net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     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 modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link __rcu **tab;
 290         struct rtnl_link *link;
 291         int msgindex;
 292
 293         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 294         msgindex = rtm_msgindex(msgtype);
 295
 296         rtnl_lock();
 297         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 298         if (!tab) {
 299                 rtnl_unlock();
 300                 return -ENOENT;
 301         }
 302
 303         link = rtnl_dereference(tab[msgindex]);
 304         rcu_assign_pointer(tab[msgindex], NULL);
 305         rtnl_unlock();
 306
 307         kfree_rcu(link, rcu);
 308
 309         return 0;
 310 }
 311 EXPORT_SYMBOL_GPL(rtnl_unregister);
 312
 313 /**
 314  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 315  * @protocol : Protocol family or PF_UNSPEC
 316  *
 317  * Identical to calling rtnl_unregster() for all registered message types
 318  * of a certain protocol family.
 319  */
 320 void rtnl_unregister_all(int protocol)
 321 {
 322         struct rtnl_link __rcu **tab;
 323         struct rtnl_link *link;
 324         int msgindex;
 325
 326         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 327
 328         rtnl_lock();
 329         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 330         if (!tab) {
 331                 rtnl_unlock();
 332                 return;
 333         }
 334         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 335         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 336                 link = rtnl_dereference(tab[msgindex]);
 337                 if (!link)
 338                         continue;
 339
 340                 rcu_assign_pointer(tab[msgindex], NULL);
 341                 kfree_rcu(link, rcu);
 342         }
 343         rtnl_unlock();
 344
 345         synchronize_net();
 346
 347         kfree(tab);
 348 }
 349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 350
 351 static LIST_HEAD(link_ops);
 352
 353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 354 {
 355         const struct rtnl_link_ops *ops;
 356
 357         list_for_each_entry(ops, &link_ops, list) {
 358                 if (!strcmp(ops->kind, kind))
 359                         return ops;
 360         }
 361         return NULL;
 362 }
 363
 364 /**
 365  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 366  * @ops: struct rtnl_link_ops * to register
 367  *
 368  * The caller must hold the rtnl_mutex. This function should be used
 369  * by drivers that create devices during module initialization. It
 370  * must be called before registering the devices.
 371  *
 372  * Returns 0 on success or a negative error code.
 373  */
 374 int __rtnl_link_register(struct rtnl_link_ops *ops)
 375 {
 376         if (rtnl_link_ops_get(ops->kind))
 377                 return -EEXIST;
 378
 379         /* The check for setup is here because if ops
 380          * does not have that filled up, it is not possible
 381          * to use the ops for creating device. So do not
 382          * fill up dellink as well. That disables rtnl_dellink.
 383          */
 384         if (ops->setup && !ops->dellink)
 385                 ops->dellink = unregister_netdevice_queue;
 386
 387         list_add_tail(&ops->list, &link_ops);
 388         return 0;
 389 }
 390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 391
 392 /**
 393  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 394  * @ops: struct rtnl_link_ops * to register
 395  *
 396  * Returns 0 on success or a negative error code.
 397  */
 398 int rtnl_link_register(struct rtnl_link_ops *ops)
 399 {
 400         int err;
 401
 402         /* Sanity-check max sizes to avoid stack buffer overflow. */
 403         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 404                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 405                 return -EINVAL;
 406
 407         rtnl_lock();
 408         err = __rtnl_link_register(ops);
 409         rtnl_unlock();
 410         return err;
 411 }
 412 EXPORT_SYMBOL_GPL(rtnl_link_register);
 413
 414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 415 {
 416         struct net_device *dev;
 417         LIST_HEAD(list_kill);
 418
 419         for_each_netdev(net, dev) {
 420                 if (dev->rtnl_link_ops == ops)
 421                         ops->dellink(dev, &list_kill);
 422         }
 423         unregister_netdevice_many(&list_kill);
 424 }
 425
 426 /**
 427  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 428  * @ops: struct rtnl_link_ops * to unregister
 429  *
 430  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 431  * integrity (hold pernet_ops_rwsem for writing to close the race
 432  * with setup_net() and cleanup_net()).
 433  */
 434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 435 {
 436         struct net *net;
 437
 438         for_each_net(net) {
 439                 __rtnl_kill_links(net, ops);
 440         }
 441         list_del(&ops->list);
 442 }
 443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 444
 445 /* Return with the rtnl_lock held when there are no network
 446  * devices unregistering in any network namespace.
 447  */
 448 static void rtnl_lock_unregistering_all(void)
 449 {
 450         struct net *net;
 451         bool unregistering;
 452         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 453
 454         add_wait_queue(&netdev_unregistering_wq, &wait);
 455         for (;;) {
 456                 unregistering = false;
 457                 rtnl_lock();
 458                 /* We held write locked pernet_ops_rwsem, and parallel
 459                  * setup_net() and cleanup_net() are not possible.
 460                  */
 461                 for_each_net(net) {
 462                         if (net->dev_unreg_count > 0) {
 463                                 unregistering = true;
 464                                 break;
 465                         }
 466                 }
 467                 if (!unregistering)
 468                         break;
 469                 __rtnl_unlock();
 470
 471                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 472         }
 473         remove_wait_queue(&netdev_unregistering_wq, &wait);
 474 }
 475
 476 /**
 477  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 478  * @ops: struct rtnl_link_ops * to unregister
 479  */
 480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 481 {
 482         /* Close the race with setup_net() and cleanup_net() */
 483         down_write(&pernet_ops_rwsem);
 484         rtnl_lock_unregistering_all();
 485         __rtnl_link_unregister(ops);
 486         rtnl_unlock();
 487         up_write(&pernet_ops_rwsem);
 488 }
 489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 490
 491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 492 {
 493         struct net_device *master_dev;
 494         const struct rtnl_link_ops *ops;
 495         size_t size = 0;
 496
 497         rcu_read_lock();
 498
 499         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 500         if (!master_dev)
 501                 goto out;
 502
 503         ops = master_dev->rtnl_link_ops;
 504         if (!ops || !ops->get_slave_size)
 505                 goto out;
 506         /* IFLA_INFO_SLAVE_DATA + nested data */
 507         size = nla_total_size(sizeof(struct nlattr)) +
 508                ops->get_slave_size(master_dev, dev);
 509
 510 out:
 511         rcu_read_unlock();
 512         return size;
 513 }
 514
 515 static size_t rtnl_link_get_size(const struct net_device *dev)
 516 {
 517         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 518         size_t size;
 519
 520         if (!ops)
 521                 return 0;
 522
 523         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 524                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 525
 526         if (ops->get_size)
 527                 /* IFLA_INFO_DATA + nested data */
 528                 size += nla_total_size(sizeof(struct nlattr)) +
 529                         ops->get_size(dev);
 530
 531         if (ops->get_xstats_size)
 532                 /* IFLA_INFO_XSTATS */
 533                 size += nla_total_size(ops->get_xstats_size(dev));
 534
 535         size += rtnl_link_get_slave_info_data_size(dev);
 536
 537         return size;
 538 }
 539
 540 static LIST_HEAD(rtnl_af_ops);
 541
 542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 543 {
 544         const struct rtnl_af_ops *ops;
 545
 546         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         nlmsg_end(skb, nlh);
1825         return 0;
1826
1827 nla_put_failure_rcu:
1828         rcu_read_unlock();
1829 nla_put_failure:
1830         nlmsg_cancel(skb, nlh);
1831         return -EMSGSIZE;
1832 }
1833
1834 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1835         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1836         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1837         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1838         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1839         [IFLA_MTU]              = { .type = NLA_U32 },
1840         [IFLA_LINK]             = { .type = NLA_U32 },
1841         [IFLA_MASTER]           = { .type = NLA_U32 },
1842         [IFLA_CARRIER]          = { .type = NLA_U8 },
1843         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1844         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1845         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1846         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1847         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1848         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1849         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1850         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1851          * allow 0-length string (needed to remove an alias).
1852          */
1853         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1854         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1855         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1856         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1857         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1858         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1859         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1860         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1861         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1862         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1863         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1864         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1865         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1866         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1867         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1868         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1869         [IFLA_XDP]              = { .type = NLA_NESTED },
1870         [IFLA_EVENT]            = { .type = NLA_U32 },
1871         [IFLA_GROUP]            = { .type = NLA_U32 },
1872         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1873         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1874         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1875         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1876         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1877         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1878         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1879                                     .len = ALTIFNAMSIZ - 1 },
1880         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1881         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1882         [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
1883 };
1884
1885 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1886         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1887         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1888         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1889         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1890 };
1891
1892 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1893         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1894         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1895         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1896         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1897         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1898         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1899         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1900         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1901         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1902         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1903         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1904         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1905         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1906 };
1907
1908 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1909         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1910         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1911                                     .len = PORT_PROFILE_MAX },
1912         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1913                                       .len = PORT_UUID_MAX },
1914         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1915                                     .len = PORT_UUID_MAX },
1916         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1917         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1918
1919         /* Unused, but we need to keep it here since user space could
1920          * fill it. It's also broken with regard to NLA_BINARY use in
1921          * combination with structs.
1922          */
1923         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1924                                     .len = sizeof(struct ifla_port_vsi) },
1925 };
1926
1927 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1928         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1929         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1930         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1931         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1932         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1933         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1934 };
1935
1936 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1937 {
1938         const struct rtnl_link_ops *ops = NULL;
1939         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1940
1941         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1942                 return NULL;
1943
1944         if (linfo[IFLA_INFO_KIND]) {
1945                 char kind[MODULE_NAME_LEN];
1946
1947                 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1948                 ops = rtnl_link_ops_get(kind);
1949         }
1950
1951         return ops;
1952 }
1953
1954 static bool link_master_filtered(struct net_device *dev, int master_idx)
1955 {
1956         struct net_device *master;
1957
1958         if (!master_idx)
1959                 return false;
1960
1961         master = netdev_master_upper_dev_get(dev);
1962         if (!master || master->ifindex != master_idx)
1963                 return true;
1964
1965         return false;
1966 }
1967
1968 static bool link_kind_filtered(const struct net_device *dev,
1969                                const struct rtnl_link_ops *kind_ops)
1970 {
1971         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1972                 return true;
1973
1974         return false;
1975 }
1976
1977 static bool link_dump_filtered(struct net_device *dev,
1978                                int master_idx,
1979                                const struct rtnl_link_ops *kind_ops)
1980 {
1981         if (link_master_filtered(dev, master_idx) ||
1982             link_kind_filtered(dev, kind_ops))
1983                 return true;
1984
1985         return false;
1986 }
1987
1988 /**
1989  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1990  * @sk: netlink socket
1991  * @netnsid: network namespace identifier
1992  *
1993  * Returns the network namespace identified by netnsid on success or an error
1994  * pointer on failure.
1995  */
1996 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1997 {
1998         struct net *net;
1999
2000         net = get_net_ns_by_id(sock_net(sk), netnsid);
2001         if (!net)
2002                 return ERR_PTR(-EINVAL);
2003
2004         /* For now, the caller is required to have CAP_NET_ADMIN in
2005          * the user namespace owning the target net ns.
2006          */
2007         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2008                 put_net(net);
2009                 return ERR_PTR(-EACCES);
2010         }
2011         return net;
2012 }
2013 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2014
2015 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2016                                       bool strict_check, struct nlattr **tb,
2017                                       struct netlink_ext_ack *extack)
2018 {
2019         int hdrlen;
2020
2021         if (strict_check) {
2022                 struct ifinfomsg *ifm;
2023
2024                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2025                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2026                         return -EINVAL;
2027                 }
2028
2029                 ifm = nlmsg_data(nlh);
2030                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2031                     ifm->ifi_change) {
2032                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2033                         return -EINVAL;
2034                 }
2035                 if (ifm->ifi_index) {
2036                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2037                         return -EINVAL;
2038                 }
2039
2040                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2041                                                      IFLA_MAX, ifla_policy,
2042                                                      extack);
2043         }
2044
2045         /* A hack to preserve kernel<->userspace interface.
2046          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2047          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2048          * what iproute2 < v3.9.0 used.
2049          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2050          * attribute, its netlink message is shorter than struct ifinfomsg.
2051          */
2052         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2053                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2054
2055         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2056                                       extack);
2057 }
2058
2059 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2060 {
2061         struct netlink_ext_ack *extack = cb->extack;
2062         const struct nlmsghdr *nlh = cb->nlh;
2063         struct net *net = sock_net(skb->sk);
2064         struct net *tgt_net = net;
2065         int h, s_h;
2066         int idx = 0, s_idx;
2067         struct net_device *dev;
2068         struct hlist_head *head;
2069         struct nlattr *tb[IFLA_MAX+1];
2070         u32 ext_filter_mask = 0;
2071         const struct rtnl_link_ops *kind_ops = NULL;
2072         unsigned int flags = NLM_F_MULTI;
2073         int master_idx = 0;
2074         int netnsid = -1;
2075         int err, i;
2076
2077         s_h = cb->args[0];
2078         s_idx = cb->args[1];
2079
2080         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2081         if (err < 0) {
2082                 if (cb->strict_check)
2083                         return err;
2084
2085                 goto walk_entries;
2086         }
2087
2088         for (i = 0; i <= IFLA_MAX; ++i) {
2089                 if (!tb[i])
2090                         continue;
2091
2092                 /* new attributes should only be added with strict checking */
2093                 switch (i) {
2094                 case IFLA_TARGET_NETNSID:
2095                         netnsid = nla_get_s32(tb[i]);
2096                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2097                         if (IS_ERR(tgt_net)) {
2098                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2099                                 return PTR_ERR(tgt_net);
2100                         }
2101                         break;
2102                 case IFLA_EXT_MASK:
2103                         ext_filter_mask = nla_get_u32(tb[i]);
2104                         break;
2105                 case IFLA_MASTER:
2106                         master_idx = nla_get_u32(tb[i]);
2107                         break;
2108                 case IFLA_LINKINFO:
2109                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2110                         break;
2111                 default:
2112                         if (cb->strict_check) {
2113                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2114                                 return -EINVAL;
2115                         }
2116                 }
2117         }
2118
2119         if (master_idx || kind_ops)
2120                 flags |= NLM_F_DUMP_FILTERED;
2121
2122 walk_entries:
2123         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2124                 idx = 0;
2125                 head = &tgt_net->dev_index_head[h];
2126                 hlist_for_each_entry(dev, head, index_hlist) {
2127                         if (link_dump_filtered(dev, master_idx, kind_ops))
2128                                 goto cont;
2129                         if (idx < s_idx)
2130                                 goto cont;
2131                         err = rtnl_fill_ifinfo(skb, dev, net,
2132                                                RTM_NEWLINK,
2133                                                NETLINK_CB(cb->skb).portid,
2134                                                nlh->nlmsg_seq, 0, flags,
2135                                                ext_filter_mask, 0, NULL, 0,
2136                                                netnsid, GFP_KERNEL);
2137
2138                         if (err < 0) {
2139                                 if (likely(skb->len))
2140                                         goto out;
2141
2142                                 goto out_err;
2143                         }
2144 cont:
2145                         idx++;
2146                 }
2147         }
2148 out:
2149         err = skb->len;
2150 out_err:
2151         cb->args[1] = idx;
2152         cb->args[0] = h;
2153         cb->seq = tgt_net->dev_base_seq;
2154         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2155         if (netnsid >= 0)
2156                 put_net(tgt_net);
2157
2158         return err;
2159 }
2160
2161 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2162                         struct netlink_ext_ack *exterr)
2163 {
2164         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2165                                     exterr);
2166 }
2167 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2168
2169 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2170 {
2171         struct net *net;
2172         /* Examine the link attributes and figure out which
2173          * network namespace we are talking about.
2174          */
2175         if (tb[IFLA_NET_NS_PID])
2176                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2177         else if (tb[IFLA_NET_NS_FD])
2178                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2179         else
2180                 net = get_net(src_net);
2181         return net;
2182 }
2183 EXPORT_SYMBOL(rtnl_link_get_net);
2184
2185 /* Figure out which network namespace we are talking about by
2186  * examining the link attributes in the following order:
2187  *
2188  * 1. IFLA_NET_NS_PID
2189  * 2. IFLA_NET_NS_FD
2190  * 3. IFLA_TARGET_NETNSID
2191  */
2192 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2193                                                struct nlattr *tb[])
2194 {
2195         struct net *net;
2196
2197         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2198                 return rtnl_link_get_net(src_net, tb);
2199
2200         if (!tb[IFLA_TARGET_NETNSID])
2201                 return get_net(src_net);
2202
2203         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2204         if (!net)
2205                 return ERR_PTR(-EINVAL);
2206
2207         return net;
2208 }
2209
2210 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2211                                              struct net *src_net,
2212                                              struct nlattr *tb[], int cap)
2213 {
2214         struct net *net;
2215
2216         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2217         if (IS_ERR(net))
2218                 return net;
2219
2220         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2221                 put_net(net);
2222                 return ERR_PTR(-EPERM);
2223         }
2224
2225         return net;
2226 }
2227
2228 /* Verify that rtnetlink requests do not pass additional properties
2229  * potentially referring to different network namespaces.
2230  */
2231 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2232                                     struct netlink_ext_ack *extack,
2233                                     bool netns_id_only)
2234 {
2235
2236         if (netns_id_only) {
2237                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2238                         return 0;
2239
2240                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2241                 return -EOPNOTSUPP;
2242         }
2243
2244         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2245                 goto invalid_attr;
2246
2247         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2248                 goto invalid_attr;
2249
2250         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2251                 goto invalid_attr;
2252
2253         return 0;
2254
2255 invalid_attr:
2256         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2257         return -EINVAL;
2258 }
2259
2260 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2261 {
2262         if (dev) {
2263                 if (tb[IFLA_ADDRESS] &&
2264                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2265                         return -EINVAL;
2266
2267                 if (tb[IFLA_BROADCAST] &&
2268                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2269                         return -EINVAL;
2270         }
2271
2272         if (tb[IFLA_AF_SPEC]) {
2273                 struct nlattr *af;
2274                 int rem, err;
2275
2276                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2277                         const struct rtnl_af_ops *af_ops;
2278
2279                         af_ops = rtnl_af_lookup(nla_type(af));
2280                         if (!af_ops)
2281                                 return -EAFNOSUPPORT;
2282
2283                         if (!af_ops->set_link_af)
2284                                 return -EOPNOTSUPP;
2285
2286                         if (af_ops->validate_link_af) {
2287                                 err = af_ops->validate_link_af(dev, af);
2288                                 if (err < 0)
2289                                         return err;
2290                         }
2291                 }
2292         }
2293
2294         return 0;
2295 }
2296
2297 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2298                                   int guid_type)
2299 {
2300         const struct net_device_ops *ops = dev->netdev_ops;
2301
2302         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2303 }
2304
2305 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2306 {
2307         if (dev->type != ARPHRD_INFINIBAND)
2308                 return -EOPNOTSUPP;
2309
2310         return handle_infiniband_guid(dev, ivt, guid_type);
2311 }
2312
2313 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2314 {
2315         const struct net_device_ops *ops = dev->netdev_ops;
2316         int err = -EINVAL;
2317
2318         if (tb[IFLA_VF_MAC]) {
2319                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2320
2321                 if (ivm->vf >= INT_MAX)
2322                         return -EINVAL;
2323                 err = -EOPNOTSUPP;
2324                 if (ops->ndo_set_vf_mac)
2325                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2326                                                   ivm->mac);
2327                 if (err < 0)
2328                         return err;
2329         }
2330
2331         if (tb[IFLA_VF_VLAN]) {
2332                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2333
2334                 if (ivv->vf >= INT_MAX)
2335                         return -EINVAL;
2336                 err = -EOPNOTSUPP;
2337                 if (ops->ndo_set_vf_vlan)
2338                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2339                                                    ivv->qos,
2340                                                    htons(ETH_P_8021Q));
2341                 if (err < 0)
2342                         return err;
2343         }
2344
2345         if (tb[IFLA_VF_VLAN_LIST]) {
2346                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2347                 struct nlattr *attr;
2348                 int rem, len = 0;
2349
2350                 err = -EOPNOTSUPP;
2351                 if (!ops->ndo_set_vf_vlan)
2352                         return err;
2353
2354                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2355                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2356                             nla_len(attr) < NLA_HDRLEN) {
2357                                 return -EINVAL;
2358                         }
2359                         if (len >= MAX_VLAN_LIST_LEN)
2360                                 return -EOPNOTSUPP;
2361                         ivvl[len] = nla_data(attr);
2362
2363                         len++;
2364                 }
2365                 if (len == 0)
2366                         return -EINVAL;
2367
2368                 if (ivvl[0]->vf >= INT_MAX)
2369                         return -EINVAL;
2370                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2371                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2372                 if (err < 0)
2373                         return err;
2374         }
2375
2376         if (tb[IFLA_VF_TX_RATE]) {
2377                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2378                 struct ifla_vf_info ivf;
2379
2380                 if (ivt->vf >= INT_MAX)
2381                         return -EINVAL;
2382                 err = -EOPNOTSUPP;
2383                 if (ops->ndo_get_vf_config)
2384                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2385                 if (err < 0)
2386                         return err;
2387
2388                 err = -EOPNOTSUPP;
2389                 if (ops->ndo_set_vf_rate)
2390                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2391                                                    ivf.min_tx_rate,
2392                                                    ivt->rate);
2393                 if (err < 0)
2394                         return err;
2395         }
2396
2397         if (tb[IFLA_VF_RATE]) {
2398                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2399
2400                 if (ivt->vf >= INT_MAX)
2401                         return -EINVAL;
2402                 err = -EOPNOTSUPP;
2403                 if (ops->ndo_set_vf_rate)
2404                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2405                                                    ivt->min_tx_rate,
2406                                                    ivt->max_tx_rate);
2407                 if (err < 0)
2408                         return err;
2409         }
2410
2411         if (tb[IFLA_VF_SPOOFCHK]) {
2412                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2413
2414                 if (ivs->vf >= INT_MAX)
2415                         return -EINVAL;
2416                 err = -EOPNOTSUPP;
2417                 if (ops->ndo_set_vf_spoofchk)
2418                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2419                                                        ivs->setting);
2420                 if (err < 0)
2421                         return err;
2422         }
2423
2424         if (tb[IFLA_VF_LINK_STATE]) {
2425                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2426
2427                 if (ivl->vf >= INT_MAX)
2428                         return -EINVAL;
2429                 err = -EOPNOTSUPP;
2430                 if (ops->ndo_set_vf_link_state)
2431                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2432                                                          ivl->link_state);
2433                 if (err < 0)
2434                         return err;
2435         }
2436
2437         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2438                 struct ifla_vf_rss_query_en *ivrssq_en;
2439
2440                 err = -EOPNOTSUPP;
2441                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2442                 if (ivrssq_en->vf >= INT_MAX)
2443                         return -EINVAL;
2444                 if (ops->ndo_set_vf_rss_query_en)
2445                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2446                                                            ivrssq_en->setting);
2447                 if (err < 0)
2448                         return err;
2449         }
2450
2451         if (tb[IFLA_VF_TRUST]) {
2452                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2453
2454                 if (ivt->vf >= INT_MAX)
2455                         return -EINVAL;
2456                 err = -EOPNOTSUPP;
2457                 if (ops->ndo_set_vf_trust)
2458                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2459                 if (err < 0)
2460                         return err;
2461         }
2462
2463         if (tb[IFLA_VF_IB_NODE_GUID]) {
2464                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2465
2466                 if (ivt->vf >= INT_MAX)
2467                         return -EINVAL;
2468                 if (!ops->ndo_set_vf_guid)
2469                         return -EOPNOTSUPP;
2470                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2471         }
2472
2473         if (tb[IFLA_VF_IB_PORT_GUID]) {
2474                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2475
2476                 if (ivt->vf >= INT_MAX)
2477                         return -EINVAL;
2478                 if (!ops->ndo_set_vf_guid)
2479                         return -EOPNOTSUPP;
2480
2481                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2482         }
2483
2484         return err;
2485 }
2486
2487 static int do_set_master(struct net_device *dev, int ifindex,
2488                          struct netlink_ext_ack *extack)
2489 {
2490         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2491         const struct net_device_ops *ops;
2492         int err;
2493
2494         if (upper_dev) {
2495                 if (upper_dev->ifindex == ifindex)
2496                         return 0;
2497                 ops = upper_dev->netdev_ops;
2498                 if (ops->ndo_del_slave) {
2499                         err = ops->ndo_del_slave(upper_dev, dev);
2500                         if (err)
2501                                 return err;
2502                 } else {
2503                         return -EOPNOTSUPP;
2504                 }
2505         }
2506
2507         if (ifindex) {
2508                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2509                 if (!upper_dev)
2510                         return -EINVAL;
2511                 ops = upper_dev->netdev_ops;
2512                 if (ops->ndo_add_slave) {
2513                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2514                         if (err)
2515                                 return err;
2516                 } else {
2517                         return -EOPNOTSUPP;
2518                 }
2519         }
2520         return 0;
2521 }
2522
2523 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2524         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2525         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2526 };
2527
2528 static int do_set_proto_down(struct net_device *dev,
2529                              struct nlattr *nl_proto_down,
2530                              struct nlattr *nl_proto_down_reason,
2531                              struct netlink_ext_ack *extack)
2532 {
2533         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2534         const struct net_device_ops *ops = dev->netdev_ops;
2535         unsigned long mask = 0;
2536         u32 value;
2537         bool proto_down;
2538         int err;
2539
2540         if (!ops->ndo_change_proto_down) {
2541                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2542                 return -EOPNOTSUPP;
2543         }
2544
2545         if (nl_proto_down_reason) {
2546                 err = nla_parse_nested_deprecated(pdreason,
2547                                                   IFLA_PROTO_DOWN_REASON_MAX,
2548                                                   nl_proto_down_reason,
2549                                                   ifla_proto_down_reason_policy,
2550                                                   NULL);
2551                 if (err < 0)
2552                         return err;
2553
2554                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2555                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2556                         return -EINVAL;
2557                 }
2558
2559                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2560
2561                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2562                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2563
2564                 dev_change_proto_down_reason(dev, mask, value);
2565         }
2566
2567         if (nl_proto_down) {
2568                 proto_down = nla_get_u8(nl_proto_down);
2569
2570                 /* Dont turn off protodown if there are active reasons */
2571                 if (!proto_down && dev->proto_down_reason) {
2572                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2573                         return -EBUSY;
2574                 }
2575                 err = dev_change_proto_down(dev,
2576                                             proto_down);
2577                 if (err)
2578                         return err;
2579         }
2580
2581         return 0;
2582 }
2583
2584 #define DO_SETLINK_MODIFIED     0x01
2585 /* notify flag means notify + modified. */
2586 #define DO_SETLINK_NOTIFY       0x03
2587 static int do_setlink(const struct sk_buff *skb,
2588                       struct net_device *dev, struct ifinfomsg *ifm,
2589                       struct netlink_ext_ack *extack,
2590                       struct nlattr **tb, char *ifname, int status)
2591 {
2592         const struct net_device_ops *ops = dev->netdev_ops;
2593         int err;
2594
2595         err = validate_linkmsg(dev, tb);
2596         if (err < 0)
2597                 return err;
2598
2599         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2600                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2601                 struct net *net;
2602                 int new_ifindex;
2603
2604                 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2605                                                 tb, CAP_NET_ADMIN);
2606                 if (IS_ERR(net)) {
2607                         err = PTR_ERR(net);
2608                         goto errout;
2609                 }
2610
2611                 if (tb[IFLA_NEW_IFINDEX])
2612                         new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2613                 else
2614                         new_ifindex = 0;
2615
2616                 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2617                 put_net(net);
2618                 if (err)
2619                         goto errout;
2620                 status |= DO_SETLINK_MODIFIED;
2621         }
2622
2623         if (tb[IFLA_MAP]) {
2624                 struct rtnl_link_ifmap *u_map;
2625                 struct ifmap k_map;
2626
2627                 if (!ops->ndo_set_config) {
2628                         err = -EOPNOTSUPP;
2629                         goto errout;
2630                 }
2631
2632                 if (!netif_device_present(dev)) {
2633                         err = -ENODEV;
2634                         goto errout;
2635                 }
2636
2637                 u_map = nla_data(tb[IFLA_MAP]);
2638                 k_map.mem_start = (unsigned long) u_map->mem_start;
2639                 k_map.mem_end = (unsigned long) u_map->mem_end;
2640                 k_map.base_addr = (unsigned short) u_map->base_addr;
2641                 k_map.irq = (unsigned char) u_map->irq;
2642                 k_map.dma = (unsigned char) u_map->dma;
2643                 k_map.port = (unsigned char) u_map->port;
2644
2645                 err = ops->ndo_set_config(dev, &k_map);
2646                 if (err < 0)
2647                         goto errout;
2648
2649                 status |= DO_SETLINK_NOTIFY;
2650         }
2651
2652         if (tb[IFLA_ADDRESS]) {
2653                 struct sockaddr *sa;
2654                 int len;
2655
2656                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2657                                                   sizeof(*sa));
2658                 sa = kmalloc(len, GFP_KERNEL);
2659                 if (!sa) {
2660                         err = -ENOMEM;
2661                         goto errout;
2662                 }
2663                 sa->sa_family = dev->type;
2664                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2665                        dev->addr_len);
2666                 err = dev_set_mac_address_user(dev, sa, extack);
2667                 kfree(sa);
2668                 if (err)
2669                         goto errout;
2670                 status |= DO_SETLINK_MODIFIED;
2671         }
2672
2673         if (tb[IFLA_MTU]) {
2674                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2675                 if (err < 0)
2676                         goto errout;
2677                 status |= DO_SETLINK_MODIFIED;
2678         }
2679
2680         if (tb[IFLA_GROUP]) {
2681                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2682                 status |= DO_SETLINK_NOTIFY;
2683         }
2684
2685         /*
2686          * Interface selected by interface index but interface
2687          * name provided implies that a name change has been
2688          * requested.
2689          */
2690         if (ifm->ifi_index > 0 && ifname[0]) {
2691                 err = dev_change_name(dev, ifname);
2692                 if (err < 0)
2693                         goto errout;
2694                 status |= DO_SETLINK_MODIFIED;
2695         }
2696
2697         if (tb[IFLA_IFALIAS]) {
2698                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2699                                     nla_len(tb[IFLA_IFALIAS]));
2700                 if (err < 0)
2701                         goto errout;
2702                 status |= DO_SETLINK_NOTIFY;
2703         }
2704
2705         if (tb[IFLA_BROADCAST]) {
2706                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2707                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2708         }
2709
2710         if (ifm->ifi_flags || ifm->ifi_change) {
2711                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2712                                        extack);
2713                 if (err < 0)
2714                         goto errout;
2715         }
2716
2717         if (tb[IFLA_MASTER]) {
2718                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2719                 if (err)
2720                         goto errout;
2721                 status |= DO_SETLINK_MODIFIED;
2722         }
2723
2724         if (tb[IFLA_CARRIER]) {
2725                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2726                 if (err)
2727                         goto errout;
2728                 status |= DO_SETLINK_MODIFIED;
2729         }
2730
2731         if (tb[IFLA_TXQLEN]) {
2732                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2733
2734                 err = dev_change_tx_queue_len(dev, value);
2735                 if (err)
2736                         goto errout;
2737                 status |= DO_SETLINK_MODIFIED;
2738         }
2739
2740         if (tb[IFLA_GSO_MAX_SIZE]) {
2741                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2742
2743                 if (max_size > GSO_MAX_SIZE) {
2744                         err = -EINVAL;
2745                         goto errout;
2746                 }
2747
2748                 if (dev->gso_max_size ^ max_size) {
2749                         netif_set_gso_max_size(dev, max_size);
2750                         status |= DO_SETLINK_MODIFIED;
2751                 }
2752         }
2753
2754         if (tb[IFLA_GSO_MAX_SEGS]) {
2755                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2756
2757                 if (max_segs > GSO_MAX_SEGS) {
2758                         err = -EINVAL;
2759                         goto errout;
2760                 }
2761
2762                 if (dev->gso_max_segs ^ max_segs) {
2763                         dev->gso_max_segs = max_segs;
2764                         status |= DO_SETLINK_MODIFIED;
2765                 }
2766         }
2767
2768         if (tb[IFLA_OPERSTATE])
2769                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2770
2771         if (tb[IFLA_LINKMODE]) {
2772                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2773
2774                 write_lock_bh(&dev_base_lock);
2775                 if (dev->link_mode ^ value)
2776                         status |= DO_SETLINK_NOTIFY;
2777                 dev->link_mode = value;
2778                 write_unlock_bh(&dev_base_lock);
2779         }
2780
2781         if (tb[IFLA_VFINFO_LIST]) {
2782                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2783                 struct nlattr *attr;
2784                 int rem;
2785
2786                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2787                         if (nla_type(attr) != IFLA_VF_INFO ||
2788                             nla_len(attr) < NLA_HDRLEN) {
2789                                 err = -EINVAL;
2790                                 goto errout;
2791                         }
2792                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2793                                                           attr,
2794                                                           ifla_vf_policy,
2795                                                           NULL);
2796                         if (err < 0)
2797                                 goto errout;
2798                         err = do_setvfinfo(dev, vfinfo);
2799                         if (err < 0)
2800                                 goto errout;
2801                         status |= DO_SETLINK_NOTIFY;
2802                 }
2803         }
2804         err = 0;
2805
2806         if (tb[IFLA_VF_PORTS]) {
2807                 struct nlattr *port[IFLA_PORT_MAX+1];
2808                 struct nlattr *attr;
2809                 int vf;
2810                 int rem;
2811
2812                 err = -EOPNOTSUPP;
2813                 if (!ops->ndo_set_vf_port)
2814                         goto errout;
2815
2816                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2817                         if (nla_type(attr) != IFLA_VF_PORT ||
2818                             nla_len(attr) < NLA_HDRLEN) {
2819                                 err = -EINVAL;
2820                                 goto errout;
2821                         }
2822                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2823                                                           attr,
2824                                                           ifla_port_policy,
2825                                                           NULL);
2826                         if (err < 0)
2827                                 goto errout;
2828                         if (!port[IFLA_PORT_VF]) {
2829                                 err = -EOPNOTSUPP;
2830                                 goto errout;
2831                         }
2832                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2833                         err = ops->ndo_set_vf_port(dev, vf, port);
2834                         if (err < 0)
2835                                 goto errout;
2836                         status |= DO_SETLINK_NOTIFY;
2837                 }
2838         }
2839         err = 0;
2840
2841         if (tb[IFLA_PORT_SELF]) {
2842                 struct nlattr *port[IFLA_PORT_MAX+1];
2843
2844                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2845                                                   tb[IFLA_PORT_SELF],
2846                                                   ifla_port_policy, NULL);
2847                 if (err < 0)
2848                         goto errout;
2849
2850                 err = -EOPNOTSUPP;
2851                 if (ops->ndo_set_vf_port)
2852                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2853                 if (err < 0)
2854                         goto errout;
2855                 status |= DO_SETLINK_NOTIFY;
2856         }
2857
2858         if (tb[IFLA_AF_SPEC]) {
2859                 struct nlattr *af;
2860                 int rem;
2861
2862                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2863                         const struct rtnl_af_ops *af_ops;
2864
2865                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2866
2867                         err = af_ops->set_link_af(dev, af, extack);
2868                         if (err < 0)
2869                                 goto errout;
2870
2871                         status |= DO_SETLINK_NOTIFY;
2872                 }
2873         }
2874         err = 0;
2875
2876         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2877                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2878                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2879                 if (err)
2880                         goto errout;
2881                 status |= DO_SETLINK_NOTIFY;
2882         }
2883
2884         if (tb[IFLA_XDP]) {
2885                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2886                 u32 xdp_flags = 0;
2887
2888                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2889                                                   tb[IFLA_XDP],
2890                                                   ifla_xdp_policy, NULL);
2891                 if (err < 0)
2892                         goto errout;
2893
2894                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2895                         err = -EINVAL;
2896                         goto errout;
2897                 }
2898
2899                 if (xdp[IFLA_XDP_FLAGS]) {
2900                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2901                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2902                                 err = -EINVAL;
2903                                 goto errout;
2904                         }
2905                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2906                                 err = -EINVAL;
2907                                 goto errout;
2908                         }
2909                 }
2910
2911                 if (xdp[IFLA_XDP_FD]) {
2912                         int expected_fd = -1;
2913
2914                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2915                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2916                                         err = -EINVAL;
2917                                         goto errout;
2918                                 }
2919                                 expected_fd =
2920                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2921                         }
2922
2923                         err = dev_change_xdp_fd(dev, extack,
2924                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2925                                                 expected_fd,
2926                                                 xdp_flags);
2927                         if (err)
2928                                 goto errout;
2929                         status |= DO_SETLINK_NOTIFY;
2930                 }
2931         }
2932
2933 errout:
2934         if (status & DO_SETLINK_MODIFIED) {
2935                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2936                         netdev_state_change(dev);
2937
2938                 if (err < 0)
2939                         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",
2940                                              dev->name);
2941         }
2942
2943         return err;
2944 }
2945
2946 static struct net_device *rtnl_dev_get(struct net *net,
2947                                        struct nlattr *ifname_attr,
2948                                        struct nlattr *altifname_attr,
2949                                        char *ifname)
2950 {
2951         char buffer[ALTIFNAMSIZ];
2952
2953         if (!ifname) {
2954                 ifname = buffer;
2955                 if (ifname_attr)
2956                         nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2957                 else if (altifname_attr)
2958                         nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2959                 else
2960                         return NULL;
2961         }
2962
2963         return __dev_get_by_name(net, ifname);
2964 }
2965
2966 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2967                         struct netlink_ext_ack *extack)
2968 {
2969         struct net *net = sock_net(skb->sk);
2970         struct ifinfomsg *ifm;
2971         struct net_device *dev;
2972         int err;
2973         struct nlattr *tb[IFLA_MAX+1];
2974         char ifname[IFNAMSIZ];
2975
2976         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2977                                      ifla_policy, extack);
2978         if (err < 0)
2979                 goto errout;
2980
2981         err = rtnl_ensure_unique_netns(tb, extack, false);
2982         if (err < 0)
2983                 goto errout;
2984
2985         if (tb[IFLA_IFNAME])
2986                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2987         else
2988                 ifname[0] = '\0';
2989
2990         err = -EINVAL;
2991         ifm = nlmsg_data(nlh);
2992         if (ifm->ifi_index > 0)
2993                 dev = __dev_get_by_index(net, ifm->ifi_index);
2994         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2995                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2996         else
2997                 goto errout;
2998
2999         if (dev == NULL) {
3000                 err = -ENODEV;
3001                 goto errout;
3002         }
3003
3004         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3005 errout:
3006         return err;
3007 }
3008
3009 static int rtnl_group_dellink(const struct net *net, int group)
3010 {
3011         struct net_device *dev, *aux;
3012         LIST_HEAD(list_kill);
3013         bool found = false;
3014
3015         if (!group)
3016                 return -EPERM;
3017
3018         for_each_netdev(net, dev) {
3019                 if (dev->group == group) {
3020                         const struct rtnl_link_ops *ops;
3021
3022                         found = true;
3023                         ops = dev->rtnl_link_ops;
3024                         if (!ops || !ops->dellink)
3025                                 return -EOPNOTSUPP;
3026                 }
3027         }
3028
3029         if (!found)
3030                 return -ENODEV;
3031
3032         for_each_netdev_safe(net, dev, aux) {
3033                 if (dev->group == group) {
3034                         const struct rtnl_link_ops *ops;
3035
3036                         ops = dev->rtnl_link_ops;
3037                         ops->dellink(dev, &list_kill);
3038                 }
3039         }
3040         unregister_netdevice_many(&list_kill);
3041
3042         return 0;
3043 }
3044
3045 int rtnl_delete_link(struct net_device *dev)
3046 {
3047         const struct rtnl_link_ops *ops;
3048         LIST_HEAD(list_kill);
3049
3050         ops = dev->rtnl_link_ops;
3051         if (!ops || !ops->dellink)
3052                 return -EOPNOTSUPP;
3053
3054         ops->dellink(dev, &list_kill);
3055         unregister_netdevice_many(&list_kill);
3056
3057         return 0;
3058 }
3059 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3060
3061 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3062                         struct netlink_ext_ack *extack)
3063 {
3064         struct net *net = sock_net(skb->sk);
3065         struct net *tgt_net = net;
3066         struct net_device *dev = NULL;
3067         struct ifinfomsg *ifm;
3068         struct nlattr *tb[IFLA_MAX+1];
3069         int err;
3070         int netnsid = -1;
3071
3072         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3073                                      ifla_policy, extack);
3074         if (err < 0)
3075                 return err;
3076
3077         err = rtnl_ensure_unique_netns(tb, extack, true);
3078         if (err < 0)
3079                 return err;
3080
3081         if (tb[IFLA_TARGET_NETNSID]) {
3082                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3083                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3084                 if (IS_ERR(tgt_net))
3085                         return PTR_ERR(tgt_net);
3086         }
3087
3088         err = -EINVAL;
3089         ifm = nlmsg_data(nlh);
3090         if (ifm->ifi_index > 0)
3091                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3092         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3093                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3094                                    tb[IFLA_ALT_IFNAME], NULL);
3095         else if (tb[IFLA_GROUP])
3096                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3097         else
3098                 goto out;
3099
3100         if (!dev) {
3101                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3102                         err = -ENODEV;
3103
3104                 goto out;
3105         }
3106
3107         err = rtnl_delete_link(dev);
3108
3109 out:
3110         if (netnsid >= 0)
3111                 put_net(tgt_net);
3112
3113         return err;
3114 }
3115
3116 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3117 {
3118         unsigned int old_flags;
3119         int err;
3120
3121         old_flags = dev->flags;
3122         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3123                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3124                                          NULL);
3125                 if (err < 0)
3126                         return err;
3127         }
3128
3129         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3130                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3131         } else {
3132                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3133                 __dev_notify_flags(dev, old_flags, ~0U);
3134         }
3135         return 0;
3136 }
3137 EXPORT_SYMBOL(rtnl_configure_link);
3138
3139 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3140                                     unsigned char name_assign_type,
3141                                     const struct rtnl_link_ops *ops,
3142                                     struct nlattr *tb[],
3143                                     struct netlink_ext_ack *extack)
3144 {
3145         struct net_device *dev;
3146         unsigned int num_tx_queues = 1;
3147         unsigned int num_rx_queues = 1;
3148
3149         if (tb[IFLA_NUM_TX_QUEUES])
3150                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3151         else if (ops->get_num_tx_queues)
3152                 num_tx_queues = ops->get_num_tx_queues();
3153
3154         if (tb[IFLA_NUM_RX_QUEUES])
3155                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3156         else if (ops->get_num_rx_queues)
3157                 num_rx_queues = ops->get_num_rx_queues();
3158
3159         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3160                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3161                 return ERR_PTR(-EINVAL);
3162         }
3163
3164         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3165                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3166                 return ERR_PTR(-EINVAL);
3167         }
3168
3169         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3170                                ops->setup, num_tx_queues, num_rx_queues);
3171         if (!dev)
3172                 return ERR_PTR(-ENOMEM);
3173
3174         dev_net_set(dev, net);
3175         dev->rtnl_link_ops = ops;
3176         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3177
3178         if (tb[IFLA_MTU]) {
3179                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3180                 int err;
3181
3182                 err = dev_validate_mtu(dev, mtu, extack);
3183                 if (err) {
3184                         free_netdev(dev);
3185                         return ERR_PTR(err);
3186                 }
3187                 dev->mtu = mtu;
3188         }
3189         if (tb[IFLA_ADDRESS]) {
3190                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3191                                 nla_len(tb[IFLA_ADDRESS]));
3192                 dev->addr_assign_type = NET_ADDR_SET;
3193         }
3194         if (tb[IFLA_BROADCAST])
3195                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3196                                 nla_len(tb[IFLA_BROADCAST]));
3197         if (tb[IFLA_TXQLEN])
3198                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3199         if (tb[IFLA_OPERSTATE])
3200                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3201         if (tb[IFLA_LINKMODE])
3202                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3203         if (tb[IFLA_GROUP])
3204                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3205         if (tb[IFLA_GSO_MAX_SIZE])
3206                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3207         if (tb[IFLA_GSO_MAX_SEGS])
3208                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3209
3210         return dev;
3211 }
3212 EXPORT_SYMBOL(rtnl_create_link);
3213
3214 static int rtnl_group_changelink(const struct sk_buff *skb,
3215                 struct net *net, int group,
3216                 struct ifinfomsg *ifm,
3217                 struct netlink_ext_ack *extack,
3218                 struct nlattr **tb)
3219 {
3220         struct net_device *dev, *aux;
3221         int err;
3222
3223         for_each_netdev_safe(net, dev, aux) {
3224                 if (dev->group == group) {
3225                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3226                         if (err < 0)
3227                                 return err;
3228                 }
3229         }
3230
3231         return 0;
3232 }
3233
3234 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3235                           struct nlattr **attr, struct netlink_ext_ack *extack)
3236 {
3237         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3238         unsigned char name_assign_type = NET_NAME_USER;
3239         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3240         const struct rtnl_link_ops *m_ops = NULL;
3241         struct net_device *master_dev = NULL;
3242         struct net *net = sock_net(skb->sk);
3243         const struct rtnl_link_ops *ops;
3244         struct nlattr *tb[IFLA_MAX + 1];
3245         struct net *dest_net, *link_net;
3246         struct nlattr **slave_data;
3247         char kind[MODULE_NAME_LEN];
3248         struct net_device *dev;
3249         struct ifinfomsg *ifm;
3250         char ifname[IFNAMSIZ];
3251         struct nlattr **data;
3252         int err;
3253
3254 #ifdef CONFIG_MODULES
3255 replay:
3256 #endif
3257         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3258                                      ifla_policy, extack);
3259         if (err < 0)
3260                 return err;
3261
3262         err = rtnl_ensure_unique_netns(tb, extack, false);
3263         if (err < 0)
3264                 return err;
3265
3266         if (tb[IFLA_IFNAME])
3267                 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3268         else
3269                 ifname[0] = '\0';
3270
3271         ifm = nlmsg_data(nlh);
3272         if (ifm->ifi_index > 0)
3273                 dev = __dev_get_by_index(net, ifm->ifi_index);
3274         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3275                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3276         else
3277                 dev = NULL;
3278
3279         if (dev) {
3280                 master_dev = netdev_master_upper_dev_get(dev);
3281                 if (master_dev)
3282                         m_ops = master_dev->rtnl_link_ops;
3283         }
3284
3285         err = validate_linkmsg(dev, tb);
3286         if (err < 0)
3287                 return err;
3288
3289         if (tb[IFLA_LINKINFO]) {
3290                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3291                                                   tb[IFLA_LINKINFO],
3292                                                   ifla_info_policy, NULL);
3293                 if (err < 0)
3294                         return err;
3295         } else
3296                 memset(linkinfo, 0, sizeof(linkinfo));
3297
3298         if (linkinfo[IFLA_INFO_KIND]) {
3299                 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3300                 ops = rtnl_link_ops_get(kind);
3301         } else {
3302                 kind[0] = '\0';
3303                 ops = NULL;
3304         }
3305
3306         data = NULL;
3307         if (ops) {
3308                 if (ops->maxtype > RTNL_MAX_TYPE)
3309                         return -EINVAL;
3310
3311                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3312                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3313                                                           linkinfo[IFLA_INFO_DATA],
3314                                                           ops->policy, extack);
3315                         if (err < 0)
3316                                 return err;
3317                         data = attr;
3318                 }
3319                 if (ops->validate) {
3320                         err = ops->validate(tb, data, extack);
3321                         if (err < 0)
3322                                 return err;
3323                 }
3324         }
3325
3326         slave_data = NULL;
3327         if (m_ops) {
3328                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3329                         return -EINVAL;
3330
3331                 if (m_ops->slave_maxtype &&
3332                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3333                         err = nla_parse_nested_deprecated(slave_attr,
3334                                                           m_ops->slave_maxtype,
3335                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3336                                                           m_ops->slave_policy,
3337                                                           extack);
3338                         if (err < 0)
3339                                 return err;
3340                         slave_data = slave_attr;
3341                 }
3342         }
3343
3344         if (dev) {
3345                 int status = 0;
3346
3347                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3348                         return -EEXIST;
3349                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3350                         return -EOPNOTSUPP;
3351
3352                 if (linkinfo[IFLA_INFO_DATA]) {
3353                         if (!ops || ops != dev->rtnl_link_ops ||
3354                             !ops->changelink)
3355                                 return -EOPNOTSUPP;
3356
3357                         err = ops->changelink(dev, tb, data, extack);
3358                         if (err < 0)
3359                                 return err;
3360                         status |= DO_SETLINK_NOTIFY;
3361                 }
3362
3363                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3364                         if (!m_ops || !m_ops->slave_changelink)
3365                                 return -EOPNOTSUPP;
3366
3367                         err = m_ops->slave_changelink(master_dev, dev, tb,
3368                                                       slave_data, extack);
3369                         if (err < 0)
3370                                 return err;
3371                         status |= DO_SETLINK_NOTIFY;
3372                 }
3373
3374                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3375         }
3376
3377         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3378                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3379                         return rtnl_group_changelink(skb, net,
3380                                                 nla_get_u32(tb[IFLA_GROUP]),
3381                                                 ifm, extack, tb);
3382                 return -ENODEV;
3383         }
3384
3385         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3386                 return -EOPNOTSUPP;
3387
3388         if (!ops) {
3389 #ifdef CONFIG_MODULES
3390                 if (kind[0]) {
3391                         __rtnl_unlock();
3392                         request_module("rtnl-link-%s", kind);
3393                         rtnl_lock();
3394                         ops = rtnl_link_ops_get(kind);
3395                         if (ops)
3396                                 goto replay;
3397                 }
3398 #endif
3399                 NL_SET_ERR_MSG(extack, "Unknown device type");
3400                 return -EOPNOTSUPP;
3401         }
3402
3403         if (!ops->setup)
3404                 return -EOPNOTSUPP;
3405
3406         if (!ifname[0]) {
3407                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3408                 name_assign_type = NET_NAME_ENUM;
3409         }
3410
3411         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3412         if (IS_ERR(dest_net))
3413                 return PTR_ERR(dest_net);
3414
3415         if (tb[IFLA_LINK_NETNSID]) {
3416                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3417
3418                 link_net = get_net_ns_by_id(dest_net, id);
3419                 if (!link_net) {
3420                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3421                         err =  -EINVAL;
3422                         goto out;
3423                 }
3424                 err = -EPERM;
3425                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3426                         goto out;
3427         } else {
3428                 link_net = NULL;
3429         }
3430
3431         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3432                                name_assign_type, ops, tb, extack);
3433         if (IS_ERR(dev)) {
3434                 err = PTR_ERR(dev);
3435                 goto out;
3436         }
3437
3438         dev->ifindex = ifm->ifi_index;
3439
3440         if (ops->newlink)
3441                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3442         else
3443                 err = register_netdevice(dev);
3444         if (err < 0) {
3445                 free_netdev(dev);
3446                 goto out;
3447         }
3448
3449         err = rtnl_configure_link(dev, ifm);
3450         if (err < 0)
3451                 goto out_unregister;
3452         if (link_net) {
3453                 err = dev_change_net_namespace(dev, dest_net, ifname);
3454                 if (err < 0)
3455                         goto out_unregister;
3456         }
3457         if (tb[IFLA_MASTER]) {
3458                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3459                 if (err)
3460                         goto out_unregister;
3461         }
3462 out:
3463         if (link_net)
3464                 put_net(link_net);
3465         put_net(dest_net);
3466         return err;
3467 out_unregister:
3468         if (ops->newlink) {
3469                 LIST_HEAD(list_kill);
3470
3471                 ops->dellink(dev, &list_kill);
3472                 unregister_netdevice_many(&list_kill);
3473         } else {
3474                 unregister_netdevice(dev);
3475         }
3476         goto out;
3477 }
3478
3479 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3480                         struct netlink_ext_ack *extack)
3481 {
3482         struct nlattr **attr;
3483         int ret;
3484
3485         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3486         if (!attr)
3487                 return -ENOMEM;
3488
3489         ret = __rtnl_newlink(skb, nlh, attr, extack);
3490         kfree(attr);
3491         return ret;
3492 }
3493
3494 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3495                                   const struct nlmsghdr *nlh,
3496                                   struct nlattr **tb,
3497                                   struct netlink_ext_ack *extack)
3498 {
3499         struct ifinfomsg *ifm;
3500         int i, err;
3501
3502         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3503                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3504                 return -EINVAL;
3505         }
3506
3507         if (!netlink_strict_get_check(skb))
3508                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3509                                               ifla_policy, extack);
3510
3511         ifm = nlmsg_data(nlh);
3512         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3513             ifm->ifi_change) {
3514                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3515                 return -EINVAL;
3516         }
3517
3518         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3519                                             ifla_policy, extack);
3520         if (err)
3521                 return err;
3522
3523         for (i = 0; i <= IFLA_MAX; i++) {
3524                 if (!tb[i])
3525                         continue;
3526
3527                 switch (i) {
3528                 case IFLA_IFNAME:
3529                 case IFLA_ALT_IFNAME:
3530                 case IFLA_EXT_MASK:
3531                 case IFLA_TARGET_NETNSID:
3532                         break;
3533                 default:
3534                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3535                         return -EINVAL;
3536                 }
3537         }
3538
3539         return 0;
3540 }
3541
3542 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3543                         struct netlink_ext_ack *extack)
3544 {
3545         struct net *net = sock_net(skb->sk);
3546         struct net *tgt_net = net;
3547         struct ifinfomsg *ifm;
3548         struct nlattr *tb[IFLA_MAX+1];
3549         struct net_device *dev = NULL;
3550         struct sk_buff *nskb;
3551         int netnsid = -1;
3552         int err;
3553         u32 ext_filter_mask = 0;
3554
3555         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3556         if (err < 0)
3557                 return err;
3558
3559         err = rtnl_ensure_unique_netns(tb, extack, true);
3560         if (err < 0)
3561                 return err;
3562
3563         if (tb[IFLA_TARGET_NETNSID]) {
3564                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3565                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3566                 if (IS_ERR(tgt_net))
3567                         return PTR_ERR(tgt_net);
3568         }
3569
3570         if (tb[IFLA_EXT_MASK])
3571                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3572
3573         err = -EINVAL;
3574         ifm = nlmsg_data(nlh);
3575         if (ifm->ifi_index > 0)
3576                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3577         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3578                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3579                                    tb[IFLA_ALT_IFNAME], NULL);
3580         else
3581                 goto out;
3582
3583         err = -ENODEV;
3584         if (dev == NULL)
3585                 goto out;
3586
3587         err = -ENOBUFS;
3588         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3589         if (nskb == NULL)
3590                 goto out;
3591
3592         err = rtnl_fill_ifinfo(nskb, dev, net,
3593                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3594                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3595                                0, NULL, 0, netnsid, GFP_KERNEL);
3596         if (err < 0) {
3597                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3598                 WARN_ON(err == -EMSGSIZE);
3599                 kfree_skb(nskb);
3600         } else
3601                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3602 out:
3603         if (netnsid >= 0)
3604                 put_net(tgt_net);
3605
3606         return err;
3607 }
3608
3609 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3610                            bool *changed, struct netlink_ext_ack *extack)
3611 {
3612         char *alt_ifname;
3613         int err;
3614
3615         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3616         if (err)
3617                 return err;
3618
3619         alt_ifname = nla_strdup(attr, GFP_KERNEL);
3620         if (!alt_ifname)
3621                 return -ENOMEM;
3622
3623         if (cmd == RTM_NEWLINKPROP) {
3624                 err = netdev_name_node_alt_create(dev, alt_ifname);
3625                 if (!err)
3626                         alt_ifname = NULL;
3627         } else if (cmd == RTM_DELLINKPROP) {
3628                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3629         } else {
3630                 WARN_ON_ONCE(1);
3631                 err = -EINVAL;
3632         }
3633
3634         kfree(alt_ifname);
3635         if (!err)
3636                 *changed = true;
3637         return err;
3638 }
3639
3640 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3641                          struct netlink_ext_ack *extack)
3642 {
3643         struct net *net = sock_net(skb->sk);
3644         struct nlattr *tb[IFLA_MAX + 1];
3645         struct net_device *dev;
3646         struct ifinfomsg *ifm;
3647         bool changed = false;
3648         struct nlattr *attr;
3649         int err, rem;
3650
3651         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3652         if (err)
3653                 return err;
3654
3655         err = rtnl_ensure_unique_netns(tb, extack, true);
3656         if (err)
3657                 return err;
3658
3659         ifm = nlmsg_data(nlh);
3660         if (ifm->ifi_index > 0)
3661                 dev = __dev_get_by_index(net, ifm->ifi_index);
3662         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3663                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3664                                    tb[IFLA_ALT_IFNAME], NULL);
3665         else
3666                 return -EINVAL;
3667
3668         if (!dev)
3669                 return -ENODEV;
3670
3671         if (!tb[IFLA_PROP_LIST])
3672                 return 0;
3673
3674         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3675                 switch (nla_type(attr)) {
3676                 case IFLA_ALT_IFNAME:
3677                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3678                         if (err)
3679                                 return err;
3680                         break;
3681                 }
3682         }
3683
3684         if (changed)
3685                 netdev_state_change(dev);
3686         return 0;
3687 }
3688
3689 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3690                             struct netlink_ext_ack *extack)
3691 {
3692         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3693 }
3694
3695 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3696                             struct netlink_ext_ack *extack)
3697 {
3698         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3699 }
3700
3701 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3702 {
3703         struct net *net = sock_net(skb->sk);
3704         size_t min_ifinfo_dump_size = 0;
3705         struct nlattr *tb[IFLA_MAX+1];
3706         u32 ext_filter_mask = 0;
3707         struct net_device *dev;
3708         int hdrlen;
3709
3710         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3711         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3712                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3713
3714         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3715                 if (tb[IFLA_EXT_MASK])
3716                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3717         }
3718
3719         if (!ext_filter_mask)
3720                 return NLMSG_GOODSIZE;
3721         /*
3722          * traverse the list of net devices and compute the minimum
3723          * buffer size based upon the filter mask.
3724          */
3725         rcu_read_lock();
3726         for_each_netdev_rcu(net, dev) {
3727                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3728                                            if_nlmsg_size(dev, ext_filter_mask));
3729         }
3730         rcu_read_unlock();
3731
3732         return nlmsg_total_size(min_ifinfo_dump_size);
3733 }
3734
3735 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3736 {
3737         int idx;
3738         int s_idx = cb->family;
3739         int type = cb->nlh->nlmsg_type - RTM_BASE;
3740         int ret = 0;
3741
3742         if (s_idx == 0)
3743                 s_idx = 1;
3744
3745         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3746                 struct rtnl_link __rcu **tab;
3747                 struct rtnl_link *link;
3748                 rtnl_dumpit_func dumpit;
3749
3750                 if (idx < s_idx || idx == PF_PACKET)
3751                         continue;
3752
3753                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3754                         continue;
3755
3756                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3757                 if (!tab)
3758                         continue;
3759
3760                 link = rcu_dereference_rtnl(tab[type]);
3761                 if (!link)
3762                         continue;
3763
3764                 dumpit = link->dumpit;
3765                 if (!dumpit)
3766                         continue;
3767
3768                 if (idx > s_idx) {
3769                         memset(&cb->args[0], 0, sizeof(cb->args));
3770                         cb->prev_seq = 0;
3771                         cb->seq = 0;
3772                 }
3773                 ret = dumpit(skb, cb);
3774                 if (ret)
3775                         break;
3776         }
3777         cb->family = idx;
3778
3779         return skb->len ? : ret;
3780 }
3781
3782 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3783                                        unsigned int change,
3784                                        u32 event, gfp_t flags, int *new_nsid,
3785                                        int new_ifindex)
3786 {
3787         struct net *net = dev_net(dev);
3788         struct sk_buff *skb;
3789         int err = -ENOBUFS;
3790         size_t if_info_size;
3791
3792         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3793         if (skb == NULL)
3794                 goto errout;
3795
3796         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3797                                type, 0, 0, change, 0, 0, event,
3798                                new_nsid, new_ifindex, -1, flags);
3799         if (err < 0) {
3800                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3801                 WARN_ON(err == -EMSGSIZE);
3802                 kfree_skb(skb);
3803                 goto errout;
3804         }
3805         return skb;
3806 errout:
3807         if (err < 0)
3808                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3809         return NULL;
3810 }
3811
3812 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3813 {
3814         struct net *net = dev_net(dev);
3815
3816         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3817 }
3818
3819 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3820                                unsigned int change, u32 event,
3821                                gfp_t flags, int *new_nsid, int new_ifindex)
3822 {
3823         struct sk_buff *skb;
3824
3825         if (dev->reg_state != NETREG_REGISTERED)
3826                 return;
3827
3828         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3829                                      new_ifindex);
3830         if (skb)
3831                 rtmsg_ifinfo_send(skb, dev, flags);
3832 }
3833
3834 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3835                   gfp_t flags)
3836 {
3837         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3838                            NULL, 0);
3839 }
3840
3841 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3842                          gfp_t flags, int *new_nsid, int new_ifindex)
3843 {
3844         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3845                            new_nsid, new_ifindex);
3846 }
3847
3848 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3849                                    struct net_device *dev,
3850                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3851                                    int type, unsigned int flags,
3852                                    int nlflags, u16 ndm_state)
3853 {
3854         struct nlmsghdr *nlh;
3855         struct ndmsg *ndm;
3856
3857         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3858         if (!nlh)
3859                 return -EMSGSIZE;
3860
3861         ndm = nlmsg_data(nlh);
3862         ndm->ndm_family  = AF_BRIDGE;
3863         ndm->ndm_pad1    = 0;
3864         ndm->ndm_pad2    = 0;
3865         ndm->ndm_flags   = flags;
3866         ndm->ndm_type    = 0;
3867         ndm->ndm_ifindex = dev->ifindex;
3868         ndm->ndm_state   = ndm_state;
3869
3870         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3871                 goto nla_put_failure;
3872         if (vid)
3873                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3874                         goto nla_put_failure;
3875
3876         nlmsg_end(skb, nlh);
3877         return 0;
3878
3879 nla_put_failure:
3880         nlmsg_cancel(skb, nlh);
3881         return -EMSGSIZE;
3882 }
3883
3884 static inline size_t rtnl_fdb_nlmsg_size(void)
3885 {
3886         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3887                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3888                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3889                0;
3890 }
3891
3892 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3893                             u16 ndm_state)
3894 {
3895         struct net *net = dev_net(dev);
3896         struct sk_buff *skb;
3897         int err = -ENOBUFS;
3898
3899         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3900         if (!skb)
3901                 goto errout;
3902
3903         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3904                                       0, 0, type, NTF_SELF, 0, ndm_state);
3905         if (err < 0) {
3906                 kfree_skb(skb);
3907                 goto errout;
3908         }
3909
3910         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3911         return;
3912 errout:
3913         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3914 }
3915
3916 /*
3917  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3918  */
3919 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3920                      struct nlattr *tb[],
3921                      struct net_device *dev,
3922                      const unsigned char *addr, u16 vid,
3923                      u16 flags)
3924 {
3925         int err = -EINVAL;
3926
3927         /* If aging addresses are supported device will need to
3928          * implement its own handler for this.
3929          */
3930         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3931                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3932                 return err;
3933         }
3934
3935         if (vid) {
3936                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3937                 return err;
3938         }
3939
3940         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3941                 err = dev_uc_add_excl(dev, addr);
3942         else if (is_multicast_ether_addr(addr))
3943                 err = dev_mc_add_excl(dev, addr);
3944
3945         /* Only return duplicate errors if NLM_F_EXCL is set */
3946         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3947                 err = 0;
3948
3949         return err;
3950 }
3951 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3952
3953 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3954                          struct netlink_ext_ack *extack)
3955 {
3956         u16 vid = 0;
3957
3958         if (vlan_attr) {
3959                 if (nla_len(vlan_attr) != sizeof(u16)) {
3960                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3961                         return -EINVAL;
3962                 }
3963
3964                 vid = nla_get_u16(vlan_attr);
3965
3966                 if (!vid || vid >= VLAN_VID_MASK) {
3967                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3968                         return -EINVAL;
3969                 }
3970         }
3971         *p_vid = vid;
3972         return 0;
3973 }
3974
3975 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3976                         struct netlink_ext_ack *extack)
3977 {
3978         struct net *net = sock_net(skb->sk);
3979         struct ndmsg *ndm;
3980         struct nlattr *tb[NDA_MAX+1];
3981         struct net_device *dev;
3982         u8 *addr;
3983         u16 vid;
3984         int err;
3985
3986         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
3987                                      extack);
3988         if (err < 0)
3989                 return err;
3990
3991         ndm = nlmsg_data(nlh);
3992         if (ndm->ndm_ifindex == 0) {
3993                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3994                 return -EINVAL;
3995         }
3996
3997         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3998         if (dev == NULL) {
3999                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4000                 return -ENODEV;
4001         }
4002
4003         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4004                 NL_SET_ERR_MSG(extack, "invalid address");
4005                 return -EINVAL;
4006         }
4007
4008         if (dev->type != ARPHRD_ETHER) {
4009                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4010                 return -EINVAL;
4011         }
4012
4013         addr = nla_data(tb[NDA_LLADDR]);
4014
4015         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4016         if (err)
4017                 return err;
4018
4019         err = -EOPNOTSUPP;
4020
4021         /* Support fdb on master device the net/bridge default case */
4022         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4023             netif_is_bridge_port(dev)) {
4024                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4025                 const struct net_device_ops *ops = br_dev->netdev_ops;
4026
4027                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4028                                        nlh->nlmsg_flags, extack);
4029                 if (err)
4030                         goto out;
4031                 else
4032                         ndm->ndm_flags &= ~NTF_MASTER;
4033         }
4034
4035         /* Embedded bridge, macvlan, and any other device support */
4036         if ((ndm->ndm_flags & NTF_SELF)) {
4037                 if (dev->netdev_ops->ndo_fdb_add)
4038                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4039                                                            vid,
4040                                                            nlh->nlmsg_flags,
4041                                                            extack);
4042                 else
4043                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4044                                                nlh->nlmsg_flags);
4045
4046                 if (!err) {
4047                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4048                                         ndm->ndm_state);
4049                         ndm->ndm_flags &= ~NTF_SELF;
4050                 }
4051         }
4052 out:
4053         return err;
4054 }
4055
4056 /*
4057  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4058  */
4059 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4060                      struct nlattr *tb[],
4061                      struct net_device *dev,
4062                      const unsigned char *addr, u16 vid)
4063 {
4064         int err = -EINVAL;
4065
4066         /* If aging addresses are supported device will need to
4067          * implement its own handler for this.
4068          */
4069         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4070                 pr_info("%s: FDB only supports static addresses\n", dev->name);
4071                 return err;
4072         }
4073
4074         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4075                 err = dev_uc_del(dev, addr);
4076         else if (is_multicast_ether_addr(addr))
4077                 err = dev_mc_del(dev, addr);
4078
4079         return err;
4080 }
4081 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4082
4083 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4084                         struct netlink_ext_ack *extack)
4085 {
4086         struct net *net = sock_net(skb->sk);
4087         struct ndmsg *ndm;
4088         struct nlattr *tb[NDA_MAX+1];
4089         struct net_device *dev;
4090         __u8 *addr;
4091         int err;
4092         u16 vid;
4093
4094         if (!netlink_capable(skb, CAP_NET_ADMIN))
4095                 return -EPERM;
4096
4097         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4098                                      extack);
4099         if (err < 0)
4100                 return err;
4101
4102         ndm = nlmsg_data(nlh);
4103         if (ndm->ndm_ifindex == 0) {
4104                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4105                 return -EINVAL;
4106         }
4107
4108         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4109         if (dev == NULL) {
4110                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4111                 return -ENODEV;
4112         }
4113
4114         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4115                 NL_SET_ERR_MSG(extack, "invalid address");
4116                 return -EINVAL;
4117         }
4118
4119         if (dev->type != ARPHRD_ETHER) {
4120                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4121                 return -EINVAL;
4122         }
4123
4124         addr = nla_data(tb[NDA_LLADDR]);
4125
4126         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4127         if (err)
4128                 return err;
4129
4130         err = -EOPNOTSUPP;
4131
4132         /* Support fdb on master device the net/bridge default case */
4133         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4134             netif_is_bridge_port(dev)) {
4135                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4136                 const struct net_device_ops *ops = br_dev->netdev_ops;
4137
4138                 if (ops->ndo_fdb_del)
4139                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4140
4141                 if (err)
4142                         goto out;
4143                 else
4144                         ndm->ndm_flags &= ~NTF_MASTER;
4145         }
4146
4147         /* Embedded bridge, macvlan, and any other device support */
4148         if (ndm->ndm_flags & NTF_SELF) {
4149                 if (dev->netdev_ops->ndo_fdb_del)
4150                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4151                                                            vid);
4152                 else
4153                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4154
4155                 if (!err) {
4156                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4157                                         ndm->ndm_state);
4158                         ndm->ndm_flags &= ~NTF_SELF;
4159                 }
4160         }
4161 out:
4162         return err;
4163 }
4164
4165 static int nlmsg_populate_fdb(struct sk_buff *skb,
4166                               struct netlink_callback *cb,
4167                               struct net_device *dev,
4168                               int *idx,
4169                               struct netdev_hw_addr_list *list)
4170 {
4171         struct netdev_hw_addr *ha;
4172         int err;
4173         u32 portid, seq;
4174
4175         portid = NETLINK_CB(cb->skb).portid;
4176         seq = cb->nlh->nlmsg_seq;
4177
4178         list_for_each_entry(ha, &list->list, list) {
4179                 if (*idx < cb->args[2])
4180                         goto skip;
4181
4182                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4183                                               portid, seq,
4184                                               RTM_NEWNEIGH, NTF_SELF,
4185                                               NLM_F_MULTI, NUD_PERMANENT);
4186                 if (err < 0)
4187                         return err;
4188 skip:
4189                 *idx += 1;
4190         }
4191         return 0;
4192 }
4193
4194 /**
4195  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4196  * @skb: socket buffer to store message in
4197  * @cb: netlink callback
4198  * @dev: netdevice
4199  * @filter_dev: ignored
4200  * @idx: the number of FDB table entries dumped is added to *@idx
4201  *
4202  * Default netdevice operation to dump the existing unicast address list.
4203  * Returns number of addresses from list put in skb.
4204  */
4205 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4206                       struct netlink_callback *cb,
4207                       struct net_device *dev,
4208                       struct net_device *filter_dev,
4209                       int *idx)
4210 {
4211         int err;
4212
4213         if (dev->type != ARPHRD_ETHER)
4214                 return -EINVAL;
4215
4216         netif_addr_lock_bh(dev);
4217         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4218         if (err)
4219                 goto out;
4220         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4221 out:
4222         netif_addr_unlock_bh(dev);
4223         return err;
4224 }
4225 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4226
4227 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4228                                  int *br_idx, int *brport_idx,
4229                                  struct netlink_ext_ack *extack)
4230 {
4231         struct nlattr *tb[NDA_MAX + 1];
4232         struct ndmsg *ndm;
4233         int err, i;
4234
4235         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4236                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4237                 return -EINVAL;
4238         }
4239
4240         ndm = nlmsg_data(nlh);
4241         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4242             ndm->ndm_flags || ndm->ndm_type) {
4243                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4244                 return -EINVAL;
4245         }
4246
4247         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4248                                             NDA_MAX, NULL, extack);
4249         if (err < 0)
4250                 return err;
4251
4252         *brport_idx = ndm->ndm_ifindex;
4253         for (i = 0; i <= NDA_MAX; ++i) {
4254                 if (!tb[i])
4255                         continue;
4256
4257                 switch (i) {
4258                 case NDA_IFINDEX:
4259                         if (nla_len(tb[i]) != sizeof(u32)) {
4260                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4261                                 return -EINVAL;
4262                         }
4263                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4264                         break;
4265                 case NDA_MASTER:
4266                         if (nla_len(tb[i]) != sizeof(u32)) {
4267                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4268                                 return -EINVAL;
4269                         }
4270                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4271                         break;
4272                 default:
4273                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4274                         return -EINVAL;
4275                 }
4276         }
4277
4278         return 0;
4279 }
4280
4281 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4282                                  int *br_idx, int *brport_idx,
4283                                  struct netlink_ext_ack *extack)
4284 {
4285         struct nlattr *tb[IFLA_MAX+1];
4286         int err;
4287
4288         /* A hack to preserve kernel<->userspace interface.
4289          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4290          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4291          * So, check for ndmsg with an optional u32 attribute (not used here).
4292          * Fortunately these sizes don't conflict with the size of ifinfomsg
4293          * with an optional attribute.
4294          */
4295         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4296             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4297              nla_attr_size(sizeof(u32)))) {
4298                 struct ifinfomsg *ifm;
4299
4300                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4301                                              tb, IFLA_MAX, ifla_policy,
4302                                              extack);
4303                 if (err < 0) {
4304                         return -EINVAL;
4305                 } else if (err == 0) {
4306                         if (tb[IFLA_MASTER])
4307                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4308                 }
4309
4310                 ifm = nlmsg_data(nlh);
4311                 *brport_idx = ifm->ifi_index;
4312         }
4313         return 0;
4314 }
4315
4316 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4317 {
4318         struct net_device *dev;
4319         struct net_device *br_dev = NULL;
4320         const struct net_device_ops *ops = NULL;
4321         const struct net_device_ops *cops = NULL;
4322         struct net *net = sock_net(skb->sk);
4323         struct hlist_head *head;
4324         int brport_idx = 0;
4325         int br_idx = 0;
4326         int h, s_h;
4327         int idx = 0, s_idx;
4328         int err = 0;
4329         int fidx = 0;
4330
4331         if (cb->strict_check)
4332                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4333                                             cb->extack);
4334         else
4335                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4336                                             cb->extack);
4337         if (err < 0)
4338                 return err;
4339
4340         if (br_idx) {
4341                 br_dev = __dev_get_by_index(net, br_idx);
4342                 if (!br_dev)
4343                         return -ENODEV;
4344
4345                 ops = br_dev->netdev_ops;
4346         }
4347
4348         s_h = cb->args[0];
4349         s_idx = cb->args[1];
4350
4351         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4352                 idx = 0;
4353                 head = &net->dev_index_head[h];
4354                 hlist_for_each_entry(dev, head, index_hlist) {
4355
4356                         if (brport_idx && (dev->ifindex != brport_idx))
4357                                 continue;
4358
4359                         if (!br_idx) { /* user did not specify a specific bridge */
4360                                 if (netif_is_bridge_port(dev)) {
4361                                         br_dev = netdev_master_upper_dev_get(dev);
4362                                         cops = br_dev->netdev_ops;
4363                                 }
4364                         } else {
4365                                 if (dev != br_dev &&
4366                                     !netif_is_bridge_port(dev))
4367                                         continue;
4368
4369                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4370                                     !(dev->priv_flags & IFF_EBRIDGE))
4371                                         continue;
4372                                 cops = ops;
4373                         }
4374
4375                         if (idx < s_idx)
4376                                 goto cont;
4377
4378                         if (netif_is_bridge_port(dev)) {
4379                                 if (cops && cops->ndo_fdb_dump) {
4380                                         err = cops->ndo_fdb_dump(skb, cb,
4381                                                                 br_dev, dev,
4382                                                                 &fidx);
4383                                         if (err == -EMSGSIZE)
4384                                                 goto out;
4385                                 }
4386                         }
4387
4388                         if (dev->netdev_ops->ndo_fdb_dump)
4389                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4390                                                                     dev, NULL,
4391                                                                     &fidx);
4392                         else
4393                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4394                                                         &fidx);
4395                         if (err == -EMSGSIZE)
4396                                 goto out;
4397
4398                         cops = NULL;
4399
4400                         /* reset fdb offset to 0 for rest of the interfaces */
4401                         cb->args[2] = 0;
4402                         fidx = 0;
4403 cont:
4404                         idx++;
4405                 }
4406         }
4407
4408 out:
4409         cb->args[0] = h;
4410         cb->args[1] = idx;
4411         cb->args[2] = fidx;
4412
4413         return skb->len;
4414 }
4415
4416 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4417                                 struct nlattr **tb, u8 *ndm_flags,
4418                                 int *br_idx, int *brport_idx, u8 **addr,
4419                                 u16 *vid, struct netlink_ext_ack *extack)
4420 {
4421         struct ndmsg *ndm;
4422         int err, i;
4423
4424         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4425                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4426                 return -EINVAL;
4427         }
4428
4429         ndm = nlmsg_data(nlh);
4430         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4431             ndm->ndm_type) {
4432                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4433                 return -EINVAL;
4434         }
4435
4436         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4437                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4438                 return -EINVAL;
4439         }
4440
4441         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4442                                             NDA_MAX, nda_policy, extack);
4443         if (err < 0)
4444                 return err;
4445
4446         *ndm_flags = ndm->ndm_flags;
4447         *brport_idx = ndm->ndm_ifindex;
4448         for (i = 0; i <= NDA_MAX; ++i) {
4449                 if (!tb[i])
4450                         continue;
4451
4452                 switch (i) {
4453                 case NDA_MASTER:
4454                         *br_idx = nla_get_u32(tb[i]);
4455                         break;
4456                 case NDA_LLADDR:
4457                         if (nla_len(tb[i]) != ETH_ALEN) {
4458                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4459                                 return -EINVAL;
4460                         }
4461                         *addr = nla_data(tb[i]);
4462                         break;
4463                 case NDA_VLAN:
4464                         err = fdb_vid_parse(tb[i], vid, extack);
4465                         if (err)
4466                                 return err;
4467                         break;
4468                 case NDA_VNI:
4469                         break;
4470                 default:
4471                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4472                         return -EINVAL;
4473                 }
4474         }
4475
4476         return 0;
4477 }
4478
4479 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4480                         struct netlink_ext_ack *extack)
4481 {
4482         struct net_device *dev = NULL, *br_dev = NULL;
4483         const struct net_device_ops *ops = NULL;
4484         struct net *net = sock_net(in_skb->sk);
4485         struct nlattr *tb[NDA_MAX + 1];
4486         struct sk_buff *skb;
4487         int brport_idx = 0;
4488         u8 ndm_flags = 0;
4489         int br_idx = 0;
4490         u8 *addr = NULL;
4491         u16 vid = 0;
4492         int err;
4493
4494         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4495                                    &brport_idx, &addr, &vid, extack);
4496         if (err < 0)
4497                 return err;
4498
4499         if (!addr) {
4500                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4501                 return -EINVAL;
4502         }
4503
4504         if (brport_idx) {
4505                 dev = __dev_get_by_index(net, brport_idx);
4506                 if (!dev) {
4507                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4508                         return -ENODEV;
4509                 }
4510         }
4511
4512         if (br_idx) {
4513                 if (dev) {
4514                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4515                         return -EINVAL;
4516                 }
4517
4518                 br_dev = __dev_get_by_index(net, br_idx);
4519                 if (!br_dev) {
4520                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4521                         return -EINVAL;
4522                 }
4523                 ops = br_dev->netdev_ops;
4524         }
4525
4526         if (dev) {
4527                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4528                         if (!netif_is_bridge_port(dev)) {
4529                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4530                                 return -EINVAL;
4531                         }
4532                         br_dev = netdev_master_upper_dev_get(dev);
4533                         if (!br_dev) {
4534                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4535                                 return -EINVAL;
4536                         }
4537                         ops = br_dev->netdev_ops;
4538                 } else {
4539                         if (!(ndm_flags & NTF_SELF)) {
4540                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4541                                 return -EINVAL;
4542                         }
4543                         ops = dev->netdev_ops;
4544                 }
4545         }
4546
4547         if (!br_dev && !dev) {
4548                 NL_SET_ERR_MSG(extack, "No device specified");
4549                 return -ENODEV;
4550         }
4551
4552         if (!ops || !ops->ndo_fdb_get) {
4553                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4554                 return -EOPNOTSUPP;
4555         }
4556
4557         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4558         if (!skb)
4559                 return -ENOBUFS;
4560
4561         if (br_dev)
4562                 dev = br_dev;
4563         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4564                                NETLINK_CB(in_skb).portid,
4565                                nlh->nlmsg_seq, extack);
4566         if (err)
4567                 goto out;
4568
4569         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4570 out:
4571         kfree_skb(skb);
4572         return err;
4573 }
4574
4575 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4576                                unsigned int attrnum, unsigned int flag)
4577 {
4578         if (mask & flag)
4579                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4580         return 0;
4581 }
4582
4583 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4584                             struct net_device *dev, u16 mode,
4585                             u32 flags, u32 mask, int nlflags,
4586                             u32 filter_mask,
4587                             int (*vlan_fill)(struct sk_buff *skb,
4588                                              struct net_device *dev,
4589                                              u32 filter_mask))
4590 {
4591         struct nlmsghdr *nlh;
4592         struct ifinfomsg *ifm;
4593         struct nlattr *br_afspec;
4594         struct nlattr *protinfo;
4595         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4596         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4597         int err = 0;
4598
4599         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4600         if (nlh == NULL)
4601                 return -EMSGSIZE;
4602
4603         ifm = nlmsg_data(nlh);
4604         ifm->ifi_family = AF_BRIDGE;
4605         ifm->__ifi_pad = 0;
4606         ifm->ifi_type = dev->type;
4607         ifm->ifi_index = dev->ifindex;
4608         ifm->ifi_flags = dev_get_flags(dev);
4609         ifm->ifi_change = 0;
4610
4611
4612         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4613             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4614             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4615             (br_dev &&
4616              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4617             (dev->addr_len &&
4618              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4619             (dev->ifindex != dev_get_iflink(dev) &&
4620              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4621                 goto nla_put_failure;
4622
4623         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4624         if (!br_afspec)
4625                 goto nla_put_failure;
4626
4627         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4628                 nla_nest_cancel(skb, br_afspec);
4629                 goto nla_put_failure;
4630         }
4631
4632         if (mode != BRIDGE_MODE_UNDEF) {
4633                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4634                         nla_nest_cancel(skb, br_afspec);
4635                         goto nla_put_failure;
4636                 }
4637         }
4638         if (vlan_fill) {
4639                 err = vlan_fill(skb, dev, filter_mask);
4640                 if (err) {
4641                         nla_nest_cancel(skb, br_afspec);
4642                         goto nla_put_failure;
4643                 }
4644         }
4645         nla_nest_end(skb, br_afspec);
4646
4647         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4648         if (!protinfo)
4649                 goto nla_put_failure;
4650
4651         if (brport_nla_put_flag(skb, flags, mask,
4652                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4653             brport_nla_put_flag(skb, flags, mask,
4654                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4655             brport_nla_put_flag(skb, flags, mask,
4656                                 IFLA_BRPORT_FAST_LEAVE,
4657                                 BR_MULTICAST_FAST_LEAVE) ||
4658             brport_nla_put_flag(skb, flags, mask,
4659                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4660             brport_nla_put_flag(skb, flags, mask,
4661                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4662             brport_nla_put_flag(skb, flags, mask,
4663                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4664             brport_nla_put_flag(skb, flags, mask,
4665                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4666             brport_nla_put_flag(skb, flags, mask,
4667                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4668             brport_nla_put_flag(skb, flags, mask,
4669                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4670             brport_nla_put_flag(skb, flags, mask,
4671                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4672                 nla_nest_cancel(skb, protinfo);
4673                 goto nla_put_failure;
4674         }
4675
4676         nla_nest_end(skb, protinfo);
4677
4678         nlmsg_end(skb, nlh);
4679         return 0;
4680 nla_put_failure:
4681         nlmsg_cancel(skb, nlh);
4682         return err ? err : -EMSGSIZE;
4683 }
4684 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4685
4686 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4687                                     bool strict_check, u32 *filter_mask,
4688                                     struct netlink_ext_ack *extack)
4689 {
4690         struct nlattr *tb[IFLA_MAX+1];
4691         int err, i;
4692
4693         if (strict_check) {
4694                 struct ifinfomsg *ifm;
4695
4696                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4697                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4698                         return -EINVAL;
4699                 }
4700
4701                 ifm = nlmsg_data(nlh);
4702                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4703                     ifm->ifi_change || ifm->ifi_index) {
4704                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4705                         return -EINVAL;
4706                 }
4707
4708                 err = nlmsg_parse_deprecated_strict(nlh,
4709                                                     sizeof(struct ifinfomsg),
4710                                                     tb, IFLA_MAX, ifla_policy,
4711                                                     extack);
4712         } else {
4713                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4714                                              tb, IFLA_MAX, ifla_policy,
4715                                              extack);
4716         }
4717         if (err < 0)
4718                 return err;
4719
4720         /* new attributes should only be added with strict checking */
4721         for (i = 0; i <= IFLA_MAX; ++i) {
4722                 if (!tb[i])
4723                         continue;
4724
4725                 switch (i) {
4726                 case IFLA_EXT_MASK:
4727                         *filter_mask = nla_get_u32(tb[i]);
4728                         break;
4729                 default:
4730                         if (strict_check) {
4731                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4732                                 return -EINVAL;
4733                         }
4734                 }
4735         }
4736
4737         return 0;
4738 }
4739
4740 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4741 {
4742         const struct nlmsghdr *nlh = cb->nlh;
4743         struct net *net = sock_net(skb->sk);
4744         struct net_device *dev;
4745         int idx = 0;
4746         u32 portid = NETLINK_CB(cb->skb).portid;
4747         u32 seq = nlh->nlmsg_seq;
4748         u32 filter_mask = 0;
4749         int err;
4750
4751         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4752                                        cb->extack);
4753         if (err < 0 && cb->strict_check)
4754                 return err;
4755
4756         rcu_read_lock();
4757         for_each_netdev_rcu(net, dev) {
4758                 const struct net_device_ops *ops = dev->netdev_ops;
4759                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4760
4761                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4762                         if (idx >= cb->args[0]) {
4763                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4764                                                 skb, portid, seq, dev,
4765                                                 filter_mask, NLM_F_MULTI);
4766                                 if (err < 0 && err != -EOPNOTSUPP) {
4767                                         if (likely(skb->len))
4768                                                 break;
4769
4770                                         goto out_err;
4771                                 }
4772                         }
4773                         idx++;
4774                 }
4775
4776                 if (ops->ndo_bridge_getlink) {
4777                         if (idx >= cb->args[0]) {
4778                                 err = ops->ndo_bridge_getlink(skb, portid,
4779                                                               seq, dev,
4780                                                               filter_mask,
4781                                                               NLM_F_MULTI);
4782                                 if (err < 0 && err != -EOPNOTSUPP) {
4783                                         if (likely(skb->len))
4784                                                 break;
4785
4786                                         goto out_err;
4787                                 }
4788                         }
4789                         idx++;
4790                 }
4791         }
4792         err = skb->len;
4793 out_err:
4794         rcu_read_unlock();
4795         cb->args[0] = idx;
4796
4797         return err;
4798 }
4799
4800 static inline size_t bridge_nlmsg_size(void)
4801 {
4802         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4803                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4804                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4805                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4806                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4807                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4808                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4809                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4810                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4811                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4812                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4813 }
4814
4815 static int rtnl_bridge_notify(struct net_device *dev)
4816 {
4817         struct net *net = dev_net(dev);
4818         struct sk_buff *skb;
4819         int err = -EOPNOTSUPP;
4820
4821         if (!dev->netdev_ops->ndo_bridge_getlink)
4822                 return 0;
4823
4824         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4825         if (!skb) {
4826                 err = -ENOMEM;
4827                 goto errout;
4828         }
4829
4830         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4831         if (err < 0)
4832                 goto errout;
4833
4834         /* Notification info is only filled for bridge ports, not the bridge
4835          * device itself. Therefore, a zero notification length is valid and
4836          * should not result in an error.
4837          */
4838         if (!skb->len)
4839                 goto errout;
4840
4841         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4842         return 0;
4843 errout:
4844         WARN_ON(err == -EMSGSIZE);
4845         kfree_skb(skb);
4846         if (err)
4847                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4848         return err;
4849 }
4850
4851 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4852                                struct netlink_ext_ack *extack)
4853 {
4854         struct net *net = sock_net(skb->sk);
4855         struct ifinfomsg *ifm;
4856         struct net_device *dev;
4857         struct nlattr *br_spec, *attr = NULL;
4858         int rem, err = -EOPNOTSUPP;
4859         u16 flags = 0;
4860         bool have_flags = false;
4861
4862         if (nlmsg_len(nlh) < sizeof(*ifm))
4863                 return -EINVAL;
4864
4865         ifm = nlmsg_data(nlh);
4866         if (ifm->ifi_family != AF_BRIDGE)
4867                 return -EPFNOSUPPORT;
4868
4869         dev = __dev_get_by_index(net, ifm->ifi_index);
4870         if (!dev) {
4871                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4872                 return -ENODEV;
4873         }
4874
4875         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4876         if (br_spec) {
4877                 nla_for_each_nested(attr, br_spec, rem) {
4878                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4879                                 if (nla_len(attr) < sizeof(flags))
4880                                         return -EINVAL;
4881
4882                                 have_flags = true;
4883                                 flags = nla_get_u16(attr);
4884                                 break;
4885                         }
4886                 }
4887         }
4888
4889         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4890                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4891
4892                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4893                         err = -EOPNOTSUPP;
4894                         goto out;
4895                 }
4896
4897                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4898                                                              extack);
4899                 if (err)
4900                         goto out;
4901
4902                 flags &= ~BRIDGE_FLAGS_MASTER;
4903         }
4904
4905         if ((flags & BRIDGE_FLAGS_SELF)) {
4906                 if (!dev->netdev_ops->ndo_bridge_setlink)
4907                         err = -EOPNOTSUPP;
4908                 else
4909                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4910                                                                   flags,
4911                                                                   extack);
4912                 if (!err) {
4913                         flags &= ~BRIDGE_FLAGS_SELF;
4914
4915                         /* Generate event to notify upper layer of bridge
4916                          * change
4917                          */
4918                         err = rtnl_bridge_notify(dev);
4919                 }
4920         }
4921
4922         if (have_flags)
4923                 memcpy(nla_data(attr), &flags, sizeof(flags));
4924 out:
4925         return err;
4926 }
4927
4928 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4929                                struct netlink_ext_ack *extack)
4930 {
4931         struct net *net = sock_net(skb->sk);
4932         struct ifinfomsg *ifm;
4933         struct net_device *dev;
4934         struct nlattr *br_spec, *attr = NULL;
4935         int rem, err = -EOPNOTSUPP;
4936         u16 flags = 0;
4937         bool have_flags = false;
4938
4939         if (nlmsg_len(nlh) < sizeof(*ifm))
4940                 return -EINVAL;
4941
4942         ifm = nlmsg_data(nlh);
4943         if (ifm->ifi_family != AF_BRIDGE)
4944                 return -EPFNOSUPPORT;
4945
4946         dev = __dev_get_by_index(net, ifm->ifi_index);
4947         if (!dev) {
4948                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4949                 return -ENODEV;
4950         }
4951
4952         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4953         if (br_spec) {
4954                 nla_for_each_nested(attr, br_spec, rem) {
4955                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4956                                 if (nla_len(attr) < sizeof(flags))
4957                                         return -EINVAL;
4958
4959                                 have_flags = true;
4960                                 flags = nla_get_u16(attr);
4961                                 break;
4962                         }
4963                 }
4964         }
4965
4966         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4967                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4968
4969                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4970                         err = -EOPNOTSUPP;
4971                         goto out;
4972                 }
4973
4974                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4975                 if (err)
4976                         goto out;
4977
4978                 flags &= ~BRIDGE_FLAGS_MASTER;
4979         }
4980
4981         if ((flags & BRIDGE_FLAGS_SELF)) {
4982                 if (!dev->netdev_ops->ndo_bridge_dellink)
4983                         err = -EOPNOTSUPP;
4984                 else
4985                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4986                                                                   flags);
4987
4988                 if (!err) {
4989                         flags &= ~BRIDGE_FLAGS_SELF;
4990
4991                         /* Generate event to notify upper layer of bridge
4992                          * change
4993                          */
4994                         err = rtnl_bridge_notify(dev);
4995                 }
4996         }
4997
4998         if (have_flags)
4999                 memcpy(nla_data(attr), &flags, sizeof(flags));
5000 out:
5001         return err;
5002 }
5003
5004 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5005 {
5006         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5007                (!idxattr || idxattr == attrid);
5008 }
5009
5010 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5011 static int rtnl_get_offload_stats_attr_size(int attr_id)
5012 {
5013         switch (attr_id) {
5014         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5015                 return sizeof(struct rtnl_link_stats64);
5016         }
5017
5018         return 0;
5019 }
5020
5021 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5022                                   int *prividx)
5023 {
5024         struct nlattr *attr = NULL;
5025         int attr_id, size;
5026         void *attr_data;
5027         int err;
5028
5029         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5030               dev->netdev_ops->ndo_get_offload_stats))
5031                 return -ENODATA;
5032
5033         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5034              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5035                 if (attr_id < *prividx)
5036                         continue;
5037
5038                 size = rtnl_get_offload_stats_attr_size(attr_id);
5039                 if (!size)
5040                         continue;
5041
5042                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5043                         continue;
5044
5045                 attr = nla_reserve_64bit(skb, attr_id, size,
5046                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5047                 if (!attr)
5048                         goto nla_put_failure;
5049
5050                 attr_data = nla_data(attr);
5051                 memset(attr_data, 0, size);
5052                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5053                                                              attr_data);
5054                 if (err)
5055                         goto get_offload_stats_failure;
5056         }
5057
5058         if (!attr)
5059                 return -ENODATA;
5060
5061         *prividx = 0;
5062         return 0;
5063
5064 nla_put_failure:
5065         err = -EMSGSIZE;
5066 get_offload_stats_failure:
5067         *prividx = attr_id;
5068         return err;
5069 }
5070
5071 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5072 {
5073         int nla_size = 0;
5074         int attr_id;
5075         int size;
5076
5077         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5078               dev->netdev_ops->ndo_get_offload_stats))
5079                 return 0;
5080
5081         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5082              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5083                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5084                         continue;
5085                 size = rtnl_get_offload_stats_attr_size(attr_id);
5086                 nla_size += nla_total_size_64bit(size);
5087         }
5088
5089         if (nla_size != 0)
5090                 nla_size += nla_total_size(0);
5091
5092         return nla_size;
5093 }
5094
5095 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5096                                int type, u32 pid, u32 seq, u32 change,
5097                                unsigned int flags, unsigned int filter_mask,
5098                                int *idxattr, int *prividx)
5099 {
5100         struct if_stats_msg *ifsm;
5101         struct nlmsghdr *nlh;
5102         struct nlattr *attr;
5103         int s_prividx = *prividx;
5104         int err;
5105
5106         ASSERT_RTNL();
5107
5108         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5109         if (!nlh)
5110                 return -EMSGSIZE;
5111
5112         ifsm = nlmsg_data(nlh);
5113         ifsm->family = PF_UNSPEC;
5114         ifsm->pad1 = 0;
5115         ifsm->pad2 = 0;
5116         ifsm->ifindex = dev->ifindex;
5117         ifsm->filter_mask = filter_mask;
5118
5119         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5120                 struct rtnl_link_stats64 *sp;
5121
5122                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5123                                          sizeof(struct rtnl_link_stats64),
5124                                          IFLA_STATS_UNSPEC);
5125                 if (!attr)
5126                         goto nla_put_failure;
5127
5128                 sp = nla_data(attr);
5129                 dev_get_stats(dev, sp);
5130         }
5131
5132         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5133                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5134
5135                 if (ops && ops->fill_linkxstats) {
5136                         *idxattr = IFLA_STATS_LINK_XSTATS;
5137                         attr = nla_nest_start_noflag(skb,
5138                                                      IFLA_STATS_LINK_XSTATS);
5139                         if (!attr)
5140                                 goto nla_put_failure;
5141
5142                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5143                         nla_nest_end(skb, attr);
5144                         if (err)
5145                                 goto nla_put_failure;
5146                         *idxattr = 0;
5147                 }
5148         }
5149
5150         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5151                              *idxattr)) {
5152                 const struct rtnl_link_ops *ops = NULL;
5153                 const struct net_device *master;
5154
5155                 master = netdev_master_upper_dev_get(dev);
5156                 if (master)
5157                         ops = master->rtnl_link_ops;
5158                 if (ops && ops->fill_linkxstats) {
5159                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5160                         attr = nla_nest_start_noflag(skb,
5161                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5162                         if (!attr)
5163                                 goto nla_put_failure;
5164
5165                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5166                         nla_nest_end(skb, attr);
5167                         if (err)
5168                                 goto nla_put_failure;
5169                         *idxattr = 0;
5170                 }
5171         }
5172
5173         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5174                              *idxattr)) {
5175                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5176                 attr = nla_nest_start_noflag(skb,
5177                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5178                 if (!attr)
5179                         goto nla_put_failure;
5180
5181                 err = rtnl_get_offload_stats(skb, dev, prividx);
5182                 if (err == -ENODATA)
5183                         nla_nest_cancel(skb, attr);
5184                 else
5185                         nla_nest_end(skb, attr);
5186
5187                 if (err && err != -ENODATA)
5188                         goto nla_put_failure;
5189                 *idxattr = 0;
5190         }
5191
5192         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5193                 struct rtnl_af_ops *af_ops;
5194
5195                 *idxattr = IFLA_STATS_AF_SPEC;
5196                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5197                 if (!attr)
5198                         goto nla_put_failure;
5199
5200                 rcu_read_lock();
5201                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5202                         if (af_ops->fill_stats_af) {
5203                                 struct nlattr *af;
5204                                 int err;
5205
5206                                 af = nla_nest_start_noflag(skb,
5207                                                            af_ops->family);
5208                                 if (!af) {
5209                                         rcu_read_unlock();
5210                                         goto nla_put_failure;
5211                                 }
5212                                 err = af_ops->fill_stats_af(skb, dev);
5213
5214                                 if (err == -ENODATA) {
5215                                         nla_nest_cancel(skb, af);
5216                                 } else if (err < 0) {
5217                                         rcu_read_unlock();
5218                                         goto nla_put_failure;
5219                                 }
5220
5221                                 nla_nest_end(skb, af);
5222                         }
5223                 }
5224                 rcu_read_unlock();
5225
5226                 nla_nest_end(skb, attr);
5227
5228                 *idxattr = 0;
5229         }
5230
5231         nlmsg_end(skb, nlh);
5232
5233         return 0;
5234
5235 nla_put_failure:
5236         /* not a multi message or no progress mean a real error */
5237         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5238                 nlmsg_cancel(skb, nlh);
5239         else
5240                 nlmsg_end(skb, nlh);
5241
5242         return -EMSGSIZE;
5243 }
5244
5245 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5246                                   u32 filter_mask)
5247 {
5248         size_t size = 0;
5249
5250         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5251                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5252
5253         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5254                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5255                 int attr = IFLA_STATS_LINK_XSTATS;
5256
5257                 if (ops && ops->get_linkxstats_size) {
5258                         size += nla_total_size(ops->get_linkxstats_size(dev,
5259                                                                         attr));
5260                         /* for IFLA_STATS_LINK_XSTATS */
5261                         size += nla_total_size(0);
5262                 }
5263         }
5264
5265         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5266                 struct net_device *_dev = (struct net_device *)dev;
5267                 const struct rtnl_link_ops *ops = NULL;
5268                 const struct net_device *master;
5269
5270                 /* netdev_master_upper_dev_get can't take const */
5271                 master = netdev_master_upper_dev_get(_dev);
5272                 if (master)
5273                         ops = master->rtnl_link_ops;
5274                 if (ops && ops->get_linkxstats_size) {
5275                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5276
5277                         size += nla_total_size(ops->get_linkxstats_size(dev,
5278                                                                         attr));
5279                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5280                         size += nla_total_size(0);
5281                 }
5282         }
5283
5284         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5285                 size += rtnl_get_offload_stats_size(dev);
5286
5287         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5288                 struct rtnl_af_ops *af_ops;
5289
5290                 /* for IFLA_STATS_AF_SPEC */
5291                 size += nla_total_size(0);
5292
5293                 rcu_read_lock();
5294                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5295                         if (af_ops->get_stats_af_size) {
5296                                 size += nla_total_size(
5297                                         af_ops->get_stats_af_size(dev));
5298
5299                                 /* for AF_* */
5300                                 size += nla_total_size(0);
5301                         }
5302                 }
5303                 rcu_read_unlock();
5304         }
5305
5306         return size;
5307 }
5308
5309 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5310                                 bool is_dump, struct netlink_ext_ack *extack)
5311 {
5312         struct if_stats_msg *ifsm;
5313
5314         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5315                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5316                 return -EINVAL;
5317         }
5318
5319         if (!strict_check)
5320                 return 0;
5321
5322         ifsm = nlmsg_data(nlh);
5323
5324         /* only requests using strict checks can pass data to influence
5325          * the dump. The legacy exception is filter_mask.
5326          */
5327         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5328                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5329                 return -EINVAL;
5330         }
5331         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5332                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5333                 return -EINVAL;
5334         }
5335         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5336                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5337                 return -EINVAL;
5338         }
5339
5340         return 0;
5341 }
5342
5343 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5344                           struct netlink_ext_ack *extack)
5345 {
5346         struct net *net = sock_net(skb->sk);
5347         struct net_device *dev = NULL;
5348         int idxattr = 0, prividx = 0;
5349         struct if_stats_msg *ifsm;
5350         struct sk_buff *nskb;
5351         u32 filter_mask;
5352         int err;
5353
5354         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5355                                    false, extack);
5356         if (err)
5357                 return err;
5358
5359         ifsm = nlmsg_data(nlh);
5360         if (ifsm->ifindex > 0)
5361                 dev = __dev_get_by_index(net, ifsm->ifindex);
5362         else
5363                 return -EINVAL;
5364
5365         if (!dev)
5366                 return -ENODEV;
5367
5368         filter_mask = ifsm->filter_mask;
5369         if (!filter_mask)
5370                 return -EINVAL;
5371
5372         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5373         if (!nskb)
5374                 return -ENOBUFS;
5375
5376         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5377                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5378                                   0, filter_mask, &idxattr, &prividx);
5379         if (err < 0) {
5380                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5381                 WARN_ON(err == -EMSGSIZE);
5382                 kfree_skb(nskb);
5383         } else {
5384                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5385         }
5386
5387         return err;
5388 }
5389
5390 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5391 {
5392         struct netlink_ext_ack *extack = cb->extack;
5393         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5394         struct net *net = sock_net(skb->sk);
5395         unsigned int flags = NLM_F_MULTI;
5396         struct if_stats_msg *ifsm;
5397         struct hlist_head *head;
5398         struct net_device *dev;
5399         u32 filter_mask = 0;
5400         int idx = 0;
5401
5402         s_h = cb->args[0];
5403         s_idx = cb->args[1];
5404         s_idxattr = cb->args[2];
5405         s_prividx = cb->args[3];
5406
5407         cb->seq = net->dev_base_seq;
5408
5409         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5410         if (err)
5411                 return err;
5412
5413         ifsm = nlmsg_data(cb->nlh);
5414         filter_mask = ifsm->filter_mask;
5415         if (!filter_mask) {
5416                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5417                 return -EINVAL;
5418         }
5419
5420         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5421                 idx = 0;
5422                 head = &net->dev_index_head[h];
5423                 hlist_for_each_entry(dev, head, index_hlist) {
5424                         if (idx < s_idx)
5425                                 goto cont;
5426                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5427                                                   NETLINK_CB(cb->skb).portid,
5428                                                   cb->nlh->nlmsg_seq, 0,
5429                                                   flags, filter_mask,
5430                                                   &s_idxattr, &s_prividx);
5431                         /* If we ran out of room on the first message,
5432                          * we're in trouble
5433                          */
5434                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5435
5436                         if (err < 0)
5437                                 goto out;
5438                         s_prividx = 0;
5439                         s_idxattr = 0;
5440                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5441 cont:
5442                         idx++;
5443                 }
5444         }
5445 out:
5446         cb->args[3] = s_prividx;
5447         cb->args[2] = s_idxattr;
5448         cb->args[1] = idx;
5449         cb->args[0] = h;
5450
5451         return skb->len;
5452 }
5453
5454 /* Process one rtnetlink message. */
5455
5456 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5457                              struct netlink_ext_ack *extack)
5458 {
5459         struct net *net = sock_net(skb->sk);
5460         struct rtnl_link *link;
5461         struct module *owner;
5462         int err = -EOPNOTSUPP;
5463         rtnl_doit_func doit;
5464         unsigned int flags;
5465         int kind;
5466         int family;
5467         int type;
5468
5469         type = nlh->nlmsg_type;
5470         if (type > RTM_MAX)
5471                 return -EOPNOTSUPP;
5472
5473         type -= RTM_BASE;
5474
5475         /* All the messages must have at least 1 byte length */
5476         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5477                 return 0;
5478
5479         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5480         kind = type&3;
5481
5482         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5483                 return -EPERM;
5484
5485         rcu_read_lock();
5486         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5487                 struct sock *rtnl;
5488                 rtnl_dumpit_func dumpit;
5489                 u32 min_dump_alloc = 0;
5490
5491                 link = rtnl_get_link(family, type);
5492                 if (!link || !link->dumpit) {
5493                         family = PF_UNSPEC;
5494                         link = rtnl_get_link(family, type);
5495                         if (!link || !link->dumpit)
5496                                 goto err_unlock;
5497                 }
5498                 owner = link->owner;
5499                 dumpit = link->dumpit;
5500
5501                 if (type == RTM_GETLINK - RTM_BASE)
5502                         min_dump_alloc = rtnl_calcit(skb, nlh);
5503
5504                 err = 0;
5505                 /* need to do this before rcu_read_unlock() */
5506                 if (!try_module_get(owner))
5507                         err = -EPROTONOSUPPORT;
5508
5509                 rcu_read_unlock();
5510
5511                 rtnl = net->rtnl;
5512                 if (err == 0) {
5513                         struct netlink_dump_control c = {
5514                                 .dump           = dumpit,
5515                                 .min_dump_alloc = min_dump_alloc,
5516                                 .module         = owner,
5517                         };
5518                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5519                         /* netlink_dump_start() will keep a reference on
5520                          * module if dump is still in progress.
5521                          */
5522                         module_put(owner);
5523                 }
5524                 return err;
5525         }
5526
5527         link = rtnl_get_link(family, type);
5528         if (!link || !link->doit) {
5529                 family = PF_UNSPEC;
5530                 link = rtnl_get_link(PF_UNSPEC, type);
5531                 if (!link || !link->doit)
5532                         goto out_unlock;
5533         }
5534
5535         owner = link->owner;
5536         if (!try_module_get(owner)) {
5537                 err = -EPROTONOSUPPORT;
5538                 goto out_unlock;
5539         }
5540
5541         flags = link->flags;
5542         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5543                 doit = link->doit;
5544                 rcu_read_unlock();
5545                 if (doit)
5546                         err = doit(skb, nlh, extack);
5547                 module_put(owner);
5548                 return err;
5549         }
5550         rcu_read_unlock();
5551
5552         rtnl_lock();
5553         link = rtnl_get_link(family, type);
5554         if (link && link->doit)
5555                 err = link->doit(skb, nlh, extack);
5556         rtnl_unlock();
5557
5558         module_put(owner);
5559
5560         return err;
5561
5562 out_unlock:
5563         rcu_read_unlock();
5564         return err;
5565
5566 err_unlock:
5567         rcu_read_unlock();
5568         return -EOPNOTSUPP;
5569 }
5570
5571 static void rtnetlink_rcv(struct sk_buff *skb)
5572 {
5573         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5574 }
5575
5576 static int rtnetlink_bind(struct net *net, int group)
5577 {
5578         switch (group) {
5579         case RTNLGRP_IPV4_MROUTE_R:
5580         case RTNLGRP_IPV6_MROUTE_R:
5581                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5582                         return -EPERM;
5583                 break;
5584         }
5585         return 0;
5586 }
5587
5588 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5589 {
5590         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5591
5592         switch (event) {
5593         case NETDEV_REBOOT:
5594         case NETDEV_CHANGEMTU:
5595         case NETDEV_CHANGEADDR:
5596         case NETDEV_CHANGENAME:
5597         case NETDEV_FEAT_CHANGE:
5598         case NETDEV_BONDING_FAILOVER:
5599         case NETDEV_POST_TYPE_CHANGE:
5600         case NETDEV_NOTIFY_PEERS:
5601         case NETDEV_CHANGEUPPER:
5602         case NETDEV_RESEND_IGMP:
5603         case NETDEV_CHANGEINFODATA:
5604         case NETDEV_CHANGELOWERSTATE:
5605         case NETDEV_CHANGE_TX_QUEUE_LEN:
5606                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5607                                    GFP_KERNEL, NULL, 0);
5608                 break;
5609         default:
5610                 break;
5611         }
5612         return NOTIFY_DONE;
5613 }
5614
5615 static struct notifier_block rtnetlink_dev_notifier = {
5616         .notifier_call  = rtnetlink_event,
5617 };
5618
5619
5620 static int __net_init rtnetlink_net_init(struct net *net)
5621 {
5622         struct sock *sk;
5623         struct netlink_kernel_cfg cfg = {
5624                 .groups         = RTNLGRP_MAX,
5625                 .input          = rtnetlink_rcv,
5626                 .cb_mutex       = &rtnl_mutex,
5627                 .flags          = NL_CFG_F_NONROOT_RECV,
5628                 .bind           = rtnetlink_bind,
5629         };
5630
5631         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5632         if (!sk)
5633                 return -ENOMEM;
5634         net->rtnl = sk;
5635         return 0;
5636 }
5637
5638 static void __net_exit rtnetlink_net_exit(struct net *net)
5639 {
5640         netlink_kernel_release(net->rtnl);
5641         net->rtnl = NULL;
5642 }
5643
5644 static struct pernet_operations rtnetlink_net_ops = {
5645         .init = rtnetlink_net_init,
5646         .exit = rtnetlink_net_exit,
5647 };
5648
5649 void __init rtnetlink_init(void)
5650 {
5651         if (register_pernet_subsys(&rtnetlink_net_ops))
5652                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5653
5654         register_netdevice_notifier(&rtnetlink_dev_notifier);
5655
5656         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5657                       rtnl_dump_ifinfo, 0);
5658         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5659         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5660         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5661
5662         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5663         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5664         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5665
5666         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5667         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5668
5669         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5670         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5671         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5672
5673         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5674         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5675         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5676
5677         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5678                       0);
5679 }