net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <[email protected]>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetic was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56 #include <net/devlink.h>
  57 #if IS_ENABLED(CONFIG_IPV6)
  58 #include <net/addrconf.h>
  59 #endif
  60
  61 #include "dev.h"
  62
  63 #define RTNL_MAX_TYPE           50
  64 #define RTNL_SLAVE_MAX_TYPE     44
  65
  66 struct rtnl_link {
  67         rtnl_doit_func          doit;
  68         rtnl_dumpit_func        dumpit;
  69         struct module           *owner;
  70         unsigned int            flags;
  71         struct rcu_head         rcu;
  72 };
  73
  74 static DEFINE_MUTEX(rtnl_mutex);
  75
  76 void rtnl_lock(void)
  77 {
  78         mutex_lock(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock);
  81
  82 int rtnl_lock_killable(void)
  83 {
  84         return mutex_lock_killable(&rtnl_mutex);
  85 }
  86 EXPORT_SYMBOL(rtnl_lock_killable);
  87
  88 static struct sk_buff *defer_kfree_skb_list;
  89 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  90 {
  91         if (head && tail) {
  92                 tail->next = defer_kfree_skb_list;
  93                 defer_kfree_skb_list = head;
  94         }
  95 }
  96 EXPORT_SYMBOL(rtnl_kfree_skbs);
  97
  98 void __rtnl_unlock(void)
  99 {
 100         struct sk_buff *head = defer_kfree_skb_list;
 101
 102         defer_kfree_skb_list = NULL;
 103
 104         /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
 105          * is used. In some places, e.g. in cfg80211, we have code that will do
 106          * something like
 107          *   rtnl_lock()
 108          *   wiphy_lock()
 109          *   ...
 110          *   rtnl_unlock()
 111          *
 112          * and because netdev_run_todo() acquires the RTNL for items on the list
 113          * we could cause a situation such as this:
 114          * Thread 1                     Thread 2
 115          *                                rtnl_lock()
 116          *                                unregister_netdevice()
 117          *                                __rtnl_unlock()
 118          * rtnl_lock()
 119          * wiphy_lock()
 120          * rtnl_unlock()
 121          *   netdev_run_todo()
 122          *     __rtnl_unlock()
 123          *
 124          *     // list not empty now
 125          *     // because of thread 2
 126          *                                rtnl_lock()
 127          *     while (!list_empty(...))
 128          *       rtnl_lock()
 129          *                                wiphy_lock()
 130          * **** DEADLOCK ****
 131          *
 132          * However, usage of __rtnl_unlock() is rare, and so we can ensure that
 133          * it's not used in cases where something is added to do the list.
 134          */
 135         WARN_ON(!list_empty(&net_todo_list));
 136
 137         mutex_unlock(&rtnl_mutex);
 138
 139         while (head) {
 140                 struct sk_buff *next = head->next;
 141
 142                 kfree_skb(head);
 143                 cond_resched();
 144                 head = next;
 145         }
 146 }
 147
 148 void rtnl_unlock(void)
 149 {
 150         /* This fellow will unlock it for us. */
 151         netdev_run_todo();
 152 }
 153 EXPORT_SYMBOL(rtnl_unlock);
 154
 155 int rtnl_trylock(void)
 156 {
 157         return mutex_trylock(&rtnl_mutex);
 158 }
 159 EXPORT_SYMBOL(rtnl_trylock);
 160
 161 int rtnl_is_locked(void)
 162 {
 163         return mutex_is_locked(&rtnl_mutex);
 164 }
 165 EXPORT_SYMBOL(rtnl_is_locked);
 166
 167 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 168 {
 169         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 170 }
 171 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 172
 173 #ifdef CONFIG_PROVE_LOCKING
 174 bool lockdep_rtnl_is_held(void)
 175 {
 176         return lockdep_is_held(&rtnl_mutex);
 177 }
 178 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 179 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 180
 181 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 182
 183 static inline int rtm_msgindex(int msgtype)
 184 {
 185         int msgindex = msgtype - RTM_BASE;
 186
 187         /*
 188          * msgindex < 0 implies someone tried to register a netlink
 189          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 190          * the message type has not been added to linux/rtnetlink.h
 191          */
 192         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 193
 194         return msgindex;
 195 }
 196
 197 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 198 {
 199         struct rtnl_link __rcu **tab;
 200
 201         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 202                 protocol = PF_UNSPEC;
 203
 204         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 205         if (!tab)
 206                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 207
 208         return rcu_dereference_rtnl(tab[msgtype]);
 209 }
 210
 211 static int rtnl_register_internal(struct module *owner,
 212                                   int protocol, int msgtype,
 213                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 214                                   unsigned int flags)
 215 {
 216         struct rtnl_link *link, *old;
 217         struct rtnl_link __rcu **tab;
 218         int msgindex;
 219         int ret = -ENOBUFS;
 220
 221         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 222         msgindex = rtm_msgindex(msgtype);
 223
 224         rtnl_lock();
 225         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 226         if (tab == NULL) {
 227                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 228                 if (!tab)
 229                         goto unlock;
 230
 231                 /* ensures we see the 0 stores */
 232                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 233         }
 234
 235         old = rtnl_dereference(tab[msgindex]);
 236         if (old) {
 237                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 238                 if (!link)
 239                         goto unlock;
 240         } else {
 241                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 242                 if (!link)
 243                         goto unlock;
 244         }
 245
 246         WARN_ON(link->owner && link->owner != owner);
 247         link->owner = owner;
 248
 249         WARN_ON(doit && link->doit && link->doit != doit);
 250         if (doit)
 251                 link->doit = doit;
 252         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 253         if (dumpit)
 254                 link->dumpit = dumpit;
 255
 256         WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
 257                 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
 258         link->flags |= flags;
 259
 260         /* publish protocol:msgtype */
 261         rcu_assign_pointer(tab[msgindex], link);
 262         ret = 0;
 263         if (old)
 264                 kfree_rcu(old, rcu);
 265 unlock:
 266         rtnl_unlock();
 267         return ret;
 268 }
 269
 270 /**
 271  * rtnl_register_module - Register a rtnetlink message type
 272  *
 273  * @owner: module registering the hook (THIS_MODULE)
 274  * @protocol: Protocol family or PF_UNSPEC
 275  * @msgtype: rtnetlink message type
 276  * @doit: Function pointer called for each request message
 277  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 278  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 279  *
 280  * Like rtnl_register, but for use by removable modules.
 281  */
 282 int rtnl_register_module(struct module *owner,
 283                          int protocol, int msgtype,
 284                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 285                          unsigned int flags)
 286 {
 287         return rtnl_register_internal(owner, protocol, msgtype,
 288                                       doit, dumpit, flags);
 289 }
 290 EXPORT_SYMBOL_GPL(rtnl_register_module);
 291
 292 /**
 293  * rtnl_register - Register a rtnetlink message type
 294  * @protocol: Protocol family or PF_UNSPEC
 295  * @msgtype: rtnetlink message type
 296  * @doit: Function pointer called for each request message
 297  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 298  * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
 299  *
 300  * Registers the specified function pointers (at least one of them has
 301  * to be non-NULL) to be called whenever a request message for the
 302  * specified protocol family and message type is received.
 303  *
 304  * The special protocol family PF_UNSPEC may be used to define fallback
 305  * function pointers for the case when no entry for the specific protocol
 306  * family exists.
 307  */
 308 void rtnl_register(int protocol, int msgtype,
 309                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 310                    unsigned int flags)
 311 {
 312         int err;
 313
 314         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 315                                      flags);
 316         if (err)
 317                 pr_err("Unable to register rtnetlink message handler, "
 318                        "protocol = %d, message type = %d\n", protocol, msgtype);
 319 }
 320
 321 /**
 322  * rtnl_unregister - Unregister a rtnetlink message type
 323  * @protocol: Protocol family or PF_UNSPEC
 324  * @msgtype: rtnetlink message type
 325  *
 326  * Returns 0 on success or a negative error code.
 327  */
 328 int rtnl_unregister(int protocol, int msgtype)
 329 {
 330         struct rtnl_link __rcu **tab;
 331         struct rtnl_link *link;
 332         int msgindex;
 333
 334         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 335         msgindex = rtm_msgindex(msgtype);
 336
 337         rtnl_lock();
 338         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 339         if (!tab) {
 340                 rtnl_unlock();
 341                 return -ENOENT;
 342         }
 343
 344         link = rtnl_dereference(tab[msgindex]);
 345         RCU_INIT_POINTER(tab[msgindex], NULL);
 346         rtnl_unlock();
 347
 348         kfree_rcu(link, rcu);
 349
 350         return 0;
 351 }
 352 EXPORT_SYMBOL_GPL(rtnl_unregister);
 353
 354 /**
 355  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 356  * @protocol : Protocol family or PF_UNSPEC
 357  *
 358  * Identical to calling rtnl_unregster() for all registered message types
 359  * of a certain protocol family.
 360  */
 361 void rtnl_unregister_all(int protocol)
 362 {
 363         struct rtnl_link __rcu **tab;
 364         struct rtnl_link *link;
 365         int msgindex;
 366
 367         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 368
 369         rtnl_lock();
 370         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 371         if (!tab) {
 372                 rtnl_unlock();
 373                 return;
 374         }
 375         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 376         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 377                 link = rtnl_dereference(tab[msgindex]);
 378                 if (!link)
 379                         continue;
 380
 381                 RCU_INIT_POINTER(tab[msgindex], NULL);
 382                 kfree_rcu(link, rcu);
 383         }
 384         rtnl_unlock();
 385
 386         synchronize_net();
 387
 388         kfree(tab);
 389 }
 390 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 391
 392 static LIST_HEAD(link_ops);
 393
 394 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 395 {
 396         const struct rtnl_link_ops *ops;
 397
 398         list_for_each_entry(ops, &link_ops, list) {
 399                 if (!strcmp(ops->kind, kind))
 400                         return ops;
 401         }
 402         return NULL;
 403 }
 404
 405 /**
 406  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 407  * @ops: struct rtnl_link_ops * to register
 408  *
 409  * The caller must hold the rtnl_mutex. This function should be used
 410  * by drivers that create devices during module initialization. It
 411  * must be called before registering the devices.
 412  *
 413  * Returns 0 on success or a negative error code.
 414  */
 415 int __rtnl_link_register(struct rtnl_link_ops *ops)
 416 {
 417         if (rtnl_link_ops_get(ops->kind))
 418                 return -EEXIST;
 419
 420         /* The check for alloc/setup is here because if ops
 421          * does not have that filled up, it is not possible
 422          * to use the ops for creating device. So do not
 423          * fill up dellink as well. That disables rtnl_dellink.
 424          */
 425         if ((ops->alloc || ops->setup) && !ops->dellink)
 426                 ops->dellink = unregister_netdevice_queue;
 427
 428         list_add_tail(&ops->list, &link_ops);
 429         return 0;
 430 }
 431 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 432
 433 /**
 434  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 435  * @ops: struct rtnl_link_ops * to register
 436  *
 437  * Returns 0 on success or a negative error code.
 438  */
 439 int rtnl_link_register(struct rtnl_link_ops *ops)
 440 {
 441         int err;
 442
 443         /* Sanity-check max sizes to avoid stack buffer overflow. */
 444         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 445                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 446                 return -EINVAL;
 447
 448         rtnl_lock();
 449         err = __rtnl_link_register(ops);
 450         rtnl_unlock();
 451         return err;
 452 }
 453 EXPORT_SYMBOL_GPL(rtnl_link_register);
 454
 455 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 456 {
 457         struct net_device *dev;
 458         LIST_HEAD(list_kill);
 459
 460         for_each_netdev(net, dev) {
 461                 if (dev->rtnl_link_ops == ops)
 462                         ops->dellink(dev, &list_kill);
 463         }
 464         unregister_netdevice_many(&list_kill);
 465 }
 466
 467 /**
 468  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 469  * @ops: struct rtnl_link_ops * to unregister
 470  *
 471  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 472  * integrity (hold pernet_ops_rwsem for writing to close the race
 473  * with setup_net() and cleanup_net()).
 474  */
 475 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 476 {
 477         struct net *net;
 478
 479         for_each_net(net) {
 480                 __rtnl_kill_links(net, ops);
 481         }
 482         list_del(&ops->list);
 483 }
 484 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 485
 486 /* Return with the rtnl_lock held when there are no network
 487  * devices unregistering in any network namespace.
 488  */
 489 static void rtnl_lock_unregistering_all(void)
 490 {
 491         struct net *net;
 492         bool unregistering;
 493         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 494
 495         add_wait_queue(&netdev_unregistering_wq, &wait);
 496         for (;;) {
 497                 unregistering = false;
 498                 rtnl_lock();
 499                 /* We held write locked pernet_ops_rwsem, and parallel
 500                  * setup_net() and cleanup_net() are not possible.
 501                  */
 502                 for_each_net(net) {
 503                         if (atomic_read(&net->dev_unreg_count) > 0) {
 504                                 unregistering = true;
 505                                 break;
 506                         }
 507                 }
 508                 if (!unregistering)
 509                         break;
 510                 __rtnl_unlock();
 511
 512                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 513         }
 514         remove_wait_queue(&netdev_unregistering_wq, &wait);
 515 }
 516
 517 /**
 518  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 519  * @ops: struct rtnl_link_ops * to unregister
 520  */
 521 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 522 {
 523         /* Close the race with setup_net() and cleanup_net() */
 524         down_write(&pernet_ops_rwsem);
 525         rtnl_lock_unregistering_all();
 526         __rtnl_link_unregister(ops);
 527         rtnl_unlock();
 528         up_write(&pernet_ops_rwsem);
 529 }
 530 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 531
 532 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 533 {
 534         struct net_device *master_dev;
 535         const struct rtnl_link_ops *ops;
 536         size_t size = 0;
 537
 538         rcu_read_lock();
 539
 540         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 541         if (!master_dev)
 542                 goto out;
 543
 544         ops = master_dev->rtnl_link_ops;
 545         if (!ops || !ops->get_slave_size)
 546                 goto out;
 547         /* IFLA_INFO_SLAVE_DATA + nested data */
 548         size = nla_total_size(sizeof(struct nlattr)) +
 549                ops->get_slave_size(master_dev, dev);
 550
 551 out:
 552         rcu_read_unlock();
 553         return size;
 554 }
 555
 556 static size_t rtnl_link_get_size(const struct net_device *dev)
 557 {
 558         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 559         size_t size;
 560
 561         if (!ops)
 562                 return 0;
 563
 564         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 565                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 566
 567         if (ops->get_size)
 568                 /* IFLA_INFO_DATA + nested data */
 569                 size += nla_total_size(sizeof(struct nlattr)) +
 570                         ops->get_size(dev);
 571
 572         if (ops->get_xstats_size)
 573                 /* IFLA_INFO_XSTATS */
 574                 size += nla_total_size(ops->get_xstats_size(dev));
 575
 576         size += rtnl_link_get_slave_info_data_size(dev);
 577
 578         return size;
 579 }
 580
 581 static LIST_HEAD(rtnl_af_ops);
 582
 583 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 584 {
 585         const struct rtnl_af_ops *ops;
 586
 587         ASSERT_RTNL();
 588
 589         list_for_each_entry(ops, &rtnl_af_ops, list) {
 590                 if (ops->family == family)
 591                         return ops;
 592         }
 593
 594         return NULL;
 595 }
 596
 597 /**
 598  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 599  * @ops: struct rtnl_af_ops * to register
 600  *
 601  * Returns 0 on success or a negative error code.
 602  */
 603 void rtnl_af_register(struct rtnl_af_ops *ops)
 604 {
 605         rtnl_lock();
 606         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 607         rtnl_unlock();
 608 }
 609 EXPORT_SYMBOL_GPL(rtnl_af_register);
 610
 611 /**
 612  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 613  * @ops: struct rtnl_af_ops * to unregister
 614  */
 615 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 616 {
 617         rtnl_lock();
 618         list_del_rcu(&ops->list);
 619         rtnl_unlock();
 620
 621         synchronize_rcu();
 622 }
 623 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 624
 625 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 626                                     u32 ext_filter_mask)
 627 {
 628         struct rtnl_af_ops *af_ops;
 629         size_t size;
 630
 631         /* IFLA_AF_SPEC */
 632         size = nla_total_size(sizeof(struct nlattr));
 633
 634         rcu_read_lock();
 635         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 636                 if (af_ops->get_link_af_size) {
 637                         /* AF_* + nested data */
 638                         size += nla_total_size(sizeof(struct nlattr)) +
 639                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 640                 }
 641         }
 642         rcu_read_unlock();
 643
 644         return size;
 645 }
 646
 647 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 648 {
 649         struct net_device *master_dev;
 650         bool ret = false;
 651
 652         rcu_read_lock();
 653
 654         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 655         if (master_dev && master_dev->rtnl_link_ops)
 656                 ret = true;
 657         rcu_read_unlock();
 658         return ret;
 659 }
 660
 661 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 662                                      const struct net_device *dev)
 663 {
 664         struct net_device *master_dev;
 665         const struct rtnl_link_ops *ops;
 666         struct nlattr *slave_data;
 667         int err;
 668
 669         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 670         if (!master_dev)
 671                 return 0;
 672         ops = master_dev->rtnl_link_ops;
 673         if (!ops)
 674                 return 0;
 675         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 676                 return -EMSGSIZE;
 677         if (ops->fill_slave_info) {
 678                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 679                 if (!slave_data)
 680                         return -EMSGSIZE;
 681                 err = ops->fill_slave_info(skb, master_dev, dev);
 682                 if (err < 0)
 683                         goto err_cancel_slave_data;
 684                 nla_nest_end(skb, slave_data);
 685         }
 686         return 0;
 687
 688 err_cancel_slave_data:
 689         nla_nest_cancel(skb, slave_data);
 690         return err;
 691 }
 692
 693 static int rtnl_link_info_fill(struct sk_buff *skb,
 694                                const struct net_device *dev)
 695 {
 696         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 697         struct nlattr *data;
 698         int err;
 699
 700         if (!ops)
 701                 return 0;
 702         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 703                 return -EMSGSIZE;
 704         if (ops->fill_xstats) {
 705                 err = ops->fill_xstats(skb, dev);
 706                 if (err < 0)
 707                         return err;
 708         }
 709         if (ops->fill_info) {
 710                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 711                 if (data == NULL)
 712                         return -EMSGSIZE;
 713                 err = ops->fill_info(skb, dev);
 714                 if (err < 0)
 715                         goto err_cancel_data;
 716                 nla_nest_end(skb, data);
 717         }
 718         return 0;
 719
 720 err_cancel_data:
 721         nla_nest_cancel(skb, data);
 722         return err;
 723 }
 724
 725 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 726 {
 727         struct nlattr *linkinfo;
 728         int err = -EMSGSIZE;
 729
 730         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 731         if (linkinfo == NULL)
 732                 goto out;
 733
 734         err = rtnl_link_info_fill(skb, dev);
 735         if (err < 0)
 736                 goto err_cancel_link;
 737
 738         err = rtnl_link_slave_info_fill(skb, dev);
 739         if (err < 0)
 740                 goto err_cancel_link;
 741
 742         nla_nest_end(skb, linkinfo);
 743         return 0;
 744
 745 err_cancel_link:
 746         nla_nest_cancel(skb, linkinfo);
 747 out:
 748         return err;
 749 }
 750
 751 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 752 {
 753         struct sock *rtnl = net->rtnl;
 754
 755         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 756 }
 757
 758 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 759 {
 760         struct sock *rtnl = net->rtnl;
 761
 762         return nlmsg_unicast(rtnl, skb, pid);
 763 }
 764 EXPORT_SYMBOL(rtnl_unicast);
 765
 766 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 767                  const struct nlmsghdr *nlh, gfp_t flags)
 768 {
 769         struct sock *rtnl = net->rtnl;
 770
 771         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 772 }
 773 EXPORT_SYMBOL(rtnl_notify);
 774
 775 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 776 {
 777         struct sock *rtnl = net->rtnl;
 778
 779         netlink_set_err(rtnl, 0, group, error);
 780 }
 781 EXPORT_SYMBOL(rtnl_set_sk_err);
 782
 783 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 784 {
 785         struct nlattr *mx;
 786         int i, valid = 0;
 787
 788         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 789         if (metrics == dst_default_metrics.metrics)
 790                 return 0;
 791
 792         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 793         if (mx == NULL)
 794                 return -ENOBUFS;
 795
 796         for (i = 0; i < RTAX_MAX; i++) {
 797                 if (metrics[i]) {
 798                         if (i == RTAX_CC_ALGO - 1) {
 799                                 char tmp[TCP_CA_NAME_MAX], *name;
 800
 801                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 802                                 if (!name)
 803                                         continue;
 804                                 if (nla_put_string(skb, i + 1, name))
 805                                         goto nla_put_failure;
 806                         } else if (i == RTAX_FEATURES - 1) {
 807                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 808
 809                                 if (!user_features)
 810                                         continue;
 811                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 812                                 if (nla_put_u32(skb, i + 1, user_features))
 813                                         goto nla_put_failure;
 814                         } else {
 815                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 816                                         goto nla_put_failure;
 817                         }
 818                         valid++;
 819                 }
 820         }
 821
 822         if (!valid) {
 823                 nla_nest_cancel(skb, mx);
 824                 return 0;
 825         }
 826
 827         return nla_nest_end(skb, mx);
 828
 829 nla_put_failure:
 830         nla_nest_cancel(skb, mx);
 831         return -EMSGSIZE;
 832 }
 833 EXPORT_SYMBOL(rtnetlink_put_metrics);
 834
 835 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 836                        long expires, u32 error)
 837 {
 838         struct rta_cacheinfo ci = {
 839                 .rta_error = error,
 840                 .rta_id =  id,
 841         };
 842
 843         if (dst) {
 844                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 845                 ci.rta_used = dst->__use;
 846                 ci.rta_clntref = rcuref_read(&dst->__rcuref);
 847         }
 848         if (expires) {
 849                 unsigned long clock;
 850
 851                 clock = jiffies_to_clock_t(abs(expires));
 852                 clock = min_t(unsigned long, clock, INT_MAX);
 853                 ci.rta_expires = (expires > 0) ? clock : -clock;
 854         }
 855         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 856 }
 857 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 858
 859 static void set_operstate(struct net_device *dev, unsigned char transition)
 860 {
 861         unsigned char operstate = dev->operstate;
 862
 863         switch (transition) {
 864         case IF_OPER_UP:
 865                 if ((operstate == IF_OPER_DORMANT ||
 866                      operstate == IF_OPER_TESTING ||
 867                      operstate == IF_OPER_UNKNOWN) &&
 868                     !netif_dormant(dev) && !netif_testing(dev))
 869                         operstate = IF_OPER_UP;
 870                 break;
 871
 872         case IF_OPER_TESTING:
 873                 if (netif_oper_up(dev))
 874                         operstate = IF_OPER_TESTING;
 875                 break;
 876
 877         case IF_OPER_DORMANT:
 878                 if (netif_oper_up(dev))
 879                         operstate = IF_OPER_DORMANT;
 880                 break;
 881         }
 882
 883         if (dev->operstate != operstate) {
 884                 write_lock(&dev_base_lock);
 885                 dev->operstate = operstate;
 886                 write_unlock(&dev_base_lock);
 887                 netdev_state_change(dev);
 888         }
 889 }
 890
 891 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 892 {
 893         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 894                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 895 }
 896
 897 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 898                                            const struct ifinfomsg *ifm)
 899 {
 900         unsigned int flags = ifm->ifi_flags;
 901
 902         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 903         if (ifm->ifi_change)
 904                 flags = (flags & ifm->ifi_change) |
 905                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 906
 907         return flags;
 908 }
 909
 910 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 911                                  const struct rtnl_link_stats64 *b)
 912 {
 913         a->rx_packets = b->rx_packets;
 914         a->tx_packets = b->tx_packets;
 915         a->rx_bytes = b->rx_bytes;
 916         a->tx_bytes = b->tx_bytes;
 917         a->rx_errors = b->rx_errors;
 918         a->tx_errors = b->tx_errors;
 919         a->rx_dropped = b->rx_dropped;
 920         a->tx_dropped = b->tx_dropped;
 921
 922         a->multicast = b->multicast;
 923         a->collisions = b->collisions;
 924
 925         a->rx_length_errors = b->rx_length_errors;
 926         a->rx_over_errors = b->rx_over_errors;
 927         a->rx_crc_errors = b->rx_crc_errors;
 928         a->rx_frame_errors = b->rx_frame_errors;
 929         a->rx_fifo_errors = b->rx_fifo_errors;
 930         a->rx_missed_errors = b->rx_missed_errors;
 931
 932         a->tx_aborted_errors = b->tx_aborted_errors;
 933         a->tx_carrier_errors = b->tx_carrier_errors;
 934         a->tx_fifo_errors = b->tx_fifo_errors;
 935         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 936         a->tx_window_errors = b->tx_window_errors;
 937
 938         a->rx_compressed = b->rx_compressed;
 939         a->tx_compressed = b->tx_compressed;
 940
 941         a->rx_nohandler = b->rx_nohandler;
 942 }
 943
 944 /* All VF info */
 945 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 946                                    u32 ext_filter_mask)
 947 {
 948         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 949                 int num_vfs = dev_num_vf(dev->dev.parent);
 950                 size_t size = nla_total_size(0);
 951                 size += num_vfs *
 952                         (nla_total_size(0) +
 953                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 954                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 955                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 956                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 957                          nla_total_size(MAX_VLAN_LIST_LEN *
 958                                         sizeof(struct ifla_vf_vlan_info)) +
 959                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 960                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 961                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 962                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 963                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 964                          nla_total_size(sizeof(struct ifla_vf_trust)));
 965                 if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
 966                         size += num_vfs *
 967                                 (nla_total_size(0) + /* nest IFLA_VF_STATS */
 968                                  /* IFLA_VF_STATS_RX_PACKETS */
 969                                  nla_total_size_64bit(sizeof(__u64)) +
 970                                  /* IFLA_VF_STATS_TX_PACKETS */
 971                                  nla_total_size_64bit(sizeof(__u64)) +
 972                                  /* IFLA_VF_STATS_RX_BYTES */
 973                                  nla_total_size_64bit(sizeof(__u64)) +
 974                                  /* IFLA_VF_STATS_TX_BYTES */
 975                                  nla_total_size_64bit(sizeof(__u64)) +
 976                                  /* IFLA_VF_STATS_BROADCAST */
 977                                  nla_total_size_64bit(sizeof(__u64)) +
 978                                  /* IFLA_VF_STATS_MULTICAST */
 979                                  nla_total_size_64bit(sizeof(__u64)) +
 980                                  /* IFLA_VF_STATS_RX_DROPPED */
 981                                  nla_total_size_64bit(sizeof(__u64)) +
 982                                  /* IFLA_VF_STATS_TX_DROPPED */
 983                                  nla_total_size_64bit(sizeof(__u64)));
 984                 }
 985                 return size;
 986         } else
 987                 return 0;
 988 }
 989
 990 static size_t rtnl_port_size(const struct net_device *dev,
 991                              u32 ext_filter_mask)
 992 {
 993         size_t port_size = nla_total_size(4)            /* PORT_VF */
 994                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 995                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 996                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 997                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 998                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 999         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
1000         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
1001                 + port_size;
1002         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1003                 + port_size;
1004
1005         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1006             !(ext_filter_mask & RTEXT_FILTER_VF))
1007                 return 0;
1008         if (dev_num_vf(dev->dev.parent))
1009                 return port_self_size + vf_ports_size +
1010                         vf_port_size * dev_num_vf(dev->dev.parent);
1011         else
1012                 return port_self_size;
1013 }
1014
1015 static size_t rtnl_xdp_size(void)
1016 {
1017         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1018                           nla_total_size(1) +   /* XDP_ATTACHED */
1019                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1020                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1021
1022         return xdp_size;
1023 }
1024
1025 static size_t rtnl_prop_list_size(const struct net_device *dev)
1026 {
1027         struct netdev_name_node *name_node;
1028         size_t size;
1029
1030         if (list_empty(&dev->name_node->list))
1031                 return 0;
1032         size = nla_total_size(0);
1033         list_for_each_entry(name_node, &dev->name_node->list, list)
1034                 size += nla_total_size(ALTIFNAMSIZ);
1035         return size;
1036 }
1037
1038 static size_t rtnl_proto_down_size(const struct net_device *dev)
1039 {
1040         size_t size = nla_total_size(1);
1041
1042         if (dev->proto_down_reason)
1043                 size += nla_total_size(0) + nla_total_size(4);
1044
1045         return size;
1046 }
1047
1048 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1049 {
1050         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1051
1052         if (dev->devlink_port)
1053                 size += devlink_nl_port_handle_size(dev->devlink_port);
1054
1055         return size;
1056 }
1057
1058 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1059                                      u32 ext_filter_mask)
1060 {
1061         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1062                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1063                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1064                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1065                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1066                + nla_total_size(sizeof(struct rtnl_link_stats))
1067                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1068                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1069                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1070                + nla_total_size(4) /* IFLA_TXQLEN */
1071                + nla_total_size(4) /* IFLA_WEIGHT */
1072                + nla_total_size(4) /* IFLA_MTU */
1073                + nla_total_size(4) /* IFLA_LINK */
1074                + nla_total_size(4) /* IFLA_MASTER */
1075                + nla_total_size(1) /* IFLA_CARRIER */
1076                + nla_total_size(4) /* IFLA_PROMISCUITY */
1077                + nla_total_size(4) /* IFLA_ALLMULTI */
1078                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1079                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1080                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1081                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1082                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1083                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1084                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1085                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1086                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1087                + nla_total_size(1) /* IFLA_OPERSTATE */
1088                + nla_total_size(1) /* IFLA_LINKMODE */
1089                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1090                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1091                + nla_total_size(4) /* IFLA_GROUP */
1092                + nla_total_size(ext_filter_mask
1093                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1094                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1095                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1096                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1097                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1098                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1099                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1100                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1101                + rtnl_xdp_size() /* IFLA_XDP */
1102                + nla_total_size(4)  /* IFLA_EVENT */
1103                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1104                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1105                + rtnl_proto_down_size(dev)  /* proto down */
1106                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1107                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1108                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1109                + nla_total_size(4)  /* IFLA_MIN_MTU */
1110                + nla_total_size(4)  /* IFLA_MAX_MTU */
1111                + rtnl_prop_list_size(dev)
1112                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1113                + rtnl_devlink_port_size(dev)
1114                + 0;
1115 }
1116
1117 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1118 {
1119         struct nlattr *vf_ports;
1120         struct nlattr *vf_port;
1121         int vf;
1122         int err;
1123
1124         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1125         if (!vf_ports)
1126                 return -EMSGSIZE;
1127
1128         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1129                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1130                 if (!vf_port)
1131                         goto nla_put_failure;
1132                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1133                         goto nla_put_failure;
1134                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1135                 if (err == -EMSGSIZE)
1136                         goto nla_put_failure;
1137                 if (err) {
1138                         nla_nest_cancel(skb, vf_port);
1139                         continue;
1140                 }
1141                 nla_nest_end(skb, vf_port);
1142         }
1143
1144         nla_nest_end(skb, vf_ports);
1145
1146         return 0;
1147
1148 nla_put_failure:
1149         nla_nest_cancel(skb, vf_ports);
1150         return -EMSGSIZE;
1151 }
1152
1153 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1154 {
1155         struct nlattr *port_self;
1156         int err;
1157
1158         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1159         if (!port_self)
1160                 return -EMSGSIZE;
1161
1162         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1163         if (err) {
1164                 nla_nest_cancel(skb, port_self);
1165                 return (err == -EMSGSIZE) ? err : 0;
1166         }
1167
1168         nla_nest_end(skb, port_self);
1169
1170         return 0;
1171 }
1172
1173 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1174                           u32 ext_filter_mask)
1175 {
1176         int err;
1177
1178         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1179             !(ext_filter_mask & RTEXT_FILTER_VF))
1180                 return 0;
1181
1182         err = rtnl_port_self_fill(skb, dev);
1183         if (err)
1184                 return err;
1185
1186         if (dev_num_vf(dev->dev.parent)) {
1187                 err = rtnl_vf_ports_fill(skb, dev);
1188                 if (err)
1189                         return err;
1190         }
1191
1192         return 0;
1193 }
1194
1195 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1196 {
1197         int err;
1198         struct netdev_phys_item_id ppid;
1199
1200         err = dev_get_phys_port_id(dev, &ppid);
1201         if (err) {
1202                 if (err == -EOPNOTSUPP)
1203                         return 0;
1204                 return err;
1205         }
1206
1207         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1208                 return -EMSGSIZE;
1209
1210         return 0;
1211 }
1212
1213 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1214 {
1215         char name[IFNAMSIZ];
1216         int err;
1217
1218         err = dev_get_phys_port_name(dev, name, sizeof(name));
1219         if (err) {
1220                 if (err == -EOPNOTSUPP)
1221                         return 0;
1222                 return err;
1223         }
1224
1225         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1226                 return -EMSGSIZE;
1227
1228         return 0;
1229 }
1230
1231 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1232 {
1233         struct netdev_phys_item_id ppid = { };
1234         int err;
1235
1236         err = dev_get_port_parent_id(dev, &ppid, false);
1237         if (err) {
1238                 if (err == -EOPNOTSUPP)
1239                         return 0;
1240                 return err;
1241         }
1242
1243         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1244                 return -EMSGSIZE;
1245
1246         return 0;
1247 }
1248
1249 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1250                                               struct net_device *dev)
1251 {
1252         struct rtnl_link_stats64 *sp;
1253         struct nlattr *attr;
1254
1255         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1256                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1257         if (!attr)
1258                 return -EMSGSIZE;
1259
1260         sp = nla_data(attr);
1261         dev_get_stats(dev, sp);
1262
1263         attr = nla_reserve(skb, IFLA_STATS,
1264                            sizeof(struct rtnl_link_stats));
1265         if (!attr)
1266                 return -EMSGSIZE;
1267
1268         copy_rtnl_link_stats(nla_data(attr), sp);
1269
1270         return 0;
1271 }
1272
1273 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1274                                                struct net_device *dev,
1275                                                int vfs_num,
1276                                                u32 ext_filter_mask)
1277 {
1278         struct ifla_vf_rss_query_en vf_rss_query_en;
1279         struct nlattr *vf, *vfstats, *vfvlanlist;
1280         struct ifla_vf_link_state vf_linkstate;
1281         struct ifla_vf_vlan_info vf_vlan_info;
1282         struct ifla_vf_spoofchk vf_spoofchk;
1283         struct ifla_vf_tx_rate vf_tx_rate;
1284         struct ifla_vf_stats vf_stats;
1285         struct ifla_vf_trust vf_trust;
1286         struct ifla_vf_vlan vf_vlan;
1287         struct ifla_vf_rate vf_rate;
1288         struct ifla_vf_mac vf_mac;
1289         struct ifla_vf_broadcast vf_broadcast;
1290         struct ifla_vf_info ivi;
1291         struct ifla_vf_guid node_guid;
1292         struct ifla_vf_guid port_guid;
1293
1294         memset(&ivi, 0, sizeof(ivi));
1295
1296         /* Not all SR-IOV capable drivers support the
1297          * spoofcheck and "RSS query enable" query.  Preset to
1298          * -1 so the user space tool can detect that the driver
1299          * didn't report anything.
1300          */
1301         ivi.spoofchk = -1;
1302         ivi.rss_query_en = -1;
1303         ivi.trusted = -1;
1304         /* The default value for VF link state is "auto"
1305          * IFLA_VF_LINK_STATE_AUTO which equals zero
1306          */
1307         ivi.linkstate = 0;
1308         /* VLAN Protocol by default is 802.1Q */
1309         ivi.vlan_proto = htons(ETH_P_8021Q);
1310         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1311                 return 0;
1312
1313         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1314         memset(&node_guid, 0, sizeof(node_guid));
1315         memset(&port_guid, 0, sizeof(port_guid));
1316
1317         vf_mac.vf =
1318                 vf_vlan.vf =
1319                 vf_vlan_info.vf =
1320                 vf_rate.vf =
1321                 vf_tx_rate.vf =
1322                 vf_spoofchk.vf =
1323                 vf_linkstate.vf =
1324                 vf_rss_query_en.vf =
1325                 vf_trust.vf =
1326                 node_guid.vf =
1327                 port_guid.vf = ivi.vf;
1328
1329         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1330         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1331         vf_vlan.vlan = ivi.vlan;
1332         vf_vlan.qos = ivi.qos;
1333         vf_vlan_info.vlan = ivi.vlan;
1334         vf_vlan_info.qos = ivi.qos;
1335         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1336         vf_tx_rate.rate = ivi.max_tx_rate;
1337         vf_rate.min_tx_rate = ivi.min_tx_rate;
1338         vf_rate.max_tx_rate = ivi.max_tx_rate;
1339         vf_spoofchk.setting = ivi.spoofchk;
1340         vf_linkstate.link_state = ivi.linkstate;
1341         vf_rss_query_en.setting = ivi.rss_query_en;
1342         vf_trust.setting = ivi.trusted;
1343         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1344         if (!vf)
1345                 return -EMSGSIZE;
1346         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1347             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1348             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1349             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1350                     &vf_rate) ||
1351             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1352                     &vf_tx_rate) ||
1353             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1354                     &vf_spoofchk) ||
1355             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1356                     &vf_linkstate) ||
1357             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1358                     sizeof(vf_rss_query_en),
1359                     &vf_rss_query_en) ||
1360             nla_put(skb, IFLA_VF_TRUST,
1361                     sizeof(vf_trust), &vf_trust))
1362                 goto nla_put_vf_failure;
1363
1364         if (dev->netdev_ops->ndo_get_vf_guid &&
1365             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1366                                               &port_guid)) {
1367                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1368                             &node_guid) ||
1369                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1370                             &port_guid))
1371                         goto nla_put_vf_failure;
1372         }
1373         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1374         if (!vfvlanlist)
1375                 goto nla_put_vf_failure;
1376         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1377                     &vf_vlan_info)) {
1378                 nla_nest_cancel(skb, vfvlanlist);
1379                 goto nla_put_vf_failure;
1380         }
1381         nla_nest_end(skb, vfvlanlist);
1382         if (~ext_filter_mask & RTEXT_FILTER_SKIP_STATS) {
1383                 memset(&vf_stats, 0, sizeof(vf_stats));
1384                 if (dev->netdev_ops->ndo_get_vf_stats)
1385                         dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1386                                                           &vf_stats);
1387                 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1388                 if (!vfstats)
1389                         goto nla_put_vf_failure;
1390                 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1391                                       vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1392                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1393                                       vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1394                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1395                                       vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1396                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1397                                       vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1398                     nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1399                                       vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1400                     nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1401                                       vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1402                     nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1403                                       vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1404                     nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1405                                       vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1406                         nla_nest_cancel(skb, vfstats);
1407                         goto nla_put_vf_failure;
1408                 }
1409                 nla_nest_end(skb, vfstats);
1410         }
1411         nla_nest_end(skb, vf);
1412         return 0;
1413
1414 nla_put_vf_failure:
1415         nla_nest_cancel(skb, vf);
1416         return -EMSGSIZE;
1417 }
1418
1419 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1420                                            struct net_device *dev,
1421                                            u32 ext_filter_mask)
1422 {
1423         struct nlattr *vfinfo;
1424         int i, num_vfs;
1425
1426         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1427                 return 0;
1428
1429         num_vfs = dev_num_vf(dev->dev.parent);
1430         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1431                 return -EMSGSIZE;
1432
1433         if (!dev->netdev_ops->ndo_get_vf_config)
1434                 return 0;
1435
1436         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1437         if (!vfinfo)
1438                 return -EMSGSIZE;
1439
1440         for (i = 0; i < num_vfs; i++) {
1441                 if (rtnl_fill_vfinfo(skb, dev, i, ext_filter_mask)) {
1442                         nla_nest_cancel(skb, vfinfo);
1443                         return -EMSGSIZE;
1444                 }
1445         }
1446
1447         nla_nest_end(skb, vfinfo);
1448         return 0;
1449 }
1450
1451 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1452 {
1453         struct rtnl_link_ifmap map;
1454
1455         memset(&map, 0, sizeof(map));
1456         map.mem_start   = dev->mem_start;
1457         map.mem_end     = dev->mem_end;
1458         map.base_addr   = dev->base_addr;
1459         map.irq         = dev->irq;
1460         map.dma         = dev->dma;
1461         map.port        = dev->if_port;
1462
1463         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1464                 return -EMSGSIZE;
1465
1466         return 0;
1467 }
1468
1469 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1470 {
1471         const struct bpf_prog *generic_xdp_prog;
1472
1473         ASSERT_RTNL();
1474
1475         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1476         if (!generic_xdp_prog)
1477                 return 0;
1478         return generic_xdp_prog->aux->id;
1479 }
1480
1481 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1482 {
1483         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1484 }
1485
1486 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1487 {
1488         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1489 }
1490
1491 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1492                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1493                                u32 (*get_prog_id)(struct net_device *dev))
1494 {
1495         u32 curr_id;
1496         int err;
1497
1498         curr_id = get_prog_id(dev);
1499         if (!curr_id)
1500                 return 0;
1501
1502         *prog_id = curr_id;
1503         err = nla_put_u32(skb, attr, curr_id);
1504         if (err)
1505                 return err;
1506
1507         if (*mode != XDP_ATTACHED_NONE)
1508                 *mode = XDP_ATTACHED_MULTI;
1509         else
1510                 *mode = tgt_mode;
1511
1512         return 0;
1513 }
1514
1515 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1516 {
1517         struct nlattr *xdp;
1518         u32 prog_id;
1519         int err;
1520         u8 mode;
1521
1522         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1523         if (!xdp)
1524                 return -EMSGSIZE;
1525
1526         prog_id = 0;
1527         mode = XDP_ATTACHED_NONE;
1528         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1529                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1530         if (err)
1531                 goto err_cancel;
1532         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1533                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1534         if (err)
1535                 goto err_cancel;
1536         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1537                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1538         if (err)
1539                 goto err_cancel;
1540
1541         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1542         if (err)
1543                 goto err_cancel;
1544
1545         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1546                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1547                 if (err)
1548                         goto err_cancel;
1549         }
1550
1551         nla_nest_end(skb, xdp);
1552         return 0;
1553
1554 err_cancel:
1555         nla_nest_cancel(skb, xdp);
1556         return err;
1557 }
1558
1559 static u32 rtnl_get_event(unsigned long event)
1560 {
1561         u32 rtnl_event_type = IFLA_EVENT_NONE;
1562
1563         switch (event) {
1564         case NETDEV_REBOOT:
1565                 rtnl_event_type = IFLA_EVENT_REBOOT;
1566                 break;
1567         case NETDEV_FEAT_CHANGE:
1568                 rtnl_event_type = IFLA_EVENT_FEATURES;
1569                 break;
1570         case NETDEV_BONDING_FAILOVER:
1571                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1572                 break;
1573         case NETDEV_NOTIFY_PEERS:
1574                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1575                 break;
1576         case NETDEV_RESEND_IGMP:
1577                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1578                 break;
1579         case NETDEV_CHANGEINFODATA:
1580                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1581                 break;
1582         default:
1583                 break;
1584         }
1585
1586         return rtnl_event_type;
1587 }
1588
1589 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1590 {
1591         const struct net_device *upper_dev;
1592         int ret = 0;
1593
1594         rcu_read_lock();
1595
1596         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1597         if (upper_dev)
1598                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1599
1600         rcu_read_unlock();
1601         return ret;
1602 }
1603
1604 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1605                           bool force)
1606 {
1607         int ifindex = dev_get_iflink(dev);
1608
1609         if (force || dev->ifindex != ifindex)
1610                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1611
1612         return 0;
1613 }
1614
1615 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1616                                               struct net_device *dev)
1617 {
1618         char buf[IFALIASZ];
1619         int ret;
1620
1621         ret = dev_get_alias(dev, buf, sizeof(buf));
1622         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1623 }
1624
1625 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1626                                   const struct net_device *dev,
1627                                   struct net *src_net, gfp_t gfp)
1628 {
1629         bool put_iflink = false;
1630
1631         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1632                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1633
1634                 if (!net_eq(dev_net(dev), link_net)) {
1635                         int id = peernet2id_alloc(src_net, link_net, gfp);
1636
1637                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1638                                 return -EMSGSIZE;
1639
1640                         put_iflink = true;
1641                 }
1642         }
1643
1644         return nla_put_iflink(skb, dev, put_iflink);
1645 }
1646
1647 static int rtnl_fill_link_af(struct sk_buff *skb,
1648                              const struct net_device *dev,
1649                              u32 ext_filter_mask)
1650 {
1651         const struct rtnl_af_ops *af_ops;
1652         struct nlattr *af_spec;
1653
1654         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1655         if (!af_spec)
1656                 return -EMSGSIZE;
1657
1658         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1659                 struct nlattr *af;
1660                 int err;
1661
1662                 if (!af_ops->fill_link_af)
1663                         continue;
1664
1665                 af = nla_nest_start_noflag(skb, af_ops->family);
1666                 if (!af)
1667                         return -EMSGSIZE;
1668
1669                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1670                 /*
1671                  * Caller may return ENODATA to indicate that there
1672                  * was no data to be dumped. This is not an error, it
1673                  * means we should trim the attribute header and
1674                  * continue.
1675                  */
1676                 if (err == -ENODATA)
1677                         nla_nest_cancel(skb, af);
1678                 else if (err < 0)
1679                         return -EMSGSIZE;
1680
1681                 nla_nest_end(skb, af);
1682         }
1683
1684         nla_nest_end(skb, af_spec);
1685         return 0;
1686 }
1687
1688 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1689                                  const struct net_device *dev)
1690 {
1691         struct netdev_name_node *name_node;
1692         int count = 0;
1693
1694         list_for_each_entry(name_node, &dev->name_node->list, list) {
1695                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1696                         return -EMSGSIZE;
1697                 count++;
1698         }
1699         return count;
1700 }
1701
1702 static int rtnl_fill_prop_list(struct sk_buff *skb,
1703                                const struct net_device *dev)
1704 {
1705         struct nlattr *prop_list;
1706         int ret;
1707
1708         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1709         if (!prop_list)
1710                 return -EMSGSIZE;
1711
1712         ret = rtnl_fill_alt_ifnames(skb, dev);
1713         if (ret <= 0)
1714                 goto nest_cancel;
1715
1716         nla_nest_end(skb, prop_list);
1717         return 0;
1718
1719 nest_cancel:
1720         nla_nest_cancel(skb, prop_list);
1721         return ret;
1722 }
1723
1724 static int rtnl_fill_proto_down(struct sk_buff *skb,
1725                                 const struct net_device *dev)
1726 {
1727         struct nlattr *pr;
1728         u32 preason;
1729
1730         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1731                 goto nla_put_failure;
1732
1733         preason = dev->proto_down_reason;
1734         if (!preason)
1735                 return 0;
1736
1737         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1738         if (!pr)
1739                 return -EMSGSIZE;
1740
1741         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1742                 nla_nest_cancel(skb, pr);
1743                 goto nla_put_failure;
1744         }
1745
1746         nla_nest_end(skb, pr);
1747         return 0;
1748
1749 nla_put_failure:
1750         return -EMSGSIZE;
1751 }
1752
1753 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1754                                   const struct net_device *dev)
1755 {
1756         struct nlattr *devlink_port_nest;
1757         int ret;
1758
1759         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1760         if (!devlink_port_nest)
1761                 return -EMSGSIZE;
1762
1763         if (dev->devlink_port) {
1764                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1765                 if (ret < 0)
1766                         goto nest_cancel;
1767         }
1768
1769         nla_nest_end(skb, devlink_port_nest);
1770         return 0;
1771
1772 nest_cancel:
1773         nla_nest_cancel(skb, devlink_port_nest);
1774         return ret;
1775 }
1776
1777 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1778                             struct net_device *dev, struct net *src_net,
1779                             int type, u32 pid, u32 seq, u32 change,
1780                             unsigned int flags, u32 ext_filter_mask,
1781                             u32 event, int *new_nsid, int new_ifindex,
1782                             int tgt_netnsid, gfp_t gfp)
1783 {
1784         struct ifinfomsg *ifm;
1785         struct nlmsghdr *nlh;
1786         struct Qdisc *qdisc;
1787
1788         ASSERT_RTNL();
1789         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1790         if (nlh == NULL)
1791                 return -EMSGSIZE;
1792
1793         ifm = nlmsg_data(nlh);
1794         ifm->ifi_family = AF_UNSPEC;
1795         ifm->__ifi_pad = 0;
1796         ifm->ifi_type = dev->type;
1797         ifm->ifi_index = dev->ifindex;
1798         ifm->ifi_flags = dev_get_flags(dev);
1799         ifm->ifi_change = change;
1800
1801         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1802                 goto nla_put_failure;
1803
1804         qdisc = rtnl_dereference(dev->qdisc);
1805         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1806             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1807             nla_put_u8(skb, IFLA_OPERSTATE,
1808                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1809             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1810             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1811             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1812             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1813             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1814             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1815             nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1816             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1817             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1818             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1819             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1820             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE, dev->gso_ipv4_max_size) ||
1821             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE, dev->gro_ipv4_max_size) ||
1822             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1823             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1824 #ifdef CONFIG_RPS
1825             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1826 #endif
1827             put_master_ifindex(skb, dev) ||
1828             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1829             (qdisc &&
1830              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1831             nla_put_ifalias(skb, dev) ||
1832             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1833                         atomic_read(&dev->carrier_up_count) +
1834                         atomic_read(&dev->carrier_down_count)) ||
1835             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1836                         atomic_read(&dev->carrier_up_count)) ||
1837             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1838                         atomic_read(&dev->carrier_down_count)))
1839                 goto nla_put_failure;
1840
1841         if (rtnl_fill_proto_down(skb, dev))
1842                 goto nla_put_failure;
1843
1844         if (event != IFLA_EVENT_NONE) {
1845                 if (nla_put_u32(skb, IFLA_EVENT, event))
1846                         goto nla_put_failure;
1847         }
1848
1849         if (rtnl_fill_link_ifmap(skb, dev))
1850                 goto nla_put_failure;
1851
1852         if (dev->addr_len) {
1853                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1854                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1855                         goto nla_put_failure;
1856         }
1857
1858         if (rtnl_phys_port_id_fill(skb, dev))
1859                 goto nla_put_failure;
1860
1861         if (rtnl_phys_port_name_fill(skb, dev))
1862                 goto nla_put_failure;
1863
1864         if (rtnl_phys_switch_id_fill(skb, dev))
1865                 goto nla_put_failure;
1866
1867         if (rtnl_fill_stats(skb, dev))
1868                 goto nla_put_failure;
1869
1870         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1871                 goto nla_put_failure;
1872
1873         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1874                 goto nla_put_failure;
1875
1876         if (rtnl_xdp_fill(skb, dev))
1877                 goto nla_put_failure;
1878
1879         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1880                 if (rtnl_link_fill(skb, dev) < 0)
1881                         goto nla_put_failure;
1882         }
1883
1884         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1885                 goto nla_put_failure;
1886
1887         if (new_nsid &&
1888             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1889                 goto nla_put_failure;
1890         if (new_ifindex &&
1891             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1892                 goto nla_put_failure;
1893
1894         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1895             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1896                 goto nla_put_failure;
1897
1898         rcu_read_lock();
1899         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1900                 goto nla_put_failure_rcu;
1901         rcu_read_unlock();
1902
1903         if (rtnl_fill_prop_list(skb, dev))
1904                 goto nla_put_failure;
1905
1906         if (dev->dev.parent &&
1907             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1908                            dev_name(dev->dev.parent)))
1909                 goto nla_put_failure;
1910
1911         if (dev->dev.parent && dev->dev.parent->bus &&
1912             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1913                            dev->dev.parent->bus->name))
1914                 goto nla_put_failure;
1915
1916         if (rtnl_fill_devlink_port(skb, dev))
1917                 goto nla_put_failure;
1918
1919         nlmsg_end(skb, nlh);
1920         return 0;
1921
1922 nla_put_failure_rcu:
1923         rcu_read_unlock();
1924 nla_put_failure:
1925         nlmsg_cancel(skb, nlh);
1926         return -EMSGSIZE;
1927 }
1928
1929 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1930         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1931         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1932         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1933         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1934         [IFLA_MTU]              = { .type = NLA_U32 },
1935         [IFLA_LINK]             = { .type = NLA_U32 },
1936         [IFLA_MASTER]           = { .type = NLA_U32 },
1937         [IFLA_CARRIER]          = { .type = NLA_U8 },
1938         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1939         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1940         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1941         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1942         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1943         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1944         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1945         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1946          * allow 0-length string (needed to remove an alias).
1947          */
1948         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1949         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1950         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1951         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1952         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1953         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1954         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1955         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1956         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1957         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1958         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1959         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1960         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1961         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1962         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1963         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1964         [IFLA_XDP]              = { .type = NLA_NESTED },
1965         [IFLA_EVENT]            = { .type = NLA_U32 },
1966         [IFLA_GROUP]            = { .type = NLA_U32 },
1967         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1968         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1969         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1970         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1971         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1972         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1973         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1974                                     .len = ALTIFNAMSIZ - 1 },
1975         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1976         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1977         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1978         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1979         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
1980         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
1981         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
1982         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
1983         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1984         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1985 };
1986
1987 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1988         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1989         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1990         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1991         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1992 };
1993
1994 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1995         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1996         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1997         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1998         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1999         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
2000         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
2001         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
2002         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
2003         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
2004         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2005         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2006         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2007         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2008 };
2009
2010 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2011         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2012         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2013                                     .len = PORT_PROFILE_MAX },
2014         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2015                                       .len = PORT_UUID_MAX },
2016         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2017                                     .len = PORT_UUID_MAX },
2018         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2019         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2020
2021         /* Unused, but we need to keep it here since user space could
2022          * fill it. It's also broken with regard to NLA_BINARY use in
2023          * combination with structs.
2024          */
2025         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2026                                     .len = sizeof(struct ifla_port_vsi) },
2027 };
2028
2029 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2030         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2031         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2032         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2033         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2034         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2035         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2036 };
2037
2038 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2039 {
2040         const struct rtnl_link_ops *ops = NULL;
2041         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2042
2043         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2044                 return NULL;
2045
2046         if (linfo[IFLA_INFO_KIND]) {
2047                 char kind[MODULE_NAME_LEN];
2048
2049                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2050                 ops = rtnl_link_ops_get(kind);
2051         }
2052
2053         return ops;
2054 }
2055
2056 static bool link_master_filtered(struct net_device *dev, int master_idx)
2057 {
2058         struct net_device *master;
2059
2060         if (!master_idx)
2061                 return false;
2062
2063         master = netdev_master_upper_dev_get(dev);
2064
2065         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2066          * another invalid value for ifindex to denote "no master".
2067          */
2068         if (master_idx == -1)
2069                 return !!master;
2070
2071         if (!master || master->ifindex != master_idx)
2072                 return true;
2073
2074         return false;
2075 }
2076
2077 static bool link_kind_filtered(const struct net_device *dev,
2078                                const struct rtnl_link_ops *kind_ops)
2079 {
2080         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2081                 return true;
2082
2083         return false;
2084 }
2085
2086 static bool link_dump_filtered(struct net_device *dev,
2087                                int master_idx,
2088                                const struct rtnl_link_ops *kind_ops)
2089 {
2090         if (link_master_filtered(dev, master_idx) ||
2091             link_kind_filtered(dev, kind_ops))
2092                 return true;
2093
2094         return false;
2095 }
2096
2097 /**
2098  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2099  * @sk: netlink socket
2100  * @netnsid: network namespace identifier
2101  *
2102  * Returns the network namespace identified by netnsid on success or an error
2103  * pointer on failure.
2104  */
2105 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2106 {
2107         struct net *net;
2108
2109         net = get_net_ns_by_id(sock_net(sk), netnsid);
2110         if (!net)
2111                 return ERR_PTR(-EINVAL);
2112
2113         /* For now, the caller is required to have CAP_NET_ADMIN in
2114          * the user namespace owning the target net ns.
2115          */
2116         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2117                 put_net(net);
2118                 return ERR_PTR(-EACCES);
2119         }
2120         return net;
2121 }
2122 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2123
2124 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2125                                       bool strict_check, struct nlattr **tb,
2126                                       struct netlink_ext_ack *extack)
2127 {
2128         int hdrlen;
2129
2130         if (strict_check) {
2131                 struct ifinfomsg *ifm;
2132
2133                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2134                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2135                         return -EINVAL;
2136                 }
2137
2138                 ifm = nlmsg_data(nlh);
2139                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2140                     ifm->ifi_change) {
2141                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2142                         return -EINVAL;
2143                 }
2144                 if (ifm->ifi_index) {
2145                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2146                         return -EINVAL;
2147                 }
2148
2149                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2150                                                      IFLA_MAX, ifla_policy,
2151                                                      extack);
2152         }
2153
2154         /* A hack to preserve kernel<->userspace interface.
2155          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2156          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2157          * what iproute2 < v3.9.0 used.
2158          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2159          * attribute, its netlink message is shorter than struct ifinfomsg.
2160          */
2161         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2162                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2163
2164         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2165                                       extack);
2166 }
2167
2168 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2169 {
2170         struct netlink_ext_ack *extack = cb->extack;
2171         const struct nlmsghdr *nlh = cb->nlh;
2172         struct net *net = sock_net(skb->sk);
2173         struct net *tgt_net = net;
2174         int h, s_h;
2175         int idx = 0, s_idx;
2176         struct net_device *dev;
2177         struct hlist_head *head;
2178         struct nlattr *tb[IFLA_MAX+1];
2179         u32 ext_filter_mask = 0;
2180         const struct rtnl_link_ops *kind_ops = NULL;
2181         unsigned int flags = NLM_F_MULTI;
2182         int master_idx = 0;
2183         int netnsid = -1;
2184         int err, i;
2185
2186         s_h = cb->args[0];
2187         s_idx = cb->args[1];
2188
2189         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2190         if (err < 0) {
2191                 if (cb->strict_check)
2192                         return err;
2193
2194                 goto walk_entries;
2195         }
2196
2197         for (i = 0; i <= IFLA_MAX; ++i) {
2198                 if (!tb[i])
2199                         continue;
2200
2201                 /* new attributes should only be added with strict checking */
2202                 switch (i) {
2203                 case IFLA_TARGET_NETNSID:
2204                         netnsid = nla_get_s32(tb[i]);
2205                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2206                         if (IS_ERR(tgt_net)) {
2207                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2208                                 return PTR_ERR(tgt_net);
2209                         }
2210                         break;
2211                 case IFLA_EXT_MASK:
2212                         ext_filter_mask = nla_get_u32(tb[i]);
2213                         break;
2214                 case IFLA_MASTER:
2215                         master_idx = nla_get_u32(tb[i]);
2216                         break;
2217                 case IFLA_LINKINFO:
2218                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2219                         break;
2220                 default:
2221                         if (cb->strict_check) {
2222                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2223                                 return -EINVAL;
2224                         }
2225                 }
2226         }
2227
2228         if (master_idx || kind_ops)
2229                 flags |= NLM_F_DUMP_FILTERED;
2230
2231 walk_entries:
2232         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2233                 idx = 0;
2234                 head = &tgt_net->dev_index_head[h];
2235                 hlist_for_each_entry(dev, head, index_hlist) {
2236                         if (link_dump_filtered(dev, master_idx, kind_ops))
2237                                 goto cont;
2238                         if (idx < s_idx)
2239                                 goto cont;
2240                         err = rtnl_fill_ifinfo(skb, dev, net,
2241                                                RTM_NEWLINK,
2242                                                NETLINK_CB(cb->skb).portid,
2243                                                nlh->nlmsg_seq, 0, flags,
2244                                                ext_filter_mask, 0, NULL, 0,
2245                                                netnsid, GFP_KERNEL);
2246
2247                         if (err < 0) {
2248                                 if (likely(skb->len))
2249                                         goto out;
2250
2251                                 goto out_err;
2252                         }
2253 cont:
2254                         idx++;
2255                 }
2256         }
2257 out:
2258         err = skb->len;
2259 out_err:
2260         cb->args[1] = idx;
2261         cb->args[0] = h;
2262         cb->seq = tgt_net->dev_base_seq;
2263         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2264         if (netnsid >= 0)
2265                 put_net(tgt_net);
2266
2267         return err;
2268 }
2269
2270 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2271                         struct netlink_ext_ack *exterr)
2272 {
2273         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2274                                     exterr);
2275 }
2276 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2277
2278 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2279 {
2280         struct net *net;
2281         /* Examine the link attributes and figure out which
2282          * network namespace we are talking about.
2283          */
2284         if (tb[IFLA_NET_NS_PID])
2285                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2286         else if (tb[IFLA_NET_NS_FD])
2287                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2288         else
2289                 net = get_net(src_net);
2290         return net;
2291 }
2292 EXPORT_SYMBOL(rtnl_link_get_net);
2293
2294 /* Figure out which network namespace we are talking about by
2295  * examining the link attributes in the following order:
2296  *
2297  * 1. IFLA_NET_NS_PID
2298  * 2. IFLA_NET_NS_FD
2299  * 3. IFLA_TARGET_NETNSID
2300  */
2301 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2302                                                struct nlattr *tb[])
2303 {
2304         struct net *net;
2305
2306         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2307                 return rtnl_link_get_net(src_net, tb);
2308
2309         if (!tb[IFLA_TARGET_NETNSID])
2310                 return get_net(src_net);
2311
2312         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2313         if (!net)
2314                 return ERR_PTR(-EINVAL);
2315
2316         return net;
2317 }
2318
2319 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2320                                              struct net *src_net,
2321                                              struct nlattr *tb[], int cap)
2322 {
2323         struct net *net;
2324
2325         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2326         if (IS_ERR(net))
2327                 return net;
2328
2329         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2330                 put_net(net);
2331                 return ERR_PTR(-EPERM);
2332         }
2333
2334         return net;
2335 }
2336
2337 /* Verify that rtnetlink requests do not pass additional properties
2338  * potentially referring to different network namespaces.
2339  */
2340 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2341                                     struct netlink_ext_ack *extack,
2342                                     bool netns_id_only)
2343 {
2344
2345         if (netns_id_only) {
2346                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2347                         return 0;
2348
2349                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2350                 return -EOPNOTSUPP;
2351         }
2352
2353         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2354                 goto invalid_attr;
2355
2356         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2357                 goto invalid_attr;
2358
2359         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2360                 goto invalid_attr;
2361
2362         return 0;
2363
2364 invalid_attr:
2365         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2366         return -EINVAL;
2367 }
2368
2369 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2370                              int max_tx_rate)
2371 {
2372         const struct net_device_ops *ops = dev->netdev_ops;
2373
2374         if (!ops->ndo_set_vf_rate)
2375                 return -EOPNOTSUPP;
2376         if (max_tx_rate && max_tx_rate < min_tx_rate)
2377                 return -EINVAL;
2378
2379         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2380 }
2381
2382 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2383                             struct netlink_ext_ack *extack)
2384 {
2385         if (tb[IFLA_ADDRESS] &&
2386             nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2387                 return -EINVAL;
2388
2389         if (tb[IFLA_BROADCAST] &&
2390             nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2391                 return -EINVAL;
2392
2393         if (tb[IFLA_GSO_MAX_SIZE] &&
2394             nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2395                 NL_SET_ERR_MSG(extack, "too big gso_max_size");
2396                 return -EINVAL;
2397         }
2398
2399         if (tb[IFLA_GSO_MAX_SEGS] &&
2400             (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2401              nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2402                 NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2403                 return -EINVAL;
2404         }
2405
2406         if (tb[IFLA_GRO_MAX_SIZE] &&
2407             nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2408                 NL_SET_ERR_MSG(extack, "too big gro_max_size");
2409                 return -EINVAL;
2410         }
2411
2412         if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2413             nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2414                 NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2415                 return -EINVAL;
2416         }
2417
2418         if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2419             nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2420                 NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2421                 return -EINVAL;
2422         }
2423
2424         if (tb[IFLA_AF_SPEC]) {
2425                 struct nlattr *af;
2426                 int rem, err;
2427
2428                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2429                         const struct rtnl_af_ops *af_ops;
2430
2431                         af_ops = rtnl_af_lookup(nla_type(af));
2432                         if (!af_ops)
2433                                 return -EAFNOSUPPORT;
2434
2435                         if (!af_ops->set_link_af)
2436                                 return -EOPNOTSUPP;
2437
2438                         if (af_ops->validate_link_af) {
2439                                 err = af_ops->validate_link_af(dev, af, extack);
2440                                 if (err < 0)
2441                                         return err;
2442                         }
2443                 }
2444         }
2445
2446         return 0;
2447 }
2448
2449 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2450                                   int guid_type)
2451 {
2452         const struct net_device_ops *ops = dev->netdev_ops;
2453
2454         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2455 }
2456
2457 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2458 {
2459         if (dev->type != ARPHRD_INFINIBAND)
2460                 return -EOPNOTSUPP;
2461
2462         return handle_infiniband_guid(dev, ivt, guid_type);
2463 }
2464
2465 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2466 {
2467         const struct net_device_ops *ops = dev->netdev_ops;
2468         int err = -EINVAL;
2469
2470         if (tb[IFLA_VF_MAC]) {
2471                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2472
2473                 if (ivm->vf >= INT_MAX)
2474                         return -EINVAL;
2475                 err = -EOPNOTSUPP;
2476                 if (ops->ndo_set_vf_mac)
2477                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2478                                                   ivm->mac);
2479                 if (err < 0)
2480                         return err;
2481         }
2482
2483         if (tb[IFLA_VF_VLAN]) {
2484                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2485
2486                 if (ivv->vf >= INT_MAX)
2487                         return -EINVAL;
2488                 err = -EOPNOTSUPP;
2489                 if (ops->ndo_set_vf_vlan)
2490                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2491                                                    ivv->qos,
2492                                                    htons(ETH_P_8021Q));
2493                 if (err < 0)
2494                         return err;
2495         }
2496
2497         if (tb[IFLA_VF_VLAN_LIST]) {
2498                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2499                 struct nlattr *attr;
2500                 int rem, len = 0;
2501
2502                 err = -EOPNOTSUPP;
2503                 if (!ops->ndo_set_vf_vlan)
2504                         return err;
2505
2506                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2507                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2508                             nla_len(attr) < NLA_HDRLEN) {
2509                                 return -EINVAL;
2510                         }
2511                         if (len >= MAX_VLAN_LIST_LEN)
2512                                 return -EOPNOTSUPP;
2513                         ivvl[len] = nla_data(attr);
2514
2515                         len++;
2516                 }
2517                 if (len == 0)
2518                         return -EINVAL;
2519
2520                 if (ivvl[0]->vf >= INT_MAX)
2521                         return -EINVAL;
2522                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2523                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2524                 if (err < 0)
2525                         return err;
2526         }
2527
2528         if (tb[IFLA_VF_TX_RATE]) {
2529                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2530                 struct ifla_vf_info ivf;
2531
2532                 if (ivt->vf >= INT_MAX)
2533                         return -EINVAL;
2534                 err = -EOPNOTSUPP;
2535                 if (ops->ndo_get_vf_config)
2536                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2537                 if (err < 0)
2538                         return err;
2539
2540                 err = rtnl_set_vf_rate(dev, ivt->vf,
2541                                        ivf.min_tx_rate, ivt->rate);
2542                 if (err < 0)
2543                         return err;
2544         }
2545
2546         if (tb[IFLA_VF_RATE]) {
2547                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2548
2549                 if (ivt->vf >= INT_MAX)
2550                         return -EINVAL;
2551
2552                 err = rtnl_set_vf_rate(dev, ivt->vf,
2553                                        ivt->min_tx_rate, ivt->max_tx_rate);
2554                 if (err < 0)
2555                         return err;
2556         }
2557
2558         if (tb[IFLA_VF_SPOOFCHK]) {
2559                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2560
2561                 if (ivs->vf >= INT_MAX)
2562                         return -EINVAL;
2563                 err = -EOPNOTSUPP;
2564                 if (ops->ndo_set_vf_spoofchk)
2565                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2566                                                        ivs->setting);
2567                 if (err < 0)
2568                         return err;
2569         }
2570
2571         if (tb[IFLA_VF_LINK_STATE]) {
2572                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2573
2574                 if (ivl->vf >= INT_MAX)
2575                         return -EINVAL;
2576                 err = -EOPNOTSUPP;
2577                 if (ops->ndo_set_vf_link_state)
2578                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2579                                                          ivl->link_state);
2580                 if (err < 0)
2581                         return err;
2582         }
2583
2584         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2585                 struct ifla_vf_rss_query_en *ivrssq_en;
2586
2587                 err = -EOPNOTSUPP;
2588                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2589                 if (ivrssq_en->vf >= INT_MAX)
2590                         return -EINVAL;
2591                 if (ops->ndo_set_vf_rss_query_en)
2592                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2593                                                            ivrssq_en->setting);
2594                 if (err < 0)
2595                         return err;
2596         }
2597
2598         if (tb[IFLA_VF_TRUST]) {
2599                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2600
2601                 if (ivt->vf >= INT_MAX)
2602                         return -EINVAL;
2603                 err = -EOPNOTSUPP;
2604                 if (ops->ndo_set_vf_trust)
2605                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2606                 if (err < 0)
2607                         return err;
2608         }
2609
2610         if (tb[IFLA_VF_IB_NODE_GUID]) {
2611                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2612
2613                 if (ivt->vf >= INT_MAX)
2614                         return -EINVAL;
2615                 if (!ops->ndo_set_vf_guid)
2616                         return -EOPNOTSUPP;
2617                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2618         }
2619
2620         if (tb[IFLA_VF_IB_PORT_GUID]) {
2621                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2622
2623                 if (ivt->vf >= INT_MAX)
2624                         return -EINVAL;
2625                 if (!ops->ndo_set_vf_guid)
2626                         return -EOPNOTSUPP;
2627
2628                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2629         }
2630
2631         return err;
2632 }
2633
2634 static int do_set_master(struct net_device *dev, int ifindex,
2635                          struct netlink_ext_ack *extack)
2636 {
2637         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2638         const struct net_device_ops *ops;
2639         int err;
2640
2641         if (upper_dev) {
2642                 if (upper_dev->ifindex == ifindex)
2643                         return 0;
2644                 ops = upper_dev->netdev_ops;
2645                 if (ops->ndo_del_slave) {
2646                         err = ops->ndo_del_slave(upper_dev, dev);
2647                         if (err)
2648                                 return err;
2649                 } else {
2650                         return -EOPNOTSUPP;
2651                 }
2652         }
2653
2654         if (ifindex) {
2655                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2656                 if (!upper_dev)
2657                         return -EINVAL;
2658                 ops = upper_dev->netdev_ops;
2659                 if (ops->ndo_add_slave) {
2660                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2661                         if (err)
2662                                 return err;
2663                 } else {
2664                         return -EOPNOTSUPP;
2665                 }
2666         }
2667         return 0;
2668 }
2669
2670 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2671         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2672         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2673 };
2674
2675 static int do_set_proto_down(struct net_device *dev,
2676                              struct nlattr *nl_proto_down,
2677                              struct nlattr *nl_proto_down_reason,
2678                              struct netlink_ext_ack *extack)
2679 {
2680         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2681         unsigned long mask = 0;
2682         u32 value;
2683         bool proto_down;
2684         int err;
2685
2686         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2687                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2688                 return -EOPNOTSUPP;
2689         }
2690
2691         if (nl_proto_down_reason) {
2692                 err = nla_parse_nested_deprecated(pdreason,
2693                                                   IFLA_PROTO_DOWN_REASON_MAX,
2694                                                   nl_proto_down_reason,
2695                                                   ifla_proto_down_reason_policy,
2696                                                   NULL);
2697                 if (err < 0)
2698                         return err;
2699
2700                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2701                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2702                         return -EINVAL;
2703                 }
2704
2705                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2706
2707                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2708                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2709
2710                 dev_change_proto_down_reason(dev, mask, value);
2711         }
2712
2713         if (nl_proto_down) {
2714                 proto_down = nla_get_u8(nl_proto_down);
2715
2716                 /* Don't turn off protodown if there are active reasons */
2717                 if (!proto_down && dev->proto_down_reason) {
2718                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2719                         return -EBUSY;
2720                 }
2721                 err = dev_change_proto_down(dev,
2722                                             proto_down);
2723                 if (err)
2724                         return err;
2725         }
2726
2727         return 0;
2728 }
2729
2730 #define DO_SETLINK_MODIFIED     0x01
2731 /* notify flag means notify + modified. */
2732 #define DO_SETLINK_NOTIFY       0x03
2733 static int do_setlink(const struct sk_buff *skb,
2734                       struct net_device *dev, struct ifinfomsg *ifm,
2735                       struct netlink_ext_ack *extack,
2736                       struct nlattr **tb, int status)
2737 {
2738         const struct net_device_ops *ops = dev->netdev_ops;
2739         char ifname[IFNAMSIZ];
2740         int err;
2741
2742         if (tb[IFLA_IFNAME])
2743                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2744         else
2745                 ifname[0] = '\0';
2746
2747         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2748                 const char *pat = ifname[0] ? ifname : NULL;
2749                 struct net *net;
2750                 int new_ifindex;
2751
2752                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2753                                                 tb, CAP_NET_ADMIN);
2754                 if (IS_ERR(net)) {
2755                         err = PTR_ERR(net);
2756                         goto errout;
2757                 }
2758
2759                 if (tb[IFLA_NEW_IFINDEX])
2760                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2761                 else
2762                         new_ifindex = 0;
2763
2764                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2765                 put_net(net);
2766                 if (err)
2767                         goto errout;
2768                 status |= DO_SETLINK_MODIFIED;
2769         }
2770
2771         if (tb[IFLA_MAP]) {
2772                 struct rtnl_link_ifmap *u_map;
2773                 struct ifmap k_map;
2774
2775                 if (!ops->ndo_set_config) {
2776                         err = -EOPNOTSUPP;
2777                         goto errout;
2778                 }
2779
2780                 if (!netif_device_present(dev)) {
2781                         err = -ENODEV;
2782                         goto errout;
2783                 }
2784
2785                 u_map = nla_data(tb[IFLA_MAP]);
2786                 k_map.mem_start = (unsigned long) u_map->mem_start;
2787                 k_map.mem_end = (unsigned long) u_map->mem_end;
2788                 k_map.base_addr = (unsigned short) u_map->base_addr;
2789                 k_map.irq = (unsigned char) u_map->irq;
2790                 k_map.dma = (unsigned char) u_map->dma;
2791                 k_map.port = (unsigned char) u_map->port;
2792
2793                 err = ops->ndo_set_config(dev, &k_map);
2794                 if (err < 0)
2795                         goto errout;
2796
2797                 status |= DO_SETLINK_NOTIFY;
2798         }
2799
2800         if (tb[IFLA_ADDRESS]) {
2801                 struct sockaddr *sa;
2802                 int len;
2803
2804                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2805                                                   sizeof(*sa));
2806                 sa = kmalloc(len, GFP_KERNEL);
2807                 if (!sa) {
2808                         err = -ENOMEM;
2809                         goto errout;
2810                 }
2811                 sa->sa_family = dev->type;
2812                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2813                        dev->addr_len);
2814                 err = dev_set_mac_address_user(dev, sa, extack);
2815                 kfree(sa);
2816                 if (err)
2817                         goto errout;
2818                 status |= DO_SETLINK_MODIFIED;
2819         }
2820
2821         if (tb[IFLA_MTU]) {
2822                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2823                 if (err < 0)
2824                         goto errout;
2825                 status |= DO_SETLINK_MODIFIED;
2826         }
2827
2828         if (tb[IFLA_GROUP]) {
2829                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2830                 status |= DO_SETLINK_NOTIFY;
2831         }
2832
2833         /*
2834          * Interface selected by interface index but interface
2835          * name provided implies that a name change has been
2836          * requested.
2837          */
2838         if (ifm->ifi_index > 0 && ifname[0]) {
2839                 err = dev_change_name(dev, ifname);
2840                 if (err < 0)
2841                         goto errout;
2842                 status |= DO_SETLINK_MODIFIED;
2843         }
2844
2845         if (tb[IFLA_IFALIAS]) {
2846                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2847                                     nla_len(tb[IFLA_IFALIAS]));
2848                 if (err < 0)
2849                         goto errout;
2850                 status |= DO_SETLINK_NOTIFY;
2851         }
2852
2853         if (tb[IFLA_BROADCAST]) {
2854                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2855                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2856         }
2857
2858         if (tb[IFLA_MASTER]) {
2859                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2860                 if (err)
2861                         goto errout;
2862                 status |= DO_SETLINK_MODIFIED;
2863         }
2864
2865         if (ifm->ifi_flags || ifm->ifi_change) {
2866                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2867                                        extack);
2868                 if (err < 0)
2869                         goto errout;
2870         }
2871
2872         if (tb[IFLA_CARRIER]) {
2873                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2874                 if (err)
2875                         goto errout;
2876                 status |= DO_SETLINK_MODIFIED;
2877         }
2878
2879         if (tb[IFLA_TXQLEN]) {
2880                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2881
2882                 err = dev_change_tx_queue_len(dev, value);
2883                 if (err)
2884                         goto errout;
2885                 status |= DO_SETLINK_MODIFIED;
2886         }
2887
2888         if (tb[IFLA_GSO_MAX_SIZE]) {
2889                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2890
2891                 if (dev->gso_max_size ^ max_size) {
2892                         netif_set_gso_max_size(dev, max_size);
2893                         status |= DO_SETLINK_MODIFIED;
2894                 }
2895         }
2896
2897         if (tb[IFLA_GSO_MAX_SEGS]) {
2898                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2899
2900                 if (dev->gso_max_segs ^ max_segs) {
2901                         netif_set_gso_max_segs(dev, max_segs);
2902                         status |= DO_SETLINK_MODIFIED;
2903                 }
2904         }
2905
2906         if (tb[IFLA_GRO_MAX_SIZE]) {
2907                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2908
2909                 if (dev->gro_max_size ^ gro_max_size) {
2910                         netif_set_gro_max_size(dev, gro_max_size);
2911                         status |= DO_SETLINK_MODIFIED;
2912                 }
2913         }
2914
2915         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2916                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2917
2918                 if (dev->gso_ipv4_max_size ^ max_size) {
2919                         netif_set_gso_ipv4_max_size(dev, max_size);
2920                         status |= DO_SETLINK_MODIFIED;
2921                 }
2922         }
2923
2924         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2925                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2926
2927                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2928                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2929                         status |= DO_SETLINK_MODIFIED;
2930                 }
2931         }
2932
2933         if (tb[IFLA_OPERSTATE])
2934                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2935
2936         if (tb[IFLA_LINKMODE]) {
2937                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2938
2939                 write_lock(&dev_base_lock);
2940                 if (dev->link_mode ^ value)
2941                         status |= DO_SETLINK_NOTIFY;
2942                 dev->link_mode = value;
2943                 write_unlock(&dev_base_lock);
2944         }
2945
2946         if (tb[IFLA_VFINFO_LIST]) {
2947                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2948                 struct nlattr *attr;
2949                 int rem;
2950
2951                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2952                         if (nla_type(attr) != IFLA_VF_INFO ||
2953                             nla_len(attr) < NLA_HDRLEN) {
2954                                 err = -EINVAL;
2955                                 goto errout;
2956                         }
2957                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2958                                                           attr,
2959                                                           ifla_vf_policy,
2960                                                           NULL);
2961                         if (err < 0)
2962                                 goto errout;
2963                         err = do_setvfinfo(dev, vfinfo);
2964                         if (err < 0)
2965                                 goto errout;
2966                         status |= DO_SETLINK_NOTIFY;
2967                 }
2968         }
2969         err = 0;
2970
2971         if (tb[IFLA_VF_PORTS]) {
2972                 struct nlattr *port[IFLA_PORT_MAX+1];
2973                 struct nlattr *attr;
2974                 int vf;
2975                 int rem;
2976
2977                 err = -EOPNOTSUPP;
2978                 if (!ops->ndo_set_vf_port)
2979                         goto errout;
2980
2981                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2982                         if (nla_type(attr) != IFLA_VF_PORT ||
2983                             nla_len(attr) < NLA_HDRLEN) {
2984                                 err = -EINVAL;
2985                                 goto errout;
2986                         }
2987                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2988                                                           attr,
2989                                                           ifla_port_policy,
2990                                                           NULL);
2991                         if (err < 0)
2992                                 goto errout;
2993                         if (!port[IFLA_PORT_VF]) {
2994                                 err = -EOPNOTSUPP;
2995                                 goto errout;
2996                         }
2997                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2998                         err = ops->ndo_set_vf_port(dev, vf, port);
2999                         if (err < 0)
3000                                 goto errout;
3001                         status |= DO_SETLINK_NOTIFY;
3002                 }
3003         }
3004         err = 0;
3005
3006         if (tb[IFLA_PORT_SELF]) {
3007                 struct nlattr *port[IFLA_PORT_MAX+1];
3008
3009                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3010                                                   tb[IFLA_PORT_SELF],
3011                                                   ifla_port_policy, NULL);
3012                 if (err < 0)
3013                         goto errout;
3014
3015                 err = -EOPNOTSUPP;
3016                 if (ops->ndo_set_vf_port)
3017                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3018                 if (err < 0)
3019                         goto errout;
3020                 status |= DO_SETLINK_NOTIFY;
3021         }
3022
3023         if (tb[IFLA_AF_SPEC]) {
3024                 struct nlattr *af;
3025                 int rem;
3026
3027                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3028                         const struct rtnl_af_ops *af_ops;
3029
3030                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3031
3032                         err = af_ops->set_link_af(dev, af, extack);
3033                         if (err < 0)
3034                                 goto errout;
3035
3036                         status |= DO_SETLINK_NOTIFY;
3037                 }
3038         }
3039         err = 0;
3040
3041         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3042                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3043                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3044                 if (err)
3045                         goto errout;
3046                 status |= DO_SETLINK_NOTIFY;
3047         }
3048
3049         if (tb[IFLA_XDP]) {
3050                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3051                 u32 xdp_flags = 0;
3052
3053                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3054                                                   tb[IFLA_XDP],
3055                                                   ifla_xdp_policy, NULL);
3056                 if (err < 0)
3057                         goto errout;
3058
3059                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3060                         err = -EINVAL;
3061                         goto errout;
3062                 }
3063
3064                 if (xdp[IFLA_XDP_FLAGS]) {
3065                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3066                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3067                                 err = -EINVAL;
3068                                 goto errout;
3069                         }
3070                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3071                                 err = -EINVAL;
3072                                 goto errout;
3073                         }
3074                 }
3075
3076                 if (xdp[IFLA_XDP_FD]) {
3077                         int expected_fd = -1;
3078
3079                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3080                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3081                                         err = -EINVAL;
3082                                         goto errout;
3083                                 }
3084                                 expected_fd =
3085                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3086                         }
3087
3088                         err = dev_change_xdp_fd(dev, extack,
3089                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3090                                                 expected_fd,
3091                                                 xdp_flags);
3092                         if (err)
3093                                 goto errout;
3094                         status |= DO_SETLINK_NOTIFY;
3095                 }
3096         }
3097
3098 errout:
3099         if (status & DO_SETLINK_MODIFIED) {
3100                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3101                         netdev_state_change(dev);
3102
3103                 if (err < 0)
3104                         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",
3105                                              dev->name);
3106         }
3107
3108         return err;
3109 }
3110
3111 static struct net_device *rtnl_dev_get(struct net *net,
3112                                        struct nlattr *tb[])
3113 {
3114         char ifname[ALTIFNAMSIZ];
3115
3116         if (tb[IFLA_IFNAME])
3117                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3118         else if (tb[IFLA_ALT_IFNAME])
3119                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3120         else
3121                 return NULL;
3122
3123         return __dev_get_by_name(net, ifname);
3124 }
3125
3126 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3127                         struct netlink_ext_ack *extack)
3128 {
3129         struct net *net = sock_net(skb->sk);
3130         struct ifinfomsg *ifm;
3131         struct net_device *dev;
3132         int err;
3133         struct nlattr *tb[IFLA_MAX+1];
3134
3135         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3136                                      ifla_policy, extack);
3137         if (err < 0)
3138                 goto errout;
3139
3140         err = rtnl_ensure_unique_netns(tb, extack, false);
3141         if (err < 0)
3142                 goto errout;
3143
3144         err = -EINVAL;
3145         ifm = nlmsg_data(nlh);
3146         if (ifm->ifi_index > 0)
3147                 dev = __dev_get_by_index(net, ifm->ifi_index);
3148         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3149                 dev = rtnl_dev_get(net, tb);
3150         else
3151                 goto errout;
3152
3153         if (dev == NULL) {
3154                 err = -ENODEV;
3155                 goto errout;
3156         }
3157
3158         err = validate_linkmsg(dev, tb, extack);
3159         if (err < 0)
3160                 goto errout;
3161
3162         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3163 errout:
3164         return err;
3165 }
3166
3167 static int rtnl_group_dellink(const struct net *net, int group)
3168 {
3169         struct net_device *dev, *aux;
3170         LIST_HEAD(list_kill);
3171         bool found = false;
3172
3173         if (!group)
3174                 return -EPERM;
3175
3176         for_each_netdev(net, dev) {
3177                 if (dev->group == group) {
3178                         const struct rtnl_link_ops *ops;
3179
3180                         found = true;
3181                         ops = dev->rtnl_link_ops;
3182                         if (!ops || !ops->dellink)
3183                                 return -EOPNOTSUPP;
3184                 }
3185         }
3186
3187         if (!found)
3188                 return -ENODEV;
3189
3190         for_each_netdev_safe(net, dev, aux) {
3191                 if (dev->group == group) {
3192                         const struct rtnl_link_ops *ops;
3193
3194                         ops = dev->rtnl_link_ops;
3195                         ops->dellink(dev, &list_kill);
3196                 }
3197         }
3198         unregister_netdevice_many(&list_kill);
3199
3200         return 0;
3201 }
3202
3203 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3204 {
3205         const struct rtnl_link_ops *ops;
3206         LIST_HEAD(list_kill);
3207
3208         ops = dev->rtnl_link_ops;
3209         if (!ops || !ops->dellink)
3210                 return -EOPNOTSUPP;
3211
3212         ops->dellink(dev, &list_kill);
3213         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3214
3215         return 0;
3216 }
3217 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3218
3219 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3220                         struct netlink_ext_ack *extack)
3221 {
3222         struct net *net = sock_net(skb->sk);
3223         u32 portid = NETLINK_CB(skb).portid;
3224         struct net *tgt_net = net;
3225         struct net_device *dev = NULL;
3226         struct ifinfomsg *ifm;
3227         struct nlattr *tb[IFLA_MAX+1];
3228         int err;
3229         int netnsid = -1;
3230
3231         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3232                                      ifla_policy, extack);
3233         if (err < 0)
3234                 return err;
3235
3236         err = rtnl_ensure_unique_netns(tb, extack, true);
3237         if (err < 0)
3238                 return err;
3239
3240         if (tb[IFLA_TARGET_NETNSID]) {
3241                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3242                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3243                 if (IS_ERR(tgt_net))
3244                         return PTR_ERR(tgt_net);
3245         }
3246
3247         err = -EINVAL;
3248         ifm = nlmsg_data(nlh);
3249         if (ifm->ifi_index > 0)
3250                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3251         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3252                 dev = rtnl_dev_get(net, tb);
3253         else if (tb[IFLA_GROUP])
3254                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3255         else
3256                 goto out;
3257
3258         if (!dev) {
3259                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3260                         err = -ENODEV;
3261
3262                 goto out;
3263         }
3264
3265         err = rtnl_delete_link(dev, portid, nlh);
3266
3267 out:
3268         if (netnsid >= 0)
3269                 put_net(tgt_net);
3270
3271         return err;
3272 }
3273
3274 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3275                         u32 portid, const struct nlmsghdr *nlh)
3276 {
3277         unsigned int old_flags;
3278         int err;
3279
3280         old_flags = dev->flags;
3281         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3282                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3283                                          NULL);
3284                 if (err < 0)
3285                         return err;
3286         }
3287
3288         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3289                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3290         } else {
3291                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3292                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3293         }
3294         return 0;
3295 }
3296 EXPORT_SYMBOL(rtnl_configure_link);
3297
3298 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3299                                     unsigned char name_assign_type,
3300                                     const struct rtnl_link_ops *ops,
3301                                     struct nlattr *tb[],
3302                                     struct netlink_ext_ack *extack)
3303 {
3304         struct net_device *dev;
3305         unsigned int num_tx_queues = 1;
3306         unsigned int num_rx_queues = 1;
3307         int err;
3308
3309         if (tb[IFLA_NUM_TX_QUEUES])
3310                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3311         else if (ops->get_num_tx_queues)
3312                 num_tx_queues = ops->get_num_tx_queues();
3313
3314         if (tb[IFLA_NUM_RX_QUEUES])
3315                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3316         else if (ops->get_num_rx_queues)
3317                 num_rx_queues = ops->get_num_rx_queues();
3318
3319         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3320                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3321                 return ERR_PTR(-EINVAL);
3322         }
3323
3324         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3325                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3326                 return ERR_PTR(-EINVAL);
3327         }
3328
3329         if (ops->alloc) {
3330                 dev = ops->alloc(tb, ifname, name_assign_type,
3331                                  num_tx_queues, num_rx_queues);
3332                 if (IS_ERR(dev))
3333                         return dev;
3334         } else {
3335                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3336                                        name_assign_type, ops->setup,
3337                                        num_tx_queues, num_rx_queues);
3338         }
3339
3340         if (!dev)
3341                 return ERR_PTR(-ENOMEM);
3342
3343         err = validate_linkmsg(dev, tb, extack);
3344         if (err < 0) {
3345                 free_netdev(dev);
3346                 return ERR_PTR(err);
3347         }
3348
3349         dev_net_set(dev, net);
3350         dev->rtnl_link_ops = ops;
3351         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3352
3353         if (tb[IFLA_MTU]) {
3354                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3355
3356                 err = dev_validate_mtu(dev, mtu, extack);
3357                 if (err) {
3358                         free_netdev(dev);
3359                         return ERR_PTR(err);
3360                 }
3361                 dev->mtu = mtu;
3362         }
3363         if (tb[IFLA_ADDRESS]) {
3364                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3365                                nla_len(tb[IFLA_ADDRESS]));
3366                 dev->addr_assign_type = NET_ADDR_SET;
3367         }
3368         if (tb[IFLA_BROADCAST])
3369                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3370                                 nla_len(tb[IFLA_BROADCAST]));
3371         if (tb[IFLA_TXQLEN])
3372                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3373         if (tb[IFLA_OPERSTATE])
3374                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3375         if (tb[IFLA_LINKMODE])
3376                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3377         if (tb[IFLA_GROUP])
3378                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3379         if (tb[IFLA_GSO_MAX_SIZE])
3380                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3381         if (tb[IFLA_GSO_MAX_SEGS])
3382                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3383         if (tb[IFLA_GRO_MAX_SIZE])
3384                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3385         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3386                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3387         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3388                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3389
3390         return dev;
3391 }
3392 EXPORT_SYMBOL(rtnl_create_link);
3393
3394 static int rtnl_group_changelink(const struct sk_buff *skb,
3395                 struct net *net, int group,
3396                 struct ifinfomsg *ifm,
3397                 struct netlink_ext_ack *extack,
3398                 struct nlattr **tb)
3399 {
3400         struct net_device *dev, *aux;
3401         int err;
3402
3403         for_each_netdev_safe(net, dev, aux) {
3404                 if (dev->group == group) {
3405                         err = validate_linkmsg(dev, tb, extack);
3406                         if (err < 0)
3407                                 return err;
3408                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3409                         if (err < 0)
3410                                 return err;
3411                 }
3412         }
3413
3414         return 0;
3415 }
3416
3417 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3418                                const struct rtnl_link_ops *ops,
3419                                const struct nlmsghdr *nlh,
3420                                struct nlattr **tb, struct nlattr **data,
3421                                struct netlink_ext_ack *extack)
3422 {
3423         unsigned char name_assign_type = NET_NAME_USER;
3424         struct net *net = sock_net(skb->sk);
3425         u32 portid = NETLINK_CB(skb).portid;
3426         struct net *dest_net, *link_net;
3427         struct net_device *dev;
3428         char ifname[IFNAMSIZ];
3429         int err;
3430
3431         if (!ops->alloc && !ops->setup)
3432                 return -EOPNOTSUPP;
3433
3434         if (tb[IFLA_IFNAME]) {
3435                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3436         } else {
3437                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3438                 name_assign_type = NET_NAME_ENUM;
3439         }
3440
3441         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3442         if (IS_ERR(dest_net))
3443                 return PTR_ERR(dest_net);
3444
3445         if (tb[IFLA_LINK_NETNSID]) {
3446                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3447
3448                 link_net = get_net_ns_by_id(dest_net, id);
3449                 if (!link_net) {
3450                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3451                         err =  -EINVAL;
3452                         goto out;
3453                 }
3454                 err = -EPERM;
3455                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3456                         goto out;
3457         } else {
3458                 link_net = NULL;
3459         }
3460
3461         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3462                                name_assign_type, ops, tb, extack);
3463         if (IS_ERR(dev)) {
3464                 err = PTR_ERR(dev);
3465                 goto out;
3466         }
3467
3468         dev->ifindex = ifm->ifi_index;
3469
3470         if (ops->newlink)
3471                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3472         else
3473                 err = register_netdevice(dev);
3474         if (err < 0) {
3475                 free_netdev(dev);
3476                 goto out;
3477         }
3478
3479         err = rtnl_configure_link(dev, ifm, portid, nlh);
3480         if (err < 0)
3481                 goto out_unregister;
3482         if (link_net) {
3483                 err = dev_change_net_namespace(dev, dest_net, ifname);
3484                 if (err < 0)
3485                         goto out_unregister;
3486         }
3487         if (tb[IFLA_MASTER]) {
3488                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3489                 if (err)
3490                         goto out_unregister;
3491         }
3492 out:
3493         if (link_net)
3494                 put_net(link_net);
3495         put_net(dest_net);
3496         return err;
3497 out_unregister:
3498         if (ops->newlink) {
3499                 LIST_HEAD(list_kill);
3500
3501                 ops->dellink(dev, &list_kill);
3502                 unregister_netdevice_many(&list_kill);
3503         } else {
3504                 unregister_netdevice(dev);
3505         }
3506         goto out;
3507 }
3508
3509 struct rtnl_newlink_tbs {
3510         struct nlattr *tb[IFLA_MAX + 1];
3511         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3512         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3513 };
3514
3515 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3516                           struct rtnl_newlink_tbs *tbs,
3517                           struct netlink_ext_ack *extack)
3518 {
3519         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3520         struct nlattr ** const tb = tbs->tb;
3521         const struct rtnl_link_ops *m_ops;
3522         struct net_device *master_dev;
3523         struct net *net = sock_net(skb->sk);
3524         const struct rtnl_link_ops *ops;
3525         struct nlattr **slave_data;
3526         char kind[MODULE_NAME_LEN];
3527         struct net_device *dev;
3528         struct ifinfomsg *ifm;
3529         struct nlattr **data;
3530         bool link_specified;
3531         int err;
3532
3533 #ifdef CONFIG_MODULES
3534 replay:
3535 #endif
3536         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3537                                      ifla_policy, extack);
3538         if (err < 0)
3539                 return err;
3540
3541         err = rtnl_ensure_unique_netns(tb, extack, false);
3542         if (err < 0)
3543                 return err;
3544
3545         ifm = nlmsg_data(nlh);
3546         if (ifm->ifi_index > 0) {
3547                 link_specified = true;
3548                 dev = __dev_get_by_index(net, ifm->ifi_index);
3549         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3550                 link_specified = true;
3551                 dev = rtnl_dev_get(net, tb);
3552         } else {
3553                 link_specified = false;
3554                 dev = NULL;
3555         }
3556
3557         master_dev = NULL;
3558         m_ops = NULL;
3559         if (dev) {
3560                 master_dev = netdev_master_upper_dev_get(dev);
3561                 if (master_dev)
3562                         m_ops = master_dev->rtnl_link_ops;
3563         }
3564
3565         if (tb[IFLA_LINKINFO]) {
3566                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3567                                                   tb[IFLA_LINKINFO],
3568                                                   ifla_info_policy, NULL);
3569                 if (err < 0)
3570                         return err;
3571         } else
3572                 memset(linkinfo, 0, sizeof(linkinfo));
3573
3574         if (linkinfo[IFLA_INFO_KIND]) {
3575                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3576                 ops = rtnl_link_ops_get(kind);
3577         } else {
3578                 kind[0] = '\0';
3579                 ops = NULL;
3580         }
3581
3582         data = NULL;
3583         if (ops) {
3584                 if (ops->maxtype > RTNL_MAX_TYPE)
3585                         return -EINVAL;
3586
3587                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3588                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3589                                                           linkinfo[IFLA_INFO_DATA],
3590                                                           ops->policy, extack);
3591                         if (err < 0)
3592                                 return err;
3593                         data = tbs->attr;
3594                 }
3595                 if (ops->validate) {
3596                         err = ops->validate(tb, data, extack);
3597                         if (err < 0)
3598                                 return err;
3599                 }
3600         }
3601
3602         slave_data = NULL;
3603         if (m_ops) {
3604                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3605                         return -EINVAL;
3606
3607                 if (m_ops->slave_maxtype &&
3608                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3609                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3610                                                           m_ops->slave_maxtype,
3611                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3612                                                           m_ops->slave_policy,
3613                                                           extack);
3614                         if (err < 0)
3615                                 return err;
3616                         slave_data = tbs->slave_attr;
3617                 }
3618         }
3619
3620         if (dev) {
3621                 int status = 0;
3622
3623                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3624                         return -EEXIST;
3625                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3626                         return -EOPNOTSUPP;
3627
3628                 err = validate_linkmsg(dev, tb, extack);
3629                 if (err < 0)
3630                         return err;
3631
3632                 if (linkinfo[IFLA_INFO_DATA]) {
3633                         if (!ops || ops != dev->rtnl_link_ops ||
3634                             !ops->changelink)
3635                                 return -EOPNOTSUPP;
3636
3637                         err = ops->changelink(dev, tb, data, extack);
3638                         if (err < 0)
3639                                 return err;
3640                         status |= DO_SETLINK_NOTIFY;
3641                 }
3642
3643                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3644                         if (!m_ops || !m_ops->slave_changelink)
3645                                 return -EOPNOTSUPP;
3646
3647                         err = m_ops->slave_changelink(master_dev, dev, tb,
3648                                                       slave_data, extack);
3649                         if (err < 0)
3650                                 return err;
3651                         status |= DO_SETLINK_NOTIFY;
3652                 }
3653
3654                 return do_setlink(skb, dev, ifm, extack, tb, status);
3655         }
3656
3657         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3658                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3659                  * or it's for a group
3660                 */
3661                 if (link_specified)
3662                         return -ENODEV;
3663                 if (tb[IFLA_GROUP])
3664                         return rtnl_group_changelink(skb, net,
3665                                                 nla_get_u32(tb[IFLA_GROUP]),
3666                                                 ifm, extack, tb);
3667                 return -ENODEV;
3668         }
3669
3670         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3671                 return -EOPNOTSUPP;
3672
3673         if (!ops) {
3674 #ifdef CONFIG_MODULES
3675                 if (kind[0]) {
3676                         __rtnl_unlock();
3677                         request_module("rtnl-link-%s", kind);
3678                         rtnl_lock();
3679                         ops = rtnl_link_ops_get(kind);
3680                         if (ops)
3681                                 goto replay;
3682                 }
3683 #endif
3684                 NL_SET_ERR_MSG(extack, "Unknown device type");
3685                 return -EOPNOTSUPP;
3686         }
3687
3688         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3689 }
3690
3691 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3692                         struct netlink_ext_ack *extack)
3693 {
3694         struct rtnl_newlink_tbs *tbs;
3695         int ret;
3696
3697         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3698         if (!tbs)
3699                 return -ENOMEM;
3700
3701         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3702         kfree(tbs);
3703         return ret;
3704 }
3705
3706 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3707                                   const struct nlmsghdr *nlh,
3708                                   struct nlattr **tb,
3709                                   struct netlink_ext_ack *extack)
3710 {
3711         struct ifinfomsg *ifm;
3712         int i, err;
3713
3714         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3715                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3716                 return -EINVAL;
3717         }
3718
3719         if (!netlink_strict_get_check(skb))
3720                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3721                                               ifla_policy, extack);
3722
3723         ifm = nlmsg_data(nlh);
3724         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3725             ifm->ifi_change) {
3726                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3727                 return -EINVAL;
3728         }
3729
3730         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3731                                             ifla_policy, extack);
3732         if (err)
3733                 return err;
3734
3735         for (i = 0; i <= IFLA_MAX; i++) {
3736                 if (!tb[i])
3737                         continue;
3738
3739                 switch (i) {
3740                 case IFLA_IFNAME:
3741                 case IFLA_ALT_IFNAME:
3742                 case IFLA_EXT_MASK:
3743                 case IFLA_TARGET_NETNSID:
3744                         break;
3745                 default:
3746                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3747                         return -EINVAL;
3748                 }
3749         }
3750
3751         return 0;
3752 }
3753
3754 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3755                         struct netlink_ext_ack *extack)
3756 {
3757         struct net *net = sock_net(skb->sk);
3758         struct net *tgt_net = net;
3759         struct ifinfomsg *ifm;
3760         struct nlattr *tb[IFLA_MAX+1];
3761         struct net_device *dev = NULL;
3762         struct sk_buff *nskb;
3763         int netnsid = -1;
3764         int err;
3765         u32 ext_filter_mask = 0;
3766
3767         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3768         if (err < 0)
3769                 return err;
3770
3771         err = rtnl_ensure_unique_netns(tb, extack, true);
3772         if (err < 0)
3773                 return err;
3774
3775         if (tb[IFLA_TARGET_NETNSID]) {
3776                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3777                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3778                 if (IS_ERR(tgt_net))
3779                         return PTR_ERR(tgt_net);
3780         }
3781
3782         if (tb[IFLA_EXT_MASK])
3783                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3784
3785         err = -EINVAL;
3786         ifm = nlmsg_data(nlh);
3787         if (ifm->ifi_index > 0)
3788                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3789         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3790                 dev = rtnl_dev_get(tgt_net, tb);
3791         else
3792                 goto out;
3793
3794         err = -ENODEV;
3795         if (dev == NULL)
3796                 goto out;
3797
3798         err = -ENOBUFS;
3799         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3800         if (nskb == NULL)
3801                 goto out;
3802
3803         err = rtnl_fill_ifinfo(nskb, dev, net,
3804                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3805                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3806                                0, NULL, 0, netnsid, GFP_KERNEL);
3807         if (err < 0) {
3808                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3809                 WARN_ON(err == -EMSGSIZE);
3810                 kfree_skb(nskb);
3811         } else
3812                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3813 out:
3814         if (netnsid >= 0)
3815                 put_net(tgt_net);
3816
3817         return err;
3818 }
3819
3820 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3821                            bool *changed, struct netlink_ext_ack *extack)
3822 {
3823         char *alt_ifname;
3824         size_t size;
3825         int err;
3826
3827         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3828         if (err)
3829                 return err;
3830
3831         if (cmd == RTM_NEWLINKPROP) {
3832                 size = rtnl_prop_list_size(dev);
3833                 size += nla_total_size(ALTIFNAMSIZ);
3834                 if (size >= U16_MAX) {
3835                         NL_SET_ERR_MSG(extack,
3836                                        "effective property list too long");
3837                         return -EINVAL;
3838                 }
3839         }
3840
3841         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3842         if (!alt_ifname)
3843                 return -ENOMEM;
3844
3845         if (cmd == RTM_NEWLINKPROP) {
3846                 err = netdev_name_node_alt_create(dev, alt_ifname);
3847                 if (!err)
3848                         alt_ifname = NULL;
3849         } else if (cmd == RTM_DELLINKPROP) {
3850                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3851         } else {
3852                 WARN_ON_ONCE(1);
3853                 err = -EINVAL;
3854         }
3855
3856         kfree(alt_ifname);
3857         if (!err)
3858                 *changed = true;
3859         return err;
3860 }
3861
3862 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3863                          struct netlink_ext_ack *extack)
3864 {
3865         struct net *net = sock_net(skb->sk);
3866         struct nlattr *tb[IFLA_MAX + 1];
3867         struct net_device *dev;
3868         struct ifinfomsg *ifm;
3869         bool changed = false;
3870         struct nlattr *attr;
3871         int err, rem;
3872
3873         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3874         if (err)
3875                 return err;
3876
3877         err = rtnl_ensure_unique_netns(tb, extack, true);
3878         if (err)
3879                 return err;
3880
3881         ifm = nlmsg_data(nlh);
3882         if (ifm->ifi_index > 0)
3883                 dev = __dev_get_by_index(net, ifm->ifi_index);
3884         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3885                 dev = rtnl_dev_get(net, tb);
3886         else
3887                 return -EINVAL;
3888
3889         if (!dev)
3890                 return -ENODEV;
3891
3892         if (!tb[IFLA_PROP_LIST])
3893                 return 0;
3894
3895         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3896                 switch (nla_type(attr)) {
3897                 case IFLA_ALT_IFNAME:
3898                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3899                         if (err)
3900                                 return err;
3901                         break;
3902                 }
3903         }
3904
3905         if (changed)
3906                 netdev_state_change(dev);
3907         return 0;
3908 }
3909
3910 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3911                             struct netlink_ext_ack *extack)
3912 {
3913         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3914 }
3915
3916 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3917                             struct netlink_ext_ack *extack)
3918 {
3919         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3920 }
3921
3922 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3923 {
3924         struct net *net = sock_net(skb->sk);
3925         size_t min_ifinfo_dump_size = 0;
3926         struct nlattr *tb[IFLA_MAX+1];
3927         u32 ext_filter_mask = 0;
3928         struct net_device *dev;
3929         int hdrlen;
3930
3931         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3932         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3933                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3934
3935         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3936                 if (tb[IFLA_EXT_MASK])
3937                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3938         }
3939
3940         if (!ext_filter_mask)
3941                 return NLMSG_GOODSIZE;
3942         /*
3943          * traverse the list of net devices and compute the minimum
3944          * buffer size based upon the filter mask.
3945          */
3946         rcu_read_lock();
3947         for_each_netdev_rcu(net, dev) {
3948                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3949                                            if_nlmsg_size(dev, ext_filter_mask));
3950         }
3951         rcu_read_unlock();
3952
3953         return nlmsg_total_size(min_ifinfo_dump_size);
3954 }
3955
3956 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3957 {
3958         int idx;
3959         int s_idx = cb->family;
3960         int type = cb->nlh->nlmsg_type - RTM_BASE;
3961         int ret = 0;
3962
3963         if (s_idx == 0)
3964                 s_idx = 1;
3965
3966         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3967                 struct rtnl_link __rcu **tab;
3968                 struct rtnl_link *link;
3969                 rtnl_dumpit_func dumpit;
3970
3971                 if (idx < s_idx || idx == PF_PACKET)
3972                         continue;
3973
3974                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3975                         continue;
3976
3977                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3978                 if (!tab)
3979                         continue;
3980
3981                 link = rcu_dereference_rtnl(tab[type]);
3982                 if (!link)
3983                         continue;
3984
3985                 dumpit = link->dumpit;
3986                 if (!dumpit)
3987                         continue;
3988
3989                 if (idx > s_idx) {
3990                         memset(&cb->args[0], 0, sizeof(cb->args));
3991                         cb->prev_seq = 0;
3992                         cb->seq = 0;
3993                 }
3994                 ret = dumpit(skb, cb);
3995                 if (ret)
3996                         break;
3997         }
3998         cb->family = idx;
3999
4000         return skb->len ? : ret;
4001 }
4002
4003 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
4004                                        unsigned int change,
4005                                        u32 event, gfp_t flags, int *new_nsid,
4006                                        int new_ifindex, u32 portid,
4007                                        const struct nlmsghdr *nlh)
4008 {
4009         struct net *net = dev_net(dev);
4010         struct sk_buff *skb;
4011         int err = -ENOBUFS;
4012         u32 seq = 0;
4013
4014         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4015         if (skb == NULL)
4016                 goto errout;
4017
4018         if (nlmsg_report(nlh))
4019                 seq = nlmsg_seq(nlh);
4020         else
4021                 portid = 0;
4022
4023         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4024                                type, portid, seq, change, 0, 0, event,
4025                                new_nsid, new_ifindex, -1, flags);
4026         if (err < 0) {
4027                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4028                 WARN_ON(err == -EMSGSIZE);
4029                 kfree_skb(skb);
4030                 goto errout;
4031         }
4032         return skb;
4033 errout:
4034         if (err < 0)
4035                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4036         return NULL;
4037 }
4038
4039 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4040                        u32 portid, const struct nlmsghdr *nlh)
4041 {
4042         struct net *net = dev_net(dev);
4043
4044         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4045 }
4046
4047 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4048                                unsigned int change, u32 event,
4049                                gfp_t flags, int *new_nsid, int new_ifindex,
4050                                u32 portid, const struct nlmsghdr *nlh)
4051 {
4052         struct sk_buff *skb;
4053
4054         if (dev->reg_state != NETREG_REGISTERED)
4055                 return;
4056
4057         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4058                                      new_ifindex, portid, nlh);
4059         if (skb)
4060                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4061 }
4062
4063 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4064                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4065 {
4066         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4067                            NULL, 0, portid, nlh);
4068 }
4069
4070 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4071                          gfp_t flags, int *new_nsid, int new_ifindex)
4072 {
4073         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4074                            new_nsid, new_ifindex, 0, NULL);
4075 }
4076
4077 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4078                                    struct net_device *dev,
4079                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4080                                    int type, unsigned int flags,
4081                                    int nlflags, u16 ndm_state)
4082 {
4083         struct nlmsghdr *nlh;
4084         struct ndmsg *ndm;
4085
4086         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4087         if (!nlh)
4088                 return -EMSGSIZE;
4089
4090         ndm = nlmsg_data(nlh);
4091         ndm->ndm_family  = AF_BRIDGE;
4092         ndm->ndm_pad1    = 0;
4093         ndm->ndm_pad2    = 0;
4094         ndm->ndm_flags   = flags;
4095         ndm->ndm_type    = 0;
4096         ndm->ndm_ifindex = dev->ifindex;
4097         ndm->ndm_state   = ndm_state;
4098
4099         if (nla_put(skb, NDA_LLADDR, dev->addr_len, addr))
4100                 goto nla_put_failure;
4101         if (vid)
4102                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4103                         goto nla_put_failure;
4104
4105         nlmsg_end(skb, nlh);
4106         return 0;
4107
4108 nla_put_failure:
4109         nlmsg_cancel(skb, nlh);
4110         return -EMSGSIZE;
4111 }
4112
4113 static inline size_t rtnl_fdb_nlmsg_size(const struct net_device *dev)
4114 {
4115         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4116                nla_total_size(dev->addr_len) +  /* NDA_LLADDR */
4117                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4118                0;
4119 }
4120
4121 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4122                             u16 ndm_state)
4123 {
4124         struct net *net = dev_net(dev);
4125         struct sk_buff *skb;
4126         int err = -ENOBUFS;
4127
4128         skb = nlmsg_new(rtnl_fdb_nlmsg_size(dev), GFP_ATOMIC);
4129         if (!skb)
4130                 goto errout;
4131
4132         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4133                                       0, 0, type, NTF_SELF, 0, ndm_state);
4134         if (err < 0) {
4135                 kfree_skb(skb);
4136                 goto errout;
4137         }
4138
4139         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4140         return;
4141 errout:
4142         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4143 }
4144
4145 /*
4146  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4147  */
4148 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4149                      struct nlattr *tb[],
4150                      struct net_device *dev,
4151                      const unsigned char *addr, u16 vid,
4152                      u16 flags)
4153 {
4154         int err = -EINVAL;
4155
4156         /* If aging addresses are supported device will need to
4157          * implement its own handler for this.
4158          */
4159         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4160                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4161                 return err;
4162         }
4163
4164         if (tb[NDA_FLAGS_EXT]) {
4165                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4166                 return err;
4167         }
4168
4169         if (vid) {
4170                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4171                 return err;
4172         }
4173
4174         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4175                 err = dev_uc_add_excl(dev, addr);
4176         else if (is_multicast_ether_addr(addr))
4177                 err = dev_mc_add_excl(dev, addr);
4178
4179         /* Only return duplicate errors if NLM_F_EXCL is set */
4180         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4181                 err = 0;
4182
4183         return err;
4184 }
4185 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4186
4187 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4188                          struct netlink_ext_ack *extack)
4189 {
4190         u16 vid = 0;
4191
4192         if (vlan_attr) {
4193                 if (nla_len(vlan_attr) != sizeof(u16)) {
4194                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4195                         return -EINVAL;
4196                 }
4197
4198                 vid = nla_get_u16(vlan_attr);
4199
4200                 if (!vid || vid >= VLAN_VID_MASK) {
4201                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4202                         return -EINVAL;
4203                 }
4204         }
4205         *p_vid = vid;
4206         return 0;
4207 }
4208
4209 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4210                         struct netlink_ext_ack *extack)
4211 {
4212         struct net *net = sock_net(skb->sk);
4213         struct ndmsg *ndm;
4214         struct nlattr *tb[NDA_MAX+1];
4215         struct net_device *dev;
4216         u8 *addr;
4217         u16 vid;
4218         int err;
4219
4220         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4221                                      extack);
4222         if (err < 0)
4223                 return err;
4224
4225         ndm = nlmsg_data(nlh);
4226         if (ndm->ndm_ifindex == 0) {
4227                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4228                 return -EINVAL;
4229         }
4230
4231         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4232         if (dev == NULL) {
4233                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4234                 return -ENODEV;
4235         }
4236
4237         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4238                 NL_SET_ERR_MSG(extack, "invalid address");
4239                 return -EINVAL;
4240         }
4241
4242         if (dev->type != ARPHRD_ETHER) {
4243                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4244                 return -EINVAL;
4245         }
4246
4247         addr = nla_data(tb[NDA_LLADDR]);
4248
4249         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4250         if (err)
4251                 return err;
4252
4253         err = -EOPNOTSUPP;
4254
4255         /* Support fdb on master device the net/bridge default case */
4256         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4257             netif_is_bridge_port(dev)) {
4258                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4259                 const struct net_device_ops *ops = br_dev->netdev_ops;
4260
4261                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4262                                        nlh->nlmsg_flags, extack);
4263                 if (err)
4264                         goto out;
4265                 else
4266                         ndm->ndm_flags &= ~NTF_MASTER;
4267         }
4268
4269         /* Embedded bridge, macvlan, and any other device support */
4270         if ((ndm->ndm_flags & NTF_SELF)) {
4271                 if (dev->netdev_ops->ndo_fdb_add)
4272                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4273                                                            vid,
4274                                                            nlh->nlmsg_flags,
4275                                                            extack);
4276                 else
4277                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4278                                                nlh->nlmsg_flags);
4279
4280                 if (!err) {
4281                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4282                                         ndm->ndm_state);
4283                         ndm->ndm_flags &= ~NTF_SELF;
4284                 }
4285         }
4286 out:
4287         return err;
4288 }
4289
4290 /*
4291  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4292  */
4293 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4294                      struct nlattr *tb[],
4295                      struct net_device *dev,
4296                      const unsigned char *addr, u16 vid)
4297 {
4298         int err = -EINVAL;
4299
4300         /* If aging addresses are supported device will need to
4301          * implement its own handler for this.
4302          */
4303         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4304                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4305                 return err;
4306         }
4307
4308         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4309                 err = dev_uc_del(dev, addr);
4310         else if (is_multicast_ether_addr(addr))
4311                 err = dev_mc_del(dev, addr);
4312
4313         return err;
4314 }
4315 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4316
4317 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4318         [NDA_VLAN]      = { .type = NLA_U16 },
4319         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4320         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4321         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4322 };
4323
4324 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4325                         struct netlink_ext_ack *extack)
4326 {
4327         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4328         struct net *net = sock_net(skb->sk);
4329         const struct net_device_ops *ops;
4330         struct ndmsg *ndm;
4331         struct nlattr *tb[NDA_MAX+1];
4332         struct net_device *dev;
4333         __u8 *addr = NULL;
4334         int err;
4335         u16 vid;
4336
4337         if (!netlink_capable(skb, CAP_NET_ADMIN))
4338                 return -EPERM;
4339
4340         if (!del_bulk) {
4341                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4342                                              NULL, extack);
4343         } else {
4344                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4345                                   fdb_del_bulk_policy, extack);
4346         }
4347         if (err < 0)
4348                 return err;
4349
4350         ndm = nlmsg_data(nlh);
4351         if (ndm->ndm_ifindex == 0) {
4352                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4353                 return -EINVAL;
4354         }
4355
4356         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4357         if (dev == NULL) {
4358                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4359                 return -ENODEV;
4360         }
4361
4362         if (!del_bulk) {
4363                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4364                         NL_SET_ERR_MSG(extack, "invalid address");
4365                         return -EINVAL;
4366                 }
4367                 addr = nla_data(tb[NDA_LLADDR]);
4368         }
4369
4370         if (dev->type != ARPHRD_ETHER) {
4371                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4372                 return -EINVAL;
4373         }
4374
4375         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4376         if (err)
4377                 return err;
4378
4379         err = -EOPNOTSUPP;
4380
4381         /* Support fdb on master device the net/bridge default case */
4382         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4383             netif_is_bridge_port(dev)) {
4384                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4385
4386                 ops = br_dev->netdev_ops;
4387                 if (!del_bulk) {
4388                         if (ops->ndo_fdb_del)
4389                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4390                 } else {
4391                         if (ops->ndo_fdb_del_bulk)
4392                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4393                                                             extack);
4394                 }
4395
4396                 if (err)
4397                         goto out;
4398                 else
4399                         ndm->ndm_flags &= ~NTF_MASTER;
4400         }
4401
4402         /* Embedded bridge, macvlan, and any other device support */
4403         if (ndm->ndm_flags & NTF_SELF) {
4404                 ops = dev->netdev_ops;
4405                 if (!del_bulk) {
4406                         if (ops->ndo_fdb_del)
4407                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4408                         else
4409                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4410                 } else {
4411                         /* in case err was cleared by NTF_MASTER call */
4412                         err = -EOPNOTSUPP;
4413                         if (ops->ndo_fdb_del_bulk)
4414                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4415                                                             extack);
4416                 }
4417
4418                 if (!err) {
4419                         if (!del_bulk)
4420                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4421                                                 ndm->ndm_state);
4422                         ndm->ndm_flags &= ~NTF_SELF;
4423                 }
4424         }
4425 out:
4426         return err;
4427 }
4428
4429 static int nlmsg_populate_fdb(struct sk_buff *skb,
4430                               struct netlink_callback *cb,
4431                               struct net_device *dev,
4432                               int *idx,
4433                               struct netdev_hw_addr_list *list)
4434 {
4435         struct netdev_hw_addr *ha;
4436         int err;
4437         u32 portid, seq;
4438
4439         portid = NETLINK_CB(cb->skb).portid;
4440         seq = cb->nlh->nlmsg_seq;
4441
4442         list_for_each_entry(ha, &list->list, list) {
4443                 if (*idx < cb->args[2])
4444                         goto skip;
4445
4446                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4447                                               portid, seq,
4448                                               RTM_NEWNEIGH, NTF_SELF,
4449                                               NLM_F_MULTI, NUD_PERMANENT);
4450                 if (err < 0)
4451                         return err;
4452 skip:
4453                 *idx += 1;
4454         }
4455         return 0;
4456 }
4457
4458 /**
4459  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4460  * @skb: socket buffer to store message in
4461  * @cb: netlink callback
4462  * @dev: netdevice
4463  * @filter_dev: ignored
4464  * @idx: the number of FDB table entries dumped is added to *@idx
4465  *
4466  * Default netdevice operation to dump the existing unicast address list.
4467  * Returns number of addresses from list put in skb.
4468  */
4469 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4470                       struct netlink_callback *cb,
4471                       struct net_device *dev,
4472                       struct net_device *filter_dev,
4473                       int *idx)
4474 {
4475         int err;
4476
4477         if (dev->type != ARPHRD_ETHER)
4478                 return -EINVAL;
4479
4480         netif_addr_lock_bh(dev);
4481         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4482         if (err)
4483                 goto out;
4484         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4485 out:
4486         netif_addr_unlock_bh(dev);
4487         return err;
4488 }
4489 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4490
4491 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4492                                  int *br_idx, int *brport_idx,
4493                                  struct netlink_ext_ack *extack)
4494 {
4495         struct nlattr *tb[NDA_MAX + 1];
4496         struct ndmsg *ndm;
4497         int err, i;
4498
4499         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4500                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4501                 return -EINVAL;
4502         }
4503
4504         ndm = nlmsg_data(nlh);
4505         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4506             ndm->ndm_flags || ndm->ndm_type) {
4507                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4508                 return -EINVAL;
4509         }
4510
4511         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4512                                             NDA_MAX, NULL, extack);
4513         if (err < 0)
4514                 return err;
4515
4516         *brport_idx = ndm->ndm_ifindex;
4517         for (i = 0; i <= NDA_MAX; ++i) {
4518                 if (!tb[i])
4519                         continue;
4520
4521                 switch (i) {
4522                 case NDA_IFINDEX:
4523                         if (nla_len(tb[i]) != sizeof(u32)) {
4524                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4525                                 return -EINVAL;
4526                         }
4527                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4528                         break;
4529                 case NDA_MASTER:
4530                         if (nla_len(tb[i]) != sizeof(u32)) {
4531                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4532                                 return -EINVAL;
4533                         }
4534                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4535                         break;
4536                 default:
4537                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4538                         return -EINVAL;
4539                 }
4540         }
4541
4542         return 0;
4543 }
4544
4545 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4546                                  int *br_idx, int *brport_idx,
4547                                  struct netlink_ext_ack *extack)
4548 {
4549         struct nlattr *tb[IFLA_MAX+1];
4550         int err;
4551
4552         /* A hack to preserve kernel<->userspace interface.
4553          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4554          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4555          * So, check for ndmsg with an optional u32 attribute (not used here).
4556          * Fortunately these sizes don't conflict with the size of ifinfomsg
4557          * with an optional attribute.
4558          */
4559         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4560             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4561              nla_attr_size(sizeof(u32)))) {
4562                 struct ifinfomsg *ifm;
4563
4564                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4565                                              tb, IFLA_MAX, ifla_policy,
4566                                              extack);
4567                 if (err < 0) {
4568                         return -EINVAL;
4569                 } else if (err == 0) {
4570                         if (tb[IFLA_MASTER])
4571                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4572                 }
4573
4574                 ifm = nlmsg_data(nlh);
4575                 *brport_idx = ifm->ifi_index;
4576         }
4577         return 0;
4578 }
4579
4580 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4581 {
4582         struct net_device *dev;
4583         struct net_device *br_dev = NULL;
4584         const struct net_device_ops *ops = NULL;
4585         const struct net_device_ops *cops = NULL;
4586         struct net *net = sock_net(skb->sk);
4587         struct hlist_head *head;
4588         int brport_idx = 0;
4589         int br_idx = 0;
4590         int h, s_h;
4591         int idx = 0, s_idx;
4592         int err = 0;
4593         int fidx = 0;
4594
4595         if (cb->strict_check)
4596                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4597                                             cb->extack);
4598         else
4599                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4600                                             cb->extack);
4601         if (err < 0)
4602                 return err;
4603
4604         if (br_idx) {
4605                 br_dev = __dev_get_by_index(net, br_idx);
4606                 if (!br_dev)
4607                         return -ENODEV;
4608
4609                 ops = br_dev->netdev_ops;
4610         }
4611
4612         s_h = cb->args[0];
4613         s_idx = cb->args[1];
4614
4615         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4616                 idx = 0;
4617                 head = &net->dev_index_head[h];
4618                 hlist_for_each_entry(dev, head, index_hlist) {
4619
4620                         if (brport_idx && (dev->ifindex != brport_idx))
4621                                 continue;
4622
4623                         if (!br_idx) { /* user did not specify a specific bridge */
4624                                 if (netif_is_bridge_port(dev)) {
4625                                         br_dev = netdev_master_upper_dev_get(dev);
4626                                         cops = br_dev->netdev_ops;
4627                                 }
4628                         } else {
4629                                 if (dev != br_dev &&
4630                                     !netif_is_bridge_port(dev))
4631                                         continue;
4632
4633                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4634                                     !netif_is_bridge_master(dev))
4635                                         continue;
4636                                 cops = ops;
4637                         }
4638
4639                         if (idx < s_idx)
4640                                 goto cont;
4641
4642                         if (netif_is_bridge_port(dev)) {
4643                                 if (cops && cops->ndo_fdb_dump) {
4644                                         err = cops->ndo_fdb_dump(skb, cb,
4645                                                                 br_dev, dev,
4646                                                                 &fidx);
4647                                         if (err == -EMSGSIZE)
4648                                                 goto out;
4649                                 }
4650                         }
4651
4652                         if (dev->netdev_ops->ndo_fdb_dump)
4653                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4654                                                                     dev, NULL,
4655                                                                     &fidx);
4656                         else
4657                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4658                                                         &fidx);
4659                         if (err == -EMSGSIZE)
4660                                 goto out;
4661
4662                         cops = NULL;
4663
4664                         /* reset fdb offset to 0 for rest of the interfaces */
4665                         cb->args[2] = 0;
4666                         fidx = 0;
4667 cont:
4668                         idx++;
4669                 }
4670         }
4671
4672 out:
4673         cb->args[0] = h;
4674         cb->args[1] = idx;
4675         cb->args[2] = fidx;
4676
4677         return skb->len;
4678 }
4679
4680 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4681                                 struct nlattr **tb, u8 *ndm_flags,
4682                                 int *br_idx, int *brport_idx, u8 **addr,
4683                                 u16 *vid, struct netlink_ext_ack *extack)
4684 {
4685         struct ndmsg *ndm;
4686         int err, i;
4687
4688         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4689                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4690                 return -EINVAL;
4691         }
4692
4693         ndm = nlmsg_data(nlh);
4694         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4695             ndm->ndm_type) {
4696                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4697                 return -EINVAL;
4698         }
4699
4700         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4701                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4702                 return -EINVAL;
4703         }
4704
4705         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4706                                             NDA_MAX, nda_policy, extack);
4707         if (err < 0)
4708                 return err;
4709
4710         *ndm_flags = ndm->ndm_flags;
4711         *brport_idx = ndm->ndm_ifindex;
4712         for (i = 0; i <= NDA_MAX; ++i) {
4713                 if (!tb[i])
4714                         continue;
4715
4716                 switch (i) {
4717                 case NDA_MASTER:
4718                         *br_idx = nla_get_u32(tb[i]);
4719                         break;
4720                 case NDA_LLADDR:
4721                         if (nla_len(tb[i]) != ETH_ALEN) {
4722                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4723                                 return -EINVAL;
4724                         }
4725                         *addr = nla_data(tb[i]);
4726                         break;
4727                 case NDA_VLAN:
4728                         err = fdb_vid_parse(tb[i], vid, extack);
4729                         if (err)
4730                                 return err;
4731                         break;
4732                 case NDA_VNI:
4733                         break;
4734                 default:
4735                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4736                         return -EINVAL;
4737                 }
4738         }
4739
4740         return 0;
4741 }
4742
4743 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4744                         struct netlink_ext_ack *extack)
4745 {
4746         struct net_device *dev = NULL, *br_dev = NULL;
4747         const struct net_device_ops *ops = NULL;
4748         struct net *net = sock_net(in_skb->sk);
4749         struct nlattr *tb[NDA_MAX + 1];
4750         struct sk_buff *skb;
4751         int brport_idx = 0;
4752         u8 ndm_flags = 0;
4753         int br_idx = 0;
4754         u8 *addr = NULL;
4755         u16 vid = 0;
4756         int err;
4757
4758         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4759                                    &brport_idx, &addr, &vid, extack);
4760         if (err < 0)
4761                 return err;
4762
4763         if (!addr) {
4764                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4765                 return -EINVAL;
4766         }
4767
4768         if (brport_idx) {
4769                 dev = __dev_get_by_index(net, brport_idx);
4770                 if (!dev) {
4771                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4772                         return -ENODEV;
4773                 }
4774         }
4775
4776         if (br_idx) {
4777                 if (dev) {
4778                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4779                         return -EINVAL;
4780                 }
4781
4782                 br_dev = __dev_get_by_index(net, br_idx);
4783                 if (!br_dev) {
4784                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4785                         return -EINVAL;
4786                 }
4787                 ops = br_dev->netdev_ops;
4788         }
4789
4790         if (dev) {
4791                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4792                         if (!netif_is_bridge_port(dev)) {
4793                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4794                                 return -EINVAL;
4795                         }
4796                         br_dev = netdev_master_upper_dev_get(dev);
4797                         if (!br_dev) {
4798                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4799                                 return -EINVAL;
4800                         }
4801                         ops = br_dev->netdev_ops;
4802                 } else {
4803                         if (!(ndm_flags & NTF_SELF)) {
4804                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4805                                 return -EINVAL;
4806                         }
4807                         ops = dev->netdev_ops;
4808                 }
4809         }
4810
4811         if (!br_dev && !dev) {
4812                 NL_SET_ERR_MSG(extack, "No device specified");
4813                 return -ENODEV;
4814         }
4815
4816         if (!ops || !ops->ndo_fdb_get) {
4817                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4818                 return -EOPNOTSUPP;
4819         }
4820
4821         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4822         if (!skb)
4823                 return -ENOBUFS;
4824
4825         if (br_dev)
4826                 dev = br_dev;
4827         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4828                                NETLINK_CB(in_skb).portid,
4829                                nlh->nlmsg_seq, extack);
4830         if (err)
4831                 goto out;
4832
4833         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4834 out:
4835         kfree_skb(skb);
4836         return err;
4837 }
4838
4839 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4840                                unsigned int attrnum, unsigned int flag)
4841 {
4842         if (mask & flag)
4843                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4844         return 0;
4845 }
4846
4847 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4848                             struct net_device *dev, u16 mode,
4849                             u32 flags, u32 mask, int nlflags,
4850                             u32 filter_mask,
4851                             int (*vlan_fill)(struct sk_buff *skb,
4852                                              struct net_device *dev,
4853                                              u32 filter_mask))
4854 {
4855         struct nlmsghdr *nlh;
4856         struct ifinfomsg *ifm;
4857         struct nlattr *br_afspec;
4858         struct nlattr *protinfo;
4859         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4860         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4861         int err = 0;
4862
4863         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4864         if (nlh == NULL)
4865                 return -EMSGSIZE;
4866
4867         ifm = nlmsg_data(nlh);
4868         ifm->ifi_family = AF_BRIDGE;
4869         ifm->__ifi_pad = 0;
4870         ifm->ifi_type = dev->type;
4871         ifm->ifi_index = dev->ifindex;
4872         ifm->ifi_flags = dev_get_flags(dev);
4873         ifm->ifi_change = 0;
4874
4875
4876         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4877             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4878             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4879             (br_dev &&
4880              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4881             (dev->addr_len &&
4882              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4883             (dev->ifindex != dev_get_iflink(dev) &&
4884              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4885                 goto nla_put_failure;
4886
4887         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4888         if (!br_afspec)
4889                 goto nla_put_failure;
4890
4891         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4892                 nla_nest_cancel(skb, br_afspec);
4893                 goto nla_put_failure;
4894         }
4895
4896         if (mode != BRIDGE_MODE_UNDEF) {
4897                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4898                         nla_nest_cancel(skb, br_afspec);
4899                         goto nla_put_failure;
4900                 }
4901         }
4902         if (vlan_fill) {
4903                 err = vlan_fill(skb, dev, filter_mask);
4904                 if (err) {
4905                         nla_nest_cancel(skb, br_afspec);
4906                         goto nla_put_failure;
4907                 }
4908         }
4909         nla_nest_end(skb, br_afspec);
4910
4911         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4912         if (!protinfo)
4913                 goto nla_put_failure;
4914
4915         if (brport_nla_put_flag(skb, flags, mask,
4916                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4917             brport_nla_put_flag(skb, flags, mask,
4918                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4919             brport_nla_put_flag(skb, flags, mask,
4920                                 IFLA_BRPORT_FAST_LEAVE,
4921                                 BR_MULTICAST_FAST_LEAVE) ||
4922             brport_nla_put_flag(skb, flags, mask,
4923                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4924             brport_nla_put_flag(skb, flags, mask,
4925                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4926             brport_nla_put_flag(skb, flags, mask,
4927                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4928             brport_nla_put_flag(skb, flags, mask,
4929                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4930             brport_nla_put_flag(skb, flags, mask,
4931                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4932             brport_nla_put_flag(skb, flags, mask,
4933                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4934             brport_nla_put_flag(skb, flags, mask,
4935                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4936                 nla_nest_cancel(skb, protinfo);
4937                 goto nla_put_failure;
4938         }
4939
4940         nla_nest_end(skb, protinfo);
4941
4942         nlmsg_end(skb, nlh);
4943         return 0;
4944 nla_put_failure:
4945         nlmsg_cancel(skb, nlh);
4946         return err ? err : -EMSGSIZE;
4947 }
4948 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4949
4950 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4951                                     bool strict_check, u32 *filter_mask,
4952                                     struct netlink_ext_ack *extack)
4953 {
4954         struct nlattr *tb[IFLA_MAX+1];
4955         int err, i;
4956
4957         if (strict_check) {
4958                 struct ifinfomsg *ifm;
4959
4960                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4961                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4962                         return -EINVAL;
4963                 }
4964
4965                 ifm = nlmsg_data(nlh);
4966                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4967                     ifm->ifi_change || ifm->ifi_index) {
4968                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4969                         return -EINVAL;
4970                 }
4971
4972                 err = nlmsg_parse_deprecated_strict(nlh,
4973                                                     sizeof(struct ifinfomsg),
4974                                                     tb, IFLA_MAX, ifla_policy,
4975                                                     extack);
4976         } else {
4977                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4978                                              tb, IFLA_MAX, ifla_policy,
4979                                              extack);
4980         }
4981         if (err < 0)
4982                 return err;
4983
4984         /* new attributes should only be added with strict checking */
4985         for (i = 0; i <= IFLA_MAX; ++i) {
4986                 if (!tb[i])
4987                         continue;
4988
4989                 switch (i) {
4990                 case IFLA_EXT_MASK:
4991                         *filter_mask = nla_get_u32(tb[i]);
4992                         break;
4993                 default:
4994                         if (strict_check) {
4995                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4996                                 return -EINVAL;
4997                         }
4998                 }
4999         }
5000
5001         return 0;
5002 }
5003
5004 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
5005 {
5006         const struct nlmsghdr *nlh = cb->nlh;
5007         struct net *net = sock_net(skb->sk);
5008         struct net_device *dev;
5009         int idx = 0;
5010         u32 portid = NETLINK_CB(cb->skb).portid;
5011         u32 seq = nlh->nlmsg_seq;
5012         u32 filter_mask = 0;
5013         int err;
5014
5015         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5016                                        cb->extack);
5017         if (err < 0 && cb->strict_check)
5018                 return err;
5019
5020         rcu_read_lock();
5021         for_each_netdev_rcu(net, dev) {
5022                 const struct net_device_ops *ops = dev->netdev_ops;
5023                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5024
5025                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5026                         if (idx >= cb->args[0]) {
5027                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5028                                                 skb, portid, seq, dev,
5029                                                 filter_mask, NLM_F_MULTI);
5030                                 if (err < 0 && err != -EOPNOTSUPP) {
5031                                         if (likely(skb->len))
5032                                                 break;
5033
5034                                         goto out_err;
5035                                 }
5036                         }
5037                         idx++;
5038                 }
5039
5040                 if (ops->ndo_bridge_getlink) {
5041                         if (idx >= cb->args[0]) {
5042                                 err = ops->ndo_bridge_getlink(skb, portid,
5043                                                               seq, dev,
5044                                                               filter_mask,
5045                                                               NLM_F_MULTI);
5046                                 if (err < 0 && err != -EOPNOTSUPP) {
5047                                         if (likely(skb->len))
5048                                                 break;
5049
5050                                         goto out_err;
5051                                 }
5052                         }
5053                         idx++;
5054                 }
5055         }
5056         err = skb->len;
5057 out_err:
5058         rcu_read_unlock();
5059         cb->args[0] = idx;
5060
5061         return err;
5062 }
5063
5064 static inline size_t bridge_nlmsg_size(void)
5065 {
5066         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5067                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5068                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5069                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5070                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5071                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5072                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5073                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5074                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5075                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5076                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5077 }
5078
5079 static int rtnl_bridge_notify(struct net_device *dev)
5080 {
5081         struct net *net = dev_net(dev);
5082         struct sk_buff *skb;
5083         int err = -EOPNOTSUPP;
5084
5085         if (!dev->netdev_ops->ndo_bridge_getlink)
5086                 return 0;
5087
5088         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5089         if (!skb) {
5090                 err = -ENOMEM;
5091                 goto errout;
5092         }
5093
5094         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5095         if (err < 0)
5096                 goto errout;
5097
5098         /* Notification info is only filled for bridge ports, not the bridge
5099          * device itself. Therefore, a zero notification length is valid and
5100          * should not result in an error.
5101          */
5102         if (!skb->len)
5103                 goto errout;
5104
5105         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5106         return 0;
5107 errout:
5108         WARN_ON(err == -EMSGSIZE);
5109         kfree_skb(skb);
5110         if (err)
5111                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5112         return err;
5113 }
5114
5115 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5116                                struct netlink_ext_ack *extack)
5117 {
5118         struct net *net = sock_net(skb->sk);
5119         struct ifinfomsg *ifm;
5120         struct net_device *dev;
5121         struct nlattr *br_spec, *attr = NULL;
5122         int rem, err = -EOPNOTSUPP;
5123         u16 flags = 0;
5124         bool have_flags = false;
5125
5126         if (nlmsg_len(nlh) < sizeof(*ifm))
5127                 return -EINVAL;
5128
5129         ifm = nlmsg_data(nlh);
5130         if (ifm->ifi_family != AF_BRIDGE)
5131                 return -EPFNOSUPPORT;
5132
5133         dev = __dev_get_by_index(net, ifm->ifi_index);
5134         if (!dev) {
5135                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5136                 return -ENODEV;
5137         }
5138
5139         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5140         if (br_spec) {
5141                 nla_for_each_nested(attr, br_spec, rem) {
5142                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS && !have_flags) {
5143                                 if (nla_len(attr) < sizeof(flags))
5144                                         return -EINVAL;
5145
5146                                 have_flags = true;
5147                                 flags = nla_get_u16(attr);
5148                         }
5149
5150                         if (nla_type(attr) == IFLA_BRIDGE_MODE) {
5151                                 if (nla_len(attr) < sizeof(u16))
5152                                         return -EINVAL;
5153                         }
5154                 }
5155         }
5156
5157         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5158                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5159
5160                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5161                         err = -EOPNOTSUPP;
5162                         goto out;
5163                 }
5164
5165                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5166                                                              extack);
5167                 if (err)
5168                         goto out;
5169
5170                 flags &= ~BRIDGE_FLAGS_MASTER;
5171         }
5172
5173         if ((flags & BRIDGE_FLAGS_SELF)) {
5174                 if (!dev->netdev_ops->ndo_bridge_setlink)
5175                         err = -EOPNOTSUPP;
5176                 else
5177                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5178                                                                   flags,
5179                                                                   extack);
5180                 if (!err) {
5181                         flags &= ~BRIDGE_FLAGS_SELF;
5182
5183                         /* Generate event to notify upper layer of bridge
5184                          * change
5185                          */
5186                         err = rtnl_bridge_notify(dev);
5187                 }
5188         }
5189
5190         if (have_flags)
5191                 memcpy(nla_data(attr), &flags, sizeof(flags));
5192 out:
5193         return err;
5194 }
5195
5196 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5197                                struct netlink_ext_ack *extack)
5198 {
5199         struct net *net = sock_net(skb->sk);
5200         struct ifinfomsg *ifm;
5201         struct net_device *dev;
5202         struct nlattr *br_spec, *attr = NULL;
5203         int rem, err = -EOPNOTSUPP;
5204         u16 flags = 0;
5205         bool have_flags = false;
5206
5207         if (nlmsg_len(nlh) < sizeof(*ifm))
5208                 return -EINVAL;
5209
5210         ifm = nlmsg_data(nlh);
5211         if (ifm->ifi_family != AF_BRIDGE)
5212                 return -EPFNOSUPPORT;
5213
5214         dev = __dev_get_by_index(net, ifm->ifi_index);
5215         if (!dev) {
5216                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5217                 return -ENODEV;
5218         }
5219
5220         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5221         if (br_spec) {
5222                 nla_for_each_nested(attr, br_spec, rem) {
5223                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5224                                 if (nla_len(attr) < sizeof(flags))
5225                                         return -EINVAL;
5226
5227                                 have_flags = true;
5228                                 flags = nla_get_u16(attr);
5229                                 break;
5230                         }
5231                 }
5232         }
5233
5234         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5235                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5236
5237                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5238                         err = -EOPNOTSUPP;
5239                         goto out;
5240                 }
5241
5242                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5243                 if (err)
5244                         goto out;
5245
5246                 flags &= ~BRIDGE_FLAGS_MASTER;
5247         }
5248
5249         if ((flags & BRIDGE_FLAGS_SELF)) {
5250                 if (!dev->netdev_ops->ndo_bridge_dellink)
5251                         err = -EOPNOTSUPP;
5252                 else
5253                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5254                                                                   flags);
5255
5256                 if (!err) {
5257                         flags &= ~BRIDGE_FLAGS_SELF;
5258
5259                         /* Generate event to notify upper layer of bridge
5260                          * change
5261                          */
5262                         err = rtnl_bridge_notify(dev);
5263                 }
5264         }
5265
5266         if (have_flags)
5267                 memcpy(nla_data(attr), &flags, sizeof(flags));
5268 out:
5269         return err;
5270 }
5271
5272 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5273 {
5274         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5275                (!idxattr || idxattr == attrid);
5276 }
5277
5278 static bool
5279 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5280 {
5281         return dev->netdev_ops &&
5282                dev->netdev_ops->ndo_has_offload_stats &&
5283                dev->netdev_ops->ndo_get_offload_stats &&
5284                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5285 }
5286
5287 static unsigned int
5288 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5289 {
5290         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5291                sizeof(struct rtnl_link_stats64) : 0;
5292 }
5293
5294 static int
5295 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5296                              struct sk_buff *skb)
5297 {
5298         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5299         struct nlattr *attr = NULL;
5300         void *attr_data;
5301         int err;
5302
5303         if (!size)
5304                 return -ENODATA;
5305
5306         attr = nla_reserve_64bit(skb, attr_id, size,
5307                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5308         if (!attr)
5309                 return -EMSGSIZE;
5310
5311         attr_data = nla_data(attr);
5312         memset(attr_data, 0, size);
5313
5314         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5315         if (err)
5316                 return err;
5317
5318         return 0;
5319 }
5320
5321 static unsigned int
5322 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5323                                    enum netdev_offload_xstats_type type)
5324 {
5325         bool enabled = netdev_offload_xstats_enabled(dev, type);
5326
5327         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5328 }
5329
5330 struct rtnl_offload_xstats_request_used {
5331         bool request;
5332         bool used;
5333 };
5334
5335 static int
5336 rtnl_offload_xstats_get_stats(struct net_device *dev,
5337                               enum netdev_offload_xstats_type type,
5338                               struct rtnl_offload_xstats_request_used *ru,
5339                               struct rtnl_hw_stats64 *stats,
5340                               struct netlink_ext_ack *extack)
5341 {
5342         bool request;
5343         bool used;
5344         int err;
5345
5346         request = netdev_offload_xstats_enabled(dev, type);
5347         if (!request) {
5348                 used = false;
5349                 goto out;
5350         }
5351
5352         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5353         if (err)
5354                 return err;
5355
5356 out:
5357         if (ru) {
5358                 ru->request = request;
5359                 ru->used = used;
5360         }
5361         return 0;
5362 }
5363
5364 static int
5365 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5366                                        struct rtnl_offload_xstats_request_used *ru)
5367 {
5368         struct nlattr *nest;
5369
5370         nest = nla_nest_start(skb, attr_id);
5371         if (!nest)
5372                 return -EMSGSIZE;
5373
5374         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5375                 goto nla_put_failure;
5376
5377         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5378                 goto nla_put_failure;
5379
5380         nla_nest_end(skb, nest);
5381         return 0;
5382
5383 nla_put_failure:
5384         nla_nest_cancel(skb, nest);
5385         return -EMSGSIZE;
5386 }
5387
5388 static int
5389 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5390                                    struct netlink_ext_ack *extack)
5391 {
5392         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5393         struct rtnl_offload_xstats_request_used ru_l3;
5394         struct nlattr *nest;
5395         int err;
5396
5397         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5398         if (err)
5399                 return err;
5400
5401         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5402         if (!nest)
5403                 return -EMSGSIZE;
5404
5405         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5406                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5407                                                    &ru_l3))
5408                 goto nla_put_failure;
5409
5410         nla_nest_end(skb, nest);
5411         return 0;
5412
5413 nla_put_failure:
5414         nla_nest_cancel(skb, nest);
5415         return -EMSGSIZE;
5416 }
5417
5418 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5419                                     int *prividx, u32 off_filter_mask,
5420                                     struct netlink_ext_ack *extack)
5421 {
5422         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5423         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5424         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5425         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5426         bool have_data = false;
5427         int err;
5428
5429         if (*prividx <= attr_id_cpu_hit &&
5430             (off_filter_mask &
5431              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5432                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5433                 if (!err) {
5434                         have_data = true;
5435                 } else if (err != -ENODATA) {
5436                         *prividx = attr_id_cpu_hit;
5437                         return err;
5438                 }
5439         }
5440
5441         if (*prividx <= attr_id_hw_s_info &&
5442             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5443                 *prividx = attr_id_hw_s_info;
5444
5445                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5446                 if (err)
5447                         return err;
5448
5449                 have_data = true;
5450                 *prividx = 0;
5451         }
5452
5453         if (*prividx <= attr_id_l3_stats &&
5454             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5455                 unsigned int size_l3;
5456                 struct nlattr *attr;
5457
5458                 *prividx = attr_id_l3_stats;
5459
5460                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5461                 if (!size_l3)
5462                         goto skip_l3_stats;
5463                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5464                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5465                 if (!attr)
5466                         return -EMSGSIZE;
5467
5468                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5469                                                     nla_data(attr), extack);
5470                 if (err)
5471                         return err;
5472
5473                 have_data = true;
5474 skip_l3_stats:
5475                 *prividx = 0;
5476         }
5477
5478         if (!have_data)
5479                 return -ENODATA;
5480
5481         *prividx = 0;
5482         return 0;
5483 }
5484
5485 static unsigned int
5486 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5487                                            enum netdev_offload_xstats_type type)
5488 {
5489         bool enabled = netdev_offload_xstats_enabled(dev, type);
5490
5491         return nla_total_size(0) +
5492                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5493                 nla_total_size(sizeof(u8)) +
5494                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5495                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5496                 0;
5497 }
5498
5499 static unsigned int
5500 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5501 {
5502         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5503
5504         return nla_total_size(0) +
5505                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5506                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5507                 0;
5508 }
5509
5510 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5511                                         u32 off_filter_mask)
5512 {
5513         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5514         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5515         int nla_size = 0;
5516         int size;
5517
5518         if (off_filter_mask &
5519             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5520                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5521                 nla_size += nla_total_size_64bit(size);
5522         }
5523
5524         if (off_filter_mask &
5525             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5526                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5527
5528         if (off_filter_mask &
5529             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5530                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5531                 nla_size += nla_total_size_64bit(size);
5532         }
5533
5534         if (nla_size != 0)
5535                 nla_size += nla_total_size(0);
5536
5537         return nla_size;
5538 }
5539
5540 struct rtnl_stats_dump_filters {
5541         /* mask[0] filters outer attributes. Then individual nests have their
5542          * filtering mask at the index of the nested attribute.
5543          */
5544         u32 mask[IFLA_STATS_MAX + 1];
5545 };
5546
5547 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5548                                int type, u32 pid, u32 seq, u32 change,
5549                                unsigned int flags,
5550                                const struct rtnl_stats_dump_filters *filters,
5551                                int *idxattr, int *prividx,
5552                                struct netlink_ext_ack *extack)
5553 {
5554         unsigned int filter_mask = filters->mask[0];
5555         struct if_stats_msg *ifsm;
5556         struct nlmsghdr *nlh;
5557         struct nlattr *attr;
5558         int s_prividx = *prividx;
5559         int err;
5560
5561         ASSERT_RTNL();
5562
5563         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5564         if (!nlh)
5565                 return -EMSGSIZE;
5566
5567         ifsm = nlmsg_data(nlh);
5568         ifsm->family = PF_UNSPEC;
5569         ifsm->pad1 = 0;
5570         ifsm->pad2 = 0;
5571         ifsm->ifindex = dev->ifindex;
5572         ifsm->filter_mask = filter_mask;
5573
5574         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5575                 struct rtnl_link_stats64 *sp;
5576
5577                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5578                                          sizeof(struct rtnl_link_stats64),
5579                                          IFLA_STATS_UNSPEC);
5580                 if (!attr) {
5581                         err = -EMSGSIZE;
5582                         goto nla_put_failure;
5583                 }
5584
5585                 sp = nla_data(attr);
5586                 dev_get_stats(dev, sp);
5587         }
5588
5589         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5590                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5591
5592                 if (ops && ops->fill_linkxstats) {
5593                         *idxattr = IFLA_STATS_LINK_XSTATS;
5594                         attr = nla_nest_start_noflag(skb,
5595                                                      IFLA_STATS_LINK_XSTATS);
5596                         if (!attr) {
5597                                 err = -EMSGSIZE;
5598                                 goto nla_put_failure;
5599                         }
5600
5601                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5602                         nla_nest_end(skb, attr);
5603                         if (err)
5604                                 goto nla_put_failure;
5605                         *idxattr = 0;
5606                 }
5607         }
5608
5609         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5610                              *idxattr)) {
5611                 const struct rtnl_link_ops *ops = NULL;
5612                 const struct net_device *master;
5613
5614                 master = netdev_master_upper_dev_get(dev);
5615                 if (master)
5616                         ops = master->rtnl_link_ops;
5617                 if (ops && ops->fill_linkxstats) {
5618                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5619                         attr = nla_nest_start_noflag(skb,
5620                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5621                         if (!attr) {
5622                                 err = -EMSGSIZE;
5623                                 goto nla_put_failure;
5624                         }
5625
5626                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5627                         nla_nest_end(skb, attr);
5628                         if (err)
5629                                 goto nla_put_failure;
5630                         *idxattr = 0;
5631                 }
5632         }
5633
5634         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5635                              *idxattr)) {
5636                 u32 off_filter_mask;
5637
5638                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5639                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5640                 attr = nla_nest_start_noflag(skb,
5641                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5642                 if (!attr) {
5643                         err = -EMSGSIZE;
5644                         goto nla_put_failure;
5645                 }
5646
5647                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5648                                                off_filter_mask, extack);
5649                 if (err == -ENODATA)
5650                         nla_nest_cancel(skb, attr);
5651                 else
5652                         nla_nest_end(skb, attr);
5653
5654                 if (err && err != -ENODATA)
5655                         goto nla_put_failure;
5656                 *idxattr = 0;
5657         }
5658
5659         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5660                 struct rtnl_af_ops *af_ops;
5661
5662                 *idxattr = IFLA_STATS_AF_SPEC;
5663                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5664                 if (!attr) {
5665                         err = -EMSGSIZE;
5666                         goto nla_put_failure;
5667                 }
5668
5669                 rcu_read_lock();
5670                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5671                         if (af_ops->fill_stats_af) {
5672                                 struct nlattr *af;
5673
5674                                 af = nla_nest_start_noflag(skb,
5675                                                            af_ops->family);
5676                                 if (!af) {
5677                                         rcu_read_unlock();
5678                                         err = -EMSGSIZE;
5679                                         goto nla_put_failure;
5680                                 }
5681                                 err = af_ops->fill_stats_af(skb, dev);
5682
5683                                 if (err == -ENODATA) {
5684                                         nla_nest_cancel(skb, af);
5685                                 } else if (err < 0) {
5686                                         rcu_read_unlock();
5687                                         goto nla_put_failure;
5688                                 }
5689
5690                                 nla_nest_end(skb, af);
5691                         }
5692                 }
5693                 rcu_read_unlock();
5694
5695                 nla_nest_end(skb, attr);
5696
5697                 *idxattr = 0;
5698         }
5699
5700         nlmsg_end(skb, nlh);
5701
5702         return 0;
5703
5704 nla_put_failure:
5705         /* not a multi message or no progress mean a real error */
5706         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5707                 nlmsg_cancel(skb, nlh);
5708         else
5709                 nlmsg_end(skb, nlh);
5710
5711         return err;
5712 }
5713
5714 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5715                                   const struct rtnl_stats_dump_filters *filters)
5716 {
5717         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5718         unsigned int filter_mask = filters->mask[0];
5719
5720         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5721                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5722
5723         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5724                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5725                 int attr = IFLA_STATS_LINK_XSTATS;
5726
5727                 if (ops && ops->get_linkxstats_size) {
5728                         size += nla_total_size(ops->get_linkxstats_size(dev,
5729                                                                         attr));
5730                         /* for IFLA_STATS_LINK_XSTATS */
5731                         size += nla_total_size(0);
5732                 }
5733         }
5734
5735         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5736                 struct net_device *_dev = (struct net_device *)dev;
5737                 const struct rtnl_link_ops *ops = NULL;
5738                 const struct net_device *master;
5739
5740                 /* netdev_master_upper_dev_get can't take const */
5741                 master = netdev_master_upper_dev_get(_dev);
5742                 if (master)
5743                         ops = master->rtnl_link_ops;
5744                 if (ops && ops->get_linkxstats_size) {
5745                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5746
5747                         size += nla_total_size(ops->get_linkxstats_size(dev,
5748                                                                         attr));
5749                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5750                         size += nla_total_size(0);
5751                 }
5752         }
5753
5754         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5755                 u32 off_filter_mask;
5756
5757                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5758                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5759         }
5760
5761         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5762                 struct rtnl_af_ops *af_ops;
5763
5764                 /* for IFLA_STATS_AF_SPEC */
5765                 size += nla_total_size(0);
5766
5767                 rcu_read_lock();
5768                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5769                         if (af_ops->get_stats_af_size) {
5770                                 size += nla_total_size(
5771                                         af_ops->get_stats_af_size(dev));
5772
5773                                 /* for AF_* */
5774                                 size += nla_total_size(0);
5775                         }
5776                 }
5777                 rcu_read_unlock();
5778         }
5779
5780         return size;
5781 }
5782
5783 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5784
5785 static const struct nla_policy
5786 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5787         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5788                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5789 };
5790
5791 static const struct nla_policy
5792 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5793         [IFLA_STATS_GET_FILTERS] =
5794                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5795 };
5796
5797 static const struct nla_policy
5798 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5799         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5800 };
5801
5802 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5803                                         struct rtnl_stats_dump_filters *filters,
5804                                         struct netlink_ext_ack *extack)
5805 {
5806         struct nlattr *tb[IFLA_STATS_MAX + 1];
5807         int err;
5808         int at;
5809
5810         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5811                                rtnl_stats_get_policy_filters, extack);
5812         if (err < 0)
5813                 return err;
5814
5815         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5816                 if (tb[at]) {
5817                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5818                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5819                                 return -EINVAL;
5820                         }
5821                         filters->mask[at] = nla_get_u32(tb[at]);
5822                 }
5823         }
5824
5825         return 0;
5826 }
5827
5828 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5829                                 u32 filter_mask,
5830                                 struct rtnl_stats_dump_filters *filters,
5831                                 struct netlink_ext_ack *extack)
5832 {
5833         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5834         int err;
5835         int i;
5836
5837         filters->mask[0] = filter_mask;
5838         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5839                 filters->mask[i] = -1U;
5840
5841         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5842                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5843         if (err < 0)
5844                 return err;
5845
5846         if (tb[IFLA_STATS_GET_FILTERS]) {
5847                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5848                                                    filters, extack);
5849                 if (err)
5850                         return err;
5851         }
5852
5853         return 0;
5854 }
5855
5856 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5857                                 bool is_dump, struct netlink_ext_ack *extack)
5858 {
5859         struct if_stats_msg *ifsm;
5860
5861         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5862                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5863                 return -EINVAL;
5864         }
5865
5866         if (!strict_check)
5867                 return 0;
5868
5869         ifsm = nlmsg_data(nlh);
5870
5871         /* only requests using strict checks can pass data to influence
5872          * the dump. The legacy exception is filter_mask.
5873          */
5874         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5875                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5876                 return -EINVAL;
5877         }
5878         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5879                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5880                 return -EINVAL;
5881         }
5882
5883         return 0;
5884 }
5885
5886 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5887                           struct netlink_ext_ack *extack)
5888 {
5889         struct rtnl_stats_dump_filters filters;
5890         struct net *net = sock_net(skb->sk);
5891         struct net_device *dev = NULL;
5892         int idxattr = 0, prividx = 0;
5893         struct if_stats_msg *ifsm;
5894         struct sk_buff *nskb;
5895         int err;
5896
5897         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5898                                    false, extack);
5899         if (err)
5900                 return err;
5901
5902         ifsm = nlmsg_data(nlh);
5903         if (ifsm->ifindex > 0)
5904                 dev = __dev_get_by_index(net, ifsm->ifindex);
5905         else
5906                 return -EINVAL;
5907
5908         if (!dev)
5909                 return -ENODEV;
5910
5911         if (!ifsm->filter_mask) {
5912                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5913                 return -EINVAL;
5914         }
5915
5916         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5917         if (err)
5918                 return err;
5919
5920         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5921         if (!nskb)
5922                 return -ENOBUFS;
5923
5924         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5925                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5926                                   0, &filters, &idxattr, &prividx, extack);
5927         if (err < 0) {
5928                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5929                 WARN_ON(err == -EMSGSIZE);
5930                 kfree_skb(nskb);
5931         } else {
5932                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5933         }
5934
5935         return err;
5936 }
5937
5938 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5939 {
5940         struct netlink_ext_ack *extack = cb->extack;
5941         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5942         struct rtnl_stats_dump_filters filters;
5943         struct net *net = sock_net(skb->sk);
5944         unsigned int flags = NLM_F_MULTI;
5945         struct if_stats_msg *ifsm;
5946         struct hlist_head *head;
5947         struct net_device *dev;
5948         int idx = 0;
5949
5950         s_h = cb->args[0];
5951         s_idx = cb->args[1];
5952         s_idxattr = cb->args[2];
5953         s_prividx = cb->args[3];
5954
5955         cb->seq = net->dev_base_seq;
5956
5957         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5958         if (err)
5959                 return err;
5960
5961         ifsm = nlmsg_data(cb->nlh);
5962         if (!ifsm->filter_mask) {
5963                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5964                 return -EINVAL;
5965         }
5966
5967         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5968                                    extack);
5969         if (err)
5970                 return err;
5971
5972         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5973                 idx = 0;
5974                 head = &net->dev_index_head[h];
5975                 hlist_for_each_entry(dev, head, index_hlist) {
5976                         if (idx < s_idx)
5977                                 goto cont;
5978                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5979                                                   NETLINK_CB(cb->skb).portid,
5980                                                   cb->nlh->nlmsg_seq, 0,
5981                                                   flags, &filters,
5982                                                   &s_idxattr, &s_prividx,
5983                                                   extack);
5984                         /* If we ran out of room on the first message,
5985                          * we're in trouble
5986                          */
5987                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5988
5989                         if (err < 0)
5990                                 goto out;
5991                         s_prividx = 0;
5992                         s_idxattr = 0;
5993                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5994 cont:
5995                         idx++;
5996                 }
5997         }
5998 out:
5999         cb->args[3] = s_prividx;
6000         cb->args[2] = s_idxattr;
6001         cb->args[1] = idx;
6002         cb->args[0] = h;
6003
6004         return skb->len;
6005 }
6006
6007 void rtnl_offload_xstats_notify(struct net_device *dev)
6008 {
6009         struct rtnl_stats_dump_filters response_filters = {};
6010         struct net *net = dev_net(dev);
6011         int idxattr = 0, prividx = 0;
6012         struct sk_buff *skb;
6013         int err = -ENOBUFS;
6014
6015         ASSERT_RTNL();
6016
6017         response_filters.mask[0] |=
6018                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6019         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6020                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6021
6022         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6023                         GFP_KERNEL);
6024         if (!skb)
6025                 goto errout;
6026
6027         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6028                                   &response_filters, &idxattr, &prividx, NULL);
6029         if (err < 0) {
6030                 kfree_skb(skb);
6031                 goto errout;
6032         }
6033
6034         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6035         return;
6036
6037 errout:
6038         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6039 }
6040 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6041
6042 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6043                           struct netlink_ext_ack *extack)
6044 {
6045         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6046         struct rtnl_stats_dump_filters response_filters = {};
6047         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6048         struct net *net = sock_net(skb->sk);
6049         struct net_device *dev = NULL;
6050         struct if_stats_msg *ifsm;
6051         bool notify = false;
6052         int err;
6053
6054         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6055                                    false, extack);
6056         if (err)
6057                 return err;
6058
6059         ifsm = nlmsg_data(nlh);
6060         if (ifsm->family != AF_UNSPEC) {
6061                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6062                 return -EINVAL;
6063         }
6064
6065         if (ifsm->ifindex > 0)
6066                 dev = __dev_get_by_index(net, ifsm->ifindex);
6067         else
6068                 return -EINVAL;
6069
6070         if (!dev)
6071                 return -ENODEV;
6072
6073         if (ifsm->filter_mask) {
6074                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6075                 return -EINVAL;
6076         }
6077
6078         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6079                           ifla_stats_set_policy, extack);
6080         if (err < 0)
6081                 return err;
6082
6083         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6084                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6085
6086                 if (req)
6087                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6088                 else
6089                         err = netdev_offload_xstats_disable(dev, t_l3);
6090
6091                 if (!err)
6092                         notify = true;
6093                 else if (err != -EALREADY)
6094                         return err;
6095
6096                 response_filters.mask[0] |=
6097                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6098                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6099                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6100         }
6101
6102         if (notify)
6103                 rtnl_offload_xstats_notify(dev);
6104
6105         return 0;
6106 }
6107
6108 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6109                                    struct netlink_ext_ack *extack)
6110 {
6111         struct br_port_msg *bpm;
6112
6113         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6114                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6115                 return -EINVAL;
6116         }
6117
6118         bpm = nlmsg_data(nlh);
6119         if (bpm->ifindex) {
6120                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6121                 return -EINVAL;
6122         }
6123         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6124                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6125                 return -EINVAL;
6126         }
6127
6128         return 0;
6129 }
6130
6131 struct rtnl_mdb_dump_ctx {
6132         long idx;
6133 };
6134
6135 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6136 {
6137         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6138         struct net *net = sock_net(skb->sk);
6139         struct net_device *dev;
6140         int idx, s_idx;
6141         int err;
6142
6143         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6144
6145         if (cb->strict_check) {
6146                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6147                 if (err)
6148                         return err;
6149         }
6150
6151         s_idx = ctx->idx;
6152         idx = 0;
6153
6154         for_each_netdev(net, dev) {
6155                 if (idx < s_idx)
6156                         goto skip;
6157                 if (!dev->netdev_ops->ndo_mdb_dump)
6158                         goto skip;
6159
6160                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6161                 if (err == -EMSGSIZE)
6162                         goto out;
6163                 /* Moving on to next device, reset markers and sequence
6164                  * counters since they are all maintained per-device.
6165                  */
6166                 memset(cb->ctx, 0, sizeof(cb->ctx));
6167                 cb->prev_seq = 0;
6168                 cb->seq = 0;
6169 skip:
6170                 idx++;
6171         }
6172
6173 out:
6174         ctx->idx = idx;
6175         return skb->len;
6176 }
6177
6178 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6179                                    struct netlink_ext_ack *extack)
6180 {
6181         struct br_mdb_entry *entry = nla_data(attr);
6182
6183         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6184                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6185                 return -EINVAL;
6186         }
6187
6188         if (entry->ifindex == 0) {
6189                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6190                 return -EINVAL;
6191         }
6192
6193         if (entry->addr.proto == htons(ETH_P_IP)) {
6194                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6195                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6196                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6197                         return -EINVAL;
6198                 }
6199                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6200                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6201                         return -EINVAL;
6202                 }
6203 #if IS_ENABLED(CONFIG_IPV6)
6204         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6205                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6206                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6207                         return -EINVAL;
6208                 }
6209 #endif
6210         } else if (entry->addr.proto == 0) {
6211                 /* L2 mdb */
6212                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6213                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6214                         return -EINVAL;
6215                 }
6216         } else {
6217                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6218                 return -EINVAL;
6219         }
6220
6221         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6222                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6223                 return -EINVAL;
6224         }
6225         if (entry->vid >= VLAN_VID_MASK) {
6226                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6227                 return -EINVAL;
6228         }
6229
6230         return 0;
6231 }
6232
6233 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6234         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6235         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6236                                                   rtnl_validate_mdb_entry,
6237                                                   sizeof(struct br_mdb_entry)),
6238         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6239 };
6240
6241 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6242                         struct netlink_ext_ack *extack)
6243 {
6244         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6245         struct net *net = sock_net(skb->sk);
6246         struct br_port_msg *bpm;
6247         struct net_device *dev;
6248         int err;
6249
6250         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6251                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6252         if (err)
6253                 return err;
6254
6255         bpm = nlmsg_data(nlh);
6256         if (!bpm->ifindex) {
6257                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6258                 return -EINVAL;
6259         }
6260
6261         dev = __dev_get_by_index(net, bpm->ifindex);
6262         if (!dev) {
6263                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6264                 return -ENODEV;
6265         }
6266
6267         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6268                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6269                 return -EINVAL;
6270         }
6271
6272         if (!dev->netdev_ops->ndo_mdb_add) {
6273                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6274                 return -EOPNOTSUPP;
6275         }
6276
6277         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6278 }
6279
6280 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6281                         struct netlink_ext_ack *extack)
6282 {
6283         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6284         struct net *net = sock_net(skb->sk);
6285         struct br_port_msg *bpm;
6286         struct net_device *dev;
6287         int err;
6288
6289         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6290                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6291         if (err)
6292                 return err;
6293
6294         bpm = nlmsg_data(nlh);
6295         if (!bpm->ifindex) {
6296                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6297                 return -EINVAL;
6298         }
6299
6300         dev = __dev_get_by_index(net, bpm->ifindex);
6301         if (!dev) {
6302                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6303                 return -ENODEV;
6304         }
6305
6306         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6307                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6308                 return -EINVAL;
6309         }
6310
6311         if (!dev->netdev_ops->ndo_mdb_del) {
6312                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6313                 return -EOPNOTSUPP;
6314         }
6315
6316         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6317 }
6318
6319 /* Process one rtnetlink message. */
6320
6321 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6322                              struct netlink_ext_ack *extack)
6323 {
6324         struct net *net = sock_net(skb->sk);
6325         struct rtnl_link *link;
6326         enum rtnl_kinds kind;
6327         struct module *owner;
6328         int err = -EOPNOTSUPP;
6329         rtnl_doit_func doit;
6330         unsigned int flags;
6331         int family;
6332         int type;
6333
6334         type = nlh->nlmsg_type;
6335         if (type > RTM_MAX)
6336                 return -EOPNOTSUPP;
6337
6338         type -= RTM_BASE;
6339
6340         /* All the messages must have at least 1 byte length */
6341         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6342                 return 0;
6343
6344         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6345         kind = rtnl_msgtype_kind(type);
6346
6347         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6348                 return -EPERM;
6349
6350         rcu_read_lock();
6351         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6352                 struct sock *rtnl;
6353                 rtnl_dumpit_func dumpit;
6354                 u32 min_dump_alloc = 0;
6355
6356                 link = rtnl_get_link(family, type);
6357                 if (!link || !link->dumpit) {
6358                         family = PF_UNSPEC;
6359                         link = rtnl_get_link(family, type);
6360                         if (!link || !link->dumpit)
6361                                 goto err_unlock;
6362                 }
6363                 owner = link->owner;
6364                 dumpit = link->dumpit;
6365
6366                 if (type == RTM_GETLINK - RTM_BASE)
6367                         min_dump_alloc = rtnl_calcit(skb, nlh);
6368
6369                 err = 0;
6370                 /* need to do this before rcu_read_unlock() */
6371                 if (!try_module_get(owner))
6372                         err = -EPROTONOSUPPORT;
6373
6374                 rcu_read_unlock();
6375
6376                 rtnl = net->rtnl;
6377                 if (err == 0) {
6378                         struct netlink_dump_control c = {
6379                                 .dump           = dumpit,
6380                                 .min_dump_alloc = min_dump_alloc,
6381                                 .module         = owner,
6382                         };
6383                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6384                         /* netlink_dump_start() will keep a reference on
6385                          * module if dump is still in progress.
6386                          */
6387                         module_put(owner);
6388                 }
6389                 return err;
6390         }
6391
6392         link = rtnl_get_link(family, type);
6393         if (!link || !link->doit) {
6394                 family = PF_UNSPEC;
6395                 link = rtnl_get_link(PF_UNSPEC, type);
6396                 if (!link || !link->doit)
6397                         goto out_unlock;
6398         }
6399
6400         owner = link->owner;
6401         if (!try_module_get(owner)) {
6402                 err = -EPROTONOSUPPORT;
6403                 goto out_unlock;
6404         }
6405
6406         flags = link->flags;
6407         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6408             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6409                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6410                 module_put(owner);
6411                 goto err_unlock;
6412         }
6413
6414         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6415                 doit = link->doit;
6416                 rcu_read_unlock();
6417                 if (doit)
6418                         err = doit(skb, nlh, extack);
6419                 module_put(owner);
6420                 return err;
6421         }
6422         rcu_read_unlock();
6423
6424         rtnl_lock();
6425         link = rtnl_get_link(family, type);
6426         if (link && link->doit)
6427                 err = link->doit(skb, nlh, extack);
6428         rtnl_unlock();
6429
6430         module_put(owner);
6431
6432         return err;
6433
6434 out_unlock:
6435         rcu_read_unlock();
6436         return err;
6437
6438 err_unlock:
6439         rcu_read_unlock();
6440         return -EOPNOTSUPP;
6441 }
6442
6443 static void rtnetlink_rcv(struct sk_buff *skb)
6444 {
6445         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6446 }
6447
6448 static int rtnetlink_bind(struct net *net, int group)
6449 {
6450         switch (group) {
6451         case RTNLGRP_IPV4_MROUTE_R:
6452         case RTNLGRP_IPV6_MROUTE_R:
6453                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6454                         return -EPERM;
6455                 break;
6456         }
6457         return 0;
6458 }
6459
6460 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6461 {
6462         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6463
6464         switch (event) {
6465         case NETDEV_REBOOT:
6466         case NETDEV_CHANGEMTU:
6467         case NETDEV_CHANGEADDR:
6468         case NETDEV_CHANGENAME:
6469         case NETDEV_FEAT_CHANGE:
6470         case NETDEV_BONDING_FAILOVER:
6471         case NETDEV_POST_TYPE_CHANGE:
6472         case NETDEV_NOTIFY_PEERS:
6473         case NETDEV_CHANGEUPPER:
6474         case NETDEV_RESEND_IGMP:
6475         case NETDEV_CHANGEINFODATA:
6476         case NETDEV_CHANGELOWERSTATE:
6477         case NETDEV_CHANGE_TX_QUEUE_LEN:
6478                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6479                                    GFP_KERNEL, NULL, 0, 0, NULL);
6480                 break;
6481         default:
6482                 break;
6483         }
6484         return NOTIFY_DONE;
6485 }
6486
6487 static struct notifier_block rtnetlink_dev_notifier = {
6488         .notifier_call  = rtnetlink_event,
6489 };
6490
6491
6492 static int __net_init rtnetlink_net_init(struct net *net)
6493 {
6494         struct sock *sk;
6495         struct netlink_kernel_cfg cfg = {
6496                 .groups         = RTNLGRP_MAX,
6497                 .input          = rtnetlink_rcv,
6498                 .cb_mutex       = &rtnl_mutex,
6499                 .flags          = NL_CFG_F_NONROOT_RECV,
6500                 .bind           = rtnetlink_bind,
6501         };
6502
6503         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6504         if (!sk)
6505                 return -ENOMEM;
6506         net->rtnl = sk;
6507         return 0;
6508 }
6509
6510 static void __net_exit rtnetlink_net_exit(struct net *net)
6511 {
6512         netlink_kernel_release(net->rtnl);
6513         net->rtnl = NULL;
6514 }
6515
6516 static struct pernet_operations rtnetlink_net_ops = {
6517         .init = rtnetlink_net_init,
6518         .exit = rtnetlink_net_exit,
6519 };
6520
6521 void __init rtnetlink_init(void)
6522 {
6523         if (register_pernet_subsys(&rtnetlink_net_ops))
6524                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6525
6526         register_netdevice_notifier(&rtnetlink_dev_notifier);
6527
6528         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6529                       rtnl_dump_ifinfo, 0);
6530         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6531         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6532         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6533
6534         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6535         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6536         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6537
6538         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6539         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6540
6541         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6542         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6543                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6544         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6545
6546         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6547         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6548         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6549
6550         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6551                       0);
6552         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6553
6554         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, rtnl_mdb_dump, 0);
6555         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6556         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL, 0);
6557 }