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