net/core/rtnetlink.c

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