079 2080 dev = rtnl_create_link(dest_net, ifname, name_assign_type, ops, tb); 2081 if (IS_ERR(dev)) { 2082 err = PTR_ERR(dev); 2083 goto out; 2084 } 2085 2086 dev->ifindex = ifm->ifi_index; 2087 2088 if (ops->newlink) { 2089 err = ops->newlink(net, dev, tb, data); 2090 /* Drivers should call free_netdev() in ->destructor 2091 * and unregister it on failure after registration 2092 * so that device could be finally freed in rtnl_unlock. 2093 */ 2094 if (err < 0) { 2095 /* If device is not registered at all, free it now */ 2096 if (dev->reg_state == NETREG_UNINITIALIZED) 2097 free_netdev(dev); 2098 goto out; 2099 } 2100 } else { 2101 err = register_netdevice(dev); 2102 if (err < 0) { 2103 free_netdev(dev); 2104 goto out; 2105 } 2106 } 2107 err = rtnl_configure_link(dev, ifm); 2108 if (err < 0) 2109 unregister_netdevice(dev); 2110 out: 2111 put_net(dest_net); 2112 return err; 2113 } 2114 } 2115 2116 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh) 2117 { 2118 struct net *net = sock_net(skb->sk); 2119 struct ifinfomsg *ifm; 2120 char ifname[IFNAMSIZ]; 2121 struct nlattr *tb[IFLA_MAX+1]; 2122 struct net_device *dev = NULL; 2123 struct sk_buff *nskb; 2124 int err; 2125 u32 ext_filter_mask = 0; 2126 2127 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy); 2128 if (err < 0) 2129 return err; 2130 2131 if (tb[IFLA_IFNAME]) 2132 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ); 2133 2134 if (tb[IFLA_EXT_MASK]) 2135 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2136 2137 ifm = nlmsg_data(nlh); 2138 if (ifm->ifi_index > 0) 2139 dev = __dev_get_by_index(net, ifm->ifi_index); 2140 else if (tb[IFLA_IFNAME]) 2141 dev = __dev_get_by_name(net, ifname); 2142 else 2143 return -EINVAL; 2144 2145 if (dev == NULL) 2146 return -ENODEV; 2147 2148 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL); 2149 if (nskb == NULL) 2150 return -ENOBUFS; 2151 2152 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid, 2153 nlh->nlmsg_seq, 0, 0, ext_filter_mask); 2154 if (err < 0) { 2155 /* -EMSGSIZE implies BUG in if_nlmsg_size */ 2156 WARN_ON(err == -EMSGSIZE); 2157 kfree_skb(nskb); 2158 } else 2159 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid); 2160 2161 return err; 2162 } 2163 2164 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh) 2165 { 2166 struct net *net = sock_net(skb->sk); 2167 struct net_device *dev; 2168 struct nlattr *tb[IFLA_MAX+1]; 2169 u32 ext_filter_mask = 0; 2170 u16 min_ifinfo_dump_size = 0; 2171 int hdrlen; 2172 2173 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */ 2174 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ? 2175 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg); 2176 2177 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) { 2178 if (tb[IFLA_EXT_MASK]) 2179 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]); 2180 } 2181 2182 if (!ext_filter_mask) 2183 return NLMSG_GOODSIZE; 2184 /* 2185 * traverse the list of net devices and compute the minimum 2186 * buffer size based upon the filter mask. 2187 */ 2188 list_for_each_entry(dev, &net->dev_base_head, dev_list) { 2189 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size, 2190 if_nlmsg_size(dev, 2191 ext_filter_mask)); 2192 } 2193 2194 return min_ifinfo_dump_size; 2195 } 2196 2197 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb) 2198 { 2199 int idx; 2200 int s_idx = cb->family; 2201 2202 if (s_idx == 0) 2203 s_idx = 1; 2204 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) { 2205 int type = cb->nlh->nlmsg_type-RTM_BASE; 2206 if (idx < s_idx || idx == PF_PACKET) 2207 continue; 2208 if (rtnl_msg_handlers[idx] == NULL || 2209 rtnl_msg_handlers[idx][type].dumpit == NULL) 2210 continue; 2211 if (idx > s_idx) { 2212 memset(&cb->args[0], 0, sizeof(cb->args)); 2213 cb->prev_seq = 0; 2214 cb->seq = 0; 2215 } 2216 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb)) 2217 break; 2218 } 2219 cb->family = idx; 2220 2221 return skb->len; 2222 } 2223 2224 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change, 2225 gfp_t flags) 2226 { 2227 struct net *net = dev_net(dev); 2228 struct sk_buff *skb; 2229 int err = -ENOBUFS; 2230 size_t if_info_size; 2231 2232 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags); 2233 if (skb == NULL) 2234 goto errout; 2235 2236 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0); 2237 if (err < 0) { 2238 /* -EMSGSIZE implies BUG in if_nlmsg_size() */ 2239 WARN_ON(err == -EMSGSIZE); 2240 kfree_skb(skb); 2241 goto errout; 2242 } 2243 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags); 2244 return; 2245 errout: 2246 if (err < 0) 2247 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2248 } 2249 EXPORT_SYMBOL(rtmsg_ifinfo); 2250 2251 static int nlmsg_populate_fdb_fill(struct sk_buff *skb, 2252 struct net_device *dev, 2253 u8 *addr, u32 pid, u32 seq, 2254 int type, unsigned int flags, 2255 int nlflags) 2256 { 2257 struct nlmsghdr *nlh; 2258 struct ndmsg *ndm; 2259 2260 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags); 2261 if (!nlh) 2262 return -EMSGSIZE; 2263 2264 ndm = nlmsg_data(nlh); 2265 ndm->ndm_family = AF_BRIDGE; 2266 ndm->ndm_pad1 = 0; 2267 ndm->ndm_pad2 = 0; 2268 ndm->ndm_flags = flags; 2269 ndm->ndm_type = 0; 2270 ndm->ndm_ifindex = dev->ifindex; 2271 ndm->ndm_state = NUD_PERMANENT; 2272 2273 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr)) 2274 goto nla_put_failure; 2275 2276 return nlmsg_end(skb, nlh); 2277 2278 nla_put_failure: 2279 nlmsg_cancel(skb, nlh); 2280 return -EMSGSIZE; 2281 } 2282 2283 static inline size_t rtnl_fdb_nlmsg_size(void) 2284 { 2285 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN); 2286 } 2287 2288 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type) 2289 { 2290 struct net *net = dev_net(dev); 2291 struct sk_buff *skb; 2292 int err = -ENOBUFS; 2293 2294 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC); 2295 if (!skb) 2296 goto errout; 2297 2298 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0); 2299 if (err < 0) { 2300 kfree_skb(skb); 2301 goto errout; 2302 } 2303 2304 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); 2305 return; 2306 errout: 2307 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); 2308 } 2309 2310 /** 2311 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry 2312 */ 2313 int ndo_dflt_fdb_add(struct ndmsg *ndm, 2314 struct nlattr *tb[], 2315 struct net_device *dev, 2316 const unsigned char *addr, u16 vid, 2317 u16 flags) 2318 { 2319 int err = -EINVAL; 2320 2321 /* If aging addresses are supported device will need to 2322 * implement its own handler for this. 2323 */ 2324 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) { 2325 pr_info("%s: FDB only supports static addresses\n", dev->name); 2326 return err; 2327 } 2328 2329 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 2330 err = dev_uc_add_excl(dev, addr); 2331 else if (is_multicast_ether_addr(addr)) 2332 err = dev_mc_add_excl(dev, addr); 2333 2334 /* Only return duplicate errors if NLM_F_EXCL is set */ 2335 if (err == -EEXIST && !(flags & NLM_F_EXCL)) 2336 err = 0; 2337 2338 return err; 2339 } 2340 EXPORT_SYMBOL(ndo_dflt_fdb_add); 2341 2342 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid) 2343 { 2344 u16 vid = 0; 2345 2346 if (vlan_attr) { 2347 if (nla_len(vlan_attr) != sizeof(u16)) { 2348 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n"); 2349 return -EINVAL; 2350 } 2351 2352 vid = nla_get_u16(vlan_attr); 2353 2354 if (!vid || vid >= VLAN_VID_MASK) { 2355 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n", 2356 vid); 2357 return -EINVAL; 2358 } 2359 } 2360 *p_vid = vid; 2361 return 0; 2362 } 2363 2364 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh) 2365 { 2366 struct net *net = sock_net(skb->sk); 2367 struct ndmsg *ndm; 2368 struct nlattr *tb[NDA_MAX+1]; 2369 struct net_device *dev; 2370 u8 *addr; 2371 u16 vid; 2372 int err; 2373 2374 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL); 2375 if (err < 0) 2376 return err; 2377 2378 ndm = nlmsg_data(nlh); 2379 if (ndm->ndm_ifindex == 0) { 2380 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n"); 2381 return -EINVAL; 2382 } 2383 2384 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 2385 if (dev == NULL) { 2386 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n"); 2387 return -ENODEV; 2388 } 2389 2390 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 2391 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n"); 2392 return -EINVAL; 2393 } 2394 2395 addr = nla_data(tb[NDA_LLADDR]); 2396 2397 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 2398 if (err) 2399 return err; 2400 2401 err = -EOPNOTSUPP; 2402 2403 /* Support fdb on master device the net/bridge default case */ 2404 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 2405 (dev->priv_flags & IFF_BRIDGE_PORT)) { 2406 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2407 const struct net_device_ops *ops = br_dev->netdev_ops; 2408 2409 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid, 2410 nlh->nlmsg_flags); 2411 if (err) 2412 goto out; 2413 else 2414 ndm->ndm_flags &= ~NTF_MASTER; 2415 } 2416 2417 /* Embedded bridge, macvlan, and any other device support */ 2418 if ((ndm->ndm_flags & NTF_SELF)) { 2419 if (dev->netdev_ops->ndo_fdb_add) 2420 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr, 2421 vid, 2422 nlh->nlmsg_flags); 2423 else 2424 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, 2425 nlh->nlmsg_flags); 2426 2427 if (!err) { 2428 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH); 2429 ndm->ndm_flags &= ~NTF_SELF; 2430 } 2431 } 2432 out: 2433 return err; 2434 } 2435 2436 /** 2437 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry 2438 */ 2439 int ndo_dflt_fdb_del(struct ndmsg *ndm, 2440 struct nlattr *tb[], 2441 struct net_device *dev, 2442 const unsigned char *addr, u16 vid) 2443 { 2444 int err = -EINVAL; 2445 2446 /* If aging addresses are supported device will need to 2447 * implement its own handler for this. 2448 */ 2449 if (!(ndm->ndm_state & NUD_PERMANENT)) { 2450 pr_info("%s: FDB only supports static addresses\n", dev->name); 2451 return err; 2452 } 2453 2454 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) 2455 err = dev_uc_del(dev, addr); 2456 else if (is_multicast_ether_addr(addr)) 2457 err = dev_mc_del(dev, addr); 2458 2459 return err; 2460 } 2461 EXPORT_SYMBOL(ndo_dflt_fdb_del); 2462 2463 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh) 2464 { 2465 struct net *net = sock_net(skb->sk); 2466 struct ndmsg *ndm; 2467 struct nlattr *tb[NDA_MAX+1]; 2468 struct net_device *dev; 2469 int err = -EINVAL; 2470 __u8 *addr; 2471 u16 vid; 2472 2473 if (!netlink_capable(skb, CAP_NET_ADMIN)) 2474 return -EPERM; 2475 2476 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL); 2477 if (err < 0) 2478 return err; 2479 2480 ndm = nlmsg_data(nlh); 2481 if (ndm->ndm_ifindex == 0) { 2482 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n"); 2483 return -EINVAL; 2484 } 2485 2486 dev = __dev_get_by_index(net, ndm->ndm_ifindex); 2487 if (dev == NULL) { 2488 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n"); 2489 return -ENODEV; 2490 } 2491 2492 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) { 2493 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n"); 2494 return -EINVAL; 2495 } 2496 2497 addr = nla_data(tb[NDA_LLADDR]); 2498 2499 err = fdb_vid_parse(tb[NDA_VLAN], &vid); 2500 if (err) 2501 return err; 2502 2503 err = -EOPNOTSUPP; 2504 2505 /* Support fdb on master device the net/bridge default case */ 2506 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) && 2507 (dev->priv_flags & IFF_BRIDGE_PORT)) { 2508 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2509 const struct net_device_ops *ops = br_dev->netdev_ops; 2510 2511 if (ops->ndo_fdb_del) 2512 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid); 2513 2514 if (err) 2515 goto out; 2516 else 2517 ndm->ndm_flags &= ~NTF_MASTER; 2518 } 2519 2520 /* Embedded bridge, macvlan, and any other device support */ 2521 if (ndm->ndm_flags & NTF_SELF) { 2522 if (dev->netdev_ops->ndo_fdb_del) 2523 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr, 2524 vid); 2525 else 2526 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid); 2527 2528 if (!err) { 2529 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH); 2530 ndm->ndm_flags &= ~NTF_SELF; 2531 } 2532 } 2533 out: 2534 return err; 2535 } 2536 2537 static int nlmsg_populate_fdb(struct sk_buff *skb, 2538 struct netlink_callback *cb, 2539 struct net_device *dev, 2540 int *idx, 2541 struct netdev_hw_addr_list *list) 2542 { 2543 struct netdev_hw_addr *ha; 2544 int err; 2545 u32 portid, seq; 2546 2547 portid = NETLINK_CB(cb->skb).portid; 2548 seq = cb->nlh->nlmsg_seq; 2549 2550 list_for_each_entry(ha, &list->list, list) { 2551 if (*idx < cb->args[0]) 2552 goto skip; 2553 2554 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 2555 portid, seq, 2556 RTM_NEWNEIGH, NTF_SELF, 2557 NLM_F_MULTI); 2558 if (err < 0) 2559 return err; 2560 skip: 2561 *idx += 1; 2562 } 2563 return 0; 2564 } 2565 2566 /** 2567 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table. 2568 * @nlh: netlink message header 2569 * @dev: netdevice 2570 * 2571 * Default netdevice operation to dump the existing unicast address list. 2572 * Returns number of addresses from list put in skb. 2573 */ 2574 int ndo_dflt_fdb_dump(struct sk_buff *skb, 2575 struct netlink_callback *cb, 2576 struct net_device *dev, 2577 struct net_device *filter_dev, 2578 int idx) 2579 { 2580 int err; 2581 2582 netif_addr_lock_bh(dev); 2583 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc); 2584 if (err) 2585 goto out; 2586 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc); 2587 out: 2588 netif_addr_unlock_bh(dev); 2589 return idx; 2590 } 2591 EXPORT_SYMBOL(ndo_dflt_fdb_dump); 2592 2593 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb) 2594 { 2595 struct net_device *dev; 2596 struct nlattr *tb[IFLA_MAX+1]; 2597 struct net_device *bdev = NULL; 2598 struct net_device *br_dev = NULL; 2599 const struct net_device_ops *ops = NULL; 2600 const struct net_device_ops *cops = NULL; 2601 struct ifinfomsg *ifm = nlmsg_data(cb->nlh); 2602 struct net *net = sock_net(skb->sk); 2603 int brport_idx = 0; 2604 int br_idx = 0; 2605 int idx = 0; 2606 2607 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX, 2608 ifla_policy) == 0) { 2609 if (tb[IFLA_MASTER]) 2610 br_idx = nla_get_u32(tb[IFLA_MASTER]); 2611 } 2612 2613 brport_idx = ifm->ifi_index; 2614 2615 if (br_idx) { 2616 br_dev = __dev_get_by_index(net, br_idx); 2617 if (!br_dev) 2618 return -ENODEV; 2619 2620 ops = br_dev->netdev_ops; 2621 bdev = br_dev; 2622 } 2623 2624 for_each_netdev(net, dev) { 2625 if (brport_idx && (dev->ifindex != brport_idx)) 2626 continue; 2627 2628 if (!br_idx) { /* user did not specify a specific bridge */ 2629 if (dev->priv_flags & IFF_BRIDGE_PORT) { 2630 br_dev = netdev_master_upper_dev_get(dev); 2631 cops = br_dev->netdev_ops; 2632 } 2633 2634 bdev = dev; 2635 } else { 2636 if (dev != br_dev && 2637 !(dev->priv_flags & IFF_BRIDGE_PORT)) 2638 continue; 2639 2640 if (br_dev != netdev_master_upper_dev_get(dev) && 2641 !(dev->priv_flags & IFF_EBRIDGE)) 2642 continue; 2643 2644 bdev = br_dev; 2645 cops = ops; 2646 } 2647 2648 if (dev->priv_flags & IFF_BRIDGE_PORT) { 2649 if (cops && cops->ndo_fdb_dump) 2650 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev, 2651 idx); 2652 } 2653 2654 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx); 2655 if (dev->netdev_ops->ndo_fdb_dump) 2656 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev, 2657 idx); 2658 2659 cops = NULL; 2660 } 2661 2662 cb->args[0] = idx; 2663 return skb->len; 2664 } 2665 2666 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, 2667 struct net_device *dev, u16 mode) 2668 { 2669 struct nlmsghdr *nlh; 2670 struct ifinfomsg *ifm; 2671 struct nlattr *br_afspec; 2672 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN; 2673 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2674 2675 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI); 2676 if (nlh == NULL) 2677 return -EMSGSIZE; 2678 2679 ifm = nlmsg_data(nlh); 2680 ifm->ifi_family = AF_BRIDGE; 2681 ifm->__ifi_pad = 0; 2682 ifm->ifi_type = dev->type; 2683 ifm->ifi_index = dev->ifindex; 2684 ifm->ifi_flags = dev_get_flags(dev); 2685 ifm->ifi_change = 0; 2686 2687 2688 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 2689 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 2690 nla_put_u8(skb, IFLA_OPERSTATE, operstate) || 2691 (br_dev && 2692 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) || 2693 (dev->addr_len && 2694 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 2695 (dev->ifindex != dev->iflink && 2696 nla_put_u32(skb, IFLA_LINK, dev->iflink))) 2697 goto nla_put_failure; 2698 2699 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC); 2700 if (!br_afspec) 2701 goto nla_put_failure; 2702 2703 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) || 2704 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) { 2705 nla_nest_cancel(skb, br_afspec); 2706 goto nla_put_failure; 2707 } 2708 nla_nest_end(skb, br_afspec); 2709 2710 return nlmsg_end(skb, nlh); 2711 nla_put_failure: 2712 nlmsg_cancel(skb, nlh); 2713 return -EMSGSIZE; 2714 } 2715 EXPORT_SYMBOL(ndo_dflt_bridge_getlink); 2716 2717 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb) 2718 { 2719 struct net *net = sock_net(skb->sk); 2720 struct net_device *dev; 2721 int idx = 0; 2722 u32 portid = NETLINK_CB(cb->skb).portid; 2723 u32 seq = cb->nlh->nlmsg_seq; 2724 u32 filter_mask = 0; 2725 2726 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) { 2727 struct nlattr *extfilt; 2728 2729 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg), 2730 IFLA_EXT_MASK); 2731 if (extfilt) { 2732 if (nla_len(extfilt) < sizeof(filter_mask)) 2733 return -EINVAL; 2734 2735 filter_mask = nla_get_u32(extfilt); 2736 } 2737 } 2738 2739 rcu_read_lock(); 2740 for_each_netdev_rcu(net, dev) { 2741 const struct net_device_ops *ops = dev->netdev_ops; 2742 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2743 2744 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 2745 if (idx >= cb->args[0] && 2746 br_dev->netdev_ops->ndo_bridge_getlink( 2747 skb, portid, seq, dev, filter_mask) < 0) 2748 break; 2749 idx++; 2750 } 2751 2752 if (ops->ndo_bridge_getlink) { 2753 if (idx >= cb->args[0] && 2754 ops->ndo_bridge_getlink(skb, portid, seq, dev, 2755 filter_mask) < 0) 2756 break; 2757 idx++; 2758 } 2759 } 2760 rcu_read_unlock(); 2761 cb->args[0] = idx; 2762 2763 return skb->len; 2764 } 2765 2766 static inline size_t bridge_nlmsg_size(void) 2767 { 2768 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 2769 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 2770 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 2771 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */ 2772 + nla_total_size(sizeof(u32)) /* IFLA_MTU */ 2773 + nla_total_size(sizeof(u32)) /* IFLA_LINK */ 2774 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */ 2775 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */ 2776 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */ 2777 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */ 2778 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */ 2779 } 2780 2781 static int rtnl_bridge_notify(struct net_device *dev, u16 flags) 2782 { 2783 struct net *net = dev_net(dev); 2784 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2785 struct sk_buff *skb; 2786 int err = -EOPNOTSUPP; 2787 2788 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC); 2789 if (!skb) { 2790 err = -ENOMEM; 2791 goto errout; 2792 } 2793 2794 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) && 2795 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) { 2796 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0); 2797 if (err < 0) 2798 goto errout; 2799 } 2800 2801 if ((flags & BRIDGE_FLAGS_SELF) && 2802 dev->netdev_ops->ndo_bridge_getlink) { 2803 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0); 2804 if (err < 0) 2805 goto errout; 2806 } 2807 2808 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC); 2809 return 0; 2810 errout: 2811 WARN_ON(err == -EMSGSIZE); 2812 kfree_skb(skb); 2813 rtnl_set_sk_err(net, RTNLGRP_LINK, err); 2814 return err; 2815 } 2816 2817 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh) 2818 { 2819 struct net *net = sock_net(skb->sk); 2820 struct ifinfomsg *ifm; 2821 struct net_device *dev; 2822 struct nlattr *br_spec, *attr = NULL; 2823 int rem, err = -EOPNOTSUPP; 2824 u16 oflags, flags = 0; 2825 bool have_flags = false; 2826 2827 if (nlmsg_len(nlh) < sizeof(*ifm)) 2828 return -EINVAL; 2829 2830 ifm = nlmsg_data(nlh); 2831 if (ifm->ifi_family != AF_BRIDGE) 2832 return -EPFNOSUPPORT; 2833 2834 dev = __dev_get_by_index(net, ifm->ifi_index); 2835 if (!dev) { 2836 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 2837 return -ENODEV; 2838 } 2839 2840 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 2841 if (br_spec) { 2842 nla_for_each_nested(attr, br_spec, rem) { 2843 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 2844 if (nla_len(attr) < sizeof(flags)) 2845 return -EINVAL; 2846 2847 have_flags = true; 2848 flags = nla_get_u16(attr); 2849 break; 2850 } 2851 } 2852 } 2853 2854 oflags = flags; 2855 2856 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 2857 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2858 2859 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) { 2860 err = -EOPNOTSUPP; 2861 goto out; 2862 } 2863 2864 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh); 2865 if (err) 2866 goto out; 2867 2868 flags &= ~BRIDGE_FLAGS_MASTER; 2869 } 2870 2871 if ((flags & BRIDGE_FLAGS_SELF)) { 2872 if (!dev->netdev_ops->ndo_bridge_setlink) 2873 err = -EOPNOTSUPP; 2874 else 2875 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh); 2876 2877 if (!err) 2878 flags &= ~BRIDGE_FLAGS_SELF; 2879 } 2880 2881 if (have_flags) 2882 memcpy(nla_data(attr), &flags, sizeof(flags)); 2883 /* Generate event to notify upper layer of bridge change */ 2884 if (!err) 2885 err = rtnl_bridge_notify(dev, oflags); 2886 out: 2887 return err; 2888 } 2889 2890 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh) 2891 { 2892 struct net *net = sock_net(skb->sk); 2893 struct ifinfomsg *ifm; 2894 struct net_device *dev; 2895 struct nlattr *br_spec, *attr = NULL; 2896 int rem, err = -EOPNOTSUPP; 2897 u16 oflags, flags = 0; 2898 bool have_flags = false; 2899 2900 if (nlmsg_len(nlh) < sizeof(*ifm)) 2901 return -EINVAL; 2902 2903 ifm = nlmsg_data(nlh); 2904 if (ifm->ifi_family != AF_BRIDGE) 2905 return -EPFNOSUPPORT; 2906 2907 dev = __dev_get_by_index(net, ifm->ifi_index); 2908 if (!dev) { 2909 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n"); 2910 return -ENODEV; 2911 } 2912 2913 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC); 2914 if (br_spec) { 2915 nla_for_each_nested(attr, br_spec, rem) { 2916 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) { 2917 if (nla_len(attr) < sizeof(flags)) 2918 return -EINVAL; 2919 2920 have_flags = true; 2921 flags = nla_get_u16(attr); 2922 break; 2923 } 2924 } 2925 } 2926 2927 oflags = flags; 2928 2929 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) { 2930 struct net_device *br_dev = netdev_master_upper_dev_get(dev); 2931 2932 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) { 2933 err = -EOPNOTSUPP; 2934 goto out; 2935 } 2936 2937 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh); 2938 if (err) 2939 goto out; 2940 2941 flags &= ~BRIDGE_FLAGS_MASTER; 2942 } 2943 2944 if ((flags & BRIDGE_FLAGS_SELF)) { 2945 if (!dev->netdev_ops->ndo_bridge_dellink) 2946 err = -EOPNOTSUPP; 2947 else 2948 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh); 2949 2950 if (!err) 2951 flags &= ~BRIDGE_FLAGS_SELF; 2952 } 2953 2954 if (have_flags) 2955 memcpy(nla_data(attr), &flags, sizeof(flags)); 2956 /* Generate event to notify upper layer of bridge change */ 2957 if (!err) 2958 err = rtnl_bridge_notify(dev, oflags); 2959 out: 2960 return err; 2961 } 2962 2963 /* Process one rtnetlink message. */ 2964 2965 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) 2966 { 2967 struct net *net = sock_net(skb->sk); 2968 rtnl_doit_func doit; 2969 int sz_idx, kind; 2970 int family; 2971 int type; 2972 int err; 2973 2974 type = nlh->nlmsg_type; 2975 if (type > RTM_MAX) 2976 return -EOPNOTSUPP; 2977 2978 type -= RTM_BASE; 2979 2980 /* All the messages must have at least 1 byte length */ 2981 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg)) 2982 return 0; 2983 2984 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family; 2985 sz_idx = type>>2; 2986 kind = type&3; 2987 2988 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN)) 2989 return -EPERM; 2990 2991 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) { 2992 struct sock *rtnl; 2993 rtnl_dumpit_func dumpit; 2994 rtnl_calcit_func calcit; 2995 u16 min_dump_alloc = 0; 2996 2997 dumpit = rtnl_get_dumpit(family, type); 2998 if (dumpit == NULL) 2999 return -EOPNOTSUPP; 3000 calcit = rtnl_get_calcit(family, type); 3001 if (calcit) 3002 min_dump_alloc = calcit(skb, nlh); 3003 3004 __rtnl_unlock(); 3005 rtnl = net->rtnl; 3006 { 3007 struct netlink_dump_control c = { 3008 .dump = dumpit, 3009 .min_dump_alloc = min_dump_alloc, 3010 }; 3011 err = netlink_dump_start(rtnl, skb, nlh, &c); 3012 } 3013 rtnl_lock(); 3014 return err; 3015 } 3016 3017 doit = rtnl_get_doit(family, type); 3018 if (doit == NULL) 3019 return -EOPNOTSUPP; 3020 3021 return doit(skb, nlh); 3022 } 3023 3024 static void rtnetlink_rcv(struct sk_buff *skb) 3025 { 3026 rtnl_lock(); 3027 netlink_rcv_skb(skb, &rtnetlink_rcv_msg); 3028 rtnl_unlock(); 3029 } 3030 3031 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr) 3032 { 3033 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3034 3035 switch (event) { 3036 case NETDEV_UP: 3037 case NETDEV_DOWN: 3038 case NETDEV_PRE_UP: 3039 case NETDEV_POST_INIT: 3040 case NETDEV_REGISTER: 3041 case NETDEV_CHANGE: 3042 case NETDEV_PRE_TYPE_CHANGE: 3043 case NETDEV_GOING_DOWN: 3044 case NETDEV_UNREGISTER: 3045 case NETDEV_UNREGISTER_FINAL: 3046 case NETDEV_RELEASE: 3047 case NETDEV_JOIN: 3048 break; 3049 default: 3050 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL); 3051 break; 3052 } 3053 return NOTIFY_DONE; 3054 } 3055 3056 static struct notifier_block rtnetlink_dev_notifier = { 3057 .notifier_call = rtnetlink_event, 3058 }; 3059 3060 3061 static int __net_init rtnetlink_net_init(struct net *net) 3062 { 3063 struct sock *sk; 3064 struct netlink_kernel_cfg cfg = { 3065 .groups = RTNLGRP_MAX, 3066 .input = rtnetlink_rcv, 3067 .cb_mutex = &rtnl_mutex, 3068 .flags = NL_CFG_F_NONROOT_RECV, 3069 }; 3070 3071 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg); 3072 if (!sk) 3073 return -ENOMEM; 3074 net->rtnl = sk; 3075 return 0; 3076 } 3077 3078 static void __net_exit rtnetlink_net_exit(struct net *net) 3079 { 3080 netlink_kernel_release(net->rtnl); 3081 net->rtnl = NULL; 3082 } 3083 3084 static struct pernet_operations rtnetlink_net_ops = { 3085 .init = rtnetlink_net_init, 3086 .exit = rtnetlink_net_exit, 3087 }; 3088 3089 void __init rtnetlink_init(void) 3090 { 3091 if (register_pernet_subsys(&rtnetlink_net_ops)) 3092 panic("rtnetlink_init: cannot initialize rtnetlink\n"); 3093 3094 register_netdevice_notifier(&rtnetlink_dev_notifier); 3095 3096 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, 3097 rtnl_dump_ifinfo, rtnl_calcit); 3098 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL); 3099 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL); 3100 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL); 3101 3102 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL); 3103 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL); 3104 3105 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL); 3106 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL); 3107 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL); 3108 3109 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL); 3110 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL); 3111 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL); 3112 } 3113 3114