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         int err = 0;
 714
 715         NETLINK_CB(skb).dst_group = group;
 716         if (echo)
 717                 refcount_inc(&skb->users);
 718         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 719         if (echo)
 720                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 721         return err;
 722 }
 723
 724 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 725 {
 726         struct sock *rtnl = net->rtnl;
 727
 728         return nlmsg_unicast(rtnl, skb, pid);
 729 }
 730 EXPORT_SYMBOL(rtnl_unicast);
 731
 732 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 733                  struct nlmsghdr *nlh, gfp_t flags)
 734 {
 735         struct sock *rtnl = net->rtnl;
 736         int report = 0;
 737
 738         if (nlh)
 739                 report = nlmsg_report(nlh);
 740
 741         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 742 }
 743 EXPORT_SYMBOL(rtnl_notify);
 744
 745 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 746 {
 747         struct sock *rtnl = net->rtnl;
 748
 749         netlink_set_err(rtnl, 0, group, error);
 750 }
 751 EXPORT_SYMBOL(rtnl_set_sk_err);
 752
 753 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 754 {
 755         struct nlattr *mx;
 756         int i, valid = 0;
 757
 758         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 759         if (metrics == dst_default_metrics.metrics)
 760                 return 0;
 761
 762         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 763         if (mx == NULL)
 764                 return -ENOBUFS;
 765
 766         for (i = 0; i < RTAX_MAX; i++) {
 767                 if (metrics[i]) {
 768                         if (i == RTAX_CC_ALGO - 1) {
 769                                 char tmp[TCP_CA_NAME_MAX], *name;
 770
 771                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 772                                 if (!name)
 773                                         continue;
 774                                 if (nla_put_string(skb, i + 1, name))
 775                                         goto nla_put_failure;
 776                         } else if (i == RTAX_FEATURES - 1) {
 777                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 778
 779                                 if (!user_features)
 780                                         continue;
 781                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 782                                 if (nla_put_u32(skb, i + 1, user_features))
 783                                         goto nla_put_failure;
 784                         } else {
 785                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 786                                         goto nla_put_failure;
 787                         }
 788                         valid++;
 789                 }
 790         }
 791
 792         if (!valid) {
 793                 nla_nest_cancel(skb, mx);
 794                 return 0;
 795         }
 796
 797         return nla_nest_end(skb, mx);
 798
 799 nla_put_failure:
 800         nla_nest_cancel(skb, mx);
 801         return -EMSGSIZE;
 802 }
 803 EXPORT_SYMBOL(rtnetlink_put_metrics);
 804
 805 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 806                        long expires, u32 error)
 807 {
 808         struct rta_cacheinfo ci = {
 809                 .rta_error = error,
 810                 .rta_id =  id,
 811         };
 812
 813         if (dst) {
 814                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 815                 ci.rta_used = dst->__use;
 816                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 817         }
 818         if (expires) {
 819                 unsigned long clock;
 820
 821                 clock = jiffies_to_clock_t(abs(expires));
 822                 clock = min_t(unsigned long, clock, INT_MAX);
 823                 ci.rta_expires = (expires > 0) ? clock : -clock;
 824         }
 825         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 826 }
 827 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 828
 829 static void set_operstate(struct net_device *dev, unsigned char transition)
 830 {
 831         unsigned char operstate = dev->operstate;
 832
 833         switch (transition) {
 834         case IF_OPER_UP:
 835                 if ((operstate == IF_OPER_DORMANT ||
 836                      operstate == IF_OPER_TESTING ||
 837                      operstate == IF_OPER_UNKNOWN) &&
 838                     !netif_dormant(dev) && !netif_testing(dev))
 839                         operstate = IF_OPER_UP;
 840                 break;
 841
 842         case IF_OPER_TESTING:
 843                 if (operstate == IF_OPER_UP ||
 844                     operstate == IF_OPER_UNKNOWN)
 845                         operstate = IF_OPER_TESTING;
 846                 break;
 847
 848         case IF_OPER_DORMANT:
 849                 if (operstate == IF_OPER_UP ||
 850                     operstate == IF_OPER_UNKNOWN)
 851                         operstate = IF_OPER_DORMANT;
 852                 break;
 853         }
 854
 855         if (dev->operstate != operstate) {
 856                 write_lock_bh(&dev_base_lock);
 857                 dev->operstate = operstate;
 858                 write_unlock_bh(&dev_base_lock);
 859                 netdev_state_change(dev);
 860         }
 861 }
 862
 863 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 864 {
 865         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 866                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 867 }
 868
 869 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 870                                            const struct ifinfomsg *ifm)
 871 {
 872         unsigned int flags = ifm->ifi_flags;
 873
 874         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 875         if (ifm->ifi_change)
 876                 flags = (flags & ifm->ifi_change) |
 877                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 878
 879         return flags;
 880 }
 881
 882 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 883                                  const struct rtnl_link_stats64 *b)
 884 {
 885         a->rx_packets = b->rx_packets;
 886         a->tx_packets = b->tx_packets;
 887         a->rx_bytes = b->rx_bytes;
 888         a->tx_bytes = b->tx_bytes;
 889         a->rx_errors = b->rx_errors;
 890         a->tx_errors = b->tx_errors;
 891         a->rx_dropped = b->rx_dropped;
 892         a->tx_dropped = b->tx_dropped;
 893
 894         a->multicast = b->multicast;
 895         a->collisions = b->collisions;
 896
 897         a->rx_length_errors = b->rx_length_errors;
 898         a->rx_over_errors = b->rx_over_errors;
 899         a->rx_crc_errors = b->rx_crc_errors;
 900         a->rx_frame_errors = b->rx_frame_errors;
 901         a->rx_fifo_errors = b->rx_fifo_errors;
 902         a->rx_missed_errors = b->rx_missed_errors;
 903
 904         a->tx_aborted_errors = b->tx_aborted_errors;
 905         a->tx_carrier_errors = b->tx_carrier_errors;
 906         a->tx_fifo_errors = b->tx_fifo_errors;
 907         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 908         a->tx_window_errors = b->tx_window_errors;
 909
 910         a->rx_compressed = b->rx_compressed;
 911         a->tx_compressed = b->tx_compressed;
 912
 913         a->rx_nohandler = b->rx_nohandler;
 914 }
 915
 916 /* All VF info */
 917 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 918                                    u32 ext_filter_mask)
 919 {
 920         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 921                 int num_vfs = dev_num_vf(dev->dev.parent);
 922                 size_t size = nla_total_size(0);
 923                 size += num_vfs *
 924                         (nla_total_size(0) +
 925                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 926                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 927                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 928                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 929                          nla_total_size(MAX_VLAN_LIST_LEN *
 930                                         sizeof(struct ifla_vf_vlan_info)) +
 931                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 932                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 933                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 934                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 935                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 936                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 937                          /* IFLA_VF_STATS_RX_PACKETS */
 938                          nla_total_size_64bit(sizeof(__u64)) +
 939                          /* IFLA_VF_STATS_TX_PACKETS */
 940                          nla_total_size_64bit(sizeof(__u64)) +
 941                          /* IFLA_VF_STATS_RX_BYTES */
 942                          nla_total_size_64bit(sizeof(__u64)) +
 943                          /* IFLA_VF_STATS_TX_BYTES */
 944                          nla_total_size_64bit(sizeof(__u64)) +
 945                          /* IFLA_VF_STATS_BROADCAST */
 946                          nla_total_size_64bit(sizeof(__u64)) +
 947                          /* IFLA_VF_STATS_MULTICAST */
 948                          nla_total_size_64bit(sizeof(__u64)) +
 949                          /* IFLA_VF_STATS_RX_DROPPED */
 950                          nla_total_size_64bit(sizeof(__u64)) +
 951                          /* IFLA_VF_STATS_TX_DROPPED */
 952                          nla_total_size_64bit(sizeof(__u64)) +
 953                          nla_total_size(sizeof(struct ifla_vf_trust)));
 954                 return size;
 955         } else
 956                 return 0;
 957 }
 958
 959 static size_t rtnl_port_size(const struct net_device *dev,
 960                              u32 ext_filter_mask)
 961 {
 962         size_t port_size = nla_total_size(4)            /* PORT_VF */
 963                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 964                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 965                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 966                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 967                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 968         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 969         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 970                 + port_size;
 971         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 972                 + port_size;
 973
 974         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 975             !(ext_filter_mask & RTEXT_FILTER_VF))
 976                 return 0;
 977         if (dev_num_vf(dev->dev.parent))
 978                 return port_self_size + vf_ports_size +
 979                         vf_port_size * dev_num_vf(dev->dev.parent);
 980         else
 981                 return port_self_size;
 982 }
 983
 984 static size_t rtnl_xdp_size(void)
 985 {
 986         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 987                           nla_total_size(1) +   /* XDP_ATTACHED */
 988                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 989                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 990
 991         return xdp_size;
 992 }
 993
 994 static size_t rtnl_prop_list_size(const struct net_device *dev)
 995 {
 996         struct netdev_name_node *name_node;
 997         size_t size;
 998
 999         if (list_empty(&dev->name_node->list))
1000                 return 0;
1001         size = nla_total_size(0);
1002         list_for_each_entry(name_node, &dev->name_node->list, list)
1003                 size += nla_total_size(ALTIFNAMSIZ);
1004         return size;
1005 }
1006
1007 static size_t rtnl_proto_down_size(const struct net_device *dev)
1008 {
1009         size_t size = nla_total_size(1);
1010
1011         if (dev->proto_down_reason)
1012                 size += nla_total_size(0) + nla_total_size(4);
1013
1014         return size;
1015 }
1016
1017 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1018                                      u32 ext_filter_mask)
1019 {
1020         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1021                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1022                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1023                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1024                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1025                + nla_total_size(sizeof(struct rtnl_link_stats))
1026                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1027                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1028                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1029                + nla_total_size(4) /* IFLA_TXQLEN */
1030                + nla_total_size(4) /* IFLA_WEIGHT */
1031                + nla_total_size(4) /* IFLA_MTU */
1032                + nla_total_size(4) /* IFLA_LINK */
1033                + nla_total_size(4) /* IFLA_MASTER */
1034                + nla_total_size(1) /* IFLA_CARRIER */
1035                + nla_total_size(4) /* IFLA_PROMISCUITY */
1036                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1037                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1038                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1039                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1040                + nla_total_size(1) /* IFLA_OPERSTATE */
1041                + nla_total_size(1) /* IFLA_LINKMODE */
1042                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1043                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1044                + nla_total_size(4) /* IFLA_GROUP */
1045                + nla_total_size(ext_filter_mask
1046                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1047                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1048                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1049                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1050                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1051                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1052                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1053                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1054                + rtnl_xdp_size() /* IFLA_XDP */
1055                + nla_total_size(4)  /* IFLA_EVENT */
1056                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1057                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1058                + rtnl_proto_down_size(dev)  /* proto down */
1059                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1060                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1061                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1062                + nla_total_size(4)  /* IFLA_MIN_MTU */
1063                + nla_total_size(4)  /* IFLA_MAX_MTU */
1064                + rtnl_prop_list_size(dev)
1065                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1066                + 0;
1067 }
1068
1069 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1070 {
1071         struct nlattr *vf_ports;
1072         struct nlattr *vf_port;
1073         int vf;
1074         int err;
1075
1076         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1077         if (!vf_ports)
1078                 return -EMSGSIZE;
1079
1080         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1081                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1082                 if (!vf_port)
1083                         goto nla_put_failure;
1084                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1085                         goto nla_put_failure;
1086                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1087                 if (err == -EMSGSIZE)
1088                         goto nla_put_failure;
1089                 if (err) {
1090                         nla_nest_cancel(skb, vf_port);
1091                         continue;
1092                 }
1093                 nla_nest_end(skb, vf_port);
1094         }
1095
1096         nla_nest_end(skb, vf_ports);
1097
1098         return 0;
1099
1100 nla_put_failure:
1101         nla_nest_cancel(skb, vf_ports);
1102         return -EMSGSIZE;
1103 }
1104
1105 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1106 {
1107         struct nlattr *port_self;
1108         int err;
1109
1110         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1111         if (!port_self)
1112                 return -EMSGSIZE;
1113
1114         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1115         if (err) {
1116                 nla_nest_cancel(skb, port_self);
1117                 return (err == -EMSGSIZE) ? err : 0;
1118         }
1119
1120         nla_nest_end(skb, port_self);
1121
1122         return 0;
1123 }
1124
1125 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1126                           u32 ext_filter_mask)
1127 {
1128         int err;
1129
1130         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1131             !(ext_filter_mask & RTEXT_FILTER_VF))
1132                 return 0;
1133
1134         err = rtnl_port_self_fill(skb, dev);
1135         if (err)
1136                 return err;
1137
1138         if (dev_num_vf(dev->dev.parent)) {
1139                 err = rtnl_vf_ports_fill(skb, dev);
1140                 if (err)
1141                         return err;
1142         }
1143
1144         return 0;
1145 }
1146
1147 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1148 {
1149         int err;
1150         struct netdev_phys_item_id ppid;
1151
1152         err = dev_get_phys_port_id(dev, &ppid);
1153         if (err) {
1154                 if (err == -EOPNOTSUPP)
1155                         return 0;
1156                 return err;
1157         }
1158
1159         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1160                 return -EMSGSIZE;
1161
1162         return 0;
1163 }
1164
1165 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1166 {
1167         char name[IFNAMSIZ];
1168         int err;
1169
1170         err = dev_get_phys_port_name(dev, name, sizeof(name));
1171         if (err) {
1172                 if (err == -EOPNOTSUPP)
1173                         return 0;
1174                 return err;
1175         }
1176
1177         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1178                 return -EMSGSIZE;
1179
1180         return 0;
1181 }
1182
1183 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1184 {
1185         struct netdev_phys_item_id ppid = { };
1186         int err;
1187
1188         err = dev_get_port_parent_id(dev, &ppid, false);
1189         if (err) {
1190                 if (err == -EOPNOTSUPP)
1191                         return 0;
1192                 return err;
1193         }
1194
1195         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1196                 return -EMSGSIZE;
1197
1198         return 0;
1199 }
1200
1201 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1202                                               struct net_device *dev)
1203 {
1204         struct rtnl_link_stats64 *sp;
1205         struct nlattr *attr;
1206
1207         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1208                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1209         if (!attr)
1210                 return -EMSGSIZE;
1211
1212         sp = nla_data(attr);
1213         dev_get_stats(dev, sp);
1214
1215         attr = nla_reserve(skb, IFLA_STATS,
1216                            sizeof(struct rtnl_link_stats));
1217         if (!attr)
1218                 return -EMSGSIZE;
1219
1220         copy_rtnl_link_stats(nla_data(attr), sp);
1221
1222         return 0;
1223 }
1224
1225 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1226                                                struct net_device *dev,
1227                                                int vfs_num,
1228                                                struct nlattr *vfinfo)
1229 {
1230         struct ifla_vf_rss_query_en vf_rss_query_en;
1231         struct nlattr *vf, *vfstats, *vfvlanlist;
1232         struct ifla_vf_link_state vf_linkstate;
1233         struct ifla_vf_vlan_info vf_vlan_info;
1234         struct ifla_vf_spoofchk vf_spoofchk;
1235         struct ifla_vf_tx_rate vf_tx_rate;
1236         struct ifla_vf_stats vf_stats;
1237         struct ifla_vf_trust vf_trust;
1238         struct ifla_vf_vlan vf_vlan;
1239         struct ifla_vf_rate vf_rate;
1240         struct ifla_vf_mac vf_mac;
1241         struct ifla_vf_broadcast vf_broadcast;
1242         struct ifla_vf_info ivi;
1243         struct ifla_vf_guid node_guid;
1244         struct ifla_vf_guid port_guid;
1245
1246         memset(&ivi, 0, sizeof(ivi));
1247
1248         /* Not all SR-IOV capable drivers support the
1249          * spoofcheck and "RSS query enable" query.  Preset to
1250          * -1 so the user space tool can detect that the driver
1251          * didn't report anything.
1252          */
1253         ivi.spoofchk = -1;
1254         ivi.rss_query_en = -1;
1255         ivi.trusted = -1;
1256         /* The default value for VF link state is "auto"
1257          * IFLA_VF_LINK_STATE_AUTO which equals zero
1258          */
1259         ivi.linkstate = 0;
1260         /* VLAN Protocol by default is 802.1Q */
1261         ivi.vlan_proto = htons(ETH_P_8021Q);
1262         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1263                 return 0;
1264
1265         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1266         memset(&node_guid, 0, sizeof(node_guid));
1267         memset(&port_guid, 0, sizeof(port_guid));
1268
1269         vf_mac.vf =
1270                 vf_vlan.vf =
1271                 vf_vlan_info.vf =
1272                 vf_rate.vf =
1273                 vf_tx_rate.vf =
1274                 vf_spoofchk.vf =
1275                 vf_linkstate.vf =
1276                 vf_rss_query_en.vf =
1277                 vf_trust.vf =
1278                 node_guid.vf =
1279                 port_guid.vf = ivi.vf;
1280
1281         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1282         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1283         vf_vlan.vlan = ivi.vlan;
1284         vf_vlan.qos = ivi.qos;
1285         vf_vlan_info.vlan = ivi.vlan;
1286         vf_vlan_info.qos = ivi.qos;
1287         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1288         vf_tx_rate.rate = ivi.max_tx_rate;
1289         vf_rate.min_tx_rate = ivi.min_tx_rate;
1290         vf_rate.max_tx_rate = ivi.max_tx_rate;
1291         vf_spoofchk.setting = ivi.spoofchk;
1292         vf_linkstate.link_state = ivi.linkstate;
1293         vf_rss_query_en.setting = ivi.rss_query_en;
1294         vf_trust.setting = ivi.trusted;
1295         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1296         if (!vf)
1297                 goto nla_put_vfinfo_failure;
1298         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1299             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1300             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1301             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1302                     &vf_rate) ||
1303             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1304                     &vf_tx_rate) ||
1305             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1306                     &vf_spoofchk) ||
1307             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1308                     &vf_linkstate) ||
1309             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1310                     sizeof(vf_rss_query_en),
1311                     &vf_rss_query_en) ||
1312             nla_put(skb, IFLA_VF_TRUST,
1313                     sizeof(vf_trust), &vf_trust))
1314                 goto nla_put_vf_failure;
1315
1316         if (dev->netdev_ops->ndo_get_vf_guid &&
1317             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1318                                               &port_guid)) {
1319                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1320                             &node_guid) ||
1321                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1322                             &port_guid))
1323                         goto nla_put_vf_failure;
1324         }
1325         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1326         if (!vfvlanlist)
1327                 goto nla_put_vf_failure;
1328         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1329                     &vf_vlan_info)) {
1330                 nla_nest_cancel(skb, vfvlanlist);
1331                 goto nla_put_vf_failure;
1332         }
1333         nla_nest_end(skb, vfvlanlist);
1334         memset(&vf_stats, 0, sizeof(vf_stats));
1335         if (dev->netdev_ops->ndo_get_vf_stats)
1336                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1337                                                 &vf_stats);
1338         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1339         if (!vfstats)
1340                 goto nla_put_vf_failure;
1341         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1342                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1343             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1344                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1345             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1346                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1347             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1348                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1349             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1350                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1351             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1352                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1353             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1354                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1355             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1356                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1357                 nla_nest_cancel(skb, vfstats);
1358                 goto nla_put_vf_failure;
1359         }
1360         nla_nest_end(skb, vfstats);
1361         nla_nest_end(skb, vf);
1362         return 0;
1363
1364 nla_put_vf_failure:
1365         nla_nest_cancel(skb, vf);
1366 nla_put_vfinfo_failure:
1367         nla_nest_cancel(skb, vfinfo);
1368         return -EMSGSIZE;
1369 }
1370
1371 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1372                                            struct net_device *dev,
1373                                            u32 ext_filter_mask)
1374 {
1375         struct nlattr *vfinfo;
1376         int i, num_vfs;
1377
1378         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1379                 return 0;
1380
1381         num_vfs = dev_num_vf(dev->dev.parent);
1382         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1383                 return -EMSGSIZE;
1384
1385         if (!dev->netdev_ops->ndo_get_vf_config)
1386                 return 0;
1387
1388         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1389         if (!vfinfo)
1390                 return -EMSGSIZE;
1391
1392         for (i = 0; i < num_vfs; i++) {
1393                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1394                         return -EMSGSIZE;
1395         }
1396
1397         nla_nest_end(skb, vfinfo);
1398         return 0;
1399 }
1400
1401 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1402 {
1403         struct rtnl_link_ifmap map;
1404
1405         memset(&map, 0, sizeof(map));
1406         map.mem_start   = dev->mem_start;
1407         map.mem_end     = dev->mem_end;
1408         map.base_addr   = dev->base_addr;
1409         map.irq         = dev->irq;
1410         map.dma         = dev->dma;
1411         map.port        = dev->if_port;
1412
1413         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1414                 return -EMSGSIZE;
1415
1416         return 0;
1417 }
1418
1419 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1420 {
1421         const struct bpf_prog *generic_xdp_prog;
1422
1423         ASSERT_RTNL();
1424
1425         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1426         if (!generic_xdp_prog)
1427                 return 0;
1428         return generic_xdp_prog->aux->id;
1429 }
1430
1431 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1432 {
1433         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1434 }
1435
1436 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1437 {
1438         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1439 }
1440
1441 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1442                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1443                                u32 (*get_prog_id)(struct net_device *dev))
1444 {
1445         u32 curr_id;
1446         int err;
1447
1448         curr_id = get_prog_id(dev);
1449         if (!curr_id)
1450                 return 0;
1451
1452         *prog_id = curr_id;
1453         err = nla_put_u32(skb, attr, curr_id);
1454         if (err)
1455                 return err;
1456
1457         if (*mode != XDP_ATTACHED_NONE)
1458                 *mode = XDP_ATTACHED_MULTI;
1459         else
1460                 *mode = tgt_mode;
1461
1462         return 0;
1463 }
1464
1465 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1466 {
1467         struct nlattr *xdp;
1468         u32 prog_id;
1469         int err;
1470         u8 mode;
1471
1472         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1473         if (!xdp)
1474                 return -EMSGSIZE;
1475
1476         prog_id = 0;
1477         mode = XDP_ATTACHED_NONE;
1478         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1479                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1480         if (err)
1481                 goto err_cancel;
1482         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1483                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1484         if (err)
1485                 goto err_cancel;
1486         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1487                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1488         if (err)
1489                 goto err_cancel;
1490
1491         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1492         if (err)
1493                 goto err_cancel;
1494
1495         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1496                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1497                 if (err)
1498                         goto err_cancel;
1499         }
1500
1501         nla_nest_end(skb, xdp);
1502         return 0;
1503
1504 err_cancel:
1505         nla_nest_cancel(skb, xdp);
1506         return err;
1507 }
1508
1509 static u32 rtnl_get_event(unsigned long event)
1510 {
1511         u32 rtnl_event_type = IFLA_EVENT_NONE;
1512
1513         switch (event) {
1514         case NETDEV_REBOOT:
1515                 rtnl_event_type = IFLA_EVENT_REBOOT;
1516                 break;
1517         case NETDEV_FEAT_CHANGE:
1518                 rtnl_event_type = IFLA_EVENT_FEATURES;
1519                 break;
1520         case NETDEV_BONDING_FAILOVER:
1521                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1522                 break;
1523         case NETDEV_NOTIFY_PEERS:
1524                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1525                 break;
1526         case NETDEV_RESEND_IGMP:
1527                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1528                 break;
1529         case NETDEV_CHANGEINFODATA:
1530                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1531                 break;
1532         default:
1533                 break;
1534         }
1535
1536         return rtnl_event_type;
1537 }
1538
1539 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1540 {
1541         const struct net_device *upper_dev;
1542         int ret = 0;
1543
1544         rcu_read_lock();
1545
1546         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1547         if (upper_dev)
1548                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1549
1550         rcu_read_unlock();
1551         return ret;
1552 }
1553
1554 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1555                           bool force)
1556 {
1557         int ifindex = dev_get_iflink(dev);
1558
1559         if (force || dev->ifindex != ifindex)
1560                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1561
1562         return 0;
1563 }
1564
1565 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1566                                               struct net_device *dev)
1567 {
1568         char buf[IFALIASZ];
1569         int ret;
1570
1571         ret = dev_get_alias(dev, buf, sizeof(buf));
1572         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1573 }
1574
1575 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1576                                   const struct net_device *dev,
1577                                   struct net *src_net, gfp_t gfp)
1578 {
1579         bool put_iflink = false;
1580
1581         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1582                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1583
1584                 if (!net_eq(dev_net(dev), link_net)) {
1585                         int id = peernet2id_alloc(src_net, link_net, gfp);
1586
1587                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1588                                 return -EMSGSIZE;
1589
1590                         put_iflink = true;
1591                 }
1592         }
1593
1594         return nla_put_iflink(skb, dev, put_iflink);
1595 }
1596
1597 static int rtnl_fill_link_af(struct sk_buff *skb,
1598                              const struct net_device *dev,
1599                              u32 ext_filter_mask)
1600 {
1601         const struct rtnl_af_ops *af_ops;
1602         struct nlattr *af_spec;
1603
1604         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1605         if (!af_spec)
1606                 return -EMSGSIZE;
1607
1608         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1609                 struct nlattr *af;
1610                 int err;
1611
1612                 if (!af_ops->fill_link_af)
1613                         continue;
1614
1615                 af = nla_nest_start_noflag(skb, af_ops->family);
1616                 if (!af)
1617                         return -EMSGSIZE;
1618
1619                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1620                 /*
1621                  * Caller may return ENODATA to indicate that there
1622                  * was no data to be dumped. This is not an error, it
1623                  * means we should trim the attribute header and
1624                  * continue.
1625                  */
1626                 if (err == -ENODATA)
1627                         nla_nest_cancel(skb, af);
1628                 else if (err < 0)
1629                         return -EMSGSIZE;
1630
1631                 nla_nest_end(skb, af);
1632         }
1633
1634         nla_nest_end(skb, af_spec);
1635         return 0;
1636 }
1637
1638 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1639                                  const struct net_device *dev)
1640 {
1641         struct netdev_name_node *name_node;
1642         int count = 0;
1643
1644         list_for_each_entry(name_node, &dev->name_node->list, list) {
1645                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1646                         return -EMSGSIZE;
1647                 count++;
1648         }
1649         return count;
1650 }
1651
1652 static int rtnl_fill_prop_list(struct sk_buff *skb,
1653                                const struct net_device *dev)
1654 {
1655         struct nlattr *prop_list;
1656         int ret;
1657
1658         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1659         if (!prop_list)
1660                 return -EMSGSIZE;
1661
1662         ret = rtnl_fill_alt_ifnames(skb, dev);
1663         if (ret <= 0)
1664                 goto nest_cancel;
1665
1666         nla_nest_end(skb, prop_list);
1667         return 0;
1668
1669 nest_cancel:
1670         nla_nest_cancel(skb, prop_list);
1671         return ret;
1672 }
1673
1674 static int rtnl_fill_proto_down(struct sk_buff *skb,
1675                                 const struct net_device *dev)
1676 {
1677         struct nlattr *pr;
1678         u32 preason;
1679
1680         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1681                 goto nla_put_failure;
1682
1683         preason = dev->proto_down_reason;
1684         if (!preason)
1685                 return 0;
1686
1687         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1688         if (!pr)
1689                 return -EMSGSIZE;
1690
1691         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1692                 nla_nest_cancel(skb, pr);
1693                 goto nla_put_failure;
1694         }
1695
1696         nla_nest_end(skb, pr);
1697         return 0;
1698
1699 nla_put_failure:
1700         return -EMSGSIZE;
1701 }
1702
1703 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1704                             struct net_device *dev, struct net *src_net,
1705                             int type, u32 pid, u32 seq, u32 change,
1706                             unsigned int flags, u32 ext_filter_mask,
1707                             u32 event, int *new_nsid, int new_ifindex,
1708                             int tgt_netnsid, gfp_t gfp)
1709 {
1710         struct ifinfomsg *ifm;
1711         struct nlmsghdr *nlh;
1712
1713         ASSERT_RTNL();
1714         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1715         if (nlh == NULL)
1716                 return -EMSGSIZE;
1717
1718         ifm = nlmsg_data(nlh);
1719         ifm->ifi_family = AF_UNSPEC;
1720         ifm->__ifi_pad = 0;
1721         ifm->ifi_type = dev->type;
1722         ifm->ifi_index = dev->ifindex;
1723         ifm->ifi_flags = dev_get_flags(dev);
1724         ifm->ifi_change = change;
1725
1726         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1727                 goto nla_put_failure;
1728
1729         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1730             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1731             nla_put_u8(skb, IFLA_OPERSTATE,
1732                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1733             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1734             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1735             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1736             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1737             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1738             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1739             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1740             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1741             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1742 #ifdef CONFIG_RPS
1743             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1744 #endif
1745             put_master_ifindex(skb, dev) ||
1746             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1747             (dev->qdisc &&
1748              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1749             nla_put_ifalias(skb, dev) ||
1750             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1751                         atomic_read(&dev->carrier_up_count) +
1752                         atomic_read(&dev->carrier_down_count)) ||
1753             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1754                         atomic_read(&dev->carrier_up_count)) ||
1755             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1756                         atomic_read(&dev->carrier_down_count)))
1757                 goto nla_put_failure;
1758
1759         if (rtnl_fill_proto_down(skb, dev))
1760                 goto nla_put_failure;
1761
1762         if (event != IFLA_EVENT_NONE) {
1763                 if (nla_put_u32(skb, IFLA_EVENT, event))
1764                         goto nla_put_failure;
1765         }
1766
1767         if (rtnl_fill_link_ifmap(skb, dev))
1768                 goto nla_put_failure;
1769
1770         if (dev->addr_len) {
1771                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1772                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1773                         goto nla_put_failure;
1774         }
1775
1776         if (rtnl_phys_port_id_fill(skb, dev))
1777                 goto nla_put_failure;
1778
1779         if (rtnl_phys_port_name_fill(skb, dev))
1780                 goto nla_put_failure;
1781
1782         if (rtnl_phys_switch_id_fill(skb, dev))
1783                 goto nla_put_failure;
1784
1785         if (rtnl_fill_stats(skb, dev))
1786                 goto nla_put_failure;
1787
1788         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1789                 goto nla_put_failure;
1790
1791         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1792                 goto nla_put_failure;
1793
1794         if (rtnl_xdp_fill(skb, dev))
1795                 goto nla_put_failure;
1796
1797         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1798                 if (rtnl_link_fill(skb, dev) < 0)
1799                         goto nla_put_failure;
1800         }
1801
1802         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1803                 goto nla_put_failure;
1804
1805         if (new_nsid &&
1806             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1807                 goto nla_put_failure;
1808         if (new_ifindex &&
1809             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1810                 goto nla_put_failure;
1811
1812         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1813             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1814                 goto nla_put_failure;
1815
1816         rcu_read_lock();
1817         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1818                 goto nla_put_failure_rcu;
1819         rcu_read_unlock();
1820
1821         if (rtnl_fill_prop_list(skb, dev))
1822                 goto nla_put_failure;
1823
1824         if (dev->dev.parent &&
1825             nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1826                            dev_name(dev->dev.parent)))
1827                 goto nla_put_failure;
1828
1829         if (dev->dev.parent && dev->dev.parent->bus &&
1830             nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1831                            dev->dev.parent->bus->name))
1832                 goto nla_put_failure;
1833
1834         nlmsg_end(skb, nlh);
1835         return 0;
1836
1837 nla_put_failure_rcu:
1838         rcu_read_unlock();
1839 nla_put_failure:
1840         nlmsg_cancel(skb, nlh);
1841         return -EMSGSIZE;
1842 }
1843
1844 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1845         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1846         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1847         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1848         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1849         [IFLA_MTU]              = { .type = NLA_U32 },
1850         [IFLA_LINK]             = { .type = NLA_U32 },
1851         [IFLA_MASTER]           = { .type = NLA_U32 },
1852         [IFLA_CARRIER]          = { .type = NLA_U8 },
1853         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1854         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1855         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1856         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1857         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1858         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1859         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1860         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1861          * allow 0-length string (needed to remove an alias).
1862          */
1863         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1864         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1865         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1866         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1867         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1868         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1869         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1870         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1871         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1872         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1873         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1874         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1875         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1876         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1877         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1878         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1879         [IFLA_XDP]              = { .type = NLA_NESTED },
1880         [IFLA_EVENT]            = { .type = NLA_U32 },
1881         [IFLA_GROUP]            = { .type = NLA_U32 },
1882         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1883         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1884         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1885         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1886         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1887         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1888         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1889                                     .len = ALTIFNAMSIZ - 1 },
1890         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1891         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1892         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1893         [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
1894 };
1895
1896 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1897         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1898         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1899         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1900         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1901 };
1902
1903 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1904         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1905         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1906         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1907         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1908         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1909         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1910         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1911         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1912         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1913         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1914         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1915         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1916         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1917 };
1918
1919 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1920         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1921         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1922                                     .len = PORT_PROFILE_MAX },
1923         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1924                                       .len = PORT_UUID_MAX },
1925         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1926                                     .len = PORT_UUID_MAX },
1927         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1928         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1929
1930         /* Unused, but we need to keep it here since user space could
1931          * fill it. It's also broken with regard to NLA_BINARY use in
1932          * combination with structs.
1933          */
1934         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1935                                     .len = sizeof(struct ifla_port_vsi) },
1936 };
1937
1938 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1939         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1940         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1941         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1942         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1943         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1944         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1945 };
1946
1947 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1948 {
1949         const struct rtnl_link_ops *ops = NULL;
1950         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1951
1952         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1953                 return NULL;
1954
1955         if (linfo[IFLA_INFO_KIND]) {
1956                 char kind[MODULE_NAME_LEN];
1957
1958                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1959                 ops = rtnl_link_ops_get(kind);
1960         }
1961
1962         return ops;
1963 }
1964
1965 static bool link_master_filtered(struct net_device *dev, int master_idx)
1966 {
1967         struct net_device *master;
1968
1969         if (!master_idx)
1970                 return false;
1971
1972         master = netdev_master_upper_dev_get(dev);
1973         if (!master || master->ifindex != master_idx)
1974                 return true;
1975
1976         return false;
1977 }
1978
1979 static bool link_kind_filtered(const struct net_device *dev,
1980                                const struct rtnl_link_ops *kind_ops)
1981 {
1982         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1983                 return true;
1984
1985         return false;
1986 }
1987
1988 static bool link_dump_filtered(struct net_device *dev,
1989                                int master_idx,
1990                                const struct rtnl_link_ops *kind_ops)
1991 {
1992         if (link_master_filtered(dev, master_idx) ||
1993             link_kind_filtered(dev, kind_ops))
1994                 return true;
1995
1996         return false;
1997 }
1998
1999 /**
2000  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2001  * @sk: netlink socket
2002  * @netnsid: network namespace identifier
2003  *
2004  * Returns the network namespace identified by netnsid on success or an error
2005  * pointer on failure.
2006  */
2007 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2008 {
2009         struct net *net;
2010
2011         net = get_net_ns_by_id(sock_net(sk), netnsid);
2012         if (!net)
2013                 return ERR_PTR(-EINVAL);
2014
2015         /* For now, the caller is required to have CAP_NET_ADMIN in
2016          * the user namespace owning the target net ns.
2017          */
2018         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2019                 put_net(net);
2020                 return ERR_PTR(-EACCES);
2021         }
2022         return net;
2023 }
2024 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2025
2026 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2027                                       bool strict_check, struct nlattr **tb,
2028                                       struct netlink_ext_ack *extack)
2029 {
2030         int hdrlen;
2031
2032         if (strict_check) {
2033                 struct ifinfomsg *ifm;
2034
2035                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2036                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2037                         return -EINVAL;
2038                 }
2039
2040                 ifm = nlmsg_data(nlh);
2041                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2042                     ifm->ifi_change) {
2043                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2044                         return -EINVAL;
2045                 }
2046                 if (ifm->ifi_index) {
2047                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2048                         return -EINVAL;
2049                 }
2050
2051                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2052                                                      IFLA_MAX, ifla_policy,
2053                                                      extack);
2054         }
2055
2056         /* A hack to preserve kernel<->userspace interface.
2057          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2058          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2059          * what iproute2 < v3.9.0 used.
2060          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2061          * attribute, its netlink message is shorter than struct ifinfomsg.
2062          */
2063         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2064                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2065
2066         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2067                                       extack);
2068 }
2069
2070 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2071 {
2072         struct netlink_ext_ack *extack = cb->extack;
2073         const struct nlmsghdr *nlh = cb->nlh;
2074         struct net *net = sock_net(skb->sk);
2075         struct net *tgt_net = net;
2076         int h, s_h;
2077         int idx = 0, s_idx;
2078         struct net_device *dev;
2079         struct hlist_head *head;
2080         struct nlattr *tb[IFLA_MAX+1];
2081         u32 ext_filter_mask = 0;
2082         const struct rtnl_link_ops *kind_ops = NULL;
2083         unsigned int flags = NLM_F_MULTI;
2084         int master_idx = 0;
2085         int netnsid = -1;
2086         int err, i;
2087
2088         s_h = cb->args[0];
2089         s_idx = cb->args[1];
2090
2091         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2092         if (err < 0) {
2093                 if (cb->strict_check)
2094                         return err;
2095
2096                 goto walk_entries;
2097         }
2098
2099         for (i = 0; i <= IFLA_MAX; ++i) {
2100                 if (!tb[i])
2101                         continue;
2102
2103                 /* new attributes should only be added with strict checking */
2104                 switch (i) {
2105                 case IFLA_TARGET_NETNSID:
2106                         netnsid = nla_get_s32(tb[i]);
2107                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2108                         if (IS_ERR(tgt_net)) {
2109                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2110                                 return PTR_ERR(tgt_net);
2111                         }
2112                         break;
2113                 case IFLA_EXT_MASK:
2114                         ext_filter_mask = nla_get_u32(tb[i]);
2115                         break;
2116                 case IFLA_MASTER:
2117                         master_idx = nla_get_u32(tb[i]);
2118                         break;
2119                 case IFLA_LINKINFO:
2120                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2121                         break;
2122                 default:
2123                         if (cb->strict_check) {
2124                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2125                                 return -EINVAL;
2126                         }
2127                 }
2128         }
2129
2130         if (master_idx || kind_ops)
2131                 flags |= NLM_F_DUMP_FILTERED;
2132
2133 walk_entries:
2134         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2135                 idx = 0;
2136                 head = &tgt_net->dev_index_head[h];
2137                 hlist_for_each_entry(dev, head, index_hlist) {
2138                         if (link_dump_filtered(dev, master_idx, kind_ops))
2139                                 goto cont;
2140                         if (idx < s_idx)
2141                                 goto cont;
2142                         err = rtnl_fill_ifinfo(skb, dev, net,
2143                                                RTM_NEWLINK,
2144                                                NETLINK_CB(cb->skb).portid,
2145                                                nlh->nlmsg_seq, 0, flags,
2146                                                ext_filter_mask, 0, NULL, 0,
2147                                                netnsid, GFP_KERNEL);
2148
2149                         if (err < 0) {
2150                                 if (likely(skb->len))
2151                                         goto out;
2152
2153                                 goto out_err;
2154                         }
2155 cont:
2156                         idx++;
2157                 }
2158         }
2159 out:
2160         err = skb->len;
2161 out_err:
2162         cb->args[1] = idx;
2163         cb->args[0] = h;
2164         cb->seq = tgt_net->dev_base_seq;
2165         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2166         if (netnsid >= 0)
2167                 put_net(tgt_net);
2168
2169         return err;
2170 }
2171
2172 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2173                         struct netlink_ext_ack *exterr)
2174 {
2175         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2176                                     exterr);
2177 }
2178 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2179
2180 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2181 {
2182         struct net *net;
2183         /* Examine the link attributes and figure out which
2184          * network namespace we are talking about.
2185          */
2186         if (tb[IFLA_NET_NS_PID])
2187                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2188         else if (tb[IFLA_NET_NS_FD])
2189                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2190         else
2191                 net = get_net(src_net);
2192         return net;
2193 }
2194 EXPORT_SYMBOL(rtnl_link_get_net);
2195
2196 /* Figure out which network namespace we are talking about by
2197  * examining the link attributes in the following order:
2198  *
2199  * 1. IFLA_NET_NS_PID
2200  * 2. IFLA_NET_NS_FD
2201  * 3. IFLA_TARGET_NETNSID
2202  */
2203 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2204                                                struct nlattr *tb[])
2205 {
2206         struct net *net;
2207
2208         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2209                 return rtnl_link_get_net(src_net, tb);
2210
2211         if (!tb[IFLA_TARGET_NETNSID])
2212                 return get_net(src_net);
2213
2214         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2215         if (!net)
2216                 return ERR_PTR(-EINVAL);
2217
2218         return net;
2219 }
2220
2221 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2222                                              struct net *src_net,
2223                                              struct nlattr *tb[], int cap)
2224 {
2225         struct net *net;
2226
2227         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2228         if (IS_ERR(net))
2229                 return net;
2230
2231         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2232                 put_net(net);
2233                 return ERR_PTR(-EPERM);
2234         }
2235
2236         return net;
2237 }
2238
2239 /* Verify that rtnetlink requests do not pass additional properties
2240  * potentially referring to different network namespaces.
2241  */
2242 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2243                                     struct netlink_ext_ack *extack,
2244                                     bool netns_id_only)
2245 {
2246
2247         if (netns_id_only) {
2248                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2249                         return 0;
2250
2251                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2252                 return -EOPNOTSUPP;
2253         }
2254
2255         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2256                 goto invalid_attr;
2257
2258         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2259                 goto invalid_attr;
2260
2261         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2262                 goto invalid_attr;
2263
2264         return 0;
2265
2266 invalid_attr:
2267         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2268         return -EINVAL;
2269 }
2270
2271 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2272 {
2273         if (dev) {
2274                 if (tb[IFLA_ADDRESS] &&
2275                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2276                         return -EINVAL;
2277
2278                 if (tb[IFLA_BROADCAST] &&
2279                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2280                         return -EINVAL;
2281         }
2282
2283         if (tb[IFLA_AF_SPEC]) {
2284                 struct nlattr *af;
2285                 int rem, err;
2286
2287                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2288                         const struct rtnl_af_ops *af_ops;
2289
2290                         af_ops = rtnl_af_lookup(nla_type(af));
2291                         if (!af_ops)
2292                                 return -EAFNOSUPPORT;
2293
2294                         if (!af_ops->set_link_af)
2295                                 return -EOPNOTSUPP;
2296
2297                         if (af_ops->validate_link_af) {
2298                                 err = af_ops->validate_link_af(dev, af);
2299                                 if (err < 0)
2300                                         return err;
2301                         }
2302                 }
2303         }
2304
2305         return 0;
2306 }
2307
2308 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2309                                   int guid_type)
2310 {
2311         const struct net_device_ops *ops = dev->netdev_ops;
2312
2313         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2314 }
2315
2316 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2317 {
2318         if (dev->type != ARPHRD_INFINIBAND)
2319                 return -EOPNOTSUPP;
2320
2321         return handle_infiniband_guid(dev, ivt, guid_type);
2322 }
2323
2324 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2325 {
2326         const struct net_device_ops *ops = dev->netdev_ops;
2327         int err = -EINVAL;
2328
2329         if (tb[IFLA_VF_MAC]) {
2330                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2331
2332                 if (ivm->vf >= INT_MAX)
2333                         return -EINVAL;
2334                 err = -EOPNOTSUPP;
2335                 if (ops->ndo_set_vf_mac)
2336                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2337                                                   ivm->mac);
2338                 if (err < 0)
2339                         return err;
2340         }
2341
2342         if (tb[IFLA_VF_VLAN]) {
2343                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2344
2345                 if (ivv->vf >= INT_MAX)
2346                         return -EINVAL;
2347                 err = -EOPNOTSUPP;
2348                 if (ops->ndo_set_vf_vlan)
2349                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2350                                                    ivv->qos,
2351                                                    htons(ETH_P_8021Q));
2352                 if (err < 0)
2353                         return err;
2354         }
2355
2356         if (tb[IFLA_VF_VLAN_LIST]) {
2357                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2358                 struct nlattr *attr;
2359                 int rem, len = 0;
2360
2361                 err = -EOPNOTSUPP;
2362                 if (!ops->ndo_set_vf_vlan)
2363                         return err;
2364
2365                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2366                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2367                             nla_len(attr) < NLA_HDRLEN) {
2368                                 return -EINVAL;
2369                         }
2370                         if (len >= MAX_VLAN_LIST_LEN)
2371                                 return -EOPNOTSUPP;
2372                         ivvl[len] = nla_data(attr);
2373
2374                         len++;
2375                 }
2376                 if (len == 0)
2377                         return -EINVAL;
2378
2379                 if (ivvl[0]->vf >= INT_MAX)
2380                         return -EINVAL;
2381                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2382                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2383                 if (err < 0)
2384                         return err;
2385         }
2386
2387         if (tb[IFLA_VF_TX_RATE]) {
2388                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2389                 struct ifla_vf_info ivf;
2390
2391                 if (ivt->vf >= INT_MAX)
2392                         return -EINVAL;
2393                 err = -EOPNOTSUPP;
2394                 if (ops->ndo_get_vf_config)
2395                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2396                 if (err < 0)
2397                         return err;
2398
2399                 err = -EOPNOTSUPP;
2400                 if (ops->ndo_set_vf_rate)
2401                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2402                                                    ivf.min_tx_rate,
2403                                                    ivt->rate);
2404                 if (err < 0)
2405                         return err;
2406         }
2407
2408         if (tb[IFLA_VF_RATE]) {
2409                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2410
2411                 if (ivt->vf >= INT_MAX)
2412                         return -EINVAL;
2413                 err = -EOPNOTSUPP;
2414                 if (ops->ndo_set_vf_rate)
2415                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2416                                                    ivt->min_tx_rate,
2417                                                    ivt->max_tx_rate);
2418                 if (err < 0)
2419                         return err;
2420         }
2421
2422         if (tb[IFLA_VF_SPOOFCHK]) {
2423                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2424
2425                 if (ivs->vf >= INT_MAX)
2426                         return -EINVAL;
2427                 err = -EOPNOTSUPP;
2428                 if (ops->ndo_set_vf_spoofchk)
2429                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2430                                                        ivs->setting);
2431                 if (err < 0)
2432                         return err;
2433         }
2434
2435         if (tb[IFLA_VF_LINK_STATE]) {
2436                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2437
2438                 if (ivl->vf >= INT_MAX)
2439                         return -EINVAL;
2440                 err = -EOPNOTSUPP;
2441                 if (ops->ndo_set_vf_link_state)
2442                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2443                                                          ivl->link_state);
2444                 if (err < 0)
2445                         return err;
2446         }
2447
2448         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2449                 struct ifla_vf_rss_query_en *ivrssq_en;
2450
2451                 err = -EOPNOTSUPP;
2452                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2453                 if (ivrssq_en->vf >= INT_MAX)
2454                         return -EINVAL;
2455                 if (ops->ndo_set_vf_rss_query_en)
2456                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2457                                                            ivrssq_en->setting);
2458                 if (err < 0)
2459                         return err;
2460         }
2461
2462         if (tb[IFLA_VF_TRUST]) {
2463                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2464
2465                 if (ivt->vf >= INT_MAX)
2466                         return -EINVAL;
2467                 err = -EOPNOTSUPP;
2468                 if (ops->ndo_set_vf_trust)
2469                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2470                 if (err < 0)
2471                         return err;
2472         }
2473
2474         if (tb[IFLA_VF_IB_NODE_GUID]) {
2475                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2476
2477                 if (ivt->vf >= INT_MAX)
2478                         return -EINVAL;
2479                 if (!ops->ndo_set_vf_guid)
2480                         return -EOPNOTSUPP;
2481                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2482         }
2483
2484         if (tb[IFLA_VF_IB_PORT_GUID]) {
2485                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2486
2487                 if (ivt->vf >= INT_MAX)
2488                         return -EINVAL;
2489                 if (!ops->ndo_set_vf_guid)
2490                         return -EOPNOTSUPP;
2491
2492                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2493         }
2494
2495         return err;
2496 }
2497
2498 static int do_set_master(struct net_device *dev, int ifindex,
2499                          struct netlink_ext_ack *extack)
2500 {
2501         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2502         const struct net_device_ops *ops;
2503         int err;
2504
2505         if (upper_dev) {
2506                 if (upper_dev->ifindex == ifindex)
2507                         return 0;
2508                 ops = upper_dev->netdev_ops;
2509                 if (ops->ndo_del_slave) {
2510                         err = ops->ndo_del_slave(upper_dev, dev);
2511                         if (err)
2512                                 return err;
2513                 } else {
2514                         return -EOPNOTSUPP;
2515                 }
2516         }
2517
2518         if (ifindex) {
2519                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2520                 if (!upper_dev)
2521                         return -EINVAL;
2522                 ops = upper_dev->netdev_ops;
2523                 if (ops->ndo_add_slave) {
2524                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2525                         if (err)
2526                                 return err;
2527                 } else {
2528                         return -EOPNOTSUPP;
2529                 }
2530         }
2531         return 0;
2532 }
2533
2534 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2535         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2536         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2537 };
2538
2539 static int do_set_proto_down(struct net_device *dev,
2540                              struct nlattr *nl_proto_down,
2541                              struct nlattr *nl_proto_down_reason,
2542                              struct netlink_ext_ack *extack)
2543 {
2544         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2545         const struct net_device_ops *ops = dev->netdev_ops;
2546         unsigned long mask = 0;
2547         u32 value;
2548         bool proto_down;
2549         int err;
2550
2551         if (!ops->ndo_change_proto_down) {
2552                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2553                 return -EOPNOTSUPP;
2554         }
2555
2556         if (nl_proto_down_reason) {
2557                 err = nla_parse_nested_deprecated(pdreason,
2558                                                   IFLA_PROTO_DOWN_REASON_MAX,
2559                                                   nl_proto_down_reason,
2560                                                   ifla_proto_down_reason_policy,
2561                                                   NULL);
2562                 if (err < 0)
2563                         return err;
2564
2565                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2566                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2567                         return -EINVAL;
2568                 }
2569
2570                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2571
2572                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2573                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2574
2575                 dev_change_proto_down_reason(dev, mask, value);
2576         }
2577
2578         if (nl_proto_down) {
2579                 proto_down = nla_get_u8(nl_proto_down);
2580
2581                 /* Don't turn off protodown if there are active reasons */
2582                 if (!proto_down && dev->proto_down_reason) {
2583                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2584                         return -EBUSY;
2585                 }
2586                 err = dev_change_proto_down(dev,
2587                                             proto_down);
2588                 if (err)
2589                         return err;
2590         }
2591
2592         return 0;
2593 }
2594
2595 #define DO_SETLINK_MODIFIED     0x01
2596 /* notify flag means notify + modified. */
2597 #define DO_SETLINK_NOTIFY       0x03
2598 static int do_setlink(const struct sk_buff *skb,
2599                       struct net_device *dev, struct ifinfomsg *ifm,
2600                       struct netlink_ext_ack *extack,
2601                       struct nlattr **tb, char *ifname, int status)
2602 {
2603         const struct net_device_ops *ops = dev->netdev_ops;
2604         int err;
2605
2606         err = validate_linkmsg(dev, tb);
2607         if (err < 0)
2608                 return err;
2609
2610         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2611                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2612                 struct net *net;
2613                 int new_ifindex;
2614
2615                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2616                                                 tb, CAP_NET_ADMIN);
2617                 if (IS_ERR(net)) {
2618                         err = PTR_ERR(net);
2619                         goto errout;
2620                 }
2621
2622                 if (tb[IFLA_NEW_IFINDEX])
2623                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2624                 else
2625                         new_ifindex = 0;
2626
2627                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2628                 put_net(net);
2629                 if (err)
2630                         goto errout;
2631                 status |= DO_SETLINK_MODIFIED;
2632         }
2633
2634         if (tb[IFLA_MAP]) {
2635                 struct rtnl_link_ifmap *u_map;
2636                 struct ifmap k_map;
2637
2638                 if (!ops->ndo_set_config) {
2639                         err = -EOPNOTSUPP;
2640                         goto errout;
2641                 }
2642
2643                 if (!netif_device_present(dev)) {
2644                         err = -ENODEV;
2645                         goto errout;
2646                 }
2647
2648                 u_map = nla_data(tb[IFLA_MAP]);
2649                 k_map.mem_start = (unsigned long) u_map->mem_start;
2650                 k_map.mem_end = (unsigned long) u_map->mem_end;
2651                 k_map.base_addr = (unsigned short) u_map->base_addr;
2652                 k_map.irq = (unsigned char) u_map->irq;
2653                 k_map.dma = (unsigned char) u_map->dma;
2654                 k_map.port = (unsigned char) u_map->port;
2655
2656                 err = ops->ndo_set_config(dev, &k_map);
2657                 if (err < 0)
2658                         goto errout;
2659
2660                 status |= DO_SETLINK_NOTIFY;
2661         }
2662
2663         if (tb[IFLA_ADDRESS]) {
2664                 struct sockaddr *sa;
2665                 int len;
2666
2667                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2668                                                   sizeof(*sa));
2669                 sa = kmalloc(len, GFP_KERNEL);
2670                 if (!sa) {
2671                         err = -ENOMEM;
2672                         goto errout;
2673                 }
2674                 sa->sa_family = dev->type;
2675                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2676                        dev->addr_len);
2677                 err = dev_set_mac_address_user(dev, sa, extack);
2678                 kfree(sa);
2679                 if (err)
2680                         goto errout;
2681                 status |= DO_SETLINK_MODIFIED;
2682         }
2683
2684         if (tb[IFLA_MTU]) {
2685                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2686                 if (err < 0)
2687                         goto errout;
2688                 status |= DO_SETLINK_MODIFIED;
2689         }
2690
2691         if (tb[IFLA_GROUP]) {
2692                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2693                 status |= DO_SETLINK_NOTIFY;
2694         }
2695
2696         /*
2697          * Interface selected by interface index but interface
2698          * name provided implies that a name change has been
2699          * requested.
2700          */
2701         if (ifm->ifi_index > 0 && ifname[0]) {
2702                 err = dev_change_name(dev, ifname);
2703                 if (err < 0)
2704                         goto errout;
2705                 status |= DO_SETLINK_MODIFIED;
2706         }
2707
2708         if (tb[IFLA_IFALIAS]) {
2709                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2710                                     nla_len(tb[IFLA_IFALIAS]));
2711                 if (err < 0)
2712                         goto errout;
2713                 status |= DO_SETLINK_NOTIFY;
2714         }
2715
2716         if (tb[IFLA_BROADCAST]) {
2717                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2718                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2719         }
2720
2721         if (ifm->ifi_flags || ifm->ifi_change) {
2722                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2723                                        extack);
2724                 if (err < 0)
2725                         goto errout;
2726         }
2727
2728         if (tb[IFLA_MASTER]) {
2729                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2730                 if (err)
2731                         goto errout;
2732                 status |= DO_SETLINK_MODIFIED;
2733         }
2734
2735         if (tb[IFLA_CARRIER]) {
2736                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2737                 if (err)
2738                         goto errout;
2739                 status |= DO_SETLINK_MODIFIED;
2740         }
2741
2742         if (tb[IFLA_TXQLEN]) {
2743                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2744
2745                 err = dev_change_tx_queue_len(dev, value);
2746                 if (err)
2747                         goto errout;
2748                 status |= DO_SETLINK_MODIFIED;
2749         }
2750
2751         if (tb[IFLA_GSO_MAX_SIZE]) {
2752                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2753
2754                 if (max_size > GSO_MAX_SIZE) {
2755                         err = -EINVAL;
2756                         goto errout;
2757                 }
2758
2759                 if (dev->gso_max_size ^ max_size) {
2760                         netif_set_gso_max_size(dev, max_size);
2761                         status |= DO_SETLINK_MODIFIED;
2762                 }
2763         }
2764
2765         if (tb[IFLA_GSO_MAX_SEGS]) {
2766                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2767
2768                 if (max_segs > GSO_MAX_SEGS) {
2769                         err = -EINVAL;
2770                         goto errout;
2771                 }
2772
2773                 if (dev->gso_max_segs ^ max_segs) {
2774                         dev->gso_max_segs = max_segs;
2775                         status |= DO_SETLINK_MODIFIED;
2776                 }
2777         }
2778
2779         if (tb[IFLA_OPERSTATE])
2780                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2781
2782         if (tb[IFLA_LINKMODE]) {
2783                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2784
2785                 write_lock_bh(&dev_base_lock);
2786                 if (dev->link_mode ^ value)
2787                         status |= DO_SETLINK_NOTIFY;
2788                 dev->link_mode = value;
2789                 write_unlock_bh(&dev_base_lock);
2790         }
2791
2792         if (tb[IFLA_VFINFO_LIST]) {
2793                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2794                 struct nlattr *attr;
2795                 int rem;
2796
2797                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2798                         if (nla_type(attr) != IFLA_VF_INFO ||
2799                             nla_len(attr) < NLA_HDRLEN) {
2800                                 err = -EINVAL;
2801                                 goto errout;
2802                         }
2803                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2804                                                           attr,
2805                                                           ifla_vf_policy,
2806                                                           NULL);
2807                         if (err < 0)
2808                                 goto errout;
2809                         err = do_setvfinfo(dev, vfinfo);
2810                         if (err < 0)
2811                                 goto errout;
2812                         status |= DO_SETLINK_NOTIFY;
2813                 }
2814         }
2815         err = 0;
2816
2817         if (tb[IFLA_VF_PORTS]) {
2818                 struct nlattr *port[IFLA_PORT_MAX+1];
2819                 struct nlattr *attr;
2820                 int vf;
2821                 int rem;
2822
2823                 err = -EOPNOTSUPP;
2824                 if (!ops->ndo_set_vf_port)
2825                         goto errout;
2826
2827                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2828                         if (nla_type(attr) != IFLA_VF_PORT ||
2829                             nla_len(attr) < NLA_HDRLEN) {
2830                                 err = -EINVAL;
2831                                 goto errout;
2832                         }
2833                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2834                                                           attr,
2835                                                           ifla_port_policy,
2836                                                           NULL);
2837                         if (err < 0)
2838                                 goto errout;
2839                         if (!port[IFLA_PORT_VF]) {
2840                                 err = -EOPNOTSUPP;
2841                                 goto errout;
2842                         }
2843                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2844                         err = ops->ndo_set_vf_port(dev, vf, port);
2845                         if (err < 0)
2846                                 goto errout;
2847                         status |= DO_SETLINK_NOTIFY;
2848                 }
2849         }
2850         err = 0;
2851
2852         if (tb[IFLA_PORT_SELF]) {
2853                 struct nlattr *port[IFLA_PORT_MAX+1];
2854
2855                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2856                                                   tb[IFLA_PORT_SELF],
2857                                                   ifla_port_policy, NULL);
2858                 if (err < 0)
2859                         goto errout;
2860
2861                 err = -EOPNOTSUPP;
2862                 if (ops->ndo_set_vf_port)
2863                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2864                 if (err < 0)
2865                         goto errout;
2866                 status |= DO_SETLINK_NOTIFY;
2867         }
2868
2869         if (tb[IFLA_AF_SPEC]) {
2870                 struct nlattr *af;
2871                 int rem;
2872
2873                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2874                         const struct rtnl_af_ops *af_ops;
2875
2876                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2877
2878                         err = af_ops->set_link_af(dev, af, extack);
2879                         if (err < 0)
2880                                 goto errout;
2881
2882                         status |= DO_SETLINK_NOTIFY;
2883                 }
2884         }
2885         err = 0;
2886
2887         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2888                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2889                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2890                 if (err)
2891                         goto errout;
2892                 status |= DO_SETLINK_NOTIFY;
2893         }
2894
2895         if (tb[IFLA_XDP]) {
2896                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2897                 u32 xdp_flags = 0;
2898
2899                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2900                                                   tb[IFLA_XDP],
2901                                                   ifla_xdp_policy, NULL);
2902                 if (err < 0)
2903                         goto errout;
2904
2905                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2906                         err = -EINVAL;
2907                         goto errout;
2908                 }
2909
2910                 if (xdp[IFLA_XDP_FLAGS]) {
2911                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2912                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2913                                 err = -EINVAL;
2914                                 goto errout;
2915                         }
2916                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2917                                 err = -EINVAL;
2918                                 goto errout;
2919                         }
2920                 }
2921
2922                 if (xdp[IFLA_XDP_FD]) {
2923                         int expected_fd = -1;
2924
2925                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2926                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2927                                         err = -EINVAL;
2928                                         goto errout;
2929                                 }
2930                                 expected_fd =
2931                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2932                         }
2933
2934                         err = dev_change_xdp_fd(dev, extack,
2935                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2936                                                 expected_fd,
2937                                                 xdp_flags);
2938                         if (err)
2939                                 goto errout;
2940                         status |= DO_SETLINK_NOTIFY;
2941                 }
2942         }
2943
2944 errout:
2945         if (status & DO_SETLINK_MODIFIED) {
2946                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2947                         netdev_state_change(dev);
2948
2949                 if (err < 0)
2950                         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",
2951                                              dev->name);
2952         }
2953
2954         return err;
2955 }
2956
2957 static struct net_device *rtnl_dev_get(struct net *net,
2958                                        struct nlattr *ifname_attr,
2959                                        struct nlattr *altifname_attr,
2960                                        char *ifname)
2961 {
2962         char buffer[ALTIFNAMSIZ];
2963
2964         if (!ifname) {
2965                 ifname = buffer;
2966                 if (ifname_attr)
2967                         nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2968                 else if (altifname_attr)
2969                         nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2970                 else
2971                         return NULL;
2972         }
2973
2974         return __dev_get_by_name(net, ifname);
2975 }
2976
2977 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2978                         struct netlink_ext_ack *extack)
2979 {
2980         struct net *net = sock_net(skb->sk);
2981         struct ifinfomsg *ifm;
2982         struct net_device *dev;
2983         int err;
2984         struct nlattr *tb[IFLA_MAX+1];
2985         char ifname[IFNAMSIZ];
2986
2987         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2988                                      ifla_policy, extack);
2989         if (err < 0)
2990                 goto errout;
2991
2992         err = rtnl_ensure_unique_netns(tb, extack, false);
2993         if (err < 0)
2994                 goto errout;
2995
2996         if (tb[IFLA_IFNAME])
2997                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2998         else
2999                 ifname[0] = '\0';
3000
3001         err = -EINVAL;
3002         ifm = nlmsg_data(nlh);
3003         if (ifm->ifi_index > 0)
3004                 dev = __dev_get_by_index(net, ifm->ifi_index);
3005         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3006                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3007         else
3008                 goto errout;
3009
3010         if (dev == NULL) {
3011                 err = -ENODEV;
3012                 goto errout;
3013         }
3014
3015         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3016 errout:
3017         return err;
3018 }
3019
3020 static int rtnl_group_dellink(const struct net *net, int group)
3021 {
3022         struct net_device *dev, *aux;
3023         LIST_HEAD(list_kill);
3024         bool found = false;
3025
3026         if (!group)
3027                 return -EPERM;
3028
3029         for_each_netdev(net, dev) {
3030                 if (dev->group == group) {
3031                         const struct rtnl_link_ops *ops;
3032
3033                         found = true;
3034                         ops = dev->rtnl_link_ops;
3035                         if (!ops || !ops->dellink)
3036                                 return -EOPNOTSUPP;
3037                 }
3038         }
3039
3040         if (!found)
3041                 return -ENODEV;
3042
3043         for_each_netdev_safe(net, dev, aux) {
3044                 if (dev->group == group) {
3045                         const struct rtnl_link_ops *ops;
3046
3047                         ops = dev->rtnl_link_ops;
3048                         ops->dellink(dev, &list_kill);
3049                 }
3050         }
3051         unregister_netdevice_many(&list_kill);
3052
3053         return 0;
3054 }
3055
3056 int rtnl_delete_link(struct net_device *dev)
3057 {
3058         const struct rtnl_link_ops *ops;
3059         LIST_HEAD(list_kill);
3060
3061         ops = dev->rtnl_link_ops;
3062         if (!ops || !ops->dellink)
3063                 return -EOPNOTSUPP;
3064
3065         ops->dellink(dev, &list_kill);
3066         unregister_netdevice_many(&list_kill);
3067
3068         return 0;
3069 }
3070 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3071
3072 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3073                         struct netlink_ext_ack *extack)
3074 {
3075         struct net *net = sock_net(skb->sk);
3076         struct net *tgt_net = net;
3077         struct net_device *dev = NULL;
3078         struct ifinfomsg *ifm;
3079         struct nlattr *tb[IFLA_MAX+1];
3080         int err;
3081         int netnsid = -1;
3082
3083         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3084                                      ifla_policy, extack);
3085         if (err < 0)
3086                 return err;
3087
3088         err = rtnl_ensure_unique_netns(tb, extack, true);
3089         if (err < 0)
3090                 return err;
3091
3092         if (tb[IFLA_TARGET_NETNSID]) {
3093                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3094                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3095                 if (IS_ERR(tgt_net))
3096                         return PTR_ERR(tgt_net);
3097         }
3098
3099         err = -EINVAL;
3100         ifm = nlmsg_data(nlh);
3101         if (ifm->ifi_index > 0)
3102                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3103         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3104                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3105                                    tb[IFLA_ALT_IFNAME], NULL);
3106         else if (tb[IFLA_GROUP])
3107                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3108         else
3109                 goto out;
3110
3111         if (!dev) {
3112                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3113                         err = -ENODEV;
3114
3115                 goto out;
3116         }
3117
3118         err = rtnl_delete_link(dev);
3119
3120 out:
3121         if (netnsid >= 0)
3122                 put_net(tgt_net);
3123
3124         return err;
3125 }
3126
3127 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3128 {
3129         unsigned int old_flags;
3130         int err;
3131
3132         old_flags = dev->flags;
3133         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3134                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3135                                          NULL);
3136                 if (err < 0)
3137                         return err;
3138         }
3139
3140         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3141                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3142         } else {
3143                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3144                 __dev_notify_flags(dev, old_flags, ~0U);
3145         }
3146         return 0;
3147 }
3148 EXPORT_SYMBOL(rtnl_configure_link);
3149
3150 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3151                                     unsigned char name_assign_type,
3152                                     const struct rtnl_link_ops *ops,
3153                                     struct nlattr *tb[],
3154                                     struct netlink_ext_ack *extack)
3155 {
3156         struct net_device *dev;
3157         unsigned int num_tx_queues = 1;
3158         unsigned int num_rx_queues = 1;
3159
3160         if (tb[IFLA_NUM_TX_QUEUES])
3161                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3162         else if (ops->get_num_tx_queues)
3163                 num_tx_queues = ops->get_num_tx_queues();
3164
3165         if (tb[IFLA_NUM_RX_QUEUES])
3166                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3167         else if (ops->get_num_rx_queues)
3168                 num_rx_queues = ops->get_num_rx_queues();
3169
3170         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3171                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3172                 return ERR_PTR(-EINVAL);
3173         }
3174
3175         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3176                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3177                 return ERR_PTR(-EINVAL);
3178         }
3179
3180         if (ops->alloc) {
3181                 dev = ops->alloc(tb, ifname, name_assign_type,
3182                                  num_tx_queues, num_rx_queues);
3183                 if (IS_ERR(dev))
3184                         return dev;
3185         } else {
3186                 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3187                                        name_assign_type, ops->setup,
3188                                        num_tx_queues, num_rx_queues);
3189         }
3190
3191         if (!dev)
3192                 return ERR_PTR(-ENOMEM);
3193
3194         dev_net_set(dev, net);
3195         dev->rtnl_link_ops = ops;
3196         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3197
3198         if (tb[IFLA_MTU]) {
3199                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3200                 int err;
3201
3202                 err = dev_validate_mtu(dev, mtu, extack);
3203                 if (err) {
3204                         free_netdev(dev);
3205                         return ERR_PTR(err);
3206                 }
3207                 dev->mtu = mtu;
3208         }
3209         if (tb[IFLA_ADDRESS]) {
3210                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3211                                 nla_len(tb[IFLA_ADDRESS]));
3212                 dev->addr_assign_type = NET_ADDR_SET;
3213         }
3214         if (tb[IFLA_BROADCAST])
3215                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3216                                 nla_len(tb[IFLA_BROADCAST]));
3217         if (tb[IFLA_TXQLEN])
3218                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3219         if (tb[IFLA_OPERSTATE])
3220                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3221         if (tb[IFLA_LINKMODE])
3222                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3223         if (tb[IFLA_GROUP])
3224                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3225         if (tb[IFLA_GSO_MAX_SIZE])
3226                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3227         if (tb[IFLA_GSO_MAX_SEGS])
3228                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3229
3230         return dev;
3231 }
3232 EXPORT_SYMBOL(rtnl_create_link);
3233
3234 static int rtnl_group_changelink(const struct sk_buff *skb,
3235                 struct net *net, int group,
3236                 struct ifinfomsg *ifm,
3237                 struct netlink_ext_ack *extack,
3238                 struct nlattr **tb)
3239 {
3240         struct net_device *dev, *aux;
3241         int err;
3242
3243         for_each_netdev_safe(net, dev, aux) {
3244                 if (dev->group == group) {
3245                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3246                         if (err < 0)
3247                                 return err;
3248                 }
3249         }
3250
3251         return 0;
3252 }
3253
3254 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3255                           struct nlattr **attr, struct netlink_ext_ack *extack)
3256 {
3257         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3258         unsigned char name_assign_type = NET_NAME_USER;
3259         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3260         const struct rtnl_link_ops *m_ops = NULL;
3261         struct net_device *master_dev = NULL;
3262         struct net *net = sock_net(skb->sk);
3263         const struct rtnl_link_ops *ops;
3264         struct nlattr *tb[IFLA_MAX + 1];
3265         struct net *dest_net, *link_net;
3266         struct nlattr **slave_data;
3267         char kind[MODULE_NAME_LEN];
3268         struct net_device *dev;
3269         struct ifinfomsg *ifm;
3270         char ifname[IFNAMSIZ];
3271         struct nlattr **data;
3272         int err;
3273
3274 #ifdef CONFIG_MODULES
3275 replay:
3276 #endif
3277         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3278                                      ifla_policy, extack);
3279         if (err < 0)
3280                 return err;
3281
3282         err = rtnl_ensure_unique_netns(tb, extack, false);
3283         if (err < 0)
3284                 return err;
3285
3286         if (tb[IFLA_IFNAME])
3287                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3288         else
3289                 ifname[0] = '\0';
3290
3291         ifm = nlmsg_data(nlh);
3292         if (ifm->ifi_index > 0)
3293                 dev = __dev_get_by_index(net, ifm->ifi_index);
3294         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3295                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3296         else
3297                 dev = NULL;
3298
3299         if (dev) {
3300                 master_dev = netdev_master_upper_dev_get(dev);
3301                 if (master_dev)
3302                         m_ops = master_dev->rtnl_link_ops;
3303         }
3304
3305         err = validate_linkmsg(dev, tb);
3306         if (err < 0)
3307                 return err;
3308
3309         if (tb[IFLA_LINKINFO]) {
3310                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3311                                                   tb[IFLA_LINKINFO],
3312                                                   ifla_info_policy, NULL);
3313                 if (err < 0)
3314                         return err;
3315         } else
3316                 memset(linkinfo, 0, sizeof(linkinfo));
3317
3318         if (linkinfo[IFLA_INFO_KIND]) {
3319                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3320                 ops = rtnl_link_ops_get(kind);
3321         } else {
3322                 kind[0] = '\0';
3323                 ops = NULL;
3324         }
3325
3326         data = NULL;
3327         if (ops) {
3328                 if (ops->maxtype > RTNL_MAX_TYPE)
3329                         return -EINVAL;
3330
3331                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3332                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3333                                                           linkinfo[IFLA_INFO_DATA],
3334                                                           ops->policy, extack);
3335                         if (err < 0)
3336                                 return err;
3337                         data = attr;
3338                 }
3339                 if (ops->validate) {
3340                         err = ops->validate(tb, data, extack);
3341                         if (err < 0)
3342                                 return err;
3343                 }
3344         }
3345
3346         slave_data = NULL;
3347         if (m_ops) {
3348                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3349                         return -EINVAL;
3350
3351                 if (m_ops->slave_maxtype &&
3352                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3353                         err = nla_parse_nested_deprecated(slave_attr,
3354                                                           m_ops->slave_maxtype,
3355                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3356                                                           m_ops->slave_policy,
3357                                                           extack);
3358                         if (err < 0)
3359                                 return err;
3360                         slave_data = slave_attr;
3361                 }
3362         }
3363
3364         if (dev) {
3365                 int status = 0;
3366
3367                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3368                         return -EEXIST;
3369                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3370                         return -EOPNOTSUPP;
3371
3372                 if (linkinfo[IFLA_INFO_DATA]) {
3373                         if (!ops || ops != dev->rtnl_link_ops ||
3374                             !ops->changelink)
3375                                 return -EOPNOTSUPP;
3376
3377                         err = ops->changelink(dev, tb, data, extack);
3378                         if (err < 0)
3379                                 return err;
3380                         status |= DO_SETLINK_NOTIFY;
3381                 }
3382
3383                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3384                         if (!m_ops || !m_ops->slave_changelink)
3385                                 return -EOPNOTSUPP;
3386
3387                         err = m_ops->slave_changelink(master_dev, dev, tb,
3388                                                       slave_data, extack);
3389                         if (err < 0)
3390                                 return err;
3391                         status |= DO_SETLINK_NOTIFY;
3392                 }
3393
3394                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3395         }
3396
3397         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3398                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3399                         return rtnl_group_changelink(skb, net,
3400                                                 nla_get_u32(tb[IFLA_GROUP]),
3401                                                 ifm, extack, tb);
3402                 return -ENODEV;
3403         }
3404
3405         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3406                 return -EOPNOTSUPP;
3407
3408         if (!ops) {
3409 #ifdef CONFIG_MODULES
3410                 if (kind[0]) {
3411                         __rtnl_unlock();
3412                         request_module("rtnl-link-%s", kind);
3413                         rtnl_lock();
3414                         ops = rtnl_link_ops_get(kind);
3415                         if (ops)
3416                                 goto replay;
3417                 }
3418 #endif
3419                 NL_SET_ERR_MSG(extack, "Unknown device type");
3420                 return -EOPNOTSUPP;
3421         }
3422
3423         if (!ops->alloc && !ops->setup)
3424                 return -EOPNOTSUPP;
3425
3426         if (!ifname[0]) {
3427                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3428                 name_assign_type = NET_NAME_ENUM;
3429         }
3430
3431         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3432         if (IS_ERR(dest_net))
3433                 return PTR_ERR(dest_net);
3434
3435         if (tb[IFLA_LINK_NETNSID]) {
3436                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3437
3438                 link_net = get_net_ns_by_id(dest_net, id);
3439                 if (!link_net) {
3440                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3441                         err =  -EINVAL;
3442                         goto out;
3443                 }
3444                 err = -EPERM;
3445                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3446                         goto out;
3447         } else {
3448                 link_net = NULL;
3449         }
3450
3451         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3452                                name_assign_type, ops, tb, extack);
3453         if (IS_ERR(dev)) {
3454                 err = PTR_ERR(dev);
3455                 goto out;
3456         }
3457
3458         dev->ifindex = ifm->ifi_index;
3459
3460         if (ops->newlink)
3461                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3462         else
3463                 err = register_netdevice(dev);
3464         if (err < 0) {
3465                 free_netdev(dev);
3466                 goto out;
3467         }
3468
3469         err = rtnl_configure_link(dev, ifm);
3470         if (err < 0)
3471                 goto out_unregister;
3472         if (link_net) {
3473                 err = dev_change_net_namespace(dev, dest_net, ifname);
3474                 if (err < 0)
3475                         goto out_unregister;
3476         }
3477         if (tb[IFLA_MASTER]) {
3478                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3479                 if (err)
3480                         goto out_unregister;
3481         }
3482 out:
3483         if (link_net)
3484                 put_net(link_net);
3485         put_net(dest_net);
3486         return err;
3487 out_unregister:
3488         if (ops->newlink) {
3489                 LIST_HEAD(list_kill);
3490
3491                 ops->dellink(dev, &list_kill);
3492                 unregister_netdevice_many(&list_kill);
3493         } else {
3494                 unregister_netdevice(dev);
3495         }
3496         goto out;
3497 }
3498
3499 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3500                         struct netlink_ext_ack *extack)
3501 {
3502         struct nlattr **attr;
3503         int ret;
3504
3505         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3506         if (!attr)
3507                 return -ENOMEM;
3508
3509         ret = __rtnl_newlink(skb, nlh, attr, extack);
3510         kfree(attr);
3511         return ret;
3512 }
3513
3514 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3515                                   const struct nlmsghdr *nlh,
3516                                   struct nlattr **tb,
3517                                   struct netlink_ext_ack *extack)
3518 {
3519         struct ifinfomsg *ifm;
3520         int i, err;
3521
3522         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3523                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3524                 return -EINVAL;
3525         }
3526
3527         if (!netlink_strict_get_check(skb))
3528                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3529                                               ifla_policy, extack);
3530
3531         ifm = nlmsg_data(nlh);
3532         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3533             ifm->ifi_change) {
3534                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3535                 return -EINVAL;
3536         }
3537
3538         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3539                                             ifla_policy, extack);
3540         if (err)
3541                 return err;
3542
3543         for (i = 0; i <= IFLA_MAX; i++) {
3544                 if (!tb[i])
3545                         continue;
3546
3547                 switch (i) {
3548                 case IFLA_IFNAME:
3549                 case IFLA_ALT_IFNAME:
3550                 case IFLA_EXT_MASK:
3551                 case IFLA_TARGET_NETNSID:
3552                         break;
3553                 default:
3554                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3555                         return -EINVAL;
3556                 }
3557         }
3558
3559         return 0;
3560 }
3561
3562 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3563                         struct netlink_ext_ack *extack)
3564 {
3565         struct net *net = sock_net(skb->sk);
3566         struct net *tgt_net = net;
3567         struct ifinfomsg *ifm;
3568         struct nlattr *tb[IFLA_MAX+1];
3569         struct net_device *dev = NULL;
3570         struct sk_buff *nskb;
3571         int netnsid = -1;
3572         int err;
3573         u32 ext_filter_mask = 0;
3574
3575         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3576         if (err < 0)
3577                 return err;
3578
3579         err = rtnl_ensure_unique_netns(tb, extack, true);
3580         if (err < 0)
3581                 return err;
3582
3583         if (tb[IFLA_TARGET_NETNSID]) {
3584                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3585                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3586                 if (IS_ERR(tgt_net))
3587                         return PTR_ERR(tgt_net);
3588         }
3589
3590         if (tb[IFLA_EXT_MASK])
3591                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3592
3593         err = -EINVAL;
3594         ifm = nlmsg_data(nlh);
3595         if (ifm->ifi_index > 0)
3596                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3597         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3598                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3599                                    tb[IFLA_ALT_IFNAME], NULL);
3600         else
3601                 goto out;
3602
3603         err = -ENODEV;
3604         if (dev == NULL)
3605                 goto out;
3606
3607         err = -ENOBUFS;
3608         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3609         if (nskb == NULL)
3610                 goto out;
3611
3612         err = rtnl_fill_ifinfo(nskb, dev, net,
3613                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3614                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3615                                0, NULL, 0, netnsid, GFP_KERNEL);
3616         if (err < 0) {
3617                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3618                 WARN_ON(err == -EMSGSIZE);
3619                 kfree_skb(nskb);
3620         } else
3621                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3622 out:
3623         if (netnsid >= 0)
3624                 put_net(tgt_net);
3625
3626         return err;
3627 }
3628
3629 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3630                            bool *changed, struct netlink_ext_ack *extack)
3631 {
3632         char *alt_ifname;
3633         int err;
3634
3635         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3636         if (err)
3637                 return err;
3638
3639         alt_ifname = nla_strdup(attr, GFP_KERNEL);
3640         if (!alt_ifname)
3641                 return -ENOMEM;
3642
3643         if (cmd == RTM_NEWLINKPROP) {
3644                 err = netdev_name_node_alt_create(dev, alt_ifname);
3645                 if (!err)
3646                         alt_ifname = NULL;
3647         } else if (cmd == RTM_DELLINKPROP) {
3648                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3649         } else {
3650                 WARN_ON_ONCE(1);
3651                 err = -EINVAL;
3652         }
3653
3654         kfree(alt_ifname);
3655         if (!err)
3656                 *changed = true;
3657         return err;
3658 }
3659
3660 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3661                          struct netlink_ext_ack *extack)
3662 {
3663         struct net *net = sock_net(skb->sk);
3664         struct nlattr *tb[IFLA_MAX + 1];
3665         struct net_device *dev;
3666         struct ifinfomsg *ifm;
3667         bool changed = false;
3668         struct nlattr *attr;
3669         int err, rem;
3670
3671         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3672         if (err)
3673                 return err;
3674
3675         err = rtnl_ensure_unique_netns(tb, extack, true);
3676         if (err)
3677                 return err;
3678
3679         ifm = nlmsg_data(nlh);
3680         if (ifm->ifi_index > 0)
3681                 dev = __dev_get_by_index(net, ifm->ifi_index);
3682         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3683                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3684                                    tb[IFLA_ALT_IFNAME], NULL);
3685         else
3686                 return -EINVAL;
3687
3688         if (!dev)
3689                 return -ENODEV;
3690
3691         if (!tb[IFLA_PROP_LIST])
3692                 return 0;
3693
3694         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3695                 switch (nla_type(attr)) {
3696                 case IFLA_ALT_IFNAME:
3697                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3698                         if (err)
3699                                 return err;
3700                         break;
3701                 }
3702         }
3703
3704         if (changed)
3705                 netdev_state_change(dev);
3706         return 0;
3707 }
3708
3709 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3710                             struct netlink_ext_ack *extack)
3711 {
3712         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3713 }
3714
3715 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3716                             struct netlink_ext_ack *extack)
3717 {
3718         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3719 }
3720
3721 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3722 {
3723         struct net *net = sock_net(skb->sk);
3724         size_t min_ifinfo_dump_size = 0;
3725         struct nlattr *tb[IFLA_MAX+1];
3726         u32 ext_filter_mask = 0;
3727         struct net_device *dev;
3728         int hdrlen;
3729
3730         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3731         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3732                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3733
3734         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3735                 if (tb[IFLA_EXT_MASK])
3736                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3737         }
3738
3739         if (!ext_filter_mask)
3740                 return NLMSG_GOODSIZE;
3741         /*
3742          * traverse the list of net devices and compute the minimum
3743          * buffer size based upon the filter mask.
3744          */
3745         rcu_read_lock();
3746         for_each_netdev_rcu(net, dev) {
3747                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3748                                            if_nlmsg_size(dev, ext_filter_mask));
3749         }
3750         rcu_read_unlock();
3751
3752         return nlmsg_total_size(min_ifinfo_dump_size);
3753 }
3754
3755 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3756 {
3757         int idx;
3758         int s_idx = cb->family;
3759         int type = cb->nlh->nlmsg_type - RTM_BASE;
3760         int ret = 0;
3761
3762         if (s_idx == 0)
3763                 s_idx = 1;
3764
3765         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3766                 struct rtnl_link __rcu **tab;
3767                 struct rtnl_link *link;
3768                 rtnl_dumpit_func dumpit;
3769
3770                 if (idx < s_idx || idx == PF_PACKET)
3771                         continue;
3772
3773                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3774                         continue;
3775
3776                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3777                 if (!tab)
3778                         continue;
3779
3780                 link = rcu_dereference_rtnl(tab[type]);
3781                 if (!link)
3782                         continue;
3783
3784                 dumpit = link->dumpit;
3785                 if (!dumpit)
3786                         continue;
3787
3788                 if (idx > s_idx) {
3789                         memset(&cb->args[0], 0, sizeof(cb->args));
3790                         cb->prev_seq = 0;
3791                         cb->seq = 0;
3792                 }
3793                 ret = dumpit(skb, cb);
3794                 if (ret)
3795                         break;
3796         }
3797         cb->family = idx;
3798
3799         return skb->len ? : ret;
3800 }
3801
3802 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3803                                        unsigned int change,
3804                                        u32 event, gfp_t flags, int *new_nsid,
3805                                        int new_ifindex)
3806 {
3807         struct net *net = dev_net(dev);
3808         struct sk_buff *skb;
3809         int err = -ENOBUFS;
3810         size_t if_info_size;
3811
3812         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3813         if (skb == NULL)
3814                 goto errout;
3815
3816         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3817                                type, 0, 0, change, 0, 0, event,
3818                                new_nsid, new_ifindex, -1, flags);
3819         if (err < 0) {
3820                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3821                 WARN_ON(err == -EMSGSIZE);
3822                 kfree_skb(skb);
3823                 goto errout;
3824         }
3825         return skb;
3826 errout:
3827         if (err < 0)
3828                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3829         return NULL;
3830 }
3831
3832 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3833 {
3834         struct net *net = dev_net(dev);
3835
3836         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3837 }
3838
3839 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3840                                unsigned int change, u32 event,
3841                                gfp_t flags, int *new_nsid, int new_ifindex)
3842 {
3843         struct sk_buff *skb;
3844
3845         if (dev->reg_state != NETREG_REGISTERED)
3846                 return;
3847
3848         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3849                                      new_ifindex);
3850         if (skb)
3851                 rtmsg_ifinfo_send(skb, dev, flags);
3852 }
3853
3854 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3855                   gfp_t flags)
3856 {
3857         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3858                            NULL, 0);
3859 }
3860
3861 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3862                          gfp_t flags, int *new_nsid, int new_ifindex)
3863 {
3864         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3865                            new_nsid, new_ifindex);
3866 }
3867
3868 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3869                                    struct net_device *dev,
3870                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3871                                    int type, unsigned int flags,
3872                                    int nlflags, u16 ndm_state)
3873 {
3874         struct nlmsghdr *nlh;
3875         struct ndmsg *ndm;
3876
3877         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3878         if (!nlh)
3879                 return -EMSGSIZE;
3880
3881         ndm = nlmsg_data(nlh);
3882         ndm->ndm_family  = AF_BRIDGE;
3883         ndm->ndm_pad1    = 0;
3884         ndm->ndm_pad2    = 0;
3885         ndm->ndm_flags   = flags;
3886         ndm->ndm_type    = 0;
3887         ndm->ndm_ifindex = dev->ifindex;
3888         ndm->ndm_state   = ndm_state;
3889
3890         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3891                 goto nla_put_failure;
3892         if (vid)
3893                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3894                         goto nla_put_failure;
3895
3896         nlmsg_end(skb, nlh);
3897         return 0;
3898
3899 nla_put_failure:
3900         nlmsg_cancel(skb, nlh);
3901         return -EMSGSIZE;
3902 }
3903
3904 static inline size_t rtnl_fdb_nlmsg_size(void)
3905 {
3906         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3907                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3908                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3909                0;
3910 }
3911
3912 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3913                             u16 ndm_state)
3914 {
3915         struct net *net = dev_net(dev);
3916         struct sk_buff *skb;
3917         int err = -ENOBUFS;
3918
3919         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3920         if (!skb)
3921                 goto errout;
3922
3923         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3924                                       0, 0, type, NTF_SELF, 0, ndm_state);
3925         if (err < 0) {
3926                 kfree_skb(skb);
3927                 goto errout;
3928         }
3929
3930         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3931         return;
3932 errout:
3933         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3934 }
3935
3936 /*
3937  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3938  */
3939 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3940                      struct nlattr *tb[],
3941                      struct net_device *dev,
3942                      const unsigned char *addr, u16 vid,
3943                      u16 flags)
3944 {
3945         int err = -EINVAL;
3946
3947         /* If aging addresses are supported device will need to
3948          * implement its own handler for this.
3949          */
3950         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3951                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
3952                 return err;
3953         }
3954
3955         if (vid) {
3956                 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
3957                 return err;
3958         }
3959
3960         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3961                 err = dev_uc_add_excl(dev, addr);
3962         else if (is_multicast_ether_addr(addr))
3963                 err = dev_mc_add_excl(dev, addr);
3964
3965         /* Only return duplicate errors if NLM_F_EXCL is set */
3966         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3967                 err = 0;
3968
3969         return err;
3970 }
3971 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3972
3973 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3974                          struct netlink_ext_ack *extack)
3975 {
3976         u16 vid = 0;
3977
3978         if (vlan_attr) {
3979                 if (nla_len(vlan_attr) != sizeof(u16)) {
3980                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3981                         return -EINVAL;
3982                 }
3983
3984                 vid = nla_get_u16(vlan_attr);
3985
3986                 if (!vid || vid >= VLAN_VID_MASK) {
3987                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3988                         return -EINVAL;
3989                 }
3990         }
3991         *p_vid = vid;
3992         return 0;
3993 }
3994
3995 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3996                         struct netlink_ext_ack *extack)
3997 {
3998         struct net *net = sock_net(skb->sk);
3999         struct ndmsg *ndm;
4000         struct nlattr *tb[NDA_MAX+1];
4001         struct net_device *dev;
4002         u8 *addr;
4003         u16 vid;
4004         int err;
4005
4006         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4007                                      extack);
4008         if (err < 0)
4009                 return err;
4010
4011         ndm = nlmsg_data(nlh);
4012         if (ndm->ndm_ifindex == 0) {
4013                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4014                 return -EINVAL;
4015         }
4016
4017         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4018         if (dev == NULL) {
4019                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4020                 return -ENODEV;
4021         }
4022
4023         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4024                 NL_SET_ERR_MSG(extack, "invalid address");
4025                 return -EINVAL;
4026         }
4027
4028         if (dev->type != ARPHRD_ETHER) {
4029                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4030                 return -EINVAL;
4031         }
4032
4033         addr = nla_data(tb[NDA_LLADDR]);
4034
4035         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4036         if (err)
4037                 return err;
4038
4039         err = -EOPNOTSUPP;
4040
4041         /* Support fdb on master device the net/bridge default case */
4042         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4043             netif_is_bridge_port(dev)) {
4044                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4045                 const struct net_device_ops *ops = br_dev->netdev_ops;
4046
4047                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4048                                        nlh->nlmsg_flags, extack);
4049                 if (err)
4050                         goto out;
4051                 else
4052                         ndm->ndm_flags &= ~NTF_MASTER;
4053         }
4054
4055         /* Embedded bridge, macvlan, and any other device support */
4056         if ((ndm->ndm_flags & NTF_SELF)) {
4057                 if (dev->netdev_ops->ndo_fdb_add)
4058                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4059                                                            vid,
4060                                                            nlh->nlmsg_flags,
4061                                                            extack);
4062                 else
4063                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4064                                                nlh->nlmsg_flags);
4065
4066                 if (!err) {
4067                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4068                                         ndm->ndm_state);
4069                         ndm->ndm_flags &= ~NTF_SELF;
4070                 }
4071         }
4072 out:
4073         return err;
4074 }
4075
4076 /*
4077  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4078  */
4079 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4080                      struct nlattr *tb[],
4081                      struct net_device *dev,
4082                      const unsigned char *addr, u16 vid)
4083 {
4084         int err = -EINVAL;
4085
4086         /* If aging addresses are supported device will need to
4087          * implement its own handler for this.
4088          */
4089         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4090                 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4091                 return err;
4092         }
4093
4094         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4095                 err = dev_uc_del(dev, addr);
4096         else if (is_multicast_ether_addr(addr))
4097                 err = dev_mc_del(dev, addr);
4098
4099         return err;
4100 }
4101 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4102
4103 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4104                         struct netlink_ext_ack *extack)
4105 {
4106         struct net *net = sock_net(skb->sk);
4107         struct ndmsg *ndm;
4108         struct nlattr *tb[NDA_MAX+1];
4109         struct net_device *dev;
4110         __u8 *addr;
4111         int err;
4112         u16 vid;
4113
4114         if (!netlink_capable(skb, CAP_NET_ADMIN))
4115                 return -EPERM;
4116
4117         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4118                                      extack);
4119         if (err < 0)
4120                 return err;
4121
4122         ndm = nlmsg_data(nlh);
4123         if (ndm->ndm_ifindex == 0) {
4124                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4125                 return -EINVAL;
4126         }
4127
4128         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4129         if (dev == NULL) {
4130                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4131                 return -ENODEV;
4132         }
4133
4134         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4135                 NL_SET_ERR_MSG(extack, "invalid address");
4136                 return -EINVAL;
4137         }
4138
4139         if (dev->type != ARPHRD_ETHER) {
4140                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4141                 return -EINVAL;
4142         }
4143
4144         addr = nla_data(tb[NDA_LLADDR]);
4145
4146         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4147         if (err)
4148                 return err;
4149
4150         err = -EOPNOTSUPP;
4151
4152         /* Support fdb on master device the net/bridge default case */
4153         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4154             netif_is_bridge_port(dev)) {
4155                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4156                 const struct net_device_ops *ops = br_dev->netdev_ops;
4157
4158                 if (ops->ndo_fdb_del)
4159                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4160
4161                 if (err)
4162                         goto out;
4163                 else
4164                         ndm->ndm_flags &= ~NTF_MASTER;
4165         }
4166
4167         /* Embedded bridge, macvlan, and any other device support */
4168         if (ndm->ndm_flags & NTF_SELF) {
4169                 if (dev->netdev_ops->ndo_fdb_del)
4170                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4171                                                            vid);
4172                 else
4173                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4174
4175                 if (!err) {
4176                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4177                                         ndm->ndm_state);
4178                         ndm->ndm_flags &= ~NTF_SELF;
4179                 }
4180         }
4181 out:
4182         return err;
4183 }
4184
4185 static int nlmsg_populate_fdb(struct sk_buff *skb,
4186                               struct netlink_callback *cb,
4187                               struct net_device *dev,
4188                               int *idx,
4189                               struct netdev_hw_addr_list *list)
4190 {
4191         struct netdev_hw_addr *ha;
4192         int err;
4193         u32 portid, seq;
4194
4195         portid = NETLINK_CB(cb->skb).portid;
4196         seq = cb->nlh->nlmsg_seq;
4197
4198         list_for_each_entry(ha, &list->list, list) {
4199                 if (*idx < cb->args[2])
4200                         goto skip;
4201
4202                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4203                                               portid, seq,
4204                                               RTM_NEWNEIGH, NTF_SELF,
4205                                               NLM_F_MULTI, NUD_PERMANENT);
4206                 if (err < 0)
4207                         return err;
4208 skip:
4209                 *idx += 1;
4210         }
4211         return 0;
4212 }
4213
4214 /**
4215  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4216  * @skb: socket buffer to store message in
4217  * @cb: netlink callback
4218  * @dev: netdevice
4219  * @filter_dev: ignored
4220  * @idx: the number of FDB table entries dumped is added to *@idx
4221  *
4222  * Default netdevice operation to dump the existing unicast address list.
4223  * Returns number of addresses from list put in skb.
4224  */
4225 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4226                       struct netlink_callback *cb,
4227                       struct net_device *dev,
4228                       struct net_device *filter_dev,
4229                       int *idx)
4230 {
4231         int err;
4232
4233         if (dev->type != ARPHRD_ETHER)
4234                 return -EINVAL;
4235
4236         netif_addr_lock_bh(dev);
4237         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4238         if (err)
4239                 goto out;
4240         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4241 out:
4242         netif_addr_unlock_bh(dev);
4243         return err;
4244 }
4245 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4246
4247 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4248                                  int *br_idx, int *brport_idx,
4249                                  struct netlink_ext_ack *extack)
4250 {
4251         struct nlattr *tb[NDA_MAX + 1];
4252         struct ndmsg *ndm;
4253         int err, i;
4254
4255         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4256                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4257                 return -EINVAL;
4258         }
4259
4260         ndm = nlmsg_data(nlh);
4261         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4262             ndm->ndm_flags || ndm->ndm_type) {
4263                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4264                 return -EINVAL;
4265         }
4266
4267         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4268                                             NDA_MAX, NULL, extack);
4269         if (err < 0)
4270                 return err;
4271
4272         *brport_idx = ndm->ndm_ifindex;
4273         for (i = 0; i <= NDA_MAX; ++i) {
4274                 if (!tb[i])
4275                         continue;
4276
4277                 switch (i) {
4278                 case NDA_IFINDEX:
4279                         if (nla_len(tb[i]) != sizeof(u32)) {
4280                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4281                                 return -EINVAL;
4282                         }
4283                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4284                         break;
4285                 case NDA_MASTER:
4286                         if (nla_len(tb[i]) != sizeof(u32)) {
4287                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4288                                 return -EINVAL;
4289                         }
4290                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4291                         break;
4292                 default:
4293                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4294                         return -EINVAL;
4295                 }
4296         }
4297
4298         return 0;
4299 }
4300
4301 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4302                                  int *br_idx, int *brport_idx,
4303                                  struct netlink_ext_ack *extack)
4304 {
4305         struct nlattr *tb[IFLA_MAX+1];
4306         int err;
4307
4308         /* A hack to preserve kernel<->userspace interface.
4309          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4310          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4311          * So, check for ndmsg with an optional u32 attribute (not used here).
4312          * Fortunately these sizes don't conflict with the size of ifinfomsg
4313          * with an optional attribute.
4314          */
4315         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4316             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4317              nla_attr_size(sizeof(u32)))) {
4318                 struct ifinfomsg *ifm;
4319
4320                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4321                                              tb, IFLA_MAX, ifla_policy,
4322                                              extack);
4323                 if (err < 0) {
4324                         return -EINVAL;
4325                 } else if (err == 0) {
4326                         if (tb[IFLA_MASTER])
4327                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4328                 }
4329
4330                 ifm = nlmsg_data(nlh);
4331                 *brport_idx = ifm->ifi_index;
4332         }
4333         return 0;
4334 }
4335
4336 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4337 {
4338         struct net_device *dev;
4339         struct net_device *br_dev = NULL;
4340         const struct net_device_ops *ops = NULL;
4341         const struct net_device_ops *cops = NULL;
4342         struct net *net = sock_net(skb->sk);
4343         struct hlist_head *head;
4344         int brport_idx = 0;
4345         int br_idx = 0;
4346         int h, s_h;
4347         int idx = 0, s_idx;
4348         int err = 0;
4349         int fidx = 0;
4350
4351         if (cb->strict_check)
4352                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4353                                             cb->extack);
4354         else
4355                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4356                                             cb->extack);
4357         if (err < 0)
4358                 return err;
4359
4360         if (br_idx) {
4361                 br_dev = __dev_get_by_index(net, br_idx);
4362                 if (!br_dev)
4363                         return -ENODEV;
4364
4365                 ops = br_dev->netdev_ops;
4366         }
4367
4368         s_h = cb->args[0];
4369         s_idx = cb->args[1];
4370
4371         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4372                 idx = 0;
4373                 head = &net->dev_index_head[h];
4374                 hlist_for_each_entry(dev, head, index_hlist) {
4375
4376                         if (brport_idx && (dev->ifindex != brport_idx))
4377                                 continue;
4378
4379                         if (!br_idx) { /* user did not specify a specific bridge */
4380                                 if (netif_is_bridge_port(dev)) {
4381                                         br_dev = netdev_master_upper_dev_get(dev);
4382                                         cops = br_dev->netdev_ops;
4383                                 }
4384                         } else {
4385                                 if (dev != br_dev &&
4386                                     !netif_is_bridge_port(dev))
4387                                         continue;
4388
4389                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4390                                     !(dev->priv_flags & IFF_EBRIDGE))
4391                                         continue;
4392                                 cops = ops;
4393                         }
4394
4395                         if (idx < s_idx)
4396                                 goto cont;
4397
4398                         if (netif_is_bridge_port(dev)) {
4399                                 if (cops && cops->ndo_fdb_dump) {
4400                                         err = cops->ndo_fdb_dump(skb, cb,
4401                                                                 br_dev, dev,
4402                                                                 &fidx);
4403                                         if (err == -EMSGSIZE)
4404                                                 goto out;
4405                                 }
4406                         }
4407
4408                         if (dev->netdev_ops->ndo_fdb_dump)
4409                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4410                                                                     dev, NULL,
4411                                                                     &fidx);
4412                         else
4413                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4414                                                         &fidx);
4415                         if (err == -EMSGSIZE)
4416                                 goto out;
4417
4418                         cops = NULL;
4419
4420                         /* reset fdb offset to 0 for rest of the interfaces */
4421                         cb->args[2] = 0;
4422                         fidx = 0;
4423 cont:
4424                         idx++;
4425                 }
4426         }
4427
4428 out:
4429         cb->args[0] = h;
4430         cb->args[1] = idx;
4431         cb->args[2] = fidx;
4432
4433         return skb->len;
4434 }
4435
4436 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4437                                 struct nlattr **tb, u8 *ndm_flags,
4438                                 int *br_idx, int *brport_idx, u8 **addr,
4439                                 u16 *vid, struct netlink_ext_ack *extack)
4440 {
4441         struct ndmsg *ndm;
4442         int err, i;
4443
4444         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4445                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4446                 return -EINVAL;
4447         }
4448
4449         ndm = nlmsg_data(nlh);
4450         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4451             ndm->ndm_type) {
4452                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4453                 return -EINVAL;
4454         }
4455
4456         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4457                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4458                 return -EINVAL;
4459         }
4460
4461         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4462                                             NDA_MAX, nda_policy, extack);
4463         if (err < 0)
4464                 return err;
4465
4466         *ndm_flags = ndm->ndm_flags;
4467         *brport_idx = ndm->ndm_ifindex;
4468         for (i = 0; i <= NDA_MAX; ++i) {
4469                 if (!tb[i])
4470                         continue;
4471
4472                 switch (i) {
4473                 case NDA_MASTER:
4474                         *br_idx = nla_get_u32(tb[i]);
4475                         break;
4476                 case NDA_LLADDR:
4477                         if (nla_len(tb[i]) != ETH_ALEN) {
4478                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4479                                 return -EINVAL;
4480                         }
4481                         *addr = nla_data(tb[i]);
4482                         break;
4483                 case NDA_VLAN:
4484                         err = fdb_vid_parse(tb[i], vid, extack);
4485                         if (err)
4486                                 return err;
4487                         break;
4488                 case NDA_VNI:
4489                         break;
4490                 default:
4491                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4492                         return -EINVAL;
4493                 }
4494         }
4495
4496         return 0;
4497 }
4498
4499 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4500                         struct netlink_ext_ack *extack)
4501 {
4502         struct net_device *dev = NULL, *br_dev = NULL;
4503         const struct net_device_ops *ops = NULL;
4504         struct net *net = sock_net(in_skb->sk);
4505         struct nlattr *tb[NDA_MAX + 1];
4506         struct sk_buff *skb;
4507         int brport_idx = 0;
4508         u8 ndm_flags = 0;
4509         int br_idx = 0;
4510         u8 *addr = NULL;
4511         u16 vid = 0;
4512         int err;
4513
4514         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4515                                    &brport_idx, &addr, &vid, extack);
4516         if (err < 0)
4517                 return err;
4518
4519         if (!addr) {
4520                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4521                 return -EINVAL;
4522         }
4523
4524         if (brport_idx) {
4525                 dev = __dev_get_by_index(net, brport_idx);
4526                 if (!dev) {
4527                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4528                         return -ENODEV;
4529                 }
4530         }
4531
4532         if (br_idx) {
4533                 if (dev) {
4534                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4535                         return -EINVAL;
4536                 }
4537
4538                 br_dev = __dev_get_by_index(net, br_idx);
4539                 if (!br_dev) {
4540                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4541                         return -EINVAL;
4542                 }
4543                 ops = br_dev->netdev_ops;
4544         }
4545
4546         if (dev) {
4547                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4548                         if (!netif_is_bridge_port(dev)) {
4549                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4550                                 return -EINVAL;
4551                         }
4552                         br_dev = netdev_master_upper_dev_get(dev);
4553                         if (!br_dev) {
4554                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4555                                 return -EINVAL;
4556                         }
4557                         ops = br_dev->netdev_ops;
4558                 } else {
4559                         if (!(ndm_flags & NTF_SELF)) {
4560                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4561                                 return -EINVAL;
4562                         }
4563                         ops = dev->netdev_ops;
4564                 }
4565         }
4566
4567         if (!br_dev && !dev) {
4568                 NL_SET_ERR_MSG(extack, "No device specified");
4569                 return -ENODEV;
4570         }
4571
4572         if (!ops || !ops->ndo_fdb_get) {
4573                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4574                 return -EOPNOTSUPP;
4575         }
4576
4577         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4578         if (!skb)
4579                 return -ENOBUFS;
4580
4581         if (br_dev)
4582                 dev = br_dev;
4583         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4584                                NETLINK_CB(in_skb).portid,
4585                                nlh->nlmsg_seq, extack);
4586         if (err)
4587                 goto out;
4588
4589         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4590 out:
4591         kfree_skb(skb);
4592         return err;
4593 }
4594
4595 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4596                                unsigned int attrnum, unsigned int flag)
4597 {
4598         if (mask & flag)
4599                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4600         return 0;
4601 }
4602
4603 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4604                             struct net_device *dev, u16 mode,
4605                             u32 flags, u32 mask, int nlflags,
4606                             u32 filter_mask,
4607                             int (*vlan_fill)(struct sk_buff *skb,
4608                                              struct net_device *dev,
4609                                              u32 filter_mask))
4610 {
4611         struct nlmsghdr *nlh;
4612         struct ifinfomsg *ifm;
4613         struct nlattr *br_afspec;
4614         struct nlattr *protinfo;
4615         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4616         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4617         int err = 0;
4618
4619         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4620         if (nlh == NULL)
4621                 return -EMSGSIZE;
4622
4623         ifm = nlmsg_data(nlh);
4624         ifm->ifi_family = AF_BRIDGE;
4625         ifm->__ifi_pad = 0;
4626         ifm->ifi_type = dev->type;
4627         ifm->ifi_index = dev->ifindex;
4628         ifm->ifi_flags = dev_get_flags(dev);
4629         ifm->ifi_change = 0;
4630
4631
4632         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4633             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4634             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4635             (br_dev &&
4636              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4637             (dev->addr_len &&
4638              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4639             (dev->ifindex != dev_get_iflink(dev) &&
4640              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4641                 goto nla_put_failure;
4642
4643         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4644         if (!br_afspec)
4645                 goto nla_put_failure;
4646
4647         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4648                 nla_nest_cancel(skb, br_afspec);
4649                 goto nla_put_failure;
4650         }
4651
4652         if (mode != BRIDGE_MODE_UNDEF) {
4653                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4654                         nla_nest_cancel(skb, br_afspec);
4655                         goto nla_put_failure;
4656                 }
4657         }
4658         if (vlan_fill) {
4659                 err = vlan_fill(skb, dev, filter_mask);
4660                 if (err) {
4661                         nla_nest_cancel(skb, br_afspec);
4662                         goto nla_put_failure;
4663                 }
4664         }
4665         nla_nest_end(skb, br_afspec);
4666
4667         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4668         if (!protinfo)
4669                 goto nla_put_failure;
4670
4671         if (brport_nla_put_flag(skb, flags, mask,
4672                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4673             brport_nla_put_flag(skb, flags, mask,
4674                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4675             brport_nla_put_flag(skb, flags, mask,
4676                                 IFLA_BRPORT_FAST_LEAVE,
4677                                 BR_MULTICAST_FAST_LEAVE) ||
4678             brport_nla_put_flag(skb, flags, mask,
4679                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4680             brport_nla_put_flag(skb, flags, mask,
4681                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4682             brport_nla_put_flag(skb, flags, mask,
4683                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4684             brport_nla_put_flag(skb, flags, mask,
4685                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4686             brport_nla_put_flag(skb, flags, mask,
4687                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4688             brport_nla_put_flag(skb, flags, mask,
4689                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4690             brport_nla_put_flag(skb, flags, mask,
4691                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4692                 nla_nest_cancel(skb, protinfo);
4693                 goto nla_put_failure;
4694         }
4695
4696         nla_nest_end(skb, protinfo);
4697
4698         nlmsg_end(skb, nlh);
4699         return 0;
4700 nla_put_failure:
4701         nlmsg_cancel(skb, nlh);
4702         return err ? err : -EMSGSIZE;
4703 }
4704 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4705
4706 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4707                                     bool strict_check, u32 *filter_mask,
4708                                     struct netlink_ext_ack *extack)
4709 {
4710         struct nlattr *tb[IFLA_MAX+1];
4711         int err, i;
4712
4713         if (strict_check) {
4714                 struct ifinfomsg *ifm;
4715
4716                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4717                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4718                         return -EINVAL;
4719                 }
4720
4721                 ifm = nlmsg_data(nlh);
4722                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4723                     ifm->ifi_change || ifm->ifi_index) {
4724                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4725                         return -EINVAL;
4726                 }
4727
4728                 err = nlmsg_parse_deprecated_strict(nlh,
4729                                                     sizeof(struct ifinfomsg),
4730                                                     tb, IFLA_MAX, ifla_policy,
4731                                                     extack);
4732         } else {
4733                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4734                                              tb, IFLA_MAX, ifla_policy,
4735                                              extack);
4736         }
4737         if (err < 0)
4738                 return err;
4739
4740         /* new attributes should only be added with strict checking */
4741         for (i = 0; i <= IFLA_MAX; ++i) {
4742                 if (!tb[i])
4743                         continue;
4744
4745                 switch (i) {
4746                 case IFLA_EXT_MASK:
4747                         *filter_mask = nla_get_u32(tb[i]);
4748                         break;
4749                 default:
4750                         if (strict_check) {
4751                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4752                                 return -EINVAL;
4753                         }
4754                 }
4755         }
4756
4757         return 0;
4758 }
4759
4760 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4761 {
4762         const struct nlmsghdr *nlh = cb->nlh;
4763         struct net *net = sock_net(skb->sk);
4764         struct net_device *dev;
4765         int idx = 0;
4766         u32 portid = NETLINK_CB(cb->skb).portid;
4767         u32 seq = nlh->nlmsg_seq;
4768         u32 filter_mask = 0;
4769         int err;
4770
4771         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4772                                        cb->extack);
4773         if (err < 0 && cb->strict_check)
4774                 return err;
4775
4776         rcu_read_lock();
4777         for_each_netdev_rcu(net, dev) {
4778                 const struct net_device_ops *ops = dev->netdev_ops;
4779                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4780
4781                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4782                         if (idx >= cb->args[0]) {
4783                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4784                                                 skb, portid, seq, dev,
4785                                                 filter_mask, NLM_F_MULTI);
4786                                 if (err < 0 && err != -EOPNOTSUPP) {
4787                                         if (likely(skb->len))
4788                                                 break;
4789
4790                                         goto out_err;
4791                                 }
4792                         }
4793                         idx++;
4794                 }
4795
4796                 if (ops->ndo_bridge_getlink) {
4797                         if (idx >= cb->args[0]) {
4798                                 err = ops->ndo_bridge_getlink(skb, portid,
4799                                                               seq, dev,
4800                                                               filter_mask,
4801                                                               NLM_F_MULTI);
4802                                 if (err < 0 && err != -EOPNOTSUPP) {
4803                                         if (likely(skb->len))
4804                                                 break;
4805
4806                                         goto out_err;
4807                                 }
4808                         }
4809                         idx++;
4810                 }
4811         }
4812         err = skb->len;
4813 out_err:
4814         rcu_read_unlock();
4815         cb->args[0] = idx;
4816
4817         return err;
4818 }
4819
4820 static inline size_t bridge_nlmsg_size(void)
4821 {
4822         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4823                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4824                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4825                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4826                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4827                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4828                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4829                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4830                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4831                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4832                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4833 }
4834
4835 static int rtnl_bridge_notify(struct net_device *dev)
4836 {
4837         struct net *net = dev_net(dev);
4838         struct sk_buff *skb;
4839         int err = -EOPNOTSUPP;
4840
4841         if (!dev->netdev_ops->ndo_bridge_getlink)
4842                 return 0;
4843
4844         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4845         if (!skb) {
4846                 err = -ENOMEM;
4847                 goto errout;
4848         }
4849
4850         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4851         if (err < 0)
4852                 goto errout;
4853
4854         /* Notification info is only filled for bridge ports, not the bridge
4855          * device itself. Therefore, a zero notification length is valid and
4856          * should not result in an error.
4857          */
4858         if (!skb->len)
4859                 goto errout;
4860
4861         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4862         return 0;
4863 errout:
4864         WARN_ON(err == -EMSGSIZE);
4865         kfree_skb(skb);
4866         if (err)
4867                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4868         return err;
4869 }
4870
4871 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4872                                struct netlink_ext_ack *extack)
4873 {
4874         struct net *net = sock_net(skb->sk);
4875         struct ifinfomsg *ifm;
4876         struct net_device *dev;
4877         struct nlattr *br_spec, *attr = NULL;
4878         int rem, err = -EOPNOTSUPP;
4879         u16 flags = 0;
4880         bool have_flags = false;
4881
4882         if (nlmsg_len(nlh) < sizeof(*ifm))
4883                 return -EINVAL;
4884
4885         ifm = nlmsg_data(nlh);
4886         if (ifm->ifi_family != AF_BRIDGE)
4887                 return -EPFNOSUPPORT;
4888
4889         dev = __dev_get_by_index(net, ifm->ifi_index);
4890         if (!dev) {
4891                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4892                 return -ENODEV;
4893         }
4894
4895         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4896         if (br_spec) {
4897                 nla_for_each_nested(attr, br_spec, rem) {
4898                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4899                                 if (nla_len(attr) < sizeof(flags))
4900                                         return -EINVAL;
4901
4902                                 have_flags = true;
4903                                 flags = nla_get_u16(attr);
4904                                 break;
4905                         }
4906                 }
4907         }
4908
4909         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4910                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4911
4912                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4913                         err = -EOPNOTSUPP;
4914                         goto out;
4915                 }
4916
4917                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4918                                                              extack);
4919                 if (err)
4920                         goto out;
4921
4922                 flags &= ~BRIDGE_FLAGS_MASTER;
4923         }
4924
4925         if ((flags & BRIDGE_FLAGS_SELF)) {
4926                 if (!dev->netdev_ops->ndo_bridge_setlink)
4927                         err = -EOPNOTSUPP;
4928                 else
4929                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4930                                                                   flags,
4931                                                                   extack);
4932                 if (!err) {
4933                         flags &= ~BRIDGE_FLAGS_SELF;
4934
4935                         /* Generate event to notify upper layer of bridge
4936                          * change
4937                          */
4938                         err = rtnl_bridge_notify(dev);
4939                 }
4940         }
4941
4942         if (have_flags)
4943                 memcpy(nla_data(attr), &flags, sizeof(flags));
4944 out:
4945         return err;
4946 }
4947
4948 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4949                                struct netlink_ext_ack *extack)
4950 {
4951         struct net *net = sock_net(skb->sk);
4952         struct ifinfomsg *ifm;
4953         struct net_device *dev;
4954         struct nlattr *br_spec, *attr = NULL;
4955         int rem, err = -EOPNOTSUPP;
4956         u16 flags = 0;
4957         bool have_flags = false;
4958
4959         if (nlmsg_len(nlh) < sizeof(*ifm))
4960                 return -EINVAL;
4961
4962         ifm = nlmsg_data(nlh);
4963         if (ifm->ifi_family != AF_BRIDGE)
4964                 return -EPFNOSUPPORT;
4965
4966         dev = __dev_get_by_index(net, ifm->ifi_index);
4967         if (!dev) {
4968                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4969                 return -ENODEV;
4970         }
4971
4972         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4973         if (br_spec) {
4974                 nla_for_each_nested(attr, br_spec, rem) {
4975                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4976                                 if (nla_len(attr) < sizeof(flags))
4977                                         return -EINVAL;
4978
4979                                 have_flags = true;
4980                                 flags = nla_get_u16(attr);
4981                                 break;
4982                         }
4983                 }
4984         }
4985
4986         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4987                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4988
4989                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4990                         err = -EOPNOTSUPP;
4991                         goto out;
4992                 }
4993
4994                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4995                 if (err)
4996                         goto out;
4997
4998                 flags &= ~BRIDGE_FLAGS_MASTER;
4999         }
5000
5001         if ((flags & BRIDGE_FLAGS_SELF)) {
5002                 if (!dev->netdev_ops->ndo_bridge_dellink)
5003                         err = -EOPNOTSUPP;
5004                 else
5005                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5006                                                                   flags);
5007
5008                 if (!err) {
5009                         flags &= ~BRIDGE_FLAGS_SELF;
5010
5011                         /* Generate event to notify upper layer of bridge
5012                          * change
5013                          */
5014                         err = rtnl_bridge_notify(dev);
5015                 }
5016         }
5017
5018         if (have_flags)
5019                 memcpy(nla_data(attr), &flags, sizeof(flags));
5020 out:
5021         return err;
5022 }
5023
5024 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5025 {
5026         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5027                (!idxattr || idxattr == attrid);
5028 }
5029
5030 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5031 static int rtnl_get_offload_stats_attr_size(int attr_id)
5032 {
5033         switch (attr_id) {
5034         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5035                 return sizeof(struct rtnl_link_stats64);
5036         }
5037
5038         return 0;
5039 }
5040
5041 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5042                                   int *prividx)
5043 {
5044         struct nlattr *attr = NULL;
5045         int attr_id, size;
5046         void *attr_data;
5047         int err;
5048
5049         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5050               dev->netdev_ops->ndo_get_offload_stats))
5051                 return -ENODATA;
5052
5053         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5054              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5055                 if (attr_id < *prividx)
5056                         continue;
5057
5058                 size = rtnl_get_offload_stats_attr_size(attr_id);
5059                 if (!size)
5060                         continue;
5061
5062                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5063                         continue;
5064
5065                 attr = nla_reserve_64bit(skb, attr_id, size,
5066                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5067                 if (!attr)
5068                         goto nla_put_failure;
5069
5070                 attr_data = nla_data(attr);
5071                 memset(attr_data, 0, size);
5072                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5073                                                              attr_data);
5074                 if (err)
5075                         goto get_offload_stats_failure;
5076         }
5077
5078         if (!attr)
5079                 return -ENODATA;
5080
5081         *prividx = 0;
5082         return 0;
5083
5084 nla_put_failure:
5085         err = -EMSGSIZE;
5086 get_offload_stats_failure:
5087         *prividx = attr_id;
5088         return err;
5089 }
5090
5091 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5092 {
5093         int nla_size = 0;
5094         int attr_id;
5095         int size;
5096
5097         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5098               dev->netdev_ops->ndo_get_offload_stats))
5099                 return 0;
5100
5101         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5102              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5103                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5104                         continue;
5105                 size = rtnl_get_offload_stats_attr_size(attr_id);
5106                 nla_size += nla_total_size_64bit(size);
5107         }
5108
5109         if (nla_size != 0)
5110                 nla_size += nla_total_size(0);
5111
5112         return nla_size;
5113 }
5114
5115 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5116                                int type, u32 pid, u32 seq, u32 change,
5117                                unsigned int flags, unsigned int filter_mask,
5118                                int *idxattr, int *prividx)
5119 {
5120         struct if_stats_msg *ifsm;
5121         struct nlmsghdr *nlh;
5122         struct nlattr *attr;
5123         int s_prividx = *prividx;
5124         int err;
5125
5126         ASSERT_RTNL();
5127
5128         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5129         if (!nlh)
5130                 return -EMSGSIZE;
5131
5132         ifsm = nlmsg_data(nlh);
5133         ifsm->family = PF_UNSPEC;
5134         ifsm->pad1 = 0;
5135         ifsm->pad2 = 0;
5136         ifsm->ifindex = dev->ifindex;
5137         ifsm->filter_mask = filter_mask;
5138
5139         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5140                 struct rtnl_link_stats64 *sp;
5141
5142                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5143                                          sizeof(struct rtnl_link_stats64),
5144                                          IFLA_STATS_UNSPEC);
5145                 if (!attr)
5146                         goto nla_put_failure;
5147
5148                 sp = nla_data(attr);
5149                 dev_get_stats(dev, sp);
5150         }
5151
5152         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5153                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5154
5155                 if (ops && ops->fill_linkxstats) {
5156                         *idxattr = IFLA_STATS_LINK_XSTATS;
5157                         attr = nla_nest_start_noflag(skb,
5158                                                      IFLA_STATS_LINK_XSTATS);
5159                         if (!attr)
5160                                 goto nla_put_failure;
5161
5162                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5163                         nla_nest_end(skb, attr);
5164                         if (err)
5165                                 goto nla_put_failure;
5166                         *idxattr = 0;
5167                 }
5168         }
5169
5170         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5171                              *idxattr)) {
5172                 const struct rtnl_link_ops *ops = NULL;
5173                 const struct net_device *master;
5174
5175                 master = netdev_master_upper_dev_get(dev);
5176                 if (master)
5177                         ops = master->rtnl_link_ops;
5178                 if (ops && ops->fill_linkxstats) {
5179                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5180                         attr = nla_nest_start_noflag(skb,
5181                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5182                         if (!attr)
5183                                 goto nla_put_failure;
5184
5185                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5186                         nla_nest_end(skb, attr);
5187                         if (err)
5188                                 goto nla_put_failure;
5189                         *idxattr = 0;
5190                 }
5191         }
5192
5193         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5194                              *idxattr)) {
5195                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5196                 attr = nla_nest_start_noflag(skb,
5197                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5198                 if (!attr)
5199                         goto nla_put_failure;
5200
5201                 err = rtnl_get_offload_stats(skb, dev, prividx);
5202                 if (err == -ENODATA)
5203                         nla_nest_cancel(skb, attr);
5204                 else
5205                         nla_nest_end(skb, attr);
5206
5207                 if (err && err != -ENODATA)
5208                         goto nla_put_failure;
5209                 *idxattr = 0;
5210         }
5211
5212         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5213                 struct rtnl_af_ops *af_ops;
5214
5215                 *idxattr = IFLA_STATS_AF_SPEC;
5216                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5217                 if (!attr)
5218                         goto nla_put_failure;
5219
5220                 rcu_read_lock();
5221                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5222                         if (af_ops->fill_stats_af) {
5223                                 struct nlattr *af;
5224                                 int err;
5225
5226                                 af = nla_nest_start_noflag(skb,
5227                                                            af_ops->family);
5228                                 if (!af) {
5229                                         rcu_read_unlock();
5230                                         goto nla_put_failure;
5231                                 }
5232                                 err = af_ops->fill_stats_af(skb, dev);
5233
5234                                 if (err == -ENODATA) {
5235                                         nla_nest_cancel(skb, af);
5236                                 } else if (err < 0) {
5237                                         rcu_read_unlock();
5238                                         goto nla_put_failure;
5239                                 }
5240
5241                                 nla_nest_end(skb, af);
5242                         }
5243                 }
5244                 rcu_read_unlock();
5245
5246                 nla_nest_end(skb, attr);
5247
5248                 *idxattr = 0;
5249         }
5250
5251         nlmsg_end(skb, nlh);
5252
5253         return 0;
5254
5255 nla_put_failure:
5256         /* not a multi message or no progress mean a real error */
5257         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5258                 nlmsg_cancel(skb, nlh);
5259         else
5260                 nlmsg_end(skb, nlh);
5261
5262         return -EMSGSIZE;
5263 }
5264
5265 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5266                                   u32 filter_mask)
5267 {
5268         size_t size = 0;
5269
5270         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5271                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5272
5273         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5274                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5275                 int attr = IFLA_STATS_LINK_XSTATS;
5276
5277                 if (ops && ops->get_linkxstats_size) {
5278                         size += nla_total_size(ops->get_linkxstats_size(dev,
5279                                                                         attr));
5280                         /* for IFLA_STATS_LINK_XSTATS */
5281                         size += nla_total_size(0);
5282                 }
5283         }
5284
5285         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5286                 struct net_device *_dev = (struct net_device *)dev;
5287                 const struct rtnl_link_ops *ops = NULL;
5288                 const struct net_device *master;
5289
5290                 /* netdev_master_upper_dev_get can't take const */
5291                 master = netdev_master_upper_dev_get(_dev);
5292                 if (master)
5293                         ops = master->rtnl_link_ops;
5294                 if (ops && ops->get_linkxstats_size) {
5295                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5296
5297                         size += nla_total_size(ops->get_linkxstats_size(dev,
5298                                                                         attr));
5299                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5300                         size += nla_total_size(0);
5301                 }
5302         }
5303
5304         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5305                 size += rtnl_get_offload_stats_size(dev);
5306
5307         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5308                 struct rtnl_af_ops *af_ops;
5309
5310                 /* for IFLA_STATS_AF_SPEC */
5311                 size += nla_total_size(0);
5312
5313                 rcu_read_lock();
5314                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5315                         if (af_ops->get_stats_af_size) {
5316                                 size += nla_total_size(
5317                                         af_ops->get_stats_af_size(dev));
5318
5319                                 /* for AF_* */
5320                                 size += nla_total_size(0);
5321                         }
5322                 }
5323                 rcu_read_unlock();
5324         }
5325
5326         return size;
5327 }
5328
5329 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5330                                 bool is_dump, struct netlink_ext_ack *extack)
5331 {
5332         struct if_stats_msg *ifsm;
5333
5334         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5335                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5336                 return -EINVAL;
5337         }
5338
5339         if (!strict_check)
5340                 return 0;
5341
5342         ifsm = nlmsg_data(nlh);
5343
5344         /* only requests using strict checks can pass data to influence
5345          * the dump. The legacy exception is filter_mask.
5346          */
5347         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5348                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5349                 return -EINVAL;
5350         }
5351         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5352                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5353                 return -EINVAL;
5354         }
5355         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5356                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5357                 return -EINVAL;
5358         }
5359
5360         return 0;
5361 }
5362
5363 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5364                           struct netlink_ext_ack *extack)
5365 {
5366         struct net *net = sock_net(skb->sk);
5367         struct net_device *dev = NULL;
5368         int idxattr = 0, prividx = 0;
5369         struct if_stats_msg *ifsm;
5370         struct sk_buff *nskb;
5371         u32 filter_mask;
5372         int err;
5373
5374         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5375                                    false, extack);
5376         if (err)
5377                 return err;
5378
5379         ifsm = nlmsg_data(nlh);
5380         if (ifsm->ifindex > 0)
5381                 dev = __dev_get_by_index(net, ifsm->ifindex);
5382         else
5383                 return -EINVAL;
5384
5385         if (!dev)
5386                 return -ENODEV;
5387
5388         filter_mask = ifsm->filter_mask;
5389         if (!filter_mask)
5390                 return -EINVAL;
5391
5392         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5393         if (!nskb)
5394                 return -ENOBUFS;
5395
5396         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5397                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5398                                   0, filter_mask, &idxattr, &prividx);
5399         if (err < 0) {
5400                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5401                 WARN_ON(err == -EMSGSIZE);
5402                 kfree_skb(nskb);
5403         } else {
5404                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5405         }
5406
5407         return err;
5408 }
5409
5410 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5411 {
5412         struct netlink_ext_ack *extack = cb->extack;
5413         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5414         struct net *net = sock_net(skb->sk);
5415         unsigned int flags = NLM_F_MULTI;
5416         struct if_stats_msg *ifsm;
5417         struct hlist_head *head;
5418         struct net_device *dev;
5419         u32 filter_mask = 0;
5420         int idx = 0;
5421
5422         s_h = cb->args[0];
5423         s_idx = cb->args[1];
5424         s_idxattr = cb->args[2];
5425         s_prividx = cb->args[3];
5426
5427         cb->seq = net->dev_base_seq;
5428
5429         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5430         if (err)
5431                 return err;
5432
5433         ifsm = nlmsg_data(cb->nlh);
5434         filter_mask = ifsm->filter_mask;
5435         if (!filter_mask) {
5436                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5437                 return -EINVAL;
5438         }
5439
5440         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5441                 idx = 0;
5442                 head = &net->dev_index_head[h];
5443                 hlist_for_each_entry(dev, head, index_hlist) {
5444                         if (idx < s_idx)
5445                                 goto cont;
5446                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5447                                                   NETLINK_CB(cb->skb).portid,
5448                                                   cb->nlh->nlmsg_seq, 0,
5449                                                   flags, filter_mask,
5450                                                   &s_idxattr, &s_prividx);
5451                         /* If we ran out of room on the first message,
5452                          * we're in trouble
5453                          */
5454                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5455
5456                         if (err < 0)
5457                                 goto out;
5458                         s_prividx = 0;
5459                         s_idxattr = 0;
5460                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5461 cont:
5462                         idx++;
5463                 }
5464         }
5465 out:
5466         cb->args[3] = s_prividx;
5467         cb->args[2] = s_idxattr;
5468         cb->args[1] = idx;
5469         cb->args[0] = h;
5470
5471         return skb->len;
5472 }
5473
5474 /* Process one rtnetlink message. */
5475
5476 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5477                              struct netlink_ext_ack *extack)
5478 {
5479         struct net *net = sock_net(skb->sk);
5480         struct rtnl_link *link;
5481         struct module *owner;
5482         int err = -EOPNOTSUPP;
5483         rtnl_doit_func doit;
5484         unsigned int flags;
5485         int kind;
5486         int family;
5487         int type;
5488
5489         type = nlh->nlmsg_type;
5490         if (type > RTM_MAX)
5491                 return -EOPNOTSUPP;
5492
5493         type -= RTM_BASE;
5494
5495         /* All the messages must have at least 1 byte length */
5496         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5497                 return 0;
5498
5499         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5500         kind = type&3;
5501
5502         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5503                 return -EPERM;
5504
5505         rcu_read_lock();
5506         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5507                 struct sock *rtnl;
5508                 rtnl_dumpit_func dumpit;
5509                 u32 min_dump_alloc = 0;
5510
5511                 link = rtnl_get_link(family, type);
5512                 if (!link || !link->dumpit) {
5513                         family = PF_UNSPEC;
5514                         link = rtnl_get_link(family, type);
5515                         if (!link || !link->dumpit)
5516                                 goto err_unlock;
5517                 }
5518                 owner = link->owner;
5519                 dumpit = link->dumpit;
5520
5521                 if (type == RTM_GETLINK - RTM_BASE)
5522                         min_dump_alloc = rtnl_calcit(skb, nlh);
5523
5524                 err = 0;
5525                 /* need to do this before rcu_read_unlock() */
5526                 if (!try_module_get(owner))
5527                         err = -EPROTONOSUPPORT;
5528
5529                 rcu_read_unlock();
5530
5531                 rtnl = net->rtnl;
5532                 if (err == 0) {
5533                         struct netlink_dump_control c = {
5534                                 .dump           = dumpit,
5535                                 .min_dump_alloc = min_dump_alloc,
5536                                 .module         = owner,
5537                         };
5538                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5539                         /* netlink_dump_start() will keep a reference on
5540                          * module if dump is still in progress.
5541                          */
5542                         module_put(owner);
5543                 }
5544                 return err;
5545         }
5546
5547         link = rtnl_get_link(family, type);
5548         if (!link || !link->doit) {
5549                 family = PF_UNSPEC;
5550                 link = rtnl_get_link(PF_UNSPEC, type);
5551                 if (!link || !link->doit)
5552                         goto out_unlock;
5553         }
5554
5555         owner = link->owner;
5556         if (!try_module_get(owner)) {
5557                 err = -EPROTONOSUPPORT;
5558                 goto out_unlock;
5559         }
5560
5561         flags = link->flags;
5562         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5563                 doit = link->doit;
5564                 rcu_read_unlock();
5565                 if (doit)
5566                         err = doit(skb, nlh, extack);
5567                 module_put(owner);
5568                 return err;
5569         }
5570         rcu_read_unlock();
5571
5572         rtnl_lock();
5573         link = rtnl_get_link(family, type);
5574         if (link && link->doit)
5575                 err = link->doit(skb, nlh, extack);
5576         rtnl_unlock();
5577
5578         module_put(owner);
5579
5580         return err;
5581
5582 out_unlock:
5583         rcu_read_unlock();
5584         return err;
5585
5586 err_unlock:
5587         rcu_read_unlock();
5588         return -EOPNOTSUPP;
5589 }
5590
5591 static void rtnetlink_rcv(struct sk_buff *skb)
5592 {
5593         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5594 }
5595
5596 static int rtnetlink_bind(struct net *net, int group)
5597 {
5598         switch (group) {
5599         case RTNLGRP_IPV4_MROUTE_R:
5600         case RTNLGRP_IPV6_MROUTE_R:
5601                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5602                         return -EPERM;
5603                 break;
5604         }
5605         return 0;
5606 }
5607
5608 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5609 {
5610         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5611
5612         switch (event) {
5613         case NETDEV_REBOOT:
5614         case NETDEV_CHANGEMTU:
5615         case NETDEV_CHANGEADDR:
5616         case NETDEV_CHANGENAME:
5617         case NETDEV_FEAT_CHANGE:
5618         case NETDEV_BONDING_FAILOVER:
5619         case NETDEV_POST_TYPE_CHANGE:
5620         case NETDEV_NOTIFY_PEERS:
5621         case NETDEV_CHANGEUPPER:
5622         case NETDEV_RESEND_IGMP:
5623         case NETDEV_CHANGEINFODATA:
5624         case NETDEV_CHANGELOWERSTATE:
5625         case NETDEV_CHANGE_TX_QUEUE_LEN:
5626                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5627                                    GFP_KERNEL, NULL, 0);
5628                 break;
5629         default:
5630                 break;
5631         }
5632         return NOTIFY_DONE;
5633 }
5634
5635 static struct notifier_block rtnetlink_dev_notifier = {
5636         .notifier_call  = rtnetlink_event,
5637 };
5638
5639
5640 static int __net_init rtnetlink_net_init(struct net *net)
5641 {
5642         struct sock *sk;
5643         struct netlink_kernel_cfg cfg = {
5644                 .groups         = RTNLGRP_MAX,
5645                 .input          = rtnetlink_rcv,
5646                 .cb_mutex       = &rtnl_mutex,
5647                 .flags          = NL_CFG_F_NONROOT_RECV,
5648                 .bind           = rtnetlink_bind,
5649         };
5650
5651         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5652         if (!sk)
5653                 return -ENOMEM;
5654         net->rtnl = sk;
5655         return 0;
5656 }
5657
5658 static void __net_exit rtnetlink_net_exit(struct net *net)
5659 {
5660         netlink_kernel_release(net->rtnl);
5661         net->rtnl = NULL;
5662 }
5663
5664 static struct pernet_operations rtnetlink_net_ops = {
5665         .init = rtnetlink_net_init,
5666         .exit = rtnetlink_net_exit,
5667 };
5668
5669 void __init rtnetlink_init(void)
5670 {
5671         if (register_pernet_subsys(&rtnetlink_net_ops))
5672                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5673
5674         register_netdevice_notifier(&rtnetlink_dev_notifier);
5675
5676         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5677                       rtnl_dump_ifinfo, 0);
5678         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5679         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5680         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5681
5682         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5683         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5684         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5685
5686         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5687         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5688
5689         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5690         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5691         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5692
5693         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5694         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5695         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5696
5697         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5698                       0);
5699 }