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