net/core/rtnetlink.c

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