net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  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 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 143
 144 static inline int rtm_msgindex(int msgtype)
 145 {
 146         int msgindex = msgtype - RTM_BASE;
 147
 148         /*
 149          * msgindex < 0 implies someone tried to register a netlink
 150          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 151          * the message type has not been added to linux/rtnetlink.h
 152          */
 153         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 154
 155         return msgindex;
 156 }
 157
 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 159 {
 160         struct rtnl_link **tab;
 161
 162         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 163                 protocol = PF_UNSPEC;
 164
 165         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 166         if (!tab)
 167                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 168
 169         return tab[msgtype];
 170 }
 171
 172 static int rtnl_register_internal(struct module *owner,
 173                                   int protocol, int msgtype,
 174                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 175                                   unsigned int flags)
 176 {
 177         struct rtnl_link *link, *old;
 178         struct rtnl_link __rcu **tab;
 179         int msgindex;
 180         int ret = -ENOBUFS;
 181
 182         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 183         msgindex = rtm_msgindex(msgtype);
 184
 185         rtnl_lock();
 186         tab = rtnl_msg_handlers[protocol];
 187         if (tab == NULL) {
 188                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 189                 if (!tab)
 190                         goto unlock;
 191
 192                 /* ensures we see the 0 stores */
 193                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 194         }
 195
 196         old = rtnl_dereference(tab[msgindex]);
 197         if (old) {
 198                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         } else {
 202                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 203                 if (!link)
 204                         goto unlock;
 205         }
 206
 207         WARN_ON(link->owner && link->owner != owner);
 208         link->owner = owner;
 209
 210         WARN_ON(doit && link->doit && link->doit != doit);
 211         if (doit)
 212                 link->doit = doit;
 213         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 214         if (dumpit)
 215                 link->dumpit = dumpit;
 216
 217         link->flags |= flags;
 218
 219         /* publish protocol:msgtype */
 220         rcu_assign_pointer(tab[msgindex], link);
 221         ret = 0;
 222         if (old)
 223                 kfree_rcu(old, rcu);
 224 unlock:
 225         rtnl_unlock();
 226         return ret;
 227 }
 228
 229 /**
 230  * rtnl_register_module - Register a rtnetlink message type
 231  *
 232  * @owner: module registering the hook (THIS_MODULE)
 233  * @protocol: Protocol family or PF_UNSPEC
 234  * @msgtype: rtnetlink message type
 235  * @doit: Function pointer called for each request message
 236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link **tab, *link;
 290         int msgindex;
 291
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293         msgindex = rtm_msgindex(msgtype);
 294
 295         rtnl_lock();
 296         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 297         if (!tab) {
 298                 rtnl_unlock();
 299                 return -ENOENT;
 300         }
 301
 302         link = tab[msgindex];
 303         rcu_assign_pointer(tab[msgindex], NULL);
 304         rtnl_unlock();
 305
 306         kfree_rcu(link, rcu);
 307
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_unregister);
 311
 312 /**
 313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 314  * @protocol : Protocol family or PF_UNSPEC
 315  *
 316  * Identical to calling rtnl_unregster() for all registered message types
 317  * of a certain protocol family.
 318  */
 319 void rtnl_unregister_all(int protocol)
 320 {
 321         struct rtnl_link **tab, *link;
 322         int msgindex;
 323
 324         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 325
 326         rtnl_lock();
 327         tab = rtnl_msg_handlers[protocol];
 328         if (!tab) {
 329                 rtnl_unlock();
 330                 return;
 331         }
 332         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 333         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 334                 link = tab[msgindex];
 335                 if (!link)
 336                         continue;
 337
 338                 rcu_assign_pointer(tab[msgindex], NULL);
 339                 kfree_rcu(link, rcu);
 340         }
 341         rtnl_unlock();
 342
 343         synchronize_net();
 344
 345         kfree(tab);
 346 }
 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 348
 349 static LIST_HEAD(link_ops);
 350
 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 352 {
 353         const struct rtnl_link_ops *ops;
 354
 355         list_for_each_entry(ops, &link_ops, list) {
 356                 if (!strcmp(ops->kind, kind))
 357                         return ops;
 358         }
 359         return NULL;
 360 }
 361
 362 /**
 363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 364  * @ops: struct rtnl_link_ops * to register
 365  *
 366  * The caller must hold the rtnl_mutex. This function should be used
 367  * by drivers that create devices during module initialization. It
 368  * must be called before registering the devices.
 369  *
 370  * Returns 0 on success or a negative error code.
 371  */
 372 int __rtnl_link_register(struct rtnl_link_ops *ops)
 373 {
 374         if (rtnl_link_ops_get(ops->kind))
 375                 return -EEXIST;
 376
 377         /* The check for setup is here because if ops
 378          * does not have that filled up, it is not possible
 379          * to use the ops for creating device. So do not
 380          * fill up dellink as well. That disables rtnl_dellink.
 381          */
 382         if (ops->setup && !ops->dellink)
 383                 ops->dellink = unregister_netdevice_queue;
 384
 385         list_add_tail(&ops->list, &link_ops);
 386         return 0;
 387 }
 388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 389
 390 /**
 391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 392  * @ops: struct rtnl_link_ops * to register
 393  *
 394  * Returns 0 on success or a negative error code.
 395  */
 396 int rtnl_link_register(struct rtnl_link_ops *ops)
 397 {
 398         int err;
 399
 400         /* Sanity-check max sizes to avoid stack buffer overflow. */
 401         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 402                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 403                 return -EINVAL;
 404
 405         rtnl_lock();
 406         err = __rtnl_link_register(ops);
 407         rtnl_unlock();
 408         return err;
 409 }
 410 EXPORT_SYMBOL_GPL(rtnl_link_register);
 411
 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 413 {
 414         struct net_device *dev;
 415         LIST_HEAD(list_kill);
 416
 417         for_each_netdev(net, dev) {
 418                 if (dev->rtnl_link_ops == ops)
 419                         ops->dellink(dev, &list_kill);
 420         }
 421         unregister_netdevice_many(&list_kill);
 422 }
 423
 424 /**
 425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 426  * @ops: struct rtnl_link_ops * to unregister
 427  *
 428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 429  * integrity (hold pernet_ops_rwsem for writing to close the race
 430  * with setup_net() and cleanup_net()).
 431  */
 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 433 {
 434         struct net *net;
 435
 436         for_each_net(net) {
 437                 __rtnl_kill_links(net, ops);
 438         }
 439         list_del(&ops->list);
 440 }
 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 442
 443 /* Return with the rtnl_lock held when there are no network
 444  * devices unregistering in any network namespace.
 445  */
 446 static void rtnl_lock_unregistering_all(void)
 447 {
 448         struct net *net;
 449         bool unregistering;
 450         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 451
 452         add_wait_queue(&netdev_unregistering_wq, &wait);
 453         for (;;) {
 454                 unregistering = false;
 455                 rtnl_lock();
 456                 /* We held write locked pernet_ops_rwsem, and parallel
 457                  * setup_net() and cleanup_net() are not possible.
 458                  */
 459                 for_each_net(net) {
 460                         if (net->dev_unreg_count > 0) {
 461                                 unregistering = true;
 462                                 break;
 463                         }
 464                 }
 465                 if (!unregistering)
 466                         break;
 467                 __rtnl_unlock();
 468
 469                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 470         }
 471         remove_wait_queue(&netdev_unregistering_wq, &wait);
 472 }
 473
 474 /**
 475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 476  * @ops: struct rtnl_link_ops * to unregister
 477  */
 478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 479 {
 480         /* Close the race with setup_net() and cleanup_net() */
 481         down_write(&pernet_ops_rwsem);
 482         rtnl_lock_unregistering_all();
 483         __rtnl_link_unregister(ops);
 484         rtnl_unlock();
 485         up_write(&pernet_ops_rwsem);
 486 }
 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 488
 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 490 {
 491         struct net_device *master_dev;
 492         const struct rtnl_link_ops *ops;
 493         size_t size = 0;
 494
 495         rcu_read_lock();
 496
 497         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 498         if (!master_dev)
 499                 goto out;
 500
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops || !ops->get_slave_size)
 503                 goto out;
 504         /* IFLA_INFO_SLAVE_DATA + nested data */
 505         size = nla_total_size(sizeof(struct nlattr)) +
 506                ops->get_slave_size(master_dev, dev);
 507
 508 out:
 509         rcu_read_unlock();
 510         return size;
 511 }
 512
 513 static size_t rtnl_link_get_size(const struct net_device *dev)
 514 {
 515         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 516         size_t size;
 517
 518         if (!ops)
 519                 return 0;
 520
 521         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 522                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 523
 524         if (ops->get_size)
 525                 /* IFLA_INFO_DATA + nested data */
 526                 size += nla_total_size(sizeof(struct nlattr)) +
 527                         ops->get_size(dev);
 528
 529         if (ops->get_xstats_size)
 530                 /* IFLA_INFO_XSTATS */
 531                 size += nla_total_size(ops->get_xstats_size(dev));
 532
 533         size += rtnl_link_get_slave_info_data_size(dev);
 534
 535         return size;
 536 }
 537
 538 static LIST_HEAD(rtnl_af_ops);
 539
 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 541 {
 542         const struct rtnl_af_ops *ops;
 543
 544         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 545                 if (ops->family == family)
 546                         return ops;
 547         }
 548
 549         return NULL;
 550 }
 551
 552 /**
 553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 554  * @ops: struct rtnl_af_ops * to register
 555  *
 556  * Returns 0 on success or a negative error code.
 557  */
 558 void rtnl_af_register(struct rtnl_af_ops *ops)
 559 {
 560         rtnl_lock();
 561         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 562         rtnl_unlock();
 563 }
 564 EXPORT_SYMBOL_GPL(rtnl_af_register);
 565
 566 /**
 567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 568  * @ops: struct rtnl_af_ops * to unregister
 569  */
 570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 571 {
 572         rtnl_lock();
 573         list_del_rcu(&ops->list);
 574         rtnl_unlock();
 575
 576         synchronize_rcu();
 577 }
 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 579
 580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 581                                     u32 ext_filter_mask)
 582 {
 583         struct rtnl_af_ops *af_ops;
 584         size_t size;
 585
 586         /* IFLA_AF_SPEC */
 587         size = nla_total_size(sizeof(struct nlattr));
 588
 589         rcu_read_lock();
 590         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 591                 if (af_ops->get_link_af_size) {
 592                         /* AF_* + nested data */
 593                         size += nla_total_size(sizeof(struct nlattr)) +
 594                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 595                 }
 596         }
 597         rcu_read_unlock();
 598
 599         return size;
 600 }
 601
 602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 603 {
 604         struct net_device *master_dev;
 605         bool ret = false;
 606
 607         rcu_read_lock();
 608
 609         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 610         if (master_dev && master_dev->rtnl_link_ops)
 611                 ret = true;
 612         rcu_read_unlock();
 613         return ret;
 614 }
 615
 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 617                                      const struct net_device *dev)
 618 {
 619         struct net_device *master_dev;
 620         const struct rtnl_link_ops *ops;
 621         struct nlattr *slave_data;
 622         int err;
 623
 624         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 625         if (!master_dev)
 626                 return 0;
 627         ops = master_dev->rtnl_link_ops;
 628         if (!ops)
 629                 return 0;
 630         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 631                 return -EMSGSIZE;
 632         if (ops->fill_slave_info) {
 633                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 634                 if (!slave_data)
 635                         return -EMSGSIZE;
 636                 err = ops->fill_slave_info(skb, master_dev, dev);
 637                 if (err < 0)
 638                         goto err_cancel_slave_data;
 639                 nla_nest_end(skb, slave_data);
 640         }
 641         return 0;
 642
 643 err_cancel_slave_data:
 644         nla_nest_cancel(skb, slave_data);
 645         return err;
 646 }
 647
 648 static int rtnl_link_info_fill(struct sk_buff *skb,
 649                                const struct net_device *dev)
 650 {
 651         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 652         struct nlattr *data;
 653         int err;
 654
 655         if (!ops)
 656                 return 0;
 657         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 658                 return -EMSGSIZE;
 659         if (ops->fill_xstats) {
 660                 err = ops->fill_xstats(skb, dev);
 661                 if (err < 0)
 662                         return err;
 663         }
 664         if (ops->fill_info) {
 665                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 666                 if (data == NULL)
 667                         return -EMSGSIZE;
 668                 err = ops->fill_info(skb, dev);
 669                 if (err < 0)
 670                         goto err_cancel_data;
 671                 nla_nest_end(skb, data);
 672         }
 673         return 0;
 674
 675 err_cancel_data:
 676         nla_nest_cancel(skb, data);
 677         return err;
 678 }
 679
 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 681 {
 682         struct nlattr *linkinfo;
 683         int err = -EMSGSIZE;
 684
 685         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 686         if (linkinfo == NULL)
 687                 goto out;
 688
 689         err = rtnl_link_info_fill(skb, dev);
 690         if (err < 0)
 691                 goto err_cancel_link;
 692
 693         err = rtnl_link_slave_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         nla_nest_end(skb, linkinfo);
 698         return 0;
 699
 700 err_cancel_link:
 701         nla_nest_cancel(skb, linkinfo);
 702 out:
 703         return err;
 704 }
 705
 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 707 {
 708         struct sock *rtnl = net->rtnl;
 709         int err = 0;
 710
 711         NETLINK_CB(skb).dst_group = group;
 712         if (echo)
 713                 refcount_inc(&skb->users);
 714         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 715         if (echo)
 716                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 717         return err;
 718 }
 719
 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 721 {
 722         struct sock *rtnl = net->rtnl;
 723
 724         return nlmsg_unicast(rtnl, skb, pid);
 725 }
 726 EXPORT_SYMBOL(rtnl_unicast);
 727
 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 729                  struct nlmsghdr *nlh, gfp_t flags)
 730 {
 731         struct sock *rtnl = net->rtnl;
 732         int report = 0;
 733
 734         if (nlh)
 735                 report = nlmsg_report(nlh);
 736
 737         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 738 }
 739 EXPORT_SYMBOL(rtnl_notify);
 740
 741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 742 {
 743         struct sock *rtnl = net->rtnl;
 744
 745         netlink_set_err(rtnl, 0, group, error);
 746 }
 747 EXPORT_SYMBOL(rtnl_set_sk_err);
 748
 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 750 {
 751         struct nlattr *mx;
 752         int i, valid = 0;
 753
 754         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 755         if (metrics == dst_default_metrics.metrics)
 756                 return 0;
 757
 758         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 759         if (mx == NULL)
 760                 return -ENOBUFS;
 761
 762         for (i = 0; i < RTAX_MAX; i++) {
 763                 if (metrics[i]) {
 764                         if (i == RTAX_CC_ALGO - 1) {
 765                                 char tmp[TCP_CA_NAME_MAX], *name;
 766
 767                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 768                                 if (!name)
 769                                         continue;
 770                                 if (nla_put_string(skb, i + 1, name))
 771                                         goto nla_put_failure;
 772                         } else if (i == RTAX_FEATURES - 1) {
 773                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 774
 775                                 if (!user_features)
 776                                         continue;
 777                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 778                                 if (nla_put_u32(skb, i + 1, user_features))
 779                                         goto nla_put_failure;
 780                         } else {
 781                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 782                                         goto nla_put_failure;
 783                         }
 784                         valid++;
 785                 }
 786         }
 787
 788         if (!valid) {
 789                 nla_nest_cancel(skb, mx);
 790                 return 0;
 791         }
 792
 793         return nla_nest_end(skb, mx);
 794
 795 nla_put_failure:
 796         nla_nest_cancel(skb, mx);
 797         return -EMSGSIZE;
 798 }
 799 EXPORT_SYMBOL(rtnetlink_put_metrics);
 800
 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 802                        long expires, u32 error)
 803 {
 804         struct rta_cacheinfo ci = {
 805                 .rta_error = error,
 806                 .rta_id =  id,
 807         };
 808
 809         if (dst) {
 810                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 811                 ci.rta_used = dst->__use;
 812                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 813         }
 814         if (expires) {
 815                 unsigned long clock;
 816
 817                 clock = jiffies_to_clock_t(abs(expires));
 818                 clock = min_t(unsigned long, clock, INT_MAX);
 819                 ci.rta_expires = (expires > 0) ? clock : -clock;
 820         }
 821         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 822 }
 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 824
 825 static void set_operstate(struct net_device *dev, unsigned char transition)
 826 {
 827         unsigned char operstate = dev->operstate;
 828
 829         switch (transition) {
 830         case IF_OPER_UP:
 831                 if ((operstate == IF_OPER_DORMANT ||
 832                      operstate == IF_OPER_UNKNOWN) &&
 833                     !netif_dormant(dev))
 834                         operstate = IF_OPER_UP;
 835                 break;
 836
 837         case IF_OPER_DORMANT:
 838                 if (operstate == IF_OPER_UP ||
 839                     operstate == IF_OPER_UNKNOWN)
 840                         operstate = IF_OPER_DORMANT;
 841                 break;
 842         }
 843
 844         if (dev->operstate != operstate) {
 845                 write_lock_bh(&dev_base_lock);
 846                 dev->operstate = operstate;
 847                 write_unlock_bh(&dev_base_lock);
 848                 netdev_state_change(dev);
 849         }
 850 }
 851
 852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 853 {
 854         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 855                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 856 }
 857
 858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 859                                            const struct ifinfomsg *ifm)
 860 {
 861         unsigned int flags = ifm->ifi_flags;
 862
 863         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 864         if (ifm->ifi_change)
 865                 flags = (flags & ifm->ifi_change) |
 866                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 867
 868         return flags;
 869 }
 870
 871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 872                                  const struct rtnl_link_stats64 *b)
 873 {
 874         a->rx_packets = b->rx_packets;
 875         a->tx_packets = b->tx_packets;
 876         a->rx_bytes = b->rx_bytes;
 877         a->tx_bytes = b->tx_bytes;
 878         a->rx_errors = b->rx_errors;
 879         a->tx_errors = b->tx_errors;
 880         a->rx_dropped = b->rx_dropped;
 881         a->tx_dropped = b->tx_dropped;
 882
 883         a->multicast = b->multicast;
 884         a->collisions = b->collisions;
 885
 886         a->rx_length_errors = b->rx_length_errors;
 887         a->rx_over_errors = b->rx_over_errors;
 888         a->rx_crc_errors = b->rx_crc_errors;
 889         a->rx_frame_errors = b->rx_frame_errors;
 890         a->rx_fifo_errors = b->rx_fifo_errors;
 891         a->rx_missed_errors = b->rx_missed_errors;
 892
 893         a->tx_aborted_errors = b->tx_aborted_errors;
 894         a->tx_carrier_errors = b->tx_carrier_errors;
 895         a->tx_fifo_errors = b->tx_fifo_errors;
 896         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 897         a->tx_window_errors = b->tx_window_errors;
 898
 899         a->rx_compressed = b->rx_compressed;
 900         a->tx_compressed = b->tx_compressed;
 901
 902         a->rx_nohandler = b->rx_nohandler;
 903 }
 904
 905 /* All VF info */
 906 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 907                                    u32 ext_filter_mask)
 908 {
 909         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 910                 int num_vfs = dev_num_vf(dev->dev.parent);
 911                 size_t size = nla_total_size(0);
 912                 size += num_vfs *
 913                         (nla_total_size(0) +
 914                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 915                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 916                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 917                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 918                          nla_total_size(MAX_VLAN_LIST_LEN *
 919                                         sizeof(struct ifla_vf_vlan_info)) +
 920                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 921                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 922                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 923                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 924                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 925                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 926                          /* IFLA_VF_STATS_RX_PACKETS */
 927                          nla_total_size_64bit(sizeof(__u64)) +
 928                          /* IFLA_VF_STATS_TX_PACKETS */
 929                          nla_total_size_64bit(sizeof(__u64)) +
 930                          /* IFLA_VF_STATS_RX_BYTES */
 931                          nla_total_size_64bit(sizeof(__u64)) +
 932                          /* IFLA_VF_STATS_TX_BYTES */
 933                          nla_total_size_64bit(sizeof(__u64)) +
 934                          /* IFLA_VF_STATS_BROADCAST */
 935                          nla_total_size_64bit(sizeof(__u64)) +
 936                          /* IFLA_VF_STATS_MULTICAST */
 937                          nla_total_size_64bit(sizeof(__u64)) +
 938                          /* IFLA_VF_STATS_RX_DROPPED */
 939                          nla_total_size_64bit(sizeof(__u64)) +
 940                          /* IFLA_VF_STATS_TX_DROPPED */
 941                          nla_total_size_64bit(sizeof(__u64)) +
 942                          nla_total_size(sizeof(struct ifla_vf_trust)));
 943                 return size;
 944         } else
 945                 return 0;
 946 }
 947
 948 static size_t rtnl_port_size(const struct net_device *dev,
 949                              u32 ext_filter_mask)
 950 {
 951         size_t port_size = nla_total_size(4)            /* PORT_VF */
 952                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 953                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 954                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 955                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 956                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 957         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 958         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 959                 + port_size;
 960         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 961                 + port_size;
 962
 963         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 964             !(ext_filter_mask & RTEXT_FILTER_VF))
 965                 return 0;
 966         if (dev_num_vf(dev->dev.parent))
 967                 return port_self_size + vf_ports_size +
 968                         vf_port_size * dev_num_vf(dev->dev.parent);
 969         else
 970                 return port_self_size;
 971 }
 972
 973 static size_t rtnl_xdp_size(void)
 974 {
 975         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 976                           nla_total_size(1) +   /* XDP_ATTACHED */
 977                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 978                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 979
 980         return xdp_size;
 981 }
 982
 983 static size_t rtnl_prop_list_size(const struct net_device *dev)
 984 {
 985         struct netdev_name_node *name_node;
 986         size_t size;
 987
 988         if (list_empty(&dev->name_node->list))
 989                 return 0;
 990         size = nla_total_size(0);
 991         list_for_each_entry(name_node, &dev->name_node->list, list)
 992                 size += nla_total_size(ALTIFNAMSIZ);
 993         return size;
 994 }
 995
 996 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 997                                      u32 ext_filter_mask)
 998 {
 999         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1000                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1001                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1002                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1003                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1004                + nla_total_size(sizeof(struct rtnl_link_stats))
1005                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1006                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1007                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1008                + nla_total_size(4) /* IFLA_TXQLEN */
1009                + nla_total_size(4) /* IFLA_WEIGHT */
1010                + nla_total_size(4) /* IFLA_MTU */
1011                + nla_total_size(4) /* IFLA_LINK */
1012                + nla_total_size(4) /* IFLA_MASTER */
1013                + nla_total_size(1) /* IFLA_CARRIER */
1014                + nla_total_size(4) /* IFLA_PROMISCUITY */
1015                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1016                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1017                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1018                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1019                + nla_total_size(1) /* IFLA_OPERSTATE */
1020                + nla_total_size(1) /* IFLA_LINKMODE */
1021                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1022                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1023                + nla_total_size(4) /* IFLA_GROUP */
1024                + nla_total_size(ext_filter_mask
1025                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1026                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1027                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1028                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1029                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1030                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1031                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1032                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1033                + rtnl_xdp_size() /* IFLA_XDP */
1034                + nla_total_size(4)  /* IFLA_EVENT */
1035                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1036                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1037                + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1038                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1039                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1040                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1041                + nla_total_size(4)  /* IFLA_MIN_MTU */
1042                + nla_total_size(4)  /* IFLA_MAX_MTU */
1043                + rtnl_prop_list_size(dev)
1044                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1045                + 0;
1046 }
1047
1048 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1049 {
1050         struct nlattr *vf_ports;
1051         struct nlattr *vf_port;
1052         int vf;
1053         int err;
1054
1055         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1056         if (!vf_ports)
1057                 return -EMSGSIZE;
1058
1059         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1060                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1061                 if (!vf_port)
1062                         goto nla_put_failure;
1063                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1064                         goto nla_put_failure;
1065                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1066                 if (err == -EMSGSIZE)
1067                         goto nla_put_failure;
1068                 if (err) {
1069                         nla_nest_cancel(skb, vf_port);
1070                         continue;
1071                 }
1072                 nla_nest_end(skb, vf_port);
1073         }
1074
1075         nla_nest_end(skb, vf_ports);
1076
1077         return 0;
1078
1079 nla_put_failure:
1080         nla_nest_cancel(skb, vf_ports);
1081         return -EMSGSIZE;
1082 }
1083
1084 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1085 {
1086         struct nlattr *port_self;
1087         int err;
1088
1089         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1090         if (!port_self)
1091                 return -EMSGSIZE;
1092
1093         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1094         if (err) {
1095                 nla_nest_cancel(skb, port_self);
1096                 return (err == -EMSGSIZE) ? err : 0;
1097         }
1098
1099         nla_nest_end(skb, port_self);
1100
1101         return 0;
1102 }
1103
1104 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1105                           u32 ext_filter_mask)
1106 {
1107         int err;
1108
1109         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1110             !(ext_filter_mask & RTEXT_FILTER_VF))
1111                 return 0;
1112
1113         err = rtnl_port_self_fill(skb, dev);
1114         if (err)
1115                 return err;
1116
1117         if (dev_num_vf(dev->dev.parent)) {
1118                 err = rtnl_vf_ports_fill(skb, dev);
1119                 if (err)
1120                         return err;
1121         }
1122
1123         return 0;
1124 }
1125
1126 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1127 {
1128         int err;
1129         struct netdev_phys_item_id ppid;
1130
1131         err = dev_get_phys_port_id(dev, &ppid);
1132         if (err) {
1133                 if (err == -EOPNOTSUPP)
1134                         return 0;
1135                 return err;
1136         }
1137
1138         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1139                 return -EMSGSIZE;
1140
1141         return 0;
1142 }
1143
1144 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1145 {
1146         char name[IFNAMSIZ];
1147         int err;
1148
1149         err = dev_get_phys_port_name(dev, name, sizeof(name));
1150         if (err) {
1151                 if (err == -EOPNOTSUPP)
1152                         return 0;
1153                 return err;
1154         }
1155
1156         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1157                 return -EMSGSIZE;
1158
1159         return 0;
1160 }
1161
1162 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1163 {
1164         struct netdev_phys_item_id ppid = { };
1165         int err;
1166
1167         err = dev_get_port_parent_id(dev, &ppid, false);
1168         if (err) {
1169                 if (err == -EOPNOTSUPP)
1170                         return 0;
1171                 return err;
1172         }
1173
1174         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1175                 return -EMSGSIZE;
1176
1177         return 0;
1178 }
1179
1180 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1181                                               struct net_device *dev)
1182 {
1183         struct rtnl_link_stats64 *sp;
1184         struct nlattr *attr;
1185
1186         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1187                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1188         if (!attr)
1189                 return -EMSGSIZE;
1190
1191         sp = nla_data(attr);
1192         dev_get_stats(dev, sp);
1193
1194         attr = nla_reserve(skb, IFLA_STATS,
1195                            sizeof(struct rtnl_link_stats));
1196         if (!attr)
1197                 return -EMSGSIZE;
1198
1199         copy_rtnl_link_stats(nla_data(attr), sp);
1200
1201         return 0;
1202 }
1203
1204 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1205                                                struct net_device *dev,
1206                                                int vfs_num,
1207                                                struct nlattr *vfinfo)
1208 {
1209         struct ifla_vf_rss_query_en vf_rss_query_en;
1210         struct nlattr *vf, *vfstats, *vfvlanlist;
1211         struct ifla_vf_link_state vf_linkstate;
1212         struct ifla_vf_vlan_info vf_vlan_info;
1213         struct ifla_vf_spoofchk vf_spoofchk;
1214         struct ifla_vf_tx_rate vf_tx_rate;
1215         struct ifla_vf_stats vf_stats;
1216         struct ifla_vf_trust vf_trust;
1217         struct ifla_vf_vlan vf_vlan;
1218         struct ifla_vf_rate vf_rate;
1219         struct ifla_vf_mac vf_mac;
1220         struct ifla_vf_broadcast vf_broadcast;
1221         struct ifla_vf_info ivi;
1222         struct ifla_vf_guid node_guid;
1223         struct ifla_vf_guid port_guid;
1224
1225         memset(&ivi, 0, sizeof(ivi));
1226
1227         /* Not all SR-IOV capable drivers support the
1228          * spoofcheck and "RSS query enable" query.  Preset to
1229          * -1 so the user space tool can detect that the driver
1230          * didn't report anything.
1231          */
1232         ivi.spoofchk = -1;
1233         ivi.rss_query_en = -1;
1234         ivi.trusted = -1;
1235         /* The default value for VF link state is "auto"
1236          * IFLA_VF_LINK_STATE_AUTO which equals zero
1237          */
1238         ivi.linkstate = 0;
1239         /* VLAN Protocol by default is 802.1Q */
1240         ivi.vlan_proto = htons(ETH_P_8021Q);
1241         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1242                 return 0;
1243
1244         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1245
1246         vf_mac.vf =
1247                 vf_vlan.vf =
1248                 vf_vlan_info.vf =
1249                 vf_rate.vf =
1250                 vf_tx_rate.vf =
1251                 vf_spoofchk.vf =
1252                 vf_linkstate.vf =
1253                 vf_rss_query_en.vf =
1254                 vf_trust.vf =
1255                 node_guid.vf =
1256                 port_guid.vf = ivi.vf;
1257
1258         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1259         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1260         vf_vlan.vlan = ivi.vlan;
1261         vf_vlan.qos = ivi.qos;
1262         vf_vlan_info.vlan = ivi.vlan;
1263         vf_vlan_info.qos = ivi.qos;
1264         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1265         vf_tx_rate.rate = ivi.max_tx_rate;
1266         vf_rate.min_tx_rate = ivi.min_tx_rate;
1267         vf_rate.max_tx_rate = ivi.max_tx_rate;
1268         vf_spoofchk.setting = ivi.spoofchk;
1269         vf_linkstate.link_state = ivi.linkstate;
1270         vf_rss_query_en.setting = ivi.rss_query_en;
1271         vf_trust.setting = ivi.trusted;
1272         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1273         if (!vf)
1274                 goto nla_put_vfinfo_failure;
1275         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1276             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1277             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1278             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1279                     &vf_rate) ||
1280             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1281                     &vf_tx_rate) ||
1282             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1283                     &vf_spoofchk) ||
1284             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1285                     &vf_linkstate) ||
1286             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1287                     sizeof(vf_rss_query_en),
1288                     &vf_rss_query_en) ||
1289             nla_put(skb, IFLA_VF_TRUST,
1290                     sizeof(vf_trust), &vf_trust))
1291                 goto nla_put_vf_failure;
1292
1293         memset(&node_guid, 0, sizeof(node_guid));
1294         memset(&port_guid, 0, sizeof(port_guid));
1295         if (dev->netdev_ops->ndo_get_vf_guid &&
1296             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1297                                               &port_guid)) {
1298                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1299                             &node_guid) ||
1300                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1301                             &port_guid))
1302                         goto nla_put_vf_failure;
1303         }
1304         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1305         if (!vfvlanlist)
1306                 goto nla_put_vf_failure;
1307         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1308                     &vf_vlan_info)) {
1309                 nla_nest_cancel(skb, vfvlanlist);
1310                 goto nla_put_vf_failure;
1311         }
1312         nla_nest_end(skb, vfvlanlist);
1313         memset(&vf_stats, 0, sizeof(vf_stats));
1314         if (dev->netdev_ops->ndo_get_vf_stats)
1315                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1316                                                 &vf_stats);
1317         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1318         if (!vfstats)
1319                 goto nla_put_vf_failure;
1320         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1321                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1322             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1323                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1324             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1325                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1326             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1327                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1328             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1329                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1330             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1331                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1332             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1333                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1334             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1335                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1336                 nla_nest_cancel(skb, vfstats);
1337                 goto nla_put_vf_failure;
1338         }
1339         nla_nest_end(skb, vfstats);
1340         nla_nest_end(skb, vf);
1341         return 0;
1342
1343 nla_put_vf_failure:
1344         nla_nest_cancel(skb, vf);
1345 nla_put_vfinfo_failure:
1346         nla_nest_cancel(skb, vfinfo);
1347         return -EMSGSIZE;
1348 }
1349
1350 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1351                                            struct net_device *dev,
1352                                            u32 ext_filter_mask)
1353 {
1354         struct nlattr *vfinfo;
1355         int i, num_vfs;
1356
1357         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1358                 return 0;
1359
1360         num_vfs = dev_num_vf(dev->dev.parent);
1361         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1362                 return -EMSGSIZE;
1363
1364         if (!dev->netdev_ops->ndo_get_vf_config)
1365                 return 0;
1366
1367         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1368         if (!vfinfo)
1369                 return -EMSGSIZE;
1370
1371         for (i = 0; i < num_vfs; i++) {
1372                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1373                         return -EMSGSIZE;
1374         }
1375
1376         nla_nest_end(skb, vfinfo);
1377         return 0;
1378 }
1379
1380 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1381 {
1382         struct rtnl_link_ifmap map;
1383
1384         memset(&map, 0, sizeof(map));
1385         map.mem_start   = dev->mem_start;
1386         map.mem_end     = dev->mem_end;
1387         map.base_addr   = dev->base_addr;
1388         map.irq         = dev->irq;
1389         map.dma         = dev->dma;
1390         map.port        = dev->if_port;
1391
1392         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1393                 return -EMSGSIZE;
1394
1395         return 0;
1396 }
1397
1398 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1399 {
1400         const struct bpf_prog *generic_xdp_prog;
1401
1402         ASSERT_RTNL();
1403
1404         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1405         if (!generic_xdp_prog)
1406                 return 0;
1407         return generic_xdp_prog->aux->id;
1408 }
1409
1410 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1411 {
1412         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1413 }
1414
1415 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1416 {
1417         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1418                                XDP_QUERY_PROG_HW);
1419 }
1420
1421 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1422                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1423                                u32 (*get_prog_id)(struct net_device *dev))
1424 {
1425         u32 curr_id;
1426         int err;
1427
1428         curr_id = get_prog_id(dev);
1429         if (!curr_id)
1430                 return 0;
1431
1432         *prog_id = curr_id;
1433         err = nla_put_u32(skb, attr, curr_id);
1434         if (err)
1435                 return err;
1436
1437         if (*mode != XDP_ATTACHED_NONE)
1438                 *mode = XDP_ATTACHED_MULTI;
1439         else
1440                 *mode = tgt_mode;
1441
1442         return 0;
1443 }
1444
1445 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1446 {
1447         struct nlattr *xdp;
1448         u32 prog_id;
1449         int err;
1450         u8 mode;
1451
1452         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1453         if (!xdp)
1454                 return -EMSGSIZE;
1455
1456         prog_id = 0;
1457         mode = XDP_ATTACHED_NONE;
1458         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1459                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1460         if (err)
1461                 goto err_cancel;
1462         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1463                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1464         if (err)
1465                 goto err_cancel;
1466         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1467                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1468         if (err)
1469                 goto err_cancel;
1470
1471         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1472         if (err)
1473                 goto err_cancel;
1474
1475         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1476                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1477                 if (err)
1478                         goto err_cancel;
1479         }
1480
1481         nla_nest_end(skb, xdp);
1482         return 0;
1483
1484 err_cancel:
1485         nla_nest_cancel(skb, xdp);
1486         return err;
1487 }
1488
1489 static u32 rtnl_get_event(unsigned long event)
1490 {
1491         u32 rtnl_event_type = IFLA_EVENT_NONE;
1492
1493         switch (event) {
1494         case NETDEV_REBOOT:
1495                 rtnl_event_type = IFLA_EVENT_REBOOT;
1496                 break;
1497         case NETDEV_FEAT_CHANGE:
1498                 rtnl_event_type = IFLA_EVENT_FEATURES;
1499                 break;
1500         case NETDEV_BONDING_FAILOVER:
1501                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1502                 break;
1503         case NETDEV_NOTIFY_PEERS:
1504                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1505                 break;
1506         case NETDEV_RESEND_IGMP:
1507                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1508                 break;
1509         case NETDEV_CHANGEINFODATA:
1510                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1511                 break;
1512         default:
1513                 break;
1514         }
1515
1516         return rtnl_event_type;
1517 }
1518
1519 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1520 {
1521         const struct net_device *upper_dev;
1522         int ret = 0;
1523
1524         rcu_read_lock();
1525
1526         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1527         if (upper_dev)
1528                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1529
1530         rcu_read_unlock();
1531         return ret;
1532 }
1533
1534 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1535                           bool force)
1536 {
1537         int ifindex = dev_get_iflink(dev);
1538
1539         if (force || dev->ifindex != ifindex)
1540                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1541
1542         return 0;
1543 }
1544
1545 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1546                                               struct net_device *dev)
1547 {
1548         char buf[IFALIASZ];
1549         int ret;
1550
1551         ret = dev_get_alias(dev, buf, sizeof(buf));
1552         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1553 }
1554
1555 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1556                                   const struct net_device *dev,
1557                                   struct net *src_net, gfp_t gfp)
1558 {
1559         bool put_iflink = false;
1560
1561         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1562                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1563
1564                 if (!net_eq(dev_net(dev), link_net)) {
1565                         int id = peernet2id_alloc(src_net, link_net, gfp);
1566
1567                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1568                                 return -EMSGSIZE;
1569
1570                         put_iflink = true;
1571                 }
1572         }
1573
1574         return nla_put_iflink(skb, dev, put_iflink);
1575 }
1576
1577 static int rtnl_fill_link_af(struct sk_buff *skb,
1578                              const struct net_device *dev,
1579                              u32 ext_filter_mask)
1580 {
1581         const struct rtnl_af_ops *af_ops;
1582         struct nlattr *af_spec;
1583
1584         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1585         if (!af_spec)
1586                 return -EMSGSIZE;
1587
1588         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1589                 struct nlattr *af;
1590                 int err;
1591
1592                 if (!af_ops->fill_link_af)
1593                         continue;
1594
1595                 af = nla_nest_start_noflag(skb, af_ops->family);
1596                 if (!af)
1597                         return -EMSGSIZE;
1598
1599                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1600                 /*
1601                  * Caller may return ENODATA to indicate that there
1602                  * was no data to be dumped. This is not an error, it
1603                  * means we should trim the attribute header and
1604                  * continue.
1605                  */
1606                 if (err == -ENODATA)
1607                         nla_nest_cancel(skb, af);
1608                 else if (err < 0)
1609                         return -EMSGSIZE;
1610
1611                 nla_nest_end(skb, af);
1612         }
1613
1614         nla_nest_end(skb, af_spec);
1615         return 0;
1616 }
1617
1618 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1619                                  const struct net_device *dev)
1620 {
1621         struct netdev_name_node *name_node;
1622         int count = 0;
1623
1624         list_for_each_entry(name_node, &dev->name_node->list, list) {
1625                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1626                         return -EMSGSIZE;
1627                 count++;
1628         }
1629         return count;
1630 }
1631
1632 static int rtnl_fill_prop_list(struct sk_buff *skb,
1633                                const struct net_device *dev)
1634 {
1635         struct nlattr *prop_list;
1636         int ret;
1637
1638         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1639         if (!prop_list)
1640                 return -EMSGSIZE;
1641
1642         ret = rtnl_fill_alt_ifnames(skb, dev);
1643         if (ret <= 0)
1644                 goto nest_cancel;
1645
1646         nla_nest_end(skb, prop_list);
1647         return 0;
1648
1649 nest_cancel:
1650         nla_nest_cancel(skb, prop_list);
1651         return ret;
1652 }
1653
1654 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1655                             struct net_device *dev, struct net *src_net,
1656                             int type, u32 pid, u32 seq, u32 change,
1657                             unsigned int flags, u32 ext_filter_mask,
1658                             u32 event, int *new_nsid, int new_ifindex,
1659                             int tgt_netnsid, gfp_t gfp)
1660 {
1661         struct ifinfomsg *ifm;
1662         struct nlmsghdr *nlh;
1663
1664         ASSERT_RTNL();
1665         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1666         if (nlh == NULL)
1667                 return -EMSGSIZE;
1668
1669         ifm = nlmsg_data(nlh);
1670         ifm->ifi_family = AF_UNSPEC;
1671         ifm->__ifi_pad = 0;
1672         ifm->ifi_type = dev->type;
1673         ifm->ifi_index = dev->ifindex;
1674         ifm->ifi_flags = dev_get_flags(dev);
1675         ifm->ifi_change = change;
1676
1677         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1678                 goto nla_put_failure;
1679
1680         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1681             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1682             nla_put_u8(skb, IFLA_OPERSTATE,
1683                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1684             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1685             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1686             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1687             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1688             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1689             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1690             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1691             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1692             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1693 #ifdef CONFIG_RPS
1694             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1695 #endif
1696             put_master_ifindex(skb, dev) ||
1697             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1698             (dev->qdisc &&
1699              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1700             nla_put_ifalias(skb, dev) ||
1701             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1702                         atomic_read(&dev->carrier_up_count) +
1703                         atomic_read(&dev->carrier_down_count)) ||
1704             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1705             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1706                         atomic_read(&dev->carrier_up_count)) ||
1707             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1708                         atomic_read(&dev->carrier_down_count)))
1709                 goto nla_put_failure;
1710
1711         if (event != IFLA_EVENT_NONE) {
1712                 if (nla_put_u32(skb, IFLA_EVENT, event))
1713                         goto nla_put_failure;
1714         }
1715
1716         if (rtnl_fill_link_ifmap(skb, dev))
1717                 goto nla_put_failure;
1718
1719         if (dev->addr_len) {
1720                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1721                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1722                         goto nla_put_failure;
1723         }
1724
1725         if (rtnl_phys_port_id_fill(skb, dev))
1726                 goto nla_put_failure;
1727
1728         if (rtnl_phys_port_name_fill(skb, dev))
1729                 goto nla_put_failure;
1730
1731         if (rtnl_phys_switch_id_fill(skb, dev))
1732                 goto nla_put_failure;
1733
1734         if (rtnl_fill_stats(skb, dev))
1735                 goto nla_put_failure;
1736
1737         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1738                 goto nla_put_failure;
1739
1740         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1741                 goto nla_put_failure;
1742
1743         if (rtnl_xdp_fill(skb, dev))
1744                 goto nla_put_failure;
1745
1746         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1747                 if (rtnl_link_fill(skb, dev) < 0)
1748                         goto nla_put_failure;
1749         }
1750
1751         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1752                 goto nla_put_failure;
1753
1754         if (new_nsid &&
1755             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1756                 goto nla_put_failure;
1757         if (new_ifindex &&
1758             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1759                 goto nla_put_failure;
1760
1761         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1762             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1763                 goto nla_put_failure;
1764
1765         rcu_read_lock();
1766         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1767                 goto nla_put_failure_rcu;
1768         rcu_read_unlock();
1769
1770         if (rtnl_fill_prop_list(skb, dev))
1771                 goto nla_put_failure;
1772
1773         nlmsg_end(skb, nlh);
1774         return 0;
1775
1776 nla_put_failure_rcu:
1777         rcu_read_unlock();
1778 nla_put_failure:
1779         nlmsg_cancel(skb, nlh);
1780         return -EMSGSIZE;
1781 }
1782
1783 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1784         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1785         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1786         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1787         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1788         [IFLA_MTU]              = { .type = NLA_U32 },
1789         [IFLA_LINK]             = { .type = NLA_U32 },
1790         [IFLA_MASTER]           = { .type = NLA_U32 },
1791         [IFLA_CARRIER]          = { .type = NLA_U8 },
1792         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1793         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1794         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1795         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1796         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1797         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1798         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1799         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1800          * allow 0-length string (needed to remove an alias).
1801          */
1802         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1803         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1804         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1805         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1806         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1807         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1808         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1809         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1810         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1811         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1812         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1813         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1814         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1815         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1816         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1817         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1818         [IFLA_XDP]              = { .type = NLA_NESTED },
1819         [IFLA_EVENT]            = { .type = NLA_U32 },
1820         [IFLA_GROUP]            = { .type = NLA_U32 },
1821         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1822         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1823         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1824         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1825         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1826         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1827         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1828                                     .len = ALTIFNAMSIZ - 1 },
1829         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1830 };
1831
1832 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1833         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1834         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1835         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1836         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1837 };
1838
1839 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1840         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1841         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1842         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1843         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1844         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1845         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1846         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1847         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1848         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1849         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1850         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1851         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1852         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1853 };
1854
1855 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1856         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1857         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1858                                     .len = PORT_PROFILE_MAX },
1859         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1860                                       .len = PORT_UUID_MAX },
1861         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1862                                     .len = PORT_UUID_MAX },
1863         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1864         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1865
1866         /* Unused, but we need to keep it here since user space could
1867          * fill it. It's also broken with regard to NLA_BINARY use in
1868          * combination with structs.
1869          */
1870         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1871                                     .len = sizeof(struct ifla_port_vsi) },
1872 };
1873
1874 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1875         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1876         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1877         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1878         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1879 };
1880
1881 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1882 {
1883         const struct rtnl_link_ops *ops = NULL;
1884         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1885
1886         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1887                 return NULL;
1888
1889         if (linfo[IFLA_INFO_KIND]) {
1890                 char kind[MODULE_NAME_LEN];
1891
1892                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1893                 ops = rtnl_link_ops_get(kind);
1894         }
1895
1896         return ops;
1897 }
1898
1899 static bool link_master_filtered(struct net_device *dev, int master_idx)
1900 {
1901         struct net_device *master;
1902
1903         if (!master_idx)
1904                 return false;
1905
1906         master = netdev_master_upper_dev_get(dev);
1907         if (!master || master->ifindex != master_idx)
1908                 return true;
1909
1910         return false;
1911 }
1912
1913 static bool link_kind_filtered(const struct net_device *dev,
1914                                const struct rtnl_link_ops *kind_ops)
1915 {
1916         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1917                 return true;
1918
1919         return false;
1920 }
1921
1922 static bool link_dump_filtered(struct net_device *dev,
1923                                int master_idx,
1924                                const struct rtnl_link_ops *kind_ops)
1925 {
1926         if (link_master_filtered(dev, master_idx) ||
1927             link_kind_filtered(dev, kind_ops))
1928                 return true;
1929
1930         return false;
1931 }
1932
1933 /**
1934  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1935  * @sk: netlink socket
1936  * @netnsid: network namespace identifier
1937  *
1938  * Returns the network namespace identified by netnsid on success or an error
1939  * pointer on failure.
1940  */
1941 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1942 {
1943         struct net *net;
1944
1945         net = get_net_ns_by_id(sock_net(sk), netnsid);
1946         if (!net)
1947                 return ERR_PTR(-EINVAL);
1948
1949         /* For now, the caller is required to have CAP_NET_ADMIN in
1950          * the user namespace owning the target net ns.
1951          */
1952         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1953                 put_net(net);
1954                 return ERR_PTR(-EACCES);
1955         }
1956         return net;
1957 }
1958 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1959
1960 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1961                                       bool strict_check, struct nlattr **tb,
1962                                       struct netlink_ext_ack *extack)
1963 {
1964         int hdrlen;
1965
1966         if (strict_check) {
1967                 struct ifinfomsg *ifm;
1968
1969                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1970                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1971                         return -EINVAL;
1972                 }
1973
1974                 ifm = nlmsg_data(nlh);
1975                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1976                     ifm->ifi_change) {
1977                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1978                         return -EINVAL;
1979                 }
1980                 if (ifm->ifi_index) {
1981                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1982                         return -EINVAL;
1983                 }
1984
1985                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1986                                                      IFLA_MAX, ifla_policy,
1987                                                      extack);
1988         }
1989
1990         /* A hack to preserve kernel<->userspace interface.
1991          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1992          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1993          * what iproute2 < v3.9.0 used.
1994          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1995          * attribute, its netlink message is shorter than struct ifinfomsg.
1996          */
1997         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1998                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1999
2000         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2001                                       extack);
2002 }
2003
2004 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2005 {
2006         struct netlink_ext_ack *extack = cb->extack;
2007         const struct nlmsghdr *nlh = cb->nlh;
2008         struct net *net = sock_net(skb->sk);
2009         struct net *tgt_net = net;
2010         int h, s_h;
2011         int idx = 0, s_idx;
2012         struct net_device *dev;
2013         struct hlist_head *head;
2014         struct nlattr *tb[IFLA_MAX+1];
2015         u32 ext_filter_mask = 0;
2016         const struct rtnl_link_ops *kind_ops = NULL;
2017         unsigned int flags = NLM_F_MULTI;
2018         int master_idx = 0;
2019         int netnsid = -1;
2020         int err, i;
2021
2022         s_h = cb->args[0];
2023         s_idx = cb->args[1];
2024
2025         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2026         if (err < 0) {
2027                 if (cb->strict_check)
2028                         return err;
2029
2030                 goto walk_entries;
2031         }
2032
2033         for (i = 0; i <= IFLA_MAX; ++i) {
2034                 if (!tb[i])
2035                         continue;
2036
2037                 /* new attributes should only be added with strict checking */
2038                 switch (i) {
2039                 case IFLA_TARGET_NETNSID:
2040                         netnsid = nla_get_s32(tb[i]);
2041                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2042                         if (IS_ERR(tgt_net)) {
2043                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2044                                 return PTR_ERR(tgt_net);
2045                         }
2046                         break;
2047                 case IFLA_EXT_MASK:
2048                         ext_filter_mask = nla_get_u32(tb[i]);
2049                         break;
2050                 case IFLA_MASTER:
2051                         master_idx = nla_get_u32(tb[i]);
2052                         break;
2053                 case IFLA_LINKINFO:
2054                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2055                         break;
2056                 default:
2057                         if (cb->strict_check) {
2058                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2059                                 return -EINVAL;
2060                         }
2061                 }
2062         }
2063
2064         if (master_idx || kind_ops)
2065                 flags |= NLM_F_DUMP_FILTERED;
2066
2067 walk_entries:
2068         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2069                 idx = 0;
2070                 head = &tgt_net->dev_index_head[h];
2071                 hlist_for_each_entry(dev, head, index_hlist) {
2072                         if (link_dump_filtered(dev, master_idx, kind_ops))
2073                                 goto cont;
2074                         if (idx < s_idx)
2075                                 goto cont;
2076                         err = rtnl_fill_ifinfo(skb, dev, net,
2077                                                RTM_NEWLINK,
2078                                                NETLINK_CB(cb->skb).portid,
2079                                                nlh->nlmsg_seq, 0, flags,
2080                                                ext_filter_mask, 0, NULL, 0,
2081                                                netnsid, GFP_KERNEL);
2082
2083                         if (err < 0) {
2084                                 if (likely(skb->len))
2085                                         goto out;
2086
2087                                 goto out_err;
2088                         }
2089 cont:
2090                         idx++;
2091                 }
2092         }
2093 out:
2094         err = skb->len;
2095 out_err:
2096         cb->args[1] = idx;
2097         cb->args[0] = h;
2098         cb->seq = net->dev_base_seq;
2099         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2100         if (netnsid >= 0)
2101                 put_net(tgt_net);
2102
2103         return err;
2104 }
2105
2106 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2107                         struct netlink_ext_ack *exterr)
2108 {
2109         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2110                                     exterr);
2111 }
2112 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2113
2114 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2115 {
2116         struct net *net;
2117         /* Examine the link attributes and figure out which
2118          * network namespace we are talking about.
2119          */
2120         if (tb[IFLA_NET_NS_PID])
2121                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2122         else if (tb[IFLA_NET_NS_FD])
2123                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2124         else
2125                 net = get_net(src_net);
2126         return net;
2127 }
2128 EXPORT_SYMBOL(rtnl_link_get_net);
2129
2130 /* Figure out which network namespace we are talking about by
2131  * examining the link attributes in the following order:
2132  *
2133  * 1. IFLA_NET_NS_PID
2134  * 2. IFLA_NET_NS_FD
2135  * 3. IFLA_TARGET_NETNSID
2136  */
2137 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2138                                                struct nlattr *tb[])
2139 {
2140         struct net *net;
2141
2142         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2143                 return rtnl_link_get_net(src_net, tb);
2144
2145         if (!tb[IFLA_TARGET_NETNSID])
2146                 return get_net(src_net);
2147
2148         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2149         if (!net)
2150                 return ERR_PTR(-EINVAL);
2151
2152         return net;
2153 }
2154
2155 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2156                                              struct net *src_net,
2157                                              struct nlattr *tb[], int cap)
2158 {
2159         struct net *net;
2160
2161         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2162         if (IS_ERR(net))
2163                 return net;
2164
2165         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2166                 put_net(net);
2167                 return ERR_PTR(-EPERM);
2168         }
2169
2170         return net;
2171 }
2172
2173 /* Verify that rtnetlink requests do not pass additional properties
2174  * potentially referring to different network namespaces.
2175  */
2176 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2177                                     struct netlink_ext_ack *extack,
2178                                     bool netns_id_only)
2179 {
2180
2181         if (netns_id_only) {
2182                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2183                         return 0;
2184
2185                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2186                 return -EOPNOTSUPP;
2187         }
2188
2189         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2190                 goto invalid_attr;
2191
2192         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2193                 goto invalid_attr;
2194
2195         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2196                 goto invalid_attr;
2197
2198         return 0;
2199
2200 invalid_attr:
2201         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2202         return -EINVAL;
2203 }
2204
2205 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2206 {
2207         if (dev) {
2208                 if (tb[IFLA_ADDRESS] &&
2209                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2210                         return -EINVAL;
2211
2212                 if (tb[IFLA_BROADCAST] &&
2213                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2214                         return -EINVAL;
2215         }
2216
2217         if (tb[IFLA_AF_SPEC]) {
2218                 struct nlattr *af;
2219                 int rem, err;
2220
2221                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2222                         const struct rtnl_af_ops *af_ops;
2223
2224                         rcu_read_lock();
2225                         af_ops = rtnl_af_lookup(nla_type(af));
2226                         if (!af_ops) {
2227                                 rcu_read_unlock();
2228                                 return -EAFNOSUPPORT;
2229                         }
2230
2231                         if (!af_ops->set_link_af) {
2232                                 rcu_read_unlock();
2233                                 return -EOPNOTSUPP;
2234                         }
2235
2236                         if (af_ops->validate_link_af) {
2237                                 err = af_ops->validate_link_af(dev, af);
2238                                 if (err < 0) {
2239                                         rcu_read_unlock();
2240                                         return err;
2241                                 }
2242                         }
2243
2244                         rcu_read_unlock();
2245                 }
2246         }
2247
2248         return 0;
2249 }
2250
2251 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2252                                   int guid_type)
2253 {
2254         const struct net_device_ops *ops = dev->netdev_ops;
2255
2256         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2257 }
2258
2259 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2260 {
2261         if (dev->type != ARPHRD_INFINIBAND)
2262                 return -EOPNOTSUPP;
2263
2264         return handle_infiniband_guid(dev, ivt, guid_type);
2265 }
2266
2267 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2268 {
2269         const struct net_device_ops *ops = dev->netdev_ops;
2270         int err = -EINVAL;
2271
2272         if (tb[IFLA_VF_MAC]) {
2273                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2274
2275                 if (ivm->vf >= INT_MAX)
2276                         return -EINVAL;
2277                 err = -EOPNOTSUPP;
2278                 if (ops->ndo_set_vf_mac)
2279                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2280                                                   ivm->mac);
2281                 if (err < 0)
2282                         return err;
2283         }
2284
2285         if (tb[IFLA_VF_VLAN]) {
2286                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2287
2288                 if (ivv->vf >= INT_MAX)
2289                         return -EINVAL;
2290                 err = -EOPNOTSUPP;
2291                 if (ops->ndo_set_vf_vlan)
2292                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2293                                                    ivv->qos,
2294                                                    htons(ETH_P_8021Q));
2295                 if (err < 0)
2296                         return err;
2297         }
2298
2299         if (tb[IFLA_VF_VLAN_LIST]) {
2300                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2301                 struct nlattr *attr;
2302                 int rem, len = 0;
2303
2304                 err = -EOPNOTSUPP;
2305                 if (!ops->ndo_set_vf_vlan)
2306                         return err;
2307
2308                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2309                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2310                             nla_len(attr) < NLA_HDRLEN) {
2311                                 return -EINVAL;
2312                         }
2313                         if (len >= MAX_VLAN_LIST_LEN)
2314                                 return -EOPNOTSUPP;
2315                         ivvl[len] = nla_data(attr);
2316
2317                         len++;
2318                 }
2319                 if (len == 0)
2320                         return -EINVAL;
2321
2322                 if (ivvl[0]->vf >= INT_MAX)
2323                         return -EINVAL;
2324                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2325                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2326                 if (err < 0)
2327                         return err;
2328         }
2329
2330         if (tb[IFLA_VF_TX_RATE]) {
2331                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2332                 struct ifla_vf_info ivf;
2333
2334                 if (ivt->vf >= INT_MAX)
2335                         return -EINVAL;
2336                 err = -EOPNOTSUPP;
2337                 if (ops->ndo_get_vf_config)
2338                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2339                 if (err < 0)
2340                         return err;
2341
2342                 err = -EOPNOTSUPP;
2343                 if (ops->ndo_set_vf_rate)
2344                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2345                                                    ivf.min_tx_rate,
2346                                                    ivt->rate);
2347                 if (err < 0)
2348                         return err;
2349         }
2350
2351         if (tb[IFLA_VF_RATE]) {
2352                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2353
2354                 if (ivt->vf >= INT_MAX)
2355                         return -EINVAL;
2356                 err = -EOPNOTSUPP;
2357                 if (ops->ndo_set_vf_rate)
2358                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2359                                                    ivt->min_tx_rate,
2360                                                    ivt->max_tx_rate);
2361                 if (err < 0)
2362                         return err;
2363         }
2364
2365         if (tb[IFLA_VF_SPOOFCHK]) {
2366                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2367
2368                 if (ivs->vf >= INT_MAX)
2369                         return -EINVAL;
2370                 err = -EOPNOTSUPP;
2371                 if (ops->ndo_set_vf_spoofchk)
2372                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2373                                                        ivs->setting);
2374                 if (err < 0)
2375                         return err;
2376         }
2377
2378         if (tb[IFLA_VF_LINK_STATE]) {
2379                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2380
2381                 if (ivl->vf >= INT_MAX)
2382                         return -EINVAL;
2383                 err = -EOPNOTSUPP;
2384                 if (ops->ndo_set_vf_link_state)
2385                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2386                                                          ivl->link_state);
2387                 if (err < 0)
2388                         return err;
2389         }
2390
2391         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2392                 struct ifla_vf_rss_query_en *ivrssq_en;
2393
2394                 err = -EOPNOTSUPP;
2395                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2396                 if (ivrssq_en->vf >= INT_MAX)
2397                         return -EINVAL;
2398                 if (ops->ndo_set_vf_rss_query_en)
2399                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2400                                                            ivrssq_en->setting);
2401                 if (err < 0)
2402                         return err;
2403         }
2404
2405         if (tb[IFLA_VF_TRUST]) {
2406                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2407
2408                 if (ivt->vf >= INT_MAX)
2409                         return -EINVAL;
2410                 err = -EOPNOTSUPP;
2411                 if (ops->ndo_set_vf_trust)
2412                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2413                 if (err < 0)
2414                         return err;
2415         }
2416
2417         if (tb[IFLA_VF_IB_NODE_GUID]) {
2418                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2419
2420                 if (ivt->vf >= INT_MAX)
2421                         return -EINVAL;
2422                 if (!ops->ndo_set_vf_guid)
2423                         return -EOPNOTSUPP;
2424                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2425         }
2426
2427         if (tb[IFLA_VF_IB_PORT_GUID]) {
2428                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2429
2430                 if (ivt->vf >= INT_MAX)
2431                         return -EINVAL;
2432                 if (!ops->ndo_set_vf_guid)
2433                         return -EOPNOTSUPP;
2434
2435                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2436         }
2437
2438         return err;
2439 }
2440
2441 static int do_set_master(struct net_device *dev, int ifindex,
2442                          struct netlink_ext_ack *extack)
2443 {
2444         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2445         const struct net_device_ops *ops;
2446         int err;
2447
2448         if (upper_dev) {
2449                 if (upper_dev->ifindex == ifindex)
2450                         return 0;
2451                 ops = upper_dev->netdev_ops;
2452                 if (ops->ndo_del_slave) {
2453                         err = ops->ndo_del_slave(upper_dev, dev);
2454                         if (err)
2455                                 return err;
2456                         netdev_update_lockdep_key(dev);
2457                 } else {
2458                         return -EOPNOTSUPP;
2459                 }
2460         }
2461
2462         if (ifindex) {
2463                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2464                 if (!upper_dev)
2465                         return -EINVAL;
2466                 ops = upper_dev->netdev_ops;
2467                 if (ops->ndo_add_slave) {
2468                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2469                         if (err)
2470                                 return err;
2471                 } else {
2472                         return -EOPNOTSUPP;
2473                 }
2474         }
2475         return 0;
2476 }
2477
2478 #define DO_SETLINK_MODIFIED     0x01
2479 /* notify flag means notify + modified. */
2480 #define DO_SETLINK_NOTIFY       0x03
2481 static int do_setlink(const struct sk_buff *skb,
2482                       struct net_device *dev, struct ifinfomsg *ifm,
2483                       struct netlink_ext_ack *extack,
2484                       struct nlattr **tb, char *ifname, int status)
2485 {
2486         const struct net_device_ops *ops = dev->netdev_ops;
2487         int err;
2488
2489         err = validate_linkmsg(dev, tb);
2490         if (err < 0)
2491                 return err;
2492
2493         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2494                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2495                                                             tb, CAP_NET_ADMIN);
2496                 if (IS_ERR(net)) {
2497                         err = PTR_ERR(net);
2498                         goto errout;
2499                 }
2500
2501                 err = dev_change_net_namespace(dev, net, ifname);
2502                 put_net(net);
2503                 if (err)
2504                         goto errout;
2505                 status |= DO_SETLINK_MODIFIED;
2506         }
2507
2508         if (tb[IFLA_MAP]) {
2509                 struct rtnl_link_ifmap *u_map;
2510                 struct ifmap k_map;
2511
2512                 if (!ops->ndo_set_config) {
2513                         err = -EOPNOTSUPP;
2514                         goto errout;
2515                 }
2516
2517                 if (!netif_device_present(dev)) {
2518                         err = -ENODEV;
2519                         goto errout;
2520                 }
2521
2522                 u_map = nla_data(tb[IFLA_MAP]);
2523                 k_map.mem_start = (unsigned long) u_map->mem_start;
2524                 k_map.mem_end = (unsigned long) u_map->mem_end;
2525                 k_map.base_addr = (unsigned short) u_map->base_addr;
2526                 k_map.irq = (unsigned char) u_map->irq;
2527                 k_map.dma = (unsigned char) u_map->dma;
2528                 k_map.port = (unsigned char) u_map->port;
2529
2530                 err = ops->ndo_set_config(dev, &k_map);
2531                 if (err < 0)
2532                         goto errout;
2533
2534                 status |= DO_SETLINK_NOTIFY;
2535         }
2536
2537         if (tb[IFLA_ADDRESS]) {
2538                 struct sockaddr *sa;
2539                 int len;
2540
2541                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2542                                                   sizeof(*sa));
2543                 sa = kmalloc(len, GFP_KERNEL);
2544                 if (!sa) {
2545                         err = -ENOMEM;
2546                         goto errout;
2547                 }
2548                 sa->sa_family = dev->type;
2549                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2550                        dev->addr_len);
2551                 err = dev_set_mac_address(dev, sa, extack);
2552                 kfree(sa);
2553                 if (err)
2554                         goto errout;
2555                 status |= DO_SETLINK_MODIFIED;
2556         }
2557
2558         if (tb[IFLA_MTU]) {
2559                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2560                 if (err < 0)
2561                         goto errout;
2562                 status |= DO_SETLINK_MODIFIED;
2563         }
2564
2565         if (tb[IFLA_GROUP]) {
2566                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2567                 status |= DO_SETLINK_NOTIFY;
2568         }
2569
2570         /*
2571          * Interface selected by interface index but interface
2572          * name provided implies that a name change has been
2573          * requested.
2574          */
2575         if (ifm->ifi_index > 0 && ifname[0]) {
2576                 err = dev_change_name(dev, ifname);
2577                 if (err < 0)
2578                         goto errout;
2579                 status |= DO_SETLINK_MODIFIED;
2580         }
2581
2582         if (tb[IFLA_IFALIAS]) {
2583                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2584                                     nla_len(tb[IFLA_IFALIAS]));
2585                 if (err < 0)
2586                         goto errout;
2587                 status |= DO_SETLINK_NOTIFY;
2588         }
2589
2590         if (tb[IFLA_BROADCAST]) {
2591                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2592                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2593         }
2594
2595         if (ifm->ifi_flags || ifm->ifi_change) {
2596                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2597                                        extack);
2598                 if (err < 0)
2599                         goto errout;
2600         }
2601
2602         if (tb[IFLA_MASTER]) {
2603                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2604                 if (err)
2605                         goto errout;
2606                 status |= DO_SETLINK_MODIFIED;
2607         }
2608
2609         if (tb[IFLA_CARRIER]) {
2610                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2611                 if (err)
2612                         goto errout;
2613                 status |= DO_SETLINK_MODIFIED;
2614         }
2615
2616         if (tb[IFLA_TXQLEN]) {
2617                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2618
2619                 err = dev_change_tx_queue_len(dev, value);
2620                 if (err)
2621                         goto errout;
2622                 status |= DO_SETLINK_MODIFIED;
2623         }
2624
2625         if (tb[IFLA_GSO_MAX_SIZE]) {
2626                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2627
2628                 if (max_size > GSO_MAX_SIZE) {
2629                         err = -EINVAL;
2630                         goto errout;
2631                 }
2632
2633                 if (dev->gso_max_size ^ max_size) {
2634                         netif_set_gso_max_size(dev, max_size);
2635                         status |= DO_SETLINK_MODIFIED;
2636                 }
2637         }
2638
2639         if (tb[IFLA_GSO_MAX_SEGS]) {
2640                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2641
2642                 if (max_segs > GSO_MAX_SEGS) {
2643                         err = -EINVAL;
2644                         goto errout;
2645                 }
2646
2647                 if (dev->gso_max_segs ^ max_segs) {
2648                         dev->gso_max_segs = max_segs;
2649                         status |= DO_SETLINK_MODIFIED;
2650                 }
2651         }
2652
2653         if (tb[IFLA_OPERSTATE])
2654                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2655
2656         if (tb[IFLA_LINKMODE]) {
2657                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2658
2659                 write_lock_bh(&dev_base_lock);
2660                 if (dev->link_mode ^ value)
2661                         status |= DO_SETLINK_NOTIFY;
2662                 dev->link_mode = value;
2663                 write_unlock_bh(&dev_base_lock);
2664         }
2665
2666         if (tb[IFLA_VFINFO_LIST]) {
2667                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2668                 struct nlattr *attr;
2669                 int rem;
2670
2671                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2672                         if (nla_type(attr) != IFLA_VF_INFO ||
2673                             nla_len(attr) < NLA_HDRLEN) {
2674                                 err = -EINVAL;
2675                                 goto errout;
2676                         }
2677                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2678                                                           attr,
2679                                                           ifla_vf_policy,
2680                                                           NULL);
2681                         if (err < 0)
2682                                 goto errout;
2683                         err = do_setvfinfo(dev, vfinfo);
2684                         if (err < 0)
2685                                 goto errout;
2686                         status |= DO_SETLINK_NOTIFY;
2687                 }
2688         }
2689         err = 0;
2690
2691         if (tb[IFLA_VF_PORTS]) {
2692                 struct nlattr *port[IFLA_PORT_MAX+1];
2693                 struct nlattr *attr;
2694                 int vf;
2695                 int rem;
2696
2697                 err = -EOPNOTSUPP;
2698                 if (!ops->ndo_set_vf_port)
2699                         goto errout;
2700
2701                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2702                         if (nla_type(attr) != IFLA_VF_PORT ||
2703                             nla_len(attr) < NLA_HDRLEN) {
2704                                 err = -EINVAL;
2705                                 goto errout;
2706                         }
2707                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2708                                                           attr,
2709                                                           ifla_port_policy,
2710                                                           NULL);
2711                         if (err < 0)
2712                                 goto errout;
2713                         if (!port[IFLA_PORT_VF]) {
2714                                 err = -EOPNOTSUPP;
2715                                 goto errout;
2716                         }
2717                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2718                         err = ops->ndo_set_vf_port(dev, vf, port);
2719                         if (err < 0)
2720                                 goto errout;
2721                         status |= DO_SETLINK_NOTIFY;
2722                 }
2723         }
2724         err = 0;
2725
2726         if (tb[IFLA_PORT_SELF]) {
2727                 struct nlattr *port[IFLA_PORT_MAX+1];
2728
2729                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2730                                                   tb[IFLA_PORT_SELF],
2731                                                   ifla_port_policy, NULL);
2732                 if (err < 0)
2733                         goto errout;
2734
2735                 err = -EOPNOTSUPP;
2736                 if (ops->ndo_set_vf_port)
2737                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2738                 if (err < 0)
2739                         goto errout;
2740                 status |= DO_SETLINK_NOTIFY;
2741         }
2742
2743         if (tb[IFLA_AF_SPEC]) {
2744                 struct nlattr *af;
2745                 int rem;
2746
2747                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2748                         const struct rtnl_af_ops *af_ops;
2749
2750                         rcu_read_lock();
2751
2752                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2753
2754                         err = af_ops->set_link_af(dev, af);
2755                         if (err < 0) {
2756                                 rcu_read_unlock();
2757                                 goto errout;
2758                         }
2759
2760                         rcu_read_unlock();
2761                         status |= DO_SETLINK_NOTIFY;
2762                 }
2763         }
2764         err = 0;
2765
2766         if (tb[IFLA_PROTO_DOWN]) {
2767                 err = dev_change_proto_down(dev,
2768                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2769                 if (err)
2770                         goto errout;
2771                 status |= DO_SETLINK_NOTIFY;
2772         }
2773
2774         if (tb[IFLA_XDP]) {
2775                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2776                 u32 xdp_flags = 0;
2777
2778                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2779                                                   tb[IFLA_XDP],
2780                                                   ifla_xdp_policy, NULL);
2781                 if (err < 0)
2782                         goto errout;
2783
2784                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2785                         err = -EINVAL;
2786                         goto errout;
2787                 }
2788
2789                 if (xdp[IFLA_XDP_FLAGS]) {
2790                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2791                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2792                                 err = -EINVAL;
2793                                 goto errout;
2794                         }
2795                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2796                                 err = -EINVAL;
2797                                 goto errout;
2798                         }
2799                 }
2800
2801                 if (xdp[IFLA_XDP_FD]) {
2802                         err = dev_change_xdp_fd(dev, extack,
2803                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2804                                                 xdp_flags);
2805                         if (err)
2806                                 goto errout;
2807                         status |= DO_SETLINK_NOTIFY;
2808                 }
2809         }
2810
2811 errout:
2812         if (status & DO_SETLINK_MODIFIED) {
2813                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2814                         netdev_state_change(dev);
2815
2816                 if (err < 0)
2817                         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",
2818                                              dev->name);
2819         }
2820
2821         return err;
2822 }
2823
2824 static struct net_device *rtnl_dev_get(struct net *net,
2825                                        struct nlattr *ifname_attr,
2826                                        struct nlattr *altifname_attr,
2827                                        char *ifname)
2828 {
2829         char buffer[ALTIFNAMSIZ];
2830
2831         if (!ifname) {
2832                 ifname = buffer;
2833                 if (ifname_attr)
2834                         nla_strlcpy(ifname, ifname_attr, IFNAMSIZ);
2835                 else if (altifname_attr)
2836                         nla_strlcpy(ifname, altifname_attr, ALTIFNAMSIZ);
2837                 else
2838                         return NULL;
2839         }
2840
2841         return __dev_get_by_name(net, ifname);
2842 }
2843
2844 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2845                         struct netlink_ext_ack *extack)
2846 {
2847         struct net *net = sock_net(skb->sk);
2848         struct ifinfomsg *ifm;
2849         struct net_device *dev;
2850         int err;
2851         struct nlattr *tb[IFLA_MAX+1];
2852         char ifname[IFNAMSIZ];
2853
2854         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2855                                      ifla_policy, extack);
2856         if (err < 0)
2857                 goto errout;
2858
2859         err = rtnl_ensure_unique_netns(tb, extack, false);
2860         if (err < 0)
2861                 goto errout;
2862
2863         if (tb[IFLA_IFNAME])
2864                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2865         else
2866                 ifname[0] = '\0';
2867
2868         err = -EINVAL;
2869         ifm = nlmsg_data(nlh);
2870         if (ifm->ifi_index > 0)
2871                 dev = __dev_get_by_index(net, ifm->ifi_index);
2872         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2873                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2874         else
2875                 goto errout;
2876
2877         if (dev == NULL) {
2878                 err = -ENODEV;
2879                 goto errout;
2880         }
2881
2882         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2883 errout:
2884         return err;
2885 }
2886
2887 static int rtnl_group_dellink(const struct net *net, int group)
2888 {
2889         struct net_device *dev, *aux;
2890         LIST_HEAD(list_kill);
2891         bool found = false;
2892
2893         if (!group)
2894                 return -EPERM;
2895
2896         for_each_netdev(net, dev) {
2897                 if (dev->group == group) {
2898                         const struct rtnl_link_ops *ops;
2899
2900                         found = true;
2901                         ops = dev->rtnl_link_ops;
2902                         if (!ops || !ops->dellink)
2903                                 return -EOPNOTSUPP;
2904                 }
2905         }
2906
2907         if (!found)
2908                 return -ENODEV;
2909
2910         for_each_netdev_safe(net, dev, aux) {
2911                 if (dev->group == group) {
2912                         const struct rtnl_link_ops *ops;
2913
2914                         ops = dev->rtnl_link_ops;
2915                         ops->dellink(dev, &list_kill);
2916                 }
2917         }
2918         unregister_netdevice_many(&list_kill);
2919
2920         return 0;
2921 }
2922
2923 int rtnl_delete_link(struct net_device *dev)
2924 {
2925         const struct rtnl_link_ops *ops;
2926         LIST_HEAD(list_kill);
2927
2928         ops = dev->rtnl_link_ops;
2929         if (!ops || !ops->dellink)
2930                 return -EOPNOTSUPP;
2931
2932         ops->dellink(dev, &list_kill);
2933         unregister_netdevice_many(&list_kill);
2934
2935         return 0;
2936 }
2937 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2938
2939 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2940                         struct netlink_ext_ack *extack)
2941 {
2942         struct net *net = sock_net(skb->sk);
2943         struct net *tgt_net = net;
2944         struct net_device *dev = NULL;
2945         struct ifinfomsg *ifm;
2946         struct nlattr *tb[IFLA_MAX+1];
2947         int err;
2948         int netnsid = -1;
2949
2950         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2951                                      ifla_policy, extack);
2952         if (err < 0)
2953                 return err;
2954
2955         err = rtnl_ensure_unique_netns(tb, extack, true);
2956         if (err < 0)
2957                 return err;
2958
2959         if (tb[IFLA_TARGET_NETNSID]) {
2960                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2961                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2962                 if (IS_ERR(tgt_net))
2963                         return PTR_ERR(tgt_net);
2964         }
2965
2966         err = -EINVAL;
2967         ifm = nlmsg_data(nlh);
2968         if (ifm->ifi_index > 0)
2969                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2970         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2971                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
2972                                    tb[IFLA_ALT_IFNAME], NULL);
2973         else if (tb[IFLA_GROUP])
2974                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2975         else
2976                 goto out;
2977
2978         if (!dev) {
2979                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2980                         err = -ENODEV;
2981
2982                 goto out;
2983         }
2984
2985         err = rtnl_delete_link(dev);
2986
2987 out:
2988         if (netnsid >= 0)
2989                 put_net(tgt_net);
2990
2991         return err;
2992 }
2993
2994 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2995 {
2996         unsigned int old_flags;
2997         int err;
2998
2999         old_flags = dev->flags;
3000         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3001                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3002                                          NULL);
3003                 if (err < 0)
3004                         return err;
3005         }
3006
3007         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3008                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3009         } else {
3010                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3011                 __dev_notify_flags(dev, old_flags, ~0U);
3012         }
3013         return 0;
3014 }
3015 EXPORT_SYMBOL(rtnl_configure_link);
3016
3017 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3018                                     unsigned char name_assign_type,
3019                                     const struct rtnl_link_ops *ops,
3020                                     struct nlattr *tb[],
3021                                     struct netlink_ext_ack *extack)
3022 {
3023         struct net_device *dev;
3024         unsigned int num_tx_queues = 1;
3025         unsigned int num_rx_queues = 1;
3026
3027         if (tb[IFLA_NUM_TX_QUEUES])
3028                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3029         else if (ops->get_num_tx_queues)
3030                 num_tx_queues = ops->get_num_tx_queues();
3031
3032         if (tb[IFLA_NUM_RX_QUEUES])
3033                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3034         else if (ops->get_num_rx_queues)
3035                 num_rx_queues = ops->get_num_rx_queues();
3036
3037         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3038                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3039                 return ERR_PTR(-EINVAL);
3040         }
3041
3042         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3043                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3044                 return ERR_PTR(-EINVAL);
3045         }
3046
3047         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3048                                ops->setup, num_tx_queues, num_rx_queues);
3049         if (!dev)
3050                 return ERR_PTR(-ENOMEM);
3051
3052         dev_net_set(dev, net);
3053         dev->rtnl_link_ops = ops;
3054         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3055
3056         if (tb[IFLA_MTU]) {
3057                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3058                 int err;
3059
3060                 err = dev_validate_mtu(dev, mtu, extack);
3061                 if (err) {
3062                         free_netdev(dev);
3063                         return ERR_PTR(err);
3064                 }
3065                 dev->mtu = mtu;
3066         }
3067         if (tb[IFLA_ADDRESS]) {
3068                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3069                                 nla_len(tb[IFLA_ADDRESS]));
3070                 dev->addr_assign_type = NET_ADDR_SET;
3071         }
3072         if (tb[IFLA_BROADCAST])
3073                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3074                                 nla_len(tb[IFLA_BROADCAST]));
3075         if (tb[IFLA_TXQLEN])
3076                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3077         if (tb[IFLA_OPERSTATE])
3078                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3079         if (tb[IFLA_LINKMODE])
3080                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3081         if (tb[IFLA_GROUP])
3082                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3083         if (tb[IFLA_GSO_MAX_SIZE])
3084                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3085         if (tb[IFLA_GSO_MAX_SEGS])
3086                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3087
3088         return dev;
3089 }
3090 EXPORT_SYMBOL(rtnl_create_link);
3091
3092 static int rtnl_group_changelink(const struct sk_buff *skb,
3093                 struct net *net, int group,
3094                 struct ifinfomsg *ifm,
3095                 struct netlink_ext_ack *extack,
3096                 struct nlattr **tb)
3097 {
3098         struct net_device *dev, *aux;
3099         int err;
3100
3101         for_each_netdev_safe(net, dev, aux) {
3102                 if (dev->group == group) {
3103                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3104                         if (err < 0)
3105                                 return err;
3106                 }
3107         }
3108
3109         return 0;
3110 }
3111
3112 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3113                           struct nlattr **attr, struct netlink_ext_ack *extack)
3114 {
3115         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3116         unsigned char name_assign_type = NET_NAME_USER;
3117         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3118         const struct rtnl_link_ops *m_ops = NULL;
3119         struct net_device *master_dev = NULL;
3120         struct net *net = sock_net(skb->sk);
3121         const struct rtnl_link_ops *ops;
3122         struct nlattr *tb[IFLA_MAX + 1];
3123         struct net *dest_net, *link_net;
3124         struct nlattr **slave_data;
3125         char kind[MODULE_NAME_LEN];
3126         struct net_device *dev;
3127         struct ifinfomsg *ifm;
3128         char ifname[IFNAMSIZ];
3129         struct nlattr **data;
3130         int err;
3131
3132 #ifdef CONFIG_MODULES
3133 replay:
3134 #endif
3135         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3136                                      ifla_policy, extack);
3137         if (err < 0)
3138                 return err;
3139
3140         err = rtnl_ensure_unique_netns(tb, extack, false);
3141         if (err < 0)
3142                 return err;
3143
3144         if (tb[IFLA_IFNAME])
3145                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3146         else
3147                 ifname[0] = '\0';
3148
3149         ifm = nlmsg_data(nlh);
3150         if (ifm->ifi_index > 0)
3151                 dev = __dev_get_by_index(net, ifm->ifi_index);
3152         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3153                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3154         else
3155                 dev = NULL;
3156
3157         if (dev) {
3158                 master_dev = netdev_master_upper_dev_get(dev);
3159                 if (master_dev)
3160                         m_ops = master_dev->rtnl_link_ops;
3161         }
3162
3163         err = validate_linkmsg(dev, tb);
3164         if (err < 0)
3165                 return err;
3166
3167         if (tb[IFLA_LINKINFO]) {
3168                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3169                                                   tb[IFLA_LINKINFO],
3170                                                   ifla_info_policy, NULL);
3171                 if (err < 0)
3172                         return err;
3173         } else
3174                 memset(linkinfo, 0, sizeof(linkinfo));
3175
3176         if (linkinfo[IFLA_INFO_KIND]) {
3177                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3178                 ops = rtnl_link_ops_get(kind);
3179         } else {
3180                 kind[0] = '\0';
3181                 ops = NULL;
3182         }
3183
3184         data = NULL;
3185         if (ops) {
3186                 if (ops->maxtype > RTNL_MAX_TYPE)
3187                         return -EINVAL;
3188
3189                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3190                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3191                                                           linkinfo[IFLA_INFO_DATA],
3192                                                           ops->policy, extack);
3193                         if (err < 0)
3194                                 return err;
3195                         data = attr;
3196                 }
3197                 if (ops->validate) {
3198                         err = ops->validate(tb, data, extack);
3199                         if (err < 0)
3200                                 return err;
3201                 }
3202         }
3203
3204         slave_data = NULL;
3205         if (m_ops) {
3206                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3207                         return -EINVAL;
3208
3209                 if (m_ops->slave_maxtype &&
3210                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3211                         err = nla_parse_nested_deprecated(slave_attr,
3212                                                           m_ops->slave_maxtype,
3213                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3214                                                           m_ops->slave_policy,
3215                                                           extack);
3216                         if (err < 0)
3217                                 return err;
3218                         slave_data = slave_attr;
3219                 }
3220         }
3221
3222         if (dev) {
3223                 int status = 0;
3224
3225                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3226                         return -EEXIST;
3227                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3228                         return -EOPNOTSUPP;
3229
3230                 if (linkinfo[IFLA_INFO_DATA]) {
3231                         if (!ops || ops != dev->rtnl_link_ops ||
3232                             !ops->changelink)
3233                                 return -EOPNOTSUPP;
3234
3235                         err = ops->changelink(dev, tb, data, extack);
3236                         if (err < 0)
3237                                 return err;
3238                         status |= DO_SETLINK_NOTIFY;
3239                 }
3240
3241                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3242                         if (!m_ops || !m_ops->slave_changelink)
3243                                 return -EOPNOTSUPP;
3244
3245                         err = m_ops->slave_changelink(master_dev, dev, tb,
3246                                                       slave_data, extack);
3247                         if (err < 0)
3248                                 return err;
3249                         status |= DO_SETLINK_NOTIFY;
3250                 }
3251
3252                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3253         }
3254
3255         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3256                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3257                         return rtnl_group_changelink(skb, net,
3258                                                 nla_get_u32(tb[IFLA_GROUP]),
3259                                                 ifm, extack, tb);
3260                 return -ENODEV;
3261         }
3262
3263         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3264                 return -EOPNOTSUPP;
3265
3266         if (!ops) {
3267 #ifdef CONFIG_MODULES
3268                 if (kind[0]) {
3269                         __rtnl_unlock();
3270                         request_module("rtnl-link-%s", kind);
3271                         rtnl_lock();
3272                         ops = rtnl_link_ops_get(kind);
3273                         if (ops)
3274                                 goto replay;
3275                 }
3276 #endif
3277                 NL_SET_ERR_MSG(extack, "Unknown device type");
3278                 return -EOPNOTSUPP;
3279         }
3280
3281         if (!ops->setup)
3282                 return -EOPNOTSUPP;
3283
3284         if (!ifname[0]) {
3285                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3286                 name_assign_type = NET_NAME_ENUM;
3287         }
3288
3289         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3290         if (IS_ERR(dest_net))
3291                 return PTR_ERR(dest_net);
3292
3293         if (tb[IFLA_LINK_NETNSID]) {
3294                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3295
3296                 link_net = get_net_ns_by_id(dest_net, id);
3297                 if (!link_net) {
3298                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3299                         err =  -EINVAL;
3300                         goto out;
3301                 }
3302                 err = -EPERM;
3303                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3304                         goto out;
3305         } else {
3306                 link_net = NULL;
3307         }
3308
3309         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3310                                name_assign_type, ops, tb, extack);
3311         if (IS_ERR(dev)) {
3312                 err = PTR_ERR(dev);
3313                 goto out;
3314         }
3315
3316         dev->ifindex = ifm->ifi_index;
3317
3318         if (ops->newlink) {
3319                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3320                 /* Drivers should call free_netdev() in ->destructor
3321                  * and unregister it on failure after registration
3322                  * so that device could be finally freed in rtnl_unlock.
3323                  */
3324                 if (err < 0) {
3325                         /* If device is not registered at all, free it now */
3326                         if (dev->reg_state == NETREG_UNINITIALIZED)
3327                                 free_netdev(dev);
3328                         goto out;
3329                 }
3330         } else {
3331                 err = register_netdevice(dev);
3332                 if (err < 0) {
3333                         free_netdev(dev);
3334                         goto out;
3335                 }
3336         }
3337         err = rtnl_configure_link(dev, ifm);
3338         if (err < 0)
3339                 goto out_unregister;
3340         if (link_net) {
3341                 err = dev_change_net_namespace(dev, dest_net, ifname);
3342                 if (err < 0)
3343                         goto out_unregister;
3344         }
3345         if (tb[IFLA_MASTER]) {
3346                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3347                 if (err)
3348                         goto out_unregister;
3349         }
3350 out:
3351         if (link_net)
3352                 put_net(link_net);
3353         put_net(dest_net);
3354         return err;
3355 out_unregister:
3356         if (ops->newlink) {
3357                 LIST_HEAD(list_kill);
3358
3359                 ops->dellink(dev, &list_kill);
3360                 unregister_netdevice_many(&list_kill);
3361         } else {
3362                 unregister_netdevice(dev);
3363         }
3364         goto out;
3365 }
3366
3367 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3368                         struct netlink_ext_ack *extack)
3369 {
3370         struct nlattr **attr;
3371         int ret;
3372
3373         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3374         if (!attr)
3375                 return -ENOMEM;
3376
3377         ret = __rtnl_newlink(skb, nlh, attr, extack);
3378         kfree(attr);
3379         return ret;
3380 }
3381
3382 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3383                                   const struct nlmsghdr *nlh,
3384                                   struct nlattr **tb,
3385                                   struct netlink_ext_ack *extack)
3386 {
3387         struct ifinfomsg *ifm;
3388         int i, err;
3389
3390         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3391                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3392                 return -EINVAL;
3393         }
3394
3395         if (!netlink_strict_get_check(skb))
3396                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3397                                               ifla_policy, extack);
3398
3399         ifm = nlmsg_data(nlh);
3400         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3401             ifm->ifi_change) {
3402                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3403                 return -EINVAL;
3404         }
3405
3406         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3407                                             ifla_policy, extack);
3408         if (err)
3409                 return err;
3410
3411         for (i = 0; i <= IFLA_MAX; i++) {
3412                 if (!tb[i])
3413                         continue;
3414
3415                 switch (i) {
3416                 case IFLA_IFNAME:
3417                 case IFLA_ALT_IFNAME:
3418                 case IFLA_EXT_MASK:
3419                 case IFLA_TARGET_NETNSID:
3420                         break;
3421                 default:
3422                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3423                         return -EINVAL;
3424                 }
3425         }
3426
3427         return 0;
3428 }
3429
3430 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3431                         struct netlink_ext_ack *extack)
3432 {
3433         struct net *net = sock_net(skb->sk);
3434         struct net *tgt_net = net;
3435         struct ifinfomsg *ifm;
3436         struct nlattr *tb[IFLA_MAX+1];
3437         struct net_device *dev = NULL;
3438         struct sk_buff *nskb;
3439         int netnsid = -1;
3440         int err;
3441         u32 ext_filter_mask = 0;
3442
3443         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3444         if (err < 0)
3445                 return err;
3446
3447         err = rtnl_ensure_unique_netns(tb, extack, true);
3448         if (err < 0)
3449                 return err;
3450
3451         if (tb[IFLA_TARGET_NETNSID]) {
3452                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3453                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3454                 if (IS_ERR(tgt_net))
3455                         return PTR_ERR(tgt_net);
3456         }
3457
3458         if (tb[IFLA_EXT_MASK])
3459                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3460
3461         err = -EINVAL;
3462         ifm = nlmsg_data(nlh);
3463         if (ifm->ifi_index > 0)
3464                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3465         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3466                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3467                                    tb[IFLA_ALT_IFNAME], NULL);
3468         else
3469                 goto out;
3470
3471         err = -ENODEV;
3472         if (dev == NULL)
3473                 goto out;
3474
3475         err = -ENOBUFS;
3476         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3477         if (nskb == NULL)
3478                 goto out;
3479
3480         err = rtnl_fill_ifinfo(nskb, dev, net,
3481                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3482                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3483                                0, NULL, 0, netnsid, GFP_KERNEL);
3484         if (err < 0) {
3485                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3486                 WARN_ON(err == -EMSGSIZE);
3487                 kfree_skb(nskb);
3488         } else
3489                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3490 out:
3491         if (netnsid >= 0)
3492                 put_net(tgt_net);
3493
3494         return err;
3495 }
3496
3497 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3498                            bool *changed, struct netlink_ext_ack *extack)
3499 {
3500         char *alt_ifname;
3501         int err;
3502
3503         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3504         if (err)
3505                 return err;
3506
3507         alt_ifname = nla_data(attr);
3508         if (cmd == RTM_NEWLINKPROP) {
3509                 alt_ifname = kstrdup(alt_ifname, GFP_KERNEL);
3510                 if (!alt_ifname)
3511                         return -ENOMEM;
3512                 err = netdev_name_node_alt_create(dev, alt_ifname);
3513                 if (err) {
3514                         kfree(alt_ifname);
3515                         return err;
3516                 }
3517         } else if (cmd == RTM_DELLINKPROP) {
3518                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3519                 if (err)
3520                         return err;
3521         } else {
3522                 WARN_ON(1);
3523                 return 0;
3524         }
3525
3526         *changed = true;
3527         return 0;
3528 }
3529
3530 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3531                          struct netlink_ext_ack *extack)
3532 {
3533         struct net *net = sock_net(skb->sk);
3534         struct nlattr *tb[IFLA_MAX + 1];
3535         struct net_device *dev;
3536         struct ifinfomsg *ifm;
3537         bool changed = false;
3538         struct nlattr *attr;
3539         int err, rem;
3540
3541         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3542         if (err)
3543                 return err;
3544
3545         err = rtnl_ensure_unique_netns(tb, extack, true);
3546         if (err)
3547                 return err;
3548
3549         ifm = nlmsg_data(nlh);
3550         if (ifm->ifi_index > 0)
3551                 dev = __dev_get_by_index(net, ifm->ifi_index);
3552         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3553                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3554                                    tb[IFLA_ALT_IFNAME], NULL);
3555         else
3556                 return -EINVAL;
3557
3558         if (!dev)
3559                 return -ENODEV;
3560
3561         if (!tb[IFLA_PROP_LIST])
3562                 return 0;
3563
3564         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3565                 switch (nla_type(attr)) {
3566                 case IFLA_ALT_IFNAME:
3567                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3568                         if (err)
3569                                 return err;
3570                         break;
3571                 }
3572         }
3573
3574         if (changed)
3575                 netdev_state_change(dev);
3576         return 0;
3577 }
3578
3579 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3580                             struct netlink_ext_ack *extack)
3581 {
3582         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3583 }
3584
3585 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3586                             struct netlink_ext_ack *extack)
3587 {
3588         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3589 }
3590
3591 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3592 {
3593         struct net *net = sock_net(skb->sk);
3594         struct net_device *dev;
3595         struct nlattr *tb[IFLA_MAX+1];
3596         u32 ext_filter_mask = 0;
3597         u16 min_ifinfo_dump_size = 0;
3598         int hdrlen;
3599
3600         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3601         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3602                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3603
3604         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3605                 if (tb[IFLA_EXT_MASK])
3606                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3607         }
3608
3609         if (!ext_filter_mask)
3610                 return NLMSG_GOODSIZE;
3611         /*
3612          * traverse the list of net devices and compute the minimum
3613          * buffer size based upon the filter mask.
3614          */
3615         rcu_read_lock();
3616         for_each_netdev_rcu(net, dev) {
3617                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3618                                              if_nlmsg_size(dev,
3619                                                            ext_filter_mask));
3620         }
3621         rcu_read_unlock();
3622
3623         return nlmsg_total_size(min_ifinfo_dump_size);
3624 }
3625
3626 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3627 {
3628         int idx;
3629         int s_idx = cb->family;
3630         int type = cb->nlh->nlmsg_type - RTM_BASE;
3631         int ret = 0;
3632
3633         if (s_idx == 0)
3634                 s_idx = 1;
3635
3636         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3637                 struct rtnl_link **tab;
3638                 struct rtnl_link *link;
3639                 rtnl_dumpit_func dumpit;
3640
3641                 if (idx < s_idx || idx == PF_PACKET)
3642                         continue;
3643
3644                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3645                         continue;
3646
3647                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3648                 if (!tab)
3649                         continue;
3650
3651                 link = tab[type];
3652                 if (!link)
3653                         continue;
3654
3655                 dumpit = link->dumpit;
3656                 if (!dumpit)
3657                         continue;
3658
3659                 if (idx > s_idx) {
3660                         memset(&cb->args[0], 0, sizeof(cb->args));
3661                         cb->prev_seq = 0;
3662                         cb->seq = 0;
3663                 }
3664                 ret = dumpit(skb, cb);
3665                 if (ret)
3666                         break;
3667         }
3668         cb->family = idx;
3669
3670         return skb->len ? : ret;
3671 }
3672
3673 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3674                                        unsigned int change,
3675                                        u32 event, gfp_t flags, int *new_nsid,
3676                                        int new_ifindex)
3677 {
3678         struct net *net = dev_net(dev);
3679         struct sk_buff *skb;
3680         int err = -ENOBUFS;
3681         size_t if_info_size;
3682
3683         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3684         if (skb == NULL)
3685                 goto errout;
3686
3687         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3688                                type, 0, 0, change, 0, 0, event,
3689                                new_nsid, new_ifindex, -1, flags);
3690         if (err < 0) {
3691                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3692                 WARN_ON(err == -EMSGSIZE);
3693                 kfree_skb(skb);
3694                 goto errout;
3695         }
3696         return skb;
3697 errout:
3698         if (err < 0)
3699                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3700         return NULL;
3701 }
3702
3703 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3704 {
3705         struct net *net = dev_net(dev);
3706
3707         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3708 }
3709
3710 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3711                                unsigned int change, u32 event,
3712                                gfp_t flags, int *new_nsid, int new_ifindex)
3713 {
3714         struct sk_buff *skb;
3715
3716         if (dev->reg_state != NETREG_REGISTERED)
3717                 return;
3718
3719         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3720                                      new_ifindex);
3721         if (skb)
3722                 rtmsg_ifinfo_send(skb, dev, flags);
3723 }
3724
3725 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3726                   gfp_t flags)
3727 {
3728         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3729                            NULL, 0);
3730 }
3731
3732 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3733                          gfp_t flags, int *new_nsid, int new_ifindex)
3734 {
3735         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3736                            new_nsid, new_ifindex);
3737 }
3738
3739 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3740                                    struct net_device *dev,
3741                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3742                                    int type, unsigned int flags,
3743                                    int nlflags, u16 ndm_state)
3744 {
3745         struct nlmsghdr *nlh;
3746         struct ndmsg *ndm;
3747
3748         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3749         if (!nlh)
3750                 return -EMSGSIZE;
3751
3752         ndm = nlmsg_data(nlh);
3753         ndm->ndm_family  = AF_BRIDGE;
3754         ndm->ndm_pad1    = 0;
3755         ndm->ndm_pad2    = 0;
3756         ndm->ndm_flags   = flags;
3757         ndm->ndm_type    = 0;
3758         ndm->ndm_ifindex = dev->ifindex;
3759         ndm->ndm_state   = ndm_state;
3760
3761         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3762                 goto nla_put_failure;
3763         if (vid)
3764                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3765                         goto nla_put_failure;
3766
3767         nlmsg_end(skb, nlh);
3768         return 0;
3769
3770 nla_put_failure:
3771         nlmsg_cancel(skb, nlh);
3772         return -EMSGSIZE;
3773 }
3774
3775 static inline size_t rtnl_fdb_nlmsg_size(void)
3776 {
3777         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3778                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3779                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3780                0;
3781 }
3782
3783 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3784                             u16 ndm_state)
3785 {
3786         struct net *net = dev_net(dev);
3787         struct sk_buff *skb;
3788         int err = -ENOBUFS;
3789
3790         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3791         if (!skb)
3792                 goto errout;
3793
3794         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3795                                       0, 0, type, NTF_SELF, 0, ndm_state);
3796         if (err < 0) {
3797                 kfree_skb(skb);
3798                 goto errout;
3799         }
3800
3801         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3802         return;
3803 errout:
3804         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3805 }
3806
3807 /*
3808  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3809  */
3810 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3811                      struct nlattr *tb[],
3812                      struct net_device *dev,
3813                      const unsigned char *addr, u16 vid,
3814                      u16 flags)
3815 {
3816         int err = -EINVAL;
3817
3818         /* If aging addresses are supported device will need to
3819          * implement its own handler for this.
3820          */
3821         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3822                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3823                 return err;
3824         }
3825
3826         if (vid) {
3827                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3828                 return err;
3829         }
3830
3831         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3832                 err = dev_uc_add_excl(dev, addr);
3833         else if (is_multicast_ether_addr(addr))
3834                 err = dev_mc_add_excl(dev, addr);
3835
3836         /* Only return duplicate errors if NLM_F_EXCL is set */
3837         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3838                 err = 0;
3839
3840         return err;
3841 }
3842 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3843
3844 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3845                          struct netlink_ext_ack *extack)
3846 {
3847         u16 vid = 0;
3848
3849         if (vlan_attr) {
3850                 if (nla_len(vlan_attr) != sizeof(u16)) {
3851                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3852                         return -EINVAL;
3853                 }
3854
3855                 vid = nla_get_u16(vlan_attr);
3856
3857                 if (!vid || vid >= VLAN_VID_MASK) {
3858                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3859                         return -EINVAL;
3860                 }
3861         }
3862         *p_vid = vid;
3863         return 0;
3864 }
3865
3866 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3867                         struct netlink_ext_ack *extack)
3868 {
3869         struct net *net = sock_net(skb->sk);
3870         struct ndmsg *ndm;
3871         struct nlattr *tb[NDA_MAX+1];
3872         struct net_device *dev;
3873         u8 *addr;
3874         u16 vid;
3875         int err;
3876
3877         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3878                                      extack);
3879         if (err < 0)
3880                 return err;
3881
3882         ndm = nlmsg_data(nlh);
3883         if (ndm->ndm_ifindex == 0) {
3884                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3885                 return -EINVAL;
3886         }
3887
3888         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3889         if (dev == NULL) {
3890                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3891                 return -ENODEV;
3892         }
3893
3894         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3895                 NL_SET_ERR_MSG(extack, "invalid address");
3896                 return -EINVAL;
3897         }
3898
3899         if (dev->type != ARPHRD_ETHER) {
3900                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3901                 return -EINVAL;
3902         }
3903
3904         addr = nla_data(tb[NDA_LLADDR]);
3905
3906         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3907         if (err)
3908                 return err;
3909
3910         err = -EOPNOTSUPP;
3911
3912         /* Support fdb on master device the net/bridge default case */
3913         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3914             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3915                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3916                 const struct net_device_ops *ops = br_dev->netdev_ops;
3917
3918                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3919                                        nlh->nlmsg_flags, extack);
3920                 if (err)
3921                         goto out;
3922                 else
3923                         ndm->ndm_flags &= ~NTF_MASTER;
3924         }
3925
3926         /* Embedded bridge, macvlan, and any other device support */
3927         if ((ndm->ndm_flags & NTF_SELF)) {
3928                 if (dev->netdev_ops->ndo_fdb_add)
3929                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3930                                                            vid,
3931                                                            nlh->nlmsg_flags,
3932                                                            extack);
3933                 else
3934                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3935                                                nlh->nlmsg_flags);
3936
3937                 if (!err) {
3938                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3939                                         ndm->ndm_state);
3940                         ndm->ndm_flags &= ~NTF_SELF;
3941                 }
3942         }
3943 out:
3944         return err;
3945 }
3946
3947 /*
3948  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3949  */
3950 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3951                      struct nlattr *tb[],
3952                      struct net_device *dev,
3953                      const unsigned char *addr, u16 vid)
3954 {
3955         int err = -EINVAL;
3956
3957         /* If aging addresses are supported device will need to
3958          * implement its own handler for this.
3959          */
3960         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3961                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3962                 return err;
3963         }
3964
3965         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3966                 err = dev_uc_del(dev, addr);
3967         else if (is_multicast_ether_addr(addr))
3968                 err = dev_mc_del(dev, addr);
3969
3970         return err;
3971 }
3972 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3973
3974 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3975                         struct netlink_ext_ack *extack)
3976 {
3977         struct net *net = sock_net(skb->sk);
3978         struct ndmsg *ndm;
3979         struct nlattr *tb[NDA_MAX+1];
3980         struct net_device *dev;
3981         int err = -EINVAL;
3982         __u8 *addr;
3983         u16 vid;
3984
3985         if (!netlink_capable(skb, CAP_NET_ADMIN))
3986                 return -EPERM;
3987
3988         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3989                                      extack);
3990         if (err < 0)
3991                 return err;
3992
3993         ndm = nlmsg_data(nlh);
3994         if (ndm->ndm_ifindex == 0) {
3995                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3996                 return -EINVAL;
3997         }
3998
3999         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4000         if (dev == NULL) {
4001                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4002                 return -ENODEV;
4003         }
4004
4005         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4006                 NL_SET_ERR_MSG(extack, "invalid address");
4007                 return -EINVAL;
4008         }
4009
4010         if (dev->type != ARPHRD_ETHER) {
4011                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4012                 return -EINVAL;
4013         }
4014
4015         addr = nla_data(tb[NDA_LLADDR]);
4016
4017         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4018         if (err)
4019                 return err;
4020
4021         err = -EOPNOTSUPP;
4022
4023         /* Support fdb on master device the net/bridge default case */
4024         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4025             (dev->priv_flags & IFF_BRIDGE_PORT)) {
4026                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4027                 const struct net_device_ops *ops = br_dev->netdev_ops;
4028
4029                 if (ops->ndo_fdb_del)
4030                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4031
4032                 if (err)
4033                         goto out;
4034                 else
4035                         ndm->ndm_flags &= ~NTF_MASTER;
4036         }
4037
4038         /* Embedded bridge, macvlan, and any other device support */
4039         if (ndm->ndm_flags & NTF_SELF) {
4040                 if (dev->netdev_ops->ndo_fdb_del)
4041                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4042                                                            vid);
4043                 else
4044                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4045
4046                 if (!err) {
4047                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4048                                         ndm->ndm_state);
4049                         ndm->ndm_flags &= ~NTF_SELF;
4050                 }
4051         }
4052 out:
4053         return err;
4054 }
4055
4056 static int nlmsg_populate_fdb(struct sk_buff *skb,
4057                               struct netlink_callback *cb,
4058                               struct net_device *dev,
4059                               int *idx,
4060                               struct netdev_hw_addr_list *list)
4061 {
4062         struct netdev_hw_addr *ha;
4063         int err;
4064         u32 portid, seq;
4065
4066         portid = NETLINK_CB(cb->skb).portid;
4067         seq = cb->nlh->nlmsg_seq;
4068
4069         list_for_each_entry(ha, &list->list, list) {
4070                 if (*idx < cb->args[2])
4071                         goto skip;
4072
4073                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4074                                               portid, seq,
4075                                               RTM_NEWNEIGH, NTF_SELF,
4076                                               NLM_F_MULTI, NUD_PERMANENT);
4077                 if (err < 0)
4078                         return err;
4079 skip:
4080                 *idx += 1;
4081         }
4082         return 0;
4083 }
4084
4085 /**
4086  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4087  * @skb: socket buffer to store message in
4088  * @cb: netlink callback
4089  * @dev: netdevice
4090  * @filter_dev: ignored
4091  * @idx: the number of FDB table entries dumped is added to *@idx
4092  *
4093  * Default netdevice operation to dump the existing unicast address list.
4094  * Returns number of addresses from list put in skb.
4095  */
4096 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4097                       struct netlink_callback *cb,
4098                       struct net_device *dev,
4099                       struct net_device *filter_dev,
4100                       int *idx)
4101 {
4102         int err;
4103
4104         if (dev->type != ARPHRD_ETHER)
4105                 return -EINVAL;
4106
4107         netif_addr_lock_bh(dev);
4108         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4109         if (err)
4110                 goto out;
4111         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4112 out:
4113         netif_addr_unlock_bh(dev);
4114         return err;
4115 }
4116 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4117
4118 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4119                                  int *br_idx, int *brport_idx,
4120                                  struct netlink_ext_ack *extack)
4121 {
4122         struct nlattr *tb[NDA_MAX + 1];
4123         struct ndmsg *ndm;
4124         int err, i;
4125
4126         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4127                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4128                 return -EINVAL;
4129         }
4130
4131         ndm = nlmsg_data(nlh);
4132         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4133             ndm->ndm_flags || ndm->ndm_type) {
4134                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4135                 return -EINVAL;
4136         }
4137
4138         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4139                                             NDA_MAX, NULL, extack);
4140         if (err < 0)
4141                 return err;
4142
4143         *brport_idx = ndm->ndm_ifindex;
4144         for (i = 0; i <= NDA_MAX; ++i) {
4145                 if (!tb[i])
4146                         continue;
4147
4148                 switch (i) {
4149                 case NDA_IFINDEX:
4150                         if (nla_len(tb[i]) != sizeof(u32)) {
4151                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4152                                 return -EINVAL;
4153                         }
4154                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4155                         break;
4156                 case NDA_MASTER:
4157                         if (nla_len(tb[i]) != sizeof(u32)) {
4158                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4159                                 return -EINVAL;
4160                         }
4161                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4162                         break;
4163                 default:
4164                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4165                         return -EINVAL;
4166                 }
4167         }
4168
4169         return 0;
4170 }
4171
4172 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4173                                  int *br_idx, int *brport_idx,
4174                                  struct netlink_ext_ack *extack)
4175 {
4176         struct nlattr *tb[IFLA_MAX+1];
4177         int err;
4178
4179         /* A hack to preserve kernel<->userspace interface.
4180          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4181          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4182          * So, check for ndmsg with an optional u32 attribute (not used here).
4183          * Fortunately these sizes don't conflict with the size of ifinfomsg
4184          * with an optional attribute.
4185          */
4186         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4187             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4188              nla_attr_size(sizeof(u32)))) {
4189                 struct ifinfomsg *ifm;
4190
4191                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4192                                              tb, IFLA_MAX, ifla_policy,
4193                                              extack);
4194                 if (err < 0) {
4195                         return -EINVAL;
4196                 } else if (err == 0) {
4197                         if (tb[IFLA_MASTER])
4198                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4199                 }
4200
4201                 ifm = nlmsg_data(nlh);
4202                 *brport_idx = ifm->ifi_index;
4203         }
4204         return 0;
4205 }
4206
4207 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4208 {
4209         struct net_device *dev;
4210         struct net_device *br_dev = NULL;
4211         const struct net_device_ops *ops = NULL;
4212         const struct net_device_ops *cops = NULL;
4213         struct net *net = sock_net(skb->sk);
4214         struct hlist_head *head;
4215         int brport_idx = 0;
4216         int br_idx = 0;
4217         int h, s_h;
4218         int idx = 0, s_idx;
4219         int err = 0;
4220         int fidx = 0;
4221
4222         if (cb->strict_check)
4223                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4224                                             cb->extack);
4225         else
4226                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4227                                             cb->extack);
4228         if (err < 0)
4229                 return err;
4230
4231         if (br_idx) {
4232                 br_dev = __dev_get_by_index(net, br_idx);
4233                 if (!br_dev)
4234                         return -ENODEV;
4235
4236                 ops = br_dev->netdev_ops;
4237         }
4238
4239         s_h = cb->args[0];
4240         s_idx = cb->args[1];
4241
4242         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4243                 idx = 0;
4244                 head = &net->dev_index_head[h];
4245                 hlist_for_each_entry(dev, head, index_hlist) {
4246
4247                         if (brport_idx && (dev->ifindex != brport_idx))
4248                                 continue;
4249
4250                         if (!br_idx) { /* user did not specify a specific bridge */
4251                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
4252                                         br_dev = netdev_master_upper_dev_get(dev);
4253                                         cops = br_dev->netdev_ops;
4254                                 }
4255                         } else {
4256                                 if (dev != br_dev &&
4257                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
4258                                         continue;
4259
4260                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4261                                     !(dev->priv_flags & IFF_EBRIDGE))
4262                                         continue;
4263                                 cops = ops;
4264                         }
4265
4266                         if (idx < s_idx)
4267                                 goto cont;
4268
4269                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
4270                                 if (cops && cops->ndo_fdb_dump) {
4271                                         err = cops->ndo_fdb_dump(skb, cb,
4272                                                                 br_dev, dev,
4273                                                                 &fidx);
4274                                         if (err == -EMSGSIZE)
4275                                                 goto out;
4276                                 }
4277                         }
4278
4279                         if (dev->netdev_ops->ndo_fdb_dump)
4280                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4281                                                                     dev, NULL,
4282                                                                     &fidx);
4283                         else
4284                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4285                                                         &fidx);
4286                         if (err == -EMSGSIZE)
4287                                 goto out;
4288
4289                         cops = NULL;
4290
4291                         /* reset fdb offset to 0 for rest of the interfaces */
4292                         cb->args[2] = 0;
4293                         fidx = 0;
4294 cont:
4295                         idx++;
4296                 }
4297         }
4298
4299 out:
4300         cb->args[0] = h;
4301         cb->args[1] = idx;
4302         cb->args[2] = fidx;
4303
4304         return skb->len;
4305 }
4306
4307 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4308                                 struct nlattr **tb, u8 *ndm_flags,
4309                                 int *br_idx, int *brport_idx, u8 **addr,
4310                                 u16 *vid, struct netlink_ext_ack *extack)
4311 {
4312         struct ndmsg *ndm;
4313         int err, i;
4314
4315         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4316                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4317                 return -EINVAL;
4318         }
4319
4320         ndm = nlmsg_data(nlh);
4321         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4322             ndm->ndm_type) {
4323                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4324                 return -EINVAL;
4325         }
4326
4327         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4328                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4329                 return -EINVAL;
4330         }
4331
4332         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4333                                             NDA_MAX, nda_policy, extack);
4334         if (err < 0)
4335                 return err;
4336
4337         *ndm_flags = ndm->ndm_flags;
4338         *brport_idx = ndm->ndm_ifindex;
4339         for (i = 0; i <= NDA_MAX; ++i) {
4340                 if (!tb[i])
4341                         continue;
4342
4343                 switch (i) {
4344                 case NDA_MASTER:
4345                         *br_idx = nla_get_u32(tb[i]);
4346                         break;
4347                 case NDA_LLADDR:
4348                         if (nla_len(tb[i]) != ETH_ALEN) {
4349                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4350                                 return -EINVAL;
4351                         }
4352                         *addr = nla_data(tb[i]);
4353                         break;
4354                 case NDA_VLAN:
4355                         err = fdb_vid_parse(tb[i], vid, extack);
4356                         if (err)
4357                                 return err;
4358                         break;
4359                 case NDA_VNI:
4360                         break;
4361                 default:
4362                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4363                         return -EINVAL;
4364                 }
4365         }
4366
4367         return 0;
4368 }
4369
4370 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4371                         struct netlink_ext_ack *extack)
4372 {
4373         struct net_device *dev = NULL, *br_dev = NULL;
4374         const struct net_device_ops *ops = NULL;
4375         struct net *net = sock_net(in_skb->sk);
4376         struct nlattr *tb[NDA_MAX + 1];
4377         struct sk_buff *skb;
4378         int brport_idx = 0;
4379         u8 ndm_flags = 0;
4380         int br_idx = 0;
4381         u8 *addr = NULL;
4382         u16 vid = 0;
4383         int err;
4384
4385         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4386                                    &brport_idx, &addr, &vid, extack);
4387         if (err < 0)
4388                 return err;
4389
4390         if (!addr) {
4391                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4392                 return -EINVAL;
4393         }
4394
4395         if (brport_idx) {
4396                 dev = __dev_get_by_index(net, brport_idx);
4397                 if (!dev) {
4398                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4399                         return -ENODEV;
4400                 }
4401         }
4402
4403         if (br_idx) {
4404                 if (dev) {
4405                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4406                         return -EINVAL;
4407                 }
4408
4409                 br_dev = __dev_get_by_index(net, br_idx);
4410                 if (!br_dev) {
4411                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4412                         return -EINVAL;
4413                 }
4414                 ops = br_dev->netdev_ops;
4415         }
4416
4417         if (dev) {
4418                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4419                         if (!(dev->priv_flags & IFF_BRIDGE_PORT)) {
4420                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4421                                 return -EINVAL;
4422                         }
4423                         br_dev = netdev_master_upper_dev_get(dev);
4424                         if (!br_dev) {
4425                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4426                                 return -EINVAL;
4427                         }
4428                         ops = br_dev->netdev_ops;
4429                 } else {
4430                         if (!(ndm_flags & NTF_SELF)) {
4431                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4432                                 return -EINVAL;
4433                         }
4434                         ops = dev->netdev_ops;
4435                 }
4436         }
4437
4438         if (!br_dev && !dev) {
4439                 NL_SET_ERR_MSG(extack, "No device specified");
4440                 return -ENODEV;
4441         }
4442
4443         if (!ops || !ops->ndo_fdb_get) {
4444                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4445                 return -EOPNOTSUPP;
4446         }
4447
4448         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4449         if (!skb)
4450                 return -ENOBUFS;
4451
4452         if (br_dev)
4453                 dev = br_dev;
4454         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4455                                NETLINK_CB(in_skb).portid,
4456                                nlh->nlmsg_seq, extack);
4457         if (err)
4458                 goto out;
4459
4460         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4461 out:
4462         kfree_skb(skb);
4463         return err;
4464 }
4465
4466 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4467                                unsigned int attrnum, unsigned int flag)
4468 {
4469         if (mask & flag)
4470                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4471         return 0;
4472 }
4473
4474 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4475                             struct net_device *dev, u16 mode,
4476                             u32 flags, u32 mask, int nlflags,
4477                             u32 filter_mask,
4478                             int (*vlan_fill)(struct sk_buff *skb,
4479                                              struct net_device *dev,
4480                                              u32 filter_mask))
4481 {
4482         struct nlmsghdr *nlh;
4483         struct ifinfomsg *ifm;
4484         struct nlattr *br_afspec;
4485         struct nlattr *protinfo;
4486         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4487         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4488         int err = 0;
4489
4490         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4491         if (nlh == NULL)
4492                 return -EMSGSIZE;
4493
4494         ifm = nlmsg_data(nlh);
4495         ifm->ifi_family = AF_BRIDGE;
4496         ifm->__ifi_pad = 0;
4497         ifm->ifi_type = dev->type;
4498         ifm->ifi_index = dev->ifindex;
4499         ifm->ifi_flags = dev_get_flags(dev);
4500         ifm->ifi_change = 0;
4501
4502
4503         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4504             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4505             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4506             (br_dev &&
4507              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4508             (dev->addr_len &&
4509              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4510             (dev->ifindex != dev_get_iflink(dev) &&
4511              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4512                 goto nla_put_failure;
4513
4514         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4515         if (!br_afspec)
4516                 goto nla_put_failure;
4517
4518         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4519                 nla_nest_cancel(skb, br_afspec);
4520                 goto nla_put_failure;
4521         }
4522
4523         if (mode != BRIDGE_MODE_UNDEF) {
4524                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4525                         nla_nest_cancel(skb, br_afspec);
4526                         goto nla_put_failure;
4527                 }
4528         }
4529         if (vlan_fill) {
4530                 err = vlan_fill(skb, dev, filter_mask);
4531                 if (err) {
4532                         nla_nest_cancel(skb, br_afspec);
4533                         goto nla_put_failure;
4534                 }
4535         }
4536         nla_nest_end(skb, br_afspec);
4537
4538         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4539         if (!protinfo)
4540                 goto nla_put_failure;
4541
4542         if (brport_nla_put_flag(skb, flags, mask,
4543                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4544             brport_nla_put_flag(skb, flags, mask,
4545                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4546             brport_nla_put_flag(skb, flags, mask,
4547                                 IFLA_BRPORT_FAST_LEAVE,
4548                                 BR_MULTICAST_FAST_LEAVE) ||
4549             brport_nla_put_flag(skb, flags, mask,
4550                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4551             brport_nla_put_flag(skb, flags, mask,
4552                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4553             brport_nla_put_flag(skb, flags, mask,
4554                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4555             brport_nla_put_flag(skb, flags, mask,
4556                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4557             brport_nla_put_flag(skb, flags, mask,
4558                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
4559                 nla_nest_cancel(skb, protinfo);
4560                 goto nla_put_failure;
4561         }
4562
4563         nla_nest_end(skb, protinfo);
4564
4565         nlmsg_end(skb, nlh);
4566         return 0;
4567 nla_put_failure:
4568         nlmsg_cancel(skb, nlh);
4569         return err ? err : -EMSGSIZE;
4570 }
4571 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4572
4573 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4574                                     bool strict_check, u32 *filter_mask,
4575                                     struct netlink_ext_ack *extack)
4576 {
4577         struct nlattr *tb[IFLA_MAX+1];
4578         int err, i;
4579
4580         if (strict_check) {
4581                 struct ifinfomsg *ifm;
4582
4583                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4584                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4585                         return -EINVAL;
4586                 }
4587
4588                 ifm = nlmsg_data(nlh);
4589                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4590                     ifm->ifi_change || ifm->ifi_index) {
4591                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4592                         return -EINVAL;
4593                 }
4594
4595                 err = nlmsg_parse_deprecated_strict(nlh,
4596                                                     sizeof(struct ifinfomsg),
4597                                                     tb, IFLA_MAX, ifla_policy,
4598                                                     extack);
4599         } else {
4600                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4601                                              tb, IFLA_MAX, ifla_policy,
4602                                              extack);
4603         }
4604         if (err < 0)
4605                 return err;
4606
4607         /* new attributes should only be added with strict checking */
4608         for (i = 0; i <= IFLA_MAX; ++i) {
4609                 if (!tb[i])
4610                         continue;
4611
4612                 switch (i) {
4613                 case IFLA_EXT_MASK:
4614                         *filter_mask = nla_get_u32(tb[i]);
4615                         break;
4616                 default:
4617                         if (strict_check) {
4618                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4619                                 return -EINVAL;
4620                         }
4621                 }
4622         }
4623
4624         return 0;
4625 }
4626
4627 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4628 {
4629         const struct nlmsghdr *nlh = cb->nlh;
4630         struct net *net = sock_net(skb->sk);
4631         struct net_device *dev;
4632         int idx = 0;
4633         u32 portid = NETLINK_CB(cb->skb).portid;
4634         u32 seq = nlh->nlmsg_seq;
4635         u32 filter_mask = 0;
4636         int err;
4637
4638         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4639                                        cb->extack);
4640         if (err < 0 && cb->strict_check)
4641                 return err;
4642
4643         rcu_read_lock();
4644         for_each_netdev_rcu(net, dev) {
4645                 const struct net_device_ops *ops = dev->netdev_ops;
4646                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4647
4648                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4649                         if (idx >= cb->args[0]) {
4650                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4651                                                 skb, portid, seq, dev,
4652                                                 filter_mask, NLM_F_MULTI);
4653                                 if (err < 0 && err != -EOPNOTSUPP) {
4654                                         if (likely(skb->len))
4655                                                 break;
4656
4657                                         goto out_err;
4658                                 }
4659                         }
4660                         idx++;
4661                 }
4662
4663                 if (ops->ndo_bridge_getlink) {
4664                         if (idx >= cb->args[0]) {
4665                                 err = ops->ndo_bridge_getlink(skb, portid,
4666                                                               seq, dev,
4667                                                               filter_mask,
4668                                                               NLM_F_MULTI);
4669                                 if (err < 0 && err != -EOPNOTSUPP) {
4670                                         if (likely(skb->len))
4671                                                 break;
4672
4673                                         goto out_err;
4674                                 }
4675                         }
4676                         idx++;
4677                 }
4678         }
4679         err = skb->len;
4680 out_err:
4681         rcu_read_unlock();
4682         cb->args[0] = idx;
4683
4684         return err;
4685 }
4686
4687 static inline size_t bridge_nlmsg_size(void)
4688 {
4689         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4690                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4691                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4692                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4693                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4694                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4695                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4696                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4697                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4698                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4699                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4700 }
4701
4702 static int rtnl_bridge_notify(struct net_device *dev)
4703 {
4704         struct net *net = dev_net(dev);
4705         struct sk_buff *skb;
4706         int err = -EOPNOTSUPP;
4707
4708         if (!dev->netdev_ops->ndo_bridge_getlink)
4709                 return 0;
4710
4711         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4712         if (!skb) {
4713                 err = -ENOMEM;
4714                 goto errout;
4715         }
4716
4717         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4718         if (err < 0)
4719                 goto errout;
4720
4721         if (!skb->len)
4722                 goto errout;
4723
4724         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4725         return 0;
4726 errout:
4727         WARN_ON(err == -EMSGSIZE);
4728         kfree_skb(skb);
4729         if (err)
4730                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4731         return err;
4732 }
4733
4734 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4735                                struct netlink_ext_ack *extack)
4736 {
4737         struct net *net = sock_net(skb->sk);
4738         struct ifinfomsg *ifm;
4739         struct net_device *dev;
4740         struct nlattr *br_spec, *attr = NULL;
4741         int rem, err = -EOPNOTSUPP;
4742         u16 flags = 0;
4743         bool have_flags = false;
4744
4745         if (nlmsg_len(nlh) < sizeof(*ifm))
4746                 return -EINVAL;
4747
4748         ifm = nlmsg_data(nlh);
4749         if (ifm->ifi_family != AF_BRIDGE)
4750                 return -EPFNOSUPPORT;
4751
4752         dev = __dev_get_by_index(net, ifm->ifi_index);
4753         if (!dev) {
4754                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4755                 return -ENODEV;
4756         }
4757
4758         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4759         if (br_spec) {
4760                 nla_for_each_nested(attr, br_spec, rem) {
4761                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4762                                 if (nla_len(attr) < sizeof(flags))
4763                                         return -EINVAL;
4764
4765                                 have_flags = true;
4766                                 flags = nla_get_u16(attr);
4767                                 break;
4768                         }
4769                 }
4770         }
4771
4772         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4773                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4774
4775                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4776                         err = -EOPNOTSUPP;
4777                         goto out;
4778                 }
4779
4780                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4781                                                              extack);
4782                 if (err)
4783                         goto out;
4784
4785                 flags &= ~BRIDGE_FLAGS_MASTER;
4786         }
4787
4788         if ((flags & BRIDGE_FLAGS_SELF)) {
4789                 if (!dev->netdev_ops->ndo_bridge_setlink)
4790                         err = -EOPNOTSUPP;
4791                 else
4792                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4793                                                                   flags,
4794                                                                   extack);
4795                 if (!err) {
4796                         flags &= ~BRIDGE_FLAGS_SELF;
4797
4798                         /* Generate event to notify upper layer of bridge
4799                          * change
4800                          */
4801                         err = rtnl_bridge_notify(dev);
4802                 }
4803         }
4804
4805         if (have_flags)
4806                 memcpy(nla_data(attr), &flags, sizeof(flags));
4807 out:
4808         return err;
4809 }
4810
4811 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4812                                struct netlink_ext_ack *extack)
4813 {
4814         struct net *net = sock_net(skb->sk);
4815         struct ifinfomsg *ifm;
4816         struct net_device *dev;
4817         struct nlattr *br_spec, *attr = NULL;
4818         int rem, err = -EOPNOTSUPP;
4819         u16 flags = 0;
4820         bool have_flags = false;
4821
4822         if (nlmsg_len(nlh) < sizeof(*ifm))
4823                 return -EINVAL;
4824
4825         ifm = nlmsg_data(nlh);
4826         if (ifm->ifi_family != AF_BRIDGE)
4827                 return -EPFNOSUPPORT;
4828
4829         dev = __dev_get_by_index(net, ifm->ifi_index);
4830         if (!dev) {
4831                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4832                 return -ENODEV;
4833         }
4834
4835         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4836         if (br_spec) {
4837                 nla_for_each_nested(attr, br_spec, rem) {
4838                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4839                                 if (nla_len(attr) < sizeof(flags))
4840                                         return -EINVAL;
4841
4842                                 have_flags = true;
4843                                 flags = nla_get_u16(attr);
4844                                 break;
4845                         }
4846                 }
4847         }
4848
4849         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4850                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4851
4852                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4853                         err = -EOPNOTSUPP;
4854                         goto out;
4855                 }
4856
4857                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4858                 if (err)
4859                         goto out;
4860
4861                 flags &= ~BRIDGE_FLAGS_MASTER;
4862         }
4863
4864         if ((flags & BRIDGE_FLAGS_SELF)) {
4865                 if (!dev->netdev_ops->ndo_bridge_dellink)
4866                         err = -EOPNOTSUPP;
4867                 else
4868                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4869                                                                   flags);
4870
4871                 if (!err) {
4872                         flags &= ~BRIDGE_FLAGS_SELF;
4873
4874                         /* Generate event to notify upper layer of bridge
4875                          * change
4876                          */
4877                         err = rtnl_bridge_notify(dev);
4878                 }
4879         }
4880
4881         if (have_flags)
4882                 memcpy(nla_data(attr), &flags, sizeof(flags));
4883 out:
4884         return err;
4885 }
4886
4887 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4888 {
4889         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4890                (!idxattr || idxattr == attrid);
4891 }
4892
4893 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4894 static int rtnl_get_offload_stats_attr_size(int attr_id)
4895 {
4896         switch (attr_id) {
4897         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4898                 return sizeof(struct rtnl_link_stats64);
4899         }
4900
4901         return 0;
4902 }
4903
4904 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4905                                   int *prividx)
4906 {
4907         struct nlattr *attr = NULL;
4908         int attr_id, size;
4909         void *attr_data;
4910         int err;
4911
4912         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4913               dev->netdev_ops->ndo_get_offload_stats))
4914                 return -ENODATA;
4915
4916         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4917              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4918                 if (attr_id < *prividx)
4919                         continue;
4920
4921                 size = rtnl_get_offload_stats_attr_size(attr_id);
4922                 if (!size)
4923                         continue;
4924
4925                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4926                         continue;
4927
4928                 attr = nla_reserve_64bit(skb, attr_id, size,
4929                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4930                 if (!attr)
4931                         goto nla_put_failure;
4932
4933                 attr_data = nla_data(attr);
4934                 memset(attr_data, 0, size);
4935                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4936                                                              attr_data);
4937                 if (err)
4938                         goto get_offload_stats_failure;
4939         }
4940
4941         if (!attr)
4942                 return -ENODATA;
4943
4944         *prividx = 0;
4945         return 0;
4946
4947 nla_put_failure:
4948         err = -EMSGSIZE;
4949 get_offload_stats_failure:
4950         *prividx = attr_id;
4951         return err;
4952 }
4953
4954 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4955 {
4956         int nla_size = 0;
4957         int attr_id;
4958         int size;
4959
4960         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4961               dev->netdev_ops->ndo_get_offload_stats))
4962                 return 0;
4963
4964         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4965              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4966                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4967                         continue;
4968                 size = rtnl_get_offload_stats_attr_size(attr_id);
4969                 nla_size += nla_total_size_64bit(size);
4970         }
4971
4972         if (nla_size != 0)
4973                 nla_size += nla_total_size(0);
4974
4975         return nla_size;
4976 }
4977
4978 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4979                                int type, u32 pid, u32 seq, u32 change,
4980                                unsigned int flags, unsigned int filter_mask,
4981                                int *idxattr, int *prividx)
4982 {
4983         struct if_stats_msg *ifsm;
4984         struct nlmsghdr *nlh;
4985         struct nlattr *attr;
4986         int s_prividx = *prividx;
4987         int err;
4988
4989         ASSERT_RTNL();
4990
4991         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4992         if (!nlh)
4993                 return -EMSGSIZE;
4994
4995         ifsm = nlmsg_data(nlh);
4996         ifsm->family = PF_UNSPEC;
4997         ifsm->pad1 = 0;
4998         ifsm->pad2 = 0;
4999         ifsm->ifindex = dev->ifindex;
5000         ifsm->filter_mask = filter_mask;
5001
5002         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5003                 struct rtnl_link_stats64 *sp;
5004
5005                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5006                                          sizeof(struct rtnl_link_stats64),
5007                                          IFLA_STATS_UNSPEC);
5008                 if (!attr)
5009                         goto nla_put_failure;
5010
5011                 sp = nla_data(attr);
5012                 dev_get_stats(dev, sp);
5013         }
5014
5015         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5016                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5017
5018                 if (ops && ops->fill_linkxstats) {
5019                         *idxattr = IFLA_STATS_LINK_XSTATS;
5020                         attr = nla_nest_start_noflag(skb,
5021                                                      IFLA_STATS_LINK_XSTATS);
5022                         if (!attr)
5023                                 goto nla_put_failure;
5024
5025                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5026                         nla_nest_end(skb, attr);
5027                         if (err)
5028                                 goto nla_put_failure;
5029                         *idxattr = 0;
5030                 }
5031         }
5032
5033         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5034                              *idxattr)) {
5035                 const struct rtnl_link_ops *ops = NULL;
5036                 const struct net_device *master;
5037
5038                 master = netdev_master_upper_dev_get(dev);
5039                 if (master)
5040                         ops = master->rtnl_link_ops;
5041                 if (ops && ops->fill_linkxstats) {
5042                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5043                         attr = nla_nest_start_noflag(skb,
5044                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5045                         if (!attr)
5046                                 goto nla_put_failure;
5047
5048                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5049                         nla_nest_end(skb, attr);
5050                         if (err)
5051                                 goto nla_put_failure;
5052                         *idxattr = 0;
5053                 }
5054         }
5055
5056         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5057                              *idxattr)) {
5058                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5059                 attr = nla_nest_start_noflag(skb,
5060                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5061                 if (!attr)
5062                         goto nla_put_failure;
5063
5064                 err = rtnl_get_offload_stats(skb, dev, prividx);
5065                 if (err == -ENODATA)
5066                         nla_nest_cancel(skb, attr);
5067                 else
5068                         nla_nest_end(skb, attr);
5069
5070                 if (err && err != -ENODATA)
5071                         goto nla_put_failure;
5072                 *idxattr = 0;
5073         }
5074
5075         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5076                 struct rtnl_af_ops *af_ops;
5077
5078                 *idxattr = IFLA_STATS_AF_SPEC;
5079                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5080                 if (!attr)
5081                         goto nla_put_failure;
5082
5083                 rcu_read_lock();
5084                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5085                         if (af_ops->fill_stats_af) {
5086                                 struct nlattr *af;
5087                                 int err;
5088
5089                                 af = nla_nest_start_noflag(skb,
5090                                                            af_ops->family);
5091                                 if (!af) {
5092                                         rcu_read_unlock();
5093                                         goto nla_put_failure;
5094                                 }
5095                                 err = af_ops->fill_stats_af(skb, dev);
5096
5097                                 if (err == -ENODATA) {
5098                                         nla_nest_cancel(skb, af);
5099                                 } else if (err < 0) {
5100                                         rcu_read_unlock();
5101                                         goto nla_put_failure;
5102                                 }
5103
5104                                 nla_nest_end(skb, af);
5105                         }
5106                 }
5107                 rcu_read_unlock();
5108
5109                 nla_nest_end(skb, attr);
5110
5111                 *idxattr = 0;
5112         }
5113
5114         nlmsg_end(skb, nlh);
5115
5116         return 0;
5117
5118 nla_put_failure:
5119         /* not a multi message or no progress mean a real error */
5120         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5121                 nlmsg_cancel(skb, nlh);
5122         else
5123                 nlmsg_end(skb, nlh);
5124
5125         return -EMSGSIZE;
5126 }
5127
5128 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5129                                   u32 filter_mask)
5130 {
5131         size_t size = 0;
5132
5133         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5134                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5135
5136         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5137                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5138                 int attr = IFLA_STATS_LINK_XSTATS;
5139
5140                 if (ops && ops->get_linkxstats_size) {
5141                         size += nla_total_size(ops->get_linkxstats_size(dev,
5142                                                                         attr));
5143                         /* for IFLA_STATS_LINK_XSTATS */
5144                         size += nla_total_size(0);
5145                 }
5146         }
5147
5148         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5149                 struct net_device *_dev = (struct net_device *)dev;
5150                 const struct rtnl_link_ops *ops = NULL;
5151                 const struct net_device *master;
5152
5153                 /* netdev_master_upper_dev_get can't take const */
5154                 master = netdev_master_upper_dev_get(_dev);
5155                 if (master)
5156                         ops = master->rtnl_link_ops;
5157                 if (ops && ops->get_linkxstats_size) {
5158                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5159
5160                         size += nla_total_size(ops->get_linkxstats_size(dev,
5161                                                                         attr));
5162                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5163                         size += nla_total_size(0);
5164                 }
5165         }
5166
5167         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5168                 size += rtnl_get_offload_stats_size(dev);
5169
5170         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5171                 struct rtnl_af_ops *af_ops;
5172
5173                 /* for IFLA_STATS_AF_SPEC */
5174                 size += nla_total_size(0);
5175
5176                 rcu_read_lock();
5177                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5178                         if (af_ops->get_stats_af_size) {
5179                                 size += nla_total_size(
5180                                         af_ops->get_stats_af_size(dev));
5181
5182                                 /* for AF_* */
5183                                 size += nla_total_size(0);
5184                         }
5185                 }
5186                 rcu_read_unlock();
5187         }
5188
5189         return size;
5190 }
5191
5192 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5193                                 bool is_dump, struct netlink_ext_ack *extack)
5194 {
5195         struct if_stats_msg *ifsm;
5196
5197         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5198                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5199                 return -EINVAL;
5200         }
5201
5202         if (!strict_check)
5203                 return 0;
5204
5205         ifsm = nlmsg_data(nlh);
5206
5207         /* only requests using strict checks can pass data to influence
5208          * the dump. The legacy exception is filter_mask.
5209          */
5210         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5211                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5212                 return -EINVAL;
5213         }
5214         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5215                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5216                 return -EINVAL;
5217         }
5218         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5219                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5220                 return -EINVAL;
5221         }
5222
5223         return 0;
5224 }
5225
5226 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5227                           struct netlink_ext_ack *extack)
5228 {
5229         struct net *net = sock_net(skb->sk);
5230         struct net_device *dev = NULL;
5231         int idxattr = 0, prividx = 0;
5232         struct if_stats_msg *ifsm;
5233         struct sk_buff *nskb;
5234         u32 filter_mask;
5235         int err;
5236
5237         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5238                                    false, extack);
5239         if (err)
5240                 return err;
5241
5242         ifsm = nlmsg_data(nlh);
5243         if (ifsm->ifindex > 0)
5244                 dev = __dev_get_by_index(net, ifsm->ifindex);
5245         else
5246                 return -EINVAL;
5247
5248         if (!dev)
5249                 return -ENODEV;
5250
5251         filter_mask = ifsm->filter_mask;
5252         if (!filter_mask)
5253                 return -EINVAL;
5254
5255         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5256         if (!nskb)
5257                 return -ENOBUFS;
5258
5259         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5260                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5261                                   0, filter_mask, &idxattr, &prividx);
5262         if (err < 0) {
5263                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5264                 WARN_ON(err == -EMSGSIZE);
5265                 kfree_skb(nskb);
5266         } else {
5267                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5268         }
5269
5270         return err;
5271 }
5272
5273 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5274 {
5275         struct netlink_ext_ack *extack = cb->extack;
5276         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5277         struct net *net = sock_net(skb->sk);
5278         unsigned int flags = NLM_F_MULTI;
5279         struct if_stats_msg *ifsm;
5280         struct hlist_head *head;
5281         struct net_device *dev;
5282         u32 filter_mask = 0;
5283         int idx = 0;
5284
5285         s_h = cb->args[0];
5286         s_idx = cb->args[1];
5287         s_idxattr = cb->args[2];
5288         s_prividx = cb->args[3];
5289
5290         cb->seq = net->dev_base_seq;
5291
5292         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5293         if (err)
5294                 return err;
5295
5296         ifsm = nlmsg_data(cb->nlh);
5297         filter_mask = ifsm->filter_mask;
5298         if (!filter_mask) {
5299                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5300                 return -EINVAL;
5301         }
5302
5303         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5304                 idx = 0;
5305                 head = &net->dev_index_head[h];
5306                 hlist_for_each_entry(dev, head, index_hlist) {
5307                         if (idx < s_idx)
5308                                 goto cont;
5309                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5310                                                   NETLINK_CB(cb->skb).portid,
5311                                                   cb->nlh->nlmsg_seq, 0,
5312                                                   flags, filter_mask,
5313                                                   &s_idxattr, &s_prividx);
5314                         /* If we ran out of room on the first message,
5315                          * we're in trouble
5316                          */
5317                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5318
5319                         if (err < 0)
5320                                 goto out;
5321                         s_prividx = 0;
5322                         s_idxattr = 0;
5323                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5324 cont:
5325                         idx++;
5326                 }
5327         }
5328 out:
5329         cb->args[3] = s_prividx;
5330         cb->args[2] = s_idxattr;
5331         cb->args[1] = idx;
5332         cb->args[0] = h;
5333
5334         return skb->len;
5335 }
5336
5337 /* Process one rtnetlink message. */
5338
5339 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5340                              struct netlink_ext_ack *extack)
5341 {
5342         struct net *net = sock_net(skb->sk);
5343         struct rtnl_link *link;
5344         struct module *owner;
5345         int err = -EOPNOTSUPP;
5346         rtnl_doit_func doit;
5347         unsigned int flags;
5348         int kind;
5349         int family;
5350         int type;
5351
5352         type = nlh->nlmsg_type;
5353         if (type > RTM_MAX)
5354                 return -EOPNOTSUPP;
5355
5356         type -= RTM_BASE;
5357
5358         /* All the messages must have at least 1 byte length */
5359         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5360                 return 0;
5361
5362         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5363         kind = type&3;
5364
5365         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5366                 return -EPERM;
5367
5368         rcu_read_lock();
5369         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5370                 struct sock *rtnl;
5371                 rtnl_dumpit_func dumpit;
5372                 u16 min_dump_alloc = 0;
5373
5374                 link = rtnl_get_link(family, type);
5375                 if (!link || !link->dumpit) {
5376                         family = PF_UNSPEC;
5377                         link = rtnl_get_link(family, type);
5378                         if (!link || !link->dumpit)
5379                                 goto err_unlock;
5380                 }
5381                 owner = link->owner;
5382                 dumpit = link->dumpit;
5383
5384                 if (type == RTM_GETLINK - RTM_BASE)
5385                         min_dump_alloc = rtnl_calcit(skb, nlh);
5386
5387                 err = 0;
5388                 /* need to do this before rcu_read_unlock() */
5389                 if (!try_module_get(owner))
5390                         err = -EPROTONOSUPPORT;
5391
5392                 rcu_read_unlock();
5393
5394                 rtnl = net->rtnl;
5395                 if (err == 0) {
5396                         struct netlink_dump_control c = {
5397                                 .dump           = dumpit,
5398                                 .min_dump_alloc = min_dump_alloc,
5399                                 .module         = owner,
5400                         };
5401                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5402                         /* netlink_dump_start() will keep a reference on
5403                          * module if dump is still in progress.
5404                          */
5405                         module_put(owner);
5406                 }
5407                 return err;
5408         }
5409
5410         link = rtnl_get_link(family, type);
5411         if (!link || !link->doit) {
5412                 family = PF_UNSPEC;
5413                 link = rtnl_get_link(PF_UNSPEC, type);
5414                 if (!link || !link->doit)
5415                         goto out_unlock;
5416         }
5417
5418         owner = link->owner;
5419         if (!try_module_get(owner)) {
5420                 err = -EPROTONOSUPPORT;
5421                 goto out_unlock;
5422         }
5423
5424         flags = link->flags;
5425         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5426                 doit = link->doit;
5427                 rcu_read_unlock();
5428                 if (doit)
5429                         err = doit(skb, nlh, extack);
5430                 module_put(owner);
5431                 return err;
5432         }
5433         rcu_read_unlock();
5434
5435         rtnl_lock();
5436         link = rtnl_get_link(family, type);
5437         if (link && link->doit)
5438                 err = link->doit(skb, nlh, extack);
5439         rtnl_unlock();
5440
5441         module_put(owner);
5442
5443         return err;
5444
5445 out_unlock:
5446         rcu_read_unlock();
5447         return err;
5448
5449 err_unlock:
5450         rcu_read_unlock();
5451         return -EOPNOTSUPP;
5452 }
5453
5454 static void rtnetlink_rcv(struct sk_buff *skb)
5455 {
5456         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5457 }
5458
5459 static int rtnetlink_bind(struct net *net, int group)
5460 {
5461         switch (group) {
5462         case RTNLGRP_IPV4_MROUTE_R:
5463         case RTNLGRP_IPV6_MROUTE_R:
5464                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5465                         return -EPERM;
5466                 break;
5467         }
5468         return 0;
5469 }
5470
5471 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5472 {
5473         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5474
5475         switch (event) {
5476         case NETDEV_REBOOT:
5477         case NETDEV_CHANGEMTU:
5478         case NETDEV_CHANGEADDR:
5479         case NETDEV_CHANGENAME:
5480         case NETDEV_FEAT_CHANGE:
5481         case NETDEV_BONDING_FAILOVER:
5482         case NETDEV_POST_TYPE_CHANGE:
5483         case NETDEV_NOTIFY_PEERS:
5484         case NETDEV_CHANGEUPPER:
5485         case NETDEV_RESEND_IGMP:
5486         case NETDEV_CHANGEINFODATA:
5487         case NETDEV_CHANGELOWERSTATE:
5488         case NETDEV_CHANGE_TX_QUEUE_LEN:
5489                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5490                                    GFP_KERNEL, NULL, 0);
5491                 break;
5492         default:
5493                 break;
5494         }
5495         return NOTIFY_DONE;
5496 }
5497
5498 static struct notifier_block rtnetlink_dev_notifier = {
5499         .notifier_call  = rtnetlink_event,
5500 };
5501
5502
5503 static int __net_init rtnetlink_net_init(struct net *net)
5504 {
5505         struct sock *sk;
5506         struct netlink_kernel_cfg cfg = {
5507                 .groups         = RTNLGRP_MAX,
5508                 .input          = rtnetlink_rcv,
5509                 .cb_mutex       = &rtnl_mutex,
5510                 .flags          = NL_CFG_F_NONROOT_RECV,
5511                 .bind           = rtnetlink_bind,
5512         };
5513
5514         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5515         if (!sk)
5516                 return -ENOMEM;
5517         net->rtnl = sk;
5518         return 0;
5519 }
5520
5521 static void __net_exit rtnetlink_net_exit(struct net *net)
5522 {
5523         netlink_kernel_release(net->rtnl);
5524         net->rtnl = NULL;
5525 }
5526
5527 static struct pernet_operations rtnetlink_net_ops = {
5528         .init = rtnetlink_net_init,
5529         .exit = rtnetlink_net_exit,
5530 };
5531
5532 void __init rtnetlink_init(void)
5533 {
5534         if (register_pernet_subsys(&rtnetlink_net_ops))
5535                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5536
5537         register_netdevice_notifier(&rtnetlink_dev_notifier);
5538
5539         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5540                       rtnl_dump_ifinfo, 0);
5541         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5542         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5543         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5544
5545         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5546         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5547         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5548
5549         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5550         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5551
5552         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5553         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5554         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5555
5556         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5557         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5558         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5559
5560         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5561                       0);
5562 }