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 arithmetic 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 #include "dev.h"
  58
  59 #define RTNL_MAX_TYPE           50
  60 #define RTNL_SLAVE_MAX_TYPE     40
  61
  62 struct rtnl_link {
  63         rtnl_doit_func          doit;
  64         rtnl_dumpit_func        dumpit;
  65         struct module           *owner;
  66         unsigned int            flags;
  67         struct rcu_head         rcu;
  68 };
  69
  70 static DEFINE_MUTEX(rtnl_mutex);
  71
  72 void rtnl_lock(void)
  73 {
  74         mutex_lock(&rtnl_mutex);
  75 }
  76 EXPORT_SYMBOL(rtnl_lock);
  77
  78 int rtnl_lock_killable(void)
  79 {
  80         return mutex_lock_killable(&rtnl_mutex);
  81 }
  82 EXPORT_SYMBOL(rtnl_lock_killable);
  83
  84 static struct sk_buff *defer_kfree_skb_list;
  85 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  86 {
  87         if (head && tail) {
  88                 tail->next = defer_kfree_skb_list;
  89                 defer_kfree_skb_list = head;
  90         }
  91 }
  92 EXPORT_SYMBOL(rtnl_kfree_skbs);
  93
  94 void __rtnl_unlock(void)
  95 {
  96         struct sk_buff *head = defer_kfree_skb_list;
  97
  98         defer_kfree_skb_list = NULL;
  99
 100         /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
 101          * is used. In some places, e.g. in cfg80211, we have code that will do
 102          * something like
 103          *   rtnl_lock()
 104          *   wiphy_lock()
 105          *   ...
 106          *   rtnl_unlock()
 107          *
 108          * and because netdev_run_todo() acquires the RTNL for items on the list
 109          * we could cause a situation such as this:
 110          * Thread 1                     Thread 2
 111          *                                rtnl_lock()
 112          *                                unregister_netdevice()
 113          *                                __rtnl_unlock()
 114          * rtnl_lock()
 115          * wiphy_lock()
 116          * rtnl_unlock()
 117          *   netdev_run_todo()
 118          *     __rtnl_unlock()
 119          *
 120          *     // list not empty now
 121          *     // because of thread 2
 122          *                                rtnl_lock()
 123          *     while (!list_empty(...))
 124          *       rtnl_lock()
 125          *                                wiphy_lock()
 126          * **** DEADLOCK ****
 127          *
 128          * However, usage of __rtnl_unlock() is rare, and so we can ensure that
 129          * it's not used in cases where something is added to do the list.
 130          */
 131         WARN_ON(!list_empty(&net_todo_list));
 132
 133         mutex_unlock(&rtnl_mutex);
 134
 135         while (head) {
 136                 struct sk_buff *next = head->next;
 137
 138                 kfree_skb(head);
 139                 cond_resched();
 140                 head = next;
 141         }
 142 }
 143
 144 void rtnl_unlock(void)
 145 {
 146         /* This fellow will unlock it for us. */
 147         netdev_run_todo();
 148 }
 149 EXPORT_SYMBOL(rtnl_unlock);
 150
 151 int rtnl_trylock(void)
 152 {
 153         return mutex_trylock(&rtnl_mutex);
 154 }
 155 EXPORT_SYMBOL(rtnl_trylock);
 156
 157 int rtnl_is_locked(void)
 158 {
 159         return mutex_is_locked(&rtnl_mutex);
 160 }
 161 EXPORT_SYMBOL(rtnl_is_locked);
 162
 163 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 164 {
 165         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 166 }
 167 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 168
 169 #ifdef CONFIG_PROVE_LOCKING
 170 bool lockdep_rtnl_is_held(void)
 171 {
 172         return lockdep_is_held(&rtnl_mutex);
 173 }
 174 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 175 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 176
 177 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 178
 179 static inline int rtm_msgindex(int msgtype)
 180 {
 181         int msgindex = msgtype - RTM_BASE;
 182
 183         /*
 184          * msgindex < 0 implies someone tried to register a netlink
 185          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 186          * the message type has not been added to linux/rtnetlink.h
 187          */
 188         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 189
 190         return msgindex;
 191 }
 192
 193 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 194 {
 195         struct rtnl_link __rcu **tab;
 196
 197         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 198                 protocol = PF_UNSPEC;
 199
 200         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 201         if (!tab)
 202                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 203
 204         return rcu_dereference_rtnl(tab[msgtype]);
 205 }
 206
 207 static int rtnl_register_internal(struct module *owner,
 208                                   int protocol, int msgtype,
 209                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 210                                   unsigned int flags)
 211 {
 212         struct rtnl_link *link, *old;
 213         struct rtnl_link __rcu **tab;
 214         int msgindex;
 215         int ret = -ENOBUFS;
 216
 217         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 218         msgindex = rtm_msgindex(msgtype);
 219
 220         rtnl_lock();
 221         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 222         if (tab == NULL) {
 223                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 224                 if (!tab)
 225                         goto unlock;
 226
 227                 /* ensures we see the 0 stores */
 228                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 229         }
 230
 231         old = rtnl_dereference(tab[msgindex]);
 232         if (old) {
 233                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 234                 if (!link)
 235                         goto unlock;
 236         } else {
 237                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 238                 if (!link)
 239                         goto unlock;
 240         }
 241
 242         WARN_ON(link->owner && link->owner != owner);
 243         link->owner = owner;
 244
 245         WARN_ON(doit && link->doit && link->doit != doit);
 246         if (doit)
 247                 link->doit = doit;
 248         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 249         if (dumpit)
 250                 link->dumpit = dumpit;
 251
 252         WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
 253                 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
 254         link->flags |= flags;
 255
 256         /* publish protocol:msgtype */
 257         rcu_assign_pointer(tab[msgindex], link);
 258         ret = 0;
 259         if (old)
 260                 kfree_rcu(old, rcu);
 261 unlock:
 262         rtnl_unlock();
 263         return ret;
 264 }
 265
 266 /**
 267  * rtnl_register_module - Register a rtnetlink message type
 268  *
 269  * @owner: module registering the hook (THIS_MODULE)
 270  * @protocol: Protocol family or PF_UNSPEC
 271  * @msgtype: rtnetlink message type
 272  * @doit: Function pointer called for each request message
 273  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 274  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 275  *
 276  * Like rtnl_register, but for use by removable modules.
 277  */
 278 int rtnl_register_module(struct module *owner,
 279                          int protocol, int msgtype,
 280                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 281                          unsigned int flags)
 282 {
 283         return rtnl_register_internal(owner, protocol, msgtype,
 284                                       doit, dumpit, flags);
 285 }
 286 EXPORT_SYMBOL_GPL(rtnl_register_module);
 287
 288 /**
 289  * rtnl_register - Register a rtnetlink message type
 290  * @protocol: Protocol family or PF_UNSPEC
 291  * @msgtype: rtnetlink message type
 292  * @doit: Function pointer called for each request message
 293  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 294  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 295  *
 296  * Registers the specified function pointers (at least one of them has
 297  * to be non-NULL) to be called whenever a request message for the
 298  * specified protocol family and message type is received.
 299  *
 300  * The special protocol family PF_UNSPEC may be used to define fallback
 301  * function pointers for the case when no entry for the specific protocol
 302  * family exists.
 303  */
 304 void rtnl_register(int protocol, int msgtype,
 305                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 306                    unsigned int flags)
 307 {
 308         int err;
 309
 310         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 311                                      flags);
 312         if (err)
 313                 pr_err("Unable to register rtnetlink message handler, "
 314                        "protocol = %d, message type = %d\n", protocol, msgtype);
 315 }
 316
 317 /**
 318  * rtnl_unregister - Unregister a rtnetlink message type
 319  * @protocol: Protocol family or PF_UNSPEC
 320  * @msgtype: rtnetlink message type
 321  *
 322  * Returns 0 on success or a negative error code.
 323  */
 324 int rtnl_unregister(int protocol, int msgtype)
 325 {
 326         struct rtnl_link __rcu **tab;
 327         struct rtnl_link *link;
 328         int msgindex;
 329
 330         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 331         msgindex = rtm_msgindex(msgtype);
 332
 333         rtnl_lock();
 334         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 335         if (!tab) {
 336                 rtnl_unlock();
 337                 return -ENOENT;
 338         }
 339
 340         link = rtnl_dereference(tab[msgindex]);
 341         RCU_INIT_POINTER(tab[msgindex], NULL);
 342         rtnl_unlock();
 343
 344         kfree_rcu(link, rcu);
 345
 346         return 0;
 347 }
 348 EXPORT_SYMBOL_GPL(rtnl_unregister);
 349
 350 /**
 351  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 352  * @protocol : Protocol family or PF_UNSPEC
 353  *
 354  * Identical to calling rtnl_unregster() for all registered message types
 355  * of a certain protocol family.
 356  */
 357 void rtnl_unregister_all(int protocol)
 358 {
 359         struct rtnl_link __rcu **tab;
 360         struct rtnl_link *link;
 361         int msgindex;
 362
 363         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 364
 365         rtnl_lock();
 366         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 367         if (!tab) {
 368                 rtnl_unlock();
 369                 return;
 370         }
 371         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 372         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 373                 link = rtnl_dereference(tab[msgindex]);
 374                 if (!link)
 375                         continue;
 376
 377                 RCU_INIT_POINTER(tab[msgindex], NULL);
 378                 kfree_rcu(link, rcu);
 379         }
 380         rtnl_unlock();
 381
 382         synchronize_net();
 383
 384         kfree(tab);
 385 }
 386 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 387
 388 static LIST_HEAD(link_ops);
 389
 390 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 391 {
 392         const struct rtnl_link_ops *ops;
 393
 394         list_for_each_entry(ops, &link_ops, list) {
 395                 if (!strcmp(ops->kind, kind))
 396                         return ops;
 397         }
 398         return NULL;
 399 }
 400
 401 /**
 402  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 403  * @ops: struct rtnl_link_ops * to register
 404  *
 405  * The caller must hold the rtnl_mutex. This function should be used
 406  * by drivers that create devices during module initialization. It
 407  * must be called before registering the devices.
 408  *
 409  * Returns 0 on success or a negative error code.
 410  */
 411 int __rtnl_link_register(struct rtnl_link_ops *ops)
 412 {
 413         if (rtnl_link_ops_get(ops->kind))
 414                 return -EEXIST;
 415
 416         /* The check for alloc/setup is here because if ops
 417          * does not have that filled up, it is not possible
 418          * to use the ops for creating device. So do not
 419          * fill up dellink as well. That disables rtnl_dellink.
 420          */
 421         if ((ops->alloc || ops->setup) && !ops->dellink)
 422                 ops->dellink = unregister_netdevice_queue;
 423
 424         list_add_tail(&ops->list, &link_ops);
 425         return 0;
 426 }
 427 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 428
 429 /**
 430  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 431  * @ops: struct rtnl_link_ops * to register
 432  *
 433  * Returns 0 on success or a negative error code.
 434  */
 435 int rtnl_link_register(struct rtnl_link_ops *ops)
 436 {
 437         int err;
 438
 439         /* Sanity-check max sizes to avoid stack buffer overflow. */
 440         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 441                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 442                 return -EINVAL;
 443
 444         rtnl_lock();
 445         err = __rtnl_link_register(ops);
 446         rtnl_unlock();
 447         return err;
 448 }
 449 EXPORT_SYMBOL_GPL(rtnl_link_register);
 450
 451 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 452 {
 453         struct net_device *dev;
 454         LIST_HEAD(list_kill);
 455
 456         for_each_netdev(net, dev) {
 457                 if (dev->rtnl_link_ops == ops)
 458                         ops->dellink(dev, &list_kill);
 459         }
 460         unregister_netdevice_many(&list_kill);
 461 }
 462
 463 /**
 464  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 465  * @ops: struct rtnl_link_ops * to unregister
 466  *
 467  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 468  * integrity (hold pernet_ops_rwsem for writing to close the race
 469  * with setup_net() and cleanup_net()).
 470  */
 471 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 472 {
 473         struct net *net;
 474
 475         for_each_net(net) {
 476                 __rtnl_kill_links(net, ops);
 477         }
 478         list_del(&ops->list);
 479 }
 480 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 481
 482 /* Return with the rtnl_lock held when there are no network
 483  * devices unregistering in any network namespace.
 484  */
 485 static void rtnl_lock_unregistering_all(void)
 486 {
 487         struct net *net;
 488         bool unregistering;
 489         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 490
 491         add_wait_queue(&netdev_unregistering_wq, &wait);
 492         for (;;) {
 493                 unregistering = false;
 494                 rtnl_lock();
 495                 /* We held write locked pernet_ops_rwsem, and parallel
 496                  * setup_net() and cleanup_net() are not possible.
 497                  */
 498                 for_each_net(net) {
 499                         if (atomic_read(&net->dev_unreg_count) > 0) {
 500                                 unregistering = true;
 501                                 break;
 502                         }
 503                 }
 504                 if (!unregistering)
 505                         break;
 506                 __rtnl_unlock();
 507
 508                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 509         }
 510         remove_wait_queue(&netdev_unregistering_wq, &wait);
 511 }
 512
 513 /**
 514  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 515  * @ops: struct rtnl_link_ops * to unregister
 516  */
 517 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 518 {
 519         /* Close the race with setup_net() and cleanup_net() */
 520         down_write(&pernet_ops_rwsem);
 521         rtnl_lock_unregistering_all();
 522         __rtnl_link_unregister(ops);
 523         rtnl_unlock();
 524         up_write(&pernet_ops_rwsem);
 525 }
 526 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 527
 528 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 529 {
 530         struct net_device *master_dev;
 531         const struct rtnl_link_ops *ops;
 532         size_t size = 0;
 533
 534         rcu_read_lock();
 535
 536         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 537         if (!master_dev)
 538                 goto out;
 539
 540         ops = master_dev->rtnl_link_ops;
 541         if (!ops || !ops->get_slave_size)
 542                 goto out;
 543         /* IFLA_INFO_SLAVE_DATA + nested data */
 544         size = nla_total_size(sizeof(struct nlattr)) +
 545                ops->get_slave_size(master_dev, dev);
 546
 547 out:
 548         rcu_read_unlock();
 549         return size;
 550 }
 551
 552 static size_t rtnl_link_get_size(const struct net_device *dev)
 553 {
 554         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 555         size_t size;
 556
 557         if (!ops)
 558                 return 0;
 559
 560         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 561                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 562
 563         if (ops->get_size)
 564                 /* IFLA_INFO_DATA + nested data */
 565                 size += nla_total_size(sizeof(struct nlattr)) +
 566                         ops->get_size(dev);
 567
 568         if (ops->get_xstats_size)
 569                 /* IFLA_INFO_XSTATS */
 570                 size += nla_total_size(ops->get_xstats_size(dev));
 571
 572         size += rtnl_link_get_slave_info_data_size(dev);
 573
 574         return size;
 575 }
 576
 577 static LIST_HEAD(rtnl_af_ops);
 578
 579 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 580 {
 581         const struct rtnl_af_ops *ops;
 582
 583         ASSERT_RTNL();
 584
 585         list_for_each_entry(ops, &rtnl_af_ops, list) {
 586                 if (ops->family == family)
 587                         return ops;
 588         }
 589
 590         return NULL;
 591 }
 592
 593 /**
 594  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 595  * @ops: struct rtnl_af_ops * to register
 596  *
 597  * Returns 0 on success or a negative error code.
 598  */
 599 void rtnl_af_register(struct rtnl_af_ops *ops)
 600 {
 601         rtnl_lock();
 602         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 603         rtnl_unlock();
 604 }
 605 EXPORT_SYMBOL_GPL(rtnl_af_register);
 606
 607 /**
 608  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 609  * @ops: struct rtnl_af_ops * to unregister
 610  */
 611 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 612 {
 613         rtnl_lock();
 614         list_del_rcu(&ops->list);
 615         rtnl_unlock();
 616
 617         synchronize_rcu();
 618 }
 619 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 620
 621 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 622                                     u32 ext_filter_mask)
 623 {
 624         struct rtnl_af_ops *af_ops;
 625         size_t size;
 626
 627         /* IFLA_AF_SPEC */
 628         size = nla_total_size(sizeof(struct nlattr));
 629
 630         rcu_read_lock();
 631         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 632                 if (af_ops->get_link_af_size) {
 633                         /* AF_* + nested data */
 634                         size += nla_total_size(sizeof(struct nlattr)) +
 635                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 636                 }
 637         }
 638         rcu_read_unlock();
 639
 640         return size;
 641 }
 642
 643 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 644 {
 645         struct net_device *master_dev;
 646         bool ret = false;
 647
 648         rcu_read_lock();
 649
 650         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 651         if (master_dev && master_dev->rtnl_link_ops)
 652                 ret = true;
 653         rcu_read_unlock();
 654         return ret;
 655 }
 656
 657 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 658                                      const struct net_device *dev)
 659 {
 660         struct net_device *master_dev;
 661         const struct rtnl_link_ops *ops;
 662         struct nlattr *slave_data;
 663         int err;
 664
 665         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 666         if (!master_dev)
 667                 return 0;
 668         ops = master_dev->rtnl_link_ops;
 669         if (!ops)
 670                 return 0;
 671         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 672                 return -EMSGSIZE;
 673         if (ops->fill_slave_info) {
 674                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 675                 if (!slave_data)
 676                         return -EMSGSIZE;
 677                 err = ops->fill_slave_info(skb, master_dev, dev);
 678                 if (err < 0)
 679                         goto err_cancel_slave_data;
 680                 nla_nest_end(skb, slave_data);
 681         }
 682         return 0;
 683
 684 err_cancel_slave_data:
 685         nla_nest_cancel(skb, slave_data);
 686         return err;
 687 }
 688
 689 static int rtnl_link_info_fill(struct sk_buff *skb,
 690                                const struct net_device *dev)
 691 {
 692         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 693         struct nlattr *data;
 694         int err;
 695
 696         if (!ops)
 697                 return 0;
 698         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 699                 return -EMSGSIZE;
 700         if (ops->fill_xstats) {
 701                 err = ops->fill_xstats(skb, dev);
 702                 if (err < 0)
 703                         return err;
 704         }
 705         if (ops->fill_info) {
 706                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 707                 if (data == NULL)
 708                         return -EMSGSIZE;
 709                 err = ops->fill_info(skb, dev);
 710                 if (err < 0)
 711                         goto err_cancel_data;
 712                 nla_nest_end(skb, data);
 713         }
 714         return 0;
 715
 716 err_cancel_data:
 717         nla_nest_cancel(skb, data);
 718         return err;
 719 }
 720
 721 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 722 {
 723         struct nlattr *linkinfo;
 724         int err = -EMSGSIZE;
 725
 726         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 727         if (linkinfo == NULL)
 728                 goto out;
 729
 730         err = rtnl_link_info_fill(skb, dev);
 731         if (err < 0)
 732                 goto err_cancel_link;
 733
 734         err = rtnl_link_slave_info_fill(skb, dev);
 735         if (err < 0)
 736                 goto err_cancel_link;
 737
 738         nla_nest_end(skb, linkinfo);
 739         return 0;
 740
 741 err_cancel_link:
 742         nla_nest_cancel(skb, linkinfo);
 743 out:
 744         return err;
 745 }
 746
 747 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 748 {
 749         struct sock *rtnl = net->rtnl;
 750
 751         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 752 }
 753
 754 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 755 {
 756         struct sock *rtnl = net->rtnl;
 757
 758         return nlmsg_unicast(rtnl, skb, pid);
 759 }
 760 EXPORT_SYMBOL(rtnl_unicast);
 761
 762 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 763                  struct nlmsghdr *nlh, gfp_t flags)
 764 {
 765         struct sock *rtnl = net->rtnl;
 766
 767         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 768 }
 769 EXPORT_SYMBOL(rtnl_notify);
 770
 771 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 772 {
 773         struct sock *rtnl = net->rtnl;
 774
 775         netlink_set_err(rtnl, 0, group, error);
 776 }
 777 EXPORT_SYMBOL(rtnl_set_sk_err);
 778
 779 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 780 {
 781         struct nlattr *mx;
 782         int i, valid = 0;
 783
 784         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 785         if (metrics == dst_default_metrics.metrics)
 786                 return 0;
 787
 788         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 789         if (mx == NULL)
 790                 return -ENOBUFS;
 791
 792         for (i = 0; i < RTAX_MAX; i++) {
 793                 if (metrics[i]) {
 794                         if (i == RTAX_CC_ALGO - 1) {
 795                                 char tmp[TCP_CA_NAME_MAX], *name;
 796
 797                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 798                                 if (!name)
 799                                         continue;
 800                                 if (nla_put_string(skb, i + 1, name))
 801                                         goto nla_put_failure;
 802                         } else if (i == RTAX_FEATURES - 1) {
 803                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 804
 805                                 if (!user_features)
 806                                         continue;
 807                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 808                                 if (nla_put_u32(skb, i + 1, user_features))
 809                                         goto nla_put_failure;
 810                         } else {
 811                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 812                                         goto nla_put_failure;
 813                         }
 814                         valid++;
 815                 }
 816         }
 817
 818         if (!valid) {
 819                 nla_nest_cancel(skb, mx);
 820                 return 0;
 821         }
 822
 823         return nla_nest_end(skb, mx);
 824
 825 nla_put_failure:
 826         nla_nest_cancel(skb, mx);
 827         return -EMSGSIZE;
 828 }
 829 EXPORT_SYMBOL(rtnetlink_put_metrics);
 830
 831 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 832                        long expires, u32 error)
 833 {
 834         struct rta_cacheinfo ci = {
 835                 .rta_error = error,
 836                 .rta_id =  id,
 837         };
 838
 839         if (dst) {
 840                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 841                 ci.rta_used = dst->__use;
 842                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 843         }
 844         if (expires) {
 845                 unsigned long clock;
 846
 847                 clock = jiffies_to_clock_t(abs(expires));
 848                 clock = min_t(unsigned long, clock, INT_MAX);
 849                 ci.rta_expires = (expires > 0) ? clock : -clock;
 850         }
 851         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 852 }
 853 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 854
 855 static void set_operstate(struct net_device *dev, unsigned char transition)
 856 {
 857         unsigned char operstate = dev->operstate;
 858
 859         switch (transition) {
 860         case IF_OPER_UP:
 861                 if ((operstate == IF_OPER_DORMANT ||
 862                      operstate == IF_OPER_TESTING ||
 863                      operstate == IF_OPER_UNKNOWN) &&
 864                     !netif_dormant(dev) && !netif_testing(dev))
 865                         operstate = IF_OPER_UP;
 866                 break;
 867
 868         case IF_OPER_TESTING:
 869                 if (operstate == IF_OPER_UP ||
 870                     operstate == IF_OPER_UNKNOWN)
 871                         operstate = IF_OPER_TESTING;
 872                 break;
 873
 874         case IF_OPER_DORMANT:
 875                 if (operstate == IF_OPER_UP ||
 876                     operstate == IF_OPER_UNKNOWN)
 877                         operstate = IF_OPER_DORMANT;
 878                 break;
 879         }
 880
 881         if (dev->operstate != operstate) {
 882                 write_lock(&dev_base_lock);
 883                 dev->operstate = operstate;
 884                 write_unlock(&dev_base_lock);
 885                 netdev_state_change(dev);
 886         }
 887 }
 888
 889 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 890 {
 891         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 892                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 893 }
 894
 895 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 896                                            const struct ifinfomsg *ifm)
 897 {
 898         unsigned int flags = ifm->ifi_flags;
 899
 900         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 901         if (ifm->ifi_change)
 902                 flags = (flags & ifm->ifi_change) |
 903                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 904
 905         return flags;
 906 }
 907
 908 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 909                                  const struct rtnl_link_stats64 *b)
 910 {
 911         a->rx_packets = b->rx_packets;
 912         a->tx_packets = b->tx_packets;
 913         a->rx_bytes = b->rx_bytes;
 914         a->tx_bytes = b->tx_bytes;
 915         a->rx_errors = b->rx_errors;
 916         a->tx_errors = b->tx_errors;
 917         a->rx_dropped = b->rx_dropped;
 918         a->tx_dropped = b->tx_dropped;
 919
 920         a->multicast = b->multicast;
 921         a->collisions = b->collisions;
 922
 923         a->rx_length_errors = b->rx_length_errors;
 924         a->rx_over_errors = b->rx_over_errors;
 925         a->rx_crc_errors = b->rx_crc_errors;
 926         a->rx_frame_errors = b->rx_frame_errors;
 927         a->rx_fifo_errors = b->rx_fifo_errors;
 928         a->rx_missed_errors = b->rx_missed_errors;
 929
 930         a->tx_aborted_errors = b->tx_aborted_errors;
 931         a->tx_carrier_errors = b->tx_carrier_errors;
 932         a->tx_fifo_errors = b->tx_fifo_errors;
 933         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 934         a->tx_window_errors = b->tx_window_errors;
 935
 936         a->rx_compressed = b->rx_compressed;
 937         a->tx_compressed = b->tx_compressed;
 938
 939         a->rx_nohandler = b->rx_nohandler;
 940 }
 941
 942 /* All VF info */
 943 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 944                                    u32 ext_filter_mask)
 945 {
 946         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 947                 int num_vfs = dev_num_vf(dev->dev.parent);
 948                 size_t size = nla_total_size(0);
 949                 size += num_vfs *
 950                         (nla_total_size(0) +
 951                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 952                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 953                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 954                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 955                          nla_total_size(MAX_VLAN_LIST_LEN *
 956                                         sizeof(struct ifla_vf_vlan_info)) +
 957                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 958                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 959                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 960                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 961                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 962                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 963                          /* IFLA_VF_STATS_RX_PACKETS */
 964                          nla_total_size_64bit(sizeof(__u64)) +
 965                          /* IFLA_VF_STATS_TX_PACKETS */
 966                          nla_total_size_64bit(sizeof(__u64)) +
 967                          /* IFLA_VF_STATS_RX_BYTES */
 968                          nla_total_size_64bit(sizeof(__u64)) +
 969                          /* IFLA_VF_STATS_TX_BYTES */
 970                          nla_total_size_64bit(sizeof(__u64)) +
 971                          /* IFLA_VF_STATS_BROADCAST */
 972                          nla_total_size_64bit(sizeof(__u64)) +
 973                          /* IFLA_VF_STATS_MULTICAST */
 974                          nla_total_size_64bit(sizeof(__u64)) +
 975                          /* IFLA_VF_STATS_RX_DROPPED */
 976                          nla_total_size_64bit(sizeof(__u64)) +
 977                          /* IFLA_VF_STATS_TX_DROPPED */
 978                          nla_total_size_64bit(sizeof(__u64)) +
 979                          nla_total_size(sizeof(struct ifla_vf_trust)));
 980                 return size;
 981         } else
 982                 return 0;
 983 }
 984
 985 static size_t rtnl_port_size(const struct net_device *dev,
 986                              u32 ext_filter_mask)
 987 {
 988         size_t port_size = nla_total_size(4)            /* PORT_VF */
 989                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 990                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 991                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 992                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 993                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 994         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 995         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 996                 + port_size;
 997         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 998                 + port_size;
 999
1000         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1001             !(ext_filter_mask & RTEXT_FILTER_VF))
1002                 return 0;
1003         if (dev_num_vf(dev->dev.parent))
1004                 return port_self_size + vf_ports_size +
1005                         vf_port_size * dev_num_vf(dev->dev.parent);
1006         else
1007                 return port_self_size;
1008 }
1009
1010 static size_t rtnl_xdp_size(void)
1011 {
1012         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1013                           nla_total_size(1) +   /* XDP_ATTACHED */
1014                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1015                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1016
1017         return xdp_size;
1018 }
1019
1020 static size_t rtnl_prop_list_size(const struct net_device *dev)
1021 {
1022         struct netdev_name_node *name_node;
1023         size_t size;
1024
1025         if (list_empty(&dev->name_node->list))
1026                 return 0;
1027         size = nla_total_size(0);
1028         list_for_each_entry(name_node, &dev->name_node->list, list)
1029                 size += nla_total_size(ALTIFNAMSIZ);
1030         return size;
1031 }
1032
1033 static size_t rtnl_proto_down_size(const struct net_device *dev)
1034 {
1035         size_t size = nla_total_size(1);
1036
1037         if (dev->proto_down_reason)
1038                 size += nla_total_size(0) + nla_total_size(4);
1039
1040         return size;
1041 }
1042
1043 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1044                                      u32 ext_filter_mask)
1045 {
1046         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1047                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1048                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1049                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1050                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1051                + nla_total_size(sizeof(struct rtnl_link_stats))
1052                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1053                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1054                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1055                + nla_total_size(4) /* IFLA_TXQLEN */
1056                + nla_total_size(4) /* IFLA_WEIGHT */
1057                + nla_total_size(4) /* IFLA_MTU */
1058                + nla_total_size(4) /* IFLA_LINK */
1059                + nla_total_size(4) /* IFLA_MASTER */
1060                + nla_total_size(1) /* IFLA_CARRIER */
1061                + nla_total_size(4) /* IFLA_PROMISCUITY */
1062                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1063                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1064                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1065                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1066                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1067                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1068                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1069                + nla_total_size(1) /* IFLA_OPERSTATE */
1070                + nla_total_size(1) /* IFLA_LINKMODE */
1071                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1072                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1073                + nla_total_size(4) /* IFLA_GROUP */
1074                + nla_total_size(ext_filter_mask
1075                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1076                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1077                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1078                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1079                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1080                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1081                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1082                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1083                + rtnl_xdp_size() /* IFLA_XDP */
1084                + nla_total_size(4)  /* IFLA_EVENT */
1085                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1086                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1087                + rtnl_proto_down_size(dev)  /* proto down */
1088                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1089                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1090                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1091                + nla_total_size(4)  /* IFLA_MIN_MTU */
1092                + nla_total_size(4)  /* IFLA_MAX_MTU */
1093                + rtnl_prop_list_size(dev)
1094                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1095                + 0;
1096 }
1097
1098 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1099 {
1100         struct nlattr *vf_ports;
1101         struct nlattr *vf_port;
1102         int vf;
1103         int err;
1104
1105         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1106         if (!vf_ports)
1107                 return -EMSGSIZE;
1108
1109         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1110                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1111                 if (!vf_port)
1112                         goto nla_put_failure;
1113                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1114                         goto nla_put_failure;
1115                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1116                 if (err == -EMSGSIZE)
1117                         goto nla_put_failure;
1118                 if (err) {
1119                         nla_nest_cancel(skb, vf_port);
1120                         continue;
1121                 }
1122                 nla_nest_end(skb, vf_port);
1123         }
1124
1125         nla_nest_end(skb, vf_ports);
1126
1127         return 0;
1128
1129 nla_put_failure:
1130         nla_nest_cancel(skb, vf_ports);
1131         return -EMSGSIZE;
1132 }
1133
1134 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1135 {
1136         struct nlattr *port_self;
1137         int err;
1138
1139         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1140         if (!port_self)
1141                 return -EMSGSIZE;
1142
1143         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1144         if (err) {
1145                 nla_nest_cancel(skb, port_self);
1146                 return (err == -EMSGSIZE) ? err : 0;
1147         }
1148
1149         nla_nest_end(skb, port_self);
1150
1151         return 0;
1152 }
1153
1154 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1155                           u32 ext_filter_mask)
1156 {
1157         int err;
1158
1159         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1160             !(ext_filter_mask & RTEXT_FILTER_VF))
1161                 return 0;
1162
1163         err = rtnl_port_self_fill(skb, dev);
1164         if (err)
1165                 return err;
1166
1167         if (dev_num_vf(dev->dev.parent)) {
1168                 err = rtnl_vf_ports_fill(skb, dev);
1169                 if (err)
1170                         return err;
1171         }
1172
1173         return 0;
1174 }
1175
1176 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1177 {
1178         int err;
1179         struct netdev_phys_item_id ppid;
1180
1181         err = dev_get_phys_port_id(dev, &ppid);
1182         if (err) {
1183                 if (err == -EOPNOTSUPP)
1184                         return 0;
1185                 return err;
1186         }
1187
1188         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1189                 return -EMSGSIZE;
1190
1191         return 0;
1192 }
1193
1194 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1195 {
1196         char name[IFNAMSIZ];
1197         int err;
1198
1199         err = dev_get_phys_port_name(dev, name, sizeof(name));
1200         if (err) {
1201                 if (err == -EOPNOTSUPP)
1202                         return 0;
1203                 return err;
1204         }
1205
1206         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1207                 return -EMSGSIZE;
1208
1209         return 0;
1210 }
1211
1212 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1213 {
1214         struct netdev_phys_item_id ppid = { };
1215         int err;
1216
1217         err = dev_get_port_parent_id(dev, &ppid, false);
1218         if (err) {
1219                 if (err == -EOPNOTSUPP)
1220                         return 0;
1221                 return err;
1222         }
1223
1224         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1225                 return -EMSGSIZE;
1226
1227         return 0;
1228 }
1229
1230 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1231                                               struct net_device *dev)
1232 {
1233         struct rtnl_link_stats64 *sp;
1234         struct nlattr *attr;
1235
1236         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1237                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1238         if (!attr)
1239                 return -EMSGSIZE;
1240
1241         sp = nla_data(attr);
1242         dev_get_stats(dev, sp);
1243
1244         attr = nla_reserve(skb, IFLA_STATS,
1245                            sizeof(struct rtnl_link_stats));
1246         if (!attr)
1247                 return -EMSGSIZE;
1248
1249         copy_rtnl_link_stats(nla_data(attr), sp);
1250
1251         return 0;
1252 }
1253
1254 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1255                                                struct net_device *dev,
1256                                                int vfs_num,
1257                                                struct nlattr *vfinfo)
1258 {
1259         struct ifla_vf_rss_query_en vf_rss_query_en;
1260         struct nlattr *vf, *vfstats, *vfvlanlist;
1261         struct ifla_vf_link_state vf_linkstate;
1262         struct ifla_vf_vlan_info vf_vlan_info;
1263         struct ifla_vf_spoofchk vf_spoofchk;
1264         struct ifla_vf_tx_rate vf_tx_rate;
1265         struct ifla_vf_stats vf_stats;
1266         struct ifla_vf_trust vf_trust;
1267         struct ifla_vf_vlan vf_vlan;
1268         struct ifla_vf_rate vf_rate;
1269         struct ifla_vf_mac vf_mac;
1270         struct ifla_vf_broadcast vf_broadcast;
1271         struct ifla_vf_info ivi;
1272         struct ifla_vf_guid node_guid;
1273         struct ifla_vf_guid port_guid;
1274
1275         memset(&ivi, 0, sizeof(ivi));
1276
1277         /* Not all SR-IOV capable drivers support the
1278          * spoofcheck and "RSS query enable" query.  Preset to
1279          * -1 so the user space tool can detect that the driver
1280          * didn't report anything.
1281          */
1282         ivi.spoofchk = -1;
1283         ivi.rss_query_en = -1;
1284         ivi.trusted = -1;
1285         /* The default value for VF link state is "auto"
1286          * IFLA_VF_LINK_STATE_AUTO which equals zero
1287          */
1288         ivi.linkstate = 0;
1289         /* VLAN Protocol by default is 802.1Q */
1290         ivi.vlan_proto = htons(ETH_P_8021Q);
1291         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1292                 return 0;
1293
1294         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1295         memset(&node_guid, 0, sizeof(node_guid));
1296         memset(&port_guid, 0, sizeof(port_guid));
1297
1298         vf_mac.vf =
1299                 vf_vlan.vf =
1300                 vf_vlan_info.vf =
1301                 vf_rate.vf =
1302                 vf_tx_rate.vf =
1303                 vf_spoofchk.vf =
1304                 vf_linkstate.vf =
1305                 vf_rss_query_en.vf =
1306                 vf_trust.vf =
1307                 node_guid.vf =
1308                 port_guid.vf = ivi.vf;
1309
1310         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1311         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1312         vf_vlan.vlan = ivi.vlan;
1313         vf_vlan.qos = ivi.qos;
1314         vf_vlan_info.vlan = ivi.vlan;
1315         vf_vlan_info.qos = ivi.qos;
1316         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1317         vf_tx_rate.rate = ivi.max_tx_rate;
1318         vf_rate.min_tx_rate = ivi.min_tx_rate;
1319         vf_rate.max_tx_rate = ivi.max_tx_rate;
1320         vf_spoofchk.setting = ivi.spoofchk;
1321         vf_linkstate.link_state = ivi.linkstate;
1322         vf_rss_query_en.setting = ivi.rss_query_en;
1323         vf_trust.setting = ivi.trusted;
1324         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1325         if (!vf)
1326                 goto nla_put_vfinfo_failure;
1327         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1328             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1329             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1330             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1331                     &vf_rate) ||
1332             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1333                     &vf_tx_rate) ||
1334             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1335                     &vf_spoofchk) ||
1336             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1337                     &vf_linkstate) ||
1338             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1339                     sizeof(vf_rss_query_en),
1340                     &vf_rss_query_en) ||
1341             nla_put(skb, IFLA_VF_TRUST,
1342                     sizeof(vf_trust), &vf_trust))
1343                 goto nla_put_vf_failure;
1344
1345         if (dev->netdev_ops->ndo_get_vf_guid &&
1346             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1347                                               &port_guid)) {
1348                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1349                             &node_guid) ||
1350                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1351                             &port_guid))
1352                         goto nla_put_vf_failure;
1353         }
1354         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1355         if (!vfvlanlist)
1356                 goto nla_put_vf_failure;
1357         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1358                     &vf_vlan_info)) {
1359                 nla_nest_cancel(skb, vfvlanlist);
1360                 goto nla_put_vf_failure;
1361         }
1362         nla_nest_end(skb, vfvlanlist);
1363         memset(&vf_stats, 0, sizeof(vf_stats));
1364         if (dev->netdev_ops->ndo_get_vf_stats)
1365                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1366                                                 &vf_stats);
1367         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1368         if (!vfstats)
1369                 goto nla_put_vf_failure;
1370         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1371                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1372             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1373                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1374             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1375                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1376             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1377                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1378             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1379                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1380             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1381                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1382             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1383                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1384             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1385                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1386                 nla_nest_cancel(skb, vfstats);
1387                 goto nla_put_vf_failure;
1388         }
1389         nla_nest_end(skb, vfstats);
1390         nla_nest_end(skb, vf);
1391         return 0;
1392
1393 nla_put_vf_failure:
1394         nla_nest_cancel(skb, vf);
1395 nla_put_vfinfo_failure:
1396         nla_nest_cancel(skb, vfinfo);
1397         return -EMSGSIZE;
1398 }
1399
1400 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1401                                            struct net_device *dev,
1402                                            u32 ext_filter_mask)
1403 {
1404         struct nlattr *vfinfo;
1405         int i, num_vfs;
1406
1407         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1408                 return 0;
1409
1410         num_vfs = dev_num_vf(dev->dev.parent);
1411         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1412                 return -EMSGSIZE;
1413
1414         if (!dev->netdev_ops->ndo_get_vf_config)
1415                 return 0;
1416
1417         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1418         if (!vfinfo)
1419                 return -EMSGSIZE;
1420
1421         for (i = 0; i < num_vfs; i++) {
1422                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1423                         return -EMSGSIZE;
1424         }
1425
1426         nla_nest_end(skb, vfinfo);
1427         return 0;
1428 }
1429
1430 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1431 {
1432         struct rtnl_link_ifmap map;
1433
1434         memset(&map, 0, sizeof(map));
1435         map.mem_start   = dev->mem_start;
1436         map.mem_end     = dev->mem_end;
1437         map.base_addr   = dev->base_addr;
1438         map.irq         = dev->irq;
1439         map.dma         = dev->dma;
1440         map.port        = dev->if_port;
1441
1442         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1443                 return -EMSGSIZE;
1444
1445         return 0;
1446 }
1447
1448 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1449 {
1450         const struct bpf_prog *generic_xdp_prog;
1451
1452         ASSERT_RTNL();
1453
1454         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1455         if (!generic_xdp_prog)
1456                 return 0;
1457         return generic_xdp_prog->aux->id;
1458 }
1459
1460 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1461 {
1462         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1463 }
1464
1465 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1466 {
1467         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1468 }
1469
1470 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1471                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1472                                u32 (*get_prog_id)(struct net_device *dev))
1473 {
1474         u32 curr_id;
1475         int err;
1476
1477         curr_id = get_prog_id(dev);
1478         if (!curr_id)
1479                 return 0;
1480
1481         *prog_id = curr_id;
1482         err = nla_put_u32(skb, attr, curr_id);
1483         if (err)
1484                 return err;
1485
1486         if (*mode != XDP_ATTACHED_NONE)
1487                 *mode = XDP_ATTACHED_MULTI;
1488         else
1489                 *mode = tgt_mode;
1490
1491         return 0;
1492 }
1493
1494 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1495 {
1496         struct nlattr *xdp;
1497         u32 prog_id;
1498         int err;
1499         u8 mode;
1500
1501         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1502         if (!xdp)
1503                 return -EMSGSIZE;
1504
1505         prog_id = 0;
1506         mode = XDP_ATTACHED_NONE;
1507         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1508                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1509         if (err)
1510                 goto err_cancel;
1511         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1512                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1513         if (err)
1514                 goto err_cancel;
1515         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1516                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1517         if (err)
1518                 goto err_cancel;
1519
1520         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1521         if (err)
1522                 goto err_cancel;
1523
1524         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1525                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1526                 if (err)
1527                         goto err_cancel;
1528         }
1529
1530         nla_nest_end(skb, xdp);
1531         return 0;
1532
1533 err_cancel:
1534         nla_nest_cancel(skb, xdp);
1535         return err;
1536 }
1537
1538 static u32 rtnl_get_event(unsigned long event)
1539 {
1540         u32 rtnl_event_type = IFLA_EVENT_NONE;
1541
1542         switch (event) {
1543         case NETDEV_REBOOT:
1544                 rtnl_event_type = IFLA_EVENT_REBOOT;
1545                 break;
1546         case NETDEV_FEAT_CHANGE:
1547                 rtnl_event_type = IFLA_EVENT_FEATURES;
1548                 break;
1549         case NETDEV_BONDING_FAILOVER:
1550                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1551                 break;
1552         case NETDEV_NOTIFY_PEERS:
1553                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1554                 break;
1555         case NETDEV_RESEND_IGMP:
1556                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1557                 break;
1558         case NETDEV_CHANGEINFODATA:
1559                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1560                 break;
1561         default:
1562                 break;
1563         }
1564
1565         return rtnl_event_type;
1566 }
1567
1568 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1569 {
1570         const struct net_device *upper_dev;
1571         int ret = 0;
1572
1573         rcu_read_lock();
1574
1575         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1576         if (upper_dev)
1577                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1578
1579         rcu_read_unlock();
1580         return ret;
1581 }
1582
1583 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1584                           bool force)
1585 {
1586         int ifindex = dev_get_iflink(dev);
1587
1588         if (force || dev->ifindex != ifindex)
1589                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1590
1591         return 0;
1592 }
1593
1594 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1595                                               struct net_device *dev)
1596 {
1597         char buf[IFALIASZ];
1598         int ret;
1599
1600         ret = dev_get_alias(dev, buf, sizeof(buf));
1601         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1602 }
1603
1604 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1605                                   const struct net_device *dev,
1606                                   struct net *src_net, gfp_t gfp)
1607 {
1608         bool put_iflink = false;
1609
1610         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1611                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1612
1613                 if (!net_eq(dev_net(dev), link_net)) {
1614                         int id = peernet2id_alloc(src_net, link_net, gfp);
1615
1616                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1617                                 return -EMSGSIZE;
1618
1619                         put_iflink = true;
1620                 }
1621         }
1622
1623         return nla_put_iflink(skb, dev, put_iflink);
1624 }
1625
1626 static int rtnl_fill_link_af(struct sk_buff *skb,
1627                              const struct net_device *dev,
1628                              u32 ext_filter_mask)
1629 {
1630         const struct rtnl_af_ops *af_ops;
1631         struct nlattr *af_spec;
1632
1633         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1634         if (!af_spec)
1635                 return -EMSGSIZE;
1636
1637         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1638                 struct nlattr *af;
1639                 int err;
1640
1641                 if (!af_ops->fill_link_af)
1642                         continue;
1643
1644                 af = nla_nest_start_noflag(skb, af_ops->family);
1645                 if (!af)
1646                         return -EMSGSIZE;
1647
1648                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1649                 /*
1650                  * Caller may return ENODATA to indicate that there
1651                  * was no data to be dumped. This is not an error, it
1652                  * means we should trim the attribute header and
1653                  * continue.
1654                  */
1655                 if (err == -ENODATA)
1656                         nla_nest_cancel(skb, af);
1657                 else if (err < 0)
1658                         return -EMSGSIZE;
1659
1660                 nla_nest_end(skb, af);
1661         }
1662
1663         nla_nest_end(skb, af_spec);
1664         return 0;
1665 }
1666
1667 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1668                                  const struct net_device *dev)
1669 {
1670         struct netdev_name_node *name_node;
1671         int count = 0;
1672
1673         list_for_each_entry(name_node, &dev->name_node->list, list) {
1674                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1675                         return -EMSGSIZE;
1676                 count++;
1677         }
1678         return count;
1679 }
1680
1681 static int rtnl_fill_prop_list(struct sk_buff *skb,
1682                                const struct net_device *dev)
1683 {
1684         struct nlattr *prop_list;
1685         int ret;
1686
1687         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1688         if (!prop_list)
1689                 return -EMSGSIZE;
1690
1691         ret = rtnl_fill_alt_ifnames(skb, dev);
1692         if (ret <= 0)
1693                 goto nest_cancel;
1694
1695         nla_nest_end(skb, prop_list);
1696         return 0;
1697
1698 nest_cancel:
1699         nla_nest_cancel(skb, prop_list);
1700         return ret;
1701 }
1702
1703 static int rtnl_fill_proto_down(struct sk_buff *skb,
1704                                 const struct net_device *dev)
1705 {
1706         struct nlattr *pr;
1707         u32 preason;
1708
1709         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1710                 goto nla_put_failure;
1711
1712         preason = dev->proto_down_reason;
1713         if (!preason)
1714                 return 0;
1715
1716         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1717         if (!pr)
1718                 return -EMSGSIZE;
1719
1720         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1721                 nla_nest_cancel(skb, pr);
1722                 goto nla_put_failure;
1723         }
1724
1725         nla_nest_end(skb, pr);
1726         return 0;
1727
1728 nla_put_failure:
1729         return -EMSGSIZE;
1730 }
1731
1732 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1733                             struct net_device *dev, struct net *src_net,
1734                             int type, u32 pid, u32 seq, u32 change,
1735                             unsigned int flags, u32 ext_filter_mask,
1736                             u32 event, int *new_nsid, int new_ifindex,
1737                             int tgt_netnsid, gfp_t gfp)
1738 {
1739         struct ifinfomsg *ifm;
1740         struct nlmsghdr *nlh;
1741         struct Qdisc *qdisc;
1742
1743         ASSERT_RTNL();
1744         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1745         if (nlh == NULL)
1746                 return -EMSGSIZE;
1747
1748         ifm = nlmsg_data(nlh);
1749         ifm->ifi_family = AF_UNSPEC;
1750         ifm->__ifi_pad = 0;
1751         ifm->ifi_type = dev->type;
1752         ifm->ifi_index = dev->ifindex;
1753         ifm->ifi_flags = dev_get_flags(dev);
1754         ifm->ifi_change = change;
1755
1756         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1757                 goto nla_put_failure;
1758
1759         qdisc = rtnl_dereference(dev->qdisc);
1760         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1761             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1762             nla_put_u8(skb, IFLA_OPERSTATE,
1763                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1764             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1765             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1766             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1767             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1768             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1769             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1770             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1771             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1772             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1773             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1774             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1775             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1776 #ifdef CONFIG_RPS
1777             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1778 #endif
1779             put_master_ifindex(skb, dev) ||
1780             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1781             (qdisc &&
1782              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1783             nla_put_ifalias(skb, dev) ||
1784             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1785                         atomic_read(&dev->carrier_up_count) +
1786                         atomic_read(&dev->carrier_down_count)) ||
1787             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1788                         atomic_read(&dev->carrier_up_count)) ||
1789             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1790                         atomic_read(&dev->carrier_down_count)))
1791                 goto nla_put_failure;
1792
1793         if (rtnl_fill_proto_down(skb, dev))
1794                 goto nla_put_failure;
1795
1796         if (event != IFLA_EVENT_NONE) {
1797                 if (nla_put_u32(skb, IFLA_EVENT, event))
1798                         goto nla_put_failure;
1799         }
1800
1801         if (rtnl_fill_link_ifmap(skb, dev))
1802                 goto nla_put_failure;
1803
1804         if (dev->addr_len) {
1805                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1806                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1807                         goto nla_put_failure;
1808         }
1809
1810         if (rtnl_phys_port_id_fill(skb, dev))
1811                 goto nla_put_failure;
1812
1813         if (rtnl_phys_port_name_fill(skb, dev))
1814                 goto nla_put_failure;
1815
1816         if (rtnl_phys_switch_id_fill(skb, dev))
1817                 goto nla_put_failure;
1818
1819         if (rtnl_fill_stats(skb, dev))
1820                 goto nla_put_failure;
1821
1822         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1823                 goto nla_put_failure;
1824
1825         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1826                 goto nla_put_failure;
1827
1828         if (rtnl_xdp_fill(skb, dev))
1829                 goto nla_put_failure;
1830
1831         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1832                 if (rtnl_link_fill(skb, dev) < 0)
1833                         goto nla_put_failure;
1834         }
1835
1836         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1837                 goto nla_put_failure;
1838
1839         if (new_nsid &&
1840             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1841                 goto nla_put_failure;
1842         if (new_ifindex &&
1843             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1844                 goto nla_put_failure;
1845
1846         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1847             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1848                 goto nla_put_failure;
1849
1850         rcu_read_lock();
1851         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1852                 goto nla_put_failure_rcu;
1853         rcu_read_unlock();
1854
1855         if (rtnl_fill_prop_list(skb, dev))
1856                 goto nla_put_failure;
1857
1858         if (dev->dev.parent &&
1859             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1860                            dev_name(dev->dev.parent)))
1861                 goto nla_put_failure;
1862
1863         if (dev->dev.parent && dev->dev.parent->bus &&
1864             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1865                            dev->dev.parent->bus->name))
1866                 goto nla_put_failure;
1867
1868         nlmsg_end(skb, nlh);
1869         return 0;
1870
1871 nla_put_failure_rcu:
1872         rcu_read_unlock();
1873 nla_put_failure:
1874         nlmsg_cancel(skb, nlh);
1875         return -EMSGSIZE;
1876 }
1877
1878 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1879         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1880         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1881         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1882         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1883         [IFLA_MTU]              = { .type = NLA_U32 },
1884         [IFLA_LINK]             = { .type = NLA_U32 },
1885         [IFLA_MASTER]           = { .type = NLA_U32 },
1886         [IFLA_CARRIER]          = { .type = NLA_U8 },
1887         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1888         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1889         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1890         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1891         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1892         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1893         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1894         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1895          * allow 0-length string (needed to remove an alias).
1896          */
1897         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1898         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1899         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1900         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1901         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1902         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1903         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1904         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1905         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1906         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1907         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1908         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1909         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1910         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1911         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1912         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1913         [IFLA_XDP]              = { .type = NLA_NESTED },
1914         [IFLA_EVENT]            = { .type = NLA_U32 },
1915         [IFLA_GROUP]            = { .type = NLA_U32 },
1916         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1917         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1918         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1919         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1920         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1921         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1922         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1923                                     .len = ALTIFNAMSIZ - 1 },
1924         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1925         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1926         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1927         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1928         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
1929         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
1930         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
1931 };
1932
1933 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1934         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1935         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1936         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1937         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1938 };
1939
1940 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1941         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1942         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1943         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1944         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1945         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1946         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1947         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1948         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1949         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1950         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1951         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1952         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1953         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1954 };
1955
1956 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1957         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1958         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1959                                     .len = PORT_PROFILE_MAX },
1960         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1961                                       .len = PORT_UUID_MAX },
1962         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1963                                     .len = PORT_UUID_MAX },
1964         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1965         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1966
1967         /* Unused, but we need to keep it here since user space could
1968          * fill it. It's also broken with regard to NLA_BINARY use in
1969          * combination with structs.
1970          */
1971         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1972                                     .len = sizeof(struct ifla_port_vsi) },
1973 };
1974
1975 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1976         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1977         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1978         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1979         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1980         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1981         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1982 };
1983
1984 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1985 {
1986         const struct rtnl_link_ops *ops = NULL;
1987         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1988
1989         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1990                 return NULL;
1991
1992         if (linfo[IFLA_INFO_KIND]) {
1993                 char kind[MODULE_NAME_LEN];
1994
1995                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1996                 ops = rtnl_link_ops_get(kind);
1997         }
1998
1999         return ops;
2000 }
2001
2002 static bool link_master_filtered(struct net_device *dev, int master_idx)
2003 {
2004         struct net_device *master;
2005
2006         if (!master_idx)
2007                 return false;
2008
2009         master = netdev_master_upper_dev_get(dev);
2010
2011         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2012          * another invalid value for ifindex to denote "no master".
2013          */
2014         if (master_idx == -1)
2015                 return !!master;
2016
2017         if (!master || master->ifindex != master_idx)
2018                 return true;
2019
2020         return false;
2021 }
2022
2023 static bool link_kind_filtered(const struct net_device *dev,
2024                                const struct rtnl_link_ops *kind_ops)
2025 {
2026         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2027                 return true;
2028
2029         return false;
2030 }
2031
2032 static bool link_dump_filtered(struct net_device *dev,
2033                                int master_idx,
2034                                const struct rtnl_link_ops *kind_ops)
2035 {
2036         if (link_master_filtered(dev, master_idx) ||
2037             link_kind_filtered(dev, kind_ops))
2038                 return true;
2039
2040         return false;
2041 }
2042
2043 /**
2044  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2045  * @sk: netlink socket
2046  * @netnsid: network namespace identifier
2047  *
2048  * Returns the network namespace identified by netnsid on success or an error
2049  * pointer on failure.
2050  */
2051 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2052 {
2053         struct net *net;
2054
2055         net = get_net_ns_by_id(sock_net(sk), netnsid);
2056         if (!net)
2057                 return ERR_PTR(-EINVAL);
2058
2059         /* For now, the caller is required to have CAP_NET_ADMIN in
2060          * the user namespace owning the target net ns.
2061          */
2062         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2063                 put_net(net);
2064                 return ERR_PTR(-EACCES);
2065         }
2066         return net;
2067 }
2068 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2069
2070 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2071                                       bool strict_check, struct nlattr **tb,
2072                                       struct netlink_ext_ack *extack)
2073 {
2074         int hdrlen;
2075
2076         if (strict_check) {
2077                 struct ifinfomsg *ifm;
2078
2079                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2080                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2081                         return -EINVAL;
2082                 }
2083
2084                 ifm = nlmsg_data(nlh);
2085                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2086                     ifm->ifi_change) {
2087                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2088                         return -EINVAL;
2089                 }
2090                 if (ifm->ifi_index) {
2091                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2092                         return -EINVAL;
2093                 }
2094
2095                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2096                                                      IFLA_MAX, ifla_policy,
2097                                                      extack);
2098         }
2099
2100         /* A hack to preserve kernel<->userspace interface.
2101          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2102          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2103          * what iproute2 < v3.9.0 used.
2104          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2105          * attribute, its netlink message is shorter than struct ifinfomsg.
2106          */
2107         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2108                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2109
2110         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2111                                       extack);
2112 }
2113
2114 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2115 {
2116         struct netlink_ext_ack *extack = cb->extack;
2117         const struct nlmsghdr *nlh = cb->nlh;
2118         struct net *net = sock_net(skb->sk);
2119         struct net *tgt_net = net;
2120         int h, s_h;
2121         int idx = 0, s_idx;
2122         struct net_device *dev;
2123         struct hlist_head *head;
2124         struct nlattr *tb[IFLA_MAX+1];
2125         u32 ext_filter_mask = 0;
2126         const struct rtnl_link_ops *kind_ops = NULL;
2127         unsigned int flags = NLM_F_MULTI;
2128         int master_idx = 0;
2129         int netnsid = -1;
2130         int err, i;
2131
2132         s_h = cb->args[0];
2133         s_idx = cb->args[1];
2134
2135         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2136         if (err < 0) {
2137                 if (cb->strict_check)
2138                         return err;
2139
2140                 goto walk_entries;
2141         }
2142
2143         for (i = 0; i <= IFLA_MAX; ++i) {
2144                 if (!tb[i])
2145                         continue;
2146
2147                 /* new attributes should only be added with strict checking */
2148                 switch (i) {
2149                 case IFLA_TARGET_NETNSID:
2150                         netnsid = nla_get_s32(tb[i]);
2151                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2152                         if (IS_ERR(tgt_net)) {
2153                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2154                                 return PTR_ERR(tgt_net);
2155                         }
2156                         break;
2157                 case IFLA_EXT_MASK:
2158                         ext_filter_mask = nla_get_u32(tb[i]);
2159                         break;
2160                 case IFLA_MASTER:
2161                         master_idx = nla_get_u32(tb[i]);
2162                         break;
2163                 case IFLA_LINKINFO:
2164                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2165                         break;
2166                 default:
2167                         if (cb->strict_check) {
2168                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2169                                 return -EINVAL;
2170                         }
2171                 }
2172         }
2173
2174         if (master_idx || kind_ops)
2175                 flags |= NLM_F_DUMP_FILTERED;
2176
2177 walk_entries:
2178         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2179                 idx = 0;
2180                 head = &tgt_net->dev_index_head[h];
2181                 hlist_for_each_entry(dev, head, index_hlist) {
2182                         if (link_dump_filtered(dev, master_idx, kind_ops))
2183                                 goto cont;
2184                         if (idx < s_idx)
2185                                 goto cont;
2186                         err = rtnl_fill_ifinfo(skb, dev, net,
2187                                                RTM_NEWLINK,
2188                                                NETLINK_CB(cb->skb).portid,
2189                                                nlh->nlmsg_seq, 0, flags,
2190                                                ext_filter_mask, 0, NULL, 0,
2191                                                netnsid, GFP_KERNEL);
2192
2193                         if (err < 0) {
2194                                 if (likely(skb->len))
2195                                         goto out;
2196
2197                                 goto out_err;
2198                         }
2199 cont:
2200                         idx++;
2201                 }
2202         }
2203 out:
2204         err = skb->len;
2205 out_err:
2206         cb->args[1] = idx;
2207         cb->args[0] = h;
2208         cb->seq = tgt_net->dev_base_seq;
2209         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2210         if (netnsid >= 0)
2211                 put_net(tgt_net);
2212
2213         return err;
2214 }
2215
2216 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2217                         struct netlink_ext_ack *exterr)
2218 {
2219         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2220                                     exterr);
2221 }
2222 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2223
2224 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2225 {
2226         struct net *net;
2227         /* Examine the link attributes and figure out which
2228          * network namespace we are talking about.
2229          */
2230         if (tb[IFLA_NET_NS_PID])
2231                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2232         else if (tb[IFLA_NET_NS_FD])
2233                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2234         else
2235                 net = get_net(src_net);
2236         return net;
2237 }
2238 EXPORT_SYMBOL(rtnl_link_get_net);
2239
2240 /* Figure out which network namespace we are talking about by
2241  * examining the link attributes in the following order:
2242  *
2243  * 1. IFLA_NET_NS_PID
2244  * 2. IFLA_NET_NS_FD
2245  * 3. IFLA_TARGET_NETNSID
2246  */
2247 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2248                                                struct nlattr *tb[])
2249 {
2250         struct net *net;
2251
2252         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2253                 return rtnl_link_get_net(src_net, tb);
2254
2255         if (!tb[IFLA_TARGET_NETNSID])
2256                 return get_net(src_net);
2257
2258         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2259         if (!net)
2260                 return ERR_PTR(-EINVAL);
2261
2262         return net;
2263 }
2264
2265 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2266                                              struct net *src_net,
2267                                              struct nlattr *tb[], int cap)
2268 {
2269         struct net *net;
2270
2271         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2272         if (IS_ERR(net))
2273                 return net;
2274
2275         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2276                 put_net(net);
2277                 return ERR_PTR(-EPERM);
2278         }
2279
2280         return net;
2281 }
2282
2283 /* Verify that rtnetlink requests do not pass additional properties
2284  * potentially referring to different network namespaces.
2285  */
2286 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2287                                     struct netlink_ext_ack *extack,
2288                                     bool netns_id_only)
2289 {
2290
2291         if (netns_id_only) {
2292                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2293                         return 0;
2294
2295                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2296                 return -EOPNOTSUPP;
2297         }
2298
2299         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2300                 goto invalid_attr;
2301
2302         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2303                 goto invalid_attr;
2304
2305         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2306                 goto invalid_attr;
2307
2308         return 0;
2309
2310 invalid_attr:
2311         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2312         return -EINVAL;
2313 }
2314
2315 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2316                              int max_tx_rate)
2317 {
2318         const struct net_device_ops *ops = dev->netdev_ops;
2319
2320         if (!ops->ndo_set_vf_rate)
2321                 return -EOPNOTSUPP;
2322         if (max_tx_rate && max_tx_rate < min_tx_rate)
2323                 return -EINVAL;
2324
2325         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2326 }
2327
2328 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2329                             struct netlink_ext_ack *extack)
2330 {
2331         if (dev) {
2332                 if (tb[IFLA_ADDRESS] &&
2333                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2334                         return -EINVAL;
2335
2336                 if (tb[IFLA_BROADCAST] &&
2337                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2338                         return -EINVAL;
2339         }
2340
2341         if (tb[IFLA_AF_SPEC]) {
2342                 struct nlattr *af;
2343                 int rem, err;
2344
2345                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2346                         const struct rtnl_af_ops *af_ops;
2347
2348                         af_ops = rtnl_af_lookup(nla_type(af));
2349                         if (!af_ops)
2350                                 return -EAFNOSUPPORT;
2351
2352                         if (!af_ops->set_link_af)
2353                                 return -EOPNOTSUPP;
2354
2355                         if (af_ops->validate_link_af) {
2356                                 err = af_ops->validate_link_af(dev, af, extack);
2357                                 if (err < 0)
2358                                         return err;
2359                         }
2360                 }
2361         }
2362
2363         return 0;
2364 }
2365
2366 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2367                                   int guid_type)
2368 {
2369         const struct net_device_ops *ops = dev->netdev_ops;
2370
2371         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2372 }
2373
2374 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2375 {
2376         if (dev->type != ARPHRD_INFINIBAND)
2377                 return -EOPNOTSUPP;
2378
2379         return handle_infiniband_guid(dev, ivt, guid_type);
2380 }
2381
2382 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2383 {
2384         const struct net_device_ops *ops = dev->netdev_ops;
2385         int err = -EINVAL;
2386
2387         if (tb[IFLA_VF_MAC]) {
2388                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2389
2390                 if (ivm->vf >= INT_MAX)
2391                         return -EINVAL;
2392                 err = -EOPNOTSUPP;
2393                 if (ops->ndo_set_vf_mac)
2394                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2395                                                   ivm->mac);
2396                 if (err < 0)
2397                         return err;
2398         }
2399
2400         if (tb[IFLA_VF_VLAN]) {
2401                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2402
2403                 if (ivv->vf >= INT_MAX)
2404                         return -EINVAL;
2405                 err = -EOPNOTSUPP;
2406                 if (ops->ndo_set_vf_vlan)
2407                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2408                                                    ivv->qos,
2409                                                    htons(ETH_P_8021Q));
2410                 if (err < 0)
2411                         return err;
2412         }
2413
2414         if (tb[IFLA_VF_VLAN_LIST]) {
2415                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2416                 struct nlattr *attr;
2417                 int rem, len = 0;
2418
2419                 err = -EOPNOTSUPP;
2420                 if (!ops->ndo_set_vf_vlan)
2421                         return err;
2422
2423                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2424                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2425                             nla_len(attr) < NLA_HDRLEN) {
2426                                 return -EINVAL;
2427                         }
2428                         if (len >= MAX_VLAN_LIST_LEN)
2429                                 return -EOPNOTSUPP;
2430                         ivvl[len] = nla_data(attr);
2431
2432                         len++;
2433                 }
2434                 if (len == 0)
2435                         return -EINVAL;
2436
2437                 if (ivvl[0]->vf >= INT_MAX)
2438                         return -EINVAL;
2439                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2440                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2441                 if (err < 0)
2442                         return err;
2443         }
2444
2445         if (tb[IFLA_VF_TX_RATE]) {
2446                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2447                 struct ifla_vf_info ivf;
2448
2449                 if (ivt->vf >= INT_MAX)
2450                         return -EINVAL;
2451                 err = -EOPNOTSUPP;
2452                 if (ops->ndo_get_vf_config)
2453                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2454                 if (err < 0)
2455                         return err;
2456
2457                 err = rtnl_set_vf_rate(dev, ivt->vf,
2458                                        ivf.min_tx_rate, ivt->rate);
2459                 if (err < 0)
2460                         return err;
2461         }
2462
2463         if (tb[IFLA_VF_RATE]) {
2464                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2465
2466                 if (ivt->vf >= INT_MAX)
2467                         return -EINVAL;
2468
2469                 err = rtnl_set_vf_rate(dev, ivt->vf,
2470                                        ivt->min_tx_rate, ivt->max_tx_rate);
2471                 if (err < 0)
2472                         return err;
2473         }
2474
2475         if (tb[IFLA_VF_SPOOFCHK]) {
2476                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2477
2478                 if (ivs->vf >= INT_MAX)
2479                         return -EINVAL;
2480                 err = -EOPNOTSUPP;
2481                 if (ops->ndo_set_vf_spoofchk)
2482                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2483                                                        ivs->setting);
2484                 if (err < 0)
2485                         return err;
2486         }
2487
2488         if (tb[IFLA_VF_LINK_STATE]) {
2489                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2490
2491                 if (ivl->vf >= INT_MAX)
2492                         return -EINVAL;
2493                 err = -EOPNOTSUPP;
2494                 if (ops->ndo_set_vf_link_state)
2495                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2496                                                          ivl->link_state);
2497                 if (err < 0)
2498                         return err;
2499         }
2500
2501         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2502                 struct ifla_vf_rss_query_en *ivrssq_en;
2503
2504                 err = -EOPNOTSUPP;
2505                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2506                 if (ivrssq_en->vf >= INT_MAX)
2507                         return -EINVAL;
2508                 if (ops->ndo_set_vf_rss_query_en)
2509                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2510                                                            ivrssq_en->setting);
2511                 if (err < 0)
2512                         return err;
2513         }
2514
2515         if (tb[IFLA_VF_TRUST]) {
2516                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2517
2518                 if (ivt->vf >= INT_MAX)
2519                         return -EINVAL;
2520                 err = -EOPNOTSUPP;
2521                 if (ops->ndo_set_vf_trust)
2522                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2523                 if (err < 0)
2524                         return err;
2525         }
2526
2527         if (tb[IFLA_VF_IB_NODE_GUID]) {
2528                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2529
2530                 if (ivt->vf >= INT_MAX)
2531                         return -EINVAL;
2532                 if (!ops->ndo_set_vf_guid)
2533                         return -EOPNOTSUPP;
2534                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2535         }
2536
2537         if (tb[IFLA_VF_IB_PORT_GUID]) {
2538                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2539
2540                 if (ivt->vf >= INT_MAX)
2541                         return -EINVAL;
2542                 if (!ops->ndo_set_vf_guid)
2543                         return -EOPNOTSUPP;
2544
2545                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2546         }
2547
2548         return err;
2549 }
2550
2551 static int do_set_master(struct net_device *dev, int ifindex,
2552                          struct netlink_ext_ack *extack)
2553 {
2554         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2555         const struct net_device_ops *ops;
2556         int err;
2557
2558         if (upper_dev) {
2559                 if (upper_dev->ifindex == ifindex)
2560                         return 0;
2561                 ops = upper_dev->netdev_ops;
2562                 if (ops->ndo_del_slave) {
2563                         err = ops->ndo_del_slave(upper_dev, dev);
2564                         if (err)
2565                                 return err;
2566                 } else {
2567                         return -EOPNOTSUPP;
2568                 }
2569         }
2570
2571         if (ifindex) {
2572                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2573                 if (!upper_dev)
2574                         return -EINVAL;
2575                 ops = upper_dev->netdev_ops;
2576                 if (ops->ndo_add_slave) {
2577                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2578                         if (err)
2579                                 return err;
2580                 } else {
2581                         return -EOPNOTSUPP;
2582                 }
2583         }
2584         return 0;
2585 }
2586
2587 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2588         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2589         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2590 };
2591
2592 static int do_set_proto_down(struct net_device *dev,
2593                              struct nlattr *nl_proto_down,
2594                              struct nlattr *nl_proto_down_reason,
2595                              struct netlink_ext_ack *extack)
2596 {
2597         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2598         unsigned long mask = 0;
2599         u32 value;
2600         bool proto_down;
2601         int err;
2602
2603         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2604                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2605                 return -EOPNOTSUPP;
2606         }
2607
2608         if (nl_proto_down_reason) {
2609                 err = nla_parse_nested_deprecated(pdreason,
2610                                                   IFLA_PROTO_DOWN_REASON_MAX,
2611                                                   nl_proto_down_reason,
2612                                                   ifla_proto_down_reason_policy,
2613                                                   NULL);
2614                 if (err < 0)
2615                         return err;
2616
2617                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2618                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2619                         return -EINVAL;
2620                 }
2621
2622                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2623
2624                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2625                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2626
2627                 dev_change_proto_down_reason(dev, mask, value);
2628         }
2629
2630         if (nl_proto_down) {
2631                 proto_down = nla_get_u8(nl_proto_down);
2632
2633                 /* Don't turn off protodown if there are active reasons */
2634                 if (!proto_down && dev->proto_down_reason) {
2635                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2636                         return -EBUSY;
2637                 }
2638                 err = dev_change_proto_down(dev,
2639                                             proto_down);
2640                 if (err)
2641                         return err;
2642         }
2643
2644         return 0;
2645 }
2646
2647 #define DO_SETLINK_MODIFIED     0x01
2648 /* notify flag means notify + modified. */
2649 #define DO_SETLINK_NOTIFY       0x03
2650 static int do_setlink(const struct sk_buff *skb,
2651                       struct net_device *dev, struct ifinfomsg *ifm,
2652                       struct netlink_ext_ack *extack,
2653                       struct nlattr **tb, int status)
2654 {
2655         const struct net_device_ops *ops = dev->netdev_ops;
2656         char ifname[IFNAMSIZ];
2657         int err;
2658
2659         err = validate_linkmsg(dev, tb, extack);
2660         if (err < 0)
2661                 return err;
2662
2663         if (tb[IFLA_IFNAME])
2664                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2665         else
2666                 ifname[0] = '\0';
2667
2668         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2669                 const char *pat = ifname[0] ? ifname : NULL;
2670                 struct net *net;
2671                 int new_ifindex;
2672
2673                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2674                                                 tb, CAP_NET_ADMIN);
2675                 if (IS_ERR(net)) {
2676                         err = PTR_ERR(net);
2677                         goto errout;
2678                 }
2679
2680                 if (tb[IFLA_NEW_IFINDEX])
2681                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2682                 else
2683                         new_ifindex = 0;
2684
2685                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2686                 put_net(net);
2687                 if (err)
2688                         goto errout;
2689                 status |= DO_SETLINK_MODIFIED;
2690         }
2691
2692         if (tb[IFLA_MAP]) {
2693                 struct rtnl_link_ifmap *u_map;
2694                 struct ifmap k_map;
2695
2696                 if (!ops->ndo_set_config) {
2697                         err = -EOPNOTSUPP;
2698                         goto errout;
2699                 }
2700
2701                 if (!netif_device_present(dev)) {
2702                         err = -ENODEV;
2703                         goto errout;
2704                 }
2705
2706                 u_map = nla_data(tb[IFLA_MAP]);
2707                 k_map.mem_start = (unsigned long) u_map->mem_start;
2708                 k_map.mem_end = (unsigned long) u_map->mem_end;
2709                 k_map.base_addr = (unsigned short) u_map->base_addr;
2710                 k_map.irq = (unsigned char) u_map->irq;
2711                 k_map.dma = (unsigned char) u_map->dma;
2712                 k_map.port = (unsigned char) u_map->port;
2713
2714                 err = ops->ndo_set_config(dev, &k_map);
2715                 if (err < 0)
2716                         goto errout;
2717
2718                 status |= DO_SETLINK_NOTIFY;
2719         }
2720
2721         if (tb[IFLA_ADDRESS]) {
2722                 struct sockaddr *sa;
2723                 int len;
2724
2725                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2726                                                   sizeof(*sa));
2727                 sa = kmalloc(len, GFP_KERNEL);
2728                 if (!sa) {
2729                         err = -ENOMEM;
2730                         goto errout;
2731                 }
2732                 sa->sa_family = dev->type;
2733                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2734                        dev->addr_len);
2735                 err = dev_set_mac_address_user(dev, sa, extack);
2736                 kfree(sa);
2737                 if (err)
2738                         goto errout;
2739                 status |= DO_SETLINK_MODIFIED;
2740         }
2741
2742         if (tb[IFLA_MTU]) {
2743                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2744                 if (err < 0)
2745                         goto errout;
2746                 status |= DO_SETLINK_MODIFIED;
2747         }
2748
2749         if (tb[IFLA_GROUP]) {
2750                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2751                 status |= DO_SETLINK_NOTIFY;
2752         }
2753
2754         /*
2755          * Interface selected by interface index but interface
2756          * name provided implies that a name change has been
2757          * requested.
2758          */
2759         if (ifm->ifi_index > 0 && ifname[0]) {
2760                 err = dev_change_name(dev, ifname);
2761                 if (err < 0)
2762                         goto errout;
2763                 status |= DO_SETLINK_MODIFIED;
2764         }
2765
2766         if (tb[IFLA_IFALIAS]) {
2767                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2768                                     nla_len(tb[IFLA_IFALIAS]));
2769                 if (err < 0)
2770                         goto errout;
2771                 status |= DO_SETLINK_NOTIFY;
2772         }
2773
2774         if (tb[IFLA_BROADCAST]) {
2775                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2776                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2777         }
2778
2779         if (ifm->ifi_flags || ifm->ifi_change) {
2780                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2781                                        extack);
2782                 if (err < 0)
2783                         goto errout;
2784         }
2785
2786         if (tb[IFLA_MASTER]) {
2787                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2788                 if (err)
2789                         goto errout;
2790                 status |= DO_SETLINK_MODIFIED;
2791         }
2792
2793         if (tb[IFLA_CARRIER]) {
2794                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2795                 if (err)
2796                         goto errout;
2797                 status |= DO_SETLINK_MODIFIED;
2798         }
2799
2800         if (tb[IFLA_TXQLEN]) {
2801                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2802
2803                 err = dev_change_tx_queue_len(dev, value);
2804                 if (err)
2805                         goto errout;
2806                 status |= DO_SETLINK_MODIFIED;
2807         }
2808
2809         if (tb[IFLA_GSO_MAX_SIZE]) {
2810                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2811
2812                 if (max_size > dev->tso_max_size) {
2813                         err = -EINVAL;
2814                         goto errout;
2815                 }
2816
2817                 if (dev->gso_max_size ^ max_size) {
2818                         netif_set_gso_max_size(dev, max_size);
2819                         status |= DO_SETLINK_MODIFIED;
2820                 }
2821         }
2822
2823         if (tb[IFLA_GSO_MAX_SEGS]) {
2824                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2825
2826                 if (max_segs > GSO_MAX_SEGS || max_segs > dev->tso_max_segs) {
2827                         err = -EINVAL;
2828                         goto errout;
2829                 }
2830
2831                 if (dev->gso_max_segs ^ max_segs) {
2832                         netif_set_gso_max_segs(dev, max_segs);
2833                         status |= DO_SETLINK_MODIFIED;
2834                 }
2835         }
2836
2837         if (tb[IFLA_GRO_MAX_SIZE]) {
2838                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2839
2840                 if (dev->gro_max_size ^ gro_max_size) {
2841                         netif_set_gro_max_size(dev, gro_max_size);
2842                         status |= DO_SETLINK_MODIFIED;
2843                 }
2844         }
2845
2846         if (tb[IFLA_OPERSTATE])
2847                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2848
2849         if (tb[IFLA_LINKMODE]) {
2850                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2851
2852                 write_lock(&dev_base_lock);
2853                 if (dev->link_mode ^ value)
2854                         status |= DO_SETLINK_NOTIFY;
2855                 dev->link_mode = value;
2856                 write_unlock(&dev_base_lock);
2857         }
2858
2859         if (tb[IFLA_VFINFO_LIST]) {
2860                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2861                 struct nlattr *attr;
2862                 int rem;
2863
2864                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2865                         if (nla_type(attr) != IFLA_VF_INFO ||
2866                             nla_len(attr) < NLA_HDRLEN) {
2867                                 err = -EINVAL;
2868                                 goto errout;
2869                         }
2870                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2871                                                           attr,
2872                                                           ifla_vf_policy,
2873                                                           NULL);
2874                         if (err < 0)
2875                                 goto errout;
2876                         err = do_setvfinfo(dev, vfinfo);
2877                         if (err < 0)
2878                                 goto errout;
2879                         status |= DO_SETLINK_NOTIFY;
2880                 }
2881         }
2882         err = 0;
2883
2884         if (tb[IFLA_VF_PORTS]) {
2885                 struct nlattr *port[IFLA_PORT_MAX+1];
2886                 struct nlattr *attr;
2887                 int vf;
2888                 int rem;
2889
2890                 err = -EOPNOTSUPP;
2891                 if (!ops->ndo_set_vf_port)
2892                         goto errout;
2893
2894                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2895                         if (nla_type(attr) != IFLA_VF_PORT ||
2896                             nla_len(attr) < NLA_HDRLEN) {
2897                                 err = -EINVAL;
2898                                 goto errout;
2899                         }
2900                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2901                                                           attr,
2902                                                           ifla_port_policy,
2903                                                           NULL);
2904                         if (err < 0)
2905                                 goto errout;
2906                         if (!port[IFLA_PORT_VF]) {
2907                                 err = -EOPNOTSUPP;
2908                                 goto errout;
2909                         }
2910                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2911                         err = ops->ndo_set_vf_port(dev, vf, port);
2912                         if (err < 0)
2913                                 goto errout;
2914                         status |= DO_SETLINK_NOTIFY;
2915                 }
2916         }
2917         err = 0;
2918
2919         if (tb[IFLA_PORT_SELF]) {
2920                 struct nlattr *port[IFLA_PORT_MAX+1];
2921
2922                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2923                                                   tb[IFLA_PORT_SELF],
2924                                                   ifla_port_policy, NULL);
2925                 if (err < 0)
2926                         goto errout;
2927
2928                 err = -EOPNOTSUPP;
2929                 if (ops->ndo_set_vf_port)
2930                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2931                 if (err < 0)
2932                         goto errout;
2933                 status |= DO_SETLINK_NOTIFY;
2934         }
2935
2936         if (tb[IFLA_AF_SPEC]) {
2937                 struct nlattr *af;
2938                 int rem;
2939
2940                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2941                         const struct rtnl_af_ops *af_ops;
2942
2943                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2944
2945                         err = af_ops->set_link_af(dev, af, extack);
2946                         if (err < 0)
2947                                 goto errout;
2948
2949                         status |= DO_SETLINK_NOTIFY;
2950                 }
2951         }
2952         err = 0;
2953
2954         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2955                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2956                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2957                 if (err)
2958                         goto errout;
2959                 status |= DO_SETLINK_NOTIFY;
2960         }
2961
2962         if (tb[IFLA_XDP]) {
2963                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2964                 u32 xdp_flags = 0;
2965
2966                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2967                                                   tb[IFLA_XDP],
2968                                                   ifla_xdp_policy, NULL);
2969                 if (err < 0)
2970                         goto errout;
2971
2972                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2973                         err = -EINVAL;
2974                         goto errout;
2975                 }
2976
2977                 if (xdp[IFLA_XDP_FLAGS]) {
2978                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2979                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2980                                 err = -EINVAL;
2981                                 goto errout;
2982                         }
2983                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2984                                 err = -EINVAL;
2985                                 goto errout;
2986                         }
2987                 }
2988
2989                 if (xdp[IFLA_XDP_FD]) {
2990                         int expected_fd = -1;
2991
2992                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2993                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2994                                         err = -EINVAL;
2995                                         goto errout;
2996                                 }
2997                                 expected_fd =
2998                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2999                         }
3000
3001                         err = dev_change_xdp_fd(dev, extack,
3002                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3003                                                 expected_fd,
3004                                                 xdp_flags);
3005                         if (err)
3006                                 goto errout;
3007                         status |= DO_SETLINK_NOTIFY;
3008                 }
3009         }
3010
3011 errout:
3012         if (status & DO_SETLINK_MODIFIED) {
3013                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3014                         netdev_state_change(dev);
3015
3016                 if (err < 0)
3017                         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",
3018                                              dev->name);
3019         }
3020
3021         return err;
3022 }
3023
3024 static struct net_device *rtnl_dev_get(struct net *net,
3025                                        struct nlattr *tb[])
3026 {
3027         char ifname[ALTIFNAMSIZ];
3028
3029         if (tb[IFLA_IFNAME])
3030                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3031         else if (tb[IFLA_ALT_IFNAME])
3032                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3033         else
3034                 return NULL;
3035
3036         return __dev_get_by_name(net, ifname);
3037 }
3038
3039 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3040                         struct netlink_ext_ack *extack)
3041 {
3042         struct net *net = sock_net(skb->sk);
3043         struct ifinfomsg *ifm;
3044         struct net_device *dev;
3045         int err;
3046         struct nlattr *tb[IFLA_MAX+1];
3047
3048         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3049                                      ifla_policy, extack);
3050         if (err < 0)
3051                 goto errout;
3052
3053         err = rtnl_ensure_unique_netns(tb, extack, false);
3054         if (err < 0)
3055                 goto errout;
3056
3057         err = -EINVAL;
3058         ifm = nlmsg_data(nlh);
3059         if (ifm->ifi_index > 0)
3060                 dev = __dev_get_by_index(net, ifm->ifi_index);
3061         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3062                 dev = rtnl_dev_get(net, tb);
3063         else
3064                 goto errout;
3065
3066         if (dev == NULL) {
3067                 err = -ENODEV;
3068                 goto errout;
3069         }
3070
3071         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3072 errout:
3073         return err;
3074 }
3075
3076 static int rtnl_group_dellink(const struct net *net, int group)
3077 {
3078         struct net_device *dev, *aux;
3079         LIST_HEAD(list_kill);
3080         bool found = false;
3081
3082         if (!group)
3083                 return -EPERM;
3084
3085         for_each_netdev(net, dev) {
3086                 if (dev->group == group) {
3087                         const struct rtnl_link_ops *ops;
3088
3089                         found = true;
3090                         ops = dev->rtnl_link_ops;
3091                         if (!ops || !ops->dellink)
3092                                 return -EOPNOTSUPP;
3093                 }
3094         }
3095
3096         if (!found)
3097                 return -ENODEV;
3098
3099         for_each_netdev_safe(net, dev, aux) {
3100                 if (dev->group == group) {
3101                         const struct rtnl_link_ops *ops;
3102
3103                         ops = dev->rtnl_link_ops;
3104                         ops->dellink(dev, &list_kill);
3105                 }
3106         }
3107         unregister_netdevice_many(&list_kill);
3108
3109         return 0;
3110 }
3111
3112 int rtnl_delete_link(struct net_device *dev)
3113 {
3114         const struct rtnl_link_ops *ops;
3115         LIST_HEAD(list_kill);
3116
3117         ops = dev->rtnl_link_ops;
3118         if (!ops || !ops->dellink)
3119                 return -EOPNOTSUPP;
3120
3121         ops->dellink(dev, &list_kill);
3122         unregister_netdevice_many(&list_kill);
3123
3124         return 0;
3125 }
3126 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3127
3128 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3129                         struct netlink_ext_ack *extack)
3130 {
3131         struct net *net = sock_net(skb->sk);
3132         struct net *tgt_net = net;
3133         struct net_device *dev = NULL;
3134         struct ifinfomsg *ifm;
3135         struct nlattr *tb[IFLA_MAX+1];
3136         int err;
3137         int netnsid = -1;
3138
3139         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3140                                      ifla_policy, extack);
3141         if (err < 0)
3142                 return err;
3143
3144         err = rtnl_ensure_unique_netns(tb, extack, true);
3145         if (err < 0)
3146                 return err;
3147
3148         if (tb[IFLA_TARGET_NETNSID]) {
3149                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3150                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3151                 if (IS_ERR(tgt_net))
3152                         return PTR_ERR(tgt_net);
3153         }
3154
3155         err = -EINVAL;
3156         ifm = nlmsg_data(nlh);
3157         if (ifm->ifi_index > 0)
3158                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3159         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3160                 dev = rtnl_dev_get(net, tb);
3161         else if (tb[IFLA_GROUP])
3162                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3163         else
3164                 goto out;
3165
3166         if (!dev) {
3167                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3168                         err = -ENODEV;
3169
3170                 goto out;
3171         }
3172
3173         err = rtnl_delete_link(dev);
3174
3175 out:
3176         if (netnsid >= 0)
3177                 put_net(tgt_net);
3178
3179         return err;
3180 }
3181
3182 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3183 {
3184         unsigned int old_flags;
3185         int err;
3186
3187         old_flags = dev->flags;
3188         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3189                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3190                                          NULL);
3191                 if (err < 0)
3192                         return err;
3193         }
3194
3195         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3196                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3197         } else {
3198                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3199                 __dev_notify_flags(dev, old_flags, ~0U);
3200         }
3201         return 0;
3202 }
3203 EXPORT_SYMBOL(rtnl_configure_link);
3204
3205 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3206                                     unsigned char name_assign_type,
3207                                     const struct rtnl_link_ops *ops,
3208                                     struct nlattr *tb[],
3209                                     struct netlink_ext_ack *extack)
3210 {
3211         struct net_device *dev;
3212         unsigned int num_tx_queues = 1;
3213         unsigned int num_rx_queues = 1;
3214
3215         if (tb[IFLA_NUM_TX_QUEUES])
3216                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3217         else if (ops->get_num_tx_queues)
3218                 num_tx_queues = ops->get_num_tx_queues();
3219
3220         if (tb[IFLA_NUM_RX_QUEUES])
3221                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3222         else if (ops->get_num_rx_queues)
3223                 num_rx_queues = ops->get_num_rx_queues();
3224
3225         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3226                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3227                 return ERR_PTR(-EINVAL);
3228         }
3229
3230         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3231                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3232                 return ERR_PTR(-EINVAL);
3233         }
3234
3235         if (ops->alloc) {
3236                 dev = ops->alloc(tb, ifname, name_assign_type,
3237                                  num_tx_queues, num_rx_queues);
3238                 if (IS_ERR(dev))
3239                         return dev;
3240         } else {
3241                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3242                                        name_assign_type, ops->setup,
3243                                        num_tx_queues, num_rx_queues);
3244         }
3245
3246         if (!dev)
3247                 return ERR_PTR(-ENOMEM);
3248
3249         dev_net_set(dev, net);
3250         dev->rtnl_link_ops = ops;
3251         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3252
3253         if (tb[IFLA_MTU]) {
3254                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3255                 int err;
3256
3257                 err = dev_validate_mtu(dev, mtu, extack);
3258                 if (err) {
3259                         free_netdev(dev);
3260                         return ERR_PTR(err);
3261                 }
3262                 dev->mtu = mtu;
3263         }
3264         if (tb[IFLA_ADDRESS]) {
3265                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3266                                nla_len(tb[IFLA_ADDRESS]));
3267                 dev->addr_assign_type = NET_ADDR_SET;
3268         }
3269         if (tb[IFLA_BROADCAST])
3270                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3271                                 nla_len(tb[IFLA_BROADCAST]));
3272         if (tb[IFLA_TXQLEN])
3273                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3274         if (tb[IFLA_OPERSTATE])
3275                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3276         if (tb[IFLA_LINKMODE])
3277                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3278         if (tb[IFLA_GROUP])
3279                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3280         if (tb[IFLA_GSO_MAX_SIZE])
3281                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3282         if (tb[IFLA_GSO_MAX_SEGS])
3283                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3284         if (tb[IFLA_GRO_MAX_SIZE])
3285                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3286
3287         return dev;
3288 }
3289 EXPORT_SYMBOL(rtnl_create_link);
3290
3291 static int rtnl_group_changelink(const struct sk_buff *skb,
3292                 struct net *net, int group,
3293                 struct ifinfomsg *ifm,
3294                 struct netlink_ext_ack *extack,
3295                 struct nlattr **tb)
3296 {
3297         struct net_device *dev, *aux;
3298         int err;
3299
3300         for_each_netdev_safe(net, dev, aux) {
3301                 if (dev->group == group) {
3302                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3303                         if (err < 0)
3304                                 return err;
3305                 }
3306         }
3307
3308         return 0;
3309 }
3310
3311 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3312                                const struct rtnl_link_ops *ops,
3313                                struct nlattr **tb, struct nlattr **data,
3314                                struct netlink_ext_ack *extack)
3315 {
3316         unsigned char name_assign_type = NET_NAME_USER;
3317         struct net *net = sock_net(skb->sk);
3318         struct net *dest_net, *link_net;
3319         struct net_device *dev;
3320         char ifname[IFNAMSIZ];
3321         int err;
3322
3323         if (!ops->alloc && !ops->setup)
3324                 return -EOPNOTSUPP;
3325
3326         if (tb[IFLA_IFNAME]) {
3327                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3328         } else {
3329                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3330                 name_assign_type = NET_NAME_ENUM;
3331         }
3332
3333         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3334         if (IS_ERR(dest_net))
3335                 return PTR_ERR(dest_net);
3336
3337         if (tb[IFLA_LINK_NETNSID]) {
3338                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3339
3340                 link_net = get_net_ns_by_id(dest_net, id);
3341                 if (!link_net) {
3342                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3343                         err =  -EINVAL;
3344                         goto out;
3345                 }
3346                 err = -EPERM;
3347                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3348                         goto out;
3349         } else {
3350                 link_net = NULL;
3351         }
3352
3353         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3354                                name_assign_type, ops, tb, extack);
3355         if (IS_ERR(dev)) {
3356                 err = PTR_ERR(dev);
3357                 goto out;
3358         }
3359
3360         dev->ifindex = ifm->ifi_index;
3361
3362         if (ops->newlink)
3363                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3364         else
3365                 err = register_netdevice(dev);
3366         if (err < 0) {
3367                 free_netdev(dev);
3368                 goto out;
3369         }
3370
3371         err = rtnl_configure_link(dev, ifm);
3372         if (err < 0)
3373                 goto out_unregister;
3374         if (link_net) {
3375                 err = dev_change_net_namespace(dev, dest_net, ifname);
3376                 if (err < 0)
3377                         goto out_unregister;
3378         }
3379         if (tb[IFLA_MASTER]) {
3380                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3381                 if (err)
3382                         goto out_unregister;
3383         }
3384 out:
3385         if (link_net)
3386                 put_net(link_net);
3387         put_net(dest_net);
3388         return err;
3389 out_unregister:
3390         if (ops->newlink) {
3391                 LIST_HEAD(list_kill);
3392
3393                 ops->dellink(dev, &list_kill);
3394                 unregister_netdevice_many(&list_kill);
3395         } else {
3396                 unregister_netdevice(dev);
3397         }
3398         goto out;
3399 }
3400
3401 struct rtnl_newlink_tbs {
3402         struct nlattr *tb[IFLA_MAX + 1];
3403         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3404         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3405 };
3406
3407 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3408                           struct rtnl_newlink_tbs *tbs,
3409                           struct netlink_ext_ack *extack)
3410 {
3411         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3412         struct nlattr ** const tb = tbs->tb;
3413         const struct rtnl_link_ops *m_ops;
3414         struct net_device *master_dev;
3415         struct net *net = sock_net(skb->sk);
3416         const struct rtnl_link_ops *ops;
3417         struct nlattr **slave_data;
3418         char kind[MODULE_NAME_LEN];
3419         struct net_device *dev;
3420         struct ifinfomsg *ifm;
3421         struct nlattr **data;
3422         bool link_specified;
3423         int err;
3424
3425 #ifdef CONFIG_MODULES
3426 replay:
3427 #endif
3428         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3429                                      ifla_policy, extack);
3430         if (err < 0)
3431                 return err;
3432
3433         err = rtnl_ensure_unique_netns(tb, extack, false);
3434         if (err < 0)
3435                 return err;
3436
3437         ifm = nlmsg_data(nlh);
3438         if (ifm->ifi_index > 0) {
3439                 link_specified = true;
3440                 dev = __dev_get_by_index(net, ifm->ifi_index);
3441         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3442                 link_specified = true;
3443                 dev = rtnl_dev_get(net, tb);
3444         } else {
3445                 link_specified = false;
3446                 dev = NULL;
3447         }
3448
3449         master_dev = NULL;
3450         m_ops = NULL;
3451         if (dev) {
3452                 master_dev = netdev_master_upper_dev_get(dev);
3453                 if (master_dev)
3454                         m_ops = master_dev->rtnl_link_ops;
3455         }
3456
3457         err = validate_linkmsg(dev, tb, extack);
3458         if (err < 0)
3459                 return err;
3460
3461         if (tb[IFLA_LINKINFO]) {
3462                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3463                                                   tb[IFLA_LINKINFO],
3464                                                   ifla_info_policy, NULL);
3465                 if (err < 0)
3466                         return err;
3467         } else
3468                 memset(linkinfo, 0, sizeof(linkinfo));
3469
3470         if (linkinfo[IFLA_INFO_KIND]) {
3471                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3472                 ops = rtnl_link_ops_get(kind);
3473         } else {
3474                 kind[0] = '\0';
3475                 ops = NULL;
3476         }
3477
3478         data = NULL;
3479         if (ops) {
3480                 if (ops->maxtype > RTNL_MAX_TYPE)
3481                         return -EINVAL;
3482
3483                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3484                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3485                                                           linkinfo[IFLA_INFO_DATA],
3486                                                           ops->policy, extack);
3487                         if (err < 0)
3488                                 return err;
3489                         data = tbs->attr;
3490                 }
3491                 if (ops->validate) {
3492                         err = ops->validate(tb, data, extack);
3493                         if (err < 0)
3494                                 return err;
3495                 }
3496         }
3497
3498         slave_data = NULL;
3499         if (m_ops) {
3500                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3501                         return -EINVAL;
3502
3503                 if (m_ops->slave_maxtype &&
3504                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3505                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3506                                                           m_ops->slave_maxtype,
3507                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3508                                                           m_ops->slave_policy,
3509                                                           extack);
3510                         if (err < 0)
3511                                 return err;
3512                         slave_data = tbs->slave_attr;
3513                 }
3514         }
3515
3516         if (dev) {
3517                 int status = 0;
3518
3519                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3520                         return -EEXIST;
3521                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3522                         return -EOPNOTSUPP;
3523
3524                 if (linkinfo[IFLA_INFO_DATA]) {
3525                         if (!ops || ops != dev->rtnl_link_ops ||
3526                             !ops->changelink)
3527                                 return -EOPNOTSUPP;
3528
3529                         err = ops->changelink(dev, tb, data, extack);
3530                         if (err < 0)
3531                                 return err;
3532                         status |= DO_SETLINK_NOTIFY;
3533                 }
3534
3535                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3536                         if (!m_ops || !m_ops->slave_changelink)
3537                                 return -EOPNOTSUPP;
3538
3539                         err = m_ops->slave_changelink(master_dev, dev, tb,
3540                                                       slave_data, extack);
3541                         if (err < 0)
3542                                 return err;
3543                         status |= DO_SETLINK_NOTIFY;
3544                 }
3545
3546                 return do_setlink(skb, dev, ifm, extack, tb, status);
3547         }
3548
3549         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3550                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3551                  * or it's for a group
3552                 */
3553                 if (link_specified)
3554                         return -ENODEV;
3555                 if (tb[IFLA_GROUP])
3556                         return rtnl_group_changelink(skb, net,
3557                                                 nla_get_u32(tb[IFLA_GROUP]),
3558                                                 ifm, extack, tb);
3559                 return -ENODEV;
3560         }
3561
3562         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3563                 return -EOPNOTSUPP;
3564
3565         if (!ops) {
3566 #ifdef CONFIG_MODULES
3567                 if (kind[0]) {
3568                         __rtnl_unlock();
3569                         request_module("rtnl-link-%s", kind);
3570                         rtnl_lock();
3571                         ops = rtnl_link_ops_get(kind);
3572                         if (ops)
3573                                 goto replay;
3574                 }
3575 #endif
3576                 NL_SET_ERR_MSG(extack, "Unknown device type");
3577                 return -EOPNOTSUPP;
3578         }
3579
3580         return rtnl_newlink_create(skb, ifm, ops, tb, data, extack);
3581 }
3582
3583 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3584                         struct netlink_ext_ack *extack)
3585 {
3586         struct rtnl_newlink_tbs *tbs;
3587         int ret;
3588
3589         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3590         if (!tbs)
3591                 return -ENOMEM;
3592
3593         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3594         kfree(tbs);
3595         return ret;
3596 }
3597
3598 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3599                                   const struct nlmsghdr *nlh,
3600                                   struct nlattr **tb,
3601                                   struct netlink_ext_ack *extack)
3602 {
3603         struct ifinfomsg *ifm;
3604         int i, err;
3605
3606         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3607                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3608                 return -EINVAL;
3609         }
3610
3611         if (!netlink_strict_get_check(skb))
3612                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3613                                               ifla_policy, extack);
3614
3615         ifm = nlmsg_data(nlh);
3616         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3617             ifm->ifi_change) {
3618                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3619                 return -EINVAL;
3620         }
3621
3622         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3623                                             ifla_policy, extack);
3624         if (err)
3625                 return err;
3626
3627         for (i = 0; i <= IFLA_MAX; i++) {
3628                 if (!tb[i])
3629                         continue;
3630
3631                 switch (i) {
3632                 case IFLA_IFNAME:
3633                 case IFLA_ALT_IFNAME:
3634                 case IFLA_EXT_MASK:
3635                 case IFLA_TARGET_NETNSID:
3636                         break;
3637                 default:
3638                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3639                         return -EINVAL;
3640                 }
3641         }
3642
3643         return 0;
3644 }
3645
3646 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3647                         struct netlink_ext_ack *extack)
3648 {
3649         struct net *net = sock_net(skb->sk);
3650         struct net *tgt_net = net;
3651         struct ifinfomsg *ifm;
3652         struct nlattr *tb[IFLA_MAX+1];
3653         struct net_device *dev = NULL;
3654         struct sk_buff *nskb;
3655         int netnsid = -1;
3656         int err;
3657         u32 ext_filter_mask = 0;
3658
3659         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3660         if (err < 0)
3661                 return err;
3662
3663         err = rtnl_ensure_unique_netns(tb, extack, true);
3664         if (err < 0)
3665                 return err;
3666
3667         if (tb[IFLA_TARGET_NETNSID]) {
3668                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3669                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3670                 if (IS_ERR(tgt_net))
3671                         return PTR_ERR(tgt_net);
3672         }
3673
3674         if (tb[IFLA_EXT_MASK])
3675                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3676
3677         err = -EINVAL;
3678         ifm = nlmsg_data(nlh);
3679         if (ifm->ifi_index > 0)
3680                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3681         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3682                 dev = rtnl_dev_get(tgt_net, tb);
3683         else
3684                 goto out;
3685
3686         err = -ENODEV;
3687         if (dev == NULL)
3688                 goto out;
3689
3690         err = -ENOBUFS;
3691         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3692         if (nskb == NULL)
3693                 goto out;
3694
3695         err = rtnl_fill_ifinfo(nskb, dev, net,
3696                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3697                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3698                                0, NULL, 0, netnsid, GFP_KERNEL);
3699         if (err < 0) {
3700                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3701                 WARN_ON(err == -EMSGSIZE);
3702                 kfree_skb(nskb);
3703         } else
3704                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3705 out:
3706         if (netnsid >= 0)
3707                 put_net(tgt_net);
3708
3709         return err;
3710 }
3711
3712 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3713                            bool *changed, struct netlink_ext_ack *extack)
3714 {
3715         char *alt_ifname;
3716         size_t size;
3717         int err;
3718
3719         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3720         if (err)
3721                 return err;
3722
3723         if (cmd == RTM_NEWLINKPROP) {
3724                 size = rtnl_prop_list_size(dev);
3725                 size += nla_total_size(ALTIFNAMSIZ);
3726                 if (size >= U16_MAX) {
3727                         NL_SET_ERR_MSG(extack,
3728                                        "effective property list too long");
3729                         return -EINVAL;
3730                 }
3731         }
3732
3733         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3734         if (!alt_ifname)
3735                 return -ENOMEM;
3736
3737         if (cmd == RTM_NEWLINKPROP) {
3738                 err = netdev_name_node_alt_create(dev, alt_ifname);
3739                 if (!err)
3740                         alt_ifname = NULL;
3741         } else if (cmd == RTM_DELLINKPROP) {
3742                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3743         } else {
3744                 WARN_ON_ONCE(1);
3745                 err = -EINVAL;
3746         }
3747
3748         kfree(alt_ifname);
3749         if (!err)
3750                 *changed = true;
3751         return err;
3752 }
3753
3754 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3755                          struct netlink_ext_ack *extack)
3756 {
3757         struct net *net = sock_net(skb->sk);
3758         struct nlattr *tb[IFLA_MAX + 1];
3759         struct net_device *dev;
3760         struct ifinfomsg *ifm;
3761         bool changed = false;
3762         struct nlattr *attr;
3763         int err, rem;
3764
3765         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3766         if (err)
3767                 return err;
3768
3769         err = rtnl_ensure_unique_netns(tb, extack, true);
3770         if (err)
3771                 return err;
3772
3773         ifm = nlmsg_data(nlh);
3774         if (ifm->ifi_index > 0)
3775                 dev = __dev_get_by_index(net, ifm->ifi_index);
3776         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3777                 dev = rtnl_dev_get(net, tb);
3778         else
3779                 return -EINVAL;
3780
3781         if (!dev)
3782                 return -ENODEV;
3783
3784         if (!tb[IFLA_PROP_LIST])
3785                 return 0;
3786
3787         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3788                 switch (nla_type(attr)) {
3789                 case IFLA_ALT_IFNAME:
3790                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3791                         if (err)
3792                                 return err;
3793                         break;
3794                 }
3795         }
3796
3797         if (changed)
3798                 netdev_state_change(dev);
3799         return 0;
3800 }
3801
3802 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3803                             struct netlink_ext_ack *extack)
3804 {
3805         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3806 }
3807
3808 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3809                             struct netlink_ext_ack *extack)
3810 {
3811         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3812 }
3813
3814 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3815 {
3816         struct net *net = sock_net(skb->sk);
3817         size_t min_ifinfo_dump_size = 0;
3818         struct nlattr *tb[IFLA_MAX+1];
3819         u32 ext_filter_mask = 0;
3820         struct net_device *dev;
3821         int hdrlen;
3822
3823         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3824         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3825                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3826
3827         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3828                 if (tb[IFLA_EXT_MASK])
3829                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3830         }
3831
3832         if (!ext_filter_mask)
3833                 return NLMSG_GOODSIZE;
3834         /*
3835          * traverse the list of net devices and compute the minimum
3836          * buffer size based upon the filter mask.
3837          */
3838         rcu_read_lock();
3839         for_each_netdev_rcu(net, dev) {
3840                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3841                                            if_nlmsg_size(dev, ext_filter_mask));
3842         }
3843         rcu_read_unlock();
3844
3845         return nlmsg_total_size(min_ifinfo_dump_size);
3846 }
3847
3848 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3849 {
3850         int idx;
3851         int s_idx = cb->family;
3852         int type = cb->nlh->nlmsg_type - RTM_BASE;
3853         int ret = 0;
3854
3855         if (s_idx == 0)
3856                 s_idx = 1;
3857
3858         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3859                 struct rtnl_link __rcu **tab;
3860                 struct rtnl_link *link;
3861                 rtnl_dumpit_func dumpit;
3862
3863                 if (idx < s_idx || idx == PF_PACKET)
3864                         continue;
3865
3866                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3867                         continue;
3868
3869                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3870                 if (!tab)
3871                         continue;
3872
3873                 link = rcu_dereference_rtnl(tab[type]);
3874                 if (!link)
3875                         continue;
3876
3877                 dumpit = link->dumpit;
3878                 if (!dumpit)
3879                         continue;
3880
3881                 if (idx > s_idx) {
3882                         memset(&cb->args[0], 0, sizeof(cb->args));
3883                         cb->prev_seq = 0;
3884                         cb->seq = 0;
3885                 }
3886                 ret = dumpit(skb, cb);
3887                 if (ret)
3888                         break;
3889         }
3890         cb->family = idx;
3891
3892         return skb->len ? : ret;
3893 }
3894
3895 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3896                                        unsigned int change,
3897                                        u32 event, gfp_t flags, int *new_nsid,
3898                                        int new_ifindex)
3899 {
3900         struct net *net = dev_net(dev);
3901         struct sk_buff *skb;
3902         int err = -ENOBUFS;
3903
3904         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3905         if (skb == NULL)
3906                 goto errout;
3907
3908         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3909                                type, 0, 0, change, 0, 0, event,
3910                                new_nsid, new_ifindex, -1, flags);
3911         if (err < 0) {
3912                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3913                 WARN_ON(err == -EMSGSIZE);
3914                 kfree_skb(skb);
3915                 goto errout;
3916         }
3917         return skb;
3918 errout:
3919         if (err < 0)
3920                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3921         return NULL;
3922 }
3923
3924 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3925 {
3926         struct net *net = dev_net(dev);
3927
3928         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3929 }
3930
3931 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3932                                unsigned int change, u32 event,
3933                                gfp_t flags, int *new_nsid, int new_ifindex)
3934 {
3935         struct sk_buff *skb;
3936
3937         if (dev->reg_state != NETREG_REGISTERED)
3938                 return;
3939
3940         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3941                                      new_ifindex);
3942         if (skb)
3943                 rtmsg_ifinfo_send(skb, dev, flags);
3944 }
3945
3946 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3947                   gfp_t flags)
3948 {
3949         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3950                            NULL, 0);
3951 }
3952
3953 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3954                          gfp_t flags, int *new_nsid, int new_ifindex)
3955 {
3956         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3957                            new_nsid, new_ifindex);
3958 }
3959
3960 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3961                                    struct net_device *dev,
3962                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3963                                    int type, unsigned int flags,
3964                                    int nlflags, u16 ndm_state)
3965 {
3966         struct nlmsghdr *nlh;
3967         struct ndmsg *ndm;
3968
3969         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3970         if (!nlh)
3971                 return -EMSGSIZE;
3972
3973         ndm = nlmsg_data(nlh);
3974         ndm->ndm_family  = AF_BRIDGE;
3975         ndm->ndm_pad1    = 0;
3976         ndm->ndm_pad2    = 0;
3977         ndm->ndm_flags   = flags;
3978         ndm->ndm_type    = 0;
3979         ndm->ndm_ifindex = dev->ifindex;
3980         ndm->ndm_state   = ndm_state;
3981
3982         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3983                 goto nla_put_failure;
3984         if (vid)
3985                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3986                         goto nla_put_failure;
3987
3988         nlmsg_end(skb, nlh);
3989         return 0;
3990
3991 nla_put_failure:
3992         nlmsg_cancel(skb, nlh);
3993         return -EMSGSIZE;
3994 }
3995
3996 static inline size_t rtnl_fdb_nlmsg_size(void)
3997 {
3998         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3999                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
4000                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4001                0;
4002 }
4003
4004 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4005                             u16 ndm_state)
4006 {
4007         struct net *net = dev_net(dev);
4008         struct sk_buff *skb;
4009         int err = -ENOBUFS;
4010
4011         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4012         if (!skb)
4013                 goto errout;
4014
4015         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4016                                       0, 0, type, NTF_SELF, 0, ndm_state);
4017         if (err < 0) {
4018                 kfree_skb(skb);
4019                 goto errout;
4020         }
4021
4022         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4023         return;
4024 errout:
4025         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4026 }
4027
4028 /*
4029  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4030  */
4031 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4032                      struct nlattr *tb[],
4033                      struct net_device *dev,
4034                      const unsigned char *addr, u16 vid,
4035                      u16 flags)
4036 {
4037         int err = -EINVAL;
4038
4039         /* If aging addresses are supported device will need to
4040          * implement its own handler for this.
4041          */
4042         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4043                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4044                 return err;
4045         }
4046
4047         if (vid) {
4048                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4049                 return err;
4050         }
4051
4052         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4053                 err = dev_uc_add_excl(dev, addr);
4054         else if (is_multicast_ether_addr(addr))
4055                 err = dev_mc_add_excl(dev, addr);
4056
4057         /* Only return duplicate errors if NLM_F_EXCL is set */
4058         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4059                 err = 0;
4060
4061         return err;
4062 }
4063 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4064
4065 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4066                          struct netlink_ext_ack *extack)
4067 {
4068         u16 vid = 0;
4069
4070         if (vlan_attr) {
4071                 if (nla_len(vlan_attr) != sizeof(u16)) {
4072                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4073                         return -EINVAL;
4074                 }
4075
4076                 vid = nla_get_u16(vlan_attr);
4077
4078                 if (!vid || vid >= VLAN_VID_MASK) {
4079                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4080                         return -EINVAL;
4081                 }
4082         }
4083         *p_vid = vid;
4084         return 0;
4085 }
4086
4087 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4088                         struct netlink_ext_ack *extack)
4089 {
4090         struct net *net = sock_net(skb->sk);
4091         struct ndmsg *ndm;
4092         struct nlattr *tb[NDA_MAX+1];
4093         struct net_device *dev;
4094         u8 *addr;
4095         u16 vid;
4096         int err;
4097
4098         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4099                                      extack);
4100         if (err < 0)
4101                 return err;
4102
4103         ndm = nlmsg_data(nlh);
4104         if (ndm->ndm_ifindex == 0) {
4105                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4106                 return -EINVAL;
4107         }
4108
4109         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4110         if (dev == NULL) {
4111                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4112                 return -ENODEV;
4113         }
4114
4115         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4116                 NL_SET_ERR_MSG(extack, "invalid address");
4117                 return -EINVAL;
4118         }
4119
4120         if (dev->type != ARPHRD_ETHER) {
4121                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4122                 return -EINVAL;
4123         }
4124
4125         addr = nla_data(tb[NDA_LLADDR]);
4126
4127         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4128         if (err)
4129                 return err;
4130
4131         err = -EOPNOTSUPP;
4132
4133         /* Support fdb on master device the net/bridge default case */
4134         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4135             netif_is_bridge_port(dev)) {
4136                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4137                 const struct net_device_ops *ops = br_dev->netdev_ops;
4138
4139                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4140                                        nlh->nlmsg_flags, extack);
4141                 if (err)
4142                         goto out;
4143                 else
4144                         ndm->ndm_flags &= ~NTF_MASTER;
4145         }
4146
4147         /* Embedded bridge, macvlan, and any other device support */
4148         if ((ndm->ndm_flags & NTF_SELF)) {
4149                 if (dev->netdev_ops->ndo_fdb_add)
4150                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4151                                                            vid,
4152                                                            nlh->nlmsg_flags,
4153                                                            extack);
4154                 else
4155                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4156                                                nlh->nlmsg_flags);
4157
4158                 if (!err) {
4159                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4160                                         ndm->ndm_state);
4161                         ndm->ndm_flags &= ~NTF_SELF;
4162                 }
4163         }
4164 out:
4165         return err;
4166 }
4167
4168 /*
4169  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4170  */
4171 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4172                      struct nlattr *tb[],
4173                      struct net_device *dev,
4174                      const unsigned char *addr, u16 vid)
4175 {
4176         int err = -EINVAL;
4177
4178         /* If aging addresses are supported device will need to
4179          * implement its own handler for this.
4180          */
4181         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4182                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4183                 return err;
4184         }
4185
4186         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4187                 err = dev_uc_del(dev, addr);
4188         else if (is_multicast_ether_addr(addr))
4189                 err = dev_mc_del(dev, addr);
4190
4191         return err;
4192 }
4193 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4194
4195 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4196         [NDA_VLAN]      = { .type = NLA_U16 },
4197         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4198         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4199         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4200 };
4201
4202 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4203                         struct netlink_ext_ack *extack)
4204 {
4205         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4206         struct net *net = sock_net(skb->sk);
4207         const struct net_device_ops *ops;
4208         struct ndmsg *ndm;
4209         struct nlattr *tb[NDA_MAX+1];
4210         struct net_device *dev;
4211         __u8 *addr = NULL;
4212         int err;
4213         u16 vid;
4214
4215         if (!netlink_capable(skb, CAP_NET_ADMIN))
4216                 return -EPERM;
4217
4218         if (!del_bulk) {
4219                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4220                                              NULL, extack);
4221         } else {
4222                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4223                                   fdb_del_bulk_policy, extack);
4224         }
4225         if (err < 0)
4226                 return err;
4227
4228         ndm = nlmsg_data(nlh);
4229         if (ndm->ndm_ifindex == 0) {
4230                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4231                 return -EINVAL;
4232         }
4233
4234         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4235         if (dev == NULL) {
4236                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4237                 return -ENODEV;
4238         }
4239
4240         if (!del_bulk) {
4241                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4242                         NL_SET_ERR_MSG(extack, "invalid address");
4243                         return -EINVAL;
4244                 }
4245                 addr = nla_data(tb[NDA_LLADDR]);
4246         }
4247
4248         if (dev->type != ARPHRD_ETHER) {
4249                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4250                 return -EINVAL;
4251         }
4252
4253         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4254         if (err)
4255                 return err;
4256
4257         err = -EOPNOTSUPP;
4258
4259         /* Support fdb on master device the net/bridge default case */
4260         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4261             netif_is_bridge_port(dev)) {
4262                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4263
4264                 ops = br_dev->netdev_ops;
4265                 if (!del_bulk) {
4266                         if (ops->ndo_fdb_del)
4267                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4268                 } else {
4269                         if (ops->ndo_fdb_del_bulk)
4270                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4271                                                             extack);
4272                 }
4273
4274                 if (err)
4275                         goto out;
4276                 else
4277                         ndm->ndm_flags &= ~NTF_MASTER;
4278         }
4279
4280         /* Embedded bridge, macvlan, and any other device support */
4281         if (ndm->ndm_flags & NTF_SELF) {
4282                 ops = dev->netdev_ops;
4283                 if (!del_bulk) {
4284                         if (ops->ndo_fdb_del)
4285                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4286                         else
4287                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4288                 } else {
4289                         /* in case err was cleared by NTF_MASTER call */
4290                         err = -EOPNOTSUPP;
4291                         if (ops->ndo_fdb_del_bulk)
4292                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4293                                                             extack);
4294                 }
4295
4296                 if (!err) {
4297                         if (!del_bulk)
4298                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4299                                                 ndm->ndm_state);
4300                         ndm->ndm_flags &= ~NTF_SELF;
4301                 }
4302         }
4303 out:
4304         return err;
4305 }
4306
4307 static int nlmsg_populate_fdb(struct sk_buff *skb,
4308                               struct netlink_callback *cb,
4309                               struct net_device *dev,
4310                               int *idx,
4311                               struct netdev_hw_addr_list *list)
4312 {
4313         struct netdev_hw_addr *ha;
4314         int err;
4315         u32 portid, seq;
4316
4317         portid = NETLINK_CB(cb->skb).portid;
4318         seq = cb->nlh->nlmsg_seq;
4319
4320         list_for_each_entry(ha, &list->list, list) {
4321                 if (*idx < cb->args[2])
4322                         goto skip;
4323
4324                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4325                                               portid, seq,
4326                                               RTM_NEWNEIGH, NTF_SELF,
4327                                               NLM_F_MULTI, NUD_PERMANENT);
4328                 if (err < 0)
4329                         return err;
4330 skip:
4331                 *idx += 1;
4332         }
4333         return 0;
4334 }
4335
4336 /**
4337  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4338  * @skb: socket buffer to store message in
4339  * @cb: netlink callback
4340  * @dev: netdevice
4341  * @filter_dev: ignored
4342  * @idx: the number of FDB table entries dumped is added to *@idx
4343  *
4344  * Default netdevice operation to dump the existing unicast address list.
4345  * Returns number of addresses from list put in skb.
4346  */
4347 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4348                       struct netlink_callback *cb,
4349                       struct net_device *dev,
4350                       struct net_device *filter_dev,
4351                       int *idx)
4352 {
4353         int err;
4354
4355         if (dev->type != ARPHRD_ETHER)
4356                 return -EINVAL;
4357
4358         netif_addr_lock_bh(dev);
4359         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4360         if (err)
4361                 goto out;
4362         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4363 out:
4364         netif_addr_unlock_bh(dev);
4365         return err;
4366 }
4367 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4368
4369 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4370                                  int *br_idx, int *brport_idx,
4371                                  struct netlink_ext_ack *extack)
4372 {
4373         struct nlattr *tb[NDA_MAX + 1];
4374         struct ndmsg *ndm;
4375         int err, i;
4376
4377         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4378                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4379                 return -EINVAL;
4380         }
4381
4382         ndm = nlmsg_data(nlh);
4383         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4384             ndm->ndm_flags || ndm->ndm_type) {
4385                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4386                 return -EINVAL;
4387         }
4388
4389         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4390                                             NDA_MAX, NULL, extack);
4391         if (err < 0)
4392                 return err;
4393
4394         *brport_idx = ndm->ndm_ifindex;
4395         for (i = 0; i <= NDA_MAX; ++i) {
4396                 if (!tb[i])
4397                         continue;
4398
4399                 switch (i) {
4400                 case NDA_IFINDEX:
4401                         if (nla_len(tb[i]) != sizeof(u32)) {
4402                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4403                                 return -EINVAL;
4404                         }
4405                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4406                         break;
4407                 case NDA_MASTER:
4408                         if (nla_len(tb[i]) != sizeof(u32)) {
4409                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4410                                 return -EINVAL;
4411                         }
4412                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4413                         break;
4414                 default:
4415                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4416                         return -EINVAL;
4417                 }
4418         }
4419
4420         return 0;
4421 }
4422
4423 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4424                                  int *br_idx, int *brport_idx,
4425                                  struct netlink_ext_ack *extack)
4426 {
4427         struct nlattr *tb[IFLA_MAX+1];
4428         int err;
4429
4430         /* A hack to preserve kernel<->userspace interface.
4431          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4432          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4433          * So, check for ndmsg with an optional u32 attribute (not used here).
4434          * Fortunately these sizes don't conflict with the size of ifinfomsg
4435          * with an optional attribute.
4436          */
4437         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4438             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4439              nla_attr_size(sizeof(u32)))) {
4440                 struct ifinfomsg *ifm;
4441
4442                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4443                                              tb, IFLA_MAX, ifla_policy,
4444                                              extack);
4445                 if (err < 0) {
4446                         return -EINVAL;
4447                 } else if (err == 0) {
4448                         if (tb[IFLA_MASTER])
4449                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4450                 }
4451
4452                 ifm = nlmsg_data(nlh);
4453                 *brport_idx = ifm->ifi_index;
4454         }
4455         return 0;
4456 }
4457
4458 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4459 {
4460         struct net_device *dev;
4461         struct net_device *br_dev = NULL;
4462         const struct net_device_ops *ops = NULL;
4463         const struct net_device_ops *cops = NULL;
4464         struct net *net = sock_net(skb->sk);
4465         struct hlist_head *head;
4466         int brport_idx = 0;
4467         int br_idx = 0;
4468         int h, s_h;
4469         int idx = 0, s_idx;
4470         int err = 0;
4471         int fidx = 0;
4472
4473         if (cb->strict_check)
4474                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4475                                             cb->extack);
4476         else
4477                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4478                                             cb->extack);
4479         if (err < 0)
4480                 return err;
4481
4482         if (br_idx) {
4483                 br_dev = __dev_get_by_index(net, br_idx);
4484                 if (!br_dev)
4485                         return -ENODEV;
4486
4487                 ops = br_dev->netdev_ops;
4488         }
4489
4490         s_h = cb->args[0];
4491         s_idx = cb->args[1];
4492
4493         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4494                 idx = 0;
4495                 head = &net->dev_index_head[h];
4496                 hlist_for_each_entry(dev, head, index_hlist) {
4497
4498                         if (brport_idx && (dev->ifindex != brport_idx))
4499                                 continue;
4500
4501                         if (!br_idx) { /* user did not specify a specific bridge */
4502                                 if (netif_is_bridge_port(dev)) {
4503                                         br_dev = netdev_master_upper_dev_get(dev);
4504                                         cops = br_dev->netdev_ops;
4505                                 }
4506                         } else {
4507                                 if (dev != br_dev &&
4508                                     !netif_is_bridge_port(dev))
4509                                         continue;
4510
4511                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4512                                     !netif_is_bridge_master(dev))
4513                                         continue;
4514                                 cops = ops;
4515                         }
4516
4517                         if (idx < s_idx)
4518                                 goto cont;
4519
4520                         if (netif_is_bridge_port(dev)) {
4521                                 if (cops && cops->ndo_fdb_dump) {
4522                                         err = cops->ndo_fdb_dump(skb, cb,
4523                                                                 br_dev, dev,
4524                                                                 &fidx);
4525                                         if (err == -EMSGSIZE)
4526                                                 goto out;
4527                                 }
4528                         }
4529
4530                         if (dev->netdev_ops->ndo_fdb_dump)
4531                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4532                                                                     dev, NULL,
4533                                                                     &fidx);
4534                         else
4535                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4536                                                         &fidx);
4537                         if (err == -EMSGSIZE)
4538                                 goto out;
4539
4540                         cops = NULL;
4541
4542                         /* reset fdb offset to 0 for rest of the interfaces */
4543                         cb->args[2] = 0;
4544                         fidx = 0;
4545 cont:
4546                         idx++;
4547                 }
4548         }
4549
4550 out:
4551         cb->args[0] = h;
4552         cb->args[1] = idx;
4553         cb->args[2] = fidx;
4554
4555         return skb->len;
4556 }
4557
4558 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4559                                 struct nlattr **tb, u8 *ndm_flags,
4560                                 int *br_idx, int *brport_idx, u8 **addr,
4561                                 u16 *vid, struct netlink_ext_ack *extack)
4562 {
4563         struct ndmsg *ndm;
4564         int err, i;
4565
4566         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4567                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4568                 return -EINVAL;
4569         }
4570
4571         ndm = nlmsg_data(nlh);
4572         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4573             ndm->ndm_type) {
4574                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4575                 return -EINVAL;
4576         }
4577
4578         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4579                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4580                 return -EINVAL;
4581         }
4582
4583         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4584                                             NDA_MAX, nda_policy, extack);
4585         if (err < 0)
4586                 return err;
4587
4588         *ndm_flags = ndm->ndm_flags;
4589         *brport_idx = ndm->ndm_ifindex;
4590         for (i = 0; i <= NDA_MAX; ++i) {
4591                 if (!tb[i])
4592                         continue;
4593
4594                 switch (i) {
4595                 case NDA_MASTER:
4596                         *br_idx = nla_get_u32(tb[i]);
4597                         break;
4598                 case NDA_LLADDR:
4599                         if (nla_len(tb[i]) != ETH_ALEN) {
4600                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4601                                 return -EINVAL;
4602                         }
4603                         *addr = nla_data(tb[i]);
4604                         break;
4605                 case NDA_VLAN:
4606                         err = fdb_vid_parse(tb[i], vid, extack);
4607                         if (err)
4608                                 return err;
4609                         break;
4610                 case NDA_VNI:
4611                         break;
4612                 default:
4613                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4614                         return -EINVAL;
4615                 }
4616         }
4617
4618         return 0;
4619 }
4620
4621 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4622                         struct netlink_ext_ack *extack)
4623 {
4624         struct net_device *dev = NULL, *br_dev = NULL;
4625         const struct net_device_ops *ops = NULL;
4626         struct net *net = sock_net(in_skb->sk);
4627         struct nlattr *tb[NDA_MAX + 1];
4628         struct sk_buff *skb;
4629         int brport_idx = 0;
4630         u8 ndm_flags = 0;
4631         int br_idx = 0;
4632         u8 *addr = NULL;
4633         u16 vid = 0;
4634         int err;
4635
4636         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4637                                    &brport_idx, &addr, &vid, extack);
4638         if (err < 0)
4639                 return err;
4640
4641         if (!addr) {
4642                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4643                 return -EINVAL;
4644         }
4645
4646         if (brport_idx) {
4647                 dev = __dev_get_by_index(net, brport_idx);
4648                 if (!dev) {
4649                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4650                         return -ENODEV;
4651                 }
4652         }
4653
4654         if (br_idx) {
4655                 if (dev) {
4656                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4657                         return -EINVAL;
4658                 }
4659
4660                 br_dev = __dev_get_by_index(net, br_idx);
4661                 if (!br_dev) {
4662                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4663                         return -EINVAL;
4664                 }
4665                 ops = br_dev->netdev_ops;
4666         }
4667
4668         if (dev) {
4669                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4670                         if (!netif_is_bridge_port(dev)) {
4671                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4672                                 return -EINVAL;
4673                         }
4674                         br_dev = netdev_master_upper_dev_get(dev);
4675                         if (!br_dev) {
4676                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4677                                 return -EINVAL;
4678                         }
4679                         ops = br_dev->netdev_ops;
4680                 } else {
4681                         if (!(ndm_flags & NTF_SELF)) {
4682                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4683                                 return -EINVAL;
4684                         }
4685                         ops = dev->netdev_ops;
4686                 }
4687         }
4688
4689         if (!br_dev && !dev) {
4690                 NL_SET_ERR_MSG(extack, "No device specified");
4691                 return -ENODEV;
4692         }
4693
4694         if (!ops || !ops->ndo_fdb_get) {
4695                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4696                 return -EOPNOTSUPP;
4697         }
4698
4699         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4700         if (!skb)
4701                 return -ENOBUFS;
4702
4703         if (br_dev)
4704                 dev = br_dev;
4705         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4706                                NETLINK_CB(in_skb).portid,
4707                                nlh->nlmsg_seq, extack);
4708         if (err)
4709                 goto out;
4710
4711         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4712 out:
4713         kfree_skb(skb);
4714         return err;
4715 }
4716
4717 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4718                                unsigned int attrnum, unsigned int flag)
4719 {
4720         if (mask & flag)
4721                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4722         return 0;
4723 }
4724
4725 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4726                             struct net_device *dev, u16 mode,
4727                             u32 flags, u32 mask, int nlflags,
4728                             u32 filter_mask,
4729                             int (*vlan_fill)(struct sk_buff *skb,
4730                                              struct net_device *dev,
4731                                              u32 filter_mask))
4732 {
4733         struct nlmsghdr *nlh;
4734         struct ifinfomsg *ifm;
4735         struct nlattr *br_afspec;
4736         struct nlattr *protinfo;
4737         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4738         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4739         int err = 0;
4740
4741         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4742         if (nlh == NULL)
4743                 return -EMSGSIZE;
4744
4745         ifm = nlmsg_data(nlh);
4746         ifm->ifi_family = AF_BRIDGE;
4747         ifm->__ifi_pad = 0;
4748         ifm->ifi_type = dev->type;
4749         ifm->ifi_index = dev->ifindex;
4750         ifm->ifi_flags = dev_get_flags(dev);
4751         ifm->ifi_change = 0;
4752
4753
4754         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4755             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4756             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4757             (br_dev &&
4758              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4759             (dev->addr_len &&
4760              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4761             (dev->ifindex != dev_get_iflink(dev) &&
4762              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4763                 goto nla_put_failure;
4764
4765         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4766         if (!br_afspec)
4767                 goto nla_put_failure;
4768
4769         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4770                 nla_nest_cancel(skb, br_afspec);
4771                 goto nla_put_failure;
4772         }
4773
4774         if (mode != BRIDGE_MODE_UNDEF) {
4775                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4776                         nla_nest_cancel(skb, br_afspec);
4777                         goto nla_put_failure;
4778                 }
4779         }
4780         if (vlan_fill) {
4781                 err = vlan_fill(skb, dev, filter_mask);
4782                 if (err) {
4783                         nla_nest_cancel(skb, br_afspec);
4784                         goto nla_put_failure;
4785                 }
4786         }
4787         nla_nest_end(skb, br_afspec);
4788
4789         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4790         if (!protinfo)
4791                 goto nla_put_failure;
4792
4793         if (brport_nla_put_flag(skb, flags, mask,
4794                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4795             brport_nla_put_flag(skb, flags, mask,
4796                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4797             brport_nla_put_flag(skb, flags, mask,
4798                                 IFLA_BRPORT_FAST_LEAVE,
4799                                 BR_MULTICAST_FAST_LEAVE) ||
4800             brport_nla_put_flag(skb, flags, mask,
4801                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4802             brport_nla_put_flag(skb, flags, mask,
4803                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4804             brport_nla_put_flag(skb, flags, mask,
4805                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4806             brport_nla_put_flag(skb, flags, mask,
4807                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4808             brport_nla_put_flag(skb, flags, mask,
4809                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4810             brport_nla_put_flag(skb, flags, mask,
4811                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4812             brport_nla_put_flag(skb, flags, mask,
4813                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4814                 nla_nest_cancel(skb, protinfo);
4815                 goto nla_put_failure;
4816         }
4817
4818         nla_nest_end(skb, protinfo);
4819
4820         nlmsg_end(skb, nlh);
4821         return 0;
4822 nla_put_failure:
4823         nlmsg_cancel(skb, nlh);
4824         return err ? err : -EMSGSIZE;
4825 }
4826 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4827
4828 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4829                                     bool strict_check, u32 *filter_mask,
4830                                     struct netlink_ext_ack *extack)
4831 {
4832         struct nlattr *tb[IFLA_MAX+1];
4833         int err, i;
4834
4835         if (strict_check) {
4836                 struct ifinfomsg *ifm;
4837
4838                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4839                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4840                         return -EINVAL;
4841                 }
4842
4843                 ifm = nlmsg_data(nlh);
4844                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4845                     ifm->ifi_change || ifm->ifi_index) {
4846                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4847                         return -EINVAL;
4848                 }
4849
4850                 err = nlmsg_parse_deprecated_strict(nlh,
4851                                                     sizeof(struct ifinfomsg),
4852                                                     tb, IFLA_MAX, ifla_policy,
4853                                                     extack);
4854         } else {
4855                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4856                                              tb, IFLA_MAX, ifla_policy,
4857                                              extack);
4858         }
4859         if (err < 0)
4860                 return err;
4861
4862         /* new attributes should only be added with strict checking */
4863         for (i = 0; i <= IFLA_MAX; ++i) {
4864                 if (!tb[i])
4865                         continue;
4866
4867                 switch (i) {
4868                 case IFLA_EXT_MASK:
4869                         *filter_mask = nla_get_u32(tb[i]);
4870                         break;
4871                 default:
4872                         if (strict_check) {
4873                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4874                                 return -EINVAL;
4875                         }
4876                 }
4877         }
4878
4879         return 0;
4880 }
4881
4882 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4883 {
4884         const struct nlmsghdr *nlh = cb->nlh;
4885         struct net *net = sock_net(skb->sk);
4886         struct net_device *dev;
4887         int idx = 0;
4888         u32 portid = NETLINK_CB(cb->skb).portid;
4889         u32 seq = nlh->nlmsg_seq;
4890         u32 filter_mask = 0;
4891         int err;
4892
4893         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4894                                        cb->extack);
4895         if (err < 0 && cb->strict_check)
4896                 return err;
4897
4898         rcu_read_lock();
4899         for_each_netdev_rcu(net, dev) {
4900                 const struct net_device_ops *ops = dev->netdev_ops;
4901                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4902
4903                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4904                         if (idx >= cb->args[0]) {
4905                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4906                                                 skb, portid, seq, dev,
4907                                                 filter_mask, NLM_F_MULTI);
4908                                 if (err < 0 && err != -EOPNOTSUPP) {
4909                                         if (likely(skb->len))
4910                                                 break;
4911
4912                                         goto out_err;
4913                                 }
4914                         }
4915                         idx++;
4916                 }
4917
4918                 if (ops->ndo_bridge_getlink) {
4919                         if (idx >= cb->args[0]) {
4920                                 err = ops->ndo_bridge_getlink(skb, portid,
4921                                                               seq, dev,
4922                                                               filter_mask,
4923                                                               NLM_F_MULTI);
4924                                 if (err < 0 && err != -EOPNOTSUPP) {
4925                                         if (likely(skb->len))
4926                                                 break;
4927
4928                                         goto out_err;
4929                                 }
4930                         }
4931                         idx++;
4932                 }
4933         }
4934         err = skb->len;
4935 out_err:
4936         rcu_read_unlock();
4937         cb->args[0] = idx;
4938
4939         return err;
4940 }
4941
4942 static inline size_t bridge_nlmsg_size(void)
4943 {
4944         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4945                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4946                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4947                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4948                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4949                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4950                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4951                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4952                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4953                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4954                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4955 }
4956
4957 static int rtnl_bridge_notify(struct net_device *dev)
4958 {
4959         struct net *net = dev_net(dev);
4960         struct sk_buff *skb;
4961         int err = -EOPNOTSUPP;
4962
4963         if (!dev->netdev_ops->ndo_bridge_getlink)
4964                 return 0;
4965
4966         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4967         if (!skb) {
4968                 err = -ENOMEM;
4969                 goto errout;
4970         }
4971
4972         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4973         if (err < 0)
4974                 goto errout;
4975
4976         /* Notification info is only filled for bridge ports, not the bridge
4977          * device itself. Therefore, a zero notification length is valid and
4978          * should not result in an error.
4979          */
4980         if (!skb->len)
4981                 goto errout;
4982
4983         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4984         return 0;
4985 errout:
4986         WARN_ON(err == -EMSGSIZE);
4987         kfree_skb(skb);
4988         if (err)
4989                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4990         return err;
4991 }
4992
4993 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4994                                struct netlink_ext_ack *extack)
4995 {
4996         struct net *net = sock_net(skb->sk);
4997         struct ifinfomsg *ifm;
4998         struct net_device *dev;
4999         struct nlattr *br_spec, *attr = NULL;
5000         int rem, err = -EOPNOTSUPP;
5001         u16 flags = 0;
5002         bool have_flags = false;
5003
5004         if (nlmsg_len(nlh) < sizeof(*ifm))
5005                 return -EINVAL;
5006
5007         ifm = nlmsg_data(nlh);
5008         if (ifm->ifi_family != AF_BRIDGE)
5009                 return -EPFNOSUPPORT;
5010
5011         dev = __dev_get_by_index(net, ifm->ifi_index);
5012         if (!dev) {
5013                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5014                 return -ENODEV;
5015         }
5016
5017         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5018         if (br_spec) {
5019                 nla_for_each_nested(attr, br_spec, rem) {
5020                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5021                                 if (nla_len(attr) < sizeof(flags))
5022                                         return -EINVAL;
5023
5024                                 have_flags = true;
5025                                 flags = nla_get_u16(attr);
5026                                 break;
5027                         }
5028                 }
5029         }
5030
5031         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5032                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5033
5034                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5035                         err = -EOPNOTSUPP;
5036                         goto out;
5037                 }
5038
5039                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5040                                                              extack);
5041                 if (err)
5042                         goto out;
5043
5044                 flags &= ~BRIDGE_FLAGS_MASTER;
5045         }
5046
5047         if ((flags & BRIDGE_FLAGS_SELF)) {
5048                 if (!dev->netdev_ops->ndo_bridge_setlink)
5049                         err = -EOPNOTSUPP;
5050                 else
5051                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5052                                                                   flags,
5053                                                                   extack);
5054                 if (!err) {
5055                         flags &= ~BRIDGE_FLAGS_SELF;
5056
5057                         /* Generate event to notify upper layer of bridge
5058                          * change
5059                          */
5060                         err = rtnl_bridge_notify(dev);
5061                 }
5062         }
5063
5064         if (have_flags)
5065                 memcpy(nla_data(attr), &flags, sizeof(flags));
5066 out:
5067         return err;
5068 }
5069
5070 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5071                                struct netlink_ext_ack *extack)
5072 {
5073         struct net *net = sock_net(skb->sk);
5074         struct ifinfomsg *ifm;
5075         struct net_device *dev;
5076         struct nlattr *br_spec, *attr = NULL;
5077         int rem, err = -EOPNOTSUPP;
5078         u16 flags = 0;
5079         bool have_flags = false;
5080
5081         if (nlmsg_len(nlh) < sizeof(*ifm))
5082                 return -EINVAL;
5083
5084         ifm = nlmsg_data(nlh);
5085         if (ifm->ifi_family != AF_BRIDGE)
5086                 return -EPFNOSUPPORT;
5087
5088         dev = __dev_get_by_index(net, ifm->ifi_index);
5089         if (!dev) {
5090                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5091                 return -ENODEV;
5092         }
5093
5094         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5095         if (br_spec) {
5096                 nla_for_each_nested(attr, br_spec, rem) {
5097                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5098                                 if (nla_len(attr) < sizeof(flags))
5099                                         return -EINVAL;
5100
5101                                 have_flags = true;
5102                                 flags = nla_get_u16(attr);
5103                                 break;
5104                         }
5105                 }
5106         }
5107
5108         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5109                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5110
5111                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5112                         err = -EOPNOTSUPP;
5113                         goto out;
5114                 }
5115
5116                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5117                 if (err)
5118                         goto out;
5119
5120                 flags &= ~BRIDGE_FLAGS_MASTER;
5121         }
5122
5123         if ((flags & BRIDGE_FLAGS_SELF)) {
5124                 if (!dev->netdev_ops->ndo_bridge_dellink)
5125                         err = -EOPNOTSUPP;
5126                 else
5127                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5128                                                                   flags);
5129
5130                 if (!err) {
5131                         flags &= ~BRIDGE_FLAGS_SELF;
5132
5133                         /* Generate event to notify upper layer of bridge
5134                          * change
5135                          */
5136                         err = rtnl_bridge_notify(dev);
5137                 }
5138         }
5139
5140         if (have_flags)
5141                 memcpy(nla_data(attr), &flags, sizeof(flags));
5142 out:
5143         return err;
5144 }
5145
5146 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5147 {
5148         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5149                (!idxattr || idxattr == attrid);
5150 }
5151
5152 static bool
5153 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5154 {
5155         return dev->netdev_ops &&
5156                dev->netdev_ops->ndo_has_offload_stats &&
5157                dev->netdev_ops->ndo_get_offload_stats &&
5158                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5159 }
5160
5161 static unsigned int
5162 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5163 {
5164         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5165                sizeof(struct rtnl_link_stats64) : 0;
5166 }
5167
5168 static int
5169 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5170                              struct sk_buff *skb)
5171 {
5172         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5173         struct nlattr *attr = NULL;
5174         void *attr_data;
5175         int err;
5176
5177         if (!size)
5178                 return -ENODATA;
5179
5180         attr = nla_reserve_64bit(skb, attr_id, size,
5181                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5182         if (!attr)
5183                 return -EMSGSIZE;
5184
5185         attr_data = nla_data(attr);
5186         memset(attr_data, 0, size);
5187
5188         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5189         if (err)
5190                 return err;
5191
5192         return 0;
5193 }
5194
5195 static unsigned int
5196 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5197                                    enum netdev_offload_xstats_type type)
5198 {
5199         bool enabled = netdev_offload_xstats_enabled(dev, type);
5200
5201         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5202 }
5203
5204 struct rtnl_offload_xstats_request_used {
5205         bool request;
5206         bool used;
5207 };
5208
5209 static int
5210 rtnl_offload_xstats_get_stats(struct net_device *dev,
5211                               enum netdev_offload_xstats_type type,
5212                               struct rtnl_offload_xstats_request_used *ru,
5213                               struct rtnl_hw_stats64 *stats,
5214                               struct netlink_ext_ack *extack)
5215 {
5216         bool request;
5217         bool used;
5218         int err;
5219
5220         request = netdev_offload_xstats_enabled(dev, type);
5221         if (!request) {
5222                 used = false;
5223                 goto out;
5224         }
5225
5226         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5227         if (err)
5228                 return err;
5229
5230 out:
5231         if (ru) {
5232                 ru->request = request;
5233                 ru->used = used;
5234         }
5235         return 0;
5236 }
5237
5238 static int
5239 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5240                                        struct rtnl_offload_xstats_request_used *ru)
5241 {
5242         struct nlattr *nest;
5243
5244         nest = nla_nest_start(skb, attr_id);
5245         if (!nest)
5246                 return -EMSGSIZE;
5247
5248         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5249                 goto nla_put_failure;
5250
5251         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5252                 goto nla_put_failure;
5253
5254         nla_nest_end(skb, nest);
5255         return 0;
5256
5257 nla_put_failure:
5258         nla_nest_cancel(skb, nest);
5259         return -EMSGSIZE;
5260 }
5261
5262 static int
5263 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5264                                    struct netlink_ext_ack *extack)
5265 {
5266         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5267         struct rtnl_offload_xstats_request_used ru_l3;
5268         struct nlattr *nest;
5269         int err;
5270
5271         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5272         if (err)
5273                 return err;
5274
5275         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5276         if (!nest)
5277                 return -EMSGSIZE;
5278
5279         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5280                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5281                                                    &ru_l3))
5282                 goto nla_put_failure;
5283
5284         nla_nest_end(skb, nest);
5285         return 0;
5286
5287 nla_put_failure:
5288         nla_nest_cancel(skb, nest);
5289         return -EMSGSIZE;
5290 }
5291
5292 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5293                                     int *prividx, u32 off_filter_mask,
5294                                     struct netlink_ext_ack *extack)
5295 {
5296         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5297         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5298         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5299         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5300         bool have_data = false;
5301         int err;
5302
5303         if (*prividx <= attr_id_cpu_hit &&
5304             (off_filter_mask &
5305              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5306                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5307                 if (!err) {
5308                         have_data = true;
5309                 } else if (err != -ENODATA) {
5310                         *prividx = attr_id_cpu_hit;
5311                         return err;
5312                 }
5313         }
5314
5315         if (*prividx <= attr_id_hw_s_info &&
5316             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5317                 *prividx = attr_id_hw_s_info;
5318
5319                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5320                 if (err)
5321                         return err;
5322
5323                 have_data = true;
5324                 *prividx = 0;
5325         }
5326
5327         if (*prividx <= attr_id_l3_stats &&
5328             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5329                 unsigned int size_l3;
5330                 struct nlattr *attr;
5331
5332                 *prividx = attr_id_l3_stats;
5333
5334                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5335                 if (!size_l3)
5336                         goto skip_l3_stats;
5337                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5338                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5339                 if (!attr)
5340                         return -EMSGSIZE;
5341
5342                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5343                                                     nla_data(attr), extack);
5344                 if (err)
5345                         return err;
5346
5347                 have_data = true;
5348 skip_l3_stats:
5349                 *prividx = 0;
5350         }
5351
5352         if (!have_data)
5353                 return -ENODATA;
5354
5355         *prividx = 0;
5356         return 0;
5357 }
5358
5359 static unsigned int
5360 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5361                                            enum netdev_offload_xstats_type type)
5362 {
5363         bool enabled = netdev_offload_xstats_enabled(dev, type);
5364
5365         return nla_total_size(0) +
5366                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5367                 nla_total_size(sizeof(u8)) +
5368                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5369                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5370                 0;
5371 }
5372
5373 static unsigned int
5374 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5375 {
5376         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5377
5378         return nla_total_size(0) +
5379                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5380                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5381                 0;
5382 }
5383
5384 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5385                                         u32 off_filter_mask)
5386 {
5387         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5388         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5389         int nla_size = 0;
5390         int size;
5391
5392         if (off_filter_mask &
5393             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5394                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5395                 nla_size += nla_total_size_64bit(size);
5396         }
5397
5398         if (off_filter_mask &
5399             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5400                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5401
5402         if (off_filter_mask &
5403             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5404                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5405                 nla_size += nla_total_size_64bit(size);
5406         }
5407
5408         if (nla_size != 0)
5409                 nla_size += nla_total_size(0);
5410
5411         return nla_size;
5412 }
5413
5414 struct rtnl_stats_dump_filters {
5415         /* mask[0] filters outer attributes. Then individual nests have their
5416          * filtering mask at the index of the nested attribute.
5417          */
5418         u32 mask[IFLA_STATS_MAX + 1];
5419 };
5420
5421 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5422                                int type, u32 pid, u32 seq, u32 change,
5423                                unsigned int flags,
5424                                const struct rtnl_stats_dump_filters *filters,
5425                                int *idxattr, int *prividx,
5426                                struct netlink_ext_ack *extack)
5427 {
5428         unsigned int filter_mask = filters->mask[0];
5429         struct if_stats_msg *ifsm;
5430         struct nlmsghdr *nlh;
5431         struct nlattr *attr;
5432         int s_prividx = *prividx;
5433         int err;
5434
5435         ASSERT_RTNL();
5436
5437         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5438         if (!nlh)
5439                 return -EMSGSIZE;
5440
5441         ifsm = nlmsg_data(nlh);
5442         ifsm->family = PF_UNSPEC;
5443         ifsm->pad1 = 0;
5444         ifsm->pad2 = 0;
5445         ifsm->ifindex = dev->ifindex;
5446         ifsm->filter_mask = filter_mask;
5447
5448         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5449                 struct rtnl_link_stats64 *sp;
5450
5451                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5452                                          sizeof(struct rtnl_link_stats64),
5453                                          IFLA_STATS_UNSPEC);
5454                 if (!attr) {
5455                         err = -EMSGSIZE;
5456                         goto nla_put_failure;
5457                 }
5458
5459                 sp = nla_data(attr);
5460                 dev_get_stats(dev, sp);
5461         }
5462
5463         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5464                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5465
5466                 if (ops && ops->fill_linkxstats) {
5467                         *idxattr = IFLA_STATS_LINK_XSTATS;
5468                         attr = nla_nest_start_noflag(skb,
5469                                                      IFLA_STATS_LINK_XSTATS);
5470                         if (!attr) {
5471                                 err = -EMSGSIZE;
5472                                 goto nla_put_failure;
5473                         }
5474
5475                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5476                         nla_nest_end(skb, attr);
5477                         if (err)
5478                                 goto nla_put_failure;
5479                         *idxattr = 0;
5480                 }
5481         }
5482
5483         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5484                              *idxattr)) {
5485                 const struct rtnl_link_ops *ops = NULL;
5486                 const struct net_device *master;
5487
5488                 master = netdev_master_upper_dev_get(dev);
5489                 if (master)
5490                         ops = master->rtnl_link_ops;
5491                 if (ops && ops->fill_linkxstats) {
5492                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5493                         attr = nla_nest_start_noflag(skb,
5494                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5495                         if (!attr) {
5496                                 err = -EMSGSIZE;
5497                                 goto nla_put_failure;
5498                         }
5499
5500                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5501                         nla_nest_end(skb, attr);
5502                         if (err)
5503                                 goto nla_put_failure;
5504                         *idxattr = 0;
5505                 }
5506         }
5507
5508         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5509                              *idxattr)) {
5510                 u32 off_filter_mask;
5511
5512                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5513                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5514                 attr = nla_nest_start_noflag(skb,
5515                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5516                 if (!attr) {
5517                         err = -EMSGSIZE;
5518                         goto nla_put_failure;
5519                 }
5520
5521                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5522                                                off_filter_mask, extack);
5523                 if (err == -ENODATA)
5524                         nla_nest_cancel(skb, attr);
5525                 else
5526                         nla_nest_end(skb, attr);
5527
5528                 if (err && err != -ENODATA)
5529                         goto nla_put_failure;
5530                 *idxattr = 0;
5531         }
5532
5533         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5534                 struct rtnl_af_ops *af_ops;
5535
5536                 *idxattr = IFLA_STATS_AF_SPEC;
5537                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5538                 if (!attr) {
5539                         err = -EMSGSIZE;
5540                         goto nla_put_failure;
5541                 }
5542
5543                 rcu_read_lock();
5544                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5545                         if (af_ops->fill_stats_af) {
5546                                 struct nlattr *af;
5547
5548                                 af = nla_nest_start_noflag(skb,
5549                                                            af_ops->family);
5550                                 if (!af) {
5551                                         rcu_read_unlock();
5552                                         err = -EMSGSIZE;
5553                                         goto nla_put_failure;
5554                                 }
5555                                 err = af_ops->fill_stats_af(skb, dev);
5556
5557                                 if (err == -ENODATA) {
5558                                         nla_nest_cancel(skb, af);
5559                                 } else if (err < 0) {
5560                                         rcu_read_unlock();
5561                                         goto nla_put_failure;
5562                                 }
5563
5564                                 nla_nest_end(skb, af);
5565                         }
5566                 }
5567                 rcu_read_unlock();
5568
5569                 nla_nest_end(skb, attr);
5570
5571                 *idxattr = 0;
5572         }
5573
5574         nlmsg_end(skb, nlh);
5575
5576         return 0;
5577
5578 nla_put_failure:
5579         /* not a multi message or no progress mean a real error */
5580         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5581                 nlmsg_cancel(skb, nlh);
5582         else
5583                 nlmsg_end(skb, nlh);
5584
5585         return err;
5586 }
5587
5588 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5589                                   const struct rtnl_stats_dump_filters *filters)
5590 {
5591         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5592         unsigned int filter_mask = filters->mask[0];
5593
5594         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5595                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5596
5597         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5598                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5599                 int attr = IFLA_STATS_LINK_XSTATS;
5600
5601                 if (ops && ops->get_linkxstats_size) {
5602                         size += nla_total_size(ops->get_linkxstats_size(dev,
5603                                                                         attr));
5604                         /* for IFLA_STATS_LINK_XSTATS */
5605                         size += nla_total_size(0);
5606                 }
5607         }
5608
5609         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5610                 struct net_device *_dev = (struct net_device *)dev;
5611                 const struct rtnl_link_ops *ops = NULL;
5612                 const struct net_device *master;
5613
5614                 /* netdev_master_upper_dev_get can't take const */
5615                 master = netdev_master_upper_dev_get(_dev);
5616                 if (master)
5617                         ops = master->rtnl_link_ops;
5618                 if (ops && ops->get_linkxstats_size) {
5619                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5620
5621                         size += nla_total_size(ops->get_linkxstats_size(dev,
5622                                                                         attr));
5623                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5624                         size += nla_total_size(0);
5625                 }
5626         }
5627
5628         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5629                 u32 off_filter_mask;
5630
5631                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5632                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5633         }
5634
5635         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5636                 struct rtnl_af_ops *af_ops;
5637
5638                 /* for IFLA_STATS_AF_SPEC */
5639                 size += nla_total_size(0);
5640
5641                 rcu_read_lock();
5642                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5643                         if (af_ops->get_stats_af_size) {
5644                                 size += nla_total_size(
5645                                         af_ops->get_stats_af_size(dev));
5646
5647                                 /* for AF_* */
5648                                 size += nla_total_size(0);
5649                         }
5650                 }
5651                 rcu_read_unlock();
5652         }
5653
5654         return size;
5655 }
5656
5657 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5658
5659 static const struct nla_policy
5660 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5661         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5662                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5663 };
5664
5665 static const struct nla_policy
5666 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5667         [IFLA_STATS_GET_FILTERS] =
5668                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5669 };
5670
5671 static const struct nla_policy
5672 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5673         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5674 };
5675
5676 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5677                                         struct rtnl_stats_dump_filters *filters,
5678                                         struct netlink_ext_ack *extack)
5679 {
5680         struct nlattr *tb[IFLA_STATS_MAX + 1];
5681         int err;
5682         int at;
5683
5684         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5685                                rtnl_stats_get_policy_filters, extack);
5686         if (err < 0)
5687                 return err;
5688
5689         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5690                 if (tb[at]) {
5691                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5692                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5693                                 return -EINVAL;
5694                         }
5695                         filters->mask[at] = nla_get_u32(tb[at]);
5696                 }
5697         }
5698
5699         return 0;
5700 }
5701
5702 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5703                                 u32 filter_mask,
5704                                 struct rtnl_stats_dump_filters *filters,
5705                                 struct netlink_ext_ack *extack)
5706 {
5707         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5708         int err;
5709         int i;
5710
5711         filters->mask[0] = filter_mask;
5712         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5713                 filters->mask[i] = -1U;
5714
5715         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5716                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5717         if (err < 0)
5718                 return err;
5719
5720         if (tb[IFLA_STATS_GET_FILTERS]) {
5721                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5722                                                    filters, extack);
5723                 if (err)
5724                         return err;
5725         }
5726
5727         return 0;
5728 }
5729
5730 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5731                                 bool is_dump, struct netlink_ext_ack *extack)
5732 {
5733         struct if_stats_msg *ifsm;
5734
5735         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5736                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5737                 return -EINVAL;
5738         }
5739
5740         if (!strict_check)
5741                 return 0;
5742
5743         ifsm = nlmsg_data(nlh);
5744
5745         /* only requests using strict checks can pass data to influence
5746          * the dump. The legacy exception is filter_mask.
5747          */
5748         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5749                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5750                 return -EINVAL;
5751         }
5752         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5753                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5754                 return -EINVAL;
5755         }
5756
5757         return 0;
5758 }
5759
5760 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5761                           struct netlink_ext_ack *extack)
5762 {
5763         struct rtnl_stats_dump_filters filters;
5764         struct net *net = sock_net(skb->sk);
5765         struct net_device *dev = NULL;
5766         int idxattr = 0, prividx = 0;
5767         struct if_stats_msg *ifsm;
5768         struct sk_buff *nskb;
5769         int err;
5770
5771         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5772                                    false, extack);
5773         if (err)
5774                 return err;
5775
5776         ifsm = nlmsg_data(nlh);
5777         if (ifsm->ifindex > 0)
5778                 dev = __dev_get_by_index(net, ifsm->ifindex);
5779         else
5780                 return -EINVAL;
5781
5782         if (!dev)
5783                 return -ENODEV;
5784
5785         if (!ifsm->filter_mask) {
5786                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5787                 return -EINVAL;
5788         }
5789
5790         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5791         if (err)
5792                 return err;
5793
5794         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5795         if (!nskb)
5796                 return -ENOBUFS;
5797
5798         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5799                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5800                                   0, &filters, &idxattr, &prividx, extack);
5801         if (err < 0) {
5802                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5803                 WARN_ON(err == -EMSGSIZE);
5804                 kfree_skb(nskb);
5805         } else {
5806                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5807         }
5808
5809         return err;
5810 }
5811
5812 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5813 {
5814         struct netlink_ext_ack *extack = cb->extack;
5815         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5816         struct rtnl_stats_dump_filters filters;
5817         struct net *net = sock_net(skb->sk);
5818         unsigned int flags = NLM_F_MULTI;
5819         struct if_stats_msg *ifsm;
5820         struct hlist_head *head;
5821         struct net_device *dev;
5822         int idx = 0;
5823
5824         s_h = cb->args[0];
5825         s_idx = cb->args[1];
5826         s_idxattr = cb->args[2];
5827         s_prividx = cb->args[3];
5828
5829         cb->seq = net->dev_base_seq;
5830
5831         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5832         if (err)
5833                 return err;
5834
5835         ifsm = nlmsg_data(cb->nlh);
5836         if (!ifsm->filter_mask) {
5837                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5838                 return -EINVAL;
5839         }
5840
5841         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5842                                    extack);
5843         if (err)
5844                 return err;
5845
5846         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5847                 idx = 0;
5848                 head = &net->dev_index_head[h];
5849                 hlist_for_each_entry(dev, head, index_hlist) {
5850                         if (idx < s_idx)
5851                                 goto cont;
5852                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5853                                                   NETLINK_CB(cb->skb).portid,
5854                                                   cb->nlh->nlmsg_seq, 0,
5855                                                   flags, &filters,
5856                                                   &s_idxattr, &s_prividx,
5857                                                   extack);
5858                         /* If we ran out of room on the first message,
5859                          * we're in trouble
5860                          */
5861                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5862
5863                         if (err < 0)
5864                                 goto out;
5865                         s_prividx = 0;
5866                         s_idxattr = 0;
5867                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5868 cont:
5869                         idx++;
5870                 }
5871         }
5872 out:
5873         cb->args[3] = s_prividx;
5874         cb->args[2] = s_idxattr;
5875         cb->args[1] = idx;
5876         cb->args[0] = h;
5877
5878         return skb->len;
5879 }
5880
5881 void rtnl_offload_xstats_notify(struct net_device *dev)
5882 {
5883         struct rtnl_stats_dump_filters response_filters = {};
5884         struct net *net = dev_net(dev);
5885         int idxattr = 0, prividx = 0;
5886         struct sk_buff *skb;
5887         int err = -ENOBUFS;
5888
5889         ASSERT_RTNL();
5890
5891         response_filters.mask[0] |=
5892                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5893         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5894                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5895
5896         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
5897                         GFP_KERNEL);
5898         if (!skb)
5899                 goto errout;
5900
5901         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
5902                                   &response_filters, &idxattr, &prividx, NULL);
5903         if (err < 0) {
5904                 kfree_skb(skb);
5905                 goto errout;
5906         }
5907
5908         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
5909         return;
5910
5911 errout:
5912         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
5913 }
5914 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
5915
5916 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
5917                           struct netlink_ext_ack *extack)
5918 {
5919         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5920         struct rtnl_stats_dump_filters response_filters = {};
5921         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5922         struct net *net = sock_net(skb->sk);
5923         struct net_device *dev = NULL;
5924         struct if_stats_msg *ifsm;
5925         bool notify = false;
5926         int err;
5927
5928         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5929                                    false, extack);
5930         if (err)
5931                 return err;
5932
5933         ifsm = nlmsg_data(nlh);
5934         if (ifsm->family != AF_UNSPEC) {
5935                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
5936                 return -EINVAL;
5937         }
5938
5939         if (ifsm->ifindex > 0)
5940                 dev = __dev_get_by_index(net, ifsm->ifindex);
5941         else
5942                 return -EINVAL;
5943
5944         if (!dev)
5945                 return -ENODEV;
5946
5947         if (ifsm->filter_mask) {
5948                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
5949                 return -EINVAL;
5950         }
5951
5952         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
5953                           ifla_stats_set_policy, extack);
5954         if (err < 0)
5955                 return err;
5956
5957         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
5958                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
5959
5960                 if (req)
5961                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
5962                 else
5963                         err = netdev_offload_xstats_disable(dev, t_l3);
5964
5965                 if (!err)
5966                         notify = true;
5967                 else if (err != -EALREADY)
5968                         return err;
5969
5970                 response_filters.mask[0] |=
5971                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5972                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5973                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5974         }
5975
5976         if (notify)
5977                 rtnl_offload_xstats_notify(dev);
5978
5979         return 0;
5980 }
5981
5982 /* Process one rtnetlink message. */
5983
5984 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5985                              struct netlink_ext_ack *extack)
5986 {
5987         struct net *net = sock_net(skb->sk);
5988         struct rtnl_link *link;
5989         enum rtnl_kinds kind;
5990         struct module *owner;
5991         int err = -EOPNOTSUPP;
5992         rtnl_doit_func doit;
5993         unsigned int flags;
5994         int family;
5995         int type;
5996
5997         type = nlh->nlmsg_type;
5998         if (type > RTM_MAX)
5999                 return -EOPNOTSUPP;
6000
6001         type -= RTM_BASE;
6002
6003         /* All the messages must have at least 1 byte length */
6004         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6005                 return 0;
6006
6007         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6008         kind = rtnl_msgtype_kind(type);
6009
6010         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6011                 return -EPERM;
6012
6013         rcu_read_lock();
6014         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6015                 struct sock *rtnl;
6016                 rtnl_dumpit_func dumpit;
6017                 u32 min_dump_alloc = 0;
6018
6019                 link = rtnl_get_link(family, type);
6020                 if (!link || !link->dumpit) {
6021                         family = PF_UNSPEC;
6022                         link = rtnl_get_link(family, type);
6023                         if (!link || !link->dumpit)
6024                                 goto err_unlock;
6025                 }
6026                 owner = link->owner;
6027                 dumpit = link->dumpit;
6028
6029                 if (type == RTM_GETLINK - RTM_BASE)
6030                         min_dump_alloc = rtnl_calcit(skb, nlh);
6031
6032                 err = 0;
6033                 /* need to do this before rcu_read_unlock() */
6034                 if (!try_module_get(owner))
6035                         err = -EPROTONOSUPPORT;
6036
6037                 rcu_read_unlock();
6038
6039                 rtnl = net->rtnl;
6040                 if (err == 0) {
6041                         struct netlink_dump_control c = {
6042                                 .dump           = dumpit,
6043                                 .min_dump_alloc = min_dump_alloc,
6044                                 .module         = owner,
6045                         };
6046                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6047                         /* netlink_dump_start() will keep a reference on
6048                          * module if dump is still in progress.
6049                          */
6050                         module_put(owner);
6051                 }
6052                 return err;
6053         }
6054
6055         link = rtnl_get_link(family, type);
6056         if (!link || !link->doit) {
6057                 family = PF_UNSPEC;
6058                 link = rtnl_get_link(PF_UNSPEC, type);
6059                 if (!link || !link->doit)
6060                         goto out_unlock;
6061         }
6062
6063         owner = link->owner;
6064         if (!try_module_get(owner)) {
6065                 err = -EPROTONOSUPPORT;
6066                 goto out_unlock;
6067         }
6068
6069         flags = link->flags;
6070         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6071             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6072                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6073                 goto err_unlock;
6074         }
6075
6076         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6077                 doit = link->doit;
6078                 rcu_read_unlock();
6079                 if (doit)
6080                         err = doit(skb, nlh, extack);
6081                 module_put(owner);
6082                 return err;
6083         }
6084         rcu_read_unlock();
6085
6086         rtnl_lock();
6087         link = rtnl_get_link(family, type);
6088         if (link && link->doit)
6089                 err = link->doit(skb, nlh, extack);
6090         rtnl_unlock();
6091
6092         module_put(owner);
6093
6094         return err;
6095
6096 out_unlock:
6097         rcu_read_unlock();
6098         return err;
6099
6100 err_unlock:
6101         rcu_read_unlock();
6102         return -EOPNOTSUPP;
6103 }
6104
6105 static void rtnetlink_rcv(struct sk_buff *skb)
6106 {
6107         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6108 }
6109
6110 static int rtnetlink_bind(struct net *net, int group)
6111 {
6112         switch (group) {
6113         case RTNLGRP_IPV4_MROUTE_R:
6114         case RTNLGRP_IPV6_MROUTE_R:
6115                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6116                         return -EPERM;
6117                 break;
6118         }
6119         return 0;
6120 }
6121
6122 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6123 {
6124         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6125
6126         switch (event) {
6127         case NETDEV_REBOOT:
6128         case NETDEV_CHANGEMTU:
6129         case NETDEV_CHANGEADDR:
6130         case NETDEV_CHANGENAME:
6131         case NETDEV_FEAT_CHANGE:
6132         case NETDEV_BONDING_FAILOVER:
6133         case NETDEV_POST_TYPE_CHANGE:
6134         case NETDEV_NOTIFY_PEERS:
6135         case NETDEV_CHANGEUPPER:
6136         case NETDEV_RESEND_IGMP:
6137         case NETDEV_CHANGEINFODATA:
6138         case NETDEV_CHANGELOWERSTATE:
6139         case NETDEV_CHANGE_TX_QUEUE_LEN:
6140                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6141                                    GFP_KERNEL, NULL, 0);
6142                 break;
6143         default:
6144                 break;
6145         }
6146         return NOTIFY_DONE;
6147 }
6148
6149 static struct notifier_block rtnetlink_dev_notifier = {
6150         .notifier_call  = rtnetlink_event,
6151 };
6152
6153
6154 static int __net_init rtnetlink_net_init(struct net *net)
6155 {
6156         struct sock *sk;
6157         struct netlink_kernel_cfg cfg = {
6158                 .groups         = RTNLGRP_MAX,
6159                 .input          = rtnetlink_rcv,
6160                 .cb_mutex       = &rtnl_mutex,
6161                 .flags          = NL_CFG_F_NONROOT_RECV,
6162                 .bind           = rtnetlink_bind,
6163         };
6164
6165         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6166         if (!sk)
6167                 return -ENOMEM;
6168         net->rtnl = sk;
6169         return 0;
6170 }
6171
6172 static void __net_exit rtnetlink_net_exit(struct net *net)
6173 {
6174         netlink_kernel_release(net->rtnl);
6175         net->rtnl = NULL;
6176 }
6177
6178 static struct pernet_operations rtnetlink_net_ops = {
6179         .init = rtnetlink_net_init,
6180         .exit = rtnetlink_net_exit,
6181 };
6182
6183 void __init rtnetlink_init(void)
6184 {
6185         if (register_pernet_subsys(&rtnetlink_net_ops))
6186                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6187
6188         register_netdevice_notifier(&rtnetlink_dev_notifier);
6189
6190         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6191                       rtnl_dump_ifinfo, 0);
6192         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6193         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6194         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6195
6196         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6197         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6198         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6199
6200         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6201         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6202
6203         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6204         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6205                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6206         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6207
6208         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6209         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6210         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6211
6212         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6213                       0);
6214         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6215 }