net/core/rtnetlink.c

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