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, <[email protected]>
  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         if (tb[IFLA_GRO_MAX_SIZE]) {
2364                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2365
2366                 if (gro_max_size > GRO_MAX_SIZE) {
2367                         NL_SET_ERR_MSG(extack, "too big gro_max_size");
2368                         return -EINVAL;
2369                 }
2370         }
2371         return 0;
2372 }
2373
2374 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2375                                   int guid_type)
2376 {
2377         const struct net_device_ops *ops = dev->netdev_ops;
2378
2379         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2380 }
2381
2382 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2383 {
2384         if (dev->type != ARPHRD_INFINIBAND)
2385                 return -EOPNOTSUPP;
2386
2387         return handle_infiniband_guid(dev, ivt, guid_type);
2388 }
2389
2390 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2391 {
2392         const struct net_device_ops *ops = dev->netdev_ops;
2393         int err = -EINVAL;
2394
2395         if (tb[IFLA_VF_MAC]) {
2396                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2397
2398                 if (ivm->vf >= INT_MAX)
2399                         return -EINVAL;
2400                 err = -EOPNOTSUPP;
2401                 if (ops->ndo_set_vf_mac)
2402                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2403                                                   ivm->mac);
2404                 if (err < 0)
2405                         return err;
2406         }
2407
2408         if (tb[IFLA_VF_VLAN]) {
2409                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2410
2411                 if (ivv->vf >= INT_MAX)
2412                         return -EINVAL;
2413                 err = -EOPNOTSUPP;
2414                 if (ops->ndo_set_vf_vlan)
2415                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2416                                                    ivv->qos,
2417                                                    htons(ETH_P_8021Q));
2418                 if (err < 0)
2419                         return err;
2420         }
2421
2422         if (tb[IFLA_VF_VLAN_LIST]) {
2423                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2424                 struct nlattr *attr;
2425                 int rem, len = 0;
2426
2427                 err = -EOPNOTSUPP;
2428                 if (!ops->ndo_set_vf_vlan)
2429                         return err;
2430
2431                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2432                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2433                             nla_len(attr) < NLA_HDRLEN) {
2434                                 return -EINVAL;
2435                         }
2436                         if (len >= MAX_VLAN_LIST_LEN)
2437                                 return -EOPNOTSUPP;
2438                         ivvl[len] = nla_data(attr);
2439
2440                         len++;
2441                 }
2442                 if (len == 0)
2443                         return -EINVAL;
2444
2445                 if (ivvl[0]->vf >= INT_MAX)
2446                         return -EINVAL;
2447                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2448                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2449                 if (err < 0)
2450                         return err;
2451         }
2452
2453         if (tb[IFLA_VF_TX_RATE]) {
2454                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2455                 struct ifla_vf_info ivf;
2456
2457                 if (ivt->vf >= INT_MAX)
2458                         return -EINVAL;
2459                 err = -EOPNOTSUPP;
2460                 if (ops->ndo_get_vf_config)
2461                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2462                 if (err < 0)
2463                         return err;
2464
2465                 err = rtnl_set_vf_rate(dev, ivt->vf,
2466                                        ivf.min_tx_rate, ivt->rate);
2467                 if (err < 0)
2468                         return err;
2469         }
2470
2471         if (tb[IFLA_VF_RATE]) {
2472                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2473
2474                 if (ivt->vf >= INT_MAX)
2475                         return -EINVAL;
2476
2477                 err = rtnl_set_vf_rate(dev, ivt->vf,
2478                                        ivt->min_tx_rate, ivt->max_tx_rate);
2479                 if (err < 0)
2480                         return err;
2481         }
2482
2483         if (tb[IFLA_VF_SPOOFCHK]) {
2484                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2485
2486                 if (ivs->vf >= INT_MAX)
2487                         return -EINVAL;
2488                 err = -EOPNOTSUPP;
2489                 if (ops->ndo_set_vf_spoofchk)
2490                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2491                                                        ivs->setting);
2492                 if (err < 0)
2493                         return err;
2494         }
2495
2496         if (tb[IFLA_VF_LINK_STATE]) {
2497                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2498
2499                 if (ivl->vf >= INT_MAX)
2500                         return -EINVAL;
2501                 err = -EOPNOTSUPP;
2502                 if (ops->ndo_set_vf_link_state)
2503                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2504                                                          ivl->link_state);
2505                 if (err < 0)
2506                         return err;
2507         }
2508
2509         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2510                 struct ifla_vf_rss_query_en *ivrssq_en;
2511
2512                 err = -EOPNOTSUPP;
2513                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2514                 if (ivrssq_en->vf >= INT_MAX)
2515                         return -EINVAL;
2516                 if (ops->ndo_set_vf_rss_query_en)
2517                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2518                                                            ivrssq_en->setting);
2519                 if (err < 0)
2520                         return err;
2521         }
2522
2523         if (tb[IFLA_VF_TRUST]) {
2524                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2525
2526                 if (ivt->vf >= INT_MAX)
2527                         return -EINVAL;
2528                 err = -EOPNOTSUPP;
2529                 if (ops->ndo_set_vf_trust)
2530                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2531                 if (err < 0)
2532                         return err;
2533         }
2534
2535         if (tb[IFLA_VF_IB_NODE_GUID]) {
2536                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2537
2538                 if (ivt->vf >= INT_MAX)
2539                         return -EINVAL;
2540                 if (!ops->ndo_set_vf_guid)
2541                         return -EOPNOTSUPP;
2542                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2543         }
2544
2545         if (tb[IFLA_VF_IB_PORT_GUID]) {
2546                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2547
2548                 if (ivt->vf >= INT_MAX)
2549                         return -EINVAL;
2550                 if (!ops->ndo_set_vf_guid)
2551                         return -EOPNOTSUPP;
2552
2553                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2554         }
2555
2556         return err;
2557 }
2558
2559 static int do_set_master(struct net_device *dev, int ifindex,
2560                          struct netlink_ext_ack *extack)
2561 {
2562         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2563         const struct net_device_ops *ops;
2564         int err;
2565
2566         if (upper_dev) {
2567                 if (upper_dev->ifindex == ifindex)
2568                         return 0;
2569                 ops = upper_dev->netdev_ops;
2570                 if (ops->ndo_del_slave) {
2571                         err = ops->ndo_del_slave(upper_dev, dev);
2572                         if (err)
2573                                 return err;
2574                 } else {
2575                         return -EOPNOTSUPP;
2576                 }
2577         }
2578
2579         if (ifindex) {
2580                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2581                 if (!upper_dev)
2582                         return -EINVAL;
2583                 ops = upper_dev->netdev_ops;
2584                 if (ops->ndo_add_slave) {
2585                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2586                         if (err)
2587                                 return err;
2588                 } else {
2589                         return -EOPNOTSUPP;
2590                 }
2591         }
2592         return 0;
2593 }
2594
2595 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2596         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2597         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2598 };
2599
2600 static int do_set_proto_down(struct net_device *dev,
2601                              struct nlattr *nl_proto_down,
2602                              struct nlattr *nl_proto_down_reason,
2603                              struct netlink_ext_ack *extack)
2604 {
2605         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2606         unsigned long mask = 0;
2607         u32 value;
2608         bool proto_down;
2609         int err;
2610
2611         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2612                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2613                 return -EOPNOTSUPP;
2614         }
2615
2616         if (nl_proto_down_reason) {
2617                 err = nla_parse_nested_deprecated(pdreason,
2618                                                   IFLA_PROTO_DOWN_REASON_MAX,
2619                                                   nl_proto_down_reason,
2620                                                   ifla_proto_down_reason_policy,
2621                                                   NULL);
2622                 if (err < 0)
2623                         return err;
2624
2625                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2626                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2627                         return -EINVAL;
2628                 }
2629
2630                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2631
2632                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2633                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2634
2635                 dev_change_proto_down_reason(dev, mask, value);
2636         }
2637
2638         if (nl_proto_down) {
2639                 proto_down = nla_get_u8(nl_proto_down);
2640
2641                 /* Don't turn off protodown if there are active reasons */
2642                 if (!proto_down && dev->proto_down_reason) {
2643                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2644                         return -EBUSY;
2645                 }
2646                 err = dev_change_proto_down(dev,
2647                                             proto_down);
2648                 if (err)
2649                         return err;
2650         }
2651
2652         return 0;
2653 }
2654
2655 #define DO_SETLINK_MODIFIED     0x01
2656 /* notify flag means notify + modified. */
2657 #define DO_SETLINK_NOTIFY       0x03
2658 static int do_setlink(const struct sk_buff *skb,
2659                       struct net_device *dev, struct ifinfomsg *ifm,
2660                       struct netlink_ext_ack *extack,
2661                       struct nlattr **tb, int status)
2662 {
2663         const struct net_device_ops *ops = dev->netdev_ops;
2664         char ifname[IFNAMSIZ];
2665         int err;
2666
2667         err = validate_linkmsg(dev, tb, extack);
2668         if (err < 0)
2669                 return err;
2670
2671         if (tb[IFLA_IFNAME])
2672                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2673         else
2674                 ifname[0] = '\0';
2675
2676         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2677                 const char *pat = ifname[0] ? ifname : NULL;
2678                 struct net *net;
2679                 int new_ifindex;
2680
2681                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2682                                                 tb, CAP_NET_ADMIN);
2683                 if (IS_ERR(net)) {
2684                         err = PTR_ERR(net);
2685                         goto errout;
2686                 }
2687
2688                 if (tb[IFLA_NEW_IFINDEX])
2689                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2690                 else
2691                         new_ifindex = 0;
2692
2693                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2694                 put_net(net);
2695                 if (err)
2696                         goto errout;
2697                 status |= DO_SETLINK_MODIFIED;
2698         }
2699
2700         if (tb[IFLA_MAP]) {
2701                 struct rtnl_link_ifmap *u_map;
2702                 struct ifmap k_map;
2703
2704                 if (!ops->ndo_set_config) {
2705                         err = -EOPNOTSUPP;
2706                         goto errout;
2707                 }
2708
2709                 if (!netif_device_present(dev)) {
2710                         err = -ENODEV;
2711                         goto errout;
2712                 }
2713
2714                 u_map = nla_data(tb[IFLA_MAP]);
2715                 k_map.mem_start = (unsigned long) u_map->mem_start;
2716                 k_map.mem_end = (unsigned long) u_map->mem_end;
2717                 k_map.base_addr = (unsigned short) u_map->base_addr;
2718                 k_map.irq = (unsigned char) u_map->irq;
2719                 k_map.dma = (unsigned char) u_map->dma;
2720                 k_map.port = (unsigned char) u_map->port;
2721
2722                 err = ops->ndo_set_config(dev, &k_map);
2723                 if (err < 0)
2724                         goto errout;
2725
2726                 status |= DO_SETLINK_NOTIFY;
2727         }
2728
2729         if (tb[IFLA_ADDRESS]) {
2730                 struct sockaddr *sa;
2731                 int len;
2732
2733                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2734                                                   sizeof(*sa));
2735                 sa = kmalloc(len, GFP_KERNEL);
2736                 if (!sa) {
2737                         err = -ENOMEM;
2738                         goto errout;
2739                 }
2740                 sa->sa_family = dev->type;
2741                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2742                        dev->addr_len);
2743                 err = dev_set_mac_address_user(dev, sa, extack);
2744                 kfree(sa);
2745                 if (err)
2746                         goto errout;
2747                 status |= DO_SETLINK_MODIFIED;
2748         }
2749
2750         if (tb[IFLA_MTU]) {
2751                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2752                 if (err < 0)
2753                         goto errout;
2754                 status |= DO_SETLINK_MODIFIED;
2755         }
2756
2757         if (tb[IFLA_GROUP]) {
2758                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2759                 status |= DO_SETLINK_NOTIFY;
2760         }
2761
2762         /*
2763          * Interface selected by interface index but interface
2764          * name provided implies that a name change has been
2765          * requested.
2766          */
2767         if (ifm->ifi_index > 0 && ifname[0]) {
2768                 err = dev_change_name(dev, ifname);
2769                 if (err < 0)
2770                         goto errout;
2771                 status |= DO_SETLINK_MODIFIED;
2772         }
2773
2774         if (tb[IFLA_IFALIAS]) {
2775                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2776                                     nla_len(tb[IFLA_IFALIAS]));
2777                 if (err < 0)
2778                         goto errout;
2779                 status |= DO_SETLINK_NOTIFY;
2780         }
2781
2782         if (tb[IFLA_BROADCAST]) {
2783                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2784                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2785         }
2786
2787         if (ifm->ifi_flags || ifm->ifi_change) {
2788                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2789                                        extack);
2790                 if (err < 0)
2791                         goto errout;
2792         }
2793
2794         if (tb[IFLA_MASTER]) {
2795                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2796                 if (err)
2797                         goto errout;
2798                 status |= DO_SETLINK_MODIFIED;
2799         }
2800
2801         if (tb[IFLA_CARRIER]) {
2802                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2803                 if (err)
2804                         goto errout;
2805                 status |= DO_SETLINK_MODIFIED;
2806         }
2807
2808         if (tb[IFLA_TXQLEN]) {
2809                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2810
2811                 err = dev_change_tx_queue_len(dev, value);
2812                 if (err)
2813                         goto errout;
2814                 status |= DO_SETLINK_MODIFIED;
2815         }
2816
2817         if (tb[IFLA_GSO_MAX_SIZE]) {
2818                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2819
2820                 if (max_size > GSO_MAX_SIZE || max_size > dev->tso_max_size) {
2821                         err = -EINVAL;
2822                         goto errout;
2823                 }
2824
2825                 if (dev->gso_max_size ^ max_size) {
2826                         netif_set_gso_max_size(dev, max_size);
2827                         status |= DO_SETLINK_MODIFIED;
2828                 }
2829         }
2830
2831         if (tb[IFLA_GSO_MAX_SEGS]) {
2832                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2833
2834                 if (max_segs > GSO_MAX_SEGS || max_segs > dev->tso_max_segs) {
2835                         err = -EINVAL;
2836                         goto errout;
2837                 }
2838
2839                 if (dev->gso_max_segs ^ max_segs) {
2840                         netif_set_gso_max_segs(dev, max_segs);
2841                         status |= DO_SETLINK_MODIFIED;
2842                 }
2843         }
2844
2845         if (tb[IFLA_GRO_MAX_SIZE]) {
2846                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2847
2848                 if (dev->gro_max_size ^ gro_max_size) {
2849                         netif_set_gro_max_size(dev, gro_max_size);
2850                         status |= DO_SETLINK_MODIFIED;
2851                 }
2852         }
2853
2854         if (tb[IFLA_OPERSTATE])
2855                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2856
2857         if (tb[IFLA_LINKMODE]) {
2858                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2859
2860                 write_lock(&dev_base_lock);
2861                 if (dev->link_mode ^ value)
2862                         status |= DO_SETLINK_NOTIFY;
2863                 dev->link_mode = value;
2864                 write_unlock(&dev_base_lock);
2865         }
2866
2867         if (tb[IFLA_VFINFO_LIST]) {
2868                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2869                 struct nlattr *attr;
2870                 int rem;
2871
2872                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2873                         if (nla_type(attr) != IFLA_VF_INFO ||
2874                             nla_len(attr) < NLA_HDRLEN) {
2875                                 err = -EINVAL;
2876                                 goto errout;
2877                         }
2878                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2879                                                           attr,
2880                                                           ifla_vf_policy,
2881                                                           NULL);
2882                         if (err < 0)
2883                                 goto errout;
2884                         err = do_setvfinfo(dev, vfinfo);
2885                         if (err < 0)
2886                                 goto errout;
2887                         status |= DO_SETLINK_NOTIFY;
2888                 }
2889         }
2890         err = 0;
2891
2892         if (tb[IFLA_VF_PORTS]) {
2893                 struct nlattr *port[IFLA_PORT_MAX+1];
2894                 struct nlattr *attr;
2895                 int vf;
2896                 int rem;
2897
2898                 err = -EOPNOTSUPP;
2899                 if (!ops->ndo_set_vf_port)
2900                         goto errout;
2901
2902                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2903                         if (nla_type(attr) != IFLA_VF_PORT ||
2904                             nla_len(attr) < NLA_HDRLEN) {
2905                                 err = -EINVAL;
2906                                 goto errout;
2907                         }
2908                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2909                                                           attr,
2910                                                           ifla_port_policy,
2911                                                           NULL);
2912                         if (err < 0)
2913                                 goto errout;
2914                         if (!port[IFLA_PORT_VF]) {
2915                                 err = -EOPNOTSUPP;
2916                                 goto errout;
2917                         }
2918                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2919                         err = ops->ndo_set_vf_port(dev, vf, port);
2920                         if (err < 0)
2921                                 goto errout;
2922                         status |= DO_SETLINK_NOTIFY;
2923                 }
2924         }
2925         err = 0;
2926
2927         if (tb[IFLA_PORT_SELF]) {
2928                 struct nlattr *port[IFLA_PORT_MAX+1];
2929
2930                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2931                                                   tb[IFLA_PORT_SELF],
2932                                                   ifla_port_policy, NULL);
2933                 if (err < 0)
2934                         goto errout;
2935
2936                 err = -EOPNOTSUPP;
2937                 if (ops->ndo_set_vf_port)
2938                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2939                 if (err < 0)
2940                         goto errout;
2941                 status |= DO_SETLINK_NOTIFY;
2942         }
2943
2944         if (tb[IFLA_AF_SPEC]) {
2945                 struct nlattr *af;
2946                 int rem;
2947
2948                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2949                         const struct rtnl_af_ops *af_ops;
2950
2951                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2952
2953                         err = af_ops->set_link_af(dev, af, extack);
2954                         if (err < 0)
2955                                 goto errout;
2956
2957                         status |= DO_SETLINK_NOTIFY;
2958                 }
2959         }
2960         err = 0;
2961
2962         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2963                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2964                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2965                 if (err)
2966                         goto errout;
2967                 status |= DO_SETLINK_NOTIFY;
2968         }
2969
2970         if (tb[IFLA_XDP]) {
2971                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2972                 u32 xdp_flags = 0;
2973
2974                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2975                                                   tb[IFLA_XDP],
2976                                                   ifla_xdp_policy, NULL);
2977                 if (err < 0)
2978                         goto errout;
2979
2980                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2981                         err = -EINVAL;
2982                         goto errout;
2983                 }
2984
2985                 if (xdp[IFLA_XDP_FLAGS]) {
2986                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2987                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2988                                 err = -EINVAL;
2989                                 goto errout;
2990                         }
2991                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2992                                 err = -EINVAL;
2993                                 goto errout;
2994                         }
2995                 }
2996
2997                 if (xdp[IFLA_XDP_FD]) {
2998                         int expected_fd = -1;
2999
3000                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3001                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3002                                         err = -EINVAL;
3003                                         goto errout;
3004                                 }
3005                                 expected_fd =
3006                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3007                         }
3008
3009                         err = dev_change_xdp_fd(dev, extack,
3010                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3011                                                 expected_fd,
3012                                                 xdp_flags);
3013                         if (err)
3014                                 goto errout;
3015                         status |= DO_SETLINK_NOTIFY;
3016                 }
3017         }
3018
3019 errout:
3020         if (status & DO_SETLINK_MODIFIED) {
3021                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3022                         netdev_state_change(dev);
3023
3024                 if (err < 0)
3025                         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",
3026                                              dev->name);
3027         }
3028
3029         return err;
3030 }
3031
3032 static struct net_device *rtnl_dev_get(struct net *net,
3033                                        struct nlattr *tb[])
3034 {
3035         char ifname[ALTIFNAMSIZ];
3036
3037         if (tb[IFLA_IFNAME])
3038                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3039         else if (tb[IFLA_ALT_IFNAME])
3040                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3041         else
3042                 return NULL;
3043
3044         return __dev_get_by_name(net, ifname);
3045 }
3046
3047 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3048                         struct netlink_ext_ack *extack)
3049 {
3050         struct net *net = sock_net(skb->sk);
3051         struct ifinfomsg *ifm;
3052         struct net_device *dev;
3053         int err;
3054         struct nlattr *tb[IFLA_MAX+1];
3055
3056         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3057                                      ifla_policy, extack);
3058         if (err < 0)
3059                 goto errout;
3060
3061         err = rtnl_ensure_unique_netns(tb, extack, false);
3062         if (err < 0)
3063                 goto errout;
3064
3065         err = -EINVAL;
3066         ifm = nlmsg_data(nlh);
3067         if (ifm->ifi_index > 0)
3068                 dev = __dev_get_by_index(net, ifm->ifi_index);
3069         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3070                 dev = rtnl_dev_get(net, tb);
3071         else
3072                 goto errout;
3073
3074         if (dev == NULL) {
3075                 err = -ENODEV;
3076                 goto errout;
3077         }
3078
3079         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3080 errout:
3081         return err;
3082 }
3083
3084 static int rtnl_group_dellink(const struct net *net, int group)
3085 {
3086         struct net_device *dev, *aux;
3087         LIST_HEAD(list_kill);
3088         bool found = false;
3089
3090         if (!group)
3091                 return -EPERM;
3092
3093         for_each_netdev(net, dev) {
3094                 if (dev->group == group) {
3095                         const struct rtnl_link_ops *ops;
3096
3097                         found = true;
3098                         ops = dev->rtnl_link_ops;
3099                         if (!ops || !ops->dellink)
3100                                 return -EOPNOTSUPP;
3101                 }
3102         }
3103
3104         if (!found)
3105                 return -ENODEV;
3106
3107         for_each_netdev_safe(net, dev, aux) {
3108                 if (dev->group == group) {
3109                         const struct rtnl_link_ops *ops;
3110
3111                         ops = dev->rtnl_link_ops;
3112                         ops->dellink(dev, &list_kill);
3113                 }
3114         }
3115         unregister_netdevice_many(&list_kill);
3116
3117         return 0;
3118 }
3119
3120 int rtnl_delete_link(struct net_device *dev)
3121 {
3122         const struct rtnl_link_ops *ops;
3123         LIST_HEAD(list_kill);
3124
3125         ops = dev->rtnl_link_ops;
3126         if (!ops || !ops->dellink)
3127                 return -EOPNOTSUPP;
3128
3129         ops->dellink(dev, &list_kill);
3130         unregister_netdevice_many(&list_kill);
3131
3132         return 0;
3133 }
3134 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3135
3136 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3137                         struct netlink_ext_ack *extack)
3138 {
3139         struct net *net = sock_net(skb->sk);
3140         struct net *tgt_net = net;
3141         struct net_device *dev = NULL;
3142         struct ifinfomsg *ifm;
3143         struct nlattr *tb[IFLA_MAX+1];
3144         int err;
3145         int netnsid = -1;
3146
3147         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3148                                      ifla_policy, extack);
3149         if (err < 0)
3150                 return err;
3151
3152         err = rtnl_ensure_unique_netns(tb, extack, true);
3153         if (err < 0)
3154                 return err;
3155
3156         if (tb[IFLA_TARGET_NETNSID]) {
3157                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3158                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3159                 if (IS_ERR(tgt_net))
3160                         return PTR_ERR(tgt_net);
3161         }
3162
3163         err = -EINVAL;
3164         ifm = nlmsg_data(nlh);
3165         if (ifm->ifi_index > 0)
3166                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3167         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3168                 dev = rtnl_dev_get(net, tb);
3169         else if (tb[IFLA_GROUP])
3170                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3171         else
3172                 goto out;
3173
3174         if (!dev) {
3175                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3176                         err = -ENODEV;
3177
3178                 goto out;
3179         }
3180
3181         err = rtnl_delete_link(dev);
3182
3183 out:
3184         if (netnsid >= 0)
3185                 put_net(tgt_net);
3186
3187         return err;
3188 }
3189
3190 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3191 {
3192         unsigned int old_flags;
3193         int err;
3194
3195         old_flags = dev->flags;
3196         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3197                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3198                                          NULL);
3199                 if (err < 0)
3200                         return err;
3201         }
3202
3203         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3204                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3205         } else {
3206                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3207                 __dev_notify_flags(dev, old_flags, ~0U);
3208         }
3209         return 0;
3210 }
3211 EXPORT_SYMBOL(rtnl_configure_link);
3212
3213 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3214                                     unsigned char name_assign_type,
3215                                     const struct rtnl_link_ops *ops,
3216                                     struct nlattr *tb[],
3217                                     struct netlink_ext_ack *extack)
3218 {
3219         struct net_device *dev;
3220         unsigned int num_tx_queues = 1;
3221         unsigned int num_rx_queues = 1;
3222
3223         if (tb[IFLA_NUM_TX_QUEUES])
3224                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3225         else if (ops->get_num_tx_queues)
3226                 num_tx_queues = ops->get_num_tx_queues();
3227
3228         if (tb[IFLA_NUM_RX_QUEUES])
3229                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3230         else if (ops->get_num_rx_queues)
3231                 num_rx_queues = ops->get_num_rx_queues();
3232
3233         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3234                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3235                 return ERR_PTR(-EINVAL);
3236         }
3237
3238         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3239                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3240                 return ERR_PTR(-EINVAL);
3241         }
3242
3243         if (ops->alloc) {
3244                 dev = ops->alloc(tb, ifname, name_assign_type,
3245                                  num_tx_queues, num_rx_queues);
3246                 if (IS_ERR(dev))
3247                         return dev;
3248         } else {
3249                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3250                                        name_assign_type, ops->setup,
3251                                        num_tx_queues, num_rx_queues);
3252         }
3253
3254         if (!dev)
3255                 return ERR_PTR(-ENOMEM);
3256
3257         dev_net_set(dev, net);
3258         dev->rtnl_link_ops = ops;
3259         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3260
3261         if (tb[IFLA_MTU]) {
3262                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3263                 int err;
3264
3265                 err = dev_validate_mtu(dev, mtu, extack);
3266                 if (err) {
3267                         free_netdev(dev);
3268                         return ERR_PTR(err);
3269                 }
3270                 dev->mtu = mtu;
3271         }
3272         if (tb[IFLA_ADDRESS]) {
3273                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3274                                nla_len(tb[IFLA_ADDRESS]));
3275                 dev->addr_assign_type = NET_ADDR_SET;
3276         }
3277         if (tb[IFLA_BROADCAST])
3278                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3279                                 nla_len(tb[IFLA_BROADCAST]));
3280         if (tb[IFLA_TXQLEN])
3281                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3282         if (tb[IFLA_OPERSTATE])
3283                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3284         if (tb[IFLA_LINKMODE])
3285                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3286         if (tb[IFLA_GROUP])
3287                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3288         if (tb[IFLA_GSO_MAX_SIZE])
3289                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3290         if (tb[IFLA_GSO_MAX_SEGS])
3291                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3292         if (tb[IFLA_GRO_MAX_SIZE])
3293                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3294
3295         return dev;
3296 }
3297 EXPORT_SYMBOL(rtnl_create_link);
3298
3299 static int rtnl_group_changelink(const struct sk_buff *skb,
3300                 struct net *net, int group,
3301                 struct ifinfomsg *ifm,
3302                 struct netlink_ext_ack *extack,
3303                 struct nlattr **tb)
3304 {
3305         struct net_device *dev, *aux;
3306         int err;
3307
3308         for_each_netdev_safe(net, dev, aux) {
3309                 if (dev->group == group) {
3310                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3311                         if (err < 0)
3312                                 return err;
3313                 }
3314         }
3315
3316         return 0;
3317 }
3318
3319 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3320                                const struct rtnl_link_ops *ops,
3321                                struct nlattr **tb, struct nlattr **data,
3322                                struct netlink_ext_ack *extack)
3323 {
3324         unsigned char name_assign_type = NET_NAME_USER;
3325         struct net *net = sock_net(skb->sk);
3326         struct net *dest_net, *link_net;
3327         struct net_device *dev;
3328         char ifname[IFNAMSIZ];
3329         int err;
3330
3331         if (!ops->alloc && !ops->setup)
3332                 return -EOPNOTSUPP;
3333
3334         if (tb[IFLA_IFNAME]) {
3335                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3336         } else {
3337                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3338                 name_assign_type = NET_NAME_ENUM;
3339         }
3340
3341         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3342         if (IS_ERR(dest_net))
3343                 return PTR_ERR(dest_net);
3344
3345         if (tb[IFLA_LINK_NETNSID]) {
3346                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3347
3348                 link_net = get_net_ns_by_id(dest_net, id);
3349                 if (!link_net) {
3350                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3351                         err =  -EINVAL;
3352                         goto out;
3353                 }
3354                 err = -EPERM;
3355                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3356                         goto out;
3357         } else {
3358                 link_net = NULL;
3359         }
3360
3361         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3362                                name_assign_type, ops, tb, extack);
3363         if (IS_ERR(dev)) {
3364                 err = PTR_ERR(dev);
3365                 goto out;
3366         }
3367
3368         dev->ifindex = ifm->ifi_index;
3369
3370         if (ops->newlink)
3371                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3372         else
3373                 err = register_netdevice(dev);
3374         if (err < 0) {
3375                 free_netdev(dev);
3376                 goto out;
3377         }
3378
3379         err = rtnl_configure_link(dev, ifm);
3380         if (err < 0)
3381                 goto out_unregister;
3382         if (link_net) {
3383                 err = dev_change_net_namespace(dev, dest_net, ifname);
3384                 if (err < 0)
3385                         goto out_unregister;
3386         }
3387         if (tb[IFLA_MASTER]) {
3388                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3389                 if (err)
3390                         goto out_unregister;
3391         }
3392 out:
3393         if (link_net)
3394                 put_net(link_net);
3395         put_net(dest_net);
3396         return err;
3397 out_unregister:
3398         if (ops->newlink) {
3399                 LIST_HEAD(list_kill);
3400
3401                 ops->dellink(dev, &list_kill);
3402                 unregister_netdevice_many(&list_kill);
3403         } else {
3404                 unregister_netdevice(dev);
3405         }
3406         goto out;
3407 }
3408
3409 struct rtnl_newlink_tbs {
3410         struct nlattr *tb[IFLA_MAX + 1];
3411         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3412         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3413 };
3414
3415 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3416                           struct rtnl_newlink_tbs *tbs,
3417                           struct netlink_ext_ack *extack)
3418 {
3419         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3420         struct nlattr ** const tb = tbs->tb;
3421         const struct rtnl_link_ops *m_ops;
3422         struct net_device *master_dev;
3423         struct net *net = sock_net(skb->sk);
3424         const struct rtnl_link_ops *ops;
3425         struct nlattr **slave_data;
3426         char kind[MODULE_NAME_LEN];
3427         struct net_device *dev;
3428         struct ifinfomsg *ifm;
3429         struct nlattr **data;
3430         bool link_specified;
3431         int err;
3432
3433 #ifdef CONFIG_MODULES
3434 replay:
3435 #endif
3436         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3437                                      ifla_policy, extack);
3438         if (err < 0)
3439                 return err;
3440
3441         err = rtnl_ensure_unique_netns(tb, extack, false);
3442         if (err < 0)
3443                 return err;
3444
3445         ifm = nlmsg_data(nlh);
3446         if (ifm->ifi_index > 0) {
3447                 link_specified = true;
3448                 dev = __dev_get_by_index(net, ifm->ifi_index);
3449         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3450                 link_specified = true;
3451                 dev = rtnl_dev_get(net, tb);
3452         } else {
3453                 link_specified = false;
3454                 dev = NULL;
3455         }
3456
3457         master_dev = NULL;
3458         m_ops = NULL;
3459         if (dev) {
3460                 master_dev = netdev_master_upper_dev_get(dev);
3461                 if (master_dev)
3462                         m_ops = master_dev->rtnl_link_ops;
3463         }
3464
3465         err = validate_linkmsg(dev, tb, extack);
3466         if (err < 0)
3467                 return err;
3468
3469         if (tb[IFLA_LINKINFO]) {
3470                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3471                                                   tb[IFLA_LINKINFO],
3472                                                   ifla_info_policy, NULL);
3473                 if (err < 0)
3474                         return err;
3475         } else
3476                 memset(linkinfo, 0, sizeof(linkinfo));
3477
3478         if (linkinfo[IFLA_INFO_KIND]) {
3479                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3480                 ops = rtnl_link_ops_get(kind);
3481         } else {
3482                 kind[0] = '\0';
3483                 ops = NULL;
3484         }
3485
3486         data = NULL;
3487         if (ops) {
3488                 if (ops->maxtype > RTNL_MAX_TYPE)
3489                         return -EINVAL;
3490
3491                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3492                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3493                                                           linkinfo[IFLA_INFO_DATA],
3494                                                           ops->policy, extack);
3495                         if (err < 0)
3496                                 return err;
3497                         data = tbs->attr;
3498                 }
3499                 if (ops->validate) {
3500                         err = ops->validate(tb, data, extack);
3501                         if (err < 0)
3502                                 return err;
3503                 }
3504         }
3505
3506         slave_data = NULL;
3507         if (m_ops) {
3508                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3509                         return -EINVAL;
3510
3511                 if (m_ops->slave_maxtype &&
3512                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3513                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3514                                                           m_ops->slave_maxtype,
3515                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3516                                                           m_ops->slave_policy,
3517                                                           extack);
3518                         if (err < 0)
3519                                 return err;
3520                         slave_data = tbs->slave_attr;
3521                 }
3522         }
3523
3524         if (dev) {
3525                 int status = 0;
3526
3527                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3528                         return -EEXIST;
3529                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3530                         return -EOPNOTSUPP;
3531
3532                 if (linkinfo[IFLA_INFO_DATA]) {
3533                         if (!ops || ops != dev->rtnl_link_ops ||
3534                             !ops->changelink)
3535                                 return -EOPNOTSUPP;
3536
3537                         err = ops->changelink(dev, tb, data, extack);
3538                         if (err < 0)
3539                                 return err;
3540                         status |= DO_SETLINK_NOTIFY;
3541                 }
3542
3543                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3544                         if (!m_ops || !m_ops->slave_changelink)
3545                                 return -EOPNOTSUPP;
3546
3547                         err = m_ops->slave_changelink(master_dev, dev, tb,
3548                                                       slave_data, extack);
3549                         if (err < 0)
3550                                 return err;
3551                         status |= DO_SETLINK_NOTIFY;
3552                 }
3553
3554                 return do_setlink(skb, dev, ifm, extack, tb, status);
3555         }
3556
3557         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3558                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3559                  * or it's for a group
3560                 */
3561                 if (link_specified)
3562                         return -ENODEV;
3563                 if (tb[IFLA_GROUP])
3564                         return rtnl_group_changelink(skb, net,
3565                                                 nla_get_u32(tb[IFLA_GROUP]),
3566                                                 ifm, extack, tb);
3567                 return -ENODEV;
3568         }
3569
3570         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3571                 return -EOPNOTSUPP;
3572
3573         if (!ops) {
3574 #ifdef CONFIG_MODULES
3575                 if (kind[0]) {
3576                         __rtnl_unlock();
3577                         request_module("rtnl-link-%s", kind);
3578                         rtnl_lock();
3579                         ops = rtnl_link_ops_get(kind);
3580                         if (ops)
3581                                 goto replay;
3582                 }
3583 #endif
3584                 NL_SET_ERR_MSG(extack, "Unknown device type");
3585                 return -EOPNOTSUPP;
3586         }
3587
3588         return rtnl_newlink_create(skb, ifm, ops, tb, data, extack);
3589 }
3590
3591 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3592                         struct netlink_ext_ack *extack)
3593 {
3594         struct rtnl_newlink_tbs *tbs;
3595         int ret;
3596
3597         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3598         if (!tbs)
3599                 return -ENOMEM;
3600
3601         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3602         kfree(tbs);
3603         return ret;
3604 }
3605
3606 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3607                                   const struct nlmsghdr *nlh,
3608                                   struct nlattr **tb,
3609                                   struct netlink_ext_ack *extack)
3610 {
3611         struct ifinfomsg *ifm;
3612         int i, err;
3613
3614         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3615                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3616                 return -EINVAL;
3617         }
3618
3619         if (!netlink_strict_get_check(skb))
3620                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3621                                               ifla_policy, extack);
3622
3623         ifm = nlmsg_data(nlh);
3624         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3625             ifm->ifi_change) {
3626                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3627                 return -EINVAL;
3628         }
3629
3630         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3631                                             ifla_policy, extack);
3632         if (err)
3633                 return err;
3634
3635         for (i = 0; i <= IFLA_MAX; i++) {
3636                 if (!tb[i])
3637                         continue;
3638
3639                 switch (i) {
3640                 case IFLA_IFNAME:
3641                 case IFLA_ALT_IFNAME:
3642                 case IFLA_EXT_MASK:
3643                 case IFLA_TARGET_NETNSID:
3644                         break;
3645                 default:
3646                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3647                         return -EINVAL;
3648                 }
3649         }
3650
3651         return 0;
3652 }
3653
3654 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3655                         struct netlink_ext_ack *extack)
3656 {
3657         struct net *net = sock_net(skb->sk);
3658         struct net *tgt_net = net;
3659         struct ifinfomsg *ifm;
3660         struct nlattr *tb[IFLA_MAX+1];
3661         struct net_device *dev = NULL;
3662         struct sk_buff *nskb;
3663         int netnsid = -1;
3664         int err;
3665         u32 ext_filter_mask = 0;
3666
3667         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3668         if (err < 0)
3669                 return err;
3670
3671         err = rtnl_ensure_unique_netns(tb, extack, true);
3672         if (err < 0)
3673                 return err;
3674
3675         if (tb[IFLA_TARGET_NETNSID]) {
3676                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3677                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3678                 if (IS_ERR(tgt_net))
3679                         return PTR_ERR(tgt_net);
3680         }
3681
3682         if (tb[IFLA_EXT_MASK])
3683                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3684
3685         err = -EINVAL;
3686         ifm = nlmsg_data(nlh);
3687         if (ifm->ifi_index > 0)
3688                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3689         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3690                 dev = rtnl_dev_get(tgt_net, tb);
3691         else
3692                 goto out;
3693
3694         err = -ENODEV;
3695         if (dev == NULL)
3696                 goto out;
3697
3698         err = -ENOBUFS;
3699         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3700         if (nskb == NULL)
3701                 goto out;
3702
3703         err = rtnl_fill_ifinfo(nskb, dev, net,
3704                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3705                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3706                                0, NULL, 0, netnsid, GFP_KERNEL);
3707         if (err < 0) {
3708                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3709                 WARN_ON(err == -EMSGSIZE);
3710                 kfree_skb(nskb);
3711         } else
3712                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3713 out:
3714         if (netnsid >= 0)
3715                 put_net(tgt_net);
3716
3717         return err;
3718 }
3719
3720 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3721                            bool *changed, struct netlink_ext_ack *extack)
3722 {
3723         char *alt_ifname;
3724         size_t size;
3725         int err;
3726
3727         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3728         if (err)
3729                 return err;
3730
3731         if (cmd == RTM_NEWLINKPROP) {
3732                 size = rtnl_prop_list_size(dev);
3733                 size += nla_total_size(ALTIFNAMSIZ);
3734                 if (size >= U16_MAX) {
3735                         NL_SET_ERR_MSG(extack,
3736                                        "effective property list too long");
3737                         return -EINVAL;
3738                 }
3739         }
3740
3741         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3742         if (!alt_ifname)
3743                 return -ENOMEM;
3744
3745         if (cmd == RTM_NEWLINKPROP) {
3746                 err = netdev_name_node_alt_create(dev, alt_ifname);
3747                 if (!err)
3748                         alt_ifname = NULL;
3749         } else if (cmd == RTM_DELLINKPROP) {
3750                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3751         } else {
3752                 WARN_ON_ONCE(1);
3753                 err = -EINVAL;
3754         }
3755
3756         kfree(alt_ifname);
3757         if (!err)
3758                 *changed = true;
3759         return err;
3760 }
3761
3762 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3763                          struct netlink_ext_ack *extack)
3764 {
3765         struct net *net = sock_net(skb->sk);
3766         struct nlattr *tb[IFLA_MAX + 1];
3767         struct net_device *dev;
3768         struct ifinfomsg *ifm;
3769         bool changed = false;
3770         struct nlattr *attr;
3771         int err, rem;
3772
3773         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3774         if (err)
3775                 return err;
3776
3777         err = rtnl_ensure_unique_netns(tb, extack, true);
3778         if (err)
3779                 return err;
3780
3781         ifm = nlmsg_data(nlh);
3782         if (ifm->ifi_index > 0)
3783                 dev = __dev_get_by_index(net, ifm->ifi_index);
3784         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3785                 dev = rtnl_dev_get(net, tb);
3786         else
3787                 return -EINVAL;
3788
3789         if (!dev)
3790                 return -ENODEV;
3791
3792         if (!tb[IFLA_PROP_LIST])
3793                 return 0;
3794
3795         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3796                 switch (nla_type(attr)) {
3797                 case IFLA_ALT_IFNAME:
3798                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3799                         if (err)
3800                                 return err;
3801                         break;
3802                 }
3803         }
3804
3805         if (changed)
3806                 netdev_state_change(dev);
3807         return 0;
3808 }
3809
3810 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3811                             struct netlink_ext_ack *extack)
3812 {
3813         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3814 }
3815
3816 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3817                             struct netlink_ext_ack *extack)
3818 {
3819         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3820 }
3821
3822 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3823 {
3824         struct net *net = sock_net(skb->sk);
3825         size_t min_ifinfo_dump_size = 0;
3826         struct nlattr *tb[IFLA_MAX+1];
3827         u32 ext_filter_mask = 0;
3828         struct net_device *dev;
3829         int hdrlen;
3830
3831         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3832         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3833                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3834
3835         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3836                 if (tb[IFLA_EXT_MASK])
3837                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3838         }
3839
3840         if (!ext_filter_mask)
3841                 return NLMSG_GOODSIZE;
3842         /*
3843          * traverse the list of net devices and compute the minimum
3844          * buffer size based upon the filter mask.
3845          */
3846         rcu_read_lock();
3847         for_each_netdev_rcu(net, dev) {
3848                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3849                                            if_nlmsg_size(dev, ext_filter_mask));
3850         }
3851         rcu_read_unlock();
3852
3853         return nlmsg_total_size(min_ifinfo_dump_size);
3854 }
3855
3856 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3857 {
3858         int idx;
3859         int s_idx = cb->family;
3860         int type = cb->nlh->nlmsg_type - RTM_BASE;
3861         int ret = 0;
3862
3863         if (s_idx == 0)
3864                 s_idx = 1;
3865
3866         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3867                 struct rtnl_link __rcu **tab;
3868                 struct rtnl_link *link;
3869                 rtnl_dumpit_func dumpit;
3870
3871                 if (idx < s_idx || idx == PF_PACKET)
3872                         continue;
3873
3874                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3875                         continue;
3876
3877                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3878                 if (!tab)
3879                         continue;
3880
3881                 link = rcu_dereference_rtnl(tab[type]);
3882                 if (!link)
3883                         continue;
3884
3885                 dumpit = link->dumpit;
3886                 if (!dumpit)
3887                         continue;
3888
3889                 if (idx > s_idx) {
3890                         memset(&cb->args[0], 0, sizeof(cb->args));
3891                         cb->prev_seq = 0;
3892                         cb->seq = 0;
3893                 }
3894                 ret = dumpit(skb, cb);
3895                 if (ret)
3896                         break;
3897         }
3898         cb->family = idx;
3899
3900         return skb->len ? : ret;
3901 }
3902
3903 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3904                                        unsigned int change,
3905                                        u32 event, gfp_t flags, int *new_nsid,
3906                                        int new_ifindex)
3907 {
3908         struct net *net = dev_net(dev);
3909         struct sk_buff *skb;
3910         int err = -ENOBUFS;
3911
3912         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3913         if (skb == NULL)
3914                 goto errout;
3915
3916         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3917                                type, 0, 0, change, 0, 0, event,
3918                                new_nsid, new_ifindex, -1, flags);
3919         if (err < 0) {
3920                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3921                 WARN_ON(err == -EMSGSIZE);
3922                 kfree_skb(skb);
3923                 goto errout;
3924         }
3925         return skb;
3926 errout:
3927         if (err < 0)
3928                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3929         return NULL;
3930 }
3931
3932 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3933 {
3934         struct net *net = dev_net(dev);
3935
3936         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3937 }
3938
3939 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3940                                unsigned int change, u32 event,
3941                                gfp_t flags, int *new_nsid, int new_ifindex)
3942 {
3943         struct sk_buff *skb;
3944
3945         if (dev->reg_state != NETREG_REGISTERED)
3946                 return;
3947
3948         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3949                                      new_ifindex);
3950         if (skb)
3951                 rtmsg_ifinfo_send(skb, dev, flags);
3952 }
3953
3954 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3955                   gfp_t flags)
3956 {
3957         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3958                            NULL, 0);
3959 }
3960
3961 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3962                          gfp_t flags, int *new_nsid, int new_ifindex)
3963 {
3964         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3965                            new_nsid, new_ifindex);
3966 }
3967
3968 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3969                                    struct net_device *dev,
3970                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3971                                    int type, unsigned int flags,
3972                                    int nlflags, u16 ndm_state)
3973 {
3974         struct nlmsghdr *nlh;
3975         struct ndmsg *ndm;
3976
3977         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3978         if (!nlh)
3979                 return -EMSGSIZE;
3980
3981         ndm = nlmsg_data(nlh);
3982         ndm->ndm_family  = AF_BRIDGE;
3983         ndm->ndm_pad1    = 0;
3984         ndm->ndm_pad2    = 0;
3985         ndm->ndm_flags   = flags;
3986         ndm->ndm_type    = 0;
3987         ndm->ndm_ifindex = dev->ifindex;
3988         ndm->ndm_state   = ndm_state;
3989
3990         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3991                 goto nla_put_failure;
3992         if (vid)
3993                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3994                         goto nla_put_failure;
3995
3996         nlmsg_end(skb, nlh);
3997         return 0;
3998
3999 nla_put_failure:
4000         nlmsg_cancel(skb, nlh);
4001         return -EMSGSIZE;
4002 }
4003
4004 static inline size_t rtnl_fdb_nlmsg_size(void)
4005 {
4006         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4007                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
4008                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4009                0;
4010 }
4011
4012 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4013                             u16 ndm_state)
4014 {
4015         struct net *net = dev_net(dev);
4016         struct sk_buff *skb;
4017         int err = -ENOBUFS;
4018
4019         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4020         if (!skb)
4021                 goto errout;
4022
4023         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4024                                       0, 0, type, NTF_SELF, 0, ndm_state);
4025         if (err < 0) {
4026                 kfree_skb(skb);
4027                 goto errout;
4028         }
4029
4030         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4031         return;
4032 errout:
4033         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4034 }
4035
4036 /*
4037  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4038  */
4039 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4040                      struct nlattr *tb[],
4041                      struct net_device *dev,
4042                      const unsigned char *addr, u16 vid,
4043                      u16 flags)
4044 {
4045         int err = -EINVAL;
4046
4047         /* If aging addresses are supported device will need to
4048          * implement its own handler for this.
4049          */
4050         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4051                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4052                 return err;
4053         }
4054
4055         if (vid) {
4056                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4057                 return err;
4058         }
4059
4060         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4061                 err = dev_uc_add_excl(dev, addr);
4062         else if (is_multicast_ether_addr(addr))
4063                 err = dev_mc_add_excl(dev, addr);
4064
4065         /* Only return duplicate errors if NLM_F_EXCL is set */
4066         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4067                 err = 0;
4068
4069         return err;
4070 }
4071 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4072
4073 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4074                          struct netlink_ext_ack *extack)
4075 {
4076         u16 vid = 0;
4077
4078         if (vlan_attr) {
4079                 if (nla_len(vlan_attr) != sizeof(u16)) {
4080                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4081                         return -EINVAL;
4082                 }
4083
4084                 vid = nla_get_u16(vlan_attr);
4085
4086                 if (!vid || vid >= VLAN_VID_MASK) {
4087                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4088                         return -EINVAL;
4089                 }
4090         }
4091         *p_vid = vid;
4092         return 0;
4093 }
4094
4095 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4096                         struct netlink_ext_ack *extack)
4097 {
4098         struct net *net = sock_net(skb->sk);
4099         struct ndmsg *ndm;
4100         struct nlattr *tb[NDA_MAX+1];
4101         struct net_device *dev;
4102         u8 *addr;
4103         u16 vid;
4104         int err;
4105
4106         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4107                                      extack);
4108         if (err < 0)
4109                 return err;
4110
4111         ndm = nlmsg_data(nlh);
4112         if (ndm->ndm_ifindex == 0) {
4113                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4114                 return -EINVAL;
4115         }
4116
4117         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4118         if (dev == NULL) {
4119                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4120                 return -ENODEV;
4121         }
4122
4123         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4124                 NL_SET_ERR_MSG(extack, "invalid address");
4125                 return -EINVAL;
4126         }
4127
4128         if (dev->type != ARPHRD_ETHER) {
4129                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4130                 return -EINVAL;
4131         }
4132
4133         addr = nla_data(tb[NDA_LLADDR]);
4134
4135         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4136         if (err)
4137                 return err;
4138
4139         err = -EOPNOTSUPP;
4140
4141         /* Support fdb on master device the net/bridge default case */
4142         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4143             netif_is_bridge_port(dev)) {
4144                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4145                 const struct net_device_ops *ops = br_dev->netdev_ops;
4146
4147                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4148                                        nlh->nlmsg_flags, extack);
4149                 if (err)
4150                         goto out;
4151                 else
4152                         ndm->ndm_flags &= ~NTF_MASTER;
4153         }
4154
4155         /* Embedded bridge, macvlan, and any other device support */
4156         if ((ndm->ndm_flags & NTF_SELF)) {
4157                 if (dev->netdev_ops->ndo_fdb_add)
4158                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4159                                                            vid,
4160                                                            nlh->nlmsg_flags,
4161                                                            extack);
4162                 else
4163                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4164                                                nlh->nlmsg_flags);
4165
4166                 if (!err) {
4167                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4168                                         ndm->ndm_state);
4169                         ndm->ndm_flags &= ~NTF_SELF;
4170                 }
4171         }
4172 out:
4173         return err;
4174 }
4175
4176 /*
4177  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4178  */
4179 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4180                      struct nlattr *tb[],
4181                      struct net_device *dev,
4182                      const unsigned char *addr, u16 vid)
4183 {
4184         int err = -EINVAL;
4185
4186         /* If aging addresses are supported device will need to
4187          * implement its own handler for this.
4188          */
4189         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4190                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4191                 return err;
4192         }
4193
4194         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4195                 err = dev_uc_del(dev, addr);
4196         else if (is_multicast_ether_addr(addr))
4197                 err = dev_mc_del(dev, addr);
4198
4199         return err;
4200 }
4201 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4202
4203 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4204         [NDA_VLAN]      = { .type = NLA_U16 },
4205         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4206         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4207         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4208 };
4209
4210 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4211                         struct netlink_ext_ack *extack)
4212 {
4213         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4214         struct net *net = sock_net(skb->sk);
4215         const struct net_device_ops *ops;
4216         struct ndmsg *ndm;
4217         struct nlattr *tb[NDA_MAX+1];
4218         struct net_device *dev;
4219         __u8 *addr = NULL;
4220         int err;
4221         u16 vid;
4222
4223         if (!netlink_capable(skb, CAP_NET_ADMIN))
4224                 return -EPERM;
4225
4226         if (!del_bulk) {
4227                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4228                                              NULL, extack);
4229         } else {
4230                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4231                                   fdb_del_bulk_policy, extack);
4232         }
4233         if (err < 0)
4234                 return err;
4235
4236         ndm = nlmsg_data(nlh);
4237         if (ndm->ndm_ifindex == 0) {
4238                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4239                 return -EINVAL;
4240         }
4241
4242         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4243         if (dev == NULL) {
4244                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4245                 return -ENODEV;
4246         }
4247
4248         if (!del_bulk) {
4249                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4250                         NL_SET_ERR_MSG(extack, "invalid address");
4251                         return -EINVAL;
4252                 }
4253                 addr = nla_data(tb[NDA_LLADDR]);
4254         }
4255
4256         if (dev->type != ARPHRD_ETHER) {
4257                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4258                 return -EINVAL;
4259         }
4260
4261         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4262         if (err)
4263                 return err;
4264
4265         err = -EOPNOTSUPP;
4266
4267         /* Support fdb on master device the net/bridge default case */
4268         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4269             netif_is_bridge_port(dev)) {
4270                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4271
4272                 ops = br_dev->netdev_ops;
4273                 if (!del_bulk) {
4274                         if (ops->ndo_fdb_del)
4275                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4276                 } else {
4277                         if (ops->ndo_fdb_del_bulk)
4278                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4279                                                             extack);
4280                 }
4281
4282                 if (err)
4283                         goto out;
4284                 else
4285                         ndm->ndm_flags &= ~NTF_MASTER;
4286         }
4287
4288         /* Embedded bridge, macvlan, and any other device support */
4289         if (ndm->ndm_flags & NTF_SELF) {
4290                 ops = dev->netdev_ops;
4291                 if (!del_bulk) {
4292                         if (ops->ndo_fdb_del)
4293                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4294                         else
4295                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4296                 } else {
4297                         /* in case err was cleared by NTF_MASTER call */
4298                         err = -EOPNOTSUPP;
4299                         if (ops->ndo_fdb_del_bulk)
4300                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4301                                                             extack);
4302                 }
4303
4304                 if (!err) {
4305                         if (!del_bulk)
4306                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4307                                                 ndm->ndm_state);
4308                         ndm->ndm_flags &= ~NTF_SELF;
4309                 }
4310         }
4311 out:
4312         return err;
4313 }
4314
4315 static int nlmsg_populate_fdb(struct sk_buff *skb,
4316                               struct netlink_callback *cb,
4317                               struct net_device *dev,
4318                               int *idx,
4319                               struct netdev_hw_addr_list *list)
4320 {
4321         struct netdev_hw_addr *ha;
4322         int err;
4323         u32 portid, seq;
4324
4325         portid = NETLINK_CB(cb->skb).portid;
4326         seq = cb->nlh->nlmsg_seq;
4327
4328         list_for_each_entry(ha, &list->list, list) {
4329                 if (*idx < cb->args[2])
4330                         goto skip;
4331
4332                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4333                                               portid, seq,
4334                                               RTM_NEWNEIGH, NTF_SELF,
4335                                               NLM_F_MULTI, NUD_PERMANENT);
4336                 if (err < 0)
4337                         return err;
4338 skip:
4339                 *idx += 1;
4340         }
4341         return 0;
4342 }
4343
4344 /**
4345  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4346  * @skb: socket buffer to store message in
4347  * @cb: netlink callback
4348  * @dev: netdevice
4349  * @filter_dev: ignored
4350  * @idx: the number of FDB table entries dumped is added to *@idx
4351  *
4352  * Default netdevice operation to dump the existing unicast address list.
4353  * Returns number of addresses from list put in skb.
4354  */
4355 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4356                       struct netlink_callback *cb,
4357                       struct net_device *dev,
4358                       struct net_device *filter_dev,
4359                       int *idx)
4360 {
4361         int err;
4362
4363         if (dev->type != ARPHRD_ETHER)
4364                 return -EINVAL;
4365
4366         netif_addr_lock_bh(dev);
4367         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4368         if (err)
4369                 goto out;
4370         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4371 out:
4372         netif_addr_unlock_bh(dev);
4373         return err;
4374 }
4375 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4376
4377 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4378                                  int *br_idx, int *brport_idx,
4379                                  struct netlink_ext_ack *extack)
4380 {
4381         struct nlattr *tb[NDA_MAX + 1];
4382         struct ndmsg *ndm;
4383         int err, i;
4384
4385         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4386                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4387                 return -EINVAL;
4388         }
4389
4390         ndm = nlmsg_data(nlh);
4391         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4392             ndm->ndm_flags || ndm->ndm_type) {
4393                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4394                 return -EINVAL;
4395         }
4396
4397         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4398                                             NDA_MAX, NULL, extack);
4399         if (err < 0)
4400                 return err;
4401
4402         *brport_idx = ndm->ndm_ifindex;
4403         for (i = 0; i <= NDA_MAX; ++i) {
4404                 if (!tb[i])
4405                         continue;
4406
4407                 switch (i) {
4408                 case NDA_IFINDEX:
4409                         if (nla_len(tb[i]) != sizeof(u32)) {
4410                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4411                                 return -EINVAL;
4412                         }
4413                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4414                         break;
4415                 case NDA_MASTER:
4416                         if (nla_len(tb[i]) != sizeof(u32)) {
4417                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4418                                 return -EINVAL;
4419                         }
4420                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4421                         break;
4422                 default:
4423                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4424                         return -EINVAL;
4425                 }
4426         }
4427
4428         return 0;
4429 }
4430
4431 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4432                                  int *br_idx, int *brport_idx,
4433                                  struct netlink_ext_ack *extack)
4434 {
4435         struct nlattr *tb[IFLA_MAX+1];
4436         int err;
4437
4438         /* A hack to preserve kernel<->userspace interface.
4439          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4440          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4441          * So, check for ndmsg with an optional u32 attribute (not used here).
4442          * Fortunately these sizes don't conflict with the size of ifinfomsg
4443          * with an optional attribute.
4444          */
4445         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4446             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4447              nla_attr_size(sizeof(u32)))) {
4448                 struct ifinfomsg *ifm;
4449
4450                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4451                                              tb, IFLA_MAX, ifla_policy,
4452                                              extack);
4453                 if (err < 0) {
4454                         return -EINVAL;
4455                 } else if (err == 0) {
4456                         if (tb[IFLA_MASTER])
4457                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4458                 }
4459
4460                 ifm = nlmsg_data(nlh);
4461                 *brport_idx = ifm->ifi_index;
4462         }
4463         return 0;
4464 }
4465
4466 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4467 {
4468         struct net_device *dev;
4469         struct net_device *br_dev = NULL;
4470         const struct net_device_ops *ops = NULL;
4471         const struct net_device_ops *cops = NULL;
4472         struct net *net = sock_net(skb->sk);
4473         struct hlist_head *head;
4474         int brport_idx = 0;
4475         int br_idx = 0;
4476         int h, s_h;
4477         int idx = 0, s_idx;
4478         int err = 0;
4479         int fidx = 0;
4480
4481         if (cb->strict_check)
4482                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4483                                             cb->extack);
4484         else
4485                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4486                                             cb->extack);
4487         if (err < 0)
4488                 return err;
4489
4490         if (br_idx) {
4491                 br_dev = __dev_get_by_index(net, br_idx);
4492                 if (!br_dev)
4493                         return -ENODEV;
4494
4495                 ops = br_dev->netdev_ops;
4496         }
4497
4498         s_h = cb->args[0];
4499         s_idx = cb->args[1];
4500
4501         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4502                 idx = 0;
4503                 head = &net->dev_index_head[h];
4504                 hlist_for_each_entry(dev, head, index_hlist) {
4505
4506                         if (brport_idx && (dev->ifindex != brport_idx))
4507                                 continue;
4508
4509                         if (!br_idx) { /* user did not specify a specific bridge */
4510                                 if (netif_is_bridge_port(dev)) {
4511                                         br_dev = netdev_master_upper_dev_get(dev);
4512                                         cops = br_dev->netdev_ops;
4513                                 }
4514                         } else {
4515                                 if (dev != br_dev &&
4516                                     !netif_is_bridge_port(dev))
4517                                         continue;
4518
4519                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4520                                     !netif_is_bridge_master(dev))
4521                                         continue;
4522                                 cops = ops;
4523                         }
4524
4525                         if (idx < s_idx)
4526                                 goto cont;
4527
4528                         if (netif_is_bridge_port(dev)) {
4529                                 if (cops && cops->ndo_fdb_dump) {
4530                                         err = cops->ndo_fdb_dump(skb, cb,
4531                                                                 br_dev, dev,
4532                                                                 &fidx);
4533                                         if (err == -EMSGSIZE)
4534                                                 goto out;
4535                                 }
4536                         }
4537
4538                         if (dev->netdev_ops->ndo_fdb_dump)
4539                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4540                                                                     dev, NULL,
4541                                                                     &fidx);
4542                         else
4543                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4544                                                         &fidx);
4545                         if (err == -EMSGSIZE)
4546                                 goto out;
4547
4548                         cops = NULL;
4549
4550                         /* reset fdb offset to 0 for rest of the interfaces */
4551                         cb->args[2] = 0;
4552                         fidx = 0;
4553 cont:
4554                         idx++;
4555                 }
4556         }
4557
4558 out:
4559         cb->args[0] = h;
4560         cb->args[1] = idx;
4561         cb->args[2] = fidx;
4562
4563         return skb->len;
4564 }
4565
4566 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4567                                 struct nlattr **tb, u8 *ndm_flags,
4568                                 int *br_idx, int *brport_idx, u8 **addr,
4569                                 u16 *vid, struct netlink_ext_ack *extack)
4570 {
4571         struct ndmsg *ndm;
4572         int err, i;
4573
4574         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4575                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4576                 return -EINVAL;
4577         }
4578
4579         ndm = nlmsg_data(nlh);
4580         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4581             ndm->ndm_type) {
4582                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4583                 return -EINVAL;
4584         }
4585
4586         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4587                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4588                 return -EINVAL;
4589         }
4590
4591         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4592                                             NDA_MAX, nda_policy, extack);
4593         if (err < 0)
4594                 return err;
4595
4596         *ndm_flags = ndm->ndm_flags;
4597         *brport_idx = ndm->ndm_ifindex;
4598         for (i = 0; i <= NDA_MAX; ++i) {
4599                 if (!tb[i])
4600                         continue;
4601
4602                 switch (i) {
4603                 case NDA_MASTER:
4604                         *br_idx = nla_get_u32(tb[i]);
4605                         break;
4606                 case NDA_LLADDR:
4607                         if (nla_len(tb[i]) != ETH_ALEN) {
4608                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4609                                 return -EINVAL;
4610                         }
4611                         *addr = nla_data(tb[i]);
4612                         break;
4613                 case NDA_VLAN:
4614                         err = fdb_vid_parse(tb[i], vid, extack);
4615                         if (err)
4616                                 return err;
4617                         break;
4618                 case NDA_VNI:
4619                         break;
4620                 default:
4621                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4622                         return -EINVAL;
4623                 }
4624         }
4625
4626         return 0;
4627 }
4628
4629 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4630                         struct netlink_ext_ack *extack)
4631 {
4632         struct net_device *dev = NULL, *br_dev = NULL;
4633         const struct net_device_ops *ops = NULL;
4634         struct net *net = sock_net(in_skb->sk);
4635         struct nlattr *tb[NDA_MAX + 1];
4636         struct sk_buff *skb;
4637         int brport_idx = 0;
4638         u8 ndm_flags = 0;
4639         int br_idx = 0;
4640         u8 *addr = NULL;
4641         u16 vid = 0;
4642         int err;
4643
4644         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4645                                    &brport_idx, &addr, &vid, extack);
4646         if (err < 0)
4647                 return err;
4648
4649         if (!addr) {
4650                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4651                 return -EINVAL;
4652         }
4653
4654         if (brport_idx) {
4655                 dev = __dev_get_by_index(net, brport_idx);
4656                 if (!dev) {
4657                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4658                         return -ENODEV;
4659                 }
4660         }
4661
4662         if (br_idx) {
4663                 if (dev) {
4664                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4665                         return -EINVAL;
4666                 }
4667
4668                 br_dev = __dev_get_by_index(net, br_idx);
4669                 if (!br_dev) {
4670                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4671                         return -EINVAL;
4672                 }
4673                 ops = br_dev->netdev_ops;
4674         }
4675
4676         if (dev) {
4677                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4678                         if (!netif_is_bridge_port(dev)) {
4679                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4680                                 return -EINVAL;
4681                         }
4682                         br_dev = netdev_master_upper_dev_get(dev);
4683                         if (!br_dev) {
4684                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4685                                 return -EINVAL;
4686                         }
4687                         ops = br_dev->netdev_ops;
4688                 } else {
4689                         if (!(ndm_flags & NTF_SELF)) {
4690                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4691                                 return -EINVAL;
4692                         }
4693                         ops = dev->netdev_ops;
4694                 }
4695         }
4696
4697         if (!br_dev && !dev) {
4698                 NL_SET_ERR_MSG(extack, "No device specified");
4699                 return -ENODEV;
4700         }
4701
4702         if (!ops || !ops->ndo_fdb_get) {
4703                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4704                 return -EOPNOTSUPP;
4705         }
4706
4707         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4708         if (!skb)
4709                 return -ENOBUFS;
4710
4711         if (br_dev)
4712                 dev = br_dev;
4713         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4714                                NETLINK_CB(in_skb).portid,
4715                                nlh->nlmsg_seq, extack);
4716         if (err)
4717                 goto out;
4718
4719         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4720 out:
4721         kfree_skb(skb);
4722         return err;
4723 }
4724
4725 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4726                                unsigned int attrnum, unsigned int flag)
4727 {
4728         if (mask & flag)
4729                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4730         return 0;
4731 }
4732
4733 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4734                             struct net_device *dev, u16 mode,
4735                             u32 flags, u32 mask, int nlflags,
4736                             u32 filter_mask,
4737                             int (*vlan_fill)(struct sk_buff *skb,
4738                                              struct net_device *dev,
4739                                              u32 filter_mask))
4740 {
4741         struct nlmsghdr *nlh;
4742         struct ifinfomsg *ifm;
4743         struct nlattr *br_afspec;
4744         struct nlattr *protinfo;
4745         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4746         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4747         int err = 0;
4748
4749         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4750         if (nlh == NULL)
4751                 return -EMSGSIZE;
4752
4753         ifm = nlmsg_data(nlh);
4754         ifm->ifi_family = AF_BRIDGE;
4755         ifm->__ifi_pad = 0;
4756         ifm->ifi_type = dev->type;
4757         ifm->ifi_index = dev->ifindex;
4758         ifm->ifi_flags = dev_get_flags(dev);
4759         ifm->ifi_change = 0;
4760
4761
4762         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4763             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4764             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4765             (br_dev &&
4766              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4767             (dev->addr_len &&
4768              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4769             (dev->ifindex != dev_get_iflink(dev) &&
4770              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4771                 goto nla_put_failure;
4772
4773         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4774         if (!br_afspec)
4775                 goto nla_put_failure;
4776
4777         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4778                 nla_nest_cancel(skb, br_afspec);
4779                 goto nla_put_failure;
4780         }
4781
4782         if (mode != BRIDGE_MODE_UNDEF) {
4783                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4784                         nla_nest_cancel(skb, br_afspec);
4785                         goto nla_put_failure;
4786                 }
4787         }
4788         if (vlan_fill) {
4789                 err = vlan_fill(skb, dev, filter_mask);
4790                 if (err) {
4791                         nla_nest_cancel(skb, br_afspec);
4792                         goto nla_put_failure;
4793                 }
4794         }
4795         nla_nest_end(skb, br_afspec);
4796
4797         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4798         if (!protinfo)
4799                 goto nla_put_failure;
4800
4801         if (brport_nla_put_flag(skb, flags, mask,
4802                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4803             brport_nla_put_flag(skb, flags, mask,
4804                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4805             brport_nla_put_flag(skb, flags, mask,
4806                                 IFLA_BRPORT_FAST_LEAVE,
4807                                 BR_MULTICAST_FAST_LEAVE) ||
4808             brport_nla_put_flag(skb, flags, mask,
4809                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4810             brport_nla_put_flag(skb, flags, mask,
4811                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4812             brport_nla_put_flag(skb, flags, mask,
4813                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4814             brport_nla_put_flag(skb, flags, mask,
4815                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4816             brport_nla_put_flag(skb, flags, mask,
4817                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4818             brport_nla_put_flag(skb, flags, mask,
4819                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4820             brport_nla_put_flag(skb, flags, mask,
4821                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4822                 nla_nest_cancel(skb, protinfo);
4823                 goto nla_put_failure;
4824         }
4825
4826         nla_nest_end(skb, protinfo);
4827
4828         nlmsg_end(skb, nlh);
4829         return 0;
4830 nla_put_failure:
4831         nlmsg_cancel(skb, nlh);
4832         return err ? err : -EMSGSIZE;
4833 }
4834 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4835
4836 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4837                                     bool strict_check, u32 *filter_mask,
4838                                     struct netlink_ext_ack *extack)
4839 {
4840         struct nlattr *tb[IFLA_MAX+1];
4841         int err, i;
4842
4843         if (strict_check) {
4844                 struct ifinfomsg *ifm;
4845
4846                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4847                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4848                         return -EINVAL;
4849                 }
4850
4851                 ifm = nlmsg_data(nlh);
4852                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4853                     ifm->ifi_change || ifm->ifi_index) {
4854                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4855                         return -EINVAL;
4856                 }
4857
4858                 err = nlmsg_parse_deprecated_strict(nlh,
4859                                                     sizeof(struct ifinfomsg),
4860                                                     tb, IFLA_MAX, ifla_policy,
4861                                                     extack);
4862         } else {
4863                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4864                                              tb, IFLA_MAX, ifla_policy,
4865                                              extack);
4866         }
4867         if (err < 0)
4868                 return err;
4869
4870         /* new attributes should only be added with strict checking */
4871         for (i = 0; i <= IFLA_MAX; ++i) {
4872                 if (!tb[i])
4873                         continue;
4874
4875                 switch (i) {
4876                 case IFLA_EXT_MASK:
4877                         *filter_mask = nla_get_u32(tb[i]);
4878                         break;
4879                 default:
4880                         if (strict_check) {
4881                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4882                                 return -EINVAL;
4883                         }
4884                 }
4885         }
4886
4887         return 0;
4888 }
4889
4890 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4891 {
4892         const struct nlmsghdr *nlh = cb->nlh;
4893         struct net *net = sock_net(skb->sk);
4894         struct net_device *dev;
4895         int idx = 0;
4896         u32 portid = NETLINK_CB(cb->skb).portid;
4897         u32 seq = nlh->nlmsg_seq;
4898         u32 filter_mask = 0;
4899         int err;
4900
4901         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4902                                        cb->extack);
4903         if (err < 0 && cb->strict_check)
4904                 return err;
4905
4906         rcu_read_lock();
4907         for_each_netdev_rcu(net, dev) {
4908                 const struct net_device_ops *ops = dev->netdev_ops;
4909                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4910
4911                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4912                         if (idx >= cb->args[0]) {
4913                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4914                                                 skb, portid, seq, dev,
4915                                                 filter_mask, NLM_F_MULTI);
4916                                 if (err < 0 && err != -EOPNOTSUPP) {
4917                                         if (likely(skb->len))
4918                                                 break;
4919
4920                                         goto out_err;
4921                                 }
4922                         }
4923                         idx++;
4924                 }
4925
4926                 if (ops->ndo_bridge_getlink) {
4927                         if (idx >= cb->args[0]) {
4928                                 err = ops->ndo_bridge_getlink(skb, portid,
4929                                                               seq, dev,
4930                                                               filter_mask,
4931                                                               NLM_F_MULTI);
4932                                 if (err < 0 && err != -EOPNOTSUPP) {
4933                                         if (likely(skb->len))
4934                                                 break;
4935
4936                                         goto out_err;
4937                                 }
4938                         }
4939                         idx++;
4940                 }
4941         }
4942         err = skb->len;
4943 out_err:
4944         rcu_read_unlock();
4945         cb->args[0] = idx;
4946
4947         return err;
4948 }
4949
4950 static inline size_t bridge_nlmsg_size(void)
4951 {
4952         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4953                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4954                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4955                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4956                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4957                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4958                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4959                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4960                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4961                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4962                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4963 }
4964
4965 static int rtnl_bridge_notify(struct net_device *dev)
4966 {
4967         struct net *net = dev_net(dev);
4968         struct sk_buff *skb;
4969         int err = -EOPNOTSUPP;
4970
4971         if (!dev->netdev_ops->ndo_bridge_getlink)
4972                 return 0;
4973
4974         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4975         if (!skb) {
4976                 err = -ENOMEM;
4977                 goto errout;
4978         }
4979
4980         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4981         if (err < 0)
4982                 goto errout;
4983
4984         /* Notification info is only filled for bridge ports, not the bridge
4985          * device itself. Therefore, a zero notification length is valid and
4986          * should not result in an error.
4987          */
4988         if (!skb->len)
4989                 goto errout;
4990
4991         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4992         return 0;
4993 errout:
4994         WARN_ON(err == -EMSGSIZE);
4995         kfree_skb(skb);
4996         if (err)
4997                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4998         return err;
4999 }
5000
5001 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5002                                struct netlink_ext_ack *extack)
5003 {
5004         struct net *net = sock_net(skb->sk);
5005         struct ifinfomsg *ifm;
5006         struct net_device *dev;
5007         struct nlattr *br_spec, *attr = NULL;
5008         int rem, err = -EOPNOTSUPP;
5009         u16 flags = 0;
5010         bool have_flags = false;
5011
5012         if (nlmsg_len(nlh) < sizeof(*ifm))
5013                 return -EINVAL;
5014
5015         ifm = nlmsg_data(nlh);
5016         if (ifm->ifi_family != AF_BRIDGE)
5017                 return -EPFNOSUPPORT;
5018
5019         dev = __dev_get_by_index(net, ifm->ifi_index);
5020         if (!dev) {
5021                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5022                 return -ENODEV;
5023         }
5024
5025         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5026         if (br_spec) {
5027                 nla_for_each_nested(attr, br_spec, rem) {
5028                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5029                                 if (nla_len(attr) < sizeof(flags))
5030                                         return -EINVAL;
5031
5032                                 have_flags = true;
5033                                 flags = nla_get_u16(attr);
5034                                 break;
5035                         }
5036                 }
5037         }
5038
5039         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5040                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5041
5042                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5043                         err = -EOPNOTSUPP;
5044                         goto out;
5045                 }
5046
5047                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5048                                                              extack);
5049                 if (err)
5050                         goto out;
5051
5052                 flags &= ~BRIDGE_FLAGS_MASTER;
5053         }
5054
5055         if ((flags & BRIDGE_FLAGS_SELF)) {
5056                 if (!dev->netdev_ops->ndo_bridge_setlink)
5057                         err = -EOPNOTSUPP;
5058                 else
5059                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5060                                                                   flags,
5061                                                                   extack);
5062                 if (!err) {
5063                         flags &= ~BRIDGE_FLAGS_SELF;
5064
5065                         /* Generate event to notify upper layer of bridge
5066                          * change
5067                          */
5068                         err = rtnl_bridge_notify(dev);
5069                 }
5070         }
5071
5072         if (have_flags)
5073                 memcpy(nla_data(attr), &flags, sizeof(flags));
5074 out:
5075         return err;
5076 }
5077
5078 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5079                                struct netlink_ext_ack *extack)
5080 {
5081         struct net *net = sock_net(skb->sk);
5082         struct ifinfomsg *ifm;
5083         struct net_device *dev;
5084         struct nlattr *br_spec, *attr = NULL;
5085         int rem, err = -EOPNOTSUPP;
5086         u16 flags = 0;
5087         bool have_flags = false;
5088
5089         if (nlmsg_len(nlh) < sizeof(*ifm))
5090                 return -EINVAL;
5091
5092         ifm = nlmsg_data(nlh);
5093         if (ifm->ifi_family != AF_BRIDGE)
5094                 return -EPFNOSUPPORT;
5095
5096         dev = __dev_get_by_index(net, ifm->ifi_index);
5097         if (!dev) {
5098                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5099                 return -ENODEV;
5100         }
5101
5102         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5103         if (br_spec) {
5104                 nla_for_each_nested(attr, br_spec, rem) {
5105                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5106                                 if (nla_len(attr) < sizeof(flags))
5107                                         return -EINVAL;
5108
5109                                 have_flags = true;
5110                                 flags = nla_get_u16(attr);
5111                                 break;
5112                         }
5113                 }
5114         }
5115
5116         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5117                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5118
5119                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5120                         err = -EOPNOTSUPP;
5121                         goto out;
5122                 }
5123
5124                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5125                 if (err)
5126                         goto out;
5127
5128                 flags &= ~BRIDGE_FLAGS_MASTER;
5129         }
5130
5131         if ((flags & BRIDGE_FLAGS_SELF)) {
5132                 if (!dev->netdev_ops->ndo_bridge_dellink)
5133                         err = -EOPNOTSUPP;
5134                 else
5135                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5136                                                                   flags);
5137
5138                 if (!err) {
5139                         flags &= ~BRIDGE_FLAGS_SELF;
5140
5141                         /* Generate event to notify upper layer of bridge
5142                          * change
5143                          */
5144                         err = rtnl_bridge_notify(dev);
5145                 }
5146         }
5147
5148         if (have_flags)
5149                 memcpy(nla_data(attr), &flags, sizeof(flags));
5150 out:
5151         return err;
5152 }
5153
5154 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5155 {
5156         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5157                (!idxattr || idxattr == attrid);
5158 }
5159
5160 static bool
5161 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5162 {
5163         return dev->netdev_ops &&
5164                dev->netdev_ops->ndo_has_offload_stats &&
5165                dev->netdev_ops->ndo_get_offload_stats &&
5166                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5167 }
5168
5169 static unsigned int
5170 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5171 {
5172         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5173                sizeof(struct rtnl_link_stats64) : 0;
5174 }
5175
5176 static int
5177 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5178                              struct sk_buff *skb)
5179 {
5180         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5181         struct nlattr *attr = NULL;
5182         void *attr_data;
5183         int err;
5184
5185         if (!size)
5186                 return -ENODATA;
5187
5188         attr = nla_reserve_64bit(skb, attr_id, size,
5189                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5190         if (!attr)
5191                 return -EMSGSIZE;
5192
5193         attr_data = nla_data(attr);
5194         memset(attr_data, 0, size);
5195
5196         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5197         if (err)
5198                 return err;
5199
5200         return 0;
5201 }
5202
5203 static unsigned int
5204 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5205                                    enum netdev_offload_xstats_type type)
5206 {
5207         bool enabled = netdev_offload_xstats_enabled(dev, type);
5208
5209         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5210 }
5211
5212 struct rtnl_offload_xstats_request_used {
5213         bool request;
5214         bool used;
5215 };
5216
5217 static int
5218 rtnl_offload_xstats_get_stats(struct net_device *dev,
5219                               enum netdev_offload_xstats_type type,
5220                               struct rtnl_offload_xstats_request_used *ru,
5221                               struct rtnl_hw_stats64 *stats,
5222                               struct netlink_ext_ack *extack)
5223 {
5224         bool request;
5225         bool used;
5226         int err;
5227
5228         request = netdev_offload_xstats_enabled(dev, type);
5229         if (!request) {
5230                 used = false;
5231                 goto out;
5232         }
5233
5234         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5235         if (err)
5236                 return err;
5237
5238 out:
5239         if (ru) {
5240                 ru->request = request;
5241                 ru->used = used;
5242         }
5243         return 0;
5244 }
5245
5246 static int
5247 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5248                                        struct rtnl_offload_xstats_request_used *ru)
5249 {
5250         struct nlattr *nest;
5251
5252         nest = nla_nest_start(skb, attr_id);
5253         if (!nest)
5254                 return -EMSGSIZE;
5255
5256         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5257                 goto nla_put_failure;
5258
5259         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5260                 goto nla_put_failure;
5261
5262         nla_nest_end(skb, nest);
5263         return 0;
5264
5265 nla_put_failure:
5266         nla_nest_cancel(skb, nest);
5267         return -EMSGSIZE;
5268 }
5269
5270 static int
5271 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5272                                    struct netlink_ext_ack *extack)
5273 {
5274         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5275         struct rtnl_offload_xstats_request_used ru_l3;
5276         struct nlattr *nest;
5277         int err;
5278
5279         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5280         if (err)
5281                 return err;
5282
5283         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5284         if (!nest)
5285                 return -EMSGSIZE;
5286
5287         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5288                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5289                                                    &ru_l3))
5290                 goto nla_put_failure;
5291
5292         nla_nest_end(skb, nest);
5293         return 0;
5294
5295 nla_put_failure:
5296         nla_nest_cancel(skb, nest);
5297         return -EMSGSIZE;
5298 }
5299
5300 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5301                                     int *prividx, u32 off_filter_mask,
5302                                     struct netlink_ext_ack *extack)
5303 {
5304         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5305         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5306         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5307         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5308         bool have_data = false;
5309         int err;
5310
5311         if (*prividx <= attr_id_cpu_hit &&
5312             (off_filter_mask &
5313              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5314                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5315                 if (!err) {
5316                         have_data = true;
5317                 } else if (err != -ENODATA) {
5318                         *prividx = attr_id_cpu_hit;
5319                         return err;
5320                 }
5321         }
5322
5323         if (*prividx <= attr_id_hw_s_info &&
5324             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5325                 *prividx = attr_id_hw_s_info;
5326
5327                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5328                 if (err)
5329                         return err;
5330
5331                 have_data = true;
5332                 *prividx = 0;
5333         }
5334
5335         if (*prividx <= attr_id_l3_stats &&
5336             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5337                 unsigned int size_l3;
5338                 struct nlattr *attr;
5339
5340                 *prividx = attr_id_l3_stats;
5341
5342                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5343                 if (!size_l3)
5344                         goto skip_l3_stats;
5345                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5346                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5347                 if (!attr)
5348                         return -EMSGSIZE;
5349
5350                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5351                                                     nla_data(attr), extack);
5352                 if (err)
5353                         return err;
5354
5355                 have_data = true;
5356 skip_l3_stats:
5357                 *prividx = 0;
5358         }
5359
5360         if (!have_data)
5361                 return -ENODATA;
5362
5363         *prividx = 0;
5364         return 0;
5365 }
5366
5367 static unsigned int
5368 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5369                                            enum netdev_offload_xstats_type type)
5370 {
5371         bool enabled = netdev_offload_xstats_enabled(dev, type);
5372
5373         return nla_total_size(0) +
5374                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5375                 nla_total_size(sizeof(u8)) +
5376                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5377                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5378                 0;
5379 }
5380
5381 static unsigned int
5382 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5383 {
5384         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5385
5386         return nla_total_size(0) +
5387                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5388                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5389                 0;
5390 }
5391
5392 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5393                                         u32 off_filter_mask)
5394 {
5395         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5396         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5397         int nla_size = 0;
5398         int size;
5399
5400         if (off_filter_mask &
5401             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5402                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5403                 nla_size += nla_total_size_64bit(size);
5404         }
5405
5406         if (off_filter_mask &
5407             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5408                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5409
5410         if (off_filter_mask &
5411             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5412                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5413                 nla_size += nla_total_size_64bit(size);
5414         }
5415
5416         if (nla_size != 0)
5417                 nla_size += nla_total_size(0);
5418
5419         return nla_size;
5420 }
5421
5422 struct rtnl_stats_dump_filters {
5423         /* mask[0] filters outer attributes. Then individual nests have their
5424          * filtering mask at the index of the nested attribute.
5425          */
5426         u32 mask[IFLA_STATS_MAX + 1];
5427 };
5428
5429 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5430                                int type, u32 pid, u32 seq, u32 change,
5431                                unsigned int flags,
5432                                const struct rtnl_stats_dump_filters *filters,
5433                                int *idxattr, int *prividx,
5434                                struct netlink_ext_ack *extack)
5435 {
5436         unsigned int filter_mask = filters->mask[0];
5437         struct if_stats_msg *ifsm;
5438         struct nlmsghdr *nlh;
5439         struct nlattr *attr;
5440         int s_prividx = *prividx;
5441         int err;
5442
5443         ASSERT_RTNL();
5444
5445         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5446         if (!nlh)
5447                 return -EMSGSIZE;
5448
5449         ifsm = nlmsg_data(nlh);
5450         ifsm->family = PF_UNSPEC;
5451         ifsm->pad1 = 0;
5452         ifsm->pad2 = 0;
5453         ifsm->ifindex = dev->ifindex;
5454         ifsm->filter_mask = filter_mask;
5455
5456         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5457                 struct rtnl_link_stats64 *sp;
5458
5459                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5460                                          sizeof(struct rtnl_link_stats64),
5461                                          IFLA_STATS_UNSPEC);
5462                 if (!attr) {
5463                         err = -EMSGSIZE;
5464                         goto nla_put_failure;
5465                 }
5466
5467                 sp = nla_data(attr);
5468                 dev_get_stats(dev, sp);
5469         }
5470
5471         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5472                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5473
5474                 if (ops && ops->fill_linkxstats) {
5475                         *idxattr = IFLA_STATS_LINK_XSTATS;
5476                         attr = nla_nest_start_noflag(skb,
5477                                                      IFLA_STATS_LINK_XSTATS);
5478                         if (!attr) {
5479                                 err = -EMSGSIZE;
5480                                 goto nla_put_failure;
5481                         }
5482
5483                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5484                         nla_nest_end(skb, attr);
5485                         if (err)
5486                                 goto nla_put_failure;
5487                         *idxattr = 0;
5488                 }
5489         }
5490
5491         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5492                              *idxattr)) {
5493                 const struct rtnl_link_ops *ops = NULL;
5494                 const struct net_device *master;
5495
5496                 master = netdev_master_upper_dev_get(dev);
5497                 if (master)
5498                         ops = master->rtnl_link_ops;
5499                 if (ops && ops->fill_linkxstats) {
5500                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5501                         attr = nla_nest_start_noflag(skb,
5502                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5503                         if (!attr) {
5504                                 err = -EMSGSIZE;
5505                                 goto nla_put_failure;
5506                         }
5507
5508                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5509                         nla_nest_end(skb, attr);
5510                         if (err)
5511                                 goto nla_put_failure;
5512                         *idxattr = 0;
5513                 }
5514         }
5515
5516         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5517                              *idxattr)) {
5518                 u32 off_filter_mask;
5519
5520                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5521                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5522                 attr = nla_nest_start_noflag(skb,
5523                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5524                 if (!attr) {
5525                         err = -EMSGSIZE;
5526                         goto nla_put_failure;
5527                 }
5528
5529                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5530                                                off_filter_mask, extack);
5531                 if (err == -ENODATA)
5532                         nla_nest_cancel(skb, attr);
5533                 else
5534                         nla_nest_end(skb, attr);
5535
5536                 if (err && err != -ENODATA)
5537                         goto nla_put_failure;
5538                 *idxattr = 0;
5539         }
5540
5541         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5542                 struct rtnl_af_ops *af_ops;
5543
5544                 *idxattr = IFLA_STATS_AF_SPEC;
5545                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5546                 if (!attr) {
5547                         err = -EMSGSIZE;
5548                         goto nla_put_failure;
5549                 }
5550
5551                 rcu_read_lock();
5552                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5553                         if (af_ops->fill_stats_af) {
5554                                 struct nlattr *af;
5555
5556                                 af = nla_nest_start_noflag(skb,
5557                                                            af_ops->family);
5558                                 if (!af) {
5559                                         rcu_read_unlock();
5560                                         err = -EMSGSIZE;
5561                                         goto nla_put_failure;
5562                                 }
5563                                 err = af_ops->fill_stats_af(skb, dev);
5564
5565                                 if (err == -ENODATA) {
5566                                         nla_nest_cancel(skb, af);
5567                                 } else if (err < 0) {
5568                                         rcu_read_unlock();
5569                                         goto nla_put_failure;
5570                                 }
5571
5572                                 nla_nest_end(skb, af);
5573                         }
5574                 }
5575                 rcu_read_unlock();
5576
5577                 nla_nest_end(skb, attr);
5578
5579                 *idxattr = 0;
5580         }
5581
5582         nlmsg_end(skb, nlh);
5583
5584         return 0;
5585
5586 nla_put_failure:
5587         /* not a multi message or no progress mean a real error */
5588         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5589                 nlmsg_cancel(skb, nlh);
5590         else
5591                 nlmsg_end(skb, nlh);
5592
5593         return err;
5594 }
5595
5596 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5597                                   const struct rtnl_stats_dump_filters *filters)
5598 {
5599         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5600         unsigned int filter_mask = filters->mask[0];
5601
5602         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5603                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5604
5605         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5606                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5607                 int attr = IFLA_STATS_LINK_XSTATS;
5608
5609                 if (ops && ops->get_linkxstats_size) {
5610                         size += nla_total_size(ops->get_linkxstats_size(dev,
5611                                                                         attr));
5612                         /* for IFLA_STATS_LINK_XSTATS */
5613                         size += nla_total_size(0);
5614                 }
5615         }
5616
5617         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5618                 struct net_device *_dev = (struct net_device *)dev;
5619                 const struct rtnl_link_ops *ops = NULL;
5620                 const struct net_device *master;
5621
5622                 /* netdev_master_upper_dev_get can't take const */
5623                 master = netdev_master_upper_dev_get(_dev);
5624                 if (master)
5625                         ops = master->rtnl_link_ops;
5626                 if (ops && ops->get_linkxstats_size) {
5627                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5628
5629                         size += nla_total_size(ops->get_linkxstats_size(dev,
5630                                                                         attr));
5631                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5632                         size += nla_total_size(0);
5633                 }
5634         }
5635
5636         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5637                 u32 off_filter_mask;
5638
5639                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5640                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5641         }
5642
5643         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5644                 struct rtnl_af_ops *af_ops;
5645
5646                 /* for IFLA_STATS_AF_SPEC */
5647                 size += nla_total_size(0);
5648
5649                 rcu_read_lock();
5650                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5651                         if (af_ops->get_stats_af_size) {
5652                                 size += nla_total_size(
5653                                         af_ops->get_stats_af_size(dev));
5654
5655                                 /* for AF_* */
5656                                 size += nla_total_size(0);
5657                         }
5658                 }
5659                 rcu_read_unlock();
5660         }
5661
5662         return size;
5663 }
5664
5665 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5666
5667 static const struct nla_policy
5668 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5669         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5670                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5671 };
5672
5673 static const struct nla_policy
5674 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5675         [IFLA_STATS_GET_FILTERS] =
5676                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5677 };
5678
5679 static const struct nla_policy
5680 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5681         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5682 };
5683
5684 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5685                                         struct rtnl_stats_dump_filters *filters,
5686                                         struct netlink_ext_ack *extack)
5687 {
5688         struct nlattr *tb[IFLA_STATS_MAX + 1];
5689         int err;
5690         int at;
5691
5692         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5693                                rtnl_stats_get_policy_filters, extack);
5694         if (err < 0)
5695                 return err;
5696
5697         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5698                 if (tb[at]) {
5699                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5700                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5701                                 return -EINVAL;
5702                         }
5703                         filters->mask[at] = nla_get_u32(tb[at]);
5704                 }
5705         }
5706
5707         return 0;
5708 }
5709
5710 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5711                                 u32 filter_mask,
5712                                 struct rtnl_stats_dump_filters *filters,
5713                                 struct netlink_ext_ack *extack)
5714 {
5715         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5716         int err;
5717         int i;
5718
5719         filters->mask[0] = filter_mask;
5720         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5721                 filters->mask[i] = -1U;
5722
5723         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5724                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5725         if (err < 0)
5726                 return err;
5727
5728         if (tb[IFLA_STATS_GET_FILTERS]) {
5729                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5730                                                    filters, extack);
5731                 if (err)
5732                         return err;
5733         }
5734
5735         return 0;
5736 }
5737
5738 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5739                                 bool is_dump, struct netlink_ext_ack *extack)
5740 {
5741         struct if_stats_msg *ifsm;
5742
5743         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5744                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5745                 return -EINVAL;
5746         }
5747
5748         if (!strict_check)
5749                 return 0;
5750
5751         ifsm = nlmsg_data(nlh);
5752
5753         /* only requests using strict checks can pass data to influence
5754          * the dump. The legacy exception is filter_mask.
5755          */
5756         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5757                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5758                 return -EINVAL;
5759         }
5760         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5761                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5762                 return -EINVAL;
5763         }
5764
5765         return 0;
5766 }
5767
5768 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5769                           struct netlink_ext_ack *extack)
5770 {
5771         struct rtnl_stats_dump_filters filters;
5772         struct net *net = sock_net(skb->sk);
5773         struct net_device *dev = NULL;
5774         int idxattr = 0, prividx = 0;
5775         struct if_stats_msg *ifsm;
5776         struct sk_buff *nskb;
5777         int err;
5778
5779         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5780                                    false, extack);
5781         if (err)
5782                 return err;
5783
5784         ifsm = nlmsg_data(nlh);
5785         if (ifsm->ifindex > 0)
5786                 dev = __dev_get_by_index(net, ifsm->ifindex);
5787         else
5788                 return -EINVAL;
5789
5790         if (!dev)
5791                 return -ENODEV;
5792
5793         if (!ifsm->filter_mask) {
5794                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5795                 return -EINVAL;
5796         }
5797
5798         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5799         if (err)
5800                 return err;
5801
5802         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5803         if (!nskb)
5804                 return -ENOBUFS;
5805
5806         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5807                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5808                                   0, &filters, &idxattr, &prividx, extack);
5809         if (err < 0) {
5810                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5811                 WARN_ON(err == -EMSGSIZE);
5812                 kfree_skb(nskb);
5813         } else {
5814                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5815         }
5816
5817         return err;
5818 }
5819
5820 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5821 {
5822         struct netlink_ext_ack *extack = cb->extack;
5823         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5824         struct rtnl_stats_dump_filters filters;
5825         struct net *net = sock_net(skb->sk);
5826         unsigned int flags = NLM_F_MULTI;
5827         struct if_stats_msg *ifsm;
5828         struct hlist_head *head;
5829         struct net_device *dev;
5830         int idx = 0;
5831
5832         s_h = cb->args[0];
5833         s_idx = cb->args[1];
5834         s_idxattr = cb->args[2];
5835         s_prividx = cb->args[3];
5836
5837         cb->seq = net->dev_base_seq;
5838
5839         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5840         if (err)
5841                 return err;
5842
5843         ifsm = nlmsg_data(cb->nlh);
5844         if (!ifsm->filter_mask) {
5845                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5846                 return -EINVAL;
5847         }
5848
5849         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5850                                    extack);
5851         if (err)
5852                 return err;
5853
5854         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5855                 idx = 0;
5856                 head = &net->dev_index_head[h];
5857                 hlist_for_each_entry(dev, head, index_hlist) {
5858                         if (idx < s_idx)
5859                                 goto cont;
5860                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5861                                                   NETLINK_CB(cb->skb).portid,
5862                                                   cb->nlh->nlmsg_seq, 0,
5863                                                   flags, &filters,
5864                                                   &s_idxattr, &s_prividx,
5865                                                   extack);
5866                         /* If we ran out of room on the first message,
5867                          * we're in trouble
5868                          */
5869                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5870
5871                         if (err < 0)
5872                                 goto out;
5873                         s_prividx = 0;
5874                         s_idxattr = 0;
5875                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5876 cont:
5877                         idx++;
5878                 }
5879         }
5880 out:
5881         cb->args[3] = s_prividx;
5882         cb->args[2] = s_idxattr;
5883         cb->args[1] = idx;
5884         cb->args[0] = h;
5885
5886         return skb->len;
5887 }
5888
5889 void rtnl_offload_xstats_notify(struct net_device *dev)
5890 {
5891         struct rtnl_stats_dump_filters response_filters = {};
5892         struct net *net = dev_net(dev);
5893         int idxattr = 0, prividx = 0;
5894         struct sk_buff *skb;
5895         int err = -ENOBUFS;
5896
5897         ASSERT_RTNL();
5898
5899         response_filters.mask[0] |=
5900                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5901         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5902                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5903
5904         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
5905                         GFP_KERNEL);
5906         if (!skb)
5907                 goto errout;
5908
5909         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
5910                                   &response_filters, &idxattr, &prividx, NULL);
5911         if (err < 0) {
5912                 kfree_skb(skb);
5913                 goto errout;
5914         }
5915
5916         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
5917         return;
5918
5919 errout:
5920         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
5921 }
5922 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
5923
5924 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
5925                           struct netlink_ext_ack *extack)
5926 {
5927         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5928         struct rtnl_stats_dump_filters response_filters = {};
5929         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5930         struct net *net = sock_net(skb->sk);
5931         struct net_device *dev = NULL;
5932         struct if_stats_msg *ifsm;
5933         bool notify = false;
5934         int err;
5935
5936         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5937                                    false, extack);
5938         if (err)
5939                 return err;
5940
5941         ifsm = nlmsg_data(nlh);
5942         if (ifsm->family != AF_UNSPEC) {
5943                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
5944                 return -EINVAL;
5945         }
5946
5947         if (ifsm->ifindex > 0)
5948                 dev = __dev_get_by_index(net, ifsm->ifindex);
5949         else
5950                 return -EINVAL;
5951
5952         if (!dev)
5953                 return -ENODEV;
5954
5955         if (ifsm->filter_mask) {
5956                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
5957                 return -EINVAL;
5958         }
5959
5960         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
5961                           ifla_stats_set_policy, extack);
5962         if (err < 0)
5963                 return err;
5964
5965         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
5966                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
5967
5968                 if (req)
5969                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
5970                 else
5971                         err = netdev_offload_xstats_disable(dev, t_l3);
5972
5973                 if (!err)
5974                         notify = true;
5975                 else if (err != -EALREADY)
5976                         return err;
5977
5978                 response_filters.mask[0] |=
5979                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5980                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5981                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5982         }
5983
5984         if (notify)
5985                 rtnl_offload_xstats_notify(dev);
5986
5987         return 0;
5988 }
5989
5990 /* Process one rtnetlink message. */
5991
5992 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5993                              struct netlink_ext_ack *extack)
5994 {
5995         struct net *net = sock_net(skb->sk);
5996         struct rtnl_link *link;
5997         enum rtnl_kinds kind;
5998         struct module *owner;
5999         int err = -EOPNOTSUPP;
6000         rtnl_doit_func doit;
6001         unsigned int flags;
6002         int family;
6003         int type;
6004
6005         type = nlh->nlmsg_type;
6006         if (type > RTM_MAX)
6007                 return -EOPNOTSUPP;
6008
6009         type -= RTM_BASE;
6010
6011         /* All the messages must have at least 1 byte length */
6012         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6013                 return 0;
6014
6015         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6016         kind = rtnl_msgtype_kind(type);
6017
6018         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6019                 return -EPERM;
6020
6021         rcu_read_lock();
6022         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6023                 struct sock *rtnl;
6024                 rtnl_dumpit_func dumpit;
6025                 u32 min_dump_alloc = 0;
6026
6027                 link = rtnl_get_link(family, type);
6028                 if (!link || !link->dumpit) {
6029                         family = PF_UNSPEC;
6030                         link = rtnl_get_link(family, type);
6031                         if (!link || !link->dumpit)
6032                                 goto err_unlock;
6033                 }
6034                 owner = link->owner;
6035                 dumpit = link->dumpit;
6036
6037                 if (type == RTM_GETLINK - RTM_BASE)
6038                         min_dump_alloc = rtnl_calcit(skb, nlh);
6039
6040                 err = 0;
6041                 /* need to do this before rcu_read_unlock() */
6042                 if (!try_module_get(owner))
6043                         err = -EPROTONOSUPPORT;
6044
6045                 rcu_read_unlock();
6046
6047                 rtnl = net->rtnl;
6048                 if (err == 0) {
6049                         struct netlink_dump_control c = {
6050                                 .dump           = dumpit,
6051                                 .min_dump_alloc = min_dump_alloc,
6052                                 .module         = owner,
6053                         };
6054                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6055                         /* netlink_dump_start() will keep a reference on
6056                          * module if dump is still in progress.
6057                          */
6058                         module_put(owner);
6059                 }
6060                 return err;
6061         }
6062
6063         link = rtnl_get_link(family, type);
6064         if (!link || !link->doit) {
6065                 family = PF_UNSPEC;
6066                 link = rtnl_get_link(PF_UNSPEC, type);
6067                 if (!link || !link->doit)
6068                         goto out_unlock;
6069         }
6070
6071         owner = link->owner;
6072         if (!try_module_get(owner)) {
6073                 err = -EPROTONOSUPPORT;
6074                 goto out_unlock;
6075         }
6076
6077         flags = link->flags;
6078         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6079             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6080                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6081                 goto err_unlock;
6082         }
6083
6084         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6085                 doit = link->doit;
6086                 rcu_read_unlock();
6087                 if (doit)
6088                         err = doit(skb, nlh, extack);
6089                 module_put(owner);
6090                 return err;
6091         }
6092         rcu_read_unlock();
6093
6094         rtnl_lock();
6095         link = rtnl_get_link(family, type);
6096         if (link && link->doit)
6097                 err = link->doit(skb, nlh, extack);
6098         rtnl_unlock();
6099
6100         module_put(owner);
6101
6102         return err;
6103
6104 out_unlock:
6105         rcu_read_unlock();
6106         return err;
6107
6108 err_unlock:
6109         rcu_read_unlock();
6110         return -EOPNOTSUPP;
6111 }
6112
6113 static void rtnetlink_rcv(struct sk_buff *skb)
6114 {
6115         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6116 }
6117
6118 static int rtnetlink_bind(struct net *net, int group)
6119 {
6120         switch (group) {
6121         case RTNLGRP_IPV4_MROUTE_R:
6122         case RTNLGRP_IPV6_MROUTE_R:
6123                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6124                         return -EPERM;
6125                 break;
6126         }
6127         return 0;
6128 }
6129
6130 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6131 {
6132         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6133
6134         switch (event) {
6135         case NETDEV_REBOOT:
6136         case NETDEV_CHANGEMTU:
6137         case NETDEV_CHANGEADDR:
6138         case NETDEV_CHANGENAME:
6139         case NETDEV_FEAT_CHANGE:
6140         case NETDEV_BONDING_FAILOVER:
6141         case NETDEV_POST_TYPE_CHANGE:
6142         case NETDEV_NOTIFY_PEERS:
6143         case NETDEV_CHANGEUPPER:
6144         case NETDEV_RESEND_IGMP:
6145         case NETDEV_CHANGEINFODATA:
6146         case NETDEV_CHANGELOWERSTATE:
6147         case NETDEV_CHANGE_TX_QUEUE_LEN:
6148                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6149                                    GFP_KERNEL, NULL, 0);
6150                 break;
6151         default:
6152                 break;
6153         }
6154         return NOTIFY_DONE;
6155 }
6156
6157 static struct notifier_block rtnetlink_dev_notifier = {
6158         .notifier_call  = rtnetlink_event,
6159 };
6160
6161
6162 static int __net_init rtnetlink_net_init(struct net *net)
6163 {
6164         struct sock *sk;
6165         struct netlink_kernel_cfg cfg = {
6166                 .groups         = RTNLGRP_MAX,
6167                 .input          = rtnetlink_rcv,
6168                 .cb_mutex       = &rtnl_mutex,
6169                 .flags          = NL_CFG_F_NONROOT_RECV,
6170                 .bind           = rtnetlink_bind,
6171         };
6172
6173         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6174         if (!sk)
6175                 return -ENOMEM;
6176         net->rtnl = sk;
6177         return 0;
6178 }
6179
6180 static void __net_exit rtnetlink_net_exit(struct net *net)
6181 {
6182         netlink_kernel_release(net->rtnl);
6183         net->rtnl = NULL;
6184 }
6185
6186 static struct pernet_operations rtnetlink_net_ops = {
6187         .init = rtnetlink_net_init,
6188         .exit = rtnetlink_net_exit,
6189 };
6190
6191 void __init rtnetlink_init(void)
6192 {
6193         if (register_pernet_subsys(&rtnetlink_net_ops))
6194                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6195
6196         register_netdevice_notifier(&rtnetlink_dev_notifier);
6197
6198         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6199                       rtnl_dump_ifinfo, 0);
6200         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6201         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6202         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6203
6204         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6205         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6206         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6207
6208         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6209         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6210
6211         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6212         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6213                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6214         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6215
6216         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6217         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6218         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6219
6220         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6221                       0);
6222         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6223 }