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         memset(&node_guid, 0, sizeof(node_guid));
1246         memset(&port_guid, 0, sizeof(port_guid));
1247
1248         vf_mac.vf =
1249                 vf_vlan.vf =
1250                 vf_vlan_info.vf =
1251                 vf_rate.vf =
1252                 vf_tx_rate.vf =
1253                 vf_spoofchk.vf =
1254                 vf_linkstate.vf =
1255                 vf_rss_query_en.vf =
1256                 vf_trust.vf =
1257                 node_guid.vf =
1258                 port_guid.vf = ivi.vf;
1259
1260         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1261         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1262         vf_vlan.vlan = ivi.vlan;
1263         vf_vlan.qos = ivi.qos;
1264         vf_vlan_info.vlan = ivi.vlan;
1265         vf_vlan_info.qos = ivi.qos;
1266         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1267         vf_tx_rate.rate = ivi.max_tx_rate;
1268         vf_rate.min_tx_rate = ivi.min_tx_rate;
1269         vf_rate.max_tx_rate = ivi.max_tx_rate;
1270         vf_spoofchk.setting = ivi.spoofchk;
1271         vf_linkstate.link_state = ivi.linkstate;
1272         vf_rss_query_en.setting = ivi.rss_query_en;
1273         vf_trust.setting = ivi.trusted;
1274         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1275         if (!vf)
1276                 goto nla_put_vfinfo_failure;
1277         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1278             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1279             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1280             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1281                     &vf_rate) ||
1282             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1283                     &vf_tx_rate) ||
1284             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1285                     &vf_spoofchk) ||
1286             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1287                     &vf_linkstate) ||
1288             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1289                     sizeof(vf_rss_query_en),
1290                     &vf_rss_query_en) ||
1291             nla_put(skb, IFLA_VF_TRUST,
1292                     sizeof(vf_trust), &vf_trust))
1293                 goto nla_put_vf_failure;
1294
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_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1876         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1877         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1878         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1879         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1880         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1881 };
1882
1883 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1884 {
1885         const struct rtnl_link_ops *ops = NULL;
1886         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1887
1888         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1889                 return NULL;
1890
1891         if (linfo[IFLA_INFO_KIND]) {
1892                 char kind[MODULE_NAME_LEN];
1893
1894                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1895                 ops = rtnl_link_ops_get(kind);
1896         }
1897
1898         return ops;
1899 }
1900
1901 static bool link_master_filtered(struct net_device *dev, int master_idx)
1902 {
1903         struct net_device *master;
1904
1905         if (!master_idx)
1906                 return false;
1907
1908         master = netdev_master_upper_dev_get(dev);
1909         if (!master || master->ifindex != master_idx)
1910                 return true;
1911
1912         return false;
1913 }
1914
1915 static bool link_kind_filtered(const struct net_device *dev,
1916                                const struct rtnl_link_ops *kind_ops)
1917 {
1918         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1919                 return true;
1920
1921         return false;
1922 }
1923
1924 static bool link_dump_filtered(struct net_device *dev,
1925                                int master_idx,
1926                                const struct rtnl_link_ops *kind_ops)
1927 {
1928         if (link_master_filtered(dev, master_idx) ||
1929             link_kind_filtered(dev, kind_ops))
1930                 return true;
1931
1932         return false;
1933 }
1934
1935 /**
1936  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1937  * @sk: netlink socket
1938  * @netnsid: network namespace identifier
1939  *
1940  * Returns the network namespace identified by netnsid on success or an error
1941  * pointer on failure.
1942  */
1943 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1944 {
1945         struct net *net;
1946
1947         net = get_net_ns_by_id(sock_net(sk), netnsid);
1948         if (!net)
1949                 return ERR_PTR(-EINVAL);
1950
1951         /* For now, the caller is required to have CAP_NET_ADMIN in
1952          * the user namespace owning the target net ns.
1953          */
1954         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1955                 put_net(net);
1956                 return ERR_PTR(-EACCES);
1957         }
1958         return net;
1959 }
1960 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
1961
1962 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
1963                                       bool strict_check, struct nlattr **tb,
1964                                       struct netlink_ext_ack *extack)
1965 {
1966         int hdrlen;
1967
1968         if (strict_check) {
1969                 struct ifinfomsg *ifm;
1970
1971                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1972                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
1973                         return -EINVAL;
1974                 }
1975
1976                 ifm = nlmsg_data(nlh);
1977                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
1978                     ifm->ifi_change) {
1979                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
1980                         return -EINVAL;
1981                 }
1982                 if (ifm->ifi_index) {
1983                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
1984                         return -EINVAL;
1985                 }
1986
1987                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
1988                                                      IFLA_MAX, ifla_policy,
1989                                                      extack);
1990         }
1991
1992         /* A hack to preserve kernel<->userspace interface.
1993          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1994          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1995          * what iproute2 < v3.9.0 used.
1996          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1997          * attribute, its netlink message is shorter than struct ifinfomsg.
1998          */
1999         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2000                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2001
2002         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2003                                       extack);
2004 }
2005
2006 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2007 {
2008         struct netlink_ext_ack *extack = cb->extack;
2009         const struct nlmsghdr *nlh = cb->nlh;
2010         struct net *net = sock_net(skb->sk);
2011         struct net *tgt_net = net;
2012         int h, s_h;
2013         int idx = 0, s_idx;
2014         struct net_device *dev;
2015         struct hlist_head *head;
2016         struct nlattr *tb[IFLA_MAX+1];
2017         u32 ext_filter_mask = 0;
2018         const struct rtnl_link_ops *kind_ops = NULL;
2019         unsigned int flags = NLM_F_MULTI;
2020         int master_idx = 0;
2021         int netnsid = -1;
2022         int err, i;
2023
2024         s_h = cb->args[0];
2025         s_idx = cb->args[1];
2026
2027         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2028         if (err < 0) {
2029                 if (cb->strict_check)
2030                         return err;
2031
2032                 goto walk_entries;
2033         }
2034
2035         for (i = 0; i <= IFLA_MAX; ++i) {
2036                 if (!tb[i])
2037                         continue;
2038
2039                 /* new attributes should only be added with strict checking */
2040                 switch (i) {
2041                 case IFLA_TARGET_NETNSID:
2042                         netnsid = nla_get_s32(tb[i]);
2043                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2044                         if (IS_ERR(tgt_net)) {
2045                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2046                                 return PTR_ERR(tgt_net);
2047                         }
2048                         break;
2049                 case IFLA_EXT_MASK:
2050                         ext_filter_mask = nla_get_u32(tb[i]);
2051                         break;
2052                 case IFLA_MASTER:
2053                         master_idx = nla_get_u32(tb[i]);
2054                         break;
2055                 case IFLA_LINKINFO:
2056                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2057                         break;
2058                 default:
2059                         if (cb->strict_check) {
2060                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2061                                 return -EINVAL;
2062                         }
2063                 }
2064         }
2065
2066         if (master_idx || kind_ops)
2067                 flags |= NLM_F_DUMP_FILTERED;
2068
2069 walk_entries:
2070         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2071                 idx = 0;
2072                 head = &tgt_net->dev_index_head[h];
2073                 hlist_for_each_entry(dev, head, index_hlist) {
2074                         if (link_dump_filtered(dev, master_idx, kind_ops))
2075                                 goto cont;
2076                         if (idx < s_idx)
2077                                 goto cont;
2078                         err = rtnl_fill_ifinfo(skb, dev, net,
2079                                                RTM_NEWLINK,
2080                                                NETLINK_CB(cb->skb).portid,
2081                                                nlh->nlmsg_seq, 0, flags,
2082                                                ext_filter_mask, 0, NULL, 0,
2083                                                netnsid, GFP_KERNEL);
2084
2085                         if (err < 0) {
2086                                 if (likely(skb->len))
2087                                         goto out;
2088
2089                                 goto out_err;
2090                         }
2091 cont:
2092                         idx++;
2093                 }
2094         }
2095 out:
2096         err = skb->len;
2097 out_err:
2098         cb->args[1] = idx;
2099         cb->args[0] = h;
2100         cb->seq = net->dev_base_seq;
2101         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2102         if (netnsid >= 0)
2103                 put_net(tgt_net);
2104
2105         return err;
2106 }
2107
2108 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2109                         struct netlink_ext_ack *exterr)
2110 {
2111         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2112                                     exterr);
2113 }
2114 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2115
2116 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2117 {
2118         struct net *net;
2119         /* Examine the link attributes and figure out which
2120          * network namespace we are talking about.
2121          */
2122         if (tb[IFLA_NET_NS_PID])
2123                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2124         else if (tb[IFLA_NET_NS_FD])
2125                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2126         else
2127                 net = get_net(src_net);
2128         return net;
2129 }
2130 EXPORT_SYMBOL(rtnl_link_get_net);
2131
2132 /* Figure out which network namespace we are talking about by
2133  * examining the link attributes in the following order:
2134  *
2135  * 1. IFLA_NET_NS_PID
2136  * 2. IFLA_NET_NS_FD
2137  * 3. IFLA_TARGET_NETNSID
2138  */
2139 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2140                                                struct nlattr *tb[])
2141 {
2142         struct net *net;
2143
2144         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2145                 return rtnl_link_get_net(src_net, tb);
2146
2147         if (!tb[IFLA_TARGET_NETNSID])
2148                 return get_net(src_net);
2149
2150         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2151         if (!net)
2152                 return ERR_PTR(-EINVAL);
2153
2154         return net;
2155 }
2156
2157 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2158                                              struct net *src_net,
2159                                              struct nlattr *tb[], int cap)
2160 {
2161         struct net *net;
2162
2163         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2164         if (IS_ERR(net))
2165                 return net;
2166
2167         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2168                 put_net(net);
2169                 return ERR_PTR(-EPERM);
2170         }
2171
2172         return net;
2173 }
2174
2175 /* Verify that rtnetlink requests do not pass additional properties
2176  * potentially referring to different network namespaces.
2177  */
2178 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2179                                     struct netlink_ext_ack *extack,
2180                                     bool netns_id_only)
2181 {
2182
2183         if (netns_id_only) {
2184                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2185                         return 0;
2186
2187                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2188                 return -EOPNOTSUPP;
2189         }
2190
2191         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2192                 goto invalid_attr;
2193
2194         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2195                 goto invalid_attr;
2196
2197         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2198                 goto invalid_attr;
2199
2200         return 0;
2201
2202 invalid_attr:
2203         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2204         return -EINVAL;
2205 }
2206
2207 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2208 {
2209         if (dev) {
2210                 if (tb[IFLA_ADDRESS] &&
2211                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2212                         return -EINVAL;
2213
2214                 if (tb[IFLA_BROADCAST] &&
2215                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2216                         return -EINVAL;
2217         }
2218
2219         if (tb[IFLA_AF_SPEC]) {
2220                 struct nlattr *af;
2221                 int rem, err;
2222
2223                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2224                         const struct rtnl_af_ops *af_ops;
2225
2226                         rcu_read_lock();
2227                         af_ops = rtnl_af_lookup(nla_type(af));
2228                         if (!af_ops) {
2229                                 rcu_read_unlock();
2230                                 return -EAFNOSUPPORT;
2231                         }
2232
2233                         if (!af_ops->set_link_af) {
2234                                 rcu_read_unlock();
2235                                 return -EOPNOTSUPP;
2236                         }
2237
2238                         if (af_ops->validate_link_af) {
2239                                 err = af_ops->validate_link_af(dev, af);
2240                                 if (err < 0) {
2241                                         rcu_read_unlock();
2242                                         return err;
2243                                 }
2244                         }
2245
2246                         rcu_read_unlock();
2247                 }
2248         }
2249
2250         return 0;
2251 }
2252
2253 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2254                                   int guid_type)
2255 {
2256         const struct net_device_ops *ops = dev->netdev_ops;
2257
2258         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2259 }
2260
2261 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2262 {
2263         if (dev->type != ARPHRD_INFINIBAND)
2264                 return -EOPNOTSUPP;
2265
2266         return handle_infiniband_guid(dev, ivt, guid_type);
2267 }
2268
2269 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2270 {
2271         const struct net_device_ops *ops = dev->netdev_ops;
2272         int err = -EINVAL;
2273
2274         if (tb[IFLA_VF_MAC]) {
2275                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2276
2277                 if (ivm->vf >= INT_MAX)
2278                         return -EINVAL;
2279                 err = -EOPNOTSUPP;
2280                 if (ops->ndo_set_vf_mac)
2281                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2282                                                   ivm->mac);
2283                 if (err < 0)
2284                         return err;
2285         }
2286
2287         if (tb[IFLA_VF_VLAN]) {
2288                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2289
2290                 if (ivv->vf >= INT_MAX)
2291                         return -EINVAL;
2292                 err = -EOPNOTSUPP;
2293                 if (ops->ndo_set_vf_vlan)
2294                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2295                                                    ivv->qos,
2296                                                    htons(ETH_P_8021Q));
2297                 if (err < 0)
2298                         return err;
2299         }
2300
2301         if (tb[IFLA_VF_VLAN_LIST]) {
2302                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2303                 struct nlattr *attr;
2304                 int rem, len = 0;
2305
2306                 err = -EOPNOTSUPP;
2307                 if (!ops->ndo_set_vf_vlan)
2308                         return err;
2309
2310                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2311                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2312                             nla_len(attr) < NLA_HDRLEN) {
2313                                 return -EINVAL;
2314                         }
2315                         if (len >= MAX_VLAN_LIST_LEN)
2316                                 return -EOPNOTSUPP;
2317                         ivvl[len] = nla_data(attr);
2318
2319                         len++;
2320                 }
2321                 if (len == 0)
2322                         return -EINVAL;
2323
2324                 if (ivvl[0]->vf >= INT_MAX)
2325                         return -EINVAL;
2326                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2327                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2328                 if (err < 0)
2329                         return err;
2330         }
2331
2332         if (tb[IFLA_VF_TX_RATE]) {
2333                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2334                 struct ifla_vf_info ivf;
2335
2336                 if (ivt->vf >= INT_MAX)
2337                         return -EINVAL;
2338                 err = -EOPNOTSUPP;
2339                 if (ops->ndo_get_vf_config)
2340                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2341                 if (err < 0)
2342                         return err;
2343
2344                 err = -EOPNOTSUPP;
2345                 if (ops->ndo_set_vf_rate)
2346                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2347                                                    ivf.min_tx_rate,
2348                                                    ivt->rate);
2349                 if (err < 0)
2350                         return err;
2351         }
2352
2353         if (tb[IFLA_VF_RATE]) {
2354                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2355
2356                 if (ivt->vf >= INT_MAX)
2357                         return -EINVAL;
2358                 err = -EOPNOTSUPP;
2359                 if (ops->ndo_set_vf_rate)
2360                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2361                                                    ivt->min_tx_rate,
2362                                                    ivt->max_tx_rate);
2363                 if (err < 0)
2364                         return err;
2365         }
2366
2367         if (tb[IFLA_VF_SPOOFCHK]) {
2368                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2369
2370                 if (ivs->vf >= INT_MAX)
2371                         return -EINVAL;
2372                 err = -EOPNOTSUPP;
2373                 if (ops->ndo_set_vf_spoofchk)
2374                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2375                                                        ivs->setting);
2376                 if (err < 0)
2377                         return err;
2378         }
2379
2380         if (tb[IFLA_VF_LINK_STATE]) {
2381                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2382
2383                 if (ivl->vf >= INT_MAX)
2384                         return -EINVAL;
2385                 err = -EOPNOTSUPP;
2386                 if (ops->ndo_set_vf_link_state)
2387                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2388                                                          ivl->link_state);
2389                 if (err < 0)
2390                         return err;
2391         }
2392
2393         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2394                 struct ifla_vf_rss_query_en *ivrssq_en;
2395
2396                 err = -EOPNOTSUPP;
2397                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2398                 if (ivrssq_en->vf >= INT_MAX)
2399                         return -EINVAL;
2400                 if (ops->ndo_set_vf_rss_query_en)
2401                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2402                                                            ivrssq_en->setting);
2403                 if (err < 0)
2404                         return err;
2405         }
2406
2407         if (tb[IFLA_VF_TRUST]) {
2408                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2409
2410                 if (ivt->vf >= INT_MAX)
2411                         return -EINVAL;
2412                 err = -EOPNOTSUPP;
2413                 if (ops->ndo_set_vf_trust)
2414                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2415                 if (err < 0)
2416                         return err;
2417         }
2418
2419         if (tb[IFLA_VF_IB_NODE_GUID]) {
2420                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2421
2422                 if (ivt->vf >= INT_MAX)
2423                         return -EINVAL;
2424                 if (!ops->ndo_set_vf_guid)
2425                         return -EOPNOTSUPP;
2426                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2427         }
2428
2429         if (tb[IFLA_VF_IB_PORT_GUID]) {
2430                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2431
2432                 if (ivt->vf >= INT_MAX)
2433                         return -EINVAL;
2434                 if (!ops->ndo_set_vf_guid)
2435                         return -EOPNOTSUPP;
2436
2437                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2438         }
2439
2440         return err;
2441 }
2442
2443 static int do_set_master(struct net_device *dev, int ifindex,
2444                          struct netlink_ext_ack *extack)
2445 {
2446         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2447         const struct net_device_ops *ops;
2448         int err;
2449
2450         if (upper_dev) {
2451                 if (upper_dev->ifindex == ifindex)
2452                         return 0;
2453                 ops = upper_dev->netdev_ops;
2454                 if (ops->ndo_del_slave) {
2455                         err = ops->ndo_del_slave(upper_dev, dev);
2456                         if (err)
2457                                 return err;
2458                         netdev_update_lockdep_key(dev);
2459                 } else {
2460                         return -EOPNOTSUPP;
2461                 }
2462         }
2463
2464         if (ifindex) {
2465                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2466                 if (!upper_dev)
2467                         return -EINVAL;
2468                 ops = upper_dev->netdev_ops;
2469                 if (ops->ndo_add_slave) {
2470                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2471                         if (err)
2472                                 return err;
2473                 } else {
2474                         return -EOPNOTSUPP;
2475                 }
2476         }
2477         return 0;
2478 }
2479
2480 #define DO_SETLINK_MODIFIED     0x01
2481 /* notify flag means notify + modified. */
2482 #define DO_SETLINK_NOTIFY       0x03
2483 static int do_setlink(const struct sk_buff *skb,
2484                       struct net_device *dev, struct ifinfomsg *ifm,
2485                       struct netlink_ext_ack *extack,
2486                       struct nlattr **tb, char *ifname, int status)
2487 {
2488         const struct net_device_ops *ops = dev->netdev_ops;
2489         int err;
2490
2491         err = validate_linkmsg(dev, tb);
2492         if (err < 0)
2493                 return err;
2494
2495         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2496                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2497                                                             tb, CAP_NET_ADMIN);
2498                 if (IS_ERR(net)) {
2499                         err = PTR_ERR(net);
2500                         goto errout;
2501                 }
2502
2503                 err = dev_change_net_namespace(dev, net, ifname);
2504                 put_net(net);
2505                 if (err)
2506                         goto errout;
2507                 status |= DO_SETLINK_MODIFIED;
2508         }
2509
2510         if (tb[IFLA_MAP]) {
2511                 struct rtnl_link_ifmap *u_map;
2512                 struct ifmap k_map;
2513
2514                 if (!ops->ndo_set_config) {
2515                         err = -EOPNOTSUPP;
2516                         goto errout;
2517                 }
2518
2519                 if (!netif_device_present(dev)) {
2520                         err = -ENODEV;
2521                         goto errout;
2522                 }
2523
2524                 u_map = nla_data(tb[IFLA_MAP]);
2525                 k_map.mem_start = (unsigned long) u_map->mem_start;
2526                 k_map.mem_end = (unsigned long) u_map->mem_end;
2527                 k_map.base_addr = (unsigned short) u_map->base_addr;
2528                 k_map.irq = (unsigned char) u_map->irq;
2529                 k_map.dma = (unsigned char) u_map->dma;
2530                 k_map.port = (unsigned char) u_map->port;
2531
2532                 err = ops->ndo_set_config(dev, &k_map);
2533                 if (err < 0)
2534                         goto errout;
2535
2536                 status |= DO_SETLINK_NOTIFY;
2537         }
2538
2539         if (tb[IFLA_ADDRESS]) {
2540                 struct sockaddr *sa;
2541                 int len;
2542
2543                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2544                                                   sizeof(*sa));
2545                 sa = kmalloc(len, GFP_KERNEL);
2546                 if (!sa) {
2547                         err = -ENOMEM;
2548                         goto errout;
2549                 }
2550                 sa->sa_family = dev->type;
2551                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2552                        dev->addr_len);
2553                 err = dev_set_mac_address(dev, sa, extack);
2554                 kfree(sa);
2555                 if (err)
2556                         goto errout;
2557                 status |= DO_SETLINK_MODIFIED;
2558         }
2559
2560         if (tb[IFLA_MTU]) {
2561                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2562                 if (err < 0)
2563                         goto errout;
2564                 status |= DO_SETLINK_MODIFIED;
2565         }
2566
2567         if (tb[IFLA_GROUP]) {
2568                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2569                 status |= DO_SETLINK_NOTIFY;
2570         }
2571
2572         /*
2573          * Interface selected by interface index but interface
2574          * name provided implies that a name change has been
2575          * requested.
2576          */
2577         if (ifm->ifi_index > 0 && ifname[0]) {
2578                 err = dev_change_name(dev, ifname);
2579                 if (err < 0)
2580                         goto errout;
2581                 status |= DO_SETLINK_MODIFIED;
2582         }
2583
2584         if (tb[IFLA_IFALIAS]) {
2585                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2586                                     nla_len(tb[IFLA_IFALIAS]));
2587                 if (err < 0)
2588                         goto errout;
2589                 status |= DO_SETLINK_NOTIFY;
2590         }
2591
2592         if (tb[IFLA_BROADCAST]) {
2593                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2594                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2595         }
2596
2597         if (ifm->ifi_flags || ifm->ifi_change) {
2598                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2599                                        extack);
2600                 if (err < 0)
2601                         goto errout;
2602         }
2603
2604         if (tb[IFLA_MASTER]) {
2605                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2606                 if (err)
2607                         goto errout;
2608                 status |= DO_SETLINK_MODIFIED;
2609         }
2610
2611         if (tb[IFLA_CARRIER]) {
2612                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2613                 if (err)
2614                         goto errout;
2615                 status |= DO_SETLINK_MODIFIED;
2616         }
2617
2618         if (tb[IFLA_TXQLEN]) {
2619                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2620
2621                 err = dev_change_tx_queue_len(dev, value);
2622                 if (err)
2623                         goto errout;
2624                 status |= DO_SETLINK_MODIFIED;
2625         }
2626
2627         if (tb[IFLA_GSO_MAX_SIZE]) {
2628                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2629
2630                 if (max_size > GSO_MAX_SIZE) {
2631                         err = -EINVAL;
2632                         goto errout;
2633                 }
2634
2635                 if (dev->gso_max_size ^ max_size) {
2636                         netif_set_gso_max_size(dev, max_size);
2637                         status |= DO_SETLINK_MODIFIED;
2638                 }
2639         }
2640
2641         if (tb[IFLA_GSO_MAX_SEGS]) {
2642                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2643
2644                 if (max_segs > GSO_MAX_SEGS) {
2645                         err = -EINVAL;
2646                         goto errout;
2647                 }
2648
2649                 if (dev->gso_max_segs ^ max_segs) {
2650                         dev->gso_max_segs = max_segs;
2651                         status |= DO_SETLINK_MODIFIED;
2652                 }
2653         }
2654
2655         if (tb[IFLA_OPERSTATE])
2656                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2657
2658         if (tb[IFLA_LINKMODE]) {
2659                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2660
2661                 write_lock_bh(&dev_base_lock);
2662                 if (dev->link_mode ^ value)
2663                         status |= DO_SETLINK_NOTIFY;
2664                 dev->link_mode = value;
2665                 write_unlock_bh(&dev_base_lock);
2666         }
2667
2668         if (tb[IFLA_VFINFO_LIST]) {
2669                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2670                 struct nlattr *attr;
2671                 int rem;
2672
2673                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2674                         if (nla_type(attr) != IFLA_VF_INFO ||
2675                             nla_len(attr) < NLA_HDRLEN) {
2676                                 err = -EINVAL;
2677                                 goto errout;
2678                         }
2679                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2680                                                           attr,
2681                                                           ifla_vf_policy,
2682                                                           NULL);
2683                         if (err < 0)
2684                                 goto errout;
2685                         err = do_setvfinfo(dev, vfinfo);
2686                         if (err < 0)
2687                                 goto errout;
2688                         status |= DO_SETLINK_NOTIFY;
2689                 }
2690         }
2691         err = 0;
2692
2693         if (tb[IFLA_VF_PORTS]) {
2694                 struct nlattr *port[IFLA_PORT_MAX+1];
2695                 struct nlattr *attr;
2696                 int vf;
2697                 int rem;
2698
2699                 err = -EOPNOTSUPP;
2700                 if (!ops->ndo_set_vf_port)
2701                         goto errout;
2702
2703                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2704                         if (nla_type(attr) != IFLA_VF_PORT ||
2705                             nla_len(attr) < NLA_HDRLEN) {
2706                                 err = -EINVAL;
2707                                 goto errout;
2708                         }
2709                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2710                                                           attr,
2711                                                           ifla_port_policy,
2712                                                           NULL);
2713                         if (err < 0)
2714                                 goto errout;
2715                         if (!port[IFLA_PORT_VF]) {
2716                                 err = -EOPNOTSUPP;
2717                                 goto errout;
2718                         }
2719                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2720                         err = ops->ndo_set_vf_port(dev, vf, port);
2721                         if (err < 0)
2722                                 goto errout;
2723                         status |= DO_SETLINK_NOTIFY;
2724                 }
2725         }
2726         err = 0;
2727
2728         if (tb[IFLA_PORT_SELF]) {
2729                 struct nlattr *port[IFLA_PORT_MAX+1];
2730
2731                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2732                                                   tb[IFLA_PORT_SELF],
2733                                                   ifla_port_policy, NULL);
2734                 if (err < 0)
2735                         goto errout;
2736
2737                 err = -EOPNOTSUPP;
2738                 if (ops->ndo_set_vf_port)
2739                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2740                 if (err < 0)
2741                         goto errout;
2742                 status |= DO_SETLINK_NOTIFY;
2743         }
2744
2745         if (tb[IFLA_AF_SPEC]) {
2746                 struct nlattr *af;
2747                 int rem;
2748
2749                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2750                         const struct rtnl_af_ops *af_ops;
2751
2752                         rcu_read_lock();
2753
2754                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2755
2756                         err = af_ops->set_link_af(dev, af);
2757                         if (err < 0) {
2758                                 rcu_read_unlock();
2759                                 goto errout;
2760                         }
2761
2762                         rcu_read_unlock();
2763                         status |= DO_SETLINK_NOTIFY;
2764                 }
2765         }
2766         err = 0;
2767
2768         if (tb[IFLA_PROTO_DOWN]) {
2769                 err = dev_change_proto_down(dev,
2770                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2771                 if (err)
2772                         goto errout;
2773                 status |= DO_SETLINK_NOTIFY;
2774         }
2775
2776         if (tb[IFLA_XDP]) {
2777                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2778                 u32 xdp_flags = 0;
2779
2780                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2781                                                   tb[IFLA_XDP],
2782                                                   ifla_xdp_policy, NULL);
2783                 if (err < 0)
2784                         goto errout;
2785
2786                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2787                         err = -EINVAL;
2788                         goto errout;
2789                 }
2790
2791                 if (xdp[IFLA_XDP_FLAGS]) {
2792                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2793                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2794                                 err = -EINVAL;
2795                                 goto errout;
2796                         }
2797                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2798                                 err = -EINVAL;
2799                                 goto errout;
2800                         }
2801                 }
2802
2803                 if (xdp[IFLA_XDP_FD]) {
2804                         int expected_fd = -1;
2805
2806                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2807                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2808                                         err = -EINVAL;
2809                                         goto errout;
2810                                 }
2811                                 expected_fd =
2812                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2813                         }
2814
2815                         err = dev_change_xdp_fd(dev, extack,
2816                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2817                                                 expected_fd,
2818                                                 xdp_flags);
2819                         if (err)
2820                                 goto errout;
2821                         status |= DO_SETLINK_NOTIFY;
2822                 }
2823         }
2824
2825 errout:
2826         if (status & DO_SETLINK_MODIFIED) {
2827                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2828                         netdev_state_change(dev);
2829
2830                 if (err < 0)
2831                         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",
2832                                              dev->name);
2833         }
2834
2835         return err;
2836 }
2837
2838 static struct net_device *rtnl_dev_get(struct net *net,
2839                                        struct nlattr *ifname_attr,
2840                                        struct nlattr *altifname_attr,
2841                                        char *ifname)
2842 {
2843         char buffer[ALTIFNAMSIZ];
2844
2845         if (!ifname) {
2846                 ifname = buffer;
2847                 if (ifname_attr)
2848                         nla_strlcpy(ifname, ifname_attr, IFNAMSIZ);
2849                 else if (altifname_attr)
2850                         nla_strlcpy(ifname, altifname_attr, ALTIFNAMSIZ);
2851                 else
2852                         return NULL;
2853         }
2854
2855         return __dev_get_by_name(net, ifname);
2856 }
2857
2858 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2859                         struct netlink_ext_ack *extack)
2860 {
2861         struct net *net = sock_net(skb->sk);
2862         struct ifinfomsg *ifm;
2863         struct net_device *dev;
2864         int err;
2865         struct nlattr *tb[IFLA_MAX+1];
2866         char ifname[IFNAMSIZ];
2867
2868         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2869                                      ifla_policy, extack);
2870         if (err < 0)
2871                 goto errout;
2872
2873         err = rtnl_ensure_unique_netns(tb, extack, false);
2874         if (err < 0)
2875                 goto errout;
2876
2877         if (tb[IFLA_IFNAME])
2878                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2879         else
2880                 ifname[0] = '\0';
2881
2882         err = -EINVAL;
2883         ifm = nlmsg_data(nlh);
2884         if (ifm->ifi_index > 0)
2885                 dev = __dev_get_by_index(net, ifm->ifi_index);
2886         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2887                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2888         else
2889                 goto errout;
2890
2891         if (dev == NULL) {
2892                 err = -ENODEV;
2893                 goto errout;
2894         }
2895
2896         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2897 errout:
2898         return err;
2899 }
2900
2901 static int rtnl_group_dellink(const struct net *net, int group)
2902 {
2903         struct net_device *dev, *aux;
2904         LIST_HEAD(list_kill);
2905         bool found = false;
2906
2907         if (!group)
2908                 return -EPERM;
2909
2910         for_each_netdev(net, dev) {
2911                 if (dev->group == group) {
2912                         const struct rtnl_link_ops *ops;
2913
2914                         found = true;
2915                         ops = dev->rtnl_link_ops;
2916                         if (!ops || !ops->dellink)
2917                                 return -EOPNOTSUPP;
2918                 }
2919         }
2920
2921         if (!found)
2922                 return -ENODEV;
2923
2924         for_each_netdev_safe(net, dev, aux) {
2925                 if (dev->group == group) {
2926                         const struct rtnl_link_ops *ops;
2927
2928                         ops = dev->rtnl_link_ops;
2929                         ops->dellink(dev, &list_kill);
2930                 }
2931         }
2932         unregister_netdevice_many(&list_kill);
2933
2934         return 0;
2935 }
2936
2937 int rtnl_delete_link(struct net_device *dev)
2938 {
2939         const struct rtnl_link_ops *ops;
2940         LIST_HEAD(list_kill);
2941
2942         ops = dev->rtnl_link_ops;
2943         if (!ops || !ops->dellink)
2944                 return -EOPNOTSUPP;
2945
2946         ops->dellink(dev, &list_kill);
2947         unregister_netdevice_many(&list_kill);
2948
2949         return 0;
2950 }
2951 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2952
2953 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2954                         struct netlink_ext_ack *extack)
2955 {
2956         struct net *net = sock_net(skb->sk);
2957         struct net *tgt_net = net;
2958         struct net_device *dev = NULL;
2959         struct ifinfomsg *ifm;
2960         struct nlattr *tb[IFLA_MAX+1];
2961         int err;
2962         int netnsid = -1;
2963
2964         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2965                                      ifla_policy, extack);
2966         if (err < 0)
2967                 return err;
2968
2969         err = rtnl_ensure_unique_netns(tb, extack, true);
2970         if (err < 0)
2971                 return err;
2972
2973         if (tb[IFLA_TARGET_NETNSID]) {
2974                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
2975                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
2976                 if (IS_ERR(tgt_net))
2977                         return PTR_ERR(tgt_net);
2978         }
2979
2980         err = -EINVAL;
2981         ifm = nlmsg_data(nlh);
2982         if (ifm->ifi_index > 0)
2983                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2984         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2985                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
2986                                    tb[IFLA_ALT_IFNAME], NULL);
2987         else if (tb[IFLA_GROUP])
2988                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2989         else
2990                 goto out;
2991
2992         if (!dev) {
2993                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2994                         err = -ENODEV;
2995
2996                 goto out;
2997         }
2998
2999         err = rtnl_delete_link(dev);
3000
3001 out:
3002         if (netnsid >= 0)
3003                 put_net(tgt_net);
3004
3005         return err;
3006 }
3007
3008 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3009 {
3010         unsigned int old_flags;
3011         int err;
3012
3013         old_flags = dev->flags;
3014         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3015                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3016                                          NULL);
3017                 if (err < 0)
3018                         return err;
3019         }
3020
3021         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3022                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3023         } else {
3024                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3025                 __dev_notify_flags(dev, old_flags, ~0U);
3026         }
3027         return 0;
3028 }
3029 EXPORT_SYMBOL(rtnl_configure_link);
3030
3031 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3032                                     unsigned char name_assign_type,
3033                                     const struct rtnl_link_ops *ops,
3034                                     struct nlattr *tb[],
3035                                     struct netlink_ext_ack *extack)
3036 {
3037         struct net_device *dev;
3038         unsigned int num_tx_queues = 1;
3039         unsigned int num_rx_queues = 1;
3040
3041         if (tb[IFLA_NUM_TX_QUEUES])
3042                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3043         else if (ops->get_num_tx_queues)
3044                 num_tx_queues = ops->get_num_tx_queues();
3045
3046         if (tb[IFLA_NUM_RX_QUEUES])
3047                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3048         else if (ops->get_num_rx_queues)
3049                 num_rx_queues = ops->get_num_rx_queues();
3050
3051         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3052                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3053                 return ERR_PTR(-EINVAL);
3054         }
3055
3056         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3057                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3058                 return ERR_PTR(-EINVAL);
3059         }
3060
3061         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3062                                ops->setup, num_tx_queues, num_rx_queues);
3063         if (!dev)
3064                 return ERR_PTR(-ENOMEM);
3065
3066         dev_net_set(dev, net);
3067         dev->rtnl_link_ops = ops;
3068         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3069
3070         if (tb[IFLA_MTU]) {
3071                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3072                 int err;
3073
3074                 err = dev_validate_mtu(dev, mtu, extack);
3075                 if (err) {
3076                         free_netdev(dev);
3077                         return ERR_PTR(err);
3078                 }
3079                 dev->mtu = mtu;
3080         }
3081         if (tb[IFLA_ADDRESS]) {
3082                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3083                                 nla_len(tb[IFLA_ADDRESS]));
3084                 dev->addr_assign_type = NET_ADDR_SET;
3085         }
3086         if (tb[IFLA_BROADCAST])
3087                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3088                                 nla_len(tb[IFLA_BROADCAST]));
3089         if (tb[IFLA_TXQLEN])
3090                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3091         if (tb[IFLA_OPERSTATE])
3092                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3093         if (tb[IFLA_LINKMODE])
3094                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3095         if (tb[IFLA_GROUP])
3096                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3097         if (tb[IFLA_GSO_MAX_SIZE])
3098                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3099         if (tb[IFLA_GSO_MAX_SEGS])
3100                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3101
3102         return dev;
3103 }
3104 EXPORT_SYMBOL(rtnl_create_link);
3105
3106 static int rtnl_group_changelink(const struct sk_buff *skb,
3107                 struct net *net, int group,
3108                 struct ifinfomsg *ifm,
3109                 struct netlink_ext_ack *extack,
3110                 struct nlattr **tb)
3111 {
3112         struct net_device *dev, *aux;
3113         int err;
3114
3115         for_each_netdev_safe(net, dev, aux) {
3116                 if (dev->group == group) {
3117                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3118                         if (err < 0)
3119                                 return err;
3120                 }
3121         }
3122
3123         return 0;
3124 }
3125
3126 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3127                           struct nlattr **attr, struct netlink_ext_ack *extack)
3128 {
3129         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3130         unsigned char name_assign_type = NET_NAME_USER;
3131         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3132         const struct rtnl_link_ops *m_ops = NULL;
3133         struct net_device *master_dev = NULL;
3134         struct net *net = sock_net(skb->sk);
3135         const struct rtnl_link_ops *ops;
3136         struct nlattr *tb[IFLA_MAX + 1];
3137         struct net *dest_net, *link_net;
3138         struct nlattr **slave_data;
3139         char kind[MODULE_NAME_LEN];
3140         struct net_device *dev;
3141         struct ifinfomsg *ifm;
3142         char ifname[IFNAMSIZ];
3143         struct nlattr **data;
3144         int err;
3145
3146 #ifdef CONFIG_MODULES
3147 replay:
3148 #endif
3149         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3150                                      ifla_policy, extack);
3151         if (err < 0)
3152                 return err;
3153
3154         err = rtnl_ensure_unique_netns(tb, extack, false);
3155         if (err < 0)
3156                 return err;
3157
3158         if (tb[IFLA_IFNAME])
3159                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3160         else
3161                 ifname[0] = '\0';
3162
3163         ifm = nlmsg_data(nlh);
3164         if (ifm->ifi_index > 0)
3165                 dev = __dev_get_by_index(net, ifm->ifi_index);
3166         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3167                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3168         else
3169                 dev = NULL;
3170
3171         if (dev) {
3172                 master_dev = netdev_master_upper_dev_get(dev);
3173                 if (master_dev)
3174                         m_ops = master_dev->rtnl_link_ops;
3175         }
3176
3177         err = validate_linkmsg(dev, tb);
3178         if (err < 0)
3179                 return err;
3180
3181         if (tb[IFLA_LINKINFO]) {
3182                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3183                                                   tb[IFLA_LINKINFO],
3184                                                   ifla_info_policy, NULL);
3185                 if (err < 0)
3186                         return err;
3187         } else
3188                 memset(linkinfo, 0, sizeof(linkinfo));
3189
3190         if (linkinfo[IFLA_INFO_KIND]) {
3191                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3192                 ops = rtnl_link_ops_get(kind);
3193         } else {
3194                 kind[0] = '\0';
3195                 ops = NULL;
3196         }
3197
3198         data = NULL;
3199         if (ops) {
3200                 if (ops->maxtype > RTNL_MAX_TYPE)
3201                         return -EINVAL;
3202
3203                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3204                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3205                                                           linkinfo[IFLA_INFO_DATA],
3206                                                           ops->policy, extack);
3207                         if (err < 0)
3208                                 return err;
3209                         data = attr;
3210                 }
3211                 if (ops->validate) {
3212                         err = ops->validate(tb, data, extack);
3213                         if (err < 0)
3214                                 return err;
3215                 }
3216         }
3217
3218         slave_data = NULL;
3219         if (m_ops) {
3220                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3221                         return -EINVAL;
3222
3223                 if (m_ops->slave_maxtype &&
3224                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3225                         err = nla_parse_nested_deprecated(slave_attr,
3226                                                           m_ops->slave_maxtype,
3227                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3228                                                           m_ops->slave_policy,
3229                                                           extack);
3230                         if (err < 0)
3231                                 return err;
3232                         slave_data = slave_attr;
3233                 }
3234         }
3235
3236         if (dev) {
3237                 int status = 0;
3238
3239                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3240                         return -EEXIST;
3241                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3242                         return -EOPNOTSUPP;
3243
3244                 if (linkinfo[IFLA_INFO_DATA]) {
3245                         if (!ops || ops != dev->rtnl_link_ops ||
3246                             !ops->changelink)
3247                                 return -EOPNOTSUPP;
3248
3249                         err = ops->changelink(dev, tb, data, extack);
3250                         if (err < 0)
3251                                 return err;
3252                         status |= DO_SETLINK_NOTIFY;
3253                 }
3254
3255                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3256                         if (!m_ops || !m_ops->slave_changelink)
3257                                 return -EOPNOTSUPP;
3258
3259                         err = m_ops->slave_changelink(master_dev, dev, tb,
3260                                                       slave_data, extack);
3261                         if (err < 0)
3262                                 return err;
3263                         status |= DO_SETLINK_NOTIFY;
3264                 }
3265
3266                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3267         }
3268
3269         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3270                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3271                         return rtnl_group_changelink(skb, net,
3272                                                 nla_get_u32(tb[IFLA_GROUP]),
3273                                                 ifm, extack, tb);
3274                 return -ENODEV;
3275         }
3276
3277         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3278                 return -EOPNOTSUPP;
3279
3280         if (!ops) {
3281 #ifdef CONFIG_MODULES
3282                 if (kind[0]) {
3283                         __rtnl_unlock();
3284                         request_module("rtnl-link-%s", kind);
3285                         rtnl_lock();
3286                         ops = rtnl_link_ops_get(kind);
3287                         if (ops)
3288                                 goto replay;
3289                 }
3290 #endif
3291                 NL_SET_ERR_MSG(extack, "Unknown device type");
3292                 return -EOPNOTSUPP;
3293         }
3294
3295         if (!ops->setup)
3296                 return -EOPNOTSUPP;
3297
3298         if (!ifname[0]) {
3299                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3300                 name_assign_type = NET_NAME_ENUM;
3301         }
3302
3303         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3304         if (IS_ERR(dest_net))
3305                 return PTR_ERR(dest_net);
3306
3307         if (tb[IFLA_LINK_NETNSID]) {
3308                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3309
3310                 link_net = get_net_ns_by_id(dest_net, id);
3311                 if (!link_net) {
3312                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3313                         err =  -EINVAL;
3314                         goto out;
3315                 }
3316                 err = -EPERM;
3317                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3318                         goto out;
3319         } else {
3320                 link_net = NULL;
3321         }
3322
3323         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3324                                name_assign_type, ops, tb, extack);
3325         if (IS_ERR(dev)) {
3326                 err = PTR_ERR(dev);
3327                 goto out;
3328         }
3329
3330         dev->ifindex = ifm->ifi_index;
3331
3332         if (ops->newlink) {
3333                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3334                 /* Drivers should call free_netdev() in ->destructor
3335                  * and unregister it on failure after registration
3336                  * so that device could be finally freed in rtnl_unlock.
3337                  */
3338                 if (err < 0) {
3339                         /* If device is not registered at all, free it now */
3340                         if (dev->reg_state == NETREG_UNINITIALIZED)
3341                                 free_netdev(dev);
3342                         goto out;
3343                 }
3344         } else {
3345                 err = register_netdevice(dev);
3346                 if (err < 0) {
3347                         free_netdev(dev);
3348                         goto out;
3349                 }
3350         }
3351         err = rtnl_configure_link(dev, ifm);
3352         if (err < 0)
3353                 goto out_unregister;
3354         if (link_net) {
3355                 err = dev_change_net_namespace(dev, dest_net, ifname);
3356                 if (err < 0)
3357                         goto out_unregister;
3358         }
3359         if (tb[IFLA_MASTER]) {
3360                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3361                 if (err)
3362                         goto out_unregister;
3363         }
3364 out:
3365         if (link_net)
3366                 put_net(link_net);
3367         put_net(dest_net);
3368         return err;
3369 out_unregister:
3370         if (ops->newlink) {
3371                 LIST_HEAD(list_kill);
3372
3373                 ops->dellink(dev, &list_kill);
3374                 unregister_netdevice_many(&list_kill);
3375         } else {
3376                 unregister_netdevice(dev);
3377         }
3378         goto out;
3379 }
3380
3381 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3382                         struct netlink_ext_ack *extack)
3383 {
3384         struct nlattr **attr;
3385         int ret;
3386
3387         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3388         if (!attr)
3389                 return -ENOMEM;
3390
3391         ret = __rtnl_newlink(skb, nlh, attr, extack);
3392         kfree(attr);
3393         return ret;
3394 }
3395
3396 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3397                                   const struct nlmsghdr *nlh,
3398                                   struct nlattr **tb,
3399                                   struct netlink_ext_ack *extack)
3400 {
3401         struct ifinfomsg *ifm;
3402         int i, err;
3403
3404         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3405                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3406                 return -EINVAL;
3407         }
3408
3409         if (!netlink_strict_get_check(skb))
3410                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3411                                               ifla_policy, extack);
3412
3413         ifm = nlmsg_data(nlh);
3414         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3415             ifm->ifi_change) {
3416                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3417                 return -EINVAL;
3418         }
3419
3420         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3421                                             ifla_policy, extack);
3422         if (err)
3423                 return err;
3424
3425         for (i = 0; i <= IFLA_MAX; i++) {
3426                 if (!tb[i])
3427                         continue;
3428
3429                 switch (i) {
3430                 case IFLA_IFNAME:
3431                 case IFLA_ALT_IFNAME:
3432                 case IFLA_EXT_MASK:
3433                 case IFLA_TARGET_NETNSID:
3434                         break;
3435                 default:
3436                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3437                         return -EINVAL;
3438                 }
3439         }
3440
3441         return 0;
3442 }
3443
3444 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3445                         struct netlink_ext_ack *extack)
3446 {
3447         struct net *net = sock_net(skb->sk);
3448         struct net *tgt_net = net;
3449         struct ifinfomsg *ifm;
3450         struct nlattr *tb[IFLA_MAX+1];
3451         struct net_device *dev = NULL;
3452         struct sk_buff *nskb;
3453         int netnsid = -1;
3454         int err;
3455         u32 ext_filter_mask = 0;
3456
3457         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3458         if (err < 0)
3459                 return err;
3460
3461         err = rtnl_ensure_unique_netns(tb, extack, true);
3462         if (err < 0)
3463                 return err;
3464
3465         if (tb[IFLA_TARGET_NETNSID]) {
3466                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3467                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3468                 if (IS_ERR(tgt_net))
3469                         return PTR_ERR(tgt_net);
3470         }
3471
3472         if (tb[IFLA_EXT_MASK])
3473                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3474
3475         err = -EINVAL;
3476         ifm = nlmsg_data(nlh);
3477         if (ifm->ifi_index > 0)
3478                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3479         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3480                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3481                                    tb[IFLA_ALT_IFNAME], NULL);
3482         else
3483                 goto out;
3484
3485         err = -ENODEV;
3486         if (dev == NULL)
3487                 goto out;
3488
3489         err = -ENOBUFS;
3490         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3491         if (nskb == NULL)
3492                 goto out;
3493
3494         err = rtnl_fill_ifinfo(nskb, dev, net,
3495                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3496                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3497                                0, NULL, 0, netnsid, GFP_KERNEL);
3498         if (err < 0) {
3499                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3500                 WARN_ON(err == -EMSGSIZE);
3501                 kfree_skb(nskb);
3502         } else
3503                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3504 out:
3505         if (netnsid >= 0)
3506                 put_net(tgt_net);
3507
3508         return err;
3509 }
3510
3511 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3512                            bool *changed, struct netlink_ext_ack *extack)
3513 {
3514         char *alt_ifname;
3515         int err;
3516
3517         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3518         if (err)
3519                 return err;
3520
3521         alt_ifname = nla_strdup(attr, GFP_KERNEL);
3522         if (!alt_ifname)
3523                 return -ENOMEM;
3524
3525         if (cmd == RTM_NEWLINKPROP) {
3526                 err = netdev_name_node_alt_create(dev, alt_ifname);
3527                 if (!err)
3528                         alt_ifname = NULL;
3529         } else if (cmd == RTM_DELLINKPROP) {
3530                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3531         } else {
3532                 WARN_ON_ONCE(1);
3533                 err = -EINVAL;
3534         }
3535
3536         kfree(alt_ifname);
3537         if (!err)
3538                 *changed = true;
3539         return err;
3540 }
3541
3542 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3543                          struct netlink_ext_ack *extack)
3544 {
3545         struct net *net = sock_net(skb->sk);
3546         struct nlattr *tb[IFLA_MAX + 1];
3547         struct net_device *dev;
3548         struct ifinfomsg *ifm;
3549         bool changed = false;
3550         struct nlattr *attr;
3551         int err, rem;
3552
3553         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3554         if (err)
3555                 return err;
3556
3557         err = rtnl_ensure_unique_netns(tb, extack, true);
3558         if (err)
3559                 return err;
3560
3561         ifm = nlmsg_data(nlh);
3562         if (ifm->ifi_index > 0)
3563                 dev = __dev_get_by_index(net, ifm->ifi_index);
3564         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3565                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3566                                    tb[IFLA_ALT_IFNAME], NULL);
3567         else
3568                 return -EINVAL;
3569
3570         if (!dev)
3571                 return -ENODEV;
3572
3573         if (!tb[IFLA_PROP_LIST])
3574                 return 0;
3575
3576         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3577                 switch (nla_type(attr)) {
3578                 case IFLA_ALT_IFNAME:
3579                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3580                         if (err)
3581                                 return err;
3582                         break;
3583                 }
3584         }
3585
3586         if (changed)
3587                 netdev_state_change(dev);
3588         return 0;
3589 }
3590
3591 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3592                             struct netlink_ext_ack *extack)
3593 {
3594         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3595 }
3596
3597 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3598                             struct netlink_ext_ack *extack)
3599 {
3600         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3601 }
3602
3603 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3604 {
3605         struct net *net = sock_net(skb->sk);
3606         struct net_device *dev;
3607         struct nlattr *tb[IFLA_MAX+1];
3608         u32 ext_filter_mask = 0;
3609         u16 min_ifinfo_dump_size = 0;
3610         int hdrlen;
3611
3612         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3613         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3614                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3615
3616         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3617                 if (tb[IFLA_EXT_MASK])
3618                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3619         }
3620
3621         if (!ext_filter_mask)
3622                 return NLMSG_GOODSIZE;
3623         /*
3624          * traverse the list of net devices and compute the minimum
3625          * buffer size based upon the filter mask.
3626          */
3627         rcu_read_lock();
3628         for_each_netdev_rcu(net, dev) {
3629                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3630                                              if_nlmsg_size(dev,
3631                                                            ext_filter_mask));
3632         }
3633         rcu_read_unlock();
3634
3635         return nlmsg_total_size(min_ifinfo_dump_size);
3636 }
3637
3638 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3639 {
3640         int idx;
3641         int s_idx = cb->family;
3642         int type = cb->nlh->nlmsg_type - RTM_BASE;
3643         int ret = 0;
3644
3645         if (s_idx == 0)
3646                 s_idx = 1;
3647
3648         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3649                 struct rtnl_link **tab;
3650                 struct rtnl_link *link;
3651                 rtnl_dumpit_func dumpit;
3652
3653                 if (idx < s_idx || idx == PF_PACKET)
3654                         continue;
3655
3656                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3657                         continue;
3658
3659                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3660                 if (!tab)
3661                         continue;
3662
3663                 link = tab[type];
3664                 if (!link)
3665                         continue;
3666
3667                 dumpit = link->dumpit;
3668                 if (!dumpit)
3669                         continue;
3670
3671                 if (idx > s_idx) {
3672                         memset(&cb->args[0], 0, sizeof(cb->args));
3673                         cb->prev_seq = 0;
3674                         cb->seq = 0;
3675                 }
3676                 ret = dumpit(skb, cb);
3677                 if (ret)
3678                         break;
3679         }
3680         cb->family = idx;
3681
3682         return skb->len ? : ret;
3683 }
3684
3685 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3686                                        unsigned int change,
3687                                        u32 event, gfp_t flags, int *new_nsid,
3688                                        int new_ifindex)
3689 {
3690         struct net *net = dev_net(dev);
3691         struct sk_buff *skb;
3692         int err = -ENOBUFS;
3693         size_t if_info_size;
3694
3695         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3696         if (skb == NULL)
3697                 goto errout;
3698
3699         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3700                                type, 0, 0, change, 0, 0, event,
3701                                new_nsid, new_ifindex, -1, flags);
3702         if (err < 0) {
3703                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3704                 WARN_ON(err == -EMSGSIZE);
3705                 kfree_skb(skb);
3706                 goto errout;
3707         }
3708         return skb;
3709 errout:
3710         if (err < 0)
3711                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3712         return NULL;
3713 }
3714
3715 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3716 {
3717         struct net *net = dev_net(dev);
3718
3719         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3720 }
3721
3722 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3723                                unsigned int change, u32 event,
3724                                gfp_t flags, int *new_nsid, int new_ifindex)
3725 {
3726         struct sk_buff *skb;
3727
3728         if (dev->reg_state != NETREG_REGISTERED)
3729                 return;
3730
3731         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3732                                      new_ifindex);
3733         if (skb)
3734                 rtmsg_ifinfo_send(skb, dev, flags);
3735 }
3736
3737 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3738                   gfp_t flags)
3739 {
3740         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3741                            NULL, 0);
3742 }
3743
3744 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3745                          gfp_t flags, int *new_nsid, int new_ifindex)
3746 {
3747         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3748                            new_nsid, new_ifindex);
3749 }
3750
3751 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3752                                    struct net_device *dev,
3753                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3754                                    int type, unsigned int flags,
3755                                    int nlflags, u16 ndm_state)
3756 {
3757         struct nlmsghdr *nlh;
3758         struct ndmsg *ndm;
3759
3760         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3761         if (!nlh)
3762                 return -EMSGSIZE;
3763
3764         ndm = nlmsg_data(nlh);
3765         ndm->ndm_family  = AF_BRIDGE;
3766         ndm->ndm_pad1    = 0;
3767         ndm->ndm_pad2    = 0;
3768         ndm->ndm_flags   = flags;
3769         ndm->ndm_type    = 0;
3770         ndm->ndm_ifindex = dev->ifindex;
3771         ndm->ndm_state   = ndm_state;
3772
3773         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3774                 goto nla_put_failure;
3775         if (vid)
3776                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3777                         goto nla_put_failure;
3778
3779         nlmsg_end(skb, nlh);
3780         return 0;
3781
3782 nla_put_failure:
3783         nlmsg_cancel(skb, nlh);
3784         return -EMSGSIZE;
3785 }
3786
3787 static inline size_t rtnl_fdb_nlmsg_size(void)
3788 {
3789         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3790                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3791                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3792                0;
3793 }
3794
3795 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3796                             u16 ndm_state)
3797 {
3798         struct net *net = dev_net(dev);
3799         struct sk_buff *skb;
3800         int err = -ENOBUFS;
3801
3802         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3803         if (!skb)
3804                 goto errout;
3805
3806         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3807                                       0, 0, type, NTF_SELF, 0, ndm_state);
3808         if (err < 0) {
3809                 kfree_skb(skb);
3810                 goto errout;
3811         }
3812
3813         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3814         return;
3815 errout:
3816         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3817 }
3818
3819 /*
3820  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3821  */
3822 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3823                      struct nlattr *tb[],
3824                      struct net_device *dev,
3825                      const unsigned char *addr, u16 vid,
3826                      u16 flags)
3827 {
3828         int err = -EINVAL;
3829
3830         /* If aging addresses are supported device will need to
3831          * implement its own handler for this.
3832          */
3833         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3834                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3835                 return err;
3836         }
3837
3838         if (vid) {
3839                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3840                 return err;
3841         }
3842
3843         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3844                 err = dev_uc_add_excl(dev, addr);
3845         else if (is_multicast_ether_addr(addr))
3846                 err = dev_mc_add_excl(dev, addr);
3847
3848         /* Only return duplicate errors if NLM_F_EXCL is set */
3849         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3850                 err = 0;
3851
3852         return err;
3853 }
3854 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3855
3856 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3857                          struct netlink_ext_ack *extack)
3858 {
3859         u16 vid = 0;
3860
3861         if (vlan_attr) {
3862                 if (nla_len(vlan_attr) != sizeof(u16)) {
3863                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3864                         return -EINVAL;
3865                 }
3866
3867                 vid = nla_get_u16(vlan_attr);
3868
3869                 if (!vid || vid >= VLAN_VID_MASK) {
3870                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3871                         return -EINVAL;
3872                 }
3873         }
3874         *p_vid = vid;
3875         return 0;
3876 }
3877
3878 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3879                         struct netlink_ext_ack *extack)
3880 {
3881         struct net *net = sock_net(skb->sk);
3882         struct ndmsg *ndm;
3883         struct nlattr *tb[NDA_MAX+1];
3884         struct net_device *dev;
3885         u8 *addr;
3886         u16 vid;
3887         int err;
3888
3889         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3890                                      extack);
3891         if (err < 0)
3892                 return err;
3893
3894         ndm = nlmsg_data(nlh);
3895         if (ndm->ndm_ifindex == 0) {
3896                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3897                 return -EINVAL;
3898         }
3899
3900         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3901         if (dev == NULL) {
3902                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3903                 return -ENODEV;
3904         }
3905
3906         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3907                 NL_SET_ERR_MSG(extack, "invalid address");
3908                 return -EINVAL;
3909         }
3910
3911         if (dev->type != ARPHRD_ETHER) {
3912                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3913                 return -EINVAL;
3914         }
3915
3916         addr = nla_data(tb[NDA_LLADDR]);
3917
3918         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3919         if (err)
3920                 return err;
3921
3922         err = -EOPNOTSUPP;
3923
3924         /* Support fdb on master device the net/bridge default case */
3925         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3926             netif_is_bridge_port(dev)) {
3927                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3928                 const struct net_device_ops *ops = br_dev->netdev_ops;
3929
3930                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3931                                        nlh->nlmsg_flags, extack);
3932                 if (err)
3933                         goto out;
3934                 else
3935                         ndm->ndm_flags &= ~NTF_MASTER;
3936         }
3937
3938         /* Embedded bridge, macvlan, and any other device support */
3939         if ((ndm->ndm_flags & NTF_SELF)) {
3940                 if (dev->netdev_ops->ndo_fdb_add)
3941                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3942                                                            vid,
3943                                                            nlh->nlmsg_flags,
3944                                                            extack);
3945                 else
3946                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3947                                                nlh->nlmsg_flags);
3948
3949                 if (!err) {
3950                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3951                                         ndm->ndm_state);
3952                         ndm->ndm_flags &= ~NTF_SELF;
3953                 }
3954         }
3955 out:
3956         return err;
3957 }
3958
3959 /*
3960  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3961  */
3962 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3963                      struct nlattr *tb[],
3964                      struct net_device *dev,
3965                      const unsigned char *addr, u16 vid)
3966 {
3967         int err = -EINVAL;
3968
3969         /* If aging addresses are supported device will need to
3970          * implement its own handler for this.
3971          */
3972         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3973                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3974                 return err;
3975         }
3976
3977         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3978                 err = dev_uc_del(dev, addr);
3979         else if (is_multicast_ether_addr(addr))
3980                 err = dev_mc_del(dev, addr);
3981
3982         return err;
3983 }
3984 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3985
3986 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3987                         struct netlink_ext_ack *extack)
3988 {
3989         struct net *net = sock_net(skb->sk);
3990         struct ndmsg *ndm;
3991         struct nlattr *tb[NDA_MAX+1];
3992         struct net_device *dev;
3993         int err = -EINVAL;
3994         __u8 *addr;
3995         u16 vid;
3996
3997         if (!netlink_capable(skb, CAP_NET_ADMIN))
3998                 return -EPERM;
3999
4000         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4001                                      extack);
4002         if (err < 0)
4003                 return err;
4004
4005         ndm = nlmsg_data(nlh);
4006         if (ndm->ndm_ifindex == 0) {
4007                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4008                 return -EINVAL;
4009         }
4010
4011         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4012         if (dev == NULL) {
4013                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4014                 return -ENODEV;
4015         }
4016
4017         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4018                 NL_SET_ERR_MSG(extack, "invalid address");
4019                 return -EINVAL;
4020         }
4021
4022         if (dev->type != ARPHRD_ETHER) {
4023                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4024                 return -EINVAL;
4025         }
4026
4027         addr = nla_data(tb[NDA_LLADDR]);
4028
4029         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4030         if (err)
4031                 return err;
4032
4033         err = -EOPNOTSUPP;
4034
4035         /* Support fdb on master device the net/bridge default case */
4036         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4037             netif_is_bridge_port(dev)) {
4038                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4039                 const struct net_device_ops *ops = br_dev->netdev_ops;
4040
4041                 if (ops->ndo_fdb_del)
4042                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4043
4044                 if (err)
4045                         goto out;
4046                 else
4047                         ndm->ndm_flags &= ~NTF_MASTER;
4048         }
4049
4050         /* Embedded bridge, macvlan, and any other device support */
4051         if (ndm->ndm_flags & NTF_SELF) {
4052                 if (dev->netdev_ops->ndo_fdb_del)
4053                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4054                                                            vid);
4055                 else
4056                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4057
4058                 if (!err) {
4059                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4060                                         ndm->ndm_state);
4061                         ndm->ndm_flags &= ~NTF_SELF;
4062                 }
4063         }
4064 out:
4065         return err;
4066 }
4067
4068 static int nlmsg_populate_fdb(struct sk_buff *skb,
4069                               struct netlink_callback *cb,
4070                               struct net_device *dev,
4071                               int *idx,
4072                               struct netdev_hw_addr_list *list)
4073 {
4074         struct netdev_hw_addr *ha;
4075         int err;
4076         u32 portid, seq;
4077
4078         portid = NETLINK_CB(cb->skb).portid;
4079         seq = cb->nlh->nlmsg_seq;
4080
4081         list_for_each_entry(ha, &list->list, list) {
4082                 if (*idx < cb->args[2])
4083                         goto skip;
4084
4085                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4086                                               portid, seq,
4087                                               RTM_NEWNEIGH, NTF_SELF,
4088                                               NLM_F_MULTI, NUD_PERMANENT);
4089                 if (err < 0)
4090                         return err;
4091 skip:
4092                 *idx += 1;
4093         }
4094         return 0;
4095 }
4096
4097 /**
4098  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4099  * @skb: socket buffer to store message in
4100  * @cb: netlink callback
4101  * @dev: netdevice
4102  * @filter_dev: ignored
4103  * @idx: the number of FDB table entries dumped is added to *@idx
4104  *
4105  * Default netdevice operation to dump the existing unicast address list.
4106  * Returns number of addresses from list put in skb.
4107  */
4108 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4109                       struct netlink_callback *cb,
4110                       struct net_device *dev,
4111                       struct net_device *filter_dev,
4112                       int *idx)
4113 {
4114         int err;
4115
4116         if (dev->type != ARPHRD_ETHER)
4117                 return -EINVAL;
4118
4119         netif_addr_lock_bh(dev);
4120         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4121         if (err)
4122                 goto out;
4123         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4124 out:
4125         netif_addr_unlock_bh(dev);
4126         return err;
4127 }
4128 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4129
4130 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4131                                  int *br_idx, int *brport_idx,
4132                                  struct netlink_ext_ack *extack)
4133 {
4134         struct nlattr *tb[NDA_MAX + 1];
4135         struct ndmsg *ndm;
4136         int err, i;
4137
4138         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4139                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4140                 return -EINVAL;
4141         }
4142
4143         ndm = nlmsg_data(nlh);
4144         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4145             ndm->ndm_flags || ndm->ndm_type) {
4146                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4147                 return -EINVAL;
4148         }
4149
4150         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4151                                             NDA_MAX, NULL, extack);
4152         if (err < 0)
4153                 return err;
4154
4155         *brport_idx = ndm->ndm_ifindex;
4156         for (i = 0; i <= NDA_MAX; ++i) {
4157                 if (!tb[i])
4158                         continue;
4159
4160                 switch (i) {
4161                 case NDA_IFINDEX:
4162                         if (nla_len(tb[i]) != sizeof(u32)) {
4163                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4164                                 return -EINVAL;
4165                         }
4166                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4167                         break;
4168                 case NDA_MASTER:
4169                         if (nla_len(tb[i]) != sizeof(u32)) {
4170                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4171                                 return -EINVAL;
4172                         }
4173                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4174                         break;
4175                 default:
4176                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4177                         return -EINVAL;
4178                 }
4179         }
4180
4181         return 0;
4182 }
4183
4184 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4185                                  int *br_idx, int *brport_idx,
4186                                  struct netlink_ext_ack *extack)
4187 {
4188         struct nlattr *tb[IFLA_MAX+1];
4189         int err;
4190
4191         /* A hack to preserve kernel<->userspace interface.
4192          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4193          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4194          * So, check for ndmsg with an optional u32 attribute (not used here).
4195          * Fortunately these sizes don't conflict with the size of ifinfomsg
4196          * with an optional attribute.
4197          */
4198         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4199             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4200              nla_attr_size(sizeof(u32)))) {
4201                 struct ifinfomsg *ifm;
4202
4203                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4204                                              tb, IFLA_MAX, ifla_policy,
4205                                              extack);
4206                 if (err < 0) {
4207                         return -EINVAL;
4208                 } else if (err == 0) {
4209                         if (tb[IFLA_MASTER])
4210                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4211                 }
4212
4213                 ifm = nlmsg_data(nlh);
4214                 *brport_idx = ifm->ifi_index;
4215         }
4216         return 0;
4217 }
4218
4219 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4220 {
4221         struct net_device *dev;
4222         struct net_device *br_dev = NULL;
4223         const struct net_device_ops *ops = NULL;
4224         const struct net_device_ops *cops = NULL;
4225         struct net *net = sock_net(skb->sk);
4226         struct hlist_head *head;
4227         int brport_idx = 0;
4228         int br_idx = 0;
4229         int h, s_h;
4230         int idx = 0, s_idx;
4231         int err = 0;
4232         int fidx = 0;
4233
4234         if (cb->strict_check)
4235                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4236                                             cb->extack);
4237         else
4238                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4239                                             cb->extack);
4240         if (err < 0)
4241                 return err;
4242
4243         if (br_idx) {
4244                 br_dev = __dev_get_by_index(net, br_idx);
4245                 if (!br_dev)
4246                         return -ENODEV;
4247
4248                 ops = br_dev->netdev_ops;
4249         }
4250
4251         s_h = cb->args[0];
4252         s_idx = cb->args[1];
4253
4254         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4255                 idx = 0;
4256                 head = &net->dev_index_head[h];
4257                 hlist_for_each_entry(dev, head, index_hlist) {
4258
4259                         if (brport_idx && (dev->ifindex != brport_idx))
4260                                 continue;
4261
4262                         if (!br_idx) { /* user did not specify a specific bridge */
4263                                 if (netif_is_bridge_port(dev)) {
4264                                         br_dev = netdev_master_upper_dev_get(dev);
4265                                         cops = br_dev->netdev_ops;
4266                                 }
4267                         } else {
4268                                 if (dev != br_dev &&
4269                                     !netif_is_bridge_port(dev))
4270                                         continue;
4271
4272                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4273                                     !(dev->priv_flags & IFF_EBRIDGE))
4274                                         continue;
4275                                 cops = ops;
4276                         }
4277
4278                         if (idx < s_idx)
4279                                 goto cont;
4280
4281                         if (netif_is_bridge_port(dev)) {
4282                                 if (cops && cops->ndo_fdb_dump) {
4283                                         err = cops->ndo_fdb_dump(skb, cb,
4284                                                                 br_dev, dev,
4285                                                                 &fidx);
4286                                         if (err == -EMSGSIZE)
4287                                                 goto out;
4288                                 }
4289                         }
4290
4291                         if (dev->netdev_ops->ndo_fdb_dump)
4292                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4293                                                                     dev, NULL,
4294                                                                     &fidx);
4295                         else
4296                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4297                                                         &fidx);
4298                         if (err == -EMSGSIZE)
4299                                 goto out;
4300
4301                         cops = NULL;
4302
4303                         /* reset fdb offset to 0 for rest of the interfaces */
4304                         cb->args[2] = 0;
4305                         fidx = 0;
4306 cont:
4307                         idx++;
4308                 }
4309         }
4310
4311 out:
4312         cb->args[0] = h;
4313         cb->args[1] = idx;
4314         cb->args[2] = fidx;
4315
4316         return skb->len;
4317 }
4318
4319 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4320                                 struct nlattr **tb, u8 *ndm_flags,
4321                                 int *br_idx, int *brport_idx, u8 **addr,
4322                                 u16 *vid, struct netlink_ext_ack *extack)
4323 {
4324         struct ndmsg *ndm;
4325         int err, i;
4326
4327         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4328                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4329                 return -EINVAL;
4330         }
4331
4332         ndm = nlmsg_data(nlh);
4333         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4334             ndm->ndm_type) {
4335                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4336                 return -EINVAL;
4337         }
4338
4339         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4340                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4341                 return -EINVAL;
4342         }
4343
4344         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4345                                             NDA_MAX, nda_policy, extack);
4346         if (err < 0)
4347                 return err;
4348
4349         *ndm_flags = ndm->ndm_flags;
4350         *brport_idx = ndm->ndm_ifindex;
4351         for (i = 0; i <= NDA_MAX; ++i) {
4352                 if (!tb[i])
4353                         continue;
4354
4355                 switch (i) {
4356                 case NDA_MASTER:
4357                         *br_idx = nla_get_u32(tb[i]);
4358                         break;
4359                 case NDA_LLADDR:
4360                         if (nla_len(tb[i]) != ETH_ALEN) {
4361                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4362                                 return -EINVAL;
4363                         }
4364                         *addr = nla_data(tb[i]);
4365                         break;
4366                 case NDA_VLAN:
4367                         err = fdb_vid_parse(tb[i], vid, extack);
4368                         if (err)
4369                                 return err;
4370                         break;
4371                 case NDA_VNI:
4372                         break;
4373                 default:
4374                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4375                         return -EINVAL;
4376                 }
4377         }
4378
4379         return 0;
4380 }
4381
4382 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4383                         struct netlink_ext_ack *extack)
4384 {
4385         struct net_device *dev = NULL, *br_dev = NULL;
4386         const struct net_device_ops *ops = NULL;
4387         struct net *net = sock_net(in_skb->sk);
4388         struct nlattr *tb[NDA_MAX + 1];
4389         struct sk_buff *skb;
4390         int brport_idx = 0;
4391         u8 ndm_flags = 0;
4392         int br_idx = 0;
4393         u8 *addr = NULL;
4394         u16 vid = 0;
4395         int err;
4396
4397         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4398                                    &brport_idx, &addr, &vid, extack);
4399         if (err < 0)
4400                 return err;
4401
4402         if (!addr) {
4403                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4404                 return -EINVAL;
4405         }
4406
4407         if (brport_idx) {
4408                 dev = __dev_get_by_index(net, brport_idx);
4409                 if (!dev) {
4410                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4411                         return -ENODEV;
4412                 }
4413         }
4414
4415         if (br_idx) {
4416                 if (dev) {
4417                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4418                         return -EINVAL;
4419                 }
4420
4421                 br_dev = __dev_get_by_index(net, br_idx);
4422                 if (!br_dev) {
4423                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4424                         return -EINVAL;
4425                 }
4426                 ops = br_dev->netdev_ops;
4427         }
4428
4429         if (dev) {
4430                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4431                         if (!netif_is_bridge_port(dev)) {
4432                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4433                                 return -EINVAL;
4434                         }
4435                         br_dev = netdev_master_upper_dev_get(dev);
4436                         if (!br_dev) {
4437                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4438                                 return -EINVAL;
4439                         }
4440                         ops = br_dev->netdev_ops;
4441                 } else {
4442                         if (!(ndm_flags & NTF_SELF)) {
4443                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4444                                 return -EINVAL;
4445                         }
4446                         ops = dev->netdev_ops;
4447                 }
4448         }
4449
4450         if (!br_dev && !dev) {
4451                 NL_SET_ERR_MSG(extack, "No device specified");
4452                 return -ENODEV;
4453         }
4454
4455         if (!ops || !ops->ndo_fdb_get) {
4456                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4457                 return -EOPNOTSUPP;
4458         }
4459
4460         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4461         if (!skb)
4462                 return -ENOBUFS;
4463
4464         if (br_dev)
4465                 dev = br_dev;
4466         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4467                                NETLINK_CB(in_skb).portid,
4468                                nlh->nlmsg_seq, extack);
4469         if (err)
4470                 goto out;
4471
4472         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4473 out:
4474         kfree_skb(skb);
4475         return err;
4476 }
4477
4478 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4479                                unsigned int attrnum, unsigned int flag)
4480 {
4481         if (mask & flag)
4482                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4483         return 0;
4484 }
4485
4486 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4487                             struct net_device *dev, u16 mode,
4488                             u32 flags, u32 mask, int nlflags,
4489                             u32 filter_mask,
4490                             int (*vlan_fill)(struct sk_buff *skb,
4491                                              struct net_device *dev,
4492                                              u32 filter_mask))
4493 {
4494         struct nlmsghdr *nlh;
4495         struct ifinfomsg *ifm;
4496         struct nlattr *br_afspec;
4497         struct nlattr *protinfo;
4498         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4499         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4500         int err = 0;
4501
4502         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4503         if (nlh == NULL)
4504                 return -EMSGSIZE;
4505
4506         ifm = nlmsg_data(nlh);
4507         ifm->ifi_family = AF_BRIDGE;
4508         ifm->__ifi_pad = 0;
4509         ifm->ifi_type = dev->type;
4510         ifm->ifi_index = dev->ifindex;
4511         ifm->ifi_flags = dev_get_flags(dev);
4512         ifm->ifi_change = 0;
4513
4514
4515         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4516             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4517             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4518             (br_dev &&
4519              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4520             (dev->addr_len &&
4521              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4522             (dev->ifindex != dev_get_iflink(dev) &&
4523              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4524                 goto nla_put_failure;
4525
4526         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4527         if (!br_afspec)
4528                 goto nla_put_failure;
4529
4530         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4531                 nla_nest_cancel(skb, br_afspec);
4532                 goto nla_put_failure;
4533         }
4534
4535         if (mode != BRIDGE_MODE_UNDEF) {
4536                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4537                         nla_nest_cancel(skb, br_afspec);
4538                         goto nla_put_failure;
4539                 }
4540         }
4541         if (vlan_fill) {
4542                 err = vlan_fill(skb, dev, filter_mask);
4543                 if (err) {
4544                         nla_nest_cancel(skb, br_afspec);
4545                         goto nla_put_failure;
4546                 }
4547         }
4548         nla_nest_end(skb, br_afspec);
4549
4550         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4551         if (!protinfo)
4552                 goto nla_put_failure;
4553
4554         if (brport_nla_put_flag(skb, flags, mask,
4555                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4556             brport_nla_put_flag(skb, flags, mask,
4557                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4558             brport_nla_put_flag(skb, flags, mask,
4559                                 IFLA_BRPORT_FAST_LEAVE,
4560                                 BR_MULTICAST_FAST_LEAVE) ||
4561             brport_nla_put_flag(skb, flags, mask,
4562                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4563             brport_nla_put_flag(skb, flags, mask,
4564                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4565             brport_nla_put_flag(skb, flags, mask,
4566                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4567             brport_nla_put_flag(skb, flags, mask,
4568                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4569             brport_nla_put_flag(skb, flags, mask,
4570                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4571             brport_nla_put_flag(skb, flags, mask,
4572                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4573             brport_nla_put_flag(skb, flags, mask,
4574                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4575                 nla_nest_cancel(skb, protinfo);
4576                 goto nla_put_failure;
4577         }
4578
4579         nla_nest_end(skb, protinfo);
4580
4581         nlmsg_end(skb, nlh);
4582         return 0;
4583 nla_put_failure:
4584         nlmsg_cancel(skb, nlh);
4585         return err ? err : -EMSGSIZE;
4586 }
4587 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4588
4589 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4590                                     bool strict_check, u32 *filter_mask,
4591                                     struct netlink_ext_ack *extack)
4592 {
4593         struct nlattr *tb[IFLA_MAX+1];
4594         int err, i;
4595
4596         if (strict_check) {
4597                 struct ifinfomsg *ifm;
4598
4599                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4600                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4601                         return -EINVAL;
4602                 }
4603
4604                 ifm = nlmsg_data(nlh);
4605                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4606                     ifm->ifi_change || ifm->ifi_index) {
4607                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4608                         return -EINVAL;
4609                 }
4610
4611                 err = nlmsg_parse_deprecated_strict(nlh,
4612                                                     sizeof(struct ifinfomsg),
4613                                                     tb, IFLA_MAX, ifla_policy,
4614                                                     extack);
4615         } else {
4616                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4617                                              tb, IFLA_MAX, ifla_policy,
4618                                              extack);
4619         }
4620         if (err < 0)
4621                 return err;
4622
4623         /* new attributes should only be added with strict checking */
4624         for (i = 0; i <= IFLA_MAX; ++i) {
4625                 if (!tb[i])
4626                         continue;
4627
4628                 switch (i) {
4629                 case IFLA_EXT_MASK:
4630                         *filter_mask = nla_get_u32(tb[i]);
4631                         break;
4632                 default:
4633                         if (strict_check) {
4634                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4635                                 return -EINVAL;
4636                         }
4637                 }
4638         }
4639
4640         return 0;
4641 }
4642
4643 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4644 {
4645         const struct nlmsghdr *nlh = cb->nlh;
4646         struct net *net = sock_net(skb->sk);
4647         struct net_device *dev;
4648         int idx = 0;
4649         u32 portid = NETLINK_CB(cb->skb).portid;
4650         u32 seq = nlh->nlmsg_seq;
4651         u32 filter_mask = 0;
4652         int err;
4653
4654         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4655                                        cb->extack);
4656         if (err < 0 && cb->strict_check)
4657                 return err;
4658
4659         rcu_read_lock();
4660         for_each_netdev_rcu(net, dev) {
4661                 const struct net_device_ops *ops = dev->netdev_ops;
4662                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4663
4664                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4665                         if (idx >= cb->args[0]) {
4666                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4667                                                 skb, portid, seq, dev,
4668                                                 filter_mask, 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                 if (ops->ndo_bridge_getlink) {
4680                         if (idx >= cb->args[0]) {
4681                                 err = ops->ndo_bridge_getlink(skb, portid,
4682                                                               seq, dev,
4683                                                               filter_mask,
4684                                                               NLM_F_MULTI);
4685                                 if (err < 0 && err != -EOPNOTSUPP) {
4686                                         if (likely(skb->len))
4687                                                 break;
4688
4689                                         goto out_err;
4690                                 }
4691                         }
4692                         idx++;
4693                 }
4694         }
4695         err = skb->len;
4696 out_err:
4697         rcu_read_unlock();
4698         cb->args[0] = idx;
4699
4700         return err;
4701 }
4702
4703 static inline size_t bridge_nlmsg_size(void)
4704 {
4705         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4706                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4707                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4708                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4709                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4710                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4711                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4712                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4713                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4714                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4715                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4716 }
4717
4718 static int rtnl_bridge_notify(struct net_device *dev)
4719 {
4720         struct net *net = dev_net(dev);
4721         struct sk_buff *skb;
4722         int err = -EOPNOTSUPP;
4723
4724         if (!dev->netdev_ops->ndo_bridge_getlink)
4725                 return 0;
4726
4727         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4728         if (!skb) {
4729                 err = -ENOMEM;
4730                 goto errout;
4731         }
4732
4733         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4734         if (err < 0)
4735                 goto errout;
4736
4737         if (!skb->len)
4738                 goto errout;
4739
4740         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4741         return 0;
4742 errout:
4743         WARN_ON(err == -EMSGSIZE);
4744         kfree_skb(skb);
4745         if (err)
4746                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4747         return err;
4748 }
4749
4750 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4751                                struct netlink_ext_ack *extack)
4752 {
4753         struct net *net = sock_net(skb->sk);
4754         struct ifinfomsg *ifm;
4755         struct net_device *dev;
4756         struct nlattr *br_spec, *attr = NULL;
4757         int rem, err = -EOPNOTSUPP;
4758         u16 flags = 0;
4759         bool have_flags = false;
4760
4761         if (nlmsg_len(nlh) < sizeof(*ifm))
4762                 return -EINVAL;
4763
4764         ifm = nlmsg_data(nlh);
4765         if (ifm->ifi_family != AF_BRIDGE)
4766                 return -EPFNOSUPPORT;
4767
4768         dev = __dev_get_by_index(net, ifm->ifi_index);
4769         if (!dev) {
4770                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4771                 return -ENODEV;
4772         }
4773
4774         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4775         if (br_spec) {
4776                 nla_for_each_nested(attr, br_spec, rem) {
4777                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4778                                 if (nla_len(attr) < sizeof(flags))
4779                                         return -EINVAL;
4780
4781                                 have_flags = true;
4782                                 flags = nla_get_u16(attr);
4783                                 break;
4784                         }
4785                 }
4786         }
4787
4788         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4789                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4790
4791                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4792                         err = -EOPNOTSUPP;
4793                         goto out;
4794                 }
4795
4796                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4797                                                              extack);
4798                 if (err)
4799                         goto out;
4800
4801                 flags &= ~BRIDGE_FLAGS_MASTER;
4802         }
4803
4804         if ((flags & BRIDGE_FLAGS_SELF)) {
4805                 if (!dev->netdev_ops->ndo_bridge_setlink)
4806                         err = -EOPNOTSUPP;
4807                 else
4808                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4809                                                                   flags,
4810                                                                   extack);
4811                 if (!err) {
4812                         flags &= ~BRIDGE_FLAGS_SELF;
4813
4814                         /* Generate event to notify upper layer of bridge
4815                          * change
4816                          */
4817                         err = rtnl_bridge_notify(dev);
4818                 }
4819         }
4820
4821         if (have_flags)
4822                 memcpy(nla_data(attr), &flags, sizeof(flags));
4823 out:
4824         return err;
4825 }
4826
4827 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4828                                struct netlink_ext_ack *extack)
4829 {
4830         struct net *net = sock_net(skb->sk);
4831         struct ifinfomsg *ifm;
4832         struct net_device *dev;
4833         struct nlattr *br_spec, *attr = NULL;
4834         int rem, err = -EOPNOTSUPP;
4835         u16 flags = 0;
4836         bool have_flags = false;
4837
4838         if (nlmsg_len(nlh) < sizeof(*ifm))
4839                 return -EINVAL;
4840
4841         ifm = nlmsg_data(nlh);
4842         if (ifm->ifi_family != AF_BRIDGE)
4843                 return -EPFNOSUPPORT;
4844
4845         dev = __dev_get_by_index(net, ifm->ifi_index);
4846         if (!dev) {
4847                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4848                 return -ENODEV;
4849         }
4850
4851         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4852         if (br_spec) {
4853                 nla_for_each_nested(attr, br_spec, rem) {
4854                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4855                                 if (nla_len(attr) < sizeof(flags))
4856                                         return -EINVAL;
4857
4858                                 have_flags = true;
4859                                 flags = nla_get_u16(attr);
4860                                 break;
4861                         }
4862                 }
4863         }
4864
4865         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4866                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4867
4868                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4869                         err = -EOPNOTSUPP;
4870                         goto out;
4871                 }
4872
4873                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4874                 if (err)
4875                         goto out;
4876
4877                 flags &= ~BRIDGE_FLAGS_MASTER;
4878         }
4879
4880         if ((flags & BRIDGE_FLAGS_SELF)) {
4881                 if (!dev->netdev_ops->ndo_bridge_dellink)
4882                         err = -EOPNOTSUPP;
4883                 else
4884                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4885                                                                   flags);
4886
4887                 if (!err) {
4888                         flags &= ~BRIDGE_FLAGS_SELF;
4889
4890                         /* Generate event to notify upper layer of bridge
4891                          * change
4892                          */
4893                         err = rtnl_bridge_notify(dev);
4894                 }
4895         }
4896
4897         if (have_flags)
4898                 memcpy(nla_data(attr), &flags, sizeof(flags));
4899 out:
4900         return err;
4901 }
4902
4903 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4904 {
4905         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4906                (!idxattr || idxattr == attrid);
4907 }
4908
4909 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4910 static int rtnl_get_offload_stats_attr_size(int attr_id)
4911 {
4912         switch (attr_id) {
4913         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4914                 return sizeof(struct rtnl_link_stats64);
4915         }
4916
4917         return 0;
4918 }
4919
4920 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4921                                   int *prividx)
4922 {
4923         struct nlattr *attr = NULL;
4924         int attr_id, size;
4925         void *attr_data;
4926         int err;
4927
4928         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4929               dev->netdev_ops->ndo_get_offload_stats))
4930                 return -ENODATA;
4931
4932         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4933              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4934                 if (attr_id < *prividx)
4935                         continue;
4936
4937                 size = rtnl_get_offload_stats_attr_size(attr_id);
4938                 if (!size)
4939                         continue;
4940
4941                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4942                         continue;
4943
4944                 attr = nla_reserve_64bit(skb, attr_id, size,
4945                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4946                 if (!attr)
4947                         goto nla_put_failure;
4948
4949                 attr_data = nla_data(attr);
4950                 memset(attr_data, 0, size);
4951                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4952                                                              attr_data);
4953                 if (err)
4954                         goto get_offload_stats_failure;
4955         }
4956
4957         if (!attr)
4958                 return -ENODATA;
4959
4960         *prividx = 0;
4961         return 0;
4962
4963 nla_put_failure:
4964         err = -EMSGSIZE;
4965 get_offload_stats_failure:
4966         *prividx = attr_id;
4967         return err;
4968 }
4969
4970 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4971 {
4972         int nla_size = 0;
4973         int attr_id;
4974         int size;
4975
4976         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4977               dev->netdev_ops->ndo_get_offload_stats))
4978                 return 0;
4979
4980         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4981              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4982                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4983                         continue;
4984                 size = rtnl_get_offload_stats_attr_size(attr_id);
4985                 nla_size += nla_total_size_64bit(size);
4986         }
4987
4988         if (nla_size != 0)
4989                 nla_size += nla_total_size(0);
4990
4991         return nla_size;
4992 }
4993
4994 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4995                                int type, u32 pid, u32 seq, u32 change,
4996                                unsigned int flags, unsigned int filter_mask,
4997                                int *idxattr, int *prividx)
4998 {
4999         struct if_stats_msg *ifsm;
5000         struct nlmsghdr *nlh;
5001         struct nlattr *attr;
5002         int s_prividx = *prividx;
5003         int err;
5004
5005         ASSERT_RTNL();
5006
5007         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5008         if (!nlh)
5009                 return -EMSGSIZE;
5010
5011         ifsm = nlmsg_data(nlh);
5012         ifsm->family = PF_UNSPEC;
5013         ifsm->pad1 = 0;
5014         ifsm->pad2 = 0;
5015         ifsm->ifindex = dev->ifindex;
5016         ifsm->filter_mask = filter_mask;
5017
5018         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5019                 struct rtnl_link_stats64 *sp;
5020
5021                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5022                                          sizeof(struct rtnl_link_stats64),
5023                                          IFLA_STATS_UNSPEC);
5024                 if (!attr)
5025                         goto nla_put_failure;
5026
5027                 sp = nla_data(attr);
5028                 dev_get_stats(dev, sp);
5029         }
5030
5031         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5032                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5033
5034                 if (ops && ops->fill_linkxstats) {
5035                         *idxattr = IFLA_STATS_LINK_XSTATS;
5036                         attr = nla_nest_start_noflag(skb,
5037                                                      IFLA_STATS_LINK_XSTATS);
5038                         if (!attr)
5039                                 goto nla_put_failure;
5040
5041                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5042                         nla_nest_end(skb, attr);
5043                         if (err)
5044                                 goto nla_put_failure;
5045                         *idxattr = 0;
5046                 }
5047         }
5048
5049         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5050                              *idxattr)) {
5051                 const struct rtnl_link_ops *ops = NULL;
5052                 const struct net_device *master;
5053
5054                 master = netdev_master_upper_dev_get(dev);
5055                 if (master)
5056                         ops = master->rtnl_link_ops;
5057                 if (ops && ops->fill_linkxstats) {
5058                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5059                         attr = nla_nest_start_noflag(skb,
5060                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5061                         if (!attr)
5062                                 goto nla_put_failure;
5063
5064                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5065                         nla_nest_end(skb, attr);
5066                         if (err)
5067                                 goto nla_put_failure;
5068                         *idxattr = 0;
5069                 }
5070         }
5071
5072         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5073                              *idxattr)) {
5074                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5075                 attr = nla_nest_start_noflag(skb,
5076                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5077                 if (!attr)
5078                         goto nla_put_failure;
5079
5080                 err = rtnl_get_offload_stats(skb, dev, prividx);
5081                 if (err == -ENODATA)
5082                         nla_nest_cancel(skb, attr);
5083                 else
5084                         nla_nest_end(skb, attr);
5085
5086                 if (err && err != -ENODATA)
5087                         goto nla_put_failure;
5088                 *idxattr = 0;
5089         }
5090
5091         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5092                 struct rtnl_af_ops *af_ops;
5093
5094                 *idxattr = IFLA_STATS_AF_SPEC;
5095                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5096                 if (!attr)
5097                         goto nla_put_failure;
5098
5099                 rcu_read_lock();
5100                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5101                         if (af_ops->fill_stats_af) {
5102                                 struct nlattr *af;
5103                                 int err;
5104
5105                                 af = nla_nest_start_noflag(skb,
5106                                                            af_ops->family);
5107                                 if (!af) {
5108                                         rcu_read_unlock();
5109                                         goto nla_put_failure;
5110                                 }
5111                                 err = af_ops->fill_stats_af(skb, dev);
5112
5113                                 if (err == -ENODATA) {
5114                                         nla_nest_cancel(skb, af);
5115                                 } else if (err < 0) {
5116                                         rcu_read_unlock();
5117                                         goto nla_put_failure;
5118                                 }
5119
5120                                 nla_nest_end(skb, af);
5121                         }
5122                 }
5123                 rcu_read_unlock();
5124
5125                 nla_nest_end(skb, attr);
5126
5127                 *idxattr = 0;
5128         }
5129
5130         nlmsg_end(skb, nlh);
5131
5132         return 0;
5133
5134 nla_put_failure:
5135         /* not a multi message or no progress mean a real error */
5136         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5137                 nlmsg_cancel(skb, nlh);
5138         else
5139                 nlmsg_end(skb, nlh);
5140
5141         return -EMSGSIZE;
5142 }
5143
5144 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5145                                   u32 filter_mask)
5146 {
5147         size_t size = 0;
5148
5149         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5150                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5151
5152         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5153                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5154                 int attr = IFLA_STATS_LINK_XSTATS;
5155
5156                 if (ops && ops->get_linkxstats_size) {
5157                         size += nla_total_size(ops->get_linkxstats_size(dev,
5158                                                                         attr));
5159                         /* for IFLA_STATS_LINK_XSTATS */
5160                         size += nla_total_size(0);
5161                 }
5162         }
5163
5164         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5165                 struct net_device *_dev = (struct net_device *)dev;
5166                 const struct rtnl_link_ops *ops = NULL;
5167                 const struct net_device *master;
5168
5169                 /* netdev_master_upper_dev_get can't take const */
5170                 master = netdev_master_upper_dev_get(_dev);
5171                 if (master)
5172                         ops = master->rtnl_link_ops;
5173                 if (ops && ops->get_linkxstats_size) {
5174                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5175
5176                         size += nla_total_size(ops->get_linkxstats_size(dev,
5177                                                                         attr));
5178                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5179                         size += nla_total_size(0);
5180                 }
5181         }
5182
5183         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5184                 size += rtnl_get_offload_stats_size(dev);
5185
5186         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5187                 struct rtnl_af_ops *af_ops;
5188
5189                 /* for IFLA_STATS_AF_SPEC */
5190                 size += nla_total_size(0);
5191
5192                 rcu_read_lock();
5193                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5194                         if (af_ops->get_stats_af_size) {
5195                                 size += nla_total_size(
5196                                         af_ops->get_stats_af_size(dev));
5197
5198                                 /* for AF_* */
5199                                 size += nla_total_size(0);
5200                         }
5201                 }
5202                 rcu_read_unlock();
5203         }
5204
5205         return size;
5206 }
5207
5208 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5209                                 bool is_dump, struct netlink_ext_ack *extack)
5210 {
5211         struct if_stats_msg *ifsm;
5212
5213         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5214                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5215                 return -EINVAL;
5216         }
5217
5218         if (!strict_check)
5219                 return 0;
5220
5221         ifsm = nlmsg_data(nlh);
5222
5223         /* only requests using strict checks can pass data to influence
5224          * the dump. The legacy exception is filter_mask.
5225          */
5226         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5227                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5228                 return -EINVAL;
5229         }
5230         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5231                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5232                 return -EINVAL;
5233         }
5234         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5235                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5236                 return -EINVAL;
5237         }
5238
5239         return 0;
5240 }
5241
5242 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5243                           struct netlink_ext_ack *extack)
5244 {
5245         struct net *net = sock_net(skb->sk);
5246         struct net_device *dev = NULL;
5247         int idxattr = 0, prividx = 0;
5248         struct if_stats_msg *ifsm;
5249         struct sk_buff *nskb;
5250         u32 filter_mask;
5251         int err;
5252
5253         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5254                                    false, extack);
5255         if (err)
5256                 return err;
5257
5258         ifsm = nlmsg_data(nlh);
5259         if (ifsm->ifindex > 0)
5260                 dev = __dev_get_by_index(net, ifsm->ifindex);
5261         else
5262                 return -EINVAL;
5263
5264         if (!dev)
5265                 return -ENODEV;
5266
5267         filter_mask = ifsm->filter_mask;
5268         if (!filter_mask)
5269                 return -EINVAL;
5270
5271         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5272         if (!nskb)
5273                 return -ENOBUFS;
5274
5275         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5276                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5277                                   0, filter_mask, &idxattr, &prividx);
5278         if (err < 0) {
5279                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5280                 WARN_ON(err == -EMSGSIZE);
5281                 kfree_skb(nskb);
5282         } else {
5283                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5284         }
5285
5286         return err;
5287 }
5288
5289 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5290 {
5291         struct netlink_ext_ack *extack = cb->extack;
5292         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5293         struct net *net = sock_net(skb->sk);
5294         unsigned int flags = NLM_F_MULTI;
5295         struct if_stats_msg *ifsm;
5296         struct hlist_head *head;
5297         struct net_device *dev;
5298         u32 filter_mask = 0;
5299         int idx = 0;
5300
5301         s_h = cb->args[0];
5302         s_idx = cb->args[1];
5303         s_idxattr = cb->args[2];
5304         s_prividx = cb->args[3];
5305
5306         cb->seq = net->dev_base_seq;
5307
5308         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5309         if (err)
5310                 return err;
5311
5312         ifsm = nlmsg_data(cb->nlh);
5313         filter_mask = ifsm->filter_mask;
5314         if (!filter_mask) {
5315                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5316                 return -EINVAL;
5317         }
5318
5319         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5320                 idx = 0;
5321                 head = &net->dev_index_head[h];
5322                 hlist_for_each_entry(dev, head, index_hlist) {
5323                         if (idx < s_idx)
5324                                 goto cont;
5325                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5326                                                   NETLINK_CB(cb->skb).portid,
5327                                                   cb->nlh->nlmsg_seq, 0,
5328                                                   flags, filter_mask,
5329                                                   &s_idxattr, &s_prividx);
5330                         /* If we ran out of room on the first message,
5331                          * we're in trouble
5332                          */
5333                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5334
5335                         if (err < 0)
5336                                 goto out;
5337                         s_prividx = 0;
5338                         s_idxattr = 0;
5339                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5340 cont:
5341                         idx++;
5342                 }
5343         }
5344 out:
5345         cb->args[3] = s_prividx;
5346         cb->args[2] = s_idxattr;
5347         cb->args[1] = idx;
5348         cb->args[0] = h;
5349
5350         return skb->len;
5351 }
5352
5353 /* Process one rtnetlink message. */
5354
5355 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5356                              struct netlink_ext_ack *extack)
5357 {
5358         struct net *net = sock_net(skb->sk);
5359         struct rtnl_link *link;
5360         struct module *owner;
5361         int err = -EOPNOTSUPP;
5362         rtnl_doit_func doit;
5363         unsigned int flags;
5364         int kind;
5365         int family;
5366         int type;
5367
5368         type = nlh->nlmsg_type;
5369         if (type > RTM_MAX)
5370                 return -EOPNOTSUPP;
5371
5372         type -= RTM_BASE;
5373
5374         /* All the messages must have at least 1 byte length */
5375         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5376                 return 0;
5377
5378         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5379         kind = type&3;
5380
5381         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5382                 return -EPERM;
5383
5384         rcu_read_lock();
5385         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5386                 struct sock *rtnl;
5387                 rtnl_dumpit_func dumpit;
5388                 u16 min_dump_alloc = 0;
5389
5390                 link = rtnl_get_link(family, type);
5391                 if (!link || !link->dumpit) {
5392                         family = PF_UNSPEC;
5393                         link = rtnl_get_link(family, type);
5394                         if (!link || !link->dumpit)
5395                                 goto err_unlock;
5396                 }
5397                 owner = link->owner;
5398                 dumpit = link->dumpit;
5399
5400                 if (type == RTM_GETLINK - RTM_BASE)
5401                         min_dump_alloc = rtnl_calcit(skb, nlh);
5402
5403                 err = 0;
5404                 /* need to do this before rcu_read_unlock() */
5405                 if (!try_module_get(owner))
5406                         err = -EPROTONOSUPPORT;
5407
5408                 rcu_read_unlock();
5409
5410                 rtnl = net->rtnl;
5411                 if (err == 0) {
5412                         struct netlink_dump_control c = {
5413                                 .dump           = dumpit,
5414                                 .min_dump_alloc = min_dump_alloc,
5415                                 .module         = owner,
5416                         };
5417                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5418                         /* netlink_dump_start() will keep a reference on
5419                          * module if dump is still in progress.
5420                          */
5421                         module_put(owner);
5422                 }
5423                 return err;
5424         }
5425
5426         link = rtnl_get_link(family, type);
5427         if (!link || !link->doit) {
5428                 family = PF_UNSPEC;
5429                 link = rtnl_get_link(PF_UNSPEC, type);
5430                 if (!link || !link->doit)
5431                         goto out_unlock;
5432         }
5433
5434         owner = link->owner;
5435         if (!try_module_get(owner)) {
5436                 err = -EPROTONOSUPPORT;
5437                 goto out_unlock;
5438         }
5439
5440         flags = link->flags;
5441         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5442                 doit = link->doit;
5443                 rcu_read_unlock();
5444                 if (doit)
5445                         err = doit(skb, nlh, extack);
5446                 module_put(owner);
5447                 return err;
5448         }
5449         rcu_read_unlock();
5450
5451         rtnl_lock();
5452         link = rtnl_get_link(family, type);
5453         if (link && link->doit)
5454                 err = link->doit(skb, nlh, extack);
5455         rtnl_unlock();
5456
5457         module_put(owner);
5458
5459         return err;
5460
5461 out_unlock:
5462         rcu_read_unlock();
5463         return err;
5464
5465 err_unlock:
5466         rcu_read_unlock();
5467         return -EOPNOTSUPP;
5468 }
5469
5470 static void rtnetlink_rcv(struct sk_buff *skb)
5471 {
5472         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5473 }
5474
5475 static int rtnetlink_bind(struct net *net, int group)
5476 {
5477         switch (group) {
5478         case RTNLGRP_IPV4_MROUTE_R:
5479         case RTNLGRP_IPV6_MROUTE_R:
5480                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5481                         return -EPERM;
5482                 break;
5483         }
5484         return 0;
5485 }
5486
5487 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5488 {
5489         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5490
5491         switch (event) {
5492         case NETDEV_REBOOT:
5493         case NETDEV_CHANGEMTU:
5494         case NETDEV_CHANGEADDR:
5495         case NETDEV_CHANGENAME:
5496         case NETDEV_FEAT_CHANGE:
5497         case NETDEV_BONDING_FAILOVER:
5498         case NETDEV_POST_TYPE_CHANGE:
5499         case NETDEV_NOTIFY_PEERS:
5500         case NETDEV_CHANGEUPPER:
5501         case NETDEV_RESEND_IGMP:
5502         case NETDEV_CHANGEINFODATA:
5503         case NETDEV_CHANGELOWERSTATE:
5504         case NETDEV_CHANGE_TX_QUEUE_LEN:
5505                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5506                                    GFP_KERNEL, NULL, 0);
5507                 break;
5508         default:
5509                 break;
5510         }
5511         return NOTIFY_DONE;
5512 }
5513
5514 static struct notifier_block rtnetlink_dev_notifier = {
5515         .notifier_call  = rtnetlink_event,
5516 };
5517
5518
5519 static int __net_init rtnetlink_net_init(struct net *net)
5520 {
5521         struct sock *sk;
5522         struct netlink_kernel_cfg cfg = {
5523                 .groups         = RTNLGRP_MAX,
5524                 .input          = rtnetlink_rcv,
5525                 .cb_mutex       = &rtnl_mutex,
5526                 .flags          = NL_CFG_F_NONROOT_RECV,
5527                 .bind           = rtnetlink_bind,
5528         };
5529
5530         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5531         if (!sk)
5532                 return -ENOMEM;
5533         net->rtnl = sk;
5534         return 0;
5535 }
5536
5537 static void __net_exit rtnetlink_net_exit(struct net *net)
5538 {
5539         netlink_kernel_release(net->rtnl);
5540         net->rtnl = NULL;
5541 }
5542
5543 static struct pernet_operations rtnetlink_net_ops = {
5544         .init = rtnetlink_net_init,
5545         .exit = rtnetlink_net_exit,
5546 };
5547
5548 void __init rtnetlink_init(void)
5549 {
5550         if (register_pernet_subsys(&rtnetlink_net_ops))
5551                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5552
5553         register_netdevice_notifier(&rtnetlink_dev_notifier);
5554
5555         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5556                       rtnl_dump_ifinfo, 0);
5557         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5558         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5559         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5560
5561         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5562         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5563         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5564
5565         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5566         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5567
5568         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5569         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5570         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5571
5572         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5573         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5574         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5575
5576         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5577                       0);
5578 }