net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetic was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     40
  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 modify 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 modify 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 alloc/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->alloc || 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         ASSERT_RTNL();
 547
 548         list_for_each_entry(ops, &rtnl_af_ops, list) {
 549                 if (ops->family == family)
 550                         return ops;
 551         }
 552
 553         return NULL;
 554 }
 555
 556 /**
 557  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 558  * @ops: struct rtnl_af_ops * to register
 559  *
 560  * Returns 0 on success or a negative error code.
 561  */
 562 void rtnl_af_register(struct rtnl_af_ops *ops)
 563 {
 564         rtnl_lock();
 565         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 566         rtnl_unlock();
 567 }
 568 EXPORT_SYMBOL_GPL(rtnl_af_register);
 569
 570 /**
 571  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 572  * @ops: struct rtnl_af_ops * to unregister
 573  */
 574 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 575 {
 576         rtnl_lock();
 577         list_del_rcu(&ops->list);
 578         rtnl_unlock();
 579
 580         synchronize_rcu();
 581 }
 582 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 583
 584 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 585                                     u32 ext_filter_mask)
 586 {
 587         struct rtnl_af_ops *af_ops;
 588         size_t size;
 589
 590         /* IFLA_AF_SPEC */
 591         size = nla_total_size(sizeof(struct nlattr));
 592
 593         rcu_read_lock();
 594         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 595                 if (af_ops->get_link_af_size) {
 596                         /* AF_* + nested data */
 597                         size += nla_total_size(sizeof(struct nlattr)) +
 598                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 599                 }
 600         }
 601         rcu_read_unlock();
 602
 603         return size;
 604 }
 605
 606 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 607 {
 608         struct net_device *master_dev;
 609         bool ret = false;
 610
 611         rcu_read_lock();
 612
 613         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 614         if (master_dev && master_dev->rtnl_link_ops)
 615                 ret = true;
 616         rcu_read_unlock();
 617         return ret;
 618 }
 619
 620 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 621                                      const struct net_device *dev)
 622 {
 623         struct net_device *master_dev;
 624         const struct rtnl_link_ops *ops;
 625         struct nlattr *slave_data;
 626         int err;
 627
 628         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 629         if (!master_dev)
 630                 return 0;
 631         ops = master_dev->rtnl_link_ops;
 632         if (!ops)
 633                 return 0;
 634         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 635                 return -EMSGSIZE;
 636         if (ops->fill_slave_info) {
 637                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 638                 if (!slave_data)
 639                         return -EMSGSIZE;
 640                 err = ops->fill_slave_info(skb, master_dev, dev);
 641                 if (err < 0)
 642                         goto err_cancel_slave_data;
 643                 nla_nest_end(skb, slave_data);
 644         }
 645         return 0;
 646
 647 err_cancel_slave_data:
 648         nla_nest_cancel(skb, slave_data);
 649         return err;
 650 }
 651
 652 static int rtnl_link_info_fill(struct sk_buff *skb,
 653                                const struct net_device *dev)
 654 {
 655         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 656         struct nlattr *data;
 657         int err;
 658
 659         if (!ops)
 660                 return 0;
 661         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 662                 return -EMSGSIZE;
 663         if (ops->fill_xstats) {
 664                 err = ops->fill_xstats(skb, dev);
 665                 if (err < 0)
 666                         return err;
 667         }
 668         if (ops->fill_info) {
 669                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 670                 if (data == NULL)
 671                         return -EMSGSIZE;
 672                 err = ops->fill_info(skb, dev);
 673                 if (err < 0)
 674                         goto err_cancel_data;
 675                 nla_nest_end(skb, data);
 676         }
 677         return 0;
 678
 679 err_cancel_data:
 680         nla_nest_cancel(skb, data);
 681         return err;
 682 }
 683
 684 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 685 {
 686         struct nlattr *linkinfo;
 687         int err = -EMSGSIZE;
 688
 689         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 690         if (linkinfo == NULL)
 691                 goto out;
 692
 693         err = rtnl_link_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         err = rtnl_link_slave_info_fill(skb, dev);
 698         if (err < 0)
 699                 goto err_cancel_link;
 700
 701         nla_nest_end(skb, linkinfo);
 702         return 0;
 703
 704 err_cancel_link:
 705         nla_nest_cancel(skb, linkinfo);
 706 out:
 707         return err;
 708 }
 709
 710 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 711 {
 712         struct sock *rtnl = net->rtnl;
 713
 714         return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
 715 }
 716
 717 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 718 {
 719         struct sock *rtnl = net->rtnl;
 720
 721         return nlmsg_unicast(rtnl, skb, pid);
 722 }
 723 EXPORT_SYMBOL(rtnl_unicast);
 724
 725 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 726                  struct nlmsghdr *nlh, gfp_t flags)
 727 {
 728         struct sock *rtnl = net->rtnl;
 729
 730         nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
 731 }
 732 EXPORT_SYMBOL(rtnl_notify);
 733
 734 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 735 {
 736         struct sock *rtnl = net->rtnl;
 737
 738         netlink_set_err(rtnl, 0, group, error);
 739 }
 740 EXPORT_SYMBOL(rtnl_set_sk_err);
 741
 742 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 743 {
 744         struct nlattr *mx;
 745         int i, valid = 0;
 746
 747         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 748         if (metrics == dst_default_metrics.metrics)
 749                 return 0;
 750
 751         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 752         if (mx == NULL)
 753                 return -ENOBUFS;
 754
 755         for (i = 0; i < RTAX_MAX; i++) {
 756                 if (metrics[i]) {
 757                         if (i == RTAX_CC_ALGO - 1) {
 758                                 char tmp[TCP_CA_NAME_MAX], *name;
 759
 760                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 761                                 if (!name)
 762                                         continue;
 763                                 if (nla_put_string(skb, i + 1, name))
 764                                         goto nla_put_failure;
 765                         } else if (i == RTAX_FEATURES - 1) {
 766                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 767
 768                                 if (!user_features)
 769                                         continue;
 770                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 771                                 if (nla_put_u32(skb, i + 1, user_features))
 772                                         goto nla_put_failure;
 773                         } else {
 774                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 775                                         goto nla_put_failure;
 776                         }
 777                         valid++;
 778                 }
 779         }
 780
 781         if (!valid) {
 782                 nla_nest_cancel(skb, mx);
 783                 return 0;
 784         }
 785
 786         return nla_nest_end(skb, mx);
 787
 788 nla_put_failure:
 789         nla_nest_cancel(skb, mx);
 790         return -EMSGSIZE;
 791 }
 792 EXPORT_SYMBOL(rtnetlink_put_metrics);
 793
 794 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 795                        long expires, u32 error)
 796 {
 797         struct rta_cacheinfo ci = {
 798                 .rta_error = error,
 799                 .rta_id =  id,
 800         };
 801
 802         if (dst) {
 803                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 804                 ci.rta_used = dst->__use;
 805                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 806         }
 807         if (expires) {
 808                 unsigned long clock;
 809
 810                 clock = jiffies_to_clock_t(abs(expires));
 811                 clock = min_t(unsigned long, clock, INT_MAX);
 812                 ci.rta_expires = (expires > 0) ? clock : -clock;
 813         }
 814         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 815 }
 816 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 817
 818 static void set_operstate(struct net_device *dev, unsigned char transition)
 819 {
 820         unsigned char operstate = dev->operstate;
 821
 822         switch (transition) {
 823         case IF_OPER_UP:
 824                 if ((operstate == IF_OPER_DORMANT ||
 825                      operstate == IF_OPER_TESTING ||
 826                      operstate == IF_OPER_UNKNOWN) &&
 827                     !netif_dormant(dev) && !netif_testing(dev))
 828                         operstate = IF_OPER_UP;
 829                 break;
 830
 831         case IF_OPER_TESTING:
 832                 if (operstate == IF_OPER_UP ||
 833                     operstate == IF_OPER_UNKNOWN)
 834                         operstate = IF_OPER_TESTING;
 835                 break;
 836
 837         case IF_OPER_DORMANT:
 838                 if (operstate == IF_OPER_UP ||
 839                     operstate == IF_OPER_UNKNOWN)
 840                         operstate = IF_OPER_DORMANT;
 841                 break;
 842         }
 843
 844         if (dev->operstate != operstate) {
 845                 write_lock_bh(&dev_base_lock);
 846                 dev->operstate = operstate;
 847                 write_unlock_bh(&dev_base_lock);
 848                 netdev_state_change(dev);
 849         }
 850 }
 851
 852 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 853 {
 854         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 855                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 856 }
 857
 858 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 859                                            const struct ifinfomsg *ifm)
 860 {
 861         unsigned int flags = ifm->ifi_flags;
 862
 863         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 864         if (ifm->ifi_change)
 865                 flags = (flags & ifm->ifi_change) |
 866                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 867
 868         return flags;
 869 }
 870
 871 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 872                                  const struct rtnl_link_stats64 *b)
 873 {
 874         a->rx_packets = b->rx_packets;
 875         a->tx_packets = b->tx_packets;
 876         a->rx_bytes = b->rx_bytes;
 877         a->tx_bytes = b->tx_bytes;
 878         a->rx_errors = b->rx_errors;
 879         a->tx_errors = b->tx_errors;
 880         a->rx_dropped = b->rx_dropped;
 881         a->tx_dropped = b->tx_dropped;
 882
 883         a->multicast = b->multicast;
 884         a->collisions = b->collisions;
 885
 886         a->rx_length_errors = b->rx_length_errors;
 887         a->rx_over_errors = b->rx_over_errors;
 888         a->rx_crc_errors = b->rx_crc_errors;
 889         a->rx_frame_errors = b->rx_frame_errors;
 890         a->rx_fifo_errors = b->rx_fifo_errors;
 891         a->rx_missed_errors = b->rx_missed_errors;
 892
 893         a->tx_aborted_errors = b->tx_aborted_errors;
 894         a->tx_carrier_errors = b->tx_carrier_errors;
 895         a->tx_fifo_errors = b->tx_fifo_errors;
 896         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 897         a->tx_window_errors = b->tx_window_errors;
 898
 899         a->rx_compressed = b->rx_compressed;
 900         a->tx_compressed = b->tx_compressed;
 901
 902         a->rx_nohandler = b->rx_nohandler;
 903 }
 904
 905 /* All VF info */
 906 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 907                                    u32 ext_filter_mask)
 908 {
 909         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 910                 int num_vfs = dev_num_vf(dev->dev.parent);
 911                 size_t size = nla_total_size(0);
 912                 size += num_vfs *
 913                         (nla_total_size(0) +
 914                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 915                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 916                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 917                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 918                          nla_total_size(MAX_VLAN_LIST_LEN *
 919                                         sizeof(struct ifla_vf_vlan_info)) +
 920                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 921                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 922                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 923                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 924                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 925                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 926                          /* IFLA_VF_STATS_RX_PACKETS */
 927                          nla_total_size_64bit(sizeof(__u64)) +
 928                          /* IFLA_VF_STATS_TX_PACKETS */
 929                          nla_total_size_64bit(sizeof(__u64)) +
 930                          /* IFLA_VF_STATS_RX_BYTES */
 931                          nla_total_size_64bit(sizeof(__u64)) +
 932                          /* IFLA_VF_STATS_TX_BYTES */
 933                          nla_total_size_64bit(sizeof(__u64)) +
 934                          /* IFLA_VF_STATS_BROADCAST */
 935                          nla_total_size_64bit(sizeof(__u64)) +
 936                          /* IFLA_VF_STATS_MULTICAST */
 937                          nla_total_size_64bit(sizeof(__u64)) +
 938                          /* IFLA_VF_STATS_RX_DROPPED */
 939                          nla_total_size_64bit(sizeof(__u64)) +
 940                          /* IFLA_VF_STATS_TX_DROPPED */
 941                          nla_total_size_64bit(sizeof(__u64)) +
 942                          nla_total_size(sizeof(struct ifla_vf_trust)));
 943                 return size;
 944         } else
 945                 return 0;
 946 }
 947
 948 static size_t rtnl_port_size(const struct net_device *dev,
 949                              u32 ext_filter_mask)
 950 {
 951         size_t port_size = nla_total_size(4)            /* PORT_VF */
 952                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 953                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 954                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 955                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 956                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 957         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 958         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 959                 + port_size;
 960         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 961                 + port_size;
 962
 963         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 964             !(ext_filter_mask & RTEXT_FILTER_VF))
 965                 return 0;
 966         if (dev_num_vf(dev->dev.parent))
 967                 return port_self_size + vf_ports_size +
 968                         vf_port_size * dev_num_vf(dev->dev.parent);
 969         else
 970                 return port_self_size;
 971 }
 972
 973 static size_t rtnl_xdp_size(void)
 974 {
 975         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 976                           nla_total_size(1) +   /* XDP_ATTACHED */
 977                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 978                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 979
 980         return xdp_size;
 981 }
 982
 983 static size_t rtnl_prop_list_size(const struct net_device *dev)
 984 {
 985         struct netdev_name_node *name_node;
 986         size_t size;
 987
 988         if (list_empty(&dev->name_node->list))
 989                 return 0;
 990         size = nla_total_size(0);
 991         list_for_each_entry(name_node, &dev->name_node->list, list)
 992                 size += nla_total_size(ALTIFNAMSIZ);
 993         return size;
 994 }
 995
 996 static size_t rtnl_proto_down_size(const struct net_device *dev)
 997 {
 998         size_t size = nla_total_size(1);
 999
1000         if (dev->proto_down_reason)
1001                 size += nla_total_size(0) + nla_total_size(4);
1002
1003         return size;
1004 }
1005
1006 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1007                                      u32 ext_filter_mask)
1008 {
1009         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1010                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1011                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1012                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1013                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1014                + nla_total_size(sizeof(struct rtnl_link_stats))
1015                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1016                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1017                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1018                + nla_total_size(4) /* IFLA_TXQLEN */
1019                + nla_total_size(4) /* IFLA_WEIGHT */
1020                + nla_total_size(4) /* IFLA_MTU */
1021                + nla_total_size(4) /* IFLA_LINK */
1022                + nla_total_size(4) /* IFLA_MASTER */
1023                + nla_total_size(1) /* IFLA_CARRIER */
1024                + nla_total_size(4) /* IFLA_PROMISCUITY */
1025                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1026                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1027                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1028                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1029                + nla_total_size(1) /* IFLA_OPERSTATE */
1030                + nla_total_size(1) /* IFLA_LINKMODE */
1031                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1032                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1033                + nla_total_size(4) /* IFLA_GROUP */
1034                + nla_total_size(ext_filter_mask
1035                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1036                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1037                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1038                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1039                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1040                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1041                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1042                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1043                + rtnl_xdp_size() /* IFLA_XDP */
1044                + nla_total_size(4)  /* IFLA_EVENT */
1045                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1046                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1047                + rtnl_proto_down_size(dev)  /* proto down */
1048                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1049                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1050                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1051                + nla_total_size(4)  /* IFLA_MIN_MTU */
1052                + nla_total_size(4)  /* IFLA_MAX_MTU */
1053                + rtnl_prop_list_size(dev)
1054                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1055                + 0;
1056 }
1057
1058 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1059 {
1060         struct nlattr *vf_ports;
1061         struct nlattr *vf_port;
1062         int vf;
1063         int err;
1064
1065         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1066         if (!vf_ports)
1067                 return -EMSGSIZE;
1068
1069         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1070                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1071                 if (!vf_port)
1072                         goto nla_put_failure;
1073                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1074                         goto nla_put_failure;
1075                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1076                 if (err == -EMSGSIZE)
1077                         goto nla_put_failure;
1078                 if (err) {
1079                         nla_nest_cancel(skb, vf_port);
1080                         continue;
1081                 }
1082                 nla_nest_end(skb, vf_port);
1083         }
1084
1085         nla_nest_end(skb, vf_ports);
1086
1087         return 0;
1088
1089 nla_put_failure:
1090         nla_nest_cancel(skb, vf_ports);
1091         return -EMSGSIZE;
1092 }
1093
1094 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1095 {
1096         struct nlattr *port_self;
1097         int err;
1098
1099         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1100         if (!port_self)
1101                 return -EMSGSIZE;
1102
1103         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1104         if (err) {
1105                 nla_nest_cancel(skb, port_self);
1106                 return (err == -EMSGSIZE) ? err : 0;
1107         }
1108
1109         nla_nest_end(skb, port_self);
1110
1111         return 0;
1112 }
1113
1114 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1115                           u32 ext_filter_mask)
1116 {
1117         int err;
1118
1119         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1120             !(ext_filter_mask & RTEXT_FILTER_VF))
1121                 return 0;
1122
1123         err = rtnl_port_self_fill(skb, dev);
1124         if (err)
1125                 return err;
1126
1127         if (dev_num_vf(dev->dev.parent)) {
1128                 err = rtnl_vf_ports_fill(skb, dev);
1129                 if (err)
1130                         return err;
1131         }
1132
1133         return 0;
1134 }
1135
1136 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1137 {
1138         int err;
1139         struct netdev_phys_item_id ppid;
1140
1141         err = dev_get_phys_port_id(dev, &ppid);
1142         if (err) {
1143                 if (err == -EOPNOTSUPP)
1144                         return 0;
1145                 return err;
1146         }
1147
1148         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1149                 return -EMSGSIZE;
1150
1151         return 0;
1152 }
1153
1154 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1155 {
1156         char name[IFNAMSIZ];
1157         int err;
1158
1159         err = dev_get_phys_port_name(dev, name, sizeof(name));
1160         if (err) {
1161                 if (err == -EOPNOTSUPP)
1162                         return 0;
1163                 return err;
1164         }
1165
1166         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1167                 return -EMSGSIZE;
1168
1169         return 0;
1170 }
1171
1172 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1173 {
1174         struct netdev_phys_item_id ppid = { };
1175         int err;
1176
1177         err = dev_get_port_parent_id(dev, &ppid, false);
1178         if (err) {
1179                 if (err == -EOPNOTSUPP)
1180                         return 0;
1181                 return err;
1182         }
1183
1184         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1185                 return -EMSGSIZE;
1186
1187         return 0;
1188 }
1189
1190 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1191                                               struct net_device *dev)
1192 {
1193         struct rtnl_link_stats64 *sp;
1194         struct nlattr *attr;
1195
1196         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1197                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1198         if (!attr)
1199                 return -EMSGSIZE;
1200
1201         sp = nla_data(attr);
1202         dev_get_stats(dev, sp);
1203
1204         attr = nla_reserve(skb, IFLA_STATS,
1205                            sizeof(struct rtnl_link_stats));
1206         if (!attr)
1207                 return -EMSGSIZE;
1208
1209         copy_rtnl_link_stats(nla_data(attr), sp);
1210
1211         return 0;
1212 }
1213
1214 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1215                                                struct net_device *dev,
1216                                                int vfs_num,
1217                                                struct nlattr *vfinfo)
1218 {
1219         struct ifla_vf_rss_query_en vf_rss_query_en;
1220         struct nlattr *vf, *vfstats, *vfvlanlist;
1221         struct ifla_vf_link_state vf_linkstate;
1222         struct ifla_vf_vlan_info vf_vlan_info;
1223         struct ifla_vf_spoofchk vf_spoofchk;
1224         struct ifla_vf_tx_rate vf_tx_rate;
1225         struct ifla_vf_stats vf_stats;
1226         struct ifla_vf_trust vf_trust;
1227         struct ifla_vf_vlan vf_vlan;
1228         struct ifla_vf_rate vf_rate;
1229         struct ifla_vf_mac vf_mac;
1230         struct ifla_vf_broadcast vf_broadcast;
1231         struct ifla_vf_info ivi;
1232         struct ifla_vf_guid node_guid;
1233         struct ifla_vf_guid port_guid;
1234
1235         memset(&ivi, 0, sizeof(ivi));
1236
1237         /* Not all SR-IOV capable drivers support the
1238          * spoofcheck and "RSS query enable" query.  Preset to
1239          * -1 so the user space tool can detect that the driver
1240          * didn't report anything.
1241          */
1242         ivi.spoofchk = -1;
1243         ivi.rss_query_en = -1;
1244         ivi.trusted = -1;
1245         /* The default value for VF link state is "auto"
1246          * IFLA_VF_LINK_STATE_AUTO which equals zero
1247          */
1248         ivi.linkstate = 0;
1249         /* VLAN Protocol by default is 802.1Q */
1250         ivi.vlan_proto = htons(ETH_P_8021Q);
1251         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1252                 return 0;
1253
1254         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1255         memset(&node_guid, 0, sizeof(node_guid));
1256         memset(&port_guid, 0, sizeof(port_guid));
1257
1258         vf_mac.vf =
1259                 vf_vlan.vf =
1260                 vf_vlan_info.vf =
1261                 vf_rate.vf =
1262                 vf_tx_rate.vf =
1263                 vf_spoofchk.vf =
1264                 vf_linkstate.vf =
1265                 vf_rss_query_en.vf =
1266                 vf_trust.vf =
1267                 node_guid.vf =
1268                 port_guid.vf = ivi.vf;
1269
1270         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1271         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1272         vf_vlan.vlan = ivi.vlan;
1273         vf_vlan.qos = ivi.qos;
1274         vf_vlan_info.vlan = ivi.vlan;
1275         vf_vlan_info.qos = ivi.qos;
1276         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1277         vf_tx_rate.rate = ivi.max_tx_rate;
1278         vf_rate.min_tx_rate = ivi.min_tx_rate;
1279         vf_rate.max_tx_rate = ivi.max_tx_rate;
1280         vf_spoofchk.setting = ivi.spoofchk;
1281         vf_linkstate.link_state = ivi.linkstate;
1282         vf_rss_query_en.setting = ivi.rss_query_en;
1283         vf_trust.setting = ivi.trusted;
1284         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1285         if (!vf)
1286                 goto nla_put_vfinfo_failure;
1287         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1288             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1289             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1290             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1291                     &vf_rate) ||
1292             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1293                     &vf_tx_rate) ||
1294             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1295                     &vf_spoofchk) ||
1296             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1297                     &vf_linkstate) ||
1298             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1299                     sizeof(vf_rss_query_en),
1300                     &vf_rss_query_en) ||
1301             nla_put(skb, IFLA_VF_TRUST,
1302                     sizeof(vf_trust), &vf_trust))
1303                 goto nla_put_vf_failure;
1304
1305         if (dev->netdev_ops->ndo_get_vf_guid &&
1306             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1307                                               &port_guid)) {
1308                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1309                             &node_guid) ||
1310                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1311                             &port_guid))
1312                         goto nla_put_vf_failure;
1313         }
1314         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1315         if (!vfvlanlist)
1316                 goto nla_put_vf_failure;
1317         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1318                     &vf_vlan_info)) {
1319                 nla_nest_cancel(skb, vfvlanlist);
1320                 goto nla_put_vf_failure;
1321         }
1322         nla_nest_end(skb, vfvlanlist);
1323         memset(&vf_stats, 0, sizeof(vf_stats));
1324         if (dev->netdev_ops->ndo_get_vf_stats)
1325                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1326                                                 &vf_stats);
1327         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1328         if (!vfstats)
1329                 goto nla_put_vf_failure;
1330         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1331                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1332             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1333                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1334             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1335                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1336             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1337                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1338             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1339                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1340             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1341                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1342             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1343                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1344             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1345                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1346                 nla_nest_cancel(skb, vfstats);
1347                 goto nla_put_vf_failure;
1348         }
1349         nla_nest_end(skb, vfstats);
1350         nla_nest_end(skb, vf);
1351         return 0;
1352
1353 nla_put_vf_failure:
1354         nla_nest_cancel(skb, vf);
1355 nla_put_vfinfo_failure:
1356         nla_nest_cancel(skb, vfinfo);
1357         return -EMSGSIZE;
1358 }
1359
1360 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1361                                            struct net_device *dev,
1362                                            u32 ext_filter_mask)
1363 {
1364         struct nlattr *vfinfo;
1365         int i, num_vfs;
1366
1367         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1368                 return 0;
1369
1370         num_vfs = dev_num_vf(dev->dev.parent);
1371         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1372                 return -EMSGSIZE;
1373
1374         if (!dev->netdev_ops->ndo_get_vf_config)
1375                 return 0;
1376
1377         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1378         if (!vfinfo)
1379                 return -EMSGSIZE;
1380
1381         for (i = 0; i < num_vfs; i++) {
1382                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1383                         return -EMSGSIZE;
1384         }
1385
1386         nla_nest_end(skb, vfinfo);
1387         return 0;
1388 }
1389
1390 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1391 {
1392         struct rtnl_link_ifmap map;
1393
1394         memset(&map, 0, sizeof(map));
1395         map.mem_start   = dev->mem_start;
1396         map.mem_end     = dev->mem_end;
1397         map.base_addr   = dev->base_addr;
1398         map.irq         = dev->irq;
1399         map.dma         = dev->dma;
1400         map.port        = dev->if_port;
1401
1402         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1403                 return -EMSGSIZE;
1404
1405         return 0;
1406 }
1407
1408 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1409 {
1410         const struct bpf_prog *generic_xdp_prog;
1411
1412         ASSERT_RTNL();
1413
1414         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1415         if (!generic_xdp_prog)
1416                 return 0;
1417         return generic_xdp_prog->aux->id;
1418 }
1419
1420 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1421 {
1422         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1423 }
1424
1425 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1426 {
1427         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1428 }
1429
1430 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1431                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1432                                u32 (*get_prog_id)(struct net_device *dev))
1433 {
1434         u32 curr_id;
1435         int err;
1436
1437         curr_id = get_prog_id(dev);
1438         if (!curr_id)
1439                 return 0;
1440
1441         *prog_id = curr_id;
1442         err = nla_put_u32(skb, attr, curr_id);
1443         if (err)
1444                 return err;
1445
1446         if (*mode != XDP_ATTACHED_NONE)
1447                 *mode = XDP_ATTACHED_MULTI;
1448         else
1449                 *mode = tgt_mode;
1450
1451         return 0;
1452 }
1453
1454 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1455 {
1456         struct nlattr *xdp;
1457         u32 prog_id;
1458         int err;
1459         u8 mode;
1460
1461         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1462         if (!xdp)
1463                 return -EMSGSIZE;
1464
1465         prog_id = 0;
1466         mode = XDP_ATTACHED_NONE;
1467         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1468                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1469         if (err)
1470                 goto err_cancel;
1471         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1472                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1473         if (err)
1474                 goto err_cancel;
1475         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1476                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1477         if (err)
1478                 goto err_cancel;
1479
1480         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1481         if (err)
1482                 goto err_cancel;
1483
1484         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1485                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1486                 if (err)
1487                         goto err_cancel;
1488         }
1489
1490         nla_nest_end(skb, xdp);
1491         return 0;
1492
1493 err_cancel:
1494         nla_nest_cancel(skb, xdp);
1495         return err;
1496 }
1497
1498 static u32 rtnl_get_event(unsigned long event)
1499 {
1500         u32 rtnl_event_type = IFLA_EVENT_NONE;
1501
1502         switch (event) {
1503         case NETDEV_REBOOT:
1504                 rtnl_event_type = IFLA_EVENT_REBOOT;
1505                 break;
1506         case NETDEV_FEAT_CHANGE:
1507                 rtnl_event_type = IFLA_EVENT_FEATURES;
1508                 break;
1509         case NETDEV_BONDING_FAILOVER:
1510                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1511                 break;
1512         case NETDEV_NOTIFY_PEERS:
1513                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1514                 break;
1515         case NETDEV_RESEND_IGMP:
1516                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1517                 break;
1518         case NETDEV_CHANGEINFODATA:
1519                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1520                 break;
1521         default:
1522                 break;
1523         }
1524
1525         return rtnl_event_type;
1526 }
1527
1528 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1529 {
1530         const struct net_device *upper_dev;
1531         int ret = 0;
1532
1533         rcu_read_lock();
1534
1535         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1536         if (upper_dev)
1537                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1538
1539         rcu_read_unlock();
1540         return ret;
1541 }
1542
1543 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1544                           bool force)
1545 {
1546         int ifindex = dev_get_iflink(dev);
1547
1548         if (force || dev->ifindex != ifindex)
1549                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1550
1551         return 0;
1552 }
1553
1554 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1555                                               struct net_device *dev)
1556 {
1557         char buf[IFALIASZ];
1558         int ret;
1559
1560         ret = dev_get_alias(dev, buf, sizeof(buf));
1561         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1562 }
1563
1564 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1565                                   const struct net_device *dev,
1566                                   struct net *src_net, gfp_t gfp)
1567 {
1568         bool put_iflink = false;
1569
1570         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1571                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1572
1573                 if (!net_eq(dev_net(dev), link_net)) {
1574                         int id = peernet2id_alloc(src_net, link_net, gfp);
1575
1576                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1577                                 return -EMSGSIZE;
1578
1579                         put_iflink = true;
1580                 }
1581         }
1582
1583         return nla_put_iflink(skb, dev, put_iflink);
1584 }
1585
1586 static int rtnl_fill_link_af(struct sk_buff *skb,
1587                              const struct net_device *dev,
1588                              u32 ext_filter_mask)
1589 {
1590         const struct rtnl_af_ops *af_ops;
1591         struct nlattr *af_spec;
1592
1593         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1594         if (!af_spec)
1595                 return -EMSGSIZE;
1596
1597         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1598                 struct nlattr *af;
1599                 int err;
1600
1601                 if (!af_ops->fill_link_af)
1602                         continue;
1603
1604                 af = nla_nest_start_noflag(skb, af_ops->family);
1605                 if (!af)
1606                         return -EMSGSIZE;
1607
1608                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1609                 /*
1610                  * Caller may return ENODATA to indicate that there
1611                  * was no data to be dumped. This is not an error, it
1612                  * means we should trim the attribute header and
1613                  * continue.
1614                  */
1615                 if (err == -ENODATA)
1616                         nla_nest_cancel(skb, af);
1617                 else if (err < 0)
1618                         return -EMSGSIZE;
1619
1620                 nla_nest_end(skb, af);
1621         }
1622
1623         nla_nest_end(skb, af_spec);
1624         return 0;
1625 }
1626
1627 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1628                                  const struct net_device *dev)
1629 {
1630         struct netdev_name_node *name_node;
1631         int count = 0;
1632
1633         list_for_each_entry(name_node, &dev->name_node->list, list) {
1634                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1635                         return -EMSGSIZE;
1636                 count++;
1637         }
1638         return count;
1639 }
1640
1641 static int rtnl_fill_prop_list(struct sk_buff *skb,
1642                                const struct net_device *dev)
1643 {
1644         struct nlattr *prop_list;
1645         int ret;
1646
1647         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1648         if (!prop_list)
1649                 return -EMSGSIZE;
1650
1651         ret = rtnl_fill_alt_ifnames(skb, dev);
1652         if (ret <= 0)
1653                 goto nest_cancel;
1654
1655         nla_nest_end(skb, prop_list);
1656         return 0;
1657
1658 nest_cancel:
1659         nla_nest_cancel(skb, prop_list);
1660         return ret;
1661 }
1662
1663 static int rtnl_fill_proto_down(struct sk_buff *skb,
1664                                 const struct net_device *dev)
1665 {
1666         struct nlattr *pr;
1667         u32 preason;
1668
1669         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1670                 goto nla_put_failure;
1671
1672         preason = dev->proto_down_reason;
1673         if (!preason)
1674                 return 0;
1675
1676         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1677         if (!pr)
1678                 return -EMSGSIZE;
1679
1680         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1681                 nla_nest_cancel(skb, pr);
1682                 goto nla_put_failure;
1683         }
1684
1685         nla_nest_end(skb, pr);
1686         return 0;
1687
1688 nla_put_failure:
1689         return -EMSGSIZE;
1690 }
1691
1692 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1693                             struct net_device *dev, struct net *src_net,
1694                             int type, u32 pid, u32 seq, u32 change,
1695                             unsigned int flags, u32 ext_filter_mask,
1696                             u32 event, int *new_nsid, int new_ifindex,
1697                             int tgt_netnsid, gfp_t gfp)
1698 {
1699         struct ifinfomsg *ifm;
1700         struct nlmsghdr *nlh;
1701
1702         ASSERT_RTNL();
1703         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1704         if (nlh == NULL)
1705                 return -EMSGSIZE;
1706
1707         ifm = nlmsg_data(nlh);
1708         ifm->ifi_family = AF_UNSPEC;
1709         ifm->__ifi_pad = 0;
1710         ifm->ifi_type = dev->type;
1711         ifm->ifi_index = dev->ifindex;
1712         ifm->ifi_flags = dev_get_flags(dev);
1713         ifm->ifi_change = change;
1714
1715         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1716                 goto nla_put_failure;
1717
1718         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1719             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1720             nla_put_u8(skb, IFLA_OPERSTATE,
1721                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1722             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1723             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1724             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1725             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1726             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1727             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1728             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1729             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1730             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1731 #ifdef CONFIG_RPS
1732             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1733 #endif
1734             put_master_ifindex(skb, dev) ||
1735             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1736             (dev->qdisc &&
1737              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1738             nla_put_ifalias(skb, dev) ||
1739             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1740                         atomic_read(&dev->carrier_up_count) +
1741                         atomic_read(&dev->carrier_down_count)) ||
1742             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1743                         atomic_read(&dev->carrier_up_count)) ||
1744             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1745                         atomic_read(&dev->carrier_down_count)))
1746                 goto nla_put_failure;
1747
1748         if (rtnl_fill_proto_down(skb, dev))
1749                 goto nla_put_failure;
1750
1751         if (event != IFLA_EVENT_NONE) {
1752                 if (nla_put_u32(skb, IFLA_EVENT, event))
1753                         goto nla_put_failure;
1754         }
1755
1756         if (rtnl_fill_link_ifmap(skb, dev))
1757                 goto nla_put_failure;
1758
1759         if (dev->addr_len) {
1760                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1761                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1762                         goto nla_put_failure;
1763         }
1764
1765         if (rtnl_phys_port_id_fill(skb, dev))
1766                 goto nla_put_failure;
1767
1768         if (rtnl_phys_port_name_fill(skb, dev))
1769                 goto nla_put_failure;
1770
1771         if (rtnl_phys_switch_id_fill(skb, dev))
1772                 goto nla_put_failure;
1773
1774         if (rtnl_fill_stats(skb, dev))
1775                 goto nla_put_failure;
1776
1777         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1778                 goto nla_put_failure;
1779
1780         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1781                 goto nla_put_failure;
1782
1783         if (rtnl_xdp_fill(skb, dev))
1784                 goto nla_put_failure;
1785
1786         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1787                 if (rtnl_link_fill(skb, dev) < 0)
1788                         goto nla_put_failure;
1789         }
1790
1791         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1792                 goto nla_put_failure;
1793
1794         if (new_nsid &&
1795             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1796                 goto nla_put_failure;
1797         if (new_ifindex &&
1798             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1799                 goto nla_put_failure;
1800
1801         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1802             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1803                 goto nla_put_failure;
1804
1805         rcu_read_lock();
1806         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1807                 goto nla_put_failure_rcu;
1808         rcu_read_unlock();
1809
1810         if (rtnl_fill_prop_list(skb, dev))
1811                 goto nla_put_failure;
1812
1813         if (dev->dev.parent &&
1814             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1815                            dev_name(dev->dev.parent)))
1816                 goto nla_put_failure;
1817
1818         if (dev->dev.parent && dev->dev.parent->bus &&
1819             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1820                            dev->dev.parent->bus->name))
1821                 goto nla_put_failure;
1822
1823         nlmsg_end(skb, nlh);
1824         return 0;
1825
1826 nla_put_failure_rcu:
1827         rcu_read_unlock();
1828 nla_put_failure:
1829         nlmsg_cancel(skb, nlh);
1830         return -EMSGSIZE;
1831 }
1832
1833 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1834         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1835         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1837         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1838         [IFLA_MTU]              = { .type = NLA_U32 },
1839         [IFLA_LINK]             = { .type = NLA_U32 },
1840         [IFLA_MASTER]           = { .type = NLA_U32 },
1841         [IFLA_CARRIER]          = { .type = NLA_U8 },
1842         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1843         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1844         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1845         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1846         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1847         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1848         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1849         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1850          * allow 0-length string (needed to remove an alias).
1851          */
1852         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1853         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1854         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1855         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1856         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1857         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1858         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1859         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1860         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1861         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1862         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1863         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1864         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1865         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1866         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1867         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1868         [IFLA_XDP]              = { .type = NLA_NESTED },
1869         [IFLA_EVENT]            = { .type = NLA_U32 },
1870         [IFLA_GROUP]            = { .type = NLA_U32 },
1871         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1872         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1873         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1874         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1875         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1876         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1877         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1878                                     .len = ALTIFNAMSIZ - 1 },
1879         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1880         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1881         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1882         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1883 };
1884
1885 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1886         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1887         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1888         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1889         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1890 };
1891
1892 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1893         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1894         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1895         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1896         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1897         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1898         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1899         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1900         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1901         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1902         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1903         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1904         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1905         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1906 };
1907
1908 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1909         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1910         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1911                                     .len = PORT_PROFILE_MAX },
1912         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1913                                       .len = PORT_UUID_MAX },
1914         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1915                                     .len = PORT_UUID_MAX },
1916         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1917         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1918
1919         /* Unused, but we need to keep it here since user space could
1920          * fill it. It's also broken with regard to NLA_BINARY use in
1921          * combination with structs.
1922          */
1923         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1924                                     .len = sizeof(struct ifla_port_vsi) },
1925 };
1926
1927 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1928         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1929         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1930         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1931         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1932         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1933         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1934 };
1935
1936 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1937 {
1938         const struct rtnl_link_ops *ops = NULL;
1939         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1940
1941         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1942                 return NULL;
1943
1944         if (linfo[IFLA_INFO_KIND]) {
1945                 char kind[MODULE_NAME_LEN];
1946
1947                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1948                 ops = rtnl_link_ops_get(kind);
1949         }
1950
1951         return ops;
1952 }
1953
1954 static bool link_master_filtered(struct net_device *dev, int master_idx)
1955 {
1956         struct net_device *master;
1957
1958         if (!master_idx)
1959                 return false;
1960
1961         master = netdev_master_upper_dev_get(dev);
1962
1963         /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
1964          * another invalid value for ifindex to denote "no master".
1965          */
1966         if (master_idx == -1)
1967                 return !!master;
1968
1969         if (!master || master->ifindex != master_idx)
1970                 return true;
1971
1972         return false;
1973 }
1974
1975 static bool link_kind_filtered(const struct net_device *dev,
1976                                const struct rtnl_link_ops *kind_ops)
1977 {
1978         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1979                 return true;
1980
1981         return false;
1982 }
1983
1984 static bool link_dump_filtered(struct net_device *dev,
1985                                int master_idx,
1986                                const struct rtnl_link_ops *kind_ops)
1987 {
1988         if (link_master_filtered(dev, master_idx) ||
1989             link_kind_filtered(dev, kind_ops))
1990                 return true;
1991
1992         return false;
1993 }
1994
1995 /**
1996  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1997  * @sk: netlink socket
1998  * @netnsid: network namespace identifier
1999  *
2000  * Returns the network namespace identified by netnsid on success or an error
2001  * pointer on failure.
2002  */
2003 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2004 {
2005         struct net *net;
2006
2007         net = get_net_ns_by_id(sock_net(sk), netnsid);
2008         if (!net)
2009                 return ERR_PTR(-EINVAL);
2010
2011         /* For now, the caller is required to have CAP_NET_ADMIN in
2012          * the user namespace owning the target net ns.
2013          */
2014         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2015                 put_net(net);
2016                 return ERR_PTR(-EACCES);
2017         }
2018         return net;
2019 }
2020 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2021
2022 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2023                                       bool strict_check, struct nlattr **tb,
2024                                       struct netlink_ext_ack *extack)
2025 {
2026         int hdrlen;
2027
2028         if (strict_check) {
2029                 struct ifinfomsg *ifm;
2030
2031                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2032                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2033                         return -EINVAL;
2034                 }
2035
2036                 ifm = nlmsg_data(nlh);
2037                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2038                     ifm->ifi_change) {
2039                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2040                         return -EINVAL;
2041                 }
2042                 if (ifm->ifi_index) {
2043                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2044                         return -EINVAL;
2045                 }
2046
2047                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2048                                                      IFLA_MAX, ifla_policy,
2049                                                      extack);
2050         }
2051
2052         /* A hack to preserve kernel<->userspace interface.
2053          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2054          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2055          * what iproute2 < v3.9.0 used.
2056          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2057          * attribute, its netlink message is shorter than struct ifinfomsg.
2058          */
2059         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2060                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2061
2062         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2063                                       extack);
2064 }
2065
2066 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2067 {
2068         struct netlink_ext_ack *extack = cb->extack;
2069         const struct nlmsghdr *nlh = cb->nlh;
2070         struct net *net = sock_net(skb->sk);
2071         struct net *tgt_net = net;
2072         int h, s_h;
2073         int idx = 0, s_idx;
2074         struct net_device *dev;
2075         struct hlist_head *head;
2076         struct nlattr *tb[IFLA_MAX+1];
2077         u32 ext_filter_mask = 0;
2078         const struct rtnl_link_ops *kind_ops = NULL;
2079         unsigned int flags = NLM_F_MULTI;
2080         int master_idx = 0;
2081         int netnsid = -1;
2082         int err, i;
2083
2084         s_h = cb->args[0];
2085         s_idx = cb->args[1];
2086
2087         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2088         if (err < 0) {
2089                 if (cb->strict_check)
2090                         return err;
2091
2092                 goto walk_entries;
2093         }
2094
2095         for (i = 0; i <= IFLA_MAX; ++i) {
2096                 if (!tb[i])
2097                         continue;
2098
2099                 /* new attributes should only be added with strict checking */
2100                 switch (i) {
2101                 case IFLA_TARGET_NETNSID:
2102                         netnsid = nla_get_s32(tb[i]);
2103                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2104                         if (IS_ERR(tgt_net)) {
2105                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2106                                 return PTR_ERR(tgt_net);
2107                         }
2108                         break;
2109                 case IFLA_EXT_MASK:
2110                         ext_filter_mask = nla_get_u32(tb[i]);
2111                         break;
2112                 case IFLA_MASTER:
2113                         master_idx = nla_get_u32(tb[i]);
2114                         break;
2115                 case IFLA_LINKINFO:
2116                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2117                         break;
2118                 default:
2119                         if (cb->strict_check) {
2120                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2121                                 return -EINVAL;
2122                         }
2123                 }
2124         }
2125
2126         if (master_idx || kind_ops)
2127                 flags |= NLM_F_DUMP_FILTERED;
2128
2129 walk_entries:
2130         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2131                 idx = 0;
2132                 head = &tgt_net->dev_index_head[h];
2133                 hlist_for_each_entry(dev, head, index_hlist) {
2134                         if (link_dump_filtered(dev, master_idx, kind_ops))
2135                                 goto cont;
2136                         if (idx < s_idx)
2137                                 goto cont;
2138                         err = rtnl_fill_ifinfo(skb, dev, net,
2139                                                RTM_NEWLINK,
2140                                                NETLINK_CB(cb->skb).portid,
2141                                                nlh->nlmsg_seq, 0, flags,
2142                                                ext_filter_mask, 0, NULL, 0,
2143                                                netnsid, GFP_KERNEL);
2144
2145                         if (err < 0) {
2146                                 if (likely(skb->len))
2147                                         goto out;
2148
2149                                 goto out_err;
2150                         }
2151 cont:
2152                         idx++;
2153                 }
2154         }
2155 out:
2156         err = skb->len;
2157 out_err:
2158         cb->args[1] = idx;
2159         cb->args[0] = h;
2160         cb->seq = tgt_net->dev_base_seq;
2161         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2162         if (netnsid >= 0)
2163                 put_net(tgt_net);
2164
2165         return err;
2166 }
2167
2168 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2169                         struct netlink_ext_ack *exterr)
2170 {
2171         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2172                                     exterr);
2173 }
2174 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2175
2176 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2177 {
2178         struct net *net;
2179         /* Examine the link attributes and figure out which
2180          * network namespace we are talking about.
2181          */
2182         if (tb[IFLA_NET_NS_PID])
2183                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2184         else if (tb[IFLA_NET_NS_FD])
2185                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2186         else
2187                 net = get_net(src_net);
2188         return net;
2189 }
2190 EXPORT_SYMBOL(rtnl_link_get_net);
2191
2192 /* Figure out which network namespace we are talking about by
2193  * examining the link attributes in the following order:
2194  *
2195  * 1. IFLA_NET_NS_PID
2196  * 2. IFLA_NET_NS_FD
2197  * 3. IFLA_TARGET_NETNSID
2198  */
2199 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2200                                                struct nlattr *tb[])
2201 {
2202         struct net *net;
2203
2204         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2205                 return rtnl_link_get_net(src_net, tb);
2206
2207         if (!tb[IFLA_TARGET_NETNSID])
2208                 return get_net(src_net);
2209
2210         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2211         if (!net)
2212                 return ERR_PTR(-EINVAL);
2213
2214         return net;
2215 }
2216
2217 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2218                                              struct net *src_net,
2219                                              struct nlattr *tb[], int cap)
2220 {
2221         struct net *net;
2222
2223         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2224         if (IS_ERR(net))
2225                 return net;
2226
2227         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2228                 put_net(net);
2229                 return ERR_PTR(-EPERM);
2230         }
2231
2232         return net;
2233 }
2234
2235 /* Verify that rtnetlink requests do not pass additional properties
2236  * potentially referring to different network namespaces.
2237  */
2238 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2239                                     struct netlink_ext_ack *extack,
2240                                     bool netns_id_only)
2241 {
2242
2243         if (netns_id_only) {
2244                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2245                         return 0;
2246
2247                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2248                 return -EOPNOTSUPP;
2249         }
2250
2251         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2252                 goto invalid_attr;
2253
2254         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2255                 goto invalid_attr;
2256
2257         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2258                 goto invalid_attr;
2259
2260         return 0;
2261
2262 invalid_attr:
2263         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2264         return -EINVAL;
2265 }
2266
2267 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2268                             struct netlink_ext_ack *extack)
2269 {
2270         if (dev) {
2271                 if (tb[IFLA_ADDRESS] &&
2272                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2273                         return -EINVAL;
2274
2275                 if (tb[IFLA_BROADCAST] &&
2276                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2277                         return -EINVAL;
2278         }
2279
2280         if (tb[IFLA_AF_SPEC]) {
2281                 struct nlattr *af;
2282                 int rem, err;
2283
2284                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2285                         const struct rtnl_af_ops *af_ops;
2286
2287                         af_ops = rtnl_af_lookup(nla_type(af));
2288                         if (!af_ops)
2289                                 return -EAFNOSUPPORT;
2290
2291                         if (!af_ops->set_link_af)
2292                                 return -EOPNOTSUPP;
2293
2294                         if (af_ops->validate_link_af) {
2295                                 err = af_ops->validate_link_af(dev, af, extack);
2296                                 if (err < 0)
2297                                         return err;
2298                         }
2299                 }
2300         }
2301
2302         return 0;
2303 }
2304
2305 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2306                                   int guid_type)
2307 {
2308         const struct net_device_ops *ops = dev->netdev_ops;
2309
2310         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2311 }
2312
2313 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2314 {
2315         if (dev->type != ARPHRD_INFINIBAND)
2316                 return -EOPNOTSUPP;
2317
2318         return handle_infiniband_guid(dev, ivt, guid_type);
2319 }
2320
2321 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2322 {
2323         const struct net_device_ops *ops = dev->netdev_ops;
2324         int err = -EINVAL;
2325
2326         if (tb[IFLA_VF_MAC]) {
2327                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2328
2329                 if (ivm->vf >= INT_MAX)
2330                         return -EINVAL;
2331                 err = -EOPNOTSUPP;
2332                 if (ops->ndo_set_vf_mac)
2333                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2334                                                   ivm->mac);
2335                 if (err < 0)
2336                         return err;
2337         }
2338
2339         if (tb[IFLA_VF_VLAN]) {
2340                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2341
2342                 if (ivv->vf >= INT_MAX)
2343                         return -EINVAL;
2344                 err = -EOPNOTSUPP;
2345                 if (ops->ndo_set_vf_vlan)
2346                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2347                                                    ivv->qos,
2348                                                    htons(ETH_P_8021Q));
2349                 if (err < 0)
2350                         return err;
2351         }
2352
2353         if (tb[IFLA_VF_VLAN_LIST]) {
2354                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2355                 struct nlattr *attr;
2356                 int rem, len = 0;
2357
2358                 err = -EOPNOTSUPP;
2359                 if (!ops->ndo_set_vf_vlan)
2360                         return err;
2361
2362                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2363                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2364                             nla_len(attr) < NLA_HDRLEN) {
2365                                 return -EINVAL;
2366                         }
2367                         if (len >= MAX_VLAN_LIST_LEN)
2368                                 return -EOPNOTSUPP;
2369                         ivvl[len] = nla_data(attr);
2370
2371                         len++;
2372                 }
2373                 if (len == 0)
2374                         return -EINVAL;
2375
2376                 if (ivvl[0]->vf >= INT_MAX)
2377                         return -EINVAL;
2378                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2379                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2380                 if (err < 0)
2381                         return err;
2382         }
2383
2384         if (tb[IFLA_VF_TX_RATE]) {
2385                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2386                 struct ifla_vf_info ivf;
2387
2388                 if (ivt->vf >= INT_MAX)
2389                         return -EINVAL;
2390                 err = -EOPNOTSUPP;
2391                 if (ops->ndo_get_vf_config)
2392                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2393                 if (err < 0)
2394                         return err;
2395
2396                 err = -EOPNOTSUPP;
2397                 if (ops->ndo_set_vf_rate)
2398                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2399                                                    ivf.min_tx_rate,
2400                                                    ivt->rate);
2401                 if (err < 0)
2402                         return err;
2403         }
2404
2405         if (tb[IFLA_VF_RATE]) {
2406                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2407
2408                 if (ivt->vf >= INT_MAX)
2409                         return -EINVAL;
2410                 err = -EOPNOTSUPP;
2411                 if (ops->ndo_set_vf_rate)
2412                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2413                                                    ivt->min_tx_rate,
2414                                                    ivt->max_tx_rate);
2415                 if (err < 0)
2416                         return err;
2417         }
2418
2419         if (tb[IFLA_VF_SPOOFCHK]) {
2420                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2421
2422                 if (ivs->vf >= INT_MAX)
2423                         return -EINVAL;
2424                 err = -EOPNOTSUPP;
2425                 if (ops->ndo_set_vf_spoofchk)
2426                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2427                                                        ivs->setting);
2428                 if (err < 0)
2429                         return err;
2430         }
2431
2432         if (tb[IFLA_VF_LINK_STATE]) {
2433                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2434
2435                 if (ivl->vf >= INT_MAX)
2436                         return -EINVAL;
2437                 err = -EOPNOTSUPP;
2438                 if (ops->ndo_set_vf_link_state)
2439                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2440                                                          ivl->link_state);
2441                 if (err < 0)
2442                         return err;
2443         }
2444
2445         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2446                 struct ifla_vf_rss_query_en *ivrssq_en;
2447
2448                 err = -EOPNOTSUPP;
2449                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2450                 if (ivrssq_en->vf >= INT_MAX)
2451                         return -EINVAL;
2452                 if (ops->ndo_set_vf_rss_query_en)
2453                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2454                                                            ivrssq_en->setting);
2455                 if (err < 0)
2456                         return err;
2457         }
2458
2459         if (tb[IFLA_VF_TRUST]) {
2460                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2461
2462                 if (ivt->vf >= INT_MAX)
2463                         return -EINVAL;
2464                 err = -EOPNOTSUPP;
2465                 if (ops->ndo_set_vf_trust)
2466                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2467                 if (err < 0)
2468                         return err;
2469         }
2470
2471         if (tb[IFLA_VF_IB_NODE_GUID]) {
2472                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2473
2474                 if (ivt->vf >= INT_MAX)
2475                         return -EINVAL;
2476                 if (!ops->ndo_set_vf_guid)
2477                         return -EOPNOTSUPP;
2478                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2479         }
2480
2481         if (tb[IFLA_VF_IB_PORT_GUID]) {
2482                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2483
2484                 if (ivt->vf >= INT_MAX)
2485                         return -EINVAL;
2486                 if (!ops->ndo_set_vf_guid)
2487                         return -EOPNOTSUPP;
2488
2489                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2490         }
2491
2492         return err;
2493 }
2494
2495 static int do_set_master(struct net_device *dev, int ifindex,
2496                          struct netlink_ext_ack *extack)
2497 {
2498         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2499         const struct net_device_ops *ops;
2500         int err;
2501
2502         if (upper_dev) {
2503                 if (upper_dev->ifindex == ifindex)
2504                         return 0;
2505                 ops = upper_dev->netdev_ops;
2506                 if (ops->ndo_del_slave) {
2507                         err = ops->ndo_del_slave(upper_dev, dev);
2508                         if (err)
2509                                 return err;
2510                 } else {
2511                         return -EOPNOTSUPP;
2512                 }
2513         }
2514
2515         if (ifindex) {
2516                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2517                 if (!upper_dev)
2518                         return -EINVAL;
2519                 ops = upper_dev->netdev_ops;
2520                 if (ops->ndo_add_slave) {
2521                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2522                         if (err)
2523                                 return err;
2524                 } else {
2525                         return -EOPNOTSUPP;
2526                 }
2527         }
2528         return 0;
2529 }
2530
2531 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2532         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2533         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2534 };
2535
2536 static int do_set_proto_down(struct net_device *dev,
2537                              struct nlattr *nl_proto_down,
2538                              struct nlattr *nl_proto_down_reason,
2539                              struct netlink_ext_ack *extack)
2540 {
2541         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2542         const struct net_device_ops *ops = dev->netdev_ops;
2543         unsigned long mask = 0;
2544         u32 value;
2545         bool proto_down;
2546         int err;
2547
2548         if (!ops->ndo_change_proto_down) {
2549                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2550                 return -EOPNOTSUPP;
2551         }
2552
2553         if (nl_proto_down_reason) {
2554                 err = nla_parse_nested_deprecated(pdreason,
2555                                                   IFLA_PROTO_DOWN_REASON_MAX,
2556                                                   nl_proto_down_reason,
2557                                                   ifla_proto_down_reason_policy,
2558                                                   NULL);
2559                 if (err < 0)
2560                         return err;
2561
2562                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2563                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2564                         return -EINVAL;
2565                 }
2566
2567                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2568
2569                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2570                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2571
2572                 dev_change_proto_down_reason(dev, mask, value);
2573         }
2574
2575         if (nl_proto_down) {
2576                 proto_down = nla_get_u8(nl_proto_down);
2577
2578                 /* Don't turn off protodown if there are active reasons */
2579                 if (!proto_down && dev->proto_down_reason) {
2580                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2581                         return -EBUSY;
2582                 }
2583                 err = dev_change_proto_down(dev,
2584                                             proto_down);
2585                 if (err)
2586                         return err;
2587         }
2588
2589         return 0;
2590 }
2591
2592 #define DO_SETLINK_MODIFIED     0x01
2593 /* notify flag means notify + modified. */
2594 #define DO_SETLINK_NOTIFY       0x03
2595 static int do_setlink(const struct sk_buff *skb,
2596                       struct net_device *dev, struct ifinfomsg *ifm,
2597                       struct netlink_ext_ack *extack,
2598                       struct nlattr **tb, char *ifname, int status)
2599 {
2600         const struct net_device_ops *ops = dev->netdev_ops;
2601         int err;
2602
2603         err = validate_linkmsg(dev, tb, extack);
2604         if (err < 0)
2605                 return err;
2606
2607         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2608                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2609                 struct net *net;
2610                 int new_ifindex;
2611
2612                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2613                                                 tb, CAP_NET_ADMIN);
2614                 if (IS_ERR(net)) {
2615                         err = PTR_ERR(net);
2616                         goto errout;
2617                 }
2618
2619                 if (tb[IFLA_NEW_IFINDEX])
2620                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2621                 else
2622                         new_ifindex = 0;
2623
2624                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2625                 put_net(net);
2626                 if (err)
2627                         goto errout;
2628                 status |= DO_SETLINK_MODIFIED;
2629         }
2630
2631         if (tb[IFLA_MAP]) {
2632                 struct rtnl_link_ifmap *u_map;
2633                 struct ifmap k_map;
2634
2635                 if (!ops->ndo_set_config) {
2636                         err = -EOPNOTSUPP;
2637                         goto errout;
2638                 }
2639
2640                 if (!netif_device_present(dev)) {
2641                         err = -ENODEV;
2642                         goto errout;
2643                 }
2644
2645                 u_map = nla_data(tb[IFLA_MAP]);
2646                 k_map.mem_start = (unsigned long) u_map->mem_start;
2647                 k_map.mem_end = (unsigned long) u_map->mem_end;
2648                 k_map.base_addr = (unsigned short) u_map->base_addr;
2649                 k_map.irq = (unsigned char) u_map->irq;
2650                 k_map.dma = (unsigned char) u_map->dma;
2651                 k_map.port = (unsigned char) u_map->port;
2652
2653                 err = ops->ndo_set_config(dev, &k_map);
2654                 if (err < 0)
2655                         goto errout;
2656
2657                 status |= DO_SETLINK_NOTIFY;
2658         }
2659
2660         if (tb[IFLA_ADDRESS]) {
2661                 struct sockaddr *sa;
2662                 int len;
2663
2664                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2665                                                   sizeof(*sa));
2666                 sa = kmalloc(len, GFP_KERNEL);
2667                 if (!sa) {
2668                         err = -ENOMEM;
2669                         goto errout;
2670                 }
2671                 sa->sa_family = dev->type;
2672                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2673                        dev->addr_len);
2674                 err = dev_set_mac_address_user(dev, sa, extack);
2675                 kfree(sa);
2676                 if (err)
2677                         goto errout;
2678                 status |= DO_SETLINK_MODIFIED;
2679         }
2680
2681         if (tb[IFLA_MTU]) {
2682                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2683                 if (err < 0)
2684                         goto errout;
2685                 status |= DO_SETLINK_MODIFIED;
2686         }
2687
2688         if (tb[IFLA_GROUP]) {
2689                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2690                 status |= DO_SETLINK_NOTIFY;
2691         }
2692
2693         /*
2694          * Interface selected by interface index but interface
2695          * name provided implies that a name change has been
2696          * requested.
2697          */
2698         if (ifm->ifi_index > 0 && ifname[0]) {
2699                 err = dev_change_name(dev, ifname);
2700                 if (err < 0)
2701                         goto errout;
2702                 status |= DO_SETLINK_MODIFIED;
2703         }
2704
2705         if (tb[IFLA_IFALIAS]) {
2706                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2707                                     nla_len(tb[IFLA_IFALIAS]));
2708                 if (err < 0)
2709                         goto errout;
2710                 status |= DO_SETLINK_NOTIFY;
2711         }
2712
2713         if (tb[IFLA_BROADCAST]) {
2714                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2715                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2716         }
2717
2718         if (ifm->ifi_flags || ifm->ifi_change) {
2719                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2720                                        extack);
2721                 if (err < 0)
2722                         goto errout;
2723         }
2724
2725         if (tb[IFLA_MASTER]) {
2726                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2727                 if (err)
2728                         goto errout;
2729                 status |= DO_SETLINK_MODIFIED;
2730         }
2731
2732         if (tb[IFLA_CARRIER]) {
2733                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2734                 if (err)
2735                         goto errout;
2736                 status |= DO_SETLINK_MODIFIED;
2737         }
2738
2739         if (tb[IFLA_TXQLEN]) {
2740                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2741
2742                 err = dev_change_tx_queue_len(dev, value);
2743                 if (err)
2744                         goto errout;
2745                 status |= DO_SETLINK_MODIFIED;
2746         }
2747
2748         if (tb[IFLA_GSO_MAX_SIZE]) {
2749                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2750
2751                 if (max_size > GSO_MAX_SIZE) {
2752                         err = -EINVAL;
2753                         goto errout;
2754                 }
2755
2756                 if (dev->gso_max_size ^ max_size) {
2757                         netif_set_gso_max_size(dev, max_size);
2758                         status |= DO_SETLINK_MODIFIED;
2759                 }
2760         }
2761
2762         if (tb[IFLA_GSO_MAX_SEGS]) {
2763                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2764
2765                 if (max_segs > GSO_MAX_SEGS) {
2766                         err = -EINVAL;
2767                         goto errout;
2768                 }
2769
2770                 if (dev->gso_max_segs ^ max_segs) {
2771                         dev->gso_max_segs = max_segs;
2772                         status |= DO_SETLINK_MODIFIED;
2773                 }
2774         }
2775
2776         if (tb[IFLA_OPERSTATE])
2777                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2778
2779         if (tb[IFLA_LINKMODE]) {
2780                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2781
2782                 write_lock_bh(&dev_base_lock);
2783                 if (dev->link_mode ^ value)
2784                         status |= DO_SETLINK_NOTIFY;
2785                 dev->link_mode = value;
2786                 write_unlock_bh(&dev_base_lock);
2787         }
2788
2789         if (tb[IFLA_VFINFO_LIST]) {
2790                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2791                 struct nlattr *attr;
2792                 int rem;
2793
2794                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2795                         if (nla_type(attr) != IFLA_VF_INFO ||
2796                             nla_len(attr) < NLA_HDRLEN) {
2797                                 err = -EINVAL;
2798                                 goto errout;
2799                         }
2800                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2801                                                           attr,
2802                                                           ifla_vf_policy,
2803                                                           NULL);
2804                         if (err < 0)
2805                                 goto errout;
2806                         err = do_setvfinfo(dev, vfinfo);
2807                         if (err < 0)
2808                                 goto errout;
2809                         status |= DO_SETLINK_NOTIFY;
2810                 }
2811         }
2812         err = 0;
2813
2814         if (tb[IFLA_VF_PORTS]) {
2815                 struct nlattr *port[IFLA_PORT_MAX+1];
2816                 struct nlattr *attr;
2817                 int vf;
2818                 int rem;
2819
2820                 err = -EOPNOTSUPP;
2821                 if (!ops->ndo_set_vf_port)
2822                         goto errout;
2823
2824                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2825                         if (nla_type(attr) != IFLA_VF_PORT ||
2826                             nla_len(attr) < NLA_HDRLEN) {
2827                                 err = -EINVAL;
2828                                 goto errout;
2829                         }
2830                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2831                                                           attr,
2832                                                           ifla_port_policy,
2833                                                           NULL);
2834                         if (err < 0)
2835                                 goto errout;
2836                         if (!port[IFLA_PORT_VF]) {
2837                                 err = -EOPNOTSUPP;
2838                                 goto errout;
2839                         }
2840                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2841                         err = ops->ndo_set_vf_port(dev, vf, port);
2842                         if (err < 0)
2843                                 goto errout;
2844                         status |= DO_SETLINK_NOTIFY;
2845                 }
2846         }
2847         err = 0;
2848
2849         if (tb[IFLA_PORT_SELF]) {
2850                 struct nlattr *port[IFLA_PORT_MAX+1];
2851
2852                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2853                                                   tb[IFLA_PORT_SELF],
2854                                                   ifla_port_policy, NULL);
2855                 if (err < 0)
2856                         goto errout;
2857
2858                 err = -EOPNOTSUPP;
2859                 if (ops->ndo_set_vf_port)
2860                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2861                 if (err < 0)
2862                         goto errout;
2863                 status |= DO_SETLINK_NOTIFY;
2864         }
2865
2866         if (tb[IFLA_AF_SPEC]) {
2867                 struct nlattr *af;
2868                 int rem;
2869
2870                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2871                         const struct rtnl_af_ops *af_ops;
2872
2873                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2874
2875                         err = af_ops->set_link_af(dev, af, extack);
2876                         if (err < 0)
2877                                 goto errout;
2878
2879                         status |= DO_SETLINK_NOTIFY;
2880                 }
2881         }
2882         err = 0;
2883
2884         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2885                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2886                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2887                 if (err)
2888                         goto errout;
2889                 status |= DO_SETLINK_NOTIFY;
2890         }
2891
2892         if (tb[IFLA_XDP]) {
2893                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2894                 u32 xdp_flags = 0;
2895
2896                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2897                                                   tb[IFLA_XDP],
2898                                                   ifla_xdp_policy, NULL);
2899                 if (err < 0)
2900                         goto errout;
2901
2902                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2903                         err = -EINVAL;
2904                         goto errout;
2905                 }
2906
2907                 if (xdp[IFLA_XDP_FLAGS]) {
2908                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2909                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2910                                 err = -EINVAL;
2911                                 goto errout;
2912                         }
2913                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2914                                 err = -EINVAL;
2915                                 goto errout;
2916                         }
2917                 }
2918
2919                 if (xdp[IFLA_XDP_FD]) {
2920                         int expected_fd = -1;
2921
2922                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2923                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2924                                         err = -EINVAL;
2925                                         goto errout;
2926                                 }
2927                                 expected_fd =
2928                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2929                         }
2930
2931                         err = dev_change_xdp_fd(dev, extack,
2932                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2933                                                 expected_fd,
2934                                                 xdp_flags);
2935                         if (err)
2936                                 goto errout;
2937                         status |= DO_SETLINK_NOTIFY;
2938                 }
2939         }
2940
2941 errout:
2942         if (status & DO_SETLINK_MODIFIED) {
2943                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2944                         netdev_state_change(dev);
2945
2946                 if (err < 0)
2947                         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",
2948                                              dev->name);
2949         }
2950
2951         return err;
2952 }
2953
2954 static struct net_device *rtnl_dev_get(struct net *net,
2955                                        struct nlattr *ifname_attr,
2956                                        struct nlattr *altifname_attr,
2957                                        char *ifname)
2958 {
2959         char buffer[ALTIFNAMSIZ];
2960
2961         if (!ifname) {
2962                 ifname = buffer;
2963                 if (ifname_attr)
2964                         nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2965                 else if (altifname_attr)
2966                         nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2967                 else
2968                         return NULL;
2969         }
2970
2971         return __dev_get_by_name(net, ifname);
2972 }
2973
2974 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2975                         struct netlink_ext_ack *extack)
2976 {
2977         struct net *net = sock_net(skb->sk);
2978         struct ifinfomsg *ifm;
2979         struct net_device *dev;
2980         int err;
2981         struct nlattr *tb[IFLA_MAX+1];
2982         char ifname[IFNAMSIZ];
2983
2984         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2985                                      ifla_policy, extack);
2986         if (err < 0)
2987                 goto errout;
2988
2989         err = rtnl_ensure_unique_netns(tb, extack, false);
2990         if (err < 0)
2991                 goto errout;
2992
2993         if (tb[IFLA_IFNAME])
2994                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2995         else
2996                 ifname[0] = '\0';
2997
2998         err = -EINVAL;
2999         ifm = nlmsg_data(nlh);
3000         if (ifm->ifi_index > 0)
3001                 dev = __dev_get_by_index(net, ifm->ifi_index);
3002         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3003                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3004         else
3005                 goto errout;
3006
3007         if (dev == NULL) {
3008                 err = -ENODEV;
3009                 goto errout;
3010         }
3011
3012         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3013 errout:
3014         return err;
3015 }
3016
3017 static int rtnl_group_dellink(const struct net *net, int group)
3018 {
3019         struct net_device *dev, *aux;
3020         LIST_HEAD(list_kill);
3021         bool found = false;
3022
3023         if (!group)
3024                 return -EPERM;
3025
3026         for_each_netdev(net, dev) {
3027                 if (dev->group == group) {
3028                         const struct rtnl_link_ops *ops;
3029
3030                         found = true;
3031                         ops = dev->rtnl_link_ops;
3032                         if (!ops || !ops->dellink)
3033                                 return -EOPNOTSUPP;
3034                 }
3035         }
3036
3037         if (!found)
3038                 return -ENODEV;
3039
3040         for_each_netdev_safe(net, dev, aux) {
3041                 if (dev->group == group) {
3042                         const struct rtnl_link_ops *ops;
3043
3044                         ops = dev->rtnl_link_ops;
3045                         ops->dellink(dev, &list_kill);
3046                 }
3047         }
3048         unregister_netdevice_many(&list_kill);
3049
3050         return 0;
3051 }
3052
3053 int rtnl_delete_link(struct net_device *dev)
3054 {
3055         const struct rtnl_link_ops *ops;
3056         LIST_HEAD(list_kill);
3057
3058         ops = dev->rtnl_link_ops;
3059         if (!ops || !ops->dellink)
3060                 return -EOPNOTSUPP;
3061
3062         ops->dellink(dev, &list_kill);
3063         unregister_netdevice_many(&list_kill);
3064
3065         return 0;
3066 }
3067 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3068
3069 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3070                         struct netlink_ext_ack *extack)
3071 {
3072         struct net *net = sock_net(skb->sk);
3073         struct net *tgt_net = net;
3074         struct net_device *dev = NULL;
3075         struct ifinfomsg *ifm;
3076         struct nlattr *tb[IFLA_MAX+1];
3077         int err;
3078         int netnsid = -1;
3079
3080         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3081                                      ifla_policy, extack);
3082         if (err < 0)
3083                 return err;
3084
3085         err = rtnl_ensure_unique_netns(tb, extack, true);
3086         if (err < 0)
3087                 return err;
3088
3089         if (tb[IFLA_TARGET_NETNSID]) {
3090                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3091                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3092                 if (IS_ERR(tgt_net))
3093                         return PTR_ERR(tgt_net);
3094         }
3095
3096         err = -EINVAL;
3097         ifm = nlmsg_data(nlh);
3098         if (ifm->ifi_index > 0)
3099                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3100         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3101                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3102                                    tb[IFLA_ALT_IFNAME], NULL);
3103         else if (tb[IFLA_GROUP])
3104                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3105         else
3106                 goto out;
3107
3108         if (!dev) {
3109                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3110                         err = -ENODEV;
3111
3112                 goto out;
3113         }
3114
3115         err = rtnl_delete_link(dev);
3116
3117 out:
3118         if (netnsid >= 0)
3119                 put_net(tgt_net);
3120
3121         return err;
3122 }
3123
3124 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3125 {
3126         unsigned int old_flags;
3127         int err;
3128
3129         old_flags = dev->flags;
3130         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3131                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3132                                          NULL);
3133                 if (err < 0)
3134                         return err;
3135         }
3136
3137         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3138                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3139         } else {
3140                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3141                 __dev_notify_flags(dev, old_flags, ~0U);
3142         }
3143         return 0;
3144 }
3145 EXPORT_SYMBOL(rtnl_configure_link);
3146
3147 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3148                                     unsigned char name_assign_type,
3149                                     const struct rtnl_link_ops *ops,
3150                                     struct nlattr *tb[],
3151                                     struct netlink_ext_ack *extack)
3152 {
3153         struct net_device *dev;
3154         unsigned int num_tx_queues = 1;
3155         unsigned int num_rx_queues = 1;
3156
3157         if (tb[IFLA_NUM_TX_QUEUES])
3158                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3159         else if (ops->get_num_tx_queues)
3160                 num_tx_queues = ops->get_num_tx_queues();
3161
3162         if (tb[IFLA_NUM_RX_QUEUES])
3163                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3164         else if (ops->get_num_rx_queues)
3165                 num_rx_queues = ops->get_num_rx_queues();
3166
3167         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3168                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3169                 return ERR_PTR(-EINVAL);
3170         }
3171
3172         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3173                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3174                 return ERR_PTR(-EINVAL);
3175         }
3176
3177         if (ops->alloc) {
3178                 dev = ops->alloc(tb, ifname, name_assign_type,
3179                                  num_tx_queues, num_rx_queues);
3180                 if (IS_ERR(dev))
3181                         return dev;
3182         } else {
3183                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3184                                        name_assign_type, ops->setup,
3185                                        num_tx_queues, num_rx_queues);
3186         }
3187
3188         if (!dev)
3189                 return ERR_PTR(-ENOMEM);
3190
3191         dev_net_set(dev, net);
3192         dev->rtnl_link_ops = ops;
3193         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3194
3195         if (tb[IFLA_MTU]) {
3196                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3197                 int err;
3198
3199                 err = dev_validate_mtu(dev, mtu, extack);
3200                 if (err) {
3201                         free_netdev(dev);
3202                         return ERR_PTR(err);
3203                 }
3204                 dev->mtu = mtu;
3205         }
3206         if (tb[IFLA_ADDRESS]) {
3207                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3208                                 nla_len(tb[IFLA_ADDRESS]));
3209                 dev->addr_assign_type = NET_ADDR_SET;
3210         }
3211         if (tb[IFLA_BROADCAST])
3212                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3213                                 nla_len(tb[IFLA_BROADCAST]));
3214         if (tb[IFLA_TXQLEN])
3215                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3216         if (tb[IFLA_OPERSTATE])
3217                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3218         if (tb[IFLA_LINKMODE])
3219                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3220         if (tb[IFLA_GROUP])
3221                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3222         if (tb[IFLA_GSO_MAX_SIZE])
3223                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3224         if (tb[IFLA_GSO_MAX_SEGS])
3225                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3226
3227         return dev;
3228 }
3229 EXPORT_SYMBOL(rtnl_create_link);
3230
3231 static int rtnl_group_changelink(const struct sk_buff *skb,
3232                 struct net *net, int group,
3233                 struct ifinfomsg *ifm,
3234                 struct netlink_ext_ack *extack,
3235                 struct nlattr **tb)
3236 {
3237         struct net_device *dev, *aux;
3238         int err;
3239
3240         for_each_netdev_safe(net, dev, aux) {
3241                 if (dev->group == group) {
3242                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3243                         if (err < 0)
3244                                 return err;
3245                 }
3246         }
3247
3248         return 0;
3249 }
3250
3251 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3252                           struct nlattr **attr, struct netlink_ext_ack *extack)
3253 {
3254         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3255         unsigned char name_assign_type = NET_NAME_USER;
3256         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3257         const struct rtnl_link_ops *m_ops = NULL;
3258         struct net_device *master_dev = NULL;
3259         struct net *net = sock_net(skb->sk);
3260         const struct rtnl_link_ops *ops;
3261         struct nlattr *tb[IFLA_MAX + 1];
3262         struct net *dest_net, *link_net;
3263         struct nlattr **slave_data;
3264         char kind[MODULE_NAME_LEN];
3265         struct net_device *dev;
3266         struct ifinfomsg *ifm;
3267         char ifname[IFNAMSIZ];
3268         struct nlattr **data;
3269         int err;
3270
3271 #ifdef CONFIG_MODULES
3272 replay:
3273 #endif
3274         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3275                                      ifla_policy, extack);
3276         if (err < 0)
3277                 return err;
3278
3279         err = rtnl_ensure_unique_netns(tb, extack, false);
3280         if (err < 0)
3281                 return err;
3282
3283         if (tb[IFLA_IFNAME])
3284                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3285         else
3286                 ifname[0] = '\0';
3287
3288         ifm = nlmsg_data(nlh);
3289         if (ifm->ifi_index > 0)
3290                 dev = __dev_get_by_index(net, ifm->ifi_index);
3291         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3292                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3293         else
3294                 dev = NULL;
3295
3296         if (dev) {
3297                 master_dev = netdev_master_upper_dev_get(dev);
3298                 if (master_dev)
3299                         m_ops = master_dev->rtnl_link_ops;
3300         }
3301
3302         err = validate_linkmsg(dev, tb, extack);
3303         if (err < 0)
3304                 return err;
3305
3306         if (tb[IFLA_LINKINFO]) {
3307                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3308                                                   tb[IFLA_LINKINFO],
3309                                                   ifla_info_policy, NULL);
3310                 if (err < 0)
3311                         return err;
3312         } else
3313                 memset(linkinfo, 0, sizeof(linkinfo));
3314
3315         if (linkinfo[IFLA_INFO_KIND]) {
3316                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3317                 ops = rtnl_link_ops_get(kind);
3318         } else {
3319                 kind[0] = '\0';
3320                 ops = NULL;
3321         }
3322
3323         data = NULL;
3324         if (ops) {
3325                 if (ops->maxtype > RTNL_MAX_TYPE)
3326                         return -EINVAL;
3327
3328                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3329                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3330                                                           linkinfo[IFLA_INFO_DATA],
3331                                                           ops->policy, extack);
3332                         if (err < 0)
3333                                 return err;
3334                         data = attr;
3335                 }
3336                 if (ops->validate) {
3337                         err = ops->validate(tb, data, extack);
3338                         if (err < 0)
3339                                 return err;
3340                 }
3341         }
3342
3343         slave_data = NULL;
3344         if (m_ops) {
3345                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3346                         return -EINVAL;
3347
3348                 if (m_ops->slave_maxtype &&
3349                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3350                         err = nla_parse_nested_deprecated(slave_attr,
3351                                                           m_ops->slave_maxtype,
3352                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3353                                                           m_ops->slave_policy,
3354                                                           extack);
3355                         if (err < 0)
3356                                 return err;
3357                         slave_data = slave_attr;
3358                 }
3359         }
3360
3361         if (dev) {
3362                 int status = 0;
3363
3364                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3365                         return -EEXIST;
3366                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3367                         return -EOPNOTSUPP;
3368
3369                 if (linkinfo[IFLA_INFO_DATA]) {
3370                         if (!ops || ops != dev->rtnl_link_ops ||
3371                             !ops->changelink)
3372                                 return -EOPNOTSUPP;
3373
3374                         err = ops->changelink(dev, tb, data, extack);
3375                         if (err < 0)
3376                                 return err;
3377                         status |= DO_SETLINK_NOTIFY;
3378                 }
3379
3380                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3381                         if (!m_ops || !m_ops->slave_changelink)
3382                                 return -EOPNOTSUPP;
3383
3384                         err = m_ops->slave_changelink(master_dev, dev, tb,
3385                                                       slave_data, extack);
3386                         if (err < 0)
3387                                 return err;
3388                         status |= DO_SETLINK_NOTIFY;
3389                 }
3390
3391                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3392         }
3393
3394         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3395                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3396                         return rtnl_group_changelink(skb, net,
3397                                                 nla_get_u32(tb[IFLA_GROUP]),
3398                                                 ifm, extack, tb);
3399                 return -ENODEV;
3400         }
3401
3402         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3403                 return -EOPNOTSUPP;
3404
3405         if (!ops) {
3406 #ifdef CONFIG_MODULES
3407                 if (kind[0]) {
3408                         __rtnl_unlock();
3409                         request_module("rtnl-link-%s", kind);
3410                         rtnl_lock();
3411                         ops = rtnl_link_ops_get(kind);
3412                         if (ops)
3413                                 goto replay;
3414                 }
3415 #endif
3416                 NL_SET_ERR_MSG(extack, "Unknown device type");
3417                 return -EOPNOTSUPP;
3418         }
3419
3420         if (!ops->alloc && !ops->setup)
3421                 return -EOPNOTSUPP;
3422
3423         if (!ifname[0]) {
3424                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3425                 name_assign_type = NET_NAME_ENUM;
3426         }
3427
3428         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3429         if (IS_ERR(dest_net))
3430                 return PTR_ERR(dest_net);
3431
3432         if (tb[IFLA_LINK_NETNSID]) {
3433                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3434
3435                 link_net = get_net_ns_by_id(dest_net, id);
3436                 if (!link_net) {
3437                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3438                         err =  -EINVAL;
3439                         goto out;
3440                 }
3441                 err = -EPERM;
3442                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3443                         goto out;
3444         } else {
3445                 link_net = NULL;
3446         }
3447
3448         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3449                                name_assign_type, ops, tb, extack);
3450         if (IS_ERR(dev)) {
3451                 err = PTR_ERR(dev);
3452                 goto out;
3453         }
3454
3455         dev->ifindex = ifm->ifi_index;
3456
3457         if (ops->newlink)
3458                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3459         else
3460                 err = register_netdevice(dev);
3461         if (err < 0) {
3462                 free_netdev(dev);
3463                 goto out;
3464         }
3465
3466         err = rtnl_configure_link(dev, ifm);
3467         if (err < 0)
3468                 goto out_unregister;
3469         if (link_net) {
3470                 err = dev_change_net_namespace(dev, dest_net, ifname);
3471                 if (err < 0)
3472                         goto out_unregister;
3473         }
3474         if (tb[IFLA_MASTER]) {
3475                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3476                 if (err)
3477                         goto out_unregister;
3478         }
3479 out:
3480         if (link_net)
3481                 put_net(link_net);
3482         put_net(dest_net);
3483         return err;
3484 out_unregister:
3485         if (ops->newlink) {
3486                 LIST_HEAD(list_kill);
3487
3488                 ops->dellink(dev, &list_kill);
3489                 unregister_netdevice_many(&list_kill);
3490         } else {
3491                 unregister_netdevice(dev);
3492         }
3493         goto out;
3494 }
3495
3496 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3497                         struct netlink_ext_ack *extack)
3498 {
3499         struct nlattr **attr;
3500         int ret;
3501
3502         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3503         if (!attr)
3504                 return -ENOMEM;
3505
3506         ret = __rtnl_newlink(skb, nlh, attr, extack);
3507         kfree(attr);
3508         return ret;
3509 }
3510
3511 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3512                                   const struct nlmsghdr *nlh,
3513                                   struct nlattr **tb,
3514                                   struct netlink_ext_ack *extack)
3515 {
3516         struct ifinfomsg *ifm;
3517         int i, err;
3518
3519         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3520                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3521                 return -EINVAL;
3522         }
3523
3524         if (!netlink_strict_get_check(skb))
3525                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3526                                               ifla_policy, extack);
3527
3528         ifm = nlmsg_data(nlh);
3529         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3530             ifm->ifi_change) {
3531                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3532                 return -EINVAL;
3533         }
3534
3535         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3536                                             ifla_policy, extack);
3537         if (err)
3538                 return err;
3539
3540         for (i = 0; i <= IFLA_MAX; i++) {
3541                 if (!tb[i])
3542                         continue;
3543
3544                 switch (i) {
3545                 case IFLA_IFNAME:
3546                 case IFLA_ALT_IFNAME:
3547                 case IFLA_EXT_MASK:
3548                 case IFLA_TARGET_NETNSID:
3549                         break;
3550                 default:
3551                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3552                         return -EINVAL;
3553                 }
3554         }
3555
3556         return 0;
3557 }
3558
3559 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3560                         struct netlink_ext_ack *extack)
3561 {
3562         struct net *net = sock_net(skb->sk);
3563         struct net *tgt_net = net;
3564         struct ifinfomsg *ifm;
3565         struct nlattr *tb[IFLA_MAX+1];
3566         struct net_device *dev = NULL;
3567         struct sk_buff *nskb;
3568         int netnsid = -1;
3569         int err;
3570         u32 ext_filter_mask = 0;
3571
3572         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3573         if (err < 0)
3574                 return err;
3575
3576         err = rtnl_ensure_unique_netns(tb, extack, true);
3577         if (err < 0)
3578                 return err;
3579
3580         if (tb[IFLA_TARGET_NETNSID]) {
3581                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3582                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3583                 if (IS_ERR(tgt_net))
3584                         return PTR_ERR(tgt_net);
3585         }
3586
3587         if (tb[IFLA_EXT_MASK])
3588                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3589
3590         err = -EINVAL;
3591         ifm = nlmsg_data(nlh);
3592         if (ifm->ifi_index > 0)
3593                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3594         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3595                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3596                                    tb[IFLA_ALT_IFNAME], NULL);
3597         else
3598                 goto out;
3599
3600         err = -ENODEV;
3601         if (dev == NULL)
3602                 goto out;
3603
3604         err = -ENOBUFS;
3605         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3606         if (nskb == NULL)
3607                 goto out;
3608
3609         err = rtnl_fill_ifinfo(nskb, dev, net,
3610                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3611                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3612                                0, NULL, 0, netnsid, GFP_KERNEL);
3613         if (err < 0) {
3614                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3615                 WARN_ON(err == -EMSGSIZE);
3616                 kfree_skb(nskb);
3617         } else
3618                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3619 out:
3620         if (netnsid >= 0)
3621                 put_net(tgt_net);
3622
3623         return err;
3624 }
3625
3626 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3627                            bool *changed, struct netlink_ext_ack *extack)
3628 {
3629         char *alt_ifname;
3630         int err;
3631
3632         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3633         if (err)
3634                 return err;
3635
3636         alt_ifname = nla_strdup(attr, GFP_KERNEL);
3637         if (!alt_ifname)
3638                 return -ENOMEM;
3639
3640         if (cmd == RTM_NEWLINKPROP) {
3641                 err = netdev_name_node_alt_create(dev, alt_ifname);
3642                 if (!err)
3643                         alt_ifname = NULL;
3644         } else if (cmd == RTM_DELLINKPROP) {
3645                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3646         } else {
3647                 WARN_ON_ONCE(1);
3648                 err = -EINVAL;
3649         }
3650
3651         kfree(alt_ifname);
3652         if (!err)
3653                 *changed = true;
3654         return err;
3655 }
3656
3657 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3658                          struct netlink_ext_ack *extack)
3659 {
3660         struct net *net = sock_net(skb->sk);
3661         struct nlattr *tb[IFLA_MAX + 1];
3662         struct net_device *dev;
3663         struct ifinfomsg *ifm;
3664         bool changed = false;
3665         struct nlattr *attr;
3666         int err, rem;
3667
3668         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3669         if (err)
3670                 return err;
3671
3672         err = rtnl_ensure_unique_netns(tb, extack, true);
3673         if (err)
3674                 return err;
3675
3676         ifm = nlmsg_data(nlh);
3677         if (ifm->ifi_index > 0)
3678                 dev = __dev_get_by_index(net, ifm->ifi_index);
3679         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3680                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3681                                    tb[IFLA_ALT_IFNAME], NULL);
3682         else
3683                 return -EINVAL;
3684
3685         if (!dev)
3686                 return -ENODEV;
3687
3688         if (!tb[IFLA_PROP_LIST])
3689                 return 0;
3690
3691         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3692                 switch (nla_type(attr)) {
3693                 case IFLA_ALT_IFNAME:
3694                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3695                         if (err)
3696                                 return err;
3697                         break;
3698                 }
3699         }
3700
3701         if (changed)
3702                 netdev_state_change(dev);
3703         return 0;
3704 }
3705
3706 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3707                             struct netlink_ext_ack *extack)
3708 {
3709         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3710 }
3711
3712 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3713                             struct netlink_ext_ack *extack)
3714 {
3715         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3716 }
3717
3718 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3719 {
3720         struct net *net = sock_net(skb->sk);
3721         size_t min_ifinfo_dump_size = 0;
3722         struct nlattr *tb[IFLA_MAX+1];
3723         u32 ext_filter_mask = 0;
3724         struct net_device *dev;
3725         int hdrlen;
3726
3727         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3728         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3729                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3730
3731         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3732                 if (tb[IFLA_EXT_MASK])
3733                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3734         }
3735
3736         if (!ext_filter_mask)
3737                 return NLMSG_GOODSIZE;
3738         /*
3739          * traverse the list of net devices and compute the minimum
3740          * buffer size based upon the filter mask.
3741          */
3742         rcu_read_lock();
3743         for_each_netdev_rcu(net, dev) {
3744                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3745                                            if_nlmsg_size(dev, ext_filter_mask));
3746         }
3747         rcu_read_unlock();
3748
3749         return nlmsg_total_size(min_ifinfo_dump_size);
3750 }
3751
3752 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3753 {
3754         int idx;
3755         int s_idx = cb->family;
3756         int type = cb->nlh->nlmsg_type - RTM_BASE;
3757         int ret = 0;
3758
3759         if (s_idx == 0)
3760                 s_idx = 1;
3761
3762         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3763                 struct rtnl_link __rcu **tab;
3764                 struct rtnl_link *link;
3765                 rtnl_dumpit_func dumpit;
3766
3767                 if (idx < s_idx || idx == PF_PACKET)
3768                         continue;
3769
3770                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3771                         continue;
3772
3773                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3774                 if (!tab)
3775                         continue;
3776
3777                 link = rcu_dereference_rtnl(tab[type]);
3778                 if (!link)
3779                         continue;
3780
3781                 dumpit = link->dumpit;
3782                 if (!dumpit)
3783                         continue;
3784
3785                 if (idx > s_idx) {
3786                         memset(&cb->args[0], 0, sizeof(cb->args));
3787                         cb->prev_seq = 0;
3788                         cb->seq = 0;
3789                 }
3790                 ret = dumpit(skb, cb);
3791                 if (ret)
3792                         break;
3793         }
3794         cb->family = idx;
3795
3796         return skb->len ? : ret;
3797 }
3798
3799 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3800                                        unsigned int change,
3801                                        u32 event, gfp_t flags, int *new_nsid,
3802                                        int new_ifindex)
3803 {
3804         struct net *net = dev_net(dev);
3805         struct sk_buff *skb;
3806         int err = -ENOBUFS;
3807         size_t if_info_size;
3808
3809         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3810         if (skb == NULL)
3811                 goto errout;
3812
3813         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3814                                type, 0, 0, change, 0, 0, event,
3815                                new_nsid, new_ifindex, -1, flags);
3816         if (err < 0) {
3817                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3818                 WARN_ON(err == -EMSGSIZE);
3819                 kfree_skb(skb);
3820                 goto errout;
3821         }
3822         return skb;
3823 errout:
3824         if (err < 0)
3825                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3826         return NULL;
3827 }
3828
3829 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3830 {
3831         struct net *net = dev_net(dev);
3832
3833         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3834 }
3835
3836 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3837                                unsigned int change, u32 event,
3838                                gfp_t flags, int *new_nsid, int new_ifindex)
3839 {
3840         struct sk_buff *skb;
3841
3842         if (dev->reg_state != NETREG_REGISTERED)
3843                 return;
3844
3845         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3846                                      new_ifindex);
3847         if (skb)
3848                 rtmsg_ifinfo_send(skb, dev, flags);
3849 }
3850
3851 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3852                   gfp_t flags)
3853 {
3854         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3855                            NULL, 0);
3856 }
3857
3858 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3859                          gfp_t flags, int *new_nsid, int new_ifindex)
3860 {
3861         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3862                            new_nsid, new_ifindex);
3863 }
3864
3865 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3866                                    struct net_device *dev,
3867                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3868                                    int type, unsigned int flags,
3869                                    int nlflags, u16 ndm_state)
3870 {
3871         struct nlmsghdr *nlh;
3872         struct ndmsg *ndm;
3873
3874         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3875         if (!nlh)
3876                 return -EMSGSIZE;
3877
3878         ndm = nlmsg_data(nlh);
3879         ndm->ndm_family  = AF_BRIDGE;
3880         ndm->ndm_pad1    = 0;
3881         ndm->ndm_pad2    = 0;
3882         ndm->ndm_flags   = flags;
3883         ndm->ndm_type    = 0;
3884         ndm->ndm_ifindex = dev->ifindex;
3885         ndm->ndm_state   = ndm_state;
3886
3887         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3888                 goto nla_put_failure;
3889         if (vid)
3890                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3891                         goto nla_put_failure;
3892
3893         nlmsg_end(skb, nlh);
3894         return 0;
3895
3896 nla_put_failure:
3897         nlmsg_cancel(skb, nlh);
3898         return -EMSGSIZE;
3899 }
3900
3901 static inline size_t rtnl_fdb_nlmsg_size(void)
3902 {
3903         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3904                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3905                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3906                0;
3907 }
3908
3909 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3910                             u16 ndm_state)
3911 {
3912         struct net *net = dev_net(dev);
3913         struct sk_buff *skb;
3914         int err = -ENOBUFS;
3915
3916         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3917         if (!skb)
3918                 goto errout;
3919
3920         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3921                                       0, 0, type, NTF_SELF, 0, ndm_state);
3922         if (err < 0) {
3923                 kfree_skb(skb);
3924                 goto errout;
3925         }
3926
3927         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3928         return;
3929 errout:
3930         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3931 }
3932
3933 /*
3934  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3935  */
3936 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3937                      struct nlattr *tb[],
3938                      struct net_device *dev,
3939                      const unsigned char *addr, u16 vid,
3940                      u16 flags)
3941 {
3942         int err = -EINVAL;
3943
3944         /* If aging addresses are supported device will need to
3945          * implement its own handler for this.
3946          */
3947         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3948                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
3949                 return err;
3950         }
3951
3952         if (vid) {
3953                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3954                 return err;
3955         }
3956
3957         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3958                 err = dev_uc_add_excl(dev, addr);
3959         else if (is_multicast_ether_addr(addr))
3960                 err = dev_mc_add_excl(dev, addr);
3961
3962         /* Only return duplicate errors if NLM_F_EXCL is set */
3963         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3964                 err = 0;
3965
3966         return err;
3967 }
3968 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3969
3970 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3971                          struct netlink_ext_ack *extack)
3972 {
3973         u16 vid = 0;
3974
3975         if (vlan_attr) {
3976                 if (nla_len(vlan_attr) != sizeof(u16)) {
3977                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3978                         return -EINVAL;
3979                 }
3980
3981                 vid = nla_get_u16(vlan_attr);
3982
3983                 if (!vid || vid >= VLAN_VID_MASK) {
3984                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3985                         return -EINVAL;
3986                 }
3987         }
3988         *p_vid = vid;
3989         return 0;
3990 }
3991
3992 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3993                         struct netlink_ext_ack *extack)
3994 {
3995         struct net *net = sock_net(skb->sk);
3996         struct ndmsg *ndm;
3997         struct nlattr *tb[NDA_MAX+1];
3998         struct net_device *dev;
3999         u8 *addr;
4000         u16 vid;
4001         int err;
4002
4003         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4004                                      extack);
4005         if (err < 0)
4006                 return err;
4007
4008         ndm = nlmsg_data(nlh);
4009         if (ndm->ndm_ifindex == 0) {
4010                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4011                 return -EINVAL;
4012         }
4013
4014         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4015         if (dev == NULL) {
4016                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4017                 return -ENODEV;
4018         }
4019
4020         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4021                 NL_SET_ERR_MSG(extack, "invalid address");
4022                 return -EINVAL;
4023         }
4024
4025         if (dev->type != ARPHRD_ETHER) {
4026                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4027                 return -EINVAL;
4028         }
4029
4030         addr = nla_data(tb[NDA_LLADDR]);
4031
4032         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4033         if (err)
4034                 return err;
4035
4036         err = -EOPNOTSUPP;
4037
4038         /* Support fdb on master device the net/bridge default case */
4039         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4040             netif_is_bridge_port(dev)) {
4041                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4042                 const struct net_device_ops *ops = br_dev->netdev_ops;
4043
4044                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4045                                        nlh->nlmsg_flags, extack);
4046                 if (err)
4047                         goto out;
4048                 else
4049                         ndm->ndm_flags &= ~NTF_MASTER;
4050         }
4051
4052         /* Embedded bridge, macvlan, and any other device support */
4053         if ((ndm->ndm_flags & NTF_SELF)) {
4054                 if (dev->netdev_ops->ndo_fdb_add)
4055                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4056                                                            vid,
4057                                                            nlh->nlmsg_flags,
4058                                                            extack);
4059                 else
4060                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4061                                                nlh->nlmsg_flags);
4062
4063                 if (!err) {
4064                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4065                                         ndm->ndm_state);
4066                         ndm->ndm_flags &= ~NTF_SELF;
4067                 }
4068         }
4069 out:
4070         return err;
4071 }
4072
4073 /*
4074  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4075  */
4076 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4077                      struct nlattr *tb[],
4078                      struct net_device *dev,
4079                      const unsigned char *addr, u16 vid)
4080 {
4081         int err = -EINVAL;
4082
4083         /* If aging addresses are supported device will need to
4084          * implement its own handler for this.
4085          */
4086         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4087                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4088                 return err;
4089         }
4090
4091         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4092                 err = dev_uc_del(dev, addr);
4093         else if (is_multicast_ether_addr(addr))
4094                 err = dev_mc_del(dev, addr);
4095
4096         return err;
4097 }
4098 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4099
4100 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4101                         struct netlink_ext_ack *extack)
4102 {
4103         struct net *net = sock_net(skb->sk);
4104         struct ndmsg *ndm;
4105         struct nlattr *tb[NDA_MAX+1];
4106         struct net_device *dev;
4107         __u8 *addr;
4108         int err;
4109         u16 vid;
4110
4111         if (!netlink_capable(skb, CAP_NET_ADMIN))
4112                 return -EPERM;
4113
4114         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4115                                      extack);
4116         if (err < 0)
4117                 return err;
4118
4119         ndm = nlmsg_data(nlh);
4120         if (ndm->ndm_ifindex == 0) {
4121                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4122                 return -EINVAL;
4123         }
4124
4125         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4126         if (dev == NULL) {
4127                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4128                 return -ENODEV;
4129         }
4130
4131         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4132                 NL_SET_ERR_MSG(extack, "invalid address");
4133                 return -EINVAL;
4134         }
4135
4136         if (dev->type != ARPHRD_ETHER) {
4137                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4138                 return -EINVAL;
4139         }
4140
4141         addr = nla_data(tb[NDA_LLADDR]);
4142
4143         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4144         if (err)
4145                 return err;
4146
4147         err = -EOPNOTSUPP;
4148
4149         /* Support fdb on master device the net/bridge default case */
4150         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4151             netif_is_bridge_port(dev)) {
4152                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4153                 const struct net_device_ops *ops = br_dev->netdev_ops;
4154
4155                 if (ops->ndo_fdb_del)
4156                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4157
4158                 if (err)
4159                         goto out;
4160                 else
4161                         ndm->ndm_flags &= ~NTF_MASTER;
4162         }
4163
4164         /* Embedded bridge, macvlan, and any other device support */
4165         if (ndm->ndm_flags & NTF_SELF) {
4166                 if (dev->netdev_ops->ndo_fdb_del)
4167                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4168                                                            vid);
4169                 else
4170                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4171
4172                 if (!err) {
4173                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4174                                         ndm->ndm_state);
4175                         ndm->ndm_flags &= ~NTF_SELF;
4176                 }
4177         }
4178 out:
4179         return err;
4180 }
4181
4182 static int nlmsg_populate_fdb(struct sk_buff *skb,
4183                               struct netlink_callback *cb,
4184                               struct net_device *dev,
4185                               int *idx,
4186                               struct netdev_hw_addr_list *list)
4187 {
4188         struct netdev_hw_addr *ha;
4189         int err;
4190         u32 portid, seq;
4191
4192         portid = NETLINK_CB(cb->skb).portid;
4193         seq = cb->nlh->nlmsg_seq;
4194
4195         list_for_each_entry(ha, &list->list, list) {
4196                 if (*idx < cb->args[2])
4197                         goto skip;
4198
4199                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4200                                               portid, seq,
4201                                               RTM_NEWNEIGH, NTF_SELF,
4202                                               NLM_F_MULTI, NUD_PERMANENT);
4203                 if (err < 0)
4204                         return err;
4205 skip:
4206                 *idx += 1;
4207         }
4208         return 0;
4209 }
4210
4211 /**
4212  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4213  * @skb: socket buffer to store message in
4214  * @cb: netlink callback
4215  * @dev: netdevice
4216  * @filter_dev: ignored
4217  * @idx: the number of FDB table entries dumped is added to *@idx
4218  *
4219  * Default netdevice operation to dump the existing unicast address list.
4220  * Returns number of addresses from list put in skb.
4221  */
4222 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4223                       struct netlink_callback *cb,
4224                       struct net_device *dev,
4225                       struct net_device *filter_dev,
4226                       int *idx)
4227 {
4228         int err;
4229
4230         if (dev->type != ARPHRD_ETHER)
4231                 return -EINVAL;
4232
4233         netif_addr_lock_bh(dev);
4234         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4235         if (err)
4236                 goto out;
4237         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4238 out:
4239         netif_addr_unlock_bh(dev);
4240         return err;
4241 }
4242 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4243
4244 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4245                                  int *br_idx, int *brport_idx,
4246                                  struct netlink_ext_ack *extack)
4247 {
4248         struct nlattr *tb[NDA_MAX + 1];
4249         struct ndmsg *ndm;
4250         int err, i;
4251
4252         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4253                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4254                 return -EINVAL;
4255         }
4256
4257         ndm = nlmsg_data(nlh);
4258         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4259             ndm->ndm_flags || ndm->ndm_type) {
4260                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4261                 return -EINVAL;
4262         }
4263
4264         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4265                                             NDA_MAX, NULL, extack);
4266         if (err < 0)
4267                 return err;
4268
4269         *brport_idx = ndm->ndm_ifindex;
4270         for (i = 0; i <= NDA_MAX; ++i) {
4271                 if (!tb[i])
4272                         continue;
4273
4274                 switch (i) {
4275                 case NDA_IFINDEX:
4276                         if (nla_len(tb[i]) != sizeof(u32)) {
4277                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4278                                 return -EINVAL;
4279                         }
4280                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4281                         break;
4282                 case NDA_MASTER:
4283                         if (nla_len(tb[i]) != sizeof(u32)) {
4284                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4285                                 return -EINVAL;
4286                         }
4287                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4288                         break;
4289                 default:
4290                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4291                         return -EINVAL;
4292                 }
4293         }
4294
4295         return 0;
4296 }
4297
4298 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4299                                  int *br_idx, int *brport_idx,
4300                                  struct netlink_ext_ack *extack)
4301 {
4302         struct nlattr *tb[IFLA_MAX+1];
4303         int err;
4304
4305         /* A hack to preserve kernel<->userspace interface.
4306          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4307          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4308          * So, check for ndmsg with an optional u32 attribute (not used here).
4309          * Fortunately these sizes don't conflict with the size of ifinfomsg
4310          * with an optional attribute.
4311          */
4312         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4313             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4314              nla_attr_size(sizeof(u32)))) {
4315                 struct ifinfomsg *ifm;
4316
4317                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4318                                              tb, IFLA_MAX, ifla_policy,
4319                                              extack);
4320                 if (err < 0) {
4321                         return -EINVAL;
4322                 } else if (err == 0) {
4323                         if (tb[IFLA_MASTER])
4324                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4325                 }
4326
4327                 ifm = nlmsg_data(nlh);
4328                 *brport_idx = ifm->ifi_index;
4329         }
4330         return 0;
4331 }
4332
4333 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4334 {
4335         struct net_device *dev;
4336         struct net_device *br_dev = NULL;
4337         const struct net_device_ops *ops = NULL;
4338         const struct net_device_ops *cops = NULL;
4339         struct net *net = sock_net(skb->sk);
4340         struct hlist_head *head;
4341         int brport_idx = 0;
4342         int br_idx = 0;
4343         int h, s_h;
4344         int idx = 0, s_idx;
4345         int err = 0;
4346         int fidx = 0;
4347
4348         if (cb->strict_check)
4349                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4350                                             cb->extack);
4351         else
4352                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4353                                             cb->extack);
4354         if (err < 0)
4355                 return err;
4356
4357         if (br_idx) {
4358                 br_dev = __dev_get_by_index(net, br_idx);
4359                 if (!br_dev)
4360                         return -ENODEV;
4361
4362                 ops = br_dev->netdev_ops;
4363         }
4364
4365         s_h = cb->args[0];
4366         s_idx = cb->args[1];
4367
4368         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4369                 idx = 0;
4370                 head = &net->dev_index_head[h];
4371                 hlist_for_each_entry(dev, head, index_hlist) {
4372
4373                         if (brport_idx && (dev->ifindex != brport_idx))
4374                                 continue;
4375
4376                         if (!br_idx) { /* user did not specify a specific bridge */
4377                                 if (netif_is_bridge_port(dev)) {
4378                                         br_dev = netdev_master_upper_dev_get(dev);
4379                                         cops = br_dev->netdev_ops;
4380                                 }
4381                         } else {
4382                                 if (dev != br_dev &&
4383                                     !netif_is_bridge_port(dev))
4384                                         continue;
4385
4386                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4387                                     !(dev->priv_flags & IFF_EBRIDGE))
4388                                         continue;
4389                                 cops = ops;
4390                         }
4391
4392                         if (idx < s_idx)
4393                                 goto cont;
4394
4395                         if (netif_is_bridge_port(dev)) {
4396                                 if (cops && cops->ndo_fdb_dump) {
4397                                         err = cops->ndo_fdb_dump(skb, cb,
4398                                                                 br_dev, dev,
4399                                                                 &fidx);
4400                                         if (err == -EMSGSIZE)
4401                                                 goto out;
4402                                 }
4403                         }
4404
4405                         if (dev->netdev_ops->ndo_fdb_dump)
4406                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4407                                                                     dev, NULL,
4408                                                                     &fidx);
4409                         else
4410                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4411                                                         &fidx);
4412                         if (err == -EMSGSIZE)
4413                                 goto out;
4414
4415                         cops = NULL;
4416
4417                         /* reset fdb offset to 0 for rest of the interfaces */
4418                         cb->args[2] = 0;
4419                         fidx = 0;
4420 cont:
4421                         idx++;
4422                 }
4423         }
4424
4425 out:
4426         cb->args[0] = h;
4427         cb->args[1] = idx;
4428         cb->args[2] = fidx;
4429
4430         return skb->len;
4431 }
4432
4433 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4434                                 struct nlattr **tb, u8 *ndm_flags,
4435                                 int *br_idx, int *brport_idx, u8 **addr,
4436                                 u16 *vid, struct netlink_ext_ack *extack)
4437 {
4438         struct ndmsg *ndm;
4439         int err, i;
4440
4441         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4442                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4443                 return -EINVAL;
4444         }
4445
4446         ndm = nlmsg_data(nlh);
4447         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4448             ndm->ndm_type) {
4449                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4450                 return -EINVAL;
4451         }
4452
4453         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4454                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4455                 return -EINVAL;
4456         }
4457
4458         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4459                                             NDA_MAX, nda_policy, extack);
4460         if (err < 0)
4461                 return err;
4462
4463         *ndm_flags = ndm->ndm_flags;
4464         *brport_idx = ndm->ndm_ifindex;
4465         for (i = 0; i <= NDA_MAX; ++i) {
4466                 if (!tb[i])
4467                         continue;
4468
4469                 switch (i) {
4470                 case NDA_MASTER:
4471                         *br_idx = nla_get_u32(tb[i]);
4472                         break;
4473                 case NDA_LLADDR:
4474                         if (nla_len(tb[i]) != ETH_ALEN) {
4475                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4476                                 return -EINVAL;
4477                         }
4478                         *addr = nla_data(tb[i]);
4479                         break;
4480                 case NDA_VLAN:
4481                         err = fdb_vid_parse(tb[i], vid, extack);
4482                         if (err)
4483                                 return err;
4484                         break;
4485                 case NDA_VNI:
4486                         break;
4487                 default:
4488                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4489                         return -EINVAL;
4490                 }
4491         }
4492
4493         return 0;
4494 }
4495
4496 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4497                         struct netlink_ext_ack *extack)
4498 {
4499         struct net_device *dev = NULL, *br_dev = NULL;
4500         const struct net_device_ops *ops = NULL;
4501         struct net *net = sock_net(in_skb->sk);
4502         struct nlattr *tb[NDA_MAX + 1];
4503         struct sk_buff *skb;
4504         int brport_idx = 0;
4505         u8 ndm_flags = 0;
4506         int br_idx = 0;
4507         u8 *addr = NULL;
4508         u16 vid = 0;
4509         int err;
4510
4511         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4512                                    &brport_idx, &addr, &vid, extack);
4513         if (err < 0)
4514                 return err;
4515
4516         if (!addr) {
4517                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4518                 return -EINVAL;
4519         }
4520
4521         if (brport_idx) {
4522                 dev = __dev_get_by_index(net, brport_idx);
4523                 if (!dev) {
4524                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4525                         return -ENODEV;
4526                 }
4527         }
4528
4529         if (br_idx) {
4530                 if (dev) {
4531                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4532                         return -EINVAL;
4533                 }
4534
4535                 br_dev = __dev_get_by_index(net, br_idx);
4536                 if (!br_dev) {
4537                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4538                         return -EINVAL;
4539                 }
4540                 ops = br_dev->netdev_ops;
4541         }
4542
4543         if (dev) {
4544                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4545                         if (!netif_is_bridge_port(dev)) {
4546                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4547                                 return -EINVAL;
4548                         }
4549                         br_dev = netdev_master_upper_dev_get(dev);
4550                         if (!br_dev) {
4551                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4552                                 return -EINVAL;
4553                         }
4554                         ops = br_dev->netdev_ops;
4555                 } else {
4556                         if (!(ndm_flags & NTF_SELF)) {
4557                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4558                                 return -EINVAL;
4559                         }
4560                         ops = dev->netdev_ops;
4561                 }
4562         }
4563
4564         if (!br_dev && !dev) {
4565                 NL_SET_ERR_MSG(extack, "No device specified");
4566                 return -ENODEV;
4567         }
4568
4569         if (!ops || !ops->ndo_fdb_get) {
4570                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4571                 return -EOPNOTSUPP;
4572         }
4573
4574         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4575         if (!skb)
4576                 return -ENOBUFS;
4577
4578         if (br_dev)
4579                 dev = br_dev;
4580         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4581                                NETLINK_CB(in_skb).portid,
4582                                nlh->nlmsg_seq, extack);
4583         if (err)
4584                 goto out;
4585
4586         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4587 out:
4588         kfree_skb(skb);
4589         return err;
4590 }
4591
4592 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4593                                unsigned int attrnum, unsigned int flag)
4594 {
4595         if (mask & flag)
4596                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4597         return 0;
4598 }
4599
4600 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4601                             struct net_device *dev, u16 mode,
4602                             u32 flags, u32 mask, int nlflags,
4603                             u32 filter_mask,
4604                             int (*vlan_fill)(struct sk_buff *skb,
4605                                              struct net_device *dev,
4606                                              u32 filter_mask))
4607 {
4608         struct nlmsghdr *nlh;
4609         struct ifinfomsg *ifm;
4610         struct nlattr *br_afspec;
4611         struct nlattr *protinfo;
4612         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4613         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4614         int err = 0;
4615
4616         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4617         if (nlh == NULL)
4618                 return -EMSGSIZE;
4619
4620         ifm = nlmsg_data(nlh);
4621         ifm->ifi_family = AF_BRIDGE;
4622         ifm->__ifi_pad = 0;
4623         ifm->ifi_type = dev->type;
4624         ifm->ifi_index = dev->ifindex;
4625         ifm->ifi_flags = dev_get_flags(dev);
4626         ifm->ifi_change = 0;
4627
4628
4629         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4630             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4631             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4632             (br_dev &&
4633              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4634             (dev->addr_len &&
4635              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4636             (dev->ifindex != dev_get_iflink(dev) &&
4637              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4638                 goto nla_put_failure;
4639
4640         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4641         if (!br_afspec)
4642                 goto nla_put_failure;
4643
4644         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4645                 nla_nest_cancel(skb, br_afspec);
4646                 goto nla_put_failure;
4647         }
4648
4649         if (mode != BRIDGE_MODE_UNDEF) {
4650                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4651                         nla_nest_cancel(skb, br_afspec);
4652                         goto nla_put_failure;
4653                 }
4654         }
4655         if (vlan_fill) {
4656                 err = vlan_fill(skb, dev, filter_mask);
4657                 if (err) {
4658                         nla_nest_cancel(skb, br_afspec);
4659                         goto nla_put_failure;
4660                 }
4661         }
4662         nla_nest_end(skb, br_afspec);
4663
4664         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4665         if (!protinfo)
4666                 goto nla_put_failure;
4667
4668         if (brport_nla_put_flag(skb, flags, mask,
4669                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4670             brport_nla_put_flag(skb, flags, mask,
4671                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4672             brport_nla_put_flag(skb, flags, mask,
4673                                 IFLA_BRPORT_FAST_LEAVE,
4674                                 BR_MULTICAST_FAST_LEAVE) ||
4675             brport_nla_put_flag(skb, flags, mask,
4676                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4677             brport_nla_put_flag(skb, flags, mask,
4678                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4679             brport_nla_put_flag(skb, flags, mask,
4680                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4681             brport_nla_put_flag(skb, flags, mask,
4682                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4683             brport_nla_put_flag(skb, flags, mask,
4684                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4685             brport_nla_put_flag(skb, flags, mask,
4686                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4687             brport_nla_put_flag(skb, flags, mask,
4688                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4689                 nla_nest_cancel(skb, protinfo);
4690                 goto nla_put_failure;
4691         }
4692
4693         nla_nest_end(skb, protinfo);
4694
4695         nlmsg_end(skb, nlh);
4696         return 0;
4697 nla_put_failure:
4698         nlmsg_cancel(skb, nlh);
4699         return err ? err : -EMSGSIZE;
4700 }
4701 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4702
4703 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4704                                     bool strict_check, u32 *filter_mask,
4705                                     struct netlink_ext_ack *extack)
4706 {
4707         struct nlattr *tb[IFLA_MAX+1];
4708         int err, i;
4709
4710         if (strict_check) {
4711                 struct ifinfomsg *ifm;
4712
4713                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4714                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4715                         return -EINVAL;
4716                 }
4717
4718                 ifm = nlmsg_data(nlh);
4719                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4720                     ifm->ifi_change || ifm->ifi_index) {
4721                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4722                         return -EINVAL;
4723                 }
4724
4725                 err = nlmsg_parse_deprecated_strict(nlh,
4726                                                     sizeof(struct ifinfomsg),
4727                                                     tb, IFLA_MAX, ifla_policy,
4728                                                     extack);
4729         } else {
4730                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4731                                              tb, IFLA_MAX, ifla_policy,
4732                                              extack);
4733         }
4734         if (err < 0)
4735                 return err;
4736
4737         /* new attributes should only be added with strict checking */
4738         for (i = 0; i <= IFLA_MAX; ++i) {
4739                 if (!tb[i])
4740                         continue;
4741
4742                 switch (i) {
4743                 case IFLA_EXT_MASK:
4744                         *filter_mask = nla_get_u32(tb[i]);
4745                         break;
4746                 default:
4747                         if (strict_check) {
4748                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4749                                 return -EINVAL;
4750                         }
4751                 }
4752         }
4753
4754         return 0;
4755 }
4756
4757 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4758 {
4759         const struct nlmsghdr *nlh = cb->nlh;
4760         struct net *net = sock_net(skb->sk);
4761         struct net_device *dev;
4762         int idx = 0;
4763         u32 portid = NETLINK_CB(cb->skb).portid;
4764         u32 seq = nlh->nlmsg_seq;
4765         u32 filter_mask = 0;
4766         int err;
4767
4768         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4769                                        cb->extack);
4770         if (err < 0 && cb->strict_check)
4771                 return err;
4772
4773         rcu_read_lock();
4774         for_each_netdev_rcu(net, dev) {
4775                 const struct net_device_ops *ops = dev->netdev_ops;
4776                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4777
4778                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4779                         if (idx >= cb->args[0]) {
4780                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4781                                                 skb, portid, seq, dev,
4782                                                 filter_mask, NLM_F_MULTI);
4783                                 if (err < 0 && err != -EOPNOTSUPP) {
4784                                         if (likely(skb->len))
4785                                                 break;
4786
4787                                         goto out_err;
4788                                 }
4789                         }
4790                         idx++;
4791                 }
4792
4793                 if (ops->ndo_bridge_getlink) {
4794                         if (idx >= cb->args[0]) {
4795                                 err = ops->ndo_bridge_getlink(skb, portid,
4796                                                               seq, dev,
4797                                                               filter_mask,
4798                                                               NLM_F_MULTI);
4799                                 if (err < 0 && err != -EOPNOTSUPP) {
4800                                         if (likely(skb->len))
4801                                                 break;
4802
4803                                         goto out_err;
4804                                 }
4805                         }
4806                         idx++;
4807                 }
4808         }
4809         err = skb->len;
4810 out_err:
4811         rcu_read_unlock();
4812         cb->args[0] = idx;
4813
4814         return err;
4815 }
4816
4817 static inline size_t bridge_nlmsg_size(void)
4818 {
4819         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4820                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4821                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4822                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4823                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4824                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4825                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4826                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4827                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4828                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4829                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4830 }
4831
4832 static int rtnl_bridge_notify(struct net_device *dev)
4833 {
4834         struct net *net = dev_net(dev);
4835         struct sk_buff *skb;
4836         int err = -EOPNOTSUPP;
4837
4838         if (!dev->netdev_ops->ndo_bridge_getlink)
4839                 return 0;
4840
4841         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4842         if (!skb) {
4843                 err = -ENOMEM;
4844                 goto errout;
4845         }
4846
4847         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4848         if (err < 0)
4849                 goto errout;
4850
4851         /* Notification info is only filled for bridge ports, not the bridge
4852          * device itself. Therefore, a zero notification length is valid and
4853          * should not result in an error.
4854          */
4855         if (!skb->len)
4856                 goto errout;
4857
4858         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4859         return 0;
4860 errout:
4861         WARN_ON(err == -EMSGSIZE);
4862         kfree_skb(skb);
4863         if (err)
4864                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4865         return err;
4866 }
4867
4868 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4869                                struct netlink_ext_ack *extack)
4870 {
4871         struct net *net = sock_net(skb->sk);
4872         struct ifinfomsg *ifm;
4873         struct net_device *dev;
4874         struct nlattr *br_spec, *attr = NULL;
4875         int rem, err = -EOPNOTSUPP;
4876         u16 flags = 0;
4877         bool have_flags = false;
4878
4879         if (nlmsg_len(nlh) < sizeof(*ifm))
4880                 return -EINVAL;
4881
4882         ifm = nlmsg_data(nlh);
4883         if (ifm->ifi_family != AF_BRIDGE)
4884                 return -EPFNOSUPPORT;
4885
4886         dev = __dev_get_by_index(net, ifm->ifi_index);
4887         if (!dev) {
4888                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4889                 return -ENODEV;
4890         }
4891
4892         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4893         if (br_spec) {
4894                 nla_for_each_nested(attr, br_spec, rem) {
4895                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4896                                 if (nla_len(attr) < sizeof(flags))
4897                                         return -EINVAL;
4898
4899                                 have_flags = true;
4900                                 flags = nla_get_u16(attr);
4901                                 break;
4902                         }
4903                 }
4904         }
4905
4906         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4907                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4908
4909                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4910                         err = -EOPNOTSUPP;
4911                         goto out;
4912                 }
4913
4914                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4915                                                              extack);
4916                 if (err)
4917                         goto out;
4918
4919                 flags &= ~BRIDGE_FLAGS_MASTER;
4920         }
4921
4922         if ((flags & BRIDGE_FLAGS_SELF)) {
4923                 if (!dev->netdev_ops->ndo_bridge_setlink)
4924                         err = -EOPNOTSUPP;
4925                 else
4926                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4927                                                                   flags,
4928                                                                   extack);
4929                 if (!err) {
4930                         flags &= ~BRIDGE_FLAGS_SELF;
4931
4932                         /* Generate event to notify upper layer of bridge
4933                          * change
4934                          */
4935                         err = rtnl_bridge_notify(dev);
4936                 }
4937         }
4938
4939         if (have_flags)
4940                 memcpy(nla_data(attr), &flags, sizeof(flags));
4941 out:
4942         return err;
4943 }
4944
4945 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4946                                struct netlink_ext_ack *extack)
4947 {
4948         struct net *net = sock_net(skb->sk);
4949         struct ifinfomsg *ifm;
4950         struct net_device *dev;
4951         struct nlattr *br_spec, *attr = NULL;
4952         int rem, err = -EOPNOTSUPP;
4953         u16 flags = 0;
4954         bool have_flags = false;
4955
4956         if (nlmsg_len(nlh) < sizeof(*ifm))
4957                 return -EINVAL;
4958
4959         ifm = nlmsg_data(nlh);
4960         if (ifm->ifi_family != AF_BRIDGE)
4961                 return -EPFNOSUPPORT;
4962
4963         dev = __dev_get_by_index(net, ifm->ifi_index);
4964         if (!dev) {
4965                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4966                 return -ENODEV;
4967         }
4968
4969         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4970         if (br_spec) {
4971                 nla_for_each_nested(attr, br_spec, rem) {
4972                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4973                                 if (nla_len(attr) < sizeof(flags))
4974                                         return -EINVAL;
4975
4976                                 have_flags = true;
4977                                 flags = nla_get_u16(attr);
4978                                 break;
4979                         }
4980                 }
4981         }
4982
4983         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4984                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4985
4986                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4987                         err = -EOPNOTSUPP;
4988                         goto out;
4989                 }
4990
4991                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4992                 if (err)
4993                         goto out;
4994
4995                 flags &= ~BRIDGE_FLAGS_MASTER;
4996         }
4997
4998         if ((flags & BRIDGE_FLAGS_SELF)) {
4999                 if (!dev->netdev_ops->ndo_bridge_dellink)
5000                         err = -EOPNOTSUPP;
5001                 else
5002                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5003                                                                   flags);
5004
5005                 if (!err) {
5006                         flags &= ~BRIDGE_FLAGS_SELF;
5007
5008                         /* Generate event to notify upper layer of bridge
5009                          * change
5010                          */
5011                         err = rtnl_bridge_notify(dev);
5012                 }
5013         }
5014
5015         if (have_flags)
5016                 memcpy(nla_data(attr), &flags, sizeof(flags));
5017 out:
5018         return err;
5019 }
5020
5021 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5022 {
5023         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5024                (!idxattr || idxattr == attrid);
5025 }
5026
5027 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5028 static int rtnl_get_offload_stats_attr_size(int attr_id)
5029 {
5030         switch (attr_id) {
5031         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5032                 return sizeof(struct rtnl_link_stats64);
5033         }
5034
5035         return 0;
5036 }
5037
5038 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5039                                   int *prividx)
5040 {
5041         struct nlattr *attr = NULL;
5042         int attr_id, size;
5043         void *attr_data;
5044         int err;
5045
5046         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5047               dev->netdev_ops->ndo_get_offload_stats))
5048                 return -ENODATA;
5049
5050         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5051              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5052                 if (attr_id < *prividx)
5053                         continue;
5054
5055                 size = rtnl_get_offload_stats_attr_size(attr_id);
5056                 if (!size)
5057                         continue;
5058
5059                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5060                         continue;
5061
5062                 attr = nla_reserve_64bit(skb, attr_id, size,
5063                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5064                 if (!attr)
5065                         goto nla_put_failure;
5066
5067                 attr_data = nla_data(attr);
5068                 memset(attr_data, 0, size);
5069                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5070                                                              attr_data);
5071                 if (err)
5072                         goto get_offload_stats_failure;
5073         }
5074
5075         if (!attr)
5076                 return -ENODATA;
5077
5078         *prividx = 0;
5079         return 0;
5080
5081 nla_put_failure:
5082         err = -EMSGSIZE;
5083 get_offload_stats_failure:
5084         *prividx = attr_id;
5085         return err;
5086 }
5087
5088 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5089 {
5090         int nla_size = 0;
5091         int attr_id;
5092         int size;
5093
5094         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5095               dev->netdev_ops->ndo_get_offload_stats))
5096                 return 0;
5097
5098         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5099              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5100                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5101                         continue;
5102                 size = rtnl_get_offload_stats_attr_size(attr_id);
5103                 nla_size += nla_total_size_64bit(size);
5104         }
5105
5106         if (nla_size != 0)
5107                 nla_size += nla_total_size(0);
5108
5109         return nla_size;
5110 }
5111
5112 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5113                                int type, u32 pid, u32 seq, u32 change,
5114                                unsigned int flags, unsigned int filter_mask,
5115                                int *idxattr, int *prividx)
5116 {
5117         struct if_stats_msg *ifsm;
5118         struct nlmsghdr *nlh;
5119         struct nlattr *attr;
5120         int s_prividx = *prividx;
5121         int err;
5122
5123         ASSERT_RTNL();
5124
5125         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5126         if (!nlh)
5127                 return -EMSGSIZE;
5128
5129         ifsm = nlmsg_data(nlh);
5130         ifsm->family = PF_UNSPEC;
5131         ifsm->pad1 = 0;
5132         ifsm->pad2 = 0;
5133         ifsm->ifindex = dev->ifindex;
5134         ifsm->filter_mask = filter_mask;
5135
5136         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5137                 struct rtnl_link_stats64 *sp;
5138
5139                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5140                                          sizeof(struct rtnl_link_stats64),
5141                                          IFLA_STATS_UNSPEC);
5142                 if (!attr)
5143                         goto nla_put_failure;
5144
5145                 sp = nla_data(attr);
5146                 dev_get_stats(dev, sp);
5147         }
5148
5149         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5150                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5151
5152                 if (ops && ops->fill_linkxstats) {
5153                         *idxattr = IFLA_STATS_LINK_XSTATS;
5154                         attr = nla_nest_start_noflag(skb,
5155                                                      IFLA_STATS_LINK_XSTATS);
5156                         if (!attr)
5157                                 goto nla_put_failure;
5158
5159                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5160                         nla_nest_end(skb, attr);
5161                         if (err)
5162                                 goto nla_put_failure;
5163                         *idxattr = 0;
5164                 }
5165         }
5166
5167         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5168                              *idxattr)) {
5169                 const struct rtnl_link_ops *ops = NULL;
5170                 const struct net_device *master;
5171
5172                 master = netdev_master_upper_dev_get(dev);
5173                 if (master)
5174                         ops = master->rtnl_link_ops;
5175                 if (ops && ops->fill_linkxstats) {
5176                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5177                         attr = nla_nest_start_noflag(skb,
5178                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5179                         if (!attr)
5180                                 goto nla_put_failure;
5181
5182                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5183                         nla_nest_end(skb, attr);
5184                         if (err)
5185                                 goto nla_put_failure;
5186                         *idxattr = 0;
5187                 }
5188         }
5189
5190         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5191                              *idxattr)) {
5192                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5193                 attr = nla_nest_start_noflag(skb,
5194                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5195                 if (!attr)
5196                         goto nla_put_failure;
5197
5198                 err = rtnl_get_offload_stats(skb, dev, prividx);
5199                 if (err == -ENODATA)
5200                         nla_nest_cancel(skb, attr);
5201                 else
5202                         nla_nest_end(skb, attr);
5203
5204                 if (err && err != -ENODATA)
5205                         goto nla_put_failure;
5206                 *idxattr = 0;
5207         }
5208
5209         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5210                 struct rtnl_af_ops *af_ops;
5211
5212                 *idxattr = IFLA_STATS_AF_SPEC;
5213                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5214                 if (!attr)
5215                         goto nla_put_failure;
5216
5217                 rcu_read_lock();
5218                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5219                         if (af_ops->fill_stats_af) {
5220                                 struct nlattr *af;
5221                                 int err;
5222
5223                                 af = nla_nest_start_noflag(skb,
5224                                                            af_ops->family);
5225                                 if (!af) {
5226                                         rcu_read_unlock();
5227                                         goto nla_put_failure;
5228                                 }
5229                                 err = af_ops->fill_stats_af(skb, dev);
5230
5231                                 if (err == -ENODATA) {
5232                                         nla_nest_cancel(skb, af);
5233                                 } else if (err < 0) {
5234                                         rcu_read_unlock();
5235                                         goto nla_put_failure;
5236                                 }
5237
5238                                 nla_nest_end(skb, af);
5239                         }
5240                 }
5241                 rcu_read_unlock();
5242
5243                 nla_nest_end(skb, attr);
5244
5245                 *idxattr = 0;
5246         }
5247
5248         nlmsg_end(skb, nlh);
5249
5250         return 0;
5251
5252 nla_put_failure:
5253         /* not a multi message or no progress mean a real error */
5254         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5255                 nlmsg_cancel(skb, nlh);
5256         else
5257                 nlmsg_end(skb, nlh);
5258
5259         return -EMSGSIZE;
5260 }
5261
5262 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5263                                   u32 filter_mask)
5264 {
5265         size_t size = 0;
5266
5267         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5268                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5269
5270         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5271                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5272                 int attr = IFLA_STATS_LINK_XSTATS;
5273
5274                 if (ops && ops->get_linkxstats_size) {
5275                         size += nla_total_size(ops->get_linkxstats_size(dev,
5276                                                                         attr));
5277                         /* for IFLA_STATS_LINK_XSTATS */
5278                         size += nla_total_size(0);
5279                 }
5280         }
5281
5282         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5283                 struct net_device *_dev = (struct net_device *)dev;
5284                 const struct rtnl_link_ops *ops = NULL;
5285                 const struct net_device *master;
5286
5287                 /* netdev_master_upper_dev_get can't take const */
5288                 master = netdev_master_upper_dev_get(_dev);
5289                 if (master)
5290                         ops = master->rtnl_link_ops;
5291                 if (ops && ops->get_linkxstats_size) {
5292                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5293
5294                         size += nla_total_size(ops->get_linkxstats_size(dev,
5295                                                                         attr));
5296                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5297                         size += nla_total_size(0);
5298                 }
5299         }
5300
5301         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5302                 size += rtnl_get_offload_stats_size(dev);
5303
5304         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5305                 struct rtnl_af_ops *af_ops;
5306
5307                 /* for IFLA_STATS_AF_SPEC */
5308                 size += nla_total_size(0);
5309
5310                 rcu_read_lock();
5311                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5312                         if (af_ops->get_stats_af_size) {
5313                                 size += nla_total_size(
5314                                         af_ops->get_stats_af_size(dev));
5315
5316                                 /* for AF_* */
5317                                 size += nla_total_size(0);
5318                         }
5319                 }
5320                 rcu_read_unlock();
5321         }
5322
5323         return size;
5324 }
5325
5326 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5327                                 bool is_dump, struct netlink_ext_ack *extack)
5328 {
5329         struct if_stats_msg *ifsm;
5330
5331         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5332                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5333                 return -EINVAL;
5334         }
5335
5336         if (!strict_check)
5337                 return 0;
5338
5339         ifsm = nlmsg_data(nlh);
5340
5341         /* only requests using strict checks can pass data to influence
5342          * the dump. The legacy exception is filter_mask.
5343          */
5344         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5345                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5346                 return -EINVAL;
5347         }
5348         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5349                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5350                 return -EINVAL;
5351         }
5352         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5353                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5354                 return -EINVAL;
5355         }
5356
5357         return 0;
5358 }
5359
5360 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5361                           struct netlink_ext_ack *extack)
5362 {
5363         struct net *net = sock_net(skb->sk);
5364         struct net_device *dev = NULL;
5365         int idxattr = 0, prividx = 0;
5366         struct if_stats_msg *ifsm;
5367         struct sk_buff *nskb;
5368         u32 filter_mask;
5369         int err;
5370
5371         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5372                                    false, extack);
5373         if (err)
5374                 return err;
5375
5376         ifsm = nlmsg_data(nlh);
5377         if (ifsm->ifindex > 0)
5378                 dev = __dev_get_by_index(net, ifsm->ifindex);
5379         else
5380                 return -EINVAL;
5381
5382         if (!dev)
5383                 return -ENODEV;
5384
5385         filter_mask = ifsm->filter_mask;
5386         if (!filter_mask)
5387                 return -EINVAL;
5388
5389         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5390         if (!nskb)
5391                 return -ENOBUFS;
5392
5393         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5394                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5395                                   0, filter_mask, &idxattr, &prividx);
5396         if (err < 0) {
5397                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5398                 WARN_ON(err == -EMSGSIZE);
5399                 kfree_skb(nskb);
5400         } else {
5401                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5402         }
5403
5404         return err;
5405 }
5406
5407 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5408 {
5409         struct netlink_ext_ack *extack = cb->extack;
5410         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5411         struct net *net = sock_net(skb->sk);
5412         unsigned int flags = NLM_F_MULTI;
5413         struct if_stats_msg *ifsm;
5414         struct hlist_head *head;
5415         struct net_device *dev;
5416         u32 filter_mask = 0;
5417         int idx = 0;
5418
5419         s_h = cb->args[0];
5420         s_idx = cb->args[1];
5421         s_idxattr = cb->args[2];
5422         s_prividx = cb->args[3];
5423
5424         cb->seq = net->dev_base_seq;
5425
5426         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5427         if (err)
5428                 return err;
5429
5430         ifsm = nlmsg_data(cb->nlh);
5431         filter_mask = ifsm->filter_mask;
5432         if (!filter_mask) {
5433                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5434                 return -EINVAL;
5435         }
5436
5437         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5438                 idx = 0;
5439                 head = &net->dev_index_head[h];
5440                 hlist_for_each_entry(dev, head, index_hlist) {
5441                         if (idx < s_idx)
5442                                 goto cont;
5443                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5444                                                   NETLINK_CB(cb->skb).portid,
5445                                                   cb->nlh->nlmsg_seq, 0,
5446                                                   flags, filter_mask,
5447                                                   &s_idxattr, &s_prividx);
5448                         /* If we ran out of room on the first message,
5449                          * we're in trouble
5450                          */
5451                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5452
5453                         if (err < 0)
5454                                 goto out;
5455                         s_prividx = 0;
5456                         s_idxattr = 0;
5457                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5458 cont:
5459                         idx++;
5460                 }
5461         }
5462 out:
5463         cb->args[3] = s_prividx;
5464         cb->args[2] = s_idxattr;
5465         cb->args[1] = idx;
5466         cb->args[0] = h;
5467
5468         return skb->len;
5469 }
5470
5471 /* Process one rtnetlink message. */
5472
5473 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5474                              struct netlink_ext_ack *extack)
5475 {
5476         struct net *net = sock_net(skb->sk);
5477         struct rtnl_link *link;
5478         struct module *owner;
5479         int err = -EOPNOTSUPP;
5480         rtnl_doit_func doit;
5481         unsigned int flags;
5482         int kind;
5483         int family;
5484         int type;
5485
5486         type = nlh->nlmsg_type;
5487         if (type > RTM_MAX)
5488                 return -EOPNOTSUPP;
5489
5490         type -= RTM_BASE;
5491
5492         /* All the messages must have at least 1 byte length */
5493         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5494                 return 0;
5495
5496         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5497         kind = type&3;
5498
5499         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5500                 return -EPERM;
5501
5502         rcu_read_lock();
5503         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5504                 struct sock *rtnl;
5505                 rtnl_dumpit_func dumpit;
5506                 u32 min_dump_alloc = 0;
5507
5508                 link = rtnl_get_link(family, type);
5509                 if (!link || !link->dumpit) {
5510                         family = PF_UNSPEC;
5511                         link = rtnl_get_link(family, type);
5512                         if (!link || !link->dumpit)
5513                                 goto err_unlock;
5514                 }
5515                 owner = link->owner;
5516                 dumpit = link->dumpit;
5517
5518                 if (type == RTM_GETLINK - RTM_BASE)
5519                         min_dump_alloc = rtnl_calcit(skb, nlh);
5520
5521                 err = 0;
5522                 /* need to do this before rcu_read_unlock() */
5523                 if (!try_module_get(owner))
5524                         err = -EPROTONOSUPPORT;
5525
5526                 rcu_read_unlock();
5527
5528                 rtnl = net->rtnl;
5529                 if (err == 0) {
5530                         struct netlink_dump_control c = {
5531                                 .dump           = dumpit,
5532                                 .min_dump_alloc = min_dump_alloc,
5533                                 .module         = owner,
5534                         };
5535                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5536                         /* netlink_dump_start() will keep a reference on
5537                          * module if dump is still in progress.
5538                          */
5539                         module_put(owner);
5540                 }
5541                 return err;
5542         }
5543
5544         link = rtnl_get_link(family, type);
5545         if (!link || !link->doit) {
5546                 family = PF_UNSPEC;
5547                 link = rtnl_get_link(PF_UNSPEC, type);
5548                 if (!link || !link->doit)
5549                         goto out_unlock;
5550         }
5551
5552         owner = link->owner;
5553         if (!try_module_get(owner)) {
5554                 err = -EPROTONOSUPPORT;
5555                 goto out_unlock;
5556         }
5557
5558         flags = link->flags;
5559         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5560                 doit = link->doit;
5561                 rcu_read_unlock();
5562                 if (doit)
5563                         err = doit(skb, nlh, extack);
5564                 module_put(owner);
5565                 return err;
5566         }
5567         rcu_read_unlock();
5568
5569         rtnl_lock();
5570         link = rtnl_get_link(family, type);
5571         if (link && link->doit)
5572                 err = link->doit(skb, nlh, extack);
5573         rtnl_unlock();
5574
5575         module_put(owner);
5576
5577         return err;
5578
5579 out_unlock:
5580         rcu_read_unlock();
5581         return err;
5582
5583 err_unlock:
5584         rcu_read_unlock();
5585         return -EOPNOTSUPP;
5586 }
5587
5588 static void rtnetlink_rcv(struct sk_buff *skb)
5589 {
5590         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5591 }
5592
5593 static int rtnetlink_bind(struct net *net, int group)
5594 {
5595         switch (group) {
5596         case RTNLGRP_IPV4_MROUTE_R:
5597         case RTNLGRP_IPV6_MROUTE_R:
5598                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5599                         return -EPERM;
5600                 break;
5601         }
5602         return 0;
5603 }
5604
5605 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5606 {
5607         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5608
5609         switch (event) {
5610         case NETDEV_REBOOT:
5611         case NETDEV_CHANGEMTU:
5612         case NETDEV_CHANGEADDR:
5613         case NETDEV_CHANGENAME:
5614         case NETDEV_FEAT_CHANGE:
5615         case NETDEV_BONDING_FAILOVER:
5616         case NETDEV_POST_TYPE_CHANGE:
5617         case NETDEV_NOTIFY_PEERS:
5618         case NETDEV_CHANGEUPPER:
5619         case NETDEV_RESEND_IGMP:
5620         case NETDEV_CHANGEINFODATA:
5621         case NETDEV_CHANGELOWERSTATE:
5622         case NETDEV_CHANGE_TX_QUEUE_LEN:
5623                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5624                                    GFP_KERNEL, NULL, 0);
5625                 break;
5626         default:
5627                 break;
5628         }
5629         return NOTIFY_DONE;
5630 }
5631
5632 static struct notifier_block rtnetlink_dev_notifier = {
5633         .notifier_call  = rtnetlink_event,
5634 };
5635
5636
5637 static int __net_init rtnetlink_net_init(struct net *net)
5638 {
5639         struct sock *sk;
5640         struct netlink_kernel_cfg cfg = {
5641                 .groups         = RTNLGRP_MAX,
5642                 .input          = rtnetlink_rcv,
5643                 .cb_mutex       = &rtnl_mutex,
5644                 .flags          = NL_CFG_F_NONROOT_RECV,
5645                 .bind           = rtnetlink_bind,
5646         };
5647
5648         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5649         if (!sk)
5650                 return -ENOMEM;
5651         net->rtnl = sk;
5652         return 0;
5653 }
5654
5655 static void __net_exit rtnetlink_net_exit(struct net *net)
5656 {
5657         netlink_kernel_release(net->rtnl);
5658         net->rtnl = NULL;
5659 }
5660
5661 static struct pernet_operations rtnetlink_net_ops = {
5662         .init = rtnetlink_net_init,
5663         .exit = rtnetlink_net_exit,
5664 };
5665
5666 void __init rtnetlink_init(void)
5667 {
5668         if (register_pernet_subsys(&rtnetlink_net_ops))
5669                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5670
5671         register_netdevice_notifier(&rtnetlink_dev_notifier);
5672
5673         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5674                       rtnl_dump_ifinfo, 0);
5675         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5676         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5677         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5678
5679         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5680         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5681         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5682
5683         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5684         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5685
5686         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5687         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5688         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5689
5690         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5691         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5692         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5693
5694         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5695                       0);
5696 }