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