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