net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetic was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56 #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     43
  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(0) + /* nest IFLA_VF_STATS */
 965                          /* IFLA_VF_STATS_RX_PACKETS */
 966                          nla_total_size_64bit(sizeof(__u64)) +
 967                          /* IFLA_VF_STATS_TX_PACKETS */
 968                          nla_total_size_64bit(sizeof(__u64)) +
 969                          /* IFLA_VF_STATS_RX_BYTES */
 970                          nla_total_size_64bit(sizeof(__u64)) +
 971                          /* IFLA_VF_STATS_TX_BYTES */
 972                          nla_total_size_64bit(sizeof(__u64)) +
 973                          /* IFLA_VF_STATS_BROADCAST */
 974                          nla_total_size_64bit(sizeof(__u64)) +
 975                          /* IFLA_VF_STATS_MULTICAST */
 976                          nla_total_size_64bit(sizeof(__u64)) +
 977                          /* IFLA_VF_STATS_RX_DROPPED */
 978                          nla_total_size_64bit(sizeof(__u64)) +
 979                          /* IFLA_VF_STATS_TX_DROPPED */
 980                          nla_total_size_64bit(sizeof(__u64)) +
 981                          nla_total_size(sizeof(struct ifla_vf_trust)));
 982                 return size;
 983         } else
 984                 return 0;
 985 }
 986
 987 static size_t rtnl_port_size(const struct net_device *dev,
 988                              u32 ext_filter_mask)
 989 {
 990         size_t port_size = nla_total_size(4)            /* PORT_VF */
 991                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 992                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 993                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 994                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 995                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 996         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 997         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 998                 + port_size;
 999         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
1000                 + port_size;
1001
1002         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1003             !(ext_filter_mask & RTEXT_FILTER_VF))
1004                 return 0;
1005         if (dev_num_vf(dev->dev.parent))
1006                 return port_self_size + vf_ports_size +
1007                         vf_port_size * dev_num_vf(dev->dev.parent);
1008         else
1009                 return port_self_size;
1010 }
1011
1012 static size_t rtnl_xdp_size(void)
1013 {
1014         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
1015                           nla_total_size(1) +   /* XDP_ATTACHED */
1016                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
1017                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
1018
1019         return xdp_size;
1020 }
1021
1022 static size_t rtnl_prop_list_size(const struct net_device *dev)
1023 {
1024         struct netdev_name_node *name_node;
1025         size_t size;
1026
1027         if (list_empty(&dev->name_node->list))
1028                 return 0;
1029         size = nla_total_size(0);
1030         list_for_each_entry(name_node, &dev->name_node->list, list)
1031                 size += nla_total_size(ALTIFNAMSIZ);
1032         return size;
1033 }
1034
1035 static size_t rtnl_proto_down_size(const struct net_device *dev)
1036 {
1037         size_t size = nla_total_size(1);
1038
1039         if (dev->proto_down_reason)
1040                 size += nla_total_size(0) + nla_total_size(4);
1041
1042         return size;
1043 }
1044
1045 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1046 {
1047         size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1048
1049         if (dev->devlink_port)
1050                 size += devlink_nl_port_handle_size(dev->devlink_port);
1051
1052         return size;
1053 }
1054
1055 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1056                                      u32 ext_filter_mask)
1057 {
1058         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1059                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1060                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1061                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1062                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1063                + nla_total_size(sizeof(struct rtnl_link_stats))
1064                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1065                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1066                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1067                + nla_total_size(4) /* IFLA_TXQLEN */
1068                + nla_total_size(4) /* IFLA_WEIGHT */
1069                + nla_total_size(4) /* IFLA_MTU */
1070                + nla_total_size(4) /* IFLA_LINK */
1071                + nla_total_size(4) /* IFLA_MASTER */
1072                + nla_total_size(1) /* IFLA_CARRIER */
1073                + nla_total_size(4) /* IFLA_PROMISCUITY */
1074                + nla_total_size(4) /* IFLA_ALLMULTI */
1075                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1076                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1077                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1078                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1079                + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1080                + nla_total_size(4) /* IFLA_GSO_IPV4_MAX_SIZE */
1081                + nla_total_size(4) /* IFLA_GRO_IPV4_MAX_SIZE */
1082                + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1083                + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1084                + nla_total_size(1) /* IFLA_OPERSTATE */
1085                + nla_total_size(1) /* IFLA_LINKMODE */
1086                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1087                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1088                + nla_total_size(4) /* IFLA_GROUP */
1089                + nla_total_size(ext_filter_mask
1090                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1091                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1092                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1093                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1094                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1095                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1096                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1097                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1098                + rtnl_xdp_size() /* IFLA_XDP */
1099                + nla_total_size(4)  /* IFLA_EVENT */
1100                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1101                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1102                + rtnl_proto_down_size(dev)  /* proto down */
1103                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1104                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1105                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1106                + nla_total_size(4)  /* IFLA_MIN_MTU */
1107                + nla_total_size(4)  /* IFLA_MAX_MTU */
1108                + rtnl_prop_list_size(dev)
1109                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1110                + rtnl_devlink_port_size(dev)
1111                + 0;
1112 }
1113
1114 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1115 {
1116         struct nlattr *vf_ports;
1117         struct nlattr *vf_port;
1118         int vf;
1119         int err;
1120
1121         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1122         if (!vf_ports)
1123                 return -EMSGSIZE;
1124
1125         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1126                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1127                 if (!vf_port)
1128                         goto nla_put_failure;
1129                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1130                         goto nla_put_failure;
1131                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1132                 if (err == -EMSGSIZE)
1133                         goto nla_put_failure;
1134                 if (err) {
1135                         nla_nest_cancel(skb, vf_port);
1136                         continue;
1137                 }
1138                 nla_nest_end(skb, vf_port);
1139         }
1140
1141         nla_nest_end(skb, vf_ports);
1142
1143         return 0;
1144
1145 nla_put_failure:
1146         nla_nest_cancel(skb, vf_ports);
1147         return -EMSGSIZE;
1148 }
1149
1150 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1151 {
1152         struct nlattr *port_self;
1153         int err;
1154
1155         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1156         if (!port_self)
1157                 return -EMSGSIZE;
1158
1159         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1160         if (err) {
1161                 nla_nest_cancel(skb, port_self);
1162                 return (err == -EMSGSIZE) ? err : 0;
1163         }
1164
1165         nla_nest_end(skb, port_self);
1166
1167         return 0;
1168 }
1169
1170 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1171                           u32 ext_filter_mask)
1172 {
1173         int err;
1174
1175         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1176             !(ext_filter_mask & RTEXT_FILTER_VF))
1177                 return 0;
1178
1179         err = rtnl_port_self_fill(skb, dev);
1180         if (err)
1181                 return err;
1182
1183         if (dev_num_vf(dev->dev.parent)) {
1184                 err = rtnl_vf_ports_fill(skb, dev);
1185                 if (err)
1186                         return err;
1187         }
1188
1189         return 0;
1190 }
1191
1192 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1193 {
1194         int err;
1195         struct netdev_phys_item_id ppid;
1196
1197         err = dev_get_phys_port_id(dev, &ppid);
1198         if (err) {
1199                 if (err == -EOPNOTSUPP)
1200                         return 0;
1201                 return err;
1202         }
1203
1204         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1205                 return -EMSGSIZE;
1206
1207         return 0;
1208 }
1209
1210 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1211 {
1212         char name[IFNAMSIZ];
1213         int err;
1214
1215         err = dev_get_phys_port_name(dev, name, sizeof(name));
1216         if (err) {
1217                 if (err == -EOPNOTSUPP)
1218                         return 0;
1219                 return err;
1220         }
1221
1222         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1223                 return -EMSGSIZE;
1224
1225         return 0;
1226 }
1227
1228 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1229 {
1230         struct netdev_phys_item_id ppid = { };
1231         int err;
1232
1233         err = dev_get_port_parent_id(dev, &ppid, false);
1234         if (err) {
1235                 if (err == -EOPNOTSUPP)
1236                         return 0;
1237                 return err;
1238         }
1239
1240         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1241                 return -EMSGSIZE;
1242
1243         return 0;
1244 }
1245
1246 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1247                                               struct net_device *dev)
1248 {
1249         struct rtnl_link_stats64 *sp;
1250         struct nlattr *attr;
1251
1252         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1253                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1254         if (!attr)
1255                 return -EMSGSIZE;
1256
1257         sp = nla_data(attr);
1258         dev_get_stats(dev, sp);
1259
1260         attr = nla_reserve(skb, IFLA_STATS,
1261                            sizeof(struct rtnl_link_stats));
1262         if (!attr)
1263                 return -EMSGSIZE;
1264
1265         copy_rtnl_link_stats(nla_data(attr), sp);
1266
1267         return 0;
1268 }
1269
1270 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1271                                                struct net_device *dev,
1272                                                int vfs_num,
1273                                                struct nlattr *vfinfo)
1274 {
1275         struct ifla_vf_rss_query_en vf_rss_query_en;
1276         struct nlattr *vf, *vfstats, *vfvlanlist;
1277         struct ifla_vf_link_state vf_linkstate;
1278         struct ifla_vf_vlan_info vf_vlan_info;
1279         struct ifla_vf_spoofchk vf_spoofchk;
1280         struct ifla_vf_tx_rate vf_tx_rate;
1281         struct ifla_vf_stats vf_stats;
1282         struct ifla_vf_trust vf_trust;
1283         struct ifla_vf_vlan vf_vlan;
1284         struct ifla_vf_rate vf_rate;
1285         struct ifla_vf_mac vf_mac;
1286         struct ifla_vf_broadcast vf_broadcast;
1287         struct ifla_vf_info ivi;
1288         struct ifla_vf_guid node_guid;
1289         struct ifla_vf_guid port_guid;
1290
1291         memset(&ivi, 0, sizeof(ivi));
1292
1293         /* Not all SR-IOV capable drivers support the
1294          * spoofcheck and "RSS query enable" query.  Preset to
1295          * -1 so the user space tool can detect that the driver
1296          * didn't report anything.
1297          */
1298         ivi.spoofchk = -1;
1299         ivi.rss_query_en = -1;
1300         ivi.trusted = -1;
1301         /* The default value for VF link state is "auto"
1302          * IFLA_VF_LINK_STATE_AUTO which equals zero
1303          */
1304         ivi.linkstate = 0;
1305         /* VLAN Protocol by default is 802.1Q */
1306         ivi.vlan_proto = htons(ETH_P_8021Q);
1307         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1308                 return 0;
1309
1310         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1311         memset(&node_guid, 0, sizeof(node_guid));
1312         memset(&port_guid, 0, sizeof(port_guid));
1313
1314         vf_mac.vf =
1315                 vf_vlan.vf =
1316                 vf_vlan_info.vf =
1317                 vf_rate.vf =
1318                 vf_tx_rate.vf =
1319                 vf_spoofchk.vf =
1320                 vf_linkstate.vf =
1321                 vf_rss_query_en.vf =
1322                 vf_trust.vf =
1323                 node_guid.vf =
1324                 port_guid.vf = ivi.vf;
1325
1326         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1327         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1328         vf_vlan.vlan = ivi.vlan;
1329         vf_vlan.qos = ivi.qos;
1330         vf_vlan_info.vlan = ivi.vlan;
1331         vf_vlan_info.qos = ivi.qos;
1332         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1333         vf_tx_rate.rate = ivi.max_tx_rate;
1334         vf_rate.min_tx_rate = ivi.min_tx_rate;
1335         vf_rate.max_tx_rate = ivi.max_tx_rate;
1336         vf_spoofchk.setting = ivi.spoofchk;
1337         vf_linkstate.link_state = ivi.linkstate;
1338         vf_rss_query_en.setting = ivi.rss_query_en;
1339         vf_trust.setting = ivi.trusted;
1340         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1341         if (!vf)
1342                 goto nla_put_vfinfo_failure;
1343         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1344             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1345             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1346             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1347                     &vf_rate) ||
1348             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1349                     &vf_tx_rate) ||
1350             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1351                     &vf_spoofchk) ||
1352             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1353                     &vf_linkstate) ||
1354             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1355                     sizeof(vf_rss_query_en),
1356                     &vf_rss_query_en) ||
1357             nla_put(skb, IFLA_VF_TRUST,
1358                     sizeof(vf_trust), &vf_trust))
1359                 goto nla_put_vf_failure;
1360
1361         if (dev->netdev_ops->ndo_get_vf_guid &&
1362             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1363                                               &port_guid)) {
1364                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1365                             &node_guid) ||
1366                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1367                             &port_guid))
1368                         goto nla_put_vf_failure;
1369         }
1370         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1371         if (!vfvlanlist)
1372                 goto nla_put_vf_failure;
1373         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1374                     &vf_vlan_info)) {
1375                 nla_nest_cancel(skb, vfvlanlist);
1376                 goto nla_put_vf_failure;
1377         }
1378         nla_nest_end(skb, vfvlanlist);
1379         memset(&vf_stats, 0, sizeof(vf_stats));
1380         if (dev->netdev_ops->ndo_get_vf_stats)
1381                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1382                                                 &vf_stats);
1383         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1384         if (!vfstats)
1385                 goto nla_put_vf_failure;
1386         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1387                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1388             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1389                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1390             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1391                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1392             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1393                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1394             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1395                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1396             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1397                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1398             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1399                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1400             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1401                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1402                 nla_nest_cancel(skb, vfstats);
1403                 goto nla_put_vf_failure;
1404         }
1405         nla_nest_end(skb, vfstats);
1406         nla_nest_end(skb, vf);
1407         return 0;
1408
1409 nla_put_vf_failure:
1410         nla_nest_cancel(skb, vf);
1411 nla_put_vfinfo_failure:
1412         nla_nest_cancel(skb, vfinfo);
1413         return -EMSGSIZE;
1414 }
1415
1416 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1417                                            struct net_device *dev,
1418                                            u32 ext_filter_mask)
1419 {
1420         struct nlattr *vfinfo;
1421         int i, num_vfs;
1422
1423         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1424                 return 0;
1425
1426         num_vfs = dev_num_vf(dev->dev.parent);
1427         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1428                 return -EMSGSIZE;
1429
1430         if (!dev->netdev_ops->ndo_get_vf_config)
1431                 return 0;
1432
1433         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1434         if (!vfinfo)
1435                 return -EMSGSIZE;
1436
1437         for (i = 0; i < num_vfs; i++) {
1438                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1439                         return -EMSGSIZE;
1440         }
1441
1442         nla_nest_end(skb, vfinfo);
1443         return 0;
1444 }
1445
1446 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1447 {
1448         struct rtnl_link_ifmap map;
1449
1450         memset(&map, 0, sizeof(map));
1451         map.mem_start   = dev->mem_start;
1452         map.mem_end     = dev->mem_end;
1453         map.base_addr   = dev->base_addr;
1454         map.irq         = dev->irq;
1455         map.dma         = dev->dma;
1456         map.port        = dev->if_port;
1457
1458         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1459                 return -EMSGSIZE;
1460
1461         return 0;
1462 }
1463
1464 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1465 {
1466         const struct bpf_prog *generic_xdp_prog;
1467
1468         ASSERT_RTNL();
1469
1470         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1471         if (!generic_xdp_prog)
1472                 return 0;
1473         return generic_xdp_prog->aux->id;
1474 }
1475
1476 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1477 {
1478         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1479 }
1480
1481 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1482 {
1483         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1484 }
1485
1486 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1487                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1488                                u32 (*get_prog_id)(struct net_device *dev))
1489 {
1490         u32 curr_id;
1491         int err;
1492
1493         curr_id = get_prog_id(dev);
1494         if (!curr_id)
1495                 return 0;
1496
1497         *prog_id = curr_id;
1498         err = nla_put_u32(skb, attr, curr_id);
1499         if (err)
1500                 return err;
1501
1502         if (*mode != XDP_ATTACHED_NONE)
1503                 *mode = XDP_ATTACHED_MULTI;
1504         else
1505                 *mode = tgt_mode;
1506
1507         return 0;
1508 }
1509
1510 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1511 {
1512         struct nlattr *xdp;
1513         u32 prog_id;
1514         int err;
1515         u8 mode;
1516
1517         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1518         if (!xdp)
1519                 return -EMSGSIZE;
1520
1521         prog_id = 0;
1522         mode = XDP_ATTACHED_NONE;
1523         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1524                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1525         if (err)
1526                 goto err_cancel;
1527         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1528                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1529         if (err)
1530                 goto err_cancel;
1531         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1532                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1533         if (err)
1534                 goto err_cancel;
1535
1536         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1537         if (err)
1538                 goto err_cancel;
1539
1540         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1541                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1542                 if (err)
1543                         goto err_cancel;
1544         }
1545
1546         nla_nest_end(skb, xdp);
1547         return 0;
1548
1549 err_cancel:
1550         nla_nest_cancel(skb, xdp);
1551         return err;
1552 }
1553
1554 static u32 rtnl_get_event(unsigned long event)
1555 {
1556         u32 rtnl_event_type = IFLA_EVENT_NONE;
1557
1558         switch (event) {
1559         case NETDEV_REBOOT:
1560                 rtnl_event_type = IFLA_EVENT_REBOOT;
1561                 break;
1562         case NETDEV_FEAT_CHANGE:
1563                 rtnl_event_type = IFLA_EVENT_FEATURES;
1564                 break;
1565         case NETDEV_BONDING_FAILOVER:
1566                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1567                 break;
1568         case NETDEV_NOTIFY_PEERS:
1569                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1570                 break;
1571         case NETDEV_RESEND_IGMP:
1572                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1573                 break;
1574         case NETDEV_CHANGEINFODATA:
1575                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1576                 break;
1577         default:
1578                 break;
1579         }
1580
1581         return rtnl_event_type;
1582 }
1583
1584 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1585 {
1586         const struct net_device *upper_dev;
1587         int ret = 0;
1588
1589         rcu_read_lock();
1590
1591         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1592         if (upper_dev)
1593                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1594
1595         rcu_read_unlock();
1596         return ret;
1597 }
1598
1599 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1600                           bool force)
1601 {
1602         int ifindex = dev_get_iflink(dev);
1603
1604         if (force || dev->ifindex != ifindex)
1605                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1606
1607         return 0;
1608 }
1609
1610 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1611                                               struct net_device *dev)
1612 {
1613         char buf[IFALIASZ];
1614         int ret;
1615
1616         ret = dev_get_alias(dev, buf, sizeof(buf));
1617         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1618 }
1619
1620 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1621                                   const struct net_device *dev,
1622                                   struct net *src_net, gfp_t gfp)
1623 {
1624         bool put_iflink = false;
1625
1626         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1627                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1628
1629                 if (!net_eq(dev_net(dev), link_net)) {
1630                         int id = peernet2id_alloc(src_net, link_net, gfp);
1631
1632                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1633                                 return -EMSGSIZE;
1634
1635                         put_iflink = true;
1636                 }
1637         }
1638
1639         return nla_put_iflink(skb, dev, put_iflink);
1640 }
1641
1642 static int rtnl_fill_link_af(struct sk_buff *skb,
1643                              const struct net_device *dev,
1644                              u32 ext_filter_mask)
1645 {
1646         const struct rtnl_af_ops *af_ops;
1647         struct nlattr *af_spec;
1648
1649         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1650         if (!af_spec)
1651                 return -EMSGSIZE;
1652
1653         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1654                 struct nlattr *af;
1655                 int err;
1656
1657                 if (!af_ops->fill_link_af)
1658                         continue;
1659
1660                 af = nla_nest_start_noflag(skb, af_ops->family);
1661                 if (!af)
1662                         return -EMSGSIZE;
1663
1664                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1665                 /*
1666                  * Caller may return ENODATA to indicate that there
1667                  * was no data to be dumped. This is not an error, it
1668                  * means we should trim the attribute header and
1669                  * continue.
1670                  */
1671                 if (err == -ENODATA)
1672                         nla_nest_cancel(skb, af);
1673                 else if (err < 0)
1674                         return -EMSGSIZE;
1675
1676                 nla_nest_end(skb, af);
1677         }
1678
1679         nla_nest_end(skb, af_spec);
1680         return 0;
1681 }
1682
1683 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1684                                  const struct net_device *dev)
1685 {
1686         struct netdev_name_node *name_node;
1687         int count = 0;
1688
1689         list_for_each_entry(name_node, &dev->name_node->list, list) {
1690                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1691                         return -EMSGSIZE;
1692                 count++;
1693         }
1694         return count;
1695 }
1696
1697 static int rtnl_fill_prop_list(struct sk_buff *skb,
1698                                const struct net_device *dev)
1699 {
1700         struct nlattr *prop_list;
1701         int ret;
1702
1703         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1704         if (!prop_list)
1705                 return -EMSGSIZE;
1706
1707         ret = rtnl_fill_alt_ifnames(skb, dev);
1708         if (ret <= 0)
1709                 goto nest_cancel;
1710
1711         nla_nest_end(skb, prop_list);
1712         return 0;
1713
1714 nest_cancel:
1715         nla_nest_cancel(skb, prop_list);
1716         return ret;
1717 }
1718
1719 static int rtnl_fill_proto_down(struct sk_buff *skb,
1720                                 const struct net_device *dev)
1721 {
1722         struct nlattr *pr;
1723         u32 preason;
1724
1725         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1726                 goto nla_put_failure;
1727
1728         preason = dev->proto_down_reason;
1729         if (!preason)
1730                 return 0;
1731
1732         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1733         if (!pr)
1734                 return -EMSGSIZE;
1735
1736         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1737                 nla_nest_cancel(skb, pr);
1738                 goto nla_put_failure;
1739         }
1740
1741         nla_nest_end(skb, pr);
1742         return 0;
1743
1744 nla_put_failure:
1745         return -EMSGSIZE;
1746 }
1747
1748 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1749                                   const struct net_device *dev)
1750 {
1751         struct nlattr *devlink_port_nest;
1752         int ret;
1753
1754         devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1755         if (!devlink_port_nest)
1756                 return -EMSGSIZE;
1757
1758         if (dev->devlink_port) {
1759                 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1760                 if (ret < 0)
1761                         goto nest_cancel;
1762         }
1763
1764         nla_nest_end(skb, devlink_port_nest);
1765         return 0;
1766
1767 nest_cancel:
1768         nla_nest_cancel(skb, devlink_port_nest);
1769         return ret;
1770 }
1771
1772 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1773                             struct net_device *dev, struct net *src_net,
1774                             int type, u32 pid, u32 seq, u32 change,
1775                             unsigned int flags, u32 ext_filter_mask,
1776                             u32 event, int *new_nsid, int new_ifindex,
1777                             int tgt_netnsid, gfp_t gfp)
1778 {
1779         struct ifinfomsg *ifm;
1780         struct nlmsghdr *nlh;
1781         struct Qdisc *qdisc;
1782
1783         ASSERT_RTNL();
1784         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1785         if (nlh == NULL)
1786                 return -EMSGSIZE;
1787
1788         ifm = nlmsg_data(nlh);
1789         ifm->ifi_family = AF_UNSPEC;
1790         ifm->__ifi_pad = 0;
1791         ifm->ifi_type = dev->type;
1792         ifm->ifi_index = dev->ifindex;
1793         ifm->ifi_flags = dev_get_flags(dev);
1794         ifm->ifi_change = change;
1795
1796         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1797                 goto nla_put_failure;
1798
1799         qdisc = rtnl_dereference(dev->qdisc);
1800         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1801             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1802             nla_put_u8(skb, IFLA_OPERSTATE,
1803                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1804             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1805             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1806             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1807             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1808             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1809             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1810             nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1811             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1812             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1813             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1814             nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1815             nla_put_u32(skb, IFLA_GSO_IPV4_MAX_SIZE, dev->gso_ipv4_max_size) ||
1816             nla_put_u32(skb, IFLA_GRO_IPV4_MAX_SIZE, dev->gro_ipv4_max_size) ||
1817             nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1818             nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1819 #ifdef CONFIG_RPS
1820             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1821 #endif
1822             put_master_ifindex(skb, dev) ||
1823             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1824             (qdisc &&
1825              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1826             nla_put_ifalias(skb, dev) ||
1827             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1828                         atomic_read(&dev->carrier_up_count) +
1829                         atomic_read(&dev->carrier_down_count)) ||
1830             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1831                         atomic_read(&dev->carrier_up_count)) ||
1832             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1833                         atomic_read(&dev->carrier_down_count)))
1834                 goto nla_put_failure;
1835
1836         if (rtnl_fill_proto_down(skb, dev))
1837                 goto nla_put_failure;
1838
1839         if (event != IFLA_EVENT_NONE) {
1840                 if (nla_put_u32(skb, IFLA_EVENT, event))
1841                         goto nla_put_failure;
1842         }
1843
1844         if (rtnl_fill_link_ifmap(skb, dev))
1845                 goto nla_put_failure;
1846
1847         if (dev->addr_len) {
1848                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1849                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1850                         goto nla_put_failure;
1851         }
1852
1853         if (rtnl_phys_port_id_fill(skb, dev))
1854                 goto nla_put_failure;
1855
1856         if (rtnl_phys_port_name_fill(skb, dev))
1857                 goto nla_put_failure;
1858
1859         if (rtnl_phys_switch_id_fill(skb, dev))
1860                 goto nla_put_failure;
1861
1862         if (rtnl_fill_stats(skb, dev))
1863                 goto nla_put_failure;
1864
1865         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1866                 goto nla_put_failure;
1867
1868         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1869                 goto nla_put_failure;
1870
1871         if (rtnl_xdp_fill(skb, dev))
1872                 goto nla_put_failure;
1873
1874         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1875                 if (rtnl_link_fill(skb, dev) < 0)
1876                         goto nla_put_failure;
1877         }
1878
1879         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1880                 goto nla_put_failure;
1881
1882         if (new_nsid &&
1883             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1884                 goto nla_put_failure;
1885         if (new_ifindex &&
1886             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1887                 goto nla_put_failure;
1888
1889         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1890             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1891                 goto nla_put_failure;
1892
1893         rcu_read_lock();
1894         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1895                 goto nla_put_failure_rcu;
1896         rcu_read_unlock();
1897
1898         if (rtnl_fill_prop_list(skb, dev))
1899                 goto nla_put_failure;
1900
1901         if (dev->dev.parent &&
1902             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1903                            dev_name(dev->dev.parent)))
1904                 goto nla_put_failure;
1905
1906         if (dev->dev.parent && dev->dev.parent->bus &&
1907             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1908                            dev->dev.parent->bus->name))
1909                 goto nla_put_failure;
1910
1911         if (rtnl_fill_devlink_port(skb, dev))
1912                 goto nla_put_failure;
1913
1914         nlmsg_end(skb, nlh);
1915         return 0;
1916
1917 nla_put_failure_rcu:
1918         rcu_read_unlock();
1919 nla_put_failure:
1920         nlmsg_cancel(skb, nlh);
1921         return -EMSGSIZE;
1922 }
1923
1924 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1925         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1926         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1927         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1928         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1929         [IFLA_MTU]              = { .type = NLA_U32 },
1930         [IFLA_LINK]             = { .type = NLA_U32 },
1931         [IFLA_MASTER]           = { .type = NLA_U32 },
1932         [IFLA_CARRIER]          = { .type = NLA_U8 },
1933         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1934         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1935         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1936         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1937         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1938         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1939         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1940         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1941          * allow 0-length string (needed to remove an alias).
1942          */
1943         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1944         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1945         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1946         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1947         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1948         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1949         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1950         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1951         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1952         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1953         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1954         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1955         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1956         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1957         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1958         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1959         [IFLA_XDP]              = { .type = NLA_NESTED },
1960         [IFLA_EVENT]            = { .type = NLA_U32 },
1961         [IFLA_GROUP]            = { .type = NLA_U32 },
1962         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1963         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1964         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1965         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1966         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1967         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1968         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1969                                     .len = ALTIFNAMSIZ - 1 },
1970         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1971         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1972         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1973         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1974         [IFLA_GRO_MAX_SIZE]     = { .type = NLA_U32 },
1975         [IFLA_TSO_MAX_SIZE]     = { .type = NLA_REJECT },
1976         [IFLA_TSO_MAX_SEGS]     = { .type = NLA_REJECT },
1977         [IFLA_ALLMULTI]         = { .type = NLA_REJECT },
1978         [IFLA_GSO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1979         [IFLA_GRO_IPV4_MAX_SIZE]        = { .type = NLA_U32 },
1980 };
1981
1982 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1983         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1984         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1985         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1986         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1987 };
1988
1989 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1990         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1991         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1992         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1993         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1994         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1995         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1996         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1997         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1998         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1999         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
2000         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
2001         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2002         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
2003 };
2004
2005 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
2006         [IFLA_PORT_VF]          = { .type = NLA_U32 },
2007         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
2008                                     .len = PORT_PROFILE_MAX },
2009         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2010                                       .len = PORT_UUID_MAX },
2011         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
2012                                     .len = PORT_UUID_MAX },
2013         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
2014         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
2015
2016         /* Unused, but we need to keep it here since user space could
2017          * fill it. It's also broken with regard to NLA_BINARY use in
2018          * combination with structs.
2019          */
2020         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
2021                                     .len = sizeof(struct ifla_port_vsi) },
2022 };
2023
2024 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2025         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2026         [IFLA_XDP_FD]           = { .type = NLA_S32 },
2027         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
2028         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
2029         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
2030         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
2031 };
2032
2033 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2034 {
2035         const struct rtnl_link_ops *ops = NULL;
2036         struct nlattr *linfo[IFLA_INFO_MAX + 1];
2037
2038         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2039                 return NULL;
2040
2041         if (linfo[IFLA_INFO_KIND]) {
2042                 char kind[MODULE_NAME_LEN];
2043
2044                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2045                 ops = rtnl_link_ops_get(kind);
2046         }
2047
2048         return ops;
2049 }
2050
2051 static bool link_master_filtered(struct net_device *dev, int master_idx)
2052 {
2053         struct net_device *master;
2054
2055         if (!master_idx)
2056                 return false;
2057
2058         master = netdev_master_upper_dev_get(dev);
2059
2060         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2061          * another invalid value for ifindex to denote "no master".
2062          */
2063         if (master_idx == -1)
2064                 return !!master;
2065
2066         if (!master || master->ifindex != master_idx)
2067                 return true;
2068
2069         return false;
2070 }
2071
2072 static bool link_kind_filtered(const struct net_device *dev,
2073                                const struct rtnl_link_ops *kind_ops)
2074 {
2075         if (kind_ops && dev->rtnl_link_ops != kind_ops)
2076                 return true;
2077
2078         return false;
2079 }
2080
2081 static bool link_dump_filtered(struct net_device *dev,
2082                                int master_idx,
2083                                const struct rtnl_link_ops *kind_ops)
2084 {
2085         if (link_master_filtered(dev, master_idx) ||
2086             link_kind_filtered(dev, kind_ops))
2087                 return true;
2088
2089         return false;
2090 }
2091
2092 /**
2093  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2094  * @sk: netlink socket
2095  * @netnsid: network namespace identifier
2096  *
2097  * Returns the network namespace identified by netnsid on success or an error
2098  * pointer on failure.
2099  */
2100 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2101 {
2102         struct net *net;
2103
2104         net = get_net_ns_by_id(sock_net(sk), netnsid);
2105         if (!net)
2106                 return ERR_PTR(-EINVAL);
2107
2108         /* For now, the caller is required to have CAP_NET_ADMIN in
2109          * the user namespace owning the target net ns.
2110          */
2111         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2112                 put_net(net);
2113                 return ERR_PTR(-EACCES);
2114         }
2115         return net;
2116 }
2117 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2118
2119 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2120                                       bool strict_check, struct nlattr **tb,
2121                                       struct netlink_ext_ack *extack)
2122 {
2123         int hdrlen;
2124
2125         if (strict_check) {
2126                 struct ifinfomsg *ifm;
2127
2128                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2129                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2130                         return -EINVAL;
2131                 }
2132
2133                 ifm = nlmsg_data(nlh);
2134                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2135                     ifm->ifi_change) {
2136                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2137                         return -EINVAL;
2138                 }
2139                 if (ifm->ifi_index) {
2140                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2141                         return -EINVAL;
2142                 }
2143
2144                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2145                                                      IFLA_MAX, ifla_policy,
2146                                                      extack);
2147         }
2148
2149         /* A hack to preserve kernel<->userspace interface.
2150          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2151          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2152          * what iproute2 < v3.9.0 used.
2153          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2154          * attribute, its netlink message is shorter than struct ifinfomsg.
2155          */
2156         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2157                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2158
2159         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2160                                       extack);
2161 }
2162
2163 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2164 {
2165         struct netlink_ext_ack *extack = cb->extack;
2166         const struct nlmsghdr *nlh = cb->nlh;
2167         struct net *net = sock_net(skb->sk);
2168         struct net *tgt_net = net;
2169         int h, s_h;
2170         int idx = 0, s_idx;
2171         struct net_device *dev;
2172         struct hlist_head *head;
2173         struct nlattr *tb[IFLA_MAX+1];
2174         u32 ext_filter_mask = 0;
2175         const struct rtnl_link_ops *kind_ops = NULL;
2176         unsigned int flags = NLM_F_MULTI;
2177         int master_idx = 0;
2178         int netnsid = -1;
2179         int err, i;
2180
2181         s_h = cb->args[0];
2182         s_idx = cb->args[1];
2183
2184         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2185         if (err < 0) {
2186                 if (cb->strict_check)
2187                         return err;
2188
2189                 goto walk_entries;
2190         }
2191
2192         for (i = 0; i <= IFLA_MAX; ++i) {
2193                 if (!tb[i])
2194                         continue;
2195
2196                 /* new attributes should only be added with strict checking */
2197                 switch (i) {
2198                 case IFLA_TARGET_NETNSID:
2199                         netnsid = nla_get_s32(tb[i]);
2200                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2201                         if (IS_ERR(tgt_net)) {
2202                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2203                                 return PTR_ERR(tgt_net);
2204                         }
2205                         break;
2206                 case IFLA_EXT_MASK:
2207                         ext_filter_mask = nla_get_u32(tb[i]);
2208                         break;
2209                 case IFLA_MASTER:
2210                         master_idx = nla_get_u32(tb[i]);
2211                         break;
2212                 case IFLA_LINKINFO:
2213                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2214                         break;
2215                 default:
2216                         if (cb->strict_check) {
2217                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2218                                 return -EINVAL;
2219                         }
2220                 }
2221         }
2222
2223         if (master_idx || kind_ops)
2224                 flags |= NLM_F_DUMP_FILTERED;
2225
2226 walk_entries:
2227         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2228                 idx = 0;
2229                 head = &tgt_net->dev_index_head[h];
2230                 hlist_for_each_entry(dev, head, index_hlist) {
2231                         if (link_dump_filtered(dev, master_idx, kind_ops))
2232                                 goto cont;
2233                         if (idx < s_idx)
2234                                 goto cont;
2235                         err = rtnl_fill_ifinfo(skb, dev, net,
2236                                                RTM_NEWLINK,
2237                                                NETLINK_CB(cb->skb).portid,
2238                                                nlh->nlmsg_seq, 0, flags,
2239                                                ext_filter_mask, 0, NULL, 0,
2240                                                netnsid, GFP_KERNEL);
2241
2242                         if (err < 0) {
2243                                 if (likely(skb->len))
2244                                         goto out;
2245
2246                                 goto out_err;
2247                         }
2248 cont:
2249                         idx++;
2250                 }
2251         }
2252 out:
2253         err = skb->len;
2254 out_err:
2255         cb->args[1] = idx;
2256         cb->args[0] = h;
2257         cb->seq = tgt_net->dev_base_seq;
2258         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2259         if (netnsid >= 0)
2260                 put_net(tgt_net);
2261
2262         return err;
2263 }
2264
2265 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2266                         struct netlink_ext_ack *exterr)
2267 {
2268         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2269                                     exterr);
2270 }
2271 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2272
2273 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2274 {
2275         struct net *net;
2276         /* Examine the link attributes and figure out which
2277          * network namespace we are talking about.
2278          */
2279         if (tb[IFLA_NET_NS_PID])
2280                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2281         else if (tb[IFLA_NET_NS_FD])
2282                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2283         else
2284                 net = get_net(src_net);
2285         return net;
2286 }
2287 EXPORT_SYMBOL(rtnl_link_get_net);
2288
2289 /* Figure out which network namespace we are talking about by
2290  * examining the link attributes in the following order:
2291  *
2292  * 1. IFLA_NET_NS_PID
2293  * 2. IFLA_NET_NS_FD
2294  * 3. IFLA_TARGET_NETNSID
2295  */
2296 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2297                                                struct nlattr *tb[])
2298 {
2299         struct net *net;
2300
2301         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2302                 return rtnl_link_get_net(src_net, tb);
2303
2304         if (!tb[IFLA_TARGET_NETNSID])
2305                 return get_net(src_net);
2306
2307         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2308         if (!net)
2309                 return ERR_PTR(-EINVAL);
2310
2311         return net;
2312 }
2313
2314 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2315                                              struct net *src_net,
2316                                              struct nlattr *tb[], int cap)
2317 {
2318         struct net *net;
2319
2320         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2321         if (IS_ERR(net))
2322                 return net;
2323
2324         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2325                 put_net(net);
2326                 return ERR_PTR(-EPERM);
2327         }
2328
2329         return net;
2330 }
2331
2332 /* Verify that rtnetlink requests do not pass additional properties
2333  * potentially referring to different network namespaces.
2334  */
2335 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2336                                     struct netlink_ext_ack *extack,
2337                                     bool netns_id_only)
2338 {
2339
2340         if (netns_id_only) {
2341                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2342                         return 0;
2343
2344                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2345                 return -EOPNOTSUPP;
2346         }
2347
2348         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2349                 goto invalid_attr;
2350
2351         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2352                 goto invalid_attr;
2353
2354         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2355                 goto invalid_attr;
2356
2357         return 0;
2358
2359 invalid_attr:
2360         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2361         return -EINVAL;
2362 }
2363
2364 static  int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2365                              int max_tx_rate)
2366 {
2367         const struct net_device_ops *ops = dev->netdev_ops;
2368
2369         if (!ops->ndo_set_vf_rate)
2370                 return -EOPNOTSUPP;
2371         if (max_tx_rate && max_tx_rate < min_tx_rate)
2372                 return -EINVAL;
2373
2374         return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2375 }
2376
2377 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2378                             struct netlink_ext_ack *extack)
2379 {
2380         if (dev) {
2381                 if (tb[IFLA_ADDRESS] &&
2382                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2383                         return -EINVAL;
2384
2385                 if (tb[IFLA_BROADCAST] &&
2386                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2387                         return -EINVAL;
2388
2389                 if (tb[IFLA_GSO_MAX_SIZE] &&
2390                     nla_get_u32(tb[IFLA_GSO_MAX_SIZE]) > dev->tso_max_size) {
2391                         NL_SET_ERR_MSG(extack, "too big gso_max_size");
2392                         return -EINVAL;
2393                 }
2394
2395                 if (tb[IFLA_GSO_MAX_SEGS] &&
2396                     (nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > GSO_MAX_SEGS ||
2397                      nla_get_u32(tb[IFLA_GSO_MAX_SEGS]) > dev->tso_max_segs)) {
2398                         NL_SET_ERR_MSG(extack, "too big gso_max_segs");
2399                         return -EINVAL;
2400                 }
2401
2402                 if (tb[IFLA_GRO_MAX_SIZE] &&
2403                     nla_get_u32(tb[IFLA_GRO_MAX_SIZE]) > GRO_MAX_SIZE) {
2404                         NL_SET_ERR_MSG(extack, "too big gro_max_size");
2405                         return -EINVAL;
2406                 }
2407
2408                 if (tb[IFLA_GSO_IPV4_MAX_SIZE] &&
2409                     nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]) > dev->tso_max_size) {
2410                         NL_SET_ERR_MSG(extack, "too big gso_ipv4_max_size");
2411                         return -EINVAL;
2412                 }
2413
2414                 if (tb[IFLA_GRO_IPV4_MAX_SIZE] &&
2415                     nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]) > GRO_MAX_SIZE) {
2416                         NL_SET_ERR_MSG(extack, "too big gro_ipv4_max_size");
2417                         return -EINVAL;
2418                 }
2419         }
2420
2421         if (tb[IFLA_AF_SPEC]) {
2422                 struct nlattr *af;
2423                 int rem, err;
2424
2425                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2426                         const struct rtnl_af_ops *af_ops;
2427
2428                         af_ops = rtnl_af_lookup(nla_type(af));
2429                         if (!af_ops)
2430                                 return -EAFNOSUPPORT;
2431
2432                         if (!af_ops->set_link_af)
2433                                 return -EOPNOTSUPP;
2434
2435                         if (af_ops->validate_link_af) {
2436                                 err = af_ops->validate_link_af(dev, af, extack);
2437                                 if (err < 0)
2438                                         return err;
2439                         }
2440                 }
2441         }
2442
2443         return 0;
2444 }
2445
2446 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2447                                   int guid_type)
2448 {
2449         const struct net_device_ops *ops = dev->netdev_ops;
2450
2451         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2452 }
2453
2454 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2455 {
2456         if (dev->type != ARPHRD_INFINIBAND)
2457                 return -EOPNOTSUPP;
2458
2459         return handle_infiniband_guid(dev, ivt, guid_type);
2460 }
2461
2462 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2463 {
2464         const struct net_device_ops *ops = dev->netdev_ops;
2465         int err = -EINVAL;
2466
2467         if (tb[IFLA_VF_MAC]) {
2468                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2469
2470                 if (ivm->vf >= INT_MAX)
2471                         return -EINVAL;
2472                 err = -EOPNOTSUPP;
2473                 if (ops->ndo_set_vf_mac)
2474                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2475                                                   ivm->mac);
2476                 if (err < 0)
2477                         return err;
2478         }
2479
2480         if (tb[IFLA_VF_VLAN]) {
2481                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2482
2483                 if (ivv->vf >= INT_MAX)
2484                         return -EINVAL;
2485                 err = -EOPNOTSUPP;
2486                 if (ops->ndo_set_vf_vlan)
2487                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2488                                                    ivv->qos,
2489                                                    htons(ETH_P_8021Q));
2490                 if (err < 0)
2491                         return err;
2492         }
2493
2494         if (tb[IFLA_VF_VLAN_LIST]) {
2495                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2496                 struct nlattr *attr;
2497                 int rem, len = 0;
2498
2499                 err = -EOPNOTSUPP;
2500                 if (!ops->ndo_set_vf_vlan)
2501                         return err;
2502
2503                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2504                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2505                             nla_len(attr) < NLA_HDRLEN) {
2506                                 return -EINVAL;
2507                         }
2508                         if (len >= MAX_VLAN_LIST_LEN)
2509                                 return -EOPNOTSUPP;
2510                         ivvl[len] = nla_data(attr);
2511
2512                         len++;
2513                 }
2514                 if (len == 0)
2515                         return -EINVAL;
2516
2517                 if (ivvl[0]->vf >= INT_MAX)
2518                         return -EINVAL;
2519                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2520                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2521                 if (err < 0)
2522                         return err;
2523         }
2524
2525         if (tb[IFLA_VF_TX_RATE]) {
2526                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2527                 struct ifla_vf_info ivf;
2528
2529                 if (ivt->vf >= INT_MAX)
2530                         return -EINVAL;
2531                 err = -EOPNOTSUPP;
2532                 if (ops->ndo_get_vf_config)
2533                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2534                 if (err < 0)
2535                         return err;
2536
2537                 err = rtnl_set_vf_rate(dev, ivt->vf,
2538                                        ivf.min_tx_rate, ivt->rate);
2539                 if (err < 0)
2540                         return err;
2541         }
2542
2543         if (tb[IFLA_VF_RATE]) {
2544                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2545
2546                 if (ivt->vf >= INT_MAX)
2547                         return -EINVAL;
2548
2549                 err = rtnl_set_vf_rate(dev, ivt->vf,
2550                                        ivt->min_tx_rate, ivt->max_tx_rate);
2551                 if (err < 0)
2552                         return err;
2553         }
2554
2555         if (tb[IFLA_VF_SPOOFCHK]) {
2556                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2557
2558                 if (ivs->vf >= INT_MAX)
2559                         return -EINVAL;
2560                 err = -EOPNOTSUPP;
2561                 if (ops->ndo_set_vf_spoofchk)
2562                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2563                                                        ivs->setting);
2564                 if (err < 0)
2565                         return err;
2566         }
2567
2568         if (tb[IFLA_VF_LINK_STATE]) {
2569                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2570
2571                 if (ivl->vf >= INT_MAX)
2572                         return -EINVAL;
2573                 err = -EOPNOTSUPP;
2574                 if (ops->ndo_set_vf_link_state)
2575                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2576                                                          ivl->link_state);
2577                 if (err < 0)
2578                         return err;
2579         }
2580
2581         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2582                 struct ifla_vf_rss_query_en *ivrssq_en;
2583
2584                 err = -EOPNOTSUPP;
2585                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2586                 if (ivrssq_en->vf >= INT_MAX)
2587                         return -EINVAL;
2588                 if (ops->ndo_set_vf_rss_query_en)
2589                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2590                                                            ivrssq_en->setting);
2591                 if (err < 0)
2592                         return err;
2593         }
2594
2595         if (tb[IFLA_VF_TRUST]) {
2596                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2597
2598                 if (ivt->vf >= INT_MAX)
2599                         return -EINVAL;
2600                 err = -EOPNOTSUPP;
2601                 if (ops->ndo_set_vf_trust)
2602                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2603                 if (err < 0)
2604                         return err;
2605         }
2606
2607         if (tb[IFLA_VF_IB_NODE_GUID]) {
2608                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2609
2610                 if (ivt->vf >= INT_MAX)
2611                         return -EINVAL;
2612                 if (!ops->ndo_set_vf_guid)
2613                         return -EOPNOTSUPP;
2614                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2615         }
2616
2617         if (tb[IFLA_VF_IB_PORT_GUID]) {
2618                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2619
2620                 if (ivt->vf >= INT_MAX)
2621                         return -EINVAL;
2622                 if (!ops->ndo_set_vf_guid)
2623                         return -EOPNOTSUPP;
2624
2625                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2626         }
2627
2628         return err;
2629 }
2630
2631 static int do_set_master(struct net_device *dev, int ifindex,
2632                          struct netlink_ext_ack *extack)
2633 {
2634         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2635         const struct net_device_ops *ops;
2636         int err;
2637
2638         if (upper_dev) {
2639                 if (upper_dev->ifindex == ifindex)
2640                         return 0;
2641                 ops = upper_dev->netdev_ops;
2642                 if (ops->ndo_del_slave) {
2643                         err = ops->ndo_del_slave(upper_dev, dev);
2644                         if (err)
2645                                 return err;
2646                 } else {
2647                         return -EOPNOTSUPP;
2648                 }
2649         }
2650
2651         if (ifindex) {
2652                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2653                 if (!upper_dev)
2654                         return -EINVAL;
2655                 ops = upper_dev->netdev_ops;
2656                 if (ops->ndo_add_slave) {
2657                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2658                         if (err)
2659                                 return err;
2660                 } else {
2661                         return -EOPNOTSUPP;
2662                 }
2663         }
2664         return 0;
2665 }
2666
2667 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2668         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2669         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2670 };
2671
2672 static int do_set_proto_down(struct net_device *dev,
2673                              struct nlattr *nl_proto_down,
2674                              struct nlattr *nl_proto_down_reason,
2675                              struct netlink_ext_ack *extack)
2676 {
2677         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2678         unsigned long mask = 0;
2679         u32 value;
2680         bool proto_down;
2681         int err;
2682
2683         if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2684                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2685                 return -EOPNOTSUPP;
2686         }
2687
2688         if (nl_proto_down_reason) {
2689                 err = nla_parse_nested_deprecated(pdreason,
2690                                                   IFLA_PROTO_DOWN_REASON_MAX,
2691                                                   nl_proto_down_reason,
2692                                                   ifla_proto_down_reason_policy,
2693                                                   NULL);
2694                 if (err < 0)
2695                         return err;
2696
2697                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2698                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2699                         return -EINVAL;
2700                 }
2701
2702                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2703
2704                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2705                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2706
2707                 dev_change_proto_down_reason(dev, mask, value);
2708         }
2709
2710         if (nl_proto_down) {
2711                 proto_down = nla_get_u8(nl_proto_down);
2712
2713                 /* Don't turn off protodown if there are active reasons */
2714                 if (!proto_down && dev->proto_down_reason) {
2715                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2716                         return -EBUSY;
2717                 }
2718                 err = dev_change_proto_down(dev,
2719                                             proto_down);
2720                 if (err)
2721                         return err;
2722         }
2723
2724         return 0;
2725 }
2726
2727 #define DO_SETLINK_MODIFIED     0x01
2728 /* notify flag means notify + modified. */
2729 #define DO_SETLINK_NOTIFY       0x03
2730 static int do_setlink(const struct sk_buff *skb,
2731                       struct net_device *dev, struct ifinfomsg *ifm,
2732                       struct netlink_ext_ack *extack,
2733                       struct nlattr **tb, int status)
2734 {
2735         const struct net_device_ops *ops = dev->netdev_ops;
2736         char ifname[IFNAMSIZ];
2737         int err;
2738
2739         err = validate_linkmsg(dev, tb, extack);
2740         if (err < 0)
2741                 return err;
2742
2743         if (tb[IFLA_IFNAME])
2744                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2745         else
2746                 ifname[0] = '\0';
2747
2748         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2749                 const char *pat = ifname[0] ? ifname : NULL;
2750                 struct net *net;
2751                 int new_ifindex;
2752
2753                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2754                                                 tb, CAP_NET_ADMIN);
2755                 if (IS_ERR(net)) {
2756                         err = PTR_ERR(net);
2757                         goto errout;
2758                 }
2759
2760                 if (tb[IFLA_NEW_IFINDEX])
2761                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2762                 else
2763                         new_ifindex = 0;
2764
2765                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2766                 put_net(net);
2767                 if (err)
2768                         goto errout;
2769                 status |= DO_SETLINK_MODIFIED;
2770         }
2771
2772         if (tb[IFLA_MAP]) {
2773                 struct rtnl_link_ifmap *u_map;
2774                 struct ifmap k_map;
2775
2776                 if (!ops->ndo_set_config) {
2777                         err = -EOPNOTSUPP;
2778                         goto errout;
2779                 }
2780
2781                 if (!netif_device_present(dev)) {
2782                         err = -ENODEV;
2783                         goto errout;
2784                 }
2785
2786                 u_map = nla_data(tb[IFLA_MAP]);
2787                 k_map.mem_start = (unsigned long) u_map->mem_start;
2788                 k_map.mem_end = (unsigned long) u_map->mem_end;
2789                 k_map.base_addr = (unsigned short) u_map->base_addr;
2790                 k_map.irq = (unsigned char) u_map->irq;
2791                 k_map.dma = (unsigned char) u_map->dma;
2792                 k_map.port = (unsigned char) u_map->port;
2793
2794                 err = ops->ndo_set_config(dev, &k_map);
2795                 if (err < 0)
2796                         goto errout;
2797
2798                 status |= DO_SETLINK_NOTIFY;
2799         }
2800
2801         if (tb[IFLA_ADDRESS]) {
2802                 struct sockaddr *sa;
2803                 int len;
2804
2805                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2806                                                   sizeof(*sa));
2807                 sa = kmalloc(len, GFP_KERNEL);
2808                 if (!sa) {
2809                         err = -ENOMEM;
2810                         goto errout;
2811                 }
2812                 sa->sa_family = dev->type;
2813                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2814                        dev->addr_len);
2815                 err = dev_set_mac_address_user(dev, sa, extack);
2816                 kfree(sa);
2817                 if (err)
2818                         goto errout;
2819                 status |= DO_SETLINK_MODIFIED;
2820         }
2821
2822         if (tb[IFLA_MTU]) {
2823                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2824                 if (err < 0)
2825                         goto errout;
2826                 status |= DO_SETLINK_MODIFIED;
2827         }
2828
2829         if (tb[IFLA_GROUP]) {
2830                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2831                 status |= DO_SETLINK_NOTIFY;
2832         }
2833
2834         /*
2835          * Interface selected by interface index but interface
2836          * name provided implies that a name change has been
2837          * requested.
2838          */
2839         if (ifm->ifi_index > 0 && ifname[0]) {
2840                 err = dev_change_name(dev, ifname);
2841                 if (err < 0)
2842                         goto errout;
2843                 status |= DO_SETLINK_MODIFIED;
2844         }
2845
2846         if (tb[IFLA_IFALIAS]) {
2847                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2848                                     nla_len(tb[IFLA_IFALIAS]));
2849                 if (err < 0)
2850                         goto errout;
2851                 status |= DO_SETLINK_NOTIFY;
2852         }
2853
2854         if (tb[IFLA_BROADCAST]) {
2855                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2856                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2857         }
2858
2859         if (tb[IFLA_MASTER]) {
2860                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2861                 if (err)
2862                         goto errout;
2863                 status |= DO_SETLINK_MODIFIED;
2864         }
2865
2866         if (ifm->ifi_flags || ifm->ifi_change) {
2867                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2868                                        extack);
2869                 if (err < 0)
2870                         goto errout;
2871         }
2872
2873         if (tb[IFLA_CARRIER]) {
2874                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2875                 if (err)
2876                         goto errout;
2877                 status |= DO_SETLINK_MODIFIED;
2878         }
2879
2880         if (tb[IFLA_TXQLEN]) {
2881                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2882
2883                 err = dev_change_tx_queue_len(dev, value);
2884                 if (err)
2885                         goto errout;
2886                 status |= DO_SETLINK_MODIFIED;
2887         }
2888
2889         if (tb[IFLA_GSO_MAX_SIZE]) {
2890                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2891
2892                 if (dev->gso_max_size ^ max_size) {
2893                         netif_set_gso_max_size(dev, max_size);
2894                         status |= DO_SETLINK_MODIFIED;
2895                 }
2896         }
2897
2898         if (tb[IFLA_GSO_MAX_SEGS]) {
2899                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2900
2901                 if (dev->gso_max_segs ^ max_segs) {
2902                         netif_set_gso_max_segs(dev, max_segs);
2903                         status |= DO_SETLINK_MODIFIED;
2904                 }
2905         }
2906
2907         if (tb[IFLA_GRO_MAX_SIZE]) {
2908                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2909
2910                 if (dev->gro_max_size ^ gro_max_size) {
2911                         netif_set_gro_max_size(dev, gro_max_size);
2912                         status |= DO_SETLINK_MODIFIED;
2913                 }
2914         }
2915
2916         if (tb[IFLA_GSO_IPV4_MAX_SIZE]) {
2917                 u32 max_size = nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]);
2918
2919                 if (dev->gso_ipv4_max_size ^ max_size) {
2920                         netif_set_gso_ipv4_max_size(dev, max_size);
2921                         status |= DO_SETLINK_MODIFIED;
2922                 }
2923         }
2924
2925         if (tb[IFLA_GRO_IPV4_MAX_SIZE]) {
2926                 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]);
2927
2928                 if (dev->gro_ipv4_max_size ^ gro_max_size) {
2929                         netif_set_gro_ipv4_max_size(dev, gro_max_size);
2930                         status |= DO_SETLINK_MODIFIED;
2931                 }
2932         }
2933
2934         if (tb[IFLA_OPERSTATE])
2935                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2936
2937         if (tb[IFLA_LINKMODE]) {
2938                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2939
2940                 write_lock(&dev_base_lock);
2941                 if (dev->link_mode ^ value)
2942                         status |= DO_SETLINK_NOTIFY;
2943                 dev->link_mode = value;
2944                 write_unlock(&dev_base_lock);
2945         }
2946
2947         if (tb[IFLA_VFINFO_LIST]) {
2948                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2949                 struct nlattr *attr;
2950                 int rem;
2951
2952                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2953                         if (nla_type(attr) != IFLA_VF_INFO ||
2954                             nla_len(attr) < NLA_HDRLEN) {
2955                                 err = -EINVAL;
2956                                 goto errout;
2957                         }
2958                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2959                                                           attr,
2960                                                           ifla_vf_policy,
2961                                                           NULL);
2962                         if (err < 0)
2963                                 goto errout;
2964                         err = do_setvfinfo(dev, vfinfo);
2965                         if (err < 0)
2966                                 goto errout;
2967                         status |= DO_SETLINK_NOTIFY;
2968                 }
2969         }
2970         err = 0;
2971
2972         if (tb[IFLA_VF_PORTS]) {
2973                 struct nlattr *port[IFLA_PORT_MAX+1];
2974                 struct nlattr *attr;
2975                 int vf;
2976                 int rem;
2977
2978                 err = -EOPNOTSUPP;
2979                 if (!ops->ndo_set_vf_port)
2980                         goto errout;
2981
2982                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2983                         if (nla_type(attr) != IFLA_VF_PORT ||
2984                             nla_len(attr) < NLA_HDRLEN) {
2985                                 err = -EINVAL;
2986                                 goto errout;
2987                         }
2988                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2989                                                           attr,
2990                                                           ifla_port_policy,
2991                                                           NULL);
2992                         if (err < 0)
2993                                 goto errout;
2994                         if (!port[IFLA_PORT_VF]) {
2995                                 err = -EOPNOTSUPP;
2996                                 goto errout;
2997                         }
2998                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2999                         err = ops->ndo_set_vf_port(dev, vf, port);
3000                         if (err < 0)
3001                                 goto errout;
3002                         status |= DO_SETLINK_NOTIFY;
3003                 }
3004         }
3005         err = 0;
3006
3007         if (tb[IFLA_PORT_SELF]) {
3008                 struct nlattr *port[IFLA_PORT_MAX+1];
3009
3010                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
3011                                                   tb[IFLA_PORT_SELF],
3012                                                   ifla_port_policy, NULL);
3013                 if (err < 0)
3014                         goto errout;
3015
3016                 err = -EOPNOTSUPP;
3017                 if (ops->ndo_set_vf_port)
3018                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
3019                 if (err < 0)
3020                         goto errout;
3021                 status |= DO_SETLINK_NOTIFY;
3022         }
3023
3024         if (tb[IFLA_AF_SPEC]) {
3025                 struct nlattr *af;
3026                 int rem;
3027
3028                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
3029                         const struct rtnl_af_ops *af_ops;
3030
3031                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
3032
3033                         err = af_ops->set_link_af(dev, af, extack);
3034                         if (err < 0)
3035                                 goto errout;
3036
3037                         status |= DO_SETLINK_NOTIFY;
3038                 }
3039         }
3040         err = 0;
3041
3042         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
3043                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
3044                                         tb[IFLA_PROTO_DOWN_REASON], extack);
3045                 if (err)
3046                         goto errout;
3047                 status |= DO_SETLINK_NOTIFY;
3048         }
3049
3050         if (tb[IFLA_XDP]) {
3051                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3052                 u32 xdp_flags = 0;
3053
3054                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3055                                                   tb[IFLA_XDP],
3056                                                   ifla_xdp_policy, NULL);
3057                 if (err < 0)
3058                         goto errout;
3059
3060                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3061                         err = -EINVAL;
3062                         goto errout;
3063                 }
3064
3065                 if (xdp[IFLA_XDP_FLAGS]) {
3066                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3067                         if (xdp_flags & ~XDP_FLAGS_MASK) {
3068                                 err = -EINVAL;
3069                                 goto errout;
3070                         }
3071                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3072                                 err = -EINVAL;
3073                                 goto errout;
3074                         }
3075                 }
3076
3077                 if (xdp[IFLA_XDP_FD]) {
3078                         int expected_fd = -1;
3079
3080                         if (xdp_flags & XDP_FLAGS_REPLACE) {
3081                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3082                                         err = -EINVAL;
3083                                         goto errout;
3084                                 }
3085                                 expected_fd =
3086                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3087                         }
3088
3089                         err = dev_change_xdp_fd(dev, extack,
3090                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
3091                                                 expected_fd,
3092                                                 xdp_flags);
3093                         if (err)
3094                                 goto errout;
3095                         status |= DO_SETLINK_NOTIFY;
3096                 }
3097         }
3098
3099 errout:
3100         if (status & DO_SETLINK_MODIFIED) {
3101                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3102                         netdev_state_change(dev);
3103
3104                 if (err < 0)
3105                         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",
3106                                              dev->name);
3107         }
3108
3109         return err;
3110 }
3111
3112 static struct net_device *rtnl_dev_get(struct net *net,
3113                                        struct nlattr *tb[])
3114 {
3115         char ifname[ALTIFNAMSIZ];
3116
3117         if (tb[IFLA_IFNAME])
3118                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3119         else if (tb[IFLA_ALT_IFNAME])
3120                 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3121         else
3122                 return NULL;
3123
3124         return __dev_get_by_name(net, ifname);
3125 }
3126
3127 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3128                         struct netlink_ext_ack *extack)
3129 {
3130         struct net *net = sock_net(skb->sk);
3131         struct ifinfomsg *ifm;
3132         struct net_device *dev;
3133         int err;
3134         struct nlattr *tb[IFLA_MAX+1];
3135
3136         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3137                                      ifla_policy, extack);
3138         if (err < 0)
3139                 goto errout;
3140
3141         err = rtnl_ensure_unique_netns(tb, extack, false);
3142         if (err < 0)
3143                 goto errout;
3144
3145         err = -EINVAL;
3146         ifm = nlmsg_data(nlh);
3147         if (ifm->ifi_index > 0)
3148                 dev = __dev_get_by_index(net, ifm->ifi_index);
3149         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3150                 dev = rtnl_dev_get(net, tb);
3151         else
3152                 goto errout;
3153
3154         if (dev == NULL) {
3155                 err = -ENODEV;
3156                 goto errout;
3157         }
3158
3159         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3160 errout:
3161         return err;
3162 }
3163
3164 static int rtnl_group_dellink(const struct net *net, int group)
3165 {
3166         struct net_device *dev, *aux;
3167         LIST_HEAD(list_kill);
3168         bool found = false;
3169
3170         if (!group)
3171                 return -EPERM;
3172
3173         for_each_netdev(net, dev) {
3174                 if (dev->group == group) {
3175                         const struct rtnl_link_ops *ops;
3176
3177                         found = true;
3178                         ops = dev->rtnl_link_ops;
3179                         if (!ops || !ops->dellink)
3180                                 return -EOPNOTSUPP;
3181                 }
3182         }
3183
3184         if (!found)
3185                 return -ENODEV;
3186
3187         for_each_netdev_safe(net, dev, aux) {
3188                 if (dev->group == group) {
3189                         const struct rtnl_link_ops *ops;
3190
3191                         ops = dev->rtnl_link_ops;
3192                         ops->dellink(dev, &list_kill);
3193                 }
3194         }
3195         unregister_netdevice_many(&list_kill);
3196
3197         return 0;
3198 }
3199
3200 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3201 {
3202         const struct rtnl_link_ops *ops;
3203         LIST_HEAD(list_kill);
3204
3205         ops = dev->rtnl_link_ops;
3206         if (!ops || !ops->dellink)
3207                 return -EOPNOTSUPP;
3208
3209         ops->dellink(dev, &list_kill);
3210         unregister_netdevice_many_notify(&list_kill, portid, nlh);
3211
3212         return 0;
3213 }
3214 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3215
3216 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3217                         struct netlink_ext_ack *extack)
3218 {
3219         struct net *net = sock_net(skb->sk);
3220         u32 portid = NETLINK_CB(skb).portid;
3221         struct net *tgt_net = net;
3222         struct net_device *dev = NULL;
3223         struct ifinfomsg *ifm;
3224         struct nlattr *tb[IFLA_MAX+1];
3225         int err;
3226         int netnsid = -1;
3227
3228         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3229                                      ifla_policy, extack);
3230         if (err < 0)
3231                 return err;
3232
3233         err = rtnl_ensure_unique_netns(tb, extack, true);
3234         if (err < 0)
3235                 return err;
3236
3237         if (tb[IFLA_TARGET_NETNSID]) {
3238                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3239                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3240                 if (IS_ERR(tgt_net))
3241                         return PTR_ERR(tgt_net);
3242         }
3243
3244         err = -EINVAL;
3245         ifm = nlmsg_data(nlh);
3246         if (ifm->ifi_index > 0)
3247                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3248         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3249                 dev = rtnl_dev_get(net, tb);
3250         else if (tb[IFLA_GROUP])
3251                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3252         else
3253                 goto out;
3254
3255         if (!dev) {
3256                 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3257                         err = -ENODEV;
3258
3259                 goto out;
3260         }
3261
3262         err = rtnl_delete_link(dev, portid, nlh);
3263
3264 out:
3265         if (netnsid >= 0)
3266                 put_net(tgt_net);
3267
3268         return err;
3269 }
3270
3271 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3272                         u32 portid, const struct nlmsghdr *nlh)
3273 {
3274         unsigned int old_flags;
3275         int err;
3276
3277         old_flags = dev->flags;
3278         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3279                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3280                                          NULL);
3281                 if (err < 0)
3282                         return err;
3283         }
3284
3285         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3286                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3287         } else {
3288                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3289                 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3290         }
3291         return 0;
3292 }
3293 EXPORT_SYMBOL(rtnl_configure_link);
3294
3295 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3296                                     unsigned char name_assign_type,
3297                                     const struct rtnl_link_ops *ops,
3298                                     struct nlattr *tb[],
3299                                     struct netlink_ext_ack *extack)
3300 {
3301         struct net_device *dev;
3302         unsigned int num_tx_queues = 1;
3303         unsigned int num_rx_queues = 1;
3304         int err;
3305
3306         if (tb[IFLA_NUM_TX_QUEUES])
3307                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3308         else if (ops->get_num_tx_queues)
3309                 num_tx_queues = ops->get_num_tx_queues();
3310
3311         if (tb[IFLA_NUM_RX_QUEUES])
3312                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3313         else if (ops->get_num_rx_queues)
3314                 num_rx_queues = ops->get_num_rx_queues();
3315
3316         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3317                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3318                 return ERR_PTR(-EINVAL);
3319         }
3320
3321         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3322                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3323                 return ERR_PTR(-EINVAL);
3324         }
3325
3326         if (ops->alloc) {
3327                 dev = ops->alloc(tb, ifname, name_assign_type,
3328                                  num_tx_queues, num_rx_queues);
3329                 if (IS_ERR(dev))
3330                         return dev;
3331         } else {
3332                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3333                                        name_assign_type, ops->setup,
3334                                        num_tx_queues, num_rx_queues);
3335         }
3336
3337         if (!dev)
3338                 return ERR_PTR(-ENOMEM);
3339
3340         err = validate_linkmsg(dev, tb, extack);
3341         if (err < 0) {
3342                 free_netdev(dev);
3343                 return ERR_PTR(err);
3344         }
3345
3346         dev_net_set(dev, net);
3347         dev->rtnl_link_ops = ops;
3348         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3349
3350         if (tb[IFLA_MTU]) {
3351                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3352
3353                 err = dev_validate_mtu(dev, mtu, extack);
3354                 if (err) {
3355                         free_netdev(dev);
3356                         return ERR_PTR(err);
3357                 }
3358                 dev->mtu = mtu;
3359         }
3360         if (tb[IFLA_ADDRESS]) {
3361                 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3362                                nla_len(tb[IFLA_ADDRESS]));
3363                 dev->addr_assign_type = NET_ADDR_SET;
3364         }
3365         if (tb[IFLA_BROADCAST])
3366                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3367                                 nla_len(tb[IFLA_BROADCAST]));
3368         if (tb[IFLA_TXQLEN])
3369                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3370         if (tb[IFLA_OPERSTATE])
3371                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3372         if (tb[IFLA_LINKMODE])
3373                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3374         if (tb[IFLA_GROUP])
3375                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3376         if (tb[IFLA_GSO_MAX_SIZE])
3377                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3378         if (tb[IFLA_GSO_MAX_SEGS])
3379                 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3380         if (tb[IFLA_GRO_MAX_SIZE])
3381                 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3382         if (tb[IFLA_GSO_IPV4_MAX_SIZE])
3383                 netif_set_gso_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GSO_IPV4_MAX_SIZE]));
3384         if (tb[IFLA_GRO_IPV4_MAX_SIZE])
3385                 netif_set_gro_ipv4_max_size(dev, nla_get_u32(tb[IFLA_GRO_IPV4_MAX_SIZE]));
3386
3387         return dev;
3388 }
3389 EXPORT_SYMBOL(rtnl_create_link);
3390
3391 static int rtnl_group_changelink(const struct sk_buff *skb,
3392                 struct net *net, int group,
3393                 struct ifinfomsg *ifm,
3394                 struct netlink_ext_ack *extack,
3395                 struct nlattr **tb)
3396 {
3397         struct net_device *dev, *aux;
3398         int err;
3399
3400         for_each_netdev_safe(net, dev, aux) {
3401                 if (dev->group == group) {
3402                         err = do_setlink(skb, dev, ifm, extack, tb, 0);
3403                         if (err < 0)
3404                                 return err;
3405                 }
3406         }
3407
3408         return 0;
3409 }
3410
3411 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3412                                const struct rtnl_link_ops *ops,
3413                                const struct nlmsghdr *nlh,
3414                                struct nlattr **tb, struct nlattr **data,
3415                                struct netlink_ext_ack *extack)
3416 {
3417         unsigned char name_assign_type = NET_NAME_USER;
3418         struct net *net = sock_net(skb->sk);
3419         u32 portid = NETLINK_CB(skb).portid;
3420         struct net *dest_net, *link_net;
3421         struct net_device *dev;
3422         char ifname[IFNAMSIZ];
3423         int err;
3424
3425         if (!ops->alloc && !ops->setup)
3426                 return -EOPNOTSUPP;
3427
3428         if (tb[IFLA_IFNAME]) {
3429                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3430         } else {
3431                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3432                 name_assign_type = NET_NAME_ENUM;
3433         }
3434
3435         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3436         if (IS_ERR(dest_net))
3437                 return PTR_ERR(dest_net);
3438
3439         if (tb[IFLA_LINK_NETNSID]) {
3440                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3441
3442                 link_net = get_net_ns_by_id(dest_net, id);
3443                 if (!link_net) {
3444                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3445                         err =  -EINVAL;
3446                         goto out;
3447                 }
3448                 err = -EPERM;
3449                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3450                         goto out;
3451         } else {
3452                 link_net = NULL;
3453         }
3454
3455         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3456                                name_assign_type, ops, tb, extack);
3457         if (IS_ERR(dev)) {
3458                 err = PTR_ERR(dev);
3459                 goto out;
3460         }
3461
3462         dev->ifindex = ifm->ifi_index;
3463
3464         if (ops->newlink)
3465                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3466         else
3467                 err = register_netdevice(dev);
3468         if (err < 0) {
3469                 free_netdev(dev);
3470                 goto out;
3471         }
3472
3473         err = rtnl_configure_link(dev, ifm, portid, nlh);
3474         if (err < 0)
3475                 goto out_unregister;
3476         if (link_net) {
3477                 err = dev_change_net_namespace(dev, dest_net, ifname);
3478                 if (err < 0)
3479                         goto out_unregister;
3480         }
3481         if (tb[IFLA_MASTER]) {
3482                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3483                 if (err)
3484                         goto out_unregister;
3485         }
3486 out:
3487         if (link_net)
3488                 put_net(link_net);
3489         put_net(dest_net);
3490         return err;
3491 out_unregister:
3492         if (ops->newlink) {
3493                 LIST_HEAD(list_kill);
3494
3495                 ops->dellink(dev, &list_kill);
3496                 unregister_netdevice_many(&list_kill);
3497         } else {
3498                 unregister_netdevice(dev);
3499         }
3500         goto out;
3501 }
3502
3503 struct rtnl_newlink_tbs {
3504         struct nlattr *tb[IFLA_MAX + 1];
3505         struct nlattr *attr[RTNL_MAX_TYPE + 1];
3506         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3507 };
3508
3509 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3510                           struct rtnl_newlink_tbs *tbs,
3511                           struct netlink_ext_ack *extack)
3512 {
3513         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3514         struct nlattr ** const tb = tbs->tb;
3515         const struct rtnl_link_ops *m_ops;
3516         struct net_device *master_dev;
3517         struct net *net = sock_net(skb->sk);
3518         const struct rtnl_link_ops *ops;
3519         struct nlattr **slave_data;
3520         char kind[MODULE_NAME_LEN];
3521         struct net_device *dev;
3522         struct ifinfomsg *ifm;
3523         struct nlattr **data;
3524         bool link_specified;
3525         int err;
3526
3527 #ifdef CONFIG_MODULES
3528 replay:
3529 #endif
3530         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3531                                      ifla_policy, extack);
3532         if (err < 0)
3533                 return err;
3534
3535         err = rtnl_ensure_unique_netns(tb, extack, false);
3536         if (err < 0)
3537                 return err;
3538
3539         ifm = nlmsg_data(nlh);
3540         if (ifm->ifi_index > 0) {
3541                 link_specified = true;
3542                 dev = __dev_get_by_index(net, ifm->ifi_index);
3543         } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3544                 link_specified = true;
3545                 dev = rtnl_dev_get(net, tb);
3546         } else {
3547                 link_specified = false;
3548                 dev = NULL;
3549         }
3550
3551         master_dev = NULL;
3552         m_ops = NULL;
3553         if (dev) {
3554                 master_dev = netdev_master_upper_dev_get(dev);
3555                 if (master_dev)
3556                         m_ops = master_dev->rtnl_link_ops;
3557         }
3558
3559         err = validate_linkmsg(dev, tb, extack);
3560         if (err < 0)
3561                 return err;
3562
3563         if (tb[IFLA_LINKINFO]) {
3564                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3565                                                   tb[IFLA_LINKINFO],
3566                                                   ifla_info_policy, NULL);
3567                 if (err < 0)
3568                         return err;
3569         } else
3570                 memset(linkinfo, 0, sizeof(linkinfo));
3571
3572         if (linkinfo[IFLA_INFO_KIND]) {
3573                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3574                 ops = rtnl_link_ops_get(kind);
3575         } else {
3576                 kind[0] = '\0';
3577                 ops = NULL;
3578         }
3579
3580         data = NULL;
3581         if (ops) {
3582                 if (ops->maxtype > RTNL_MAX_TYPE)
3583                         return -EINVAL;
3584
3585                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3586                         err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3587                                                           linkinfo[IFLA_INFO_DATA],
3588                                                           ops->policy, extack);
3589                         if (err < 0)
3590                                 return err;
3591                         data = tbs->attr;
3592                 }
3593                 if (ops->validate) {
3594                         err = ops->validate(tb, data, extack);
3595                         if (err < 0)
3596                                 return err;
3597                 }
3598         }
3599
3600         slave_data = NULL;
3601         if (m_ops) {
3602                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3603                         return -EINVAL;
3604
3605                 if (m_ops->slave_maxtype &&
3606                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3607                         err = nla_parse_nested_deprecated(tbs->slave_attr,
3608                                                           m_ops->slave_maxtype,
3609                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3610                                                           m_ops->slave_policy,
3611                                                           extack);
3612                         if (err < 0)
3613                                 return err;
3614                         slave_data = tbs->slave_attr;
3615                 }
3616         }
3617
3618         if (dev) {
3619                 int status = 0;
3620
3621                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3622                         return -EEXIST;
3623                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3624                         return -EOPNOTSUPP;
3625
3626                 if (linkinfo[IFLA_INFO_DATA]) {
3627                         if (!ops || ops != dev->rtnl_link_ops ||
3628                             !ops->changelink)
3629                                 return -EOPNOTSUPP;
3630
3631                         err = ops->changelink(dev, tb, data, extack);
3632                         if (err < 0)
3633                                 return err;
3634                         status |= DO_SETLINK_NOTIFY;
3635                 }
3636
3637                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3638                         if (!m_ops || !m_ops->slave_changelink)
3639                                 return -EOPNOTSUPP;
3640
3641                         err = m_ops->slave_changelink(master_dev, dev, tb,
3642                                                       slave_data, extack);
3643                         if (err < 0)
3644                                 return err;
3645                         status |= DO_SETLINK_NOTIFY;
3646                 }
3647
3648                 return do_setlink(skb, dev, ifm, extack, tb, status);
3649         }
3650
3651         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3652                 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3653                  * or it's for a group
3654                 */
3655                 if (link_specified)
3656                         return -ENODEV;
3657                 if (tb[IFLA_GROUP])
3658                         return rtnl_group_changelink(skb, net,
3659                                                 nla_get_u32(tb[IFLA_GROUP]),
3660                                                 ifm, extack, tb);
3661                 return -ENODEV;
3662         }
3663
3664         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3665                 return -EOPNOTSUPP;
3666
3667         if (!ops) {
3668 #ifdef CONFIG_MODULES
3669                 if (kind[0]) {
3670                         __rtnl_unlock();
3671                         request_module("rtnl-link-%s", kind);
3672                         rtnl_lock();
3673                         ops = rtnl_link_ops_get(kind);
3674                         if (ops)
3675                                 goto replay;
3676                 }
3677 #endif
3678                 NL_SET_ERR_MSG(extack, "Unknown device type");
3679                 return -EOPNOTSUPP;
3680         }
3681
3682         return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3683 }
3684
3685 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3686                         struct netlink_ext_ack *extack)
3687 {
3688         struct rtnl_newlink_tbs *tbs;
3689         int ret;
3690
3691         tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3692         if (!tbs)
3693                 return -ENOMEM;
3694
3695         ret = __rtnl_newlink(skb, nlh, tbs, extack);
3696         kfree(tbs);
3697         return ret;
3698 }
3699
3700 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3701                                   const struct nlmsghdr *nlh,
3702                                   struct nlattr **tb,
3703                                   struct netlink_ext_ack *extack)
3704 {
3705         struct ifinfomsg *ifm;
3706         int i, err;
3707
3708         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3709                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3710                 return -EINVAL;
3711         }
3712
3713         if (!netlink_strict_get_check(skb))
3714                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3715                                               ifla_policy, extack);
3716
3717         ifm = nlmsg_data(nlh);
3718         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3719             ifm->ifi_change) {
3720                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3721                 return -EINVAL;
3722         }
3723
3724         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3725                                             ifla_policy, extack);
3726         if (err)
3727                 return err;
3728
3729         for (i = 0; i <= IFLA_MAX; i++) {
3730                 if (!tb[i])
3731                         continue;
3732
3733                 switch (i) {
3734                 case IFLA_IFNAME:
3735                 case IFLA_ALT_IFNAME:
3736                 case IFLA_EXT_MASK:
3737                 case IFLA_TARGET_NETNSID:
3738                         break;
3739                 default:
3740                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3741                         return -EINVAL;
3742                 }
3743         }
3744
3745         return 0;
3746 }
3747
3748 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3749                         struct netlink_ext_ack *extack)
3750 {
3751         struct net *net = sock_net(skb->sk);
3752         struct net *tgt_net = net;
3753         struct ifinfomsg *ifm;
3754         struct nlattr *tb[IFLA_MAX+1];
3755         struct net_device *dev = NULL;
3756         struct sk_buff *nskb;
3757         int netnsid = -1;
3758         int err;
3759         u32 ext_filter_mask = 0;
3760
3761         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3762         if (err < 0)
3763                 return err;
3764
3765         err = rtnl_ensure_unique_netns(tb, extack, true);
3766         if (err < 0)
3767                 return err;
3768
3769         if (tb[IFLA_TARGET_NETNSID]) {
3770                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3771                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3772                 if (IS_ERR(tgt_net))
3773                         return PTR_ERR(tgt_net);
3774         }
3775
3776         if (tb[IFLA_EXT_MASK])
3777                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3778
3779         err = -EINVAL;
3780         ifm = nlmsg_data(nlh);
3781         if (ifm->ifi_index > 0)
3782                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3783         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3784                 dev = rtnl_dev_get(tgt_net, tb);
3785         else
3786                 goto out;
3787
3788         err = -ENODEV;
3789         if (dev == NULL)
3790                 goto out;
3791
3792         err = -ENOBUFS;
3793         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3794         if (nskb == NULL)
3795                 goto out;
3796
3797         err = rtnl_fill_ifinfo(nskb, dev, net,
3798                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3799                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3800                                0, NULL, 0, netnsid, GFP_KERNEL);
3801         if (err < 0) {
3802                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3803                 WARN_ON(err == -EMSGSIZE);
3804                 kfree_skb(nskb);
3805         } else
3806                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3807 out:
3808         if (netnsid >= 0)
3809                 put_net(tgt_net);
3810
3811         return err;
3812 }
3813
3814 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3815                            bool *changed, struct netlink_ext_ack *extack)
3816 {
3817         char *alt_ifname;
3818         size_t size;
3819         int err;
3820
3821         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3822         if (err)
3823                 return err;
3824
3825         if (cmd == RTM_NEWLINKPROP) {
3826                 size = rtnl_prop_list_size(dev);
3827                 size += nla_total_size(ALTIFNAMSIZ);
3828                 if (size >= U16_MAX) {
3829                         NL_SET_ERR_MSG(extack,
3830                                        "effective property list too long");
3831                         return -EINVAL;
3832                 }
3833         }
3834
3835         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3836         if (!alt_ifname)
3837                 return -ENOMEM;
3838
3839         if (cmd == RTM_NEWLINKPROP) {
3840                 err = netdev_name_node_alt_create(dev, alt_ifname);
3841                 if (!err)
3842                         alt_ifname = NULL;
3843         } else if (cmd == RTM_DELLINKPROP) {
3844                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3845         } else {
3846                 WARN_ON_ONCE(1);
3847                 err = -EINVAL;
3848         }
3849
3850         kfree(alt_ifname);
3851         if (!err)
3852                 *changed = true;
3853         return err;
3854 }
3855
3856 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3857                          struct netlink_ext_ack *extack)
3858 {
3859         struct net *net = sock_net(skb->sk);
3860         struct nlattr *tb[IFLA_MAX + 1];
3861         struct net_device *dev;
3862         struct ifinfomsg *ifm;
3863         bool changed = false;
3864         struct nlattr *attr;
3865         int err, rem;
3866
3867         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3868         if (err)
3869                 return err;
3870
3871         err = rtnl_ensure_unique_netns(tb, extack, true);
3872         if (err)
3873                 return err;
3874
3875         ifm = nlmsg_data(nlh);
3876         if (ifm->ifi_index > 0)
3877                 dev = __dev_get_by_index(net, ifm->ifi_index);
3878         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3879                 dev = rtnl_dev_get(net, tb);
3880         else
3881                 return -EINVAL;
3882
3883         if (!dev)
3884                 return -ENODEV;
3885
3886         if (!tb[IFLA_PROP_LIST])
3887                 return 0;
3888
3889         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3890                 switch (nla_type(attr)) {
3891                 case IFLA_ALT_IFNAME:
3892                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3893                         if (err)
3894                                 return err;
3895                         break;
3896                 }
3897         }
3898
3899         if (changed)
3900                 netdev_state_change(dev);
3901         return 0;
3902 }
3903
3904 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3905                             struct netlink_ext_ack *extack)
3906 {
3907         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3908 }
3909
3910 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3911                             struct netlink_ext_ack *extack)
3912 {
3913         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3914 }
3915
3916 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3917 {
3918         struct net *net = sock_net(skb->sk);
3919         size_t min_ifinfo_dump_size = 0;
3920         struct nlattr *tb[IFLA_MAX+1];
3921         u32 ext_filter_mask = 0;
3922         struct net_device *dev;
3923         int hdrlen;
3924
3925         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3926         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3927                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3928
3929         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3930                 if (tb[IFLA_EXT_MASK])
3931                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3932         }
3933
3934         if (!ext_filter_mask)
3935                 return NLMSG_GOODSIZE;
3936         /*
3937          * traverse the list of net devices and compute the minimum
3938          * buffer size based upon the filter mask.
3939          */
3940         rcu_read_lock();
3941         for_each_netdev_rcu(net, dev) {
3942                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3943                                            if_nlmsg_size(dev, ext_filter_mask));
3944         }
3945         rcu_read_unlock();
3946
3947         return nlmsg_total_size(min_ifinfo_dump_size);
3948 }
3949
3950 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3951 {
3952         int idx;
3953         int s_idx = cb->family;
3954         int type = cb->nlh->nlmsg_type - RTM_BASE;
3955         int ret = 0;
3956
3957         if (s_idx == 0)
3958                 s_idx = 1;
3959
3960         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3961                 struct rtnl_link __rcu **tab;
3962                 struct rtnl_link *link;
3963                 rtnl_dumpit_func dumpit;
3964
3965                 if (idx < s_idx || idx == PF_PACKET)
3966                         continue;
3967
3968                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3969                         continue;
3970
3971                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3972                 if (!tab)
3973                         continue;
3974
3975                 link = rcu_dereference_rtnl(tab[type]);
3976                 if (!link)
3977                         continue;
3978
3979                 dumpit = link->dumpit;
3980                 if (!dumpit)
3981                         continue;
3982
3983                 if (idx > s_idx) {
3984                         memset(&cb->args[0], 0, sizeof(cb->args));
3985                         cb->prev_seq = 0;
3986                         cb->seq = 0;
3987                 }
3988                 ret = dumpit(skb, cb);
3989                 if (ret)
3990                         break;
3991         }
3992         cb->family = idx;
3993
3994         return skb->len ? : ret;
3995 }
3996
3997 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3998                                        unsigned int change,
3999                                        u32 event, gfp_t flags, int *new_nsid,
4000                                        int new_ifindex, u32 portid,
4001                                        const struct nlmsghdr *nlh)
4002 {
4003         struct net *net = dev_net(dev);
4004         struct sk_buff *skb;
4005         int err = -ENOBUFS;
4006         u32 seq = 0;
4007
4008         skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
4009         if (skb == NULL)
4010                 goto errout;
4011
4012         if (nlmsg_report(nlh))
4013                 seq = nlmsg_seq(nlh);
4014         else
4015                 portid = 0;
4016
4017         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
4018                                type, portid, seq, change, 0, 0, event,
4019                                new_nsid, new_ifindex, -1, flags);
4020         if (err < 0) {
4021                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
4022                 WARN_ON(err == -EMSGSIZE);
4023                 kfree_skb(skb);
4024                 goto errout;
4025         }
4026         return skb;
4027 errout:
4028         if (err < 0)
4029                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4030         return NULL;
4031 }
4032
4033 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
4034                        u32 portid, const struct nlmsghdr *nlh)
4035 {
4036         struct net *net = dev_net(dev);
4037
4038         rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
4039 }
4040
4041 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
4042                                unsigned int change, u32 event,
4043                                gfp_t flags, int *new_nsid, int new_ifindex,
4044                                u32 portid, const struct nlmsghdr *nlh)
4045 {
4046         struct sk_buff *skb;
4047
4048         if (dev->reg_state != NETREG_REGISTERED)
4049                 return;
4050
4051         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
4052                                      new_ifindex, portid, nlh);
4053         if (skb)
4054                 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
4055 }
4056
4057 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
4058                   gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
4059 {
4060         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4061                            NULL, 0, portid, nlh);
4062 }
4063
4064 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4065                          gfp_t flags, int *new_nsid, int new_ifindex)
4066 {
4067         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4068                            new_nsid, new_ifindex, 0, NULL);
4069 }
4070
4071 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4072                                    struct net_device *dev,
4073                                    u8 *addr, u16 vid, u32 pid, u32 seq,
4074                                    int type, unsigned int flags,
4075                                    int nlflags, u16 ndm_state)
4076 {
4077         struct nlmsghdr *nlh;
4078         struct ndmsg *ndm;
4079
4080         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4081         if (!nlh)
4082                 return -EMSGSIZE;
4083
4084         ndm = nlmsg_data(nlh);
4085         ndm->ndm_family  = AF_BRIDGE;
4086         ndm->ndm_pad1    = 0;
4087         ndm->ndm_pad2    = 0;
4088         ndm->ndm_flags   = flags;
4089         ndm->ndm_type    = 0;
4090         ndm->ndm_ifindex = dev->ifindex;
4091         ndm->ndm_state   = ndm_state;
4092
4093         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
4094                 goto nla_put_failure;
4095         if (vid)
4096                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4097                         goto nla_put_failure;
4098
4099         nlmsg_end(skb, nlh);
4100         return 0;
4101
4102 nla_put_failure:
4103         nlmsg_cancel(skb, nlh);
4104         return -EMSGSIZE;
4105 }
4106
4107 static inline size_t rtnl_fdb_nlmsg_size(void)
4108 {
4109         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4110                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
4111                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
4112                0;
4113 }
4114
4115 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4116                             u16 ndm_state)
4117 {
4118         struct net *net = dev_net(dev);
4119         struct sk_buff *skb;
4120         int err = -ENOBUFS;
4121
4122         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4123         if (!skb)
4124                 goto errout;
4125
4126         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4127                                       0, 0, type, NTF_SELF, 0, ndm_state);
4128         if (err < 0) {
4129                 kfree_skb(skb);
4130                 goto errout;
4131         }
4132
4133         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4134         return;
4135 errout:
4136         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4137 }
4138
4139 /*
4140  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4141  */
4142 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4143                      struct nlattr *tb[],
4144                      struct net_device *dev,
4145                      const unsigned char *addr, u16 vid,
4146                      u16 flags)
4147 {
4148         int err = -EINVAL;
4149
4150         /* If aging addresses are supported device will need to
4151          * implement its own handler for this.
4152          */
4153         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4154                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4155                 return err;
4156         }
4157
4158         if (tb[NDA_FLAGS_EXT]) {
4159                 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4160                 return err;
4161         }
4162
4163         if (vid) {
4164                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4165                 return err;
4166         }
4167
4168         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4169                 err = dev_uc_add_excl(dev, addr);
4170         else if (is_multicast_ether_addr(addr))
4171                 err = dev_mc_add_excl(dev, addr);
4172
4173         /* Only return duplicate errors if NLM_F_EXCL is set */
4174         if (err == -EEXIST && !(flags & NLM_F_EXCL))
4175                 err = 0;
4176
4177         return err;
4178 }
4179 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4180
4181 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4182                          struct netlink_ext_ack *extack)
4183 {
4184         u16 vid = 0;
4185
4186         if (vlan_attr) {
4187                 if (nla_len(vlan_attr) != sizeof(u16)) {
4188                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4189                         return -EINVAL;
4190                 }
4191
4192                 vid = nla_get_u16(vlan_attr);
4193
4194                 if (!vid || vid >= VLAN_VID_MASK) {
4195                         NL_SET_ERR_MSG(extack, "invalid vlan id");
4196                         return -EINVAL;
4197                 }
4198         }
4199         *p_vid = vid;
4200         return 0;
4201 }
4202
4203 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4204                         struct netlink_ext_ack *extack)
4205 {
4206         struct net *net = sock_net(skb->sk);
4207         struct ndmsg *ndm;
4208         struct nlattr *tb[NDA_MAX+1];
4209         struct net_device *dev;
4210         u8 *addr;
4211         u16 vid;
4212         int err;
4213
4214         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4215                                      extack);
4216         if (err < 0)
4217                 return err;
4218
4219         ndm = nlmsg_data(nlh);
4220         if (ndm->ndm_ifindex == 0) {
4221                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4222                 return -EINVAL;
4223         }
4224
4225         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4226         if (dev == NULL) {
4227                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4228                 return -ENODEV;
4229         }
4230
4231         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4232                 NL_SET_ERR_MSG(extack, "invalid address");
4233                 return -EINVAL;
4234         }
4235
4236         if (dev->type != ARPHRD_ETHER) {
4237                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4238                 return -EINVAL;
4239         }
4240
4241         addr = nla_data(tb[NDA_LLADDR]);
4242
4243         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4244         if (err)
4245                 return err;
4246
4247         err = -EOPNOTSUPP;
4248
4249         /* Support fdb on master device the net/bridge default case */
4250         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4251             netif_is_bridge_port(dev)) {
4252                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4253                 const struct net_device_ops *ops = br_dev->netdev_ops;
4254
4255                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4256                                        nlh->nlmsg_flags, extack);
4257                 if (err)
4258                         goto out;
4259                 else
4260                         ndm->ndm_flags &= ~NTF_MASTER;
4261         }
4262
4263         /* Embedded bridge, macvlan, and any other device support */
4264         if ((ndm->ndm_flags & NTF_SELF)) {
4265                 if (dev->netdev_ops->ndo_fdb_add)
4266                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4267                                                            vid,
4268                                                            nlh->nlmsg_flags,
4269                                                            extack);
4270                 else
4271                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4272                                                nlh->nlmsg_flags);
4273
4274                 if (!err) {
4275                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4276                                         ndm->ndm_state);
4277                         ndm->ndm_flags &= ~NTF_SELF;
4278                 }
4279         }
4280 out:
4281         return err;
4282 }
4283
4284 /*
4285  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4286  */
4287 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4288                      struct nlattr *tb[],
4289                      struct net_device *dev,
4290                      const unsigned char *addr, u16 vid)
4291 {
4292         int err = -EINVAL;
4293
4294         /* If aging addresses are supported device will need to
4295          * implement its own handler for this.
4296          */
4297         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4298                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4299                 return err;
4300         }
4301
4302         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4303                 err = dev_uc_del(dev, addr);
4304         else if (is_multicast_ether_addr(addr))
4305                 err = dev_mc_del(dev, addr);
4306
4307         return err;
4308 }
4309 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4310
4311 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4312         [NDA_VLAN]      = { .type = NLA_U16 },
4313         [NDA_IFINDEX]   = NLA_POLICY_MIN(NLA_S32, 1),
4314         [NDA_NDM_STATE_MASK]    = { .type = NLA_U16  },
4315         [NDA_NDM_FLAGS_MASK]    = { .type = NLA_U8 },
4316 };
4317
4318 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4319                         struct netlink_ext_ack *extack)
4320 {
4321         bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4322         struct net *net = sock_net(skb->sk);
4323         const struct net_device_ops *ops;
4324         struct ndmsg *ndm;
4325         struct nlattr *tb[NDA_MAX+1];
4326         struct net_device *dev;
4327         __u8 *addr = NULL;
4328         int err;
4329         u16 vid;
4330
4331         if (!netlink_capable(skb, CAP_NET_ADMIN))
4332                 return -EPERM;
4333
4334         if (!del_bulk) {
4335                 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4336                                              NULL, extack);
4337         } else {
4338                 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4339                                   fdb_del_bulk_policy, extack);
4340         }
4341         if (err < 0)
4342                 return err;
4343
4344         ndm = nlmsg_data(nlh);
4345         if (ndm->ndm_ifindex == 0) {
4346                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4347                 return -EINVAL;
4348         }
4349
4350         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4351         if (dev == NULL) {
4352                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4353                 return -ENODEV;
4354         }
4355
4356         if (!del_bulk) {
4357                 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4358                         NL_SET_ERR_MSG(extack, "invalid address");
4359                         return -EINVAL;
4360                 }
4361                 addr = nla_data(tb[NDA_LLADDR]);
4362         }
4363
4364         if (dev->type != ARPHRD_ETHER) {
4365                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4366                 return -EINVAL;
4367         }
4368
4369         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4370         if (err)
4371                 return err;
4372
4373         err = -EOPNOTSUPP;
4374
4375         /* Support fdb on master device the net/bridge default case */
4376         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4377             netif_is_bridge_port(dev)) {
4378                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4379
4380                 ops = br_dev->netdev_ops;
4381                 if (!del_bulk) {
4382                         if (ops->ndo_fdb_del)
4383                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4384                 } else {
4385                         if (ops->ndo_fdb_del_bulk)
4386                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4387                                                             extack);
4388                 }
4389
4390                 if (err)
4391                         goto out;
4392                 else
4393                         ndm->ndm_flags &= ~NTF_MASTER;
4394         }
4395
4396         /* Embedded bridge, macvlan, and any other device support */
4397         if (ndm->ndm_flags & NTF_SELF) {
4398                 ops = dev->netdev_ops;
4399                 if (!del_bulk) {
4400                         if (ops->ndo_fdb_del)
4401                                 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4402                         else
4403                                 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4404                 } else {
4405                         /* in case err was cleared by NTF_MASTER call */
4406                         err = -EOPNOTSUPP;
4407                         if (ops->ndo_fdb_del_bulk)
4408                                 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4409                                                             extack);
4410                 }
4411
4412                 if (!err) {
4413                         if (!del_bulk)
4414                                 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4415                                                 ndm->ndm_state);
4416                         ndm->ndm_flags &= ~NTF_SELF;
4417                 }
4418         }
4419 out:
4420         return err;
4421 }
4422
4423 static int nlmsg_populate_fdb(struct sk_buff *skb,
4424                               struct netlink_callback *cb,
4425                               struct net_device *dev,
4426                               int *idx,
4427                               struct netdev_hw_addr_list *list)
4428 {
4429         struct netdev_hw_addr *ha;
4430         int err;
4431         u32 portid, seq;
4432
4433         portid = NETLINK_CB(cb->skb).portid;
4434         seq = cb->nlh->nlmsg_seq;
4435
4436         list_for_each_entry(ha, &list->list, list) {
4437                 if (*idx < cb->args[2])
4438                         goto skip;
4439
4440                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4441                                               portid, seq,
4442                                               RTM_NEWNEIGH, NTF_SELF,
4443                                               NLM_F_MULTI, NUD_PERMANENT);
4444                 if (err < 0)
4445                         return err;
4446 skip:
4447                 *idx += 1;
4448         }
4449         return 0;
4450 }
4451
4452 /**
4453  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4454  * @skb: socket buffer to store message in
4455  * @cb: netlink callback
4456  * @dev: netdevice
4457  * @filter_dev: ignored
4458  * @idx: the number of FDB table entries dumped is added to *@idx
4459  *
4460  * Default netdevice operation to dump the existing unicast address list.
4461  * Returns number of addresses from list put in skb.
4462  */
4463 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4464                       struct netlink_callback *cb,
4465                       struct net_device *dev,
4466                       struct net_device *filter_dev,
4467                       int *idx)
4468 {
4469         int err;
4470
4471         if (dev->type != ARPHRD_ETHER)
4472                 return -EINVAL;
4473
4474         netif_addr_lock_bh(dev);
4475         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4476         if (err)
4477                 goto out;
4478         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4479 out:
4480         netif_addr_unlock_bh(dev);
4481         return err;
4482 }
4483 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4484
4485 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4486                                  int *br_idx, int *brport_idx,
4487                                  struct netlink_ext_ack *extack)
4488 {
4489         struct nlattr *tb[NDA_MAX + 1];
4490         struct ndmsg *ndm;
4491         int err, i;
4492
4493         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4494                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4495                 return -EINVAL;
4496         }
4497
4498         ndm = nlmsg_data(nlh);
4499         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4500             ndm->ndm_flags || ndm->ndm_type) {
4501                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4502                 return -EINVAL;
4503         }
4504
4505         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4506                                             NDA_MAX, NULL, extack);
4507         if (err < 0)
4508                 return err;
4509
4510         *brport_idx = ndm->ndm_ifindex;
4511         for (i = 0; i <= NDA_MAX; ++i) {
4512                 if (!tb[i])
4513                         continue;
4514
4515                 switch (i) {
4516                 case NDA_IFINDEX:
4517                         if (nla_len(tb[i]) != sizeof(u32)) {
4518                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4519                                 return -EINVAL;
4520                         }
4521                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4522                         break;
4523                 case NDA_MASTER:
4524                         if (nla_len(tb[i]) != sizeof(u32)) {
4525                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4526                                 return -EINVAL;
4527                         }
4528                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4529                         break;
4530                 default:
4531                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4532                         return -EINVAL;
4533                 }
4534         }
4535
4536         return 0;
4537 }
4538
4539 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4540                                  int *br_idx, int *brport_idx,
4541                                  struct netlink_ext_ack *extack)
4542 {
4543         struct nlattr *tb[IFLA_MAX+1];
4544         int err;
4545
4546         /* A hack to preserve kernel<->userspace interface.
4547          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4548          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4549          * So, check for ndmsg with an optional u32 attribute (not used here).
4550          * Fortunately these sizes don't conflict with the size of ifinfomsg
4551          * with an optional attribute.
4552          */
4553         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4554             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4555              nla_attr_size(sizeof(u32)))) {
4556                 struct ifinfomsg *ifm;
4557
4558                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4559                                              tb, IFLA_MAX, ifla_policy,
4560                                              extack);
4561                 if (err < 0) {
4562                         return -EINVAL;
4563                 } else if (err == 0) {
4564                         if (tb[IFLA_MASTER])
4565                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4566                 }
4567
4568                 ifm = nlmsg_data(nlh);
4569                 *brport_idx = ifm->ifi_index;
4570         }
4571         return 0;
4572 }
4573
4574 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4575 {
4576         struct net_device *dev;
4577         struct net_device *br_dev = NULL;
4578         const struct net_device_ops *ops = NULL;
4579         const struct net_device_ops *cops = NULL;
4580         struct net *net = sock_net(skb->sk);
4581         struct hlist_head *head;
4582         int brport_idx = 0;
4583         int br_idx = 0;
4584         int h, s_h;
4585         int idx = 0, s_idx;
4586         int err = 0;
4587         int fidx = 0;
4588
4589         if (cb->strict_check)
4590                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4591                                             cb->extack);
4592         else
4593                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4594                                             cb->extack);
4595         if (err < 0)
4596                 return err;
4597
4598         if (br_idx) {
4599                 br_dev = __dev_get_by_index(net, br_idx);
4600                 if (!br_dev)
4601                         return -ENODEV;
4602
4603                 ops = br_dev->netdev_ops;
4604         }
4605
4606         s_h = cb->args[0];
4607         s_idx = cb->args[1];
4608
4609         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4610                 idx = 0;
4611                 head = &net->dev_index_head[h];
4612                 hlist_for_each_entry(dev, head, index_hlist) {
4613
4614                         if (brport_idx && (dev->ifindex != brport_idx))
4615                                 continue;
4616
4617                         if (!br_idx) { /* user did not specify a specific bridge */
4618                                 if (netif_is_bridge_port(dev)) {
4619                                         br_dev = netdev_master_upper_dev_get(dev);
4620                                         cops = br_dev->netdev_ops;
4621                                 }
4622                         } else {
4623                                 if (dev != br_dev &&
4624                                     !netif_is_bridge_port(dev))
4625                                         continue;
4626
4627                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4628                                     !netif_is_bridge_master(dev))
4629                                         continue;
4630                                 cops = ops;
4631                         }
4632
4633                         if (idx < s_idx)
4634                                 goto cont;
4635
4636                         if (netif_is_bridge_port(dev)) {
4637                                 if (cops && cops->ndo_fdb_dump) {
4638                                         err = cops->ndo_fdb_dump(skb, cb,
4639                                                                 br_dev, dev,
4640                                                                 &fidx);
4641                                         if (err == -EMSGSIZE)
4642                                                 goto out;
4643                                 }
4644                         }
4645
4646                         if (dev->netdev_ops->ndo_fdb_dump)
4647                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4648                                                                     dev, NULL,
4649                                                                     &fidx);
4650                         else
4651                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4652                                                         &fidx);
4653                         if (err == -EMSGSIZE)
4654                                 goto out;
4655
4656                         cops = NULL;
4657
4658                         /* reset fdb offset to 0 for rest of the interfaces */
4659                         cb->args[2] = 0;
4660                         fidx = 0;
4661 cont:
4662                         idx++;
4663                 }
4664         }
4665
4666 out:
4667         cb->args[0] = h;
4668         cb->args[1] = idx;
4669         cb->args[2] = fidx;
4670
4671         return skb->len;
4672 }
4673
4674 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4675                                 struct nlattr **tb, u8 *ndm_flags,
4676                                 int *br_idx, int *brport_idx, u8 **addr,
4677                                 u16 *vid, struct netlink_ext_ack *extack)
4678 {
4679         struct ndmsg *ndm;
4680         int err, i;
4681
4682         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4683                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4684                 return -EINVAL;
4685         }
4686
4687         ndm = nlmsg_data(nlh);
4688         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4689             ndm->ndm_type) {
4690                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4691                 return -EINVAL;
4692         }
4693
4694         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4695                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4696                 return -EINVAL;
4697         }
4698
4699         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4700                                             NDA_MAX, nda_policy, extack);
4701         if (err < 0)
4702                 return err;
4703
4704         *ndm_flags = ndm->ndm_flags;
4705         *brport_idx = ndm->ndm_ifindex;
4706         for (i = 0; i <= NDA_MAX; ++i) {
4707                 if (!tb[i])
4708                         continue;
4709
4710                 switch (i) {
4711                 case NDA_MASTER:
4712                         *br_idx = nla_get_u32(tb[i]);
4713                         break;
4714                 case NDA_LLADDR:
4715                         if (nla_len(tb[i]) != ETH_ALEN) {
4716                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4717                                 return -EINVAL;
4718                         }
4719                         *addr = nla_data(tb[i]);
4720                         break;
4721                 case NDA_VLAN:
4722                         err = fdb_vid_parse(tb[i], vid, extack);
4723                         if (err)
4724                                 return err;
4725                         break;
4726                 case NDA_VNI:
4727                         break;
4728                 default:
4729                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4730                         return -EINVAL;
4731                 }
4732         }
4733
4734         return 0;
4735 }
4736
4737 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4738                         struct netlink_ext_ack *extack)
4739 {
4740         struct net_device *dev = NULL, *br_dev = NULL;
4741         const struct net_device_ops *ops = NULL;
4742         struct net *net = sock_net(in_skb->sk);
4743         struct nlattr *tb[NDA_MAX + 1];
4744         struct sk_buff *skb;
4745         int brport_idx = 0;
4746         u8 ndm_flags = 0;
4747         int br_idx = 0;
4748         u8 *addr = NULL;
4749         u16 vid = 0;
4750         int err;
4751
4752         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4753                                    &brport_idx, &addr, &vid, extack);
4754         if (err < 0)
4755                 return err;
4756
4757         if (!addr) {
4758                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4759                 return -EINVAL;
4760         }
4761
4762         if (brport_idx) {
4763                 dev = __dev_get_by_index(net, brport_idx);
4764                 if (!dev) {
4765                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4766                         return -ENODEV;
4767                 }
4768         }
4769
4770         if (br_idx) {
4771                 if (dev) {
4772                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4773                         return -EINVAL;
4774                 }
4775
4776                 br_dev = __dev_get_by_index(net, br_idx);
4777                 if (!br_dev) {
4778                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4779                         return -EINVAL;
4780                 }
4781                 ops = br_dev->netdev_ops;
4782         }
4783
4784         if (dev) {
4785                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4786                         if (!netif_is_bridge_port(dev)) {
4787                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4788                                 return -EINVAL;
4789                         }
4790                         br_dev = netdev_master_upper_dev_get(dev);
4791                         if (!br_dev) {
4792                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4793                                 return -EINVAL;
4794                         }
4795                         ops = br_dev->netdev_ops;
4796                 } else {
4797                         if (!(ndm_flags & NTF_SELF)) {
4798                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4799                                 return -EINVAL;
4800                         }
4801                         ops = dev->netdev_ops;
4802                 }
4803         }
4804
4805         if (!br_dev && !dev) {
4806                 NL_SET_ERR_MSG(extack, "No device specified");
4807                 return -ENODEV;
4808         }
4809
4810         if (!ops || !ops->ndo_fdb_get) {
4811                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4812                 return -EOPNOTSUPP;
4813         }
4814
4815         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4816         if (!skb)
4817                 return -ENOBUFS;
4818
4819         if (br_dev)
4820                 dev = br_dev;
4821         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4822                                NETLINK_CB(in_skb).portid,
4823                                nlh->nlmsg_seq, extack);
4824         if (err)
4825                 goto out;
4826
4827         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4828 out:
4829         kfree_skb(skb);
4830         return err;
4831 }
4832
4833 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4834                                unsigned int attrnum, unsigned int flag)
4835 {
4836         if (mask & flag)
4837                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4838         return 0;
4839 }
4840
4841 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4842                             struct net_device *dev, u16 mode,
4843                             u32 flags, u32 mask, int nlflags,
4844                             u32 filter_mask,
4845                             int (*vlan_fill)(struct sk_buff *skb,
4846                                              struct net_device *dev,
4847                                              u32 filter_mask))
4848 {
4849         struct nlmsghdr *nlh;
4850         struct ifinfomsg *ifm;
4851         struct nlattr *br_afspec;
4852         struct nlattr *protinfo;
4853         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4854         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4855         int err = 0;
4856
4857         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4858         if (nlh == NULL)
4859                 return -EMSGSIZE;
4860
4861         ifm = nlmsg_data(nlh);
4862         ifm->ifi_family = AF_BRIDGE;
4863         ifm->__ifi_pad = 0;
4864         ifm->ifi_type = dev->type;
4865         ifm->ifi_index = dev->ifindex;
4866         ifm->ifi_flags = dev_get_flags(dev);
4867         ifm->ifi_change = 0;
4868
4869
4870         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4871             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4872             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4873             (br_dev &&
4874              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4875             (dev->addr_len &&
4876              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4877             (dev->ifindex != dev_get_iflink(dev) &&
4878              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4879                 goto nla_put_failure;
4880
4881         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4882         if (!br_afspec)
4883                 goto nla_put_failure;
4884
4885         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4886                 nla_nest_cancel(skb, br_afspec);
4887                 goto nla_put_failure;
4888         }
4889
4890         if (mode != BRIDGE_MODE_UNDEF) {
4891                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4892                         nla_nest_cancel(skb, br_afspec);
4893                         goto nla_put_failure;
4894                 }
4895         }
4896         if (vlan_fill) {
4897                 err = vlan_fill(skb, dev, filter_mask);
4898                 if (err) {
4899                         nla_nest_cancel(skb, br_afspec);
4900                         goto nla_put_failure;
4901                 }
4902         }
4903         nla_nest_end(skb, br_afspec);
4904
4905         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4906         if (!protinfo)
4907                 goto nla_put_failure;
4908
4909         if (brport_nla_put_flag(skb, flags, mask,
4910                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4911             brport_nla_put_flag(skb, flags, mask,
4912                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4913             brport_nla_put_flag(skb, flags, mask,
4914                                 IFLA_BRPORT_FAST_LEAVE,
4915                                 BR_MULTICAST_FAST_LEAVE) ||
4916             brport_nla_put_flag(skb, flags, mask,
4917                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4918             brport_nla_put_flag(skb, flags, mask,
4919                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4920             brport_nla_put_flag(skb, flags, mask,
4921                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4922             brport_nla_put_flag(skb, flags, mask,
4923                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4924             brport_nla_put_flag(skb, flags, mask,
4925                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4926             brport_nla_put_flag(skb, flags, mask,
4927                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4928             brport_nla_put_flag(skb, flags, mask,
4929                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4930                 nla_nest_cancel(skb, protinfo);
4931                 goto nla_put_failure;
4932         }
4933
4934         nla_nest_end(skb, protinfo);
4935
4936         nlmsg_end(skb, nlh);
4937         return 0;
4938 nla_put_failure:
4939         nlmsg_cancel(skb, nlh);
4940         return err ? err : -EMSGSIZE;
4941 }
4942 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4943
4944 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4945                                     bool strict_check, u32 *filter_mask,
4946                                     struct netlink_ext_ack *extack)
4947 {
4948         struct nlattr *tb[IFLA_MAX+1];
4949         int err, i;
4950
4951         if (strict_check) {
4952                 struct ifinfomsg *ifm;
4953
4954                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4955                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4956                         return -EINVAL;
4957                 }
4958
4959                 ifm = nlmsg_data(nlh);
4960                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4961                     ifm->ifi_change || ifm->ifi_index) {
4962                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4963                         return -EINVAL;
4964                 }
4965
4966                 err = nlmsg_parse_deprecated_strict(nlh,
4967                                                     sizeof(struct ifinfomsg),
4968                                                     tb, IFLA_MAX, ifla_policy,
4969                                                     extack);
4970         } else {
4971                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4972                                              tb, IFLA_MAX, ifla_policy,
4973                                              extack);
4974         }
4975         if (err < 0)
4976                 return err;
4977
4978         /* new attributes should only be added with strict checking */
4979         for (i = 0; i <= IFLA_MAX; ++i) {
4980                 if (!tb[i])
4981                         continue;
4982
4983                 switch (i) {
4984                 case IFLA_EXT_MASK:
4985                         *filter_mask = nla_get_u32(tb[i]);
4986                         break;
4987                 default:
4988                         if (strict_check) {
4989                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4990                                 return -EINVAL;
4991                         }
4992                 }
4993         }
4994
4995         return 0;
4996 }
4997
4998 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4999 {
5000         const struct nlmsghdr *nlh = cb->nlh;
5001         struct net *net = sock_net(skb->sk);
5002         struct net_device *dev;
5003         int idx = 0;
5004         u32 portid = NETLINK_CB(cb->skb).portid;
5005         u32 seq = nlh->nlmsg_seq;
5006         u32 filter_mask = 0;
5007         int err;
5008
5009         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
5010                                        cb->extack);
5011         if (err < 0 && cb->strict_check)
5012                 return err;
5013
5014         rcu_read_lock();
5015         for_each_netdev_rcu(net, dev) {
5016                 const struct net_device_ops *ops = dev->netdev_ops;
5017                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5018
5019                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
5020                         if (idx >= cb->args[0]) {
5021                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
5022                                                 skb, portid, seq, dev,
5023                                                 filter_mask, NLM_F_MULTI);
5024                                 if (err < 0 && err != -EOPNOTSUPP) {
5025                                         if (likely(skb->len))
5026                                                 break;
5027
5028                                         goto out_err;
5029                                 }
5030                         }
5031                         idx++;
5032                 }
5033
5034                 if (ops->ndo_bridge_getlink) {
5035                         if (idx >= cb->args[0]) {
5036                                 err = ops->ndo_bridge_getlink(skb, portid,
5037                                                               seq, dev,
5038                                                               filter_mask,
5039                                                               NLM_F_MULTI);
5040                                 if (err < 0 && err != -EOPNOTSUPP) {
5041                                         if (likely(skb->len))
5042                                                 break;
5043
5044                                         goto out_err;
5045                                 }
5046                         }
5047                         idx++;
5048                 }
5049         }
5050         err = skb->len;
5051 out_err:
5052         rcu_read_unlock();
5053         cb->args[0] = idx;
5054
5055         return err;
5056 }
5057
5058 static inline size_t bridge_nlmsg_size(void)
5059 {
5060         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
5061                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
5062                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
5063                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
5064                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
5065                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
5066                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
5067                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
5068                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5069                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
5070                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
5071 }
5072
5073 static int rtnl_bridge_notify(struct net_device *dev)
5074 {
5075         struct net *net = dev_net(dev);
5076         struct sk_buff *skb;
5077         int err = -EOPNOTSUPP;
5078
5079         if (!dev->netdev_ops->ndo_bridge_getlink)
5080                 return 0;
5081
5082         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5083         if (!skb) {
5084                 err = -ENOMEM;
5085                 goto errout;
5086         }
5087
5088         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5089         if (err < 0)
5090                 goto errout;
5091
5092         /* Notification info is only filled for bridge ports, not the bridge
5093          * device itself. Therefore, a zero notification length is valid and
5094          * should not result in an error.
5095          */
5096         if (!skb->len)
5097                 goto errout;
5098
5099         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5100         return 0;
5101 errout:
5102         WARN_ON(err == -EMSGSIZE);
5103         kfree_skb(skb);
5104         if (err)
5105                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5106         return err;
5107 }
5108
5109 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5110                                struct netlink_ext_ack *extack)
5111 {
5112         struct net *net = sock_net(skb->sk);
5113         struct ifinfomsg *ifm;
5114         struct net_device *dev;
5115         struct nlattr *br_spec, *attr = NULL;
5116         int rem, err = -EOPNOTSUPP;
5117         u16 flags = 0;
5118         bool have_flags = false;
5119
5120         if (nlmsg_len(nlh) < sizeof(*ifm))
5121                 return -EINVAL;
5122
5123         ifm = nlmsg_data(nlh);
5124         if (ifm->ifi_family != AF_BRIDGE)
5125                 return -EPFNOSUPPORT;
5126
5127         dev = __dev_get_by_index(net, ifm->ifi_index);
5128         if (!dev) {
5129                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5130                 return -ENODEV;
5131         }
5132
5133         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5134         if (br_spec) {
5135                 nla_for_each_nested(attr, br_spec, rem) {
5136                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5137                                 if (nla_len(attr) < sizeof(flags))
5138                                         return -EINVAL;
5139
5140                                 have_flags = true;
5141                                 flags = nla_get_u16(attr);
5142                                 break;
5143                         }
5144                 }
5145         }
5146
5147         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5148                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5149
5150                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5151                         err = -EOPNOTSUPP;
5152                         goto out;
5153                 }
5154
5155                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5156                                                              extack);
5157                 if (err)
5158                         goto out;
5159
5160                 flags &= ~BRIDGE_FLAGS_MASTER;
5161         }
5162
5163         if ((flags & BRIDGE_FLAGS_SELF)) {
5164                 if (!dev->netdev_ops->ndo_bridge_setlink)
5165                         err = -EOPNOTSUPP;
5166                 else
5167                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5168                                                                   flags,
5169                                                                   extack);
5170                 if (!err) {
5171                         flags &= ~BRIDGE_FLAGS_SELF;
5172
5173                         /* Generate event to notify upper layer of bridge
5174                          * change
5175                          */
5176                         err = rtnl_bridge_notify(dev);
5177                 }
5178         }
5179
5180         if (have_flags)
5181                 memcpy(nla_data(attr), &flags, sizeof(flags));
5182 out:
5183         return err;
5184 }
5185
5186 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5187                                struct netlink_ext_ack *extack)
5188 {
5189         struct net *net = sock_net(skb->sk);
5190         struct ifinfomsg *ifm;
5191         struct net_device *dev;
5192         struct nlattr *br_spec, *attr = NULL;
5193         int rem, err = -EOPNOTSUPP;
5194         u16 flags = 0;
5195         bool have_flags = false;
5196
5197         if (nlmsg_len(nlh) < sizeof(*ifm))
5198                 return -EINVAL;
5199
5200         ifm = nlmsg_data(nlh);
5201         if (ifm->ifi_family != AF_BRIDGE)
5202                 return -EPFNOSUPPORT;
5203
5204         dev = __dev_get_by_index(net, ifm->ifi_index);
5205         if (!dev) {
5206                 NL_SET_ERR_MSG(extack, "unknown ifindex");
5207                 return -ENODEV;
5208         }
5209
5210         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5211         if (br_spec) {
5212                 nla_for_each_nested(attr, br_spec, rem) {
5213                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5214                                 if (nla_len(attr) < sizeof(flags))
5215                                         return -EINVAL;
5216
5217                                 have_flags = true;
5218                                 flags = nla_get_u16(attr);
5219                                 break;
5220                         }
5221                 }
5222         }
5223
5224         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5225                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5226
5227                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5228                         err = -EOPNOTSUPP;
5229                         goto out;
5230                 }
5231
5232                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5233                 if (err)
5234                         goto out;
5235
5236                 flags &= ~BRIDGE_FLAGS_MASTER;
5237         }
5238
5239         if ((flags & BRIDGE_FLAGS_SELF)) {
5240                 if (!dev->netdev_ops->ndo_bridge_dellink)
5241                         err = -EOPNOTSUPP;
5242                 else
5243                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5244                                                                   flags);
5245
5246                 if (!err) {
5247                         flags &= ~BRIDGE_FLAGS_SELF;
5248
5249                         /* Generate event to notify upper layer of bridge
5250                          * change
5251                          */
5252                         err = rtnl_bridge_notify(dev);
5253                 }
5254         }
5255
5256         if (have_flags)
5257                 memcpy(nla_data(attr), &flags, sizeof(flags));
5258 out:
5259         return err;
5260 }
5261
5262 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5263 {
5264         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5265                (!idxattr || idxattr == attrid);
5266 }
5267
5268 static bool
5269 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5270 {
5271         return dev->netdev_ops &&
5272                dev->netdev_ops->ndo_has_offload_stats &&
5273                dev->netdev_ops->ndo_get_offload_stats &&
5274                dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5275 }
5276
5277 static unsigned int
5278 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5279 {
5280         return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5281                sizeof(struct rtnl_link_stats64) : 0;
5282 }
5283
5284 static int
5285 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5286                              struct sk_buff *skb)
5287 {
5288         unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5289         struct nlattr *attr = NULL;
5290         void *attr_data;
5291         int err;
5292
5293         if (!size)
5294                 return -ENODATA;
5295
5296         attr = nla_reserve_64bit(skb, attr_id, size,
5297                                  IFLA_OFFLOAD_XSTATS_UNSPEC);
5298         if (!attr)
5299                 return -EMSGSIZE;
5300
5301         attr_data = nla_data(attr);
5302         memset(attr_data, 0, size);
5303
5304         err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5305         if (err)
5306                 return err;
5307
5308         return 0;
5309 }
5310
5311 static unsigned int
5312 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5313                                    enum netdev_offload_xstats_type type)
5314 {
5315         bool enabled = netdev_offload_xstats_enabled(dev, type);
5316
5317         return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5318 }
5319
5320 struct rtnl_offload_xstats_request_used {
5321         bool request;
5322         bool used;
5323 };
5324
5325 static int
5326 rtnl_offload_xstats_get_stats(struct net_device *dev,
5327                               enum netdev_offload_xstats_type type,
5328                               struct rtnl_offload_xstats_request_used *ru,
5329                               struct rtnl_hw_stats64 *stats,
5330                               struct netlink_ext_ack *extack)
5331 {
5332         bool request;
5333         bool used;
5334         int err;
5335
5336         request = netdev_offload_xstats_enabled(dev, type);
5337         if (!request) {
5338                 used = false;
5339                 goto out;
5340         }
5341
5342         err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5343         if (err)
5344                 return err;
5345
5346 out:
5347         if (ru) {
5348                 ru->request = request;
5349                 ru->used = used;
5350         }
5351         return 0;
5352 }
5353
5354 static int
5355 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5356                                        struct rtnl_offload_xstats_request_used *ru)
5357 {
5358         struct nlattr *nest;
5359
5360         nest = nla_nest_start(skb, attr_id);
5361         if (!nest)
5362                 return -EMSGSIZE;
5363
5364         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5365                 goto nla_put_failure;
5366
5367         if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5368                 goto nla_put_failure;
5369
5370         nla_nest_end(skb, nest);
5371         return 0;
5372
5373 nla_put_failure:
5374         nla_nest_cancel(skb, nest);
5375         return -EMSGSIZE;
5376 }
5377
5378 static int
5379 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5380                                    struct netlink_ext_ack *extack)
5381 {
5382         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5383         struct rtnl_offload_xstats_request_used ru_l3;
5384         struct nlattr *nest;
5385         int err;
5386
5387         err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5388         if (err)
5389                 return err;
5390
5391         nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5392         if (!nest)
5393                 return -EMSGSIZE;
5394
5395         if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5396                                                    IFLA_OFFLOAD_XSTATS_L3_STATS,
5397                                                    &ru_l3))
5398                 goto nla_put_failure;
5399
5400         nla_nest_end(skb, nest);
5401         return 0;
5402
5403 nla_put_failure:
5404         nla_nest_cancel(skb, nest);
5405         return -EMSGSIZE;
5406 }
5407
5408 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5409                                     int *prividx, u32 off_filter_mask,
5410                                     struct netlink_ext_ack *extack)
5411 {
5412         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5413         int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5414         int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5415         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5416         bool have_data = false;
5417         int err;
5418
5419         if (*prividx <= attr_id_cpu_hit &&
5420             (off_filter_mask &
5421              IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5422                 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5423                 if (!err) {
5424                         have_data = true;
5425                 } else if (err != -ENODATA) {
5426                         *prividx = attr_id_cpu_hit;
5427                         return err;
5428                 }
5429         }
5430
5431         if (*prividx <= attr_id_hw_s_info &&
5432             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5433                 *prividx = attr_id_hw_s_info;
5434
5435                 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5436                 if (err)
5437                         return err;
5438
5439                 have_data = true;
5440                 *prividx = 0;
5441         }
5442
5443         if (*prividx <= attr_id_l3_stats &&
5444             (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5445                 unsigned int size_l3;
5446                 struct nlattr *attr;
5447
5448                 *prividx = attr_id_l3_stats;
5449
5450                 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5451                 if (!size_l3)
5452                         goto skip_l3_stats;
5453                 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5454                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5455                 if (!attr)
5456                         return -EMSGSIZE;
5457
5458                 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5459                                                     nla_data(attr), extack);
5460                 if (err)
5461                         return err;
5462
5463                 have_data = true;
5464 skip_l3_stats:
5465                 *prividx = 0;
5466         }
5467
5468         if (!have_data)
5469                 return -ENODATA;
5470
5471         *prividx = 0;
5472         return 0;
5473 }
5474
5475 static unsigned int
5476 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5477                                            enum netdev_offload_xstats_type type)
5478 {
5479         bool enabled = netdev_offload_xstats_enabled(dev, type);
5480
5481         return nla_total_size(0) +
5482                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5483                 nla_total_size(sizeof(u8)) +
5484                 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5485                 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5486                 0;
5487 }
5488
5489 static unsigned int
5490 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5491 {
5492         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5493
5494         return nla_total_size(0) +
5495                 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5496                 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5497                 0;
5498 }
5499
5500 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5501                                         u32 off_filter_mask)
5502 {
5503         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5504         int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5505         int nla_size = 0;
5506         int size;
5507
5508         if (off_filter_mask &
5509             IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5510                 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5511                 nla_size += nla_total_size_64bit(size);
5512         }
5513
5514         if (off_filter_mask &
5515             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5516                 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5517
5518         if (off_filter_mask &
5519             IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5520                 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5521                 nla_size += nla_total_size_64bit(size);
5522         }
5523
5524         if (nla_size != 0)
5525                 nla_size += nla_total_size(0);
5526
5527         return nla_size;
5528 }
5529
5530 struct rtnl_stats_dump_filters {
5531         /* mask[0] filters outer attributes. Then individual nests have their
5532          * filtering mask at the index of the nested attribute.
5533          */
5534         u32 mask[IFLA_STATS_MAX + 1];
5535 };
5536
5537 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5538                                int type, u32 pid, u32 seq, u32 change,
5539                                unsigned int flags,
5540                                const struct rtnl_stats_dump_filters *filters,
5541                                int *idxattr, int *prividx,
5542                                struct netlink_ext_ack *extack)
5543 {
5544         unsigned int filter_mask = filters->mask[0];
5545         struct if_stats_msg *ifsm;
5546         struct nlmsghdr *nlh;
5547         struct nlattr *attr;
5548         int s_prividx = *prividx;
5549         int err;
5550
5551         ASSERT_RTNL();
5552
5553         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5554         if (!nlh)
5555                 return -EMSGSIZE;
5556
5557         ifsm = nlmsg_data(nlh);
5558         ifsm->family = PF_UNSPEC;
5559         ifsm->pad1 = 0;
5560         ifsm->pad2 = 0;
5561         ifsm->ifindex = dev->ifindex;
5562         ifsm->filter_mask = filter_mask;
5563
5564         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5565                 struct rtnl_link_stats64 *sp;
5566
5567                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5568                                          sizeof(struct rtnl_link_stats64),
5569                                          IFLA_STATS_UNSPEC);
5570                 if (!attr) {
5571                         err = -EMSGSIZE;
5572                         goto nla_put_failure;
5573                 }
5574
5575                 sp = nla_data(attr);
5576                 dev_get_stats(dev, sp);
5577         }
5578
5579         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5580                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5581
5582                 if (ops && ops->fill_linkxstats) {
5583                         *idxattr = IFLA_STATS_LINK_XSTATS;
5584                         attr = nla_nest_start_noflag(skb,
5585                                                      IFLA_STATS_LINK_XSTATS);
5586                         if (!attr) {
5587                                 err = -EMSGSIZE;
5588                                 goto nla_put_failure;
5589                         }
5590
5591                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5592                         nla_nest_end(skb, attr);
5593                         if (err)
5594                                 goto nla_put_failure;
5595                         *idxattr = 0;
5596                 }
5597         }
5598
5599         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5600                              *idxattr)) {
5601                 const struct rtnl_link_ops *ops = NULL;
5602                 const struct net_device *master;
5603
5604                 master = netdev_master_upper_dev_get(dev);
5605                 if (master)
5606                         ops = master->rtnl_link_ops;
5607                 if (ops && ops->fill_linkxstats) {
5608                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5609                         attr = nla_nest_start_noflag(skb,
5610                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5611                         if (!attr) {
5612                                 err = -EMSGSIZE;
5613                                 goto nla_put_failure;
5614                         }
5615
5616                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5617                         nla_nest_end(skb, attr);
5618                         if (err)
5619                                 goto nla_put_failure;
5620                         *idxattr = 0;
5621                 }
5622         }
5623
5624         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5625                              *idxattr)) {
5626                 u32 off_filter_mask;
5627
5628                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5629                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5630                 attr = nla_nest_start_noflag(skb,
5631                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5632                 if (!attr) {
5633                         err = -EMSGSIZE;
5634                         goto nla_put_failure;
5635                 }
5636
5637                 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5638                                                off_filter_mask, extack);
5639                 if (err == -ENODATA)
5640                         nla_nest_cancel(skb, attr);
5641                 else
5642                         nla_nest_end(skb, attr);
5643
5644                 if (err && err != -ENODATA)
5645                         goto nla_put_failure;
5646                 *idxattr = 0;
5647         }
5648
5649         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5650                 struct rtnl_af_ops *af_ops;
5651
5652                 *idxattr = IFLA_STATS_AF_SPEC;
5653                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5654                 if (!attr) {
5655                         err = -EMSGSIZE;
5656                         goto nla_put_failure;
5657                 }
5658
5659                 rcu_read_lock();
5660                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5661                         if (af_ops->fill_stats_af) {
5662                                 struct nlattr *af;
5663
5664                                 af = nla_nest_start_noflag(skb,
5665                                                            af_ops->family);
5666                                 if (!af) {
5667                                         rcu_read_unlock();
5668                                         err = -EMSGSIZE;
5669                                         goto nla_put_failure;
5670                                 }
5671                                 err = af_ops->fill_stats_af(skb, dev);
5672
5673                                 if (err == -ENODATA) {
5674                                         nla_nest_cancel(skb, af);
5675                                 } else if (err < 0) {
5676                                         rcu_read_unlock();
5677                                         goto nla_put_failure;
5678                                 }
5679
5680                                 nla_nest_end(skb, af);
5681                         }
5682                 }
5683                 rcu_read_unlock();
5684
5685                 nla_nest_end(skb, attr);
5686
5687                 *idxattr = 0;
5688         }
5689
5690         nlmsg_end(skb, nlh);
5691
5692         return 0;
5693
5694 nla_put_failure:
5695         /* not a multi message or no progress mean a real error */
5696         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5697                 nlmsg_cancel(skb, nlh);
5698         else
5699                 nlmsg_end(skb, nlh);
5700
5701         return err;
5702 }
5703
5704 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5705                                   const struct rtnl_stats_dump_filters *filters)
5706 {
5707         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5708         unsigned int filter_mask = filters->mask[0];
5709
5710         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5711                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5712
5713         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5714                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5715                 int attr = IFLA_STATS_LINK_XSTATS;
5716
5717                 if (ops && ops->get_linkxstats_size) {
5718                         size += nla_total_size(ops->get_linkxstats_size(dev,
5719                                                                         attr));
5720                         /* for IFLA_STATS_LINK_XSTATS */
5721                         size += nla_total_size(0);
5722                 }
5723         }
5724
5725         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5726                 struct net_device *_dev = (struct net_device *)dev;
5727                 const struct rtnl_link_ops *ops = NULL;
5728                 const struct net_device *master;
5729
5730                 /* netdev_master_upper_dev_get can't take const */
5731                 master = netdev_master_upper_dev_get(_dev);
5732                 if (master)
5733                         ops = master->rtnl_link_ops;
5734                 if (ops && ops->get_linkxstats_size) {
5735                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5736
5737                         size += nla_total_size(ops->get_linkxstats_size(dev,
5738                                                                         attr));
5739                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5740                         size += nla_total_size(0);
5741                 }
5742         }
5743
5744         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5745                 u32 off_filter_mask;
5746
5747                 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5748                 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5749         }
5750
5751         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5752                 struct rtnl_af_ops *af_ops;
5753
5754                 /* for IFLA_STATS_AF_SPEC */
5755                 size += nla_total_size(0);
5756
5757                 rcu_read_lock();
5758                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5759                         if (af_ops->get_stats_af_size) {
5760                                 size += nla_total_size(
5761                                         af_ops->get_stats_af_size(dev));
5762
5763                                 /* for AF_* */
5764                                 size += nla_total_size(0);
5765                         }
5766                 }
5767                 rcu_read_unlock();
5768         }
5769
5770         return size;
5771 }
5772
5773 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5774
5775 static const struct nla_policy
5776 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5777         [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5778                     NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5779 };
5780
5781 static const struct nla_policy
5782 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5783         [IFLA_STATS_GET_FILTERS] =
5784                     NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5785 };
5786
5787 static const struct nla_policy
5788 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5789         [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5790 };
5791
5792 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5793                                         struct rtnl_stats_dump_filters *filters,
5794                                         struct netlink_ext_ack *extack)
5795 {
5796         struct nlattr *tb[IFLA_STATS_MAX + 1];
5797         int err;
5798         int at;
5799
5800         err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5801                                rtnl_stats_get_policy_filters, extack);
5802         if (err < 0)
5803                 return err;
5804
5805         for (at = 1; at <= IFLA_STATS_MAX; at++) {
5806                 if (tb[at]) {
5807                         if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5808                                 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5809                                 return -EINVAL;
5810                         }
5811                         filters->mask[at] = nla_get_u32(tb[at]);
5812                 }
5813         }
5814
5815         return 0;
5816 }
5817
5818 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5819                                 u32 filter_mask,
5820                                 struct rtnl_stats_dump_filters *filters,
5821                                 struct netlink_ext_ack *extack)
5822 {
5823         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5824         int err;
5825         int i;
5826
5827         filters->mask[0] = filter_mask;
5828         for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5829                 filters->mask[i] = -1U;
5830
5831         err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5832                           IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5833         if (err < 0)
5834                 return err;
5835
5836         if (tb[IFLA_STATS_GET_FILTERS]) {
5837                 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5838                                                    filters, extack);
5839                 if (err)
5840                         return err;
5841         }
5842
5843         return 0;
5844 }
5845
5846 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5847                                 bool is_dump, struct netlink_ext_ack *extack)
5848 {
5849         struct if_stats_msg *ifsm;
5850
5851         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5852                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5853                 return -EINVAL;
5854         }
5855
5856         if (!strict_check)
5857                 return 0;
5858
5859         ifsm = nlmsg_data(nlh);
5860
5861         /* only requests using strict checks can pass data to influence
5862          * the dump. The legacy exception is filter_mask.
5863          */
5864         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5865                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5866                 return -EINVAL;
5867         }
5868         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5869                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5870                 return -EINVAL;
5871         }
5872
5873         return 0;
5874 }
5875
5876 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5877                           struct netlink_ext_ack *extack)
5878 {
5879         struct rtnl_stats_dump_filters filters;
5880         struct net *net = sock_net(skb->sk);
5881         struct net_device *dev = NULL;
5882         int idxattr = 0, prividx = 0;
5883         struct if_stats_msg *ifsm;
5884         struct sk_buff *nskb;
5885         int err;
5886
5887         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5888                                    false, extack);
5889         if (err)
5890                 return err;
5891
5892         ifsm = nlmsg_data(nlh);
5893         if (ifsm->ifindex > 0)
5894                 dev = __dev_get_by_index(net, ifsm->ifindex);
5895         else
5896                 return -EINVAL;
5897
5898         if (!dev)
5899                 return -ENODEV;
5900
5901         if (!ifsm->filter_mask) {
5902                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5903                 return -EINVAL;
5904         }
5905
5906         err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5907         if (err)
5908                 return err;
5909
5910         nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5911         if (!nskb)
5912                 return -ENOBUFS;
5913
5914         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5915                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5916                                   0, &filters, &idxattr, &prividx, extack);
5917         if (err < 0) {
5918                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5919                 WARN_ON(err == -EMSGSIZE);
5920                 kfree_skb(nskb);
5921         } else {
5922                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5923         }
5924
5925         return err;
5926 }
5927
5928 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5929 {
5930         struct netlink_ext_ack *extack = cb->extack;
5931         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5932         struct rtnl_stats_dump_filters filters;
5933         struct net *net = sock_net(skb->sk);
5934         unsigned int flags = NLM_F_MULTI;
5935         struct if_stats_msg *ifsm;
5936         struct hlist_head *head;
5937         struct net_device *dev;
5938         int idx = 0;
5939
5940         s_h = cb->args[0];
5941         s_idx = cb->args[1];
5942         s_idxattr = cb->args[2];
5943         s_prividx = cb->args[3];
5944
5945         cb->seq = net->dev_base_seq;
5946
5947         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5948         if (err)
5949                 return err;
5950
5951         ifsm = nlmsg_data(cb->nlh);
5952         if (!ifsm->filter_mask) {
5953                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5954                 return -EINVAL;
5955         }
5956
5957         err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5958                                    extack);
5959         if (err)
5960                 return err;
5961
5962         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5963                 idx = 0;
5964                 head = &net->dev_index_head[h];
5965                 hlist_for_each_entry(dev, head, index_hlist) {
5966                         if (idx < s_idx)
5967                                 goto cont;
5968                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5969                                                   NETLINK_CB(cb->skb).portid,
5970                                                   cb->nlh->nlmsg_seq, 0,
5971                                                   flags, &filters,
5972                                                   &s_idxattr, &s_prividx,
5973                                                   extack);
5974                         /* If we ran out of room on the first message,
5975                          * we're in trouble
5976                          */
5977                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5978
5979                         if (err < 0)
5980                                 goto out;
5981                         s_prividx = 0;
5982                         s_idxattr = 0;
5983                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5984 cont:
5985                         idx++;
5986                 }
5987         }
5988 out:
5989         cb->args[3] = s_prividx;
5990         cb->args[2] = s_idxattr;
5991         cb->args[1] = idx;
5992         cb->args[0] = h;
5993
5994         return skb->len;
5995 }
5996
5997 void rtnl_offload_xstats_notify(struct net_device *dev)
5998 {
5999         struct rtnl_stats_dump_filters response_filters = {};
6000         struct net *net = dev_net(dev);
6001         int idxattr = 0, prividx = 0;
6002         struct sk_buff *skb;
6003         int err = -ENOBUFS;
6004
6005         ASSERT_RTNL();
6006
6007         response_filters.mask[0] |=
6008                 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6009         response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6010                 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6011
6012         skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
6013                         GFP_KERNEL);
6014         if (!skb)
6015                 goto errout;
6016
6017         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
6018                                   &response_filters, &idxattr, &prividx, NULL);
6019         if (err < 0) {
6020                 kfree_skb(skb);
6021                 goto errout;
6022         }
6023
6024         rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
6025         return;
6026
6027 errout:
6028         rtnl_set_sk_err(net, RTNLGRP_STATS, err);
6029 }
6030 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
6031
6032 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
6033                           struct netlink_ext_ack *extack)
6034 {
6035         enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
6036         struct rtnl_stats_dump_filters response_filters = {};
6037         struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
6038         struct net *net = sock_net(skb->sk);
6039         struct net_device *dev = NULL;
6040         struct if_stats_msg *ifsm;
6041         bool notify = false;
6042         int err;
6043
6044         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
6045                                    false, extack);
6046         if (err)
6047                 return err;
6048
6049         ifsm = nlmsg_data(nlh);
6050         if (ifsm->family != AF_UNSPEC) {
6051                 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
6052                 return -EINVAL;
6053         }
6054
6055         if (ifsm->ifindex > 0)
6056                 dev = __dev_get_by_index(net, ifsm->ifindex);
6057         else
6058                 return -EINVAL;
6059
6060         if (!dev)
6061                 return -ENODEV;
6062
6063         if (ifsm->filter_mask) {
6064                 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6065                 return -EINVAL;
6066         }
6067
6068         err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6069                           ifla_stats_set_policy, extack);
6070         if (err < 0)
6071                 return err;
6072
6073         if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6074                 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6075
6076                 if (req)
6077                         err = netdev_offload_xstats_enable(dev, t_l3, extack);
6078                 else
6079                         err = netdev_offload_xstats_disable(dev, t_l3);
6080
6081                 if (!err)
6082                         notify = true;
6083                 else if (err != -EALREADY)
6084                         return err;
6085
6086                 response_filters.mask[0] |=
6087                         IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6088                 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6089                         IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6090         }
6091
6092         if (notify)
6093                 rtnl_offload_xstats_notify(dev);
6094
6095         return 0;
6096 }
6097
6098 static int rtnl_mdb_valid_dump_req(const struct nlmsghdr *nlh,
6099                                    struct netlink_ext_ack *extack)
6100 {
6101         struct br_port_msg *bpm;
6102
6103         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*bpm))) {
6104                 NL_SET_ERR_MSG(extack, "Invalid header for mdb dump request");
6105                 return -EINVAL;
6106         }
6107
6108         bpm = nlmsg_data(nlh);
6109         if (bpm->ifindex) {
6110                 NL_SET_ERR_MSG(extack, "Filtering by device index is not supported for mdb dump request");
6111                 return -EINVAL;
6112         }
6113         if (nlmsg_attrlen(nlh, sizeof(*bpm))) {
6114                 NL_SET_ERR_MSG(extack, "Invalid data after header in mdb dump request");
6115                 return -EINVAL;
6116         }
6117
6118         return 0;
6119 }
6120
6121 struct rtnl_mdb_dump_ctx {
6122         long idx;
6123 };
6124
6125 static int rtnl_mdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
6126 {
6127         struct rtnl_mdb_dump_ctx *ctx = (void *)cb->ctx;
6128         struct net *net = sock_net(skb->sk);
6129         struct net_device *dev;
6130         int idx, s_idx;
6131         int err;
6132
6133         NL_ASSERT_DUMP_CTX_FITS(struct rtnl_mdb_dump_ctx);
6134
6135         if (cb->strict_check) {
6136                 err = rtnl_mdb_valid_dump_req(cb->nlh, cb->extack);
6137                 if (err)
6138                         return err;
6139         }
6140
6141         s_idx = ctx->idx;
6142         idx = 0;
6143
6144         for_each_netdev(net, dev) {
6145                 if (idx < s_idx)
6146                         goto skip;
6147                 if (!dev->netdev_ops->ndo_mdb_dump)
6148                         goto skip;
6149
6150                 err = dev->netdev_ops->ndo_mdb_dump(dev, skb, cb);
6151                 if (err == -EMSGSIZE)
6152                         goto out;
6153                 /* Moving on to next device, reset markers and sequence
6154                  * counters since they are all maintained per-device.
6155                  */
6156                 memset(cb->ctx, 0, sizeof(cb->ctx));
6157                 cb->prev_seq = 0;
6158                 cb->seq = 0;
6159 skip:
6160                 idx++;
6161         }
6162
6163 out:
6164         ctx->idx = idx;
6165         return skb->len;
6166 }
6167
6168 static int rtnl_validate_mdb_entry(const struct nlattr *attr,
6169                                    struct netlink_ext_ack *extack)
6170 {
6171         struct br_mdb_entry *entry = nla_data(attr);
6172
6173         if (nla_len(attr) != sizeof(struct br_mdb_entry)) {
6174                 NL_SET_ERR_MSG_ATTR(extack, attr, "Invalid attribute length");
6175                 return -EINVAL;
6176         }
6177
6178         if (entry->ifindex == 0) {
6179                 NL_SET_ERR_MSG(extack, "Zero entry ifindex is not allowed");
6180                 return -EINVAL;
6181         }
6182
6183         if (entry->addr.proto == htons(ETH_P_IP)) {
6184                 if (!ipv4_is_multicast(entry->addr.u.ip4) &&
6185                     !ipv4_is_zeronet(entry->addr.u.ip4)) {
6186                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is not multicast or 0.0.0.0");
6187                         return -EINVAL;
6188                 }
6189                 if (ipv4_is_local_multicast(entry->addr.u.ip4)) {
6190                         NL_SET_ERR_MSG(extack, "IPv4 entry group address is local multicast");
6191                         return -EINVAL;
6192                 }
6193 #if IS_ENABLED(CONFIG_IPV6)
6194         } else if (entry->addr.proto == htons(ETH_P_IPV6)) {
6195                 if (ipv6_addr_is_ll_all_nodes(&entry->addr.u.ip6)) {
6196                         NL_SET_ERR_MSG(extack, "IPv6 entry group address is link-local all nodes");
6197                         return -EINVAL;
6198                 }
6199 #endif
6200         } else if (entry->addr.proto == 0) {
6201                 /* L2 mdb */
6202                 if (!is_multicast_ether_addr(entry->addr.u.mac_addr)) {
6203                         NL_SET_ERR_MSG(extack, "L2 entry group is not multicast");
6204                         return -EINVAL;
6205                 }
6206         } else {
6207                 NL_SET_ERR_MSG(extack, "Unknown entry protocol");
6208                 return -EINVAL;
6209         }
6210
6211         if (entry->state != MDB_PERMANENT && entry->state != MDB_TEMPORARY) {
6212                 NL_SET_ERR_MSG(extack, "Unknown entry state");
6213                 return -EINVAL;
6214         }
6215         if (entry->vid >= VLAN_VID_MASK) {
6216                 NL_SET_ERR_MSG(extack, "Invalid entry VLAN id");
6217                 return -EINVAL;
6218         }
6219
6220         return 0;
6221 }
6222
6223 static const struct nla_policy mdba_policy[MDBA_SET_ENTRY_MAX + 1] = {
6224         [MDBA_SET_ENTRY_UNSPEC] = { .strict_start_type = MDBA_SET_ENTRY_ATTRS + 1 },
6225         [MDBA_SET_ENTRY] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
6226                                                   rtnl_validate_mdb_entry,
6227                                                   sizeof(struct br_mdb_entry)),
6228         [MDBA_SET_ENTRY_ATTRS] = { .type = NLA_NESTED },
6229 };
6230
6231 static int rtnl_mdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
6232                         struct netlink_ext_ack *extack)
6233 {
6234         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6235         struct net *net = sock_net(skb->sk);
6236         struct br_port_msg *bpm;
6237         struct net_device *dev;
6238         int err;
6239
6240         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6241                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6242         if (err)
6243                 return err;
6244
6245         bpm = nlmsg_data(nlh);
6246         if (!bpm->ifindex) {
6247                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6248                 return -EINVAL;
6249         }
6250
6251         dev = __dev_get_by_index(net, bpm->ifindex);
6252         if (!dev) {
6253                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6254                 return -ENODEV;
6255         }
6256
6257         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6258                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6259                 return -EINVAL;
6260         }
6261
6262         if (!dev->netdev_ops->ndo_mdb_add) {
6263                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6264                 return -EOPNOTSUPP;
6265         }
6266
6267         return dev->netdev_ops->ndo_mdb_add(dev, tb, nlh->nlmsg_flags, extack);
6268 }
6269
6270 static int rtnl_mdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
6271                         struct netlink_ext_ack *extack)
6272 {
6273         struct nlattr *tb[MDBA_SET_ENTRY_MAX + 1];
6274         struct net *net = sock_net(skb->sk);
6275         struct br_port_msg *bpm;
6276         struct net_device *dev;
6277         int err;
6278
6279         err = nlmsg_parse_deprecated(nlh, sizeof(*bpm), tb,
6280                                      MDBA_SET_ENTRY_MAX, mdba_policy, extack);
6281         if (err)
6282                 return err;
6283
6284         bpm = nlmsg_data(nlh);
6285         if (!bpm->ifindex) {
6286                 NL_SET_ERR_MSG(extack, "Invalid ifindex");
6287                 return -EINVAL;
6288         }
6289
6290         dev = __dev_get_by_index(net, bpm->ifindex);
6291         if (!dev) {
6292                 NL_SET_ERR_MSG(extack, "Device doesn't exist");
6293                 return -ENODEV;
6294         }
6295
6296         if (NL_REQ_ATTR_CHECK(extack, NULL, tb, MDBA_SET_ENTRY)) {
6297                 NL_SET_ERR_MSG(extack, "Missing MDBA_SET_ENTRY attribute");
6298                 return -EINVAL;
6299         }
6300
6301         if (!dev->netdev_ops->ndo_mdb_del) {
6302                 NL_SET_ERR_MSG(extack, "Device does not support MDB operations");
6303                 return -EOPNOTSUPP;
6304         }
6305
6306         return dev->netdev_ops->ndo_mdb_del(dev, tb, extack);
6307 }
6308
6309 /* Process one rtnetlink message. */
6310
6311 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6312                              struct netlink_ext_ack *extack)
6313 {
6314         struct net *net = sock_net(skb->sk);
6315         struct rtnl_link *link;
6316         enum rtnl_kinds kind;
6317         struct module *owner;
6318         int err = -EOPNOTSUPP;
6319         rtnl_doit_func doit;
6320         unsigned int flags;
6321         int family;
6322         int type;
6323
6324         type = nlh->nlmsg_type;
6325         if (type > RTM_MAX)
6326                 return -EOPNOTSUPP;
6327
6328         type -= RTM_BASE;
6329
6330         /* All the messages must have at least 1 byte length */
6331         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6332                 return 0;
6333
6334         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6335         kind = rtnl_msgtype_kind(type);
6336
6337         if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6338                 return -EPERM;
6339
6340         rcu_read_lock();
6341         if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6342                 struct sock *rtnl;
6343                 rtnl_dumpit_func dumpit;
6344                 u32 min_dump_alloc = 0;
6345
6346                 link = rtnl_get_link(family, type);
6347                 if (!link || !link->dumpit) {
6348                         family = PF_UNSPEC;
6349                         link = rtnl_get_link(family, type);
6350                         if (!link || !link->dumpit)
6351                                 goto err_unlock;
6352                 }
6353                 owner = link->owner;
6354                 dumpit = link->dumpit;
6355
6356                 if (type == RTM_GETLINK - RTM_BASE)
6357                         min_dump_alloc = rtnl_calcit(skb, nlh);
6358
6359                 err = 0;
6360                 /* need to do this before rcu_read_unlock() */
6361                 if (!try_module_get(owner))
6362                         err = -EPROTONOSUPPORT;
6363
6364                 rcu_read_unlock();
6365
6366                 rtnl = net->rtnl;
6367                 if (err == 0) {
6368                         struct netlink_dump_control c = {
6369                                 .dump           = dumpit,
6370                                 .min_dump_alloc = min_dump_alloc,
6371                                 .module         = owner,
6372                         };
6373                         err = netlink_dump_start(rtnl, skb, nlh, &c);
6374                         /* netlink_dump_start() will keep a reference on
6375                          * module if dump is still in progress.
6376                          */
6377                         module_put(owner);
6378                 }
6379                 return err;
6380         }
6381
6382         link = rtnl_get_link(family, type);
6383         if (!link || !link->doit) {
6384                 family = PF_UNSPEC;
6385                 link = rtnl_get_link(PF_UNSPEC, type);
6386                 if (!link || !link->doit)
6387                         goto out_unlock;
6388         }
6389
6390         owner = link->owner;
6391         if (!try_module_get(owner)) {
6392                 err = -EPROTONOSUPPORT;
6393                 goto out_unlock;
6394         }
6395
6396         flags = link->flags;
6397         if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6398             !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6399                 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6400                 module_put(owner);
6401                 goto err_unlock;
6402         }
6403
6404         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6405                 doit = link->doit;
6406                 rcu_read_unlock();
6407                 if (doit)
6408                         err = doit(skb, nlh, extack);
6409                 module_put(owner);
6410                 return err;
6411         }
6412         rcu_read_unlock();
6413
6414         rtnl_lock();
6415         link = rtnl_get_link(family, type);
6416         if (link && link->doit)
6417                 err = link->doit(skb, nlh, extack);
6418         rtnl_unlock();
6419
6420         module_put(owner);
6421
6422         return err;
6423
6424 out_unlock:
6425         rcu_read_unlock();
6426         return err;
6427
6428 err_unlock:
6429         rcu_read_unlock();
6430         return -EOPNOTSUPP;
6431 }
6432
6433 static void rtnetlink_rcv(struct sk_buff *skb)
6434 {
6435         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6436 }
6437
6438 static int rtnetlink_bind(struct net *net, int group)
6439 {
6440         switch (group) {
6441         case RTNLGRP_IPV4_MROUTE_R:
6442         case RTNLGRP_IPV6_MROUTE_R:
6443                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6444                         return -EPERM;
6445                 break;
6446         }
6447         return 0;
6448 }
6449
6450 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6451 {
6452         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6453
6454         switch (event) {
6455         case NETDEV_REBOOT:
6456         case NETDEV_CHANGEMTU:
6457         case NETDEV_CHANGEADDR:
6458         case NETDEV_CHANGENAME:
6459         case NETDEV_FEAT_CHANGE:
6460         case NETDEV_BONDING_FAILOVER:
6461         case NETDEV_POST_TYPE_CHANGE:
6462         case NETDEV_NOTIFY_PEERS:
6463         case NETDEV_CHANGEUPPER:
6464         case NETDEV_RESEND_IGMP:
6465         case NETDEV_CHANGEINFODATA:
6466         case NETDEV_CHANGELOWERSTATE:
6467         case NETDEV_CHANGE_TX_QUEUE_LEN:
6468                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6469                                    GFP_KERNEL, NULL, 0, 0, NULL);
6470                 break;
6471         default:
6472                 break;
6473         }
6474         return NOTIFY_DONE;
6475 }
6476
6477 static struct notifier_block rtnetlink_dev_notifier = {
6478         .notifier_call  = rtnetlink_event,
6479 };
6480
6481
6482 static int __net_init rtnetlink_net_init(struct net *net)
6483 {
6484         struct sock *sk;
6485         struct netlink_kernel_cfg cfg = {
6486                 .groups         = RTNLGRP_MAX,
6487                 .input          = rtnetlink_rcv,
6488                 .cb_mutex       = &rtnl_mutex,
6489                 .flags          = NL_CFG_F_NONROOT_RECV,
6490                 .bind           = rtnetlink_bind,
6491         };
6492
6493         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6494         if (!sk)
6495                 return -ENOMEM;
6496         net->rtnl = sk;
6497         return 0;
6498 }
6499
6500 static void __net_exit rtnetlink_net_exit(struct net *net)
6501 {
6502         netlink_kernel_release(net->rtnl);
6503         net->rtnl = NULL;
6504 }
6505
6506 static struct pernet_operations rtnetlink_net_ops = {
6507         .init = rtnetlink_net_init,
6508         .exit = rtnetlink_net_exit,
6509 };
6510
6511 void __init rtnetlink_init(void)
6512 {
6513         if (register_pernet_subsys(&rtnetlink_net_ops))
6514                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6515
6516         register_netdevice_notifier(&rtnetlink_dev_notifier);
6517
6518         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6519                       rtnl_dump_ifinfo, 0);
6520         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6521         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6522         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6523
6524         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6525         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6526         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6527
6528         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6529         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6530
6531         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6532         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6533                       RTNL_FLAG_BULK_DEL_SUPPORTED);
6534         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6535
6536         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6537         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6538         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6539
6540         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6541                       0);
6542         rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);
6543
6544         rtnl_register(PF_BRIDGE, RTM_GETMDB, NULL, rtnl_mdb_dump, 0);
6545         rtnl_register(PF_BRIDGE, RTM_NEWMDB, rtnl_mdb_add, NULL, 0);
6546         rtnl_register(PF_BRIDGE, RTM_DELMDB, rtnl_mdb_del, NULL, 0);
6547 }