net/core/rtnetlink.c

   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/pci.h>
  40 #include <linux/etherdevice.h>
  41
  42 #include <asm/uaccess.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 struct rtnl_link {
  58         rtnl_doit_func          doit;
  59         rtnl_dumpit_func        dumpit;
  60         rtnl_calcit_func        calcit;
  61 };
  62
  63 static DEFINE_MUTEX(rtnl_mutex);
  64
  65 void rtnl_lock(void)
  66 {
  67         mutex_lock(&rtnl_mutex);
  68 }
  69 EXPORT_SYMBOL(rtnl_lock);
  70
  71 void __rtnl_unlock(void)
  72 {
  73         mutex_unlock(&rtnl_mutex);
  74 }
  75
  76 void rtnl_unlock(void)
  77 {
  78         /* This fellow will unlock it for us. */
  79         netdev_run_todo();
  80 }
  81 EXPORT_SYMBOL(rtnl_unlock);
  82
  83 int rtnl_trylock(void)
  84 {
  85         return mutex_trylock(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_trylock);
  88
  89 int rtnl_is_locked(void)
  90 {
  91         return mutex_is_locked(&rtnl_mutex);
  92 }
  93 EXPORT_SYMBOL(rtnl_is_locked);
  94
  95 #ifdef CONFIG_PROVE_LOCKING
  96 int lockdep_rtnl_is_held(void)
  97 {
  98         return lockdep_is_held(&rtnl_mutex);
  99 }
 100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 102
 103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 104
 105 static inline int rtm_msgindex(int msgtype)
 106 {
 107         int msgindex = msgtype - RTM_BASE;
 108
 109         /*
 110          * msgindex < 0 implies someone tried to register a netlink
 111          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 112          * the message type has not been added to linux/rtnetlink.h
 113          */
 114         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 115
 116         return msgindex;
 117 }
 118
 119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 120 {
 121         struct rtnl_link *tab;
 122
 123         if (protocol <= RTNL_FAMILY_MAX)
 124                 tab = rtnl_msg_handlers[protocol];
 125         else
 126                 tab = NULL;
 127
 128         if (tab == NULL || tab[msgindex].doit == NULL)
 129                 tab = rtnl_msg_handlers[PF_UNSPEC];
 130
 131         return tab[msgindex].doit;
 132 }
 133
 134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 135 {
 136         struct rtnl_link *tab;
 137
 138         if (protocol <= RTNL_FAMILY_MAX)
 139                 tab = rtnl_msg_handlers[protocol];
 140         else
 141                 tab = NULL;
 142
 143         if (tab == NULL || tab[msgindex].dumpit == NULL)
 144                 tab = rtnl_msg_handlers[PF_UNSPEC];
 145
 146         return tab[msgindex].dumpit;
 147 }
 148
 149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 150 {
 151         struct rtnl_link *tab;
 152
 153         if (protocol <= RTNL_FAMILY_MAX)
 154                 tab = rtnl_msg_handlers[protocol];
 155         else
 156                 tab = NULL;
 157
 158         if (tab == NULL || tab[msgindex].calcit == NULL)
 159                 tab = rtnl_msg_handlers[PF_UNSPEC];
 160
 161         return tab[msgindex].calcit;
 162 }
 163
 164 /**
 165  * __rtnl_register - Register a rtnetlink message type
 166  * @protocol: Protocol family or PF_UNSPEC
 167  * @msgtype: rtnetlink message type
 168  * @doit: Function pointer called for each request message
 169  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 170  * @calcit: Function pointer to calc size of dump message
 171  *
 172  * Registers the specified function pointers (at least one of them has
 173  * to be non-NULL) to be called whenever a request message for the
 174  * specified protocol family and message type is received.
 175  *
 176  * The special protocol family PF_UNSPEC may be used to define fallback
 177  * function pointers for the case when no entry for the specific protocol
 178  * family exists.
 179  *
 180  * Returns 0 on success or a negative error code.
 181  */
 182 int __rtnl_register(int protocol, int msgtype,
 183                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 184                     rtnl_calcit_func calcit)
 185 {
 186         struct rtnl_link *tab;
 187         int msgindex;
 188
 189         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 190         msgindex = rtm_msgindex(msgtype);
 191
 192         tab = rtnl_msg_handlers[protocol];
 193         if (tab == NULL) {
 194                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 195                 if (tab == NULL)
 196                         return -ENOBUFS;
 197
 198                 rtnl_msg_handlers[protocol] = tab;
 199         }
 200
 201         if (doit)
 202                 tab[msgindex].doit = doit;
 203
 204         if (dumpit)
 205                 tab[msgindex].dumpit = dumpit;
 206
 207         if (calcit)
 208                 tab[msgindex].calcit = calcit;
 209
 210         return 0;
 211 }
 212 EXPORT_SYMBOL_GPL(__rtnl_register);
 213
 214 /**
 215  * rtnl_register - Register a rtnetlink message type
 216  *
 217  * Identical to __rtnl_register() but panics on failure. This is useful
 218  * as failure of this function is very unlikely, it can only happen due
 219  * to lack of memory when allocating the chain to store all message
 220  * handlers for a protocol. Meant for use in init functions where lack
 221  * of memory implies no sense in continuing.
 222  */
 223 void rtnl_register(int protocol, int msgtype,
 224                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 225                    rtnl_calcit_func calcit)
 226 {
 227         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 228                 panic("Unable to register rtnetlink message handler, "
 229                       "protocol = %d, message type = %d\n",
 230                       protocol, msgtype);
 231 }
 232 EXPORT_SYMBOL_GPL(rtnl_register);
 233
 234 /**
 235  * rtnl_unregister - Unregister a rtnetlink message type
 236  * @protocol: Protocol family or PF_UNSPEC
 237  * @msgtype: rtnetlink message type
 238  *
 239  * Returns 0 on success or a negative error code.
 240  */
 241 int rtnl_unregister(int protocol, int msgtype)
 242 {
 243         int msgindex;
 244
 245         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 246         msgindex = rtm_msgindex(msgtype);
 247
 248         if (rtnl_msg_handlers[protocol] == NULL)
 249                 return -ENOENT;
 250
 251         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 252         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 253
 254         return 0;
 255 }
 256 EXPORT_SYMBOL_GPL(rtnl_unregister);
 257
 258 /**
 259  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 260  * @protocol : Protocol family or PF_UNSPEC
 261  *
 262  * Identical to calling rtnl_unregster() for all registered message types
 263  * of a certain protocol family.
 264  */
 265 void rtnl_unregister_all(int protocol)
 266 {
 267         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 268
 269         kfree(rtnl_msg_handlers[protocol]);
 270         rtnl_msg_handlers[protocol] = NULL;
 271 }
 272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 273
 274 static LIST_HEAD(link_ops);
 275
 276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 277 {
 278         const struct rtnl_link_ops *ops;
 279
 280         list_for_each_entry(ops, &link_ops, list) {
 281                 if (!strcmp(ops->kind, kind))
 282                         return ops;
 283         }
 284         return NULL;
 285 }
 286
 287 /**
 288  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 289  * @ops: struct rtnl_link_ops * to register
 290  *
 291  * The caller must hold the rtnl_mutex. This function should be used
 292  * by drivers that create devices during module initialization. It
 293  * must be called before registering the devices.
 294  *
 295  * Returns 0 on success or a negative error code.
 296  */
 297 int __rtnl_link_register(struct rtnl_link_ops *ops)
 298 {
 299         if (rtnl_link_ops_get(ops->kind))
 300                 return -EEXIST;
 301
 302         /* The check for setup is here because if ops
 303          * does not have that filled up, it is not possible
 304          * to use the ops for creating device. So do not
 305          * fill up dellink as well. That disables rtnl_dellink.
 306          */
 307         if (ops->setup && !ops->dellink)
 308                 ops->dellink = unregister_netdevice_queue;
 309
 310         list_add_tail(&ops->list, &link_ops);
 311         return 0;
 312 }
 313 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 314
 315 /**
 316  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 317  * @ops: struct rtnl_link_ops * to register
 318  *
 319  * Returns 0 on success or a negative error code.
 320  */
 321 int rtnl_link_register(struct rtnl_link_ops *ops)
 322 {
 323         int err;
 324
 325         rtnl_lock();
 326         err = __rtnl_link_register(ops);
 327         rtnl_unlock();
 328         return err;
 329 }
 330 EXPORT_SYMBOL_GPL(rtnl_link_register);
 331
 332 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 333 {
 334         struct net_device *dev;
 335         LIST_HEAD(list_kill);
 336
 337         for_each_netdev(net, dev) {
 338                 if (dev->rtnl_link_ops == ops)
 339                         ops->dellink(dev, &list_kill);
 340         }
 341         unregister_netdevice_many(&list_kill);
 342 }
 343
 344 /**
 345  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 346  * @ops: struct rtnl_link_ops * to unregister
 347  *
 348  * The caller must hold the rtnl_mutex.
 349  */
 350 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 351 {
 352         struct net *net;
 353
 354         for_each_net(net) {
 355                 __rtnl_kill_links(net, ops);
 356         }
 357         list_del(&ops->list);
 358 }
 359 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 360
 361 /* Return with the rtnl_lock held when there are no network
 362  * devices unregistering in any network namespace.
 363  */
 364 static void rtnl_lock_unregistering_all(void)
 365 {
 366         struct net *net;
 367         bool unregistering;
 368         DEFINE_WAIT(wait);
 369
 370         for (;;) {
 371                 prepare_to_wait(&netdev_unregistering_wq, &wait,
 372                                 TASK_UNINTERRUPTIBLE);
 373                 unregistering = false;
 374                 rtnl_lock();
 375                 for_each_net(net) {
 376                         if (net->dev_unreg_count > 0) {
 377                                 unregistering = true;
 378                                 break;
 379                         }
 380                 }
 381                 if (!unregistering)
 382                         break;
 383                 __rtnl_unlock();
 384                 schedule();
 385         }
 386         finish_wait(&netdev_unregistering_wq, &wait);
 387 }
 388
 389 /**
 390  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 391  * @ops: struct rtnl_link_ops * to unregister
 392  */
 393 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 394 {
 395         /* Close the race with cleanup_net() */
 396         mutex_lock(&net_mutex);
 397         rtnl_lock_unregistering_all();
 398         __rtnl_link_unregister(ops);
 399         rtnl_unlock();
 400         mutex_unlock(&net_mutex);
 401 }
 402 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 403
 404 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 405 {
 406         struct net_device *master_dev;
 407         const struct rtnl_link_ops *ops;
 408
 409         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 410         if (!master_dev)
 411                 return 0;
 412         ops = master_dev->rtnl_link_ops;
 413         if (!ops || !ops->get_slave_size)
 414                 return 0;
 415         /* IFLA_INFO_SLAVE_DATA + nested data */
 416         return nla_total_size(sizeof(struct nlattr)) +
 417                ops->get_slave_size(master_dev, dev);
 418 }
 419
 420 static size_t rtnl_link_get_size(const struct net_device *dev)
 421 {
 422         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 423         size_t size;
 424
 425         if (!ops)
 426                 return 0;
 427
 428         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 429                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 430
 431         if (ops->get_size)
 432                 /* IFLA_INFO_DATA + nested data */
 433                 size += nla_total_size(sizeof(struct nlattr)) +
 434                         ops->get_size(dev);
 435
 436         if (ops->get_xstats_size)
 437                 /* IFLA_INFO_XSTATS */
 438                 size += nla_total_size(ops->get_xstats_size(dev));
 439
 440         size += rtnl_link_get_slave_info_data_size(dev);
 441
 442         return size;
 443 }
 444
 445 static LIST_HEAD(rtnl_af_ops);
 446
 447 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 448 {
 449         const struct rtnl_af_ops *ops;
 450
 451         list_for_each_entry(ops, &rtnl_af_ops, list) {
 452                 if (ops->family == family)
 453                         return ops;
 454         }
 455
 456         return NULL;
 457 }
 458
 459 /**
 460  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 461  * @ops: struct rtnl_af_ops * to register
 462  *
 463  * Returns 0 on success or a negative error code.
 464  */
 465 void rtnl_af_register(struct rtnl_af_ops *ops)
 466 {
 467         rtnl_lock();
 468         list_add_tail(&ops->list, &rtnl_af_ops);
 469         rtnl_unlock();
 470 }
 471 EXPORT_SYMBOL_GPL(rtnl_af_register);
 472
 473 /**
 474  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 475  * @ops: struct rtnl_af_ops * to unregister
 476  *
 477  * The caller must hold the rtnl_mutex.
 478  */
 479 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 480 {
 481         list_del(&ops->list);
 482 }
 483 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 484
 485 /**
 486  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 487  * @ops: struct rtnl_af_ops * to unregister
 488  */
 489 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 490 {
 491         rtnl_lock();
 492         __rtnl_af_unregister(ops);
 493         rtnl_unlock();
 494 }
 495 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 496
 497 static size_t rtnl_link_get_af_size(const struct net_device *dev)
 498 {
 499         struct rtnl_af_ops *af_ops;
 500         size_t size;
 501
 502         /* IFLA_AF_SPEC */
 503         size = nla_total_size(sizeof(struct nlattr));
 504
 505         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 506                 if (af_ops->get_link_af_size) {
 507                         /* AF_* + nested data */
 508                         size += nla_total_size(sizeof(struct nlattr)) +
 509                                 af_ops->get_link_af_size(dev);
 510                 }
 511         }
 512
 513         return size;
 514 }
 515
 516 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 517 {
 518         struct net_device *master_dev;
 519
 520         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 521         if (master_dev && master_dev->rtnl_link_ops)
 522                 return true;
 523         return false;
 524 }
 525
 526 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 527                                      const struct net_device *dev)
 528 {
 529         struct net_device *master_dev;
 530         const struct rtnl_link_ops *ops;
 531         struct nlattr *slave_data;
 532         int err;
 533
 534         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 535         if (!master_dev)
 536                 return 0;
 537         ops = master_dev->rtnl_link_ops;
 538         if (!ops)
 539                 return 0;
 540         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 541                 return -EMSGSIZE;
 542         if (ops->fill_slave_info) {
 543                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 544                 if (!slave_data)
 545                         return -EMSGSIZE;
 546                 err = ops->fill_slave_info(skb, master_dev, dev);
 547                 if (err < 0)
 548                         goto err_cancel_slave_data;
 549                 nla_nest_end(skb, slave_data);
 550         }
 551         return 0;
 552
 553 err_cancel_slave_data:
 554         nla_nest_cancel(skb, slave_data);
 555         return err;
 556 }
 557
 558 static int rtnl_link_info_fill(struct sk_buff *skb,
 559                                const struct net_device *dev)
 560 {
 561         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 562         struct nlattr *data;
 563         int err;
 564
 565         if (!ops)
 566                 return 0;
 567         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 568                 return -EMSGSIZE;
 569         if (ops->fill_xstats) {
 570                 err = ops->fill_xstats(skb, dev);
 571                 if (err < 0)
 572                         return err;
 573         }
 574         if (ops->fill_info) {
 575                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 576                 if (data == NULL)
 577                         return -EMSGSIZE;
 578                 err = ops->fill_info(skb, dev);
 579                 if (err < 0)
 580                         goto err_cancel_data;
 581                 nla_nest_end(skb, data);
 582         }
 583         return 0;
 584
 585 err_cancel_data:
 586         nla_nest_cancel(skb, data);
 587         return err;
 588 }
 589
 590 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 591 {
 592         struct nlattr *linkinfo;
 593         int err = -EMSGSIZE;
 594
 595         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 596         if (linkinfo == NULL)
 597                 goto out;
 598
 599         err = rtnl_link_info_fill(skb, dev);
 600         if (err < 0)
 601                 goto err_cancel_link;
 602
 603         err = rtnl_link_slave_info_fill(skb, dev);
 604         if (err < 0)
 605                 goto err_cancel_link;
 606
 607         nla_nest_end(skb, linkinfo);
 608         return 0;
 609
 610 err_cancel_link:
 611         nla_nest_cancel(skb, linkinfo);
 612 out:
 613         return err;
 614 }
 615
 616 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 617 {
 618         struct sock *rtnl = net->rtnl;
 619         int err = 0;
 620
 621         NETLINK_CB(skb).dst_group = group;
 622         if (echo)
 623                 atomic_inc(&skb->users);
 624         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 625         if (echo)
 626                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 627         return err;
 628 }
 629
 630 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 631 {
 632         struct sock *rtnl = net->rtnl;
 633
 634         return nlmsg_unicast(rtnl, skb, pid);
 635 }
 636 EXPORT_SYMBOL(rtnl_unicast);
 637
 638 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 639                  struct nlmsghdr *nlh, gfp_t flags)
 640 {
 641         struct sock *rtnl = net->rtnl;
 642         int report = 0;
 643
 644         if (nlh)
 645                 report = nlmsg_report(nlh);
 646
 647         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 648 }
 649 EXPORT_SYMBOL(rtnl_notify);
 650
 651 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 652 {
 653         struct sock *rtnl = net->rtnl;
 654
 655         netlink_set_err(rtnl, 0, group, error);
 656 }
 657 EXPORT_SYMBOL(rtnl_set_sk_err);
 658
 659 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 660 {
 661         struct nlattr *mx;
 662         int i, valid = 0;
 663
 664         mx = nla_nest_start(skb, RTA_METRICS);
 665         if (mx == NULL)
 666                 return -ENOBUFS;
 667
 668         for (i = 0; i < RTAX_MAX; i++) {
 669                 if (metrics[i]) {
 670                         valid++;
 671                         if (nla_put_u32(skb, i+1, metrics[i]))
 672                                 goto nla_put_failure;
 673                 }
 674         }
 675
 676         if (!valid) {
 677                 nla_nest_cancel(skb, mx);
 678                 return 0;
 679         }
 680
 681         return nla_nest_end(skb, mx);
 682
 683 nla_put_failure:
 684         nla_nest_cancel(skb, mx);
 685         return -EMSGSIZE;
 686 }
 687 EXPORT_SYMBOL(rtnetlink_put_metrics);
 688
 689 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 690                        long expires, u32 error)
 691 {
 692         struct rta_cacheinfo ci = {
 693                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 694                 .rta_used = dst->__use,
 695                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 696                 .rta_error = error,
 697                 .rta_id =  id,
 698         };
 699
 700         if (expires) {
 701                 unsigned long clock;
 702
 703                 clock = jiffies_to_clock_t(abs(expires));
 704                 clock = min_t(unsigned long, clock, INT_MAX);
 705                 ci.rta_expires = (expires > 0) ? clock : -clock;
 706         }
 707         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 708 }
 709 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 710
 711 static void set_operstate(struct net_device *dev, unsigned char transition)
 712 {
 713         unsigned char operstate = dev->operstate;
 714
 715         switch (transition) {
 716         case IF_OPER_UP:
 717                 if ((operstate == IF_OPER_DORMANT ||
 718                      operstate == IF_OPER_UNKNOWN) &&
 719                     !netif_dormant(dev))
 720                         operstate = IF_OPER_UP;
 721                 break;
 722
 723         case IF_OPER_DORMANT:
 724                 if (operstate == IF_OPER_UP ||
 725                     operstate == IF_OPER_UNKNOWN)
 726                         operstate = IF_OPER_DORMANT;
 727                 break;
 728         }
 729
 730         if (dev->operstate != operstate) {
 731                 write_lock_bh(&dev_base_lock);
 732                 dev->operstate = operstate;
 733                 write_unlock_bh(&dev_base_lock);
 734                 netdev_state_change(dev);
 735         }
 736 }
 737
 738 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 739 {
 740         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 741                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 742 }
 743
 744 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 745                                            const struct ifinfomsg *ifm)
 746 {
 747         unsigned int flags = ifm->ifi_flags;
 748
 749         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 750         if (ifm->ifi_change)
 751                 flags = (flags & ifm->ifi_change) |
 752                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 753
 754         return flags;
 755 }
 756
 757 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 758                                  const struct rtnl_link_stats64 *b)
 759 {
 760         a->rx_packets = b->rx_packets;
 761         a->tx_packets = b->tx_packets;
 762         a->rx_bytes = b->rx_bytes;
 763         a->tx_bytes = b->tx_bytes;
 764         a->rx_errors = b->rx_errors;
 765         a->tx_errors = b->tx_errors;
 766         a->rx_dropped = b->rx_dropped;
 767         a->tx_dropped = b->tx_dropped;
 768
 769         a->multicast = b->multicast;
 770         a->collisions = b->collisions;
 771
 772         a->rx_length_errors = b->rx_length_errors;
 773         a->rx_over_errors = b->rx_over_errors;
 774         a->rx_crc_errors = b->rx_crc_errors;
 775         a->rx_frame_errors = b->rx_frame_errors;
 776         a->rx_fifo_errors = b->rx_fifo_errors;
 777         a->rx_missed_errors = b->rx_missed_errors;
 778
 779         a->tx_aborted_errors = b->tx_aborted_errors;
 780         a->tx_carrier_errors = b->tx_carrier_errors;
 781         a->tx_fifo_errors = b->tx_fifo_errors;
 782         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 783         a->tx_window_errors = b->tx_window_errors;
 784
 785         a->rx_compressed = b->rx_compressed;
 786         a->tx_compressed = b->tx_compressed;
 787 }
 788
 789 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
 790 {
 791         memcpy(v, b, sizeof(*b));
 792 }
 793
 794 /* All VF info */
 795 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 796                                    u32 ext_filter_mask)
 797 {
 798         if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
 799             (ext_filter_mask & RTEXT_FILTER_VF)) {
 800                 int num_vfs = dev_num_vf(dev->dev.parent);
 801                 size_t size = nla_total_size(sizeof(struct nlattr));
 802                 size += nla_total_size(num_vfs * sizeof(struct nlattr));
 803                 size += num_vfs *
 804                         (nla_total_size(sizeof(struct ifla_vf_mac)) +
 805                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 806                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 807                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 808                          nla_total_size(sizeof(struct ifla_vf_link_state)));
 809                 return size;
 810         } else
 811                 return 0;
 812 }
 813
 814 static size_t rtnl_port_size(const struct net_device *dev,
 815                              u32 ext_filter_mask)
 816 {
 817         size_t port_size = nla_total_size(4)            /* PORT_VF */
 818                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 819                 + nla_total_size(sizeof(struct ifla_port_vsi))
 820                                                         /* PORT_VSI_TYPE */
 821                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 822                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 823                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 824                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 825         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 826         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 827                 + port_size;
 828         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 829                 + port_size;
 830
 831         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 832             !(ext_filter_mask & RTEXT_FILTER_VF))
 833                 return 0;
 834         if (dev_num_vf(dev->dev.parent))
 835                 return port_self_size + vf_ports_size +
 836                         vf_port_size * dev_num_vf(dev->dev.parent);
 837         else
 838                 return port_self_size;
 839 }
 840
 841 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 842                                      u32 ext_filter_mask)
 843 {
 844         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 845                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 846                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 847                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 848                + nla_total_size(sizeof(struct rtnl_link_ifmap))
 849                + nla_total_size(sizeof(struct rtnl_link_stats))
 850                + nla_total_size(sizeof(struct rtnl_link_stats64))
 851                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 852                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 853                + nla_total_size(4) /* IFLA_TXQLEN */
 854                + nla_total_size(4) /* IFLA_WEIGHT */
 855                + nla_total_size(4) /* IFLA_MTU */
 856                + nla_total_size(4) /* IFLA_LINK */
 857                + nla_total_size(4) /* IFLA_MASTER */
 858                + nla_total_size(1) /* IFLA_CARRIER */
 859                + nla_total_size(4) /* IFLA_PROMISCUITY */
 860                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 861                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 862                + nla_total_size(1) /* IFLA_OPERSTATE */
 863                + nla_total_size(1) /* IFLA_LINKMODE */
 864                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 865                + nla_total_size(ext_filter_mask
 866                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 867                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 868                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 869                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 870                + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
 871                + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
 872 }
 873
 874 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 875 {
 876         struct nlattr *vf_ports;
 877         struct nlattr *vf_port;
 878         int vf;
 879         int err;
 880
 881         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 882         if (!vf_ports)
 883                 return -EMSGSIZE;
 884
 885         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 886                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 887                 if (!vf_port)
 888                         goto nla_put_failure;
 889                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 890                         goto nla_put_failure;
 891                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 892                 if (err == -EMSGSIZE)
 893                         goto nla_put_failure;
 894                 if (err) {
 895                         nla_nest_cancel(skb, vf_port);
 896                         continue;
 897                 }
 898                 nla_nest_end(skb, vf_port);
 899         }
 900
 901         nla_nest_end(skb, vf_ports);
 902
 903         return 0;
 904
 905 nla_put_failure:
 906         nla_nest_cancel(skb, vf_ports);
 907         return -EMSGSIZE;
 908 }
 909
 910 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 911 {
 912         struct nlattr *port_self;
 913         int err;
 914
 915         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 916         if (!port_self)
 917                 return -EMSGSIZE;
 918
 919         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 920         if (err) {
 921                 nla_nest_cancel(skb, port_self);
 922                 return (err == -EMSGSIZE) ? err : 0;
 923         }
 924
 925         nla_nest_end(skb, port_self);
 926
 927         return 0;
 928 }
 929
 930 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
 931                           u32 ext_filter_mask)
 932 {
 933         int err;
 934
 935         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 936             !(ext_filter_mask & RTEXT_FILTER_VF))
 937                 return 0;
 938
 939         err = rtnl_port_self_fill(skb, dev);
 940         if (err)
 941                 return err;
 942
 943         if (dev_num_vf(dev->dev.parent)) {
 944                 err = rtnl_vf_ports_fill(skb, dev);
 945                 if (err)
 946                         return err;
 947         }
 948
 949         return 0;
 950 }
 951
 952 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
 953 {
 954         int err;
 955         struct netdev_phys_port_id ppid;
 956
 957         err = dev_get_phys_port_id(dev, &ppid);
 958         if (err) {
 959                 if (err == -EOPNOTSUPP)
 960                         return 0;
 961                 return err;
 962         }
 963
 964         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
 965                 return -EMSGSIZE;
 966
 967         return 0;
 968 }
 969
 970 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 971                             int type, u32 pid, u32 seq, u32 change,
 972                             unsigned int flags, u32 ext_filter_mask)
 973 {
 974         struct ifinfomsg *ifm;
 975         struct nlmsghdr *nlh;
 976         struct rtnl_link_stats64 temp;
 977         const struct rtnl_link_stats64 *stats;
 978         struct nlattr *attr, *af_spec;
 979         struct rtnl_af_ops *af_ops;
 980         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
 981
 982         ASSERT_RTNL();
 983         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 984         if (nlh == NULL)
 985                 return -EMSGSIZE;
 986
 987         ifm = nlmsg_data(nlh);
 988         ifm->ifi_family = AF_UNSPEC;
 989         ifm->__ifi_pad = 0;
 990         ifm->ifi_type = dev->type;
 991         ifm->ifi_index = dev->ifindex;
 992         ifm->ifi_flags = dev_get_flags(dev);
 993         ifm->ifi_change = change;
 994
 995         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
 996             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
 997             nla_put_u8(skb, IFLA_OPERSTATE,
 998                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
 999             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1000             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1001             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1002             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1003             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1004 #ifdef CONFIG_RPS
1005             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1006 #endif
1007             (dev->ifindex != dev->iflink &&
1008              nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1009             (upper_dev &&
1010              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1011             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1012             (dev->qdisc &&
1013              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1014             (dev->ifalias &&
1015              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1016             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1017                         atomic_read(&dev->carrier_changes)))
1018                 goto nla_put_failure;
1019
1020         if (1) {
1021                 struct rtnl_link_ifmap map = {
1022                         .mem_start   = dev->mem_start,
1023                         .mem_end     = dev->mem_end,
1024                         .base_addr   = dev->base_addr,
1025                         .irq         = dev->irq,
1026                         .dma         = dev->dma,
1027                         .port        = dev->if_port,
1028                 };
1029                 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1030                         goto nla_put_failure;
1031         }
1032
1033         if (dev->addr_len) {
1034                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1035                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1036                         goto nla_put_failure;
1037         }
1038
1039         if (rtnl_phys_port_id_fill(skb, dev))
1040                 goto nla_put_failure;
1041
1042         attr = nla_reserve(skb, IFLA_STATS,
1043                         sizeof(struct rtnl_link_stats));
1044         if (attr == NULL)
1045                 goto nla_put_failure;
1046
1047         stats = dev_get_stats(dev, &temp);
1048         copy_rtnl_link_stats(nla_data(attr), stats);
1049
1050         attr = nla_reserve(skb, IFLA_STATS64,
1051                         sizeof(struct rtnl_link_stats64));
1052         if (attr == NULL)
1053                 goto nla_put_failure;
1054         copy_rtnl_link_stats64(nla_data(attr), stats);
1055
1056         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1057             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1058                 goto nla_put_failure;
1059
1060         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1061             && (ext_filter_mask & RTEXT_FILTER_VF)) {
1062                 int i;
1063
1064                 struct nlattr *vfinfo, *vf;
1065                 int num_vfs = dev_num_vf(dev->dev.parent);
1066
1067                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1068                 if (!vfinfo)
1069                         goto nla_put_failure;
1070                 for (i = 0; i < num_vfs; i++) {
1071                         struct ifla_vf_info ivi;
1072                         struct ifla_vf_mac vf_mac;
1073                         struct ifla_vf_vlan vf_vlan;
1074                         struct ifla_vf_rate vf_rate;
1075                         struct ifla_vf_tx_rate vf_tx_rate;
1076                         struct ifla_vf_spoofchk vf_spoofchk;
1077                         struct ifla_vf_link_state vf_linkstate;
1078
1079                         /*
1080                          * Not all SR-IOV capable drivers support the
1081                          * spoofcheck query.  Preset to -1 so the user
1082                          * space tool can detect that the driver didn't
1083                          * report anything.
1084                          */
1085                         ivi.spoofchk = -1;
1086                         memset(ivi.mac, 0, sizeof(ivi.mac));
1087                         /* The default value for VF link state is "auto"
1088                          * IFLA_VF_LINK_STATE_AUTO which equals zero
1089                          */
1090                         ivi.linkstate = 0;
1091                         if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1092                                 break;
1093                         vf_mac.vf =
1094                                 vf_vlan.vf =
1095                                 vf_rate.vf =
1096                                 vf_tx_rate.vf =
1097                                 vf_spoofchk.vf =
1098                                 vf_linkstate.vf = ivi.vf;
1099
1100                         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1101                         vf_vlan.vlan = ivi.vlan;
1102                         vf_vlan.qos = ivi.qos;
1103                         vf_tx_rate.rate = ivi.max_tx_rate;
1104                         vf_rate.min_tx_rate = ivi.min_tx_rate;
1105                         vf_rate.max_tx_rate = ivi.max_tx_rate;
1106                         vf_spoofchk.setting = ivi.spoofchk;
1107                         vf_linkstate.link_state = ivi.linkstate;
1108                         vf = nla_nest_start(skb, IFLA_VF_INFO);
1109                         if (!vf) {
1110                                 nla_nest_cancel(skb, vfinfo);
1111                                 goto nla_put_failure;
1112                         }
1113                         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1114                             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1115                             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1116                                     &vf_rate) ||
1117                             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1118                                     &vf_tx_rate) ||
1119                             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1120                                     &vf_spoofchk) ||
1121                             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1122                                     &vf_linkstate))
1123                                 goto nla_put_failure;
1124                         nla_nest_end(skb, vf);
1125                 }
1126                 nla_nest_end(skb, vfinfo);
1127         }
1128
1129         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1130                 goto nla_put_failure;
1131
1132         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1133                 if (rtnl_link_fill(skb, dev) < 0)
1134                         goto nla_put_failure;
1135         }
1136
1137         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1138                 goto nla_put_failure;
1139
1140         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1141                 if (af_ops->fill_link_af) {
1142                         struct nlattr *af;
1143                         int err;
1144
1145                         if (!(af = nla_nest_start(skb, af_ops->family)))
1146                                 goto nla_put_failure;
1147
1148                         err = af_ops->fill_link_af(skb, dev);
1149
1150                         /*
1151                          * Caller may return ENODATA to indicate that there
1152                          * was no data to be dumped. This is not an error, it
1153                          * means we should trim the attribute header and
1154                          * continue.
1155                          */
1156                         if (err == -ENODATA)
1157                                 nla_nest_cancel(skb, af);
1158                         else if (err < 0)
1159                                 goto nla_put_failure;
1160
1161                         nla_nest_end(skb, af);
1162                 }
1163         }
1164
1165         nla_nest_end(skb, af_spec);
1166
1167         return nlmsg_end(skb, nlh);
1168
1169 nla_put_failure:
1170         nlmsg_cancel(skb, nlh);
1171         return -EMSGSIZE;
1172 }
1173
1174 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1175         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1176         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1177         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1178         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1179         [IFLA_MTU]              = { .type = NLA_U32 },
1180         [IFLA_LINK]             = { .type = NLA_U32 },
1181         [IFLA_MASTER]           = { .type = NLA_U32 },
1182         [IFLA_CARRIER]          = { .type = NLA_U8 },
1183         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1184         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1185         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1186         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1187         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1188         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1189         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1190         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1191         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1192         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1193         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1194         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1195         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1196         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1197         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1198         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1199         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1200         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1201 };
1202
1203 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1204         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1205         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1206         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1207         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1208 };
1209
1210 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1211         [IFLA_VF_INFO]          = { .type = NLA_NESTED },
1212 };
1213
1214 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1215         [IFLA_VF_MAC]           = { .type = NLA_BINARY,
1216                                     .len = sizeof(struct ifla_vf_mac) },
1217         [IFLA_VF_VLAN]          = { .type = NLA_BINARY,
1218                                     .len = sizeof(struct ifla_vf_vlan) },
1219         [IFLA_VF_TX_RATE]       = { .type = NLA_BINARY,
1220                                     .len = sizeof(struct ifla_vf_tx_rate) },
1221         [IFLA_VF_SPOOFCHK]      = { .type = NLA_BINARY,
1222                                     .len = sizeof(struct ifla_vf_spoofchk) },
1223         [IFLA_VF_RATE]          = { .type = NLA_BINARY,
1224                                     .len = sizeof(struct ifla_vf_rate) },
1225         [IFLA_VF_LINK_STATE]    = { .type = NLA_BINARY,
1226                                     .len = sizeof(struct ifla_vf_link_state) },
1227 };
1228
1229 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1230         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1231         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1232                                     .len = PORT_PROFILE_MAX },
1233         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1234                                     .len = sizeof(struct ifla_port_vsi)},
1235         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1236                                       .len = PORT_UUID_MAX },
1237         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1238                                     .len = PORT_UUID_MAX },
1239         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1240         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1241 };
1242
1243 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1244 {
1245         struct net *net = sock_net(skb->sk);
1246         int h, s_h;
1247         int idx = 0, s_idx;
1248         struct net_device *dev;
1249         struct hlist_head *head;
1250         struct nlattr *tb[IFLA_MAX+1];
1251         u32 ext_filter_mask = 0;
1252         int err;
1253         int hdrlen;
1254
1255         s_h = cb->args[0];
1256         s_idx = cb->args[1];
1257
1258         rcu_read_lock();
1259         cb->seq = net->dev_base_seq;
1260
1261         /* A hack to preserve kernel<->userspace interface.
1262          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1263          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1264          * what iproute2 < v3.9.0 used.
1265          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1266          * attribute, its netlink message is shorter than struct ifinfomsg.
1267          */
1268         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1269                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1270
1271         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1272
1273                 if (tb[IFLA_EXT_MASK])
1274                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1275         }
1276
1277         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1278                 idx = 0;
1279                 head = &net->dev_index_head[h];
1280                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1281                         if (idx < s_idx)
1282                                 goto cont;
1283                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1284                                                NETLINK_CB(cb->skb).portid,
1285                                                cb->nlh->nlmsg_seq, 0,
1286                                                NLM_F_MULTI,
1287                                                ext_filter_mask);
1288                         /* If we ran out of room on the first message,
1289                          * we're in trouble
1290                          */
1291                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1292
1293                         if (err <= 0)
1294                                 goto out;
1295
1296                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1297 cont:
1298                         idx++;
1299                 }
1300         }
1301 out:
1302         rcu_read_unlock();
1303         cb->args[1] = idx;
1304         cb->args[0] = h;
1305
1306         return skb->len;
1307 }
1308
1309 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1310 {
1311         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1312 }
1313 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1314
1315 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1316 {
1317         struct net *net;
1318         /* Examine the link attributes and figure out which
1319          * network namespace we are talking about.
1320          */
1321         if (tb[IFLA_NET_NS_PID])
1322                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1323         else if (tb[IFLA_NET_NS_FD])
1324                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1325         else
1326                 net = get_net(src_net);
1327         return net;
1328 }
1329 EXPORT_SYMBOL(rtnl_link_get_net);
1330
1331 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1332 {
1333         if (dev) {
1334                 if (tb[IFLA_ADDRESS] &&
1335                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1336                         return -EINVAL;
1337
1338                 if (tb[IFLA_BROADCAST] &&
1339                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1340                         return -EINVAL;
1341         }
1342
1343         if (tb[IFLA_AF_SPEC]) {
1344                 struct nlattr *af;
1345                 int rem, err;
1346
1347                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1348                         const struct rtnl_af_ops *af_ops;
1349
1350                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1351                                 return -EAFNOSUPPORT;
1352
1353                         if (!af_ops->set_link_af)
1354                                 return -EOPNOTSUPP;
1355
1356                         if (af_ops->validate_link_af) {
1357                                 err = af_ops->validate_link_af(dev, af);
1358                                 if (err < 0)
1359                                         return err;
1360                         }
1361                 }
1362         }
1363
1364         return 0;
1365 }
1366
1367 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1368 {
1369         int rem, err = -EINVAL;
1370         struct nlattr *vf;
1371         const struct net_device_ops *ops = dev->netdev_ops;
1372
1373         nla_for_each_nested(vf, attr, rem) {
1374                 switch (nla_type(vf)) {
1375                 case IFLA_VF_MAC: {
1376                         struct ifla_vf_mac *ivm;
1377                         ivm = nla_data(vf);
1378                         err = -EOPNOTSUPP;
1379                         if (ops->ndo_set_vf_mac)
1380                                 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1381                                                           ivm->mac);
1382                         break;
1383                 }
1384                 case IFLA_VF_VLAN: {
1385                         struct ifla_vf_vlan *ivv;
1386                         ivv = nla_data(vf);
1387                         err = -EOPNOTSUPP;
1388                         if (ops->ndo_set_vf_vlan)
1389                                 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1390                                                            ivv->vlan,
1391                                                            ivv->qos);
1392                         break;
1393                 }
1394                 case IFLA_VF_TX_RATE: {
1395                         struct ifla_vf_tx_rate *ivt;
1396                         struct ifla_vf_info ivf;
1397                         ivt = nla_data(vf);
1398                         err = -EOPNOTSUPP;
1399                         if (ops->ndo_get_vf_config)
1400                                 err = ops->ndo_get_vf_config(dev, ivt->vf,
1401                                                              &ivf);
1402                         if (err)
1403                                 break;
1404                         err = -EOPNOTSUPP;
1405                         if (ops->ndo_set_vf_rate)
1406                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1407                                                            ivf.min_tx_rate,
1408                                                            ivt->rate);
1409                         break;
1410                 }
1411                 case IFLA_VF_RATE: {
1412                         struct ifla_vf_rate *ivt;
1413                         ivt = nla_data(vf);
1414                         err = -EOPNOTSUPP;
1415                         if (ops->ndo_set_vf_rate)
1416                                 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1417                                                            ivt->min_tx_rate,
1418                                                            ivt->max_tx_rate);
1419                         break;
1420                 }
1421                 case IFLA_VF_SPOOFCHK: {
1422                         struct ifla_vf_spoofchk *ivs;
1423                         ivs = nla_data(vf);
1424                         err = -EOPNOTSUPP;
1425                         if (ops->ndo_set_vf_spoofchk)
1426                                 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1427                                                                ivs->setting);
1428                         break;
1429                 }
1430                 case IFLA_VF_LINK_STATE: {
1431                         struct ifla_vf_link_state *ivl;
1432                         ivl = nla_data(vf);
1433                         err = -EOPNOTSUPP;
1434                         if (ops->ndo_set_vf_link_state)
1435                                 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1436                                                                  ivl->link_state);
1437                         break;
1438                 }
1439                 default:
1440                         err = -EINVAL;
1441                         break;
1442                 }
1443                 if (err)
1444                         break;
1445         }
1446         return err;
1447 }
1448
1449 static int do_set_master(struct net_device *dev, int ifindex)
1450 {
1451         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1452         const struct net_device_ops *ops;
1453         int err;
1454
1455         if (upper_dev) {
1456                 if (upper_dev->ifindex == ifindex)
1457                         return 0;
1458                 ops = upper_dev->netdev_ops;
1459                 if (ops->ndo_del_slave) {
1460                         err = ops->ndo_del_slave(upper_dev, dev);
1461                         if (err)
1462                                 return err;
1463                 } else {
1464                         return -EOPNOTSUPP;
1465                 }
1466         }
1467
1468         if (ifindex) {
1469                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1470                 if (!upper_dev)
1471                         return -EINVAL;
1472                 ops = upper_dev->netdev_ops;
1473                 if (ops->ndo_add_slave) {
1474                         err = ops->ndo_add_slave(upper_dev, dev);
1475                         if (err)
1476                                 return err;
1477                 } else {
1478                         return -EOPNOTSUPP;
1479                 }
1480         }
1481         return 0;
1482 }
1483
1484 #define DO_SETLINK_MODIFIED     0x01
1485 /* notify flag means notify + modified. */
1486 #define DO_SETLINK_NOTIFY       0x03
1487 static int do_setlink(const struct sk_buff *skb,
1488                       struct net_device *dev, struct ifinfomsg *ifm,
1489                       struct nlattr **tb, char *ifname, int status)
1490 {
1491         const struct net_device_ops *ops = dev->netdev_ops;
1492         int err;
1493
1494         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1495                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1496                 if (IS_ERR(net)) {
1497                         err = PTR_ERR(net);
1498                         goto errout;
1499                 }
1500                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1501                         put_net(net);
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,
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 rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2343 {
2344         struct net *net = sock_net(skb->sk);
2345         struct ndmsg *ndm;
2346         struct nlattr *tb[NDA_MAX+1];
2347         struct net_device *dev;
2348         u8 *addr;
2349         int err;
2350
2351         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2352         if (err < 0)
2353                 return err;
2354
2355         ndm = nlmsg_data(nlh);
2356         if (ndm->ndm_ifindex == 0) {
2357                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2358                 return -EINVAL;
2359         }
2360
2361         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2362         if (dev == NULL) {
2363                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2364                 return -ENODEV;
2365         }
2366
2367         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2368                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2369                 return -EINVAL;
2370         }
2371
2372         addr = nla_data(tb[NDA_LLADDR]);
2373
2374         err = -EOPNOTSUPP;
2375
2376         /* Support fdb on master device the net/bridge default case */
2377         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2378             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2379                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2380                 const struct net_device_ops *ops = br_dev->netdev_ops;
2381
2382                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2383                 if (err)
2384                         goto out;
2385                 else
2386                         ndm->ndm_flags &= ~NTF_MASTER;
2387         }
2388
2389         /* Embedded bridge, macvlan, and any other device support */
2390         if ((ndm->ndm_flags & NTF_SELF)) {
2391                 if (dev->netdev_ops->ndo_fdb_add)
2392                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2393                                                            nlh->nlmsg_flags);
2394                 else
2395                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2396                                                nlh->nlmsg_flags);
2397
2398                 if (!err) {
2399                         rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2400                         ndm->ndm_flags &= ~NTF_SELF;
2401                 }
2402         }
2403 out:
2404         return err;
2405 }
2406
2407 /**
2408  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2409  */
2410 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2411                      struct nlattr *tb[],
2412                      struct net_device *dev,
2413                      const unsigned char *addr)
2414 {
2415         int err = -EINVAL;
2416
2417         /* If aging addresses are supported device will need to
2418          * implement its own handler for this.
2419          */
2420         if (!(ndm->ndm_state & NUD_PERMANENT)) {
2421                 pr_info("%s: FDB only supports static addresses\n", dev->name);
2422                 return err;
2423         }
2424
2425         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2426                 err = dev_uc_del(dev, addr);
2427         else if (is_multicast_ether_addr(addr))
2428                 err = dev_mc_del(dev, addr);
2429
2430         return err;
2431 }
2432 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2433
2434 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2435 {
2436         struct net *net = sock_net(skb->sk);
2437         struct ndmsg *ndm;
2438         struct nlattr *tb[NDA_MAX+1];
2439         struct net_device *dev;
2440         int err = -EINVAL;
2441         __u8 *addr;
2442
2443         if (!netlink_capable(skb, CAP_NET_ADMIN))
2444                 return -EPERM;
2445
2446         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2447         if (err < 0)
2448                 return err;
2449
2450         ndm = nlmsg_data(nlh);
2451         if (ndm->ndm_ifindex == 0) {
2452                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2453                 return -EINVAL;
2454         }
2455
2456         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2457         if (dev == NULL) {
2458                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2459                 return -ENODEV;
2460         }
2461
2462         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2463                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2464                 return -EINVAL;
2465         }
2466
2467         addr = nla_data(tb[NDA_LLADDR]);
2468
2469         err = -EOPNOTSUPP;
2470
2471         /* Support fdb on master device the net/bridge default case */
2472         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2473             (dev->priv_flags & IFF_BRIDGE_PORT)) {
2474                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2475                 const struct net_device_ops *ops = br_dev->netdev_ops;
2476
2477                 if (ops->ndo_fdb_del)
2478                         err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2479
2480                 if (err)
2481                         goto out;
2482                 else
2483                         ndm->ndm_flags &= ~NTF_MASTER;
2484         }
2485
2486         /* Embedded bridge, macvlan, and any other device support */
2487         if (ndm->ndm_flags & NTF_SELF) {
2488                 if (dev->netdev_ops->ndo_fdb_del)
2489                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2490                 else
2491                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2492
2493                 if (!err) {
2494                         rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2495                         ndm->ndm_flags &= ~NTF_SELF;
2496                 }
2497         }
2498 out:
2499         return err;
2500 }
2501
2502 static int nlmsg_populate_fdb(struct sk_buff *skb,
2503                               struct netlink_callback *cb,
2504                               struct net_device *dev,
2505                               int *idx,
2506                               struct netdev_hw_addr_list *list)
2507 {
2508         struct netdev_hw_addr *ha;
2509         int err;
2510         u32 portid, seq;
2511
2512         portid = NETLINK_CB(cb->skb).portid;
2513         seq = cb->nlh->nlmsg_seq;
2514
2515         list_for_each_entry(ha, &list->list, list) {
2516                 if (*idx < cb->args[0])
2517                         goto skip;
2518
2519                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2520                                               portid, seq,
2521                                               RTM_NEWNEIGH, NTF_SELF,
2522                                               NLM_F_MULTI);
2523                 if (err < 0)
2524                         return err;
2525 skip:
2526                 *idx += 1;
2527         }
2528         return 0;
2529 }
2530
2531 /**
2532  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2533  * @nlh: netlink message header
2534  * @dev: netdevice
2535  *
2536  * Default netdevice operation to dump the existing unicast address list.
2537  * Returns number of addresses from list put in skb.
2538  */
2539 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2540                       struct netlink_callback *cb,
2541                       struct net_device *dev,
2542                       struct net_device *filter_dev,
2543                       int idx)
2544 {
2545         int err;
2546
2547         netif_addr_lock_bh(dev);
2548         err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2549         if (err)
2550                 goto out;
2551         nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2552 out:
2553         netif_addr_unlock_bh(dev);
2554         return idx;
2555 }
2556 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2557
2558 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2559 {
2560         struct net_device *dev;
2561         struct nlattr *tb[IFLA_MAX+1];
2562         struct net_device *bdev = NULL;
2563         struct net_device *br_dev = NULL;
2564         const struct net_device_ops *ops = NULL;
2565         const struct net_device_ops *cops = NULL;
2566         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2567         struct net *net = sock_net(skb->sk);
2568         int brport_idx = 0;
2569         int br_idx = 0;
2570         int idx = 0;
2571
2572         if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2573                         ifla_policy) == 0) {
2574                 if (tb[IFLA_MASTER])
2575                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
2576         }
2577
2578         brport_idx = ifm->ifi_index;
2579
2580         if (br_idx) {
2581                 br_dev = __dev_get_by_index(net, br_idx);
2582                 if (!br_dev)
2583                         return -ENODEV;
2584
2585                 ops = br_dev->netdev_ops;
2586                 bdev = br_dev;
2587         }
2588
2589         for_each_netdev(net, dev) {
2590                 if (brport_idx && (dev->ifindex != brport_idx))
2591                         continue;
2592
2593                 if (!br_idx) { /* user did not specify a specific bridge */
2594                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
2595                                 br_dev = netdev_master_upper_dev_get(dev);
2596                                 cops = br_dev->netdev_ops;
2597                         }
2598
2599                         bdev = dev;
2600                 } else {
2601                         if (dev != br_dev &&
2602                             !(dev->priv_flags & IFF_BRIDGE_PORT))
2603                                 continue;
2604
2605                         if (br_dev != netdev_master_upper_dev_get(dev) &&
2606                             !(dev->priv_flags & IFF_EBRIDGE))
2607                                 continue;
2608
2609                         bdev = br_dev;
2610                         cops = ops;
2611                 }
2612
2613                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2614                         if (cops && cops->ndo_fdb_dump)
2615                                 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2616                                                          idx);
2617                 }
2618
2619                 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2620                 if (dev->netdev_ops->ndo_fdb_dump)
2621                         idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev,
2622                                                             idx);
2623
2624                 cops = NULL;
2625         }
2626
2627         cb->args[0] = idx;
2628         return skb->len;
2629 }
2630
2631 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2632                             struct net_device *dev, u16 mode)
2633 {
2634         struct nlmsghdr *nlh;
2635         struct ifinfomsg *ifm;
2636         struct nlattr *br_afspec;
2637         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2638         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2639
2640         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2641         if (nlh == NULL)
2642                 return -EMSGSIZE;
2643
2644         ifm = nlmsg_data(nlh);
2645         ifm->ifi_family = AF_BRIDGE;
2646         ifm->__ifi_pad = 0;
2647         ifm->ifi_type = dev->type;
2648         ifm->ifi_index = dev->ifindex;
2649         ifm->ifi_flags = dev_get_flags(dev);
2650         ifm->ifi_change = 0;
2651
2652
2653         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2654             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2655             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2656             (br_dev &&
2657              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2658             (dev->addr_len &&
2659              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2660             (dev->ifindex != dev->iflink &&
2661              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2662                 goto nla_put_failure;
2663
2664         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2665         if (!br_afspec)
2666                 goto nla_put_failure;
2667
2668         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2669             nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2670                 nla_nest_cancel(skb, br_afspec);
2671                 goto nla_put_failure;
2672         }
2673         nla_nest_end(skb, br_afspec);
2674
2675         return nlmsg_end(skb, nlh);
2676 nla_put_failure:
2677         nlmsg_cancel(skb, nlh);
2678         return -EMSGSIZE;
2679 }
2680 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2681
2682 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2683 {
2684         struct net *net = sock_net(skb->sk);
2685         struct net_device *dev;
2686         int idx = 0;
2687         u32 portid = NETLINK_CB(cb->skb).portid;
2688         u32 seq = cb->nlh->nlmsg_seq;
2689         u32 filter_mask = 0;
2690
2691         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2692                 struct nlattr *extfilt;
2693
2694                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2695                                           IFLA_EXT_MASK);
2696                 if (extfilt) {
2697                         if (nla_len(extfilt) < sizeof(filter_mask))
2698                                 return -EINVAL;
2699
2700                         filter_mask = nla_get_u32(extfilt);
2701                 }
2702         }
2703
2704         rcu_read_lock();
2705         for_each_netdev_rcu(net, dev) {
2706                 const struct net_device_ops *ops = dev->netdev_ops;
2707                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2708
2709                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2710                         if (idx >= cb->args[0] &&
2711                             br_dev->netdev_ops->ndo_bridge_getlink(
2712                                     skb, portid, seq, dev, filter_mask) < 0)
2713                                 break;
2714                         idx++;
2715                 }
2716
2717                 if (ops->ndo_bridge_getlink) {
2718                         if (idx >= cb->args[0] &&
2719                             ops->ndo_bridge_getlink(skb, portid, seq, dev,
2720                                                     filter_mask) < 0)
2721                                 break;
2722                         idx++;
2723                 }
2724         }
2725         rcu_read_unlock();
2726         cb->args[0] = idx;
2727
2728         return skb->len;
2729 }
2730
2731 static inline size_t bridge_nlmsg_size(void)
2732 {
2733         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2734                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
2735                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
2736                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
2737                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
2738                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
2739                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
2740                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
2741                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2742                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
2743                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
2744 }
2745
2746 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2747 {
2748         struct net *net = dev_net(dev);
2749         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2750         struct sk_buff *skb;
2751         int err = -EOPNOTSUPP;
2752
2753         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2754         if (!skb) {
2755                 err = -ENOMEM;
2756                 goto errout;
2757         }
2758
2759         if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2760             br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2761                 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2762                 if (err < 0)
2763                         goto errout;
2764         }
2765
2766         if ((flags & BRIDGE_FLAGS_SELF) &&
2767             dev->netdev_ops->ndo_bridge_getlink) {
2768                 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2769                 if (err < 0)
2770                         goto errout;
2771         }
2772
2773         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2774         return 0;
2775 errout:
2776         WARN_ON(err == -EMSGSIZE);
2777         kfree_skb(skb);
2778         rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2779         return err;
2780 }
2781
2782 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2783 {
2784         struct net *net = sock_net(skb->sk);
2785         struct ifinfomsg *ifm;
2786         struct net_device *dev;
2787         struct nlattr *br_spec, *attr = NULL;
2788         int rem, err = -EOPNOTSUPP;
2789         u16 oflags, flags = 0;
2790         bool have_flags = false;
2791
2792         if (nlmsg_len(nlh) < sizeof(*ifm))
2793                 return -EINVAL;
2794
2795         ifm = nlmsg_data(nlh);
2796         if (ifm->ifi_family != AF_BRIDGE)
2797                 return -EPFNOSUPPORT;
2798
2799         dev = __dev_get_by_index(net, ifm->ifi_index);
2800         if (!dev) {
2801                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2802                 return -ENODEV;
2803         }
2804
2805         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2806         if (br_spec) {
2807                 nla_for_each_nested(attr, br_spec, rem) {
2808                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2809                                 if (nla_len(attr) < sizeof(flags))
2810                                         return -EINVAL;
2811
2812                                 have_flags = true;
2813                                 flags = nla_get_u16(attr);
2814                                 break;
2815                         }
2816                 }
2817         }
2818
2819         oflags = flags;
2820
2821         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2822                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2823
2824                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2825                         err = -EOPNOTSUPP;
2826                         goto out;
2827                 }
2828
2829                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2830                 if (err)
2831                         goto out;
2832
2833                 flags &= ~BRIDGE_FLAGS_MASTER;
2834         }
2835
2836         if ((flags & BRIDGE_FLAGS_SELF)) {
2837                 if (!dev->netdev_ops->ndo_bridge_setlink)
2838                         err = -EOPNOTSUPP;
2839                 else
2840                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2841
2842                 if (!err)
2843                         flags &= ~BRIDGE_FLAGS_SELF;
2844         }
2845
2846         if (have_flags)
2847                 memcpy(nla_data(attr), &flags, sizeof(flags));
2848         /* Generate event to notify upper layer of bridge change */
2849         if (!err)
2850                 err = rtnl_bridge_notify(dev, oflags);
2851 out:
2852         return err;
2853 }
2854
2855 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2856 {
2857         struct net *net = sock_net(skb->sk);
2858         struct ifinfomsg *ifm;
2859         struct net_device *dev;
2860         struct nlattr *br_spec, *attr = NULL;
2861         int rem, err = -EOPNOTSUPP;
2862         u16 oflags, flags = 0;
2863         bool have_flags = false;
2864
2865         if (nlmsg_len(nlh) < sizeof(*ifm))
2866                 return -EINVAL;
2867
2868         ifm = nlmsg_data(nlh);
2869         if (ifm->ifi_family != AF_BRIDGE)
2870                 return -EPFNOSUPPORT;
2871
2872         dev = __dev_get_by_index(net, ifm->ifi_index);
2873         if (!dev) {
2874                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2875                 return -ENODEV;
2876         }
2877
2878         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2879         if (br_spec) {
2880                 nla_for_each_nested(attr, br_spec, rem) {
2881                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2882                                 if (nla_len(attr) < sizeof(flags))
2883                                         return -EINVAL;
2884
2885                                 have_flags = true;
2886                                 flags = nla_get_u16(attr);
2887                                 break;
2888                         }
2889                 }
2890         }
2891
2892         oflags = flags;
2893
2894         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2895                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2896
2897                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2898                         err = -EOPNOTSUPP;
2899                         goto out;
2900                 }
2901
2902                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2903                 if (err)
2904                         goto out;
2905
2906                 flags &= ~BRIDGE_FLAGS_MASTER;
2907         }
2908
2909         if ((flags & BRIDGE_FLAGS_SELF)) {
2910                 if (!dev->netdev_ops->ndo_bridge_dellink)
2911                         err = -EOPNOTSUPP;
2912                 else
2913                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2914
2915                 if (!err)
2916                         flags &= ~BRIDGE_FLAGS_SELF;
2917         }
2918
2919         if (have_flags)
2920                 memcpy(nla_data(attr), &flags, sizeof(flags));
2921         /* Generate event to notify upper layer of bridge change */
2922         if (!err)
2923                 err = rtnl_bridge_notify(dev, oflags);
2924 out:
2925         return err;
2926 }
2927
2928 /* Process one rtnetlink message. */
2929
2930 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2931 {
2932         struct net *net = sock_net(skb->sk);
2933         rtnl_doit_func doit;
2934         int sz_idx, kind;
2935         int family;
2936         int type;
2937         int err;
2938
2939         type = nlh->nlmsg_type;
2940         if (type > RTM_MAX)
2941                 return -EOPNOTSUPP;
2942
2943         type -= RTM_BASE;
2944
2945         /* All the messages must have at least 1 byte length */
2946         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2947                 return 0;
2948
2949         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2950         sz_idx = type>>2;
2951         kind = type&3;
2952
2953         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2954                 return -EPERM;
2955
2956         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2957                 struct sock *rtnl;
2958                 rtnl_dumpit_func dumpit;
2959                 rtnl_calcit_func calcit;
2960                 u16 min_dump_alloc = 0;
2961
2962                 dumpit = rtnl_get_dumpit(family, type);
2963                 if (dumpit == NULL)
2964                         return -EOPNOTSUPP;
2965                 calcit = rtnl_get_calcit(family, type);
2966                 if (calcit)
2967                         min_dump_alloc = calcit(skb, nlh);
2968
2969                 __rtnl_unlock();
2970                 rtnl = net->rtnl;
2971                 {
2972                         struct netlink_dump_control c = {
2973                                 .dump           = dumpit,
2974                                 .min_dump_alloc = min_dump_alloc,
2975                         };
2976                         err = netlink_dump_start(rtnl, skb, nlh, &c);
2977                 }
2978                 rtnl_lock();
2979                 return err;
2980         }
2981
2982         doit = rtnl_get_doit(family, type);
2983         if (doit == NULL)
2984                 return -EOPNOTSUPP;
2985
2986         return doit(skb, nlh);
2987 }
2988
2989 static void rtnetlink_rcv(struct sk_buff *skb)
2990 {
2991         rtnl_lock();
2992         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2993         rtnl_unlock();
2994 }
2995
2996 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2997 {
2998         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2999
3000         switch (event) {
3001         case NETDEV_UP:
3002         case NETDEV_DOWN:
3003         case NETDEV_PRE_UP:
3004         case NETDEV_POST_INIT:
3005         case NETDEV_REGISTER:
3006         case NETDEV_CHANGE:
3007         case NETDEV_PRE_TYPE_CHANGE:
3008         case NETDEV_GOING_DOWN:
3009         case NETDEV_UNREGISTER:
3010         case NETDEV_UNREGISTER_FINAL:
3011         case NETDEV_RELEASE:
3012         case NETDEV_JOIN:
3013                 break;
3014         default:
3015                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3016                 break;
3017         }
3018         return NOTIFY_DONE;
3019 }
3020
3021 static struct notifier_block rtnetlink_dev_notifier = {
3022         .notifier_call  = rtnetlink_event,
3023 };
3024
3025
3026 static int __net_init rtnetlink_net_init(struct net *net)
3027 {
3028         struct sock *sk;
3029         struct netlink_kernel_cfg cfg = {
3030                 .groups         = RTNLGRP_MAX,
3031                 .input          = rtnetlink_rcv,
3032                 .cb_mutex       = &rtnl_mutex,
3033                 .flags          = NL_CFG_F_NONROOT_RECV,
3034         };
3035
3036         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3037         if (!sk)
3038                 return -ENOMEM;
3039         net->rtnl = sk;
3040         return 0;
3041 }
3042
3043 static void __net_exit rtnetlink_net_exit(struct net *net)
3044 {
3045         netlink_kernel_release(net->rtnl);
3046         net->rtnl = NULL;
3047 }
3048
3049 static struct pernet_operations rtnetlink_net_ops = {
3050         .init = rtnetlink_net_init,
3051         .exit = rtnetlink_net_exit,
3052 };
3053
3054 void __init rtnetlink_init(void)
3055 {
3056         if (register_pernet_subsys(&rtnetlink_net_ops))
3057                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3058
3059         register_netdevice_notifier(&rtnetlink_dev_notifier);
3060
3061         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3062                       rtnl_dump_ifinfo, rtnl_calcit);
3063         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3064         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3065         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3066
3067         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3068         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3069
3070         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3071         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3072         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3073
3074         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3075         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3076         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
3077 }
3078