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, <kuznet@ms2.inr.ac.ru>
   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           48
  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 #ifdef CONFIG_PROVE_LOCKING
 134 bool lockdep_rtnl_is_held(void)
 135 {
 136         return lockdep_is_held(&rtnl_mutex);
 137 }
 138 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 139 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 140
 141 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 142
 143 static inline int rtm_msgindex(int msgtype)
 144 {
 145         int msgindex = msgtype - RTM_BASE;
 146
 147         /*
 148          * msgindex < 0 implies someone tried to register a netlink
 149          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 150          * the message type has not been added to linux/rtnetlink.h
 151          */
 152         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 153
 154         return msgindex;
 155 }
 156
 157 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 158 {
 159         struct rtnl_link **tab;
 160
 161         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 162                 protocol = PF_UNSPEC;
 163
 164         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 165         if (!tab)
 166                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 167
 168         return tab[msgtype];
 169 }
 170
 171 static int rtnl_register_internal(struct module *owner,
 172                                   int protocol, int msgtype,
 173                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 174                                   unsigned int flags)
 175 {
 176         struct rtnl_link *link, *old;
 177         struct rtnl_link __rcu **tab;
 178         int msgindex;
 179         int ret = -ENOBUFS;
 180
 181         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 182         msgindex = rtm_msgindex(msgtype);
 183
 184         rtnl_lock();
 185         tab = rtnl_msg_handlers[protocol];
 186         if (tab == NULL) {
 187                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 188                 if (!tab)
 189                         goto unlock;
 190
 191                 /* ensures we see the 0 stores */
 192                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 193         }
 194
 195         old = rtnl_dereference(tab[msgindex]);
 196         if (old) {
 197                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 198                 if (!link)
 199                         goto unlock;
 200         } else {
 201                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 202                 if (!link)
 203                         goto unlock;
 204         }
 205
 206         WARN_ON(link->owner && link->owner != owner);
 207         link->owner = owner;
 208
 209         WARN_ON(doit && link->doit && link->doit != doit);
 210         if (doit)
 211                 link->doit = doit;
 212         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 213         if (dumpit)
 214                 link->dumpit = dumpit;
 215
 216         link->flags |= flags;
 217
 218         /* publish protocol:msgtype */
 219         rcu_assign_pointer(tab[msgindex], link);
 220         ret = 0;
 221         if (old)
 222                 kfree_rcu(old, rcu);
 223 unlock:
 224         rtnl_unlock();
 225         return ret;
 226 }
 227
 228 /**
 229  * rtnl_register_module - Register a rtnetlink message type
 230  *
 231  * @owner: module registering the hook (THIS_MODULE)
 232  * @protocol: Protocol family or PF_UNSPEC
 233  * @msgtype: rtnetlink message type
 234  * @doit: Function pointer called for each request message
 235  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 236  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 237  *
 238  * Like rtnl_register, but for use by removable modules.
 239  */
 240 int rtnl_register_module(struct module *owner,
 241                          int protocol, int msgtype,
 242                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 243                          unsigned int flags)
 244 {
 245         return rtnl_register_internal(owner, protocol, msgtype,
 246                                       doit, dumpit, flags);
 247 }
 248 EXPORT_SYMBOL_GPL(rtnl_register_module);
 249
 250 /**
 251  * rtnl_register - Register a rtnetlink message type
 252  * @protocol: Protocol family or PF_UNSPEC
 253  * @msgtype: rtnetlink message type
 254  * @doit: Function pointer called for each request message
 255  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 256  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 257  *
 258  * Registers the specified function pointers (at least one of them has
 259  * to be non-NULL) to be called whenever a request message for the
 260  * specified protocol family and message type is received.
 261  *
 262  * The special protocol family PF_UNSPEC may be used to define fallback
 263  * function pointers for the case when no entry for the specific protocol
 264  * family exists.
 265  */
 266 void rtnl_register(int protocol, int msgtype,
 267                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 268                    unsigned int flags)
 269 {
 270         int err;
 271
 272         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 273                                      flags);
 274         if (err)
 275                 pr_err("Unable to register rtnetlink message handler, "
 276                        "protocol = %d, message type = %d\n", protocol, msgtype);
 277 }
 278
 279 /**
 280  * rtnl_unregister - Unregister a rtnetlink message type
 281  * @protocol: Protocol family or PF_UNSPEC
 282  * @msgtype: rtnetlink message type
 283  *
 284  * Returns 0 on success or a negative error code.
 285  */
 286 int rtnl_unregister(int protocol, int msgtype)
 287 {
 288         struct rtnl_link **tab, *link;
 289         int msgindex;
 290
 291         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 292         msgindex = rtm_msgindex(msgtype);
 293
 294         rtnl_lock();
 295         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 296         if (!tab) {
 297                 rtnl_unlock();
 298                 return -ENOENT;
 299         }
 300
 301         link = tab[msgindex];
 302         rcu_assign_pointer(tab[msgindex], NULL);
 303         rtnl_unlock();
 304
 305         kfree_rcu(link, rcu);
 306
 307         return 0;
 308 }
 309 EXPORT_SYMBOL_GPL(rtnl_unregister);
 310
 311 /**
 312  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 313  * @protocol : Protocol family or PF_UNSPEC
 314  *
 315  * Identical to calling rtnl_unregster() for all registered message types
 316  * of a certain protocol family.
 317  */
 318 void rtnl_unregister_all(int protocol)
 319 {
 320         struct rtnl_link **tab, *link;
 321         int msgindex;
 322
 323         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 324
 325         rtnl_lock();
 326         tab = rtnl_msg_handlers[protocol];
 327         if (!tab) {
 328                 rtnl_unlock();
 329                 return;
 330         }
 331         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 332         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 333                 link = tab[msgindex];
 334                 if (!link)
 335                         continue;
 336
 337                 rcu_assign_pointer(tab[msgindex], NULL);
 338                 kfree_rcu(link, rcu);
 339         }
 340         rtnl_unlock();
 341
 342         synchronize_net();
 343
 344         kfree(tab);
 345 }
 346 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 347
 348 static LIST_HEAD(link_ops);
 349
 350 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 351 {
 352         const struct rtnl_link_ops *ops;
 353
 354         list_for_each_entry(ops, &link_ops, list) {
 355                 if (!strcmp(ops->kind, kind))
 356                         return ops;
 357         }
 358         return NULL;
 359 }
 360
 361 /**
 362  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 363  * @ops: struct rtnl_link_ops * to register
 364  *
 365  * The caller must hold the rtnl_mutex. This function should be used
 366  * by drivers that create devices during module initialization. It
 367  * must be called before registering the devices.
 368  *
 369  * Returns 0 on success or a negative error code.
 370  */
 371 int __rtnl_link_register(struct rtnl_link_ops *ops)
 372 {
 373         if (rtnl_link_ops_get(ops->kind))
 374                 return -EEXIST;
 375
 376         /* The check for setup is here because if ops
 377          * does not have that filled up, it is not possible
 378          * to use the ops for creating device. So do not
 379          * fill up dellink as well. That disables rtnl_dellink.
 380          */
 381         if (ops->setup && !ops->dellink)
 382                 ops->dellink = unregister_netdevice_queue;
 383
 384         list_add_tail(&ops->list, &link_ops);
 385         return 0;
 386 }
 387 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 388
 389 /**
 390  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 391  * @ops: struct rtnl_link_ops * to register
 392  *
 393  * Returns 0 on success or a negative error code.
 394  */
 395 int rtnl_link_register(struct rtnl_link_ops *ops)
 396 {
 397         int err;
 398
 399         /* Sanity-check max sizes to avoid stack buffer overflow. */
 400         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 401                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 402                 return -EINVAL;
 403
 404         rtnl_lock();
 405         err = __rtnl_link_register(ops);
 406         rtnl_unlock();
 407         return err;
 408 }
 409 EXPORT_SYMBOL_GPL(rtnl_link_register);
 410
 411 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 412 {
 413         struct net_device *dev;
 414         LIST_HEAD(list_kill);
 415
 416         for_each_netdev(net, dev) {
 417                 if (dev->rtnl_link_ops == ops)
 418                         ops->dellink(dev, &list_kill);
 419         }
 420         unregister_netdevice_many(&list_kill);
 421 }
 422
 423 /**
 424  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 425  * @ops: struct rtnl_link_ops * to unregister
 426  *
 427  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 428  * integrity (hold pernet_ops_rwsem for writing to close the race
 429  * with setup_net() and cleanup_net()).
 430  */
 431 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 432 {
 433         struct net *net;
 434
 435         for_each_net(net) {
 436                 __rtnl_kill_links(net, ops);
 437         }
 438         list_del(&ops->list);
 439 }
 440 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 441
 442 /* Return with the rtnl_lock held when there are no network
 443  * devices unregistering in any network namespace.
 444  */
 445 static void rtnl_lock_unregistering_all(void)
 446 {
 447         struct net *net;
 448         bool unregistering;
 449         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 450
 451         add_wait_queue(&netdev_unregistering_wq, &wait);
 452         for (;;) {
 453                 unregistering = false;
 454                 rtnl_lock();
 455                 /* We held write locked pernet_ops_rwsem, and parallel
 456                  * setup_net() and cleanup_net() are not possible.
 457                  */
 458                 for_each_net(net) {
 459                         if (net->dev_unreg_count > 0) {
 460                                 unregistering = true;
 461                                 break;
 462                         }
 463                 }
 464                 if (!unregistering)
 465                         break;
 466                 __rtnl_unlock();
 467
 468                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 469         }
 470         remove_wait_queue(&netdev_unregistering_wq, &wait);
 471 }
 472
 473 /**
 474  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 475  * @ops: struct rtnl_link_ops * to unregister
 476  */
 477 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 478 {
 479         /* Close the race with setup_net() and cleanup_net() */
 480         down_write(&pernet_ops_rwsem);
 481         rtnl_lock_unregistering_all();
 482         __rtnl_link_unregister(ops);
 483         rtnl_unlock();
 484         up_write(&pernet_ops_rwsem);
 485 }
 486 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 487
 488 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 489 {
 490         struct net_device *master_dev;
 491         const struct rtnl_link_ops *ops;
 492         size_t size = 0;
 493
 494         rcu_read_lock();
 495
 496         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 497         if (!master_dev)
 498                 goto out;
 499
 500         ops = master_dev->rtnl_link_ops;
 501         if (!ops || !ops->get_slave_size)
 502                 goto out;
 503         /* IFLA_INFO_SLAVE_DATA + nested data */
 504         size = nla_total_size(sizeof(struct nlattr)) +
 505                ops->get_slave_size(master_dev, dev);
 506
 507 out:
 508         rcu_read_unlock();
 509         return size;
 510 }
 511
 512 static size_t rtnl_link_get_size(const struct net_device *dev)
 513 {
 514         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 515         size_t size;
 516
 517         if (!ops)
 518                 return 0;
 519
 520         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 521                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 522
 523         if (ops->get_size)
 524                 /* IFLA_INFO_DATA + nested data */
 525                 size += nla_total_size(sizeof(struct nlattr)) +
 526                         ops->get_size(dev);
 527
 528         if (ops->get_xstats_size)
 529                 /* IFLA_INFO_XSTATS */
 530                 size += nla_total_size(ops->get_xstats_size(dev));
 531
 532         size += rtnl_link_get_slave_info_data_size(dev);
 533
 534         return size;
 535 }
 536
 537 static LIST_HEAD(rtnl_af_ops);
 538
 539 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 540 {
 541         const struct rtnl_af_ops *ops;
 542
 543         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 544                 if (ops->family == family)
 545                         return ops;
 546         }
 547
 548         return NULL;
 549 }
 550
 551 /**
 552  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 553  * @ops: struct rtnl_af_ops * to register
 554  *
 555  * Returns 0 on success or a negative error code.
 556  */
 557 void rtnl_af_register(struct rtnl_af_ops *ops)
 558 {
 559         rtnl_lock();
 560         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 561         rtnl_unlock();
 562 }
 563 EXPORT_SYMBOL_GPL(rtnl_af_register);
 564
 565 /**
 566  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 567  * @ops: struct rtnl_af_ops * to unregister
 568  */
 569 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 570 {
 571         rtnl_lock();
 572         list_del_rcu(&ops->list);
 573         rtnl_unlock();
 574
 575         synchronize_rcu();
 576 }
 577 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 578
 579 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 580                                     u32 ext_filter_mask)
 581 {
 582         struct rtnl_af_ops *af_ops;
 583         size_t size;
 584
 585         /* IFLA_AF_SPEC */
 586         size = nla_total_size(sizeof(struct nlattr));
 587
 588         rcu_read_lock();
 589         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 590                 if (af_ops->get_link_af_size) {
 591                         /* AF_* + nested data */
 592                         size += nla_total_size(sizeof(struct nlattr)) +
 593                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 594                 }
 595         }
 596         rcu_read_unlock();
 597
 598         return size;
 599 }
 600
 601 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 602 {
 603         struct net_device *master_dev;
 604         bool ret = false;
 605
 606         rcu_read_lock();
 607
 608         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 609         if (master_dev && master_dev->rtnl_link_ops)
 610                 ret = true;
 611         rcu_read_unlock();
 612         return ret;
 613 }
 614
 615 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 616                                      const struct net_device *dev)
 617 {
 618         struct net_device *master_dev;
 619         const struct rtnl_link_ops *ops;
 620         struct nlattr *slave_data;
 621         int err;
 622
 623         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 624         if (!master_dev)
 625                 return 0;
 626         ops = master_dev->rtnl_link_ops;
 627         if (!ops)
 628                 return 0;
 629         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 630                 return -EMSGSIZE;
 631         if (ops->fill_slave_info) {
 632                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 633                 if (!slave_data)
 634                         return -EMSGSIZE;
 635                 err = ops->fill_slave_info(skb, master_dev, dev);
 636                 if (err < 0)
 637                         goto err_cancel_slave_data;
 638                 nla_nest_end(skb, slave_data);
 639         }
 640         return 0;
 641
 642 err_cancel_slave_data:
 643         nla_nest_cancel(skb, slave_data);
 644         return err;
 645 }
 646
 647 static int rtnl_link_info_fill(struct sk_buff *skb,
 648                                const struct net_device *dev)
 649 {
 650         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 651         struct nlattr *data;
 652         int err;
 653
 654         if (!ops)
 655                 return 0;
 656         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 657                 return -EMSGSIZE;
 658         if (ops->fill_xstats) {
 659                 err = ops->fill_xstats(skb, dev);
 660                 if (err < 0)
 661                         return err;
 662         }
 663         if (ops->fill_info) {
 664                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 665                 if (data == NULL)
 666                         return -EMSGSIZE;
 667                 err = ops->fill_info(skb, dev);
 668                 if (err < 0)
 669                         goto err_cancel_data;
 670                 nla_nest_end(skb, data);
 671         }
 672         return 0;
 673
 674 err_cancel_data:
 675         nla_nest_cancel(skb, data);
 676         return err;
 677 }
 678
 679 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 680 {
 681         struct nlattr *linkinfo;
 682         int err = -EMSGSIZE;
 683
 684         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 685         if (linkinfo == NULL)
 686                 goto out;
 687
 688         err = rtnl_link_info_fill(skb, dev);
 689         if (err < 0)
 690                 goto err_cancel_link;
 691
 692         err = rtnl_link_slave_info_fill(skb, dev);
 693         if (err < 0)
 694                 goto err_cancel_link;
 695
 696         nla_nest_end(skb, linkinfo);
 697         return 0;
 698
 699 err_cancel_link:
 700         nla_nest_cancel(skb, linkinfo);
 701 out:
 702         return err;
 703 }
 704
 705 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 706 {
 707         struct sock *rtnl = net->rtnl;
 708         int err = 0;
 709
 710         NETLINK_CB(skb).dst_group = group;
 711         if (echo)
 712                 refcount_inc(&skb->users);
 713         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 714         if (echo)
 715                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 716         return err;
 717 }
 718
 719 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 720 {
 721         struct sock *rtnl = net->rtnl;
 722
 723         return nlmsg_unicast(rtnl, skb, pid);
 724 }
 725 EXPORT_SYMBOL(rtnl_unicast);
 726
 727 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 728                  struct nlmsghdr *nlh, gfp_t flags)
 729 {
 730         struct sock *rtnl = net->rtnl;
 731         int report = 0;
 732
 733         if (nlh)
 734                 report = nlmsg_report(nlh);
 735
 736         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 737 }
 738 EXPORT_SYMBOL(rtnl_notify);
 739
 740 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 741 {
 742         struct sock *rtnl = net->rtnl;
 743
 744         netlink_set_err(rtnl, 0, group, error);
 745 }
 746 EXPORT_SYMBOL(rtnl_set_sk_err);
 747
 748 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 749 {
 750         struct nlattr *mx;
 751         int i, valid = 0;
 752
 753         mx = nla_nest_start(skb, RTA_METRICS);
 754         if (mx == NULL)
 755                 return -ENOBUFS;
 756
 757         for (i = 0; i < RTAX_MAX; i++) {
 758                 if (metrics[i]) {
 759                         if (i == RTAX_CC_ALGO - 1) {
 760                                 char tmp[TCP_CA_NAME_MAX], *name;
 761
 762                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 763                                 if (!name)
 764                                         continue;
 765                                 if (nla_put_string(skb, i + 1, name))
 766                                         goto nla_put_failure;
 767                         } else if (i == RTAX_FEATURES - 1) {
 768                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 769
 770                                 if (!user_features)
 771                                         continue;
 772                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 773                                 if (nla_put_u32(skb, i + 1, user_features))
 774                                         goto nla_put_failure;
 775                         } else {
 776                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 777                                         goto nla_put_failure;
 778                         }
 779                         valid++;
 780                 }
 781         }
 782
 783         if (!valid) {
 784                 nla_nest_cancel(skb, mx);
 785                 return 0;
 786         }
 787
 788         return nla_nest_end(skb, mx);
 789
 790 nla_put_failure:
 791         nla_nest_cancel(skb, mx);
 792         return -EMSGSIZE;
 793 }
 794 EXPORT_SYMBOL(rtnetlink_put_metrics);
 795
 796 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 797                        long expires, u32 error)
 798 {
 799         struct rta_cacheinfo ci = {
 800                 .rta_error = error,
 801                 .rta_id =  id,
 802         };
 803
 804         if (dst) {
 805                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 806                 ci.rta_used = dst->__use;
 807                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 808         }
 809         if (expires) {
 810                 unsigned long clock;
 811
 812                 clock = jiffies_to_clock_t(abs(expires));
 813                 clock = min_t(unsigned long, clock, INT_MAX);
 814                 ci.rta_expires = (expires > 0) ? clock : -clock;
 815         }
 816         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 817 }
 818 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 819
 820 static void set_operstate(struct net_device *dev, unsigned char transition)
 821 {
 822         unsigned char operstate = dev->operstate;
 823
 824         switch (transition) {
 825         case IF_OPER_UP:
 826                 if ((operstate == IF_OPER_DORMANT ||
 827                      operstate == IF_OPER_UNKNOWN) &&
 828                     !netif_dormant(dev))
 829                         operstate = IF_OPER_UP;
 830                 break;
 831
 832         case IF_OPER_DORMANT:
 833                 if (operstate == IF_OPER_UP ||
 834                     operstate == IF_OPER_UNKNOWN)
 835                         operstate = IF_OPER_DORMANT;
 836                 break;
 837         }
 838
 839         if (dev->operstate != operstate) {
 840                 write_lock_bh(&dev_base_lock);
 841                 dev->operstate = operstate;
 842                 write_unlock_bh(&dev_base_lock);
 843                 netdev_state_change(dev);
 844         }
 845 }
 846
 847 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 848 {
 849         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 850                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 851 }
 852
 853 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 854                                            const struct ifinfomsg *ifm)
 855 {
 856         unsigned int flags = ifm->ifi_flags;
 857
 858         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 859         if (ifm->ifi_change)
 860                 flags = (flags & ifm->ifi_change) |
 861                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 862
 863         return flags;
 864 }
 865
 866 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 867                                  const struct rtnl_link_stats64 *b)
 868 {
 869         a->rx_packets = b->rx_packets;
 870         a->tx_packets = b->tx_packets;
 871         a->rx_bytes = b->rx_bytes;
 872         a->tx_bytes = b->tx_bytes;
 873         a->rx_errors = b->rx_errors;
 874         a->tx_errors = b->tx_errors;
 875         a->rx_dropped = b->rx_dropped;
 876         a->tx_dropped = b->tx_dropped;
 877
 878         a->multicast = b->multicast;
 879         a->collisions = b->collisions;
 880
 881         a->rx_length_errors = b->rx_length_errors;
 882         a->rx_over_errors = b->rx_over_errors;
 883         a->rx_crc_errors = b->rx_crc_errors;
 884         a->rx_frame_errors = b->rx_frame_errors;
 885         a->rx_fifo_errors = b->rx_fifo_errors;
 886         a->rx_missed_errors = b->rx_missed_errors;
 887
 888         a->tx_aborted_errors = b->tx_aborted_errors;
 889         a->tx_carrier_errors = b->tx_carrier_errors;
 890         a->tx_fifo_errors = b->tx_fifo_errors;
 891         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 892         a->tx_window_errors = b->tx_window_errors;
 893
 894         a->rx_compressed = b->rx_compressed;
 895         a->tx_compressed = b->tx_compressed;
 896
 897         a->rx_nohandler = b->rx_nohandler;
 898 }
 899
 900 /* All VF info */
 901 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 902                                    u32 ext_filter_mask)
 903 {
 904         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 905                 int num_vfs = dev_num_vf(dev->dev.parent);
 906                 size_t size = nla_total_size(0);
 907                 size += num_vfs *
 908                         (nla_total_size(0) +
 909                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 910                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 911                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 912                          nla_total_size(MAX_VLAN_LIST_LEN *
 913                                         sizeof(struct ifla_vf_vlan_info)) +
 914                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 915                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 916                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 917                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 918                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 919                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 920                          /* IFLA_VF_STATS_RX_PACKETS */
 921                          nla_total_size_64bit(sizeof(__u64)) +
 922                          /* IFLA_VF_STATS_TX_PACKETS */
 923                          nla_total_size_64bit(sizeof(__u64)) +
 924                          /* IFLA_VF_STATS_RX_BYTES */
 925                          nla_total_size_64bit(sizeof(__u64)) +
 926                          /* IFLA_VF_STATS_TX_BYTES */
 927                          nla_total_size_64bit(sizeof(__u64)) +
 928                          /* IFLA_VF_STATS_BROADCAST */
 929                          nla_total_size_64bit(sizeof(__u64)) +
 930                          /* IFLA_VF_STATS_MULTICAST */
 931                          nla_total_size_64bit(sizeof(__u64)) +
 932                          /* IFLA_VF_STATS_RX_DROPPED */
 933                          nla_total_size_64bit(sizeof(__u64)) +
 934                          /* IFLA_VF_STATS_TX_DROPPED */
 935                          nla_total_size_64bit(sizeof(__u64)) +
 936                          nla_total_size(sizeof(struct ifla_vf_trust)));
 937                 return size;
 938         } else
 939                 return 0;
 940 }
 941
 942 static size_t rtnl_port_size(const struct net_device *dev,
 943                              u32 ext_filter_mask)
 944 {
 945         size_t port_size = nla_total_size(4)            /* PORT_VF */
 946                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 947                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 948                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 949                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 950                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 951         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 952         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 953                 + port_size;
 954         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 955                 + port_size;
 956
 957         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 958             !(ext_filter_mask & RTEXT_FILTER_VF))
 959                 return 0;
 960         if (dev_num_vf(dev->dev.parent))
 961                 return port_self_size + vf_ports_size +
 962                         vf_port_size * dev_num_vf(dev->dev.parent);
 963         else
 964                 return port_self_size;
 965 }
 966
 967 static size_t rtnl_xdp_size(void)
 968 {
 969         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 970                           nla_total_size(1) +   /* XDP_ATTACHED */
 971                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 972                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 973
 974         return xdp_size;
 975 }
 976
 977 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 978                                      u32 ext_filter_mask)
 979 {
 980         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 981                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 982                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 983                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 984                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 985                + nla_total_size(sizeof(struct rtnl_link_stats))
 986                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 987                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 988                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 989                + nla_total_size(4) /* IFLA_TXQLEN */
 990                + nla_total_size(4) /* IFLA_WEIGHT */
 991                + nla_total_size(4) /* IFLA_MTU */
 992                + nla_total_size(4) /* IFLA_LINK */
 993                + nla_total_size(4) /* IFLA_MASTER */
 994                + nla_total_size(1) /* IFLA_CARRIER */
 995                + nla_total_size(4) /* IFLA_PROMISCUITY */
 996                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 997                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 998                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 999                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1000                + nla_total_size(1) /* IFLA_OPERSTATE */
1001                + nla_total_size(1) /* IFLA_LINKMODE */
1002                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1003                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1004                + nla_total_size(4) /* IFLA_GROUP */
1005                + nla_total_size(ext_filter_mask
1006                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1007                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1008                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1009                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1010                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1011                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1012                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1013                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1014                + rtnl_xdp_size() /* IFLA_XDP */
1015                + nla_total_size(4)  /* IFLA_EVENT */
1016                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1017                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1018                + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1019                + nla_total_size(4)  /* IFLA_IF_NETNSID */
1020                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1021                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1022                + nla_total_size(4)  /* IFLA_MIN_MTU */
1023                + nla_total_size(4)  /* IFLA_MAX_MTU */
1024                + 0;
1025 }
1026
1027 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1028 {
1029         struct nlattr *vf_ports;
1030         struct nlattr *vf_port;
1031         int vf;
1032         int err;
1033
1034         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
1035         if (!vf_ports)
1036                 return -EMSGSIZE;
1037
1038         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1039                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
1040                 if (!vf_port)
1041                         goto nla_put_failure;
1042                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1043                         goto nla_put_failure;
1044                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1045                 if (err == -EMSGSIZE)
1046                         goto nla_put_failure;
1047                 if (err) {
1048                         nla_nest_cancel(skb, vf_port);
1049                         continue;
1050                 }
1051                 nla_nest_end(skb, vf_port);
1052         }
1053
1054         nla_nest_end(skb, vf_ports);
1055
1056         return 0;
1057
1058 nla_put_failure:
1059         nla_nest_cancel(skb, vf_ports);
1060         return -EMSGSIZE;
1061 }
1062
1063 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1064 {
1065         struct nlattr *port_self;
1066         int err;
1067
1068         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1069         if (!port_self)
1070                 return -EMSGSIZE;
1071
1072         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1073         if (err) {
1074                 nla_nest_cancel(skb, port_self);
1075                 return (err == -EMSGSIZE) ? err : 0;
1076         }
1077
1078         nla_nest_end(skb, port_self);
1079
1080         return 0;
1081 }
1082
1083 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1084                           u32 ext_filter_mask)
1085 {
1086         int err;
1087
1088         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1089             !(ext_filter_mask & RTEXT_FILTER_VF))
1090                 return 0;
1091
1092         err = rtnl_port_self_fill(skb, dev);
1093         if (err)
1094                 return err;
1095
1096         if (dev_num_vf(dev->dev.parent)) {
1097                 err = rtnl_vf_ports_fill(skb, dev);
1098                 if (err)
1099                         return err;
1100         }
1101
1102         return 0;
1103 }
1104
1105 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1106 {
1107         int err;
1108         struct netdev_phys_item_id ppid;
1109
1110         err = dev_get_phys_port_id(dev, &ppid);
1111         if (err) {
1112                 if (err == -EOPNOTSUPP)
1113                         return 0;
1114                 return err;
1115         }
1116
1117         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1118                 return -EMSGSIZE;
1119
1120         return 0;
1121 }
1122
1123 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1124 {
1125         char name[IFNAMSIZ];
1126         int err;
1127
1128         err = dev_get_phys_port_name(dev, name, sizeof(name));
1129         if (err) {
1130                 if (err == -EOPNOTSUPP)
1131                         return 0;
1132                 return err;
1133         }
1134
1135         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1136                 return -EMSGSIZE;
1137
1138         return 0;
1139 }
1140
1141 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1142 {
1143         int err;
1144         struct switchdev_attr attr = {
1145                 .orig_dev = dev,
1146                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1147                 .flags = SWITCHDEV_F_NO_RECURSE,
1148         };
1149
1150         err = switchdev_port_attr_get(dev, &attr);
1151         if (err) {
1152                 if (err == -EOPNOTSUPP)
1153                         return 0;
1154                 return err;
1155         }
1156
1157         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1158                     attr.u.ppid.id))
1159                 return -EMSGSIZE;
1160
1161         return 0;
1162 }
1163
1164 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1165                                               struct net_device *dev)
1166 {
1167         struct rtnl_link_stats64 *sp;
1168         struct nlattr *attr;
1169
1170         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1171                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1172         if (!attr)
1173                 return -EMSGSIZE;
1174
1175         sp = nla_data(attr);
1176         dev_get_stats(dev, sp);
1177
1178         attr = nla_reserve(skb, IFLA_STATS,
1179                            sizeof(struct rtnl_link_stats));
1180         if (!attr)
1181                 return -EMSGSIZE;
1182
1183         copy_rtnl_link_stats(nla_data(attr), sp);
1184
1185         return 0;
1186 }
1187
1188 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1189                                                struct net_device *dev,
1190                                                int vfs_num,
1191                                                struct nlattr *vfinfo)
1192 {
1193         struct ifla_vf_rss_query_en vf_rss_query_en;
1194         struct nlattr *vf, *vfstats, *vfvlanlist;
1195         struct ifla_vf_link_state vf_linkstate;
1196         struct ifla_vf_vlan_info vf_vlan_info;
1197         struct ifla_vf_spoofchk vf_spoofchk;
1198         struct ifla_vf_tx_rate vf_tx_rate;
1199         struct ifla_vf_stats vf_stats;
1200         struct ifla_vf_trust vf_trust;
1201         struct ifla_vf_vlan vf_vlan;
1202         struct ifla_vf_rate vf_rate;
1203         struct ifla_vf_mac vf_mac;
1204         struct ifla_vf_info ivi;
1205
1206         memset(&ivi, 0, sizeof(ivi));
1207
1208         /* Not all SR-IOV capable drivers support the
1209          * spoofcheck and "RSS query enable" query.  Preset to
1210          * -1 so the user space tool can detect that the driver
1211          * didn't report anything.
1212          */
1213         ivi.spoofchk = -1;
1214         ivi.rss_query_en = -1;
1215         ivi.trusted = -1;
1216         /* The default value for VF link state is "auto"
1217          * IFLA_VF_LINK_STATE_AUTO which equals zero
1218          */
1219         ivi.linkstate = 0;
1220         /* VLAN Protocol by default is 802.1Q */
1221         ivi.vlan_proto = htons(ETH_P_8021Q);
1222         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1223                 return 0;
1224
1225         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1226
1227         vf_mac.vf =
1228                 vf_vlan.vf =
1229                 vf_vlan_info.vf =
1230                 vf_rate.vf =
1231                 vf_tx_rate.vf =
1232                 vf_spoofchk.vf =
1233                 vf_linkstate.vf =
1234                 vf_rss_query_en.vf =
1235                 vf_trust.vf = ivi.vf;
1236
1237         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1238         vf_vlan.vlan = ivi.vlan;
1239         vf_vlan.qos = ivi.qos;
1240         vf_vlan_info.vlan = ivi.vlan;
1241         vf_vlan_info.qos = ivi.qos;
1242         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1243         vf_tx_rate.rate = ivi.max_tx_rate;
1244         vf_rate.min_tx_rate = ivi.min_tx_rate;
1245         vf_rate.max_tx_rate = ivi.max_tx_rate;
1246         vf_spoofchk.setting = ivi.spoofchk;
1247         vf_linkstate.link_state = ivi.linkstate;
1248         vf_rss_query_en.setting = ivi.rss_query_en;
1249         vf_trust.setting = ivi.trusted;
1250         vf = nla_nest_start(skb, IFLA_VF_INFO);
1251         if (!vf)
1252                 goto nla_put_vfinfo_failure;
1253         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1254             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1255             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1256                     &vf_rate) ||
1257             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1258                     &vf_tx_rate) ||
1259             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1260                     &vf_spoofchk) ||
1261             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1262                     &vf_linkstate) ||
1263             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1264                     sizeof(vf_rss_query_en),
1265                     &vf_rss_query_en) ||
1266             nla_put(skb, IFLA_VF_TRUST,
1267                     sizeof(vf_trust), &vf_trust))
1268                 goto nla_put_vf_failure;
1269         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1270         if (!vfvlanlist)
1271                 goto nla_put_vf_failure;
1272         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1273                     &vf_vlan_info)) {
1274                 nla_nest_cancel(skb, vfvlanlist);
1275                 goto nla_put_vf_failure;
1276         }
1277         nla_nest_end(skb, vfvlanlist);
1278         memset(&vf_stats, 0, sizeof(vf_stats));
1279         if (dev->netdev_ops->ndo_get_vf_stats)
1280                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1281                                                 &vf_stats);
1282         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1283         if (!vfstats)
1284                 goto nla_put_vf_failure;
1285         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1286                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1287             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1288                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1289             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1290                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1291             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1292                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1293             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1294                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1295             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1296                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1297             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1298                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1299             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1300                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1301                 nla_nest_cancel(skb, vfstats);
1302                 goto nla_put_vf_failure;
1303         }
1304         nla_nest_end(skb, vfstats);
1305         nla_nest_end(skb, vf);
1306         return 0;
1307
1308 nla_put_vf_failure:
1309         nla_nest_cancel(skb, vf);
1310 nla_put_vfinfo_failure:
1311         nla_nest_cancel(skb, vfinfo);
1312         return -EMSGSIZE;
1313 }
1314
1315 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1316                                            struct net_device *dev,
1317                                            u32 ext_filter_mask)
1318 {
1319         struct nlattr *vfinfo;
1320         int i, num_vfs;
1321
1322         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1323                 return 0;
1324
1325         num_vfs = dev_num_vf(dev->dev.parent);
1326         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1327                 return -EMSGSIZE;
1328
1329         if (!dev->netdev_ops->ndo_get_vf_config)
1330                 return 0;
1331
1332         vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1333         if (!vfinfo)
1334                 return -EMSGSIZE;
1335
1336         for (i = 0; i < num_vfs; i++) {
1337                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1338                         return -EMSGSIZE;
1339         }
1340
1341         nla_nest_end(skb, vfinfo);
1342         return 0;
1343 }
1344
1345 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1346 {
1347         struct rtnl_link_ifmap map;
1348
1349         memset(&map, 0, sizeof(map));
1350         map.mem_start   = dev->mem_start;
1351         map.mem_end     = dev->mem_end;
1352         map.base_addr   = dev->base_addr;
1353         map.irq         = dev->irq;
1354         map.dma         = dev->dma;
1355         map.port        = dev->if_port;
1356
1357         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1358                 return -EMSGSIZE;
1359
1360         return 0;
1361 }
1362
1363 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1364 {
1365         const struct bpf_prog *generic_xdp_prog;
1366
1367         ASSERT_RTNL();
1368
1369         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1370         if (!generic_xdp_prog)
1371                 return 0;
1372         return generic_xdp_prog->aux->id;
1373 }
1374
1375 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1376 {
1377         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1378 }
1379
1380 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1381 {
1382         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1383                                XDP_QUERY_PROG_HW);
1384 }
1385
1386 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1387                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1388                                u32 (*get_prog_id)(struct net_device *dev))
1389 {
1390         u32 curr_id;
1391         int err;
1392
1393         curr_id = get_prog_id(dev);
1394         if (!curr_id)
1395                 return 0;
1396
1397         *prog_id = curr_id;
1398         err = nla_put_u32(skb, attr, curr_id);
1399         if (err)
1400                 return err;
1401
1402         if (*mode != XDP_ATTACHED_NONE)
1403                 *mode = XDP_ATTACHED_MULTI;
1404         else
1405                 *mode = tgt_mode;
1406
1407         return 0;
1408 }
1409
1410 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1411 {
1412         struct nlattr *xdp;
1413         u32 prog_id;
1414         int err;
1415         u8 mode;
1416
1417         xdp = nla_nest_start(skb, IFLA_XDP);
1418         if (!xdp)
1419                 return -EMSGSIZE;
1420
1421         prog_id = 0;
1422         mode = XDP_ATTACHED_NONE;
1423         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1424                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1425         if (err)
1426                 goto err_cancel;
1427         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1428                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1429         if (err)
1430                 goto err_cancel;
1431         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1432                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1433         if (err)
1434                 goto err_cancel;
1435
1436         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1437         if (err)
1438                 goto err_cancel;
1439
1440         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1441                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1442                 if (err)
1443                         goto err_cancel;
1444         }
1445
1446         nla_nest_end(skb, xdp);
1447         return 0;
1448
1449 err_cancel:
1450         nla_nest_cancel(skb, xdp);
1451         return err;
1452 }
1453
1454 static u32 rtnl_get_event(unsigned long event)
1455 {
1456         u32 rtnl_event_type = IFLA_EVENT_NONE;
1457
1458         switch (event) {
1459         case NETDEV_REBOOT:
1460                 rtnl_event_type = IFLA_EVENT_REBOOT;
1461                 break;
1462         case NETDEV_FEAT_CHANGE:
1463                 rtnl_event_type = IFLA_EVENT_FEATURES;
1464                 break;
1465         case NETDEV_BONDING_FAILOVER:
1466                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1467                 break;
1468         case NETDEV_NOTIFY_PEERS:
1469                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1470                 break;
1471         case NETDEV_RESEND_IGMP:
1472                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1473                 break;
1474         case NETDEV_CHANGEINFODATA:
1475                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1476                 break;
1477         default:
1478                 break;
1479         }
1480
1481         return rtnl_event_type;
1482 }
1483
1484 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1485 {
1486         const struct net_device *upper_dev;
1487         int ret = 0;
1488
1489         rcu_read_lock();
1490
1491         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1492         if (upper_dev)
1493                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1494
1495         rcu_read_unlock();
1496         return ret;
1497 }
1498
1499 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1500 {
1501         int ifindex = dev_get_iflink(dev);
1502
1503         if (dev->ifindex == ifindex)
1504                 return 0;
1505
1506         return nla_put_u32(skb, IFLA_LINK, ifindex);
1507 }
1508
1509 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1510                                               struct net_device *dev)
1511 {
1512         char buf[IFALIASZ];
1513         int ret;
1514
1515         ret = dev_get_alias(dev, buf, sizeof(buf));
1516         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1517 }
1518
1519 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1520                                   const struct net_device *dev,
1521                                   struct net *src_net)
1522 {
1523         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1524                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1525
1526                 if (!net_eq(dev_net(dev), link_net)) {
1527                         int id = peernet2id_alloc(src_net, link_net);
1528
1529                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1530                                 return -EMSGSIZE;
1531                 }
1532         }
1533
1534         return 0;
1535 }
1536
1537 static int rtnl_fill_link_af(struct sk_buff *skb,
1538                              const struct net_device *dev,
1539                              u32 ext_filter_mask)
1540 {
1541         const struct rtnl_af_ops *af_ops;
1542         struct nlattr *af_spec;
1543
1544         af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1545         if (!af_spec)
1546                 return -EMSGSIZE;
1547
1548         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1549                 struct nlattr *af;
1550                 int err;
1551
1552                 if (!af_ops->fill_link_af)
1553                         continue;
1554
1555                 af = nla_nest_start(skb, af_ops->family);
1556                 if (!af)
1557                         return -EMSGSIZE;
1558
1559                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1560                 /*
1561                  * Caller may return ENODATA to indicate that there
1562                  * was no data to be dumped. This is not an error, it
1563                  * means we should trim the attribute header and
1564                  * continue.
1565                  */
1566                 if (err == -ENODATA)
1567                         nla_nest_cancel(skb, af);
1568                 else if (err < 0)
1569                         return -EMSGSIZE;
1570
1571                 nla_nest_end(skb, af);
1572         }
1573
1574         nla_nest_end(skb, af_spec);
1575         return 0;
1576 }
1577
1578 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1579                             struct net_device *dev, struct net *src_net,
1580                             int type, u32 pid, u32 seq, u32 change,
1581                             unsigned int flags, u32 ext_filter_mask,
1582                             u32 event, int *new_nsid, int new_ifindex,
1583                             int tgt_netnsid)
1584 {
1585         struct ifinfomsg *ifm;
1586         struct nlmsghdr *nlh;
1587
1588         ASSERT_RTNL();
1589         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1590         if (nlh == NULL)
1591                 return -EMSGSIZE;
1592
1593         ifm = nlmsg_data(nlh);
1594         ifm->ifi_family = AF_UNSPEC;
1595         ifm->__ifi_pad = 0;
1596         ifm->ifi_type = dev->type;
1597         ifm->ifi_index = dev->ifindex;
1598         ifm->ifi_flags = dev_get_flags(dev);
1599         ifm->ifi_change = change;
1600
1601         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1602                 goto nla_put_failure;
1603
1604         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1605             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1606             nla_put_u8(skb, IFLA_OPERSTATE,
1607                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1608             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1609             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1610             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1611             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1612             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1613             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1614             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1615             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1616             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1617 #ifdef CONFIG_RPS
1618             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1619 #endif
1620             nla_put_iflink(skb, dev) ||
1621             put_master_ifindex(skb, dev) ||
1622             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1623             (dev->qdisc &&
1624              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1625             nla_put_ifalias(skb, dev) ||
1626             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1627                         atomic_read(&dev->carrier_up_count) +
1628                         atomic_read(&dev->carrier_down_count)) ||
1629             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1630             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1631                         atomic_read(&dev->carrier_up_count)) ||
1632             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1633                         atomic_read(&dev->carrier_down_count)))
1634                 goto nla_put_failure;
1635
1636         if (event != IFLA_EVENT_NONE) {
1637                 if (nla_put_u32(skb, IFLA_EVENT, event))
1638                         goto nla_put_failure;
1639         }
1640
1641         if (rtnl_fill_link_ifmap(skb, dev))
1642                 goto nla_put_failure;
1643
1644         if (dev->addr_len) {
1645                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1646                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1647                         goto nla_put_failure;
1648         }
1649
1650         if (rtnl_phys_port_id_fill(skb, dev))
1651                 goto nla_put_failure;
1652
1653         if (rtnl_phys_port_name_fill(skb, dev))
1654                 goto nla_put_failure;
1655
1656         if (rtnl_phys_switch_id_fill(skb, dev))
1657                 goto nla_put_failure;
1658
1659         if (rtnl_fill_stats(skb, dev))
1660                 goto nla_put_failure;
1661
1662         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1663                 goto nla_put_failure;
1664
1665         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1666                 goto nla_put_failure;
1667
1668         if (rtnl_xdp_fill(skb, dev))
1669                 goto nla_put_failure;
1670
1671         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1672                 if (rtnl_link_fill(skb, dev) < 0)
1673                         goto nla_put_failure;
1674         }
1675
1676         if (rtnl_fill_link_netnsid(skb, dev, src_net))
1677                 goto nla_put_failure;
1678
1679         if (new_nsid &&
1680             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1681                 goto nla_put_failure;
1682         if (new_ifindex &&
1683             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1684                 goto nla_put_failure;
1685
1686
1687         rcu_read_lock();
1688         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1689                 goto nla_put_failure_rcu;
1690         rcu_read_unlock();
1691
1692         nlmsg_end(skb, nlh);
1693         return 0;
1694
1695 nla_put_failure_rcu:
1696         rcu_read_unlock();
1697 nla_put_failure:
1698         nlmsg_cancel(skb, nlh);
1699         return -EMSGSIZE;
1700 }
1701
1702 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1703         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1704         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1705         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1706         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1707         [IFLA_MTU]              = { .type = NLA_U32 },
1708         [IFLA_LINK]             = { .type = NLA_U32 },
1709         [IFLA_MASTER]           = { .type = NLA_U32 },
1710         [IFLA_CARRIER]          = { .type = NLA_U8 },
1711         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1712         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1713         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1714         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1715         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1716         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1717         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1718         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1719          * allow 0-length string (needed to remove an alias).
1720          */
1721         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1722         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1723         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1724         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1725         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1726         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1727         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1728         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1729         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1730         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1731         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1732         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1733         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1734         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1735         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1736         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1737         [IFLA_XDP]              = { .type = NLA_NESTED },
1738         [IFLA_EVENT]            = { .type = NLA_U32 },
1739         [IFLA_GROUP]            = { .type = NLA_U32 },
1740         [IFLA_IF_NETNSID]       = { .type = NLA_S32 },
1741         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1742         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1743         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1744         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1745 };
1746
1747 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1748         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1749         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1750         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1751         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1752 };
1753
1754 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1755         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1756         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1757         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1758         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1759         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1760         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1761         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1762         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1763         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1764         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1765         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1766         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1767 };
1768
1769 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1770         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1771         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1772                                     .len = PORT_PROFILE_MAX },
1773         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1774                                       .len = PORT_UUID_MAX },
1775         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1776                                     .len = PORT_UUID_MAX },
1777         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1778         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1779
1780         /* Unused, but we need to keep it here since user space could
1781          * fill it. It's also broken with regard to NLA_BINARY use in
1782          * combination with structs.
1783          */
1784         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1785                                     .len = sizeof(struct ifla_port_vsi) },
1786 };
1787
1788 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1789         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1790         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1791         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1792         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1793 };
1794
1795 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1796 {
1797         const struct rtnl_link_ops *ops = NULL;
1798         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1799
1800         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1801                              ifla_info_policy, NULL) < 0)
1802                 return NULL;
1803
1804         if (linfo[IFLA_INFO_KIND]) {
1805                 char kind[MODULE_NAME_LEN];
1806
1807                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1808                 ops = rtnl_link_ops_get(kind);
1809         }
1810
1811         return ops;
1812 }
1813
1814 static bool link_master_filtered(struct net_device *dev, int master_idx)
1815 {
1816         struct net_device *master;
1817
1818         if (!master_idx)
1819                 return false;
1820
1821         master = netdev_master_upper_dev_get(dev);
1822         if (!master || master->ifindex != master_idx)
1823                 return true;
1824
1825         return false;
1826 }
1827
1828 static bool link_kind_filtered(const struct net_device *dev,
1829                                const struct rtnl_link_ops *kind_ops)
1830 {
1831         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1832                 return true;
1833
1834         return false;
1835 }
1836
1837 static bool link_dump_filtered(struct net_device *dev,
1838                                int master_idx,
1839                                const struct rtnl_link_ops *kind_ops)
1840 {
1841         if (link_master_filtered(dev, master_idx) ||
1842             link_kind_filtered(dev, kind_ops))
1843                 return true;
1844
1845         return false;
1846 }
1847
1848 static struct net *get_target_net(struct sock *sk, int netnsid)
1849 {
1850         struct net *net;
1851
1852         net = get_net_ns_by_id(sock_net(sk), netnsid);
1853         if (!net)
1854                 return ERR_PTR(-EINVAL);
1855
1856         /* For now, the caller is required to have CAP_NET_ADMIN in
1857          * the user namespace owning the target net ns.
1858          */
1859         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1860                 put_net(net);
1861                 return ERR_PTR(-EACCES);
1862         }
1863         return net;
1864 }
1865
1866 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1867 {
1868         struct net *net = sock_net(skb->sk);
1869         struct net *tgt_net = net;
1870         int h, s_h;
1871         int idx = 0, s_idx;
1872         struct net_device *dev;
1873         struct hlist_head *head;
1874         struct nlattr *tb[IFLA_MAX+1];
1875         u32 ext_filter_mask = 0;
1876         const struct rtnl_link_ops *kind_ops = NULL;
1877         unsigned int flags = NLM_F_MULTI;
1878         int master_idx = 0;
1879         int netnsid = -1;
1880         int err;
1881         int hdrlen;
1882
1883         s_h = cb->args[0];
1884         s_idx = cb->args[1];
1885
1886         /* A hack to preserve kernel<->userspace interface.
1887          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1888          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1889          * what iproute2 < v3.9.0 used.
1890          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1891          * attribute, its netlink message is shorter than struct ifinfomsg.
1892          */
1893         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1894                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1895
1896         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1897                         ifla_policy, NULL) >= 0) {
1898                 if (tb[IFLA_IF_NETNSID]) {
1899                         netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1900                         tgt_net = get_target_net(skb->sk, netnsid);
1901                         if (IS_ERR(tgt_net))
1902                                 return PTR_ERR(tgt_net);
1903                 }
1904
1905                 if (tb[IFLA_EXT_MASK])
1906                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1907
1908                 if (tb[IFLA_MASTER])
1909                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1910
1911                 if (tb[IFLA_LINKINFO])
1912                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1913
1914                 if (master_idx || kind_ops)
1915                         flags |= NLM_F_DUMP_FILTERED;
1916         }
1917
1918         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1919                 idx = 0;
1920                 head = &tgt_net->dev_index_head[h];
1921                 hlist_for_each_entry(dev, head, index_hlist) {
1922                         if (link_dump_filtered(dev, master_idx, kind_ops))
1923                                 goto cont;
1924                         if (idx < s_idx)
1925                                 goto cont;
1926                         err = rtnl_fill_ifinfo(skb, dev, net,
1927                                                RTM_NEWLINK,
1928                                                NETLINK_CB(cb->skb).portid,
1929                                                cb->nlh->nlmsg_seq, 0,
1930                                                flags,
1931                                                ext_filter_mask, 0, NULL, 0,
1932                                                netnsid);
1933
1934                         if (err < 0) {
1935                                 if (likely(skb->len))
1936                                         goto out;
1937
1938                                 goto out_err;
1939                         }
1940 cont:
1941                         idx++;
1942                 }
1943         }
1944 out:
1945         err = skb->len;
1946 out_err:
1947         cb->args[1] = idx;
1948         cb->args[0] = h;
1949         cb->seq = net->dev_base_seq;
1950         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1951         if (netnsid >= 0)
1952                 put_net(tgt_net);
1953
1954         return err;
1955 }
1956
1957 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1958                         struct netlink_ext_ack *exterr)
1959 {
1960         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1961 }
1962 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1963
1964 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1965 {
1966         struct net *net;
1967         /* Examine the link attributes and figure out which
1968          * network namespace we are talking about.
1969          */
1970         if (tb[IFLA_NET_NS_PID])
1971                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1972         else if (tb[IFLA_NET_NS_FD])
1973                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1974         else
1975                 net = get_net(src_net);
1976         return net;
1977 }
1978 EXPORT_SYMBOL(rtnl_link_get_net);
1979
1980 /* Figure out which network namespace we are talking about by
1981  * examining the link attributes in the following order:
1982  *
1983  * 1. IFLA_NET_NS_PID
1984  * 2. IFLA_NET_NS_FD
1985  * 3. IFLA_IF_NETNSID
1986  */
1987 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1988                                                struct nlattr *tb[])
1989 {
1990         struct net *net;
1991
1992         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1993                 return rtnl_link_get_net(src_net, tb);
1994
1995         if (!tb[IFLA_IF_NETNSID])
1996                 return get_net(src_net);
1997
1998         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1999         if (!net)
2000                 return ERR_PTR(-EINVAL);
2001
2002         return net;
2003 }
2004
2005 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2006                                              struct net *src_net,
2007                                              struct nlattr *tb[], int cap)
2008 {
2009         struct net *net;
2010
2011         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2012         if (IS_ERR(net))
2013                 return net;
2014
2015         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2016                 put_net(net);
2017                 return ERR_PTR(-EPERM);
2018         }
2019
2020         return net;
2021 }
2022
2023 /* Verify that rtnetlink requests do not pass additional properties
2024  * potentially referring to different network namespaces.
2025  */
2026 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2027                                     struct netlink_ext_ack *extack,
2028                                     bool netns_id_only)
2029 {
2030
2031         if (netns_id_only) {
2032                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2033                         return 0;
2034
2035                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2036                 return -EOPNOTSUPP;
2037         }
2038
2039         if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2040                 goto invalid_attr;
2041
2042         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
2043                 goto invalid_attr;
2044
2045         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
2046                 goto invalid_attr;
2047
2048         return 0;
2049
2050 invalid_attr:
2051         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2052         return -EINVAL;
2053 }
2054
2055 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2056 {
2057         if (dev) {
2058                 if (tb[IFLA_ADDRESS] &&
2059                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2060                         return -EINVAL;
2061
2062                 if (tb[IFLA_BROADCAST] &&
2063                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2064                         return -EINVAL;
2065         }
2066
2067         if (tb[IFLA_AF_SPEC]) {
2068                 struct nlattr *af;
2069                 int rem, err;
2070
2071                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2072                         const struct rtnl_af_ops *af_ops;
2073
2074                         rcu_read_lock();
2075                         af_ops = rtnl_af_lookup(nla_type(af));
2076                         if (!af_ops) {
2077                                 rcu_read_unlock();
2078                                 return -EAFNOSUPPORT;
2079                         }
2080
2081                         if (!af_ops->set_link_af) {
2082                                 rcu_read_unlock();
2083                                 return -EOPNOTSUPP;
2084                         }
2085
2086                         if (af_ops->validate_link_af) {
2087                                 err = af_ops->validate_link_af(dev, af);
2088                                 if (err < 0) {
2089                                         rcu_read_unlock();
2090                                         return err;
2091                                 }
2092                         }
2093
2094                         rcu_read_unlock();
2095                 }
2096         }
2097
2098         return 0;
2099 }
2100
2101 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2102                                   int guid_type)
2103 {
2104         const struct net_device_ops *ops = dev->netdev_ops;
2105
2106         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2107 }
2108
2109 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2110 {
2111         if (dev->type != ARPHRD_INFINIBAND)
2112                 return -EOPNOTSUPP;
2113
2114         return handle_infiniband_guid(dev, ivt, guid_type);
2115 }
2116
2117 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2118 {
2119         const struct net_device_ops *ops = dev->netdev_ops;
2120         int err = -EINVAL;
2121
2122         if (tb[IFLA_VF_MAC]) {
2123                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2124
2125                 err = -EOPNOTSUPP;
2126                 if (ops->ndo_set_vf_mac)
2127                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2128                                                   ivm->mac);
2129                 if (err < 0)
2130                         return err;
2131         }
2132
2133         if (tb[IFLA_VF_VLAN]) {
2134                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2135
2136                 err = -EOPNOTSUPP;
2137                 if (ops->ndo_set_vf_vlan)
2138                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2139                                                    ivv->qos,
2140                                                    htons(ETH_P_8021Q));
2141                 if (err < 0)
2142                         return err;
2143         }
2144
2145         if (tb[IFLA_VF_VLAN_LIST]) {
2146                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2147                 struct nlattr *attr;
2148                 int rem, len = 0;
2149
2150                 err = -EOPNOTSUPP;
2151                 if (!ops->ndo_set_vf_vlan)
2152                         return err;
2153
2154                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2155                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2156                             nla_len(attr) < NLA_HDRLEN) {
2157                                 return -EINVAL;
2158                         }
2159                         if (len >= MAX_VLAN_LIST_LEN)
2160                                 return -EOPNOTSUPP;
2161                         ivvl[len] = nla_data(attr);
2162
2163                         len++;
2164                 }
2165                 if (len == 0)
2166                         return -EINVAL;
2167
2168                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2169                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2170                 if (err < 0)
2171                         return err;
2172         }
2173
2174         if (tb[IFLA_VF_TX_RATE]) {
2175                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2176                 struct ifla_vf_info ivf;
2177
2178                 err = -EOPNOTSUPP;
2179                 if (ops->ndo_get_vf_config)
2180                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2181                 if (err < 0)
2182                         return err;
2183
2184                 err = -EOPNOTSUPP;
2185                 if (ops->ndo_set_vf_rate)
2186                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2187                                                    ivf.min_tx_rate,
2188                                                    ivt->rate);
2189                 if (err < 0)
2190                         return err;
2191         }
2192
2193         if (tb[IFLA_VF_RATE]) {
2194                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2195
2196                 err = -EOPNOTSUPP;
2197                 if (ops->ndo_set_vf_rate)
2198                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2199                                                    ivt->min_tx_rate,
2200                                                    ivt->max_tx_rate);
2201                 if (err < 0)
2202                         return err;
2203         }
2204
2205         if (tb[IFLA_VF_SPOOFCHK]) {
2206                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2207
2208                 err = -EOPNOTSUPP;
2209                 if (ops->ndo_set_vf_spoofchk)
2210                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2211                                                        ivs->setting);
2212                 if (err < 0)
2213                         return err;
2214         }
2215
2216         if (tb[IFLA_VF_LINK_STATE]) {
2217                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2218
2219                 err = -EOPNOTSUPP;
2220                 if (ops->ndo_set_vf_link_state)
2221                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2222                                                          ivl->link_state);
2223                 if (err < 0)
2224                         return err;
2225         }
2226
2227         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2228                 struct ifla_vf_rss_query_en *ivrssq_en;
2229
2230                 err = -EOPNOTSUPP;
2231                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2232                 if (ops->ndo_set_vf_rss_query_en)
2233                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2234                                                            ivrssq_en->setting);
2235                 if (err < 0)
2236                         return err;
2237         }
2238
2239         if (tb[IFLA_VF_TRUST]) {
2240                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2241
2242                 err = -EOPNOTSUPP;
2243                 if (ops->ndo_set_vf_trust)
2244                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2245                 if (err < 0)
2246                         return err;
2247         }
2248
2249         if (tb[IFLA_VF_IB_NODE_GUID]) {
2250                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2251
2252                 if (!ops->ndo_set_vf_guid)
2253                         return -EOPNOTSUPP;
2254
2255                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2256         }
2257
2258         if (tb[IFLA_VF_IB_PORT_GUID]) {
2259                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2260
2261                 if (!ops->ndo_set_vf_guid)
2262                         return -EOPNOTSUPP;
2263
2264                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2265         }
2266
2267         return err;
2268 }
2269
2270 static int do_set_master(struct net_device *dev, int ifindex,
2271                          struct netlink_ext_ack *extack)
2272 {
2273         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2274         const struct net_device_ops *ops;
2275         int err;
2276
2277         if (upper_dev) {
2278                 if (upper_dev->ifindex == ifindex)
2279                         return 0;
2280                 ops = upper_dev->netdev_ops;
2281                 if (ops->ndo_del_slave) {
2282                         err = ops->ndo_del_slave(upper_dev, dev);
2283                         if (err)
2284                                 return err;
2285                 } else {
2286                         return -EOPNOTSUPP;
2287                 }
2288         }
2289
2290         if (ifindex) {
2291                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2292                 if (!upper_dev)
2293                         return -EINVAL;
2294                 ops = upper_dev->netdev_ops;
2295                 if (ops->ndo_add_slave) {
2296                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2297                         if (err)
2298                                 return err;
2299                 } else {
2300                         return -EOPNOTSUPP;
2301                 }
2302         }
2303         return 0;
2304 }
2305
2306 #define DO_SETLINK_MODIFIED     0x01
2307 /* notify flag means notify + modified. */
2308 #define DO_SETLINK_NOTIFY       0x03
2309 static int do_setlink(const struct sk_buff *skb,
2310                       struct net_device *dev, struct ifinfomsg *ifm,
2311                       struct netlink_ext_ack *extack,
2312                       struct nlattr **tb, char *ifname, int status)
2313 {
2314         const struct net_device_ops *ops = dev->netdev_ops;
2315         int err;
2316
2317         err = validate_linkmsg(dev, tb);
2318         if (err < 0)
2319                 return err;
2320
2321         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2322                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2323                                                             tb, CAP_NET_ADMIN);
2324                 if (IS_ERR(net)) {
2325                         err = PTR_ERR(net);
2326                         goto errout;
2327                 }
2328
2329                 err = dev_change_net_namespace(dev, net, ifname);
2330                 put_net(net);
2331                 if (err)
2332                         goto errout;
2333                 status |= DO_SETLINK_MODIFIED;
2334         }
2335
2336         if (tb[IFLA_MAP]) {
2337                 struct rtnl_link_ifmap *u_map;
2338                 struct ifmap k_map;
2339
2340                 if (!ops->ndo_set_config) {
2341                         err = -EOPNOTSUPP;
2342                         goto errout;
2343                 }
2344
2345                 if (!netif_device_present(dev)) {
2346                         err = -ENODEV;
2347                         goto errout;
2348                 }
2349
2350                 u_map = nla_data(tb[IFLA_MAP]);
2351                 k_map.mem_start = (unsigned long) u_map->mem_start;
2352                 k_map.mem_end = (unsigned long) u_map->mem_end;
2353                 k_map.base_addr = (unsigned short) u_map->base_addr;
2354                 k_map.irq = (unsigned char) u_map->irq;
2355                 k_map.dma = (unsigned char) u_map->dma;
2356                 k_map.port = (unsigned char) u_map->port;
2357
2358                 err = ops->ndo_set_config(dev, &k_map);
2359                 if (err < 0)
2360                         goto errout;
2361
2362                 status |= DO_SETLINK_NOTIFY;
2363         }
2364
2365         if (tb[IFLA_ADDRESS]) {
2366                 struct sockaddr *sa;
2367                 int len;
2368
2369                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2370                                                   sizeof(*sa));
2371                 sa = kmalloc(len, GFP_KERNEL);
2372                 if (!sa) {
2373                         err = -ENOMEM;
2374                         goto errout;
2375                 }
2376                 sa->sa_family = dev->type;
2377                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2378                        dev->addr_len);
2379                 err = dev_set_mac_address(dev, sa);
2380                 kfree(sa);
2381                 if (err)
2382                         goto errout;
2383                 status |= DO_SETLINK_MODIFIED;
2384         }
2385
2386         if (tb[IFLA_MTU]) {
2387                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2388                 if (err < 0)
2389                         goto errout;
2390                 status |= DO_SETLINK_MODIFIED;
2391         }
2392
2393         if (tb[IFLA_GROUP]) {
2394                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2395                 status |= DO_SETLINK_NOTIFY;
2396         }
2397
2398         /*
2399          * Interface selected by interface index but interface
2400          * name provided implies that a name change has been
2401          * requested.
2402          */
2403         if (ifm->ifi_index > 0 && ifname[0]) {
2404                 err = dev_change_name(dev, ifname);
2405                 if (err < 0)
2406                         goto errout;
2407                 status |= DO_SETLINK_MODIFIED;
2408         }
2409
2410         if (tb[IFLA_IFALIAS]) {
2411                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2412                                     nla_len(tb[IFLA_IFALIAS]));
2413                 if (err < 0)
2414                         goto errout;
2415                 status |= DO_SETLINK_NOTIFY;
2416         }
2417
2418         if (tb[IFLA_BROADCAST]) {
2419                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2420                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2421         }
2422
2423         if (ifm->ifi_flags || ifm->ifi_change) {
2424                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2425                 if (err < 0)
2426                         goto errout;
2427         }
2428
2429         if (tb[IFLA_MASTER]) {
2430                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2431                 if (err)
2432                         goto errout;
2433                 status |= DO_SETLINK_MODIFIED;
2434         }
2435
2436         if (tb[IFLA_CARRIER]) {
2437                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2438                 if (err)
2439                         goto errout;
2440                 status |= DO_SETLINK_MODIFIED;
2441         }
2442
2443         if (tb[IFLA_TXQLEN]) {
2444                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2445
2446                 err = dev_change_tx_queue_len(dev, value);
2447                 if (err)
2448                         goto errout;
2449                 status |= DO_SETLINK_MODIFIED;
2450         }
2451
2452         if (tb[IFLA_GSO_MAX_SIZE]) {
2453                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2454
2455                 if (max_size > GSO_MAX_SIZE) {
2456                         err = -EINVAL;
2457                         goto errout;
2458                 }
2459
2460                 if (dev->gso_max_size ^ max_size) {
2461                         netif_set_gso_max_size(dev, max_size);
2462                         status |= DO_SETLINK_MODIFIED;
2463                 }
2464         }
2465
2466         if (tb[IFLA_GSO_MAX_SEGS]) {
2467                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2468
2469                 if (max_segs > GSO_MAX_SEGS) {
2470                         err = -EINVAL;
2471                         goto errout;
2472                 }
2473
2474                 if (dev->gso_max_segs ^ max_segs) {
2475                         dev->gso_max_segs = max_segs;
2476                         status |= DO_SETLINK_MODIFIED;
2477                 }
2478         }
2479
2480         if (tb[IFLA_OPERSTATE])
2481                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2482
2483         if (tb[IFLA_LINKMODE]) {
2484                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2485
2486                 write_lock_bh(&dev_base_lock);
2487                 if (dev->link_mode ^ value)
2488                         status |= DO_SETLINK_NOTIFY;
2489                 dev->link_mode = value;
2490                 write_unlock_bh(&dev_base_lock);
2491         }
2492
2493         if (tb[IFLA_VFINFO_LIST]) {
2494                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2495                 struct nlattr *attr;
2496                 int rem;
2497
2498                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2499                         if (nla_type(attr) != IFLA_VF_INFO ||
2500                             nla_len(attr) < NLA_HDRLEN) {
2501                                 err = -EINVAL;
2502                                 goto errout;
2503                         }
2504                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2505                                                ifla_vf_policy, NULL);
2506                         if (err < 0)
2507                                 goto errout;
2508                         err = do_setvfinfo(dev, vfinfo);
2509                         if (err < 0)
2510                                 goto errout;
2511                         status |= DO_SETLINK_NOTIFY;
2512                 }
2513         }
2514         err = 0;
2515
2516         if (tb[IFLA_VF_PORTS]) {
2517                 struct nlattr *port[IFLA_PORT_MAX+1];
2518                 struct nlattr *attr;
2519                 int vf;
2520                 int rem;
2521
2522                 err = -EOPNOTSUPP;
2523                 if (!ops->ndo_set_vf_port)
2524                         goto errout;
2525
2526                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2527                         if (nla_type(attr) != IFLA_VF_PORT ||
2528                             nla_len(attr) < NLA_HDRLEN) {
2529                                 err = -EINVAL;
2530                                 goto errout;
2531                         }
2532                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2533                                                ifla_port_policy, NULL);
2534                         if (err < 0)
2535                                 goto errout;
2536                         if (!port[IFLA_PORT_VF]) {
2537                                 err = -EOPNOTSUPP;
2538                                 goto errout;
2539                         }
2540                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2541                         err = ops->ndo_set_vf_port(dev, vf, port);
2542                         if (err < 0)
2543                                 goto errout;
2544                         status |= DO_SETLINK_NOTIFY;
2545                 }
2546         }
2547         err = 0;
2548
2549         if (tb[IFLA_PORT_SELF]) {
2550                 struct nlattr *port[IFLA_PORT_MAX+1];
2551
2552                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2553                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2554                                        NULL);
2555                 if (err < 0)
2556                         goto errout;
2557
2558                 err = -EOPNOTSUPP;
2559                 if (ops->ndo_set_vf_port)
2560                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2561                 if (err < 0)
2562                         goto errout;
2563                 status |= DO_SETLINK_NOTIFY;
2564         }
2565
2566         if (tb[IFLA_AF_SPEC]) {
2567                 struct nlattr *af;
2568                 int rem;
2569
2570                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2571                         const struct rtnl_af_ops *af_ops;
2572
2573                         rcu_read_lock();
2574
2575                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2576
2577                         err = af_ops->set_link_af(dev, af);
2578                         if (err < 0) {
2579                                 rcu_read_unlock();
2580                                 goto errout;
2581                         }
2582
2583                         rcu_read_unlock();
2584                         status |= DO_SETLINK_NOTIFY;
2585                 }
2586         }
2587         err = 0;
2588
2589         if (tb[IFLA_PROTO_DOWN]) {
2590                 err = dev_change_proto_down(dev,
2591                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2592                 if (err)
2593                         goto errout;
2594                 status |= DO_SETLINK_NOTIFY;
2595         }
2596
2597         if (tb[IFLA_XDP]) {
2598                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2599                 u32 xdp_flags = 0;
2600
2601                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2602                                        ifla_xdp_policy, NULL);
2603                 if (err < 0)
2604                         goto errout;
2605
2606                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2607                         err = -EINVAL;
2608                         goto errout;
2609                 }
2610
2611                 if (xdp[IFLA_XDP_FLAGS]) {
2612                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2613                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2614                                 err = -EINVAL;
2615                                 goto errout;
2616                         }
2617                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2618                                 err = -EINVAL;
2619                                 goto errout;
2620                         }
2621                 }
2622
2623                 if (xdp[IFLA_XDP_FD]) {
2624                         err = dev_change_xdp_fd(dev, extack,
2625                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2626                                                 xdp_flags);
2627                         if (err)
2628                                 goto errout;
2629                         status |= DO_SETLINK_NOTIFY;
2630                 }
2631         }
2632
2633 errout:
2634         if (status & DO_SETLINK_MODIFIED) {
2635                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2636                         netdev_state_change(dev);
2637
2638                 if (err < 0)
2639                         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",
2640                                              dev->name);
2641         }
2642
2643         return err;
2644 }
2645
2646 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2647                         struct netlink_ext_ack *extack)
2648 {
2649         struct net *net = sock_net(skb->sk);
2650         struct ifinfomsg *ifm;
2651         struct net_device *dev;
2652         int err;
2653         struct nlattr *tb[IFLA_MAX+1];
2654         char ifname[IFNAMSIZ];
2655
2656         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2657                           extack);
2658         if (err < 0)
2659                 goto errout;
2660
2661         err = rtnl_ensure_unique_netns(tb, extack, false);
2662         if (err < 0)
2663                 goto errout;
2664
2665         if (tb[IFLA_IFNAME])
2666                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2667         else
2668                 ifname[0] = '\0';
2669
2670         err = -EINVAL;
2671         ifm = nlmsg_data(nlh);
2672         if (ifm->ifi_index > 0)
2673                 dev = __dev_get_by_index(net, ifm->ifi_index);
2674         else if (tb[IFLA_IFNAME])
2675                 dev = __dev_get_by_name(net, ifname);
2676         else
2677                 goto errout;
2678
2679         if (dev == NULL) {
2680                 err = -ENODEV;
2681                 goto errout;
2682         }
2683
2684         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2685 errout:
2686         return err;
2687 }
2688
2689 static int rtnl_group_dellink(const struct net *net, int group)
2690 {
2691         struct net_device *dev, *aux;
2692         LIST_HEAD(list_kill);
2693         bool found = false;
2694
2695         if (!group)
2696                 return -EPERM;
2697
2698         for_each_netdev(net, dev) {
2699                 if (dev->group == group) {
2700                         const struct rtnl_link_ops *ops;
2701
2702                         found = true;
2703                         ops = dev->rtnl_link_ops;
2704                         if (!ops || !ops->dellink)
2705                                 return -EOPNOTSUPP;
2706                 }
2707         }
2708
2709         if (!found)
2710                 return -ENODEV;
2711
2712         for_each_netdev_safe(net, dev, aux) {
2713                 if (dev->group == group) {
2714                         const struct rtnl_link_ops *ops;
2715
2716                         ops = dev->rtnl_link_ops;
2717                         ops->dellink(dev, &list_kill);
2718                 }
2719         }
2720         unregister_netdevice_many(&list_kill);
2721
2722         return 0;
2723 }
2724
2725 int rtnl_delete_link(struct net_device *dev)
2726 {
2727         const struct rtnl_link_ops *ops;
2728         LIST_HEAD(list_kill);
2729
2730         ops = dev->rtnl_link_ops;
2731         if (!ops || !ops->dellink)
2732                 return -EOPNOTSUPP;
2733
2734         ops->dellink(dev, &list_kill);
2735         unregister_netdevice_many(&list_kill);
2736
2737         return 0;
2738 }
2739 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2740
2741 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2742                         struct netlink_ext_ack *extack)
2743 {
2744         struct net *net = sock_net(skb->sk);
2745         struct net *tgt_net = net;
2746         struct net_device *dev = NULL;
2747         struct ifinfomsg *ifm;
2748         char ifname[IFNAMSIZ];
2749         struct nlattr *tb[IFLA_MAX+1];
2750         int err;
2751         int netnsid = -1;
2752
2753         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2754         if (err < 0)
2755                 return err;
2756
2757         err = rtnl_ensure_unique_netns(tb, extack, true);
2758         if (err < 0)
2759                 return err;
2760
2761         if (tb[IFLA_IFNAME])
2762                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2763
2764         if (tb[IFLA_IF_NETNSID]) {
2765                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2766                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2767                 if (IS_ERR(tgt_net))
2768                         return PTR_ERR(tgt_net);
2769         }
2770
2771         err = -EINVAL;
2772         ifm = nlmsg_data(nlh);
2773         if (ifm->ifi_index > 0)
2774                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2775         else if (tb[IFLA_IFNAME])
2776                 dev = __dev_get_by_name(tgt_net, ifname);
2777         else if (tb[IFLA_GROUP])
2778                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2779         else
2780                 goto out;
2781
2782         if (!dev) {
2783                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2784                         err = -ENODEV;
2785
2786                 goto out;
2787         }
2788
2789         err = rtnl_delete_link(dev);
2790
2791 out:
2792         if (netnsid >= 0)
2793                 put_net(tgt_net);
2794
2795         return err;
2796 }
2797
2798 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2799 {
2800         unsigned int old_flags;
2801         int err;
2802
2803         old_flags = dev->flags;
2804         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2805                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2806                 if (err < 0)
2807                         return err;
2808         }
2809
2810         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2811                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2812         } else {
2813                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2814                 __dev_notify_flags(dev, old_flags, ~0U);
2815         }
2816         return 0;
2817 }
2818 EXPORT_SYMBOL(rtnl_configure_link);
2819
2820 struct net_device *rtnl_create_link(struct net *net,
2821         const char *ifname, unsigned char name_assign_type,
2822         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2823 {
2824         struct net_device *dev;
2825         unsigned int num_tx_queues = 1;
2826         unsigned int num_rx_queues = 1;
2827
2828         if (tb[IFLA_NUM_TX_QUEUES])
2829                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2830         else if (ops->get_num_tx_queues)
2831                 num_tx_queues = ops->get_num_tx_queues();
2832
2833         if (tb[IFLA_NUM_RX_QUEUES])
2834                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2835         else if (ops->get_num_rx_queues)
2836                 num_rx_queues = ops->get_num_rx_queues();
2837
2838         if (num_tx_queues < 1 || num_tx_queues > 4096)
2839                 return ERR_PTR(-EINVAL);
2840
2841         if (num_rx_queues < 1 || num_rx_queues > 4096)
2842                 return ERR_PTR(-EINVAL);
2843
2844         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2845                                ops->setup, num_tx_queues, num_rx_queues);
2846         if (!dev)
2847                 return ERR_PTR(-ENOMEM);
2848
2849         dev_net_set(dev, net);
2850         dev->rtnl_link_ops = ops;
2851         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2852
2853         if (tb[IFLA_MTU])
2854                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2855         if (tb[IFLA_ADDRESS]) {
2856                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2857                                 nla_len(tb[IFLA_ADDRESS]));
2858                 dev->addr_assign_type = NET_ADDR_SET;
2859         }
2860         if (tb[IFLA_BROADCAST])
2861                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2862                                 nla_len(tb[IFLA_BROADCAST]));
2863         if (tb[IFLA_TXQLEN])
2864                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2865         if (tb[IFLA_OPERSTATE])
2866                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2867         if (tb[IFLA_LINKMODE])
2868                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2869         if (tb[IFLA_GROUP])
2870                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2871         if (tb[IFLA_GSO_MAX_SIZE])
2872                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2873         if (tb[IFLA_GSO_MAX_SEGS])
2874                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2875
2876         return dev;
2877 }
2878 EXPORT_SYMBOL(rtnl_create_link);
2879
2880 static int rtnl_group_changelink(const struct sk_buff *skb,
2881                 struct net *net, int group,
2882                 struct ifinfomsg *ifm,
2883                 struct netlink_ext_ack *extack,
2884                 struct nlattr **tb)
2885 {
2886         struct net_device *dev, *aux;
2887         int err;
2888
2889         for_each_netdev_safe(net, dev, aux) {
2890                 if (dev->group == group) {
2891                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2892                         if (err < 0)
2893                                 return err;
2894                 }
2895         }
2896
2897         return 0;
2898 }
2899
2900 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2901                         struct netlink_ext_ack *extack)
2902 {
2903         struct net *net = sock_net(skb->sk);
2904         const struct rtnl_link_ops *ops;
2905         const struct rtnl_link_ops *m_ops = NULL;
2906         struct net_device *dev;
2907         struct net_device *master_dev = NULL;
2908         struct ifinfomsg *ifm;
2909         char kind[MODULE_NAME_LEN];
2910         char ifname[IFNAMSIZ];
2911         struct nlattr *tb[IFLA_MAX+1];
2912         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2913         unsigned char name_assign_type = NET_NAME_USER;
2914         int err;
2915
2916 #ifdef CONFIG_MODULES
2917 replay:
2918 #endif
2919         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2920         if (err < 0)
2921                 return err;
2922
2923         err = rtnl_ensure_unique_netns(tb, extack, false);
2924         if (err < 0)
2925                 return err;
2926
2927         if (tb[IFLA_IFNAME])
2928                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2929         else
2930                 ifname[0] = '\0';
2931
2932         ifm = nlmsg_data(nlh);
2933         if (ifm->ifi_index > 0)
2934                 dev = __dev_get_by_index(net, ifm->ifi_index);
2935         else {
2936                 if (ifname[0])
2937                         dev = __dev_get_by_name(net, ifname);
2938                 else
2939                         dev = NULL;
2940         }
2941
2942         if (dev) {
2943                 master_dev = netdev_master_upper_dev_get(dev);
2944                 if (master_dev)
2945                         m_ops = master_dev->rtnl_link_ops;
2946         }
2947
2948         err = validate_linkmsg(dev, tb);
2949         if (err < 0)
2950                 return err;
2951
2952         if (tb[IFLA_LINKINFO]) {
2953                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2954                                        tb[IFLA_LINKINFO], ifla_info_policy,
2955                                        NULL);
2956                 if (err < 0)
2957                         return err;
2958         } else
2959                 memset(linkinfo, 0, sizeof(linkinfo));
2960
2961         if (linkinfo[IFLA_INFO_KIND]) {
2962                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2963                 ops = rtnl_link_ops_get(kind);
2964         } else {
2965                 kind[0] = '\0';
2966                 ops = NULL;
2967         }
2968
2969         if (1) {
2970                 struct nlattr *attr[RTNL_MAX_TYPE + 1];
2971                 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2972                 struct nlattr **data = NULL;
2973                 struct nlattr **slave_data = NULL;
2974                 struct net *dest_net, *link_net = NULL;
2975
2976                 if (ops) {
2977                         if (ops->maxtype > RTNL_MAX_TYPE)
2978                                 return -EINVAL;
2979
2980                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2981                                 err = nla_parse_nested(attr, ops->maxtype,
2982                                                        linkinfo[IFLA_INFO_DATA],
2983                                                        ops->policy, NULL);
2984                                 if (err < 0)
2985                                         return err;
2986                                 data = attr;
2987                         }
2988                         if (ops->validate) {
2989                                 err = ops->validate(tb, data, extack);
2990                                 if (err < 0)
2991                                         return err;
2992                         }
2993                 }
2994
2995                 if (m_ops) {
2996                         if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
2997                                 return -EINVAL;
2998
2999                         if (m_ops->slave_maxtype &&
3000                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
3001                                 err = nla_parse_nested(slave_attr,
3002                                                        m_ops->slave_maxtype,
3003                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
3004                                                        m_ops->slave_policy,
3005                                                        NULL);
3006                                 if (err < 0)
3007                                         return err;
3008                                 slave_data = slave_attr;
3009                         }
3010                 }
3011
3012                 if (dev) {
3013                         int status = 0;
3014
3015                         if (nlh->nlmsg_flags & NLM_F_EXCL)
3016                                 return -EEXIST;
3017                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
3018                                 return -EOPNOTSUPP;
3019
3020                         if (linkinfo[IFLA_INFO_DATA]) {
3021                                 if (!ops || ops != dev->rtnl_link_ops ||
3022                                     !ops->changelink)
3023                                         return -EOPNOTSUPP;
3024
3025                                 err = ops->changelink(dev, tb, data, extack);
3026                                 if (err < 0)
3027                                         return err;
3028                                 status |= DO_SETLINK_NOTIFY;
3029                         }
3030
3031                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3032                                 if (!m_ops || !m_ops->slave_changelink)
3033                                         return -EOPNOTSUPP;
3034
3035                                 err = m_ops->slave_changelink(master_dev, dev,
3036                                                               tb, slave_data,
3037                                                               extack);
3038                                 if (err < 0)
3039                                         return err;
3040                                 status |= DO_SETLINK_NOTIFY;
3041                         }
3042
3043                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
3044                                           status);
3045                 }
3046
3047                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3048                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3049                                 return rtnl_group_changelink(skb, net,
3050                                                 nla_get_u32(tb[IFLA_GROUP]),
3051                                                 ifm, extack, tb);
3052                         return -ENODEV;
3053                 }
3054
3055                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3056                         return -EOPNOTSUPP;
3057
3058                 if (!ops) {
3059 #ifdef CONFIG_MODULES
3060                         if (kind[0]) {
3061                                 __rtnl_unlock();
3062                                 request_module("rtnl-link-%s", kind);
3063                                 rtnl_lock();
3064                                 ops = rtnl_link_ops_get(kind);
3065                                 if (ops)
3066                                         goto replay;
3067                         }
3068 #endif
3069                         return -EOPNOTSUPP;
3070                 }
3071
3072                 if (!ops->setup)
3073                         return -EOPNOTSUPP;
3074
3075                 if (!ifname[0]) {
3076                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3077                         name_assign_type = NET_NAME_ENUM;
3078                 }
3079
3080                 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3081                 if (IS_ERR(dest_net))
3082                         return PTR_ERR(dest_net);
3083
3084                 if (tb[IFLA_LINK_NETNSID]) {
3085                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3086
3087                         link_net = get_net_ns_by_id(dest_net, id);
3088                         if (!link_net) {
3089                                 err =  -EINVAL;
3090                                 goto out;
3091                         }
3092                         err = -EPERM;
3093                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3094                                 goto out;
3095                 }
3096
3097                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3098                                        name_assign_type, ops, tb);
3099                 if (IS_ERR(dev)) {
3100                         err = PTR_ERR(dev);
3101                         goto out;
3102                 }
3103
3104                 dev->ifindex = ifm->ifi_index;
3105
3106                 if (ops->newlink) {
3107                         err = ops->newlink(link_net ? : net, dev, tb, data,
3108                                            extack);
3109                         /* Drivers should call free_netdev() in ->destructor
3110                          * and unregister it on failure after registration
3111                          * so that device could be finally freed in rtnl_unlock.
3112                          */
3113                         if (err < 0) {
3114                                 /* If device is not registered at all, free it now */
3115                                 if (dev->reg_state == NETREG_UNINITIALIZED)
3116                                         free_netdev(dev);
3117                                 goto out;
3118                         }
3119                 } else {
3120                         err = register_netdevice(dev);
3121                         if (err < 0) {
3122                                 free_netdev(dev);
3123                                 goto out;
3124                         }
3125                 }
3126                 err = rtnl_configure_link(dev, ifm);
3127                 if (err < 0)
3128                         goto out_unregister;
3129                 if (link_net) {
3130                         err = dev_change_net_namespace(dev, dest_net, ifname);
3131                         if (err < 0)
3132                                 goto out_unregister;
3133                 }
3134                 if (tb[IFLA_MASTER]) {
3135                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
3136                                             extack);
3137                         if (err)
3138                                 goto out_unregister;
3139                 }
3140 out:
3141                 if (link_net)
3142                         put_net(link_net);
3143                 put_net(dest_net);
3144                 return err;
3145 out_unregister:
3146                 if (ops->newlink) {
3147                         LIST_HEAD(list_kill);
3148
3149                         ops->dellink(dev, &list_kill);
3150                         unregister_netdevice_many(&list_kill);
3151                 } else {
3152                         unregister_netdevice(dev);
3153                 }
3154                 goto out;
3155         }
3156 }
3157
3158 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3159                         struct netlink_ext_ack *extack)
3160 {
3161         struct net *net = sock_net(skb->sk);
3162         struct net *tgt_net = net;
3163         struct ifinfomsg *ifm;
3164         char ifname[IFNAMSIZ];
3165         struct nlattr *tb[IFLA_MAX+1];
3166         struct net_device *dev = NULL;
3167         struct sk_buff *nskb;
3168         int netnsid = -1;
3169         int err;
3170         u32 ext_filter_mask = 0;
3171
3172         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3173         if (err < 0)
3174                 return err;
3175
3176         err = rtnl_ensure_unique_netns(tb, extack, true);
3177         if (err < 0)
3178                 return err;
3179
3180         if (tb[IFLA_IF_NETNSID]) {
3181                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3182                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3183                 if (IS_ERR(tgt_net))
3184                         return PTR_ERR(tgt_net);
3185         }
3186
3187         if (tb[IFLA_IFNAME])
3188                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3189
3190         if (tb[IFLA_EXT_MASK])
3191                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3192
3193         err = -EINVAL;
3194         ifm = nlmsg_data(nlh);
3195         if (ifm->ifi_index > 0)
3196                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3197         else if (tb[IFLA_IFNAME])
3198                 dev = __dev_get_by_name(tgt_net, ifname);
3199         else
3200                 goto out;
3201
3202         err = -ENODEV;
3203         if (dev == NULL)
3204                 goto out;
3205
3206         err = -ENOBUFS;
3207         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3208         if (nskb == NULL)
3209                 goto out;
3210
3211         err = rtnl_fill_ifinfo(nskb, dev, net,
3212                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3213                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3214                                0, NULL, 0, netnsid);
3215         if (err < 0) {
3216                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3217                 WARN_ON(err == -EMSGSIZE);
3218                 kfree_skb(nskb);
3219         } else
3220                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3221 out:
3222         if (netnsid >= 0)
3223                 put_net(tgt_net);
3224
3225         return err;
3226 }
3227
3228 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3229 {
3230         struct net *net = sock_net(skb->sk);
3231         struct net_device *dev;
3232         struct nlattr *tb[IFLA_MAX+1];
3233         u32 ext_filter_mask = 0;
3234         u16 min_ifinfo_dump_size = 0;
3235         int hdrlen;
3236
3237         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3238         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3239                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3240
3241         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3242                 if (tb[IFLA_EXT_MASK])
3243                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3244         }
3245
3246         if (!ext_filter_mask)
3247                 return NLMSG_GOODSIZE;
3248         /*
3249          * traverse the list of net devices and compute the minimum
3250          * buffer size based upon the filter mask.
3251          */
3252         rcu_read_lock();
3253         for_each_netdev_rcu(net, dev) {
3254                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3255                                              if_nlmsg_size(dev,
3256                                                            ext_filter_mask));
3257         }
3258         rcu_read_unlock();
3259
3260         return nlmsg_total_size(min_ifinfo_dump_size);
3261 }
3262
3263 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3264 {
3265         int idx;
3266         int s_idx = cb->family;
3267
3268         if (s_idx == 0)
3269                 s_idx = 1;
3270
3271         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3272                 struct rtnl_link **tab;
3273                 int type = cb->nlh->nlmsg_type-RTM_BASE;
3274                 struct rtnl_link *link;
3275                 rtnl_dumpit_func dumpit;
3276
3277                 if (idx < s_idx || idx == PF_PACKET)
3278                         continue;
3279
3280                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3281                         continue;
3282
3283                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3284                 if (!tab)
3285                         continue;
3286
3287                 link = tab[type];
3288                 if (!link)
3289                         continue;
3290
3291                 dumpit = link->dumpit;
3292                 if (!dumpit)
3293                         continue;
3294
3295                 if (idx > s_idx) {
3296                         memset(&cb->args[0], 0, sizeof(cb->args));
3297                         cb->prev_seq = 0;
3298                         cb->seq = 0;
3299                 }
3300                 if (dumpit(skb, cb))
3301                         break;
3302         }
3303         cb->family = idx;
3304
3305         return skb->len;
3306 }
3307
3308 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3309                                        unsigned int change,
3310                                        u32 event, gfp_t flags, int *new_nsid,
3311                                        int new_ifindex)
3312 {
3313         struct net *net = dev_net(dev);
3314         struct sk_buff *skb;
3315         int err = -ENOBUFS;
3316         size_t if_info_size;
3317
3318         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3319         if (skb == NULL)
3320                 goto errout;
3321
3322         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3323                                type, 0, 0, change, 0, 0, event,
3324                                new_nsid, new_ifindex, -1);
3325         if (err < 0) {
3326                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3327                 WARN_ON(err == -EMSGSIZE);
3328                 kfree_skb(skb);
3329                 goto errout;
3330         }
3331         return skb;
3332 errout:
3333         if (err < 0)
3334                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3335         return NULL;
3336 }
3337
3338 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3339 {
3340         struct net *net = dev_net(dev);
3341
3342         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3343 }
3344
3345 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3346                                unsigned int change, u32 event,
3347                                gfp_t flags, int *new_nsid, int new_ifindex)
3348 {
3349         struct sk_buff *skb;
3350
3351         if (dev->reg_state != NETREG_REGISTERED)
3352                 return;
3353
3354         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3355                                      new_ifindex);
3356         if (skb)
3357                 rtmsg_ifinfo_send(skb, dev, flags);
3358 }
3359
3360 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3361                   gfp_t flags)
3362 {
3363         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3364                            NULL, 0);
3365 }
3366
3367 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3368                          gfp_t flags, int *new_nsid, int new_ifindex)
3369 {
3370         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3371                            new_nsid, new_ifindex);
3372 }
3373
3374 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3375                                    struct net_device *dev,
3376                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3377                                    int type, unsigned int flags,
3378                                    int nlflags, u16 ndm_state)
3379 {
3380         struct nlmsghdr *nlh;
3381         struct ndmsg *ndm;
3382
3383         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3384         if (!nlh)
3385                 return -EMSGSIZE;
3386
3387         ndm = nlmsg_data(nlh);
3388         ndm->ndm_family  = AF_BRIDGE;
3389         ndm->ndm_pad1    = 0;
3390         ndm->ndm_pad2    = 0;
3391         ndm->ndm_flags   = flags;
3392         ndm->ndm_type    = 0;
3393         ndm->ndm_ifindex = dev->ifindex;
3394         ndm->ndm_state   = ndm_state;
3395
3396         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3397                 goto nla_put_failure;
3398         if (vid)
3399                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3400                         goto nla_put_failure;
3401
3402         nlmsg_end(skb, nlh);
3403         return 0;
3404
3405 nla_put_failure:
3406         nlmsg_cancel(skb, nlh);
3407         return -EMSGSIZE;
3408 }
3409
3410 static inline size_t rtnl_fdb_nlmsg_size(void)
3411 {
3412         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3413                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3414                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3415                0;
3416 }
3417
3418 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3419                             u16 ndm_state)
3420 {
3421         struct net *net = dev_net(dev);
3422         struct sk_buff *skb;
3423         int err = -ENOBUFS;
3424
3425         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3426         if (!skb)
3427                 goto errout;
3428
3429         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3430                                       0, 0, type, NTF_SELF, 0, ndm_state);
3431         if (err < 0) {
3432                 kfree_skb(skb);
3433                 goto errout;
3434         }
3435
3436         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3437         return;
3438 errout:
3439         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3440 }
3441
3442 /**
3443  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3444  */
3445 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3446                      struct nlattr *tb[],
3447                      struct net_device *dev,
3448                      const unsigned char *addr, u16 vid,
3449                      u16 flags)
3450 {
3451         int err = -EINVAL;
3452
3453         /* If aging addresses are supported device will need to
3454          * implement its own handler for this.
3455          */
3456         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3457                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3458                 return err;
3459         }
3460
3461         if (vid) {
3462                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3463                 return err;
3464         }
3465
3466         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3467                 err = dev_uc_add_excl(dev, addr);
3468         else if (is_multicast_ether_addr(addr))
3469                 err = dev_mc_add_excl(dev, addr);
3470
3471         /* Only return duplicate errors if NLM_F_EXCL is set */
3472         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3473                 err = 0;
3474
3475         return err;
3476 }
3477 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3478
3479 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3480                          struct netlink_ext_ack *extack)
3481 {
3482         u16 vid = 0;
3483
3484         if (vlan_attr) {
3485                 if (nla_len(vlan_attr) != sizeof(u16)) {
3486                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3487                         return -EINVAL;
3488                 }
3489
3490                 vid = nla_get_u16(vlan_attr);
3491
3492                 if (!vid || vid >= VLAN_VID_MASK) {
3493                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3494                         return -EINVAL;
3495                 }
3496         }
3497         *p_vid = vid;
3498         return 0;
3499 }
3500
3501 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3502                         struct netlink_ext_ack *extack)
3503 {
3504         struct net *net = sock_net(skb->sk);
3505         struct ndmsg *ndm;
3506         struct nlattr *tb[NDA_MAX+1];
3507         struct net_device *dev;
3508         u8 *addr;
3509         u16 vid;
3510         int err;
3511
3512         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3513         if (err < 0)
3514                 return err;
3515
3516         ndm = nlmsg_data(nlh);
3517         if (ndm->ndm_ifindex == 0) {
3518                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3519                 return -EINVAL;
3520         }
3521
3522         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3523         if (dev == NULL) {
3524                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3525                 return -ENODEV;
3526         }
3527
3528         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3529                 NL_SET_ERR_MSG(extack, "invalid address");
3530                 return -EINVAL;
3531         }
3532
3533         addr = nla_data(tb[NDA_LLADDR]);
3534
3535         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3536         if (err)
3537                 return err;
3538
3539         err = -EOPNOTSUPP;
3540
3541         /* Support fdb on master device the net/bridge default case */
3542         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3543             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3544                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3545                 const struct net_device_ops *ops = br_dev->netdev_ops;
3546
3547                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3548                                        nlh->nlmsg_flags);
3549                 if (err)
3550                         goto out;
3551                 else
3552                         ndm->ndm_flags &= ~NTF_MASTER;
3553         }
3554
3555         /* Embedded bridge, macvlan, and any other device support */
3556         if ((ndm->ndm_flags & NTF_SELF)) {
3557                 if (dev->netdev_ops->ndo_fdb_add)
3558                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3559                                                            vid,
3560                                                            nlh->nlmsg_flags);
3561                 else
3562                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3563                                                nlh->nlmsg_flags);
3564
3565                 if (!err) {
3566                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3567                                         ndm->ndm_state);
3568                         ndm->ndm_flags &= ~NTF_SELF;
3569                 }
3570         }
3571 out:
3572         return err;
3573 }
3574
3575 /**
3576  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3577  */
3578 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3579                      struct nlattr *tb[],
3580                      struct net_device *dev,
3581                      const unsigned char *addr, u16 vid)
3582 {
3583         int err = -EINVAL;
3584
3585         /* If aging addresses are supported device will need to
3586          * implement its own handler for this.
3587          */
3588         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3589                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3590                 return err;
3591         }
3592
3593         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3594                 err = dev_uc_del(dev, addr);
3595         else if (is_multicast_ether_addr(addr))
3596                 err = dev_mc_del(dev, addr);
3597
3598         return err;
3599 }
3600 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3601
3602 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3603                         struct netlink_ext_ack *extack)
3604 {
3605         struct net *net = sock_net(skb->sk);
3606         struct ndmsg *ndm;
3607         struct nlattr *tb[NDA_MAX+1];
3608         struct net_device *dev;
3609         int err = -EINVAL;
3610         __u8 *addr;
3611         u16 vid;
3612
3613         if (!netlink_capable(skb, CAP_NET_ADMIN))
3614                 return -EPERM;
3615
3616         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3617         if (err < 0)
3618                 return err;
3619
3620         ndm = nlmsg_data(nlh);
3621         if (ndm->ndm_ifindex == 0) {
3622                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3623                 return -EINVAL;
3624         }
3625
3626         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3627         if (dev == NULL) {
3628                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3629                 return -ENODEV;
3630         }
3631
3632         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3633                 NL_SET_ERR_MSG(extack, "invalid address");
3634                 return -EINVAL;
3635         }
3636
3637         addr = nla_data(tb[NDA_LLADDR]);
3638
3639         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3640         if (err)
3641                 return err;
3642
3643         err = -EOPNOTSUPP;
3644
3645         /* Support fdb on master device the net/bridge default case */
3646         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3647             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3648                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3649                 const struct net_device_ops *ops = br_dev->netdev_ops;
3650
3651                 if (ops->ndo_fdb_del)
3652                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3653
3654                 if (err)
3655                         goto out;
3656                 else
3657                         ndm->ndm_flags &= ~NTF_MASTER;
3658         }
3659
3660         /* Embedded bridge, macvlan, and any other device support */
3661         if (ndm->ndm_flags & NTF_SELF) {
3662                 if (dev->netdev_ops->ndo_fdb_del)
3663                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3664                                                            vid);
3665                 else
3666                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3667
3668                 if (!err) {
3669                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3670                                         ndm->ndm_state);
3671                         ndm->ndm_flags &= ~NTF_SELF;
3672                 }
3673         }
3674 out:
3675         return err;
3676 }
3677
3678 static int nlmsg_populate_fdb(struct sk_buff *skb,
3679                               struct netlink_callback *cb,
3680                               struct net_device *dev,
3681                               int *idx,
3682                               struct netdev_hw_addr_list *list)
3683 {
3684         struct netdev_hw_addr *ha;
3685         int err;
3686         u32 portid, seq;
3687
3688         portid = NETLINK_CB(cb->skb).portid;
3689         seq = cb->nlh->nlmsg_seq;
3690
3691         list_for_each_entry(ha, &list->list, list) {
3692                 if (*idx < cb->args[2])
3693                         goto skip;
3694
3695                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3696                                               portid, seq,
3697                                               RTM_NEWNEIGH, NTF_SELF,
3698                                               NLM_F_MULTI, NUD_PERMANENT);
3699                 if (err < 0)
3700                         return err;
3701 skip:
3702                 *idx += 1;
3703         }
3704         return 0;
3705 }
3706
3707 /**
3708  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3709  * @nlh: netlink message header
3710  * @dev: netdevice
3711  *
3712  * Default netdevice operation to dump the existing unicast address list.
3713  * Returns number of addresses from list put in skb.
3714  */
3715 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3716                       struct netlink_callback *cb,
3717                       struct net_device *dev,
3718                       struct net_device *filter_dev,
3719                       int *idx)
3720 {
3721         int err;
3722
3723         netif_addr_lock_bh(dev);
3724         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3725         if (err)
3726                 goto out;
3727         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3728 out:
3729         netif_addr_unlock_bh(dev);
3730         return err;
3731 }
3732 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3733
3734 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3735 {
3736         struct net_device *dev;
3737         struct nlattr *tb[IFLA_MAX+1];
3738         struct net_device *br_dev = NULL;
3739         const struct net_device_ops *ops = NULL;
3740         const struct net_device_ops *cops = NULL;
3741         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3742         struct net *net = sock_net(skb->sk);
3743         struct hlist_head *head;
3744         int brport_idx = 0;
3745         int br_idx = 0;
3746         int h, s_h;
3747         int idx = 0, s_idx;
3748         int err = 0;
3749         int fidx = 0;
3750
3751         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3752                           IFLA_MAX, ifla_policy, NULL);
3753         if (err < 0) {
3754                 return -EINVAL;
3755         } else if (err == 0) {
3756                 if (tb[IFLA_MASTER])
3757                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3758         }
3759
3760         brport_idx = ifm->ifi_index;
3761
3762         if (br_idx) {
3763                 br_dev = __dev_get_by_index(net, br_idx);
3764                 if (!br_dev)
3765                         return -ENODEV;
3766
3767                 ops = br_dev->netdev_ops;
3768         }
3769
3770         s_h = cb->args[0];
3771         s_idx = cb->args[1];
3772
3773         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3774                 idx = 0;
3775                 head = &net->dev_index_head[h];
3776                 hlist_for_each_entry(dev, head, index_hlist) {
3777
3778                         if (brport_idx && (dev->ifindex != brport_idx))
3779                                 continue;
3780
3781                         if (!br_idx) { /* user did not specify a specific bridge */
3782                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3783                                         br_dev = netdev_master_upper_dev_get(dev);
3784                                         cops = br_dev->netdev_ops;
3785                                 }
3786                         } else {
3787                                 if (dev != br_dev &&
3788                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3789                                         continue;
3790
3791                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3792                                     !(dev->priv_flags & IFF_EBRIDGE))
3793                                         continue;
3794                                 cops = ops;
3795                         }
3796
3797                         if (idx < s_idx)
3798                                 goto cont;
3799
3800                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3801                                 if (cops && cops->ndo_fdb_dump) {
3802                                         err = cops->ndo_fdb_dump(skb, cb,
3803                                                                 br_dev, dev,
3804                                                                 &fidx);
3805                                         if (err == -EMSGSIZE)
3806                                                 goto out;
3807                                 }
3808                         }
3809
3810                         if (dev->netdev_ops->ndo_fdb_dump)
3811                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3812                                                                     dev, NULL,
3813                                                                     &fidx);
3814                         else
3815                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3816                                                         &fidx);
3817                         if (err == -EMSGSIZE)
3818                                 goto out;
3819
3820                         cops = NULL;
3821
3822                         /* reset fdb offset to 0 for rest of the interfaces */
3823                         cb->args[2] = 0;
3824                         fidx = 0;
3825 cont:
3826                         idx++;
3827                 }
3828         }
3829
3830 out:
3831         cb->args[0] = h;
3832         cb->args[1] = idx;
3833         cb->args[2] = fidx;
3834
3835         return skb->len;
3836 }
3837
3838 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3839                                unsigned int attrnum, unsigned int flag)
3840 {
3841         if (mask & flag)
3842                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3843         return 0;
3844 }
3845
3846 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3847                             struct net_device *dev, u16 mode,
3848                             u32 flags, u32 mask, int nlflags,
3849                             u32 filter_mask,
3850                             int (*vlan_fill)(struct sk_buff *skb,
3851                                              struct net_device *dev,
3852                                              u32 filter_mask))
3853 {
3854         struct nlmsghdr *nlh;
3855         struct ifinfomsg *ifm;
3856         struct nlattr *br_afspec;
3857         struct nlattr *protinfo;
3858         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3859         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3860         int err = 0;
3861
3862         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3863         if (nlh == NULL)
3864                 return -EMSGSIZE;
3865
3866         ifm = nlmsg_data(nlh);
3867         ifm->ifi_family = AF_BRIDGE;
3868         ifm->__ifi_pad = 0;
3869         ifm->ifi_type = dev->type;
3870         ifm->ifi_index = dev->ifindex;
3871         ifm->ifi_flags = dev_get_flags(dev);
3872         ifm->ifi_change = 0;
3873
3874
3875         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3876             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3877             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3878             (br_dev &&
3879              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3880             (dev->addr_len &&
3881              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3882             (dev->ifindex != dev_get_iflink(dev) &&
3883              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3884                 goto nla_put_failure;
3885
3886         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3887         if (!br_afspec)
3888                 goto nla_put_failure;
3889
3890         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3891                 nla_nest_cancel(skb, br_afspec);
3892                 goto nla_put_failure;
3893         }
3894
3895         if (mode != BRIDGE_MODE_UNDEF) {
3896                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3897                         nla_nest_cancel(skb, br_afspec);
3898                         goto nla_put_failure;
3899                 }
3900         }
3901         if (vlan_fill) {
3902                 err = vlan_fill(skb, dev, filter_mask);
3903                 if (err) {
3904                         nla_nest_cancel(skb, br_afspec);
3905                         goto nla_put_failure;
3906                 }
3907         }
3908         nla_nest_end(skb, br_afspec);
3909
3910         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3911         if (!protinfo)
3912                 goto nla_put_failure;
3913
3914         if (brport_nla_put_flag(skb, flags, mask,
3915                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3916             brport_nla_put_flag(skb, flags, mask,
3917                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3918             brport_nla_put_flag(skb, flags, mask,
3919                                 IFLA_BRPORT_FAST_LEAVE,
3920                                 BR_MULTICAST_FAST_LEAVE) ||
3921             brport_nla_put_flag(skb, flags, mask,
3922                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3923             brport_nla_put_flag(skb, flags, mask,
3924                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3925             brport_nla_put_flag(skb, flags, mask,
3926                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3927             brport_nla_put_flag(skb, flags, mask,
3928                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3929             brport_nla_put_flag(skb, flags, mask,
3930                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3931                 nla_nest_cancel(skb, protinfo);
3932                 goto nla_put_failure;
3933         }
3934
3935         nla_nest_end(skb, protinfo);
3936
3937         nlmsg_end(skb, nlh);
3938         return 0;
3939 nla_put_failure:
3940         nlmsg_cancel(skb, nlh);
3941         return err ? err : -EMSGSIZE;
3942 }
3943 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3944
3945 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3946 {
3947         struct net *net = sock_net(skb->sk);
3948         struct net_device *dev;
3949         int idx = 0;
3950         u32 portid = NETLINK_CB(cb->skb).portid;
3951         u32 seq = cb->nlh->nlmsg_seq;
3952         u32 filter_mask = 0;
3953         int err;
3954
3955         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3956                 struct nlattr *extfilt;
3957
3958                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3959                                           IFLA_EXT_MASK);
3960                 if (extfilt) {
3961                         if (nla_len(extfilt) < sizeof(filter_mask))
3962                                 return -EINVAL;
3963
3964                         filter_mask = nla_get_u32(extfilt);
3965                 }
3966         }
3967
3968         rcu_read_lock();
3969         for_each_netdev_rcu(net, dev) {
3970                 const struct net_device_ops *ops = dev->netdev_ops;
3971                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3972
3973                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3974                         if (idx >= cb->args[0]) {
3975                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3976                                                 skb, portid, seq, dev,
3977                                                 filter_mask, NLM_F_MULTI);
3978                                 if (err < 0 && err != -EOPNOTSUPP) {
3979                                         if (likely(skb->len))
3980                                                 break;
3981
3982                                         goto out_err;
3983                                 }
3984                         }
3985                         idx++;
3986                 }
3987
3988                 if (ops->ndo_bridge_getlink) {
3989                         if (idx >= cb->args[0]) {
3990                                 err = ops->ndo_bridge_getlink(skb, portid,
3991                                                               seq, dev,
3992                                                               filter_mask,
3993                                                               NLM_F_MULTI);
3994                                 if (err < 0 && err != -EOPNOTSUPP) {
3995                                         if (likely(skb->len))
3996                                                 break;
3997
3998                                         goto out_err;
3999                                 }
4000                         }
4001                         idx++;
4002                 }
4003         }
4004         err = skb->len;
4005 out_err:
4006         rcu_read_unlock();
4007         cb->args[0] = idx;
4008
4009         return err;
4010 }
4011
4012 static inline size_t bridge_nlmsg_size(void)
4013 {
4014         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4015                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4016                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4017                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4018                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4019                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4020                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4021                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4022                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4023                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4024                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4025 }
4026
4027 static int rtnl_bridge_notify(struct net_device *dev)
4028 {
4029         struct net *net = dev_net(dev);
4030         struct sk_buff *skb;
4031         int err = -EOPNOTSUPP;
4032
4033         if (!dev->netdev_ops->ndo_bridge_getlink)
4034                 return 0;
4035
4036         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4037         if (!skb) {
4038                 err = -ENOMEM;
4039                 goto errout;
4040         }
4041
4042         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4043         if (err < 0)
4044                 goto errout;
4045
4046         if (!skb->len)
4047                 goto errout;
4048
4049         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4050         return 0;
4051 errout:
4052         WARN_ON(err == -EMSGSIZE);
4053         kfree_skb(skb);
4054         if (err)
4055                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4056         return err;
4057 }
4058
4059 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4060                                struct netlink_ext_ack *extack)
4061 {
4062         struct net *net = sock_net(skb->sk);
4063         struct ifinfomsg *ifm;
4064         struct net_device *dev;
4065         struct nlattr *br_spec, *attr = NULL;
4066         int rem, err = -EOPNOTSUPP;
4067         u16 flags = 0;
4068         bool have_flags = false;
4069
4070         if (nlmsg_len(nlh) < sizeof(*ifm))
4071                 return -EINVAL;
4072
4073         ifm = nlmsg_data(nlh);
4074         if (ifm->ifi_family != AF_BRIDGE)
4075                 return -EPFNOSUPPORT;
4076
4077         dev = __dev_get_by_index(net, ifm->ifi_index);
4078         if (!dev) {
4079                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4080                 return -ENODEV;
4081         }
4082
4083         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4084         if (br_spec) {
4085                 nla_for_each_nested(attr, br_spec, rem) {
4086                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4087                                 if (nla_len(attr) < sizeof(flags))
4088                                         return -EINVAL;
4089
4090                                 have_flags = true;
4091                                 flags = nla_get_u16(attr);
4092                                 break;
4093                         }
4094                 }
4095         }
4096
4097         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4098                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4099
4100                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4101                         err = -EOPNOTSUPP;
4102                         goto out;
4103                 }
4104
4105                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
4106                 if (err)
4107                         goto out;
4108
4109                 flags &= ~BRIDGE_FLAGS_MASTER;
4110         }
4111
4112         if ((flags & BRIDGE_FLAGS_SELF)) {
4113                 if (!dev->netdev_ops->ndo_bridge_setlink)
4114                         err = -EOPNOTSUPP;
4115                 else
4116                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4117                                                                   flags);
4118                 if (!err) {
4119                         flags &= ~BRIDGE_FLAGS_SELF;
4120
4121                         /* Generate event to notify upper layer of bridge
4122                          * change
4123                          */
4124                         err = rtnl_bridge_notify(dev);
4125                 }
4126         }
4127
4128         if (have_flags)
4129                 memcpy(nla_data(attr), &flags, sizeof(flags));
4130 out:
4131         return err;
4132 }
4133
4134 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4135                                struct netlink_ext_ack *extack)
4136 {
4137         struct net *net = sock_net(skb->sk);
4138         struct ifinfomsg *ifm;
4139         struct net_device *dev;
4140         struct nlattr *br_spec, *attr = NULL;
4141         int rem, err = -EOPNOTSUPP;
4142         u16 flags = 0;
4143         bool have_flags = false;
4144
4145         if (nlmsg_len(nlh) < sizeof(*ifm))
4146                 return -EINVAL;
4147
4148         ifm = nlmsg_data(nlh);
4149         if (ifm->ifi_family != AF_BRIDGE)
4150                 return -EPFNOSUPPORT;
4151
4152         dev = __dev_get_by_index(net, ifm->ifi_index);
4153         if (!dev) {
4154                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4155                 return -ENODEV;
4156         }
4157
4158         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4159         if (br_spec) {
4160                 nla_for_each_nested(attr, br_spec, rem) {
4161                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4162                                 if (nla_len(attr) < sizeof(flags))
4163                                         return -EINVAL;
4164
4165                                 have_flags = true;
4166                                 flags = nla_get_u16(attr);
4167                                 break;
4168                         }
4169                 }
4170         }
4171
4172         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4173                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4174
4175                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4176                         err = -EOPNOTSUPP;
4177                         goto out;
4178                 }
4179
4180                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4181                 if (err)
4182                         goto out;
4183
4184                 flags &= ~BRIDGE_FLAGS_MASTER;
4185         }
4186
4187         if ((flags & BRIDGE_FLAGS_SELF)) {
4188                 if (!dev->netdev_ops->ndo_bridge_dellink)
4189                         err = -EOPNOTSUPP;
4190                 else
4191                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4192                                                                   flags);
4193
4194                 if (!err) {
4195                         flags &= ~BRIDGE_FLAGS_SELF;
4196
4197                         /* Generate event to notify upper layer of bridge
4198                          * change
4199                          */
4200                         err = rtnl_bridge_notify(dev);
4201                 }
4202         }
4203
4204         if (have_flags)
4205                 memcpy(nla_data(attr), &flags, sizeof(flags));
4206 out:
4207         return err;
4208 }
4209
4210 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4211 {
4212         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4213                (!idxattr || idxattr == attrid);
4214 }
4215
4216 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4217 static int rtnl_get_offload_stats_attr_size(int attr_id)
4218 {
4219         switch (attr_id) {
4220         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4221                 return sizeof(struct rtnl_link_stats64);
4222         }
4223
4224         return 0;
4225 }
4226
4227 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4228                                   int *prividx)
4229 {
4230         struct nlattr *attr = NULL;
4231         int attr_id, size;
4232         void *attr_data;
4233         int err;
4234
4235         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4236               dev->netdev_ops->ndo_get_offload_stats))
4237                 return -ENODATA;
4238
4239         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4240              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4241                 if (attr_id < *prividx)
4242                         continue;
4243
4244                 size = rtnl_get_offload_stats_attr_size(attr_id);
4245                 if (!size)
4246                         continue;
4247
4248                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4249                         continue;
4250
4251                 attr = nla_reserve_64bit(skb, attr_id, size,
4252                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4253                 if (!attr)
4254                         goto nla_put_failure;
4255
4256                 attr_data = nla_data(attr);
4257                 memset(attr_data, 0, size);
4258                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4259                                                              attr_data);
4260                 if (err)
4261                         goto get_offload_stats_failure;
4262         }
4263
4264         if (!attr)
4265                 return -ENODATA;
4266
4267         *prividx = 0;
4268         return 0;
4269
4270 nla_put_failure:
4271         err = -EMSGSIZE;
4272 get_offload_stats_failure:
4273         *prividx = attr_id;
4274         return err;
4275 }
4276
4277 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4278 {
4279         int nla_size = 0;
4280         int attr_id;
4281         int size;
4282
4283         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4284               dev->netdev_ops->ndo_get_offload_stats))
4285                 return 0;
4286
4287         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4288              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4289                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4290                         continue;
4291                 size = rtnl_get_offload_stats_attr_size(attr_id);
4292                 nla_size += nla_total_size_64bit(size);
4293         }
4294
4295         if (nla_size != 0)
4296                 nla_size += nla_total_size(0);
4297
4298         return nla_size;
4299 }
4300
4301 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4302                                int type, u32 pid, u32 seq, u32 change,
4303                                unsigned int flags, unsigned int filter_mask,
4304                                int *idxattr, int *prividx)
4305 {
4306         struct if_stats_msg *ifsm;
4307         struct nlmsghdr *nlh;
4308         struct nlattr *attr;
4309         int s_prividx = *prividx;
4310         int err;
4311
4312         ASSERT_RTNL();
4313
4314         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4315         if (!nlh)
4316                 return -EMSGSIZE;
4317
4318         ifsm = nlmsg_data(nlh);
4319         ifsm->family = PF_UNSPEC;
4320         ifsm->pad1 = 0;
4321         ifsm->pad2 = 0;
4322         ifsm->ifindex = dev->ifindex;
4323         ifsm->filter_mask = filter_mask;
4324
4325         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4326                 struct rtnl_link_stats64 *sp;
4327
4328                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4329                                          sizeof(struct rtnl_link_stats64),
4330                                          IFLA_STATS_UNSPEC);
4331                 if (!attr)
4332                         goto nla_put_failure;
4333
4334                 sp = nla_data(attr);
4335                 dev_get_stats(dev, sp);
4336         }
4337
4338         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4339                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4340
4341                 if (ops && ops->fill_linkxstats) {
4342                         *idxattr = IFLA_STATS_LINK_XSTATS;
4343                         attr = nla_nest_start(skb,
4344                                               IFLA_STATS_LINK_XSTATS);
4345                         if (!attr)
4346                                 goto nla_put_failure;
4347
4348                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4349                         nla_nest_end(skb, attr);
4350                         if (err)
4351                                 goto nla_put_failure;
4352                         *idxattr = 0;
4353                 }
4354         }
4355
4356         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4357                              *idxattr)) {
4358                 const struct rtnl_link_ops *ops = NULL;
4359                 const struct net_device *master;
4360
4361                 master = netdev_master_upper_dev_get(dev);
4362                 if (master)
4363                         ops = master->rtnl_link_ops;
4364                 if (ops && ops->fill_linkxstats) {
4365                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4366                         attr = nla_nest_start(skb,
4367                                               IFLA_STATS_LINK_XSTATS_SLAVE);
4368                         if (!attr)
4369                                 goto nla_put_failure;
4370
4371                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4372                         nla_nest_end(skb, attr);
4373                         if (err)
4374                                 goto nla_put_failure;
4375                         *idxattr = 0;
4376                 }
4377         }
4378
4379         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4380                              *idxattr)) {
4381                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4382                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4383                 if (!attr)
4384                         goto nla_put_failure;
4385
4386                 err = rtnl_get_offload_stats(skb, dev, prividx);
4387                 if (err == -ENODATA)
4388                         nla_nest_cancel(skb, attr);
4389                 else
4390                         nla_nest_end(skb, attr);
4391
4392                 if (err && err != -ENODATA)
4393                         goto nla_put_failure;
4394                 *idxattr = 0;
4395         }
4396
4397         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4398                 struct rtnl_af_ops *af_ops;
4399
4400                 *idxattr = IFLA_STATS_AF_SPEC;
4401                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4402                 if (!attr)
4403                         goto nla_put_failure;
4404
4405                 rcu_read_lock();
4406                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4407                         if (af_ops->fill_stats_af) {
4408                                 struct nlattr *af;
4409                                 int err;
4410
4411                                 af = nla_nest_start(skb, af_ops->family);
4412                                 if (!af) {
4413                                         rcu_read_unlock();
4414                                         goto nla_put_failure;
4415                                 }
4416                                 err = af_ops->fill_stats_af(skb, dev);
4417
4418                                 if (err == -ENODATA) {
4419                                         nla_nest_cancel(skb, af);
4420                                 } else if (err < 0) {
4421                                         rcu_read_unlock();
4422                                         goto nla_put_failure;
4423                                 }
4424
4425                                 nla_nest_end(skb, af);
4426                         }
4427                 }
4428                 rcu_read_unlock();
4429
4430                 nla_nest_end(skb, attr);
4431
4432                 *idxattr = 0;
4433         }
4434
4435         nlmsg_end(skb, nlh);
4436
4437         return 0;
4438
4439 nla_put_failure:
4440         /* not a multi message or no progress mean a real error */
4441         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4442                 nlmsg_cancel(skb, nlh);
4443         else
4444                 nlmsg_end(skb, nlh);
4445
4446         return -EMSGSIZE;
4447 }
4448
4449 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4450                                   u32 filter_mask)
4451 {
4452         size_t size = 0;
4453
4454         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4455                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4456
4457         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4458                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4459                 int attr = IFLA_STATS_LINK_XSTATS;
4460
4461                 if (ops && ops->get_linkxstats_size) {
4462                         size += nla_total_size(ops->get_linkxstats_size(dev,
4463                                                                         attr));
4464                         /* for IFLA_STATS_LINK_XSTATS */
4465                         size += nla_total_size(0);
4466                 }
4467         }
4468
4469         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4470                 struct net_device *_dev = (struct net_device *)dev;
4471                 const struct rtnl_link_ops *ops = NULL;
4472                 const struct net_device *master;
4473
4474                 /* netdev_master_upper_dev_get can't take const */
4475                 master = netdev_master_upper_dev_get(_dev);
4476                 if (master)
4477                         ops = master->rtnl_link_ops;
4478                 if (ops && ops->get_linkxstats_size) {
4479                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4480
4481                         size += nla_total_size(ops->get_linkxstats_size(dev,
4482                                                                         attr));
4483                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4484                         size += nla_total_size(0);
4485                 }
4486         }
4487
4488         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4489                 size += rtnl_get_offload_stats_size(dev);
4490
4491         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4492                 struct rtnl_af_ops *af_ops;
4493
4494                 /* for IFLA_STATS_AF_SPEC */
4495                 size += nla_total_size(0);
4496
4497                 rcu_read_lock();
4498                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4499                         if (af_ops->get_stats_af_size) {
4500                                 size += nla_total_size(
4501                                         af_ops->get_stats_af_size(dev));
4502
4503                                 /* for AF_* */
4504                                 size += nla_total_size(0);
4505                         }
4506                 }
4507                 rcu_read_unlock();
4508         }
4509
4510         return size;
4511 }
4512
4513 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4514                           struct netlink_ext_ack *extack)
4515 {
4516         struct net *net = sock_net(skb->sk);
4517         struct net_device *dev = NULL;
4518         int idxattr = 0, prividx = 0;
4519         struct if_stats_msg *ifsm;
4520         struct sk_buff *nskb;
4521         u32 filter_mask;
4522         int err;
4523
4524         if (nlmsg_len(nlh) < sizeof(*ifsm))
4525                 return -EINVAL;
4526
4527         ifsm = nlmsg_data(nlh);
4528         if (ifsm->ifindex > 0)
4529                 dev = __dev_get_by_index(net, ifsm->ifindex);
4530         else
4531                 return -EINVAL;
4532
4533         if (!dev)
4534                 return -ENODEV;
4535
4536         filter_mask = ifsm->filter_mask;
4537         if (!filter_mask)
4538                 return -EINVAL;
4539
4540         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4541         if (!nskb)
4542                 return -ENOBUFS;
4543
4544         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4545                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4546                                   0, filter_mask, &idxattr, &prividx);
4547         if (err < 0) {
4548                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4549                 WARN_ON(err == -EMSGSIZE);
4550                 kfree_skb(nskb);
4551         } else {
4552                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4553         }
4554
4555         return err;
4556 }
4557
4558 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4559 {
4560         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4561         struct net *net = sock_net(skb->sk);
4562         unsigned int flags = NLM_F_MULTI;
4563         struct if_stats_msg *ifsm;
4564         struct hlist_head *head;
4565         struct net_device *dev;
4566         u32 filter_mask = 0;
4567         int idx = 0;
4568
4569         s_h = cb->args[0];
4570         s_idx = cb->args[1];
4571         s_idxattr = cb->args[2];
4572         s_prividx = cb->args[3];
4573
4574         cb->seq = net->dev_base_seq;
4575
4576         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4577                 return -EINVAL;
4578
4579         ifsm = nlmsg_data(cb->nlh);
4580         filter_mask = ifsm->filter_mask;
4581         if (!filter_mask)
4582                 return -EINVAL;
4583
4584         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4585                 idx = 0;
4586                 head = &net->dev_index_head[h];
4587                 hlist_for_each_entry(dev, head, index_hlist) {
4588                         if (idx < s_idx)
4589                                 goto cont;
4590                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4591                                                   NETLINK_CB(cb->skb).portid,
4592                                                   cb->nlh->nlmsg_seq, 0,
4593                                                   flags, filter_mask,
4594                                                   &s_idxattr, &s_prividx);
4595                         /* If we ran out of room on the first message,
4596                          * we're in trouble
4597                          */
4598                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4599
4600                         if (err < 0)
4601                                 goto out;
4602                         s_prividx = 0;
4603                         s_idxattr = 0;
4604                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4605 cont:
4606                         idx++;
4607                 }
4608         }
4609 out:
4610         cb->args[3] = s_prividx;
4611         cb->args[2] = s_idxattr;
4612         cb->args[1] = idx;
4613         cb->args[0] = h;
4614
4615         return skb->len;
4616 }
4617
4618 /* Process one rtnetlink message. */
4619
4620 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4621                              struct netlink_ext_ack *extack)
4622 {
4623         struct net *net = sock_net(skb->sk);
4624         struct rtnl_link *link;
4625         struct module *owner;
4626         int err = -EOPNOTSUPP;
4627         rtnl_doit_func doit;
4628         unsigned int flags;
4629         int kind;
4630         int family;
4631         int type;
4632
4633         type = nlh->nlmsg_type;
4634         if (type > RTM_MAX)
4635                 return -EOPNOTSUPP;
4636
4637         type -= RTM_BASE;
4638
4639         /* All the messages must have at least 1 byte length */
4640         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4641                 return 0;
4642
4643         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4644         kind = type&3;
4645
4646         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4647                 return -EPERM;
4648
4649         rcu_read_lock();
4650         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4651                 struct sock *rtnl;
4652                 rtnl_dumpit_func dumpit;
4653                 u16 min_dump_alloc = 0;
4654
4655                 link = rtnl_get_link(family, type);
4656                 if (!link || !link->dumpit) {
4657                         family = PF_UNSPEC;
4658                         link = rtnl_get_link(family, type);
4659                         if (!link || !link->dumpit)
4660                                 goto err_unlock;
4661                 }
4662                 owner = link->owner;
4663                 dumpit = link->dumpit;
4664
4665                 if (type == RTM_GETLINK - RTM_BASE)
4666                         min_dump_alloc = rtnl_calcit(skb, nlh);
4667
4668                 err = 0;
4669                 /* need to do this before rcu_read_unlock() */
4670                 if (!try_module_get(owner))
4671                         err = -EPROTONOSUPPORT;
4672
4673                 rcu_read_unlock();
4674
4675                 rtnl = net->rtnl;
4676                 if (err == 0) {
4677                         struct netlink_dump_control c = {
4678                                 .dump           = dumpit,
4679                                 .min_dump_alloc = min_dump_alloc,
4680                                 .module         = owner,
4681                         };
4682                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4683                         /* netlink_dump_start() will keep a reference on
4684                          * module if dump is still in progress.
4685                          */
4686                         module_put(owner);
4687                 }
4688                 return err;
4689         }
4690
4691         link = rtnl_get_link(family, type);
4692         if (!link || !link->doit) {
4693                 family = PF_UNSPEC;
4694                 link = rtnl_get_link(PF_UNSPEC, type);
4695                 if (!link || !link->doit)
4696                         goto out_unlock;
4697         }
4698
4699         owner = link->owner;
4700         if (!try_module_get(owner)) {
4701                 err = -EPROTONOSUPPORT;
4702                 goto out_unlock;
4703         }
4704
4705         flags = link->flags;
4706         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4707                 doit = link->doit;
4708                 rcu_read_unlock();
4709                 if (doit)
4710                         err = doit(skb, nlh, extack);
4711                 module_put(owner);
4712                 return err;
4713         }
4714         rcu_read_unlock();
4715
4716         rtnl_lock();
4717         link = rtnl_get_link(family, type);
4718         if (link && link->doit)
4719                 err = link->doit(skb, nlh, extack);
4720         rtnl_unlock();
4721
4722         module_put(owner);
4723
4724         return err;
4725
4726 out_unlock:
4727         rcu_read_unlock();
4728         return err;
4729
4730 err_unlock:
4731         rcu_read_unlock();
4732         return -EOPNOTSUPP;
4733 }
4734
4735 static void rtnetlink_rcv(struct sk_buff *skb)
4736 {
4737         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4738 }
4739
4740 static int rtnetlink_bind(struct net *net, int group)
4741 {
4742         switch (group) {
4743         case RTNLGRP_IPV4_MROUTE_R:
4744         case RTNLGRP_IPV6_MROUTE_R:
4745                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4746                         return -EPERM;
4747                 break;
4748         }
4749         return 0;
4750 }
4751
4752 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4753 {
4754         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4755
4756         switch (event) {
4757         case NETDEV_REBOOT:
4758         case NETDEV_CHANGEMTU:
4759         case NETDEV_CHANGEADDR:
4760         case NETDEV_CHANGENAME:
4761         case NETDEV_FEAT_CHANGE:
4762         case NETDEV_BONDING_FAILOVER:
4763         case NETDEV_POST_TYPE_CHANGE:
4764         case NETDEV_NOTIFY_PEERS:
4765         case NETDEV_CHANGEUPPER:
4766         case NETDEV_RESEND_IGMP:
4767         case NETDEV_CHANGEINFODATA:
4768         case NETDEV_CHANGELOWERSTATE:
4769         case NETDEV_CHANGE_TX_QUEUE_LEN:
4770                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4771                                    GFP_KERNEL, NULL, 0);
4772                 break;
4773         default:
4774                 break;
4775         }
4776         return NOTIFY_DONE;
4777 }
4778
4779 static struct notifier_block rtnetlink_dev_notifier = {
4780         .notifier_call  = rtnetlink_event,
4781 };
4782
4783
4784 static int __net_init rtnetlink_net_init(struct net *net)
4785 {
4786         struct sock *sk;
4787         struct netlink_kernel_cfg cfg = {
4788                 .groups         = RTNLGRP_MAX,
4789                 .input          = rtnetlink_rcv,
4790                 .cb_mutex       = &rtnl_mutex,
4791                 .flags          = NL_CFG_F_NONROOT_RECV,
4792                 .bind           = rtnetlink_bind,
4793         };
4794
4795         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4796         if (!sk)
4797                 return -ENOMEM;
4798         net->rtnl = sk;
4799         return 0;
4800 }
4801
4802 static void __net_exit rtnetlink_net_exit(struct net *net)
4803 {
4804         netlink_kernel_release(net->rtnl);
4805         net->rtnl = NULL;
4806 }
4807
4808 static struct pernet_operations rtnetlink_net_ops = {
4809         .init = rtnetlink_net_init,
4810         .exit = rtnetlink_net_exit,
4811 };
4812
4813 void __init rtnetlink_init(void)
4814 {
4815         if (register_pernet_subsys(&rtnetlink_net_ops))
4816                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4817
4818         register_netdevice_notifier(&rtnetlink_dev_notifier);
4819
4820         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4821                       rtnl_dump_ifinfo, 0);
4822         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4823         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4824         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4825
4826         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4827         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4828         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4829
4830         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4831         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4832         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4833
4834         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4835         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4836         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4837
4838         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4839                       0);
4840 }