net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 143
 144 static inline int rtm_msgindex(int msgtype)
 145 {
 146         int msgindex = msgtype - RTM_BASE;
 147
 148         /*
 149          * msgindex < 0 implies someone tried to register a netlink
 150          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 151          * the message type has not been added to linux/rtnetlink.h
 152          */
 153         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 154
 155         return msgindex;
 156 }
 157
 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 159 {
 160         struct rtnl_link __rcu **tab;
 161
 162         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 163                 protocol = PF_UNSPEC;
 164
 165         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 166         if (!tab)
 167                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 168
 169         return rcu_dereference_rtnl(tab[msgtype]);
 170 }
 171
 172 static int rtnl_register_internal(struct module *owner,
 173                                   int protocol, int msgtype,
 174                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 175                                   unsigned int flags)
 176 {
 177         struct rtnl_link *link, *old;
 178         struct rtnl_link __rcu **tab;
 179         int msgindex;
 180         int ret = -ENOBUFS;
 181
 182         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 183         msgindex = rtm_msgindex(msgtype);
 184
 185         rtnl_lock();
 186         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 187         if (tab == NULL) {
 188                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 189                 if (!tab)
 190                         goto unlock;
 191
 192                 /* ensures we see the 0 stores */
 193                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 194         }
 195
 196         old = rtnl_dereference(tab[msgindex]);
 197         if (old) {
 198                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         } else {
 202                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 203                 if (!link)
 204                         goto unlock;
 205         }
 206
 207         WARN_ON(link->owner && link->owner != owner);
 208         link->owner = owner;
 209
 210         WARN_ON(doit && link->doit && link->doit != doit);
 211         if (doit)
 212                 link->doit = doit;
 213         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 214         if (dumpit)
 215                 link->dumpit = dumpit;
 216
 217         link->flags |= flags;
 218
 219         /* publish protocol:msgtype */
 220         rcu_assign_pointer(tab[msgindex], link);
 221         ret = 0;
 222         if (old)
 223                 kfree_rcu(old, rcu);
 224 unlock:
 225         rtnl_unlock();
 226         return ret;
 227 }
 228
 229 /**
 230  * rtnl_register_module - Register a rtnetlink message type
 231  *
 232  * @owner: module registering the hook (THIS_MODULE)
 233  * @protocol: Protocol family or PF_UNSPEC
 234  * @msgtype: rtnetlink message type
 235  * @doit: Function pointer called for each request message
 236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link __rcu **tab;
 290         struct rtnl_link *link;
 291         int msgindex;
 292
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294         msgindex = rtm_msgindex(msgtype);
 295
 296         rtnl_lock();
 297         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 298         if (!tab) {
 299                 rtnl_unlock();
 300                 return -ENOENT;
 301         }
 302
 303         link = rtnl_dereference(tab[msgindex]);
 304         rcu_assign_pointer(tab[msgindex], NULL);
 305         rtnl_unlock();
 306
 307         kfree_rcu(link, rcu);
 308
 309         return 0;
 310 }
 311 EXPORT_SYMBOL_GPL(rtnl_unregister);
 312
 313 /**
 314  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 315  * @protocol : Protocol family or PF_UNSPEC
 316  *
 317  * Identical to calling rtnl_unregster() for all registered message types
 318  * of a certain protocol family.
 319  */
 320 void rtnl_unregister_all(int protocol)
 321 {
 322         struct rtnl_link __rcu **tab;
 323         struct rtnl_link *link;
 324         int msgindex;
 325
 326         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 327
 328         rtnl_lock();
 329         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 330         if (!tab) {
 331                 rtnl_unlock();
 332                 return;
 333         }
 334         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 335         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 336                 link = rtnl_dereference(tab[msgindex]);
 337                 if (!link)
 338                         continue;
 339
 340                 rcu_assign_pointer(tab[msgindex], NULL);
 341                 kfree_rcu(link, rcu);
 342         }
 343         rtnl_unlock();
 344
 345         synchronize_net();
 346
 347         kfree(tab);
 348 }
 349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 350
 351 static LIST_HEAD(link_ops);
 352
 353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 354 {
 355         const struct rtnl_link_ops *ops;
 356
 357         list_for_each_entry(ops, &link_ops, list) {
 358                 if (!strcmp(ops->kind, kind))
 359                         return ops;
 360         }
 361         return NULL;
 362 }
 363
 364 /**
 365  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 366  * @ops: struct rtnl_link_ops * to register
 367  *
 368  * The caller must hold the rtnl_mutex. This function should be used
 369  * by drivers that create devices during module initialization. It
 370  * must be called before registering the devices.
 371  *
 372  * Returns 0 on success or a negative error code.
 373  */
 374 int __rtnl_link_register(struct rtnl_link_ops *ops)
 375 {
 376         if (rtnl_link_ops_get(ops->kind))
 377                 return -EEXIST;
 378
 379         /* The check for setup is here because if ops
 380          * does not have that filled up, it is not possible
 381          * to use the ops for creating device. So do not
 382          * fill up dellink as well. That disables rtnl_dellink.
 383          */
 384         if (ops->setup && !ops->dellink)
 385                 ops->dellink = unregister_netdevice_queue;
 386
 387         list_add_tail(&ops->list, &link_ops);
 388         return 0;
 389 }
 390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 391
 392 /**
 393  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 394  * @ops: struct rtnl_link_ops * to register
 395  *
 396  * Returns 0 on success or a negative error code.
 397  */
 398 int rtnl_link_register(struct rtnl_link_ops *ops)
 399 {
 400         int err;
 401
 402         /* Sanity-check max sizes to avoid stack buffer overflow. */
 403         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 404                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 405                 return -EINVAL;
 406
 407         rtnl_lock();
 408         err = __rtnl_link_register(ops);
 409         rtnl_unlock();
 410         return err;
 411 }
 412 EXPORT_SYMBOL_GPL(rtnl_link_register);
 413
 414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 415 {
 416         struct net_device *dev;
 417         LIST_HEAD(list_kill);
 418
 419         for_each_netdev(net, dev) {
 420                 if (dev->rtnl_link_ops == ops)
 421                         ops->dellink(dev, &list_kill);
 422         }
 423         unregister_netdevice_many(&list_kill);
 424 }
 425
 426 /**
 427  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 428  * @ops: struct rtnl_link_ops * to unregister
 429  *
 430  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 431  * integrity (hold pernet_ops_rwsem for writing to close the race
 432  * with setup_net() and cleanup_net()).
 433  */
 434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 435 {
 436         struct net *net;
 437
 438         for_each_net(net) {
 439                 __rtnl_kill_links(net, ops);
 440         }
 441         list_del(&ops->list);
 442 }
 443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 444
 445 /* Return with the rtnl_lock held when there are no network
 446  * devices unregistering in any network namespace.
 447  */
 448 static void rtnl_lock_unregistering_all(void)
 449 {
 450         struct net *net;
 451         bool unregistering;
 452         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 453
 454         add_wait_queue(&netdev_unregistering_wq, &wait);
 455         for (;;) {
 456                 unregistering = false;
 457                 rtnl_lock();
 458                 /* We held write locked pernet_ops_rwsem, and parallel
 459                  * setup_net() and cleanup_net() are not possible.
 460                  */
 461                 for_each_net(net) {
 462                         if (net->dev_unreg_count > 0) {
 463                                 unregistering = true;
 464                                 break;
 465                         }
 466                 }
 467                 if (!unregistering)
 468                         break;
 469                 __rtnl_unlock();
 470
 471                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 472         }
 473         remove_wait_queue(&netdev_unregistering_wq, &wait);
 474 }
 475
 476 /**
 477  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 478  * @ops: struct rtnl_link_ops * to unregister
 479  */
 480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 481 {
 482         /* Close the race with setup_net() and cleanup_net() */
 483         down_write(&pernet_ops_rwsem);
 484         rtnl_lock_unregistering_all();
 485         __rtnl_link_unregister(ops);
 486         rtnl_unlock();
 487         up_write(&pernet_ops_rwsem);
 488 }
 489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 490
 491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 492 {
 493         struct net_device *master_dev;
 494         const struct rtnl_link_ops *ops;
 495         size_t size = 0;
 496
 497         rcu_read_lock();
 498
 499         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 500         if (!master_dev)
 501                 goto out;
 502
 503         ops = master_dev->rtnl_link_ops;
 504         if (!ops || !ops->get_slave_size)
 505                 goto out;
 506         /* IFLA_INFO_SLAVE_DATA + nested data */
 507         size = nla_total_size(sizeof(struct nlattr)) +
 508                ops->get_slave_size(master_dev, dev);
 509
 510 out:
 511         rcu_read_unlock();
 512         return size;
 513 }
 514
 515 static size_t rtnl_link_get_size(const struct net_device *dev)
 516 {
 517         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 518         size_t size;
 519
 520         if (!ops)
 521                 return 0;
 522
 523         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 524                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 525
 526         if (ops->get_size)
 527                 /* IFLA_INFO_DATA + nested data */
 528                 size += nla_total_size(sizeof(struct nlattr)) +
 529                         ops->get_size(dev);
 530
 531         if (ops->get_xstats_size)
 532                 /* IFLA_INFO_XSTATS */
 533                 size += nla_total_size(ops->get_xstats_size(dev));
 534
 535         size += rtnl_link_get_slave_info_data_size(dev);
 536
 537         return size;
 538 }
 539
 540 static LIST_HEAD(rtnl_af_ops);
 541
 542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 543 {
 544         const struct rtnl_af_ops *ops;
 545
 546         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 547                 if (ops->family == family)
 548                         return ops;
 549         }
 550
 551         return NULL;
 552 }
 553
 554 /**
 555  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 556  * @ops: struct rtnl_af_ops * to register
 557  *
 558  * Returns 0 on success or a negative error code.
 559  */
 560 void rtnl_af_register(struct rtnl_af_ops *ops)
 561 {
 562         rtnl_lock();
 563         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 564         rtnl_unlock();
 565 }
 566 EXPORT_SYMBOL_GPL(rtnl_af_register);
 567
 568 /**
 569  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 570  * @ops: struct rtnl_af_ops * to unregister
 571  */
 572 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 573 {
 574         rtnl_lock();
 575         list_del_rcu(&ops->list);
 576         rtnl_unlock();
 577
 578         synchronize_rcu();
 579 }
 580 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 581
 582 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 583                                     u32 ext_filter_mask)
 584 {
 585         struct rtnl_af_ops *af_ops;
 586         size_t size;
 587
 588         /* IFLA_AF_SPEC */
 589         size = nla_total_size(sizeof(struct nlattr));
 590
 591         rcu_read_lock();
 592         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 593                 if (af_ops->get_link_af_size) {
 594                         /* AF_* + nested data */
 595                         size += nla_total_size(sizeof(struct nlattr)) +
 596                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 597                 }
 598         }
 599         rcu_read_unlock();
 600
 601         return size;
 602 }
 603
 604 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 605 {
 606         struct net_device *master_dev;
 607         bool ret = false;
 608
 609         rcu_read_lock();
 610
 611         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 612         if (master_dev && master_dev->rtnl_link_ops)
 613                 ret = true;
 614         rcu_read_unlock();
 615         return ret;
 616 }
 617
 618 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 619                                      const struct net_device *dev)
 620 {
 621         struct net_device *master_dev;
 622         const struct rtnl_link_ops *ops;
 623         struct nlattr *slave_data;
 624         int err;
 625
 626         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 627         if (!master_dev)
 628                 return 0;
 629         ops = master_dev->rtnl_link_ops;
 630         if (!ops)
 631                 return 0;
 632         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 633                 return -EMSGSIZE;
 634         if (ops->fill_slave_info) {
 635                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 636                 if (!slave_data)
 637                         return -EMSGSIZE;
 638                 err = ops->fill_slave_info(skb, master_dev, dev);
 639                 if (err < 0)
 640                         goto err_cancel_slave_data;
 641                 nla_nest_end(skb, slave_data);
 642         }
 643         return 0;
 644
 645 err_cancel_slave_data:
 646         nla_nest_cancel(skb, slave_data);
 647         return err;
 648 }
 649
 650 static int rtnl_link_info_fill(struct sk_buff *skb,
 651                                const struct net_device *dev)
 652 {
 653         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 654         struct nlattr *data;
 655         int err;
 656
 657         if (!ops)
 658                 return 0;
 659         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 660                 return -EMSGSIZE;
 661         if (ops->fill_xstats) {
 662                 err = ops->fill_xstats(skb, dev);
 663                 if (err < 0)
 664                         return err;
 665         }
 666         if (ops->fill_info) {
 667                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 668                 if (data == NULL)
 669                         return -EMSGSIZE;
 670                 err = ops->fill_info(skb, dev);
 671                 if (err < 0)
 672                         goto err_cancel_data;
 673                 nla_nest_end(skb, data);
 674         }
 675         return 0;
 676
 677 err_cancel_data:
 678         nla_nest_cancel(skb, data);
 679         return err;
 680 }
 681
 682 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 683 {
 684         struct nlattr *linkinfo;
 685         int err = -EMSGSIZE;
 686
 687         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 688         if (linkinfo == NULL)
 689                 goto out;
 690
 691         err = rtnl_link_info_fill(skb, dev);
 692         if (err < 0)
 693                 goto err_cancel_link;
 694
 695         err = rtnl_link_slave_info_fill(skb, dev);
 696         if (err < 0)
 697                 goto err_cancel_link;
 698
 699         nla_nest_end(skb, linkinfo);
 700         return 0;
 701
 702 err_cancel_link:
 703         nla_nest_cancel(skb, linkinfo);
 704 out:
 705         return err;
 706 }
 707
 708 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 709 {
 710         struct sock *rtnl = net->rtnl;
 711         int err = 0;
 712
 713         NETLINK_CB(skb).dst_group = group;
 714         if (echo)
 715                 refcount_inc(&skb->users);
 716         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 717         if (echo)
 718                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 719         return err;
 720 }
 721
 722 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 723 {
 724         struct sock *rtnl = net->rtnl;
 725
 726         return nlmsg_unicast(rtnl, skb, pid);
 727 }
 728 EXPORT_SYMBOL(rtnl_unicast);
 729
 730 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 731                  struct nlmsghdr *nlh, gfp_t flags)
 732 {
 733         struct sock *rtnl = net->rtnl;
 734         int report = 0;
 735
 736         if (nlh)
 737                 report = nlmsg_report(nlh);
 738
 739         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 740 }
 741 EXPORT_SYMBOL(rtnl_notify);
 742
 743 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 744 {
 745         struct sock *rtnl = net->rtnl;
 746
 747         netlink_set_err(rtnl, 0, group, error);
 748 }
 749 EXPORT_SYMBOL(rtnl_set_sk_err);
 750
 751 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 752 {
 753         struct nlattr *mx;
 754         int i, valid = 0;
 755
 756         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 757         if (metrics == dst_default_metrics.metrics)
 758                 return 0;
 759
 760         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 761         if (mx == NULL)
 762                 return -ENOBUFS;
 763
 764         for (i = 0; i < RTAX_MAX; i++) {
 765                 if (metrics[i]) {
 766                         if (i == RTAX_CC_ALGO - 1) {
 767                                 char tmp[TCP_CA_NAME_MAX], *name;
 768
 769                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 770                                 if (!name)
 771                                         continue;
 772                                 if (nla_put_string(skb, i + 1, name))
 773                                         goto nla_put_failure;
 774                         } else if (i == RTAX_FEATURES - 1) {
 775                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 776
 777                                 if (!user_features)
 778                                         continue;
 779                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 780                                 if (nla_put_u32(skb, i + 1, user_features))
 781                                         goto nla_put_failure;
 782                         } else {
 783                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 784                                         goto nla_put_failure;
 785                         }
 786                         valid++;
 787                 }
 788         }
 789
 790         if (!valid) {
 791                 nla_nest_cancel(skb, mx);
 792                 return 0;
 793         }
 794
 795         return nla_nest_end(skb, mx);
 796
 797 nla_put_failure:
 798         nla_nest_cancel(skb, mx);
 799         return -EMSGSIZE;
 800 }
 801 EXPORT_SYMBOL(rtnetlink_put_metrics);
 802
 803 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 804                        long expires, u32 error)
 805 {
 806         struct rta_cacheinfo ci = {
 807                 .rta_error = error,
 808                 .rta_id =  id,
 809         };
 810
 811         if (dst) {
 812                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 813                 ci.rta_used = dst->__use;
 814                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 815         }
 816         if (expires) {
 817                 unsigned long clock;
 818
 819                 clock = jiffies_to_clock_t(abs(expires));
 820                 clock = min_t(unsigned long, clock, INT_MAX);
 821                 ci.rta_expires = (expires > 0) ? clock : -clock;
 822         }
 823         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 824 }
 825 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 826
 827 static void set_operstate(struct net_device *dev, unsigned char transition)
 828 {
 829         unsigned char operstate = dev->operstate;
 830
 831         switch (transition) {
 832         case IF_OPER_UP:
 833                 if ((operstate == IF_OPER_DORMANT ||
 834                      operstate == IF_OPER_TESTING ||
 835                      operstate == IF_OPER_UNKNOWN) &&
 836                     !netif_dormant(dev) && !netif_testing(dev))
 837                         operstate = IF_OPER_UP;
 838                 break;
 839
 840         case IF_OPER_TESTING:
 841                 if (operstate == IF_OPER_UP ||
 842                     operstate == IF_OPER_UNKNOWN)
 843                         operstate = IF_OPER_TESTING;
 844                 break;
 845
 846         case IF_OPER_DORMANT:
 847                 if (operstate == IF_OPER_UP ||
 848                     operstate == IF_OPER_UNKNOWN)
 849                         operstate = IF_OPER_DORMANT;
 850                 break;
 851         }
 852
 853         if (dev->operstate != operstate) {
 854                 write_lock_bh(&dev_base_lock);
 855                 dev->operstate = operstate;
 856                 write_unlock_bh(&dev_base_lock);
 857                 netdev_state_change(dev);
 858         }
 859 }
 860
 861 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 862 {
 863         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 864                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 865 }
 866
 867 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 868                                            const struct ifinfomsg *ifm)
 869 {
 870         unsigned int flags = ifm->ifi_flags;
 871
 872         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 873         if (ifm->ifi_change)
 874                 flags = (flags & ifm->ifi_change) |
 875                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 876
 877         return flags;
 878 }
 879
 880 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 881                                  const struct rtnl_link_stats64 *b)
 882 {
 883         a->rx_packets = b->rx_packets;
 884         a->tx_packets = b->tx_packets;
 885         a->rx_bytes = b->rx_bytes;
 886         a->tx_bytes = b->tx_bytes;
 887         a->rx_errors = b->rx_errors;
 888         a->tx_errors = b->tx_errors;
 889         a->rx_dropped = b->rx_dropped;
 890         a->tx_dropped = b->tx_dropped;
 891
 892         a->multicast = b->multicast;
 893         a->collisions = b->collisions;
 894
 895         a->rx_length_errors = b->rx_length_errors;
 896         a->rx_over_errors = b->rx_over_errors;
 897         a->rx_crc_errors = b->rx_crc_errors;
 898         a->rx_frame_errors = b->rx_frame_errors;
 899         a->rx_fifo_errors = b->rx_fifo_errors;
 900         a->rx_missed_errors = b->rx_missed_errors;
 901
 902         a->tx_aborted_errors = b->tx_aborted_errors;
 903         a->tx_carrier_errors = b->tx_carrier_errors;
 904         a->tx_fifo_errors = b->tx_fifo_errors;
 905         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 906         a->tx_window_errors = b->tx_window_errors;
 907
 908         a->rx_compressed = b->rx_compressed;
 909         a->tx_compressed = b->tx_compressed;
 910
 911         a->rx_nohandler = b->rx_nohandler;
 912 }
 913
 914 /* All VF info */
 915 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 916                                    u32 ext_filter_mask)
 917 {
 918         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 919                 int num_vfs = dev_num_vf(dev->dev.parent);
 920                 size_t size = nla_total_size(0);
 921                 size += num_vfs *
 922                         (nla_total_size(0) +
 923                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 924                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 925                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 926                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 927                          nla_total_size(MAX_VLAN_LIST_LEN *
 928                                         sizeof(struct ifla_vf_vlan_info)) +
 929                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 930                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 931                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 932                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 933                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 934                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 935                          /* IFLA_VF_STATS_RX_PACKETS */
 936                          nla_total_size_64bit(sizeof(__u64)) +
 937                          /* IFLA_VF_STATS_TX_PACKETS */
 938                          nla_total_size_64bit(sizeof(__u64)) +
 939                          /* IFLA_VF_STATS_RX_BYTES */
 940                          nla_total_size_64bit(sizeof(__u64)) +
 941                          /* IFLA_VF_STATS_TX_BYTES */
 942                          nla_total_size_64bit(sizeof(__u64)) +
 943                          /* IFLA_VF_STATS_BROADCAST */
 944                          nla_total_size_64bit(sizeof(__u64)) +
 945                          /* IFLA_VF_STATS_MULTICAST */
 946                          nla_total_size_64bit(sizeof(__u64)) +
 947                          /* IFLA_VF_STATS_RX_DROPPED */
 948                          nla_total_size_64bit(sizeof(__u64)) +
 949                          /* IFLA_VF_STATS_TX_DROPPED */
 950                          nla_total_size_64bit(sizeof(__u64)) +
 951                          nla_total_size(sizeof(struct ifla_vf_trust)));
 952                 return size;
 953         } else
 954                 return 0;
 955 }
 956
 957 static size_t rtnl_port_size(const struct net_device *dev,
 958                              u32 ext_filter_mask)
 959 {
 960         size_t port_size = nla_total_size(4)            /* PORT_VF */
 961                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 962                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 963                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 964                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 965                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 966         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 967         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 968                 + port_size;
 969         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 970                 + port_size;
 971
 972         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 973             !(ext_filter_mask & RTEXT_FILTER_VF))
 974                 return 0;
 975         if (dev_num_vf(dev->dev.parent))
 976                 return port_self_size + vf_ports_size +
 977                         vf_port_size * dev_num_vf(dev->dev.parent);
 978         else
 979                 return port_self_size;
 980 }
 981
 982 static size_t rtnl_xdp_size(void)
 983 {
 984         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 985                           nla_total_size(1) +   /* XDP_ATTACHED */
 986                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 987                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 988
 989         return xdp_size;
 990 }
 991
 992 static size_t rtnl_prop_list_size(const struct net_device *dev)
 993 {
 994         struct netdev_name_node *name_node;
 995         size_t size;
 996
 997         if (list_empty(&dev->name_node->list))
 998                 return 0;
 999         size = nla_total_size(0);
1000         list_for_each_entry(name_node, &dev->name_node->list, list)
1001                 size += nla_total_size(ALTIFNAMSIZ);
1002         return size;
1003 }
1004
1005 static size_t rtnl_proto_down_size(const struct net_device *dev)
1006 {
1007         size_t size = nla_total_size(1);
1008
1009         if (dev->proto_down_reason)
1010                 size += nla_total_size(0) + nla_total_size(4);
1011
1012         return size;
1013 }
1014
1015 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1016                                      u32 ext_filter_mask)
1017 {
1018         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1019                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1020                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1021                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1022                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1023                + nla_total_size(sizeof(struct rtnl_link_stats))
1024                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1025                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1026                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1027                + nla_total_size(4) /* IFLA_TXQLEN */
1028                + nla_total_size(4) /* IFLA_WEIGHT */
1029                + nla_total_size(4) /* IFLA_MTU */
1030                + nla_total_size(4) /* IFLA_LINK */
1031                + nla_total_size(4) /* IFLA_MASTER */
1032                + nla_total_size(1) /* IFLA_CARRIER */
1033                + nla_total_size(4) /* IFLA_PROMISCUITY */
1034                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1035                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1036                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1037                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1038                + nla_total_size(1) /* IFLA_OPERSTATE */
1039                + nla_total_size(1) /* IFLA_LINKMODE */
1040                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1041                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1042                + nla_total_size(4) /* IFLA_GROUP */
1043                + nla_total_size(ext_filter_mask
1044                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1045                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1046                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1047                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1048                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1049                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1050                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1051                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1052                + rtnl_xdp_size() /* IFLA_XDP */
1053                + nla_total_size(4)  /* IFLA_EVENT */
1054                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1055                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1056                + rtnl_proto_down_size(dev)  /* proto down */
1057                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1058                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1059                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1060                + nla_total_size(4)  /* IFLA_MIN_MTU */
1061                + nla_total_size(4)  /* IFLA_MAX_MTU */
1062                + rtnl_prop_list_size(dev)
1063                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1064                + 0;
1065 }
1066
1067 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1068 {
1069         struct nlattr *vf_ports;
1070         struct nlattr *vf_port;
1071         int vf;
1072         int err;
1073
1074         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1075         if (!vf_ports)
1076                 return -EMSGSIZE;
1077
1078         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1079                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1080                 if (!vf_port)
1081                         goto nla_put_failure;
1082                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1083                         goto nla_put_failure;
1084                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1085                 if (err == -EMSGSIZE)
1086                         goto nla_put_failure;
1087                 if (err) {
1088                         nla_nest_cancel(skb, vf_port);
1089                         continue;
1090                 }
1091                 nla_nest_end(skb, vf_port);
1092         }
1093
1094         nla_nest_end(skb, vf_ports);
1095
1096         return 0;
1097
1098 nla_put_failure:
1099         nla_nest_cancel(skb, vf_ports);
1100         return -EMSGSIZE;
1101 }
1102
1103 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1104 {
1105         struct nlattr *port_self;
1106         int err;
1107
1108         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1109         if (!port_self)
1110                 return -EMSGSIZE;
1111
1112         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1113         if (err) {
1114                 nla_nest_cancel(skb, port_self);
1115                 return (err == -EMSGSIZE) ? err : 0;
1116         }
1117
1118         nla_nest_end(skb, port_self);
1119
1120         return 0;
1121 }
1122
1123 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1124                           u32 ext_filter_mask)
1125 {
1126         int err;
1127
1128         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1129             !(ext_filter_mask & RTEXT_FILTER_VF))
1130                 return 0;
1131
1132         err = rtnl_port_self_fill(skb, dev);
1133         if (err)
1134                 return err;
1135
1136         if (dev_num_vf(dev->dev.parent)) {
1137                 err = rtnl_vf_ports_fill(skb, dev);
1138                 if (err)
1139                         return err;
1140         }
1141
1142         return 0;
1143 }
1144
1145 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1146 {
1147         int err;
1148         struct netdev_phys_item_id ppid;
1149
1150         err = dev_get_phys_port_id(dev, &ppid);
1151         if (err) {
1152                 if (err == -EOPNOTSUPP)
1153                         return 0;
1154                 return err;
1155         }
1156
1157         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1158                 return -EMSGSIZE;
1159
1160         return 0;
1161 }
1162
1163 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1164 {
1165         char name[IFNAMSIZ];
1166         int err;
1167
1168         err = dev_get_phys_port_name(dev, name, sizeof(name));
1169         if (err) {
1170                 if (err == -EOPNOTSUPP)
1171                         return 0;
1172                 return err;
1173         }
1174
1175         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1176                 return -EMSGSIZE;
1177
1178         return 0;
1179 }
1180
1181 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1182 {
1183         struct netdev_phys_item_id ppid = { };
1184         int err;
1185
1186         err = dev_get_port_parent_id(dev, &ppid, false);
1187         if (err) {
1188                 if (err == -EOPNOTSUPP)
1189                         return 0;
1190                 return err;
1191         }
1192
1193         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1194                 return -EMSGSIZE;
1195
1196         return 0;
1197 }
1198
1199 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1200                                               struct net_device *dev)
1201 {
1202         struct rtnl_link_stats64 *sp;
1203         struct nlattr *attr;
1204
1205         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1206                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1207         if (!attr)
1208                 return -EMSGSIZE;
1209
1210         sp = nla_data(attr);
1211         dev_get_stats(dev, sp);
1212
1213         attr = nla_reserve(skb, IFLA_STATS,
1214                            sizeof(struct rtnl_link_stats));
1215         if (!attr)
1216                 return -EMSGSIZE;
1217
1218         copy_rtnl_link_stats(nla_data(attr), sp);
1219
1220         return 0;
1221 }
1222
1223 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1224                                                struct net_device *dev,
1225                                                int vfs_num,
1226                                                struct nlattr *vfinfo)
1227 {
1228         struct ifla_vf_rss_query_en vf_rss_query_en;
1229         struct nlattr *vf, *vfstats, *vfvlanlist;
1230         struct ifla_vf_link_state vf_linkstate;
1231         struct ifla_vf_vlan_info vf_vlan_info;
1232         struct ifla_vf_spoofchk vf_spoofchk;
1233         struct ifla_vf_tx_rate vf_tx_rate;
1234         struct ifla_vf_stats vf_stats;
1235         struct ifla_vf_trust vf_trust;
1236         struct ifla_vf_vlan vf_vlan;
1237         struct ifla_vf_rate vf_rate;
1238         struct ifla_vf_mac vf_mac;
1239         struct ifla_vf_broadcast vf_broadcast;
1240         struct ifla_vf_info ivi;
1241         struct ifla_vf_guid node_guid;
1242         struct ifla_vf_guid port_guid;
1243
1244         memset(&ivi, 0, sizeof(ivi));
1245
1246         /* Not all SR-IOV capable drivers support the
1247          * spoofcheck and "RSS query enable" query.  Preset to
1248          * -1 so the user space tool can detect that the driver
1249          * didn't report anything.
1250          */
1251         ivi.spoofchk = -1;
1252         ivi.rss_query_en = -1;
1253         ivi.trusted = -1;
1254         /* The default value for VF link state is "auto"
1255          * IFLA_VF_LINK_STATE_AUTO which equals zero
1256          */
1257         ivi.linkstate = 0;
1258         /* VLAN Protocol by default is 802.1Q */
1259         ivi.vlan_proto = htons(ETH_P_8021Q);
1260         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1261                 return 0;
1262
1263         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1264         memset(&node_guid, 0, sizeof(node_guid));
1265         memset(&port_guid, 0, sizeof(port_guid));
1266
1267         vf_mac.vf =
1268                 vf_vlan.vf =
1269                 vf_vlan_info.vf =
1270                 vf_rate.vf =
1271                 vf_tx_rate.vf =
1272                 vf_spoofchk.vf =
1273                 vf_linkstate.vf =
1274                 vf_rss_query_en.vf =
1275                 vf_trust.vf =
1276                 node_guid.vf =
1277                 port_guid.vf = ivi.vf;
1278
1279         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1280         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1281         vf_vlan.vlan = ivi.vlan;
1282         vf_vlan.qos = ivi.qos;
1283         vf_vlan_info.vlan = ivi.vlan;
1284         vf_vlan_info.qos = ivi.qos;
1285         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1286         vf_tx_rate.rate = ivi.max_tx_rate;
1287         vf_rate.min_tx_rate = ivi.min_tx_rate;
1288         vf_rate.max_tx_rate = ivi.max_tx_rate;
1289         vf_spoofchk.setting = ivi.spoofchk;
1290         vf_linkstate.link_state = ivi.linkstate;
1291         vf_rss_query_en.setting = ivi.rss_query_en;
1292         vf_trust.setting = ivi.trusted;
1293         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1294         if (!vf)
1295                 goto nla_put_vfinfo_failure;
1296         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1297             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1298             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1299             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1300                     &vf_rate) ||
1301             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1302                     &vf_tx_rate) ||
1303             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1304                     &vf_spoofchk) ||
1305             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1306                     &vf_linkstate) ||
1307             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1308                     sizeof(vf_rss_query_en),
1309                     &vf_rss_query_en) ||
1310             nla_put(skb, IFLA_VF_TRUST,
1311                     sizeof(vf_trust), &vf_trust))
1312                 goto nla_put_vf_failure;
1313
1314         if (dev->netdev_ops->ndo_get_vf_guid &&
1315             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1316                                               &port_guid)) {
1317                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1318                             &node_guid) ||
1319                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1320                             &port_guid))
1321                         goto nla_put_vf_failure;
1322         }
1323         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1324         if (!vfvlanlist)
1325                 goto nla_put_vf_failure;
1326         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1327                     &vf_vlan_info)) {
1328                 nla_nest_cancel(skb, vfvlanlist);
1329                 goto nla_put_vf_failure;
1330         }
1331         nla_nest_end(skb, vfvlanlist);
1332         memset(&vf_stats, 0, sizeof(vf_stats));
1333         if (dev->netdev_ops->ndo_get_vf_stats)
1334                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1335                                                 &vf_stats);
1336         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1337         if (!vfstats)
1338                 goto nla_put_vf_failure;
1339         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1340                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1341             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1342                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1343             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1344                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1345             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1346                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1347             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1348                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1349             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1350                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1351             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1352                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1353             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1354                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1355                 nla_nest_cancel(skb, vfstats);
1356                 goto nla_put_vf_failure;
1357         }
1358         nla_nest_end(skb, vfstats);
1359         nla_nest_end(skb, vf);
1360         return 0;
1361
1362 nla_put_vf_failure:
1363         nla_nest_cancel(skb, vf);
1364 nla_put_vfinfo_failure:
1365         nla_nest_cancel(skb, vfinfo);
1366         return -EMSGSIZE;
1367 }
1368
1369 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1370                                            struct net_device *dev,
1371                                            u32 ext_filter_mask)
1372 {
1373         struct nlattr *vfinfo;
1374         int i, num_vfs;
1375
1376         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1377                 return 0;
1378
1379         num_vfs = dev_num_vf(dev->dev.parent);
1380         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1381                 return -EMSGSIZE;
1382
1383         if (!dev->netdev_ops->ndo_get_vf_config)
1384                 return 0;
1385
1386         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1387         if (!vfinfo)
1388                 return -EMSGSIZE;
1389
1390         for (i = 0; i < num_vfs; i++) {
1391                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1392                         return -EMSGSIZE;
1393         }
1394
1395         nla_nest_end(skb, vfinfo);
1396         return 0;
1397 }
1398
1399 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1400 {
1401         struct rtnl_link_ifmap map;
1402
1403         memset(&map, 0, sizeof(map));
1404         map.mem_start   = dev->mem_start;
1405         map.mem_end     = dev->mem_end;
1406         map.base_addr   = dev->base_addr;
1407         map.irq         = dev->irq;
1408         map.dma         = dev->dma;
1409         map.port        = dev->if_port;
1410
1411         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1412                 return -EMSGSIZE;
1413
1414         return 0;
1415 }
1416
1417 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1418 {
1419         const struct bpf_prog *generic_xdp_prog;
1420
1421         ASSERT_RTNL();
1422
1423         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1424         if (!generic_xdp_prog)
1425                 return 0;
1426         return generic_xdp_prog->aux->id;
1427 }
1428
1429 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1430 {
1431         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1432 }
1433
1434 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1435 {
1436         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1437 }
1438
1439 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1440                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1441                                u32 (*get_prog_id)(struct net_device *dev))
1442 {
1443         u32 curr_id;
1444         int err;
1445
1446         curr_id = get_prog_id(dev);
1447         if (!curr_id)
1448                 return 0;
1449
1450         *prog_id = curr_id;
1451         err = nla_put_u32(skb, attr, curr_id);
1452         if (err)
1453                 return err;
1454
1455         if (*mode != XDP_ATTACHED_NONE)
1456                 *mode = XDP_ATTACHED_MULTI;
1457         else
1458                 *mode = tgt_mode;
1459
1460         return 0;
1461 }
1462
1463 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1464 {
1465         struct nlattr *xdp;
1466         u32 prog_id;
1467         int err;
1468         u8 mode;
1469
1470         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1471         if (!xdp)
1472                 return -EMSGSIZE;
1473
1474         prog_id = 0;
1475         mode = XDP_ATTACHED_NONE;
1476         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1477                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1478         if (err)
1479                 goto err_cancel;
1480         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1481                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1482         if (err)
1483                 goto err_cancel;
1484         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1485                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1486         if (err)
1487                 goto err_cancel;
1488
1489         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1490         if (err)
1491                 goto err_cancel;
1492
1493         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1494                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1495                 if (err)
1496                         goto err_cancel;
1497         }
1498
1499         nla_nest_end(skb, xdp);
1500         return 0;
1501
1502 err_cancel:
1503         nla_nest_cancel(skb, xdp);
1504         return err;
1505 }
1506
1507 static u32 rtnl_get_event(unsigned long event)
1508 {
1509         u32 rtnl_event_type = IFLA_EVENT_NONE;
1510
1511         switch (event) {
1512         case NETDEV_REBOOT:
1513                 rtnl_event_type = IFLA_EVENT_REBOOT;
1514                 break;
1515         case NETDEV_FEAT_CHANGE:
1516                 rtnl_event_type = IFLA_EVENT_FEATURES;
1517                 break;
1518         case NETDEV_BONDING_FAILOVER:
1519                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1520                 break;
1521         case NETDEV_NOTIFY_PEERS:
1522                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1523                 break;
1524         case NETDEV_RESEND_IGMP:
1525                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1526                 break;
1527         case NETDEV_CHANGEINFODATA:
1528                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1529                 break;
1530         default:
1531                 break;
1532         }
1533
1534         return rtnl_event_type;
1535 }
1536
1537 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1538 {
1539         const struct net_device *upper_dev;
1540         int ret = 0;
1541
1542         rcu_read_lock();
1543
1544         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1545         if (upper_dev)
1546                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1547
1548         rcu_read_unlock();
1549         return ret;
1550 }
1551
1552 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1553                           bool force)
1554 {
1555         int ifindex = dev_get_iflink(dev);
1556
1557         if (force || dev->ifindex != ifindex)
1558                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1559
1560         return 0;
1561 }
1562
1563 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1564                                               struct net_device *dev)
1565 {
1566         char buf[IFALIASZ];
1567         int ret;
1568
1569         ret = dev_get_alias(dev, buf, sizeof(buf));
1570         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1571 }
1572
1573 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1574                                   const struct net_device *dev,
1575                                   struct net *src_net, gfp_t gfp)
1576 {
1577         bool put_iflink = false;
1578
1579         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1580                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1581
1582                 if (!net_eq(dev_net(dev), link_net)) {
1583                         int id = peernet2id_alloc(src_net, link_net, gfp);
1584
1585                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1586                                 return -EMSGSIZE;
1587
1588                         put_iflink = true;
1589                 }
1590         }
1591
1592         return nla_put_iflink(skb, dev, put_iflink);
1593 }
1594
1595 static int rtnl_fill_link_af(struct sk_buff *skb,
1596                              const struct net_device *dev,
1597                              u32 ext_filter_mask)
1598 {
1599         const struct rtnl_af_ops *af_ops;
1600         struct nlattr *af_spec;
1601
1602         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1603         if (!af_spec)
1604                 return -EMSGSIZE;
1605
1606         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1607                 struct nlattr *af;
1608                 int err;
1609
1610                 if (!af_ops->fill_link_af)
1611                         continue;
1612
1613                 af = nla_nest_start_noflag(skb, af_ops->family);
1614                 if (!af)
1615                         return -EMSGSIZE;
1616
1617                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1618                 /*
1619                  * Caller may return ENODATA to indicate that there
1620                  * was no data to be dumped. This is not an error, it
1621                  * means we should trim the attribute header and
1622                  * continue.
1623                  */
1624                 if (err == -ENODATA)
1625                         nla_nest_cancel(skb, af);
1626                 else if (err < 0)
1627                         return -EMSGSIZE;
1628
1629                 nla_nest_end(skb, af);
1630         }
1631
1632         nla_nest_end(skb, af_spec);
1633         return 0;
1634 }
1635
1636 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1637                                  const struct net_device *dev)
1638 {
1639         struct netdev_name_node *name_node;
1640         int count = 0;
1641
1642         list_for_each_entry(name_node, &dev->name_node->list, list) {
1643                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1644                         return -EMSGSIZE;
1645                 count++;
1646         }
1647         return count;
1648 }
1649
1650 static int rtnl_fill_prop_list(struct sk_buff *skb,
1651                                const struct net_device *dev)
1652 {
1653         struct nlattr *prop_list;
1654         int ret;
1655
1656         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1657         if (!prop_list)
1658                 return -EMSGSIZE;
1659
1660         ret = rtnl_fill_alt_ifnames(skb, dev);
1661         if (ret <= 0)
1662                 goto nest_cancel;
1663
1664         nla_nest_end(skb, prop_list);
1665         return 0;
1666
1667 nest_cancel:
1668         nla_nest_cancel(skb, prop_list);
1669         return ret;
1670 }
1671
1672 static int rtnl_fill_proto_down(struct sk_buff *skb,
1673                                 const struct net_device *dev)
1674 {
1675         struct nlattr *pr;
1676         u32 preason;
1677
1678         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1679                 goto nla_put_failure;
1680
1681         preason = dev->proto_down_reason;
1682         if (!preason)
1683                 return 0;
1684
1685         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1686         if (!pr)
1687                 return -EMSGSIZE;
1688
1689         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1690                 nla_nest_cancel(skb, pr);
1691                 goto nla_put_failure;
1692         }
1693
1694         nla_nest_end(skb, pr);
1695         return 0;
1696
1697 nla_put_failure:
1698         return -EMSGSIZE;
1699 }
1700
1701 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1702                             struct net_device *dev, struct net *src_net,
1703                             int type, u32 pid, u32 seq, u32 change,
1704                             unsigned int flags, u32 ext_filter_mask,
1705                             u32 event, int *new_nsid, int new_ifindex,
1706                             int tgt_netnsid, gfp_t gfp)
1707 {
1708         struct ifinfomsg *ifm;
1709         struct nlmsghdr *nlh;
1710
1711         ASSERT_RTNL();
1712         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1713         if (nlh == NULL)
1714                 return -EMSGSIZE;
1715
1716         ifm = nlmsg_data(nlh);
1717         ifm->ifi_family = AF_UNSPEC;
1718         ifm->__ifi_pad = 0;
1719         ifm->ifi_type = dev->type;
1720         ifm->ifi_index = dev->ifindex;
1721         ifm->ifi_flags = dev_get_flags(dev);
1722         ifm->ifi_change = change;
1723
1724         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1725                 goto nla_put_failure;
1726
1727         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1728             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1729             nla_put_u8(skb, IFLA_OPERSTATE,
1730                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1731             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1732             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1733             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1734             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1735             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1736             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1737             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1738             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1739             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1740 #ifdef CONFIG_RPS
1741             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1742 #endif
1743             put_master_ifindex(skb, dev) ||
1744             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1745             (dev->qdisc &&
1746              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1747             nla_put_ifalias(skb, dev) ||
1748             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1749                         atomic_read(&dev->carrier_up_count) +
1750                         atomic_read(&dev->carrier_down_count)) ||
1751             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1752                         atomic_read(&dev->carrier_up_count)) ||
1753             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1754                         atomic_read(&dev->carrier_down_count)))
1755                 goto nla_put_failure;
1756
1757         if (rtnl_fill_proto_down(skb, dev))
1758                 goto nla_put_failure;
1759
1760         if (event != IFLA_EVENT_NONE) {
1761                 if (nla_put_u32(skb, IFLA_EVENT, event))
1762                         goto nla_put_failure;
1763         }
1764
1765         if (rtnl_fill_link_ifmap(skb, dev))
1766                 goto nla_put_failure;
1767
1768         if (dev->addr_len) {
1769                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1770                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1771                         goto nla_put_failure;
1772         }
1773
1774         if (rtnl_phys_port_id_fill(skb, dev))
1775                 goto nla_put_failure;
1776
1777         if (rtnl_phys_port_name_fill(skb, dev))
1778                 goto nla_put_failure;
1779
1780         if (rtnl_phys_switch_id_fill(skb, dev))
1781                 goto nla_put_failure;
1782
1783         if (rtnl_fill_stats(skb, dev))
1784                 goto nla_put_failure;
1785
1786         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1787                 goto nla_put_failure;
1788
1789         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1790                 goto nla_put_failure;
1791
1792         if (rtnl_xdp_fill(skb, dev))
1793                 goto nla_put_failure;
1794
1795         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1796                 if (rtnl_link_fill(skb, dev) < 0)
1797                         goto nla_put_failure;
1798         }
1799
1800         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1801                 goto nla_put_failure;
1802
1803         if (new_nsid &&
1804             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1805                 goto nla_put_failure;
1806         if (new_ifindex &&
1807             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1808                 goto nla_put_failure;
1809
1810         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1811             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1812                 goto nla_put_failure;
1813
1814         rcu_read_lock();
1815         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1816                 goto nla_put_failure_rcu;
1817         rcu_read_unlock();
1818
1819         if (rtnl_fill_prop_list(skb, dev))
1820                 goto nla_put_failure;
1821
1822         nlmsg_end(skb, nlh);
1823         return 0;
1824
1825 nla_put_failure_rcu:
1826         rcu_read_unlock();
1827 nla_put_failure:
1828         nlmsg_cancel(skb, nlh);
1829         return -EMSGSIZE;
1830 }
1831
1832 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1833         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1834         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1835         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1837         [IFLA_MTU]              = { .type = NLA_U32 },
1838         [IFLA_LINK]             = { .type = NLA_U32 },
1839         [IFLA_MASTER]           = { .type = NLA_U32 },
1840         [IFLA_CARRIER]          = { .type = NLA_U8 },
1841         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1842         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1843         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1844         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1845         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1846         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1847         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1848         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1849          * allow 0-length string (needed to remove an alias).
1850          */
1851         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1852         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1853         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1854         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1855         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1856         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1857         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1858         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1859         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1860         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1861         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1862         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1863         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1864         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1865         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1866         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1867         [IFLA_XDP]              = { .type = NLA_NESTED },
1868         [IFLA_EVENT]            = { .type = NLA_U32 },
1869         [IFLA_GROUP]            = { .type = NLA_U32 },
1870         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1871         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1872         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1873         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1874         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1875         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1876         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1877                                     .len = ALTIFNAMSIZ - 1 },
1878         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1879         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1880 };
1881
1882 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1883         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1884         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1885         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1886         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1887 };
1888
1889 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1890         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1891         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1892         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1893         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1894         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1895         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1896         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1897         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1898         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1899         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1900         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1901         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1902         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1903 };
1904
1905 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1906         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1907         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1908                                     .len = PORT_PROFILE_MAX },
1909         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1910                                       .len = PORT_UUID_MAX },
1911         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1912                                     .len = PORT_UUID_MAX },
1913         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1914         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1915
1916         /* Unused, but we need to keep it here since user space could
1917          * fill it. It's also broken with regard to NLA_BINARY use in
1918          * combination with structs.
1919          */
1920         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1921                                     .len = sizeof(struct ifla_port_vsi) },
1922 };
1923
1924 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1925         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1926         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1927         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1928         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1929         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1930         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1931 };
1932
1933 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1934 {
1935         const struct rtnl_link_ops *ops = NULL;
1936         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1937
1938         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1939                 return NULL;
1940
1941         if (linfo[IFLA_INFO_KIND]) {
1942                 char kind[MODULE_NAME_LEN];
1943
1944                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1945                 ops = rtnl_link_ops_get(kind);
1946         }
1947
1948         return ops;
1949 }
1950
1951 static bool link_master_filtered(struct net_device *dev, int master_idx)
1952 {
1953         struct net_device *master;
1954
1955         if (!master_idx)
1956                 return false;
1957
1958         master = netdev_master_upper_dev_get(dev);
1959         if (!master || master->ifindex != master_idx)
1960                 return true;
1961
1962         return false;
1963 }
1964
1965 static bool link_kind_filtered(const struct net_device *dev,
1966                                const struct rtnl_link_ops *kind_ops)
1967 {
1968         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1969                 return true;
1970
1971         return false;
1972 }
1973
1974 static bool link_dump_filtered(struct net_device *dev,
1975                                int master_idx,
1976                                const struct rtnl_link_ops *kind_ops)
1977 {
1978         if (link_master_filtered(dev, master_idx) ||
1979             link_kind_filtered(dev, kind_ops))
1980                 return true;
1981
1982         return false;
1983 }
1984
1985 /**
1986  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1987  * @sk: netlink socket
1988  * @netnsid: network namespace identifier
1989  *
1990  * Returns the network namespace identified by netnsid on success or an error
1991  * pointer on failure.
1992  */
1993 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1994 {
1995         struct net *net;
1996
1997         net = get_net_ns_by_id(sock_net(sk), netnsid);
1998         if (!net)
1999                 return ERR_PTR(-EINVAL);
2000
2001         /* For now, the caller is required to have CAP_NET_ADMIN in
2002          * the user namespace owning the target net ns.
2003          */
2004         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2005                 put_net(net);
2006                 return ERR_PTR(-EACCES);
2007         }
2008         return net;
2009 }
2010 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2011
2012 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2013                                       bool strict_check, struct nlattr **tb,
2014                                       struct netlink_ext_ack *extack)
2015 {
2016         int hdrlen;
2017
2018         if (strict_check) {
2019                 struct ifinfomsg *ifm;
2020
2021                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2022                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2023                         return -EINVAL;
2024                 }
2025
2026                 ifm = nlmsg_data(nlh);
2027                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2028                     ifm->ifi_change) {
2029                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2030                         return -EINVAL;
2031                 }
2032                 if (ifm->ifi_index) {
2033                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2034                         return -EINVAL;
2035                 }
2036
2037                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2038                                                      IFLA_MAX, ifla_policy,
2039                                                      extack);
2040         }
2041
2042         /* A hack to preserve kernel<->userspace interface.
2043          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2044          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2045          * what iproute2 < v3.9.0 used.
2046          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2047          * attribute, its netlink message is shorter than struct ifinfomsg.
2048          */
2049         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2050                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2051
2052         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2053                                       extack);
2054 }
2055
2056 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2057 {
2058         struct netlink_ext_ack *extack = cb->extack;
2059         const struct nlmsghdr *nlh = cb->nlh;
2060         struct net *net = sock_net(skb->sk);
2061         struct net *tgt_net = net;
2062         int h, s_h;
2063         int idx = 0, s_idx;
2064         struct net_device *dev;
2065         struct hlist_head *head;
2066         struct nlattr *tb[IFLA_MAX+1];
2067         u32 ext_filter_mask = 0;
2068         const struct rtnl_link_ops *kind_ops = NULL;
2069         unsigned int flags = NLM_F_MULTI;
2070         int master_idx = 0;
2071         int netnsid = -1;
2072         int err, i;
2073
2074         s_h = cb->args[0];
2075         s_idx = cb->args[1];
2076
2077         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2078         if (err < 0) {
2079                 if (cb->strict_check)
2080                         return err;
2081
2082                 goto walk_entries;
2083         }
2084
2085         for (i = 0; i <= IFLA_MAX; ++i) {
2086                 if (!tb[i])
2087                         continue;
2088
2089                 /* new attributes should only be added with strict checking */
2090                 switch (i) {
2091                 case IFLA_TARGET_NETNSID:
2092                         netnsid = nla_get_s32(tb[i]);
2093                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2094                         if (IS_ERR(tgt_net)) {
2095                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2096                                 return PTR_ERR(tgt_net);
2097                         }
2098                         break;
2099                 case IFLA_EXT_MASK:
2100                         ext_filter_mask = nla_get_u32(tb[i]);
2101                         break;
2102                 case IFLA_MASTER:
2103                         master_idx = nla_get_u32(tb[i]);
2104                         break;
2105                 case IFLA_LINKINFO:
2106                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2107                         break;
2108                 default:
2109                         if (cb->strict_check) {
2110                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2111                                 return -EINVAL;
2112                         }
2113                 }
2114         }
2115
2116         if (master_idx || kind_ops)
2117                 flags |= NLM_F_DUMP_FILTERED;
2118
2119 walk_entries:
2120         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2121                 idx = 0;
2122                 head = &tgt_net->dev_index_head[h];
2123                 hlist_for_each_entry(dev, head, index_hlist) {
2124                         if (link_dump_filtered(dev, master_idx, kind_ops))
2125                                 goto cont;
2126                         if (idx < s_idx)
2127                                 goto cont;
2128                         err = rtnl_fill_ifinfo(skb, dev, net,
2129                                                RTM_NEWLINK,
2130                                                NETLINK_CB(cb->skb).portid,
2131                                                nlh->nlmsg_seq, 0, flags,
2132                                                ext_filter_mask, 0, NULL, 0,
2133                                                netnsid, GFP_KERNEL);
2134
2135                         if (err < 0) {
2136                                 if (likely(skb->len))
2137                                         goto out;
2138
2139                                 goto out_err;
2140                         }
2141 cont:
2142                         idx++;
2143                 }
2144         }
2145 out:
2146         err = skb->len;
2147 out_err:
2148         cb->args[1] = idx;
2149         cb->args[0] = h;
2150         cb->seq = net->dev_base_seq;
2151         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2152         if (netnsid >= 0)
2153                 put_net(tgt_net);
2154
2155         return err;
2156 }
2157
2158 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2159                         struct netlink_ext_ack *exterr)
2160 {
2161         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2162                                     exterr);
2163 }
2164 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2165
2166 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2167 {
2168         struct net *net;
2169         /* Examine the link attributes and figure out which
2170          * network namespace we are talking about.
2171          */
2172         if (tb[IFLA_NET_NS_PID])
2173                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2174         else if (tb[IFLA_NET_NS_FD])
2175                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2176         else
2177                 net = get_net(src_net);
2178         return net;
2179 }
2180 EXPORT_SYMBOL(rtnl_link_get_net);
2181
2182 /* Figure out which network namespace we are talking about by
2183  * examining the link attributes in the following order:
2184  *
2185  * 1. IFLA_NET_NS_PID
2186  * 2. IFLA_NET_NS_FD
2187  * 3. IFLA_TARGET_NETNSID
2188  */
2189 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2190                                                struct nlattr *tb[])
2191 {
2192         struct net *net;
2193
2194         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2195                 return rtnl_link_get_net(src_net, tb);
2196
2197         if (!tb[IFLA_TARGET_NETNSID])
2198                 return get_net(src_net);
2199
2200         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2201         if (!net)
2202                 return ERR_PTR(-EINVAL);
2203
2204         return net;
2205 }
2206
2207 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2208                                              struct net *src_net,
2209                                              struct nlattr *tb[], int cap)
2210 {
2211         struct net *net;
2212
2213         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2214         if (IS_ERR(net))
2215                 return net;
2216
2217         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2218                 put_net(net);
2219                 return ERR_PTR(-EPERM);
2220         }
2221
2222         return net;
2223 }
2224
2225 /* Verify that rtnetlink requests do not pass additional properties
2226  * potentially referring to different network namespaces.
2227  */
2228 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2229                                     struct netlink_ext_ack *extack,
2230                                     bool netns_id_only)
2231 {
2232
2233         if (netns_id_only) {
2234                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2235                         return 0;
2236
2237                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2238                 return -EOPNOTSUPP;
2239         }
2240
2241         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2242                 goto invalid_attr;
2243
2244         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2245                 goto invalid_attr;
2246
2247         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2248                 goto invalid_attr;
2249
2250         return 0;
2251
2252 invalid_attr:
2253         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2254         return -EINVAL;
2255 }
2256
2257 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2258 {
2259         if (dev) {
2260                 if (tb[IFLA_ADDRESS] &&
2261                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2262                         return -EINVAL;
2263
2264                 if (tb[IFLA_BROADCAST] &&
2265                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2266                         return -EINVAL;
2267         }
2268
2269         if (tb[IFLA_AF_SPEC]) {
2270                 struct nlattr *af;
2271                 int rem, err;
2272
2273                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2274                         const struct rtnl_af_ops *af_ops;
2275
2276                         rcu_read_lock();
2277                         af_ops = rtnl_af_lookup(nla_type(af));
2278                         if (!af_ops) {
2279                                 rcu_read_unlock();
2280                                 return -EAFNOSUPPORT;
2281                         }
2282
2283                         if (!af_ops->set_link_af) {
2284                                 rcu_read_unlock();
2285                                 return -EOPNOTSUPP;
2286                         }
2287
2288                         if (af_ops->validate_link_af) {
2289                                 err = af_ops->validate_link_af(dev, af);
2290                                 if (err < 0) {
2291                                         rcu_read_unlock();
2292                                         return err;
2293                                 }
2294                         }
2295
2296                         rcu_read_unlock();
2297                 }
2298         }
2299
2300         return 0;
2301 }
2302
2303 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2304                                   int guid_type)
2305 {
2306         const struct net_device_ops *ops = dev->netdev_ops;
2307
2308         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2309 }
2310
2311 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2312 {
2313         if (dev->type != ARPHRD_INFINIBAND)
2314                 return -EOPNOTSUPP;
2315
2316         return handle_infiniband_guid(dev, ivt, guid_type);
2317 }
2318
2319 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2320 {
2321         const struct net_device_ops *ops = dev->netdev_ops;
2322         int err = -EINVAL;
2323
2324         if (tb[IFLA_VF_MAC]) {
2325                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2326
2327                 if (ivm->vf >= INT_MAX)
2328                         return -EINVAL;
2329                 err = -EOPNOTSUPP;
2330                 if (ops->ndo_set_vf_mac)
2331                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2332                                                   ivm->mac);
2333                 if (err < 0)
2334                         return err;
2335         }
2336
2337         if (tb[IFLA_VF_VLAN]) {
2338                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2339
2340                 if (ivv->vf >= INT_MAX)
2341                         return -EINVAL;
2342                 err = -EOPNOTSUPP;
2343                 if (ops->ndo_set_vf_vlan)
2344                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2345                                                    ivv->qos,
2346                                                    htons(ETH_P_8021Q));
2347                 if (err < 0)
2348                         return err;
2349         }
2350
2351         if (tb[IFLA_VF_VLAN_LIST]) {
2352                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2353                 struct nlattr *attr;
2354                 int rem, len = 0;
2355
2356                 err = -EOPNOTSUPP;
2357                 if (!ops->ndo_set_vf_vlan)
2358                         return err;
2359
2360                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2361                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2362                             nla_len(attr) < NLA_HDRLEN) {
2363                                 return -EINVAL;
2364                         }
2365                         if (len >= MAX_VLAN_LIST_LEN)
2366                                 return -EOPNOTSUPP;
2367                         ivvl[len] = nla_data(attr);
2368
2369                         len++;
2370                 }
2371                 if (len == 0)
2372                         return -EINVAL;
2373
2374                 if (ivvl[0]->vf >= INT_MAX)
2375                         return -EINVAL;
2376                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2377                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2378                 if (err < 0)
2379                         return err;
2380         }
2381
2382         if (tb[IFLA_VF_TX_RATE]) {
2383                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2384                 struct ifla_vf_info ivf;
2385
2386                 if (ivt->vf >= INT_MAX)
2387                         return -EINVAL;
2388                 err = -EOPNOTSUPP;
2389                 if (ops->ndo_get_vf_config)
2390                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2391                 if (err < 0)
2392                         return err;
2393
2394                 err = -EOPNOTSUPP;
2395                 if (ops->ndo_set_vf_rate)
2396                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2397                                                    ivf.min_tx_rate,
2398                                                    ivt->rate);
2399                 if (err < 0)
2400                         return err;
2401         }
2402
2403         if (tb[IFLA_VF_RATE]) {
2404                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2405
2406                 if (ivt->vf >= INT_MAX)
2407                         return -EINVAL;
2408                 err = -EOPNOTSUPP;
2409                 if (ops->ndo_set_vf_rate)
2410                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2411                                                    ivt->min_tx_rate,
2412                                                    ivt->max_tx_rate);
2413                 if (err < 0)
2414                         return err;
2415         }
2416
2417         if (tb[IFLA_VF_SPOOFCHK]) {
2418                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2419
2420                 if (ivs->vf >= INT_MAX)
2421                         return -EINVAL;
2422                 err = -EOPNOTSUPP;
2423                 if (ops->ndo_set_vf_spoofchk)
2424                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2425                                                        ivs->setting);
2426                 if (err < 0)
2427                         return err;
2428         }
2429
2430         if (tb[IFLA_VF_LINK_STATE]) {
2431                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2432
2433                 if (ivl->vf >= INT_MAX)
2434                         return -EINVAL;
2435                 err = -EOPNOTSUPP;
2436                 if (ops->ndo_set_vf_link_state)
2437                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2438                                                          ivl->link_state);
2439                 if (err < 0)
2440                         return err;
2441         }
2442
2443         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2444                 struct ifla_vf_rss_query_en *ivrssq_en;
2445
2446                 err = -EOPNOTSUPP;
2447                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2448                 if (ivrssq_en->vf >= INT_MAX)
2449                         return -EINVAL;
2450                 if (ops->ndo_set_vf_rss_query_en)
2451                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2452                                                            ivrssq_en->setting);
2453                 if (err < 0)
2454                         return err;
2455         }
2456
2457         if (tb[IFLA_VF_TRUST]) {
2458                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2459
2460                 if (ivt->vf >= INT_MAX)
2461                         return -EINVAL;
2462                 err = -EOPNOTSUPP;
2463                 if (ops->ndo_set_vf_trust)
2464                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2465                 if (err < 0)
2466                         return err;
2467         }
2468
2469         if (tb[IFLA_VF_IB_NODE_GUID]) {
2470                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2471
2472                 if (ivt->vf >= INT_MAX)
2473                         return -EINVAL;
2474                 if (!ops->ndo_set_vf_guid)
2475                         return -EOPNOTSUPP;
2476                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2477         }
2478
2479         if (tb[IFLA_VF_IB_PORT_GUID]) {
2480                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2481
2482                 if (ivt->vf >= INT_MAX)
2483                         return -EINVAL;
2484                 if (!ops->ndo_set_vf_guid)
2485                         return -EOPNOTSUPP;
2486
2487                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2488         }
2489
2490         return err;
2491 }
2492
2493 static int do_set_master(struct net_device *dev, int ifindex,
2494                          struct netlink_ext_ack *extack)
2495 {
2496         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2497         const struct net_device_ops *ops;
2498         int err;
2499
2500         if (upper_dev) {
2501                 if (upper_dev->ifindex == ifindex)
2502                         return 0;
2503                 ops = upper_dev->netdev_ops;
2504                 if (ops->ndo_del_slave) {
2505                         err = ops->ndo_del_slave(upper_dev, dev);
2506                         if (err)
2507                                 return err;
2508                 } else {
2509                         return -EOPNOTSUPP;
2510                 }
2511         }
2512
2513         if (ifindex) {
2514                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2515                 if (!upper_dev)
2516                         return -EINVAL;
2517                 ops = upper_dev->netdev_ops;
2518                 if (ops->ndo_add_slave) {
2519                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2520                         if (err)
2521                                 return err;
2522                 } else {
2523                         return -EOPNOTSUPP;
2524                 }
2525         }
2526         return 0;
2527 }
2528
2529 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2530         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2531         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2532 };
2533
2534 static int do_set_proto_down(struct net_device *dev,
2535                              struct nlattr *nl_proto_down,
2536                              struct nlattr *nl_proto_down_reason,
2537                              struct netlink_ext_ack *extack)
2538 {
2539         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2540         const struct net_device_ops *ops = dev->netdev_ops;
2541         unsigned long mask = 0;
2542         u32 value;
2543         bool proto_down;
2544         int err;
2545
2546         if (!ops->ndo_change_proto_down) {
2547                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2548                 return -EOPNOTSUPP;
2549         }
2550
2551         if (nl_proto_down_reason) {
2552                 err = nla_parse_nested_deprecated(pdreason,
2553                                                   IFLA_PROTO_DOWN_REASON_MAX,
2554                                                   nl_proto_down_reason,
2555                                                   ifla_proto_down_reason_policy,
2556                                                   NULL);
2557                 if (err < 0)
2558                         return err;
2559
2560                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2561                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2562                         return -EINVAL;
2563                 }
2564
2565                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2566
2567                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2568                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2569
2570                 dev_change_proto_down_reason(dev, mask, value);
2571         }
2572
2573         if (nl_proto_down) {
2574                 proto_down = nla_get_u8(nl_proto_down);
2575
2576                 /* Dont turn off protodown if there are active reasons */
2577                 if (!proto_down && dev->proto_down_reason) {
2578                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2579                         return -EBUSY;
2580                 }
2581                 err = dev_change_proto_down(dev,
2582                                             proto_down);
2583                 if (err)
2584                         return err;
2585         }
2586
2587         return 0;
2588 }
2589
2590 #define DO_SETLINK_MODIFIED     0x01
2591 /* notify flag means notify + modified. */
2592 #define DO_SETLINK_NOTIFY       0x03
2593 static int do_setlink(const struct sk_buff *skb,
2594                       struct net_device *dev, struct ifinfomsg *ifm,
2595                       struct netlink_ext_ack *extack,
2596                       struct nlattr **tb, char *ifname, int status)
2597 {
2598         const struct net_device_ops *ops = dev->netdev_ops;
2599         int err;
2600
2601         err = validate_linkmsg(dev, tb);
2602         if (err < 0)
2603                 return err;
2604
2605         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2606                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2607                                                             tb, CAP_NET_ADMIN);
2608                 if (IS_ERR(net)) {
2609                         err = PTR_ERR(net);
2610                         goto errout;
2611                 }
2612
2613                 err = dev_change_net_namespace(dev, net, ifname);
2614                 put_net(net);
2615                 if (err)
2616                         goto errout;
2617                 status |= DO_SETLINK_MODIFIED;
2618         }
2619
2620         if (tb[IFLA_MAP]) {
2621                 struct rtnl_link_ifmap *u_map;
2622                 struct ifmap k_map;
2623
2624                 if (!ops->ndo_set_config) {
2625                         err = -EOPNOTSUPP;
2626                         goto errout;
2627                 }
2628
2629                 if (!netif_device_present(dev)) {
2630                         err = -ENODEV;
2631                         goto errout;
2632                 }
2633
2634                 u_map = nla_data(tb[IFLA_MAP]);
2635                 k_map.mem_start = (unsigned long) u_map->mem_start;
2636                 k_map.mem_end = (unsigned long) u_map->mem_end;
2637                 k_map.base_addr = (unsigned short) u_map->base_addr;
2638                 k_map.irq = (unsigned char) u_map->irq;
2639                 k_map.dma = (unsigned char) u_map->dma;
2640                 k_map.port = (unsigned char) u_map->port;
2641
2642                 err = ops->ndo_set_config(dev, &k_map);
2643                 if (err < 0)
2644                         goto errout;
2645
2646                 status |= DO_SETLINK_NOTIFY;
2647         }
2648
2649         if (tb[IFLA_ADDRESS]) {
2650                 struct sockaddr *sa;
2651                 int len;
2652
2653                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2654                                                   sizeof(*sa));
2655                 sa = kmalloc(len, GFP_KERNEL);
2656                 if (!sa) {
2657                         err = -ENOMEM;
2658                         goto errout;
2659                 }
2660                 sa->sa_family = dev->type;
2661                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2662                        dev->addr_len);
2663                 err = dev_set_mac_address(dev, sa, extack);
2664                 kfree(sa);
2665                 if (err)
2666                         goto errout;
2667                 status |= DO_SETLINK_MODIFIED;
2668         }
2669
2670         if (tb[IFLA_MTU]) {
2671                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2672                 if (err < 0)
2673                         goto errout;
2674                 status |= DO_SETLINK_MODIFIED;
2675         }
2676
2677         if (tb[IFLA_GROUP]) {
2678                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2679                 status |= DO_SETLINK_NOTIFY;
2680         }
2681
2682         /*
2683          * Interface selected by interface index but interface
2684          * name provided implies that a name change has been
2685          * requested.
2686          */
2687         if (ifm->ifi_index > 0 && ifname[0]) {
2688                 err = dev_change_name(dev, ifname);
2689                 if (err < 0)
2690                         goto errout;
2691                 status |= DO_SETLINK_MODIFIED;
2692         }
2693
2694         if (tb[IFLA_IFALIAS]) {
2695                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2696                                     nla_len(tb[IFLA_IFALIAS]));
2697                 if (err < 0)
2698                         goto errout;
2699                 status |= DO_SETLINK_NOTIFY;
2700         }
2701
2702         if (tb[IFLA_BROADCAST]) {
2703                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2704                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2705         }
2706
2707         if (ifm->ifi_flags || ifm->ifi_change) {
2708                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2709                                        extack);
2710                 if (err < 0)
2711                         goto errout;
2712         }
2713
2714         if (tb[IFLA_MASTER]) {
2715                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2716                 if (err)
2717                         goto errout;
2718                 status |= DO_SETLINK_MODIFIED;
2719         }
2720
2721         if (tb[IFLA_CARRIER]) {
2722                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2723                 if (err)
2724                         goto errout;
2725                 status |= DO_SETLINK_MODIFIED;
2726         }
2727
2728         if (tb[IFLA_TXQLEN]) {
2729                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2730
2731                 err = dev_change_tx_queue_len(dev, value);
2732                 if (err)
2733                         goto errout;
2734                 status |= DO_SETLINK_MODIFIED;
2735         }
2736
2737         if (tb[IFLA_GSO_MAX_SIZE]) {
2738                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2739
2740                 if (max_size > GSO_MAX_SIZE) {
2741                         err = -EINVAL;
2742                         goto errout;
2743                 }
2744
2745                 if (dev->gso_max_size ^ max_size) {
2746                         netif_set_gso_max_size(dev, max_size);
2747                         status |= DO_SETLINK_MODIFIED;
2748                 }
2749         }
2750
2751         if (tb[IFLA_GSO_MAX_SEGS]) {
2752                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2753
2754                 if (max_segs > GSO_MAX_SEGS) {
2755                         err = -EINVAL;
2756                         goto errout;
2757                 }
2758
2759                 if (dev->gso_max_segs ^ max_segs) {
2760                         dev->gso_max_segs = max_segs;
2761                         status |= DO_SETLINK_MODIFIED;
2762                 }
2763         }
2764
2765         if (tb[IFLA_OPERSTATE])
2766                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2767
2768         if (tb[IFLA_LINKMODE]) {
2769                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2770
2771                 write_lock_bh(&dev_base_lock);
2772                 if (dev->link_mode ^ value)
2773                         status |= DO_SETLINK_NOTIFY;
2774                 dev->link_mode = value;
2775                 write_unlock_bh(&dev_base_lock);
2776         }
2777
2778         if (tb[IFLA_VFINFO_LIST]) {
2779                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2780                 struct nlattr *attr;
2781                 int rem;
2782
2783                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2784                         if (nla_type(attr) != IFLA_VF_INFO ||
2785                             nla_len(attr) < NLA_HDRLEN) {
2786                                 err = -EINVAL;
2787                                 goto errout;
2788                         }
2789                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2790                                                           attr,
2791                                                           ifla_vf_policy,
2792                                                           NULL);
2793                         if (err < 0)
2794                                 goto errout;
2795                         err = do_setvfinfo(dev, vfinfo);
2796                         if (err < 0)
2797                                 goto errout;
2798                         status |= DO_SETLINK_NOTIFY;
2799                 }
2800         }
2801         err = 0;
2802
2803         if (tb[IFLA_VF_PORTS]) {
2804                 struct nlattr *port[IFLA_PORT_MAX+1];
2805                 struct nlattr *attr;
2806                 int vf;
2807                 int rem;
2808
2809                 err = -EOPNOTSUPP;
2810                 if (!ops->ndo_set_vf_port)
2811                         goto errout;
2812
2813                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2814                         if (nla_type(attr) != IFLA_VF_PORT ||
2815                             nla_len(attr) < NLA_HDRLEN) {
2816                                 err = -EINVAL;
2817                                 goto errout;
2818                         }
2819                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2820                                                           attr,
2821                                                           ifla_port_policy,
2822                                                           NULL);
2823                         if (err < 0)
2824                                 goto errout;
2825                         if (!port[IFLA_PORT_VF]) {
2826                                 err = -EOPNOTSUPP;
2827                                 goto errout;
2828                         }
2829                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2830                         err = ops->ndo_set_vf_port(dev, vf, port);
2831                         if (err < 0)
2832                                 goto errout;
2833                         status |= DO_SETLINK_NOTIFY;
2834                 }
2835         }
2836         err = 0;
2837
2838         if (tb[IFLA_PORT_SELF]) {
2839                 struct nlattr *port[IFLA_PORT_MAX+1];
2840
2841                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2842                                                   tb[IFLA_PORT_SELF],
2843                                                   ifla_port_policy, NULL);
2844                 if (err < 0)
2845                         goto errout;
2846
2847                 err = -EOPNOTSUPP;
2848                 if (ops->ndo_set_vf_port)
2849                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2850                 if (err < 0)
2851                         goto errout;
2852                 status |= DO_SETLINK_NOTIFY;
2853         }
2854
2855         if (tb[IFLA_AF_SPEC]) {
2856                 struct nlattr *af;
2857                 int rem;
2858
2859                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2860                         const struct rtnl_af_ops *af_ops;
2861
2862                         rcu_read_lock();
2863
2864                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2865
2866                         err = af_ops->set_link_af(dev, af);
2867                         if (err < 0) {
2868                                 rcu_read_unlock();
2869                                 goto errout;
2870                         }
2871
2872                         rcu_read_unlock();
2873                         status |= DO_SETLINK_NOTIFY;
2874                 }
2875         }
2876         err = 0;
2877
2878         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2879                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2880                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2881                 if (err)
2882                         goto errout;
2883                 status |= DO_SETLINK_NOTIFY;
2884         }
2885
2886         if (tb[IFLA_XDP]) {
2887                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2888                 u32 xdp_flags = 0;
2889
2890                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2891                                                   tb[IFLA_XDP],
2892                                                   ifla_xdp_policy, NULL);
2893                 if (err < 0)
2894                         goto errout;
2895
2896                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2897                         err = -EINVAL;
2898                         goto errout;
2899                 }
2900
2901                 if (xdp[IFLA_XDP_FLAGS]) {
2902                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2903                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2904                                 err = -EINVAL;
2905                                 goto errout;
2906                         }
2907                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2908                                 err = -EINVAL;
2909                                 goto errout;
2910                         }
2911                 }
2912
2913                 if (xdp[IFLA_XDP_FD]) {
2914                         int expected_fd = -1;
2915
2916                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2917                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2918                                         err = -EINVAL;
2919                                         goto errout;
2920                                 }
2921                                 expected_fd =
2922                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2923                         }
2924
2925                         err = dev_change_xdp_fd(dev, extack,
2926                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2927                                                 expected_fd,
2928                                                 xdp_flags);
2929                         if (err)
2930                                 goto errout;
2931                         status |= DO_SETLINK_NOTIFY;
2932                 }
2933         }
2934
2935 errout:
2936         if (status & DO_SETLINK_MODIFIED) {
2937                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2938                         netdev_state_change(dev);
2939
2940                 if (err < 0)
2941                         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",
2942                                              dev->name);
2943         }
2944
2945         return err;
2946 }
2947
2948 static struct net_device *rtnl_dev_get(struct net *net,
2949                                        struct nlattr *ifname_attr,
2950                                        struct nlattr *altifname_attr,
2951                                        char *ifname)
2952 {
2953         char buffer[ALTIFNAMSIZ];
2954
2955         if (!ifname) {
2956                 ifname = buffer;
2957                 if (ifname_attr)
2958                         nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2959                 else if (altifname_attr)
2960                         nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2961                 else
2962                         return NULL;
2963         }
2964
2965         return __dev_get_by_name(net, ifname);
2966 }
2967
2968 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2969                         struct netlink_ext_ack *extack)
2970 {
2971         struct net *net = sock_net(skb->sk);
2972         struct ifinfomsg *ifm;
2973         struct net_device *dev;
2974         int err;
2975         struct nlattr *tb[IFLA_MAX+1];
2976         char ifname[IFNAMSIZ];
2977
2978         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2979                                      ifla_policy, extack);
2980         if (err < 0)
2981                 goto errout;
2982
2983         err = rtnl_ensure_unique_netns(tb, extack, false);
2984         if (err < 0)
2985                 goto errout;
2986
2987         if (tb[IFLA_IFNAME])
2988                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2989         else
2990                 ifname[0] = '\0';
2991
2992         err = -EINVAL;
2993         ifm = nlmsg_data(nlh);
2994         if (ifm->ifi_index > 0)
2995                 dev = __dev_get_by_index(net, ifm->ifi_index);
2996         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2997                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2998         else
2999                 goto errout;
3000
3001         if (dev == NULL) {
3002                 err = -ENODEV;
3003                 goto errout;
3004         }
3005
3006         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3007 errout:
3008         return err;
3009 }
3010
3011 static int rtnl_group_dellink(const struct net *net, int group)
3012 {
3013         struct net_device *dev, *aux;
3014         LIST_HEAD(list_kill);
3015         bool found = false;
3016
3017         if (!group)
3018                 return -EPERM;
3019
3020         for_each_netdev(net, dev) {
3021                 if (dev->group == group) {
3022                         const struct rtnl_link_ops *ops;
3023
3024                         found = true;
3025                         ops = dev->rtnl_link_ops;
3026                         if (!ops || !ops->dellink)
3027                                 return -EOPNOTSUPP;
3028                 }
3029         }
3030
3031         if (!found)
3032                 return -ENODEV;
3033
3034         for_each_netdev_safe(net, dev, aux) {
3035                 if (dev->group == group) {
3036                         const struct rtnl_link_ops *ops;
3037
3038                         ops = dev->rtnl_link_ops;
3039                         ops->dellink(dev, &list_kill);
3040                 }
3041         }
3042         unregister_netdevice_many(&list_kill);
3043
3044         return 0;
3045 }
3046
3047 int rtnl_delete_link(struct net_device *dev)
3048 {
3049         const struct rtnl_link_ops *ops;
3050         LIST_HEAD(list_kill);
3051
3052         ops = dev->rtnl_link_ops;
3053         if (!ops || !ops->dellink)
3054                 return -EOPNOTSUPP;
3055
3056         ops->dellink(dev, &list_kill);
3057         unregister_netdevice_many(&list_kill);
3058
3059         return 0;
3060 }
3061 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3062
3063 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3064                         struct netlink_ext_ack *extack)
3065 {
3066         struct net *net = sock_net(skb->sk);
3067         struct net *tgt_net = net;
3068         struct net_device *dev = NULL;
3069         struct ifinfomsg *ifm;
3070         struct nlattr *tb[IFLA_MAX+1];
3071         int err;
3072         int netnsid = -1;
3073
3074         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3075                                      ifla_policy, extack);
3076         if (err < 0)
3077                 return err;
3078
3079         err = rtnl_ensure_unique_netns(tb, extack, true);
3080         if (err < 0)
3081                 return err;
3082
3083         if (tb[IFLA_TARGET_NETNSID]) {
3084                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3085                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3086                 if (IS_ERR(tgt_net))
3087                         return PTR_ERR(tgt_net);
3088         }
3089
3090         err = -EINVAL;
3091         ifm = nlmsg_data(nlh);
3092         if (ifm->ifi_index > 0)
3093                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3094         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3095                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3096                                    tb[IFLA_ALT_IFNAME], NULL);
3097         else if (tb[IFLA_GROUP])
3098                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3099         else
3100                 goto out;
3101
3102         if (!dev) {
3103                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3104                         err = -ENODEV;
3105
3106                 goto out;
3107         }
3108
3109         err = rtnl_delete_link(dev);
3110
3111 out:
3112         if (netnsid >= 0)
3113                 put_net(tgt_net);
3114
3115         return err;
3116 }
3117
3118 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3119 {
3120         unsigned int old_flags;
3121         int err;
3122
3123         old_flags = dev->flags;
3124         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3125                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3126                                          NULL);
3127                 if (err < 0)
3128                         return err;
3129         }
3130
3131         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3132                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3133         } else {
3134                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3135                 __dev_notify_flags(dev, old_flags, ~0U);
3136         }
3137         return 0;
3138 }
3139 EXPORT_SYMBOL(rtnl_configure_link);
3140
3141 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3142                                     unsigned char name_assign_type,
3143                                     const struct rtnl_link_ops *ops,
3144                                     struct nlattr *tb[],
3145                                     struct netlink_ext_ack *extack)
3146 {
3147         struct net_device *dev;
3148         unsigned int num_tx_queues = 1;
3149         unsigned int num_rx_queues = 1;
3150
3151         if (tb[IFLA_NUM_TX_QUEUES])
3152                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3153         else if (ops->get_num_tx_queues)
3154                 num_tx_queues = ops->get_num_tx_queues();
3155
3156         if (tb[IFLA_NUM_RX_QUEUES])
3157                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3158         else if (ops->get_num_rx_queues)
3159                 num_rx_queues = ops->get_num_rx_queues();
3160
3161         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3162                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3163                 return ERR_PTR(-EINVAL);
3164         }
3165
3166         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3167                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3168                 return ERR_PTR(-EINVAL);
3169         }
3170
3171         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3172                                ops->setup, num_tx_queues, num_rx_queues);
3173         if (!dev)
3174                 return ERR_PTR(-ENOMEM);
3175
3176         dev_net_set(dev, net);
3177         dev->rtnl_link_ops = ops;
3178         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3179
3180         if (tb[IFLA_MTU]) {
3181                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3182                 int err;
3183
3184                 err = dev_validate_mtu(dev, mtu, extack);
3185                 if (err) {
3186                         free_netdev(dev);
3187                         return ERR_PTR(err);
3188                 }
3189                 dev->mtu = mtu;
3190         }
3191         if (tb[IFLA_ADDRESS]) {
3192                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3193                                 nla_len(tb[IFLA_ADDRESS]));
3194                 dev->addr_assign_type = NET_ADDR_SET;
3195         }
3196         if (tb[IFLA_BROADCAST])
3197                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3198                                 nla_len(tb[IFLA_BROADCAST]));
3199         if (tb[IFLA_TXQLEN])
3200                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3201         if (tb[IFLA_OPERSTATE])
3202                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3203         if (tb[IFLA_LINKMODE])
3204                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3205         if (tb[IFLA_GROUP])
3206                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3207         if (tb[IFLA_GSO_MAX_SIZE])
3208                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3209         if (tb[IFLA_GSO_MAX_SEGS])
3210                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3211
3212         return dev;
3213 }
3214 EXPORT_SYMBOL(rtnl_create_link);
3215
3216 static int rtnl_group_changelink(const struct sk_buff *skb,
3217                 struct net *net, int group,
3218                 struct ifinfomsg *ifm,
3219                 struct netlink_ext_ack *extack,
3220                 struct nlattr **tb)
3221 {
3222         struct net_device *dev, *aux;
3223         int err;
3224
3225         for_each_netdev_safe(net, dev, aux) {
3226                 if (dev->group == group) {
3227                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3228                         if (err < 0)
3229                                 return err;
3230                 }
3231         }
3232
3233         return 0;
3234 }
3235
3236 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3237                           struct nlattr **attr, struct netlink_ext_ack *extack)
3238 {
3239         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3240         unsigned char name_assign_type = NET_NAME_USER;
3241         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3242         const struct rtnl_link_ops *m_ops = NULL;
3243         struct net_device *master_dev = NULL;
3244         struct net *net = sock_net(skb->sk);
3245         const struct rtnl_link_ops *ops;
3246         struct nlattr *tb[IFLA_MAX + 1];
3247         struct net *dest_net, *link_net;
3248         struct nlattr **slave_data;
3249         char kind[MODULE_NAME_LEN];
3250         struct net_device *dev;
3251         struct ifinfomsg *ifm;
3252         char ifname[IFNAMSIZ];
3253         struct nlattr **data;
3254         int err;
3255
3256 #ifdef CONFIG_MODULES
3257 replay:
3258 #endif
3259         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3260                                      ifla_policy, extack);
3261         if (err < 0)
3262                 return err;
3263
3264         err = rtnl_ensure_unique_netns(tb, extack, false);
3265         if (err < 0)
3266                 return err;
3267
3268         if (tb[IFLA_IFNAME])
3269                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3270         else
3271                 ifname[0] = '\0';
3272
3273         ifm = nlmsg_data(nlh);
3274         if (ifm->ifi_index > 0)
3275                 dev = __dev_get_by_index(net, ifm->ifi_index);
3276         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3277                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3278         else
3279                 dev = NULL;
3280
3281         if (dev) {
3282                 master_dev = netdev_master_upper_dev_get(dev);
3283                 if (master_dev)
3284                         m_ops = master_dev->rtnl_link_ops;
3285         }
3286
3287         err = validate_linkmsg(dev, tb);
3288         if (err < 0)
3289                 return err;
3290
3291         if (tb[IFLA_LINKINFO]) {
3292                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3293                                                   tb[IFLA_LINKINFO],
3294                                                   ifla_info_policy, NULL);
3295                 if (err < 0)
3296                         return err;
3297         } else
3298                 memset(linkinfo, 0, sizeof(linkinfo));
3299
3300         if (linkinfo[IFLA_INFO_KIND]) {
3301                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3302                 ops = rtnl_link_ops_get(kind);
3303         } else {
3304                 kind[0] = '\0';
3305                 ops = NULL;
3306         }
3307
3308         data = NULL;
3309         if (ops) {
3310                 if (ops->maxtype > RTNL_MAX_TYPE)
3311                         return -EINVAL;
3312
3313                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3314                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3315                                                           linkinfo[IFLA_INFO_DATA],
3316                                                           ops->policy, extack);
3317                         if (err < 0)
3318                                 return err;
3319                         data = attr;
3320                 }
3321                 if (ops->validate) {
3322                         err = ops->validate(tb, data, extack);
3323                         if (err < 0)
3324                                 return err;
3325                 }
3326         }
3327
3328         slave_data = NULL;
3329         if (m_ops) {
3330                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3331                         return -EINVAL;
3332
3333                 if (m_ops->slave_maxtype &&
3334                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3335                         err = nla_parse_nested_deprecated(slave_attr,
3336                                                           m_ops->slave_maxtype,
3337                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3338                                                           m_ops->slave_policy,
3339                                                           extack);
3340                         if (err < 0)
3341                                 return err;
3342                         slave_data = slave_attr;
3343                 }
3344         }
3345
3346         if (dev) {
3347                 int status = 0;
3348
3349                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3350                         return -EEXIST;
3351                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3352                         return -EOPNOTSUPP;
3353
3354                 if (linkinfo[IFLA_INFO_DATA]) {
3355                         if (!ops || ops != dev->rtnl_link_ops ||
3356                             !ops->changelink)
3357                                 return -EOPNOTSUPP;
3358
3359                         err = ops->changelink(dev, tb, data, extack);
3360                         if (err < 0)
3361                                 return err;
3362                         status |= DO_SETLINK_NOTIFY;
3363                 }
3364
3365                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3366                         if (!m_ops || !m_ops->slave_changelink)
3367                                 return -EOPNOTSUPP;
3368
3369                         err = m_ops->slave_changelink(master_dev, dev, tb,
3370                                                       slave_data, extack);
3371                         if (err < 0)
3372                                 return err;
3373                         status |= DO_SETLINK_NOTIFY;
3374                 }
3375
3376                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3377         }
3378
3379         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3380                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3381                         return rtnl_group_changelink(skb, net,
3382                                                 nla_get_u32(tb[IFLA_GROUP]),
3383                                                 ifm, extack, tb);
3384                 return -ENODEV;
3385         }
3386
3387         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3388                 return -EOPNOTSUPP;
3389
3390         if (!ops) {
3391 #ifdef CONFIG_MODULES
3392                 if (kind[0]) {
3393                         __rtnl_unlock();
3394                         request_module("rtnl-link-%s", kind);
3395                         rtnl_lock();
3396                         ops = rtnl_link_ops_get(kind);
3397                         if (ops)
3398                                 goto replay;
3399                 }
3400 #endif
3401                 NL_SET_ERR_MSG(extack, "Unknown device type");
3402                 return -EOPNOTSUPP;
3403         }
3404
3405         if (!ops->setup)
3406                 return -EOPNOTSUPP;
3407
3408         if (!ifname[0]) {
3409                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3410                 name_assign_type = NET_NAME_ENUM;
3411         }
3412
3413         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3414         if (IS_ERR(dest_net))
3415                 return PTR_ERR(dest_net);
3416
3417         if (tb[IFLA_LINK_NETNSID]) {
3418                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3419
3420                 link_net = get_net_ns_by_id(dest_net, id);
3421                 if (!link_net) {
3422                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3423                         err =  -EINVAL;
3424                         goto out;
3425                 }
3426                 err = -EPERM;
3427                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3428                         goto out;
3429         } else {
3430                 link_net = NULL;
3431         }
3432
3433         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3434                                name_assign_type, ops, tb, extack);
3435         if (IS_ERR(dev)) {
3436                 err = PTR_ERR(dev);
3437                 goto out;
3438         }
3439
3440         dev->ifindex = ifm->ifi_index;
3441
3442         if (ops->newlink)
3443                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3444         else
3445                 err = register_netdevice(dev);
3446         if (err < 0) {
3447                 free_netdev(dev);
3448                 goto out;
3449         }
3450
3451         err = rtnl_configure_link(dev, ifm);
3452         if (err < 0)
3453                 goto out_unregister;
3454         if (link_net) {
3455                 err = dev_change_net_namespace(dev, dest_net, ifname);
3456                 if (err < 0)
3457                         goto out_unregister;
3458         }
3459         if (tb[IFLA_MASTER]) {
3460                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3461                 if (err)
3462                         goto out_unregister;
3463         }
3464 out:
3465         if (link_net)
3466                 put_net(link_net);
3467         put_net(dest_net);
3468         return err;
3469 out_unregister:
3470         if (ops->newlink) {
3471                 LIST_HEAD(list_kill);
3472
3473                 ops->dellink(dev, &list_kill);
3474                 unregister_netdevice_many(&list_kill);
3475         } else {
3476                 unregister_netdevice(dev);
3477         }
3478         goto out;
3479 }
3480
3481 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3482                         struct netlink_ext_ack *extack)
3483 {
3484         struct nlattr **attr;
3485         int ret;
3486
3487         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3488         if (!attr)
3489                 return -ENOMEM;
3490
3491         ret = __rtnl_newlink(skb, nlh, attr, extack);
3492         kfree(attr);
3493         return ret;
3494 }
3495
3496 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3497                                   const struct nlmsghdr *nlh,
3498                                   struct nlattr **tb,
3499                                   struct netlink_ext_ack *extack)
3500 {
3501         struct ifinfomsg *ifm;
3502         int i, err;
3503
3504         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3505                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3506                 return -EINVAL;
3507         }
3508
3509         if (!netlink_strict_get_check(skb))
3510                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3511                                               ifla_policy, extack);
3512
3513         ifm = nlmsg_data(nlh);
3514         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3515             ifm->ifi_change) {
3516                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3517                 return -EINVAL;
3518         }
3519
3520         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3521                                             ifla_policy, extack);
3522         if (err)
3523                 return err;
3524
3525         for (i = 0; i <= IFLA_MAX; i++) {
3526                 if (!tb[i])
3527                         continue;
3528
3529                 switch (i) {
3530                 case IFLA_IFNAME:
3531                 case IFLA_ALT_IFNAME:
3532                 case IFLA_EXT_MASK:
3533                 case IFLA_TARGET_NETNSID:
3534                         break;
3535                 default:
3536                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3537                         return -EINVAL;
3538                 }
3539         }
3540
3541         return 0;
3542 }
3543
3544 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3545                         struct netlink_ext_ack *extack)
3546 {
3547         struct net *net = sock_net(skb->sk);
3548         struct net *tgt_net = net;
3549         struct ifinfomsg *ifm;
3550         struct nlattr *tb[IFLA_MAX+1];
3551         struct net_device *dev = NULL;
3552         struct sk_buff *nskb;
3553         int netnsid = -1;
3554         int err;
3555         u32 ext_filter_mask = 0;
3556
3557         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3558         if (err < 0)
3559                 return err;
3560
3561         err = rtnl_ensure_unique_netns(tb, extack, true);
3562         if (err < 0)
3563                 return err;
3564
3565         if (tb[IFLA_TARGET_NETNSID]) {
3566                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3567                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3568                 if (IS_ERR(tgt_net))
3569                         return PTR_ERR(tgt_net);
3570         }
3571
3572         if (tb[IFLA_EXT_MASK])
3573                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3574
3575         err = -EINVAL;
3576         ifm = nlmsg_data(nlh);
3577         if (ifm->ifi_index > 0)
3578                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3579         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3580                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3581                                    tb[IFLA_ALT_IFNAME], NULL);
3582         else
3583                 goto out;
3584
3585         err = -ENODEV;
3586         if (dev == NULL)
3587                 goto out;
3588
3589         err = -ENOBUFS;
3590         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3591         if (nskb == NULL)
3592                 goto out;
3593
3594         err = rtnl_fill_ifinfo(nskb, dev, net,
3595                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3596                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3597                                0, NULL, 0, netnsid, GFP_KERNEL);
3598         if (err < 0) {
3599                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3600                 WARN_ON(err == -EMSGSIZE);
3601                 kfree_skb(nskb);
3602         } else
3603                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3604 out:
3605         if (netnsid >= 0)
3606                 put_net(tgt_net);
3607
3608         return err;
3609 }
3610
3611 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3612                            bool *changed, struct netlink_ext_ack *extack)
3613 {
3614         char *alt_ifname;
3615         int err;
3616
3617         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3618         if (err)
3619                 return err;
3620
3621         alt_ifname = nla_strdup(attr, GFP_KERNEL);
3622         if (!alt_ifname)
3623                 return -ENOMEM;
3624
3625         if (cmd == RTM_NEWLINKPROP) {
3626                 err = netdev_name_node_alt_create(dev, alt_ifname);
3627                 if (!err)
3628                         alt_ifname = NULL;
3629         } else if (cmd == RTM_DELLINKPROP) {
3630                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3631         } else {
3632                 WARN_ON_ONCE(1);
3633                 err = -EINVAL;
3634         }
3635
3636         kfree(alt_ifname);
3637         if (!err)
3638                 *changed = true;
3639         return err;
3640 }
3641
3642 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3643                          struct netlink_ext_ack *extack)
3644 {
3645         struct net *net = sock_net(skb->sk);
3646         struct nlattr *tb[IFLA_MAX + 1];
3647         struct net_device *dev;
3648         struct ifinfomsg *ifm;
3649         bool changed = false;
3650         struct nlattr *attr;
3651         int err, rem;
3652
3653         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3654         if (err)
3655                 return err;
3656
3657         err = rtnl_ensure_unique_netns(tb, extack, true);
3658         if (err)
3659                 return err;
3660
3661         ifm = nlmsg_data(nlh);
3662         if (ifm->ifi_index > 0)
3663                 dev = __dev_get_by_index(net, ifm->ifi_index);
3664         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3665                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3666                                    tb[IFLA_ALT_IFNAME], NULL);
3667         else
3668                 return -EINVAL;
3669
3670         if (!dev)
3671                 return -ENODEV;
3672
3673         if (!tb[IFLA_PROP_LIST])
3674                 return 0;
3675
3676         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3677                 switch (nla_type(attr)) {
3678                 case IFLA_ALT_IFNAME:
3679                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3680                         if (err)
3681                                 return err;
3682                         break;
3683                 }
3684         }
3685
3686         if (changed)
3687                 netdev_state_change(dev);
3688         return 0;
3689 }
3690
3691 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3692                             struct netlink_ext_ack *extack)
3693 {
3694         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3695 }
3696
3697 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3698                             struct netlink_ext_ack *extack)
3699 {
3700         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3701 }
3702
3703 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3704 {
3705         struct net *net = sock_net(skb->sk);
3706         size_t min_ifinfo_dump_size = 0;
3707         struct nlattr *tb[IFLA_MAX+1];
3708         u32 ext_filter_mask = 0;
3709         struct net_device *dev;
3710         int hdrlen;
3711
3712         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3713         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3714                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3715
3716         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3717                 if (tb[IFLA_EXT_MASK])
3718                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3719         }
3720
3721         if (!ext_filter_mask)
3722                 return NLMSG_GOODSIZE;
3723         /*
3724          * traverse the list of net devices and compute the minimum
3725          * buffer size based upon the filter mask.
3726          */
3727         rcu_read_lock();
3728         for_each_netdev_rcu(net, dev) {
3729                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3730                                            if_nlmsg_size(dev, ext_filter_mask));
3731         }
3732         rcu_read_unlock();
3733
3734         return nlmsg_total_size(min_ifinfo_dump_size);
3735 }
3736
3737 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3738 {
3739         int idx;
3740         int s_idx = cb->family;
3741         int type = cb->nlh->nlmsg_type - RTM_BASE;
3742         int ret = 0;
3743
3744         if (s_idx == 0)
3745                 s_idx = 1;
3746
3747         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3748                 struct rtnl_link __rcu **tab;
3749                 struct rtnl_link *link;
3750                 rtnl_dumpit_func dumpit;
3751
3752                 if (idx < s_idx || idx == PF_PACKET)
3753                         continue;
3754
3755                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3756                         continue;
3757
3758                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3759                 if (!tab)
3760                         continue;
3761
3762                 link = rcu_dereference_rtnl(tab[type]);
3763                 if (!link)
3764                         continue;
3765
3766                 dumpit = link->dumpit;
3767                 if (!dumpit)
3768                         continue;
3769
3770                 if (idx > s_idx) {
3771                         memset(&cb->args[0], 0, sizeof(cb->args));
3772                         cb->prev_seq = 0;
3773                         cb->seq = 0;
3774                 }
3775                 ret = dumpit(skb, cb);
3776                 if (ret)
3777                         break;
3778         }
3779         cb->family = idx;
3780
3781         return skb->len ? : ret;
3782 }
3783
3784 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3785                                        unsigned int change,
3786                                        u32 event, gfp_t flags, int *new_nsid,
3787                                        int new_ifindex)
3788 {
3789         struct net *net = dev_net(dev);
3790         struct sk_buff *skb;
3791         int err = -ENOBUFS;
3792         size_t if_info_size;
3793
3794         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3795         if (skb == NULL)
3796                 goto errout;
3797
3798         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3799                                type, 0, 0, change, 0, 0, event,
3800                                new_nsid, new_ifindex, -1, flags);
3801         if (err < 0) {
3802                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3803                 WARN_ON(err == -EMSGSIZE);
3804                 kfree_skb(skb);
3805                 goto errout;
3806         }
3807         return skb;
3808 errout:
3809         if (err < 0)
3810                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3811         return NULL;
3812 }
3813
3814 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3815 {
3816         struct net *net = dev_net(dev);
3817
3818         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3819 }
3820
3821 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3822                                unsigned int change, u32 event,
3823                                gfp_t flags, int *new_nsid, int new_ifindex)
3824 {
3825         struct sk_buff *skb;
3826
3827         if (dev->reg_state != NETREG_REGISTERED)
3828                 return;
3829
3830         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3831                                      new_ifindex);
3832         if (skb)
3833                 rtmsg_ifinfo_send(skb, dev, flags);
3834 }
3835
3836 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3837                   gfp_t flags)
3838 {
3839         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3840                            NULL, 0);
3841 }
3842
3843 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3844                          gfp_t flags, int *new_nsid, int new_ifindex)
3845 {
3846         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3847                            new_nsid, new_ifindex);
3848 }
3849
3850 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3851                                    struct net_device *dev,
3852                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3853                                    int type, unsigned int flags,
3854                                    int nlflags, u16 ndm_state)
3855 {
3856         struct nlmsghdr *nlh;
3857         struct ndmsg *ndm;
3858
3859         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3860         if (!nlh)
3861                 return -EMSGSIZE;
3862
3863         ndm = nlmsg_data(nlh);
3864         ndm->ndm_family  = AF_BRIDGE;
3865         ndm->ndm_pad1    = 0;
3866         ndm->ndm_pad2    = 0;
3867         ndm->ndm_flags   = flags;
3868         ndm->ndm_type    = 0;
3869         ndm->ndm_ifindex = dev->ifindex;
3870         ndm->ndm_state   = ndm_state;
3871
3872         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3873                 goto nla_put_failure;
3874         if (vid)
3875                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3876                         goto nla_put_failure;
3877
3878         nlmsg_end(skb, nlh);
3879         return 0;
3880
3881 nla_put_failure:
3882         nlmsg_cancel(skb, nlh);
3883         return -EMSGSIZE;
3884 }
3885
3886 static inline size_t rtnl_fdb_nlmsg_size(void)
3887 {
3888         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3889                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3890                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3891                0;
3892 }
3893
3894 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3895                             u16 ndm_state)
3896 {
3897         struct net *net = dev_net(dev);
3898         struct sk_buff *skb;
3899         int err = -ENOBUFS;
3900
3901         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3902         if (!skb)
3903                 goto errout;
3904
3905         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3906                                       0, 0, type, NTF_SELF, 0, ndm_state);
3907         if (err < 0) {
3908                 kfree_skb(skb);
3909                 goto errout;
3910         }
3911
3912         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3913         return;
3914 errout:
3915         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3916 }
3917
3918 /*
3919  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3920  */
3921 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3922                      struct nlattr *tb[],
3923                      struct net_device *dev,
3924                      const unsigned char *addr, u16 vid,
3925                      u16 flags)
3926 {
3927         int err = -EINVAL;
3928
3929         /* If aging addresses are supported device will need to
3930          * implement its own handler for this.
3931          */
3932         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3933                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3934                 return err;
3935         }
3936
3937         if (vid) {
3938                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3939                 return err;
3940         }
3941
3942         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3943                 err = dev_uc_add_excl(dev, addr);
3944         else if (is_multicast_ether_addr(addr))
3945                 err = dev_mc_add_excl(dev, addr);
3946
3947         /* Only return duplicate errors if NLM_F_EXCL is set */
3948         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3949                 err = 0;
3950
3951         return err;
3952 }
3953 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3954
3955 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3956                          struct netlink_ext_ack *extack)
3957 {
3958         u16 vid = 0;
3959
3960         if (vlan_attr) {
3961                 if (nla_len(vlan_attr) != sizeof(u16)) {
3962                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3963                         return -EINVAL;
3964                 }
3965
3966                 vid = nla_get_u16(vlan_attr);
3967
3968                 if (!vid || vid >= VLAN_VID_MASK) {
3969                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3970                         return -EINVAL;
3971                 }
3972         }
3973         *p_vid = vid;
3974         return 0;
3975 }
3976
3977 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3978                         struct netlink_ext_ack *extack)
3979 {
3980         struct net *net = sock_net(skb->sk);
3981         struct ndmsg *ndm;
3982         struct nlattr *tb[NDA_MAX+1];
3983         struct net_device *dev;
3984         u8 *addr;
3985         u16 vid;
3986         int err;
3987
3988         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3989                                      extack);
3990         if (err < 0)
3991                 return err;
3992
3993         ndm = nlmsg_data(nlh);
3994         if (ndm->ndm_ifindex == 0) {
3995                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3996                 return -EINVAL;
3997         }
3998
3999         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4000         if (dev == NULL) {
4001                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4002                 return -ENODEV;
4003         }
4004
4005         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4006                 NL_SET_ERR_MSG(extack, "invalid address");
4007                 return -EINVAL;
4008         }
4009
4010         if (dev->type != ARPHRD_ETHER) {
4011                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4012                 return -EINVAL;
4013         }
4014
4015         addr = nla_data(tb[NDA_LLADDR]);
4016
4017         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4018         if (err)
4019                 return err;
4020
4021         err = -EOPNOTSUPP;
4022
4023         /* Support fdb on master device the net/bridge default case */
4024         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4025             netif_is_bridge_port(dev)) {
4026                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4027                 const struct net_device_ops *ops = br_dev->netdev_ops;
4028
4029                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4030                                        nlh->nlmsg_flags, extack);
4031                 if (err)
4032                         goto out;
4033                 else
4034                         ndm->ndm_flags &= ~NTF_MASTER;
4035         }
4036
4037         /* Embedded bridge, macvlan, and any other device support */
4038         if ((ndm->ndm_flags & NTF_SELF)) {
4039                 if (dev->netdev_ops->ndo_fdb_add)
4040                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4041                                                            vid,
4042                                                            nlh->nlmsg_flags,
4043                                                            extack);
4044                 else
4045                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4046                                                nlh->nlmsg_flags);
4047
4048                 if (!err) {
4049                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4050                                         ndm->ndm_state);
4051                         ndm->ndm_flags &= ~NTF_SELF;
4052                 }
4053         }
4054 out:
4055         return err;
4056 }
4057
4058 /*
4059  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4060  */
4061 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4062                      struct nlattr *tb[],
4063                      struct net_device *dev,
4064                      const unsigned char *addr, u16 vid)
4065 {
4066         int err = -EINVAL;
4067
4068         /* If aging addresses are supported device will need to
4069          * implement its own handler for this.
4070          */
4071         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4072                 pr_info("%s: FDB only supports static addresses\n", dev->name);
4073                 return err;
4074         }
4075
4076         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4077                 err = dev_uc_del(dev, addr);
4078         else if (is_multicast_ether_addr(addr))
4079                 err = dev_mc_del(dev, addr);
4080
4081         return err;
4082 }
4083 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4084
4085 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4086                         struct netlink_ext_ack *extack)
4087 {
4088         struct net *net = sock_net(skb->sk);
4089         struct ndmsg *ndm;
4090         struct nlattr *tb[NDA_MAX+1];
4091         struct net_device *dev;
4092         __u8 *addr;
4093         int err;
4094         u16 vid;
4095
4096         if (!netlink_capable(skb, CAP_NET_ADMIN))
4097                 return -EPERM;
4098
4099         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4100                                      extack);
4101         if (err < 0)
4102                 return err;
4103
4104         ndm = nlmsg_data(nlh);
4105         if (ndm->ndm_ifindex == 0) {
4106                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4107                 return -EINVAL;
4108         }
4109
4110         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4111         if (dev == NULL) {
4112                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4113                 return -ENODEV;
4114         }
4115
4116         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4117                 NL_SET_ERR_MSG(extack, "invalid address");
4118                 return -EINVAL;
4119         }
4120
4121         if (dev->type != ARPHRD_ETHER) {
4122                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4123                 return -EINVAL;
4124         }
4125
4126         addr = nla_data(tb[NDA_LLADDR]);
4127
4128         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4129         if (err)
4130                 return err;
4131
4132         err = -EOPNOTSUPP;
4133
4134         /* Support fdb on master device the net/bridge default case */
4135         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4136             netif_is_bridge_port(dev)) {
4137                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4138                 const struct net_device_ops *ops = br_dev->netdev_ops;
4139
4140                 if (ops->ndo_fdb_del)
4141                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4142
4143                 if (err)
4144                         goto out;
4145                 else
4146                         ndm->ndm_flags &= ~NTF_MASTER;
4147         }
4148
4149         /* Embedded bridge, macvlan, and any other device support */
4150         if (ndm->ndm_flags & NTF_SELF) {
4151                 if (dev->netdev_ops->ndo_fdb_del)
4152                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4153                                                            vid);
4154                 else
4155                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4156
4157                 if (!err) {
4158                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4159                                         ndm->ndm_state);
4160                         ndm->ndm_flags &= ~NTF_SELF;
4161                 }
4162         }
4163 out:
4164         return err;
4165 }
4166
4167 static int nlmsg_populate_fdb(struct sk_buff *skb,
4168                               struct netlink_callback *cb,
4169                               struct net_device *dev,
4170                               int *idx,
4171                               struct netdev_hw_addr_list *list)
4172 {
4173         struct netdev_hw_addr *ha;
4174         int err;
4175         u32 portid, seq;
4176
4177         portid = NETLINK_CB(cb->skb).portid;
4178         seq = cb->nlh->nlmsg_seq;
4179
4180         list_for_each_entry(ha, &list->list, list) {
4181                 if (*idx < cb->args[2])
4182                         goto skip;
4183
4184                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4185                                               portid, seq,
4186                                               RTM_NEWNEIGH, NTF_SELF,
4187                                               NLM_F_MULTI, NUD_PERMANENT);
4188                 if (err < 0)
4189                         return err;
4190 skip:
4191                 *idx += 1;
4192         }
4193         return 0;
4194 }
4195
4196 /**
4197  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4198  * @skb: socket buffer to store message in
4199  * @cb: netlink callback
4200  * @dev: netdevice
4201  * @filter_dev: ignored
4202  * @idx: the number of FDB table entries dumped is added to *@idx
4203  *
4204  * Default netdevice operation to dump the existing unicast address list.
4205  * Returns number of addresses from list put in skb.
4206  */
4207 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4208                       struct netlink_callback *cb,
4209                       struct net_device *dev,
4210                       struct net_device *filter_dev,
4211                       int *idx)
4212 {
4213         int err;
4214
4215         if (dev->type != ARPHRD_ETHER)
4216                 return -EINVAL;
4217
4218         netif_addr_lock_bh(dev);
4219         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4220         if (err)
4221                 goto out;
4222         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4223 out:
4224         netif_addr_unlock_bh(dev);
4225         return err;
4226 }
4227 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4228
4229 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4230                                  int *br_idx, int *brport_idx,
4231                                  struct netlink_ext_ack *extack)
4232 {
4233         struct nlattr *tb[NDA_MAX + 1];
4234         struct ndmsg *ndm;
4235         int err, i;
4236
4237         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4238                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4239                 return -EINVAL;
4240         }
4241
4242         ndm = nlmsg_data(nlh);
4243         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4244             ndm->ndm_flags || ndm->ndm_type) {
4245                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4246                 return -EINVAL;
4247         }
4248
4249         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4250                                             NDA_MAX, NULL, extack);
4251         if (err < 0)
4252                 return err;
4253
4254         *brport_idx = ndm->ndm_ifindex;
4255         for (i = 0; i <= NDA_MAX; ++i) {
4256                 if (!tb[i])
4257                         continue;
4258
4259                 switch (i) {
4260                 case NDA_IFINDEX:
4261                         if (nla_len(tb[i]) != sizeof(u32)) {
4262                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4263                                 return -EINVAL;
4264                         }
4265                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4266                         break;
4267                 case NDA_MASTER:
4268                         if (nla_len(tb[i]) != sizeof(u32)) {
4269                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4270                                 return -EINVAL;
4271                         }
4272                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4273                         break;
4274                 default:
4275                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4276                         return -EINVAL;
4277                 }
4278         }
4279
4280         return 0;
4281 }
4282
4283 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4284                                  int *br_idx, int *brport_idx,
4285                                  struct netlink_ext_ack *extack)
4286 {
4287         struct nlattr *tb[IFLA_MAX+1];
4288         int err;
4289
4290         /* A hack to preserve kernel<->userspace interface.
4291          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4292          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4293          * So, check for ndmsg with an optional u32 attribute (not used here).
4294          * Fortunately these sizes don't conflict with the size of ifinfomsg
4295          * with an optional attribute.
4296          */
4297         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4298             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4299              nla_attr_size(sizeof(u32)))) {
4300                 struct ifinfomsg *ifm;
4301
4302                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4303                                              tb, IFLA_MAX, ifla_policy,
4304                                              extack);
4305                 if (err < 0) {
4306                         return -EINVAL;
4307                 } else if (err == 0) {
4308                         if (tb[IFLA_MASTER])
4309                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4310                 }
4311
4312                 ifm = nlmsg_data(nlh);
4313                 *brport_idx = ifm->ifi_index;
4314         }
4315         return 0;
4316 }
4317
4318 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4319 {
4320         struct net_device *dev;
4321         struct net_device *br_dev = NULL;
4322         const struct net_device_ops *ops = NULL;
4323         const struct net_device_ops *cops = NULL;
4324         struct net *net = sock_net(skb->sk);
4325         struct hlist_head *head;
4326         int brport_idx = 0;
4327         int br_idx = 0;
4328         int h, s_h;
4329         int idx = 0, s_idx;
4330         int err = 0;
4331         int fidx = 0;
4332
4333         if (cb->strict_check)
4334                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4335                                             cb->extack);
4336         else
4337                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4338                                             cb->extack);
4339         if (err < 0)
4340                 return err;
4341
4342         if (br_idx) {
4343                 br_dev = __dev_get_by_index(net, br_idx);
4344                 if (!br_dev)
4345                         return -ENODEV;
4346
4347                 ops = br_dev->netdev_ops;
4348         }
4349
4350         s_h = cb->args[0];
4351         s_idx = cb->args[1];
4352
4353         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4354                 idx = 0;
4355                 head = &net->dev_index_head[h];
4356                 hlist_for_each_entry(dev, head, index_hlist) {
4357
4358                         if (brport_idx && (dev->ifindex != brport_idx))
4359                                 continue;
4360
4361                         if (!br_idx) { /* user did not specify a specific bridge */
4362                                 if (netif_is_bridge_port(dev)) {
4363                                         br_dev = netdev_master_upper_dev_get(dev);
4364                                         cops = br_dev->netdev_ops;
4365                                 }
4366                         } else {
4367                                 if (dev != br_dev &&
4368                                     !netif_is_bridge_port(dev))
4369                                         continue;
4370
4371                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4372                                     !(dev->priv_flags & IFF_EBRIDGE))
4373                                         continue;
4374                                 cops = ops;
4375                         }
4376
4377                         if (idx < s_idx)
4378                                 goto cont;
4379
4380                         if (netif_is_bridge_port(dev)) {
4381                                 if (cops && cops->ndo_fdb_dump) {
4382                                         err = cops->ndo_fdb_dump(skb, cb,
4383                                                                 br_dev, dev,
4384                                                                 &fidx);
4385                                         if (err == -EMSGSIZE)
4386                                                 goto out;
4387                                 }
4388                         }
4389
4390                         if (dev->netdev_ops->ndo_fdb_dump)
4391                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4392                                                                     dev, NULL,
4393                                                                     &fidx);
4394                         else
4395                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4396                                                         &fidx);
4397                         if (err == -EMSGSIZE)
4398                                 goto out;
4399
4400                         cops = NULL;
4401
4402                         /* reset fdb offset to 0 for rest of the interfaces */
4403                         cb->args[2] = 0;
4404                         fidx = 0;
4405 cont:
4406                         idx++;
4407                 }
4408         }
4409
4410 out:
4411         cb->args[0] = h;
4412         cb->args[1] = idx;
4413         cb->args[2] = fidx;
4414
4415         return skb->len;
4416 }
4417
4418 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4419                                 struct nlattr **tb, u8 *ndm_flags,
4420                                 int *br_idx, int *brport_idx, u8 **addr,
4421                                 u16 *vid, struct netlink_ext_ack *extack)
4422 {
4423         struct ndmsg *ndm;
4424         int err, i;
4425
4426         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4427                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4428                 return -EINVAL;
4429         }
4430
4431         ndm = nlmsg_data(nlh);
4432         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4433             ndm->ndm_type) {
4434                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4435                 return -EINVAL;
4436         }
4437
4438         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4439                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4440                 return -EINVAL;
4441         }
4442
4443         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4444                                             NDA_MAX, nda_policy, extack);
4445         if (err < 0)
4446                 return err;
4447
4448         *ndm_flags = ndm->ndm_flags;
4449         *brport_idx = ndm->ndm_ifindex;
4450         for (i = 0; i <= NDA_MAX; ++i) {
4451                 if (!tb[i])
4452                         continue;
4453
4454                 switch (i) {
4455                 case NDA_MASTER:
4456                         *br_idx = nla_get_u32(tb[i]);
4457                         break;
4458                 case NDA_LLADDR:
4459                         if (nla_len(tb[i]) != ETH_ALEN) {
4460                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4461                                 return -EINVAL;
4462                         }
4463                         *addr = nla_data(tb[i]);
4464                         break;
4465                 case NDA_VLAN:
4466                         err = fdb_vid_parse(tb[i], vid, extack);
4467                         if (err)
4468                                 return err;
4469                         break;
4470                 case NDA_VNI:
4471                         break;
4472                 default:
4473                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4474                         return -EINVAL;
4475                 }
4476         }
4477
4478         return 0;
4479 }
4480
4481 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4482                         struct netlink_ext_ack *extack)
4483 {
4484         struct net_device *dev = NULL, *br_dev = NULL;
4485         const struct net_device_ops *ops = NULL;
4486         struct net *net = sock_net(in_skb->sk);
4487         struct nlattr *tb[NDA_MAX + 1];
4488         struct sk_buff *skb;
4489         int brport_idx = 0;
4490         u8 ndm_flags = 0;
4491         int br_idx = 0;
4492         u8 *addr = NULL;
4493         u16 vid = 0;
4494         int err;
4495
4496         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4497                                    &brport_idx, &addr, &vid, extack);
4498         if (err < 0)
4499                 return err;
4500
4501         if (!addr) {
4502                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4503                 return -EINVAL;
4504         }
4505
4506         if (brport_idx) {
4507                 dev = __dev_get_by_index(net, brport_idx);
4508                 if (!dev) {
4509                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4510                         return -ENODEV;
4511                 }
4512         }
4513
4514         if (br_idx) {
4515                 if (dev) {
4516                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4517                         return -EINVAL;
4518                 }
4519
4520                 br_dev = __dev_get_by_index(net, br_idx);
4521                 if (!br_dev) {
4522                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4523                         return -EINVAL;
4524                 }
4525                 ops = br_dev->netdev_ops;
4526         }
4527
4528         if (dev) {
4529                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4530                         if (!netif_is_bridge_port(dev)) {
4531                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4532                                 return -EINVAL;
4533                         }
4534                         br_dev = netdev_master_upper_dev_get(dev);
4535                         if (!br_dev) {
4536                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4537                                 return -EINVAL;
4538                         }
4539                         ops = br_dev->netdev_ops;
4540                 } else {
4541                         if (!(ndm_flags & NTF_SELF)) {
4542                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4543                                 return -EINVAL;
4544                         }
4545                         ops = dev->netdev_ops;
4546                 }
4547         }
4548
4549         if (!br_dev && !dev) {
4550                 NL_SET_ERR_MSG(extack, "No device specified");
4551                 return -ENODEV;
4552         }
4553
4554         if (!ops || !ops->ndo_fdb_get) {
4555                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4556                 return -EOPNOTSUPP;
4557         }
4558
4559         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4560         if (!skb)
4561                 return -ENOBUFS;
4562
4563         if (br_dev)
4564                 dev = br_dev;
4565         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4566                                NETLINK_CB(in_skb).portid,
4567                                nlh->nlmsg_seq, extack);
4568         if (err)
4569                 goto out;
4570
4571         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4572 out:
4573         kfree_skb(skb);
4574         return err;
4575 }
4576
4577 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4578                                unsigned int attrnum, unsigned int flag)
4579 {
4580         if (mask & flag)
4581                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4582         return 0;
4583 }
4584
4585 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4586                             struct net_device *dev, u16 mode,
4587                             u32 flags, u32 mask, int nlflags,
4588                             u32 filter_mask,
4589                             int (*vlan_fill)(struct sk_buff *skb,
4590                                              struct net_device *dev,
4591                                              u32 filter_mask))
4592 {
4593         struct nlmsghdr *nlh;
4594         struct ifinfomsg *ifm;
4595         struct nlattr *br_afspec;
4596         struct nlattr *protinfo;
4597         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4598         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4599         int err = 0;
4600
4601         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4602         if (nlh == NULL)
4603                 return -EMSGSIZE;
4604
4605         ifm = nlmsg_data(nlh);
4606         ifm->ifi_family = AF_BRIDGE;
4607         ifm->__ifi_pad = 0;
4608         ifm->ifi_type = dev->type;
4609         ifm->ifi_index = dev->ifindex;
4610         ifm->ifi_flags = dev_get_flags(dev);
4611         ifm->ifi_change = 0;
4612
4613
4614         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4615             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4616             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4617             (br_dev &&
4618              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4619             (dev->addr_len &&
4620              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4621             (dev->ifindex != dev_get_iflink(dev) &&
4622              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4623                 goto nla_put_failure;
4624
4625         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4626         if (!br_afspec)
4627                 goto nla_put_failure;
4628
4629         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4630                 nla_nest_cancel(skb, br_afspec);
4631                 goto nla_put_failure;
4632         }
4633
4634         if (mode != BRIDGE_MODE_UNDEF) {
4635                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4636                         nla_nest_cancel(skb, br_afspec);
4637                         goto nla_put_failure;
4638                 }
4639         }
4640         if (vlan_fill) {
4641                 err = vlan_fill(skb, dev, filter_mask);
4642                 if (err) {
4643                         nla_nest_cancel(skb, br_afspec);
4644                         goto nla_put_failure;
4645                 }
4646         }
4647         nla_nest_end(skb, br_afspec);
4648
4649         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4650         if (!protinfo)
4651                 goto nla_put_failure;
4652
4653         if (brport_nla_put_flag(skb, flags, mask,
4654                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4655             brport_nla_put_flag(skb, flags, mask,
4656                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4657             brport_nla_put_flag(skb, flags, mask,
4658                                 IFLA_BRPORT_FAST_LEAVE,
4659                                 BR_MULTICAST_FAST_LEAVE) ||
4660             brport_nla_put_flag(skb, flags, mask,
4661                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4662             brport_nla_put_flag(skb, flags, mask,
4663                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4664             brport_nla_put_flag(skb, flags, mask,
4665                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4666             brport_nla_put_flag(skb, flags, mask,
4667                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4668             brport_nla_put_flag(skb, flags, mask,
4669                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4670             brport_nla_put_flag(skb, flags, mask,
4671                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4672             brport_nla_put_flag(skb, flags, mask,
4673                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4674                 nla_nest_cancel(skb, protinfo);
4675                 goto nla_put_failure;
4676         }
4677
4678         nla_nest_end(skb, protinfo);
4679
4680         nlmsg_end(skb, nlh);
4681         return 0;
4682 nla_put_failure:
4683         nlmsg_cancel(skb, nlh);
4684         return err ? err : -EMSGSIZE;
4685 }
4686 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4687
4688 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4689                                     bool strict_check, u32 *filter_mask,
4690                                     struct netlink_ext_ack *extack)
4691 {
4692         struct nlattr *tb[IFLA_MAX+1];
4693         int err, i;
4694
4695         if (strict_check) {
4696                 struct ifinfomsg *ifm;
4697
4698                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4699                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4700                         return -EINVAL;
4701                 }
4702
4703                 ifm = nlmsg_data(nlh);
4704                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4705                     ifm->ifi_change || ifm->ifi_index) {
4706                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4707                         return -EINVAL;
4708                 }
4709
4710                 err = nlmsg_parse_deprecated_strict(nlh,
4711                                                     sizeof(struct ifinfomsg),
4712                                                     tb, IFLA_MAX, ifla_policy,
4713                                                     extack);
4714         } else {
4715                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4716                                              tb, IFLA_MAX, ifla_policy,
4717                                              extack);
4718         }
4719         if (err < 0)
4720                 return err;
4721
4722         /* new attributes should only be added with strict checking */
4723         for (i = 0; i <= IFLA_MAX; ++i) {
4724                 if (!tb[i])
4725                         continue;
4726
4727                 switch (i) {
4728                 case IFLA_EXT_MASK:
4729                         *filter_mask = nla_get_u32(tb[i]);
4730                         break;
4731                 default:
4732                         if (strict_check) {
4733                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4734                                 return -EINVAL;
4735                         }
4736                 }
4737         }
4738
4739         return 0;
4740 }
4741
4742 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4743 {
4744         const struct nlmsghdr *nlh = cb->nlh;
4745         struct net *net = sock_net(skb->sk);
4746         struct net_device *dev;
4747         int idx = 0;
4748         u32 portid = NETLINK_CB(cb->skb).portid;
4749         u32 seq = nlh->nlmsg_seq;
4750         u32 filter_mask = 0;
4751         int err;
4752
4753         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4754                                        cb->extack);
4755         if (err < 0 && cb->strict_check)
4756                 return err;
4757
4758         rcu_read_lock();
4759         for_each_netdev_rcu(net, dev) {
4760                 const struct net_device_ops *ops = dev->netdev_ops;
4761                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4762
4763                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4764                         if (idx >= cb->args[0]) {
4765                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4766                                                 skb, portid, seq, dev,
4767                                                 filter_mask, NLM_F_MULTI);
4768                                 if (err < 0 && err != -EOPNOTSUPP) {
4769                                         if (likely(skb->len))
4770                                                 break;
4771
4772                                         goto out_err;
4773                                 }
4774                         }
4775                         idx++;
4776                 }
4777
4778                 if (ops->ndo_bridge_getlink) {
4779                         if (idx >= cb->args[0]) {
4780                                 err = ops->ndo_bridge_getlink(skb, portid,
4781                                                               seq, dev,
4782                                                               filter_mask,
4783                                                               NLM_F_MULTI);
4784                                 if (err < 0 && err != -EOPNOTSUPP) {
4785                                         if (likely(skb->len))
4786                                                 break;
4787
4788                                         goto out_err;
4789                                 }
4790                         }
4791                         idx++;
4792                 }
4793         }
4794         err = skb->len;
4795 out_err:
4796         rcu_read_unlock();
4797         cb->args[0] = idx;
4798
4799         return err;
4800 }
4801
4802 static inline size_t bridge_nlmsg_size(void)
4803 {
4804         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4805                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4806                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4807                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4808                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4809                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4810                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4811                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4812                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4813                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4814                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4815 }
4816
4817 static int rtnl_bridge_notify(struct net_device *dev)
4818 {
4819         struct net *net = dev_net(dev);
4820         struct sk_buff *skb;
4821         int err = -EOPNOTSUPP;
4822
4823         if (!dev->netdev_ops->ndo_bridge_getlink)
4824                 return 0;
4825
4826         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4827         if (!skb) {
4828                 err = -ENOMEM;
4829                 goto errout;
4830         }
4831
4832         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4833         if (err < 0)
4834                 goto errout;
4835
4836         if (!skb->len)
4837                 goto errout;
4838
4839         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4840         return 0;
4841 errout:
4842         WARN_ON(err == -EMSGSIZE);
4843         kfree_skb(skb);
4844         if (err)
4845                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4846         return err;
4847 }
4848
4849 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4850                                struct netlink_ext_ack *extack)
4851 {
4852         struct net *net = sock_net(skb->sk);
4853         struct ifinfomsg *ifm;
4854         struct net_device *dev;
4855         struct nlattr *br_spec, *attr = NULL;
4856         int rem, err = -EOPNOTSUPP;
4857         u16 flags = 0;
4858         bool have_flags = false;
4859
4860         if (nlmsg_len(nlh) < sizeof(*ifm))
4861                 return -EINVAL;
4862
4863         ifm = nlmsg_data(nlh);
4864         if (ifm->ifi_family != AF_BRIDGE)
4865                 return -EPFNOSUPPORT;
4866
4867         dev = __dev_get_by_index(net, ifm->ifi_index);
4868         if (!dev) {
4869                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4870                 return -ENODEV;
4871         }
4872
4873         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4874         if (br_spec) {
4875                 nla_for_each_nested(attr, br_spec, rem) {
4876                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4877                                 if (nla_len(attr) < sizeof(flags))
4878                                         return -EINVAL;
4879
4880                                 have_flags = true;
4881                                 flags = nla_get_u16(attr);
4882                                 break;
4883                         }
4884                 }
4885         }
4886
4887         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4888                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4889
4890                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4891                         err = -EOPNOTSUPP;
4892                         goto out;
4893                 }
4894
4895                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4896                                                              extack);
4897                 if (err)
4898                         goto out;
4899
4900                 flags &= ~BRIDGE_FLAGS_MASTER;
4901         }
4902
4903         if ((flags & BRIDGE_FLAGS_SELF)) {
4904                 if (!dev->netdev_ops->ndo_bridge_setlink)
4905                         err = -EOPNOTSUPP;
4906                 else
4907                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4908                                                                   flags,
4909                                                                   extack);
4910                 if (!err) {
4911                         flags &= ~BRIDGE_FLAGS_SELF;
4912
4913                         /* Generate event to notify upper layer of bridge
4914                          * change
4915                          */
4916                         err = rtnl_bridge_notify(dev);
4917                 }
4918         }
4919
4920         if (have_flags)
4921                 memcpy(nla_data(attr), &flags, sizeof(flags));
4922 out:
4923         return err;
4924 }
4925
4926 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4927                                struct netlink_ext_ack *extack)
4928 {
4929         struct net *net = sock_net(skb->sk);
4930         struct ifinfomsg *ifm;
4931         struct net_device *dev;
4932         struct nlattr *br_spec, *attr = NULL;
4933         int rem, err = -EOPNOTSUPP;
4934         u16 flags = 0;
4935         bool have_flags = false;
4936
4937         if (nlmsg_len(nlh) < sizeof(*ifm))
4938                 return -EINVAL;
4939
4940         ifm = nlmsg_data(nlh);
4941         if (ifm->ifi_family != AF_BRIDGE)
4942                 return -EPFNOSUPPORT;
4943
4944         dev = __dev_get_by_index(net, ifm->ifi_index);
4945         if (!dev) {
4946                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4947                 return -ENODEV;
4948         }
4949
4950         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4951         if (br_spec) {
4952                 nla_for_each_nested(attr, br_spec, rem) {
4953                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4954                                 if (nla_len(attr) < sizeof(flags))
4955                                         return -EINVAL;
4956
4957                                 have_flags = true;
4958                                 flags = nla_get_u16(attr);
4959                                 break;
4960                         }
4961                 }
4962         }
4963
4964         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4965                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4966
4967                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4968                         err = -EOPNOTSUPP;
4969                         goto out;
4970                 }
4971
4972                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4973                 if (err)
4974                         goto out;
4975
4976                 flags &= ~BRIDGE_FLAGS_MASTER;
4977         }
4978
4979         if ((flags & BRIDGE_FLAGS_SELF)) {
4980                 if (!dev->netdev_ops->ndo_bridge_dellink)
4981                         err = -EOPNOTSUPP;
4982                 else
4983                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4984                                                                   flags);
4985
4986                 if (!err) {
4987                         flags &= ~BRIDGE_FLAGS_SELF;
4988
4989                         /* Generate event to notify upper layer of bridge
4990                          * change
4991                          */
4992                         err = rtnl_bridge_notify(dev);
4993                 }
4994         }
4995
4996         if (have_flags)
4997                 memcpy(nla_data(attr), &flags, sizeof(flags));
4998 out:
4999         return err;
5000 }
5001
5002 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5003 {
5004         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5005                (!idxattr || idxattr == attrid);
5006 }
5007
5008 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5009 static int rtnl_get_offload_stats_attr_size(int attr_id)
5010 {
5011         switch (attr_id) {
5012         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5013                 return sizeof(struct rtnl_link_stats64);
5014         }
5015
5016         return 0;
5017 }
5018
5019 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5020                                   int *prividx)
5021 {
5022         struct nlattr *attr = NULL;
5023         int attr_id, size;
5024         void *attr_data;
5025         int err;
5026
5027         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5028               dev->netdev_ops->ndo_get_offload_stats))
5029                 return -ENODATA;
5030
5031         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5032              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5033                 if (attr_id < *prividx)
5034                         continue;
5035
5036                 size = rtnl_get_offload_stats_attr_size(attr_id);
5037                 if (!size)
5038                         continue;
5039
5040                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5041                         continue;
5042
5043                 attr = nla_reserve_64bit(skb, attr_id, size,
5044                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5045                 if (!attr)
5046                         goto nla_put_failure;
5047
5048                 attr_data = nla_data(attr);
5049                 memset(attr_data, 0, size);
5050                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5051                                                              attr_data);
5052                 if (err)
5053                         goto get_offload_stats_failure;
5054         }
5055
5056         if (!attr)
5057                 return -ENODATA;
5058
5059         *prividx = 0;
5060         return 0;
5061
5062 nla_put_failure:
5063         err = -EMSGSIZE;
5064 get_offload_stats_failure:
5065         *prividx = attr_id;
5066         return err;
5067 }
5068
5069 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5070 {
5071         int nla_size = 0;
5072         int attr_id;
5073         int size;
5074
5075         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5076               dev->netdev_ops->ndo_get_offload_stats))
5077                 return 0;
5078
5079         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5080              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5081                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5082                         continue;
5083                 size = rtnl_get_offload_stats_attr_size(attr_id);
5084                 nla_size += nla_total_size_64bit(size);
5085         }
5086
5087         if (nla_size != 0)
5088                 nla_size += nla_total_size(0);
5089
5090         return nla_size;
5091 }
5092
5093 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5094                                int type, u32 pid, u32 seq, u32 change,
5095                                unsigned int flags, unsigned int filter_mask,
5096                                int *idxattr, int *prividx)
5097 {
5098         struct if_stats_msg *ifsm;
5099         struct nlmsghdr *nlh;
5100         struct nlattr *attr;
5101         int s_prividx = *prividx;
5102         int err;
5103
5104         ASSERT_RTNL();
5105
5106         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5107         if (!nlh)
5108                 return -EMSGSIZE;
5109
5110         ifsm = nlmsg_data(nlh);
5111         ifsm->family = PF_UNSPEC;
5112         ifsm->pad1 = 0;
5113         ifsm->pad2 = 0;
5114         ifsm->ifindex = dev->ifindex;
5115         ifsm->filter_mask = filter_mask;
5116
5117         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5118                 struct rtnl_link_stats64 *sp;
5119
5120                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5121                                          sizeof(struct rtnl_link_stats64),
5122                                          IFLA_STATS_UNSPEC);
5123                 if (!attr)
5124                         goto nla_put_failure;
5125
5126                 sp = nla_data(attr);
5127                 dev_get_stats(dev, sp);
5128         }
5129
5130         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5131                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5132
5133                 if (ops && ops->fill_linkxstats) {
5134                         *idxattr = IFLA_STATS_LINK_XSTATS;
5135                         attr = nla_nest_start_noflag(skb,
5136                                                      IFLA_STATS_LINK_XSTATS);
5137                         if (!attr)
5138                                 goto nla_put_failure;
5139
5140                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5141                         nla_nest_end(skb, attr);
5142                         if (err)
5143                                 goto nla_put_failure;
5144                         *idxattr = 0;
5145                 }
5146         }
5147
5148         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5149                              *idxattr)) {
5150                 const struct rtnl_link_ops *ops = NULL;
5151                 const struct net_device *master;
5152
5153                 master = netdev_master_upper_dev_get(dev);
5154                 if (master)
5155                         ops = master->rtnl_link_ops;
5156                 if (ops && ops->fill_linkxstats) {
5157                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5158                         attr = nla_nest_start_noflag(skb,
5159                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5160                         if (!attr)
5161                                 goto nla_put_failure;
5162
5163                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5164                         nla_nest_end(skb, attr);
5165                         if (err)
5166                                 goto nla_put_failure;
5167                         *idxattr = 0;
5168                 }
5169         }
5170
5171         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5172                              *idxattr)) {
5173                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5174                 attr = nla_nest_start_noflag(skb,
5175                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5176                 if (!attr)
5177                         goto nla_put_failure;
5178
5179                 err = rtnl_get_offload_stats(skb, dev, prividx);
5180                 if (err == -ENODATA)
5181                         nla_nest_cancel(skb, attr);
5182                 else
5183                         nla_nest_end(skb, attr);
5184
5185                 if (err && err != -ENODATA)
5186                         goto nla_put_failure;
5187                 *idxattr = 0;
5188         }
5189
5190         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5191                 struct rtnl_af_ops *af_ops;
5192
5193                 *idxattr = IFLA_STATS_AF_SPEC;
5194                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5195                 if (!attr)
5196                         goto nla_put_failure;
5197
5198                 rcu_read_lock();
5199                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5200                         if (af_ops->fill_stats_af) {
5201                                 struct nlattr *af;
5202                                 int err;
5203
5204                                 af = nla_nest_start_noflag(skb,
5205                                                            af_ops->family);
5206                                 if (!af) {
5207                                         rcu_read_unlock();
5208                                         goto nla_put_failure;
5209                                 }
5210                                 err = af_ops->fill_stats_af(skb, dev);
5211
5212                                 if (err == -ENODATA) {
5213                                         nla_nest_cancel(skb, af);
5214                                 } else if (err < 0) {
5215                                         rcu_read_unlock();
5216                                         goto nla_put_failure;
5217                                 }
5218
5219                                 nla_nest_end(skb, af);
5220                         }
5221                 }
5222                 rcu_read_unlock();
5223
5224                 nla_nest_end(skb, attr);
5225
5226                 *idxattr = 0;
5227         }
5228
5229         nlmsg_end(skb, nlh);
5230
5231         return 0;
5232
5233 nla_put_failure:
5234         /* not a multi message or no progress mean a real error */
5235         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5236                 nlmsg_cancel(skb, nlh);
5237         else
5238                 nlmsg_end(skb, nlh);
5239
5240         return -EMSGSIZE;
5241 }
5242
5243 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5244                                   u32 filter_mask)
5245 {
5246         size_t size = 0;
5247
5248         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5249                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5250
5251         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5252                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5253                 int attr = IFLA_STATS_LINK_XSTATS;
5254
5255                 if (ops && ops->get_linkxstats_size) {
5256                         size += nla_total_size(ops->get_linkxstats_size(dev,
5257                                                                         attr));
5258                         /* for IFLA_STATS_LINK_XSTATS */
5259                         size += nla_total_size(0);
5260                 }
5261         }
5262
5263         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5264                 struct net_device *_dev = (struct net_device *)dev;
5265                 const struct rtnl_link_ops *ops = NULL;
5266                 const struct net_device *master;
5267
5268                 /* netdev_master_upper_dev_get can't take const */
5269                 master = netdev_master_upper_dev_get(_dev);
5270                 if (master)
5271                         ops = master->rtnl_link_ops;
5272                 if (ops && ops->get_linkxstats_size) {
5273                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5274
5275                         size += nla_total_size(ops->get_linkxstats_size(dev,
5276                                                                         attr));
5277                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5278                         size += nla_total_size(0);
5279                 }
5280         }
5281
5282         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5283                 size += rtnl_get_offload_stats_size(dev);
5284
5285         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5286                 struct rtnl_af_ops *af_ops;
5287
5288                 /* for IFLA_STATS_AF_SPEC */
5289                 size += nla_total_size(0);
5290
5291                 rcu_read_lock();
5292                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5293                         if (af_ops->get_stats_af_size) {
5294                                 size += nla_total_size(
5295                                         af_ops->get_stats_af_size(dev));
5296
5297                                 /* for AF_* */
5298                                 size += nla_total_size(0);
5299                         }
5300                 }
5301                 rcu_read_unlock();
5302         }
5303
5304         return size;
5305 }
5306
5307 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5308                                 bool is_dump, struct netlink_ext_ack *extack)
5309 {
5310         struct if_stats_msg *ifsm;
5311
5312         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5313                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5314                 return -EINVAL;
5315         }
5316
5317         if (!strict_check)
5318                 return 0;
5319
5320         ifsm = nlmsg_data(nlh);
5321
5322         /* only requests using strict checks can pass data to influence
5323          * the dump. The legacy exception is filter_mask.
5324          */
5325         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5326                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5327                 return -EINVAL;
5328         }
5329         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5330                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5331                 return -EINVAL;
5332         }
5333         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5334                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5335                 return -EINVAL;
5336         }
5337
5338         return 0;
5339 }
5340
5341 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5342                           struct netlink_ext_ack *extack)
5343 {
5344         struct net *net = sock_net(skb->sk);
5345         struct net_device *dev = NULL;
5346         int idxattr = 0, prividx = 0;
5347         struct if_stats_msg *ifsm;
5348         struct sk_buff *nskb;
5349         u32 filter_mask;
5350         int err;
5351
5352         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5353                                    false, extack);
5354         if (err)
5355                 return err;
5356
5357         ifsm = nlmsg_data(nlh);
5358         if (ifsm->ifindex > 0)
5359                 dev = __dev_get_by_index(net, ifsm->ifindex);
5360         else
5361                 return -EINVAL;
5362
5363         if (!dev)
5364                 return -ENODEV;
5365
5366         filter_mask = ifsm->filter_mask;
5367         if (!filter_mask)
5368                 return -EINVAL;
5369
5370         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5371         if (!nskb)
5372                 return -ENOBUFS;
5373
5374         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5375                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5376                                   0, filter_mask, &idxattr, &prividx);
5377         if (err < 0) {
5378                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5379                 WARN_ON(err == -EMSGSIZE);
5380                 kfree_skb(nskb);
5381         } else {
5382                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5383         }
5384
5385         return err;
5386 }
5387
5388 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5389 {
5390         struct netlink_ext_ack *extack = cb->extack;
5391         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5392         struct net *net = sock_net(skb->sk);
5393         unsigned int flags = NLM_F_MULTI;
5394         struct if_stats_msg *ifsm;
5395         struct hlist_head *head;
5396         struct net_device *dev;
5397         u32 filter_mask = 0;
5398         int idx = 0;
5399
5400         s_h = cb->args[0];
5401         s_idx = cb->args[1];
5402         s_idxattr = cb->args[2];
5403         s_prividx = cb->args[3];
5404
5405         cb->seq = net->dev_base_seq;
5406
5407         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5408         if (err)
5409                 return err;
5410
5411         ifsm = nlmsg_data(cb->nlh);
5412         filter_mask = ifsm->filter_mask;
5413         if (!filter_mask) {
5414                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5415                 return -EINVAL;
5416         }
5417
5418         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5419                 idx = 0;
5420                 head = &net->dev_index_head[h];
5421                 hlist_for_each_entry(dev, head, index_hlist) {
5422                         if (idx < s_idx)
5423                                 goto cont;
5424                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5425                                                   NETLINK_CB(cb->skb).portid,
5426                                                   cb->nlh->nlmsg_seq, 0,
5427                                                   flags, filter_mask,
5428                                                   &s_idxattr, &s_prividx);
5429                         /* If we ran out of room on the first message,
5430                          * we're in trouble
5431                          */
5432                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5433
5434                         if (err < 0)
5435                                 goto out;
5436                         s_prividx = 0;
5437                         s_idxattr = 0;
5438                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5439 cont:
5440                         idx++;
5441                 }
5442         }
5443 out:
5444         cb->args[3] = s_prividx;
5445         cb->args[2] = s_idxattr;
5446         cb->args[1] = idx;
5447         cb->args[0] = h;
5448
5449         return skb->len;
5450 }
5451
5452 /* Process one rtnetlink message. */
5453
5454 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5455                              struct netlink_ext_ack *extack)
5456 {
5457         struct net *net = sock_net(skb->sk);
5458         struct rtnl_link *link;
5459         struct module *owner;
5460         int err = -EOPNOTSUPP;
5461         rtnl_doit_func doit;
5462         unsigned int flags;
5463         int kind;
5464         int family;
5465         int type;
5466
5467         type = nlh->nlmsg_type;
5468         if (type > RTM_MAX)
5469                 return -EOPNOTSUPP;
5470
5471         type -= RTM_BASE;
5472
5473         /* All the messages must have at least 1 byte length */
5474         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5475                 return 0;
5476
5477         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5478         kind = type&3;
5479
5480         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5481                 return -EPERM;
5482
5483         rcu_read_lock();
5484         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5485                 struct sock *rtnl;
5486                 rtnl_dumpit_func dumpit;
5487                 u32 min_dump_alloc = 0;
5488
5489                 link = rtnl_get_link(family, type);
5490                 if (!link || !link->dumpit) {
5491                         family = PF_UNSPEC;
5492                         link = rtnl_get_link(family, type);
5493                         if (!link || !link->dumpit)
5494                                 goto err_unlock;
5495                 }
5496                 owner = link->owner;
5497                 dumpit = link->dumpit;
5498
5499                 if (type == RTM_GETLINK - RTM_BASE)
5500                         min_dump_alloc = rtnl_calcit(skb, nlh);
5501
5502                 err = 0;
5503                 /* need to do this before rcu_read_unlock() */
5504                 if (!try_module_get(owner))
5505                         err = -EPROTONOSUPPORT;
5506
5507                 rcu_read_unlock();
5508
5509                 rtnl = net->rtnl;
5510                 if (err == 0) {
5511                         struct netlink_dump_control c = {
5512                                 .dump           = dumpit,
5513                                 .min_dump_alloc = min_dump_alloc,
5514                                 .module         = owner,
5515                         };
5516                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5517                         /* netlink_dump_start() will keep a reference on
5518                          * module if dump is still in progress.
5519                          */
5520                         module_put(owner);
5521                 }
5522                 return err;
5523         }
5524
5525         link = rtnl_get_link(family, type);
5526         if (!link || !link->doit) {
5527                 family = PF_UNSPEC;
5528                 link = rtnl_get_link(PF_UNSPEC, type);
5529                 if (!link || !link->doit)
5530                         goto out_unlock;
5531         }
5532
5533         owner = link->owner;
5534         if (!try_module_get(owner)) {
5535                 err = -EPROTONOSUPPORT;
5536                 goto out_unlock;
5537         }
5538
5539         flags = link->flags;
5540         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5541                 doit = link->doit;
5542                 rcu_read_unlock();
5543                 if (doit)
5544                         err = doit(skb, nlh, extack);
5545                 module_put(owner);
5546                 return err;
5547         }
5548         rcu_read_unlock();
5549
5550         rtnl_lock();
5551         link = rtnl_get_link(family, type);
5552         if (link && link->doit)
5553                 err = link->doit(skb, nlh, extack);
5554         rtnl_unlock();
5555
5556         module_put(owner);
5557
5558         return err;
5559
5560 out_unlock:
5561         rcu_read_unlock();
5562         return err;
5563
5564 err_unlock:
5565         rcu_read_unlock();
5566         return -EOPNOTSUPP;
5567 }
5568
5569 static void rtnetlink_rcv(struct sk_buff *skb)
5570 {
5571         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5572 }
5573
5574 static int rtnetlink_bind(struct net *net, int group)
5575 {
5576         switch (group) {
5577         case RTNLGRP_IPV4_MROUTE_R:
5578         case RTNLGRP_IPV6_MROUTE_R:
5579                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5580                         return -EPERM;
5581                 break;
5582         }
5583         return 0;
5584 }
5585
5586 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5587 {
5588         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5589
5590         switch (event) {
5591         case NETDEV_REBOOT:
5592         case NETDEV_CHANGEMTU:
5593         case NETDEV_CHANGEADDR:
5594         case NETDEV_CHANGENAME:
5595         case NETDEV_FEAT_CHANGE:
5596         case NETDEV_BONDING_FAILOVER:
5597         case NETDEV_POST_TYPE_CHANGE:
5598         case NETDEV_NOTIFY_PEERS:
5599         case NETDEV_CHANGEUPPER:
5600         case NETDEV_RESEND_IGMP:
5601         case NETDEV_CHANGEINFODATA:
5602         case NETDEV_CHANGELOWERSTATE:
5603         case NETDEV_CHANGE_TX_QUEUE_LEN:
5604                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5605                                    GFP_KERNEL, NULL, 0);
5606                 break;
5607         default:
5608                 break;
5609         }
5610         return NOTIFY_DONE;
5611 }
5612
5613 static struct notifier_block rtnetlink_dev_notifier = {
5614         .notifier_call  = rtnetlink_event,
5615 };
5616
5617
5618 static int __net_init rtnetlink_net_init(struct net *net)
5619 {
5620         struct sock *sk;
5621         struct netlink_kernel_cfg cfg = {
5622                 .groups         = RTNLGRP_MAX,
5623                 .input          = rtnetlink_rcv,
5624                 .cb_mutex       = &rtnl_mutex,
5625                 .flags          = NL_CFG_F_NONROOT_RECV,
5626                 .bind           = rtnetlink_bind,
5627         };
5628
5629         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5630         if (!sk)
5631                 return -ENOMEM;
5632         net->rtnl = sk;
5633         return 0;
5634 }
5635
5636 static void __net_exit rtnetlink_net_exit(struct net *net)
5637 {
5638         netlink_kernel_release(net->rtnl);
5639         net->rtnl = NULL;
5640 }
5641
5642 static struct pernet_operations rtnetlink_net_ops = {
5643         .init = rtnetlink_net_init,
5644         .exit = rtnetlink_net_exit,
5645 };
5646
5647 void __init rtnetlink_init(void)
5648 {
5649         if (register_pernet_subsys(&rtnetlink_net_ops))
5650                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5651
5652         register_netdevice_notifier(&rtnetlink_dev_notifier);
5653
5654         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5655                       rtnl_dump_ifinfo, 0);
5656         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5657         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5658         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5659
5660         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5661         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5662         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5663
5664         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5665         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5666
5667         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5668         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5669         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5670
5671         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5672         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5673         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5674
5675         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5676                       0);
5677 }