net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetic was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56 #include <net/devlink.h>
  57 #if IS_ENABLED(CONFIG_IPV6)
  58 #include <net/addrconf.h>
  59 #endif
  60 #include <linux/dpll.h>
  61
  62 #include "dev.h"
  63
  64 #define RTNL_MAX_TYPE           50
  65 #define RTNL_SLAVE_MAX_TYPE     44
  66
  67 struct rtnl_link {
  68         rtnl_doit_func          doit;
  69         rtnl_dumpit_func        dumpit;
  70         struct module           *owner;
  71         unsigned int            flags;
  72         struct rcu_head         rcu;
  73 };
  74
  75 static DEFINE_MUTEX(rtnl_mutex);
  76
  77 void rtnl_lock(void)
  78 {
  79         mutex_lock(&rtnl_mutex);
  80 }
  81 EXPORT_SYMBOL(rtnl_lock);
  82
  83 int rtnl_lock_killable(void)
  84 {
  85         return mutex_lock_killable(&rtnl_mutex);
  86 }
  87 EXPORT_SYMBOL(rtnl_lock_killable);
  88
  89 static struct sk_buff *defer_kfree_skb_list;
  90 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  91 {
  92         if (head && tail) {
  93                 tail->next = defer_kfree_skb_list;
  94                 defer_kfree_skb_list = head;
  95         }
  96 }
  97 EXPORT_SYMBOL(rtnl_kfree_skbs);
  98
  99 void __rtnl_unlock(void)
 100 {
 101         struct sk_buff *head = defer_kfree_skb_list;
 102
 103         defer_kfree_skb_list = NULL;
 104
 105         /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
 106          * is used. In some places, e.g. in cfg80211, we have code that will do
 107          * something like
 108          *   rtnl_lock()
 109          *   wiphy_lock()
 110          *   ...
 111          *   rtnl_unlock()
 112          *
 113          * and because netdev_run_todo() acquires the RTNL for items on the list
 114          * we could cause a situation such as this:
 115          * Thread 1                     Thread 2
 116          *                                rtnl_lock()
 117          *                                unregister_netdevice()
 118          *                                __rtnl_unlock()
 119          * rtnl_lock()
 120          * wiphy_lock()
 121          * rtnl_unlock()
 122          *   netdev_run_todo()
 123          *     __rtnl_unlock()
 124          *
 125          *     // list not empty now
 126          *     // because of thread 2
 127          *                                rtnl_lock()
 128          *     while (!list_empty(...))
 129          *       rtnl_lock()
 130          *                                wiphy_lock()
 131          * **** DEADLOCK ****
 132          *
 133          * However, usage of __rtnl_unlock() is rare, and so we can ensure that
 134          * it's not used in cases where something is added to do the list.
 135          */
 136         WARN_ON(!list_empty(&net_todo_list));
 137
 138         mutex_unlock(&rtnl_mutex);
 139
 140         while (head) {
 141                 struct sk_buff *next = head->next;
 142
 143                 kfree_skb(head);
 144                 cond_resched();
 145                 head = next;
 146         }
 147 }
 148
 149 void rtnl_unlock(void)
 150 {
 151         /* This fellow will unlock it for us. */
 152         netdev_run_todo();
 153 }
 154 EXPORT_SYMBOL(rtnl_unlock);
 155
 156 int rtnl_trylock(void)
 157 {
 158         return mutex_trylock(&rtnl_mutex);
 159 }
 160 EXPORT_SYMBOL(rtnl_trylock);
 161
 162 int rtnl_is_locked(void)
 163 {
 164         return mutex_is_locked(&rtnl_mutex);
 165 }
 166 EXPORT_SYMBOL(rtnl_is_locked);
 167
 168 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 169 {
 170         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 171 }
 172 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 173
 174 #ifdef CONFIG_PROVE_LOCKING
 175 bool lockdep_rtnl_is_held(void)
 176 {
 177         return lockdep_is_held(&rtnl_mutex);
 178 }
 179 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 180 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 181
 182 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 183
 184 static inline int rtm_msgindex(int msgtype)
 185 {
 186         int msgindex = msgtype - RTM_BASE;
 187
 188         /*
 189          * msgindex < 0 implies someone tried to register a netlink
 190          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 191          * the message type has not been added to linux/rtnetlink.h
 192          */
 193         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 194
 195         return msgindex;
 196 }
 197
 198 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 199 {
 200         struct rtnl_link __rcu **tab;
 201
 202         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 203                 protocol = PF_UNSPEC;
 204
 205         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 206         if (!tab)
 207                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 208
 209         return rcu_dereference_rtnl(tab[msgtype]);
 210 }
 211
 212 static int rtnl_register_internal(struct module *owner,
 213                                   int protocol, int msgtype,
 214                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 215                                   unsigned int flags)
 216 {
 217         struct rtnl_link *link, *old;
 218         struct rtnl_link __rcu **tab;
 219         int msgindex;
 220         int ret = -ENOBUFS;
 221
 222         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 223         msgindex = rtm_msgindex(msgtype);
 224
 225         rtnl_lock();
 226         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 227         if (tab == NULL) {
 228                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 229                 if (!tab)
 230                         goto unlock;
 231
 232                 /* ensures we see the 0 stores */
 233                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 234         }
 235
 236         old = rtnl_dereference(tab[msgindex]);
 237         if (old) {
 238                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 239                 if (!link)
 240                         goto unlock;
 241         } else {
 242                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 243                 if (!link)
 244                         goto unlock;
 245         }
 246
 247         WARN_ON(link->owner && link->owner != owner);
 248         link->owner = owner;
 249
 250         WARN_ON(doit && link->doit && link->doit != doit);
 251         if (doit)
 252                 link->doit = doit;
 253         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 254         if (dumpit)
 255                 link->dumpit = dumpit;
 256
 257         WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
 258                 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
 259         link->flags |= flags;
 260
 261         /* publish protocol:msgtype */
 262         rcu_assign_pointer(tab[msgindex], link);
 263         ret = 0;
 264         if (old)
 265                 kfree_rcu(old, rcu);
 266 unlock:
 267         rtnl_unlock();
 268         return ret;
 269 }
 270
 271 /**
 272  * rtnl_register_module - Register a rtnetlink message type
 273  *
 274  * @owner: module registering the hook (THIS_MODULE)
 275  * @protocol: Protocol family or PF_UNSPEC
 276  * @msgtype: rtnetlink message type
 277  * @doit: Function pointer called for each request message
 278  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 279  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 280  *
 281  * Like rtnl_register, but for use by removable modules.
 282  */
 283 int rtnl_register_module(struct module *owner,
 284                          int protocol, int msgtype,
 285                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 286                          unsigned int flags)
 287 {
 288         return rtnl_register_internal(owner, protocol, msgtype,
 289                                       doit, dumpit, flags);
 290 }
 291 EXPORT_SYMBOL_GPL(rtnl_register_module);
 292
 293 /**
 294  * rtnl_register - Register a rtnetlink message type
 295  * @protocol: Protocol family or PF_UNSPEC
 296  * @msgtype: rtnetlink message type
 297  * @doit: Function pointer called for each request message
 298  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 299  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 300  *
 301  * Registers the specified function pointers (at least one of them has
 302  * to be non-NULL) to be called whenever a request message for the
 303  * specified protocol family and message type is received.
 304  *
 305  * The special protocol family PF_UNSPEC may be used to define fallback
 306  * function pointers for the case when no entry for the specific protocol
 307  * family exists.
 308  */
 309 void rtnl_register(int protocol, int msgtype,
 310                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 311                    unsigned int flags)
 312 {
 313         int err;
 314
 315         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 316                                      flags);
 317         if (err)
 318                 pr_err("Unable to register rtnetlink message handler, "
 319                        "protocol = %d, message type = %d\n", protocol, msgtype);
 320 }
 321
 322 /**
 323  * rtnl_unregister - Unregister a rtnetlink message type
 324  * @protocol: Protocol family or PF_UNSPEC
 325  * @msgtype: rtnetlink message type
 326  *
 327  * Returns 0 on success or a negative error code.
 328  */
 329 int rtnl_unregister(int protocol, int msgtype)
 330 {
 331         struct rtnl_link __rcu **tab;
 332         struct rtnl_link *link;
 333         int msgindex;
 334
 335         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 336         msgindex = rtm_msgindex(msgtype);
 337
 338         rtnl_lock();
 339         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 340         if (!tab) {
 341                 rtnl_unlock();
 342                 return -ENOENT;
 343         }
 344
 345         link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
 346         rtnl_unlock();
 347
 348         kfree_rcu(link, rcu);
 349
 350         return 0;
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_unregister);
 353
 354 /**
 355  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 356  * @protocol : Protocol family or PF_UNSPEC
 357  *
 358  * Identical to calling rtnl_unregster() for all registered message types
 359  * of a certain protocol family.
 360  */
 361 void rtnl_unregister_all(int protocol)
 362 {
 363         struct rtnl_link __rcu **tab;
 364         struct rtnl_link *link;
 365         int msgindex;
 366
 367         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 368
 369         rtnl_lock();
 370         tab = rcu_replace_pointer_rtnl(rtnl_msg_handlers[protocol], NULL);
 371         if (!tab) {
 372                 rtnl_unlock();
 373                 return;
 374         }
 375         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 376                 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
 377                 kfree_rcu(link, rcu);
 378         }
 379         rtnl_unlock();
 380
 381         synchronize_net();
 382
 383         kfree(tab);
 384 }
 385 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 386
 387 static LIST_HEAD(link_ops);
 388
 389 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 390 {
 391         const struct rtnl_link_ops *ops;
 392
 393         list_for_each_entry(ops, &link_ops, list) {
 394                 if (!strcmp(ops->kind, kind))
 395                         return ops;
 396         }
 397         return NULL;
 398 }
 399
 400 /**
 401  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 402  * @ops: struct rtnl_link_ops * to register
 403  *
 404  * The caller must hold the rtnl_mutex. This function should be used
 405  * by drivers that create devices during module initialization. It
 406  * must be called before registering the devices.
 407  *
 408  * Returns 0 on success or a negative error code.
 409  */
 410 int __rtnl_link_register(struct rtnl_link_ops *ops)
 411 {
 412         if (rtnl_link_ops_get(ops->kind))
 413                 return -EEXIST;
 414
 415         /* The check for alloc/setup is here because if ops
 416          * does not have that filled up, it is not possible
 417          * to use the ops for creating device. So do not
 418          * fill up dellink as well. That disables rtnl_dellink.
 419          */
 420         if ((ops->alloc || ops->setup) && !ops->dellink)
 421                 ops->dellink = unregister_netdevice_queue;
 422
 423         list_add_tail(&ops->list, &link_ops);
 424         return 0;
 425 }
 426 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 427
 428 /**
 429  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 430  * @ops: struct rtnl_link_ops * to register
 431  *
 432  * Returns 0 on success or a negative error code.
 433  */
 434 int rtnl_link_register(struct rtnl_link_ops *ops)
 435 {
 436         int err;
 437
 438         /* Sanity-check max sizes to avoid stack buffer overflow. */
 439         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 440                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 441                 return -EINVAL;
 442
 443         rtnl_lock();
 444         err = __rtnl_link_register(ops);
 445         rtnl_unlock();
 446         return err;
 447 }
 448 EXPORT_SYMBOL_GPL(rtnl_link_register);
 449
 450 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 451 {
 452         struct net_device *dev;
 453         LIST_HEAD(list_kill);
 454
 455         for_each_netdev(net, dev) {
 456                 if (dev->rtnl_link_ops == ops)
 457                         ops->dellink(dev, &list_kill);
 458         }
 459         unregister_netdevice_many(&list_kill);
 460 }
 461
 462 /**
 463  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 464  * @ops: struct rtnl_link_ops * to unregister
 465  *
 466  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 467  * integrity (hold pernet_ops_rwsem for writing to close the race
 468  * with setup_net() and cleanup_net()).
 469  */
 470 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 471 {
 472         struct net *net;
 473
 474         for_each_net(net) {
 475                 __rtnl_kill_links(net, ops);
 476         }
 477         list_del(&ops->list);
 478 }
 479 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 480
 481 /* Return with the rtnl_lock held when there are no network
 482  * devices unregistering in any network namespace.
 483  */
 484 static void rtnl_lock_unregistering_all(void)
 485 {
 486         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 487
 488         add_wait_queue(&netdev_unregistering_wq, &wait);
 489         for (;;) {
 490                 rtnl_lock();
 491                 /* We held write locked pernet_ops_rwsem, and parallel
 492                  * setup_net() and cleanup_net() are not possible.
 493                  */
 494                 if (!atomic_read(&dev_unreg_count))
 495                         break;
 496                 __rtnl_unlock();
 497
 498                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 499         }
 500         remove_wait_queue(&netdev_unregistering_wq, &wait);
 501 }
 502
 503 /**
 504  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 505  * @ops: struct rtnl_link_ops * to unregister
 506  */
 507 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 508 {
 509         /* Close the race with setup_net() and cleanup_net() */
 510         down_write(&pernet_ops_rwsem);
 511         rtnl_lock_unregistering_all();
 512         __rtnl_link_unregister(ops);
 513         rtnl_unlock();
 514         up_write(&pernet_ops_rwsem);
 515 }
 516 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 517
 518 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 519 {
 520         struct net_device *master_dev;
 521         const struct rtnl_link_ops *ops;
 522         size_t size = 0;
 523
 524         rcu_read_lock();
 525
 526         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 527         if (!master_dev)
 528                 goto out;
 529
 530         ops = master_dev->rtnl_link_ops;
 531         if (!ops || !ops->get_slave_size)
 532                 goto out;
 533         /* IFLA_INFO_SLAVE_DATA + nested data */
 534         size = nla_total_size(sizeof(struct nlattr)) +
 535                ops->get_slave_size(master_dev, dev);
 536
 537 out:
 538         rcu_read_unlock();
 539         return size;
 540 }
 541
 542 static size_t rtnl_link_get_size(const struct net_device *dev)
 543 {
 544         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 545         size_t size;
 546
 547         if (!ops)
 548                 return 0;
 549
 550         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 551                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 552
 553         if (ops->get_size)
 554                 /* IFLA_INFO_DATA + nested data */
 555                 size += nla_total_size(sizeof(struct nlattr)) +
 556                         ops->get_size(dev);
 557
 558         if (ops->get_xstats_size)
 559                 /* IFLA_INFO_XSTATS */
 560                 size += nla_total_size(ops->get_xstats_size(dev));
 561
 562         size += rtnl_link_get_slave_info_data_size(dev);
 563
 564         return size;
 565 }
 566
 567 static LIST_HEAD(rtnl_af_ops);
 568
 569 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 570 {
 571         const struct rtnl_af_ops *ops;
 572
 573         ASSERT_RTNL();
 574
 575         list_for_each_entry(ops, &rtnl_af_ops, list) {
 576                 if (ops->family == family)
 577                         return ops;
 578         }
 579
 580         return NULL;
 581 }
 582
 583 /**
 584  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 585  * @ops: struct rtnl_af_ops * to register
 586  *
 587  * Returns 0 on success or a negative error code.
 588  */
 589 void rtnl_af_register(struct rtnl_af_ops *ops)
 590 {
 591         rtnl_lock();
 592         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 593         rtnl_unlock();
 594 }
 595 EXPORT_SYMBOL_GPL(rtnl_af_register);
 596
 597 /**
 598  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 599  * @ops: struct rtnl_af_ops * to unregister
 600  */
 601 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 602 {
 603         rtnl_lock();
 604         list_del_rcu(&ops->list);
 605         rtnl_unlock();
 606
 607         synchronize_rcu();
 608 }
 609 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 610
 611 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 612                                     u32 ext_filter_mask)
 613 {
 614         struct rtnl_af_ops *af_ops;
 615         size_t size;
 616
 617         /* IFLA_AF_SPEC */
 618         size = nla_total_size(sizeof(struct nlattr));
 619
 620         rcu_read_lock();
 621         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 622                 if (af_ops->get_link_af_size) {
 623                         /* AF_* + nested data */
 624                         size += nla_total_size(sizeof(struct nlattr)) +
 625                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 626                 }
 627         }
 628         rcu_read_unlock();
 629
 630         return size;
 631 }
 632
 633 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 634 {
 635         struct net_device *master_dev;
 636         bool ret = false;
 637
 638         rcu_read_lock();
 639
 640         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 641         if (master_dev && master_dev->rtnl_link_ops)
 642                 ret = true;
 643         rcu_read_unlock();
 644         return ret;
 645 }
 646
 647 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 648                                      const struct net_device *dev)
 649 {
 650         struct net_device *master_dev;
 651         const struct rtnl_link_ops *ops;
 652         struct nlattr *slave_data;
 653         int err;
 654
 655         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 656         if (!master_dev)
 657                 return 0;
 658         ops = master_dev->rtnl_link_ops;
 659         if (!ops)
 660                 return 0;
 661         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 662                 return -EMSGSIZE;
 663         if (ops->fill_slave_info) {
 664                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 665                 if (!slave_data)
 666                         return -EMSGSIZE;
 667                 err = ops->fill_slave_info(skb, master_dev, dev);
 668                 if (err < 0)
 669                         goto err_cancel_slave_data;
 670                 nla_nest_end(skb, slave_data);
 671         }
 672         return 0;
 673
 674 err_cancel_slave_data:
 675         nla_nest_cancel(skb, slave_data);
 676         return err;
 677 }
 678
 679 static int rtnl_link_info_fill(struct sk_buff *skb,
 680                                const struct net_device *dev)
 681 {
 682         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 683         struct nlattr *data;
 684         int err;
 685
 686         if (!ops)
 687                 return 0;
 688         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 689                 return -EMSGSIZE;
 690         if (ops->fill_xstats) {
 691                 err = ops->fill_xstats(skb, dev);
 692                 if (err < 0)
 693                         return err;
 694         }
 695         if (ops->fill_info) {
 696                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 697                 if (data == NULL)
 698                         return -EMSGSIZE;
 699                 err = ops->fill_info(skb, dev);
 700                 if (err < 0)
 701                         goto err_cancel_data;
 702                 nla_nest_end(skb, data);
 703         }
 704         return 0;
 705
 706 err_cancel_data:
 707         nla_nest_cancel(skb, data);
 708         return err;
 709 }
 710
 711 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 712 {
 713         struct nlattr *linkinfo;
 714         int err = -EMSGSIZE;
 715
 716         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 717         if (linkinfo == NULL)
 718                 goto out;
 719
 720         err = rtnl_link_info_fill(skb, dev);
 721         if (err < 0)
 722                 goto err_cancel_link;
 723
 724         err = rtnl_link_slave_info_fill(skb, dev);
 725         if (err < 0)
 726                 goto err_cancel_link;
 727
 728         nla_nest_end(skb, linkinfo);
 729         return 0;
 730
 731 err_cancel_link:
 732         nla_nest_cancel(skb, linkinfo);
 733 out:
 734         return err;
 735 }
 736
 737 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 738 {
 739         struct sock *rtnl = net->rtnl;
 740
 741         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 742 }
 743
 744 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 745 {
 746         struct sock *rtnl = net->rtnl;
 747
 748         return nlmsg_unicast(rtnl, skb, pid);
 749 }
 750 EXPORT_SYMBOL(rtnl_unicast);
 751
 752 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 753                  const struct nlmsghdr *nlh, gfp_t flags)
 754 {
 755         struct sock *rtnl = net->rtnl;
 756
 757         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 758 }
 759 EXPORT_SYMBOL(rtnl_notify);
 760
 761 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 762 {
 763         struct sock *rtnl = net->rtnl;
 764
 765         netlink_set_err(rtnl, 0, group, error);
 766 }
 767 EXPORT_SYMBOL(rtnl_set_sk_err);
 768
 769 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 770 {
 771         struct nlattr *mx;
 772         int i, valid = 0;
 773
 774         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 775         if (metrics == dst_default_metrics.metrics)
 776                 return 0;
 777
 778         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 779         if (mx == NULL)
 780                 return -ENOBUFS;
 781
 782         for (i = 0; i < RTAX_MAX; i++) {
 783                 if (metrics[i]) {
 784                         if (i == RTAX_CC_ALGO - 1) {
 785                                 char tmp[TCP_CA_NAME_MAX], *name;
 786
 787                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 788                                 if (!name)
 789                                         continue;
 790                                 if (nla_put_string(skb, i + 1, name))
 791                                         goto nla_put_failure;
 792                         } else if (i == RTAX_FEATURES - 1) {
 793                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 794
 795                                 if (!user_features)
 796                                         continue;
 797                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 798                                 if (nla_put_u32(skb, i + 1, user_features))
 799                                         goto nla_put_failure;
 800                         } else {
 801                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 802                                         goto nla_put_failure;
 803                         }
 804                         valid++;
 805                 }
 806         }
 807
 808         if (!valid) {
 809                 nla_nest_cancel(skb, mx);
 810                 return 0;
 811         }
 812
 813         return nla_nest_end(skb, mx);
 814
 815 nla_put_failure:
 816         nla_nest_cancel(skb, mx);
 817         return -EMSGSIZE;
 818 }
 819 EXPORT_SYMBOL(rtnetlink_put_metrics);
 820
 821 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 822                        long expires, u32 error)
 823 {
 824         struct rta_cacheinfo ci = {
 825                 .rta_error = error,
 826                 .rta_id =  id,
 827         };
 828
 829         if (dst) {
 830                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 831                 ci.rta_used = dst->__use;
 832                 ci.rta_clntref = rcuref_read(&dst->__rcuref);
 833         }
 834         if (expires) {
 835                 unsigned long clock;
 836
 837                 clock = jiffies_to_clock_t(abs(expires));
 838                 clock = min_t(unsigned long, clock, INT_MAX);
 839                 ci.rta_expires = (expires > 0) ? clock : -clock;
 840         }
 841         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 842 }
 843 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 844
 845 void netdev_set_operstate(struct net_device *dev, int newstate)
 846 {
 847         unsigned int old = READ_ONCE(dev->operstate);
 848
 849         do {
 850                 if (old == newstate)
 851                         return;
 852         } while (!try_cmpxchg(&dev->operstate, &old, newstate));
 853
 854         netdev_state_change(dev);
 855 }
 856 EXPORT_SYMBOL(netdev_set_operstate);
 857
 858 static void set_operstate(struct net_device *dev, unsigned char transition)
 859 {
 860         unsigned char operstate = READ_ONCE(dev->operstate);
 861
 862         switch (transition) {
 863         case IF_OPER_UP:
 864                 if ((operstate == IF_OPER_DORMANT ||
 865                      operstate == IF_OPER_TESTING ||
 866                      operstate == IF_OPER_UNKNOWN) &&
 867                     !netif_dormant(dev) && !netif_testing(dev))
 868                         operstate = IF_OPER_UP;
 869                 break;
 870
 871         case IF_OPER_TESTING:
 872                 if (netif_oper_up(dev))
 873                         operstate = IF_OPER_TESTING;
 874                 break;
 875
 876         case IF_OPER_DORMANT:
 877                 if (netif_oper_up(dev))
 878                         operstate = IF_OPER_DORMANT;
 879                 break;
 880         }
 881
 882         netdev_set_operstate(dev, operstate);
 883 }
 884
 885 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 886 {
 887         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 888                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 889 }
 890
 891 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 892                                            const struct ifinfomsg *ifm)
 893 {
 894         unsigned int flags = ifm->ifi_flags;
 895
 896         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 897         if (ifm->ifi_change)
 898                 flags = (flags & ifm->ifi_change) |
 899                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 900
 901         return flags;
 902 }
 903
 904 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 905                                  const struct rtnl_link_stats64 *b)
 906 {
 907         a->rx_packets = b->rx_packets;
 908         a->tx_packets = b->tx_packets;
 909         a->rx_bytes = b->rx_bytes;
 910         a->tx_bytes = b->tx_bytes;
 911         a->rx_errors = b->rx_errors;
 912         a->tx_errors = b->tx_errors;
 913         a->rx_dropped = b->rx_dropped;
 914         a->tx_dropped = b->tx_dropped;
 915
 916         a->multicast = b->multicast;
 917         a->collisions = b->collisions;
 918
 919         a->rx_length_errors = b->rx_length_errors;
 920         a->rx_over_errors = b->rx_over_errors;
 921         a->rx_crc_errors = b->rx_crc_errors;
 922         a->rx_frame_errors = b->rx_frame_errors;
 923         a->rx_fifo_errors = b->rx_fifo_errors;
 924         a->rx_missed_errors = b->rx_missed_errors;
 925
 926         a->tx_aborted_errors = b->tx_aborted_errors;
 927         a->tx_carrier_errors = b->tx_carrier_errors;
 928         a->tx_fifo_errors = b->tx_fifo_errors;
 929         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 930         a->tx_window_errors = b->tx_window_errors;
 931
 932         a->rx_compressed = b->rx_compressed;
 933         a->tx_compressed = b->tx_compressed;
 934
 935         a->rx_nohandler = b->rx_nohandler;
 936 }
 937
 938 /* All VF info */
 939 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 940                                    u32 ext_filter_mask)
 941 {
 942         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 943                 int num_vfs = dev_num_vf(dev->dev.parent);
 944                 size_t size = nla_total_size(0);
 945                 size += num_vfs *
 946                         (nla_total_size(0) +
 947                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 948                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 949                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 950                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 951                          nla_total_size(MAX_VLAN_LIST_LEN *
 952                                         sizeof(struct ifla_vf_vlan_info)) +
 953                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 954                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 955                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 956                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 957                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 958                          nla_total_size(sizeof(struct ifla_vf_trust)));
 959                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 960                         size += num_vfs *
 961                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 962                                  /* IFLA_VF_STATS_RX_PACKETS */
 963                                  nla_total_size_64bit(sizeof(__u64)) +
 964                                  /* IFLA_VF_STATS_TX_PACKETS */
 965                                  nla_total_size_64bit(sizeof(__u64)) +
 966                                  /* IFLA_VF_STATS_RX_BYTES */
 967                                  nla_total_size_64bit(sizeof(__u64)) +
 968                                  /* IFLA_VF_STATS_TX_BYTES */
 969                                  nla_total_size_64bit(sizeof(__u64)) +
 970                                  /* IFLA_VF_STATS_BROADCAST */
 971                                  nla_total_size_64bit(sizeof(__u64)) +
 972                                  /* IFLA_VF_STATS_MULTICAST */
 973                                  nla_total_size_64bit(sizeof(__u64)) +
 974                                  /* IFLA_VF_STATS_RX_DROPPED */
 975                                  nla_total_size_64bit(sizeof(__u64)) +
 976                                  /* IFLA_VF_STATS_TX_DROPPED */
 977                                  nla_total_size_64bit(sizeof(__u64)));
 978                 }
 979                 return size;
 980         } else
 981                 return 0;
 982 }
 983
 984 static size_t rtnl_port_size(const struct net_device *dev,
 985                              u32 ext_filter_mask)
 986 {
 987         size_t port_size = nla_total_size(4)            /* PORT_VF */
 988                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 989                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 990                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 991                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 992                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 993         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 994         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 995                 + port_size;
 996         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 997                 + port_size;
 998
 999         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1000             !(ext_filter_mask & RTEXT_FILTER_VF))
1001                 return 0;
1002         if (dev_num_vf(dev->dev.parent))
1003                 return port_self_size + vf_ports_size +
1004                         vf_port_size * dev_num_vf(dev->dev.parent);
1005         else
1006                 return port_self_size;
1007 }
1008
1009 static size_t rtnl_xdp_size(void)
1010 {
1011         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1012                           nla_total_size(1) +   /* XDP_ATTACHED */
1013                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1014                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1015
1016         return xdp_size;
1017 }
1018
1019 static size_t rtnl_prop_list_size(const struct net_device *dev)
1020 {
1021         struct netdev_name_node *name_node;
1022         unsigned int cnt = 0;
1023
1024         rcu_read_lock();
1025         list_for_each_entry_rcu(name_node, &dev->name_node->list, list)
1026                 cnt++;
1027         rcu_read_unlock();
1028
1029         if (!cnt)
1030                 return 0;
1031
1032         return nla_total_size(0) + cnt * nla_total_size(ALTIFNAMSIZ);
1033 }
1034
1035 static size_t rtnl_proto_down_size(const struct net_device *dev)
1036 {
1037         size_t size = nla_total_size(1);
1038
1039         if (dev->proto_down_reason)
1040                 size += nla_total_size(0) + nla_total_size(4);
1041
1042         return size;
1043 }
1044
1045 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1046 {
1047         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1048
1049         if (dev->devlink_port)
1050                 size += devlink_nl_port_handle_size(dev->devlink_port);
1051
1052         return size;
1053 }
1054
1055 static size_t rtnl_dpll_pin_size(const struct net_device *dev)
1056 {
1057         size_t size = nla_total_size(0); /* nest IFLA_DPLL_PIN */
1058
1059         size += dpll_msg_pin_handle_size(netdev_dpll_pin(dev));
1060
1061         return size;
1062 }
1063
1064 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1065                                      u32 ext_filter_mask)
1066 {
1067         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1068                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1069                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1070                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1071                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1072                + nla_total_size(sizeof(struct rtnl_link_stats))
1073                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1074                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1075                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1076                + nla_total_size(4) /* IFLA_TXQLEN */
1077                + nla_total_size(4) /* IFLA_WEIGHT */
1078                + nla_total_size(4) /* IFLA_MTU */
1079                + nla_total_size(4) /* IFLA_LINK */
1080                + nla_total_size(4) /* IFLA_MASTER */
1081                + nla_total_size(1) /* IFLA_CARRIER */
1082                + nla_total_size(4) /* IFLA_PROMISCUITY */
1083                + nla_total_size(4) /* IFLA_ALLMULTI */
1084                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1085                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1086                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1087                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1088                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1089                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1090                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1091                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1092                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1093                + nla_total_size(1) /* IFLA_OPERSTATE */
1094                + nla_total_size(1) /* IFLA_LINKMODE */
1095                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1096                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1097                + nla_total_size(4) /* IFLA_GROUP */
1098                + nla_total_size(ext_filter_mask
1099                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1100                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1101                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1102                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1103                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1104                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1105                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1106                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1107                + rtnl_xdp_size() /* IFLA_XDP */
1108                + nla_total_size(4)  /* IFLA_EVENT */
1109                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1110                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1111                + rtnl_proto_down_size(dev)  /* proto down */
1112                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1113                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1114                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1115                + nla_total_size(4)  /* IFLA_MIN_MTU */
1116                + nla_total_size(4)  /* IFLA_MAX_MTU */
1117                + rtnl_prop_list_size(dev)
1118                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1119                + rtnl_devlink_port_size(dev)
1120                + rtnl_dpll_pin_size(dev)
1121                + 0;
1122 }
1123
1124 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1125 {
1126         struct nlattr *vf_ports;
1127         struct nlattr *vf_port;
1128         int vf;
1129         int err;
1130
1131         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1132         if (!vf_ports)
1133                 return -EMSGSIZE;
1134
1135         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1136                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1137                 if (!vf_port)
1138                         goto nla_put_failure;
1139                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1140                         goto nla_put_failure;
1141                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1142                 if (err == -EMSGSIZE)
1143                         goto nla_put_failure;
1144                 if (err) {
1145                         nla_nest_cancel(skb, vf_port);
1146                         continue;
1147                 }
1148                 nla_nest_end(skb, vf_port);
1149         }
1150
1151         nla_nest_end(skb, vf_ports);
1152
1153         return 0;
1154
1155 nla_put_failure:
1156         nla_nest_cancel(skb, vf_ports);
1157         return -EMSGSIZE;
1158 }
1159
1160 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1161 {
1162         struct nlattr *port_self;
1163         int err;
1164
1165         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1166         if (!port_self)
1167                 return -EMSGSIZE;
1168
1169         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1170         if (err) {
1171                 nla_nest_cancel(skb, port_self);
1172                 return (err == -EMSGSIZE) ? err : 0;
1173         }
1174
1175         nla_nest_end(skb, port_self);
1176
1177         return 0;
1178 }
1179
1180 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1181                           u32 ext_filter_mask)
1182 {
1183         int err;
1184
1185         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1186             !(ext_filter_mask & RTEXT_FILTER_VF))
1187                 return 0;
1188
1189         err = rtnl_port_self_fill(skb, dev);
1190         if (err)
1191                 return err;
1192
1193         if (dev_num_vf(dev->dev.parent)) {
1194                 err = rtnl_vf_ports_fill(skb, dev);
1195                 if (err)
1196                         return err;
1197         }
1198
1199         return 0;
1200 }
1201
1202 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1203 {
1204         int err;
1205         struct netdev_phys_item_id ppid;
1206
1207         err = dev_get_phys_port_id(dev, &ppid);
1208         if (err) {
1209                 if (err == -EOPNOTSUPP)
1210                         return 0;
1211                 return err;
1212         }
1213
1214         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1215                 return -EMSGSIZE;
1216
1217         return 0;
1218 }
1219
1220 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1221 {
1222         char name[IFNAMSIZ];
1223         int err;
1224
1225         err = dev_get_phys_port_name(dev, name, sizeof(name));
1226         if (err) {
1227                 if (err == -EOPNOTSUPP)
1228                         return 0;
1229                 return err;
1230         }
1231
1232         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1233                 return -EMSGSIZE;
1234
1235         return 0;
1236 }
1237
1238 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1239 {
1240         struct netdev_phys_item_id ppid = { };
1241         int err;
1242
1243         err = dev_get_port_parent_id(dev, &ppid, false);
1244         if (err) {
1245                 if (err == -EOPNOTSUPP)
1246                         return 0;
1247                 return err;
1248         }
1249
1250         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1251                 return -EMSGSIZE;
1252
1253         return 0;
1254 }
1255
1256 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1257                                               struct net_device *dev)
1258 {
1259         struct rtnl_link_stats64 *sp;
1260         struct nlattr *attr;
1261
1262         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1263                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1264         if (!attr)
1265                 return -EMSGSIZE;
1266
1267         sp = nla_data(attr);
1268         dev_get_stats(dev, sp);
1269
1270         attr = nla_reserve(skb, IFLA_STATS,
1271                            sizeof(struct rtnl_link_stats));
1272         if (!attr)
1273                 return -EMSGSIZE;
1274
1275         copy_rtnl_link_stats(nla_data(attr), sp);
1276
1277         return 0;
1278 }
1279
1280 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1281                                                struct net_device *dev,
1282                                                int vfs_num,
1283                                                u32 ext_filter_mask)
1284 {
1285         struct ifla_vf_rss_query_en vf_rss_query_en;
1286         struct nlattr *vf, *vfstats, *vfvlanlist;
1287         struct ifla_vf_link_state vf_linkstate;
1288         struct ifla_vf_vlan_info vf_vlan_info;
1289         struct ifla_vf_spoofchk vf_spoofchk;
1290         struct ifla_vf_tx_rate vf_tx_rate;
1291         struct ifla_vf_stats vf_stats;
1292         struct ifla_vf_trust vf_trust;
1293         struct ifla_vf_vlan vf_vlan;
1294         struct ifla_vf_rate vf_rate;
1295         struct ifla_vf_mac vf_mac;
1296         struct ifla_vf_broadcast vf_broadcast;
1297         struct ifla_vf_info ivi;
1298         struct ifla_vf_guid node_guid;
1299         struct ifla_vf_guid port_guid;
1300
1301         memset(&ivi, 0, sizeof(ivi));
1302
1303         /* Not all SR-IOV capable drivers support the
1304          * spoofcheck and "RSS query enable" query.  Preset to
1305          * -1 so the user space tool can detect that the driver
1306          * didn't report anything.
1307          */
1308         ivi.spoofchk = -1;
1309         ivi.rss_query_en = -1;
1310         ivi.trusted = -1;
1311         /* The default value for VF link state is "auto"
1312          * IFLA_VF_LINK_STATE_AUTO which equals zero
1313          */
1314         ivi.linkstate = 0;
1315         /* VLAN Protocol by default is 802.1Q */
1316         ivi.vlan_proto = htons(ETH_P_8021Q);
1317         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1318                 return 0;
1319
1320         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1321         memset(&node_guid, 0, sizeof(node_guid));
1322         memset(&port_guid, 0, sizeof(port_guid));
1323
1324         vf_mac.vf =
1325                 vf_vlan.vf =
1326                 vf_vlan_info.vf =
1327                 vf_rate.vf =
1328                 vf_tx_rate.vf =
1329                 vf_spoofchk.vf =
1330                 vf_linkstate.vf =
1331                 vf_rss_query_en.vf =
1332                 vf_trust.vf =
1333                 node_guid.vf =
1334                 port_guid.vf = ivi.vf;
1335
1336         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1337         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1338         vf_vlan.vlan = ivi.vlan;
1339         vf_vlan.qos = ivi.qos;
1340         vf_vlan_info.vlan = ivi.vlan;
1341         vf_vlan_info.qos = ivi.qos;
1342         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1343         vf_tx_rate.rate = ivi.max_tx_rate;
1344         vf_rate.min_tx_rate = ivi.min_tx_rate;
1345         vf_rate.max_tx_rate = ivi.max_tx_rate;
1346         vf_spoofchk.setting = ivi.spoofchk;
1347         vf_linkstate.link_state = ivi.linkstate;
1348         vf_rss_query_en.setting = ivi.rss_query_en;
1349         vf_trust.setting = ivi.trusted;
1350         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1351         if (!vf)
1352                 return -EMSGSIZE;
1353         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1354             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1355             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1356             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1357                     &vf_rate) ||
1358             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1359                     &vf_tx_rate) ||
1360             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1361                     &vf_spoofchk) ||
1362             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1363                     &vf_linkstate) ||
1364             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1365                     sizeof(vf_rss_query_en),
1366                     &vf_rss_query_en) ||
1367             nla_put(skb, IFLA_VF_TRUST,
1368                     sizeof(vf_trust), &vf_trust))
1369                 goto nla_put_vf_failure;
1370
1371         if (dev->netdev_ops->ndo_get_vf_guid &&
1372             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1373                                               &port_guid)) {
1374                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1375                             &node_guid) ||
1376                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1377                             &port_guid))
1378                         goto nla_put_vf_failure;
1379         }
1380         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1381         if (!vfvlanlist)
1382                 goto nla_put_vf_failure;
1383         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1384                     &vf_vlan_info)) {
1385                 nla_nest_cancel(skb, vfvlanlist);
1386                 goto nla_put_vf_failure;
1387         }
1388         nla_nest_end(skb, vfvlanlist);
1389         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1390                 memset(&vf_stats, 0, sizeof(vf_stats));
1391                 if (dev->netdev_ops->ndo_get_vf_stats)
1392                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1393                                                           &vf_stats);
1394                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1395                 if (!vfstats)
1396                         goto nla_put_vf_failure;
1397                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1398                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1399                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1400                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1401                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1402                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1403                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1404                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1405                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1406                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1407                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1408                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1409                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1410                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1411                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1412                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1413                         nla_nest_cancel(skb, vfstats);
1414                         goto nla_put_vf_failure;
1415                 }
1416                 nla_nest_end(skb, vfstats);
1417         }
1418         nla_nest_end(skb, vf);
1419         return 0;
1420
1421 nla_put_vf_failure:
1422         nla_nest_cancel(skb, vf);
1423         return -EMSGSIZE;
1424 }
1425
1426 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1427                                            struct net_device *dev,
1428                                            u32 ext_filter_mask)
1429 {
1430         struct nlattr *vfinfo;
1431         int i, num_vfs;
1432
1433         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1434                 return 0;
1435
1436         num_vfs = dev_num_vf(dev->dev.parent);
1437         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1438                 return -EMSGSIZE;
1439
1440         if (!dev->netdev_ops->ndo_get_vf_config)
1441                 return 0;
1442
1443         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1444         if (!vfinfo)
1445                 return -EMSGSIZE;
1446
1447         for (i = 0; i < num_vfs; i++) {
1448                 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1449                         nla_nest_cancel(skb, vfinfo);
1450                         return -EMSGSIZE;
1451                 }
1452         }
1453
1454         nla_nest_end(skb, vfinfo);
1455         return 0;
1456 }
1457
1458 static int rtnl_fill_link_ifmap(struct sk_buff *skb,
1459                                 const struct net_device *dev)
1460 {
1461         struct rtnl_link_ifmap map;
1462
1463         memset(&map, 0, sizeof(map));
1464         map.mem_start = READ_ONCE(dev->mem_start);
1465         map.mem_end   = READ_ONCE(dev->mem_end);
1466         map.base_addr = READ_ONCE(dev->base_addr);
1467         map.irq       = READ_ONCE(dev->irq);
1468         map.dma       = READ_ONCE(dev->dma);
1469         map.port      = READ_ONCE(dev->if_port);
1470
1471         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1472                 return -EMSGSIZE;
1473
1474         return 0;
1475 }
1476
1477 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1478 {
1479         const struct bpf_prog *generic_xdp_prog;
1480
1481         ASSERT_RTNL();
1482
1483         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1484         if (!generic_xdp_prog)
1485                 return 0;
1486         return generic_xdp_prog->aux->id;
1487 }
1488
1489 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1490 {
1491         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1492 }
1493
1494 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1495 {
1496         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1497 }
1498
1499 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1500                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1501                                u32 (*get_prog_id)(struct net_device *dev))
1502 {
1503         u32 curr_id;
1504         int err;
1505
1506         curr_id = get_prog_id(dev);
1507         if (!curr_id)
1508                 return 0;
1509
1510         *prog_id = curr_id;
1511         err = nla_put_u32(skb, attr, curr_id);
1512         if (err)
1513                 return err;
1514
1515         if (*mode != XDP_ATTACHED_NONE)
1516                 *mode = XDP_ATTACHED_MULTI;
1517         else
1518                 *mode = tgt_mode;
1519
1520         return 0;
1521 }
1522
1523 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1524 {
1525         struct nlattr *xdp;
1526         u32 prog_id;
1527         int err;
1528         u8 mode;
1529
1530         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1531         if (!xdp)
1532                 return -EMSGSIZE;
1533
1534         prog_id = 0;
1535         mode = XDP_ATTACHED_NONE;
1536         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1537                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1538         if (err)
1539                 goto err_cancel;
1540         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1541                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1542         if (err)
1543                 goto err_cancel;
1544         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1545                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1546         if (err)
1547                 goto err_cancel;
1548
1549         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1550         if (err)
1551                 goto err_cancel;
1552
1553         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1554                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1555                 if (err)
1556                         goto err_cancel;
1557         }
1558
1559         nla_nest_end(skb, xdp);
1560         return 0;
1561
1562 err_cancel:
1563         nla_nest_cancel(skb, xdp);
1564         return err;
1565 }
1566
1567 static u32 rtnl_get_event(unsigned long event)
1568 {
1569         u32 rtnl_event_type = IFLA_EVENT_NONE;
1570
1571         switch (event) {
1572         case NETDEV_REBOOT:
1573                 rtnl_event_type = IFLA_EVENT_REBOOT;
1574                 break;
1575         case NETDEV_FEAT_CHANGE:
1576                 rtnl_event_type = IFLA_EVENT_FEATURES;
1577                 break;
1578         case NETDEV_BONDING_FAILOVER:
1579                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1580                 break;
1581         case NETDEV_NOTIFY_PEERS:
1582                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1583                 break;
1584         case NETDEV_RESEND_IGMP:
1585                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1586                 break;
1587         case NETDEV_CHANGEINFODATA:
1588                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1589                 break;
1590         default:
1591                 break;
1592         }
1593
1594         return rtnl_event_type;
1595 }
1596
1597 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1598 {
1599         const struct net_device *upper_dev;
1600         int ret = 0;
1601
1602         rcu_read_lock();
1603
1604         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1605         if (upper_dev)
1606                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1607
1608         rcu_read_unlock();
1609         return ret;
1610 }
1611
1612 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1613                           bool force)
1614 {
1615         int iflink = dev_get_iflink(dev);
1616
1617         if (force || READ_ONCE(dev->ifindex) != iflink)
1618                 return nla_put_u32(skb, IFLA_LINK, iflink);
1619
1620         return 0;
1621 }
1622
1623 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1624                                               struct net_device *dev)
1625 {
1626         char buf[IFALIASZ];
1627         int ret;
1628
1629         ret = dev_get_alias(dev, buf, sizeof(buf));
1630         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1631 }
1632
1633 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1634                                   const struct net_device *dev,
1635                                   struct net *src_net, gfp_t gfp)
1636 {
1637         bool put_iflink = false;
1638
1639         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1640                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1641
1642                 if (!net_eq(dev_net(dev), link_net)) {
1643                         int id = peernet2id_alloc(src_net, link_net, gfp);
1644
1645                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1646                                 return -EMSGSIZE;
1647
1648                         put_iflink = true;
1649                 }
1650         }
1651
1652         return nla_put_iflink(skb, dev, put_iflink);
1653 }
1654
1655 static int rtnl_fill_link_af(struct sk_buff *skb,
1656                              const struct net_device *dev,
1657                              u32 ext_filter_mask)
1658 {
1659         const struct rtnl_af_ops *af_ops;
1660         struct nlattr *af_spec;
1661
1662         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1663         if (!af_spec)
1664                 return -EMSGSIZE;
1665
1666         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1667                 struct nlattr *af;
1668                 int err;
1669
1670                 if (!af_ops->fill_link_af)
1671                         continue;
1672
1673                 af = nla_nest_start_noflag(skb, af_ops->family);
1674                 if (!af)
1675                         return -EMSGSIZE;
1676
1677                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1678                 /*
1679                  * Caller may return ENODATA to indicate that there
1680                  * was no data to be dumped. This is not an error, it
1681                  * means we should trim the attribute header and
1682                  * continue.
1683                  */
1684                 if (err == -ENODATA)
1685                         nla_nest_cancel(skb, af);
1686                 else if (err < 0)
1687                         return -EMSGSIZE;
1688
1689                 nla_nest_end(skb, af);
1690         }
1691
1692         nla_nest_end(skb, af_spec);
1693         return 0;
1694 }
1695
1696 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1697                                  const struct net_device *dev)
1698 {
1699         struct netdev_name_node *name_node;
1700         int count = 0;
1701
1702         list_for_each_entry_rcu(name_node, &dev->name_node->list, list) {
1703                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1704                         return -EMSGSIZE;
1705                 count++;
1706         }
1707         return count;
1708 }
1709
1710 /* RCU protected. */
1711 static int rtnl_fill_prop_list(struct sk_buff *skb,
1712                                const struct net_device *dev)
1713 {
1714         struct nlattr *prop_list;
1715         int ret;
1716
1717         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1718         if (!prop_list)
1719                 return -EMSGSIZE;
1720
1721         ret = rtnl_fill_alt_ifnames(skb, dev);
1722         if (ret <= 0)
1723                 goto nest_cancel;
1724
1725         nla_nest_end(skb, prop_list);
1726         return 0;
1727
1728 nest_cancel:
1729         nla_nest_cancel(skb, prop_list);
1730         return ret;
1731 }
1732
1733 static int rtnl_fill_proto_down(struct sk_buff *skb,
1734                                 const struct net_device *dev)
1735 {
1736         struct nlattr *pr;
1737         u32 preason;
1738
1739         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1740                 goto nla_put_failure;
1741
1742         preason = dev->proto_down_reason;
1743         if (!preason)
1744                 return 0;
1745
1746         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1747         if (!pr)
1748                 return -EMSGSIZE;
1749
1750         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1751                 nla_nest_cancel(skb, pr);
1752                 goto nla_put_failure;
1753         }
1754
1755         nla_nest_end(skb, pr);
1756         return 0;
1757
1758 nla_put_failure:
1759         return -EMSGSIZE;
1760 }
1761
1762 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1763                                   const struct net_device *dev)
1764 {
1765         struct nlattr *devlink_port_nest;
1766         int ret;
1767
1768         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1769         if (!devlink_port_nest)
1770                 return -EMSGSIZE;
1771
1772         if (dev->devlink_port) {
1773                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1774                 if (ret < 0)
1775                         goto nest_cancel;
1776         }
1777
1778         nla_nest_end(skb, devlink_port_nest);
1779         return 0;
1780
1781 nest_cancel:
1782         nla_nest_cancel(skb, devlink_port_nest);
1783         return ret;
1784 }
1785
1786 static int rtnl_fill_dpll_pin(struct sk_buff *skb,
1787                               const struct net_device *dev)
1788 {
1789         struct nlattr *dpll_pin_nest;
1790         int ret;
1791
1792         dpll_pin_nest = nla_nest_start(skb, IFLA_DPLL_PIN);
1793         if (!dpll_pin_nest)
1794                 return -EMSGSIZE;
1795
1796         ret = dpll_msg_add_pin_handle(skb, netdev_dpll_pin(dev));
1797         if (ret < 0)
1798                 goto nest_cancel;
1799
1800         nla_nest_end(skb, dpll_pin_nest);
1801         return 0;
1802
1803 nest_cancel:
1804         nla_nest_cancel(skb, dpll_pin_nest);
1805         return ret;
1806 }
1807
1808 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1809                             struct net_device *dev, struct net *src_net,
1810                             int type, u32 pid, u32 seq, u32 change,
1811                             unsigned int flags, u32 ext_filter_mask,
1812                             u32 event, int *new_nsid, int new_ifindex,
1813                             int tgt_netnsid, gfp_t gfp)
1814 {
1815         struct ifinfomsg *ifm;
1816         struct nlmsghdr *nlh;
1817         struct Qdisc *qdisc;
1818
1819         ASSERT_RTNL();
1820         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1821         if (nlh == NULL)
1822                 return -EMSGSIZE;
1823
1824         ifm = nlmsg_data(nlh);
1825         ifm->ifi_family = AF_UNSPEC;
1826         ifm->__ifi_pad = 0;
1827         ifm->ifi_type = dev->type;
1828         ifm->ifi_index = dev->ifindex;
1829         ifm->ifi_flags = dev_get_flags(dev);
1830         ifm->ifi_change = change;
1831
1832         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1833                 goto nla_put_failure;
1834
1835         qdisc = rtnl_dereference(dev->qdisc);
1836         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1837             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1838             nla_put_u8(skb, IFLA_OPERSTATE,
1839                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1840             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1841             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1842             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1843             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1844             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1845             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1846             nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1847             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1848             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1849             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1850             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1851             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE, dev->gso_ipv4_max_size) ||
1852             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE, dev->gro_ipv4_max_size) ||
1853             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1854             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1855 #ifdef CONFIG_RPS
1856             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1857 #endif
1858             put_master_ifindex(skb, dev) ||
1859             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1860             (qdisc &&
1861              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1862             nla_put_ifalias(skb, dev) ||
1863             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1864                         atomic_read(&dev->carrier_up_count) +
1865                         atomic_read(&dev->carrier_down_count)) ||
1866             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1867                         atomic_read(&dev->carrier_up_count)) ||
1868             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1869                         atomic_read(&dev->carrier_down_count)))
1870                 goto nla_put_failure;
1871
1872         if (rtnl_fill_proto_down(skb, dev))
1873                 goto nla_put_failure;
1874
1875         if (event != IFLA_EVENT_NONE) {
1876                 if (nla_put_u32(skb, IFLA_EVENT, event))
1877                         goto nla_put_failure;
1878         }
1879
1880         if (dev->addr_len) {
1881                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1882                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1883                         goto nla_put_failure;
1884         }
1885
1886         if (rtnl_phys_port_id_fill(skb, dev))
1887                 goto nla_put_failure;
1888
1889         if (rtnl_phys_port_name_fill(skb, dev))
1890                 goto nla_put_failure;
1891
1892         if (rtnl_phys_switch_id_fill(skb, dev))
1893                 goto nla_put_failure;
1894
1895         if (rtnl_fill_stats(skb, dev))
1896                 goto nla_put_failure;
1897
1898         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1899                 goto nla_put_failure;
1900
1901         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1902                 goto nla_put_failure;
1903
1904         if (rtnl_xdp_fill(skb, dev))
1905                 goto nla_put_failure;
1906
1907         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1908                 if (rtnl_link_fill(skb, dev) < 0)
1909                         goto nla_put_failure;
1910         }
1911
1912         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1913                 goto nla_put_failure;
1914
1915         if (new_nsid &&
1916             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1917                 goto nla_put_failure;
1918         if (new_ifindex &&
1919             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1920                 goto nla_put_failure;
1921
1922         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1923             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1924                 goto nla_put_failure;
1925
1926         rcu_read_lock();
1927         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1928                 goto nla_put_failure_rcu;
1929         if (rtnl_fill_link_ifmap(skb, dev))
1930                 goto nla_put_failure_rcu;
1931         if (rtnl_fill_prop_list(skb, dev))
1932                 goto nla_put_failure_rcu;
1933         rcu_read_unlock();
1934
1935         if (dev->dev.parent &&
1936             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1937                            dev_name(dev->dev.parent)))
1938                 goto nla_put_failure;
1939
1940         if (dev->dev.parent && dev->dev.parent->bus &&
1941             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1942                            dev->dev.parent->bus->name))
1943                 goto nla_put_failure;
1944
1945         if (rtnl_fill_devlink_port(skb, dev))
1946                 goto nla_put_failure;
1947
1948         if (rtnl_fill_dpll_pin(skb, dev))
1949                 goto nla_put_failure;
1950
1951         nlmsg_end(skb, nlh);
1952         return 0;
1953
1954 nla_put_failure_rcu:
1955         rcu_read_unlock();
1956 nla_put_failure:
1957         nlmsg_cancel(skb, nlh);
1958         return -EMSGSIZE;
1959 }
1960
1961 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1962         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1963         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1964         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1965         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1966         [IFLA_MTU]              = { .type = NLA_U32 },
1967         [IFLA_LINK]             = { .type = NLA_U32 },
1968         [IFLA_MASTER]           = { .type = NLA_U32 },
1969         [IFLA_CARRIER]          = { .type = NLA_U8 },
1970         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1971         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1972         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1973         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1974         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1975         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1976         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1977         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1978          * allow 0-length string (needed to remove an alias).
1979          */
1980         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1981         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1982         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1983         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1984         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1985         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1986         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1987         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1988         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1989         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1990         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1991         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1992         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1993         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1994         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1995         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1996         [IFLA_XDP]              = { .type = NLA_NESTED },
1997         [IFLA_EVENT]            = { .type = NLA_U32 },
1998         [IFLA_GROUP]            = { .type = NLA_U32 },
1999         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
2000         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
2001         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
2002         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
2003         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
2004         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
2005         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
2006                                     .len = ALTIFNAMSIZ - 1 },
2007         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
2008         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
2009         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
2010         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
2011         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
2012         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
2013         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
2014         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
2015         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
2016         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
2017 };
2018
2019 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
2020         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
2021         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
2022         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
2023         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
2024 };
2025
2026 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
2027         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
2028         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
2029         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
2030         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
2031         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
2032         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
2033         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
2034         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
2035         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
2036         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2037         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2038         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2039         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2040 };
2041
2042 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2043         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2044         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2045                                     .len = PORT_PROFILE_MAX },
2046         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2047                                       .len = PORT_UUID_MAX },
2048         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2049                                     .len = PORT_UUID_MAX },
2050         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2051         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2052
2053         /* Unused, but we need to keep it here since user space could
2054          * fill it. It's also broken with regard to NLA_BINARY use in
2055          * combination with structs.
2056          */
2057         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2058                                     .len = sizeof(struct ifla_port_vsi) },
2059 };
2060
2061 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2062         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2063         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2064         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2065         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2066         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2067         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2068 };
2069
2070 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2071 {
2072         const struct rtnl_link_ops *ops = NULL;
2073         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2074
2075         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2076                 return NULL;
2077
2078         if (linfo[IFLA_INFO_KIND]) {
2079                 char kind[MODULE_NAME_LEN];
2080
2081                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2082                 ops = rtnl_link_ops_get(kind);
2083         }
2084
2085         return ops;
2086 }
2087
2088 static bool link_master_filtered(struct net_device *dev, int master_idx)
2089 {
2090         struct net_device *master;
2091
2092         if (!master_idx)
2093                 return false;
2094
2095         master = netdev_master_upper_dev_get(dev);
2096
2097         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2098          * another invalid value for ifindex to denote "no master".
2099          */
2100         if (master_idx == -1)
2101                 return !!master;
2102
2103         if (!master || master->ifindex != master_idx)
2104                 return true;
2105
2106         return false;
2107 }
2108
2109 static bool link_kind_filtered(const struct net_device *dev,
2110                                const struct rtnl_link_ops *kind_ops)
2111 {
2112         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2113                 return true;
2114
2115         return false;
2116 }
2117
2118 static bool link_dump_filtered(struct net_device *dev,
2119                                int master_idx,
2120                                const struct rtnl_link_ops *kind_ops)
2121 {
2122         if (link_master_filtered(dev, master_idx) ||
2123             link_kind_filtered(dev, kind_ops))
2124                 return true;
2125
2126         return false;
2127 }
2128
2129 /**
2130  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2131  * @sk: netlink socket
2132  * @netnsid: network namespace identifier
2133  *
2134  * Returns the network namespace identified by netnsid on success or an error
2135  * pointer on failure.
2136  */
2137 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2138 {
2139         struct net *net;
2140
2141         net = get_net_ns_by_id(sock_net(sk), netnsid);
2142         if (!net)
2143                 return ERR_PTR(-EINVAL);
2144
2145         /* For now, the caller is required to have CAP_NET_ADMIN in
2146          * the user namespace owning the target net ns.
2147          */
2148         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2149                 put_net(net);
2150                 return ERR_PTR(-EACCES);
2151         }
2152         return net;
2153 }
2154 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2155
2156 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2157                                       bool strict_check, struct nlattr **tb,
2158                                       struct netlink_ext_ack *extack)
2159 {
2160         int hdrlen;
2161
2162         if (strict_check) {
2163                 struct ifinfomsg *ifm;
2164
2165                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2166                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2167                         return -EINVAL;
2168                 }
2169
2170                 ifm = nlmsg_data(nlh);
2171                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2172                     ifm->ifi_change) {
2173                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2174                         return -EINVAL;
2175                 }
2176                 if (ifm->ifi_index) {
2177                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2178                         return -EINVAL;
2179                 }
2180
2181                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2182                                                      IFLA_MAX, ifla_policy,
2183                                                      extack);
2184         }
2185
2186         /* A hack to preserve kernel<->userspace interface.
2187          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2188          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2189          * what iproute2 < v3.9.0 used.
2190          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2191          * attribute, its netlink message is shorter than struct ifinfomsg.
2192          */
2193         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2194                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2195
2196         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2197                                       extack);
2198 }
2199
2200 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2201 {
2202         const struct rtnl_link_ops *kind_ops = NULL;
2203         struct netlink_ext_ack *extack = cb->extack;
2204         const struct nlmsghdr *nlh = cb->nlh;
2205         struct net *net = sock_net(skb->sk);
2206         unsigned int flags = NLM_F_MULTI;
2207         struct nlattr *tb[IFLA_MAX+1];
2208         struct {
2209                 unsigned long ifindex;
2210         } *ctx = (void *)cb->ctx;
2211         struct net *tgt_net = net;
2212         u32 ext_filter_mask = 0;
2213         struct net_device *dev;
2214         int master_idx = 0;
2215         int netnsid = -1;
2216         int err, i;
2217
2218         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2219         if (err < 0) {
2220                 if (cb->strict_check)
2221                         return err;
2222
2223                 goto walk_entries;
2224         }
2225
2226         for (i = 0; i <= IFLA_MAX; ++i) {
2227                 if (!tb[i])
2228                         continue;
2229
2230                 /* new attributes should only be added with strict checking */
2231                 switch (i) {
2232                 case IFLA_TARGET_NETNSID:
2233                         netnsid = nla_get_s32(tb[i]);
2234                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2235                         if (IS_ERR(tgt_net)) {
2236                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2237                                 return PTR_ERR(tgt_net);
2238                         }
2239                         break;
2240                 case IFLA_EXT_MASK:
2241                         ext_filter_mask = nla_get_u32(tb[i]);
2242                         break;
2243                 case IFLA_MASTER:
2244                         master_idx = nla_get_u32(tb[i]);
2245                         break;
2246                 case IFLA_LINKINFO:
2247                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2248                         break;
2249                 default:
2250                         if (cb->strict_check) {
2251                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2252                                 return -EINVAL;
2253                         }
2254                 }
2255         }
2256
2257         if (master_idx || kind_ops)
2258                 flags |= NLM_F_DUMP_FILTERED;
2259
2260 walk_entries:
2261         err = 0;
2262         for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
2263                 if (link_dump_filtered(dev, master_idx, kind_ops))
2264                         continue;
2265                 err = rtnl_fill_ifinfo(skb, dev, net, RTM_NEWLINK,
2266                                        NETLINK_CB(cb->skb).portid,
2267                                        nlh->nlmsg_seq, 0, flags,
2268                                        ext_filter_mask, 0, NULL, 0,
2269                                        netnsid, GFP_KERNEL);
2270                 if (err < 0) {
2271                         if (likely(skb->len))
2272                                 err = skb->len;
2273                         break;
2274                 }
2275         }
2276         cb->seq = tgt_net->dev_base_seq;
2277         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2278         if (netnsid >= 0)
2279                 put_net(tgt_net);
2280
2281         return err;
2282 }
2283
2284 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2285                              struct netlink_ext_ack *exterr)
2286 {
2287         const struct ifinfomsg *ifmp;
2288         const struct nlattr *attrs;
2289         size_t len;
2290
2291         ifmp = nla_data(nla_peer);
2292         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2293         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2294
2295         if (ifmp->ifi_index < 0) {
2296                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2297                                     "ifindex can't be negative");
2298                 return -EINVAL;
2299         }
2300
2301         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2302                                     exterr);
2303 }
2304 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2305
2306 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2307 {
2308         struct net *net;
2309         /* Examine the link attributes and figure out which
2310          * network namespace we are talking about.
2311          */
2312         if (tb[IFLA_NET_NS_PID])
2313                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2314         else if (tb[IFLA_NET_NS_FD])
2315                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2316         else
2317                 net = get_net(src_net);
2318         return net;
2319 }
2320 EXPORT_SYMBOL(rtnl_link_get_net);
2321
2322 /* Figure out which network namespace we are talking about by
2323  * examining the link attributes in the following order:
2324  *
2325  * 1. IFLA_NET_NS_PID
2326  * 2. IFLA_NET_NS_FD
2327  * 3. IFLA_TARGET_NETNSID
2328  */
2329 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2330                                                struct nlattr *tb[])
2331 {
2332         struct net *net;
2333
2334         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2335                 return rtnl_link_get_net(src_net, tb);
2336
2337         if (!tb[IFLA_TARGET_NETNSID])
2338                 return get_net(src_net);
2339
2340         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2341         if (!net)
2342                 return ERR_PTR(-EINVAL);
2343
2344         return net;
2345 }
2346
2347 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2348                                              struct net *src_net,
2349                                              struct nlattr *tb[], int cap)
2350 {
2351         struct net *net;
2352
2353         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2354         if (IS_ERR(net))
2355                 return net;
2356
2357         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2358                 put_net(net);
2359                 return ERR_PTR(-EPERM);
2360         }
2361
2362         return net;
2363 }
2364
2365 /* Verify that rtnetlink requests do not pass additional properties
2366  * potentially referring to different network namespaces.
2367  */
2368 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2369                                     struct netlink_ext_ack *extack,
2370                                     bool netns_id_only)
2371 {
2372
2373         if (netns_id_only) {
2374                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2375                         return 0;
2376
2377                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2378                 return -EOPNOTSUPP;
2379         }
2380
2381         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2382                 goto invalid_attr;
2383
2384         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2385                 goto invalid_attr;
2386
2387         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2388                 goto invalid_attr;
2389
2390         return 0;
2391
2392 invalid_attr:
2393         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2394         return -EINVAL;
2395 }
2396
2397 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2398                              int max_tx_rate)
2399 {
2400         const struct net_device_ops *ops = dev->netdev_ops;
2401
2402         if (!ops->ndo_set_vf_rate)
2403                 return -EOPNOTSUPP;
2404         if (max_tx_rate && max_tx_rate < min_tx_rate)
2405                 return -EINVAL;
2406
2407         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2408 }
2409
2410 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2411                             struct netlink_ext_ack *extack)
2412 {
2413         if (tb[IFLA_ADDRESS] &&
2414             nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2415                 return -EINVAL;
2416
2417         if (tb[IFLA_BROADCAST] &&
2418             nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2419                 return -EINVAL;
2420
2421         if (tb[IFLA_GSO_MAX_SIZE] &&
2422             nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2423                 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2424                 return -EINVAL;
2425         }
2426
2427         if (tb[IFLA_GSO_MAX_SEGS] &&
2428             (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2429              nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2430                 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2431                 return -EINVAL;
2432         }
2433
2434         if (tb[IFLA_GRO_MAX_SIZE] &&
2435             nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2436                 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2437                 return -EINVAL;
2438         }
2439
2440         if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2441             nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2442                 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2443                 return -EINVAL;
2444         }
2445
2446         if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2447             nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2448                 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2449                 return -EINVAL;
2450         }
2451
2452         if (tb[IFLA_AF_SPEC]) {
2453                 struct nlattr *af;
2454                 int rem, err;
2455
2456                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2457                         const struct rtnl_af_ops *af_ops;
2458
2459                         af_ops = rtnl_af_lookup(nla_type(af));
2460                         if (!af_ops)
2461                                 return -EAFNOSUPPORT;
2462
2463                         if (!af_ops->set_link_af)
2464                                 return -EOPNOTSUPP;
2465
2466                         if (af_ops->validate_link_af) {
2467                                 err = af_ops->validate_link_af(dev, af, extack);
2468                                 if (err < 0)
2469                                         return err;
2470                         }
2471                 }
2472         }
2473
2474         return 0;
2475 }
2476
2477 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2478                                   int guid_type)
2479 {
2480         const struct net_device_ops *ops = dev->netdev_ops;
2481
2482         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2483 }
2484
2485 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2486 {
2487         if (dev->type != ARPHRD_INFINIBAND)
2488                 return -EOPNOTSUPP;
2489
2490         return handle_infiniband_guid(dev, ivt, guid_type);
2491 }
2492
2493 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2494 {
2495         const struct net_device_ops *ops = dev->netdev_ops;
2496         int err = -EINVAL;
2497
2498         if (tb[IFLA_VF_MAC]) {
2499                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2500
2501                 if (ivm->vf >= INT_MAX)
2502                         return -EINVAL;
2503                 err = -EOPNOTSUPP;
2504                 if (ops->ndo_set_vf_mac)
2505                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2506                                                   ivm->mac);
2507                 if (err < 0)
2508                         return err;
2509         }
2510
2511         if (tb[IFLA_VF_VLAN]) {
2512                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2513
2514                 if (ivv->vf >= INT_MAX)
2515                         return -EINVAL;
2516                 err = -EOPNOTSUPP;
2517                 if (ops->ndo_set_vf_vlan)
2518                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2519                                                    ivv->qos,
2520                                                    htons(ETH_P_8021Q));
2521                 if (err < 0)
2522                         return err;
2523         }
2524
2525         if (tb[IFLA_VF_VLAN_LIST]) {
2526                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2527                 struct nlattr *attr;
2528                 int rem, len = 0;
2529
2530                 err = -EOPNOTSUPP;
2531                 if (!ops->ndo_set_vf_vlan)
2532                         return err;
2533
2534                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2535                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2536                             nla_len(attr) < NLA_HDRLEN) {
2537                                 return -EINVAL;
2538                         }
2539                         if (len >= MAX_VLAN_LIST_LEN)
2540                                 return -EOPNOTSUPP;
2541                         ivvl[len] = nla_data(attr);
2542
2543                         len++;
2544                 }
2545                 if (len == 0)
2546                         return -EINVAL;
2547
2548                 if (ivvl[0]->vf >= INT_MAX)
2549                         return -EINVAL;
2550                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2551                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2552                 if (err < 0)
2553                         return err;
2554         }
2555
2556         if (tb[IFLA_VF_TX_RATE]) {
2557                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2558                 struct ifla_vf_info ivf;
2559
2560                 if (ivt->vf >= INT_MAX)
2561                         return -EINVAL;
2562                 err = -EOPNOTSUPP;
2563                 if (ops->ndo_get_vf_config)
2564                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2565                 if (err < 0)
2566                         return err;
2567
2568                 err = rtnl_set_vf_rate(dev, ivt->vf,
2569                                        ivf.min_tx_rate, ivt->rate);
2570                 if (err < 0)
2571                         return err;
2572         }
2573
2574         if (tb[IFLA_VF_RATE]) {
2575                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2576
2577                 if (ivt->vf >= INT_MAX)
2578                         return -EINVAL;
2579
2580                 err = rtnl_set_vf_rate(dev, ivt->vf,
2581                                        ivt->min_tx_rate, ivt->max_tx_rate);
2582                 if (err < 0)
2583                         return err;
2584         }
2585
2586         if (tb[IFLA_VF_SPOOFCHK]) {
2587                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2588
2589                 if (ivs->vf >= INT_MAX)
2590                         return -EINVAL;
2591                 err = -EOPNOTSUPP;
2592                 if (ops->ndo_set_vf_spoofchk)
2593                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2594                                                        ivs->setting);
2595                 if (err < 0)
2596                         return err;
2597         }
2598
2599         if (tb[IFLA_VF_LINK_STATE]) {
2600                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2601
2602                 if (ivl->vf >= INT_MAX)
2603                         return -EINVAL;
2604                 err = -EOPNOTSUPP;
2605                 if (ops->ndo_set_vf_link_state)
2606                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2607                                                          ivl->link_state);
2608                 if (err < 0)
2609                         return err;
2610         }
2611
2612         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2613                 struct ifla_vf_rss_query_en *ivrssq_en;
2614
2615                 err = -EOPNOTSUPP;
2616                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2617                 if (ivrssq_en->vf >= INT_MAX)
2618                         return -EINVAL;
2619                 if (ops->ndo_set_vf_rss_query_en)
2620                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2621                                                            ivrssq_en->setting);
2622                 if (err < 0)
2623                         return err;
2624         }
2625
2626         if (tb[IFLA_VF_TRUST]) {
2627                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2628
2629                 if (ivt->vf >= INT_MAX)
2630                         return -EINVAL;
2631                 err = -EOPNOTSUPP;
2632                 if (ops->ndo_set_vf_trust)
2633                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2634                 if (err < 0)
2635                         return err;
2636         }
2637
2638         if (tb[IFLA_VF_IB_NODE_GUID]) {
2639                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2640
2641                 if (ivt->vf >= INT_MAX)
2642                         return -EINVAL;
2643                 if (!ops->ndo_set_vf_guid)
2644                         return -EOPNOTSUPP;
2645                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2646         }
2647
2648         if (tb[IFLA_VF_IB_PORT_GUID]) {
2649                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2650
2651                 if (ivt->vf >= INT_MAX)
2652                         return -EINVAL;
2653                 if (!ops->ndo_set_vf_guid)
2654                         return -EOPNOTSUPP;
2655
2656                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2657         }
2658
2659         return err;
2660 }
2661
2662 static int do_set_master(struct net_device *dev, int ifindex,
2663                          struct netlink_ext_ack *extack)
2664 {
2665         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2666         const struct net_device_ops *ops;
2667         int err;
2668
2669         if (upper_dev) {
2670                 if (upper_dev->ifindex == ifindex)
2671                         return 0;
2672                 ops = upper_dev->netdev_ops;
2673                 if (ops->ndo_del_slave) {
2674                         err = ops->ndo_del_slave(upper_dev, dev);
2675                         if (err)
2676                                 return err;
2677                 } else {
2678                         return -EOPNOTSUPP;
2679                 }
2680         }
2681
2682         if (ifindex) {
2683                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2684                 if (!upper_dev)
2685                         return -EINVAL;
2686                 ops = upper_dev->netdev_ops;
2687                 if (ops->ndo_add_slave) {
2688                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2689                         if (err)
2690                                 return err;
2691                 } else {
2692                         return -EOPNOTSUPP;
2693                 }
2694         }
2695         return 0;
2696 }
2697
2698 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2699         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2700         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2701 };
2702
2703 static int do_set_proto_down(struct net_device *dev,
2704                              struct nlattr *nl_proto_down,
2705                              struct nlattr *nl_proto_down_reason,
2706                              struct netlink_ext_ack *extack)
2707 {
2708         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2709         unsigned long mask = 0;
2710         u32 value;
2711         bool proto_down;
2712         int err;
2713
2714         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2715                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2716                 return -EOPNOTSUPP;
2717         }
2718
2719         if (nl_proto_down_reason) {
2720                 err = nla_parse_nested_deprecated(pdreason,
2721                                                   IFLA_PROTO_DOWN_REASON_MAX,
2722                                                   nl_proto_down_reason,
2723                                                   ifla_proto_down_reason_policy,
2724                                                   NULL);
2725                 if (err < 0)
2726                         return err;
2727
2728                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2729                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2730                         return -EINVAL;
2731                 }
2732
2733                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2734
2735                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2736                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2737
2738                 dev_change_proto_down_reason(dev, mask, value);
2739         }
2740
2741         if (nl_proto_down) {
2742                 proto_down = nla_get_u8(nl_proto_down);
2743
2744                 /* Don't turn off protodown if there are active reasons */
2745                 if (!proto_down && dev->proto_down_reason) {
2746                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2747                         return -EBUSY;
2748                 }
2749                 err = dev_change_proto_down(dev,
2750                                             proto_down);
2751                 if (err)
2752                         return err;
2753         }
2754
2755         return 0;
2756 }
2757
2758 #define DO_SETLINK_MODIFIED     0x01
2759 /* notify flag means notify + modified. */
2760 #define DO_SETLINK_NOTIFY       0x03
2761 static int do_setlink(const struct sk_buff *skb,
2762                       struct net_device *dev, struct ifinfomsg *ifm,
2763                       struct netlink_ext_ack *extack,
2764                       struct nlattr **tb, int status)
2765 {
2766         const struct net_device_ops *ops = dev->netdev_ops;
2767         char ifname[IFNAMSIZ];
2768         int err;
2769
2770         if (tb[IFLA_IFNAME])
2771                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2772         else
2773                 ifname[0] = '\0';
2774
2775         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2776                 const char *pat = ifname[0] ? ifname : NULL;
2777                 struct net *net;
2778                 int new_ifindex;
2779
2780                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2781                                                 tb, CAP_NET_ADMIN);
2782                 if (IS_ERR(net)) {
2783                         err = PTR_ERR(net);
2784                         goto errout;
2785                 }
2786
2787                 if (tb[IFLA_NEW_IFINDEX])
2788                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2789                 else
2790                         new_ifindex = 0;
2791
2792                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2793                 put_net(net);
2794                 if (err)
2795                         goto errout;
2796                 status |= DO_SETLINK_MODIFIED;
2797         }
2798
2799         if (tb[IFLA_MAP]) {
2800                 struct rtnl_link_ifmap *u_map;
2801                 struct ifmap k_map;
2802
2803                 if (!ops->ndo_set_config) {
2804                         err = -EOPNOTSUPP;
2805                         goto errout;
2806                 }
2807
2808                 if (!netif_device_present(dev)) {
2809                         err = -ENODEV;
2810                         goto errout;
2811                 }
2812
2813                 u_map = nla_data(tb[IFLA_MAP]);
2814                 k_map.mem_start = (unsigned long) u_map->mem_start;
2815                 k_map.mem_end = (unsigned long) u_map->mem_end;
2816                 k_map.base_addr = (unsigned short) u_map->base_addr;
2817                 k_map.irq = (unsigned char) u_map->irq;
2818                 k_map.dma = (unsigned char) u_map->dma;
2819                 k_map.port = (unsigned char) u_map->port;
2820
2821                 err = ops->ndo_set_config(dev, &k_map);
2822                 if (err < 0)
2823                         goto errout;
2824
2825                 status |= DO_SETLINK_NOTIFY;
2826         }
2827
2828         if (tb[IFLA_ADDRESS]) {
2829                 struct sockaddr *sa;
2830                 int len;
2831
2832                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2833                                                   sizeof(*sa));
2834                 sa = kmalloc(len, GFP_KERNEL);
2835                 if (!sa) {
2836                         err = -ENOMEM;
2837                         goto errout;
2838                 }
2839                 sa->sa_family = dev->type;
2840                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2841                        dev->addr_len);
2842                 err = dev_set_mac_address_user(dev, sa, extack);
2843                 kfree(sa);
2844                 if (err)
2845                         goto errout;
2846                 status |= DO_SETLINK_MODIFIED;
2847         }
2848
2849         if (tb[IFLA_MTU]) {
2850                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2851                 if (err < 0)
2852                         goto errout;
2853                 status |= DO_SETLINK_MODIFIED;
2854         }
2855
2856         if (tb[IFLA_GROUP]) {
2857                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2858                 status |= DO_SETLINK_NOTIFY;
2859         }
2860
2861         /*
2862          * Interface selected by interface index but interface
2863          * name provided implies that a name change has been
2864          * requested.
2865          */
2866         if (ifm->ifi_index > 0 && ifname[0]) {
2867                 err = dev_change_name(dev, ifname);
2868                 if (err < 0)
2869                         goto errout;
2870                 status |= DO_SETLINK_MODIFIED;
2871         }
2872
2873         if (tb[IFLA_IFALIAS]) {
2874                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2875                                     nla_len(tb[IFLA_IFALIAS]));
2876                 if (err < 0)
2877                         goto errout;
2878                 status |= DO_SETLINK_NOTIFY;
2879         }
2880
2881         if (tb[IFLA_BROADCAST]) {
2882                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2883                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2884         }
2885
2886         if (ifm->ifi_flags || ifm->ifi_change) {
2887                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2888                                        extack);
2889                 if (err < 0)
2890                         goto errout;
2891         }
2892
2893         if (tb[IFLA_MASTER]) {
2894                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2895                 if (err)
2896                         goto errout;
2897                 status |= DO_SETLINK_MODIFIED;
2898         }
2899
2900         if (tb[IFLA_CARRIER]) {
2901                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2902                 if (err)
2903                         goto errout;
2904                 status |= DO_SETLINK_MODIFIED;
2905         }
2906
2907         if (tb[IFLA_TXQLEN]) {
2908                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2909
2910                 err = dev_change_tx_queue_len(dev, value);
2911                 if (err)
2912                         goto errout;
2913                 status |= DO_SETLINK_MODIFIED;
2914         }
2915
2916         if (tb[IFLA_GSO_MAX_SIZE]) {
2917                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2918
2919                 if (dev->gso_max_size ^ max_size) {
2920                         netif_set_gso_max_size(dev, max_size);
2921                         status |= DO_SETLINK_MODIFIED;
2922                 }
2923         }
2924
2925         if (tb[IFLA_GSO_MAX_SEGS]) {
2926                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2927
2928                 if (dev->gso_max_segs ^ max_segs) {
2929                         netif_set_gso_max_segs(dev, max_segs);
2930                         status |= DO_SETLINK_MODIFIED;
2931                 }
2932         }
2933
2934         if (tb[IFLA_GRO_MAX_SIZE]) {
2935                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2936
2937                 if (dev->gro_max_size ^ gro_max_size) {
2938                         netif_set_gro_max_size(dev, gro_max_size);
2939                         status |= DO_SETLINK_MODIFIED;
2940                 }
2941         }
2942
2943         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2944                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2945
2946                 if (dev->gso_ipv4_max_size ^ max_size) {
2947                         netif_set_gso_ipv4_max_size(dev, max_size);
2948                         status |= DO_SETLINK_MODIFIED;
2949                 }
2950         }
2951
2952         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2953                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2954
2955                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2956                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2957                         status |= DO_SETLINK_MODIFIED;
2958                 }
2959         }
2960
2961         if (tb[IFLA_OPERSTATE])
2962                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2963
2964         if (tb[IFLA_LINKMODE]) {
2965                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2966
2967                 if (dev->link_mode ^ value)
2968                         status |= DO_SETLINK_NOTIFY;
2969                 WRITE_ONCE(dev->link_mode, value);
2970         }
2971
2972         if (tb[IFLA_VFINFO_LIST]) {
2973                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2974                 struct nlattr *attr;
2975                 int rem;
2976
2977                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2978                         if (nla_type(attr) != IFLA_VF_INFO ||
2979                             nla_len(attr) < NLA_HDRLEN) {
2980                                 err = -EINVAL;
2981                                 goto errout;
2982                         }
2983                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2984                                                           attr,
2985                                                           ifla_vf_policy,
2986                                                           NULL);
2987                         if (err < 0)
2988                                 goto errout;
2989                         err = do_setvfinfo(dev, vfinfo);
2990                         if (err < 0)
2991                                 goto errout;
2992                         status |= DO_SETLINK_NOTIFY;
2993                 }
2994         }
2995         err = 0;
2996
2997         if (tb[IFLA_VF_PORTS]) {
2998                 struct nlattr *port[IFLA_PORT_MAX+1];
2999                 struct nlattr *attr;
3000                 int vf;
3001                 int rem;
3002
3003                 err = -EOPNOTSUPP;
3004                 if (!ops->ndo_set_vf_port)
3005                         goto errout;
3006
3007                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
3008                         if (nla_type(attr) != IFLA_VF_PORT ||
3009                             nla_len(attr) < NLA_HDRLEN) {
3010                                 err = -EINVAL;
3011                                 goto errout;
3012                         }
3013                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3014                                                           attr,
3015                                                           ifla_port_policy,
3016                                                           NULL);
3017                         if (err < 0)
3018                                 goto errout;
3019                         if (!port[IFLA_PORT_VF]) {
3020                                 err = -EOPNOTSUPP;
3021                                 goto errout;
3022                         }
3023                         vf = nla_get_u32(port[IFLA_PORT_VF]);
3024                         err = ops->ndo_set_vf_port(dev, vf, port);
3025                         if (err < 0)
3026                                 goto errout;
3027                         status |= DO_SETLINK_NOTIFY;
3028                 }
3029         }
3030         err = 0;
3031
3032         if (tb[IFLA_PORT_SELF]) {
3033                 struct nlattr *port[IFLA_PORT_MAX+1];
3034
3035                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3036                                                   tb[IFLA_PORT_SELF],
3037                                                   ifla_port_policy, NULL);
3038                 if (err < 0)
3039                         goto errout;
3040
3041                 err = -EOPNOTSUPP;
3042                 if (ops->ndo_set_vf_port)
3043                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3044                 if (err < 0)
3045                         goto errout;
3046                 status |= DO_SETLINK_NOTIFY;
3047         }
3048
3049         if (tb[IFLA_AF_SPEC]) {
3050                 struct nlattr *af;
3051                 int rem;
3052
3053                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3054                         const struct rtnl_af_ops *af_ops;
3055
3056                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3057
3058                         err = af_ops->set_link_af(dev, af, extack);
3059                         if (err < 0)
3060                                 goto errout;
3061
3062                         status |= DO_SETLINK_NOTIFY;
3063                 }
3064         }
3065         err = 0;
3066
3067         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3068                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3069                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3070                 if (err)
3071                         goto errout;
3072                 status |= DO_SETLINK_NOTIFY;
3073         }
3074
3075         if (tb[IFLA_XDP]) {
3076                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3077                 u32 xdp_flags = 0;
3078
3079                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3080                                                   tb[IFLA_XDP],
3081                                                   ifla_xdp_policy, NULL);
3082                 if (err < 0)
3083                         goto errout;
3084
3085                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3086                         err = -EINVAL;
3087                         goto errout;
3088                 }
3089
3090                 if (xdp[IFLA_XDP_FLAGS]) {
3091                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3092                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3093                                 err = -EINVAL;
3094                                 goto errout;
3095                         }
3096                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3097                                 err = -EINVAL;
3098                                 goto errout;
3099                         }
3100                 }
3101
3102                 if (xdp[IFLA_XDP_FD]) {
3103                         int expected_fd = -1;
3104
3105                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3106                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3107                                         err = -EINVAL;
3108                                         goto errout;
3109                                 }
3110                                 expected_fd =
3111                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3112                         }
3113
3114                         err = dev_change_xdp_fd(dev, extack,
3115                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3116                                                 expected_fd,
3117                                                 xdp_flags);
3118                         if (err)
3119                                 goto errout;
3120                         status |= DO_SETLINK_NOTIFY;
3121                 }
3122         }
3123
3124 errout:
3125         if (status & DO_SETLINK_MODIFIED) {
3126                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3127                         netdev_state_change(dev);
3128
3129                 if (err < 0)
3130                         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",
3131                                              dev->name);
3132         }
3133
3134         return err;
3135 }
3136
3137 static struct net_device *rtnl_dev_get(struct net *net,
3138                                        struct nlattr *tb[])
3139 {
3140         char ifname[ALTIFNAMSIZ];
3141
3142         if (tb[IFLA_IFNAME])
3143                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3144         else if (tb[IFLA_ALT_IFNAME])
3145                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3146         else
3147                 return NULL;
3148
3149         return __dev_get_by_name(net, ifname);
3150 }
3151
3152 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3153                         struct netlink_ext_ack *extack)
3154 {
3155         struct net *net = sock_net(skb->sk);
3156         struct ifinfomsg *ifm;
3157         struct net_device *dev;
3158         int err;
3159         struct nlattr *tb[IFLA_MAX+1];
3160
3161         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3162                                      ifla_policy, extack);
3163         if (err < 0)
3164                 goto errout;
3165
3166         err = rtnl_ensure_unique_netns(tb, extack, false);
3167         if (err < 0)
3168                 goto errout;
3169
3170         err = -EINVAL;
3171         ifm = nlmsg_data(nlh);
3172         if (ifm->ifi_index > 0)
3173                 dev = __dev_get_by_index(net, ifm->ifi_index);
3174         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3175                 dev = rtnl_dev_get(net, tb);
3176         else
3177                 goto errout;
3178
3179         if (dev == NULL) {
3180                 err = -ENODEV;
3181                 goto errout;
3182         }
3183
3184         err = validate_linkmsg(dev, tb, extack);
3185         if (err < 0)
3186                 goto errout;
3187
3188         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3189 errout:
3190         return err;
3191 }
3192
3193 static int rtnl_group_dellink(const struct net *net, int group)
3194 {
3195         struct net_device *dev, *aux;
3196         LIST_HEAD(list_kill);
3197         bool found = false;
3198
3199         if (!group)
3200                 return -EPERM;
3201
3202         for_each_netdev(net, dev) {
3203                 if (dev->group == group) {
3204                         const struct rtnl_link_ops *ops;
3205
3206                         found = true;
3207                         ops = dev->rtnl_link_ops;
3208                         if (!ops || !ops->dellink)
3209                                 return -EOPNOTSUPP;
3210                 }
3211         }
3212
3213         if (!found)
3214                 return -ENODEV;
3215
3216         for_each_netdev_safe(net, dev, aux) {
3217                 if (dev->group == group) {
3218                         const struct rtnl_link_ops *ops;
3219
3220                         ops = dev->rtnl_link_ops;
3221                         ops->dellink(dev, &list_kill);
3222                 }
3223         }
3224         unregister_netdevice_many(&list_kill);
3225
3226         return 0;
3227 }
3228
3229 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3230 {
3231         const struct rtnl_link_ops *ops;
3232         LIST_HEAD(list_kill);
3233
3234         ops = dev->rtnl_link_ops;
3235         if (!ops || !ops->dellink)
3236                 return -EOPNOTSUPP;
3237
3238         ops->dellink(dev, &list_kill);
3239         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3240
3241         return 0;
3242 }
3243 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3244
3245 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3246                         struct netlink_ext_ack *extack)
3247 {
3248         struct net *net = sock_net(skb->sk);
3249         u32 portid = NETLINK_CB(skb).portid;
3250         struct net *tgt_net = net;
3251         struct net_device *dev = NULL;
3252         struct ifinfomsg *ifm;
3253         struct nlattr *tb[IFLA_MAX+1];
3254         int err;
3255         int netnsid = -1;
3256
3257         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3258                                      ifla_policy, extack);
3259         if (err < 0)
3260                 return err;
3261
3262         err = rtnl_ensure_unique_netns(tb, extack, true);
3263         if (err < 0)
3264                 return err;
3265
3266         if (tb[IFLA_TARGET_NETNSID]) {
3267                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3268                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3269                 if (IS_ERR(tgt_net))
3270                         return PTR_ERR(tgt_net);
3271         }
3272
3273         err = -EINVAL;
3274         ifm = nlmsg_data(nlh);
3275         if (ifm->ifi_index > 0)
3276                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3277         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3278                 dev = rtnl_dev_get(net, tb);
3279         else if (tb[IFLA_GROUP])
3280                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3281         else
3282                 goto out;
3283
3284         if (!dev) {
3285                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3286                         err = -ENODEV;
3287
3288                 goto out;
3289         }
3290
3291         err = rtnl_delete_link(dev, portid, nlh);
3292
3293 out:
3294         if (netnsid >= 0)
3295                 put_net(tgt_net);
3296
3297         return err;
3298 }
3299
3300 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3301                         u32 portid, const struct nlmsghdr *nlh)
3302 {
3303         unsigned int old_flags;
3304         int err;
3305
3306         old_flags = dev->flags;
3307         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3308                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3309                                          NULL);
3310                 if (err < 0)
3311                         return err;
3312         }
3313
3314         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3315                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3316         } else {
3317                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3318                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3319         }
3320         return 0;
3321 }
3322 EXPORT_SYMBOL(rtnl_configure_link);
3323
3324 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3325                                     unsigned char name_assign_type,
3326                                     const struct rtnl_link_ops *ops,
3327                                     struct nlattr *tb[],
3328                                     struct netlink_ext_ack *extack)
3329 {
3330         struct net_device *dev;
3331         unsigned int num_tx_queues = 1;
3332         unsigned int num_rx_queues = 1;
3333         int err;
3334
3335         if (tb[IFLA_NUM_TX_QUEUES])
3336                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3337         else if (ops->get_num_tx_queues)
3338                 num_tx_queues = ops->get_num_tx_queues();
3339
3340         if (tb[IFLA_NUM_RX_QUEUES])
3341                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3342         else if (ops->get_num_rx_queues)
3343                 num_rx_queues = ops->get_num_rx_queues();
3344
3345         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3346                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3347                 return ERR_PTR(-EINVAL);
3348         }
3349
3350         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3351                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3352                 return ERR_PTR(-EINVAL);
3353         }
3354
3355         if (ops->alloc) {
3356                 dev = ops->alloc(tb, ifname, name_assign_type,
3357                                  num_tx_queues, num_rx_queues);
3358                 if (IS_ERR(dev))
3359                         return dev;
3360         } else {
3361                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3362                                        name_assign_type, ops->setup,
3363                                        num_tx_queues, num_rx_queues);
3364         }
3365
3366         if (!dev)
3367                 return ERR_PTR(-ENOMEM);
3368
3369         err = validate_linkmsg(dev, tb, extack);
3370         if (err < 0) {
3371                 free_netdev(dev);
3372                 return ERR_PTR(err);
3373         }
3374
3375         dev_net_set(dev, net);
3376         dev->rtnl_link_ops = ops;
3377         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3378
3379         if (tb[IFLA_MTU]) {
3380                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3381
3382                 err = dev_validate_mtu(dev, mtu, extack);
3383                 if (err) {
3384                         free_netdev(dev);
3385                         return ERR_PTR(err);
3386                 }
3387                 dev->mtu = mtu;
3388         }
3389         if (tb[IFLA_ADDRESS]) {
3390                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3391                                nla_len(tb[IFLA_ADDRESS]));
3392                 dev->addr_assign_type = NET_ADDR_SET;
3393         }
3394         if (tb[IFLA_BROADCAST])
3395                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3396                                 nla_len(tb[IFLA_BROADCAST]));
3397         if (tb[IFLA_TXQLEN])
3398                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3399         if (tb[IFLA_OPERSTATE])
3400                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3401         if (tb[IFLA_LINKMODE])
3402                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3403         if (tb[IFLA_GROUP])
3404                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3405         if (tb[IFLA_GSO_MAX_SIZE])
3406                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3407         if (tb[IFLA_GSO_MAX_SEGS])
3408                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3409         if (tb[IFLA_GRO_MAX_SIZE])
3410                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3411         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3412                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3413         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3414                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3415
3416         return dev;
3417 }
3418 EXPORT_SYMBOL(rtnl_create_link);
3419
3420 static int rtnl_group_changelink(const struct sk_buff *skb,
3421                 struct net *net, int group,
3422                 struct ifinfomsg *ifm,
3423                 struct netlink_ext_ack *extack,
3424                 struct nlattr **tb)
3425 {
3426         struct net_device *dev, *aux;
3427         int err;
3428
3429         for_each_netdev_safe(net, dev, aux) {
3430                 if (dev->group == group) {
3431                         err = validate_linkmsg(dev, tb, extack);
3432                         if (err < 0)
3433                                 return err;
3434                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3435                         if (err < 0)
3436                                 return err;
3437                 }
3438         }
3439
3440         return 0;
3441 }
3442
3443 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3444                                const struct rtnl_link_ops *ops,
3445                                const struct nlmsghdr *nlh,
3446                                struct nlattr **tb, struct nlattr **data,
3447                                struct netlink_ext_ack *extack)
3448 {
3449         unsigned char name_assign_type = NET_NAME_USER;
3450         struct net *net = sock_net(skb->sk);
3451         u32 portid = NETLINK_CB(skb).portid;
3452         struct net *dest_net, *link_net;
3453         struct net_device *dev;
3454         char ifname[IFNAMSIZ];
3455         int err;
3456
3457         if (!ops->alloc && !ops->setup)
3458                 return -EOPNOTSUPP;
3459
3460         if (tb[IFLA_IFNAME]) {
3461                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3462         } else {
3463                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3464                 name_assign_type = NET_NAME_ENUM;
3465         }
3466
3467         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3468         if (IS_ERR(dest_net))
3469                 return PTR_ERR(dest_net);
3470
3471         if (tb[IFLA_LINK_NETNSID]) {
3472                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3473
3474                 link_net = get_net_ns_by_id(dest_net, id);
3475                 if (!link_net) {
3476                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3477                         err =  -EINVAL;
3478                         goto out;
3479                 }
3480                 err = -EPERM;
3481                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3482                         goto out;
3483         } else {
3484                 link_net = NULL;
3485         }
3486
3487         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3488                                name_assign_type, ops, tb, extack);
3489         if (IS_ERR(dev)) {
3490                 err = PTR_ERR(dev);
3491                 goto out;
3492         }
3493
3494         dev->ifindex = ifm->ifi_index;
3495
3496         if (ops->newlink)
3497                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3498         else
3499                 err = register_netdevice(dev);
3500         if (err < 0) {
3501                 free_netdev(dev);
3502                 goto out;
3503         }
3504
3505         err = rtnl_configure_link(dev, ifm, portid, nlh);
3506         if (err < 0)
3507                 goto out_unregister;
3508         if (link_net) {
3509                 err = dev_change_net_namespace(dev, dest_net, ifname);
3510                 if (err < 0)
3511                         goto out_unregister;
3512         }
3513         if (tb[IFLA_MASTER]) {
3514                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3515                 if (err)
3516                         goto out_unregister;
3517         }
3518 out:
3519         if (link_net)
3520                 put_net(link_net);
3521         put_net(dest_net);
3522         return err;
3523 out_unregister:
3524         if (ops->newlink) {
3525                 LIST_HEAD(list_kill);
3526
3527                 ops->dellink(dev, &list_kill);
3528                 unregister_netdevice_many(&list_kill);
3529         } else {
3530                 unregister_netdevice(dev);
3531         }
3532         goto out;
3533 }
3534
3535 struct rtnl_newlink_tbs {
3536         struct nlattr *tb[IFLA_MAX + 1];
3537         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3538         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3539 };
3540
3541 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3542                           struct rtnl_newlink_tbs *tbs,
3543                           struct netlink_ext_ack *extack)
3544 {
3545         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3546         struct nlattr ** const tb = tbs->tb;
3547         const struct rtnl_link_ops *m_ops;
3548         struct net_device *master_dev;
3549         struct net *net = sock_net(skb->sk);
3550         const struct rtnl_link_ops *ops;
3551         struct nlattr **slave_data;
3552         char kind[MODULE_NAME_LEN];
3553         struct net_device *dev;
3554         struct ifinfomsg *ifm;
3555         struct nlattr **data;
3556         bool link_specified;
3557         int err;
3558
3559 #ifdef CONFIG_MODULES
3560 replay:
3561 #endif
3562         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3563                                      ifla_policy, extack);
3564         if (err < 0)
3565                 return err;
3566
3567         err = rtnl_ensure_unique_netns(tb, extack, false);
3568         if (err < 0)
3569                 return err;
3570
3571         ifm = nlmsg_data(nlh);
3572         if (ifm->ifi_index > 0) {
3573                 link_specified = true;
3574                 dev = __dev_get_by_index(net, ifm->ifi_index);
3575         } else if (ifm->ifi_index < 0) {
3576                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3577                 return -EINVAL;
3578         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3579                 link_specified = true;
3580                 dev = rtnl_dev_get(net, tb);
3581         } else {
3582                 link_specified = false;
3583                 dev = NULL;
3584         }
3585
3586         master_dev = NULL;
3587         m_ops = NULL;
3588         if (dev) {
3589                 master_dev = netdev_master_upper_dev_get(dev);
3590                 if (master_dev)
3591                         m_ops = master_dev->rtnl_link_ops;
3592         }
3593
3594         if (tb[IFLA_LINKINFO]) {
3595                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3596                                                   tb[IFLA_LINKINFO],
3597                                                   ifla_info_policy, NULL);
3598                 if (err < 0)
3599                         return err;
3600         } else
3601                 memset(linkinfo, 0, sizeof(linkinfo));
3602
3603         if (linkinfo[IFLA_INFO_KIND]) {
3604                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3605                 ops = rtnl_link_ops_get(kind);
3606         } else {
3607                 kind[0] = '\0';
3608                 ops = NULL;
3609         }
3610
3611         data = NULL;
3612         if (ops) {
3613                 if (ops->maxtype > RTNL_MAX_TYPE)
3614                         return -EINVAL;
3615
3616                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3617                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3618                                                           linkinfo[IFLA_INFO_DATA],
3619                                                           ops->policy, extack);
3620                         if (err < 0)
3621                                 return err;
3622                         data = tbs->attr;
3623                 }
3624                 if (ops->validate) {
3625                         err = ops->validate(tb, data, extack);
3626                         if (err < 0)
3627                                 return err;
3628                 }
3629         }
3630
3631         slave_data = NULL;
3632         if (m_ops) {
3633                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3634                         return -EINVAL;
3635
3636                 if (m_ops->slave_maxtype &&
3637                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3638                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3639                                                           m_ops->slave_maxtype,
3640                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3641                                                           m_ops->slave_policy,
3642                                                           extack);
3643                         if (err < 0)
3644                                 return err;
3645                         slave_data = tbs->slave_attr;
3646                 }
3647         }
3648
3649         if (dev) {
3650                 int status = 0;
3651
3652                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3653                         return -EEXIST;
3654                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3655                         return -EOPNOTSUPP;
3656
3657                 err = validate_linkmsg(dev, tb, extack);
3658                 if (err < 0)
3659                         return err;
3660
3661                 if (linkinfo[IFLA_INFO_DATA]) {
3662                         if (!ops || ops != dev->rtnl_link_ops ||
3663                             !ops->changelink)
3664                                 return -EOPNOTSUPP;
3665
3666                         err = ops->changelink(dev, tb, data, extack);
3667                         if (err < 0)
3668                                 return err;
3669                         status |= DO_SETLINK_NOTIFY;
3670                 }
3671
3672                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3673                         if (!m_ops || !m_ops->slave_changelink)
3674                                 return -EOPNOTSUPP;
3675
3676                         err = m_ops->slave_changelink(master_dev, dev, tb,
3677                                                       slave_data, extack);
3678                         if (err < 0)
3679                                 return err;
3680                         status |= DO_SETLINK_NOTIFY;
3681                 }
3682
3683                 return do_setlink(skb, dev, ifm, extack, tb, status);
3684         }
3685
3686         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3687                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3688                  * or it's for a group
3689                 */
3690                 if (link_specified)
3691                         return -ENODEV;
3692                 if (tb[IFLA_GROUP])
3693                         return rtnl_group_changelink(skb, net,
3694                                                 nla_get_u32(tb[IFLA_GROUP]),
3695                                                 ifm, extack, tb);
3696                 return -ENODEV;
3697         }
3698
3699         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3700                 return -EOPNOTSUPP;
3701
3702         if (!ops) {
3703 #ifdef CONFIG_MODULES
3704                 if (kind[0]) {
3705                         __rtnl_unlock();
3706                         request_module("rtnl-link-%s", kind);
3707                         rtnl_lock();
3708                         ops = rtnl_link_ops_get(kind);
3709                         if (ops)
3710                                 goto replay;
3711                 }
3712 #endif
3713                 NL_SET_ERR_MSG(extack, "Unknown device type");
3714                 return -EOPNOTSUPP;
3715         }
3716
3717         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3718 }
3719
3720 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3721                         struct netlink_ext_ack *extack)
3722 {
3723         struct rtnl_newlink_tbs *tbs;
3724         int ret;
3725
3726         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3727         if (!tbs)
3728                 return -ENOMEM;
3729
3730         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3731         kfree(tbs);
3732         return ret;
3733 }
3734
3735 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3736                                   const struct nlmsghdr *nlh,
3737                                   struct nlattr **tb,
3738                                   struct netlink_ext_ack *extack)
3739 {
3740         struct ifinfomsg *ifm;
3741         int i, err;
3742
3743         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3744                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3745                 return -EINVAL;
3746         }
3747
3748         if (!netlink_strict_get_check(skb))
3749                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3750                                               ifla_policy, extack);
3751
3752         ifm = nlmsg_data(nlh);
3753         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3754             ifm->ifi_change) {
3755                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3756                 return -EINVAL;
3757         }
3758
3759         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3760                                             ifla_policy, extack);
3761         if (err)
3762                 return err;
3763
3764         for (i = 0; i <= IFLA_MAX; i++) {
3765                 if (!tb[i])
3766                         continue;
3767
3768                 switch (i) {
3769                 case IFLA_IFNAME:
3770                 case IFLA_ALT_IFNAME:
3771                 case IFLA_EXT_MASK:
3772                 case IFLA_TARGET_NETNSID:
3773                         break;
3774                 default:
3775                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3776                         return -EINVAL;
3777                 }
3778         }
3779
3780         return 0;
3781 }
3782
3783 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3784                         struct netlink_ext_ack *extack)
3785 {
3786         struct net *net = sock_net(skb->sk);
3787         struct net *tgt_net = net;
3788         struct ifinfomsg *ifm;
3789         struct nlattr *tb[IFLA_MAX+1];
3790         struct net_device *dev = NULL;
3791         struct sk_buff *nskb;
3792         int netnsid = -1;
3793         int err;
3794         u32 ext_filter_mask = 0;
3795
3796         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3797         if (err < 0)
3798                 return err;
3799
3800         err = rtnl_ensure_unique_netns(tb, extack, true);
3801         if (err < 0)
3802                 return err;
3803
3804         if (tb[IFLA_TARGET_NETNSID]) {
3805                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3806                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3807                 if (IS_ERR(tgt_net))
3808                         return PTR_ERR(tgt_net);
3809         }
3810
3811         if (tb[IFLA_EXT_MASK])
3812                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3813
3814         err = -EINVAL;
3815         ifm = nlmsg_data(nlh);
3816         if (ifm->ifi_index > 0)
3817                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3818         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3819                 dev = rtnl_dev_get(tgt_net, tb);
3820         else
3821                 goto out;
3822
3823         err = -ENODEV;
3824         if (dev == NULL)
3825                 goto out;
3826
3827         err = -ENOBUFS;
3828         nskb = nlmsg_new_large(if_nlmsg_size(dev, ext_filter_mask));
3829         if (nskb == NULL)
3830                 goto out;
3831
3832         /* Synchronize the carrier state so we don't report a state
3833          * that we're not actually going to honour immediately; if
3834          * the driver just did a carrier off->on transition, we can
3835          * only TX if link watch work has run, but without this we'd
3836          * already report carrier on, even if it doesn't work yet.
3837          */
3838         linkwatch_sync_dev(dev);
3839
3840         err = rtnl_fill_ifinfo(nskb, dev, net,
3841                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3842                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3843                                0, NULL, 0, netnsid, GFP_KERNEL);
3844         if (err < 0) {
3845                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3846                 WARN_ON(err == -EMSGSIZE);
3847                 kfree_skb(nskb);
3848         } else
3849                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3850 out:
3851         if (netnsid >= 0)
3852                 put_net(tgt_net);
3853
3854         return err;
3855 }
3856
3857 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3858                            bool *changed, struct netlink_ext_ack *extack)
3859 {
3860         char *alt_ifname;
3861         size_t size;
3862         int err;
3863
3864         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3865         if (err)
3866                 return err;
3867
3868         if (cmd == RTM_NEWLINKPROP) {
3869                 size = rtnl_prop_list_size(dev);
3870                 size += nla_total_size(ALTIFNAMSIZ);
3871                 if (size >= U16_MAX) {
3872                         NL_SET_ERR_MSG(extack,
3873                                        "effective property list too long");
3874                         return -EINVAL;
3875                 }
3876         }
3877
3878         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3879         if (!alt_ifname)
3880                 return -ENOMEM;
3881
3882         if (cmd == RTM_NEWLINKPROP) {
3883                 err = netdev_name_node_alt_create(dev, alt_ifname);
3884                 if (!err)
3885                         alt_ifname = NULL;
3886         } else if (cmd == RTM_DELLINKPROP) {
3887                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3888         } else {
3889                 WARN_ON_ONCE(1);
3890                 err = -EINVAL;
3891         }
3892
3893         kfree(alt_ifname);
3894         if (!err)
3895                 *changed = true;
3896         return err;
3897 }
3898
3899 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3900                          struct netlink_ext_ack *extack)
3901 {
3902         struct net *net = sock_net(skb->sk);
3903         struct nlattr *tb[IFLA_MAX + 1];
3904         struct net_device *dev;
3905         struct ifinfomsg *ifm;
3906         bool changed = false;
3907         struct nlattr *attr;
3908         int err, rem;
3909
3910         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3911         if (err)
3912                 return err;
3913
3914         err = rtnl_ensure_unique_netns(tb, extack, true);
3915         if (err)
3916                 return err;
3917
3918         ifm = nlmsg_data(nlh);
3919         if (ifm->ifi_index > 0)
3920                 dev = __dev_get_by_index(net, ifm->ifi_index);
3921         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3922                 dev = rtnl_dev_get(net, tb);
3923         else
3924                 return -EINVAL;
3925
3926         if (!dev)
3927                 return -ENODEV;
3928
3929         if (!tb[IFLA_PROP_LIST])
3930                 return 0;
3931
3932         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3933                 switch (nla_type(attr)) {
3934                 case IFLA_ALT_IFNAME:
3935                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3936                         if (err)
3937                                 return err;
3938                         break;
3939                 }
3940         }
3941
3942         if (changed)
3943                 netdev_state_change(dev);
3944         return 0;
3945 }
3946
3947 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3948                             struct netlink_ext_ack *extack)
3949 {
3950         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3951 }
3952
3953 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3954                             struct netlink_ext_ack *extack)
3955 {
3956         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3957 }
3958
3959 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3960 {
3961         struct net *net = sock_net(skb->sk);
3962         size_t min_ifinfo_dump_size = 0;
3963         struct nlattr *tb[IFLA_MAX+1];
3964         u32 ext_filter_mask = 0;
3965         struct net_device *dev;
3966         int hdrlen;
3967
3968         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3969         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3970                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3971
3972         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3973                 if (tb[IFLA_EXT_MASK])
3974                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3975         }
3976
3977         if (!ext_filter_mask)
3978                 return NLMSG_GOODSIZE;
3979         /*
3980          * traverse the list of net devices and compute the minimum
3981          * buffer size based upon the filter mask.
3982          */
3983         rcu_read_lock();
3984         for_each_netdev_rcu(net, dev) {
3985                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3986                                            if_nlmsg_size(dev, ext_filter_mask));
3987         }
3988         rcu_read_unlock();
3989
3990         return nlmsg_total_size(min_ifinfo_dump_size);
3991 }
3992
3993 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3994 {
3995         int idx;
3996         int s_idx = cb->family;
3997         int type = cb->nlh->nlmsg_type - RTM_BASE;
3998         int ret = 0;
3999
4000         if (s_idx == 0)
4001                 s_idx = 1;
4002
4003         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
4004                 struct rtnl_link __rcu **tab;
4005                 struct rtnl_link *link;
4006                 rtnl_dumpit_func dumpit;
4007
4008                 if (idx < s_idx || idx == PF_PACKET)
4009                         continue;
4010
4011                 if (type < 0 || type >= RTM_NR_MSGTYPES)
4012                         continue;
4013
4014                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
4015                 if (!tab)
4016                         continue;
4017
4018                 link = rcu_dereference_rtnl(tab[type]);
4019                 if (!link)
4020                         continue;
4021
4022                 dumpit = link->dumpit;
4023                 if (!dumpit)
4024                         continue;
4025
4026                 if (idx > s_idx) {
4027                         memset(&cb->args[0], 0, sizeof(cb->args));
4028                         cb->prev_seq = 0;
4029                         cb->seq = 0;
4030                 }
4031                 ret = dumpit(skb, cb);
4032                 if (ret)
4033                         break;
4034         }
4035         cb->family = idx;
4036
4037         return skb->len ? : ret;
4038 }
4039
4040 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4041                                        unsigned int change,
4042                                        u32 event, gfp_t flags, int *new_nsid,
4043                                        int new_ifindex, u32 portid,
4044                                        const struct nlmsghdr *nlh)
4045 {
4046         struct net *net = dev_net(dev);
4047         struct sk_buff *skb;
4048         int err = -ENOBUFS;
4049         u32 seq = 0;
4050
4051         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4052         if (skb == NULL)
4053                 goto errout;
4054
4055         if (nlmsg_report(nlh))
4056                 seq = nlmsg_seq(nlh);
4057         else
4058                 portid = 0;
4059
4060         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4061                                type, portid, seq, change, 0, 0, event,
4062                                new_nsid, new_ifindex, -1, flags);
4063         if (err < 0) {
4064                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4065                 WARN_ON(err == -EMSGSIZE);
4066                 kfree_skb(skb);
4067                 goto errout;
4068         }
4069         return skb;
4070 errout:
4071         if (err < 0)
4072                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4073         return NULL;
4074 }
4075
4076 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4077                        u32 portid, const struct nlmsghdr *nlh)
4078 {
4079         struct net *net = dev_net(dev);
4080
4081         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4082 }
4083
4084 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4085                                unsigned int change, u32 event,
4086                                gfp_t flags, int *new_nsid, int new_ifindex,
4087                                u32 portid, const struct nlmsghdr *nlh)
4088 {
4089         struct sk_buff *skb;
4090
4091         if (dev->reg_state != NETREG_REGISTERED)
4092                 return;
4093
4094         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4095                                      new_ifindex, portid, nlh);
4096         if (skb)
4097                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4098 }
4099
4100 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4101                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4102 {
4103         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4104                            NULL, 0, portid, nlh);
4105 }
4106
4107 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4108                          gfp_t flags, int *new_nsid, int new_ifindex)
4109 {
4110         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4111                            new_nsid, new_ifindex, 0, NULL);
4112 }
4113
4114 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4115                                    struct net_device *dev,
4116                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4117                                    int type, unsigned int flags,
4118                                    int nlflags, u16 ndm_state)
4119 {
4120         struct nlmsghdr *nlh;
4121         struct ndmsg *ndm;
4122
4123         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4124         if (!nlh)
4125                 return -EMSGSIZE;
4126
4127         ndm = nlmsg_data(nlh);
4128         ndm->ndm_family  = AF_BRIDGE;
4129         ndm->ndm_pad1    = 0;
4130         ndm->ndm_pad2    = 0;
4131         ndm->ndm_flags   = flags;
4132         ndm->ndm_type    = 0;
4133         ndm->ndm_ifindex = dev->ifindex;
4134         ndm->ndm_state   = ndm_state;
4135
4136         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4137                 goto nla_put_failure;
4138         if (vid)
4139                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4140                         goto nla_put_failure;
4141
4142         nlmsg_end(skb, nlh);
4143         return 0;
4144
4145 nla_put_failure:
4146         nlmsg_cancel(skb, nlh);
4147         return -EMSGSIZE;
4148 }
4149
4150 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4151 {
4152         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4153                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
4154                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4155                0;
4156 }
4157
4158 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4159                             u16 ndm_state)
4160 {
4161         struct net *net = dev_net(dev);
4162         struct sk_buff *skb;
4163         int err = -ENOBUFS;
4164
4165         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4166         if (!skb)
4167                 goto errout;
4168
4169         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4170                                       0, 0, type, NTF_SELF, 0, ndm_state);
4171         if (err < 0) {
4172                 kfree_skb(skb);
4173                 goto errout;
4174         }
4175
4176         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4177         return;
4178 errout:
4179         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4180 }
4181
4182 /*
4183  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4184  */
4185 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4186                      struct nlattr *tb[],
4187                      struct net_device *dev,
4188                      const unsigned char *addr, u16 vid,
4189                      u16 flags)
4190 {
4191         int err = -EINVAL;
4192
4193         /* If aging addresses are supported device will need to
4194          * implement its own handler for this.
4195          */
4196         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4197                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4198                 return err;
4199         }
4200
4201         if (tb[NDA_FLAGS_EXT]) {
4202                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4203                 return err;
4204         }
4205
4206         if (vid) {
4207                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4208                 return err;
4209         }
4210
4211         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4212                 err = dev_uc_add_excl(dev, addr);
4213         else if (is_multicast_ether_addr(addr))
4214                 err = dev_mc_add_excl(dev, addr);
4215
4216         /* Only return duplicate errors if NLM_F_EXCL is set */
4217         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4218                 err = 0;
4219
4220         return err;
4221 }
4222 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4223
4224 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4225                          struct netlink_ext_ack *extack)
4226 {
4227         u16 vid = 0;
4228
4229         if (vlan_attr) {
4230                 if (nla_len(vlan_attr) != sizeof(u16)) {
4231                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4232                         return -EINVAL;
4233                 }
4234
4235                 vid = nla_get_u16(vlan_attr);
4236
4237                 if (!vid || vid >= VLAN_VID_MASK) {
4238                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4239                         return -EINVAL;
4240                 }
4241         }
4242         *p_vid = vid;
4243         return 0;
4244 }
4245
4246 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4247                         struct netlink_ext_ack *extack)
4248 {
4249         struct net *net = sock_net(skb->sk);
4250         struct ndmsg *ndm;
4251         struct nlattr *tb[NDA_MAX+1];
4252         struct net_device *dev;
4253         u8 *addr;
4254         u16 vid;
4255         int err;
4256
4257         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4258                                      extack);
4259         if (err < 0)
4260                 return err;
4261
4262         ndm = nlmsg_data(nlh);
4263         if (ndm->ndm_ifindex == 0) {
4264                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4265                 return -EINVAL;
4266         }
4267
4268         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4269         if (dev == NULL) {
4270                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4271                 return -ENODEV;
4272         }
4273
4274         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4275                 NL_SET_ERR_MSG(extack, "invalid address");
4276                 return -EINVAL;
4277         }
4278
4279         if (dev->type != ARPHRD_ETHER) {
4280                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4281                 return -EINVAL;
4282         }
4283
4284         addr = nla_data(tb[NDA_LLADDR]);
4285
4286         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4287         if (err)
4288                 return err;
4289
4290         err = -EOPNOTSUPP;
4291
4292         /* Support fdb on master device the net/bridge default case */
4293         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4294             netif_is_bridge_port(dev)) {
4295                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4296                 const struct net_device_ops *ops = br_dev->netdev_ops;
4297
4298                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4299                                        nlh->nlmsg_flags, extack);
4300                 if (err)
4301                         goto out;
4302                 else
4303                         ndm->ndm_flags &= ~NTF_MASTER;
4304         }
4305
4306         /* Embedded bridge, macvlan, and any other device support */
4307         if ((ndm->ndm_flags & NTF_SELF)) {
4308                 if (dev->netdev_ops->ndo_fdb_add)
4309                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4310                                                            vid,
4311                                                            nlh->nlmsg_flags,
4312                                                            extack);
4313                 else
4314                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4315                                                nlh->nlmsg_flags);
4316
4317                 if (!err) {
4318                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4319                                         ndm->ndm_state);
4320                         ndm->ndm_flags &= ~NTF_SELF;
4321                 }
4322         }
4323 out:
4324         return err;
4325 }
4326
4327 /*
4328  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4329  */
4330 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4331                      struct nlattr *tb[],
4332                      struct net_device *dev,
4333                      const unsigned char *addr, u16 vid)
4334 {
4335         int err = -EINVAL;
4336
4337         /* If aging addresses are supported device will need to
4338          * implement its own handler for this.
4339          */
4340         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4341                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4342                 return err;
4343         }
4344
4345         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4346                 err = dev_uc_del(dev, addr);
4347         else if (is_multicast_ether_addr(addr))
4348                 err = dev_mc_del(dev, addr);
4349
4350         return err;
4351 }
4352 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4353
4354 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4355                         struct netlink_ext_ack *extack)
4356 {
4357         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4358         struct net *net = sock_net(skb->sk);
4359         const struct net_device_ops *ops;
4360         struct ndmsg *ndm;
4361         struct nlattr *tb[NDA_MAX+1];
4362         struct net_device *dev;
4363         __u8 *addr = NULL;
4364         int err;
4365         u16 vid;
4366
4367         if (!netlink_capable(skb, CAP_NET_ADMIN))
4368                 return -EPERM;
4369
4370         if (!del_bulk) {
4371                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4372                                              NULL, extack);
4373         } else {
4374                 /* For bulk delete, the drivers will parse the message with
4375                  * policy.
4376                  */
4377                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
4378         }
4379         if (err < 0)
4380                 return err;
4381
4382         ndm = nlmsg_data(nlh);
4383         if (ndm->ndm_ifindex == 0) {
4384                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4385                 return -EINVAL;
4386         }
4387
4388         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4389         if (dev == NULL) {
4390                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4391                 return -ENODEV;
4392         }
4393
4394         if (!del_bulk) {
4395                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4396                         NL_SET_ERR_MSG(extack, "invalid address");
4397                         return -EINVAL;
4398                 }
4399                 addr = nla_data(tb[NDA_LLADDR]);
4400
4401                 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4402                 if (err)
4403                         return err;
4404         }
4405
4406         if (dev->type != ARPHRD_ETHER) {
4407                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4408                 return -EINVAL;
4409         }
4410
4411         err = -EOPNOTSUPP;
4412
4413         /* Support fdb on master device the net/bridge default case */
4414         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4415             netif_is_bridge_port(dev)) {
4416                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4417
4418                 ops = br_dev->netdev_ops;
4419                 if (!del_bulk) {
4420                         if (ops->ndo_fdb_del)
4421                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4422                 } else {
4423                         if (ops->ndo_fdb_del_bulk)
4424                                 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4425                 }
4426
4427                 if (err)
4428                         goto out;
4429                 else
4430                         ndm->ndm_flags &= ~NTF_MASTER;
4431         }
4432
4433         /* Embedded bridge, macvlan, and any other device support */
4434         if (ndm->ndm_flags & NTF_SELF) {
4435                 ops = dev->netdev_ops;
4436                 if (!del_bulk) {
4437                         if (ops->ndo_fdb_del)
4438                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4439                         else
4440                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4441                 } else {
4442                         /* in case err was cleared by NTF_MASTER call */
4443                         err = -EOPNOTSUPP;
4444                         if (ops->ndo_fdb_del_bulk)
4445                                 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4446                 }
4447
4448                 if (!err) {
4449                         if (!del_bulk)
4450                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4451                                                 ndm->ndm_state);
4452                         ndm->ndm_flags &= ~NTF_SELF;
4453                 }
4454         }
4455 out:
4456         return err;
4457 }
4458
4459 static int nlmsg_populate_fdb(struct sk_buff *skb,
4460                               struct netlink_callback *cb,
4461                               struct net_device *dev,
4462                               int *idx,
4463                               struct netdev_hw_addr_list *list)
4464 {
4465         struct netdev_hw_addr *ha;
4466         int err;
4467         u32 portid, seq;
4468
4469         portid = NETLINK_CB(cb->skb).portid;
4470         seq = cb->nlh->nlmsg_seq;
4471
4472         list_for_each_entry(ha, &list->list, list) {
4473                 if (*idx < cb->args[2])
4474                         goto skip;
4475
4476                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4477                                               portid, seq,
4478                                               RTM_NEWNEIGH, NTF_SELF,
4479                                               NLM_F_MULTI, NUD_PERMANENT);
4480                 if (err < 0)
4481                         return err;
4482 skip:
4483                 *idx += 1;
4484         }
4485         return 0;
4486 }
4487
4488 /**
4489  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4490  * @skb: socket buffer to store message in
4491  * @cb: netlink callback
4492  * @dev: netdevice
4493  * @filter_dev: ignored
4494  * @idx: the number of FDB table entries dumped is added to *@idx
4495  *
4496  * Default netdevice operation to dump the existing unicast address list.
4497  * Returns number of addresses from list put in skb.
4498  */
4499 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4500                       struct netlink_callback *cb,
4501                       struct net_device *dev,
4502                       struct net_device *filter_dev,
4503                       int *idx)
4504 {
4505         int err;
4506
4507         if (dev->type != ARPHRD_ETHER)
4508                 return -EINVAL;
4509
4510         netif_addr_lock_bh(dev);
4511         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4512         if (err)
4513                 goto out;
4514         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4515 out:
4516         netif_addr_unlock_bh(dev);
4517         return err;
4518 }
4519 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4520
4521 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4522                                  int *br_idx, int *brport_idx,
4523                                  struct netlink_ext_ack *extack)
4524 {
4525         struct nlattr *tb[NDA_MAX + 1];
4526         struct ndmsg *ndm;
4527         int err, i;
4528
4529         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4530                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4531                 return -EINVAL;
4532         }
4533
4534         ndm = nlmsg_data(nlh);
4535         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4536             ndm->ndm_flags || ndm->ndm_type) {
4537                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4538                 return -EINVAL;
4539         }
4540
4541         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4542                                             NDA_MAX, NULL, extack);
4543         if (err < 0)
4544                 return err;
4545
4546         *brport_idx = ndm->ndm_ifindex;
4547         for (i = 0; i <= NDA_MAX; ++i) {
4548                 if (!tb[i])
4549                         continue;
4550
4551                 switch (i) {
4552                 case NDA_IFINDEX:
4553                         if (nla_len(tb[i]) != sizeof(u32)) {
4554                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4555                                 return -EINVAL;
4556                         }
4557                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4558                         break;
4559                 case NDA_MASTER:
4560                         if (nla_len(tb[i]) != sizeof(u32)) {
4561                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4562                                 return -EINVAL;
4563                         }
4564                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4565                         break;
4566                 default:
4567                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4568                         return -EINVAL;
4569                 }
4570         }
4571
4572         return 0;
4573 }
4574
4575 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4576                                  int *br_idx, int *brport_idx,
4577                                  struct netlink_ext_ack *extack)
4578 {
4579         struct nlattr *tb[IFLA_MAX+1];
4580         int err;
4581
4582         /* A hack to preserve kernel<->userspace interface.
4583          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4584          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4585          * So, check for ndmsg with an optional u32 attribute (not used here).
4586          * Fortunately these sizes don't conflict with the size of ifinfomsg
4587          * with an optional attribute.
4588          */
4589         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4590             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4591              nla_attr_size(sizeof(u32)))) {
4592                 struct ifinfomsg *ifm;
4593
4594                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4595                                              tb, IFLA_MAX, ifla_policy,
4596                                              extack);
4597                 if (err < 0) {
4598                         return -EINVAL;
4599                 } else if (err == 0) {
4600                         if (tb[IFLA_MASTER])
4601                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4602                 }
4603
4604                 ifm = nlmsg_data(nlh);
4605                 *brport_idx = ifm->ifi_index;
4606         }
4607         return 0;
4608 }
4609
4610 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4611 {
4612         struct net_device *dev;
4613         struct net_device *br_dev = NULL;
4614         const struct net_device_ops *ops = NULL;
4615         const struct net_device_ops *cops = NULL;
4616         struct net *net = sock_net(skb->sk);
4617         struct hlist_head *head;
4618         int brport_idx = 0;
4619         int br_idx = 0;
4620         int h, s_h;
4621         int idx = 0, s_idx;
4622         int err = 0;
4623         int fidx = 0;
4624
4625         if (cb->strict_check)
4626                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4627                                             cb->extack);
4628         else
4629                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4630                                             cb->extack);
4631         if (err < 0)
4632                 return err;
4633
4634         if (br_idx) {
4635                 br_dev = __dev_get_by_index(net, br_idx);
4636                 if (!br_dev)
4637                         return -ENODEV;
4638
4639                 ops = br_dev->netdev_ops;
4640         }
4641
4642         s_h = cb->args[0];
4643         s_idx = cb->args[1];
4644
4645         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4646                 idx = 0;
4647                 head = &net->dev_index_head[h];
4648                 hlist_for_each_entry(dev, head, index_hlist) {
4649
4650                         if (brport_idx && (dev->ifindex != brport_idx))
4651                                 continue;
4652
4653                         if (!br_idx) { /* user did not specify a specific bridge */
4654                                 if (netif_is_bridge_port(dev)) {
4655                                         br_dev = netdev_master_upper_dev_get(dev);
4656                                         cops = br_dev->netdev_ops;
4657                                 }
4658                         } else {
4659                                 if (dev != br_dev &&
4660                                     !netif_is_bridge_port(dev))
4661                                         continue;
4662
4663                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4664                                     !netif_is_bridge_master(dev))
4665                                         continue;
4666                                 cops = ops;
4667                         }
4668
4669                         if (idx < s_idx)
4670                                 goto cont;
4671
4672                         if (netif_is_bridge_port(dev)) {
4673                                 if (cops && cops->ndo_fdb_dump) {
4674                                         err = cops->ndo_fdb_dump(skb, cb,
4675                                                                 br_dev, dev,
4676                                                                 &fidx);
4677                                         if (err == -EMSGSIZE)
4678                                                 goto out;
4679                                 }
4680                         }
4681
4682                         if (dev->netdev_ops->ndo_fdb_dump)
4683                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4684                                                                     dev, NULL,
4685                                                                     &fidx);
4686                         else
4687                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4688                                                         &fidx);
4689                         if (err == -EMSGSIZE)
4690                                 goto out;
4691
4692                         cops = NULL;
4693
4694                         /* reset fdb offset to 0 for rest of the interfaces */
4695                         cb->args[2] = 0;
4696                         fidx = 0;
4697 cont:
4698                         idx++;
4699                 }
4700         }
4701
4702 out:
4703         cb->args[0] = h;
4704         cb->args[1] = idx;
4705         cb->args[2] = fidx;
4706
4707         return skb->len;
4708 }
4709
4710 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4711                                 struct nlattr **tb, u8 *ndm_flags,
4712                                 int *br_idx, int *brport_idx, u8 **addr,
4713                                 u16 *vid, struct netlink_ext_ack *extack)
4714 {
4715         struct ndmsg *ndm;
4716         int err, i;
4717
4718         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4719                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4720                 return -EINVAL;
4721         }
4722
4723         ndm = nlmsg_data(nlh);
4724         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4725             ndm->ndm_type) {
4726                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4727                 return -EINVAL;
4728         }
4729
4730         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4731                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4732                 return -EINVAL;
4733         }
4734
4735         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4736                                             NDA_MAX, nda_policy, extack);
4737         if (err < 0)
4738                 return err;
4739
4740         *ndm_flags = ndm->ndm_flags;
4741         *brport_idx = ndm->ndm_ifindex;
4742         for (i = 0; i <= NDA_MAX; ++i) {
4743                 if (!tb[i])
4744                         continue;
4745
4746                 switch (i) {
4747                 case NDA_MASTER:
4748                         *br_idx = nla_get_u32(tb[i]);
4749                         break;
4750                 case NDA_LLADDR:
4751                         if (nla_len(tb[i]) != ETH_ALEN) {
4752                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4753                                 return -EINVAL;
4754                         }
4755                         *addr = nla_data(tb[i]);
4756                         break;
4757                 case NDA_VLAN:
4758                         err = fdb_vid_parse(tb[i], vid, extack);
4759                         if (err)
4760                                 return err;
4761                         break;
4762                 case NDA_VNI:
4763                         break;
4764                 default:
4765                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4766                         return -EINVAL;
4767                 }
4768         }
4769
4770         return 0;
4771 }
4772
4773 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4774                         struct netlink_ext_ack *extack)
4775 {
4776         struct net_device *dev = NULL, *br_dev = NULL;
4777         const struct net_device_ops *ops = NULL;
4778         struct net *net = sock_net(in_skb->sk);
4779         struct nlattr *tb[NDA_MAX + 1];
4780         struct sk_buff *skb;
4781         int brport_idx = 0;
4782         u8 ndm_flags = 0;
4783         int br_idx = 0;
4784         u8 *addr = NULL;
4785         u16 vid = 0;
4786         int err;
4787
4788         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4789                                    &brport_idx, &addr, &vid, extack);
4790         if (err < 0)
4791                 return err;
4792
4793         if (!addr) {
4794                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4795                 return -EINVAL;
4796         }
4797
4798         if (brport_idx) {
4799                 dev = __dev_get_by_index(net, brport_idx);
4800                 if (!dev) {
4801                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4802                         return -ENODEV;
4803                 }
4804         }
4805
4806         if (br_idx) {
4807                 if (dev) {
4808                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4809                         return -EINVAL;
4810                 }
4811
4812                 br_dev = __dev_get_by_index(net, br_idx);
4813                 if (!br_dev) {
4814                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4815                         return -EINVAL;
4816                 }
4817                 ops = br_dev->netdev_ops;
4818         }
4819
4820         if (dev) {
4821                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4822                         if (!netif_is_bridge_port(dev)) {
4823                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4824                                 return -EINVAL;
4825                         }
4826                         br_dev = netdev_master_upper_dev_get(dev);
4827                         if (!br_dev) {
4828                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4829                                 return -EINVAL;
4830                         }
4831                         ops = br_dev->netdev_ops;
4832                 } else {
4833                         if (!(ndm_flags & NTF_SELF)) {
4834                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4835                                 return -EINVAL;
4836                         }
4837                         ops = dev->netdev_ops;
4838                 }
4839         }
4840
4841         if (!br_dev && !dev) {
4842                 NL_SET_ERR_MSG(extack, "No device specified");
4843                 return -ENODEV;
4844         }
4845
4846         if (!ops || !ops->ndo_fdb_get) {
4847                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4848                 return -EOPNOTSUPP;
4849         }
4850
4851         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4852         if (!skb)
4853                 return -ENOBUFS;
4854
4855         if (br_dev)
4856                 dev = br_dev;
4857         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4858                                NETLINK_CB(in_skb).portid,
4859                                nlh->nlmsg_seq, extack);
4860         if (err)
4861                 goto out;
4862
4863         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4864 out:
4865         kfree_skb(skb);
4866         return err;
4867 }
4868
4869 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4870                                unsigned int attrnum, unsigned int flag)
4871 {
4872         if (mask & flag)
4873                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4874         return 0;
4875 }
4876
4877 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4878                             struct net_device *dev, u16 mode,
4879                             u32 flags, u32 mask, int nlflags,
4880                             u32 filter_mask,
4881                             int (*vlan_fill)(struct sk_buff *skb,
4882                                              struct net_device *dev,
4883                                              u32 filter_mask))
4884 {
4885         struct nlmsghdr *nlh;
4886         struct ifinfomsg *ifm;
4887         struct nlattr *br_afspec;
4888         struct nlattr *protinfo;
4889         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4890         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4891         int err = 0;
4892
4893         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4894         if (nlh == NULL)
4895                 return -EMSGSIZE;
4896
4897         ifm = nlmsg_data(nlh);
4898         ifm->ifi_family = AF_BRIDGE;
4899         ifm->__ifi_pad = 0;
4900         ifm->ifi_type = dev->type;
4901         ifm->ifi_index = dev->ifindex;
4902         ifm->ifi_flags = dev_get_flags(dev);
4903         ifm->ifi_change = 0;
4904
4905
4906         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4907             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4908             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4909             (br_dev &&
4910              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4911             (dev->addr_len &&
4912              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4913             (dev->ifindex != dev_get_iflink(dev) &&
4914              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4915                 goto nla_put_failure;
4916
4917         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4918         if (!br_afspec)
4919                 goto nla_put_failure;
4920
4921         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4922                 nla_nest_cancel(skb, br_afspec);
4923                 goto nla_put_failure;
4924         }
4925
4926         if (mode != BRIDGE_MODE_UNDEF) {
4927                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4928                         nla_nest_cancel(skb, br_afspec);
4929                         goto nla_put_failure;
4930                 }
4931         }
4932         if (vlan_fill) {
4933                 err = vlan_fill(skb, dev, filter_mask);
4934                 if (err) {
4935                         nla_nest_cancel(skb, br_afspec);
4936                         goto nla_put_failure;
4937                 }
4938         }
4939         nla_nest_end(skb, br_afspec);
4940
4941         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4942         if (!protinfo)
4943                 goto nla_put_failure;
4944
4945         if (brport_nla_put_flag(skb, flags, mask,
4946                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4947             brport_nla_put_flag(skb, flags, mask,
4948                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4949             brport_nla_put_flag(skb, flags, mask,
4950                                 IFLA_BRPORT_FAST_LEAVE,
4951                                 BR_MULTICAST_FAST_LEAVE) ||
4952             brport_nla_put_flag(skb, flags, mask,
4953                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4954             brport_nla_put_flag(skb, flags, mask,
4955                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4956             brport_nla_put_flag(skb, flags, mask,
4957                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4958             brport_nla_put_flag(skb, flags, mask,
4959                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4960             brport_nla_put_flag(skb, flags, mask,
4961                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4962             brport_nla_put_flag(skb, flags, mask,
4963                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4964             brport_nla_put_flag(skb, flags, mask,
4965                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4966                 nla_nest_cancel(skb, protinfo);
4967                 goto nla_put_failure;
4968         }
4969
4970         nla_nest_end(skb, protinfo);
4971
4972         nlmsg_end(skb, nlh);
4973         return 0;
4974 nla_put_failure:
4975         nlmsg_cancel(skb, nlh);
4976         return err ? err : -EMSGSIZE;
4977 }
4978 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4979
4980 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4981                                     bool strict_check, u32 *filter_mask,
4982                                     struct netlink_ext_ack *extack)
4983 {
4984         struct nlattr *tb[IFLA_MAX+1];
4985         int err, i;
4986
4987         if (strict_check) {
4988                 struct ifinfomsg *ifm;
4989
4990                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4991                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4992                         return -EINVAL;
4993                 }
4994
4995                 ifm = nlmsg_data(nlh);
4996                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4997                     ifm->ifi_change || ifm->ifi_index) {
4998                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4999                         return -EINVAL;
5000                 }
5001
5002                 err = nlmsg_parse_deprecated_strict(nlh,
5003                                                     sizeof(struct ifinfomsg),
5004                                                     tb, IFLA_MAX, ifla_policy,
5005                                                     extack);
5006         } else {
5007                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
5008                                              tb, IFLA_MAX, ifla_policy,
5009                                              extack);
5010         }
5011         if (err < 0)
5012                 return err;
5013
5014         /* new attributes should only be added with strict checking */
5015         for (i = 0; i <= IFLA_MAX; ++i) {
5016                 if (!tb[i])
5017                         continue;
5018
5019                 switch (i) {
5020                 case IFLA_EXT_MASK:
5021                         *filter_mask = nla_get_u32(tb[i]);
5022                         break;
5023                 default:
5024                         if (strict_check) {
5025                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5026                                 return -EINVAL;
5027                         }
5028                 }
5029         }
5030
5031         return 0;
5032 }
5033
5034 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5035 {
5036         const struct nlmsghdr *nlh = cb->nlh;
5037         struct net *net = sock_net(skb->sk);
5038         struct net_device *dev;
5039         int idx = 0;
5040         u32 portid = NETLINK_CB(cb->skb).portid;
5041         u32 seq = nlh->nlmsg_seq;
5042         u32 filter_mask = 0;
5043         int err;
5044
5045         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5046                                        cb->extack);
5047         if (err < 0 && cb->strict_check)
5048                 return err;
5049
5050         rcu_read_lock();
5051         for_each_netdev_rcu(net, dev) {
5052                 const struct net_device_ops *ops = dev->netdev_ops;
5053                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5054
5055                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5056                         if (idx >= cb->args[0]) {
5057                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5058                                                 skb, portid, seq, dev,
5059                                                 filter_mask, NLM_F_MULTI);
5060                                 if (err < 0 && err != -EOPNOTSUPP) {
5061                                         if (likely(skb->len))
5062                                                 break;
5063
5064                                         goto out_err;
5065                                 }
5066                         }
5067                         idx++;
5068                 }
5069
5070                 if (ops->ndo_bridge_getlink) {
5071                         if (idx >= cb->args[0]) {
5072                                 err = ops->ndo_bridge_getlink(skb, portid,
5073                                                               seq, dev,
5074                                                               filter_mask,
5075                                                               NLM_F_MULTI);
5076                                 if (err < 0 && err != -EOPNOTSUPP) {
5077                                         if (likely(skb->len))
5078                                                 break;
5079
5080                                         goto out_err;
5081                                 }
5082                         }
5083                         idx++;
5084                 }
5085         }
5086         err = skb->len;
5087 out_err:
5088         rcu_read_unlock();
5089         cb->args[0] = idx;
5090
5091         return err;
5092 }
5093
5094 static inline size_t bridge_nlmsg_size(void)
5095 {
5096         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5097                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5098                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5099                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5100                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5101                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5102                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5103                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5104                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5105                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5106                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5107 }
5108
5109 static int rtnl_bridge_notify(struct net_device *dev)
5110 {
5111         struct net *net = dev_net(dev);
5112         struct sk_buff *skb;
5113         int err = -EOPNOTSUPP;
5114
5115         if (!dev->netdev_ops->ndo_bridge_getlink)
5116                 return 0;
5117
5118         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5119         if (!skb) {
5120                 err = -ENOMEM;
5121                 goto errout;
5122         }
5123
5124         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5125         if (err < 0)
5126                 goto errout;
5127
5128         /* Notification info is only filled for bridge ports, not the bridge
5129          * device itself. Therefore, a zero notification length is valid and
5130          * should not result in an error.
5131          */
5132         if (!skb->len)
5133                 goto errout;
5134
5135         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5136         return 0;
5137 errout:
5138         WARN_ON(err == -EMSGSIZE);
5139         kfree_skb(skb);
5140         if (err)
5141                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5142         return err;
5143 }
5144
5145 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5146                                struct netlink_ext_ack *extack)
5147 {
5148         struct net *net = sock_net(skb->sk);
5149         struct ifinfomsg *ifm;
5150         struct net_device *dev;
5151         struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
5152         int rem, err = -EOPNOTSUPP;
5153         u16 flags = 0;
5154
5155         if (nlmsg_len(nlh) < sizeof(*ifm))
5156                 return -EINVAL;
5157
5158         ifm = nlmsg_data(nlh);
5159         if (ifm->ifi_family != AF_BRIDGE)
5160                 return -EPFNOSUPPORT;
5161
5162         dev = __dev_get_by_index(net, ifm->ifi_index);
5163         if (!dev) {
5164                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5165                 return -ENODEV;
5166         }
5167
5168         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5169         if (br_spec) {
5170                 nla_for_each_nested(attr, br_spec, rem) {
5171                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
5172                                 if (nla_len(attr) < sizeof(flags))
5173                                         return -EINVAL;
5174
5175                                 br_flags_attr = attr;
5176                                 flags = nla_get_u16(attr);
5177                         }
5178
5179                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5180                                 if (nla_len(attr) < sizeof(u16))
5181                                         return -EINVAL;
5182                         }
5183                 }
5184         }
5185
5186         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5187                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5188
5189                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5190                         err = -EOPNOTSUPP;
5191                         goto out;
5192                 }
5193
5194                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5195                                                              extack);
5196                 if (err)
5197                         goto out;
5198
5199                 flags &= ~BRIDGE_FLAGS_MASTER;
5200         }
5201
5202         if ((flags & BRIDGE_FLAGS_SELF)) {
5203                 if (!dev->netdev_ops->ndo_bridge_setlink)
5204                         err = -EOPNOTSUPP;
5205                 else
5206                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5207                                                                   flags,
5208                                                                   extack);
5209                 if (!err) {
5210                         flags &= ~BRIDGE_FLAGS_SELF;
5211
5212                         /* Generate event to notify upper layer of bridge
5213                          * change
5214                          */
5215                         err = rtnl_bridge_notify(dev);
5216                 }
5217         }
5218
5219         if (br_flags_attr)
5220                 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
5221 out:
5222         return err;
5223 }
5224
5225 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5226                                struct netlink_ext_ack *extack)
5227 {
5228         struct net *net = sock_net(skb->sk);
5229         struct ifinfomsg *ifm;
5230         struct net_device *dev;
5231         struct nlattr *br_spec, *attr = NULL;
5232         int rem, err = -EOPNOTSUPP;
5233         u16 flags = 0;
5234         bool have_flags = false;
5235
5236         if (nlmsg_len(nlh) < sizeof(*ifm))
5237                 return -EINVAL;
5238
5239         ifm = nlmsg_data(nlh);
5240         if (ifm->ifi_family != AF_BRIDGE)
5241                 return -EPFNOSUPPORT;
5242
5243         dev = __dev_get_by_index(net, ifm->ifi_index);
5244         if (!dev) {
5245                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5246                 return -ENODEV;
5247         }
5248
5249         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5250         if (br_spec) {
5251                 nla_for_each_nested(attr, br_spec, rem) {
5252                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5253                                 if (nla_len(attr) < sizeof(flags))
5254                                         return -EINVAL;
5255
5256                                 have_flags = true;
5257                                 flags = nla_get_u16(attr);
5258                                 break;
5259                         }
5260                 }
5261         }
5262
5263         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5264                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5265
5266                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5267                         err = -EOPNOTSUPP;
5268                         goto out;
5269                 }
5270
5271                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5272                 if (err)
5273                         goto out;
5274
5275                 flags &= ~BRIDGE_FLAGS_MASTER;
5276         }
5277
5278         if ((flags & BRIDGE_FLAGS_SELF)) {
5279                 if (!dev->netdev_ops->ndo_bridge_dellink)
5280                         err = -EOPNOTSUPP;
5281                 else
5282                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5283                                                                   flags);
5284
5285                 if (!err) {
5286                         flags &= ~BRIDGE_FLAGS_SELF;
5287
5288                         /* Generate event to notify upper layer of bridge
5289                          * change
5290                          */
5291                         err = rtnl_bridge_notify(dev);
5292                 }
5293         }
5294
5295         if (have_flags)
5296                 memcpy(nla_data(attr), &flags, sizeof(flags));
5297 out:
5298         return err;
5299 }
5300
5301 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5302 {
5303         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5304                (!idxattr || idxattr == attrid);
5305 }
5306
5307 static bool
5308 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5309 {
5310         return dev->netdev_ops &&
5311                dev->netdev_ops->ndo_has_offload_stats &&
5312                dev->netdev_ops->ndo_get_offload_stats &&
5313                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5314 }
5315
5316 static unsigned int
5317 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5318 {
5319         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5320                sizeof(struct rtnl_link_stats64) : 0;
5321 }
5322
5323 static int
5324 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5325                              struct sk_buff *skb)
5326 {
5327         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5328         struct nlattr *attr = NULL;
5329         void *attr_data;
5330         int err;
5331
5332         if (!size)
5333                 return -ENODATA;
5334
5335         attr = nla_reserve_64bit(skb, attr_id, size,
5336                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5337         if (!attr)
5338                 return -EMSGSIZE;
5339
5340         attr_data = nla_data(attr);
5341         memset(attr_data, 0, size);
5342
5343         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5344         if (err)
5345                 return err;
5346
5347         return 0;
5348 }
5349
5350 static unsigned int
5351 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5352                                    enum netdev_offload_xstats_type type)
5353 {
5354         bool enabled = netdev_offload_xstats_enabled(dev, type);
5355
5356         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5357 }
5358
5359 struct rtnl_offload_xstats_request_used {
5360         bool request;
5361         bool used;
5362 };
5363
5364 static int
5365 rtnl_offload_xstats_get_stats(struct net_device *dev,
5366                               enum netdev_offload_xstats_type type,
5367                               struct rtnl_offload_xstats_request_used *ru,
5368                               struct rtnl_hw_stats64 *stats,
5369                               struct netlink_ext_ack *extack)
5370 {
5371         bool request;
5372         bool used;
5373         int err;
5374
5375         request = netdev_offload_xstats_enabled(dev, type);
5376         if (!request) {
5377                 used = false;
5378                 goto out;
5379         }
5380
5381         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5382         if (err)
5383                 return err;
5384
5385 out:
5386         if (ru) {
5387                 ru->request = request;
5388                 ru->used = used;
5389         }
5390         return 0;
5391 }
5392
5393 static int
5394 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5395                                        struct rtnl_offload_xstats_request_used *ru)
5396 {
5397         struct nlattr *nest;
5398
5399         nest = nla_nest_start(skb, attr_id);
5400         if (!nest)
5401                 return -EMSGSIZE;
5402
5403         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5404                 goto nla_put_failure;
5405
5406         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5407                 goto nla_put_failure;
5408
5409         nla_nest_end(skb, nest);
5410         return 0;
5411
5412 nla_put_failure:
5413         nla_nest_cancel(skb, nest);
5414         return -EMSGSIZE;
5415 }
5416
5417 static int
5418 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5419                                    struct netlink_ext_ack *extack)
5420 {
5421         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5422         struct rtnl_offload_xstats_request_used ru_l3;
5423         struct nlattr *nest;
5424         int err;
5425
5426         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5427         if (err)
5428                 return err;
5429
5430         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5431         if (!nest)
5432                 return -EMSGSIZE;
5433
5434         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5435                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5436                                                    &ru_l3))
5437                 goto nla_put_failure;
5438
5439         nla_nest_end(skb, nest);
5440         return 0;
5441
5442 nla_put_failure:
5443         nla_nest_cancel(skb, nest);
5444         return -EMSGSIZE;
5445 }
5446
5447 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5448                                     int *prividx, u32 off_filter_mask,
5449                                     struct netlink_ext_ack *extack)
5450 {
5451         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5452         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5453         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5454         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5455         bool have_data = false;
5456         int err;
5457
5458         if (*prividx <= attr_id_cpu_hit &&
5459             (off_filter_mask &
5460              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5461                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5462                 if (!err) {
5463                         have_data = true;
5464                 } else if (err != -ENODATA) {
5465                         *prividx = attr_id_cpu_hit;
5466                         return err;
5467                 }
5468         }
5469
5470         if (*prividx <= attr_id_hw_s_info &&
5471             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5472                 *prividx = attr_id_hw_s_info;
5473
5474                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5475                 if (err)
5476                         return err;
5477
5478                 have_data = true;
5479                 *prividx = 0;
5480         }
5481
5482         if (*prividx <= attr_id_l3_stats &&
5483             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5484                 unsigned int size_l3;
5485                 struct nlattr *attr;
5486
5487                 *prividx = attr_id_l3_stats;
5488
5489                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5490                 if (!size_l3)
5491                         goto skip_l3_stats;
5492                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5493                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5494                 if (!attr)
5495                         return -EMSGSIZE;
5496
5497                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5498                                                     nla_data(attr), extack);
5499                 if (err)
5500                         return err;
5501
5502                 have_data = true;
5503 skip_l3_stats:
5504                 *prividx = 0;
5505         }
5506
5507         if (!have_data)
5508                 return -ENODATA;
5509
5510         *prividx = 0;
5511         return 0;
5512 }
5513
5514 static unsigned int
5515 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5516                                            enum netdev_offload_xstats_type type)
5517 {
5518         return nla_total_size(0) +
5519                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5520                 nla_total_size(sizeof(u8)) +
5521                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5522                 nla_total_size(sizeof(u8)) +
5523                 0;
5524 }
5525
5526 static unsigned int
5527 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5528 {
5529         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5530
5531         return nla_total_size(0) +
5532                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5533                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5534                 0;
5535 }
5536
5537 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5538                                         u32 off_filter_mask)
5539 {
5540         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5541         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5542         int nla_size = 0;
5543         int size;
5544
5545         if (off_filter_mask &
5546             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5547                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5548                 nla_size += nla_total_size_64bit(size);
5549         }
5550
5551         if (off_filter_mask &
5552             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5553                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5554
5555         if (off_filter_mask &
5556             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5557                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5558                 nla_size += nla_total_size_64bit(size);
5559         }
5560
5561         if (nla_size != 0)
5562                 nla_size += nla_total_size(0);
5563
5564         return nla_size;
5565 }
5566
5567 struct rtnl_stats_dump_filters {
5568         /* mask[0] filters outer attributes. Then individual nests have their
5569          * filtering mask at the index of the nested attribute.
5570          */
5571         u32 mask[IFLA_STATS_MAX + 1];
5572 };
5573
5574 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5575                                int type, u32 pid, u32 seq, u32 change,
5576                                unsigned int flags,
5577                                const struct rtnl_stats_dump_filters *filters,
5578                                int *idxattr, int *prividx,
5579                                struct netlink_ext_ack *extack)
5580 {
5581         unsigned int filter_mask = filters->mask[0];
5582         struct if_stats_msg *ifsm;
5583         struct nlmsghdr *nlh;
5584         struct nlattr *attr;
5585         int s_prividx = *prividx;
5586         int err;
5587
5588         ASSERT_RTNL();
5589
5590         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5591         if (!nlh)
5592                 return -EMSGSIZE;
5593
5594         ifsm = nlmsg_data(nlh);
5595         ifsm->family = PF_UNSPEC;
5596         ifsm->pad1 = 0;
5597         ifsm->pad2 = 0;
5598         ifsm->ifindex = dev->ifindex;
5599         ifsm->filter_mask = filter_mask;
5600
5601         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5602                 struct rtnl_link_stats64 *sp;
5603
5604                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5605                                          sizeof(struct rtnl_link_stats64),
5606                                          IFLA_STATS_UNSPEC);
5607                 if (!attr) {
5608                         err = -EMSGSIZE;
5609                         goto nla_put_failure;
5610                 }
5611
5612                 sp = nla_data(attr);
5613                 dev_get_stats(dev, sp);
5614         }
5615
5616         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5617                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5618
5619                 if (ops && ops->fill_linkxstats) {
5620                         *idxattr = IFLA_STATS_LINK_XSTATS;
5621                         attr = nla_nest_start_noflag(skb,
5622                                                      IFLA_STATS_LINK_XSTATS);
5623                         if (!attr) {
5624                                 err = -EMSGSIZE;
5625                                 goto nla_put_failure;
5626                         }
5627
5628                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5629                         nla_nest_end(skb, attr);
5630                         if (err)
5631                                 goto nla_put_failure;
5632                         *idxattr = 0;
5633                 }
5634         }
5635
5636         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5637                              *idxattr)) {
5638                 const struct rtnl_link_ops *ops = NULL;
5639                 const struct net_device *master;
5640
5641                 master = netdev_master_upper_dev_get(dev);
5642                 if (master)
5643                         ops = master->rtnl_link_ops;
5644                 if (ops && ops->fill_linkxstats) {
5645                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5646                         attr = nla_nest_start_noflag(skb,
5647                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5648                         if (!attr) {
5649                                 err = -EMSGSIZE;
5650                                 goto nla_put_failure;
5651                         }
5652
5653                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5654                         nla_nest_end(skb, attr);
5655                         if (err)
5656                                 goto nla_put_failure;
5657                         *idxattr = 0;
5658                 }
5659         }
5660
5661         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5662                              *idxattr)) {
5663                 u32 off_filter_mask;
5664
5665                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5666                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5667                 attr = nla_nest_start_noflag(skb,
5668                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5669                 if (!attr) {
5670                         err = -EMSGSIZE;
5671                         goto nla_put_failure;
5672                 }
5673
5674                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5675                                                off_filter_mask, extack);
5676                 if (err == -ENODATA)
5677                         nla_nest_cancel(skb, attr);
5678                 else
5679                         nla_nest_end(skb, attr);
5680
5681                 if (err && err != -ENODATA)
5682                         goto nla_put_failure;
5683                 *idxattr = 0;
5684         }
5685
5686         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5687                 struct rtnl_af_ops *af_ops;
5688
5689                 *idxattr = IFLA_STATS_AF_SPEC;
5690                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5691                 if (!attr) {
5692                         err = -EMSGSIZE;
5693                         goto nla_put_failure;
5694                 }
5695
5696                 rcu_read_lock();
5697                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5698                         if (af_ops->fill_stats_af) {
5699                                 struct nlattr *af;
5700
5701                                 af = nla_nest_start_noflag(skb,
5702                                                            af_ops->family);
5703                                 if (!af) {
5704                                         rcu_read_unlock();
5705                                         err = -EMSGSIZE;
5706                                         goto nla_put_failure;
5707                                 }
5708                                 err = af_ops->fill_stats_af(skb, dev);
5709
5710                                 if (err == -ENODATA) {
5711                                         nla_nest_cancel(skb, af);
5712                                 } else if (err < 0) {
5713                                         rcu_read_unlock();
5714                                         goto nla_put_failure;
5715                                 }
5716
5717                                 nla_nest_end(skb, af);
5718                         }
5719                 }
5720                 rcu_read_unlock();
5721
5722                 nla_nest_end(skb, attr);
5723
5724                 *idxattr = 0;
5725         }
5726
5727         nlmsg_end(skb, nlh);
5728
5729         return 0;
5730
5731 nla_put_failure:
5732         /* not a multi message or no progress mean a real error */
5733         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5734                 nlmsg_cancel(skb, nlh);
5735         else
5736                 nlmsg_end(skb, nlh);
5737
5738         return err;
5739 }
5740
5741 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5742                                   const struct rtnl_stats_dump_filters *filters)
5743 {
5744         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5745         unsigned int filter_mask = filters->mask[0];
5746
5747         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5748                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5749
5750         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5751                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5752                 int attr = IFLA_STATS_LINK_XSTATS;
5753
5754                 if (ops && ops->get_linkxstats_size) {
5755                         size += nla_total_size(ops->get_linkxstats_size(dev,
5756                                                                         attr));
5757                         /* for IFLA_STATS_LINK_XSTATS */
5758                         size += nla_total_size(0);
5759                 }
5760         }
5761
5762         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5763                 struct net_device *_dev = (struct net_device *)dev;
5764                 const struct rtnl_link_ops *ops = NULL;
5765                 const struct net_device *master;
5766
5767                 /* netdev_master_upper_dev_get can't take const */
5768                 master = netdev_master_upper_dev_get(_dev);
5769                 if (master)
5770                         ops = master->rtnl_link_ops;
5771                 if (ops && ops->get_linkxstats_size) {
5772                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5773
5774                         size += nla_total_size(ops->get_linkxstats_size(dev,
5775                                                                         attr));
5776                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5777                         size += nla_total_size(0);
5778                 }
5779         }
5780
5781         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5782                 u32 off_filter_mask;
5783
5784                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5785                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5786         }
5787
5788         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5789                 struct rtnl_af_ops *af_ops;
5790
5791                 /* for IFLA_STATS_AF_SPEC */
5792                 size += nla_total_size(0);
5793
5794                 rcu_read_lock();
5795                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5796                         if (af_ops->get_stats_af_size) {
5797                                 size += nla_total_size(
5798                                         af_ops->get_stats_af_size(dev));
5799
5800                                 /* for AF_* */
5801                                 size += nla_total_size(0);
5802                         }
5803                 }
5804                 rcu_read_unlock();
5805         }
5806
5807         return size;
5808 }
5809
5810 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5811
5812 static const struct nla_policy
5813 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5814         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5815                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5816 };
5817
5818 static const struct nla_policy
5819 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5820         [IFLA_STATS_GET_FILTERS] =
5821                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5822 };
5823
5824 static const struct nla_policy
5825 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5826         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5827 };
5828
5829 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5830                                         struct rtnl_stats_dump_filters *filters,
5831                                         struct netlink_ext_ack *extack)
5832 {
5833         struct nlattr *tb[IFLA_STATS_MAX + 1];
5834         int err;
5835         int at;
5836
5837         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5838                                rtnl_stats_get_policy_filters, extack);
5839         if (err < 0)
5840                 return err;
5841
5842         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5843                 if (tb[at]) {
5844                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5845                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5846                                 return -EINVAL;
5847                         }
5848                         filters->mask[at] = nla_get_u32(tb[at]);
5849                 }
5850         }
5851
5852         return 0;
5853 }
5854
5855 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5856                                 u32 filter_mask,
5857                                 struct rtnl_stats_dump_filters *filters,
5858                                 struct netlink_ext_ack *extack)
5859 {
5860         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5861         int err;
5862         int i;
5863
5864         filters->mask[0] = filter_mask;
5865         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5866                 filters->mask[i] = -1U;
5867
5868         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5869                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5870         if (err < 0)
5871                 return err;
5872
5873         if (tb[IFLA_STATS_GET_FILTERS]) {
5874                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5875                                                    filters, extack);
5876                 if (err)
5877                         return err;
5878         }
5879
5880         return 0;
5881 }
5882
5883 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5884                                 bool is_dump, struct netlink_ext_ack *extack)
5885 {
5886         struct if_stats_msg *ifsm;
5887
5888         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5889                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5890                 return -EINVAL;
5891         }
5892
5893         if (!strict_check)
5894                 return 0;
5895
5896         ifsm = nlmsg_data(nlh);
5897
5898         /* only requests using strict checks can pass data to influence
5899          * the dump. The legacy exception is filter_mask.
5900          */
5901         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5902                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5903                 return -EINVAL;
5904         }
5905         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5906                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5907                 return -EINVAL;
5908         }
5909
5910         return 0;
5911 }
5912
5913 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5914                           struct netlink_ext_ack *extack)
5915 {
5916         struct rtnl_stats_dump_filters filters;
5917         struct net *net = sock_net(skb->sk);
5918         struct net_device *dev = NULL;
5919         int idxattr = 0, prividx = 0;
5920         struct if_stats_msg *ifsm;
5921         struct sk_buff *nskb;
5922         int err;
5923
5924         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5925                                    false, extack);
5926         if (err)
5927                 return err;
5928
5929         ifsm = nlmsg_data(nlh);
5930         if (ifsm->ifindex > 0)
5931                 dev = __dev_get_by_index(net, ifsm->ifindex);
5932         else
5933                 return -EINVAL;
5934
5935         if (!dev)
5936                 return -ENODEV;
5937
5938         if (!ifsm->filter_mask) {
5939                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5940                 return -EINVAL;
5941         }
5942
5943         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5944         if (err)
5945                 return err;
5946
5947         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5948         if (!nskb)
5949                 return -ENOBUFS;
5950
5951         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5952                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5953                                   0, &filters, &idxattr, &prividx, extack);
5954         if (err < 0) {
5955                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5956                 WARN_ON(err == -EMSGSIZE);
5957                 kfree_skb(nskb);
5958         } else {
5959                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5960         }
5961
5962         return err;
5963 }
5964
5965 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5966 {
5967         struct netlink_ext_ack *extack = cb->extack;
5968         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5969         struct rtnl_stats_dump_filters filters;
5970         struct net *net = sock_net(skb->sk);
5971         unsigned int flags = NLM_F_MULTI;
5972         struct if_stats_msg *ifsm;
5973         struct hlist_head *head;
5974         struct net_device *dev;
5975         int idx = 0;
5976
5977         s_h = cb->args[0];
5978         s_idx = cb->args[1];
5979         s_idxattr = cb->args[2];
5980         s_prividx = cb->args[3];
5981
5982         cb->seq = net->dev_base_seq;
5983
5984         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5985         if (err)
5986                 return err;
5987
5988         ifsm = nlmsg_data(cb->nlh);
5989         if (!ifsm->filter_mask) {
5990                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5991                 return -EINVAL;
5992         }
5993
5994         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5995                                    extack);
5996         if (err)
5997                 return err;
5998
5999         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
6000                 idx = 0;
6001                 head = &net->dev_index_head[h];
6002                 hlist_for_each_entry(dev, head, index_hlist) {
6003                         if (idx < s_idx)
6004                                 goto cont;
6005                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
6006                                                   NETLINK_CB(cb->skb).portid,
6007                                                   cb->nlh->nlmsg_seq, 0,
6008                                                   flags, &filters,
6009                                                   &s_idxattr, &s_prividx,
6010                                                   extack);
6011                         /* If we ran out of room on the first message,
6012                          * we're in trouble
6013                          */
6014                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6015
6016                         if (err < 0)
6017                                 goto out;
6018                         s_prividx = 0;
6019                         s_idxattr = 0;
6020                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6021 cont:
6022                         idx++;
6023                 }
6024         }
6025 out:
6026         cb->args[3] = s_prividx;
6027         cb->args[2] = s_idxattr;
6028         cb->args[1] = idx;
6029         cb->args[0] = h;
6030
6031         return skb->len;
6032 }
6033
6034 void rtnl_offload_xstats_notify(struct net_device *dev)
6035 {
6036         struct rtnl_stats_dump_filters response_filters = {};
6037         struct net *net = dev_net(dev);
6038         int idxattr = 0, prividx = 0;
6039         struct sk_buff *skb;
6040         int err = -ENOBUFS;
6041
6042         ASSERT_RTNL();
6043
6044         response_filters.mask[0] |=
6045                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6046         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6047                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6048
6049         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6050                         GFP_KERNEL);
6051         if (!skb)
6052                 goto errout;
6053
6054         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6055                                   &response_filters, &idxattr, &prividx, NULL);
6056         if (err < 0) {
6057                 kfree_skb(skb);
6058                 goto errout;
6059         }
6060
6061         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6062         return;
6063
6064 errout:
6065         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6066 }
6067 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6068
6069 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6070                           struct netlink_ext_ack *extack)
6071 {
6072         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6073         struct rtnl_stats_dump_filters response_filters = {};
6074         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6075         struct net *net = sock_net(skb->sk);
6076         struct net_device *dev = NULL;
6077         struct if_stats_msg *ifsm;
6078         bool notify = false;
6079         int err;
6080
6081         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6082                                    false, extack);
6083         if (err)
6084                 return err;
6085
6086         ifsm = nlmsg_data(nlh);
6087         if (ifsm->family != AF_UNSPEC) {
6088                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6089                 return -EINVAL;
6090         }
6091
6092         if (ifsm->ifindex > 0)
6093                 dev = __dev_get_by_index(net, ifsm->ifindex);
6094         else
6095                 return -EINVAL;
6096
6097         if (!dev)
6098                 return -ENODEV;
6099
6100         if (ifsm->filter_mask) {
6101                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6102                 return -EINVAL;
6103         }
6104
6105         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6106                           ifla_stats_set_policy, extack);
6107         if (err < 0)
6108                 return err;
6109
6110         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6111                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6112
6113                 if (req)
6114                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6115                 else
6116                         err = netdev_offload_xstats_disable(dev, t_l3);
6117
6118                 if (!err)
6119                         notify = true;
6120                 else if (err != -EALREADY)
6121                         return err;
6122
6123                 response_filters.mask[0] |=
6124                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6125                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6126                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6127         }
6128
6129         if (notify)
6130                 rtnl_offload_xstats_notify(dev);
6131
6132         return 0;
6133 }
6134
6135 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6136                                    struct netlink_ext_ack *extack)
6137 {
6138         struct br_port_msg *bpm;
6139
6140         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6141                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6142                 return -EINVAL;
6143         }
6144
6145         bpm = nlmsg_data(nlh);
6146         if (bpm->ifindex) {
6147                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6148                 return -EINVAL;
6149         }
6150         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6151                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6152                 return -EINVAL;
6153         }
6154
6155         return 0;
6156 }
6157
6158 struct rtnl_mdb_dump_ctx {
6159         long idx;
6160 };
6161
6162 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6163 {
6164         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6165         struct net *net = sock_net(skb->sk);
6166         struct net_device *dev;
6167         int idx, s_idx;
6168         int err;
6169
6170         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6171
6172         if (cb->strict_check) {
6173                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6174                 if (err)
6175                         return err;
6176         }
6177
6178         s_idx = ctx->idx;
6179         idx = 0;
6180
6181         for_each_netdev(net, dev) {
6182                 if (idx < s_idx)
6183                         goto skip;
6184                 if (!dev->netdev_ops->ndo_mdb_dump)
6185                         goto skip;
6186
6187                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6188                 if (err == -EMSGSIZE)
6189                         goto out;
6190                 /* Moving on to next device, reset markers and sequence
6191                  * counters since they are all maintained per-device.
6192                  */
6193                 memset(cb->ctx, 0, sizeof(cb->ctx));
6194                 cb->prev_seq = 0;
6195                 cb->seq = 0;
6196 skip:
6197                 idx++;
6198         }
6199
6200 out:
6201         ctx->idx = idx;
6202         return skb->len;
6203 }
6204
6205 static int rtnl_validate_mdb_entry_get(const struct nlattr *attr,
6206                                        struct netlink_ext_ack *extack)
6207 {
6208         struct br_mdb_entry *entry = nla_data(attr);
6209
6210         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6211                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6212                 return -EINVAL;
6213         }
6214
6215         if (entry->ifindex) {
6216                 NL_SET_ERR_MSG(extack, "Entry ifindex cannot be specified");
6217                 return -EINVAL;
6218         }
6219
6220         if (entry->state) {
6221                 NL_SET_ERR_MSG(extack, "Entry state cannot be specified");
6222                 return -EINVAL;
6223         }
6224
6225         if (entry->flags) {
6226                 NL_SET_ERR_MSG(extack, "Entry flags cannot be specified");
6227                 return -EINVAL;
6228         }
6229
6230         if (entry->vid >= VLAN_VID_MASK) {
6231                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6232                 return -EINVAL;
6233         }
6234
6235         if (entry->addr.proto != htons(ETH_P_IP) &&
6236             entry->addr.proto != htons(ETH_P_IPV6) &&
6237             entry->addr.proto != 0) {
6238                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6239                 return -EINVAL;
6240         }
6241
6242         return 0;
6243 }
6244
6245 static const struct nla_policy mdba_get_policy[MDBA_GET_ENTRY_MAX + 1] = {
6246         [MDBA_GET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6247                                                   rtnl_validate_mdb_entry_get,
6248                                                   sizeof(struct br_mdb_entry)),
6249         [MDBA_GET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6250 };
6251
6252 static int rtnl_mdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6253                         struct netlink_ext_ack *extack)
6254 {
6255         struct nlattr *tb[MDBA_GET_ENTRY_MAX + 1];
6256         struct net *net = sock_net(in_skb->sk);
6257         struct br_port_msg *bpm;
6258         struct net_device *dev;
6259         int err;
6260
6261         err = nlmsg_parse(nlh, sizeof(struct br_port_msg), tb,
6262                           MDBA_GET_ENTRY_MAX, mdba_get_policy, extack);
6263         if (err)
6264                 return err;
6265
6266         bpm = nlmsg_data(nlh);
6267         if (!bpm->ifindex) {
6268                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6269                 return -EINVAL;
6270         }
6271
6272         dev = __dev_get_by_index(net, bpm->ifindex);
6273         if (!dev) {
6274                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6275                 return -ENODEV;
6276         }
6277
6278         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_GET_ENTRY)) {
6279                 NL_SET_ERR_MSG(extack, "Missing MDBA_GET_ENTRY attribute");
6280                 return -EINVAL;
6281         }
6282
6283         if (!dev->netdev_ops->ndo_mdb_get) {
6284                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6285                 return -EOPNOTSUPP;
6286         }
6287
6288         return dev->netdev_ops->ndo_mdb_get(dev, tb, NETLINK_CB(in_skb).portid,
6289                                             nlh->nlmsg_seq, extack);
6290 }
6291
6292 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6293                                    struct netlink_ext_ack *extack)
6294 {
6295         struct br_mdb_entry *entry = nla_data(attr);
6296
6297         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6298                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6299                 return -EINVAL;
6300         }
6301
6302         if (entry->ifindex == 0) {
6303                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6304                 return -EINVAL;
6305         }
6306
6307         if (entry->addr.proto == htons(ETH_P_IP)) {
6308                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6309                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6310                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6311                         return -EINVAL;
6312                 }
6313                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6314                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6315                         return -EINVAL;
6316                 }
6317 #if IS_ENABLED(CONFIG_IPV6)
6318         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6319                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6320                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6321                         return -EINVAL;
6322                 }
6323 #endif
6324         } else if (entry->addr.proto == 0) {
6325                 /* L2 mdb */
6326                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6327                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6328                         return -EINVAL;
6329                 }
6330         } else {
6331                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6332                 return -EINVAL;
6333         }
6334
6335         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6336                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6337                 return -EINVAL;
6338         }
6339         if (entry->vid >= VLAN_VID_MASK) {
6340                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6341                 return -EINVAL;
6342         }
6343
6344         return 0;
6345 }
6346
6347 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6348         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6349         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6350                                                   rtnl_validate_mdb_entry,
6351                                                   sizeof(struct br_mdb_entry)),
6352         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6353 };
6354
6355 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6356                         struct netlink_ext_ack *extack)
6357 {
6358         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6359         struct net *net = sock_net(skb->sk);
6360         struct br_port_msg *bpm;
6361         struct net_device *dev;
6362         int err;
6363
6364         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6365                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6366         if (err)
6367                 return err;
6368
6369         bpm = nlmsg_data(nlh);
6370         if (!bpm->ifindex) {
6371                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6372                 return -EINVAL;
6373         }
6374
6375         dev = __dev_get_by_index(net, bpm->ifindex);
6376         if (!dev) {
6377                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6378                 return -ENODEV;
6379         }
6380
6381         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6382                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6383                 return -EINVAL;
6384         }
6385
6386         if (!dev->netdev_ops->ndo_mdb_add) {
6387                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6388                 return -EOPNOTSUPP;
6389         }
6390
6391         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6392 }
6393
6394 static int rtnl_validate_mdb_entry_del_bulk(const struct nlattr *attr,
6395                                             struct netlink_ext_ack *extack)
6396 {
6397         struct br_mdb_entry *entry = nla_data(attr);
6398         struct br_mdb_entry zero_entry = {};
6399
6400         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6401                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6402                 return -EINVAL;
6403         }
6404
6405         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6406                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6407                 return -EINVAL;
6408         }
6409
6410         if (entry->flags) {
6411                 NL_SET_ERR_MSG(extack, "Entry flags cannot be set");
6412                 return -EINVAL;
6413         }
6414
6415         if (entry->vid >= VLAN_N_VID - 1) {
6416                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6417                 return -EINVAL;
6418         }
6419
6420         if (memcmp(&entry->addr, &zero_entry.addr, sizeof(entry->addr))) {
6421                 NL_SET_ERR_MSG(extack, "Entry address cannot be set");
6422                 return -EINVAL;
6423         }
6424
6425         return 0;
6426 }
6427
6428 static const struct nla_policy mdba_del_bulk_policy[MDBA_SET_ENTRY_MAX + 1] = {
6429         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6430                                                   rtnl_validate_mdb_entry_del_bulk,
6431                                                   sizeof(struct br_mdb_entry)),
6432         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6433 };
6434
6435 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6436                         struct netlink_ext_ack *extack)
6437 {
6438         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
6439         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6440         struct net *net = sock_net(skb->sk);
6441         struct br_port_msg *bpm;
6442         struct net_device *dev;
6443         int err;
6444
6445         if (!del_bulk)
6446                 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6447                                              MDBA_SET_ENTRY_MAX, mdba_policy,
6448                                              extack);
6449         else
6450                 err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX,
6451                                   mdba_del_bulk_policy, extack);
6452         if (err)
6453                 return err;
6454
6455         bpm = nlmsg_data(nlh);
6456         if (!bpm->ifindex) {
6457                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6458                 return -EINVAL;
6459         }
6460
6461         dev = __dev_get_by_index(net, bpm->ifindex);
6462         if (!dev) {
6463                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6464                 return -ENODEV;
6465         }
6466
6467         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6468                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6469                 return -EINVAL;
6470         }
6471
6472         if (del_bulk) {
6473                 if (!dev->netdev_ops->ndo_mdb_del_bulk) {
6474                         NL_SET_ERR_MSG(extack, "Device does not support MDB bulk deletion");
6475                         return -EOPNOTSUPP;
6476                 }
6477                 return dev->netdev_ops->ndo_mdb_del_bulk(dev, tb, extack);
6478         }
6479
6480         if (!dev->netdev_ops->ndo_mdb_del) {
6481                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6482                 return -EOPNOTSUPP;
6483         }
6484
6485         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6486 }
6487
6488 /* Process one rtnetlink message. */
6489
6490 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6491                              struct netlink_ext_ack *extack)
6492 {
6493         struct net *net = sock_net(skb->sk);
6494         struct rtnl_link *link;
6495         enum rtnl_kinds kind;
6496         struct module *owner;
6497         int err = -EOPNOTSUPP;
6498         rtnl_doit_func doit;
6499         unsigned int flags;
6500         int family;
6501         int type;
6502
6503         type = nlh->nlmsg_type;
6504         if (type > RTM_MAX)
6505                 return -EOPNOTSUPP;
6506
6507         type -= RTM_BASE;
6508
6509         /* All the messages must have at least 1 byte length */
6510         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6511                 return 0;
6512
6513         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6514         kind = rtnl_msgtype_kind(type);
6515
6516         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6517                 return -EPERM;
6518
6519         rcu_read_lock();
6520         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6521                 struct sock *rtnl;
6522                 rtnl_dumpit_func dumpit;
6523                 u32 min_dump_alloc = 0;
6524
6525                 link = rtnl_get_link(family, type);
6526                 if (!link || !link->dumpit) {
6527                         family = PF_UNSPEC;
6528                         link = rtnl_get_link(family, type);
6529                         if (!link || !link->dumpit)
6530                                 goto err_unlock;
6531                 }
6532                 owner = link->owner;
6533                 dumpit = link->dumpit;
6534                 flags = link->flags;
6535
6536                 if (type == RTM_GETLINK - RTM_BASE)
6537                         min_dump_alloc = rtnl_calcit(skb, nlh);
6538
6539                 err = 0;
6540                 /* need to do this before rcu_read_unlock() */
6541                 if (!try_module_get(owner))
6542                         err = -EPROTONOSUPPORT;
6543
6544                 rcu_read_unlock();
6545
6546                 rtnl = net->rtnl;
6547                 if (err == 0) {
6548                         struct netlink_dump_control c = {
6549                                 .dump           = dumpit,
6550                                 .min_dump_alloc = min_dump_alloc,
6551                                 .module         = owner,
6552                                 .flags          = flags,
6553                         };
6554                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6555                         /* netlink_dump_start() will keep a reference on
6556                          * module if dump is still in progress.
6557                          */
6558                         module_put(owner);
6559                 }
6560                 return err;
6561         }
6562
6563         link = rtnl_get_link(family, type);
6564         if (!link || !link->doit) {
6565                 family = PF_UNSPEC;
6566                 link = rtnl_get_link(PF_UNSPEC, type);
6567                 if (!link || !link->doit)
6568                         goto out_unlock;
6569         }
6570
6571         owner = link->owner;
6572         if (!try_module_get(owner)) {
6573                 err = -EPROTONOSUPPORT;
6574                 goto out_unlock;
6575         }
6576
6577         flags = link->flags;
6578         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6579             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6580                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6581                 module_put(owner);
6582                 goto err_unlock;
6583         }
6584
6585         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6586                 doit = link->doit;
6587                 rcu_read_unlock();
6588                 if (doit)
6589                         err = doit(skb, nlh, extack);
6590                 module_put(owner);
6591                 return err;
6592         }
6593         rcu_read_unlock();
6594
6595         rtnl_lock();
6596         link = rtnl_get_link(family, type);
6597         if (link && link->doit)
6598                 err = link->doit(skb, nlh, extack);
6599         rtnl_unlock();
6600
6601         module_put(owner);
6602
6603         return err;
6604
6605 out_unlock:
6606         rcu_read_unlock();
6607         return err;
6608
6609 err_unlock:
6610         rcu_read_unlock();
6611         return -EOPNOTSUPP;
6612 }
6613
6614 static void rtnetlink_rcv(struct sk_buff *skb)
6615 {
6616         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6617 }
6618
6619 static int rtnetlink_bind(struct net *net, int group)
6620 {
6621         switch (group) {
6622         case RTNLGRP_IPV4_MROUTE_R:
6623         case RTNLGRP_IPV6_MROUTE_R:
6624                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6625                         return -EPERM;
6626                 break;
6627         }
6628         return 0;
6629 }
6630
6631 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6632 {
6633         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6634
6635         switch (event) {
6636         case NETDEV_REBOOT:
6637         case NETDEV_CHANGEMTU:
6638         case NETDEV_CHANGEADDR:
6639         case NETDEV_CHANGENAME:
6640         case NETDEV_FEAT_CHANGE:
6641         case NETDEV_BONDING_FAILOVER:
6642         case NETDEV_POST_TYPE_CHANGE:
6643         case NETDEV_NOTIFY_PEERS:
6644         case NETDEV_CHANGEUPPER:
6645         case NETDEV_RESEND_IGMP:
6646         case NETDEV_CHANGEINFODATA:
6647         case NETDEV_CHANGELOWERSTATE:
6648         case NETDEV_CHANGE_TX_QUEUE_LEN:
6649                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6650                                    GFP_KERNEL, NULL, 0, 0, NULL);
6651                 break;
6652         default:
6653                 break;
6654         }
6655         return NOTIFY_DONE;
6656 }
6657
6658 static struct notifier_block rtnetlink_dev_notifier = {
6659         .notifier_call  = rtnetlink_event,
6660 };
6661
6662
6663 static int __net_init rtnetlink_net_init(struct net *net)
6664 {
6665         struct sock *sk;
6666         struct netlink_kernel_cfg cfg = {
6667                 .groups         = RTNLGRP_MAX,
6668                 .input          = rtnetlink_rcv,
6669                 .cb_mutex       = &rtnl_mutex,
6670                 .flags          = NL_CFG_F_NONROOT_RECV,
6671                 .bind           = rtnetlink_bind,
6672         };
6673
6674         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6675         if (!sk)
6676                 return -ENOMEM;
6677         net->rtnl = sk;
6678         return 0;
6679 }
6680
6681 static void __net_exit rtnetlink_net_exit(struct net *net)
6682 {
6683         netlink_kernel_release(net->rtnl);
6684         net->rtnl = NULL;
6685 }
6686
6687 static struct pernet_operations rtnetlink_net_ops = {
6688         .init = rtnetlink_net_init,
6689         .exit = rtnetlink_net_exit,
6690 };
6691
6692 void __init rtnetlink_init(void)
6693 {
6694         if (register_pernet_subsys(&rtnetlink_net_ops))
6695                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6696
6697         register_netdevice_notifier(&rtnetlink_dev_notifier);
6698
6699         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6700                       rtnl_dump_ifinfo, 0);
6701         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6702         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6703         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6704
6705         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6706         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6707         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6708
6709         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6710         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6711
6712         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6713         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6714                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6715         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6716
6717         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6718         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6719         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6720
6721         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6722                       0);
6723         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6724
6725         rtnl_register(PF_BRIDGE, RTM_GETMDB, rtnl_mdb_get, rtnl_mdb_dump, 0);
6726         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6727         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL,
6728                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6729 }