core/rtnetlink.c

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