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/bitops.h>
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/kernel.h>
  25 #include <linux/timer.h>
  26 #include <linux/string.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/fcntl.h>
  30 #include <linux/mm.h>
  31 #include <linux/slab.h>
  32 #include <linux/interrupt.h>
  33 #include <linux/capability.h>
  34 #include <linux/skbuff.h>
  35 #include <linux/init.h>
  36 #include <linux/security.h>
  37 #include <linux/mutex.h>
  38 #include <linux/if_addr.h>
  39 #include <linux/if_bridge.h>
  40 #include <linux/if_vlan.h>
  41 #include <linux/pci.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/bpf.h>
  44
  45 #include <linux/uaccess.h>
  46
  47 #include <linux/inet.h>
  48 #include <linux/netdevice.h>
  49 #include <net/switchdev.h>
  50 #include <net/ip.h>
  51 #include <net/protocol.h>
  52 #include <net/arp.h>
  53 #include <net/route.h>
  54 #include <net/udp.h>
  55 #include <net/tcp.h>
  56 #include <net/sock.h>
  57 #include <net/pkt_sched.h>
  58 #include <net/fib_rules.h>
  59 #include <net/rtnetlink.h>
  60 #include <net/net_namespace.h>
  61
  62 struct rtnl_link {
  63         rtnl_doit_func          doit;
  64         rtnl_dumpit_func        dumpit;
  65         rtnl_calcit_func        calcit;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 static struct sk_buff *defer_kfree_skb_list;
  77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  78 {
  79         if (head && tail) {
  80                 tail->next = defer_kfree_skb_list;
  81                 defer_kfree_skb_list = head;
  82         }
  83 }
  84 EXPORT_SYMBOL(rtnl_kfree_skbs);
  85
  86 void __rtnl_unlock(void)
  87 {
  88         struct sk_buff *head = defer_kfree_skb_list;
  89
  90         defer_kfree_skb_list = NULL;
  91
  92         mutex_unlock(&rtnl_mutex);
  93
  94         while (head) {
  95                 struct sk_buff *next = head->next;
  96
  97                 kfree_skb(head);
  98                 cond_resched();
  99                 head = next;
 100         }
 101 }
 102
 103 void rtnl_unlock(void)
 104 {
 105         /* This fellow will unlock it for us. */
 106         netdev_run_todo();
 107 }
 108 EXPORT_SYMBOL(rtnl_unlock);
 109
 110 int rtnl_trylock(void)
 111 {
 112         return mutex_trylock(&rtnl_mutex);
 113 }
 114 EXPORT_SYMBOL(rtnl_trylock);
 115
 116 int rtnl_is_locked(void)
 117 {
 118         return mutex_is_locked(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_is_locked);
 121
 122 #ifdef CONFIG_PROVE_LOCKING
 123 bool lockdep_rtnl_is_held(void)
 124 {
 125         return lockdep_is_held(&rtnl_mutex);
 126 }
 127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 129
 130 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 131
 132 static inline int rtm_msgindex(int msgtype)
 133 {
 134         int msgindex = msgtype - RTM_BASE;
 135
 136         /*
 137          * msgindex < 0 implies someone tried to register a netlink
 138          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 139          * the message type has not been added to linux/rtnetlink.h
 140          */
 141         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 142
 143         return msgindex;
 144 }
 145
 146 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 147 {
 148         struct rtnl_link *tab;
 149
 150         if (protocol <= RTNL_FAMILY_MAX)
 151                 tab = rtnl_msg_handlers[protocol];
 152         else
 153                 tab = NULL;
 154
 155         if (tab == NULL || tab[msgindex].doit == NULL)
 156                 tab = rtnl_msg_handlers[PF_UNSPEC];
 157
 158         return tab[msgindex].doit;
 159 }
 160
 161 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 162 {
 163         struct rtnl_link *tab;
 164
 165         if (protocol <= RTNL_FAMILY_MAX)
 166                 tab = rtnl_msg_handlers[protocol];
 167         else
 168                 tab = NULL;
 169
 170         if (tab == NULL || tab[msgindex].dumpit == NULL)
 171                 tab = rtnl_msg_handlers[PF_UNSPEC];
 172
 173         return tab[msgindex].dumpit;
 174 }
 175
 176 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
 177 {
 178         struct rtnl_link *tab;
 179
 180         if (protocol <= RTNL_FAMILY_MAX)
 181                 tab = rtnl_msg_handlers[protocol];
 182         else
 183                 tab = NULL;
 184
 185         if (tab == NULL || tab[msgindex].calcit == NULL)
 186                 tab = rtnl_msg_handlers[PF_UNSPEC];
 187
 188         return tab[msgindex].calcit;
 189 }
 190
 191 /**
 192  * __rtnl_register - Register a rtnetlink message type
 193  * @protocol: Protocol family or PF_UNSPEC
 194  * @msgtype: rtnetlink message type
 195  * @doit: Function pointer called for each request message
 196  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 197  * @calcit: Function pointer to calc size of dump message
 198  *
 199  * Registers the specified function pointers (at least one of them has
 200  * to be non-NULL) to be called whenever a request message for the
 201  * specified protocol family and message type is received.
 202  *
 203  * The special protocol family PF_UNSPEC may be used to define fallback
 204  * function pointers for the case when no entry for the specific protocol
 205  * family exists.
 206  *
 207  * Returns 0 on success or a negative error code.
 208  */
 209 int __rtnl_register(int protocol, int msgtype,
 210                     rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 211                     rtnl_calcit_func calcit)
 212 {
 213         struct rtnl_link *tab;
 214         int msgindex;
 215
 216         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 217         msgindex = rtm_msgindex(msgtype);
 218
 219         tab = rtnl_msg_handlers[protocol];
 220         if (tab == NULL) {
 221                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 222                 if (tab == NULL)
 223                         return -ENOBUFS;
 224
 225                 rtnl_msg_handlers[protocol] = tab;
 226         }
 227
 228         if (doit)
 229                 tab[msgindex].doit = doit;
 230
 231         if (dumpit)
 232                 tab[msgindex].dumpit = dumpit;
 233
 234         if (calcit)
 235                 tab[msgindex].calcit = calcit;
 236
 237         return 0;
 238 }
 239 EXPORT_SYMBOL_GPL(__rtnl_register);
 240
 241 /**
 242  * rtnl_register - Register a rtnetlink message type
 243  *
 244  * Identical to __rtnl_register() but panics on failure. This is useful
 245  * as failure of this function is very unlikely, it can only happen due
 246  * to lack of memory when allocating the chain to store all message
 247  * handlers for a protocol. Meant for use in init functions where lack
 248  * of memory implies no sense in continuing.
 249  */
 250 void rtnl_register(int protocol, int msgtype,
 251                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 252                    rtnl_calcit_func calcit)
 253 {
 254         if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
 255                 panic("Unable to register rtnetlink message handler, "
 256                       "protocol = %d, message type = %d\n",
 257                       protocol, msgtype);
 258 }
 259 EXPORT_SYMBOL_GPL(rtnl_register);
 260
 261 /**
 262  * rtnl_unregister - Unregister a rtnetlink message type
 263  * @protocol: Protocol family or PF_UNSPEC
 264  * @msgtype: rtnetlink message type
 265  *
 266  * Returns 0 on success or a negative error code.
 267  */
 268 int rtnl_unregister(int protocol, int msgtype)
 269 {
 270         int msgindex;
 271
 272         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 273         msgindex = rtm_msgindex(msgtype);
 274
 275         if (rtnl_msg_handlers[protocol] == NULL)
 276                 return -ENOENT;
 277
 278         rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 279         rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 280         rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
 281
 282         return 0;
 283 }
 284 EXPORT_SYMBOL_GPL(rtnl_unregister);
 285
 286 /**
 287  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 288  * @protocol : Protocol family or PF_UNSPEC
 289  *
 290  * Identical to calling rtnl_unregster() for all registered message types
 291  * of a certain protocol family.
 292  */
 293 void rtnl_unregister_all(int protocol)
 294 {
 295         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 296
 297         kfree(rtnl_msg_handlers[protocol]);
 298         rtnl_msg_handlers[protocol] = NULL;
 299 }
 300 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 301
 302 static LIST_HEAD(link_ops);
 303
 304 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 305 {
 306         const struct rtnl_link_ops *ops;
 307
 308         list_for_each_entry(ops, &link_ops, list) {
 309                 if (!strcmp(ops->kind, kind))
 310                         return ops;
 311         }
 312         return NULL;
 313 }
 314
 315 /**
 316  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 317  * @ops: struct rtnl_link_ops * to register
 318  *
 319  * The caller must hold the rtnl_mutex. This function should be used
 320  * by drivers that create devices during module initialization. It
 321  * must be called before registering the devices.
 322  *
 323  * Returns 0 on success or a negative error code.
 324  */
 325 int __rtnl_link_register(struct rtnl_link_ops *ops)
 326 {
 327         if (rtnl_link_ops_get(ops->kind))
 328                 return -EEXIST;
 329
 330         /* The check for setup is here because if ops
 331          * does not have that filled up, it is not possible
 332          * to use the ops for creating device. So do not
 333          * fill up dellink as well. That disables rtnl_dellink.
 334          */
 335         if (ops->setup && !ops->dellink)
 336                 ops->dellink = unregister_netdevice_queue;
 337
 338         list_add_tail(&ops->list, &link_ops);
 339         return 0;
 340 }
 341 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 342
 343 /**
 344  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 345  * @ops: struct rtnl_link_ops * to register
 346  *
 347  * Returns 0 on success or a negative error code.
 348  */
 349 int rtnl_link_register(struct rtnl_link_ops *ops)
 350 {
 351         int err;
 352
 353         rtnl_lock();
 354         err = __rtnl_link_register(ops);
 355         rtnl_unlock();
 356         return err;
 357 }
 358 EXPORT_SYMBOL_GPL(rtnl_link_register);
 359
 360 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 361 {
 362         struct net_device *dev;
 363         LIST_HEAD(list_kill);
 364
 365         for_each_netdev(net, dev) {
 366                 if (dev->rtnl_link_ops == ops)
 367                         ops->dellink(dev, &list_kill);
 368         }
 369         unregister_netdevice_many(&list_kill);
 370 }
 371
 372 /**
 373  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 374  * @ops: struct rtnl_link_ops * to unregister
 375  *
 376  * The caller must hold the rtnl_mutex.
 377  */
 378 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 379 {
 380         struct net *net;
 381
 382         for_each_net(net) {
 383                 __rtnl_kill_links(net, ops);
 384         }
 385         list_del(&ops->list);
 386 }
 387 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 388
 389 /* Return with the rtnl_lock held when there are no network
 390  * devices unregistering in any network namespace.
 391  */
 392 static void rtnl_lock_unregistering_all(void)
 393 {
 394         struct net *net;
 395         bool unregistering;
 396         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 397
 398         add_wait_queue(&netdev_unregistering_wq, &wait);
 399         for (;;) {
 400                 unregistering = false;
 401                 rtnl_lock();
 402                 for_each_net(net) {
 403                         if (net->dev_unreg_count > 0) {
 404                                 unregistering = true;
 405                                 break;
 406                         }
 407                 }
 408                 if (!unregistering)
 409                         break;
 410                 __rtnl_unlock();
 411
 412                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 413         }
 414         remove_wait_queue(&netdev_unregistering_wq, &wait);
 415 }
 416
 417 /**
 418  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 419  * @ops: struct rtnl_link_ops * to unregister
 420  */
 421 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 422 {
 423         /* Close the race with cleanup_net() */
 424         mutex_lock(&net_mutex);
 425         rtnl_lock_unregistering_all();
 426         __rtnl_link_unregister(ops);
 427         rtnl_unlock();
 428         mutex_unlock(&net_mutex);
 429 }
 430 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 431
 432 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 433 {
 434         struct net_device *master_dev;
 435         const struct rtnl_link_ops *ops;
 436
 437         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 438         if (!master_dev)
 439                 return 0;
 440         ops = master_dev->rtnl_link_ops;
 441         if (!ops || !ops->get_slave_size)
 442                 return 0;
 443         /* IFLA_INFO_SLAVE_DATA + nested data */
 444         return nla_total_size(sizeof(struct nlattr)) +
 445                ops->get_slave_size(master_dev, dev);
 446 }
 447
 448 static size_t rtnl_link_get_size(const struct net_device *dev)
 449 {
 450         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 451         size_t size;
 452
 453         if (!ops)
 454                 return 0;
 455
 456         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 457                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 458
 459         if (ops->get_size)
 460                 /* IFLA_INFO_DATA + nested data */
 461                 size += nla_total_size(sizeof(struct nlattr)) +
 462                         ops->get_size(dev);
 463
 464         if (ops->get_xstats_size)
 465                 /* IFLA_INFO_XSTATS */
 466                 size += nla_total_size(ops->get_xstats_size(dev));
 467
 468         size += rtnl_link_get_slave_info_data_size(dev);
 469
 470         return size;
 471 }
 472
 473 static LIST_HEAD(rtnl_af_ops);
 474
 475 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 476 {
 477         const struct rtnl_af_ops *ops;
 478
 479         list_for_each_entry(ops, &rtnl_af_ops, list) {
 480                 if (ops->family == family)
 481                         return ops;
 482         }
 483
 484         return NULL;
 485 }
 486
 487 /**
 488  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 489  * @ops: struct rtnl_af_ops * to register
 490  *
 491  * Returns 0 on success or a negative error code.
 492  */
 493 void rtnl_af_register(struct rtnl_af_ops *ops)
 494 {
 495         rtnl_lock();
 496         list_add_tail(&ops->list, &rtnl_af_ops);
 497         rtnl_unlock();
 498 }
 499 EXPORT_SYMBOL_GPL(rtnl_af_register);
 500
 501 /**
 502  * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 503  * @ops: struct rtnl_af_ops * to unregister
 504  *
 505  * The caller must hold the rtnl_mutex.
 506  */
 507 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
 508 {
 509         list_del(&ops->list);
 510 }
 511 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
 512
 513 /**
 514  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 515  * @ops: struct rtnl_af_ops * to unregister
 516  */
 517 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 518 {
 519         rtnl_lock();
 520         __rtnl_af_unregister(ops);
 521         rtnl_unlock();
 522 }
 523 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 524
 525 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 526                                     u32 ext_filter_mask)
 527 {
 528         struct rtnl_af_ops *af_ops;
 529         size_t size;
 530
 531         /* IFLA_AF_SPEC */
 532         size = nla_total_size(sizeof(struct nlattr));
 533
 534         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
 535                 if (af_ops->get_link_af_size) {
 536                         /* AF_* + nested data */
 537                         size += nla_total_size(sizeof(struct nlattr)) +
 538                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 539                 }
 540         }
 541
 542         return size;
 543 }
 544
 545 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 546 {
 547         struct net_device *master_dev;
 548
 549         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 550         if (master_dev && master_dev->rtnl_link_ops)
 551                 return true;
 552         return false;
 553 }
 554
 555 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 556                                      const struct net_device *dev)
 557 {
 558         struct net_device *master_dev;
 559         const struct rtnl_link_ops *ops;
 560         struct nlattr *slave_data;
 561         int err;
 562
 563         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 564         if (!master_dev)
 565                 return 0;
 566         ops = master_dev->rtnl_link_ops;
 567         if (!ops)
 568                 return 0;
 569         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 570                 return -EMSGSIZE;
 571         if (ops->fill_slave_info) {
 572                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 573                 if (!slave_data)
 574                         return -EMSGSIZE;
 575                 err = ops->fill_slave_info(skb, master_dev, dev);
 576                 if (err < 0)
 577                         goto err_cancel_slave_data;
 578                 nla_nest_end(skb, slave_data);
 579         }
 580         return 0;
 581
 582 err_cancel_slave_data:
 583         nla_nest_cancel(skb, slave_data);
 584         return err;
 585 }
 586
 587 static int rtnl_link_info_fill(struct sk_buff *skb,
 588                                const struct net_device *dev)
 589 {
 590         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 591         struct nlattr *data;
 592         int err;
 593
 594         if (!ops)
 595                 return 0;
 596         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 597                 return -EMSGSIZE;
 598         if (ops->fill_xstats) {
 599                 err = ops->fill_xstats(skb, dev);
 600                 if (err < 0)
 601                         return err;
 602         }
 603         if (ops->fill_info) {
 604                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 605                 if (data == NULL)
 606                         return -EMSGSIZE;
 607                 err = ops->fill_info(skb, dev);
 608                 if (err < 0)
 609                         goto err_cancel_data;
 610                 nla_nest_end(skb, data);
 611         }
 612         return 0;
 613
 614 err_cancel_data:
 615         nla_nest_cancel(skb, data);
 616         return err;
 617 }
 618
 619 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 620 {
 621         struct nlattr *linkinfo;
 622         int err = -EMSGSIZE;
 623
 624         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 625         if (linkinfo == NULL)
 626                 goto out;
 627
 628         err = rtnl_link_info_fill(skb, dev);
 629         if (err < 0)
 630                 goto err_cancel_link;
 631
 632         err = rtnl_link_slave_info_fill(skb, dev);
 633         if (err < 0)
 634                 goto err_cancel_link;
 635
 636         nla_nest_end(skb, linkinfo);
 637         return 0;
 638
 639 err_cancel_link:
 640         nla_nest_cancel(skb, linkinfo);
 641 out:
 642         return err;
 643 }
 644
 645 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 646 {
 647         struct sock *rtnl = net->rtnl;
 648         int err = 0;
 649
 650         NETLINK_CB(skb).dst_group = group;
 651         if (echo)
 652                 refcount_inc(&skb->users);
 653         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 654         if (echo)
 655                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 656         return err;
 657 }
 658
 659 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 660 {
 661         struct sock *rtnl = net->rtnl;
 662
 663         return nlmsg_unicast(rtnl, skb, pid);
 664 }
 665 EXPORT_SYMBOL(rtnl_unicast);
 666
 667 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 668                  struct nlmsghdr *nlh, gfp_t flags)
 669 {
 670         struct sock *rtnl = net->rtnl;
 671         int report = 0;
 672
 673         if (nlh)
 674                 report = nlmsg_report(nlh);
 675
 676         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 677 }
 678 EXPORT_SYMBOL(rtnl_notify);
 679
 680 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 681 {
 682         struct sock *rtnl = net->rtnl;
 683
 684         netlink_set_err(rtnl, 0, group, error);
 685 }
 686 EXPORT_SYMBOL(rtnl_set_sk_err);
 687
 688 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 689 {
 690         struct nlattr *mx;
 691         int i, valid = 0;
 692
 693         mx = nla_nest_start(skb, RTA_METRICS);
 694         if (mx == NULL)
 695                 return -ENOBUFS;
 696
 697         for (i = 0; i < RTAX_MAX; i++) {
 698                 if (metrics[i]) {
 699                         if (i == RTAX_CC_ALGO - 1) {
 700                                 char tmp[TCP_CA_NAME_MAX], *name;
 701
 702                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 703                                 if (!name)
 704                                         continue;
 705                                 if (nla_put_string(skb, i + 1, name))
 706                                         goto nla_put_failure;
 707                         } else if (i == RTAX_FEATURES - 1) {
 708                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 709
 710                                 if (!user_features)
 711                                         continue;
 712                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 713                                 if (nla_put_u32(skb, i + 1, user_features))
 714                                         goto nla_put_failure;
 715                         } else {
 716                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 717                                         goto nla_put_failure;
 718                         }
 719                         valid++;
 720                 }
 721         }
 722
 723         if (!valid) {
 724                 nla_nest_cancel(skb, mx);
 725                 return 0;
 726         }
 727
 728         return nla_nest_end(skb, mx);
 729
 730 nla_put_failure:
 731         nla_nest_cancel(skb, mx);
 732         return -EMSGSIZE;
 733 }
 734 EXPORT_SYMBOL(rtnetlink_put_metrics);
 735
 736 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 737                        long expires, u32 error)
 738 {
 739         struct rta_cacheinfo ci = {
 740                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 741                 .rta_used = dst->__use,
 742                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 743                 .rta_error = error,
 744                 .rta_id =  id,
 745         };
 746
 747         if (expires) {
 748                 unsigned long clock;
 749
 750                 clock = jiffies_to_clock_t(abs(expires));
 751                 clock = min_t(unsigned long, clock, INT_MAX);
 752                 ci.rta_expires = (expires > 0) ? clock : -clock;
 753         }
 754         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 755 }
 756 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 757
 758 static void set_operstate(struct net_device *dev, unsigned char transition)
 759 {
 760         unsigned char operstate = dev->operstate;
 761
 762         switch (transition) {
 763         case IF_OPER_UP:
 764                 if ((operstate == IF_OPER_DORMANT ||
 765                      operstate == IF_OPER_UNKNOWN) &&
 766                     !netif_dormant(dev))
 767                         operstate = IF_OPER_UP;
 768                 break;
 769
 770         case IF_OPER_DORMANT:
 771                 if (operstate == IF_OPER_UP ||
 772                     operstate == IF_OPER_UNKNOWN)
 773                         operstate = IF_OPER_DORMANT;
 774                 break;
 775         }
 776
 777         if (dev->operstate != operstate) {
 778                 write_lock_bh(&dev_base_lock);
 779                 dev->operstate = operstate;
 780                 write_unlock_bh(&dev_base_lock);
 781                 netdev_state_change(dev);
 782         }
 783 }
 784
 785 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 786 {
 787         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 788                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 789 }
 790
 791 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 792                                            const struct ifinfomsg *ifm)
 793 {
 794         unsigned int flags = ifm->ifi_flags;
 795
 796         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 797         if (ifm->ifi_change)
 798                 flags = (flags & ifm->ifi_change) |
 799                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 800
 801         return flags;
 802 }
 803
 804 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 805                                  const struct rtnl_link_stats64 *b)
 806 {
 807         a->rx_packets = b->rx_packets;
 808         a->tx_packets = b->tx_packets;
 809         a->rx_bytes = b->rx_bytes;
 810         a->tx_bytes = b->tx_bytes;
 811         a->rx_errors = b->rx_errors;
 812         a->tx_errors = b->tx_errors;
 813         a->rx_dropped = b->rx_dropped;
 814         a->tx_dropped = b->tx_dropped;
 815
 816         a->multicast = b->multicast;
 817         a->collisions = b->collisions;
 818
 819         a->rx_length_errors = b->rx_length_errors;
 820         a->rx_over_errors = b->rx_over_errors;
 821         a->rx_crc_errors = b->rx_crc_errors;
 822         a->rx_frame_errors = b->rx_frame_errors;
 823         a->rx_fifo_errors = b->rx_fifo_errors;
 824         a->rx_missed_errors = b->rx_missed_errors;
 825
 826         a->tx_aborted_errors = b->tx_aborted_errors;
 827         a->tx_carrier_errors = b->tx_carrier_errors;
 828         a->tx_fifo_errors = b->tx_fifo_errors;
 829         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 830         a->tx_window_errors = b->tx_window_errors;
 831
 832         a->rx_compressed = b->rx_compressed;
 833         a->tx_compressed = b->tx_compressed;
 834
 835         a->rx_nohandler = b->rx_nohandler;
 836 }
 837
 838 /* All VF info */
 839 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 840                                    u32 ext_filter_mask)
 841 {
 842         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 843                 int num_vfs = dev_num_vf(dev->dev.parent);
 844                 size_t size = nla_total_size(0);
 845                 size += num_vfs *
 846                         (nla_total_size(0) +
 847                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 848                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 849                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 850                          nla_total_size(MAX_VLAN_LIST_LEN *
 851                                         sizeof(struct ifla_vf_vlan_info)) +
 852                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 853                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 854                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 855                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 856                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 857                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 858                          /* IFLA_VF_STATS_RX_PACKETS */
 859                          nla_total_size_64bit(sizeof(__u64)) +
 860                          /* IFLA_VF_STATS_TX_PACKETS */
 861                          nla_total_size_64bit(sizeof(__u64)) +
 862                          /* IFLA_VF_STATS_RX_BYTES */
 863                          nla_total_size_64bit(sizeof(__u64)) +
 864                          /* IFLA_VF_STATS_TX_BYTES */
 865                          nla_total_size_64bit(sizeof(__u64)) +
 866                          /* IFLA_VF_STATS_BROADCAST */
 867                          nla_total_size_64bit(sizeof(__u64)) +
 868                          /* IFLA_VF_STATS_MULTICAST */
 869                          nla_total_size_64bit(sizeof(__u64)) +
 870                          nla_total_size(sizeof(struct ifla_vf_trust)));
 871                 return size;
 872         } else
 873                 return 0;
 874 }
 875
 876 static size_t rtnl_port_size(const struct net_device *dev,
 877                              u32 ext_filter_mask)
 878 {
 879         size_t port_size = nla_total_size(4)            /* PORT_VF */
 880                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 881                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 882                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 883                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 884                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 885         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 886         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 887                 + port_size;
 888         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 889                 + port_size;
 890
 891         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 892             !(ext_filter_mask & RTEXT_FILTER_VF))
 893                 return 0;
 894         if (dev_num_vf(dev->dev.parent))
 895                 return port_self_size + vf_ports_size +
 896                         vf_port_size * dev_num_vf(dev->dev.parent);
 897         else
 898                 return port_self_size;
 899 }
 900
 901 static size_t rtnl_xdp_size(void)
 902 {
 903         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 904                           nla_total_size(1) +   /* XDP_ATTACHED */
 905                           nla_total_size(4);    /* XDP_PROG_ID */
 906
 907         return xdp_size;
 908 }
 909
 910 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 911                                      u32 ext_filter_mask)
 912 {
 913         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 914                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 915                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 916                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 917                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 918                + nla_total_size(sizeof(struct rtnl_link_stats))
 919                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 920                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 921                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 922                + nla_total_size(4) /* IFLA_TXQLEN */
 923                + nla_total_size(4) /* IFLA_WEIGHT */
 924                + nla_total_size(4) /* IFLA_MTU */
 925                + nla_total_size(4) /* IFLA_LINK */
 926                + nla_total_size(4) /* IFLA_MASTER */
 927                + nla_total_size(1) /* IFLA_CARRIER */
 928                + nla_total_size(4) /* IFLA_PROMISCUITY */
 929                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 930                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 931                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 932                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 933                + nla_total_size(1) /* IFLA_OPERSTATE */
 934                + nla_total_size(1) /* IFLA_LINKMODE */
 935                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 936                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 937                + nla_total_size(4) /* IFLA_GROUP */
 938                + nla_total_size(ext_filter_mask
 939                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 940                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 941                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 942                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 943                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 944                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 945                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 946                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 947                + rtnl_xdp_size() /* IFLA_XDP */
 948                + nla_total_size(4)  /* IFLA_EVENT */
 949                + nla_total_size(1); /* IFLA_PROTO_DOWN */
 950
 951 }
 952
 953 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
 954 {
 955         struct nlattr *vf_ports;
 956         struct nlattr *vf_port;
 957         int vf;
 958         int err;
 959
 960         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
 961         if (!vf_ports)
 962                 return -EMSGSIZE;
 963
 964         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
 965                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
 966                 if (!vf_port)
 967                         goto nla_put_failure;
 968                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
 969                         goto nla_put_failure;
 970                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
 971                 if (err == -EMSGSIZE)
 972                         goto nla_put_failure;
 973                 if (err) {
 974                         nla_nest_cancel(skb, vf_port);
 975                         continue;
 976                 }
 977                 nla_nest_end(skb, vf_port);
 978         }
 979
 980         nla_nest_end(skb, vf_ports);
 981
 982         return 0;
 983
 984 nla_put_failure:
 985         nla_nest_cancel(skb, vf_ports);
 986         return -EMSGSIZE;
 987 }
 988
 989 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
 990 {
 991         struct nlattr *port_self;
 992         int err;
 993
 994         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
 995         if (!port_self)
 996                 return -EMSGSIZE;
 997
 998         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
 999         if (err) {
1000                 nla_nest_cancel(skb, port_self);
1001                 return (err == -EMSGSIZE) ? err : 0;
1002         }
1003
1004         nla_nest_end(skb, port_self);
1005
1006         return 0;
1007 }
1008
1009 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1010                           u32 ext_filter_mask)
1011 {
1012         int err;
1013
1014         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1015             !(ext_filter_mask & RTEXT_FILTER_VF))
1016                 return 0;
1017
1018         err = rtnl_port_self_fill(skb, dev);
1019         if (err)
1020                 return err;
1021
1022         if (dev_num_vf(dev->dev.parent)) {
1023                 err = rtnl_vf_ports_fill(skb, dev);
1024                 if (err)
1025                         return err;
1026         }
1027
1028         return 0;
1029 }
1030
1031 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 {
1033         int err;
1034         struct netdev_phys_item_id ppid;
1035
1036         err = dev_get_phys_port_id(dev, &ppid);
1037         if (err) {
1038                 if (err == -EOPNOTSUPP)
1039                         return 0;
1040                 return err;
1041         }
1042
1043         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1044                 return -EMSGSIZE;
1045
1046         return 0;
1047 }
1048
1049 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1050 {
1051         char name[IFNAMSIZ];
1052         int err;
1053
1054         err = dev_get_phys_port_name(dev, name, sizeof(name));
1055         if (err) {
1056                 if (err == -EOPNOTSUPP)
1057                         return 0;
1058                 return err;
1059         }
1060
1061         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1062                 return -EMSGSIZE;
1063
1064         return 0;
1065 }
1066
1067 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1068 {
1069         int err;
1070         struct switchdev_attr attr = {
1071                 .orig_dev = dev,
1072                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1073                 .flags = SWITCHDEV_F_NO_RECURSE,
1074         };
1075
1076         err = switchdev_port_attr_get(dev, &attr);
1077         if (err) {
1078                 if (err == -EOPNOTSUPP)
1079                         return 0;
1080                 return err;
1081         }
1082
1083         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1084                     attr.u.ppid.id))
1085                 return -EMSGSIZE;
1086
1087         return 0;
1088 }
1089
1090 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1091                                               struct net_device *dev)
1092 {
1093         struct rtnl_link_stats64 *sp;
1094         struct nlattr *attr;
1095
1096         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1097                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1098         if (!attr)
1099                 return -EMSGSIZE;
1100
1101         sp = nla_data(attr);
1102         dev_get_stats(dev, sp);
1103
1104         attr = nla_reserve(skb, IFLA_STATS,
1105                            sizeof(struct rtnl_link_stats));
1106         if (!attr)
1107                 return -EMSGSIZE;
1108
1109         copy_rtnl_link_stats(nla_data(attr), sp);
1110
1111         return 0;
1112 }
1113
1114 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1115                                                struct net_device *dev,
1116                                                int vfs_num,
1117                                                struct nlattr *vfinfo)
1118 {
1119         struct ifla_vf_rss_query_en vf_rss_query_en;
1120         struct nlattr *vf, *vfstats, *vfvlanlist;
1121         struct ifla_vf_link_state vf_linkstate;
1122         struct ifla_vf_vlan_info vf_vlan_info;
1123         struct ifla_vf_spoofchk vf_spoofchk;
1124         struct ifla_vf_tx_rate vf_tx_rate;
1125         struct ifla_vf_stats vf_stats;
1126         struct ifla_vf_trust vf_trust;
1127         struct ifla_vf_vlan vf_vlan;
1128         struct ifla_vf_rate vf_rate;
1129         struct ifla_vf_mac vf_mac;
1130         struct ifla_vf_info ivi;
1131
1132         memset(&ivi, 0, sizeof(ivi));
1133
1134         /* Not all SR-IOV capable drivers support the
1135          * spoofcheck and "RSS query enable" query.  Preset to
1136          * -1 so the user space tool can detect that the driver
1137          * didn't report anything.
1138          */
1139         ivi.spoofchk = -1;
1140         ivi.rss_query_en = -1;
1141         ivi.trusted = -1;
1142         /* The default value for VF link state is "auto"
1143          * IFLA_VF_LINK_STATE_AUTO which equals zero
1144          */
1145         ivi.linkstate = 0;
1146         /* VLAN Protocol by default is 802.1Q */
1147         ivi.vlan_proto = htons(ETH_P_8021Q);
1148         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1149                 return 0;
1150
1151         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1152
1153         vf_mac.vf =
1154                 vf_vlan.vf =
1155                 vf_vlan_info.vf =
1156                 vf_rate.vf =
1157                 vf_tx_rate.vf =
1158                 vf_spoofchk.vf =
1159                 vf_linkstate.vf =
1160                 vf_rss_query_en.vf =
1161                 vf_trust.vf = ivi.vf;
1162
1163         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1164         vf_vlan.vlan = ivi.vlan;
1165         vf_vlan.qos = ivi.qos;
1166         vf_vlan_info.vlan = ivi.vlan;
1167         vf_vlan_info.qos = ivi.qos;
1168         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1169         vf_tx_rate.rate = ivi.max_tx_rate;
1170         vf_rate.min_tx_rate = ivi.min_tx_rate;
1171         vf_rate.max_tx_rate = ivi.max_tx_rate;
1172         vf_spoofchk.setting = ivi.spoofchk;
1173         vf_linkstate.link_state = ivi.linkstate;
1174         vf_rss_query_en.setting = ivi.rss_query_en;
1175         vf_trust.setting = ivi.trusted;
1176         vf = nla_nest_start(skb, IFLA_VF_INFO);
1177         if (!vf)
1178                 goto nla_put_vfinfo_failure;
1179         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1180             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1181             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1182                     &vf_rate) ||
1183             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1184                     &vf_tx_rate) ||
1185             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1186                     &vf_spoofchk) ||
1187             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1188                     &vf_linkstate) ||
1189             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1190                     sizeof(vf_rss_query_en),
1191                     &vf_rss_query_en) ||
1192             nla_put(skb, IFLA_VF_TRUST,
1193                     sizeof(vf_trust), &vf_trust))
1194                 goto nla_put_vf_failure;
1195         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1196         if (!vfvlanlist)
1197                 goto nla_put_vf_failure;
1198         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1199                     &vf_vlan_info)) {
1200                 nla_nest_cancel(skb, vfvlanlist);
1201                 goto nla_put_vf_failure;
1202         }
1203         nla_nest_end(skb, vfvlanlist);
1204         memset(&vf_stats, 0, sizeof(vf_stats));
1205         if (dev->netdev_ops->ndo_get_vf_stats)
1206                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1207                                                 &vf_stats);
1208         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1209         if (!vfstats)
1210                 goto nla_put_vf_failure;
1211         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1212                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1213             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1214                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1215             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1216                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1217             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1218                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1219             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1220                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1221             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1222                               vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1223                 nla_nest_cancel(skb, vfstats);
1224                 goto nla_put_vf_failure;
1225         }
1226         nla_nest_end(skb, vfstats);
1227         nla_nest_end(skb, vf);
1228         return 0;
1229
1230 nla_put_vf_failure:
1231         nla_nest_cancel(skb, vf);
1232 nla_put_vfinfo_failure:
1233         nla_nest_cancel(skb, vfinfo);
1234         return -EMSGSIZE;
1235 }
1236
1237 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1238 {
1239         struct rtnl_link_ifmap map;
1240
1241         memset(&map, 0, sizeof(map));
1242         map.mem_start   = dev->mem_start;
1243         map.mem_end     = dev->mem_end;
1244         map.base_addr   = dev->base_addr;
1245         map.irq         = dev->irq;
1246         map.dma         = dev->dma;
1247         map.port        = dev->if_port;
1248
1249         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1250                 return -EMSGSIZE;
1251
1252         return 0;
1253 }
1254
1255 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1256 {
1257         const struct net_device_ops *ops = dev->netdev_ops;
1258         const struct bpf_prog *generic_xdp_prog;
1259
1260         ASSERT_RTNL();
1261
1262         *prog_id = 0;
1263         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1264         if (generic_xdp_prog) {
1265                 *prog_id = generic_xdp_prog->aux->id;
1266                 return XDP_ATTACHED_SKB;
1267         }
1268         if (!ops->ndo_xdp)
1269                 return XDP_ATTACHED_NONE;
1270
1271         return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1272 }
1273
1274 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1275 {
1276         struct nlattr *xdp;
1277         u32 prog_id;
1278         int err;
1279
1280         xdp = nla_nest_start(skb, IFLA_XDP);
1281         if (!xdp)
1282                 return -EMSGSIZE;
1283
1284         err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1285                          rtnl_xdp_attached_mode(dev, &prog_id));
1286         if (err)
1287                 goto err_cancel;
1288
1289         if (prog_id) {
1290                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1291                 if (err)
1292                         goto err_cancel;
1293         }
1294
1295         nla_nest_end(skb, xdp);
1296         return 0;
1297
1298 err_cancel:
1299         nla_nest_cancel(skb, xdp);
1300         return err;
1301 }
1302
1303 static u32 rtnl_get_event(unsigned long event)
1304 {
1305         u32 rtnl_event_type = IFLA_EVENT_NONE;
1306
1307         switch (event) {
1308         case NETDEV_REBOOT:
1309                 rtnl_event_type = IFLA_EVENT_REBOOT;
1310                 break;
1311         case NETDEV_FEAT_CHANGE:
1312                 rtnl_event_type = IFLA_EVENT_FEATURES;
1313                 break;
1314         case NETDEV_BONDING_FAILOVER:
1315                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1316                 break;
1317         case NETDEV_NOTIFY_PEERS:
1318                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1319                 break;
1320         case NETDEV_RESEND_IGMP:
1321                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1322                 break;
1323         case NETDEV_CHANGEINFODATA:
1324                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1325                 break;
1326         default:
1327                 break;
1328         }
1329
1330         return rtnl_event_type;
1331 }
1332
1333 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1334                             int type, u32 pid, u32 seq, u32 change,
1335                             unsigned int flags, u32 ext_filter_mask,
1336                             u32 event)
1337 {
1338         struct ifinfomsg *ifm;
1339         struct nlmsghdr *nlh;
1340         struct nlattr *af_spec;
1341         struct rtnl_af_ops *af_ops;
1342         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1343
1344         ASSERT_RTNL();
1345         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1346         if (nlh == NULL)
1347                 return -EMSGSIZE;
1348
1349         ifm = nlmsg_data(nlh);
1350         ifm->ifi_family = AF_UNSPEC;
1351         ifm->__ifi_pad = 0;
1352         ifm->ifi_type = dev->type;
1353         ifm->ifi_index = dev->ifindex;
1354         ifm->ifi_flags = dev_get_flags(dev);
1355         ifm->ifi_change = change;
1356
1357         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1358             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1359             nla_put_u8(skb, IFLA_OPERSTATE,
1360                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1361             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1362             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1363             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1364             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1365             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1366             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1367             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1368 #ifdef CONFIG_RPS
1369             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1370 #endif
1371             (dev->ifindex != dev_get_iflink(dev) &&
1372              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1373             (upper_dev &&
1374              nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1375             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1376             (dev->qdisc &&
1377              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1378             (dev->ifalias &&
1379              nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1380             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1381                         atomic_read(&dev->carrier_changes)) ||
1382             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1383                 goto nla_put_failure;
1384
1385         if (event != IFLA_EVENT_NONE) {
1386                 if (nla_put_u32(skb, IFLA_EVENT, event))
1387                         goto nla_put_failure;
1388         }
1389
1390         if (rtnl_fill_link_ifmap(skb, dev))
1391                 goto nla_put_failure;
1392
1393         if (dev->addr_len) {
1394                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1395                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1396                         goto nla_put_failure;
1397         }
1398
1399         if (rtnl_phys_port_id_fill(skb, dev))
1400                 goto nla_put_failure;
1401
1402         if (rtnl_phys_port_name_fill(skb, dev))
1403                 goto nla_put_failure;
1404
1405         if (rtnl_phys_switch_id_fill(skb, dev))
1406                 goto nla_put_failure;
1407
1408         if (rtnl_fill_stats(skb, dev))
1409                 goto nla_put_failure;
1410
1411         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1412             nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1413                 goto nla_put_failure;
1414
1415         if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1416             ext_filter_mask & RTEXT_FILTER_VF) {
1417                 int i;
1418                 struct nlattr *vfinfo;
1419                 int num_vfs = dev_num_vf(dev->dev.parent);
1420
1421                 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1422                 if (!vfinfo)
1423                         goto nla_put_failure;
1424                 for (i = 0; i < num_vfs; i++) {
1425                         if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1426                                 goto nla_put_failure;
1427                 }
1428
1429                 nla_nest_end(skb, vfinfo);
1430         }
1431
1432         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1433                 goto nla_put_failure;
1434
1435         if (rtnl_xdp_fill(skb, dev))
1436                 goto nla_put_failure;
1437
1438         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1439                 if (rtnl_link_fill(skb, dev) < 0)
1440                         goto nla_put_failure;
1441         }
1442
1443         if (dev->rtnl_link_ops &&
1444             dev->rtnl_link_ops->get_link_net) {
1445                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1446
1447                 if (!net_eq(dev_net(dev), link_net)) {
1448                         int id = peernet2id_alloc(dev_net(dev), link_net);
1449
1450                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1451                                 goto nla_put_failure;
1452                 }
1453         }
1454
1455         if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1456                 goto nla_put_failure;
1457
1458         list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1459                 if (af_ops->fill_link_af) {
1460                         struct nlattr *af;
1461                         int err;
1462
1463                         if (!(af = nla_nest_start(skb, af_ops->family)))
1464                                 goto nla_put_failure;
1465
1466                         err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1467
1468                         /*
1469                          * Caller may return ENODATA to indicate that there
1470                          * was no data to be dumped. This is not an error, it
1471                          * means we should trim the attribute header and
1472                          * continue.
1473                          */
1474                         if (err == -ENODATA)
1475                                 nla_nest_cancel(skb, af);
1476                         else if (err < 0)
1477                                 goto nla_put_failure;
1478
1479                         nla_nest_end(skb, af);
1480                 }
1481         }
1482
1483         nla_nest_end(skb, af_spec);
1484
1485         nlmsg_end(skb, nlh);
1486         return 0;
1487
1488 nla_put_failure:
1489         nlmsg_cancel(skb, nlh);
1490         return -EMSGSIZE;
1491 }
1492
1493 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1494         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1495         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1496         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1497         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1498         [IFLA_MTU]              = { .type = NLA_U32 },
1499         [IFLA_LINK]             = { .type = NLA_U32 },
1500         [IFLA_MASTER]           = { .type = NLA_U32 },
1501         [IFLA_CARRIER]          = { .type = NLA_U8 },
1502         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1503         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1504         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1505         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1506         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1507         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1508         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1509         [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
1510         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1511         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1512         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1513         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1514         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1515         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1516         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1517         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1518         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1519         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1520         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1521         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1522         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1523         [IFLA_XDP]              = { .type = NLA_NESTED },
1524         [IFLA_EVENT]            = { .type = NLA_U32 },
1525         [IFLA_GROUP]            = { .type = NLA_U32 },
1526 };
1527
1528 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1529         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1530         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1531         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1532         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1533 };
1534
1535 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1536         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1537         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1538         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1539         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1540         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1541         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1542         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1543         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1544         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1545         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1546         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1547         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1548 };
1549
1550 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1551         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1552         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1553                                     .len = PORT_PROFILE_MAX },
1554         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1555                                       .len = PORT_UUID_MAX },
1556         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1557                                     .len = PORT_UUID_MAX },
1558         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1559         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1560
1561         /* Unused, but we need to keep it here since user space could
1562          * fill it. It's also broken with regard to NLA_BINARY use in
1563          * combination with structs.
1564          */
1565         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1566                                     .len = sizeof(struct ifla_port_vsi) },
1567 };
1568
1569 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1570         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1571         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1572         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1573         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1574 };
1575
1576 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1577 {
1578         const struct rtnl_link_ops *ops = NULL;
1579         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1580
1581         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1582                              ifla_info_policy, NULL) < 0)
1583                 return NULL;
1584
1585         if (linfo[IFLA_INFO_KIND]) {
1586                 char kind[MODULE_NAME_LEN];
1587
1588                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1589                 ops = rtnl_link_ops_get(kind);
1590         }
1591
1592         return ops;
1593 }
1594
1595 static bool link_master_filtered(struct net_device *dev, int master_idx)
1596 {
1597         struct net_device *master;
1598
1599         if (!master_idx)
1600                 return false;
1601
1602         master = netdev_master_upper_dev_get(dev);
1603         if (!master || master->ifindex != master_idx)
1604                 return true;
1605
1606         return false;
1607 }
1608
1609 static bool link_kind_filtered(const struct net_device *dev,
1610                                const struct rtnl_link_ops *kind_ops)
1611 {
1612         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1613                 return true;
1614
1615         return false;
1616 }
1617
1618 static bool link_dump_filtered(struct net_device *dev,
1619                                int master_idx,
1620                                const struct rtnl_link_ops *kind_ops)
1621 {
1622         if (link_master_filtered(dev, master_idx) ||
1623             link_kind_filtered(dev, kind_ops))
1624                 return true;
1625
1626         return false;
1627 }
1628
1629 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1630 {
1631         struct net *net = sock_net(skb->sk);
1632         int h, s_h;
1633         int idx = 0, s_idx;
1634         struct net_device *dev;
1635         struct hlist_head *head;
1636         struct nlattr *tb[IFLA_MAX+1];
1637         u32 ext_filter_mask = 0;
1638         const struct rtnl_link_ops *kind_ops = NULL;
1639         unsigned int flags = NLM_F_MULTI;
1640         int master_idx = 0;
1641         int err;
1642         int hdrlen;
1643
1644         s_h = cb->args[0];
1645         s_idx = cb->args[1];
1646
1647         cb->seq = net->dev_base_seq;
1648
1649         /* A hack to preserve kernel<->userspace interface.
1650          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1651          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1652          * what iproute2 < v3.9.0 used.
1653          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1654          * attribute, its netlink message is shorter than struct ifinfomsg.
1655          */
1656         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1657                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1658
1659         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1660                         ifla_policy, NULL) >= 0) {
1661                 if (tb[IFLA_EXT_MASK])
1662                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1663
1664                 if (tb[IFLA_MASTER])
1665                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1666
1667                 if (tb[IFLA_LINKINFO])
1668                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1669
1670                 if (master_idx || kind_ops)
1671                         flags |= NLM_F_DUMP_FILTERED;
1672         }
1673
1674         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1675                 idx = 0;
1676                 head = &net->dev_index_head[h];
1677                 hlist_for_each_entry(dev, head, index_hlist) {
1678                         if (link_dump_filtered(dev, master_idx, kind_ops))
1679                                 goto cont;
1680                         if (idx < s_idx)
1681                                 goto cont;
1682                         err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1683                                                NETLINK_CB(cb->skb).portid,
1684                                                cb->nlh->nlmsg_seq, 0,
1685                                                flags,
1686                                                ext_filter_mask, 0);
1687
1688                         if (err < 0) {
1689                                 if (likely(skb->len))
1690                                         goto out;
1691
1692                                 goto out_err;
1693                         }
1694
1695                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1696 cont:
1697                         idx++;
1698                 }
1699         }
1700 out:
1701         err = skb->len;
1702 out_err:
1703         cb->args[1] = idx;
1704         cb->args[0] = h;
1705
1706         return err;
1707 }
1708
1709 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1710                         struct netlink_ext_ack *exterr)
1711 {
1712         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1713 }
1714 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1715
1716 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1717 {
1718         struct net *net;
1719         /* Examine the link attributes and figure out which
1720          * network namespace we are talking about.
1721          */
1722         if (tb[IFLA_NET_NS_PID])
1723                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1724         else if (tb[IFLA_NET_NS_FD])
1725                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1726         else
1727                 net = get_net(src_net);
1728         return net;
1729 }
1730 EXPORT_SYMBOL(rtnl_link_get_net);
1731
1732 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1733 {
1734         if (dev) {
1735                 if (tb[IFLA_ADDRESS] &&
1736                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1737                         return -EINVAL;
1738
1739                 if (tb[IFLA_BROADCAST] &&
1740                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1741                         return -EINVAL;
1742         }
1743
1744         if (tb[IFLA_AF_SPEC]) {
1745                 struct nlattr *af;
1746                 int rem, err;
1747
1748                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1749                         const struct rtnl_af_ops *af_ops;
1750
1751                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1752                                 return -EAFNOSUPPORT;
1753
1754                         if (!af_ops->set_link_af)
1755                                 return -EOPNOTSUPP;
1756
1757                         if (af_ops->validate_link_af) {
1758                                 err = af_ops->validate_link_af(dev, af);
1759                                 if (err < 0)
1760                                         return err;
1761                         }
1762                 }
1763         }
1764
1765         return 0;
1766 }
1767
1768 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1769                                   int guid_type)
1770 {
1771         const struct net_device_ops *ops = dev->netdev_ops;
1772
1773         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1774 }
1775
1776 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1777 {
1778         if (dev->type != ARPHRD_INFINIBAND)
1779                 return -EOPNOTSUPP;
1780
1781         return handle_infiniband_guid(dev, ivt, guid_type);
1782 }
1783
1784 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1785 {
1786         const struct net_device_ops *ops = dev->netdev_ops;
1787         int err = -EINVAL;
1788
1789         if (tb[IFLA_VF_MAC]) {
1790                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1791
1792                 err = -EOPNOTSUPP;
1793                 if (ops->ndo_set_vf_mac)
1794                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
1795                                                   ivm->mac);
1796                 if (err < 0)
1797                         return err;
1798         }
1799
1800         if (tb[IFLA_VF_VLAN]) {
1801                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1802
1803                 err = -EOPNOTSUPP;
1804                 if (ops->ndo_set_vf_vlan)
1805                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1806                                                    ivv->qos,
1807                                                    htons(ETH_P_8021Q));
1808                 if (err < 0)
1809                         return err;
1810         }
1811
1812         if (tb[IFLA_VF_VLAN_LIST]) {
1813                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1814                 struct nlattr *attr;
1815                 int rem, len = 0;
1816
1817                 err = -EOPNOTSUPP;
1818                 if (!ops->ndo_set_vf_vlan)
1819                         return err;
1820
1821                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1822                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1823                             nla_len(attr) < NLA_HDRLEN) {
1824                                 return -EINVAL;
1825                         }
1826                         if (len >= MAX_VLAN_LIST_LEN)
1827                                 return -EOPNOTSUPP;
1828                         ivvl[len] = nla_data(attr);
1829
1830                         len++;
1831                 }
1832                 if (len == 0)
1833                         return -EINVAL;
1834
1835                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1836                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
1837                 if (err < 0)
1838                         return err;
1839         }
1840
1841         if (tb[IFLA_VF_TX_RATE]) {
1842                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1843                 struct ifla_vf_info ivf;
1844
1845                 err = -EOPNOTSUPP;
1846                 if (ops->ndo_get_vf_config)
1847                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1848                 if (err < 0)
1849                         return err;
1850
1851                 err = -EOPNOTSUPP;
1852                 if (ops->ndo_set_vf_rate)
1853                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1854                                                    ivf.min_tx_rate,
1855                                                    ivt->rate);
1856                 if (err < 0)
1857                         return err;
1858         }
1859
1860         if (tb[IFLA_VF_RATE]) {
1861                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1862
1863                 err = -EOPNOTSUPP;
1864                 if (ops->ndo_set_vf_rate)
1865                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
1866                                                    ivt->min_tx_rate,
1867                                                    ivt->max_tx_rate);
1868                 if (err < 0)
1869                         return err;
1870         }
1871
1872         if (tb[IFLA_VF_SPOOFCHK]) {
1873                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1874
1875                 err = -EOPNOTSUPP;
1876                 if (ops->ndo_set_vf_spoofchk)
1877                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1878                                                        ivs->setting);
1879                 if (err < 0)
1880                         return err;
1881         }
1882
1883         if (tb[IFLA_VF_LINK_STATE]) {
1884                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1885
1886                 err = -EOPNOTSUPP;
1887                 if (ops->ndo_set_vf_link_state)
1888                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1889                                                          ivl->link_state);
1890                 if (err < 0)
1891                         return err;
1892         }
1893
1894         if (tb[IFLA_VF_RSS_QUERY_EN]) {
1895                 struct ifla_vf_rss_query_en *ivrssq_en;
1896
1897                 err = -EOPNOTSUPP;
1898                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1899                 if (ops->ndo_set_vf_rss_query_en)
1900                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1901                                                            ivrssq_en->setting);
1902                 if (err < 0)
1903                         return err;
1904         }
1905
1906         if (tb[IFLA_VF_TRUST]) {
1907                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1908
1909                 err = -EOPNOTSUPP;
1910                 if (ops->ndo_set_vf_trust)
1911                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1912                 if (err < 0)
1913                         return err;
1914         }
1915
1916         if (tb[IFLA_VF_IB_NODE_GUID]) {
1917                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1918
1919                 if (!ops->ndo_set_vf_guid)
1920                         return -EOPNOTSUPP;
1921
1922                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1923         }
1924
1925         if (tb[IFLA_VF_IB_PORT_GUID]) {
1926                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1927
1928                 if (!ops->ndo_set_vf_guid)
1929                         return -EOPNOTSUPP;
1930
1931                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1932         }
1933
1934         return err;
1935 }
1936
1937 static int do_set_master(struct net_device *dev, int ifindex)
1938 {
1939         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1940         const struct net_device_ops *ops;
1941         int err;
1942
1943         if (upper_dev) {
1944                 if (upper_dev->ifindex == ifindex)
1945                         return 0;
1946                 ops = upper_dev->netdev_ops;
1947                 if (ops->ndo_del_slave) {
1948                         err = ops->ndo_del_slave(upper_dev, dev);
1949                         if (err)
1950                                 return err;
1951                 } else {
1952                         return -EOPNOTSUPP;
1953                 }
1954         }
1955
1956         if (ifindex) {
1957                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1958                 if (!upper_dev)
1959                         return -EINVAL;
1960                 ops = upper_dev->netdev_ops;
1961                 if (ops->ndo_add_slave) {
1962                         err = ops->ndo_add_slave(upper_dev, dev);
1963                         if (err)
1964                                 return err;
1965                 } else {
1966                         return -EOPNOTSUPP;
1967                 }
1968         }
1969         return 0;
1970 }
1971
1972 #define DO_SETLINK_MODIFIED     0x01
1973 /* notify flag means notify + modified. */
1974 #define DO_SETLINK_NOTIFY       0x03
1975 static int do_setlink(const struct sk_buff *skb,
1976                       struct net_device *dev, struct ifinfomsg *ifm,
1977                       struct netlink_ext_ack *extack,
1978                       struct nlattr **tb, char *ifname, int status)
1979 {
1980         const struct net_device_ops *ops = dev->netdev_ops;
1981         int err;
1982
1983         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1984                 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1985                 if (IS_ERR(net)) {
1986                         err = PTR_ERR(net);
1987                         goto errout;
1988                 }
1989                 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1990                         put_net(net);
1991                         err = -EPERM;
1992                         goto errout;
1993                 }
1994                 err = dev_change_net_namespace(dev, net, ifname);
1995                 put_net(net);
1996                 if (err)
1997                         goto errout;
1998                 status |= DO_SETLINK_MODIFIED;
1999         }
2000
2001         if (tb[IFLA_MAP]) {
2002                 struct rtnl_link_ifmap *u_map;
2003                 struct ifmap k_map;
2004
2005                 if (!ops->ndo_set_config) {
2006                         err = -EOPNOTSUPP;
2007                         goto errout;
2008                 }
2009
2010                 if (!netif_device_present(dev)) {
2011                         err = -ENODEV;
2012                         goto errout;
2013                 }
2014
2015                 u_map = nla_data(tb[IFLA_MAP]);
2016                 k_map.mem_start = (unsigned long) u_map->mem_start;
2017                 k_map.mem_end = (unsigned long) u_map->mem_end;
2018                 k_map.base_addr = (unsigned short) u_map->base_addr;
2019                 k_map.irq = (unsigned char) u_map->irq;
2020                 k_map.dma = (unsigned char) u_map->dma;
2021                 k_map.port = (unsigned char) u_map->port;
2022
2023                 err = ops->ndo_set_config(dev, &k_map);
2024                 if (err < 0)
2025                         goto errout;
2026
2027                 status |= DO_SETLINK_NOTIFY;
2028         }
2029
2030         if (tb[IFLA_ADDRESS]) {
2031                 struct sockaddr *sa;
2032                 int len;
2033
2034                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2035                                                   sizeof(*sa));
2036                 sa = kmalloc(len, GFP_KERNEL);
2037                 if (!sa) {
2038                         err = -ENOMEM;
2039                         goto errout;
2040                 }
2041                 sa->sa_family = dev->type;
2042                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2043                        dev->addr_len);
2044                 err = dev_set_mac_address(dev, sa);
2045                 kfree(sa);
2046                 if (err)
2047                         goto errout;
2048                 status |= DO_SETLINK_MODIFIED;
2049         }
2050
2051         if (tb[IFLA_MTU]) {
2052                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2053                 if (err < 0)
2054                         goto errout;
2055                 status |= DO_SETLINK_MODIFIED;
2056         }
2057
2058         if (tb[IFLA_GROUP]) {
2059                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2060                 status |= DO_SETLINK_NOTIFY;
2061         }
2062
2063         /*
2064          * Interface selected by interface index but interface
2065          * name provided implies that a name change has been
2066          * requested.
2067          */
2068         if (ifm->ifi_index > 0 && ifname[0]) {
2069                 err = dev_change_name(dev, ifname);
2070                 if (err < 0)
2071                         goto errout;
2072                 status |= DO_SETLINK_MODIFIED;
2073         }
2074
2075         if (tb[IFLA_IFALIAS]) {
2076                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2077                                     nla_len(tb[IFLA_IFALIAS]));
2078                 if (err < 0)
2079                         goto errout;
2080                 status |= DO_SETLINK_NOTIFY;
2081         }
2082
2083         if (tb[IFLA_BROADCAST]) {
2084                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2085                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2086         }
2087
2088         if (ifm->ifi_flags || ifm->ifi_change) {
2089                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2090                 if (err < 0)
2091                         goto errout;
2092         }
2093
2094         if (tb[IFLA_MASTER]) {
2095                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2096                 if (err)
2097                         goto errout;
2098                 status |= DO_SETLINK_MODIFIED;
2099         }
2100
2101         if (tb[IFLA_CARRIER]) {
2102                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2103                 if (err)
2104                         goto errout;
2105                 status |= DO_SETLINK_MODIFIED;
2106         }
2107
2108         if (tb[IFLA_TXQLEN]) {
2109                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2110                 unsigned int orig_len = dev->tx_queue_len;
2111
2112                 if (dev->tx_queue_len ^ value) {
2113                         dev->tx_queue_len = value;
2114                         err = call_netdevice_notifiers(
2115                               NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2116                         err = notifier_to_errno(err);
2117                         if (err) {
2118                                 dev->tx_queue_len = orig_len;
2119                                 goto errout;
2120                         }
2121                         status |= DO_SETLINK_NOTIFY;
2122                 }
2123         }
2124
2125         if (tb[IFLA_OPERSTATE])
2126                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2127
2128         if (tb[IFLA_LINKMODE]) {
2129                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2130
2131                 write_lock_bh(&dev_base_lock);
2132                 if (dev->link_mode ^ value)
2133                         status |= DO_SETLINK_NOTIFY;
2134                 dev->link_mode = value;
2135                 write_unlock_bh(&dev_base_lock);
2136         }
2137
2138         if (tb[IFLA_VFINFO_LIST]) {
2139                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2140                 struct nlattr *attr;
2141                 int rem;
2142
2143                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2144                         if (nla_type(attr) != IFLA_VF_INFO ||
2145                             nla_len(attr) < NLA_HDRLEN) {
2146                                 err = -EINVAL;
2147                                 goto errout;
2148                         }
2149                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2150                                                ifla_vf_policy, NULL);
2151                         if (err < 0)
2152                                 goto errout;
2153                         err = do_setvfinfo(dev, vfinfo);
2154                         if (err < 0)
2155                                 goto errout;
2156                         status |= DO_SETLINK_NOTIFY;
2157                 }
2158         }
2159         err = 0;
2160
2161         if (tb[IFLA_VF_PORTS]) {
2162                 struct nlattr *port[IFLA_PORT_MAX+1];
2163                 struct nlattr *attr;
2164                 int vf;
2165                 int rem;
2166
2167                 err = -EOPNOTSUPP;
2168                 if (!ops->ndo_set_vf_port)
2169                         goto errout;
2170
2171                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2172                         if (nla_type(attr) != IFLA_VF_PORT ||
2173                             nla_len(attr) < NLA_HDRLEN) {
2174                                 err = -EINVAL;
2175                                 goto errout;
2176                         }
2177                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2178                                                ifla_port_policy, NULL);
2179                         if (err < 0)
2180                                 goto errout;
2181                         if (!port[IFLA_PORT_VF]) {
2182                                 err = -EOPNOTSUPP;
2183                                 goto errout;
2184                         }
2185                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2186                         err = ops->ndo_set_vf_port(dev, vf, port);
2187                         if (err < 0)
2188                                 goto errout;
2189                         status |= DO_SETLINK_NOTIFY;
2190                 }
2191         }
2192         err = 0;
2193
2194         if (tb[IFLA_PORT_SELF]) {
2195                 struct nlattr *port[IFLA_PORT_MAX+1];
2196
2197                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2198                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2199                                        NULL);
2200                 if (err < 0)
2201                         goto errout;
2202
2203                 err = -EOPNOTSUPP;
2204                 if (ops->ndo_set_vf_port)
2205                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2206                 if (err < 0)
2207                         goto errout;
2208                 status |= DO_SETLINK_NOTIFY;
2209         }
2210
2211         if (tb[IFLA_AF_SPEC]) {
2212                 struct nlattr *af;
2213                 int rem;
2214
2215                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2216                         const struct rtnl_af_ops *af_ops;
2217
2218                         if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2219                                 BUG();
2220
2221                         err = af_ops->set_link_af(dev, af);
2222                         if (err < 0)
2223                                 goto errout;
2224
2225                         status |= DO_SETLINK_NOTIFY;
2226                 }
2227         }
2228         err = 0;
2229
2230         if (tb[IFLA_PROTO_DOWN]) {
2231                 err = dev_change_proto_down(dev,
2232                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2233                 if (err)
2234                         goto errout;
2235                 status |= DO_SETLINK_NOTIFY;
2236         }
2237
2238         if (tb[IFLA_XDP]) {
2239                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2240                 u32 xdp_flags = 0;
2241
2242                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2243                                        ifla_xdp_policy, NULL);
2244                 if (err < 0)
2245                         goto errout;
2246
2247                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2248                         err = -EINVAL;
2249                         goto errout;
2250                 }
2251
2252                 if (xdp[IFLA_XDP_FLAGS]) {
2253                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2254                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2255                                 err = -EINVAL;
2256                                 goto errout;
2257                         }
2258                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2259                                 err = -EINVAL;
2260                                 goto errout;
2261                         }
2262                 }
2263
2264                 if (xdp[IFLA_XDP_FD]) {
2265                         err = dev_change_xdp_fd(dev, extack,
2266                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2267                                                 xdp_flags);
2268                         if (err)
2269                                 goto errout;
2270                         status |= DO_SETLINK_NOTIFY;
2271                 }
2272         }
2273
2274 errout:
2275         if (status & DO_SETLINK_MODIFIED) {
2276                 if (status & DO_SETLINK_NOTIFY)
2277                         netdev_state_change(dev);
2278
2279                 if (err < 0)
2280                         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",
2281                                              dev->name);
2282         }
2283
2284         return err;
2285 }
2286
2287 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2288                         struct netlink_ext_ack *extack)
2289 {
2290         struct net *net = sock_net(skb->sk);
2291         struct ifinfomsg *ifm;
2292         struct net_device *dev;
2293         int err;
2294         struct nlattr *tb[IFLA_MAX+1];
2295         char ifname[IFNAMSIZ];
2296
2297         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2298                           extack);
2299         if (err < 0)
2300                 goto errout;
2301
2302         if (tb[IFLA_IFNAME])
2303                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2304         else
2305                 ifname[0] = '\0';
2306
2307         err = -EINVAL;
2308         ifm = nlmsg_data(nlh);
2309         if (ifm->ifi_index > 0)
2310                 dev = __dev_get_by_index(net, ifm->ifi_index);
2311         else if (tb[IFLA_IFNAME])
2312                 dev = __dev_get_by_name(net, ifname);
2313         else
2314                 goto errout;
2315
2316         if (dev == NULL) {
2317                 err = -ENODEV;
2318                 goto errout;
2319         }
2320
2321         err = validate_linkmsg(dev, tb);
2322         if (err < 0)
2323                 goto errout;
2324
2325         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2326 errout:
2327         return err;
2328 }
2329
2330 static int rtnl_group_dellink(const struct net *net, int group)
2331 {
2332         struct net_device *dev, *aux;
2333         LIST_HEAD(list_kill);
2334         bool found = false;
2335
2336         if (!group)
2337                 return -EPERM;
2338
2339         for_each_netdev(net, dev) {
2340                 if (dev->group == group) {
2341                         const struct rtnl_link_ops *ops;
2342
2343                         found = true;
2344                         ops = dev->rtnl_link_ops;
2345                         if (!ops || !ops->dellink)
2346                                 return -EOPNOTSUPP;
2347                 }
2348         }
2349
2350         if (!found)
2351                 return -ENODEV;
2352
2353         for_each_netdev_safe(net, dev, aux) {
2354                 if (dev->group == group) {
2355                         const struct rtnl_link_ops *ops;
2356
2357                         ops = dev->rtnl_link_ops;
2358                         ops->dellink(dev, &list_kill);
2359                 }
2360         }
2361         unregister_netdevice_many(&list_kill);
2362
2363         return 0;
2364 }
2365
2366 int rtnl_delete_link(struct net_device *dev)
2367 {
2368         const struct rtnl_link_ops *ops;
2369         LIST_HEAD(list_kill);
2370
2371         ops = dev->rtnl_link_ops;
2372         if (!ops || !ops->dellink)
2373                 return -EOPNOTSUPP;
2374
2375         ops->dellink(dev, &list_kill);
2376         unregister_netdevice_many(&list_kill);
2377
2378         return 0;
2379 }
2380 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2381
2382 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2383                         struct netlink_ext_ack *extack)
2384 {
2385         struct net *net = sock_net(skb->sk);
2386         struct net_device *dev;
2387         struct ifinfomsg *ifm;
2388         char ifname[IFNAMSIZ];
2389         struct nlattr *tb[IFLA_MAX+1];
2390         int err;
2391
2392         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2393         if (err < 0)
2394                 return err;
2395
2396         if (tb[IFLA_IFNAME])
2397                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2398
2399         ifm = nlmsg_data(nlh);
2400         if (ifm->ifi_index > 0)
2401                 dev = __dev_get_by_index(net, ifm->ifi_index);
2402         else if (tb[IFLA_IFNAME])
2403                 dev = __dev_get_by_name(net, ifname);
2404         else if (tb[IFLA_GROUP])
2405                 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2406         else
2407                 return -EINVAL;
2408
2409         if (!dev)
2410                 return -ENODEV;
2411
2412         return rtnl_delete_link(dev);
2413 }
2414
2415 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2416 {
2417         unsigned int old_flags;
2418         int err;
2419
2420         old_flags = dev->flags;
2421         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2422                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2423                 if (err < 0)
2424                         return err;
2425         }
2426
2427         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2428
2429         __dev_notify_flags(dev, old_flags, ~0U);
2430         return 0;
2431 }
2432 EXPORT_SYMBOL(rtnl_configure_link);
2433
2434 struct net_device *rtnl_create_link(struct net *net,
2435         const char *ifname, unsigned char name_assign_type,
2436         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2437 {
2438         struct net_device *dev;
2439         unsigned int num_tx_queues = 1;
2440         unsigned int num_rx_queues = 1;
2441
2442         if (tb[IFLA_NUM_TX_QUEUES])
2443                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2444         else if (ops->get_num_tx_queues)
2445                 num_tx_queues = ops->get_num_tx_queues();
2446
2447         if (tb[IFLA_NUM_RX_QUEUES])
2448                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2449         else if (ops->get_num_rx_queues)
2450                 num_rx_queues = ops->get_num_rx_queues();
2451
2452         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2453                                ops->setup, num_tx_queues, num_rx_queues);
2454         if (!dev)
2455                 return ERR_PTR(-ENOMEM);
2456
2457         dev_net_set(dev, net);
2458         dev->rtnl_link_ops = ops;
2459         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2460
2461         if (tb[IFLA_MTU])
2462                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2463         if (tb[IFLA_ADDRESS]) {
2464                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2465                                 nla_len(tb[IFLA_ADDRESS]));
2466                 dev->addr_assign_type = NET_ADDR_SET;
2467         }
2468         if (tb[IFLA_BROADCAST])
2469                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2470                                 nla_len(tb[IFLA_BROADCAST]));
2471         if (tb[IFLA_TXQLEN])
2472                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2473         if (tb[IFLA_OPERSTATE])
2474                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2475         if (tb[IFLA_LINKMODE])
2476                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2477         if (tb[IFLA_GROUP])
2478                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2479
2480         return dev;
2481 }
2482 EXPORT_SYMBOL(rtnl_create_link);
2483
2484 static int rtnl_group_changelink(const struct sk_buff *skb,
2485                 struct net *net, int group,
2486                 struct ifinfomsg *ifm,
2487                 struct netlink_ext_ack *extack,
2488                 struct nlattr **tb)
2489 {
2490         struct net_device *dev, *aux;
2491         int err;
2492
2493         for_each_netdev_safe(net, dev, aux) {
2494                 if (dev->group == group) {
2495                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2496                         if (err < 0)
2497                                 return err;
2498                 }
2499         }
2500
2501         return 0;
2502 }
2503
2504 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2505                         struct netlink_ext_ack *extack)
2506 {
2507         struct net *net = sock_net(skb->sk);
2508         const struct rtnl_link_ops *ops;
2509         const struct rtnl_link_ops *m_ops = NULL;
2510         struct net_device *dev;
2511         struct net_device *master_dev = NULL;
2512         struct ifinfomsg *ifm;
2513         char kind[MODULE_NAME_LEN];
2514         char ifname[IFNAMSIZ];
2515         struct nlattr *tb[IFLA_MAX+1];
2516         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2517         unsigned char name_assign_type = NET_NAME_USER;
2518         int err;
2519
2520 #ifdef CONFIG_MODULES
2521 replay:
2522 #endif
2523         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2524         if (err < 0)
2525                 return err;
2526
2527         if (tb[IFLA_IFNAME])
2528                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2529         else
2530                 ifname[0] = '\0';
2531
2532         ifm = nlmsg_data(nlh);
2533         if (ifm->ifi_index > 0)
2534                 dev = __dev_get_by_index(net, ifm->ifi_index);
2535         else {
2536                 if (ifname[0])
2537                         dev = __dev_get_by_name(net, ifname);
2538                 else
2539                         dev = NULL;
2540         }
2541
2542         if (dev) {
2543                 master_dev = netdev_master_upper_dev_get(dev);
2544                 if (master_dev)
2545                         m_ops = master_dev->rtnl_link_ops;
2546         }
2547
2548         err = validate_linkmsg(dev, tb);
2549         if (err < 0)
2550                 return err;
2551
2552         if (tb[IFLA_LINKINFO]) {
2553                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2554                                        tb[IFLA_LINKINFO], ifla_info_policy,
2555                                        NULL);
2556                 if (err < 0)
2557                         return err;
2558         } else
2559                 memset(linkinfo, 0, sizeof(linkinfo));
2560
2561         if (linkinfo[IFLA_INFO_KIND]) {
2562                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2563                 ops = rtnl_link_ops_get(kind);
2564         } else {
2565                 kind[0] = '\0';
2566                 ops = NULL;
2567         }
2568
2569         if (1) {
2570                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2571                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2572                 struct nlattr **data = NULL;
2573                 struct nlattr **slave_data = NULL;
2574                 struct net *dest_net, *link_net = NULL;
2575
2576                 if (ops) {
2577                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2578                                 err = nla_parse_nested(attr, ops->maxtype,
2579                                                        linkinfo[IFLA_INFO_DATA],
2580                                                        ops->policy, NULL);
2581                                 if (err < 0)
2582                                         return err;
2583                                 data = attr;
2584                         }
2585                         if (ops->validate) {
2586                                 err = ops->validate(tb, data, extack);
2587                                 if (err < 0)
2588                                         return err;
2589                         }
2590                 }
2591
2592                 if (m_ops) {
2593                         if (m_ops->slave_maxtype &&
2594                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2595                                 err = nla_parse_nested(slave_attr,
2596                                                        m_ops->slave_maxtype,
2597                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2598                                                        m_ops->slave_policy,
2599                                                        NULL);
2600                                 if (err < 0)
2601                                         return err;
2602                                 slave_data = slave_attr;
2603                         }
2604                         if (m_ops->slave_validate) {
2605                                 err = m_ops->slave_validate(tb, slave_data,
2606                                                             extack);
2607                                 if (err < 0)
2608                                         return err;
2609                         }
2610                 }
2611
2612                 if (dev) {
2613                         int status = 0;
2614
2615                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2616                                 return -EEXIST;
2617                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2618                                 return -EOPNOTSUPP;
2619
2620                         if (linkinfo[IFLA_INFO_DATA]) {
2621                                 if (!ops || ops != dev->rtnl_link_ops ||
2622                                     !ops->changelink)
2623                                         return -EOPNOTSUPP;
2624
2625                                 err = ops->changelink(dev, tb, data, extack);
2626                                 if (err < 0)
2627                                         return err;
2628                                 status |= DO_SETLINK_NOTIFY;
2629                         }
2630
2631                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2632                                 if (!m_ops || !m_ops->slave_changelink)
2633                                         return -EOPNOTSUPP;
2634
2635                                 err = m_ops->slave_changelink(master_dev, dev,
2636                                                               tb, slave_data,
2637                                                               extack);
2638                                 if (err < 0)
2639                                         return err;
2640                                 status |= DO_SETLINK_NOTIFY;
2641                         }
2642
2643                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2644                                           status);
2645                 }
2646
2647                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2648                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2649                                 return rtnl_group_changelink(skb, net,
2650                                                 nla_get_u32(tb[IFLA_GROUP]),
2651                                                 ifm, extack, tb);
2652                         return -ENODEV;
2653                 }
2654
2655                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2656                         return -EOPNOTSUPP;
2657
2658                 if (!ops) {
2659 #ifdef CONFIG_MODULES
2660                         if (kind[0]) {
2661                                 __rtnl_unlock();
2662                                 request_module("rtnl-link-%s", kind);
2663                                 rtnl_lock();
2664                                 ops = rtnl_link_ops_get(kind);
2665                                 if (ops)
2666                                         goto replay;
2667                         }
2668 #endif
2669                         return -EOPNOTSUPP;
2670                 }
2671
2672                 if (!ops->setup)
2673                         return -EOPNOTSUPP;
2674
2675                 if (!ifname[0]) {
2676                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2677                         name_assign_type = NET_NAME_ENUM;
2678                 }
2679
2680                 dest_net = rtnl_link_get_net(net, tb);
2681                 if (IS_ERR(dest_net))
2682                         return PTR_ERR(dest_net);
2683
2684                 err = -EPERM;
2685                 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2686                         goto out;
2687
2688                 if (tb[IFLA_LINK_NETNSID]) {
2689                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2690
2691                         link_net = get_net_ns_by_id(dest_net, id);
2692                         if (!link_net) {
2693                                 err =  -EINVAL;
2694                                 goto out;
2695                         }
2696                         err = -EPERM;
2697                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2698                                 goto out;
2699                 }
2700
2701                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2702                                        name_assign_type, ops, tb);
2703                 if (IS_ERR(dev)) {
2704                         err = PTR_ERR(dev);
2705                         goto out;
2706                 }
2707
2708                 dev->ifindex = ifm->ifi_index;
2709
2710                 if (ops->newlink) {
2711                         err = ops->newlink(link_net ? : net, dev, tb, data,
2712                                            extack);
2713                         /* Drivers should call free_netdev() in ->destructor
2714                          * and unregister it on failure after registration
2715                          * so that device could be finally freed in rtnl_unlock.
2716                          */
2717                         if (err < 0) {
2718                                 /* If device is not registered at all, free it now */
2719                                 if (dev->reg_state == NETREG_UNINITIALIZED)
2720                                         free_netdev(dev);
2721                                 goto out;
2722                         }
2723                 } else {
2724                         err = register_netdevice(dev);
2725                         if (err < 0) {
2726                                 free_netdev(dev);
2727                                 goto out;
2728                         }
2729                 }
2730                 err = rtnl_configure_link(dev, ifm);
2731                 if (err < 0)
2732                         goto out_unregister;
2733                 if (link_net) {
2734                         err = dev_change_net_namespace(dev, dest_net, ifname);
2735                         if (err < 0)
2736                                 goto out_unregister;
2737                 }
2738                 if (tb[IFLA_MASTER]) {
2739                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2740                         if (err)
2741                                 goto out_unregister;
2742                 }
2743 out:
2744                 if (link_net)
2745                         put_net(link_net);
2746                 put_net(dest_net);
2747                 return err;
2748 out_unregister:
2749                 if (ops->newlink) {
2750                         LIST_HEAD(list_kill);
2751
2752                         ops->dellink(dev, &list_kill);
2753                         unregister_netdevice_many(&list_kill);
2754                 } else {
2755                         unregister_netdevice(dev);
2756                 }
2757                 goto out;
2758         }
2759 }
2760
2761 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2762                         struct netlink_ext_ack *extack)
2763 {
2764         struct net *net = sock_net(skb->sk);
2765         struct ifinfomsg *ifm;
2766         char ifname[IFNAMSIZ];
2767         struct nlattr *tb[IFLA_MAX+1];
2768         struct net_device *dev = NULL;
2769         struct sk_buff *nskb;
2770         int err;
2771         u32 ext_filter_mask = 0;
2772
2773         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2774         if (err < 0)
2775                 return err;
2776
2777         if (tb[IFLA_IFNAME])
2778                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2779
2780         if (tb[IFLA_EXT_MASK])
2781                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2782
2783         ifm = nlmsg_data(nlh);
2784         if (ifm->ifi_index > 0)
2785                 dev = __dev_get_by_index(net, ifm->ifi_index);
2786         else if (tb[IFLA_IFNAME])
2787                 dev = __dev_get_by_name(net, ifname);
2788         else
2789                 return -EINVAL;
2790
2791         if (dev == NULL)
2792                 return -ENODEV;
2793
2794         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2795         if (nskb == NULL)
2796                 return -ENOBUFS;
2797
2798         err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2799                                nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2800         if (err < 0) {
2801                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2802                 WARN_ON(err == -EMSGSIZE);
2803                 kfree_skb(nskb);
2804         } else
2805                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2806
2807         return err;
2808 }
2809
2810 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2811 {
2812         struct net *net = sock_net(skb->sk);
2813         struct net_device *dev;
2814         struct nlattr *tb[IFLA_MAX+1];
2815         u32 ext_filter_mask = 0;
2816         u16 min_ifinfo_dump_size = 0;
2817         int hdrlen;
2818
2819         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2820         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2821                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2822
2823         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2824                 if (tb[IFLA_EXT_MASK])
2825                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2826         }
2827
2828         if (!ext_filter_mask)
2829                 return NLMSG_GOODSIZE;
2830         /*
2831          * traverse the list of net devices and compute the minimum
2832          * buffer size based upon the filter mask.
2833          */
2834         list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2835                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2836                                              if_nlmsg_size(dev,
2837                                                            ext_filter_mask));
2838         }
2839
2840         return nlmsg_total_size(min_ifinfo_dump_size);
2841 }
2842
2843 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2844 {
2845         int idx;
2846         int s_idx = cb->family;
2847
2848         if (s_idx == 0)
2849                 s_idx = 1;
2850         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2851                 int type = cb->nlh->nlmsg_type-RTM_BASE;
2852                 if (idx < s_idx || idx == PF_PACKET)
2853                         continue;
2854                 if (rtnl_msg_handlers[idx] == NULL ||
2855                     rtnl_msg_handlers[idx][type].dumpit == NULL)
2856                         continue;
2857                 if (idx > s_idx) {
2858                         memset(&cb->args[0], 0, sizeof(cb->args));
2859                         cb->prev_seq = 0;
2860                         cb->seq = 0;
2861                 }
2862                 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2863                         break;
2864         }
2865         cb->family = idx;
2866
2867         return skb->len;
2868 }
2869
2870 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2871                                        unsigned int change,
2872                                        u32 event, gfp_t flags)
2873 {
2874         struct net *net = dev_net(dev);
2875         struct sk_buff *skb;
2876         int err = -ENOBUFS;
2877         size_t if_info_size;
2878
2879         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2880         if (skb == NULL)
2881                 goto errout;
2882
2883         err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2884         if (err < 0) {
2885                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2886                 WARN_ON(err == -EMSGSIZE);
2887                 kfree_skb(skb);
2888                 goto errout;
2889         }
2890         return skb;
2891 errout:
2892         if (err < 0)
2893                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2894         return NULL;
2895 }
2896
2897 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2898 {
2899         struct net *net = dev_net(dev);
2900
2901         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2902 }
2903
2904 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2905                                unsigned int change, u32 event,
2906                                gfp_t flags)
2907 {
2908         struct sk_buff *skb;
2909
2910         if (dev->reg_state != NETREG_REGISTERED)
2911                 return;
2912
2913         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2914         if (skb)
2915                 rtmsg_ifinfo_send(skb, dev, flags);
2916 }
2917
2918 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2919                   gfp_t flags)
2920 {
2921         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2922 }
2923 EXPORT_SYMBOL(rtmsg_ifinfo);
2924
2925 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2926                                    struct net_device *dev,
2927                                    u8 *addr, u16 vid, u32 pid, u32 seq,
2928                                    int type, unsigned int flags,
2929                                    int nlflags, u16 ndm_state)
2930 {
2931         struct nlmsghdr *nlh;
2932         struct ndmsg *ndm;
2933
2934         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2935         if (!nlh)
2936                 return -EMSGSIZE;
2937
2938         ndm = nlmsg_data(nlh);
2939         ndm->ndm_family  = AF_BRIDGE;
2940         ndm->ndm_pad1    = 0;
2941         ndm->ndm_pad2    = 0;
2942         ndm->ndm_flags   = flags;
2943         ndm->ndm_type    = 0;
2944         ndm->ndm_ifindex = dev->ifindex;
2945         ndm->ndm_state   = ndm_state;
2946
2947         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2948                 goto nla_put_failure;
2949         if (vid)
2950                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2951                         goto nla_put_failure;
2952
2953         nlmsg_end(skb, nlh);
2954         return 0;
2955
2956 nla_put_failure:
2957         nlmsg_cancel(skb, nlh);
2958         return -EMSGSIZE;
2959 }
2960
2961 static inline size_t rtnl_fdb_nlmsg_size(void)
2962 {
2963         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2964                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
2965                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
2966                0;
2967 }
2968
2969 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2970                             u16 ndm_state)
2971 {
2972         struct net *net = dev_net(dev);
2973         struct sk_buff *skb;
2974         int err = -ENOBUFS;
2975
2976         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2977         if (!skb)
2978                 goto errout;
2979
2980         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2981                                       0, 0, type, NTF_SELF, 0, ndm_state);
2982         if (err < 0) {
2983                 kfree_skb(skb);
2984                 goto errout;
2985         }
2986
2987         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2988         return;
2989 errout:
2990         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2991 }
2992
2993 /**
2994  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2995  */
2996 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2997                      struct nlattr *tb[],
2998                      struct net_device *dev,
2999                      const unsigned char *addr, u16 vid,
3000                      u16 flags)
3001 {
3002         int err = -EINVAL;
3003
3004         /* If aging addresses are supported device will need to
3005          * implement its own handler for this.
3006          */
3007         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3008                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3009                 return err;
3010         }
3011
3012         if (vid) {
3013                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3014                 return err;
3015         }
3016
3017         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3018                 err = dev_uc_add_excl(dev, addr);
3019         else if (is_multicast_ether_addr(addr))
3020                 err = dev_mc_add_excl(dev, addr);
3021
3022         /* Only return duplicate errors if NLM_F_EXCL is set */
3023         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3024                 err = 0;
3025
3026         return err;
3027 }
3028 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3029
3030 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3031 {
3032         u16 vid = 0;
3033
3034         if (vlan_attr) {
3035                 if (nla_len(vlan_attr) != sizeof(u16)) {
3036                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3037                         return -EINVAL;
3038                 }
3039
3040                 vid = nla_get_u16(vlan_attr);
3041
3042                 if (!vid || vid >= VLAN_VID_MASK) {
3043                         pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3044                                 vid);
3045                         return -EINVAL;
3046                 }
3047         }
3048         *p_vid = vid;
3049         return 0;
3050 }
3051
3052 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3053                         struct netlink_ext_ack *extack)
3054 {
3055         struct net *net = sock_net(skb->sk);
3056         struct ndmsg *ndm;
3057         struct nlattr *tb[NDA_MAX+1];
3058         struct net_device *dev;
3059         u8 *addr;
3060         u16 vid;
3061         int err;
3062
3063         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3064         if (err < 0)
3065                 return err;
3066
3067         ndm = nlmsg_data(nlh);
3068         if (ndm->ndm_ifindex == 0) {
3069                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3070                 return -EINVAL;
3071         }
3072
3073         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3074         if (dev == NULL) {
3075                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3076                 return -ENODEV;
3077         }
3078
3079         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3080                 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3081                 return -EINVAL;
3082         }
3083
3084         addr = nla_data(tb[NDA_LLADDR]);
3085
3086         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3087         if (err)
3088                 return err;
3089
3090         err = -EOPNOTSUPP;
3091
3092         /* Support fdb on master device the net/bridge default case */
3093         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3094             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3095                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3096                 const struct net_device_ops *ops = br_dev->netdev_ops;
3097
3098                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3099                                        nlh->nlmsg_flags);
3100                 if (err)
3101                         goto out;
3102                 else
3103                         ndm->ndm_flags &= ~NTF_MASTER;
3104         }
3105
3106         /* Embedded bridge, macvlan, and any other device support */
3107         if ((ndm->ndm_flags & NTF_SELF)) {
3108                 if (dev->netdev_ops->ndo_fdb_add)
3109                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3110                                                            vid,
3111                                                            nlh->nlmsg_flags);
3112                 else
3113                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3114                                                nlh->nlmsg_flags);
3115
3116                 if (!err) {
3117                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3118                                         ndm->ndm_state);
3119                         ndm->ndm_flags &= ~NTF_SELF;
3120                 }
3121         }
3122 out:
3123         return err;
3124 }
3125
3126 /**
3127  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3128  */
3129 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3130                      struct nlattr *tb[],
3131                      struct net_device *dev,
3132                      const unsigned char *addr, u16 vid)
3133 {
3134         int err = -EINVAL;
3135
3136         /* If aging addresses are supported device will need to
3137          * implement its own handler for this.
3138          */
3139         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3140                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3141                 return err;
3142         }
3143
3144         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3145                 err = dev_uc_del(dev, addr);
3146         else if (is_multicast_ether_addr(addr))
3147                 err = dev_mc_del(dev, addr);
3148
3149         return err;
3150 }
3151 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3152
3153 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3154                         struct netlink_ext_ack *extack)
3155 {
3156         struct net *net = sock_net(skb->sk);
3157         struct ndmsg *ndm;
3158         struct nlattr *tb[NDA_MAX+1];
3159         struct net_device *dev;
3160         int err = -EINVAL;
3161         __u8 *addr;
3162         u16 vid;
3163
3164         if (!netlink_capable(skb, CAP_NET_ADMIN))
3165                 return -EPERM;
3166
3167         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3168         if (err < 0)
3169                 return err;
3170
3171         ndm = nlmsg_data(nlh);
3172         if (ndm->ndm_ifindex == 0) {
3173                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3174                 return -EINVAL;
3175         }
3176
3177         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3178         if (dev == NULL) {
3179                 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3180                 return -ENODEV;
3181         }
3182
3183         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3184                 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3185                 return -EINVAL;
3186         }
3187
3188         addr = nla_data(tb[NDA_LLADDR]);
3189
3190         err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3191         if (err)
3192                 return err;
3193
3194         err = -EOPNOTSUPP;
3195
3196         /* Support fdb on master device the net/bridge default case */
3197         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3198             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3199                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3200                 const struct net_device_ops *ops = br_dev->netdev_ops;
3201
3202                 if (ops->ndo_fdb_del)
3203                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3204
3205                 if (err)
3206                         goto out;
3207                 else
3208                         ndm->ndm_flags &= ~NTF_MASTER;
3209         }
3210
3211         /* Embedded bridge, macvlan, and any other device support */
3212         if (ndm->ndm_flags & NTF_SELF) {
3213                 if (dev->netdev_ops->ndo_fdb_del)
3214                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3215                                                            vid);
3216                 else
3217                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3218
3219                 if (!err) {
3220                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3221                                         ndm->ndm_state);
3222                         ndm->ndm_flags &= ~NTF_SELF;
3223                 }
3224         }
3225 out:
3226         return err;
3227 }
3228
3229 static int nlmsg_populate_fdb(struct sk_buff *skb,
3230                               struct netlink_callback *cb,
3231                               struct net_device *dev,
3232                               int *idx,
3233                               struct netdev_hw_addr_list *list)
3234 {
3235         struct netdev_hw_addr *ha;
3236         int err;
3237         u32 portid, seq;
3238
3239         portid = NETLINK_CB(cb->skb).portid;
3240         seq = cb->nlh->nlmsg_seq;
3241
3242         list_for_each_entry(ha, &list->list, list) {
3243                 if (*idx < cb->args[2])
3244                         goto skip;
3245
3246                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3247                                               portid, seq,
3248                                               RTM_NEWNEIGH, NTF_SELF,
3249                                               NLM_F_MULTI, NUD_PERMANENT);
3250                 if (err < 0)
3251                         return err;
3252 skip:
3253                 *idx += 1;
3254         }
3255         return 0;
3256 }
3257
3258 /**
3259  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3260  * @nlh: netlink message header
3261  * @dev: netdevice
3262  *
3263  * Default netdevice operation to dump the existing unicast address list.
3264  * Returns number of addresses from list put in skb.
3265  */
3266 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3267                       struct netlink_callback *cb,
3268                       struct net_device *dev,
3269                       struct net_device *filter_dev,
3270                       int *idx)
3271 {
3272         int err;
3273
3274         netif_addr_lock_bh(dev);
3275         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3276         if (err)
3277                 goto out;
3278         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3279 out:
3280         netif_addr_unlock_bh(dev);
3281         return err;
3282 }
3283 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3284
3285 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3286 {
3287         struct net_device *dev;
3288         struct nlattr *tb[IFLA_MAX+1];
3289         struct net_device *br_dev = NULL;
3290         const struct net_device_ops *ops = NULL;
3291         const struct net_device_ops *cops = NULL;
3292         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3293         struct net *net = sock_net(skb->sk);
3294         struct hlist_head *head;
3295         int brport_idx = 0;
3296         int br_idx = 0;
3297         int h, s_h;
3298         int idx = 0, s_idx;
3299         int err = 0;
3300         int fidx = 0;
3301
3302         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3303                           IFLA_MAX, ifla_policy, NULL);
3304         if (err < 0) {
3305                 return -EINVAL;
3306         } else if (err == 0) {
3307                 if (tb[IFLA_MASTER])
3308                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3309         }
3310
3311         brport_idx = ifm->ifi_index;
3312
3313         if (br_idx) {
3314                 br_dev = __dev_get_by_index(net, br_idx);
3315                 if (!br_dev)
3316                         return -ENODEV;
3317
3318                 ops = br_dev->netdev_ops;
3319         }
3320
3321         s_h = cb->args[0];
3322         s_idx = cb->args[1];
3323
3324         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3325                 idx = 0;
3326                 head = &net->dev_index_head[h];
3327                 hlist_for_each_entry(dev, head, index_hlist) {
3328
3329                         if (brport_idx && (dev->ifindex != brport_idx))
3330                                 continue;
3331
3332                         if (!br_idx) { /* user did not specify a specific bridge */
3333                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3334                                         br_dev = netdev_master_upper_dev_get(dev);
3335                                         cops = br_dev->netdev_ops;
3336                                 }
3337                         } else {
3338                                 if (dev != br_dev &&
3339                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3340                                         continue;
3341
3342                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3343                                     !(dev->priv_flags & IFF_EBRIDGE))
3344                                         continue;
3345                                 cops = ops;
3346                         }
3347
3348                         if (idx < s_idx)
3349                                 goto cont;
3350
3351                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3352                                 if (cops && cops->ndo_fdb_dump) {
3353                                         err = cops->ndo_fdb_dump(skb, cb,
3354                                                                 br_dev, dev,
3355                                                                 &fidx);
3356                                         if (err == -EMSGSIZE)
3357                                                 goto out;
3358                                 }
3359                         }
3360
3361                         if (dev->netdev_ops->ndo_fdb_dump)
3362                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3363                                                                     dev, NULL,
3364                                                                     &fidx);
3365                         else
3366                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3367                                                         &fidx);
3368                         if (err == -EMSGSIZE)
3369                                 goto out;
3370
3371                         cops = NULL;
3372
3373                         /* reset fdb offset to 0 for rest of the interfaces */
3374                         cb->args[2] = 0;
3375                         fidx = 0;
3376 cont:
3377                         idx++;
3378                 }
3379         }
3380
3381 out:
3382         cb->args[0] = h;
3383         cb->args[1] = idx;
3384         cb->args[2] = fidx;
3385
3386         return skb->len;
3387 }
3388
3389 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3390                                unsigned int attrnum, unsigned int flag)
3391 {
3392         if (mask & flag)
3393                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3394         return 0;
3395 }
3396
3397 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3398                             struct net_device *dev, u16 mode,
3399                             u32 flags, u32 mask, int nlflags,
3400                             u32 filter_mask,
3401                             int (*vlan_fill)(struct sk_buff *skb,
3402                                              struct net_device *dev,
3403                                              u32 filter_mask))
3404 {
3405         struct nlmsghdr *nlh;
3406         struct ifinfomsg *ifm;
3407         struct nlattr *br_afspec;
3408         struct nlattr *protinfo;
3409         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3410         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3411         int err = 0;
3412
3413         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3414         if (nlh == NULL)
3415                 return -EMSGSIZE;
3416
3417         ifm = nlmsg_data(nlh);
3418         ifm->ifi_family = AF_BRIDGE;
3419         ifm->__ifi_pad = 0;
3420         ifm->ifi_type = dev->type;
3421         ifm->ifi_index = dev->ifindex;
3422         ifm->ifi_flags = dev_get_flags(dev);
3423         ifm->ifi_change = 0;
3424
3425
3426         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3427             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3428             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3429             (br_dev &&
3430              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3431             (dev->addr_len &&
3432              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3433             (dev->ifindex != dev_get_iflink(dev) &&
3434              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3435                 goto nla_put_failure;
3436
3437         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3438         if (!br_afspec)
3439                 goto nla_put_failure;
3440
3441         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3442                 nla_nest_cancel(skb, br_afspec);
3443                 goto nla_put_failure;
3444         }
3445
3446         if (mode != BRIDGE_MODE_UNDEF) {
3447                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3448                         nla_nest_cancel(skb, br_afspec);
3449                         goto nla_put_failure;
3450                 }
3451         }
3452         if (vlan_fill) {
3453                 err = vlan_fill(skb, dev, filter_mask);
3454                 if (err) {
3455                         nla_nest_cancel(skb, br_afspec);
3456                         goto nla_put_failure;
3457                 }
3458         }
3459         nla_nest_end(skb, br_afspec);
3460
3461         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3462         if (!protinfo)
3463                 goto nla_put_failure;
3464
3465         if (brport_nla_put_flag(skb, flags, mask,
3466                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3467             brport_nla_put_flag(skb, flags, mask,
3468                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3469             brport_nla_put_flag(skb, flags, mask,
3470                                 IFLA_BRPORT_FAST_LEAVE,
3471                                 BR_MULTICAST_FAST_LEAVE) ||
3472             brport_nla_put_flag(skb, flags, mask,
3473                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3474             brport_nla_put_flag(skb, flags, mask,
3475                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3476             brport_nla_put_flag(skb, flags, mask,
3477                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3478             brport_nla_put_flag(skb, flags, mask,
3479                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3480             brport_nla_put_flag(skb, flags, mask,
3481                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3482                 nla_nest_cancel(skb, protinfo);
3483                 goto nla_put_failure;
3484         }
3485
3486         nla_nest_end(skb, protinfo);
3487
3488         nlmsg_end(skb, nlh);
3489         return 0;
3490 nla_put_failure:
3491         nlmsg_cancel(skb, nlh);
3492         return err ? err : -EMSGSIZE;
3493 }
3494 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3495
3496 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3497 {
3498         struct net *net = sock_net(skb->sk);
3499         struct net_device *dev;
3500         int idx = 0;
3501         u32 portid = NETLINK_CB(cb->skb).portid;
3502         u32 seq = cb->nlh->nlmsg_seq;
3503         u32 filter_mask = 0;
3504         int err;
3505
3506         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3507                 struct nlattr *extfilt;
3508
3509                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3510                                           IFLA_EXT_MASK);
3511                 if (extfilt) {
3512                         if (nla_len(extfilt) < sizeof(filter_mask))
3513                                 return -EINVAL;
3514
3515                         filter_mask = nla_get_u32(extfilt);
3516                 }
3517         }
3518
3519         rcu_read_lock();
3520         for_each_netdev_rcu(net, dev) {
3521                 const struct net_device_ops *ops = dev->netdev_ops;
3522                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3523
3524                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3525                         if (idx >= cb->args[0]) {
3526                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3527                                                 skb, portid, seq, dev,
3528                                                 filter_mask, NLM_F_MULTI);
3529                                 if (err < 0 && err != -EOPNOTSUPP) {
3530                                         if (likely(skb->len))
3531                                                 break;
3532
3533                                         goto out_err;
3534                                 }
3535                         }
3536                         idx++;
3537                 }
3538
3539                 if (ops->ndo_bridge_getlink) {
3540                         if (idx >= cb->args[0]) {
3541                                 err = ops->ndo_bridge_getlink(skb, portid,
3542                                                               seq, dev,
3543                                                               filter_mask,
3544                                                               NLM_F_MULTI);
3545                                 if (err < 0 && err != -EOPNOTSUPP) {
3546                                         if (likely(skb->len))
3547                                                 break;
3548
3549                                         goto out_err;
3550                                 }
3551                         }
3552                         idx++;
3553                 }
3554         }
3555         err = skb->len;
3556 out_err:
3557         rcu_read_unlock();
3558         cb->args[0] = idx;
3559
3560         return err;
3561 }
3562
3563 static inline size_t bridge_nlmsg_size(void)
3564 {
3565         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3566                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3567                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3568                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3569                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3570                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3571                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3572                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3573                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3574                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3575                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3576 }
3577
3578 static int rtnl_bridge_notify(struct net_device *dev)
3579 {
3580         struct net *net = dev_net(dev);
3581         struct sk_buff *skb;
3582         int err = -EOPNOTSUPP;
3583
3584         if (!dev->netdev_ops->ndo_bridge_getlink)
3585                 return 0;
3586
3587         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3588         if (!skb) {
3589                 err = -ENOMEM;
3590                 goto errout;
3591         }
3592
3593         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3594         if (err < 0)
3595                 goto errout;
3596
3597         if (!skb->len)
3598                 goto errout;
3599
3600         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3601         return 0;
3602 errout:
3603         WARN_ON(err == -EMSGSIZE);
3604         kfree_skb(skb);
3605         if (err)
3606                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3607         return err;
3608 }
3609
3610 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3611                                struct netlink_ext_ack *extack)
3612 {
3613         struct net *net = sock_net(skb->sk);
3614         struct ifinfomsg *ifm;
3615         struct net_device *dev;
3616         struct nlattr *br_spec, *attr = NULL;
3617         int rem, err = -EOPNOTSUPP;
3618         u16 flags = 0;
3619         bool have_flags = false;
3620
3621         if (nlmsg_len(nlh) < sizeof(*ifm))
3622                 return -EINVAL;
3623
3624         ifm = nlmsg_data(nlh);
3625         if (ifm->ifi_family != AF_BRIDGE)
3626                 return -EPFNOSUPPORT;
3627
3628         dev = __dev_get_by_index(net, ifm->ifi_index);
3629         if (!dev) {
3630                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3631                 return -ENODEV;
3632         }
3633
3634         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3635         if (br_spec) {
3636                 nla_for_each_nested(attr, br_spec, rem) {
3637                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3638                                 if (nla_len(attr) < sizeof(flags))
3639                                         return -EINVAL;
3640
3641                                 have_flags = true;
3642                                 flags = nla_get_u16(attr);
3643                                 break;
3644                         }
3645                 }
3646         }
3647
3648         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3649                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3650
3651                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3652                         err = -EOPNOTSUPP;
3653                         goto out;
3654                 }
3655
3656                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3657                 if (err)
3658                         goto out;
3659
3660                 flags &= ~BRIDGE_FLAGS_MASTER;
3661         }
3662
3663         if ((flags & BRIDGE_FLAGS_SELF)) {
3664                 if (!dev->netdev_ops->ndo_bridge_setlink)
3665                         err = -EOPNOTSUPP;
3666                 else
3667                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3668                                                                   flags);
3669                 if (!err) {
3670                         flags &= ~BRIDGE_FLAGS_SELF;
3671
3672                         /* Generate event to notify upper layer of bridge
3673                          * change
3674                          */
3675                         err = rtnl_bridge_notify(dev);
3676                 }
3677         }
3678
3679         if (have_flags)
3680                 memcpy(nla_data(attr), &flags, sizeof(flags));
3681 out:
3682         return err;
3683 }
3684
3685 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3686                                struct netlink_ext_ack *extack)
3687 {
3688         struct net *net = sock_net(skb->sk);
3689         struct ifinfomsg *ifm;
3690         struct net_device *dev;
3691         struct nlattr *br_spec, *attr = NULL;
3692         int rem, err = -EOPNOTSUPP;
3693         u16 flags = 0;
3694         bool have_flags = false;
3695
3696         if (nlmsg_len(nlh) < sizeof(*ifm))
3697                 return -EINVAL;
3698
3699         ifm = nlmsg_data(nlh);
3700         if (ifm->ifi_family != AF_BRIDGE)
3701                 return -EPFNOSUPPORT;
3702
3703         dev = __dev_get_by_index(net, ifm->ifi_index);
3704         if (!dev) {
3705                 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3706                 return -ENODEV;
3707         }
3708
3709         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3710         if (br_spec) {
3711                 nla_for_each_nested(attr, br_spec, rem) {
3712                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3713                                 if (nla_len(attr) < sizeof(flags))
3714                                         return -EINVAL;
3715
3716                                 have_flags = true;
3717                                 flags = nla_get_u16(attr);
3718                                 break;
3719                         }
3720                 }
3721         }
3722
3723         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3724                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3725
3726                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3727                         err = -EOPNOTSUPP;
3728                         goto out;
3729                 }
3730
3731                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3732                 if (err)
3733                         goto out;
3734
3735                 flags &= ~BRIDGE_FLAGS_MASTER;
3736         }
3737
3738         if ((flags & BRIDGE_FLAGS_SELF)) {
3739                 if (!dev->netdev_ops->ndo_bridge_dellink)
3740                         err = -EOPNOTSUPP;
3741                 else
3742                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3743                                                                   flags);
3744
3745                 if (!err) {
3746                         flags &= ~BRIDGE_FLAGS_SELF;
3747
3748                         /* Generate event to notify upper layer of bridge
3749                          * change
3750                          */
3751                         err = rtnl_bridge_notify(dev);
3752                 }
3753         }
3754
3755         if (have_flags)
3756                 memcpy(nla_data(attr), &flags, sizeof(flags));
3757 out:
3758         return err;
3759 }
3760
3761 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3762 {
3763         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3764                (!idxattr || idxattr == attrid);
3765 }
3766
3767 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3768 static int rtnl_get_offload_stats_attr_size(int attr_id)
3769 {
3770         switch (attr_id) {
3771         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3772                 return sizeof(struct rtnl_link_stats64);
3773         }
3774
3775         return 0;
3776 }
3777
3778 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3779                                   int *prividx)
3780 {
3781         struct nlattr *attr = NULL;
3782         int attr_id, size;
3783         void *attr_data;
3784         int err;
3785
3786         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3787               dev->netdev_ops->ndo_get_offload_stats))
3788                 return -ENODATA;
3789
3790         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3791              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3792                 if (attr_id < *prividx)
3793                         continue;
3794
3795                 size = rtnl_get_offload_stats_attr_size(attr_id);
3796                 if (!size)
3797                         continue;
3798
3799                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3800                         continue;
3801
3802                 attr = nla_reserve_64bit(skb, attr_id, size,
3803                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
3804                 if (!attr)
3805                         goto nla_put_failure;
3806
3807                 attr_data = nla_data(attr);
3808                 memset(attr_data, 0, size);
3809                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3810                                                              attr_data);
3811                 if (err)
3812                         goto get_offload_stats_failure;
3813         }
3814
3815         if (!attr)
3816                 return -ENODATA;
3817
3818         *prividx = 0;
3819         return 0;
3820
3821 nla_put_failure:
3822         err = -EMSGSIZE;
3823 get_offload_stats_failure:
3824         *prividx = attr_id;
3825         return err;
3826 }
3827
3828 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3829 {
3830         int nla_size = 0;
3831         int attr_id;
3832         int size;
3833
3834         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3835               dev->netdev_ops->ndo_get_offload_stats))
3836                 return 0;
3837
3838         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3839              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3840                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3841                         continue;
3842                 size = rtnl_get_offload_stats_attr_size(attr_id);
3843                 nla_size += nla_total_size_64bit(size);
3844         }
3845
3846         if (nla_size != 0)
3847                 nla_size += nla_total_size(0);
3848
3849         return nla_size;
3850 }
3851
3852 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3853                                int type, u32 pid, u32 seq, u32 change,
3854                                unsigned int flags, unsigned int filter_mask,
3855                                int *idxattr, int *prividx)
3856 {
3857         struct if_stats_msg *ifsm;
3858         struct nlmsghdr *nlh;
3859         struct nlattr *attr;
3860         int s_prividx = *prividx;
3861         int err;
3862
3863         ASSERT_RTNL();
3864
3865         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3866         if (!nlh)
3867                 return -EMSGSIZE;
3868
3869         ifsm = nlmsg_data(nlh);
3870         ifsm->ifindex = dev->ifindex;
3871         ifsm->filter_mask = filter_mask;
3872
3873         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3874                 struct rtnl_link_stats64 *sp;
3875
3876                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3877                                          sizeof(struct rtnl_link_stats64),
3878                                          IFLA_STATS_UNSPEC);
3879                 if (!attr)
3880                         goto nla_put_failure;
3881
3882                 sp = nla_data(attr);
3883                 dev_get_stats(dev, sp);
3884         }
3885
3886         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3887                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3888
3889                 if (ops && ops->fill_linkxstats) {
3890                         *idxattr = IFLA_STATS_LINK_XSTATS;
3891                         attr = nla_nest_start(skb,
3892                                               IFLA_STATS_LINK_XSTATS);
3893                         if (!attr)
3894                                 goto nla_put_failure;
3895
3896                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3897                         nla_nest_end(skb, attr);
3898                         if (err)
3899                                 goto nla_put_failure;
3900                         *idxattr = 0;
3901                 }
3902         }
3903
3904         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3905                              *idxattr)) {
3906                 const struct rtnl_link_ops *ops = NULL;
3907                 const struct net_device *master;
3908
3909                 master = netdev_master_upper_dev_get(dev);
3910                 if (master)
3911                         ops = master->rtnl_link_ops;
3912                 if (ops && ops->fill_linkxstats) {
3913                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3914                         attr = nla_nest_start(skb,
3915                                               IFLA_STATS_LINK_XSTATS_SLAVE);
3916                         if (!attr)
3917                                 goto nla_put_failure;
3918
3919                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3920                         nla_nest_end(skb, attr);
3921                         if (err)
3922                                 goto nla_put_failure;
3923                         *idxattr = 0;
3924                 }
3925         }
3926
3927         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3928                              *idxattr)) {
3929                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3930                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3931                 if (!attr)
3932                         goto nla_put_failure;
3933
3934                 err = rtnl_get_offload_stats(skb, dev, prividx);
3935                 if (err == -ENODATA)
3936                         nla_nest_cancel(skb, attr);
3937                 else
3938                         nla_nest_end(skb, attr);
3939
3940                 if (err && err != -ENODATA)
3941                         goto nla_put_failure;
3942                 *idxattr = 0;
3943         }
3944
3945         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3946                 struct rtnl_af_ops *af_ops;
3947
3948                 *idxattr = IFLA_STATS_AF_SPEC;
3949                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3950                 if (!attr)
3951                         goto nla_put_failure;
3952
3953                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3954                         if (af_ops->fill_stats_af) {
3955                                 struct nlattr *af;
3956                                 int err;
3957
3958                                 af = nla_nest_start(skb, af_ops->family);
3959                                 if (!af)
3960                                         goto nla_put_failure;
3961
3962                                 err = af_ops->fill_stats_af(skb, dev);
3963
3964                                 if (err == -ENODATA)
3965                                         nla_nest_cancel(skb, af);
3966                                 else if (err < 0)
3967                                         goto nla_put_failure;
3968
3969                                 nla_nest_end(skb, af);
3970                         }
3971                 }
3972
3973                 nla_nest_end(skb, attr);
3974
3975                 *idxattr = 0;
3976         }
3977
3978         nlmsg_end(skb, nlh);
3979
3980         return 0;
3981
3982 nla_put_failure:
3983         /* not a multi message or no progress mean a real error */
3984         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3985                 nlmsg_cancel(skb, nlh);
3986         else
3987                 nlmsg_end(skb, nlh);
3988
3989         return -EMSGSIZE;
3990 }
3991
3992 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3993                                   u32 filter_mask)
3994 {
3995         size_t size = 0;
3996
3997         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3998                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3999
4000         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4001                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4002                 int attr = IFLA_STATS_LINK_XSTATS;
4003
4004                 if (ops && ops->get_linkxstats_size) {
4005                         size += nla_total_size(ops->get_linkxstats_size(dev,
4006                                                                         attr));
4007                         /* for IFLA_STATS_LINK_XSTATS */
4008                         size += nla_total_size(0);
4009                 }
4010         }
4011
4012         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4013                 struct net_device *_dev = (struct net_device *)dev;
4014                 const struct rtnl_link_ops *ops = NULL;
4015                 const struct net_device *master;
4016
4017                 /* netdev_master_upper_dev_get can't take const */
4018                 master = netdev_master_upper_dev_get(_dev);
4019                 if (master)
4020                         ops = master->rtnl_link_ops;
4021                 if (ops && ops->get_linkxstats_size) {
4022                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4023
4024                         size += nla_total_size(ops->get_linkxstats_size(dev,
4025                                                                         attr));
4026                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4027                         size += nla_total_size(0);
4028                 }
4029         }
4030
4031         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4032                 size += rtnl_get_offload_stats_size(dev);
4033
4034         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4035                 struct rtnl_af_ops *af_ops;
4036
4037                 /* for IFLA_STATS_AF_SPEC */
4038                 size += nla_total_size(0);
4039
4040                 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4041                         if (af_ops->get_stats_af_size) {
4042                                 size += nla_total_size(
4043                                         af_ops->get_stats_af_size(dev));
4044
4045                                 /* for AF_* */
4046                                 size += nla_total_size(0);
4047                         }
4048                 }
4049         }
4050
4051         return size;
4052 }
4053
4054 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4055                           struct netlink_ext_ack *extack)
4056 {
4057         struct net *net = sock_net(skb->sk);
4058         struct net_device *dev = NULL;
4059         int idxattr = 0, prividx = 0;
4060         struct if_stats_msg *ifsm;
4061         struct sk_buff *nskb;
4062         u32 filter_mask;
4063         int err;
4064
4065         if (nlmsg_len(nlh) < sizeof(*ifsm))
4066                 return -EINVAL;
4067
4068         ifsm = nlmsg_data(nlh);
4069         if (ifsm->ifindex > 0)
4070                 dev = __dev_get_by_index(net, ifsm->ifindex);
4071         else
4072                 return -EINVAL;
4073
4074         if (!dev)
4075                 return -ENODEV;
4076
4077         filter_mask = ifsm->filter_mask;
4078         if (!filter_mask)
4079                 return -EINVAL;
4080
4081         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4082         if (!nskb)
4083                 return -ENOBUFS;
4084
4085         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4086                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4087                                   0, filter_mask, &idxattr, &prividx);
4088         if (err < 0) {
4089                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4090                 WARN_ON(err == -EMSGSIZE);
4091                 kfree_skb(nskb);
4092         } else {
4093                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4094         }
4095
4096         return err;
4097 }
4098
4099 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4100 {
4101         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4102         struct net *net = sock_net(skb->sk);
4103         unsigned int flags = NLM_F_MULTI;
4104         struct if_stats_msg *ifsm;
4105         struct hlist_head *head;
4106         struct net_device *dev;
4107         u32 filter_mask = 0;
4108         int idx = 0;
4109
4110         s_h = cb->args[0];
4111         s_idx = cb->args[1];
4112         s_idxattr = cb->args[2];
4113         s_prividx = cb->args[3];
4114
4115         cb->seq = net->dev_base_seq;
4116
4117         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4118                 return -EINVAL;
4119
4120         ifsm = nlmsg_data(cb->nlh);
4121         filter_mask = ifsm->filter_mask;
4122         if (!filter_mask)
4123                 return -EINVAL;
4124
4125         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4126                 idx = 0;
4127                 head = &net->dev_index_head[h];
4128                 hlist_for_each_entry(dev, head, index_hlist) {
4129                         if (idx < s_idx)
4130                                 goto cont;
4131                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4132                                                   NETLINK_CB(cb->skb).portid,
4133                                                   cb->nlh->nlmsg_seq, 0,
4134                                                   flags, filter_mask,
4135                                                   &s_idxattr, &s_prividx);
4136                         /* If we ran out of room on the first message,
4137                          * we're in trouble
4138                          */
4139                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4140
4141                         if (err < 0)
4142                                 goto out;
4143                         s_prividx = 0;
4144                         s_idxattr = 0;
4145                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4146 cont:
4147                         idx++;
4148                 }
4149         }
4150 out:
4151         cb->args[3] = s_prividx;
4152         cb->args[2] = s_idxattr;
4153         cb->args[1] = idx;
4154         cb->args[0] = h;
4155
4156         return skb->len;
4157 }
4158
4159 /* Process one rtnetlink message. */
4160
4161 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4162                              struct netlink_ext_ack *extack)
4163 {
4164         struct net *net = sock_net(skb->sk);
4165         rtnl_doit_func doit;
4166         int kind;
4167         int family;
4168         int type;
4169         int err;
4170
4171         type = nlh->nlmsg_type;
4172         if (type > RTM_MAX)
4173                 return -EOPNOTSUPP;
4174
4175         type -= RTM_BASE;
4176
4177         /* All the messages must have at least 1 byte length */
4178         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4179                 return 0;
4180
4181         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4182         kind = type&3;
4183
4184         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4185                 return -EPERM;
4186
4187         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4188                 struct sock *rtnl;
4189                 rtnl_dumpit_func dumpit;
4190                 rtnl_calcit_func calcit;
4191                 u16 min_dump_alloc = 0;
4192
4193                 dumpit = rtnl_get_dumpit(family, type);
4194                 if (dumpit == NULL)
4195                         return -EOPNOTSUPP;
4196                 calcit = rtnl_get_calcit(family, type);
4197                 if (calcit)
4198                         min_dump_alloc = calcit(skb, nlh);
4199
4200                 __rtnl_unlock();
4201                 rtnl = net->rtnl;
4202                 {
4203                         struct netlink_dump_control c = {
4204                                 .dump           = dumpit,
4205                                 .min_dump_alloc = min_dump_alloc,
4206                         };
4207                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4208                 }
4209                 rtnl_lock();
4210                 return err;
4211         }
4212
4213         doit = rtnl_get_doit(family, type);
4214         if (doit == NULL)
4215                 return -EOPNOTSUPP;
4216
4217         return doit(skb, nlh, extack);
4218 }
4219
4220 static void rtnetlink_rcv(struct sk_buff *skb)
4221 {
4222         rtnl_lock();
4223         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4224         rtnl_unlock();
4225 }
4226
4227 static int rtnetlink_bind(struct net *net, int group)
4228 {
4229         switch (group) {
4230         case RTNLGRP_IPV4_MROUTE_R:
4231         case RTNLGRP_IPV6_MROUTE_R:
4232                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4233                         return -EPERM;
4234                 break;
4235         }
4236         return 0;
4237 }
4238
4239 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4240 {
4241         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4242
4243         switch (event) {
4244         case NETDEV_REBOOT:
4245         case NETDEV_CHANGEADDR:
4246         case NETDEV_CHANGENAME:
4247         case NETDEV_FEAT_CHANGE:
4248         case NETDEV_BONDING_FAILOVER:
4249         case NETDEV_NOTIFY_PEERS:
4250         case NETDEV_RESEND_IGMP:
4251         case NETDEV_CHANGEINFODATA:
4252                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4253                                    GFP_KERNEL);
4254                 break;
4255         default:
4256                 break;
4257         }
4258         return NOTIFY_DONE;
4259 }
4260
4261 static struct notifier_block rtnetlink_dev_notifier = {
4262         .notifier_call  = rtnetlink_event,
4263 };
4264
4265
4266 static int __net_init rtnetlink_net_init(struct net *net)
4267 {
4268         struct sock *sk;
4269         struct netlink_kernel_cfg cfg = {
4270                 .groups         = RTNLGRP_MAX,
4271                 .input          = rtnetlink_rcv,
4272                 .cb_mutex       = &rtnl_mutex,
4273                 .flags          = NL_CFG_F_NONROOT_RECV,
4274                 .bind           = rtnetlink_bind,
4275         };
4276
4277         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4278         if (!sk)
4279                 return -ENOMEM;
4280         net->rtnl = sk;
4281         return 0;
4282 }
4283
4284 static void __net_exit rtnetlink_net_exit(struct net *net)
4285 {
4286         netlink_kernel_release(net->rtnl);
4287         net->rtnl = NULL;
4288 }
4289
4290 static struct pernet_operations rtnetlink_net_ops = {
4291         .init = rtnetlink_net_init,
4292         .exit = rtnetlink_net_exit,
4293 };
4294
4295 void __init rtnetlink_init(void)
4296 {
4297         if (register_pernet_subsys(&rtnetlink_net_ops))
4298                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4299
4300         register_netdevice_notifier(&rtnetlink_dev_notifier);
4301
4302         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4303                       rtnl_dump_ifinfo, rtnl_calcit);
4304         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4305         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4306         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4307
4308         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4309         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4310         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, NULL);
4311
4312         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4313         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4314         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4315
4316         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4317         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4318         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4319
4320         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4321                       NULL);
4322 }