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 = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
 346         rtnl_unlock();
 347
 348         kfree_rcu(link, rcu);
 349
 350         return 0;
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_unregister);
 353
 354 /**
 355  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 356  * @protocol : Protocol family or PF_UNSPEC
 357  *
 358  * Identical to calling rtnl_unregster() for all registered message types
 359  * of a certain protocol family.
 360  */
 361 void rtnl_unregister_all(int protocol)
 362 {
 363         struct rtnl_link __rcu **tab;
 364         struct rtnl_link *link;
 365         int msgindex;
 366
 367         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 368
 369         rtnl_lock();
 370         tab = rcu_replace_pointer_rtnl(rtnl_msg_handlers[protocol], NULL);
 371         if (!tab) {
 372                 rtnl_unlock();
 373                 return;
 374         }
 375         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 376                 link = rcu_replace_pointer_rtnl(tab[msgindex], NULL);
 377                 kfree_rcu(link, rcu);
 378         }
 379         rtnl_unlock();
 380
 381         synchronize_net();
 382
 383         kfree(tab);
 384 }
 385 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 386
 387 static LIST_HEAD(link_ops);
 388
 389 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 390 {
 391         const struct rtnl_link_ops *ops;
 392
 393         list_for_each_entry(ops, &link_ops, list) {
 394                 if (!strcmp(ops->kind, kind))
 395                         return ops;
 396         }
 397         return NULL;
 398 }
 399
 400 /**
 401  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 402  * @ops: struct rtnl_link_ops * to register
 403  *
 404  * The caller must hold the rtnl_mutex. This function should be used
 405  * by drivers that create devices during module initialization. It
 406  * must be called before registering the devices.
 407  *
 408  * Returns 0 on success or a negative error code.
 409  */
 410 int __rtnl_link_register(struct rtnl_link_ops *ops)
 411 {
 412         if (rtnl_link_ops_get(ops->kind))
 413                 return -EEXIST;
 414
 415         /* The check for alloc/setup is here because if ops
 416          * does not have that filled up, it is not possible
 417          * to use the ops for creating device. So do not
 418          * fill up dellink as well. That disables rtnl_dellink.
 419          */
 420         if ((ops->alloc || ops->setup) && !ops->dellink)
 421                 ops->dellink = unregister_netdevice_queue;
 422
 423         list_add_tail(&ops->list, &link_ops);
 424         return 0;
 425 }
 426 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 427
 428 /**
 429  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 430  * @ops: struct rtnl_link_ops * to register
 431  *
 432  * Returns 0 on success or a negative error code.
 433  */
 434 int rtnl_link_register(struct rtnl_link_ops *ops)
 435 {
 436         int err;
 437
 438         /* Sanity-check max sizes to avoid stack buffer overflow. */
 439         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 440                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 441                 return -EINVAL;
 442
 443         rtnl_lock();
 444         err = __rtnl_link_register(ops);
 445         rtnl_unlock();
 446         return err;
 447 }
 448 EXPORT_SYMBOL_GPL(rtnl_link_register);
 449
 450 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 451 {
 452         struct net_device *dev;
 453         LIST_HEAD(list_kill);
 454
 455         for_each_netdev(net, dev) {
 456                 if (dev->rtnl_link_ops == ops)
 457                         ops->dellink(dev, &list_kill);
 458         }
 459         unregister_netdevice_many(&list_kill);
 460 }
 461
 462 /**
 463  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 464  * @ops: struct rtnl_link_ops * to unregister
 465  *
 466  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 467  * integrity (hold pernet_ops_rwsem for writing to close the race
 468  * with setup_net() and cleanup_net()).
 469  */
 470 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 471 {
 472         struct net *net;
 473
 474         for_each_net(net) {
 475                 __rtnl_kill_links(net, ops);
 476         }
 477         list_del(&ops->list);
 478 }
 479 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 480
 481 /* Return with the rtnl_lock held when there are no network
 482  * devices unregistering in any network namespace.
 483  */
 484 static void rtnl_lock_unregistering_all(void)
 485 {
 486         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 487
 488         add_wait_queue(&netdev_unregistering_wq, &wait);
 489         for (;;) {
 490                 rtnl_lock();
 491                 /* We held write locked pernet_ops_rwsem, and parallel
 492                  * setup_net() and cleanup_net() are not possible.
 493                  */
 494                 if (!atomic_read(&dev_unreg_count))
 495                         break;
 496                 __rtnl_unlock();
 497
 498                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 499         }
 500         remove_wait_queue(&netdev_unregistering_wq, &wait);
 501 }
 502
 503 /**
 504  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 505  * @ops: struct rtnl_link_ops * to unregister
 506  */
 507 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 508 {
 509         /* Close the race with setup_net() and cleanup_net() */
 510         down_write(&pernet_ops_rwsem);
 511         rtnl_lock_unregistering_all();
 512         __rtnl_link_unregister(ops);
 513         rtnl_unlock();
 514         up_write(&pernet_ops_rwsem);
 515 }
 516 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 517
 518 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 519 {
 520         struct net_device *master_dev;
 521         const struct rtnl_link_ops *ops;
 522         size_t size = 0;
 523
 524         rcu_read_lock();
 525
 526         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 527         if (!master_dev)
 528                 goto out;
 529
 530         ops = master_dev->rtnl_link_ops;
 531         if (!ops || !ops->get_slave_size)
 532                 goto out;
 533         /* IFLA_INFO_SLAVE_DATA + nested data */
 534         size = nla_total_size(sizeof(struct nlattr)) +
 535                ops->get_slave_size(master_dev, dev);
 536
 537 out:
 538         rcu_read_unlock();
 539         return size;
 540 }
 541
 542 static size_t rtnl_link_get_size(const struct net_device *dev)
 543 {
 544         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 545         size_t size;
 546
 547         if (!ops)
 548                 return 0;
 549
 550         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 551                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 552
 553         if (ops->get_size)
 554                 /* IFLA_INFO_DATA + nested data */
 555                 size += nla_total_size(sizeof(struct nlattr)) +
 556                         ops->get_size(dev);
 557
 558         if (ops->get_xstats_size)
 559                 /* IFLA_INFO_XSTATS */
 560                 size += nla_total_size(ops->get_xstats_size(dev));
 561
 562         size += rtnl_link_get_slave_info_data_size(dev);
 563
 564         return size;
 565 }
 566
 567 static LIST_HEAD(rtnl_af_ops);
 568
 569 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 570 {
 571         const struct rtnl_af_ops *ops;
 572
 573         ASSERT_RTNL();
 574
 575         list_for_each_entry(ops, &rtnl_af_ops, list) {
 576                 if (ops->family == family)
 577                         return ops;
 578         }
 579
 580         return NULL;
 581 }
 582
 583 /**
 584  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 585  * @ops: struct rtnl_af_ops * to register
 586  *
 587  * Returns 0 on success or a negative error code.
 588  */
 589 void rtnl_af_register(struct rtnl_af_ops *ops)
 590 {
 591         rtnl_lock();
 592         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 593         rtnl_unlock();
 594 }
 595 EXPORT_SYMBOL_GPL(rtnl_af_register);
 596
 597 /**
 598  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 599  * @ops: struct rtnl_af_ops * to unregister
 600  */
 601 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 602 {
 603         rtnl_lock();
 604         list_del_rcu(&ops->list);
 605         rtnl_unlock();
 606
 607         synchronize_rcu();
 608 }
 609 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 610
 611 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 612                                     u32 ext_filter_mask)
 613 {
 614         struct rtnl_af_ops *af_ops;
 615         size_t size;
 616
 617         /* IFLA_AF_SPEC */
 618         size = nla_total_size(sizeof(struct nlattr));
 619
 620         rcu_read_lock();
 621         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 622                 if (af_ops->get_link_af_size) {
 623                         /* AF_* + nested data */
 624                         size += nla_total_size(sizeof(struct nlattr)) +
 625                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 626                 }
 627         }
 628         rcu_read_unlock();
 629
 630         return size;
 631 }
 632
 633 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 634 {
 635         struct net_device *master_dev;
 636         bool ret = false;
 637
 638         rcu_read_lock();
 639
 640         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 641         if (master_dev && master_dev->rtnl_link_ops)
 642                 ret = true;
 643         rcu_read_unlock();
 644         return ret;
 645 }
 646
 647 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 648                                      const struct net_device *dev)
 649 {
 650         struct net_device *master_dev;
 651         const struct rtnl_link_ops *ops;
 652         struct nlattr *slave_data;
 653         int err;
 654
 655         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 656         if (!master_dev)
 657                 return 0;
 658         ops = master_dev->rtnl_link_ops;
 659         if (!ops)
 660                 return 0;
 661         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 662                 return -EMSGSIZE;
 663         if (ops->fill_slave_info) {
 664                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 665                 if (!slave_data)
 666                         return -EMSGSIZE;
 667                 err = ops->fill_slave_info(skb, master_dev, dev);
 668                 if (err < 0)
 669                         goto err_cancel_slave_data;
 670                 nla_nest_end(skb, slave_data);
 671         }
 672         return 0;
 673
 674 err_cancel_slave_data:
 675         nla_nest_cancel(skb, slave_data);
 676         return err;
 677 }
 678
 679 static int rtnl_link_info_fill(struct sk_buff *skb,
 680                                const struct net_device *dev)
 681 {
 682         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 683         struct nlattr *data;
 684         int err;
 685
 686         if (!ops)
 687                 return 0;
 688         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 689                 return -EMSGSIZE;
 690         if (ops->fill_xstats) {
 691                 err = ops->fill_xstats(skb, dev);
 692                 if (err < 0)
 693                         return err;
 694         }
 695         if (ops->fill_info) {
 696                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 697                 if (data == NULL)
 698                         return -EMSGSIZE;
 699                 err = ops->fill_info(skb, dev);
 700                 if (err < 0)
 701                         goto err_cancel_data;
 702                 nla_nest_end(skb, data);
 703         }
 704         return 0;
 705
 706 err_cancel_data:
 707         nla_nest_cancel(skb, data);
 708         return err;
 709 }
 710
 711 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 712 {
 713         struct nlattr *linkinfo;
 714         int err = -EMSGSIZE;
 715
 716         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 717         if (linkinfo == NULL)
 718                 goto out;
 719
 720         err = rtnl_link_info_fill(skb, dev);
 721         if (err < 0)
 722                 goto err_cancel_link;
 723
 724         err = rtnl_link_slave_info_fill(skb, dev);
 725         if (err < 0)
 726                 goto err_cancel_link;
 727
 728         nla_nest_end(skb, linkinfo);
 729         return 0;
 730
 731 err_cancel_link:
 732         nla_nest_cancel(skb, linkinfo);
 733 out:
 734         return err;
 735 }
 736
 737 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 738 {
 739         struct sock *rtnl = net->rtnl;
 740
 741         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 742 }
 743
 744 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 745 {
 746         struct sock *rtnl = net->rtnl;
 747
 748         return nlmsg_unicast(rtnl, skb, pid);
 749 }
 750 EXPORT_SYMBOL(rtnl_unicast);
 751
 752 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 753                  const struct nlmsghdr *nlh, gfp_t flags)
 754 {
 755         struct sock *rtnl = net->rtnl;
 756
 757         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 758 }
 759 EXPORT_SYMBOL(rtnl_notify);
 760
 761 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 762 {
 763         struct sock *rtnl = net->rtnl;
 764
 765         netlink_set_err(rtnl, 0, group, error);
 766 }
 767 EXPORT_SYMBOL(rtnl_set_sk_err);
 768
 769 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 770 {
 771         struct nlattr *mx;
 772         int i, valid = 0;
 773
 774         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 775         if (metrics == dst_default_metrics.metrics)
 776                 return 0;
 777
 778         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 779         if (mx == NULL)
 780                 return -ENOBUFS;
 781
 782         for (i = 0; i < RTAX_MAX; i++) {
 783                 if (metrics[i]) {
 784                         if (i == RTAX_CC_ALGO - 1) {
 785                                 char tmp[TCP_CA_NAME_MAX], *name;
 786
 787                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 788                                 if (!name)
 789                                         continue;
 790                                 if (nla_put_string(skb, i + 1, name))
 791                                         goto nla_put_failure;
 792                         } else if (i == RTAX_FEATURES - 1) {
 793                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 794
 795                                 if (!user_features)
 796                                         continue;
 797                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 798                                 if (nla_put_u32(skb, i + 1, user_features))
 799                                         goto nla_put_failure;
 800                         } else {
 801                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 802                                         goto nla_put_failure;
 803                         }
 804                         valid++;
 805                 }
 806         }
 807
 808         if (!valid) {
 809                 nla_nest_cancel(skb, mx);
 810                 return 0;
 811         }
 812
 813         return nla_nest_end(skb, mx);
 814
 815 nla_put_failure:
 816         nla_nest_cancel(skb, mx);
 817         return -EMSGSIZE;
 818 }
 819 EXPORT_SYMBOL(rtnetlink_put_metrics);
 820
 821 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 822                        long expires, u32 error)
 823 {
 824         struct rta_cacheinfo ci = {
 825                 .rta_error = error,
 826                 .rta_id =  id,
 827         };
 828
 829         if (dst) {
 830                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 831                 ci.rta_used = dst->__use;
 832                 ci.rta_clntref = rcuref_read(&dst->__rcuref);
 833         }
 834         if (expires) {
 835                 unsigned long clock;
 836
 837                 clock = jiffies_to_clock_t(abs(expires));
 838                 clock = min_t(unsigned long, clock, INT_MAX);
 839                 ci.rta_expires = (expires > 0) ? clock : -clock;
 840         }
 841         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 842 }
 843 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 844
 845 void netdev_set_operstate(struct net_device *dev, int newstate)
 846 {
 847         unsigned int old = READ_ONCE(dev->operstate);
 848
 849         do {
 850                 if (old == newstate)
 851                         return;
 852         } while (!try_cmpxchg(&dev->operstate, &old, newstate));
 853
 854         netdev_state_change(dev);
 855 }
 856 EXPORT_SYMBOL(netdev_set_operstate);
 857
 858 static void set_operstate(struct net_device *dev, unsigned char transition)
 859 {
 860         unsigned char operstate = READ_ONCE(dev->operstate);
 861
 862         switch (transition) {
 863         case IF_OPER_UP:
 864                 if ((operstate == IF_OPER_DORMANT ||
 865                      operstate == IF_OPER_TESTING ||
 866                      operstate == IF_OPER_UNKNOWN) &&
 867                     !netif_dormant(dev) && !netif_testing(dev))
 868                         operstate = IF_OPER_UP;
 869                 break;
 870
 871         case IF_OPER_TESTING:
 872                 if (netif_oper_up(dev))
 873                         operstate = IF_OPER_TESTING;
 874                 break;
 875
 876         case IF_OPER_DORMANT:
 877                 if (netif_oper_up(dev))
 878                         operstate = IF_OPER_DORMANT;
 879                 break;
 880         }
 881
 882         netdev_set_operstate(dev, operstate);
 883 }
 884
 885 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 886 {
 887         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 888                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 889 }
 890
 891 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 892                                            const struct ifinfomsg *ifm)
 893 {
 894         unsigned int flags = ifm->ifi_flags;
 895
 896         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 897         if (ifm->ifi_change)
 898                 flags = (flags & ifm->ifi_change) |
 899                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 900
 901         return flags;
 902 }
 903
 904 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 905                                  const struct rtnl_link_stats64 *b)
 906 {
 907         a->rx_packets = b->rx_packets;
 908         a->tx_packets = b->tx_packets;
 909         a->rx_bytes = b->rx_bytes;
 910         a->tx_bytes = b->tx_bytes;
 911         a->rx_errors = b->rx_errors;
 912         a->tx_errors = b->tx_errors;
 913         a->rx_dropped = b->rx_dropped;
 914         a->tx_dropped = b->tx_dropped;
 915
 916         a->multicast = b->multicast;
 917         a->collisions = b->collisions;
 918
 919         a->rx_length_errors = b->rx_length_errors;
 920         a->rx_over_errors = b->rx_over_errors;
 921         a->rx_crc_errors = b->rx_crc_errors;
 922         a->rx_frame_errors = b->rx_frame_errors;
 923         a->rx_fifo_errors = b->rx_fifo_errors;
 924         a->rx_missed_errors = b->rx_missed_errors;
 925
 926         a->tx_aborted_errors = b->tx_aborted_errors;
 927         a->tx_carrier_errors = b->tx_carrier_errors;
 928         a->tx_fifo_errors = b->tx_fifo_errors;
 929         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 930         a->tx_window_errors = b->tx_window_errors;
 931
 932         a->rx_compressed = b->rx_compressed;
 933         a->tx_compressed = b->tx_compressed;
 934
 935         a->rx_nohandler = b->rx_nohandler;
 936 }
 937
 938 /* All VF info */
 939 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 940                                    u32 ext_filter_mask)
 941 {
 942         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 943                 int num_vfs = dev_num_vf(dev->dev.parent);
 944                 size_t size = nla_total_size(0);
 945                 size += num_vfs *
 946                         (nla_total_size(0) +
 947                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 948                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 949                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 950                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 951                          nla_total_size(MAX_VLAN_LIST_LEN *
 952                                         sizeof(struct ifla_vf_vlan_info)) +
 953                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 954                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 955                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 956                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 957                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 958                          nla_total_size(sizeof(struct ifla_vf_trust)));
 959                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 960                         size += num_vfs *
 961                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 962                                  /* IFLA_VF_STATS_RX_PACKETS */
 963                                  nla_total_size_64bit(sizeof(__u64)) +
 964                                  /* IFLA_VF_STATS_TX_PACKETS */
 965                                  nla_total_size_64bit(sizeof(__u64)) +
 966                                  /* IFLA_VF_STATS_RX_BYTES */
 967                                  nla_total_size_64bit(sizeof(__u64)) +
 968                                  /* IFLA_VF_STATS_TX_BYTES */
 969                                  nla_total_size_64bit(sizeof(__u64)) +
 970                                  /* IFLA_VF_STATS_BROADCAST */
 971                                  nla_total_size_64bit(sizeof(__u64)) +
 972                                  /* IFLA_VF_STATS_MULTICAST */
 973                                  nla_total_size_64bit(sizeof(__u64)) +
 974                                  /* IFLA_VF_STATS_RX_DROPPED */
 975                                  nla_total_size_64bit(sizeof(__u64)) +
 976                                  /* IFLA_VF_STATS_TX_DROPPED */
 977                                  nla_total_size_64bit(sizeof(__u64)));
 978                 }
 979                 return size;
 980         } else
 981                 return 0;
 982 }
 983
 984 static size_t rtnl_port_size(const struct net_device *dev,
 985                              u32 ext_filter_mask)
 986 {
 987         size_t port_size = nla_total_size(4)            /* PORT_VF */
 988                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 989                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 990                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 991                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 992                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 993         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 994         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 995                 + port_size;
 996         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 997                 + port_size;
 998
 999         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1000             !(ext_filter_mask & RTEXT_FILTER_VF))
1001                 return 0;
1002         if (dev_num_vf(dev->dev.parent))
1003                 return port_self_size + vf_ports_size +
1004                         vf_port_size * dev_num_vf(dev->dev.parent);
1005         else
1006                 return port_self_size;
1007 }
1008
1009 static size_t rtnl_xdp_size(void)
1010 {
1011         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1012                           nla_total_size(1) +   /* XDP_ATTACHED */
1013                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1014                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1015
1016         return xdp_size;
1017 }
1018
1019 static size_t rtnl_prop_list_size(const struct net_device *dev)
1020 {
1021         struct netdev_name_node *name_node;
1022         unsigned int cnt = 0;
1023
1024         rcu_read_lock();
1025         list_for_each_entry_rcu(name_node, &dev->name_node->list, list)
1026                 cnt++;
1027         rcu_read_unlock();
1028
1029         if (!cnt)
1030                 return 0;
1031
1032         return nla_total_size(0) + cnt * nla_total_size(ALTIFNAMSIZ);
1033 }
1034
1035 static size_t rtnl_proto_down_size(const struct net_device *dev)
1036 {
1037         size_t size = nla_total_size(1);
1038
1039         /* Assume dev->proto_down_reason is not zero. */
1040         size += nla_total_size(0) + nla_total_size(4);
1041
1042         return size;
1043 }
1044
1045 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1046 {
1047         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1048
1049         if (dev->devlink_port)
1050                 size += devlink_nl_port_handle_size(dev->devlink_port);
1051
1052         return size;
1053 }
1054
1055 static size_t rtnl_dpll_pin_size(const struct net_device *dev)
1056 {
1057         size_t size = nla_total_size(0); /* nest IFLA_DPLL_PIN */
1058
1059         size += dpll_netdev_pin_handle_size(dev);
1060
1061         return size;
1062 }
1063
1064 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1065                                      u32 ext_filter_mask)
1066 {
1067         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1068                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1069                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1070                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1071                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1072                + nla_total_size(sizeof(struct rtnl_link_stats))
1073                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1074                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1075                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1076                + nla_total_size(4) /* IFLA_TXQLEN */
1077                + nla_total_size(4) /* IFLA_WEIGHT */
1078                + nla_total_size(4) /* IFLA_MTU */
1079                + nla_total_size(4) /* IFLA_LINK */
1080                + nla_total_size(4) /* IFLA_MASTER */
1081                + nla_total_size(1) /* IFLA_CARRIER */
1082                + nla_total_size(4) /* IFLA_PROMISCUITY */
1083                + nla_total_size(4) /* IFLA_ALLMULTI */
1084                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1085                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1086                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1087                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1088                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1089                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1090                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1091                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1092                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1093                + nla_total_size(1) /* IFLA_OPERSTATE */
1094                + nla_total_size(1) /* IFLA_LINKMODE */
1095                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1096                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1097                + nla_total_size(4) /* IFLA_GROUP */
1098                + nla_total_size(ext_filter_mask
1099                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1100                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1101                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1102                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1103                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1104                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1105                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1106                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1107                + rtnl_xdp_size() /* IFLA_XDP */
1108                + nla_total_size(4)  /* IFLA_EVENT */
1109                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1110                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1111                + rtnl_proto_down_size(dev)  /* proto down */
1112                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1113                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1114                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1115                + nla_total_size(4)  /* IFLA_MIN_MTU */
1116                + nla_total_size(4)  /* IFLA_MAX_MTU */
1117                + rtnl_prop_list_size(dev)
1118                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1119                + rtnl_devlink_port_size(dev)
1120                + rtnl_dpll_pin_size(dev)
1121                + 0;
1122 }
1123
1124 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1125 {
1126         struct nlattr *vf_ports;
1127         struct nlattr *vf_port;
1128         int vf;
1129         int err;
1130
1131         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1132         if (!vf_ports)
1133                 return -EMSGSIZE;
1134
1135         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1136                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1137                 if (!vf_port)
1138                         goto nla_put_failure;
1139                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1140                         goto nla_put_failure;
1141                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1142                 if (err == -EMSGSIZE)
1143                         goto nla_put_failure;
1144                 if (err) {
1145                         nla_nest_cancel(skb, vf_port);
1146                         continue;
1147                 }
1148                 nla_nest_end(skb, vf_port);
1149         }
1150
1151         nla_nest_end(skb, vf_ports);
1152
1153         return 0;
1154
1155 nla_put_failure:
1156         nla_nest_cancel(skb, vf_ports);
1157         return -EMSGSIZE;
1158 }
1159
1160 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1161 {
1162         struct nlattr *port_self;
1163         int err;
1164
1165         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1166         if (!port_self)
1167                 return -EMSGSIZE;
1168
1169         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1170         if (err) {
1171                 nla_nest_cancel(skb, port_self);
1172                 return (err == -EMSGSIZE) ? err : 0;
1173         }
1174
1175         nla_nest_end(skb, port_self);
1176
1177         return 0;
1178 }
1179
1180 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1181                           u32 ext_filter_mask)
1182 {
1183         int err;
1184
1185         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1186             !(ext_filter_mask & RTEXT_FILTER_VF))
1187                 return 0;
1188
1189         err = rtnl_port_self_fill(skb, dev);
1190         if (err)
1191                 return err;
1192
1193         if (dev_num_vf(dev->dev.parent)) {
1194                 err = rtnl_vf_ports_fill(skb, dev);
1195                 if (err)
1196                         return err;
1197         }
1198
1199         return 0;
1200 }
1201
1202 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1203 {
1204         int err;
1205         struct netdev_phys_item_id ppid;
1206
1207         err = dev_get_phys_port_id(dev, &ppid);
1208         if (err) {
1209                 if (err == -EOPNOTSUPP)
1210                         return 0;
1211                 return err;
1212         }
1213
1214         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1215                 return -EMSGSIZE;
1216
1217         return 0;
1218 }
1219
1220 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1221 {
1222         char name[IFNAMSIZ];
1223         int err;
1224
1225         err = dev_get_phys_port_name(dev, name, sizeof(name));
1226         if (err) {
1227                 if (err == -EOPNOTSUPP)
1228                         return 0;
1229                 return err;
1230         }
1231
1232         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1233                 return -EMSGSIZE;
1234
1235         return 0;
1236 }
1237
1238 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1239 {
1240         struct netdev_phys_item_id ppid = { };
1241         int err;
1242
1243         err = dev_get_port_parent_id(dev, &ppid, false);
1244         if (err) {
1245                 if (err == -EOPNOTSUPP)
1246                         return 0;
1247                 return err;
1248         }
1249
1250         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1251                 return -EMSGSIZE;
1252
1253         return 0;
1254 }
1255
1256 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1257                                               struct net_device *dev)
1258 {
1259         struct rtnl_link_stats64 *sp;
1260         struct nlattr *attr;
1261
1262         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1263                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1264         if (!attr)
1265                 return -EMSGSIZE;
1266
1267         sp = nla_data(attr);
1268         dev_get_stats(dev, sp);
1269
1270         attr = nla_reserve(skb, IFLA_STATS,
1271                            sizeof(struct rtnl_link_stats));
1272         if (!attr)
1273                 return -EMSGSIZE;
1274
1275         copy_rtnl_link_stats(nla_data(attr), sp);
1276
1277         return 0;
1278 }
1279
1280 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1281                                                struct net_device *dev,
1282                                                int vfs_num,
1283                                                u32 ext_filter_mask)
1284 {
1285         struct ifla_vf_rss_query_en vf_rss_query_en;
1286         struct nlattr *vf, *vfstats, *vfvlanlist;
1287         struct ifla_vf_link_state vf_linkstate;
1288         struct ifla_vf_vlan_info vf_vlan_info;
1289         struct ifla_vf_spoofchk vf_spoofchk;
1290         struct ifla_vf_tx_rate vf_tx_rate;
1291         struct ifla_vf_stats vf_stats;
1292         struct ifla_vf_trust vf_trust;
1293         struct ifla_vf_vlan vf_vlan;
1294         struct ifla_vf_rate vf_rate;
1295         struct ifla_vf_mac vf_mac;
1296         struct ifla_vf_broadcast vf_broadcast;
1297         struct ifla_vf_info ivi;
1298         struct ifla_vf_guid node_guid;
1299         struct ifla_vf_guid port_guid;
1300
1301         memset(&ivi, 0, sizeof(ivi));
1302
1303         /* Not all SR-IOV capable drivers support the
1304          * spoofcheck and "RSS query enable" query.  Preset to
1305          * -1 so the user space tool can detect that the driver
1306          * didn't report anything.
1307          */
1308         ivi.spoofchk = -1;
1309         ivi.rss_query_en = -1;
1310         ivi.trusted = -1;
1311         /* The default value for VF link state is "auto"
1312          * IFLA_VF_LINK_STATE_AUTO which equals zero
1313          */
1314         ivi.linkstate = 0;
1315         /* VLAN Protocol by default is 802.1Q */
1316         ivi.vlan_proto = htons(ETH_P_8021Q);
1317         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1318                 return 0;
1319
1320         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1321         memset(&node_guid, 0, sizeof(node_guid));
1322         memset(&port_guid, 0, sizeof(port_guid));
1323
1324         vf_mac.vf =
1325                 vf_vlan.vf =
1326                 vf_vlan_info.vf =
1327                 vf_rate.vf =
1328                 vf_tx_rate.vf =
1329                 vf_spoofchk.vf =
1330                 vf_linkstate.vf =
1331                 vf_rss_query_en.vf =
1332                 vf_trust.vf =
1333                 node_guid.vf =
1334                 port_guid.vf = ivi.vf;
1335
1336         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1337         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1338         vf_vlan.vlan = ivi.vlan;
1339         vf_vlan.qos = ivi.qos;
1340         vf_vlan_info.vlan = ivi.vlan;
1341         vf_vlan_info.qos = ivi.qos;
1342         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1343         vf_tx_rate.rate = ivi.max_tx_rate;
1344         vf_rate.min_tx_rate = ivi.min_tx_rate;
1345         vf_rate.max_tx_rate = ivi.max_tx_rate;
1346         vf_spoofchk.setting = ivi.spoofchk;
1347         vf_linkstate.link_state = ivi.linkstate;
1348         vf_rss_query_en.setting = ivi.rss_query_en;
1349         vf_trust.setting = ivi.trusted;
1350         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1351         if (!vf)
1352                 return -EMSGSIZE;
1353         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1354             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1355             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1356             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1357                     &vf_rate) ||
1358             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1359                     &vf_tx_rate) ||
1360             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1361                     &vf_spoofchk) ||
1362             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1363                     &vf_linkstate) ||
1364             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1365                     sizeof(vf_rss_query_en),
1366                     &vf_rss_query_en) ||
1367             nla_put(skb, IFLA_VF_TRUST,
1368                     sizeof(vf_trust), &vf_trust))
1369                 goto nla_put_vf_failure;
1370
1371         if (dev->netdev_ops->ndo_get_vf_guid &&
1372             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1373                                               &port_guid)) {
1374                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1375                             &node_guid) ||
1376                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1377                             &port_guid))
1378                         goto nla_put_vf_failure;
1379         }
1380         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1381         if (!vfvlanlist)
1382                 goto nla_put_vf_failure;
1383         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1384                     &vf_vlan_info)) {
1385                 nla_nest_cancel(skb, vfvlanlist);
1386                 goto nla_put_vf_failure;
1387         }
1388         nla_nest_end(skb, vfvlanlist);
1389         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1390                 memset(&vf_stats, 0, sizeof(vf_stats));
1391                 if (dev->netdev_ops->ndo_get_vf_stats)
1392                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1393                                                           &vf_stats);
1394                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1395                 if (!vfstats)
1396                         goto nla_put_vf_failure;
1397                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1398                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1399                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1400                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1401                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1402                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1403                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1404                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1405                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1406                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1407                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1408                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1409                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1410                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1411                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1412                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1413                         nla_nest_cancel(skb, vfstats);
1414                         goto nla_put_vf_failure;
1415                 }
1416                 nla_nest_end(skb, vfstats);
1417         }
1418         nla_nest_end(skb, vf);
1419         return 0;
1420
1421 nla_put_vf_failure:
1422         nla_nest_cancel(skb, vf);
1423         return -EMSGSIZE;
1424 }
1425
1426 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1427                                            struct net_device *dev,
1428                                            u32 ext_filter_mask)
1429 {
1430         struct nlattr *vfinfo;
1431         int i, num_vfs;
1432
1433         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1434                 return 0;
1435
1436         num_vfs = dev_num_vf(dev->dev.parent);
1437         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1438                 return -EMSGSIZE;
1439
1440         if (!dev->netdev_ops->ndo_get_vf_config)
1441                 return 0;
1442
1443         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1444         if (!vfinfo)
1445                 return -EMSGSIZE;
1446
1447         for (i = 0; i < num_vfs; i++) {
1448                 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1449                         nla_nest_cancel(skb, vfinfo);
1450                         return -EMSGSIZE;
1451                 }
1452         }
1453
1454         nla_nest_end(skb, vfinfo);
1455         return 0;
1456 }
1457
1458 static int rtnl_fill_link_ifmap(struct sk_buff *skb,
1459                                 const struct net_device *dev)
1460 {
1461         struct rtnl_link_ifmap map;
1462
1463         memset(&map, 0, sizeof(map));
1464         map.mem_start = READ_ONCE(dev->mem_start);
1465         map.mem_end   = READ_ONCE(dev->mem_end);
1466         map.base_addr = READ_ONCE(dev->base_addr);
1467         map.irq       = READ_ONCE(dev->irq);
1468         map.dma       = READ_ONCE(dev->dma);
1469         map.port      = READ_ONCE(dev->if_port);
1470
1471         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1472                 return -EMSGSIZE;
1473
1474         return 0;
1475 }
1476
1477 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1478 {
1479         const struct bpf_prog *generic_xdp_prog;
1480         u32 res = 0;
1481
1482         rcu_read_lock();
1483         generic_xdp_prog = rcu_dereference(dev->xdp_prog);
1484         if (generic_xdp_prog)
1485                 res = generic_xdp_prog->aux->id;
1486         rcu_read_unlock();
1487
1488         return res;
1489 }
1490
1491 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1492 {
1493         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1494 }
1495
1496 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1497 {
1498         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1499 }
1500
1501 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1502                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1503                                u32 (*get_prog_id)(struct net_device *dev))
1504 {
1505         u32 curr_id;
1506         int err;
1507
1508         curr_id = get_prog_id(dev);
1509         if (!curr_id)
1510                 return 0;
1511
1512         *prog_id = curr_id;
1513         err = nla_put_u32(skb, attr, curr_id);
1514         if (err)
1515                 return err;
1516
1517         if (*mode != XDP_ATTACHED_NONE)
1518                 *mode = XDP_ATTACHED_MULTI;
1519         else
1520                 *mode = tgt_mode;
1521
1522         return 0;
1523 }
1524
1525 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1526 {
1527         struct nlattr *xdp;
1528         u32 prog_id;
1529         int err;
1530         u8 mode;
1531
1532         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1533         if (!xdp)
1534                 return -EMSGSIZE;
1535
1536         prog_id = 0;
1537         mode = XDP_ATTACHED_NONE;
1538         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1539                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1540         if (err)
1541                 goto err_cancel;
1542         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1543                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1544         if (err)
1545                 goto err_cancel;
1546         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1547                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1548         if (err)
1549                 goto err_cancel;
1550
1551         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1552         if (err)
1553                 goto err_cancel;
1554
1555         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1556                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1557                 if (err)
1558                         goto err_cancel;
1559         }
1560
1561         nla_nest_end(skb, xdp);
1562         return 0;
1563
1564 err_cancel:
1565         nla_nest_cancel(skb, xdp);
1566         return err;
1567 }
1568
1569 static u32 rtnl_get_event(unsigned long event)
1570 {
1571         u32 rtnl_event_type = IFLA_EVENT_NONE;
1572
1573         switch (event) {
1574         case NETDEV_REBOOT:
1575                 rtnl_event_type = IFLA_EVENT_REBOOT;
1576                 break;
1577         case NETDEV_FEAT_CHANGE:
1578                 rtnl_event_type = IFLA_EVENT_FEATURES;
1579                 break;
1580         case NETDEV_BONDING_FAILOVER:
1581                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1582                 break;
1583         case NETDEV_NOTIFY_PEERS:
1584                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1585                 break;
1586         case NETDEV_RESEND_IGMP:
1587                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1588                 break;
1589         case NETDEV_CHANGEINFODATA:
1590                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1591                 break;
1592         default:
1593                 break;
1594         }
1595
1596         return rtnl_event_type;
1597 }
1598
1599 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1600 {
1601         const struct net_device *upper_dev;
1602         int ret = 0;
1603
1604         rcu_read_lock();
1605
1606         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1607         if (upper_dev)
1608                 ret = nla_put_u32(skb, IFLA_MASTER,
1609                                   READ_ONCE(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 iflink = dev_get_iflink(dev);
1619
1620         if (force || READ_ONCE(dev->ifindex) != iflink)
1621                 return nla_put_u32(skb, IFLA_LINK, iflink);
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_rcu(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 /* RCU protected. */
1714 static int rtnl_fill_prop_list(struct sk_buff *skb,
1715                                const struct net_device *dev)
1716 {
1717         struct nlattr *prop_list;
1718         int ret;
1719
1720         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1721         if (!prop_list)
1722                 return -EMSGSIZE;
1723
1724         ret = rtnl_fill_alt_ifnames(skb, dev);
1725         if (ret <= 0)
1726                 goto nest_cancel;
1727
1728         nla_nest_end(skb, prop_list);
1729         return 0;
1730
1731 nest_cancel:
1732         nla_nest_cancel(skb, prop_list);
1733         return ret;
1734 }
1735
1736 static int rtnl_fill_proto_down(struct sk_buff *skb,
1737                                 const struct net_device *dev)
1738 {
1739         struct nlattr *pr;
1740         u32 preason;
1741
1742         if (nla_put_u8(skb, IFLA_PROTO_DOWN, READ_ONCE(dev->proto_down)))
1743                 goto nla_put_failure;
1744
1745         preason = READ_ONCE(dev->proto_down_reason);
1746         if (!preason)
1747                 return 0;
1748
1749         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1750         if (!pr)
1751                 return -EMSGSIZE;
1752
1753         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1754                 nla_nest_cancel(skb, pr);
1755                 goto nla_put_failure;
1756         }
1757
1758         nla_nest_end(skb, pr);
1759         return 0;
1760
1761 nla_put_failure:
1762         return -EMSGSIZE;
1763 }
1764
1765 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1766                                   const struct net_device *dev)
1767 {
1768         struct nlattr *devlink_port_nest;
1769         int ret;
1770
1771         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1772         if (!devlink_port_nest)
1773                 return -EMSGSIZE;
1774
1775         if (dev->devlink_port) {
1776                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1777                 if (ret < 0)
1778                         goto nest_cancel;
1779         }
1780
1781         nla_nest_end(skb, devlink_port_nest);
1782         return 0;
1783
1784 nest_cancel:
1785         nla_nest_cancel(skb, devlink_port_nest);
1786         return ret;
1787 }
1788
1789 static int rtnl_fill_dpll_pin(struct sk_buff *skb,
1790                               const struct net_device *dev)
1791 {
1792         struct nlattr *dpll_pin_nest;
1793         int ret;
1794
1795         dpll_pin_nest = nla_nest_start(skb, IFLA_DPLL_PIN);
1796         if (!dpll_pin_nest)
1797                 return -EMSGSIZE;
1798
1799         ret = dpll_netdev_add_pin_handle(skb, dev);
1800         if (ret < 0)
1801                 goto nest_cancel;
1802
1803         nla_nest_end(skb, dpll_pin_nest);
1804         return 0;
1805
1806 nest_cancel:
1807         nla_nest_cancel(skb, dpll_pin_nest);
1808         return ret;
1809 }
1810
1811 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1812                             struct net_device *dev, struct net *src_net,
1813                             int type, u32 pid, u32 seq, u32 change,
1814                             unsigned int flags, u32 ext_filter_mask,
1815                             u32 event, int *new_nsid, int new_ifindex,
1816                             int tgt_netnsid, gfp_t gfp)
1817 {
1818         char devname[IFNAMSIZ];
1819         struct ifinfomsg *ifm;
1820         struct nlmsghdr *nlh;
1821         struct Qdisc *qdisc;
1822
1823         ASSERT_RTNL();
1824         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1825         if (nlh == NULL)
1826                 return -EMSGSIZE;
1827
1828         ifm = nlmsg_data(nlh);
1829         ifm->ifi_family = AF_UNSPEC;
1830         ifm->__ifi_pad = 0;
1831         ifm->ifi_type = READ_ONCE(dev->type);
1832         ifm->ifi_index = READ_ONCE(dev->ifindex);
1833         ifm->ifi_flags = dev_get_flags(dev);
1834         ifm->ifi_change = change;
1835
1836         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1837                 goto nla_put_failure;
1838
1839         netdev_copy_name(dev, devname);
1840         if (nla_put_string(skb, IFLA_IFNAME, devname))
1841                 goto nla_put_failure;
1842
1843         if (nla_put_u32(skb, IFLA_TXQLEN, READ_ONCE(dev->tx_queue_len)) ||
1844             nla_put_u8(skb, IFLA_OPERSTATE,
1845                        netif_running(dev) ? READ_ONCE(dev->operstate) :
1846                                             IF_OPER_DOWN) ||
1847             nla_put_u8(skb, IFLA_LINKMODE, READ_ONCE(dev->link_mode)) ||
1848             nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) ||
1849             nla_put_u32(skb, IFLA_MIN_MTU, READ_ONCE(dev->min_mtu)) ||
1850             nla_put_u32(skb, IFLA_MAX_MTU, READ_ONCE(dev->max_mtu)) ||
1851             nla_put_u32(skb, IFLA_GROUP, READ_ONCE(dev->group)) ||
1852             nla_put_u32(skb, IFLA_PROMISCUITY, READ_ONCE(dev->promiscuity)) ||
1853             nla_put_u32(skb, IFLA_ALLMULTI, READ_ONCE(dev->allmulti)) ||
1854             nla_put_u32(skb, IFLA_NUM_TX_QUEUES,
1855                         READ_ONCE(dev->num_tx_queues)) ||
1856             nla_put_u32(skb, IFLA_GSO_MAX_SEGS,
1857                         READ_ONCE(dev->gso_max_segs)) ||
1858             nla_put_u32(skb, IFLA_GSO_MAX_SIZE,
1859                         READ_ONCE(dev->gso_max_size)) ||
1860             nla_put_u32(skb, IFLA_GRO_MAX_SIZE,
1861                         READ_ONCE(dev->gro_max_size)) ||
1862             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE,
1863                         READ_ONCE(dev->gso_ipv4_max_size)) ||
1864             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE,
1865                         READ_ONCE(dev->gro_ipv4_max_size)) ||
1866             nla_put_u32(skb, IFLA_TSO_MAX_SIZE,
1867                         READ_ONCE(dev->tso_max_size)) ||
1868             nla_put_u32(skb, IFLA_TSO_MAX_SEGS,
1869                         READ_ONCE(dev->tso_max_segs)) ||
1870 #ifdef CONFIG_RPS
1871             nla_put_u32(skb, IFLA_NUM_RX_QUEUES,
1872                         READ_ONCE(dev->num_rx_queues)) ||
1873 #endif
1874             put_master_ifindex(skb, dev) ||
1875             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1876             nla_put_ifalias(skb, dev) ||
1877             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1878                         atomic_read(&dev->carrier_up_count) +
1879                         atomic_read(&dev->carrier_down_count)) ||
1880             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1881                         atomic_read(&dev->carrier_up_count)) ||
1882             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1883                         atomic_read(&dev->carrier_down_count)))
1884                 goto nla_put_failure;
1885
1886         if (rtnl_fill_proto_down(skb, dev))
1887                 goto nla_put_failure;
1888
1889         if (event != IFLA_EVENT_NONE) {
1890                 if (nla_put_u32(skb, IFLA_EVENT, event))
1891                         goto nla_put_failure;
1892         }
1893
1894         if (dev->addr_len) {
1895                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1896                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1897                         goto nla_put_failure;
1898         }
1899
1900         if (rtnl_phys_port_id_fill(skb, dev))
1901                 goto nla_put_failure;
1902
1903         if (rtnl_phys_port_name_fill(skb, dev))
1904                 goto nla_put_failure;
1905
1906         if (rtnl_phys_switch_id_fill(skb, dev))
1907                 goto nla_put_failure;
1908
1909         if (rtnl_fill_stats(skb, dev))
1910                 goto nla_put_failure;
1911
1912         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1913                 goto nla_put_failure;
1914
1915         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1916                 goto nla_put_failure;
1917
1918         if (rtnl_xdp_fill(skb, dev))
1919                 goto nla_put_failure;
1920
1921         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1922                 if (rtnl_link_fill(skb, dev) < 0)
1923                         goto nla_put_failure;
1924         }
1925
1926         if (new_nsid &&
1927             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1928                 goto nla_put_failure;
1929         if (new_ifindex &&
1930             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1931                 goto nla_put_failure;
1932
1933         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1934             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1935                 goto nla_put_failure;
1936
1937         rcu_read_lock();
1938         if (rtnl_fill_link_netnsid(skb, dev, src_net, GFP_ATOMIC))
1939                 goto nla_put_failure_rcu;
1940         qdisc = rcu_dereference(dev->qdisc);
1941         if (qdisc && nla_put_string(skb, IFLA_QDISC, qdisc->ops->id))
1942                 goto nla_put_failure_rcu;
1943         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1944                 goto nla_put_failure_rcu;
1945         if (rtnl_fill_link_ifmap(skb, dev))
1946                 goto nla_put_failure_rcu;
1947         if (rtnl_fill_prop_list(skb, dev))
1948                 goto nla_put_failure_rcu;
1949         rcu_read_unlock();
1950
1951         if (dev->dev.parent &&
1952             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1953                            dev_name(dev->dev.parent)))
1954                 goto nla_put_failure;
1955
1956         if (dev->dev.parent && dev->dev.parent->bus &&
1957             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1958                            dev->dev.parent->bus->name))
1959                 goto nla_put_failure;
1960
1961         if (rtnl_fill_devlink_port(skb, dev))
1962                 goto nla_put_failure;
1963
1964         if (rtnl_fill_dpll_pin(skb, dev))
1965                 goto nla_put_failure;
1966
1967         nlmsg_end(skb, nlh);
1968         return 0;
1969
1970 nla_put_failure_rcu:
1971         rcu_read_unlock();
1972 nla_put_failure:
1973         nlmsg_cancel(skb, nlh);
1974         return -EMSGSIZE;
1975 }
1976
1977 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1978         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1979         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1980         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1981         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1982         [IFLA_MTU]              = { .type = NLA_U32 },
1983         [IFLA_LINK]             = { .type = NLA_U32 },
1984         [IFLA_MASTER]           = { .type = NLA_U32 },
1985         [IFLA_CARRIER]          = { .type = NLA_U8 },
1986         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1987         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1988         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1989         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1990         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1991         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1992         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1993         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1994          * allow 0-length string (needed to remove an alias).
1995          */
1996         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1997         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1998         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1999         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
2000         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
2001         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
2002         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
2003         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
2004         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
2005         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
2006         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
2007         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2008         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
2009         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
2010         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
2011         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
2012         [IFLA_XDP]              = { .type = NLA_NESTED },
2013         [IFLA_EVENT]            = { .type = NLA_U32 },
2014         [IFLA_GROUP]            = { .type = NLA_U32 },
2015         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
2016         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
2017         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
2018         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
2019         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
2020         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
2021         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
2022                                     .len = ALTIFNAMSIZ - 1 },
2023         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
2024         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
2025         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
2026         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
2027         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
2028         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
2029         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
2030         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
2031         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
2032         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
2033 };
2034
2035 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
2036         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
2037         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
2038         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
2039         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
2040 };
2041
2042 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
2043         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
2044         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
2045         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
2046         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
2047         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
2048         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
2049         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
2050         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
2051         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
2052         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2053         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2054         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2055         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2056 };
2057
2058 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2059         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2060         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2061                                     .len = PORT_PROFILE_MAX },
2062         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2063                                       .len = PORT_UUID_MAX },
2064         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2065                                     .len = PORT_UUID_MAX },
2066         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2067         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2068
2069         /* Unused, but we need to keep it here since user space could
2070          * fill it. It's also broken with regard to NLA_BINARY use in
2071          * combination with structs.
2072          */
2073         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2074                                     .len = sizeof(struct ifla_port_vsi) },
2075 };
2076
2077 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2078         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2079         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2080         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2081         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2082         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2083         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2084 };
2085
2086 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2087 {
2088         const struct rtnl_link_ops *ops = NULL;
2089         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2090
2091         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2092                 return NULL;
2093
2094         if (linfo[IFLA_INFO_KIND]) {
2095                 char kind[MODULE_NAME_LEN];
2096
2097                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2098                 ops = rtnl_link_ops_get(kind);
2099         }
2100
2101         return ops;
2102 }
2103
2104 static bool link_master_filtered(struct net_device *dev, int master_idx)
2105 {
2106         struct net_device *master;
2107
2108         if (!master_idx)
2109                 return false;
2110
2111         master = netdev_master_upper_dev_get(dev);
2112
2113         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2114          * another invalid value for ifindex to denote "no master".
2115          */
2116         if (master_idx == -1)
2117                 return !!master;
2118
2119         if (!master || master->ifindex != master_idx)
2120                 return true;
2121
2122         return false;
2123 }
2124
2125 static bool link_kind_filtered(const struct net_device *dev,
2126                                const struct rtnl_link_ops *kind_ops)
2127 {
2128         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2129                 return true;
2130
2131         return false;
2132 }
2133
2134 static bool link_dump_filtered(struct net_device *dev,
2135                                int master_idx,
2136                                const struct rtnl_link_ops *kind_ops)
2137 {
2138         if (link_master_filtered(dev, master_idx) ||
2139             link_kind_filtered(dev, kind_ops))
2140                 return true;
2141
2142         return false;
2143 }
2144
2145 /**
2146  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2147  * @sk: netlink socket
2148  * @netnsid: network namespace identifier
2149  *
2150  * Returns the network namespace identified by netnsid on success or an error
2151  * pointer on failure.
2152  */
2153 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2154 {
2155         struct net *net;
2156
2157         net = get_net_ns_by_id(sock_net(sk), netnsid);
2158         if (!net)
2159                 return ERR_PTR(-EINVAL);
2160
2161         /* For now, the caller is required to have CAP_NET_ADMIN in
2162          * the user namespace owning the target net ns.
2163          */
2164         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2165                 put_net(net);
2166                 return ERR_PTR(-EACCES);
2167         }
2168         return net;
2169 }
2170 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2171
2172 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2173                                       bool strict_check, struct nlattr **tb,
2174                                       struct netlink_ext_ack *extack)
2175 {
2176         int hdrlen;
2177
2178         if (strict_check) {
2179                 struct ifinfomsg *ifm;
2180
2181                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2182                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2183                         return -EINVAL;
2184                 }
2185
2186                 ifm = nlmsg_data(nlh);
2187                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2188                     ifm->ifi_change) {
2189                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2190                         return -EINVAL;
2191                 }
2192                 if (ifm->ifi_index) {
2193                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2194                         return -EINVAL;
2195                 }
2196
2197                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2198                                                      IFLA_MAX, ifla_policy,
2199                                                      extack);
2200         }
2201
2202         /* A hack to preserve kernel<->userspace interface.
2203          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2204          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2205          * what iproute2 < v3.9.0 used.
2206          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2207          * attribute, its netlink message is shorter than struct ifinfomsg.
2208          */
2209         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2210                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2211
2212         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2213                                       extack);
2214 }
2215
2216 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2217 {
2218         const struct rtnl_link_ops *kind_ops = NULL;
2219         struct netlink_ext_ack *extack = cb->extack;
2220         const struct nlmsghdr *nlh = cb->nlh;
2221         struct net *net = sock_net(skb->sk);
2222         unsigned int flags = NLM_F_MULTI;
2223         struct nlattr *tb[IFLA_MAX+1];
2224         struct {
2225                 unsigned long ifindex;
2226         } *ctx = (void *)cb->ctx;
2227         struct net *tgt_net = net;
2228         u32 ext_filter_mask = 0;
2229         struct net_device *dev;
2230         int master_idx = 0;
2231         int netnsid = -1;
2232         int err, i;
2233
2234         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2235         if (err < 0) {
2236                 if (cb->strict_check)
2237                         return err;
2238
2239                 goto walk_entries;
2240         }
2241
2242         for (i = 0; i <= IFLA_MAX; ++i) {
2243                 if (!tb[i])
2244                         continue;
2245
2246                 /* new attributes should only be added with strict checking */
2247                 switch (i) {
2248                 case IFLA_TARGET_NETNSID:
2249                         netnsid = nla_get_s32(tb[i]);
2250                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2251                         if (IS_ERR(tgt_net)) {
2252                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2253                                 return PTR_ERR(tgt_net);
2254                         }
2255                         break;
2256                 case IFLA_EXT_MASK:
2257                         ext_filter_mask = nla_get_u32(tb[i]);
2258                         break;
2259                 case IFLA_MASTER:
2260                         master_idx = nla_get_u32(tb[i]);
2261                         break;
2262                 case IFLA_LINKINFO:
2263                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2264                         break;
2265                 default:
2266                         if (cb->strict_check) {
2267                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2268                                 return -EINVAL;
2269                         }
2270                 }
2271         }
2272
2273         if (master_idx || kind_ops)
2274                 flags |= NLM_F_DUMP_FILTERED;
2275
2276 walk_entries:
2277         err = 0;
2278         for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
2279                 if (link_dump_filtered(dev, master_idx, kind_ops))
2280                         continue;
2281                 err = rtnl_fill_ifinfo(skb, dev, net, RTM_NEWLINK,
2282                                        NETLINK_CB(cb->skb).portid,
2283                                        nlh->nlmsg_seq, 0, flags,
2284                                        ext_filter_mask, 0, NULL, 0,
2285                                        netnsid, GFP_KERNEL);
2286                 if (err < 0)
2287                         break;
2288         }
2289         cb->seq = tgt_net->dev_base_seq;
2290         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2291         if (netnsid >= 0)
2292                 put_net(tgt_net);
2293
2294         return err;
2295 }
2296
2297 int rtnl_nla_parse_ifinfomsg(struct nlattr **tb, const struct nlattr *nla_peer,
2298                              struct netlink_ext_ack *exterr)
2299 {
2300         const struct ifinfomsg *ifmp;
2301         const struct nlattr *attrs;
2302         size_t len;
2303
2304         ifmp = nla_data(nla_peer);
2305         attrs = nla_data(nla_peer) + sizeof(struct ifinfomsg);
2306         len = nla_len(nla_peer) - sizeof(struct ifinfomsg);
2307
2308         if (ifmp->ifi_index < 0) {
2309                 NL_SET_ERR_MSG_ATTR(exterr, nla_peer,
2310                                     "ifindex can't be negative");
2311                 return -EINVAL;
2312         }
2313
2314         return nla_parse_deprecated(tb, IFLA_MAX, attrs, len, ifla_policy,
2315                                     exterr);
2316 }
2317 EXPORT_SYMBOL(rtnl_nla_parse_ifinfomsg);
2318
2319 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2320 {
2321         struct net *net;
2322         /* Examine the link attributes and figure out which
2323          * network namespace we are talking about.
2324          */
2325         if (tb[IFLA_NET_NS_PID])
2326                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2327         else if (tb[IFLA_NET_NS_FD])
2328                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2329         else
2330                 net = get_net(src_net);
2331         return net;
2332 }
2333 EXPORT_SYMBOL(rtnl_link_get_net);
2334
2335 /* Figure out which network namespace we are talking about by
2336  * examining the link attributes in the following order:
2337  *
2338  * 1. IFLA_NET_NS_PID
2339  * 2. IFLA_NET_NS_FD
2340  * 3. IFLA_TARGET_NETNSID
2341  */
2342 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2343                                                struct nlattr *tb[])
2344 {
2345         struct net *net;
2346
2347         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2348                 return rtnl_link_get_net(src_net, tb);
2349
2350         if (!tb[IFLA_TARGET_NETNSID])
2351                 return get_net(src_net);
2352
2353         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2354         if (!net)
2355                 return ERR_PTR(-EINVAL);
2356
2357         return net;
2358 }
2359
2360 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2361                                              struct net *src_net,
2362                                              struct nlattr *tb[], int cap)
2363 {
2364         struct net *net;
2365
2366         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2367         if (IS_ERR(net))
2368                 return net;
2369
2370         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2371                 put_net(net);
2372                 return ERR_PTR(-EPERM);
2373         }
2374
2375         return net;
2376 }
2377
2378 /* Verify that rtnetlink requests do not pass additional properties
2379  * potentially referring to different network namespaces.
2380  */
2381 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2382                                     struct netlink_ext_ack *extack,
2383                                     bool netns_id_only)
2384 {
2385
2386         if (netns_id_only) {
2387                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2388                         return 0;
2389
2390                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2391                 return -EOPNOTSUPP;
2392         }
2393
2394         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2395                 goto invalid_attr;
2396
2397         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2398                 goto invalid_attr;
2399
2400         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2401                 goto invalid_attr;
2402
2403         return 0;
2404
2405 invalid_attr:
2406         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2407         return -EINVAL;
2408 }
2409
2410 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2411                              int max_tx_rate)
2412 {
2413         const struct net_device_ops *ops = dev->netdev_ops;
2414
2415         if (!ops->ndo_set_vf_rate)
2416                 return -EOPNOTSUPP;
2417         if (max_tx_rate && max_tx_rate < min_tx_rate)
2418                 return -EINVAL;
2419
2420         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2421 }
2422
2423 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2424                             struct netlink_ext_ack *extack)
2425 {
2426         if (tb[IFLA_ADDRESS] &&
2427             nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2428                 return -EINVAL;
2429
2430         if (tb[IFLA_BROADCAST] &&
2431             nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2432                 return -EINVAL;
2433
2434         if (tb[IFLA_GSO_MAX_SIZE] &&
2435             nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2436                 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2437                 return -EINVAL;
2438         }
2439
2440         if (tb[IFLA_GSO_MAX_SEGS] &&
2441             (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2442              nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2443                 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2444                 return -EINVAL;
2445         }
2446
2447         if (tb[IFLA_GRO_MAX_SIZE] &&
2448             nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2449                 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2450                 return -EINVAL;
2451         }
2452
2453         if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2454             nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2455                 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2456                 return -EINVAL;
2457         }
2458
2459         if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2460             nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2461                 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2462                 return -EINVAL;
2463         }
2464
2465         if (tb[IFLA_AF_SPEC]) {
2466                 struct nlattr *af;
2467                 int rem, err;
2468
2469                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2470                         const struct rtnl_af_ops *af_ops;
2471
2472                         af_ops = rtnl_af_lookup(nla_type(af));
2473                         if (!af_ops)
2474                                 return -EAFNOSUPPORT;
2475
2476                         if (!af_ops->set_link_af)
2477                                 return -EOPNOTSUPP;
2478
2479                         if (af_ops->validate_link_af) {
2480                                 err = af_ops->validate_link_af(dev, af, extack);
2481                                 if (err < 0)
2482                                         return err;
2483                         }
2484                 }
2485         }
2486
2487         return 0;
2488 }
2489
2490 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2491                                   int guid_type)
2492 {
2493         const struct net_device_ops *ops = dev->netdev_ops;
2494
2495         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2496 }
2497
2498 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2499 {
2500         if (dev->type != ARPHRD_INFINIBAND)
2501                 return -EOPNOTSUPP;
2502
2503         return handle_infiniband_guid(dev, ivt, guid_type);
2504 }
2505
2506 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2507 {
2508         const struct net_device_ops *ops = dev->netdev_ops;
2509         int err = -EINVAL;
2510
2511         if (tb[IFLA_VF_MAC]) {
2512                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2513
2514                 if (ivm->vf >= INT_MAX)
2515                         return -EINVAL;
2516                 err = -EOPNOTSUPP;
2517                 if (ops->ndo_set_vf_mac)
2518                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2519                                                   ivm->mac);
2520                 if (err < 0)
2521                         return err;
2522         }
2523
2524         if (tb[IFLA_VF_VLAN]) {
2525                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2526
2527                 if (ivv->vf >= INT_MAX)
2528                         return -EINVAL;
2529                 err = -EOPNOTSUPP;
2530                 if (ops->ndo_set_vf_vlan)
2531                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2532                                                    ivv->qos,
2533                                                    htons(ETH_P_8021Q));
2534                 if (err < 0)
2535                         return err;
2536         }
2537
2538         if (tb[IFLA_VF_VLAN_LIST]) {
2539                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2540                 struct nlattr *attr;
2541                 int rem, len = 0;
2542
2543                 err = -EOPNOTSUPP;
2544                 if (!ops->ndo_set_vf_vlan)
2545                         return err;
2546
2547                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2548                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2549                             nla_len(attr) < sizeof(struct ifla_vf_vlan_info)) {
2550                                 return -EINVAL;
2551                         }
2552                         if (len >= MAX_VLAN_LIST_LEN)
2553                                 return -EOPNOTSUPP;
2554                         ivvl[len] = nla_data(attr);
2555
2556                         len++;
2557                 }
2558                 if (len == 0)
2559                         return -EINVAL;
2560
2561                 if (ivvl[0]->vf >= INT_MAX)
2562                         return -EINVAL;
2563                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2564                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2565                 if (err < 0)
2566                         return err;
2567         }
2568
2569         if (tb[IFLA_VF_TX_RATE]) {
2570                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2571                 struct ifla_vf_info ivf;
2572
2573                 if (ivt->vf >= INT_MAX)
2574                         return -EINVAL;
2575                 err = -EOPNOTSUPP;
2576                 if (ops->ndo_get_vf_config)
2577                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2578                 if (err < 0)
2579                         return err;
2580
2581                 err = rtnl_set_vf_rate(dev, ivt->vf,
2582                                        ivf.min_tx_rate, ivt->rate);
2583                 if (err < 0)
2584                         return err;
2585         }
2586
2587         if (tb[IFLA_VF_RATE]) {
2588                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2589
2590                 if (ivt->vf >= INT_MAX)
2591                         return -EINVAL;
2592
2593                 err = rtnl_set_vf_rate(dev, ivt->vf,
2594                                        ivt->min_tx_rate, ivt->max_tx_rate);
2595                 if (err < 0)
2596                         return err;
2597         }
2598
2599         if (tb[IFLA_VF_SPOOFCHK]) {
2600                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2601
2602                 if (ivs->vf >= INT_MAX)
2603                         return -EINVAL;
2604                 err = -EOPNOTSUPP;
2605                 if (ops->ndo_set_vf_spoofchk)
2606                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2607                                                        ivs->setting);
2608                 if (err < 0)
2609                         return err;
2610         }
2611
2612         if (tb[IFLA_VF_LINK_STATE]) {
2613                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2614
2615                 if (ivl->vf >= INT_MAX)
2616                         return -EINVAL;
2617                 err = -EOPNOTSUPP;
2618                 if (ops->ndo_set_vf_link_state)
2619                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2620                                                          ivl->link_state);
2621                 if (err < 0)
2622                         return err;
2623         }
2624
2625         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2626                 struct ifla_vf_rss_query_en *ivrssq_en;
2627
2628                 err = -EOPNOTSUPP;
2629                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2630                 if (ivrssq_en->vf >= INT_MAX)
2631                         return -EINVAL;
2632                 if (ops->ndo_set_vf_rss_query_en)
2633                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2634                                                            ivrssq_en->setting);
2635                 if (err < 0)
2636                         return err;
2637         }
2638
2639         if (tb[IFLA_VF_TRUST]) {
2640                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2641
2642                 if (ivt->vf >= INT_MAX)
2643                         return -EINVAL;
2644                 err = -EOPNOTSUPP;
2645                 if (ops->ndo_set_vf_trust)
2646                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2647                 if (err < 0)
2648                         return err;
2649         }
2650
2651         if (tb[IFLA_VF_IB_NODE_GUID]) {
2652                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2653
2654                 if (ivt->vf >= INT_MAX)
2655                         return -EINVAL;
2656                 if (!ops->ndo_set_vf_guid)
2657                         return -EOPNOTSUPP;
2658                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2659         }
2660
2661         if (tb[IFLA_VF_IB_PORT_GUID]) {
2662                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2663
2664                 if (ivt->vf >= INT_MAX)
2665                         return -EINVAL;
2666                 if (!ops->ndo_set_vf_guid)
2667                         return -EOPNOTSUPP;
2668
2669                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2670         }
2671
2672         return err;
2673 }
2674
2675 static int do_set_master(struct net_device *dev, int ifindex,
2676                          struct netlink_ext_ack *extack)
2677 {
2678         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2679         const struct net_device_ops *ops;
2680         int err;
2681
2682         if (upper_dev) {
2683                 if (upper_dev->ifindex == ifindex)
2684                         return 0;
2685                 ops = upper_dev->netdev_ops;
2686                 if (ops->ndo_del_slave) {
2687                         err = ops->ndo_del_slave(upper_dev, dev);
2688                         if (err)
2689                                 return err;
2690                 } else {
2691                         return -EOPNOTSUPP;
2692                 }
2693         }
2694
2695         if (ifindex) {
2696                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2697                 if (!upper_dev)
2698                         return -EINVAL;
2699                 ops = upper_dev->netdev_ops;
2700                 if (ops->ndo_add_slave) {
2701                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2702                         if (err)
2703                                 return err;
2704                 } else {
2705                         return -EOPNOTSUPP;
2706                 }
2707         }
2708         return 0;
2709 }
2710
2711 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2712         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2713         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2714 };
2715
2716 static int do_set_proto_down(struct net_device *dev,
2717                              struct nlattr *nl_proto_down,
2718                              struct nlattr *nl_proto_down_reason,
2719                              struct netlink_ext_ack *extack)
2720 {
2721         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2722         unsigned long mask = 0;
2723         u32 value;
2724         bool proto_down;
2725         int err;
2726
2727         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2728                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2729                 return -EOPNOTSUPP;
2730         }
2731
2732         if (nl_proto_down_reason) {
2733                 err = nla_parse_nested_deprecated(pdreason,
2734                                                   IFLA_PROTO_DOWN_REASON_MAX,
2735                                                   nl_proto_down_reason,
2736                                                   ifla_proto_down_reason_policy,
2737                                                   NULL);
2738                 if (err < 0)
2739                         return err;
2740
2741                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2742                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2743                         return -EINVAL;
2744                 }
2745
2746                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2747
2748                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2749                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2750
2751                 dev_change_proto_down_reason(dev, mask, value);
2752         }
2753
2754         if (nl_proto_down) {
2755                 proto_down = nla_get_u8(nl_proto_down);
2756
2757                 /* Don't turn off protodown if there are active reasons */
2758                 if (!proto_down && dev->proto_down_reason) {
2759                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2760                         return -EBUSY;
2761                 }
2762                 err = dev_change_proto_down(dev,
2763                                             proto_down);
2764                 if (err)
2765                         return err;
2766         }
2767
2768         return 0;
2769 }
2770
2771 #define DO_SETLINK_MODIFIED     0x01
2772 /* notify flag means notify + modified. */
2773 #define DO_SETLINK_NOTIFY       0x03
2774 static int do_setlink(const struct sk_buff *skb,
2775                       struct net_device *dev, struct ifinfomsg *ifm,
2776                       struct netlink_ext_ack *extack,
2777                       struct nlattr **tb, int status)
2778 {
2779         const struct net_device_ops *ops = dev->netdev_ops;
2780         char ifname[IFNAMSIZ];
2781         int err;
2782
2783         if (tb[IFLA_IFNAME])
2784                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2785         else
2786                 ifname[0] = '\0';
2787
2788         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2789                 const char *pat = ifname[0] ? ifname : NULL;
2790                 struct net *net;
2791                 int new_ifindex;
2792
2793                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2794                                                 tb, CAP_NET_ADMIN);
2795                 if (IS_ERR(net)) {
2796                         err = PTR_ERR(net);
2797                         goto errout;
2798                 }
2799
2800                 if (tb[IFLA_NEW_IFINDEX])
2801                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2802                 else
2803                         new_ifindex = 0;
2804
2805                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2806                 put_net(net);
2807                 if (err)
2808                         goto errout;
2809                 status |= DO_SETLINK_MODIFIED;
2810         }
2811
2812         if (tb[IFLA_MAP]) {
2813                 struct rtnl_link_ifmap *u_map;
2814                 struct ifmap k_map;
2815
2816                 if (!ops->ndo_set_config) {
2817                         err = -EOPNOTSUPP;
2818                         goto errout;
2819                 }
2820
2821                 if (!netif_device_present(dev)) {
2822                         err = -ENODEV;
2823                         goto errout;
2824                 }
2825
2826                 u_map = nla_data(tb[IFLA_MAP]);
2827                 k_map.mem_start = (unsigned long) u_map->mem_start;
2828                 k_map.mem_end = (unsigned long) u_map->mem_end;
2829                 k_map.base_addr = (unsigned short) u_map->base_addr;
2830                 k_map.irq = (unsigned char) u_map->irq;
2831                 k_map.dma = (unsigned char) u_map->dma;
2832                 k_map.port = (unsigned char) u_map->port;
2833
2834                 err = ops->ndo_set_config(dev, &k_map);
2835                 if (err < 0)
2836                         goto errout;
2837
2838                 status |= DO_SETLINK_NOTIFY;
2839         }
2840
2841         if (tb[IFLA_ADDRESS]) {
2842                 struct sockaddr *sa;
2843                 int len;
2844
2845                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2846                                                   sizeof(*sa));
2847                 sa = kmalloc(len, GFP_KERNEL);
2848                 if (!sa) {
2849                         err = -ENOMEM;
2850                         goto errout;
2851                 }
2852                 sa->sa_family = dev->type;
2853                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2854                        dev->addr_len);
2855                 err = dev_set_mac_address_user(dev, sa, extack);
2856                 kfree(sa);
2857                 if (err)
2858                         goto errout;
2859                 status |= DO_SETLINK_MODIFIED;
2860         }
2861
2862         if (tb[IFLA_MTU]) {
2863                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2864                 if (err < 0)
2865                         goto errout;
2866                 status |= DO_SETLINK_MODIFIED;
2867         }
2868
2869         if (tb[IFLA_GROUP]) {
2870                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2871                 status |= DO_SETLINK_NOTIFY;
2872         }
2873
2874         /*
2875          * Interface selected by interface index but interface
2876          * name provided implies that a name change has been
2877          * requested.
2878          */
2879         if (ifm->ifi_index > 0 && ifname[0]) {
2880                 err = dev_change_name(dev, ifname);
2881                 if (err < 0)
2882                         goto errout;
2883                 status |= DO_SETLINK_MODIFIED;
2884         }
2885
2886         if (tb[IFLA_IFALIAS]) {
2887                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2888                                     nla_len(tb[IFLA_IFALIAS]));
2889                 if (err < 0)
2890                         goto errout;
2891                 status |= DO_SETLINK_NOTIFY;
2892         }
2893
2894         if (tb[IFLA_BROADCAST]) {
2895                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2896                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2897         }
2898
2899         if (ifm->ifi_flags || ifm->ifi_change) {
2900                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2901                                        extack);
2902                 if (err < 0)
2903                         goto errout;
2904         }
2905
2906         if (tb[IFLA_MASTER]) {
2907                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2908                 if (err)
2909                         goto errout;
2910                 status |= DO_SETLINK_MODIFIED;
2911         }
2912
2913         if (tb[IFLA_CARRIER]) {
2914                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2915                 if (err)
2916                         goto errout;
2917                 status |= DO_SETLINK_MODIFIED;
2918         }
2919
2920         if (tb[IFLA_TXQLEN]) {
2921                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2922
2923                 err = dev_change_tx_queue_len(dev, value);
2924                 if (err)
2925                         goto errout;
2926                 status |= DO_SETLINK_MODIFIED;
2927         }
2928
2929         if (tb[IFLA_GSO_MAX_SIZE]) {
2930                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2931
2932                 if (dev->gso_max_size ^ max_size) {
2933                         netif_set_gso_max_size(dev, max_size);
2934                         status |= DO_SETLINK_MODIFIED;
2935                 }
2936         }
2937
2938         if (tb[IFLA_GSO_MAX_SEGS]) {
2939                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2940
2941                 if (dev->gso_max_segs ^ max_segs) {
2942                         netif_set_gso_max_segs(dev, max_segs);
2943                         status |= DO_SETLINK_MODIFIED;
2944                 }
2945         }
2946
2947         if (tb[IFLA_GRO_MAX_SIZE]) {
2948                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2949
2950                 if (dev->gro_max_size ^ gro_max_size) {
2951                         netif_set_gro_max_size(dev, gro_max_size);
2952                         status |= DO_SETLINK_MODIFIED;
2953                 }
2954         }
2955
2956         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2957                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2958
2959                 if (dev->gso_ipv4_max_size ^ max_size) {
2960                         netif_set_gso_ipv4_max_size(dev, max_size);
2961                         status |= DO_SETLINK_MODIFIED;
2962                 }
2963         }
2964
2965         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2966                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2967
2968                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2969                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2970                         status |= DO_SETLINK_MODIFIED;
2971                 }
2972         }
2973
2974         if (tb[IFLA_OPERSTATE])
2975                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2976
2977         if (tb[IFLA_LINKMODE]) {
2978                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2979
2980                 if (dev->link_mode ^ value)
2981                         status |= DO_SETLINK_NOTIFY;
2982                 WRITE_ONCE(dev->link_mode, value);
2983         }
2984
2985         if (tb[IFLA_VFINFO_LIST]) {
2986                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2987                 struct nlattr *attr;
2988                 int rem;
2989
2990                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2991                         if (nla_type(attr) != IFLA_VF_INFO ||
2992                             nla_len(attr) < NLA_HDRLEN) {
2993                                 err = -EINVAL;
2994                                 goto errout;
2995                         }
2996                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2997                                                           attr,
2998                                                           ifla_vf_policy,
2999                                                           NULL);
3000                         if (err < 0)
3001                                 goto errout;
3002                         err = do_setvfinfo(dev, vfinfo);
3003                         if (err < 0)
3004                                 goto errout;
3005                         status |= DO_SETLINK_NOTIFY;
3006                 }
3007         }
3008         err = 0;
3009
3010         if (tb[IFLA_VF_PORTS]) {
3011                 struct nlattr *port[IFLA_PORT_MAX+1];
3012                 struct nlattr *attr;
3013                 int vf;
3014                 int rem;
3015
3016                 err = -EOPNOTSUPP;
3017                 if (!ops->ndo_set_vf_port)
3018                         goto errout;
3019
3020                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
3021                         if (nla_type(attr) != IFLA_VF_PORT ||
3022                             nla_len(attr) < NLA_HDRLEN) {
3023                                 err = -EINVAL;
3024                                 goto errout;
3025                         }
3026                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3027                                                           attr,
3028                                                           ifla_port_policy,
3029                                                           NULL);
3030                         if (err < 0)
3031                                 goto errout;
3032                         if (!port[IFLA_PORT_VF]) {
3033                                 err = -EOPNOTSUPP;
3034                                 goto errout;
3035                         }
3036                         vf = nla_get_u32(port[IFLA_PORT_VF]);
3037                         err = ops->ndo_set_vf_port(dev, vf, port);
3038                         if (err < 0)
3039                                 goto errout;
3040                         status |= DO_SETLINK_NOTIFY;
3041                 }
3042         }
3043         err = 0;
3044
3045         if (tb[IFLA_PORT_SELF]) {
3046                 struct nlattr *port[IFLA_PORT_MAX+1];
3047
3048                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3049                                                   tb[IFLA_PORT_SELF],
3050                                                   ifla_port_policy, NULL);
3051                 if (err < 0)
3052                         goto errout;
3053
3054                 err = -EOPNOTSUPP;
3055                 if (ops->ndo_set_vf_port)
3056                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3057                 if (err < 0)
3058                         goto errout;
3059                 status |= DO_SETLINK_NOTIFY;
3060         }
3061
3062         if (tb[IFLA_AF_SPEC]) {
3063                 struct nlattr *af;
3064                 int rem;
3065
3066                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3067                         const struct rtnl_af_ops *af_ops;
3068
3069                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3070
3071                         err = af_ops->set_link_af(dev, af, extack);
3072                         if (err < 0)
3073                                 goto errout;
3074
3075                         status |= DO_SETLINK_NOTIFY;
3076                 }
3077         }
3078         err = 0;
3079
3080         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3081                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3082                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3083                 if (err)
3084                         goto errout;
3085                 status |= DO_SETLINK_NOTIFY;
3086         }
3087
3088         if (tb[IFLA_XDP]) {
3089                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3090                 u32 xdp_flags = 0;
3091
3092                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3093                                                   tb[IFLA_XDP],
3094                                                   ifla_xdp_policy, NULL);
3095                 if (err < 0)
3096                         goto errout;
3097
3098                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3099                         err = -EINVAL;
3100                         goto errout;
3101                 }
3102
3103                 if (xdp[IFLA_XDP_FLAGS]) {
3104                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3105                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3106                                 err = -EINVAL;
3107                                 goto errout;
3108                         }
3109                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3110                                 err = -EINVAL;
3111                                 goto errout;
3112                         }
3113                 }
3114
3115                 if (xdp[IFLA_XDP_FD]) {
3116                         int expected_fd = -1;
3117
3118                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3119                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3120                                         err = -EINVAL;
3121                                         goto errout;
3122                                 }
3123                                 expected_fd =
3124                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3125                         }
3126
3127                         err = dev_change_xdp_fd(dev, extack,
3128                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3129                                                 expected_fd,
3130                                                 xdp_flags);
3131                         if (err)
3132                                 goto errout;
3133                         status |= DO_SETLINK_NOTIFY;
3134                 }
3135         }
3136
3137 errout:
3138         if (status & DO_SETLINK_MODIFIED) {
3139                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3140                         netdev_state_change(dev);
3141
3142                 if (err < 0)
3143                         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",
3144                                              dev->name);
3145         }
3146
3147         return err;
3148 }
3149
3150 static struct net_device *rtnl_dev_get(struct net *net,
3151                                        struct nlattr *tb[])
3152 {
3153         char ifname[ALTIFNAMSIZ];
3154
3155         if (tb[IFLA_IFNAME])
3156                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3157         else if (tb[IFLA_ALT_IFNAME])
3158                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3159         else
3160                 return NULL;
3161
3162         return __dev_get_by_name(net, ifname);
3163 }
3164
3165 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3166                         struct netlink_ext_ack *extack)
3167 {
3168         struct net *net = sock_net(skb->sk);
3169         struct ifinfomsg *ifm;
3170         struct net_device *dev;
3171         int err;
3172         struct nlattr *tb[IFLA_MAX+1];
3173
3174         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3175                                      ifla_policy, extack);
3176         if (err < 0)
3177                 goto errout;
3178
3179         err = rtnl_ensure_unique_netns(tb, extack, false);
3180         if (err < 0)
3181                 goto errout;
3182
3183         err = -EINVAL;
3184         ifm = nlmsg_data(nlh);
3185         if (ifm->ifi_index > 0)
3186                 dev = __dev_get_by_index(net, ifm->ifi_index);
3187         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3188                 dev = rtnl_dev_get(net, tb);
3189         else
3190                 goto errout;
3191
3192         if (dev == NULL) {
3193                 err = -ENODEV;
3194                 goto errout;
3195         }
3196
3197         err = validate_linkmsg(dev, tb, extack);
3198         if (err < 0)
3199                 goto errout;
3200
3201         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3202 errout:
3203         return err;
3204 }
3205
3206 static int rtnl_group_dellink(const struct net *net, int group)
3207 {
3208         struct net_device *dev, *aux;
3209         LIST_HEAD(list_kill);
3210         bool found = false;
3211
3212         if (!group)
3213                 return -EPERM;
3214
3215         for_each_netdev(net, dev) {
3216                 if (dev->group == group) {
3217                         const struct rtnl_link_ops *ops;
3218
3219                         found = true;
3220                         ops = dev->rtnl_link_ops;
3221                         if (!ops || !ops->dellink)
3222                                 return -EOPNOTSUPP;
3223                 }
3224         }
3225
3226         if (!found)
3227                 return -ENODEV;
3228
3229         for_each_netdev_safe(net, dev, aux) {
3230                 if (dev->group == group) {
3231                         const struct rtnl_link_ops *ops;
3232
3233                         ops = dev->rtnl_link_ops;
3234                         ops->dellink(dev, &list_kill);
3235                 }
3236         }
3237         unregister_netdevice_many(&list_kill);
3238
3239         return 0;
3240 }
3241
3242 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3243 {
3244         const struct rtnl_link_ops *ops;
3245         LIST_HEAD(list_kill);
3246
3247         ops = dev->rtnl_link_ops;
3248         if (!ops || !ops->dellink)
3249                 return -EOPNOTSUPP;
3250
3251         ops->dellink(dev, &list_kill);
3252         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3253
3254         return 0;
3255 }
3256 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3257
3258 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3259                         struct netlink_ext_ack *extack)
3260 {
3261         struct net *net = sock_net(skb->sk);
3262         u32 portid = NETLINK_CB(skb).portid;
3263         struct net *tgt_net = net;
3264         struct net_device *dev = NULL;
3265         struct ifinfomsg *ifm;
3266         struct nlattr *tb[IFLA_MAX+1];
3267         int err;
3268         int netnsid = -1;
3269
3270         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3271                                      ifla_policy, extack);
3272         if (err < 0)
3273                 return err;
3274
3275         err = rtnl_ensure_unique_netns(tb, extack, true);
3276         if (err < 0)
3277                 return err;
3278
3279         if (tb[IFLA_TARGET_NETNSID]) {
3280                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3281                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3282                 if (IS_ERR(tgt_net))
3283                         return PTR_ERR(tgt_net);
3284         }
3285
3286         err = -EINVAL;
3287         ifm = nlmsg_data(nlh);
3288         if (ifm->ifi_index > 0)
3289                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3290         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3291                 dev = rtnl_dev_get(tgt_net, tb);
3292         else if (tb[IFLA_GROUP])
3293                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3294         else
3295                 goto out;
3296
3297         if (!dev) {
3298                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3299                         err = -ENODEV;
3300
3301                 goto out;
3302         }
3303
3304         err = rtnl_delete_link(dev, portid, nlh);
3305
3306 out:
3307         if (netnsid >= 0)
3308                 put_net(tgt_net);
3309
3310         return err;
3311 }
3312
3313 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3314                         u32 portid, const struct nlmsghdr *nlh)
3315 {
3316         unsigned int old_flags;
3317         int err;
3318
3319         old_flags = dev->flags;
3320         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3321                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3322                                          NULL);
3323                 if (err < 0)
3324                         return err;
3325         }
3326
3327         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3328                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3329         } else {
3330                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3331                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3332         }
3333         return 0;
3334 }
3335 EXPORT_SYMBOL(rtnl_configure_link);
3336
3337 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3338                                     unsigned char name_assign_type,
3339                                     const struct rtnl_link_ops *ops,
3340                                     struct nlattr *tb[],
3341                                     struct netlink_ext_ack *extack)
3342 {
3343         struct net_device *dev;
3344         unsigned int num_tx_queues = 1;
3345         unsigned int num_rx_queues = 1;
3346         int err;
3347
3348         if (tb[IFLA_NUM_TX_QUEUES])
3349                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3350         else if (ops->get_num_tx_queues)
3351                 num_tx_queues = ops->get_num_tx_queues();
3352
3353         if (tb[IFLA_NUM_RX_QUEUES])
3354                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3355         else if (ops->get_num_rx_queues)
3356                 num_rx_queues = ops->get_num_rx_queues();
3357
3358         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3359                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3360                 return ERR_PTR(-EINVAL);
3361         }
3362
3363         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3364                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3365                 return ERR_PTR(-EINVAL);
3366         }
3367
3368         if (ops->alloc) {
3369                 dev = ops->alloc(tb, ifname, name_assign_type,
3370                                  num_tx_queues, num_rx_queues);
3371                 if (IS_ERR(dev))
3372                         return dev;
3373         } else {
3374                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3375                                        name_assign_type, ops->setup,
3376                                        num_tx_queues, num_rx_queues);
3377         }
3378
3379         if (!dev)
3380                 return ERR_PTR(-ENOMEM);
3381
3382         err = validate_linkmsg(dev, tb, extack);
3383         if (err < 0) {
3384                 free_netdev(dev);
3385                 return ERR_PTR(err);
3386         }
3387
3388         dev_net_set(dev, net);
3389         dev->rtnl_link_ops = ops;
3390         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3391
3392         if (tb[IFLA_MTU]) {
3393                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3394
3395                 err = dev_validate_mtu(dev, mtu, extack);
3396                 if (err) {
3397                         free_netdev(dev);
3398                         return ERR_PTR(err);
3399                 }
3400                 dev->mtu = mtu;
3401         }
3402         if (tb[IFLA_ADDRESS]) {
3403                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3404                                nla_len(tb[IFLA_ADDRESS]));
3405                 dev->addr_assign_type = NET_ADDR_SET;
3406         }
3407         if (tb[IFLA_BROADCAST])
3408                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3409                                 nla_len(tb[IFLA_BROADCAST]));
3410         if (tb[IFLA_TXQLEN])
3411                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3412         if (tb[IFLA_OPERSTATE])
3413                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3414         if (tb[IFLA_LINKMODE])
3415                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3416         if (tb[IFLA_GROUP])
3417                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3418         if (tb[IFLA_GSO_MAX_SIZE])
3419                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3420         if (tb[IFLA_GSO_MAX_SEGS])
3421                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3422         if (tb[IFLA_GRO_MAX_SIZE])
3423                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3424         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3425                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3426         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3427                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3428
3429         return dev;
3430 }
3431 EXPORT_SYMBOL(rtnl_create_link);
3432
3433 static int rtnl_group_changelink(const struct sk_buff *skb,
3434                 struct net *net, int group,
3435                 struct ifinfomsg *ifm,
3436                 struct netlink_ext_ack *extack,
3437                 struct nlattr **tb)
3438 {
3439         struct net_device *dev, *aux;
3440         int err;
3441
3442         for_each_netdev_safe(net, dev, aux) {
3443                 if (dev->group == group) {
3444                         err = validate_linkmsg(dev, tb, extack);
3445                         if (err < 0)
3446                                 return err;
3447                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3448                         if (err < 0)
3449                                 return err;
3450                 }
3451         }
3452
3453         return 0;
3454 }
3455
3456 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3457                                const struct rtnl_link_ops *ops,
3458                                const struct nlmsghdr *nlh,
3459                                struct nlattr **tb, struct nlattr **data,
3460                                struct netlink_ext_ack *extack)
3461 {
3462         unsigned char name_assign_type = NET_NAME_USER;
3463         struct net *net = sock_net(skb->sk);
3464         u32 portid = NETLINK_CB(skb).portid;
3465         struct net *dest_net, *link_net;
3466         struct net_device *dev;
3467         char ifname[IFNAMSIZ];
3468         int err;
3469
3470         if (!ops->alloc && !ops->setup)
3471                 return -EOPNOTSUPP;
3472
3473         if (tb[IFLA_IFNAME]) {
3474                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3475         } else {
3476                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3477                 name_assign_type = NET_NAME_ENUM;
3478         }
3479
3480         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3481         if (IS_ERR(dest_net))
3482                 return PTR_ERR(dest_net);
3483
3484         if (tb[IFLA_LINK_NETNSID]) {
3485                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3486
3487                 link_net = get_net_ns_by_id(dest_net, id);
3488                 if (!link_net) {
3489                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3490                         err =  -EINVAL;
3491                         goto out;
3492                 }
3493                 err = -EPERM;
3494                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3495                         goto out;
3496         } else {
3497                 link_net = NULL;
3498         }
3499
3500         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3501                                name_assign_type, ops, tb, extack);
3502         if (IS_ERR(dev)) {
3503                 err = PTR_ERR(dev);
3504                 goto out;
3505         }
3506
3507         dev->ifindex = ifm->ifi_index;
3508
3509         if (ops->newlink)
3510                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3511         else
3512                 err = register_netdevice(dev);
3513         if (err < 0) {
3514                 free_netdev(dev);
3515                 goto out;
3516         }
3517
3518         err = rtnl_configure_link(dev, ifm, portid, nlh);
3519         if (err < 0)
3520                 goto out_unregister;
3521         if (link_net) {
3522                 err = dev_change_net_namespace(dev, dest_net, ifname);
3523                 if (err < 0)
3524                         goto out_unregister;
3525         }
3526         if (tb[IFLA_MASTER]) {
3527                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3528                 if (err)
3529                         goto out_unregister;
3530         }
3531 out:
3532         if (link_net)
3533                 put_net(link_net);
3534         put_net(dest_net);
3535         return err;
3536 out_unregister:
3537         if (ops->newlink) {
3538                 LIST_HEAD(list_kill);
3539
3540                 ops->dellink(dev, &list_kill);
3541                 unregister_netdevice_many(&list_kill);
3542         } else {
3543                 unregister_netdevice(dev);
3544         }
3545         goto out;
3546 }
3547
3548 struct rtnl_newlink_tbs {
3549         struct nlattr *tb[IFLA_MAX + 1];
3550         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3551         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3552 };
3553
3554 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3555                           struct rtnl_newlink_tbs *tbs,
3556                           struct netlink_ext_ack *extack)
3557 {
3558         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3559         struct nlattr ** const tb = tbs->tb;
3560         const struct rtnl_link_ops *m_ops;
3561         struct net_device *master_dev;
3562         struct net *net = sock_net(skb->sk);
3563         const struct rtnl_link_ops *ops;
3564         struct nlattr **slave_data;
3565         char kind[MODULE_NAME_LEN];
3566         struct net_device *dev;
3567         struct ifinfomsg *ifm;
3568         struct nlattr **data;
3569         bool link_specified;
3570         int err;
3571
3572 #ifdef CONFIG_MODULES
3573 replay:
3574 #endif
3575         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3576                                      ifla_policy, extack);
3577         if (err < 0)
3578                 return err;
3579
3580         err = rtnl_ensure_unique_netns(tb, extack, false);
3581         if (err < 0)
3582                 return err;
3583
3584         ifm = nlmsg_data(nlh);
3585         if (ifm->ifi_index > 0) {
3586                 link_specified = true;
3587                 dev = __dev_get_by_index(net, ifm->ifi_index);
3588         } else if (ifm->ifi_index < 0) {
3589                 NL_SET_ERR_MSG(extack, "ifindex can't be negative");
3590                 return -EINVAL;
3591         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3592                 link_specified = true;
3593                 dev = rtnl_dev_get(net, tb);
3594         } else {
3595                 link_specified = false;
3596                 dev = NULL;
3597         }
3598
3599         master_dev = NULL;
3600         m_ops = NULL;
3601         if (dev) {
3602                 master_dev = netdev_master_upper_dev_get(dev);
3603                 if (master_dev)
3604                         m_ops = master_dev->rtnl_link_ops;
3605         }
3606
3607         if (tb[IFLA_LINKINFO]) {
3608                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3609                                                   tb[IFLA_LINKINFO],
3610                                                   ifla_info_policy, NULL);
3611                 if (err < 0)
3612                         return err;
3613         } else
3614                 memset(linkinfo, 0, sizeof(linkinfo));
3615
3616         if (linkinfo[IFLA_INFO_KIND]) {
3617                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3618                 ops = rtnl_link_ops_get(kind);
3619         } else {
3620                 kind[0] = '\0';
3621                 ops = NULL;
3622         }
3623
3624         data = NULL;
3625         if (ops) {
3626                 if (ops->maxtype > RTNL_MAX_TYPE)
3627                         return -EINVAL;
3628
3629                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3630                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3631                                                           linkinfo[IFLA_INFO_DATA],
3632                                                           ops->policy, extack);
3633                         if (err < 0)
3634                                 return err;
3635                         data = tbs->attr;
3636                 }
3637                 if (ops->validate) {
3638                         err = ops->validate(tb, data, extack);
3639                         if (err < 0)
3640                                 return err;
3641                 }
3642         }
3643
3644         slave_data = NULL;
3645         if (m_ops) {
3646                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3647                         return -EINVAL;
3648
3649                 if (m_ops->slave_maxtype &&
3650                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3651                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3652                                                           m_ops->slave_maxtype,
3653                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3654                                                           m_ops->slave_policy,
3655                                                           extack);
3656                         if (err < 0)
3657                                 return err;
3658                         slave_data = tbs->slave_attr;
3659                 }
3660         }
3661
3662         if (dev) {
3663                 int status = 0;
3664
3665                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3666                         return -EEXIST;
3667                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3668                         return -EOPNOTSUPP;
3669
3670                 err = validate_linkmsg(dev, tb, extack);
3671                 if (err < 0)
3672                         return err;
3673
3674                 if (linkinfo[IFLA_INFO_DATA]) {
3675                         if (!ops || ops != dev->rtnl_link_ops ||
3676                             !ops->changelink)
3677                                 return -EOPNOTSUPP;
3678
3679                         err = ops->changelink(dev, tb, data, extack);
3680                         if (err < 0)
3681                                 return err;
3682                         status |= DO_SETLINK_NOTIFY;
3683                 }
3684
3685                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3686                         if (!m_ops || !m_ops->slave_changelink)
3687                                 return -EOPNOTSUPP;
3688
3689                         err = m_ops->slave_changelink(master_dev, dev, tb,
3690                                                       slave_data, extack);
3691                         if (err < 0)
3692                                 return err;
3693                         status |= DO_SETLINK_NOTIFY;
3694                 }
3695
3696                 return do_setlink(skb, dev, ifm, extack, tb, status);
3697         }
3698
3699         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3700                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3701                  * or it's for a group
3702                 */
3703                 if (link_specified)
3704                         return -ENODEV;
3705                 if (tb[IFLA_GROUP])
3706                         return rtnl_group_changelink(skb, net,
3707                                                 nla_get_u32(tb[IFLA_GROUP]),
3708                                                 ifm, extack, tb);
3709                 return -ENODEV;
3710         }
3711
3712         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3713                 return -EOPNOTSUPP;
3714
3715         if (!ops) {
3716 #ifdef CONFIG_MODULES
3717                 if (kind[0]) {
3718                         __rtnl_unlock();
3719                         request_module("rtnl-link-%s", kind);
3720                         rtnl_lock();
3721                         ops = rtnl_link_ops_get(kind);
3722                         if (ops)
3723                                 goto replay;
3724                 }
3725 #endif
3726                 NL_SET_ERR_MSG(extack, "Unknown device type");
3727                 return -EOPNOTSUPP;
3728         }
3729
3730         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3731 }
3732
3733 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3734                         struct netlink_ext_ack *extack)
3735 {
3736         struct rtnl_newlink_tbs *tbs;
3737         int ret;
3738
3739         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3740         if (!tbs)
3741                 return -ENOMEM;
3742
3743         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3744         kfree(tbs);
3745         return ret;
3746 }
3747
3748 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3749                                   const struct nlmsghdr *nlh,
3750                                   struct nlattr **tb,
3751                                   struct netlink_ext_ack *extack)
3752 {
3753         struct ifinfomsg *ifm;
3754         int i, err;
3755
3756         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3757                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3758                 return -EINVAL;
3759         }
3760
3761         if (!netlink_strict_get_check(skb))
3762                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3763                                               ifla_policy, extack);
3764
3765         ifm = nlmsg_data(nlh);
3766         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3767             ifm->ifi_change) {
3768                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3769                 return -EINVAL;
3770         }
3771
3772         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3773                                             ifla_policy, extack);
3774         if (err)
3775                 return err;
3776
3777         for (i = 0; i <= IFLA_MAX; i++) {
3778                 if (!tb[i])
3779                         continue;
3780
3781                 switch (i) {
3782                 case IFLA_IFNAME:
3783                 case IFLA_ALT_IFNAME:
3784                 case IFLA_EXT_MASK:
3785                 case IFLA_TARGET_NETNSID:
3786                         break;
3787                 default:
3788                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3789                         return -EINVAL;
3790                 }
3791         }
3792
3793         return 0;
3794 }
3795
3796 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3797                         struct netlink_ext_ack *extack)
3798 {
3799         struct net *net = sock_net(skb->sk);
3800         struct net *tgt_net = net;
3801         struct ifinfomsg *ifm;
3802         struct nlattr *tb[IFLA_MAX+1];
3803         struct net_device *dev = NULL;
3804         struct sk_buff *nskb;
3805         int netnsid = -1;
3806         int err;
3807         u32 ext_filter_mask = 0;
3808
3809         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3810         if (err < 0)
3811                 return err;
3812
3813         err = rtnl_ensure_unique_netns(tb, extack, true);
3814         if (err < 0)
3815                 return err;
3816
3817         if (tb[IFLA_TARGET_NETNSID]) {
3818                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3819                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3820                 if (IS_ERR(tgt_net))
3821                         return PTR_ERR(tgt_net);
3822         }
3823
3824         if (tb[IFLA_EXT_MASK])
3825                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3826
3827         err = -EINVAL;
3828         ifm = nlmsg_data(nlh);
3829         if (ifm->ifi_index > 0)
3830                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3831         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3832                 dev = rtnl_dev_get(tgt_net, tb);
3833         else
3834                 goto out;
3835
3836         err = -ENODEV;
3837         if (dev == NULL)
3838                 goto out;
3839
3840         err = -ENOBUFS;
3841         nskb = nlmsg_new_large(if_nlmsg_size(dev, ext_filter_mask));
3842         if (nskb == NULL)
3843                 goto out;
3844
3845         /* Synchronize the carrier state so we don't report a state
3846          * that we're not actually going to honour immediately; if
3847          * the driver just did a carrier off->on transition, we can
3848          * only TX if link watch work has run, but without this we'd
3849          * already report carrier on, even if it doesn't work yet.
3850          */
3851         linkwatch_sync_dev(dev);
3852
3853         err = rtnl_fill_ifinfo(nskb, dev, net,
3854                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3855                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3856                                0, NULL, 0, netnsid, GFP_KERNEL);
3857         if (err < 0) {
3858                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3859                 WARN_ON(err == -EMSGSIZE);
3860                 kfree_skb(nskb);
3861         } else
3862                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3863 out:
3864         if (netnsid >= 0)
3865                 put_net(tgt_net);
3866
3867         return err;
3868 }
3869
3870 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3871                            bool *changed, struct netlink_ext_ack *extack)
3872 {
3873         char *alt_ifname;
3874         size_t size;
3875         int err;
3876
3877         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3878         if (err)
3879                 return err;
3880
3881         if (cmd == RTM_NEWLINKPROP) {
3882                 size = rtnl_prop_list_size(dev);
3883                 size += nla_total_size(ALTIFNAMSIZ);
3884                 if (size >= U16_MAX) {
3885                         NL_SET_ERR_MSG(extack,
3886                                        "effective property list too long");
3887                         return -EINVAL;
3888                 }
3889         }
3890
3891         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3892         if (!alt_ifname)
3893                 return -ENOMEM;
3894
3895         if (cmd == RTM_NEWLINKPROP) {
3896                 err = netdev_name_node_alt_create(dev, alt_ifname);
3897                 if (!err)
3898                         alt_ifname = NULL;
3899         } else if (cmd == RTM_DELLINKPROP) {
3900                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3901         } else {
3902                 WARN_ON_ONCE(1);
3903                 err = -EINVAL;
3904         }
3905
3906         kfree(alt_ifname);
3907         if (!err)
3908                 *changed = true;
3909         return err;
3910 }
3911
3912 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3913                          struct netlink_ext_ack *extack)
3914 {
3915         struct net *net = sock_net(skb->sk);
3916         struct nlattr *tb[IFLA_MAX + 1];
3917         struct net_device *dev;
3918         struct ifinfomsg *ifm;
3919         bool changed = false;
3920         struct nlattr *attr;
3921         int err, rem;
3922
3923         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3924         if (err)
3925                 return err;
3926
3927         err = rtnl_ensure_unique_netns(tb, extack, true);
3928         if (err)
3929                 return err;
3930
3931         ifm = nlmsg_data(nlh);
3932         if (ifm->ifi_index > 0)
3933                 dev = __dev_get_by_index(net, ifm->ifi_index);
3934         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3935                 dev = rtnl_dev_get(net, tb);
3936         else
3937                 return -EINVAL;
3938
3939         if (!dev)
3940                 return -ENODEV;
3941
3942         if (!tb[IFLA_PROP_LIST])
3943                 return 0;
3944
3945         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3946                 switch (nla_type(attr)) {
3947                 case IFLA_ALT_IFNAME:
3948                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3949                         if (err)
3950                                 return err;
3951                         break;
3952                 }
3953         }
3954
3955         if (changed)
3956                 netdev_state_change(dev);
3957         return 0;
3958 }
3959
3960 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3961                             struct netlink_ext_ack *extack)
3962 {
3963         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3964 }
3965
3966 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3967                             struct netlink_ext_ack *extack)
3968 {
3969         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3970 }
3971
3972 static noinline_for_stack u32 rtnl_calcit(struct sk_buff *skb,
3973                                           struct nlmsghdr *nlh)
3974 {
3975         struct net *net = sock_net(skb->sk);
3976         size_t min_ifinfo_dump_size = 0;
3977         u32 ext_filter_mask = 0;
3978         struct net_device *dev;
3979         struct nlattr *nla;
3980         int hdrlen, rem;
3981
3982         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3983         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3984                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3985
3986         if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
3987                 return NLMSG_GOODSIZE;
3988
3989         nla_for_each_attr_type(nla, IFLA_EXT_MASK,
3990                                nlmsg_attrdata(nlh, hdrlen),
3991                                nlmsg_attrlen(nlh, hdrlen), rem) {
3992                 if (nla_len(nla) == sizeof(u32))
3993                         ext_filter_mask = nla_get_u32(nla);
3994         }
3995
3996         if (!ext_filter_mask)
3997                 return NLMSG_GOODSIZE;
3998         /*
3999          * traverse the list of net devices and compute the minimum
4000          * buffer size based upon the filter mask.
4001          */
4002         rcu_read_lock();
4003         for_each_netdev_rcu(net, dev) {
4004                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
4005                                            if_nlmsg_size(dev, ext_filter_mask));
4006         }
4007         rcu_read_unlock();
4008
4009         return nlmsg_total_size(min_ifinfo_dump_size);
4010 }
4011
4012 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
4013 {
4014         int idx;
4015         int s_idx = cb->family;
4016         int type = cb->nlh->nlmsg_type - RTM_BASE;
4017         int ret = 0;
4018
4019         if (s_idx == 0)
4020                 s_idx = 1;
4021
4022         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
4023                 struct rtnl_link __rcu **tab;
4024                 struct rtnl_link *link;
4025                 rtnl_dumpit_func dumpit;
4026
4027                 if (idx < s_idx || idx == PF_PACKET)
4028                         continue;
4029
4030                 if (type < 0 || type >= RTM_NR_MSGTYPES)
4031                         continue;
4032
4033                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
4034                 if (!tab)
4035                         continue;
4036
4037                 link = rcu_dereference_rtnl(tab[type]);
4038                 if (!link)
4039                         continue;
4040
4041                 dumpit = link->dumpit;
4042                 if (!dumpit)
4043                         continue;
4044
4045                 if (idx > s_idx) {
4046                         memset(&cb->args[0], 0, sizeof(cb->args));
4047                         cb->prev_seq = 0;
4048                         cb->seq = 0;
4049                 }
4050                 ret = dumpit(skb, cb);
4051                 if (ret)
4052                         break;
4053         }
4054         cb->family = idx;
4055
4056         return skb->len ? : ret;
4057 }
4058
4059 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4060                                        unsigned int change,
4061                                        u32 event, gfp_t flags, int *new_nsid,
4062                                        int new_ifindex, u32 portid,
4063                                        const struct nlmsghdr *nlh)
4064 {
4065         struct net *net = dev_net(dev);
4066         struct sk_buff *skb;
4067         int err = -ENOBUFS;
4068         u32 seq = 0;
4069
4070         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4071         if (skb == NULL)
4072                 goto errout;
4073
4074         if (nlmsg_report(nlh))
4075                 seq = nlmsg_seq(nlh);
4076         else
4077                 portid = 0;
4078
4079         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4080                                type, portid, seq, change, 0, 0, event,
4081                                new_nsid, new_ifindex, -1, flags);
4082         if (err < 0) {
4083                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4084                 WARN_ON(err == -EMSGSIZE);
4085                 kfree_skb(skb);
4086                 goto errout;
4087         }
4088         return skb;
4089 errout:
4090         if (err < 0)
4091                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4092         return NULL;
4093 }
4094
4095 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4096                        u32 portid, const struct nlmsghdr *nlh)
4097 {
4098         struct net *net = dev_net(dev);
4099
4100         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4101 }
4102
4103 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4104                                unsigned int change, u32 event,
4105                                gfp_t flags, int *new_nsid, int new_ifindex,
4106                                u32 portid, const struct nlmsghdr *nlh)
4107 {
4108         struct sk_buff *skb;
4109
4110         if (dev->reg_state != NETREG_REGISTERED)
4111                 return;
4112
4113         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4114                                      new_ifindex, portid, nlh);
4115         if (skb)
4116                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4117 }
4118
4119 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4120                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4121 {
4122         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4123                            NULL, 0, portid, nlh);
4124 }
4125
4126 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4127                          gfp_t flags, int *new_nsid, int new_ifindex)
4128 {
4129         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4130                            new_nsid, new_ifindex, 0, NULL);
4131 }
4132
4133 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4134                                    struct net_device *dev,
4135                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4136                                    int type, unsigned int flags,
4137                                    int nlflags, u16 ndm_state)
4138 {
4139         struct nlmsghdr *nlh;
4140         struct ndmsg *ndm;
4141
4142         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4143         if (!nlh)
4144                 return -EMSGSIZE;
4145
4146         ndm = nlmsg_data(nlh);
4147         ndm->ndm_family  = AF_BRIDGE;
4148         ndm->ndm_pad1    = 0;
4149         ndm->ndm_pad2    = 0;
4150         ndm->ndm_flags   = flags;
4151         ndm->ndm_type    = 0;
4152         ndm->ndm_ifindex = dev->ifindex;
4153         ndm->ndm_state   = ndm_state;
4154
4155         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4156                 goto nla_put_failure;
4157         if (vid)
4158                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4159                         goto nla_put_failure;
4160
4161         nlmsg_end(skb, nlh);
4162         return 0;
4163
4164 nla_put_failure:
4165         nlmsg_cancel(skb, nlh);
4166         return -EMSGSIZE;
4167 }
4168
4169 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4170 {
4171         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4172                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
4173                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4174                0;
4175 }
4176
4177 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4178                             u16 ndm_state)
4179 {
4180         struct net *net = dev_net(dev);
4181         struct sk_buff *skb;
4182         int err = -ENOBUFS;
4183
4184         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4185         if (!skb)
4186                 goto errout;
4187
4188         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4189                                       0, 0, type, NTF_SELF, 0, ndm_state);
4190         if (err < 0) {
4191                 kfree_skb(skb);
4192                 goto errout;
4193         }
4194
4195         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4196         return;
4197 errout:
4198         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4199 }
4200
4201 /*
4202  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4203  */
4204 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4205                      struct nlattr *tb[],
4206                      struct net_device *dev,
4207                      const unsigned char *addr, u16 vid,
4208                      u16 flags)
4209 {
4210         int err = -EINVAL;
4211
4212         /* If aging addresses are supported device will need to
4213          * implement its own handler for this.
4214          */
4215         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4216                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4217                 return err;
4218         }
4219
4220         if (tb[NDA_FLAGS_EXT]) {
4221                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4222                 return err;
4223         }
4224
4225         if (vid) {
4226                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4227                 return err;
4228         }
4229
4230         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4231                 err = dev_uc_add_excl(dev, addr);
4232         else if (is_multicast_ether_addr(addr))
4233                 err = dev_mc_add_excl(dev, addr);
4234
4235         /* Only return duplicate errors if NLM_F_EXCL is set */
4236         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4237                 err = 0;
4238
4239         return err;
4240 }
4241 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4242
4243 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4244                          struct netlink_ext_ack *extack)
4245 {
4246         u16 vid = 0;
4247
4248         if (vlan_attr) {
4249                 if (nla_len(vlan_attr) != sizeof(u16)) {
4250                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4251                         return -EINVAL;
4252                 }
4253
4254                 vid = nla_get_u16(vlan_attr);
4255
4256                 if (!vid || vid >= VLAN_VID_MASK) {
4257                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4258                         return -EINVAL;
4259                 }
4260         }
4261         *p_vid = vid;
4262         return 0;
4263 }
4264
4265 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4266                         struct netlink_ext_ack *extack)
4267 {
4268         struct net *net = sock_net(skb->sk);
4269         struct ndmsg *ndm;
4270         struct nlattr *tb[NDA_MAX+1];
4271         struct net_device *dev;
4272         u8 *addr;
4273         u16 vid;
4274         int err;
4275
4276         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4277                                      extack);
4278         if (err < 0)
4279                 return err;
4280
4281         ndm = nlmsg_data(nlh);
4282         if (ndm->ndm_ifindex == 0) {
4283                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4284                 return -EINVAL;
4285         }
4286
4287         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4288         if (dev == NULL) {
4289                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4290                 return -ENODEV;
4291         }
4292
4293         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4294                 NL_SET_ERR_MSG(extack, "invalid address");
4295                 return -EINVAL;
4296         }
4297
4298         if (dev->type != ARPHRD_ETHER) {
4299                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4300                 return -EINVAL;
4301         }
4302
4303         addr = nla_data(tb[NDA_LLADDR]);
4304
4305         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4306         if (err)
4307                 return err;
4308
4309         err = -EOPNOTSUPP;
4310
4311         /* Support fdb on master device the net/bridge default case */
4312         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4313             netif_is_bridge_port(dev)) {
4314                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4315                 const struct net_device_ops *ops = br_dev->netdev_ops;
4316
4317                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4318                                        nlh->nlmsg_flags, extack);
4319                 if (err)
4320                         goto out;
4321                 else
4322                         ndm->ndm_flags &= ~NTF_MASTER;
4323         }
4324
4325         /* Embedded bridge, macvlan, and any other device support */
4326         if ((ndm->ndm_flags & NTF_SELF)) {
4327                 if (dev->netdev_ops->ndo_fdb_add)
4328                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4329                                                            vid,
4330                                                            nlh->nlmsg_flags,
4331                                                            extack);
4332                 else
4333                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4334                                                nlh->nlmsg_flags);
4335
4336                 if (!err) {
4337                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4338                                         ndm->ndm_state);
4339                         ndm->ndm_flags &= ~NTF_SELF;
4340                 }
4341         }
4342 out:
4343         return err;
4344 }
4345
4346 /*
4347  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4348  */
4349 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4350                      struct nlattr *tb[],
4351                      struct net_device *dev,
4352                      const unsigned char *addr, u16 vid)
4353 {
4354         int err = -EINVAL;
4355
4356         /* If aging addresses are supported device will need to
4357          * implement its own handler for this.
4358          */
4359         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4360                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4361                 return err;
4362         }
4363
4364         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4365                 err = dev_uc_del(dev, addr);
4366         else if (is_multicast_ether_addr(addr))
4367                 err = dev_mc_del(dev, addr);
4368
4369         return err;
4370 }
4371 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4372
4373 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4374                         struct netlink_ext_ack *extack)
4375 {
4376         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4377         struct net *net = sock_net(skb->sk);
4378         const struct net_device_ops *ops;
4379         struct ndmsg *ndm;
4380         struct nlattr *tb[NDA_MAX+1];
4381         struct net_device *dev;
4382         __u8 *addr = NULL;
4383         int err;
4384         u16 vid;
4385
4386         if (!netlink_capable(skb, CAP_NET_ADMIN))
4387                 return -EPERM;
4388
4389         if (!del_bulk) {
4390                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4391                                              NULL, extack);
4392         } else {
4393                 /* For bulk delete, the drivers will parse the message with
4394                  * policy.
4395                  */
4396                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
4397         }
4398         if (err < 0)
4399                 return err;
4400
4401         ndm = nlmsg_data(nlh);
4402         if (ndm->ndm_ifindex == 0) {
4403                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4404                 return -EINVAL;
4405         }
4406
4407         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4408         if (dev == NULL) {
4409                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4410                 return -ENODEV;
4411         }
4412
4413         if (!del_bulk) {
4414                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4415                         NL_SET_ERR_MSG(extack, "invalid address");
4416                         return -EINVAL;
4417                 }
4418                 addr = nla_data(tb[NDA_LLADDR]);
4419
4420                 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4421                 if (err)
4422                         return err;
4423         }
4424
4425         if (dev->type != ARPHRD_ETHER) {
4426                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4427                 return -EINVAL;
4428         }
4429
4430         err = -EOPNOTSUPP;
4431
4432         /* Support fdb on master device the net/bridge default case */
4433         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4434             netif_is_bridge_port(dev)) {
4435                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4436
4437                 ops = br_dev->netdev_ops;
4438                 if (!del_bulk) {
4439                         if (ops->ndo_fdb_del)
4440                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4441                 } else {
4442                         if (ops->ndo_fdb_del_bulk)
4443                                 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4444                 }
4445
4446                 if (err)
4447                         goto out;
4448                 else
4449                         ndm->ndm_flags &= ~NTF_MASTER;
4450         }
4451
4452         /* Embedded bridge, macvlan, and any other device support */
4453         if (ndm->ndm_flags & NTF_SELF) {
4454                 ops = dev->netdev_ops;
4455                 if (!del_bulk) {
4456                         if (ops->ndo_fdb_del)
4457                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4458                         else
4459                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4460                 } else {
4461                         /* in case err was cleared by NTF_MASTER call */
4462                         err = -EOPNOTSUPP;
4463                         if (ops->ndo_fdb_del_bulk)
4464                                 err = ops->ndo_fdb_del_bulk(nlh, dev, extack);
4465                 }
4466
4467                 if (!err) {
4468                         if (!del_bulk)
4469                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4470                                                 ndm->ndm_state);
4471                         ndm->ndm_flags &= ~NTF_SELF;
4472                 }
4473         }
4474 out:
4475         return err;
4476 }
4477
4478 static int nlmsg_populate_fdb(struct sk_buff *skb,
4479                               struct netlink_callback *cb,
4480                               struct net_device *dev,
4481                               int *idx,
4482                               struct netdev_hw_addr_list *list)
4483 {
4484         struct netdev_hw_addr *ha;
4485         int err;
4486         u32 portid, seq;
4487
4488         portid = NETLINK_CB(cb->skb).portid;
4489         seq = cb->nlh->nlmsg_seq;
4490
4491         list_for_each_entry(ha, &list->list, list) {
4492                 if (*idx < cb->args[2])
4493                         goto skip;
4494
4495                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4496                                               portid, seq,
4497                                               RTM_NEWNEIGH, NTF_SELF,
4498                                               NLM_F_MULTI, NUD_PERMANENT);
4499                 if (err < 0)
4500                         return err;
4501 skip:
4502                 *idx += 1;
4503         }
4504         return 0;
4505 }
4506
4507 /**
4508  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4509  * @skb: socket buffer to store message in
4510  * @cb: netlink callback
4511  * @dev: netdevice
4512  * @filter_dev: ignored
4513  * @idx: the number of FDB table entries dumped is added to *@idx
4514  *
4515  * Default netdevice operation to dump the existing unicast address list.
4516  * Returns number of addresses from list put in skb.
4517  */
4518 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4519                       struct netlink_callback *cb,
4520                       struct net_device *dev,
4521                       struct net_device *filter_dev,
4522                       int *idx)
4523 {
4524         int err;
4525
4526         if (dev->type != ARPHRD_ETHER)
4527                 return -EINVAL;
4528
4529         netif_addr_lock_bh(dev);
4530         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4531         if (err)
4532                 goto out;
4533         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4534 out:
4535         netif_addr_unlock_bh(dev);
4536         return err;
4537 }
4538 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4539
4540 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4541                                  int *br_idx, int *brport_idx,
4542                                  struct netlink_ext_ack *extack)
4543 {
4544         struct nlattr *tb[NDA_MAX + 1];
4545         struct ndmsg *ndm;
4546         int err, i;
4547
4548         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4549                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4550                 return -EINVAL;
4551         }
4552
4553         ndm = nlmsg_data(nlh);
4554         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4555             ndm->ndm_flags || ndm->ndm_type) {
4556                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4557                 return -EINVAL;
4558         }
4559
4560         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4561                                             NDA_MAX, NULL, extack);
4562         if (err < 0)
4563                 return err;
4564
4565         *brport_idx = ndm->ndm_ifindex;
4566         for (i = 0; i <= NDA_MAX; ++i) {
4567                 if (!tb[i])
4568                         continue;
4569
4570                 switch (i) {
4571                 case NDA_IFINDEX:
4572                         if (nla_len(tb[i]) != sizeof(u32)) {
4573                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4574                                 return -EINVAL;
4575                         }
4576                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4577                         break;
4578                 case NDA_MASTER:
4579                         if (nla_len(tb[i]) != sizeof(u32)) {
4580                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4581                                 return -EINVAL;
4582                         }
4583                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4584                         break;
4585                 default:
4586                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4587                         return -EINVAL;
4588                 }
4589         }
4590
4591         return 0;
4592 }
4593
4594 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4595                                  int *br_idx, int *brport_idx,
4596                                  struct netlink_ext_ack *extack)
4597 {
4598         struct nlattr *tb[IFLA_MAX+1];
4599         int err;
4600
4601         /* A hack to preserve kernel<->userspace interface.
4602          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4603          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4604          * So, check for ndmsg with an optional u32 attribute (not used here).
4605          * Fortunately these sizes don't conflict with the size of ifinfomsg
4606          * with an optional attribute.
4607          */
4608         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4609             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4610              nla_attr_size(sizeof(u32)))) {
4611                 struct ifinfomsg *ifm;
4612
4613                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4614                                              tb, IFLA_MAX, ifla_policy,
4615                                              extack);
4616                 if (err < 0) {
4617                         return -EINVAL;
4618                 } else if (err == 0) {
4619                         if (tb[IFLA_MASTER])
4620                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4621                 }
4622
4623                 ifm = nlmsg_data(nlh);
4624                 *brport_idx = ifm->ifi_index;
4625         }
4626         return 0;
4627 }
4628
4629 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4630 {
4631         struct net_device *dev;
4632         struct net_device *br_dev = NULL;
4633         const struct net_device_ops *ops = NULL;
4634         const struct net_device_ops *cops = NULL;
4635         struct net *net = sock_net(skb->sk);
4636         struct hlist_head *head;
4637         int brport_idx = 0;
4638         int br_idx = 0;
4639         int h, s_h;
4640         int idx = 0, s_idx;
4641         int err = 0;
4642         int fidx = 0;
4643
4644         if (cb->strict_check)
4645                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4646                                             cb->extack);
4647         else
4648                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4649                                             cb->extack);
4650         if (err < 0)
4651                 return err;
4652
4653         if (br_idx) {
4654                 br_dev = __dev_get_by_index(net, br_idx);
4655                 if (!br_dev)
4656                         return -ENODEV;
4657
4658                 ops = br_dev->netdev_ops;
4659         }
4660
4661         s_h = cb->args[0];
4662         s_idx = cb->args[1];
4663
4664         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4665                 idx = 0;
4666                 head = &net->dev_index_head[h];
4667                 hlist_for_each_entry(dev, head, index_hlist) {
4668
4669                         if (brport_idx && (dev->ifindex != brport_idx))
4670                                 continue;
4671
4672                         if (!br_idx) { /* user did not specify a specific bridge */
4673                                 if (netif_is_bridge_port(dev)) {
4674                                         br_dev = netdev_master_upper_dev_get(dev);
4675                                         cops = br_dev->netdev_ops;
4676                                 }
4677                         } else {
4678                                 if (dev != br_dev &&
4679                                     !netif_is_bridge_port(dev))
4680                                         continue;
4681
4682                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4683                                     !netif_is_bridge_master(dev))
4684                                         continue;
4685                                 cops = ops;
4686                         }
4687
4688                         if (idx < s_idx)
4689                                 goto cont;
4690
4691                         if (netif_is_bridge_port(dev)) {
4692                                 if (cops && cops->ndo_fdb_dump) {
4693                                         err = cops->ndo_fdb_dump(skb, cb,
4694                                                                 br_dev, dev,
4695                                                                 &fidx);
4696                                         if (err == -EMSGSIZE)
4697                                                 goto out;
4698                                 }
4699                         }
4700
4701                         if (dev->netdev_ops->ndo_fdb_dump)
4702                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4703                                                                     dev, NULL,
4704                                                                     &fidx);
4705                         else
4706                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4707                                                         &fidx);
4708                         if (err == -EMSGSIZE)
4709                                 goto out;
4710
4711                         cops = NULL;
4712
4713                         /* reset fdb offset to 0 for rest of the interfaces */
4714                         cb->args[2] = 0;
4715                         fidx = 0;
4716 cont:
4717                         idx++;
4718                 }
4719         }
4720
4721 out:
4722         cb->args[0] = h;
4723         cb->args[1] = idx;
4724         cb->args[2] = fidx;
4725
4726         return skb->len;
4727 }
4728
4729 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4730                                 struct nlattr **tb, u8 *ndm_flags,
4731                                 int *br_idx, int *brport_idx, u8 **addr,
4732                                 u16 *vid, struct netlink_ext_ack *extack)
4733 {
4734         struct ndmsg *ndm;
4735         int err, i;
4736
4737         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4738                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4739                 return -EINVAL;
4740         }
4741
4742         ndm = nlmsg_data(nlh);
4743         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4744             ndm->ndm_type) {
4745                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4746                 return -EINVAL;
4747         }
4748
4749         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4750                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4751                 return -EINVAL;
4752         }
4753
4754         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4755                                             NDA_MAX, nda_policy, extack);
4756         if (err < 0)
4757                 return err;
4758
4759         *ndm_flags = ndm->ndm_flags;
4760         *brport_idx = ndm->ndm_ifindex;
4761         for (i = 0; i <= NDA_MAX; ++i) {
4762                 if (!tb[i])
4763                         continue;
4764
4765                 switch (i) {
4766                 case NDA_MASTER:
4767                         *br_idx = nla_get_u32(tb[i]);
4768                         break;
4769                 case NDA_LLADDR:
4770                         if (nla_len(tb[i]) != ETH_ALEN) {
4771                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4772                                 return -EINVAL;
4773                         }
4774                         *addr = nla_data(tb[i]);
4775                         break;
4776                 case NDA_VLAN:
4777                         err = fdb_vid_parse(tb[i], vid, extack);
4778                         if (err)
4779                                 return err;
4780                         break;
4781                 case NDA_VNI:
4782                         break;
4783                 default:
4784                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4785                         return -EINVAL;
4786                 }
4787         }
4788
4789         return 0;
4790 }
4791
4792 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4793                         struct netlink_ext_ack *extack)
4794 {
4795         struct net_device *dev = NULL, *br_dev = NULL;
4796         const struct net_device_ops *ops = NULL;
4797         struct net *net = sock_net(in_skb->sk);
4798         struct nlattr *tb[NDA_MAX + 1];
4799         struct sk_buff *skb;
4800         int brport_idx = 0;
4801         u8 ndm_flags = 0;
4802         int br_idx = 0;
4803         u8 *addr = NULL;
4804         u16 vid = 0;
4805         int err;
4806
4807         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4808                                    &brport_idx, &addr, &vid, extack);
4809         if (err < 0)
4810                 return err;
4811
4812         if (!addr) {
4813                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4814                 return -EINVAL;
4815         }
4816
4817         if (brport_idx) {
4818                 dev = __dev_get_by_index(net, brport_idx);
4819                 if (!dev) {
4820                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4821                         return -ENODEV;
4822                 }
4823         }
4824
4825         if (br_idx) {
4826                 if (dev) {
4827                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4828                         return -EINVAL;
4829                 }
4830
4831                 br_dev = __dev_get_by_index(net, br_idx);
4832                 if (!br_dev) {
4833                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4834                         return -EINVAL;
4835                 }
4836                 ops = br_dev->netdev_ops;
4837         }
4838
4839         if (dev) {
4840                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4841                         if (!netif_is_bridge_port(dev)) {
4842                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4843                                 return -EINVAL;
4844                         }
4845                         br_dev = netdev_master_upper_dev_get(dev);
4846                         if (!br_dev) {
4847                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4848                                 return -EINVAL;
4849                         }
4850                         ops = br_dev->netdev_ops;
4851                 } else {
4852                         if (!(ndm_flags & NTF_SELF)) {
4853                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4854                                 return -EINVAL;
4855                         }
4856                         ops = dev->netdev_ops;
4857                 }
4858         }
4859
4860         if (!br_dev && !dev) {
4861                 NL_SET_ERR_MSG(extack, "No device specified");
4862                 return -ENODEV;
4863         }
4864
4865         if (!ops || !ops->ndo_fdb_get) {
4866                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4867                 return -EOPNOTSUPP;
4868         }
4869
4870         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4871         if (!skb)
4872                 return -ENOBUFS;
4873
4874         if (br_dev)
4875                 dev = br_dev;
4876         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4877                                NETLINK_CB(in_skb).portid,
4878                                nlh->nlmsg_seq, extack);
4879         if (err)
4880                 goto out;
4881
4882         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4883 out:
4884         kfree_skb(skb);
4885         return err;
4886 }
4887
4888 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4889                                unsigned int attrnum, unsigned int flag)
4890 {
4891         if (mask & flag)
4892                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4893         return 0;
4894 }
4895
4896 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4897                             struct net_device *dev, u16 mode,
4898                             u32 flags, u32 mask, int nlflags,
4899                             u32 filter_mask,
4900                             int (*vlan_fill)(struct sk_buff *skb,
4901                                              struct net_device *dev,
4902                                              u32 filter_mask))
4903 {
4904         struct nlmsghdr *nlh;
4905         struct ifinfomsg *ifm;
4906         struct nlattr *br_afspec;
4907         struct nlattr *protinfo;
4908         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4909         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4910         int err = 0;
4911
4912         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4913         if (nlh == NULL)
4914                 return -EMSGSIZE;
4915
4916         ifm = nlmsg_data(nlh);
4917         ifm->ifi_family = AF_BRIDGE;
4918         ifm->__ifi_pad = 0;
4919         ifm->ifi_type = dev->type;
4920         ifm->ifi_index = dev->ifindex;
4921         ifm->ifi_flags = dev_get_flags(dev);
4922         ifm->ifi_change = 0;
4923
4924
4925         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4926             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4927             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4928             (br_dev &&
4929              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4930             (dev->addr_len &&
4931              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4932             (dev->ifindex != dev_get_iflink(dev) &&
4933              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4934                 goto nla_put_failure;
4935
4936         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4937         if (!br_afspec)
4938                 goto nla_put_failure;
4939
4940         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4941                 nla_nest_cancel(skb, br_afspec);
4942                 goto nla_put_failure;
4943         }
4944
4945         if (mode != BRIDGE_MODE_UNDEF) {
4946                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4947                         nla_nest_cancel(skb, br_afspec);
4948                         goto nla_put_failure;
4949                 }
4950         }
4951         if (vlan_fill) {
4952                 err = vlan_fill(skb, dev, filter_mask);
4953                 if (err) {
4954                         nla_nest_cancel(skb, br_afspec);
4955                         goto nla_put_failure;
4956                 }
4957         }
4958         nla_nest_end(skb, br_afspec);
4959
4960         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4961         if (!protinfo)
4962                 goto nla_put_failure;
4963
4964         if (brport_nla_put_flag(skb, flags, mask,
4965                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4966             brport_nla_put_flag(skb, flags, mask,
4967                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4968             brport_nla_put_flag(skb, flags, mask,
4969                                 IFLA_BRPORT_FAST_LEAVE,
4970                                 BR_MULTICAST_FAST_LEAVE) ||
4971             brport_nla_put_flag(skb, flags, mask,
4972                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4973             brport_nla_put_flag(skb, flags, mask,
4974                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4975             brport_nla_put_flag(skb, flags, mask,
4976                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4977             brport_nla_put_flag(skb, flags, mask,
4978                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4979             brport_nla_put_flag(skb, flags, mask,
4980                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4981             brport_nla_put_flag(skb, flags, mask,
4982                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4983             brport_nla_put_flag(skb, flags, mask,
4984                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4985                 nla_nest_cancel(skb, protinfo);
4986                 goto nla_put_failure;
4987         }
4988
4989         nla_nest_end(skb, protinfo);
4990
4991         nlmsg_end(skb, nlh);
4992         return 0;
4993 nla_put_failure:
4994         nlmsg_cancel(skb, nlh);
4995         return err ? err : -EMSGSIZE;
4996 }
4997 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4998
4999 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
5000                                     bool strict_check, u32 *filter_mask,
5001                                     struct netlink_ext_ack *extack)
5002 {
5003         struct nlattr *tb[IFLA_MAX+1];
5004         int err, i;
5005
5006         if (strict_check) {
5007                 struct ifinfomsg *ifm;
5008
5009                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
5010                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
5011                         return -EINVAL;
5012                 }
5013
5014                 ifm = nlmsg_data(nlh);
5015                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
5016                     ifm->ifi_change || ifm->ifi_index) {
5017                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
5018                         return -EINVAL;
5019                 }
5020
5021                 err = nlmsg_parse_deprecated_strict(nlh,
5022                                                     sizeof(struct ifinfomsg),
5023                                                     tb, IFLA_MAX, ifla_policy,
5024                                                     extack);
5025         } else {
5026                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
5027                                              tb, IFLA_MAX, ifla_policy,
5028                                              extack);
5029         }
5030         if (err < 0)
5031                 return err;
5032
5033         /* new attributes should only be added with strict checking */
5034         for (i = 0; i <= IFLA_MAX; ++i) {
5035                 if (!tb[i])
5036                         continue;
5037
5038                 switch (i) {
5039                 case IFLA_EXT_MASK:
5040                         *filter_mask = nla_get_u32(tb[i]);
5041                         break;
5042                 default:
5043                         if (strict_check) {
5044                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
5045                                 return -EINVAL;
5046                         }
5047                 }
5048         }
5049
5050         return 0;
5051 }
5052
5053 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5054 {
5055         const struct nlmsghdr *nlh = cb->nlh;
5056         struct net *net = sock_net(skb->sk);
5057         struct net_device *dev;
5058         int idx = 0;
5059         u32 portid = NETLINK_CB(cb->skb).portid;
5060         u32 seq = nlh->nlmsg_seq;
5061         u32 filter_mask = 0;
5062         int err;
5063
5064         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5065                                        cb->extack);
5066         if (err < 0 && cb->strict_check)
5067                 return err;
5068
5069         rcu_read_lock();
5070         for_each_netdev_rcu(net, dev) {
5071                 const struct net_device_ops *ops = dev->netdev_ops;
5072                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5073
5074                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5075                         if (idx >= cb->args[0]) {
5076                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5077                                                 skb, portid, seq, dev,
5078                                                 filter_mask, NLM_F_MULTI);
5079                                 if (err < 0 && err != -EOPNOTSUPP) {
5080                                         if (likely(skb->len))
5081                                                 break;
5082
5083                                         goto out_err;
5084                                 }
5085                         }
5086                         idx++;
5087                 }
5088
5089                 if (ops->ndo_bridge_getlink) {
5090                         if (idx >= cb->args[0]) {
5091                                 err = ops->ndo_bridge_getlink(skb, portid,
5092                                                               seq, dev,
5093                                                               filter_mask,
5094                                                               NLM_F_MULTI);
5095                                 if (err < 0 && err != -EOPNOTSUPP) {
5096                                         if (likely(skb->len))
5097                                                 break;
5098
5099                                         goto out_err;
5100                                 }
5101                         }
5102                         idx++;
5103                 }
5104         }
5105         err = skb->len;
5106 out_err:
5107         rcu_read_unlock();
5108         cb->args[0] = idx;
5109
5110         return err;
5111 }
5112
5113 static inline size_t bridge_nlmsg_size(void)
5114 {
5115         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5116                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5117                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5118                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5119                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5120                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5121                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5122                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5123                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5124                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5125                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5126 }
5127
5128 static int rtnl_bridge_notify(struct net_device *dev)
5129 {
5130         struct net *net = dev_net(dev);
5131         struct sk_buff *skb;
5132         int err = -EOPNOTSUPP;
5133
5134         if (!dev->netdev_ops->ndo_bridge_getlink)
5135                 return 0;
5136
5137         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5138         if (!skb) {
5139                 err = -ENOMEM;
5140                 goto errout;
5141         }
5142
5143         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5144         if (err < 0)
5145                 goto errout;
5146
5147         /* Notification info is only filled for bridge ports, not the bridge
5148          * device itself. Therefore, a zero notification length is valid and
5149          * should not result in an error.
5150          */
5151         if (!skb->len)
5152                 goto errout;
5153
5154         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5155         return 0;
5156 errout:
5157         WARN_ON(err == -EMSGSIZE);
5158         kfree_skb(skb);
5159         if (err)
5160                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5161         return err;
5162 }
5163
5164 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5165                                struct netlink_ext_ack *extack)
5166 {
5167         struct net *net = sock_net(skb->sk);
5168         struct ifinfomsg *ifm;
5169         struct net_device *dev;
5170         struct nlattr *br_spec, *attr, *br_flags_attr = NULL;
5171         int rem, err = -EOPNOTSUPP;
5172         u16 flags = 0;
5173
5174         if (nlmsg_len(nlh) < sizeof(*ifm))
5175                 return -EINVAL;
5176
5177         ifm = nlmsg_data(nlh);
5178         if (ifm->ifi_family != AF_BRIDGE)
5179                 return -EPFNOSUPPORT;
5180
5181         dev = __dev_get_by_index(net, ifm->ifi_index);
5182         if (!dev) {
5183                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5184                 return -ENODEV;
5185         }
5186
5187         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5188         if (br_spec) {
5189                 nla_for_each_nested(attr, br_spec, rem) {
5190                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !br_flags_attr) {
5191                                 if (nla_len(attr) < sizeof(flags))
5192                                         return -EINVAL;
5193
5194                                 br_flags_attr = attr;
5195                                 flags = nla_get_u16(attr);
5196                         }
5197
5198                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5199                                 if (nla_len(attr) < sizeof(u16))
5200                                         return -EINVAL;
5201                         }
5202                 }
5203         }
5204
5205         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5206                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5207
5208                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5209                         err = -EOPNOTSUPP;
5210                         goto out;
5211                 }
5212
5213                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5214                                                              extack);
5215                 if (err)
5216                         goto out;
5217
5218                 flags &= ~BRIDGE_FLAGS_MASTER;
5219         }
5220
5221         if ((flags & BRIDGE_FLAGS_SELF)) {
5222                 if (!dev->netdev_ops->ndo_bridge_setlink)
5223                         err = -EOPNOTSUPP;
5224                 else
5225                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5226                                                                   flags,
5227                                                                   extack);
5228                 if (!err) {
5229                         flags &= ~BRIDGE_FLAGS_SELF;
5230
5231                         /* Generate event to notify upper layer of bridge
5232                          * change
5233                          */
5234                         err = rtnl_bridge_notify(dev);
5235                 }
5236         }
5237
5238         if (br_flags_attr)
5239                 memcpy(nla_data(br_flags_attr), &flags, sizeof(flags));
5240 out:
5241         return err;
5242 }
5243
5244 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5245                                struct netlink_ext_ack *extack)
5246 {
5247         struct net *net = sock_net(skb->sk);
5248         struct ifinfomsg *ifm;
5249         struct net_device *dev;
5250         struct nlattr *br_spec, *attr = NULL;
5251         int rem, err = -EOPNOTSUPP;
5252         u16 flags = 0;
5253         bool have_flags = false;
5254
5255         if (nlmsg_len(nlh) < sizeof(*ifm))
5256                 return -EINVAL;
5257
5258         ifm = nlmsg_data(nlh);
5259         if (ifm->ifi_family != AF_BRIDGE)
5260                 return -EPFNOSUPPORT;
5261
5262         dev = __dev_get_by_index(net, ifm->ifi_index);
5263         if (!dev) {
5264                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5265                 return -ENODEV;
5266         }
5267
5268         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5269         if (br_spec) {
5270                 nla_for_each_nested_type(attr, IFLA_BRIDGE_FLAGS, br_spec,
5271                                          rem) {
5272                         if (nla_len(attr) < sizeof(flags))
5273                                 return -EINVAL;
5274
5275                         have_flags = true;
5276                         flags = nla_get_u16(attr);
5277                         break;
5278                 }
5279         }
5280
5281         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5282                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5283
5284                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5285                         err = -EOPNOTSUPP;
5286                         goto out;
5287                 }
5288
5289                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5290                 if (err)
5291                         goto out;
5292
5293                 flags &= ~BRIDGE_FLAGS_MASTER;
5294         }
5295
5296         if ((flags & BRIDGE_FLAGS_SELF)) {
5297                 if (!dev->netdev_ops->ndo_bridge_dellink)
5298                         err = -EOPNOTSUPP;
5299                 else
5300                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5301                                                                   flags);
5302
5303                 if (!err) {
5304                         flags &= ~BRIDGE_FLAGS_SELF;
5305
5306                         /* Generate event to notify upper layer of bridge
5307                          * change
5308                          */
5309                         err = rtnl_bridge_notify(dev);
5310                 }
5311         }
5312
5313         if (have_flags)
5314                 memcpy(nla_data(attr), &flags, sizeof(flags));
5315 out:
5316         return err;
5317 }
5318
5319 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5320 {
5321         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5322                (!idxattr || idxattr == attrid);
5323 }
5324
5325 static bool
5326 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5327 {
5328         return dev->netdev_ops &&
5329                dev->netdev_ops->ndo_has_offload_stats &&
5330                dev->netdev_ops->ndo_get_offload_stats &&
5331                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5332 }
5333
5334 static unsigned int
5335 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5336 {
5337         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5338                sizeof(struct rtnl_link_stats64) : 0;
5339 }
5340
5341 static int
5342 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5343                              struct sk_buff *skb)
5344 {
5345         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5346         struct nlattr *attr = NULL;
5347         void *attr_data;
5348         int err;
5349
5350         if (!size)
5351                 return -ENODATA;
5352
5353         attr = nla_reserve_64bit(skb, attr_id, size,
5354                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5355         if (!attr)
5356                 return -EMSGSIZE;
5357
5358         attr_data = nla_data(attr);
5359         memset(attr_data, 0, size);
5360
5361         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5362         if (err)
5363                 return err;
5364
5365         return 0;
5366 }
5367
5368 static unsigned int
5369 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5370                                    enum netdev_offload_xstats_type type)
5371 {
5372         bool enabled = netdev_offload_xstats_enabled(dev, type);
5373
5374         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5375 }
5376
5377 struct rtnl_offload_xstats_request_used {
5378         bool request;
5379         bool used;
5380 };
5381
5382 static int
5383 rtnl_offload_xstats_get_stats(struct net_device *dev,
5384                               enum netdev_offload_xstats_type type,
5385                               struct rtnl_offload_xstats_request_used *ru,
5386                               struct rtnl_hw_stats64 *stats,
5387                               struct netlink_ext_ack *extack)
5388 {
5389         bool request;
5390         bool used;
5391         int err;
5392
5393         request = netdev_offload_xstats_enabled(dev, type);
5394         if (!request) {
5395                 used = false;
5396                 goto out;
5397         }
5398
5399         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5400         if (err)
5401                 return err;
5402
5403 out:
5404         if (ru) {
5405                 ru->request = request;
5406                 ru->used = used;
5407         }
5408         return 0;
5409 }
5410
5411 static int
5412 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5413                                        struct rtnl_offload_xstats_request_used *ru)
5414 {
5415         struct nlattr *nest;
5416
5417         nest = nla_nest_start(skb, attr_id);
5418         if (!nest)
5419                 return -EMSGSIZE;
5420
5421         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5422                 goto nla_put_failure;
5423
5424         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5425                 goto nla_put_failure;
5426
5427         nla_nest_end(skb, nest);
5428         return 0;
5429
5430 nla_put_failure:
5431         nla_nest_cancel(skb, nest);
5432         return -EMSGSIZE;
5433 }
5434
5435 static int
5436 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5437                                    struct netlink_ext_ack *extack)
5438 {
5439         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5440         struct rtnl_offload_xstats_request_used ru_l3;
5441         struct nlattr *nest;
5442         int err;
5443
5444         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5445         if (err)
5446                 return err;
5447
5448         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5449         if (!nest)
5450                 return -EMSGSIZE;
5451
5452         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5453                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5454                                                    &ru_l3))
5455                 goto nla_put_failure;
5456
5457         nla_nest_end(skb, nest);
5458         return 0;
5459
5460 nla_put_failure:
5461         nla_nest_cancel(skb, nest);
5462         return -EMSGSIZE;
5463 }
5464
5465 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5466                                     int *prividx, u32 off_filter_mask,
5467                                     struct netlink_ext_ack *extack)
5468 {
5469         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5470         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5471         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5472         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5473         bool have_data = false;
5474         int err;
5475
5476         if (*prividx <= attr_id_cpu_hit &&
5477             (off_filter_mask &
5478              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5479                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5480                 if (!err) {
5481                         have_data = true;
5482                 } else if (err != -ENODATA) {
5483                         *prividx = attr_id_cpu_hit;
5484                         return err;
5485                 }
5486         }
5487
5488         if (*prividx <= attr_id_hw_s_info &&
5489             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5490                 *prividx = attr_id_hw_s_info;
5491
5492                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5493                 if (err)
5494                         return err;
5495
5496                 have_data = true;
5497                 *prividx = 0;
5498         }
5499
5500         if (*prividx <= attr_id_l3_stats &&
5501             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5502                 unsigned int size_l3;
5503                 struct nlattr *attr;
5504
5505                 *prividx = attr_id_l3_stats;
5506
5507                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5508                 if (!size_l3)
5509                         goto skip_l3_stats;
5510                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5511                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5512                 if (!attr)
5513                         return -EMSGSIZE;
5514
5515                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5516                                                     nla_data(attr), extack);
5517                 if (err)
5518                         return err;
5519
5520                 have_data = true;
5521 skip_l3_stats:
5522                 *prividx = 0;
5523         }
5524
5525         if (!have_data)
5526                 return -ENODATA;
5527
5528         *prividx = 0;
5529         return 0;
5530 }
5531
5532 static unsigned int
5533 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5534                                            enum netdev_offload_xstats_type type)
5535 {
5536         return nla_total_size(0) +
5537                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5538                 nla_total_size(sizeof(u8)) +
5539                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5540                 nla_total_size(sizeof(u8)) +
5541                 0;
5542 }
5543
5544 static unsigned int
5545 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5546 {
5547         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5548
5549         return nla_total_size(0) +
5550                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5551                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5552                 0;
5553 }
5554
5555 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5556                                         u32 off_filter_mask)
5557 {
5558         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5559         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5560         int nla_size = 0;
5561         int size;
5562
5563         if (off_filter_mask &
5564             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5565                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5566                 nla_size += nla_total_size_64bit(size);
5567         }
5568
5569         if (off_filter_mask &
5570             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5571                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5572
5573         if (off_filter_mask &
5574             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5575                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5576                 nla_size += nla_total_size_64bit(size);
5577         }
5578
5579         if (nla_size != 0)
5580                 nla_size += nla_total_size(0);
5581
5582         return nla_size;
5583 }
5584
5585 struct rtnl_stats_dump_filters {
5586         /* mask[0] filters outer attributes. Then individual nests have their
5587          * filtering mask at the index of the nested attribute.
5588          */
5589         u32 mask[IFLA_STATS_MAX + 1];
5590 };
5591
5592 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5593                                int type, u32 pid, u32 seq, u32 change,
5594                                unsigned int flags,
5595                                const struct rtnl_stats_dump_filters *filters,
5596                                int *idxattr, int *prividx,
5597                                struct netlink_ext_ack *extack)
5598 {
5599         unsigned int filter_mask = filters->mask[0];
5600         struct if_stats_msg *ifsm;
5601         struct nlmsghdr *nlh;
5602         struct nlattr *attr;
5603         int s_prividx = *prividx;
5604         int err;
5605
5606         ASSERT_RTNL();
5607
5608         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5609         if (!nlh)
5610                 return -EMSGSIZE;
5611
5612         ifsm = nlmsg_data(nlh);
5613         ifsm->family = PF_UNSPEC;
5614         ifsm->pad1 = 0;
5615         ifsm->pad2 = 0;
5616         ifsm->ifindex = dev->ifindex;
5617         ifsm->filter_mask = filter_mask;
5618
5619         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5620                 struct rtnl_link_stats64 *sp;
5621
5622                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5623                                          sizeof(struct rtnl_link_stats64),
5624                                          IFLA_STATS_UNSPEC);
5625                 if (!attr) {
5626                         err = -EMSGSIZE;
5627                         goto nla_put_failure;
5628                 }
5629
5630                 sp = nla_data(attr);
5631                 dev_get_stats(dev, sp);
5632         }
5633
5634         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5635                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5636
5637                 if (ops && ops->fill_linkxstats) {
5638                         *idxattr = IFLA_STATS_LINK_XSTATS;
5639                         attr = nla_nest_start_noflag(skb,
5640                                                      IFLA_STATS_LINK_XSTATS);
5641                         if (!attr) {
5642                                 err = -EMSGSIZE;
5643                                 goto nla_put_failure;
5644                         }
5645
5646                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5647                         nla_nest_end(skb, attr);
5648                         if (err)
5649                                 goto nla_put_failure;
5650                         *idxattr = 0;
5651                 }
5652         }
5653
5654         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5655                              *idxattr)) {
5656                 const struct rtnl_link_ops *ops = NULL;
5657                 const struct net_device *master;
5658
5659                 master = netdev_master_upper_dev_get(dev);
5660                 if (master)
5661                         ops = master->rtnl_link_ops;
5662                 if (ops && ops->fill_linkxstats) {
5663                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5664                         attr = nla_nest_start_noflag(skb,
5665                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5666                         if (!attr) {
5667                                 err = -EMSGSIZE;
5668                                 goto nla_put_failure;
5669                         }
5670
5671                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5672                         nla_nest_end(skb, attr);
5673                         if (err)
5674                                 goto nla_put_failure;
5675                         *idxattr = 0;
5676                 }
5677         }
5678
5679         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5680                              *idxattr)) {
5681                 u32 off_filter_mask;
5682
5683                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5684                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5685                 attr = nla_nest_start_noflag(skb,
5686                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5687                 if (!attr) {
5688                         err = -EMSGSIZE;
5689                         goto nla_put_failure;
5690                 }
5691
5692                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5693                                                off_filter_mask, extack);
5694                 if (err == -ENODATA)
5695                         nla_nest_cancel(skb, attr);
5696                 else
5697                         nla_nest_end(skb, attr);
5698
5699                 if (err && err != -ENODATA)
5700                         goto nla_put_failure;
5701                 *idxattr = 0;
5702         }
5703
5704         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5705                 struct rtnl_af_ops *af_ops;
5706
5707                 *idxattr = IFLA_STATS_AF_SPEC;
5708                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5709                 if (!attr) {
5710                         err = -EMSGSIZE;
5711                         goto nla_put_failure;
5712                 }
5713
5714                 rcu_read_lock();
5715                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5716                         if (af_ops->fill_stats_af) {
5717                                 struct nlattr *af;
5718
5719                                 af = nla_nest_start_noflag(skb,
5720                                                            af_ops->family);
5721                                 if (!af) {
5722                                         rcu_read_unlock();
5723                                         err = -EMSGSIZE;
5724                                         goto nla_put_failure;
5725                                 }
5726                                 err = af_ops->fill_stats_af(skb, dev);
5727
5728                                 if (err == -ENODATA) {
5729                                         nla_nest_cancel(skb, af);
5730                                 } else if (err < 0) {
5731                                         rcu_read_unlock();
5732                                         goto nla_put_failure;
5733                                 }
5734
5735                                 nla_nest_end(skb, af);
5736                         }
5737                 }
5738                 rcu_read_unlock();
5739
5740                 nla_nest_end(skb, attr);
5741
5742                 *idxattr = 0;
5743         }
5744
5745         nlmsg_end(skb, nlh);
5746
5747         return 0;
5748
5749 nla_put_failure:
5750         /* not a multi message or no progress mean a real error */
5751         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5752                 nlmsg_cancel(skb, nlh);
5753         else
5754                 nlmsg_end(skb, nlh);
5755
5756         return err;
5757 }
5758
5759 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5760                                   const struct rtnl_stats_dump_filters *filters)
5761 {
5762         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5763         unsigned int filter_mask = filters->mask[0];
5764
5765         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5766                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5767
5768         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5769                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5770                 int attr = IFLA_STATS_LINK_XSTATS;
5771
5772                 if (ops && ops->get_linkxstats_size) {
5773                         size += nla_total_size(ops->get_linkxstats_size(dev,
5774                                                                         attr));
5775                         /* for IFLA_STATS_LINK_XSTATS */
5776                         size += nla_total_size(0);
5777                 }
5778         }
5779
5780         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5781                 struct net_device *_dev = (struct net_device *)dev;
5782                 const struct rtnl_link_ops *ops = NULL;
5783                 const struct net_device *master;
5784
5785                 /* netdev_master_upper_dev_get can't take const */
5786                 master = netdev_master_upper_dev_get(_dev);
5787                 if (master)
5788                         ops = master->rtnl_link_ops;
5789                 if (ops && ops->get_linkxstats_size) {
5790                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5791
5792                         size += nla_total_size(ops->get_linkxstats_size(dev,
5793                                                                         attr));
5794                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5795                         size += nla_total_size(0);
5796                 }
5797         }
5798
5799         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5800                 u32 off_filter_mask;
5801
5802                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5803                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5804         }
5805
5806         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5807                 struct rtnl_af_ops *af_ops;
5808
5809                 /* for IFLA_STATS_AF_SPEC */
5810                 size += nla_total_size(0);
5811
5812                 rcu_read_lock();
5813                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5814                         if (af_ops->get_stats_af_size) {
5815                                 size += nla_total_size(
5816                                         af_ops->get_stats_af_size(dev));
5817
5818                                 /* for AF_* */
5819                                 size += nla_total_size(0);
5820                         }
5821                 }
5822                 rcu_read_unlock();
5823         }
5824
5825         return size;
5826 }
5827
5828 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5829
5830 static const struct nla_policy
5831 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5832         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5833                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5834 };
5835
5836 static const struct nla_policy
5837 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5838         [IFLA_STATS_GET_FILTERS] =
5839                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5840 };
5841
5842 static const struct nla_policy
5843 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5844         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5845 };
5846
5847 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5848                                         struct rtnl_stats_dump_filters *filters,
5849                                         struct netlink_ext_ack *extack)
5850 {
5851         struct nlattr *tb[IFLA_STATS_MAX + 1];
5852         int err;
5853         int at;
5854
5855         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5856                                rtnl_stats_get_policy_filters, extack);
5857         if (err < 0)
5858                 return err;
5859
5860         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5861                 if (tb[at]) {
5862                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5863                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5864                                 return -EINVAL;
5865                         }
5866                         filters->mask[at] = nla_get_u32(tb[at]);
5867                 }
5868         }
5869
5870         return 0;
5871 }
5872
5873 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5874                                 u32 filter_mask,
5875                                 struct rtnl_stats_dump_filters *filters,
5876                                 struct netlink_ext_ack *extack)
5877 {
5878         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5879         int err;
5880         int i;
5881
5882         filters->mask[0] = filter_mask;
5883         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5884                 filters->mask[i] = -1U;
5885
5886         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5887                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5888         if (err < 0)
5889                 return err;
5890
5891         if (tb[IFLA_STATS_GET_FILTERS]) {
5892                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5893                                                    filters, extack);
5894                 if (err)
5895                         return err;
5896         }
5897
5898         return 0;
5899 }
5900
5901 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5902                                 bool is_dump, struct netlink_ext_ack *extack)
5903 {
5904         struct if_stats_msg *ifsm;
5905
5906         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5907                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5908                 return -EINVAL;
5909         }
5910
5911         if (!strict_check)
5912                 return 0;
5913
5914         ifsm = nlmsg_data(nlh);
5915
5916         /* only requests using strict checks can pass data to influence
5917          * the dump. The legacy exception is filter_mask.
5918          */
5919         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5920                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5921                 return -EINVAL;
5922         }
5923         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5924                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5925                 return -EINVAL;
5926         }
5927
5928         return 0;
5929 }
5930
5931 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5932                           struct netlink_ext_ack *extack)
5933 {
5934         struct rtnl_stats_dump_filters filters;
5935         struct net *net = sock_net(skb->sk);
5936         struct net_device *dev = NULL;
5937         int idxattr = 0, prividx = 0;
5938         struct if_stats_msg *ifsm;
5939         struct sk_buff *nskb;
5940         int err;
5941
5942         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5943                                    false, extack);
5944         if (err)
5945                 return err;
5946
5947         ifsm = nlmsg_data(nlh);
5948         if (ifsm->ifindex > 0)
5949                 dev = __dev_get_by_index(net, ifsm->ifindex);
5950         else
5951                 return -EINVAL;
5952
5953         if (!dev)
5954                 return -ENODEV;
5955
5956         if (!ifsm->filter_mask) {
5957                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5958                 return -EINVAL;
5959         }
5960
5961         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5962         if (err)
5963                 return err;
5964
5965         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5966         if (!nskb)
5967                 return -ENOBUFS;
5968
5969         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5970                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5971                                   0, &filters, &idxattr, &prividx, extack);
5972         if (err < 0) {
5973                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5974                 WARN_ON(err == -EMSGSIZE);
5975                 kfree_skb(nskb);
5976         } else {
5977                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5978         }
5979
5980         return err;
5981 }
5982
5983 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5984 {
5985         struct netlink_ext_ack *extack = cb->extack;
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 {
5991                 unsigned long ifindex;
5992                 int idxattr;
5993                 int prividx;
5994         } *ctx = (void *)cb->ctx;
5995         struct net_device *dev;
5996         int err;
5997
5998         cb->seq = net->dev_base_seq;
5999
6000         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
6001         if (err)
6002                 return err;
6003
6004         ifsm = nlmsg_data(cb->nlh);
6005         if (!ifsm->filter_mask) {
6006                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
6007                 return -EINVAL;
6008         }
6009
6010         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
6011                                    extack);
6012         if (err)
6013                 return err;
6014
6015         for_each_netdev_dump(net, dev, ctx->ifindex) {
6016                 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
6017                                           NETLINK_CB(cb->skb).portid,
6018                                           cb->nlh->nlmsg_seq, 0,
6019                                           flags, &filters,
6020                                           &ctx->idxattr, &ctx->prividx,
6021                                           extack);
6022                 /* If we ran out of room on the first message,
6023                  * we're in trouble.
6024                  */
6025                 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
6026
6027                 if (err < 0)
6028                         break;
6029                 ctx->prividx = 0;
6030                 ctx->idxattr = 0;
6031                 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
6032         }
6033
6034         return err;
6035 }
6036
6037 void rtnl_offload_xstats_notify(struct net_device *dev)
6038 {
6039         struct rtnl_stats_dump_filters response_filters = {};
6040         struct net *net = dev_net(dev);
6041         int idxattr = 0, prividx = 0;
6042         struct sk_buff *skb;
6043         int err = -ENOBUFS;
6044
6045         ASSERT_RTNL();
6046
6047         response_filters.mask[0] |=
6048                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6049         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6050                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6051
6052         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6053                         GFP_KERNEL);
6054         if (!skb)
6055                 goto errout;
6056
6057         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6058                                   &response_filters, &idxattr, &prividx, NULL);
6059         if (err < 0) {
6060                 kfree_skb(skb);
6061                 goto errout;
6062         }
6063
6064         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6065         return;
6066
6067 errout:
6068         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6069 }
6070 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6071
6072 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6073                           struct netlink_ext_ack *extack)
6074 {
6075         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6076         struct rtnl_stats_dump_filters response_filters = {};
6077         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6078         struct net *net = sock_net(skb->sk);
6079         struct net_device *dev = NULL;
6080         struct if_stats_msg *ifsm;
6081         bool notify = false;
6082         int err;
6083
6084         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6085                                    false, extack);
6086         if (err)
6087                 return err;
6088
6089         ifsm = nlmsg_data(nlh);
6090         if (ifsm->family != AF_UNSPEC) {
6091                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6092                 return -EINVAL;
6093         }
6094
6095         if (ifsm->ifindex > 0)
6096                 dev = __dev_get_by_index(net, ifsm->ifindex);
6097         else
6098                 return -EINVAL;
6099
6100         if (!dev)
6101                 return -ENODEV;
6102
6103         if (ifsm->filter_mask) {
6104                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6105                 return -EINVAL;
6106         }
6107
6108         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6109                           ifla_stats_set_policy, extack);
6110         if (err < 0)
6111                 return err;
6112
6113         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6114                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6115
6116                 if (req)
6117                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6118                 else
6119                         err = netdev_offload_xstats_disable(dev, t_l3);
6120
6121                 if (!err)
6122                         notify = true;
6123                 else if (err != -EALREADY)
6124                         return err;
6125
6126                 response_filters.mask[0] |=
6127                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6128                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6129                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6130         }
6131
6132         if (notify)
6133                 rtnl_offload_xstats_notify(dev);
6134
6135         return 0;
6136 }
6137
6138 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6139                                    struct netlink_ext_ack *extack)
6140 {
6141         struct br_port_msg *bpm;
6142
6143         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6144                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6145                 return -EINVAL;
6146         }
6147
6148         bpm = nlmsg_data(nlh);
6149         if (bpm->ifindex) {
6150                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6151                 return -EINVAL;
6152         }
6153         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6154                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6155                 return -EINVAL;
6156         }
6157
6158         return 0;
6159 }
6160
6161 struct rtnl_mdb_dump_ctx {
6162         long idx;
6163 };
6164
6165 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6166 {
6167         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6168         struct net *net = sock_net(skb->sk);
6169         struct net_device *dev;
6170         int idx, s_idx;
6171         int err;
6172
6173         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6174
6175         if (cb->strict_check) {
6176                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6177                 if (err)
6178                         return err;
6179         }
6180
6181         s_idx = ctx->idx;
6182         idx = 0;
6183
6184         for_each_netdev(net, dev) {
6185                 if (idx < s_idx)
6186                         goto skip;
6187                 if (!dev->netdev_ops->ndo_mdb_dump)
6188                         goto skip;
6189
6190                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6191                 if (err == -EMSGSIZE)
6192                         goto out;
6193                 /* Moving on to next device, reset markers and sequence
6194                  * counters since they are all maintained per-device.
6195                  */
6196                 memset(cb->ctx, 0, sizeof(cb->ctx));
6197                 cb->prev_seq = 0;
6198                 cb->seq = 0;
6199 skip:
6200                 idx++;
6201         }
6202
6203 out:
6204         ctx->idx = idx;
6205         return skb->len;
6206 }
6207
6208 static int rtnl_validate_mdb_entry_get(const struct nlattr *attr,
6209                                        struct netlink_ext_ack *extack)
6210 {
6211         struct br_mdb_entry *entry = nla_data(attr);
6212
6213         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6214                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6215                 return -EINVAL;
6216         }
6217
6218         if (entry->ifindex) {
6219                 NL_SET_ERR_MSG(extack, "Entry ifindex cannot be specified");
6220                 return -EINVAL;
6221         }
6222
6223         if (entry->state) {
6224                 NL_SET_ERR_MSG(extack, "Entry state cannot be specified");
6225                 return -EINVAL;
6226         }
6227
6228         if (entry->flags) {
6229                 NL_SET_ERR_MSG(extack, "Entry flags cannot be specified");
6230                 return -EINVAL;
6231         }
6232
6233         if (entry->vid >= VLAN_VID_MASK) {
6234                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6235                 return -EINVAL;
6236         }
6237
6238         if (entry->addr.proto != htons(ETH_P_IP) &&
6239             entry->addr.proto != htons(ETH_P_IPV6) &&
6240             entry->addr.proto != 0) {
6241                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6242                 return -EINVAL;
6243         }
6244
6245         return 0;
6246 }
6247
6248 static const struct nla_policy mdba_get_policy[MDBA_GET_ENTRY_MAX + 1] = {
6249         [MDBA_GET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6250                                                   rtnl_validate_mdb_entry_get,
6251                                                   sizeof(struct br_mdb_entry)),
6252         [MDBA_GET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6253 };
6254
6255 static int rtnl_mdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6256                         struct netlink_ext_ack *extack)
6257 {
6258         struct nlattr *tb[MDBA_GET_ENTRY_MAX + 1];
6259         struct net *net = sock_net(in_skb->sk);
6260         struct br_port_msg *bpm;
6261         struct net_device *dev;
6262         int err;
6263
6264         err = nlmsg_parse(nlh, sizeof(struct br_port_msg), tb,
6265                           MDBA_GET_ENTRY_MAX, mdba_get_policy, extack);
6266         if (err)
6267                 return err;
6268
6269         bpm = nlmsg_data(nlh);
6270         if (!bpm->ifindex) {
6271                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6272                 return -EINVAL;
6273         }
6274
6275         dev = __dev_get_by_index(net, bpm->ifindex);
6276         if (!dev) {
6277                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6278                 return -ENODEV;
6279         }
6280
6281         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_GET_ENTRY)) {
6282                 NL_SET_ERR_MSG(extack, "Missing MDBA_GET_ENTRY attribute");
6283                 return -EINVAL;
6284         }
6285
6286         if (!dev->netdev_ops->ndo_mdb_get) {
6287                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6288                 return -EOPNOTSUPP;
6289         }
6290
6291         return dev->netdev_ops->ndo_mdb_get(dev, tb, NETLINK_CB(in_skb).portid,
6292                                             nlh->nlmsg_seq, extack);
6293 }
6294
6295 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6296                                    struct netlink_ext_ack *extack)
6297 {
6298         struct br_mdb_entry *entry = nla_data(attr);
6299
6300         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6301                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6302                 return -EINVAL;
6303         }
6304
6305         if (entry->ifindex == 0) {
6306                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6307                 return -EINVAL;
6308         }
6309
6310         if (entry->addr.proto == htons(ETH_P_IP)) {
6311                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6312                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6313                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6314                         return -EINVAL;
6315                 }
6316                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6317                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6318                         return -EINVAL;
6319                 }
6320 #if IS_ENABLED(CONFIG_IPV6)
6321         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6322                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6323                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6324                         return -EINVAL;
6325                 }
6326 #endif
6327         } else if (entry->addr.proto == 0) {
6328                 /* L2 mdb */
6329                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6330                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6331                         return -EINVAL;
6332                 }
6333         } else {
6334                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6335                 return -EINVAL;
6336         }
6337
6338         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6339                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6340                 return -EINVAL;
6341         }
6342         if (entry->vid >= VLAN_VID_MASK) {
6343                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6344                 return -EINVAL;
6345         }
6346
6347         return 0;
6348 }
6349
6350 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6351         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6352         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6353                                                   rtnl_validate_mdb_entry,
6354                                                   sizeof(struct br_mdb_entry)),
6355         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6356 };
6357
6358 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6359                         struct netlink_ext_ack *extack)
6360 {
6361         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6362         struct net *net = sock_net(skb->sk);
6363         struct br_port_msg *bpm;
6364         struct net_device *dev;
6365         int err;
6366
6367         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6368                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6369         if (err)
6370                 return err;
6371
6372         bpm = nlmsg_data(nlh);
6373         if (!bpm->ifindex) {
6374                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6375                 return -EINVAL;
6376         }
6377
6378         dev = __dev_get_by_index(net, bpm->ifindex);
6379         if (!dev) {
6380                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6381                 return -ENODEV;
6382         }
6383
6384         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6385                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6386                 return -EINVAL;
6387         }
6388
6389         if (!dev->netdev_ops->ndo_mdb_add) {
6390                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6391                 return -EOPNOTSUPP;
6392         }
6393
6394         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6395 }
6396
6397 static int rtnl_validate_mdb_entry_del_bulk(const struct nlattr *attr,
6398                                             struct netlink_ext_ack *extack)
6399 {
6400         struct br_mdb_entry *entry = nla_data(attr);
6401         struct br_mdb_entry zero_entry = {};
6402
6403         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6404                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6405                 return -EINVAL;
6406         }
6407
6408         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6409                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6410                 return -EINVAL;
6411         }
6412
6413         if (entry->flags) {
6414                 NL_SET_ERR_MSG(extack, "Entry flags cannot be set");
6415                 return -EINVAL;
6416         }
6417
6418         if (entry->vid >= VLAN_N_VID - 1) {
6419                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6420                 return -EINVAL;
6421         }
6422
6423         if (memcmp(&entry->addr, &zero_entry.addr, sizeof(entry->addr))) {
6424                 NL_SET_ERR_MSG(extack, "Entry address cannot be set");
6425                 return -EINVAL;
6426         }
6427
6428         return 0;
6429 }
6430
6431 static const struct nla_policy mdba_del_bulk_policy[MDBA_SET_ENTRY_MAX + 1] = {
6432         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6433                                                   rtnl_validate_mdb_entry_del_bulk,
6434                                                   sizeof(struct br_mdb_entry)),
6435         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6436 };
6437
6438 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6439                         struct netlink_ext_ack *extack)
6440 {
6441         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
6442         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6443         struct net *net = sock_net(skb->sk);
6444         struct br_port_msg *bpm;
6445         struct net_device *dev;
6446         int err;
6447
6448         if (!del_bulk)
6449                 err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6450                                              MDBA_SET_ENTRY_MAX, mdba_policy,
6451                                              extack);
6452         else
6453                 err = nlmsg_parse(nlh, sizeof(*bpm), tb, MDBA_SET_ENTRY_MAX,
6454                                   mdba_del_bulk_policy, extack);
6455         if (err)
6456                 return err;
6457
6458         bpm = nlmsg_data(nlh);
6459         if (!bpm->ifindex) {
6460                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6461                 return -EINVAL;
6462         }
6463
6464         dev = __dev_get_by_index(net, bpm->ifindex);
6465         if (!dev) {
6466                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6467                 return -ENODEV;
6468         }
6469
6470         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6471                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6472                 return -EINVAL;
6473         }
6474
6475         if (del_bulk) {
6476                 if (!dev->netdev_ops->ndo_mdb_del_bulk) {
6477                         NL_SET_ERR_MSG(extack, "Device does not support MDB bulk deletion");
6478                         return -EOPNOTSUPP;
6479                 }
6480                 return dev->netdev_ops->ndo_mdb_del_bulk(dev, tb, extack);
6481         }
6482
6483         if (!dev->netdev_ops->ndo_mdb_del) {
6484                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6485                 return -EOPNOTSUPP;
6486         }
6487
6488         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6489 }
6490
6491 /* Process one rtnetlink message. */
6492
6493 static int rtnl_dumpit(struct sk_buff *skb, struct netlink_callback *cb)
6494 {
6495         const bool needs_lock = !(cb->flags & RTNL_FLAG_DUMP_UNLOCKED);
6496         rtnl_dumpit_func dumpit = cb->data;
6497         int err;
6498
6499         /* Previous iteration have already finished, avoid calling->dumpit()
6500          * again, it may not expect to be called after it reached the end.
6501          */
6502         if (!dumpit)
6503                 return 0;
6504
6505         if (needs_lock)
6506                 rtnl_lock();
6507         err = dumpit(skb, cb);
6508         if (needs_lock)
6509                 rtnl_unlock();
6510
6511         /* Old dump handlers used to send NLM_DONE as in a separate recvmsg().
6512          * Some applications which parse netlink manually depend on this.
6513          */
6514         if (cb->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE) {
6515                 if (err < 0 && err != -EMSGSIZE)
6516                         return err;
6517                 if (!err)
6518                         cb->data = NULL;
6519
6520                 return skb->len;
6521         }
6522         return err;
6523 }
6524
6525 static int rtnetlink_dump_start(struct sock *ssk, struct sk_buff *skb,
6526                                 const struct nlmsghdr *nlh,
6527                                 struct netlink_dump_control *control)
6528 {
6529         if (control->flags & RTNL_FLAG_DUMP_SPLIT_NLM_DONE ||
6530             !(control->flags & RTNL_FLAG_DUMP_UNLOCKED)) {
6531                 WARN_ON(control->data);
6532                 control->data = control->dump;
6533                 control->dump = rtnl_dumpit;
6534         }
6535
6536         return netlink_dump_start(ssk, skb, nlh, control);
6537 }
6538
6539 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6540                              struct netlink_ext_ack *extack)
6541 {
6542         struct net *net = sock_net(skb->sk);
6543         struct rtnl_link *link;
6544         enum rtnl_kinds kind;
6545         struct module *owner;
6546         int err = -EOPNOTSUPP;
6547         rtnl_doit_func doit;
6548         unsigned int flags;
6549         int family;
6550         int type;
6551
6552         type = nlh->nlmsg_type;
6553         if (type > RTM_MAX)
6554                 return -EOPNOTSUPP;
6555
6556         type -= RTM_BASE;
6557
6558         /* All the messages must have at least 1 byte length */
6559         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6560                 return 0;
6561
6562         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6563         kind = rtnl_msgtype_kind(type);
6564
6565         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6566                 return -EPERM;
6567
6568         rcu_read_lock();
6569         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6570                 struct sock *rtnl;
6571                 rtnl_dumpit_func dumpit;
6572                 u32 min_dump_alloc = 0;
6573
6574                 link = rtnl_get_link(family, type);
6575                 if (!link || !link->dumpit) {
6576                         family = PF_UNSPEC;
6577                         link = rtnl_get_link(family, type);
6578                         if (!link || !link->dumpit)
6579                                 goto err_unlock;
6580                 }
6581                 owner = link->owner;
6582                 dumpit = link->dumpit;
6583                 flags = link->flags;
6584
6585                 if (type == RTM_GETLINK - RTM_BASE)
6586                         min_dump_alloc = rtnl_calcit(skb, nlh);
6587
6588                 err = 0;
6589                 /* need to do this before rcu_read_unlock() */
6590                 if (!try_module_get(owner))
6591                         err = -EPROTONOSUPPORT;
6592
6593                 rcu_read_unlock();
6594
6595                 rtnl = net->rtnl;
6596                 if (err == 0) {
6597                         struct netlink_dump_control c = {
6598                                 .dump           = dumpit,
6599                                 .min_dump_alloc = min_dump_alloc,
6600                                 .module         = owner,
6601                                 .flags          = flags,
6602                         };
6603                         err = rtnetlink_dump_start(rtnl, skb, nlh, &c);
6604                         /* netlink_dump_start() will keep a reference on
6605                          * module if dump is still in progress.
6606                          */
6607                         module_put(owner);
6608                 }
6609                 return err;
6610         }
6611
6612         link = rtnl_get_link(family, type);
6613         if (!link || !link->doit) {
6614                 family = PF_UNSPEC;
6615                 link = rtnl_get_link(PF_UNSPEC, type);
6616                 if (!link || !link->doit)
6617                         goto out_unlock;
6618         }
6619
6620         owner = link->owner;
6621         if (!try_module_get(owner)) {
6622                 err = -EPROTONOSUPPORT;
6623                 goto out_unlock;
6624         }
6625
6626         flags = link->flags;
6627         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6628             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6629                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6630                 module_put(owner);
6631                 goto err_unlock;
6632         }
6633
6634         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6635                 doit = link->doit;
6636                 rcu_read_unlock();
6637                 if (doit)
6638                         err = doit(skb, nlh, extack);
6639                 module_put(owner);
6640                 return err;
6641         }
6642         rcu_read_unlock();
6643
6644         rtnl_lock();
6645         link = rtnl_get_link(family, type);
6646         if (link && link->doit)
6647                 err = link->doit(skb, nlh, extack);
6648         rtnl_unlock();
6649
6650         module_put(owner);
6651
6652         return err;
6653
6654 out_unlock:
6655         rcu_read_unlock();
6656         return err;
6657
6658 err_unlock:
6659         rcu_read_unlock();
6660         return -EOPNOTSUPP;
6661 }
6662
6663 static void rtnetlink_rcv(struct sk_buff *skb)
6664 {
6665         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6666 }
6667
6668 static int rtnetlink_bind(struct net *net, int group)
6669 {
6670         switch (group) {
6671         case RTNLGRP_IPV4_MROUTE_R:
6672         case RTNLGRP_IPV6_MROUTE_R:
6673                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6674                         return -EPERM;
6675                 break;
6676         }
6677         return 0;
6678 }
6679
6680 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6681 {
6682         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6683
6684         switch (event) {
6685         case NETDEV_REBOOT:
6686         case NETDEV_CHANGEMTU:
6687         case NETDEV_CHANGEADDR:
6688         case NETDEV_CHANGENAME:
6689         case NETDEV_FEAT_CHANGE:
6690         case NETDEV_BONDING_FAILOVER:
6691         case NETDEV_POST_TYPE_CHANGE:
6692         case NETDEV_NOTIFY_PEERS:
6693         case NETDEV_CHANGEUPPER:
6694         case NETDEV_RESEND_IGMP:
6695         case NETDEV_CHANGEINFODATA:
6696         case NETDEV_CHANGELOWERSTATE:
6697         case NETDEV_CHANGE_TX_QUEUE_LEN:
6698                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6699                                    GFP_KERNEL, NULL, 0, 0, NULL);
6700                 break;
6701         default:
6702                 break;
6703         }
6704         return NOTIFY_DONE;
6705 }
6706
6707 static struct notifier_block rtnetlink_dev_notifier = {
6708         .notifier_call  = rtnetlink_event,
6709 };
6710
6711
6712 static int __net_init rtnetlink_net_init(struct net *net)
6713 {
6714         struct sock *sk;
6715         struct netlink_kernel_cfg cfg = {
6716                 .groups         = RTNLGRP_MAX,
6717                 .input          = rtnetlink_rcv,
6718                 .flags          = NL_CFG_F_NONROOT_RECV,
6719                 .bind           = rtnetlink_bind,
6720         };
6721
6722         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6723         if (!sk)
6724                 return -ENOMEM;
6725         net->rtnl = sk;
6726         return 0;
6727 }
6728
6729 static void __net_exit rtnetlink_net_exit(struct net *net)
6730 {
6731         netlink_kernel_release(net->rtnl);
6732         net->rtnl = NULL;
6733 }
6734
6735 static struct pernet_operations rtnetlink_net_ops = {
6736         .init = rtnetlink_net_init,
6737         .exit = rtnetlink_net_exit,
6738 };
6739
6740 void __init rtnetlink_init(void)
6741 {
6742         if (register_pernet_subsys(&rtnetlink_net_ops))
6743                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6744
6745         register_netdevice_notifier(&rtnetlink_dev_notifier);
6746
6747         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6748                       rtnl_dump_ifinfo, RTNL_FLAG_DUMP_SPLIT_NLM_DONE);
6749         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6750         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6751         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6752
6753         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6754         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6755         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6756
6757         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6758         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6759
6760         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6761         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6762                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6763         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6764
6765         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6766         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6767         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6768
6769         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6770                       0);
6771         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6772
6773         rtnl_register(PF_BRIDGE, RTM_GETMDB, rtnl_mdb_get, rtnl_mdb_dump, 0);
6774         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6775         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL,
6776                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6777 }