net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link *__rcu *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 **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 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_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 **tab, *link;
 290         int msgindex;
 291
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293         msgindex = rtm_msgindex(msgtype);
 294
 295         rtnl_lock();
 296         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 297         if (!tab) {
 298                 rtnl_unlock();
 299                 return -ENOENT;
 300         }
 301
 302         link = tab[msgindex];
 303         rcu_assign_pointer(tab[msgindex], NULL);
 304         rtnl_unlock();
 305
 306         kfree_rcu(link, rcu);
 307
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_unregister);
 311
 312 /**
 313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 314  * @protocol : Protocol family or PF_UNSPEC
 315  *
 316  * Identical to calling rtnl_unregster() for all registered message types
 317  * of a certain protocol family.
 318  */
 319 void rtnl_unregister_all(int protocol)
 320 {
 321         struct rtnl_link **tab, *link;
 322         int msgindex;
 323
 324         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 325
 326         rtnl_lock();
 327         tab = rtnl_msg_handlers[protocol];
 328         if (!tab) {
 329                 rtnl_unlock();
 330                 return;
 331         }
 332         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 333         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 334                 link = tab[msgindex];
 335                 if (!link)
 336                         continue;
 337
 338                 rcu_assign_pointer(tab[msgindex], NULL);
 339                 kfree_rcu(link, rcu);
 340         }
 341         rtnl_unlock();
 342
 343         synchronize_net();
 344
 345         kfree(tab);
 346 }
 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 348
 349 static LIST_HEAD(link_ops);
 350
 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 352 {
 353         const struct rtnl_link_ops *ops;
 354
 355         list_for_each_entry(ops, &link_ops, list) {
 356                 if (!strcmp(ops->kind, kind))
 357                         return ops;
 358         }
 359         return NULL;
 360 }
 361
 362 /**
 363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 364  * @ops: struct rtnl_link_ops * to register
 365  *
 366  * The caller must hold the rtnl_mutex. This function should be used
 367  * by drivers that create devices during module initialization. It
 368  * must be called before registering the devices.
 369  *
 370  * Returns 0 on success or a negative error code.
 371  */
 372 int __rtnl_link_register(struct rtnl_link_ops *ops)
 373 {
 374         if (rtnl_link_ops_get(ops->kind))
 375                 return -EEXIST;
 376
 377         /* The check for setup is here because if ops
 378          * does not have that filled up, it is not possible
 379          * to use the ops for creating device. So do not
 380          * fill up dellink as well. That disables rtnl_dellink.
 381          */
 382         if (ops->setup && !ops->dellink)
 383                 ops->dellink = unregister_netdevice_queue;
 384
 385         list_add_tail(&ops->list, &link_ops);
 386         return 0;
 387 }
 388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 389
 390 /**
 391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 392  * @ops: struct rtnl_link_ops * to register
 393  *
 394  * Returns 0 on success or a negative error code.
 395  */
 396 int rtnl_link_register(struct rtnl_link_ops *ops)
 397 {
 398         int err;
 399
 400         /* Sanity-check max sizes to avoid stack buffer overflow. */
 401         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 402                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 403                 return -EINVAL;
 404
 405         rtnl_lock();
 406         err = __rtnl_link_register(ops);
 407         rtnl_unlock();
 408         return err;
 409 }
 410 EXPORT_SYMBOL_GPL(rtnl_link_register);
 411
 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 413 {
 414         struct net_device *dev;
 415         LIST_HEAD(list_kill);
 416
 417         for_each_netdev(net, dev) {
 418                 if (dev->rtnl_link_ops == ops)
 419                         ops->dellink(dev, &list_kill);
 420         }
 421         unregister_netdevice_many(&list_kill);
 422 }
 423
 424 /**
 425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 426  * @ops: struct rtnl_link_ops * to unregister
 427  *
 428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 429  * integrity (hold pernet_ops_rwsem for writing to close the race
 430  * with setup_net() and cleanup_net()).
 431  */
 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 433 {
 434         struct net *net;
 435
 436         for_each_net(net) {
 437                 __rtnl_kill_links(net, ops);
 438         }
 439         list_del(&ops->list);
 440 }
 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 442
 443 /* Return with the rtnl_lock held when there are no network
 444  * devices unregistering in any network namespace.
 445  */
 446 static void rtnl_lock_unregistering_all(void)
 447 {
 448         struct net *net;
 449         bool unregistering;
 450         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 451
 452         add_wait_queue(&netdev_unregistering_wq, &wait);
 453         for (;;) {
 454                 unregistering = false;
 455                 rtnl_lock();
 456                 /* We held write locked pernet_ops_rwsem, and parallel
 457                  * setup_net() and cleanup_net() are not possible.
 458                  */
 459                 for_each_net(net) {
 460                         if (net->dev_unreg_count > 0) {
 461                                 unregistering = true;
 462                                 break;
 463                         }
 464                 }
 465                 if (!unregistering)
 466                         break;
 467                 __rtnl_unlock();
 468
 469                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 470         }
 471         remove_wait_queue(&netdev_unregistering_wq, &wait);
 472 }
 473
 474 /**
 475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 476  * @ops: struct rtnl_link_ops * to unregister
 477  */
 478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 479 {
 480         /* Close the race with setup_net() and cleanup_net() */
 481         down_write(&pernet_ops_rwsem);
 482         rtnl_lock_unregistering_all();
 483         __rtnl_link_unregister(ops);
 484         rtnl_unlock();
 485         up_write(&pernet_ops_rwsem);
 486 }
 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 488
 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 490 {
 491         struct net_device *master_dev;
 492         const struct rtnl_link_ops *ops;
 493         size_t size = 0;
 494
 495         rcu_read_lock();
 496
 497         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 498         if (!master_dev)
 499                 goto out;
 500
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops || !ops->get_slave_size)
 503                 goto out;
 504         /* IFLA_INFO_SLAVE_DATA + nested data */
 505         size = nla_total_size(sizeof(struct nlattr)) +
 506                ops->get_slave_size(master_dev, dev);
 507
 508 out:
 509         rcu_read_unlock();
 510         return size;
 511 }
 512
 513 static size_t rtnl_link_get_size(const struct net_device *dev)
 514 {
 515         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 516         size_t size;
 517
 518         if (!ops)
 519                 return 0;
 520
 521         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 522                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 523
 524         if (ops->get_size)
 525                 /* IFLA_INFO_DATA + nested data */
 526                 size += nla_total_size(sizeof(struct nlattr)) +
 527                         ops->get_size(dev);
 528
 529         if (ops->get_xstats_size)
 530                 /* IFLA_INFO_XSTATS */
 531                 size += nla_total_size(ops->get_xstats_size(dev));
 532
 533         size += rtnl_link_get_slave_info_data_size(dev);
 534
 535         return size;
 536 }
 537
 538 static LIST_HEAD(rtnl_af_ops);
 539
 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 541 {
 542         const struct rtnl_af_ops *ops;
 543
 544         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 545                 if (ops->family == family)
 546                         return ops;
 547         }
 548
 549         return NULL;
 550 }
 551
 552 /**
 553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 554  * @ops: struct rtnl_af_ops * to register
 555  *
 556  * Returns 0 on success or a negative error code.
 557  */
 558 void rtnl_af_register(struct rtnl_af_ops *ops)
 559 {
 560         rtnl_lock();
 561         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 562         rtnl_unlock();
 563 }
 564 EXPORT_SYMBOL_GPL(rtnl_af_register);
 565
 566 /**
 567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 568  * @ops: struct rtnl_af_ops * to unregister
 569  */
 570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 571 {
 572         rtnl_lock();
 573         list_del_rcu(&ops->list);
 574         rtnl_unlock();
 575
 576         synchronize_rcu();
 577 }
 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 579
 580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 581                                     u32 ext_filter_mask)
 582 {
 583         struct rtnl_af_ops *af_ops;
 584         size_t size;
 585
 586         /* IFLA_AF_SPEC */
 587         size = nla_total_size(sizeof(struct nlattr));
 588
 589         rcu_read_lock();
 590         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 591                 if (af_ops->get_link_af_size) {
 592                         /* AF_* + nested data */
 593                         size += nla_total_size(sizeof(struct nlattr)) +
 594                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 595                 }
 596         }
 597         rcu_read_unlock();
 598
 599         return size;
 600 }
 601
 602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 603 {
 604         struct net_device *master_dev;
 605         bool ret = false;
 606
 607         rcu_read_lock();
 608
 609         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 610         if (master_dev && master_dev->rtnl_link_ops)
 611                 ret = true;
 612         rcu_read_unlock();
 613         return ret;
 614 }
 615
 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 617                                      const struct net_device *dev)
 618 {
 619         struct net_device *master_dev;
 620         const struct rtnl_link_ops *ops;
 621         struct nlattr *slave_data;
 622         int err;
 623
 624         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 625         if (!master_dev)
 626                 return 0;
 627         ops = master_dev->rtnl_link_ops;
 628         if (!ops)
 629                 return 0;
 630         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 631                 return -EMSGSIZE;
 632         if (ops->fill_slave_info) {
 633                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 634                 if (!slave_data)
 635                         return -EMSGSIZE;
 636                 err = ops->fill_slave_info(skb, master_dev, dev);
 637                 if (err < 0)
 638                         goto err_cancel_slave_data;
 639                 nla_nest_end(skb, slave_data);
 640         }
 641         return 0;
 642
 643 err_cancel_slave_data:
 644         nla_nest_cancel(skb, slave_data);
 645         return err;
 646 }
 647
 648 static int rtnl_link_info_fill(struct sk_buff *skb,
 649                                const struct net_device *dev)
 650 {
 651         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 652         struct nlattr *data;
 653         int err;
 654
 655         if (!ops)
 656                 return 0;
 657         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 658                 return -EMSGSIZE;
 659         if (ops->fill_xstats) {
 660                 err = ops->fill_xstats(skb, dev);
 661                 if (err < 0)
 662                         return err;
 663         }
 664         if (ops->fill_info) {
 665                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 666                 if (data == NULL)
 667                         return -EMSGSIZE;
 668                 err = ops->fill_info(skb, dev);
 669                 if (err < 0)
 670                         goto err_cancel_data;
 671                 nla_nest_end(skb, data);
 672         }
 673         return 0;
 674
 675 err_cancel_data:
 676         nla_nest_cancel(skb, data);
 677         return err;
 678 }
 679
 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 681 {
 682         struct nlattr *linkinfo;
 683         int err = -EMSGSIZE;
 684
 685         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 686         if (linkinfo == NULL)
 687                 goto out;
 688
 689         err = rtnl_link_info_fill(skb, dev);
 690         if (err < 0)
 691                 goto err_cancel_link;
 692
 693         err = rtnl_link_slave_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         nla_nest_end(skb, linkinfo);
 698         return 0;
 699
 700 err_cancel_link:
 701         nla_nest_cancel(skb, linkinfo);
 702 out:
 703         return err;
 704 }
 705
 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 707 {
 708         struct sock *rtnl = net->rtnl;
 709         int err = 0;
 710
 711         NETLINK_CB(skb).dst_group = group;
 712         if (echo)
 713                 refcount_inc(&skb->users);
 714         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 715         if (echo)
 716                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 717         return err;
 718 }
 719
 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 721 {
 722         struct sock *rtnl = net->rtnl;
 723
 724         return nlmsg_unicast(rtnl, skb, pid);
 725 }
 726 EXPORT_SYMBOL(rtnl_unicast);
 727
 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 729                  struct nlmsghdr *nlh, gfp_t flags)
 730 {
 731         struct sock *rtnl = net->rtnl;
 732         int report = 0;
 733
 734         if (nlh)
 735                 report = nlmsg_report(nlh);
 736
 737         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 738 }
 739 EXPORT_SYMBOL(rtnl_notify);
 740
 741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 742 {
 743         struct sock *rtnl = net->rtnl;
 744
 745         netlink_set_err(rtnl, 0, group, error);
 746 }
 747 EXPORT_SYMBOL(rtnl_set_sk_err);
 748
 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 750 {
 751         struct nlattr *mx;
 752         int i, valid = 0;
 753
 754         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 755         if (metrics == dst_default_metrics.metrics)
 756                 return 0;
 757
 758         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 759         if (mx == NULL)
 760                 return -ENOBUFS;
 761
 762         for (i = 0; i < RTAX_MAX; i++) {
 763                 if (metrics[i]) {
 764                         if (i == RTAX_CC_ALGO - 1) {
 765                                 char tmp[TCP_CA_NAME_MAX], *name;
 766
 767                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 768                                 if (!name)
 769                                         continue;
 770                                 if (nla_put_string(skb, i + 1, name))
 771                                         goto nla_put_failure;
 772                         } else if (i == RTAX_FEATURES - 1) {
 773                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 774
 775                                 if (!user_features)
 776                                         continue;
 777                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 778                                 if (nla_put_u32(skb, i + 1, user_features))
 779                                         goto nla_put_failure;
 780                         } else {
 781                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 782                                         goto nla_put_failure;
 783                         }
 784                         valid++;
 785                 }
 786         }
 787
 788         if (!valid) {
 789                 nla_nest_cancel(skb, mx);
 790                 return 0;
 791         }
 792
 793         return nla_nest_end(skb, mx);
 794
 795 nla_put_failure:
 796         nla_nest_cancel(skb, mx);
 797         return -EMSGSIZE;
 798 }
 799 EXPORT_SYMBOL(rtnetlink_put_metrics);
 800
 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 802                        long expires, u32 error)
 803 {
 804         struct rta_cacheinfo ci = {
 805                 .rta_error = error,
 806                 .rta_id =  id,
 807         };
 808
 809         if (dst) {
 810                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 811                 ci.rta_used = dst->__use;
 812                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 813         }
 814         if (expires) {
 815                 unsigned long clock;
 816
 817                 clock = jiffies_to_clock_t(abs(expires));
 818                 clock = min_t(unsigned long, clock, INT_MAX);
 819                 ci.rta_expires = (expires > 0) ? clock : -clock;
 820         }
 821         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 822 }
 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 824
 825 static void set_operstate(struct net_device *dev, unsigned char transition)
 826 {
 827         unsigned char operstate = dev->operstate;
 828
 829         switch (transition) {
 830         case IF_OPER_UP:
 831                 if ((operstate == IF_OPER_DORMANT ||
 832                      operstate == IF_OPER_TESTING ||
 833                      operstate == IF_OPER_UNKNOWN) &&
 834                     !netif_dormant(dev) && !netif_testing(dev))
 835                         operstate = IF_OPER_UP;
 836                 break;
 837
 838         case IF_OPER_TESTING:
 839                 if (operstate == IF_OPER_UP ||
 840                     operstate == IF_OPER_UNKNOWN)
 841                         operstate = IF_OPER_TESTING;
 842                 break;
 843
 844         case IF_OPER_DORMANT:
 845                 if (operstate == IF_OPER_UP ||
 846                     operstate == IF_OPER_UNKNOWN)
 847                         operstate = IF_OPER_DORMANT;
 848                 break;
 849         }
 850
 851         if (dev->operstate != operstate) {
 852                 write_lock_bh(&dev_base_lock);
 853                 dev->operstate = operstate;
 854                 write_unlock_bh(&dev_base_lock);
 855                 netdev_state_change(dev);
 856         }
 857 }
 858
 859 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 860 {
 861         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 862                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 863 }
 864
 865 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 866                                            const struct ifinfomsg *ifm)
 867 {
 868         unsigned int flags = ifm->ifi_flags;
 869
 870         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 871         if (ifm->ifi_change)
 872                 flags = (flags & ifm->ifi_change) |
 873                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 874
 875         return flags;
 876 }
 877
 878 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 879                                  const struct rtnl_link_stats64 *b)
 880 {
 881         a->rx_packets = b->rx_packets;
 882         a->tx_packets = b->tx_packets;
 883         a->rx_bytes = b->rx_bytes;
 884         a->tx_bytes = b->tx_bytes;
 885         a->rx_errors = b->rx_errors;
 886         a->tx_errors = b->tx_errors;
 887         a->rx_dropped = b->rx_dropped;
 888         a->tx_dropped = b->tx_dropped;
 889
 890         a->multicast = b->multicast;
 891         a->collisions = b->collisions;
 892
 893         a->rx_length_errors = b->rx_length_errors;
 894         a->rx_over_errors = b->rx_over_errors;
 895         a->rx_crc_errors = b->rx_crc_errors;
 896         a->rx_frame_errors = b->rx_frame_errors;
 897         a->rx_fifo_errors = b->rx_fifo_errors;
 898         a->rx_missed_errors = b->rx_missed_errors;
 899
 900         a->tx_aborted_errors = b->tx_aborted_errors;
 901         a->tx_carrier_errors = b->tx_carrier_errors;
 902         a->tx_fifo_errors = b->tx_fifo_errors;
 903         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 904         a->tx_window_errors = b->tx_window_errors;
 905
 906         a->rx_compressed = b->rx_compressed;
 907         a->tx_compressed = b->tx_compressed;
 908
 909         a->rx_nohandler = b->rx_nohandler;
 910 }
 911
 912 /* All VF info */
 913 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 914                                    u32 ext_filter_mask)
 915 {
 916         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 917                 int num_vfs = dev_num_vf(dev->dev.parent);
 918                 size_t size = nla_total_size(0);
 919                 size += num_vfs *
 920                         (nla_total_size(0) +
 921                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 922                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 923                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 924                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 925                          nla_total_size(MAX_VLAN_LIST_LEN *
 926                                         sizeof(struct ifla_vf_vlan_info)) +
 927                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 928                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 929                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 930                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 931                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 932                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 933                          /* IFLA_VF_STATS_RX_PACKETS */
 934                          nla_total_size_64bit(sizeof(__u64)) +
 935                          /* IFLA_VF_STATS_TX_PACKETS */
 936                          nla_total_size_64bit(sizeof(__u64)) +
 937                          /* IFLA_VF_STATS_RX_BYTES */
 938                          nla_total_size_64bit(sizeof(__u64)) +
 939                          /* IFLA_VF_STATS_TX_BYTES */
 940                          nla_total_size_64bit(sizeof(__u64)) +
 941                          /* IFLA_VF_STATS_BROADCAST */
 942                          nla_total_size_64bit(sizeof(__u64)) +
 943                          /* IFLA_VF_STATS_MULTICAST */
 944                          nla_total_size_64bit(sizeof(__u64)) +
 945                          /* IFLA_VF_STATS_RX_DROPPED */
 946                          nla_total_size_64bit(sizeof(__u64)) +
 947                          /* IFLA_VF_STATS_TX_DROPPED */
 948                          nla_total_size_64bit(sizeof(__u64)) +
 949                          nla_total_size(sizeof(struct ifla_vf_trust)));
 950                 return size;
 951         } else
 952                 return 0;
 953 }
 954
 955 static size_t rtnl_port_size(const struct net_device *dev,
 956                              u32 ext_filter_mask)
 957 {
 958         size_t port_size = nla_total_size(4)            /* PORT_VF */
 959                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 960                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 961                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 962                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 963                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 964         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 965         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 966                 + port_size;
 967         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 968                 + port_size;
 969
 970         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 971             !(ext_filter_mask & RTEXT_FILTER_VF))
 972                 return 0;
 973         if (dev_num_vf(dev->dev.parent))
 974                 return port_self_size + vf_ports_size +
 975                         vf_port_size * dev_num_vf(dev->dev.parent);
 976         else
 977                 return port_self_size;
 978 }
 979
 980 static size_t rtnl_xdp_size(void)
 981 {
 982         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 983                           nla_total_size(1) +   /* XDP_ATTACHED */
 984                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 985                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 986
 987         return xdp_size;
 988 }
 989
 990 static size_t rtnl_prop_list_size(const struct net_device *dev)
 991 {
 992         struct netdev_name_node *name_node;
 993         size_t size;
 994
 995         if (list_empty(&dev->name_node->list))
 996                 return 0;
 997         size = nla_total_size(0);
 998         list_for_each_entry(name_node, &dev->name_node->list, list)
 999                 size += nla_total_size(ALTIFNAMSIZ);
1000         return size;
1001 }
1002
1003 static size_t rtnl_proto_down_size(const struct net_device *dev)
1004 {
1005         size_t size = nla_total_size(1);
1006
1007         if (dev->proto_down_reason)
1008                 size += nla_total_size(0) + nla_total_size(4);
1009
1010         return size;
1011 }
1012
1013 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1014                                      u32 ext_filter_mask)
1015 {
1016         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1017                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1018                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1019                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1020                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1021                + nla_total_size(sizeof(struct rtnl_link_stats))
1022                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1023                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1024                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1025                + nla_total_size(4) /* IFLA_TXQLEN */
1026                + nla_total_size(4) /* IFLA_WEIGHT */
1027                + nla_total_size(4) /* IFLA_MTU */
1028                + nla_total_size(4) /* IFLA_LINK */
1029                + nla_total_size(4) /* IFLA_MASTER */
1030                + nla_total_size(1) /* IFLA_CARRIER */
1031                + nla_total_size(4) /* IFLA_PROMISCUITY */
1032                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1033                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1034                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1035                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1036                + nla_total_size(1) /* IFLA_OPERSTATE */
1037                + nla_total_size(1) /* IFLA_LINKMODE */
1038                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1039                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1040                + nla_total_size(4) /* IFLA_GROUP */
1041                + nla_total_size(ext_filter_mask
1042                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1043                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1044                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1045                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1046                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1047                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1048                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1049                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1050                + rtnl_xdp_size() /* IFLA_XDP */
1051                + nla_total_size(4)  /* IFLA_EVENT */
1052                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1053                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1054                + rtnl_proto_down_size(dev)  /* proto down */
1055                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1056                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1057                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1058                + nla_total_size(4)  /* IFLA_MIN_MTU */
1059                + nla_total_size(4)  /* IFLA_MAX_MTU */
1060                + rtnl_prop_list_size(dev)
1061                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1062                + 0;
1063 }
1064
1065 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1066 {
1067         struct nlattr *vf_ports;
1068         struct nlattr *vf_port;
1069         int vf;
1070         int err;
1071
1072         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1073         if (!vf_ports)
1074                 return -EMSGSIZE;
1075
1076         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1077                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1078                 if (!vf_port)
1079                         goto nla_put_failure;
1080                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1081                         goto nla_put_failure;
1082                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1083                 if (err == -EMSGSIZE)
1084                         goto nla_put_failure;
1085                 if (err) {
1086                         nla_nest_cancel(skb, vf_port);
1087                         continue;
1088                 }
1089                 nla_nest_end(skb, vf_port);
1090         }
1091
1092         nla_nest_end(skb, vf_ports);
1093
1094         return 0;
1095
1096 nla_put_failure:
1097         nla_nest_cancel(skb, vf_ports);
1098         return -EMSGSIZE;
1099 }
1100
1101 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1102 {
1103         struct nlattr *port_self;
1104         int err;
1105
1106         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1107         if (!port_self)
1108                 return -EMSGSIZE;
1109
1110         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1111         if (err) {
1112                 nla_nest_cancel(skb, port_self);
1113                 return (err == -EMSGSIZE) ? err : 0;
1114         }
1115
1116         nla_nest_end(skb, port_self);
1117
1118         return 0;
1119 }
1120
1121 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1122                           u32 ext_filter_mask)
1123 {
1124         int err;
1125
1126         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1127             !(ext_filter_mask & RTEXT_FILTER_VF))
1128                 return 0;
1129
1130         err = rtnl_port_self_fill(skb, dev);
1131         if (err)
1132                 return err;
1133
1134         if (dev_num_vf(dev->dev.parent)) {
1135                 err = rtnl_vf_ports_fill(skb, dev);
1136                 if (err)
1137                         return err;
1138         }
1139
1140         return 0;
1141 }
1142
1143 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1144 {
1145         int err;
1146         struct netdev_phys_item_id ppid;
1147
1148         err = dev_get_phys_port_id(dev, &ppid);
1149         if (err) {
1150                 if (err == -EOPNOTSUPP)
1151                         return 0;
1152                 return err;
1153         }
1154
1155         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1156                 return -EMSGSIZE;
1157
1158         return 0;
1159 }
1160
1161 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1162 {
1163         char name[IFNAMSIZ];
1164         int err;
1165
1166         err = dev_get_phys_port_name(dev, name, sizeof(name));
1167         if (err) {
1168                 if (err == -EOPNOTSUPP)
1169                         return 0;
1170                 return err;
1171         }
1172
1173         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1174                 return -EMSGSIZE;
1175
1176         return 0;
1177 }
1178
1179 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1180 {
1181         struct netdev_phys_item_id ppid = { };
1182         int err;
1183
1184         err = dev_get_port_parent_id(dev, &ppid, false);
1185         if (err) {
1186                 if (err == -EOPNOTSUPP)
1187                         return 0;
1188                 return err;
1189         }
1190
1191         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1192                 return -EMSGSIZE;
1193
1194         return 0;
1195 }
1196
1197 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1198                                               struct net_device *dev)
1199 {
1200         struct rtnl_link_stats64 *sp;
1201         struct nlattr *attr;
1202
1203         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1204                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1205         if (!attr)
1206                 return -EMSGSIZE;
1207
1208         sp = nla_data(attr);
1209         dev_get_stats(dev, sp);
1210
1211         attr = nla_reserve(skb, IFLA_STATS,
1212                            sizeof(struct rtnl_link_stats));
1213         if (!attr)
1214                 return -EMSGSIZE;
1215
1216         copy_rtnl_link_stats(nla_data(attr), sp);
1217
1218         return 0;
1219 }
1220
1221 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1222                                                struct net_device *dev,
1223                                                int vfs_num,
1224                                                struct nlattr *vfinfo)
1225 {
1226         struct ifla_vf_rss_query_en vf_rss_query_en;
1227         struct nlattr *vf, *vfstats, *vfvlanlist;
1228         struct ifla_vf_link_state vf_linkstate;
1229         struct ifla_vf_vlan_info vf_vlan_info;
1230         struct ifla_vf_spoofchk vf_spoofchk;
1231         struct ifla_vf_tx_rate vf_tx_rate;
1232         struct ifla_vf_stats vf_stats;
1233         struct ifla_vf_trust vf_trust;
1234         struct ifla_vf_vlan vf_vlan;
1235         struct ifla_vf_rate vf_rate;
1236         struct ifla_vf_mac vf_mac;
1237         struct ifla_vf_broadcast vf_broadcast;
1238         struct ifla_vf_info ivi;
1239         struct ifla_vf_guid node_guid;
1240         struct ifla_vf_guid port_guid;
1241
1242         memset(&ivi, 0, sizeof(ivi));
1243
1244         /* Not all SR-IOV capable drivers support the
1245          * spoofcheck and "RSS query enable" query.  Preset to
1246          * -1 so the user space tool can detect that the driver
1247          * didn't report anything.
1248          */
1249         ivi.spoofchk = -1;
1250         ivi.rss_query_en = -1;
1251         ivi.trusted = -1;
1252         /* The default value for VF link state is "auto"
1253          * IFLA_VF_LINK_STATE_AUTO which equals zero
1254          */
1255         ivi.linkstate = 0;
1256         /* VLAN Protocol by default is 802.1Q */
1257         ivi.vlan_proto = htons(ETH_P_8021Q);
1258         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1259                 return 0;
1260
1261         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1262         memset(&node_guid, 0, sizeof(node_guid));
1263         memset(&port_guid, 0, sizeof(port_guid));
1264
1265         vf_mac.vf =
1266                 vf_vlan.vf =
1267                 vf_vlan_info.vf =
1268                 vf_rate.vf =
1269                 vf_tx_rate.vf =
1270                 vf_spoofchk.vf =
1271                 vf_linkstate.vf =
1272                 vf_rss_query_en.vf =
1273                 vf_trust.vf =
1274                 node_guid.vf =
1275                 port_guid.vf = ivi.vf;
1276
1277         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1278         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1279         vf_vlan.vlan = ivi.vlan;
1280         vf_vlan.qos = ivi.qos;
1281         vf_vlan_info.vlan = ivi.vlan;
1282         vf_vlan_info.qos = ivi.qos;
1283         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1284         vf_tx_rate.rate = ivi.max_tx_rate;
1285         vf_rate.min_tx_rate = ivi.min_tx_rate;
1286         vf_rate.max_tx_rate = ivi.max_tx_rate;
1287         vf_spoofchk.setting = ivi.spoofchk;
1288         vf_linkstate.link_state = ivi.linkstate;
1289         vf_rss_query_en.setting = ivi.rss_query_en;
1290         vf_trust.setting = ivi.trusted;
1291         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1292         if (!vf)
1293                 goto nla_put_vfinfo_failure;
1294         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1295             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1296             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1297             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1298                     &vf_rate) ||
1299             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1300                     &vf_tx_rate) ||
1301             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1302                     &vf_spoofchk) ||
1303             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1304                     &vf_linkstate) ||
1305             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1306                     sizeof(vf_rss_query_en),
1307                     &vf_rss_query_en) ||
1308             nla_put(skb, IFLA_VF_TRUST,
1309                     sizeof(vf_trust), &vf_trust))
1310                 goto nla_put_vf_failure;
1311
1312         if (dev->netdev_ops->ndo_get_vf_guid &&
1313             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1314                                               &port_guid)) {
1315                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1316                             &node_guid) ||
1317                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1318                             &port_guid))
1319                         goto nla_put_vf_failure;
1320         }
1321         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1322         if (!vfvlanlist)
1323                 goto nla_put_vf_failure;
1324         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1325                     &vf_vlan_info)) {
1326                 nla_nest_cancel(skb, vfvlanlist);
1327                 goto nla_put_vf_failure;
1328         }
1329         nla_nest_end(skb, vfvlanlist);
1330         memset(&vf_stats, 0, sizeof(vf_stats));
1331         if (dev->netdev_ops->ndo_get_vf_stats)
1332                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1333                                                 &vf_stats);
1334         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1335         if (!vfstats)
1336                 goto nla_put_vf_failure;
1337         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1338                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1339             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1340                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1341             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1342                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1343             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1344                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1345             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1346                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1347             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1348                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1349             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1350                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1351             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1352                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1353                 nla_nest_cancel(skb, vfstats);
1354                 goto nla_put_vf_failure;
1355         }
1356         nla_nest_end(skb, vfstats);
1357         nla_nest_end(skb, vf);
1358         return 0;
1359
1360 nla_put_vf_failure:
1361         nla_nest_cancel(skb, vf);
1362 nla_put_vfinfo_failure:
1363         nla_nest_cancel(skb, vfinfo);
1364         return -EMSGSIZE;
1365 }
1366
1367 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1368                                            struct net_device *dev,
1369                                            u32 ext_filter_mask)
1370 {
1371         struct nlattr *vfinfo;
1372         int i, num_vfs;
1373
1374         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1375                 return 0;
1376
1377         num_vfs = dev_num_vf(dev->dev.parent);
1378         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1379                 return -EMSGSIZE;
1380
1381         if (!dev->netdev_ops->ndo_get_vf_config)
1382                 return 0;
1383
1384         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1385         if (!vfinfo)
1386                 return -EMSGSIZE;
1387
1388         for (i = 0; i < num_vfs; i++) {
1389                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1390                         return -EMSGSIZE;
1391         }
1392
1393         nla_nest_end(skb, vfinfo);
1394         return 0;
1395 }
1396
1397 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1398 {
1399         struct rtnl_link_ifmap map;
1400
1401         memset(&map, 0, sizeof(map));
1402         map.mem_start   = dev->mem_start;
1403         map.mem_end     = dev->mem_end;
1404         map.base_addr   = dev->base_addr;
1405         map.irq         = dev->irq;
1406         map.dma         = dev->dma;
1407         map.port        = dev->if_port;
1408
1409         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1410                 return -EMSGSIZE;
1411
1412         return 0;
1413 }
1414
1415 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1416 {
1417         const struct bpf_prog *generic_xdp_prog;
1418
1419         ASSERT_RTNL();
1420
1421         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1422         if (!generic_xdp_prog)
1423                 return 0;
1424         return generic_xdp_prog->aux->id;
1425 }
1426
1427 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1428 {
1429         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1430 }
1431
1432 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1433 {
1434         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1435 }
1436
1437 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1438                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1439                                u32 (*get_prog_id)(struct net_device *dev))
1440 {
1441         u32 curr_id;
1442         int err;
1443
1444         curr_id = get_prog_id(dev);
1445         if (!curr_id)
1446                 return 0;
1447
1448         *prog_id = curr_id;
1449         err = nla_put_u32(skb, attr, curr_id);
1450         if (err)
1451                 return err;
1452
1453         if (*mode != XDP_ATTACHED_NONE)
1454                 *mode = XDP_ATTACHED_MULTI;
1455         else
1456                 *mode = tgt_mode;
1457
1458         return 0;
1459 }
1460
1461 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1462 {
1463         struct nlattr *xdp;
1464         u32 prog_id;
1465         int err;
1466         u8 mode;
1467
1468         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1469         if (!xdp)
1470                 return -EMSGSIZE;
1471
1472         prog_id = 0;
1473         mode = XDP_ATTACHED_NONE;
1474         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1475                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1476         if (err)
1477                 goto err_cancel;
1478         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1479                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1480         if (err)
1481                 goto err_cancel;
1482         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1483                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1484         if (err)
1485                 goto err_cancel;
1486
1487         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1488         if (err)
1489                 goto err_cancel;
1490
1491         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1492                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1493                 if (err)
1494                         goto err_cancel;
1495         }
1496
1497         nla_nest_end(skb, xdp);
1498         return 0;
1499
1500 err_cancel:
1501         nla_nest_cancel(skb, xdp);
1502         return err;
1503 }
1504
1505 static u32 rtnl_get_event(unsigned long event)
1506 {
1507         u32 rtnl_event_type = IFLA_EVENT_NONE;
1508
1509         switch (event) {
1510         case NETDEV_REBOOT:
1511                 rtnl_event_type = IFLA_EVENT_REBOOT;
1512                 break;
1513         case NETDEV_FEAT_CHANGE:
1514                 rtnl_event_type = IFLA_EVENT_FEATURES;
1515                 break;
1516         case NETDEV_BONDING_FAILOVER:
1517                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1518                 break;
1519         case NETDEV_NOTIFY_PEERS:
1520                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1521                 break;
1522         case NETDEV_RESEND_IGMP:
1523                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1524                 break;
1525         case NETDEV_CHANGEINFODATA:
1526                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1527                 break;
1528         default:
1529                 break;
1530         }
1531
1532         return rtnl_event_type;
1533 }
1534
1535 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1536 {
1537         const struct net_device *upper_dev;
1538         int ret = 0;
1539
1540         rcu_read_lock();
1541
1542         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1543         if (upper_dev)
1544                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1545
1546         rcu_read_unlock();
1547         return ret;
1548 }
1549
1550 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1551                           bool force)
1552 {
1553         int ifindex = dev_get_iflink(dev);
1554
1555         if (force || dev->ifindex != ifindex)
1556                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1557
1558         return 0;
1559 }
1560
1561 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1562                                               struct net_device *dev)
1563 {
1564         char buf[IFALIASZ];
1565         int ret;
1566
1567         ret = dev_get_alias(dev, buf, sizeof(buf));
1568         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1569 }
1570
1571 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1572                                   const struct net_device *dev,
1573                                   struct net *src_net, gfp_t gfp)
1574 {
1575         bool put_iflink = false;
1576
1577         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1578                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1579
1580                 if (!net_eq(dev_net(dev), link_net)) {
1581                         int id = peernet2id_alloc(src_net, link_net, gfp);
1582
1583                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1584                                 return -EMSGSIZE;
1585
1586                         put_iflink = true;
1587                 }
1588         }
1589
1590         return nla_put_iflink(skb, dev, put_iflink);
1591 }
1592
1593 static int rtnl_fill_link_af(struct sk_buff *skb,
1594                              const struct net_device *dev,
1595                              u32 ext_filter_mask)
1596 {
1597         const struct rtnl_af_ops *af_ops;
1598         struct nlattr *af_spec;
1599
1600         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1601         if (!af_spec)
1602                 return -EMSGSIZE;
1603
1604         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1605                 struct nlattr *af;
1606                 int err;
1607
1608                 if (!af_ops->fill_link_af)
1609                         continue;
1610
1611                 af = nla_nest_start_noflag(skb, af_ops->family);
1612                 if (!af)
1613                         return -EMSGSIZE;
1614
1615                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1616                 /*
1617                  * Caller may return ENODATA to indicate that there
1618                  * was no data to be dumped. This is not an error, it
1619                  * means we should trim the attribute header and
1620                  * continue.
1621                  */
1622                 if (err == -ENODATA)
1623                         nla_nest_cancel(skb, af);
1624                 else if (err < 0)
1625                         return -EMSGSIZE;
1626
1627                 nla_nest_end(skb, af);
1628         }
1629
1630         nla_nest_end(skb, af_spec);
1631         return 0;
1632 }
1633
1634 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1635                                  const struct net_device *dev)
1636 {
1637         struct netdev_name_node *name_node;
1638         int count = 0;
1639
1640         list_for_each_entry(name_node, &dev->name_node->list, list) {
1641                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1642                         return -EMSGSIZE;
1643                 count++;
1644         }
1645         return count;
1646 }
1647
1648 static int rtnl_fill_prop_list(struct sk_buff *skb,
1649                                const struct net_device *dev)
1650 {
1651         struct nlattr *prop_list;
1652         int ret;
1653
1654         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1655         if (!prop_list)
1656                 return -EMSGSIZE;
1657
1658         ret = rtnl_fill_alt_ifnames(skb, dev);
1659         if (ret <= 0)
1660                 goto nest_cancel;
1661
1662         nla_nest_end(skb, prop_list);
1663         return 0;
1664
1665 nest_cancel:
1666         nla_nest_cancel(skb, prop_list);
1667         return ret;
1668 }
1669
1670 static int rtnl_fill_proto_down(struct sk_buff *skb,
1671                                 const struct net_device *dev)
1672 {
1673         struct nlattr *pr;
1674         u32 preason;
1675
1676         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1677                 goto nla_put_failure;
1678
1679         preason = dev->proto_down_reason;
1680         if (!preason)
1681                 return 0;
1682
1683         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1684         if (!pr)
1685                 return -EMSGSIZE;
1686
1687         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1688                 nla_nest_cancel(skb, pr);
1689                 goto nla_put_failure;
1690         }
1691
1692         nla_nest_end(skb, pr);
1693         return 0;
1694
1695 nla_put_failure:
1696         return -EMSGSIZE;
1697 }
1698
1699 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1700                             struct net_device *dev, struct net *src_net,
1701                             int type, u32 pid, u32 seq, u32 change,
1702                             unsigned int flags, u32 ext_filter_mask,
1703                             u32 event, int *new_nsid, int new_ifindex,
1704                             int tgt_netnsid, gfp_t gfp)
1705 {
1706         struct ifinfomsg *ifm;
1707         struct nlmsghdr *nlh;
1708
1709         ASSERT_RTNL();
1710         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1711         if (nlh == NULL)
1712                 return -EMSGSIZE;
1713
1714         ifm = nlmsg_data(nlh);
1715         ifm->ifi_family = AF_UNSPEC;
1716         ifm->__ifi_pad = 0;
1717         ifm->ifi_type = dev->type;
1718         ifm->ifi_index = dev->ifindex;
1719         ifm->ifi_flags = dev_get_flags(dev);
1720         ifm->ifi_change = change;
1721
1722         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1723                 goto nla_put_failure;
1724
1725         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1726             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1727             nla_put_u8(skb, IFLA_OPERSTATE,
1728                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1729             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1730             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1731             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1732             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1733             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1734             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1735             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1736             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1737             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1738 #ifdef CONFIG_RPS
1739             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1740 #endif
1741             put_master_ifindex(skb, dev) ||
1742             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1743             (dev->qdisc &&
1744              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1745             nla_put_ifalias(skb, dev) ||
1746             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1747                         atomic_read(&dev->carrier_up_count) +
1748                         atomic_read(&dev->carrier_down_count)) ||
1749             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1750                         atomic_read(&dev->carrier_up_count)) ||
1751             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1752                         atomic_read(&dev->carrier_down_count)))
1753                 goto nla_put_failure;
1754
1755         if (rtnl_fill_proto_down(skb, dev))
1756                 goto nla_put_failure;
1757
1758         if (event != IFLA_EVENT_NONE) {
1759                 if (nla_put_u32(skb, IFLA_EVENT, event))
1760                         goto nla_put_failure;
1761         }
1762
1763         if (rtnl_fill_link_ifmap(skb, dev))
1764                 goto nla_put_failure;
1765
1766         if (dev->addr_len) {
1767                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1768                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1769                         goto nla_put_failure;
1770         }
1771
1772         if (rtnl_phys_port_id_fill(skb, dev))
1773                 goto nla_put_failure;
1774
1775         if (rtnl_phys_port_name_fill(skb, dev))
1776                 goto nla_put_failure;
1777
1778         if (rtnl_phys_switch_id_fill(skb, dev))
1779                 goto nla_put_failure;
1780
1781         if (rtnl_fill_stats(skb, dev))
1782                 goto nla_put_failure;
1783
1784         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1785                 goto nla_put_failure;
1786
1787         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1788                 goto nla_put_failure;
1789
1790         if (rtnl_xdp_fill(skb, dev))
1791                 goto nla_put_failure;
1792
1793         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1794                 if (rtnl_link_fill(skb, dev) < 0)
1795                         goto nla_put_failure;
1796         }
1797
1798         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1799                 goto nla_put_failure;
1800
1801         if (new_nsid &&
1802             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1803                 goto nla_put_failure;
1804         if (new_ifindex &&
1805             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1806                 goto nla_put_failure;
1807
1808         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1809             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1810                 goto nla_put_failure;
1811
1812         rcu_read_lock();
1813         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1814                 goto nla_put_failure_rcu;
1815         rcu_read_unlock();
1816
1817         if (rtnl_fill_prop_list(skb, dev))
1818                 goto nla_put_failure;
1819
1820         nlmsg_end(skb, nlh);
1821         return 0;
1822
1823 nla_put_failure_rcu:
1824         rcu_read_unlock();
1825 nla_put_failure:
1826         nlmsg_cancel(skb, nlh);
1827         return -EMSGSIZE;
1828 }
1829
1830 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1831         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1832         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1833         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1834         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1835         [IFLA_MTU]              = { .type = NLA_U32 },
1836         [IFLA_LINK]             = { .type = NLA_U32 },
1837         [IFLA_MASTER]           = { .type = NLA_U32 },
1838         [IFLA_CARRIER]          = { .type = NLA_U8 },
1839         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1840         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1841         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1842         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1843         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1844         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1845         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1846         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1847          * allow 0-length string (needed to remove an alias).
1848          */
1849         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1850         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1851         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1852         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1853         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1854         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1855         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1856         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1857         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1858         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1859         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1860         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1861         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1862         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1863         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1864         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1865         [IFLA_XDP]              = { .type = NLA_NESTED },
1866         [IFLA_EVENT]            = { .type = NLA_U32 },
1867         [IFLA_GROUP]            = { .type = NLA_U32 },
1868         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1869         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1870         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1871         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1872         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1873         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1874         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1875                                     .len = ALTIFNAMSIZ - 1 },
1876         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1877         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1878 };
1879
1880 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1881         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1882         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1883         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1884         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1885 };
1886
1887 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1888         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1889         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1890         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1891         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1892         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1893         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1894         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1895         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1896         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1897         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1898         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1899         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1900         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1901 };
1902
1903 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1904         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1905         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1906                                     .len = PORT_PROFILE_MAX },
1907         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1908                                       .len = PORT_UUID_MAX },
1909         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1910                                     .len = PORT_UUID_MAX },
1911         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1912         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1913
1914         /* Unused, but we need to keep it here since user space could
1915          * fill it. It's also broken with regard to NLA_BINARY use in
1916          * combination with structs.
1917          */
1918         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1919                                     .len = sizeof(struct ifla_port_vsi) },
1920 };
1921
1922 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1923         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1924         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1925         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1926         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1927         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1928         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1929 };
1930
1931 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1932 {
1933         const struct rtnl_link_ops *ops = NULL;
1934         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1935
1936         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1937                 return NULL;
1938
1939         if (linfo[IFLA_INFO_KIND]) {
1940                 char kind[MODULE_NAME_LEN];
1941
1942                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1943                 ops = rtnl_link_ops_get(kind);
1944         }
1945
1946         return ops;
1947 }
1948
1949 static bool link_master_filtered(struct net_device *dev, int master_idx)
1950 {
1951         struct net_device *master;
1952
1953         if (!master_idx)
1954                 return false;
1955
1956         master = netdev_master_upper_dev_get(dev);
1957         if (!master || master->ifindex != master_idx)
1958                 return true;
1959
1960         return false;
1961 }
1962
1963 static bool link_kind_filtered(const struct net_device *dev,
1964                                const struct rtnl_link_ops *kind_ops)
1965 {
1966         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1967                 return true;
1968
1969         return false;
1970 }
1971
1972 static bool link_dump_filtered(struct net_device *dev,
1973                                int master_idx,
1974                                const struct rtnl_link_ops *kind_ops)
1975 {
1976         if (link_master_filtered(dev, master_idx) ||
1977             link_kind_filtered(dev, kind_ops))
1978                 return true;
1979
1980         return false;
1981 }
1982
1983 /**
1984  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1985  * @sk: netlink socket
1986  * @netnsid: network namespace identifier
1987  *
1988  * Returns the network namespace identified by netnsid on success or an error
1989  * pointer on failure.
1990  */
1991 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1992 {
1993         struct net *net;
1994
1995         net = get_net_ns_by_id(sock_net(sk), netnsid);
1996         if (!net)
1997                 return ERR_PTR(-EINVAL);
1998
1999         /* For now, the caller is required to have CAP_NET_ADMIN in
2000          * the user namespace owning the target net ns.
2001          */
2002         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2003                 put_net(net);
2004                 return ERR_PTR(-EACCES);
2005         }
2006         return net;
2007 }
2008 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2009
2010 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2011                                       bool strict_check, struct nlattr **tb,
2012                                       struct netlink_ext_ack *extack)
2013 {
2014         int hdrlen;
2015
2016         if (strict_check) {
2017                 struct ifinfomsg *ifm;
2018
2019                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2020                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2021                         return -EINVAL;
2022                 }
2023
2024                 ifm = nlmsg_data(nlh);
2025                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2026                     ifm->ifi_change) {
2027                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2028                         return -EINVAL;
2029                 }
2030                 if (ifm->ifi_index) {
2031                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2032                         return -EINVAL;
2033                 }
2034
2035                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2036                                                      IFLA_MAX, ifla_policy,
2037                                                      extack);
2038         }
2039
2040         /* A hack to preserve kernel<->userspace interface.
2041          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2042          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2043          * what iproute2 < v3.9.0 used.
2044          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2045          * attribute, its netlink message is shorter than struct ifinfomsg.
2046          */
2047         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2048                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2049
2050         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2051                                       extack);
2052 }
2053
2054 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2055 {
2056         struct netlink_ext_ack *extack = cb->extack;
2057         const struct nlmsghdr *nlh = cb->nlh;
2058         struct net *net = sock_net(skb->sk);
2059         struct net *tgt_net = net;
2060         int h, s_h;
2061         int idx = 0, s_idx;
2062         struct net_device *dev;
2063         struct hlist_head *head;
2064         struct nlattr *tb[IFLA_MAX+1];
2065         u32 ext_filter_mask = 0;
2066         const struct rtnl_link_ops *kind_ops = NULL;
2067         unsigned int flags = NLM_F_MULTI;
2068         int master_idx = 0;
2069         int netnsid = -1;
2070         int err, i;
2071
2072         s_h = cb->args[0];
2073         s_idx = cb->args[1];
2074
2075         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2076         if (err < 0) {
2077                 if (cb->strict_check)
2078                         return err;
2079
2080                 goto walk_entries;
2081         }
2082
2083         for (i = 0; i <= IFLA_MAX; ++i) {
2084                 if (!tb[i])
2085                         continue;
2086
2087                 /* new attributes should only be added with strict checking */
2088                 switch (i) {
2089                 case IFLA_TARGET_NETNSID:
2090                         netnsid = nla_get_s32(tb[i]);
2091                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2092                         if (IS_ERR(tgt_net)) {
2093                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2094                                 return PTR_ERR(tgt_net);
2095                         }
2096                         break;
2097                 case IFLA_EXT_MASK:
2098                         ext_filter_mask = nla_get_u32(tb[i]);
2099                         break;
2100                 case IFLA_MASTER:
2101                         master_idx = nla_get_u32(tb[i]);
2102                         break;
2103                 case IFLA_LINKINFO:
2104                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2105                         break;
2106                 default:
2107                         if (cb->strict_check) {
2108                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2109                                 return -EINVAL;
2110                         }
2111                 }
2112         }
2113
2114         if (master_idx || kind_ops)
2115                 flags |= NLM_F_DUMP_FILTERED;
2116
2117 walk_entries:
2118         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2119                 idx = 0;
2120                 head = &tgt_net->dev_index_head[h];
2121                 hlist_for_each_entry(dev, head, index_hlist) {
2122                         if (link_dump_filtered(dev, master_idx, kind_ops))
2123                                 goto cont;
2124                         if (idx < s_idx)
2125                                 goto cont;
2126                         err = rtnl_fill_ifinfo(skb, dev, net,
2127                                                RTM_NEWLINK,
2128                                                NETLINK_CB(cb->skb).portid,
2129                                                nlh->nlmsg_seq, 0, flags,
2130                                                ext_filter_mask, 0, NULL, 0,
2131                                                netnsid, GFP_KERNEL);
2132
2133                         if (err < 0) {
2134                                 if (likely(skb->len))
2135                                         goto out;
2136
2137                                 goto out_err;
2138                         }
2139 cont:
2140                         idx++;
2141                 }
2142         }
2143 out:
2144         err = skb->len;
2145 out_err:
2146         cb->args[1] = idx;
2147         cb->args[0] = h;
2148         cb->seq = net->dev_base_seq;
2149         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2150         if (netnsid >= 0)
2151                 put_net(tgt_net);
2152
2153         return err;
2154 }
2155
2156 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2157                         struct netlink_ext_ack *exterr)
2158 {
2159         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2160                                     exterr);
2161 }
2162 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2163
2164 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2165 {
2166         struct net *net;
2167         /* Examine the link attributes and figure out which
2168          * network namespace we are talking about.
2169          */
2170         if (tb[IFLA_NET_NS_PID])
2171                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2172         else if (tb[IFLA_NET_NS_FD])
2173                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2174         else
2175                 net = get_net(src_net);
2176         return net;
2177 }
2178 EXPORT_SYMBOL(rtnl_link_get_net);
2179
2180 /* Figure out which network namespace we are talking about by
2181  * examining the link attributes in the following order:
2182  *
2183  * 1. IFLA_NET_NS_PID
2184  * 2. IFLA_NET_NS_FD
2185  * 3. IFLA_TARGET_NETNSID
2186  */
2187 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2188                                                struct nlattr *tb[])
2189 {
2190         struct net *net;
2191
2192         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2193                 return rtnl_link_get_net(src_net, tb);
2194
2195         if (!tb[IFLA_TARGET_NETNSID])
2196                 return get_net(src_net);
2197
2198         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2199         if (!net)
2200                 return ERR_PTR(-EINVAL);
2201
2202         return net;
2203 }
2204
2205 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2206                                              struct net *src_net,
2207                                              struct nlattr *tb[], int cap)
2208 {
2209         struct net *net;
2210
2211         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2212         if (IS_ERR(net))
2213                 return net;
2214
2215         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2216                 put_net(net);
2217                 return ERR_PTR(-EPERM);
2218         }
2219
2220         return net;
2221 }
2222
2223 /* Verify that rtnetlink requests do not pass additional properties
2224  * potentially referring to different network namespaces.
2225  */
2226 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2227                                     struct netlink_ext_ack *extack,
2228                                     bool netns_id_only)
2229 {
2230
2231         if (netns_id_only) {
2232                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2233                         return 0;
2234
2235                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2236                 return -EOPNOTSUPP;
2237         }
2238
2239         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2240                 goto invalid_attr;
2241
2242         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2243                 goto invalid_attr;
2244
2245         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2246                 goto invalid_attr;
2247
2248         return 0;
2249
2250 invalid_attr:
2251         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2252         return -EINVAL;
2253 }
2254
2255 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2256 {
2257         if (dev) {
2258                 if (tb[IFLA_ADDRESS] &&
2259                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2260                         return -EINVAL;
2261
2262                 if (tb[IFLA_BROADCAST] &&
2263                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2264                         return -EINVAL;
2265         }
2266
2267         if (tb[IFLA_AF_SPEC]) {
2268                 struct nlattr *af;
2269                 int rem, err;
2270
2271                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2272                         const struct rtnl_af_ops *af_ops;
2273
2274                         rcu_read_lock();
2275                         af_ops = rtnl_af_lookup(nla_type(af));
2276                         if (!af_ops) {
2277                                 rcu_read_unlock();
2278                                 return -EAFNOSUPPORT;
2279                         }
2280
2281                         if (!af_ops->set_link_af) {
2282                                 rcu_read_unlock();
2283                                 return -EOPNOTSUPP;
2284                         }
2285
2286                         if (af_ops->validate_link_af) {
2287                                 err = af_ops->validate_link_af(dev, af);
2288                                 if (err < 0) {
2289                                         rcu_read_unlock();
2290                                         return err;
2291                                 }
2292                         }
2293
2294                         rcu_read_unlock();
2295                 }
2296         }
2297
2298         return 0;
2299 }
2300
2301 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2302                                   int guid_type)
2303 {
2304         const struct net_device_ops *ops = dev->netdev_ops;
2305
2306         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2307 }
2308
2309 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2310 {
2311         if (dev->type != ARPHRD_INFINIBAND)
2312                 return -EOPNOTSUPP;
2313
2314         return handle_infiniband_guid(dev, ivt, guid_type);
2315 }
2316
2317 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2318 {
2319         const struct net_device_ops *ops = dev->netdev_ops;
2320         int err = -EINVAL;
2321
2322         if (tb[IFLA_VF_MAC]) {
2323                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2324
2325                 if (ivm->vf >= INT_MAX)
2326                         return -EINVAL;
2327                 err = -EOPNOTSUPP;
2328                 if (ops->ndo_set_vf_mac)
2329                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2330                                                   ivm->mac);
2331                 if (err < 0)
2332                         return err;
2333         }
2334
2335         if (tb[IFLA_VF_VLAN]) {
2336                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2337
2338                 if (ivv->vf >= INT_MAX)
2339                         return -EINVAL;
2340                 err = -EOPNOTSUPP;
2341                 if (ops->ndo_set_vf_vlan)
2342                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2343                                                    ivv->qos,
2344                                                    htons(ETH_P_8021Q));
2345                 if (err < 0)
2346                         return err;
2347         }
2348
2349         if (tb[IFLA_VF_VLAN_LIST]) {
2350                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2351                 struct nlattr *attr;
2352                 int rem, len = 0;
2353
2354                 err = -EOPNOTSUPP;
2355                 if (!ops->ndo_set_vf_vlan)
2356                         return err;
2357
2358                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2359                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2360                             nla_len(attr) < NLA_HDRLEN) {
2361                                 return -EINVAL;
2362                         }
2363                         if (len >= MAX_VLAN_LIST_LEN)
2364                                 return -EOPNOTSUPP;
2365                         ivvl[len] = nla_data(attr);
2366
2367                         len++;
2368                 }
2369                 if (len == 0)
2370                         return -EINVAL;
2371
2372                 if (ivvl[0]->vf >= INT_MAX)
2373                         return -EINVAL;
2374                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2375                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2376                 if (err < 0)
2377                         return err;
2378         }
2379
2380         if (tb[IFLA_VF_TX_RATE]) {
2381                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2382                 struct ifla_vf_info ivf;
2383
2384                 if (ivt->vf >= INT_MAX)
2385                         return -EINVAL;
2386                 err = -EOPNOTSUPP;
2387                 if (ops->ndo_get_vf_config)
2388                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2389                 if (err < 0)
2390                         return err;
2391
2392                 err = -EOPNOTSUPP;
2393                 if (ops->ndo_set_vf_rate)
2394                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2395                                                    ivf.min_tx_rate,
2396                                                    ivt->rate);
2397                 if (err < 0)
2398                         return err;
2399         }
2400
2401         if (tb[IFLA_VF_RATE]) {
2402                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2403
2404                 if (ivt->vf >= INT_MAX)
2405                         return -EINVAL;
2406                 err = -EOPNOTSUPP;
2407                 if (ops->ndo_set_vf_rate)
2408                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2409                                                    ivt->min_tx_rate,
2410                                                    ivt->max_tx_rate);
2411                 if (err < 0)
2412                         return err;
2413         }
2414
2415         if (tb[IFLA_VF_SPOOFCHK]) {
2416                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2417
2418                 if (ivs->vf >= INT_MAX)
2419                         return -EINVAL;
2420                 err = -EOPNOTSUPP;
2421                 if (ops->ndo_set_vf_spoofchk)
2422                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2423                                                        ivs->setting);
2424                 if (err < 0)
2425                         return err;
2426         }
2427
2428         if (tb[IFLA_VF_LINK_STATE]) {
2429                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2430
2431                 if (ivl->vf >= INT_MAX)
2432                         return -EINVAL;
2433                 err = -EOPNOTSUPP;
2434                 if (ops->ndo_set_vf_link_state)
2435                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2436                                                          ivl->link_state);
2437                 if (err < 0)
2438                         return err;
2439         }
2440
2441         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2442                 struct ifla_vf_rss_query_en *ivrssq_en;
2443
2444                 err = -EOPNOTSUPP;
2445                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2446                 if (ivrssq_en->vf >= INT_MAX)
2447                         return -EINVAL;
2448                 if (ops->ndo_set_vf_rss_query_en)
2449                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2450                                                            ivrssq_en->setting);
2451                 if (err < 0)
2452                         return err;
2453         }
2454
2455         if (tb[IFLA_VF_TRUST]) {
2456                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2457
2458                 if (ivt->vf >= INT_MAX)
2459                         return -EINVAL;
2460                 err = -EOPNOTSUPP;
2461                 if (ops->ndo_set_vf_trust)
2462                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2463                 if (err < 0)
2464                         return err;
2465         }
2466
2467         if (tb[IFLA_VF_IB_NODE_GUID]) {
2468                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2469
2470                 if (ivt->vf >= INT_MAX)
2471                         return -EINVAL;
2472                 if (!ops->ndo_set_vf_guid)
2473                         return -EOPNOTSUPP;
2474                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2475         }
2476
2477         if (tb[IFLA_VF_IB_PORT_GUID]) {
2478                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2479
2480                 if (ivt->vf >= INT_MAX)
2481                         return -EINVAL;
2482                 if (!ops->ndo_set_vf_guid)
2483                         return -EOPNOTSUPP;
2484
2485                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2486         }
2487
2488         return err;
2489 }
2490
2491 static int do_set_master(struct net_device *dev, int ifindex,
2492                          struct netlink_ext_ack *extack)
2493 {
2494         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2495         const struct net_device_ops *ops;
2496         int err;
2497
2498         if (upper_dev) {
2499                 if (upper_dev->ifindex == ifindex)
2500                         return 0;
2501                 ops = upper_dev->netdev_ops;
2502                 if (ops->ndo_del_slave) {
2503                         err = ops->ndo_del_slave(upper_dev, dev);
2504                         if (err)
2505                                 return err;
2506                 } else {
2507                         return -EOPNOTSUPP;
2508                 }
2509         }
2510
2511         if (ifindex) {
2512                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2513                 if (!upper_dev)
2514                         return -EINVAL;
2515                 ops = upper_dev->netdev_ops;
2516                 if (ops->ndo_add_slave) {
2517                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2518                         if (err)
2519                                 return err;
2520                 } else {
2521                         return -EOPNOTSUPP;
2522                 }
2523         }
2524         return 0;
2525 }
2526
2527 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2528         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2529         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2530 };
2531
2532 static int do_set_proto_down(struct net_device *dev,
2533                              struct nlattr *nl_proto_down,
2534                              struct nlattr *nl_proto_down_reason,
2535                              struct netlink_ext_ack *extack)
2536 {
2537         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2538         const struct net_device_ops *ops = dev->netdev_ops;
2539         unsigned long mask = 0;
2540         u32 value;
2541         bool proto_down;
2542         int err;
2543
2544         if (!ops->ndo_change_proto_down) {
2545                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2546                 return -EOPNOTSUPP;
2547         }
2548
2549         if (nl_proto_down_reason) {
2550                 err = nla_parse_nested_deprecated(pdreason,
2551                                                   IFLA_PROTO_DOWN_REASON_MAX,
2552                                                   nl_proto_down_reason,
2553                                                   ifla_proto_down_reason_policy,
2554                                                   NULL);
2555                 if (err < 0)
2556                         return err;
2557
2558                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2559                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2560                         return -EINVAL;
2561                 }
2562
2563                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2564
2565                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2566                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2567
2568                 dev_change_proto_down_reason(dev, mask, value);
2569         }
2570
2571         if (nl_proto_down) {
2572                 proto_down = nla_get_u8(nl_proto_down);
2573
2574                 /* Dont turn off protodown if there are active reasons */
2575                 if (!proto_down && dev->proto_down_reason) {
2576                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2577                         return -EBUSY;
2578                 }
2579                 err = dev_change_proto_down(dev,
2580                                             proto_down);
2581                 if (err)
2582                         return err;
2583         }
2584
2585         return 0;
2586 }
2587
2588 #define DO_SETLINK_MODIFIED     0x01
2589 /* notify flag means notify + modified. */
2590 #define DO_SETLINK_NOTIFY       0x03
2591 static int do_setlink(const struct sk_buff *skb,
2592                       struct net_device *dev, struct ifinfomsg *ifm,
2593                       struct netlink_ext_ack *extack,
2594                       struct nlattr **tb, char *ifname, int status)
2595 {
2596         const struct net_device_ops *ops = dev->netdev_ops;
2597         int err;
2598
2599         err = validate_linkmsg(dev, tb);
2600         if (err < 0)
2601                 return err;
2602
2603         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2604                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2605                                                             tb, CAP_NET_ADMIN);
2606                 if (IS_ERR(net)) {
2607                         err = PTR_ERR(net);
2608                         goto errout;
2609                 }
2610
2611                 err = dev_change_net_namespace(dev, net, ifname);
2612                 put_net(net);
2613                 if (err)
2614                         goto errout;
2615                 status |= DO_SETLINK_MODIFIED;
2616         }
2617
2618         if (tb[IFLA_MAP]) {
2619                 struct rtnl_link_ifmap *u_map;
2620                 struct ifmap k_map;
2621
2622                 if (!ops->ndo_set_config) {
2623                         err = -EOPNOTSUPP;
2624                         goto errout;
2625                 }
2626
2627                 if (!netif_device_present(dev)) {
2628                         err = -ENODEV;
2629                         goto errout;
2630                 }
2631
2632                 u_map = nla_data(tb[IFLA_MAP]);
2633                 k_map.mem_start = (unsigned long) u_map->mem_start;
2634                 k_map.mem_end = (unsigned long) u_map->mem_end;
2635                 k_map.base_addr = (unsigned short) u_map->base_addr;
2636                 k_map.irq = (unsigned char) u_map->irq;
2637                 k_map.dma = (unsigned char) u_map->dma;
2638                 k_map.port = (unsigned char) u_map->port;
2639
2640                 err = ops->ndo_set_config(dev, &k_map);
2641                 if (err < 0)
2642                         goto errout;
2643
2644                 status |= DO_SETLINK_NOTIFY;
2645         }
2646
2647         if (tb[IFLA_ADDRESS]) {
2648                 struct sockaddr *sa;
2649                 int len;
2650
2651                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2652                                                   sizeof(*sa));
2653                 sa = kmalloc(len, GFP_KERNEL);
2654                 if (!sa) {
2655                         err = -ENOMEM;
2656                         goto errout;
2657                 }
2658                 sa->sa_family = dev->type;
2659                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2660                        dev->addr_len);
2661                 err = dev_set_mac_address(dev, sa, extack);
2662                 kfree(sa);
2663                 if (err)
2664                         goto errout;
2665                 status |= DO_SETLINK_MODIFIED;
2666         }
2667
2668         if (tb[IFLA_MTU]) {
2669                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2670                 if (err < 0)
2671                         goto errout;
2672                 status |= DO_SETLINK_MODIFIED;
2673         }
2674
2675         if (tb[IFLA_GROUP]) {
2676                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2677                 status |= DO_SETLINK_NOTIFY;
2678         }
2679
2680         /*
2681          * Interface selected by interface index but interface
2682          * name provided implies that a name change has been
2683          * requested.
2684          */
2685         if (ifm->ifi_index > 0 && ifname[0]) {
2686                 err = dev_change_name(dev, ifname);
2687                 if (err < 0)
2688                         goto errout;
2689                 status |= DO_SETLINK_MODIFIED;
2690         }
2691
2692         if (tb[IFLA_IFALIAS]) {
2693                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2694                                     nla_len(tb[IFLA_IFALIAS]));
2695                 if (err < 0)
2696                         goto errout;
2697                 status |= DO_SETLINK_NOTIFY;
2698         }
2699
2700         if (tb[IFLA_BROADCAST]) {
2701                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2702                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2703         }
2704
2705         if (ifm->ifi_flags || ifm->ifi_change) {
2706                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2707                                        extack);
2708                 if (err < 0)
2709                         goto errout;
2710         }
2711
2712         if (tb[IFLA_MASTER]) {
2713                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2714                 if (err)
2715                         goto errout;
2716                 status |= DO_SETLINK_MODIFIED;
2717         }
2718
2719         if (tb[IFLA_CARRIER]) {
2720                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2721                 if (err)
2722                         goto errout;
2723                 status |= DO_SETLINK_MODIFIED;
2724         }
2725
2726         if (tb[IFLA_TXQLEN]) {
2727                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2728
2729                 err = dev_change_tx_queue_len(dev, value);
2730                 if (err)
2731                         goto errout;
2732                 status |= DO_SETLINK_MODIFIED;
2733         }
2734
2735         if (tb[IFLA_GSO_MAX_SIZE]) {
2736                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2737
2738                 if (max_size > GSO_MAX_SIZE) {
2739                         err = -EINVAL;
2740                         goto errout;
2741                 }
2742
2743                 if (dev->gso_max_size ^ max_size) {
2744                         netif_set_gso_max_size(dev, max_size);
2745                         status |= DO_SETLINK_MODIFIED;
2746                 }
2747         }
2748
2749         if (tb[IFLA_GSO_MAX_SEGS]) {
2750                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2751
2752                 if (max_segs > GSO_MAX_SEGS) {
2753                         err = -EINVAL;
2754                         goto errout;
2755                 }
2756
2757                 if (dev->gso_max_segs ^ max_segs) {
2758                         dev->gso_max_segs = max_segs;
2759                         status |= DO_SETLINK_MODIFIED;
2760                 }
2761         }
2762
2763         if (tb[IFLA_OPERSTATE])
2764                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2765
2766         if (tb[IFLA_LINKMODE]) {
2767                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2768
2769                 write_lock_bh(&dev_base_lock);
2770                 if (dev->link_mode ^ value)
2771                         status |= DO_SETLINK_NOTIFY;
2772                 dev->link_mode = value;
2773                 write_unlock_bh(&dev_base_lock);
2774         }
2775
2776         if (tb[IFLA_VFINFO_LIST]) {
2777                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2778                 struct nlattr *attr;
2779                 int rem;
2780
2781                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2782                         if (nla_type(attr) != IFLA_VF_INFO ||
2783                             nla_len(attr) < NLA_HDRLEN) {
2784                                 err = -EINVAL;
2785                                 goto errout;
2786                         }
2787                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2788                                                           attr,
2789                                                           ifla_vf_policy,
2790                                                           NULL);
2791                         if (err < 0)
2792                                 goto errout;
2793                         err = do_setvfinfo(dev, vfinfo);
2794                         if (err < 0)
2795                                 goto errout;
2796                         status |= DO_SETLINK_NOTIFY;
2797                 }
2798         }
2799         err = 0;
2800
2801         if (tb[IFLA_VF_PORTS]) {
2802                 struct nlattr *port[IFLA_PORT_MAX+1];
2803                 struct nlattr *attr;
2804                 int vf;
2805                 int rem;
2806
2807                 err = -EOPNOTSUPP;
2808                 if (!ops->ndo_set_vf_port)
2809                         goto errout;
2810
2811                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2812                         if (nla_type(attr) != IFLA_VF_PORT ||
2813                             nla_len(attr) < NLA_HDRLEN) {
2814                                 err = -EINVAL;
2815                                 goto errout;
2816                         }
2817                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2818                                                           attr,
2819                                                           ifla_port_policy,
2820                                                           NULL);
2821                         if (err < 0)
2822                                 goto errout;
2823                         if (!port[IFLA_PORT_VF]) {
2824                                 err = -EOPNOTSUPP;
2825                                 goto errout;
2826                         }
2827                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2828                         err = ops->ndo_set_vf_port(dev, vf, port);
2829                         if (err < 0)
2830                                 goto errout;
2831                         status |= DO_SETLINK_NOTIFY;
2832                 }
2833         }
2834         err = 0;
2835
2836         if (tb[IFLA_PORT_SELF]) {
2837                 struct nlattr *port[IFLA_PORT_MAX+1];
2838
2839                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2840                                                   tb[IFLA_PORT_SELF],
2841                                                   ifla_port_policy, NULL);
2842                 if (err < 0)
2843                         goto errout;
2844
2845                 err = -EOPNOTSUPP;
2846                 if (ops->ndo_set_vf_port)
2847                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2848                 if (err < 0)
2849                         goto errout;
2850                 status |= DO_SETLINK_NOTIFY;
2851         }
2852
2853         if (tb[IFLA_AF_SPEC]) {
2854                 struct nlattr *af;
2855                 int rem;
2856
2857                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2858                         const struct rtnl_af_ops *af_ops;
2859
2860                         rcu_read_lock();
2861
2862                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2863
2864                         err = af_ops->set_link_af(dev, af);
2865                         if (err < 0) {
2866                                 rcu_read_unlock();
2867                                 goto errout;
2868                         }
2869
2870                         rcu_read_unlock();
2871                         status |= DO_SETLINK_NOTIFY;
2872                 }
2873         }
2874         err = 0;
2875
2876         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2877                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2878                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2879                 if (err)
2880                         goto errout;
2881                 status |= DO_SETLINK_NOTIFY;
2882         }
2883
2884         if (tb[IFLA_XDP]) {
2885                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2886                 u32 xdp_flags = 0;
2887
2888                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2889                                                   tb[IFLA_XDP],
2890                                                   ifla_xdp_policy, NULL);
2891                 if (err < 0)
2892                         goto errout;
2893
2894                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2895                         err = -EINVAL;
2896                         goto errout;
2897                 }
2898
2899                 if (xdp[IFLA_XDP_FLAGS]) {
2900                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2901                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2902                                 err = -EINVAL;
2903                                 goto errout;
2904                         }
2905                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2906                                 err = -EINVAL;
2907                                 goto errout;
2908                         }
2909                 }
2910
2911                 if (xdp[IFLA_XDP_FD]) {
2912                         int expected_fd = -1;
2913
2914                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2915                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2916                                         err = -EINVAL;
2917                                         goto errout;
2918                                 }
2919                                 expected_fd =
2920                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2921                         }
2922
2923                         err = dev_change_xdp_fd(dev, extack,
2924                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2925                                                 expected_fd,
2926                                                 xdp_flags);
2927                         if (err)
2928                                 goto errout;
2929                         status |= DO_SETLINK_NOTIFY;
2930                 }
2931         }
2932
2933 errout:
2934         if (status & DO_SETLINK_MODIFIED) {
2935                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2936                         netdev_state_change(dev);
2937
2938                 if (err < 0)
2939                         net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
2940                                              dev->name);
2941         }
2942
2943         return err;
2944 }
2945
2946 static struct net_device *rtnl_dev_get(struct net *net,
2947                                        struct nlattr *ifname_attr,
2948                                        struct nlattr *altifname_attr,
2949                                        char *ifname)
2950 {
2951         char buffer[ALTIFNAMSIZ];
2952
2953         if (!ifname) {
2954                 ifname = buffer;
2955                 if (ifname_attr)
2956                         nla_strlcpy(ifname, ifname_attr, IFNAMSIZ);
2957                 else if (altifname_attr)
2958                         nla_strlcpy(ifname, altifname_attr, ALTIFNAMSIZ);
2959                 else
2960                         return NULL;
2961         }
2962
2963         return __dev_get_by_name(net, ifname);
2964 }
2965
2966 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2967                         struct netlink_ext_ack *extack)
2968 {
2969         struct net *net = sock_net(skb->sk);
2970         struct ifinfomsg *ifm;
2971         struct net_device *dev;
2972         int err;
2973         struct nlattr *tb[IFLA_MAX+1];
2974         char ifname[IFNAMSIZ];
2975
2976         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2977                                      ifla_policy, extack);
2978         if (err < 0)
2979                 goto errout;
2980
2981         err = rtnl_ensure_unique_netns(tb, extack, false);
2982         if (err < 0)
2983                 goto errout;
2984
2985         if (tb[IFLA_IFNAME])
2986                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2987         else
2988                 ifname[0] = '\0';
2989
2990         err = -EINVAL;
2991         ifm = nlmsg_data(nlh);
2992         if (ifm->ifi_index > 0)
2993                 dev = __dev_get_by_index(net, ifm->ifi_index);
2994         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2995                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2996         else
2997                 goto errout;
2998
2999         if (dev == NULL) {
3000                 err = -ENODEV;
3001                 goto errout;
3002         }
3003
3004         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3005 errout:
3006         return err;
3007 }
3008
3009 static int rtnl_group_dellink(const struct net *net, int group)
3010 {
3011         struct net_device *dev, *aux;
3012         LIST_HEAD(list_kill);
3013         bool found = false;
3014
3015         if (!group)
3016                 return -EPERM;
3017
3018         for_each_netdev(net, dev) {
3019                 if (dev->group == group) {
3020                         const struct rtnl_link_ops *ops;
3021
3022                         found = true;
3023                         ops = dev->rtnl_link_ops;
3024                         if (!ops || !ops->dellink)
3025                                 return -EOPNOTSUPP;
3026                 }
3027         }
3028
3029         if (!found)
3030                 return -ENODEV;
3031
3032         for_each_netdev_safe(net, dev, aux) {
3033                 if (dev->group == group) {
3034                         const struct rtnl_link_ops *ops;
3035
3036                         ops = dev->rtnl_link_ops;
3037                         ops->dellink(dev, &list_kill);
3038                 }
3039         }
3040         unregister_netdevice_many(&list_kill);
3041
3042         return 0;
3043 }
3044
3045 int rtnl_delete_link(struct net_device *dev)
3046 {
3047         const struct rtnl_link_ops *ops;
3048         LIST_HEAD(list_kill);
3049
3050         ops = dev->rtnl_link_ops;
3051         if (!ops || !ops->dellink)
3052                 return -EOPNOTSUPP;
3053
3054         ops->dellink(dev, &list_kill);
3055         unregister_netdevice_many(&list_kill);
3056
3057         return 0;
3058 }
3059 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3060
3061 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3062                         struct netlink_ext_ack *extack)
3063 {
3064         struct net *net = sock_net(skb->sk);
3065         struct net *tgt_net = net;
3066         struct net_device *dev = NULL;
3067         struct ifinfomsg *ifm;
3068         struct nlattr *tb[IFLA_MAX+1];
3069         int err;
3070         int netnsid = -1;
3071
3072         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3073                                      ifla_policy, extack);
3074         if (err < 0)
3075                 return err;
3076
3077         err = rtnl_ensure_unique_netns(tb, extack, true);
3078         if (err < 0)
3079                 return err;
3080
3081         if (tb[IFLA_TARGET_NETNSID]) {
3082                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3083                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3084                 if (IS_ERR(tgt_net))
3085                         return PTR_ERR(tgt_net);
3086         }
3087
3088         err = -EINVAL;
3089         ifm = nlmsg_data(nlh);
3090         if (ifm->ifi_index > 0)
3091                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3092         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3093                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3094                                    tb[IFLA_ALT_IFNAME], NULL);
3095         else if (tb[IFLA_GROUP])
3096                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3097         else
3098                 goto out;
3099
3100         if (!dev) {
3101                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3102                         err = -ENODEV;
3103
3104                 goto out;
3105         }
3106
3107         err = rtnl_delete_link(dev);
3108
3109 out:
3110         if (netnsid >= 0)
3111                 put_net(tgt_net);
3112
3113         return err;
3114 }
3115
3116 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3117 {
3118         unsigned int old_flags;
3119         int err;
3120
3121         old_flags = dev->flags;
3122         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3123                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3124                                          NULL);
3125                 if (err < 0)
3126                         return err;
3127         }
3128
3129         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3130                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3131         } else {
3132                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3133                 __dev_notify_flags(dev, old_flags, ~0U);
3134         }
3135         return 0;
3136 }
3137 EXPORT_SYMBOL(rtnl_configure_link);
3138
3139 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3140                                     unsigned char name_assign_type,
3141                                     const struct rtnl_link_ops *ops,
3142                                     struct nlattr *tb[],
3143                                     struct netlink_ext_ack *extack)
3144 {
3145         struct net_device *dev;
3146         unsigned int num_tx_queues = 1;
3147         unsigned int num_rx_queues = 1;
3148
3149         if (tb[IFLA_NUM_TX_QUEUES])
3150                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3151         else if (ops->get_num_tx_queues)
3152                 num_tx_queues = ops->get_num_tx_queues();
3153
3154         if (tb[IFLA_NUM_RX_QUEUES])
3155                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3156         else if (ops->get_num_rx_queues)
3157                 num_rx_queues = ops->get_num_rx_queues();
3158
3159         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3160                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3161                 return ERR_PTR(-EINVAL);
3162         }
3163
3164         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3165                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3166                 return ERR_PTR(-EINVAL);
3167         }
3168
3169         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3170                                ops->setup, num_tx_queues, num_rx_queues);
3171         if (!dev)
3172                 return ERR_PTR(-ENOMEM);
3173
3174         dev_net_set(dev, net);
3175         dev->rtnl_link_ops = ops;
3176         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3177
3178         if (tb[IFLA_MTU]) {
3179                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3180                 int err;
3181
3182                 err = dev_validate_mtu(dev, mtu, extack);
3183                 if (err) {
3184                         free_netdev(dev);
3185                         return ERR_PTR(err);
3186                 }
3187                 dev->mtu = mtu;
3188         }
3189         if (tb[IFLA_ADDRESS]) {
3190                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3191                                 nla_len(tb[IFLA_ADDRESS]));
3192                 dev->addr_assign_type = NET_ADDR_SET;
3193         }
3194         if (tb[IFLA_BROADCAST])
3195                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3196                                 nla_len(tb[IFLA_BROADCAST]));
3197         if (tb[IFLA_TXQLEN])
3198                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3199         if (tb[IFLA_OPERSTATE])
3200                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3201         if (tb[IFLA_LINKMODE])
3202                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3203         if (tb[IFLA_GROUP])
3204                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3205         if (tb[IFLA_GSO_MAX_SIZE])
3206                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3207         if (tb[IFLA_GSO_MAX_SEGS])
3208                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3209
3210         return dev;
3211 }
3212 EXPORT_SYMBOL(rtnl_create_link);
3213
3214 static int rtnl_group_changelink(const struct sk_buff *skb,
3215                 struct net *net, int group,
3216                 struct ifinfomsg *ifm,
3217                 struct netlink_ext_ack *extack,
3218                 struct nlattr **tb)
3219 {
3220         struct net_device *dev, *aux;
3221         int err;
3222
3223         for_each_netdev_safe(net, dev, aux) {
3224                 if (dev->group == group) {
3225                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3226                         if (err < 0)
3227                                 return err;
3228                 }
3229         }
3230
3231         return 0;
3232 }
3233
3234 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3235                           struct nlattr **attr, struct netlink_ext_ack *extack)
3236 {
3237         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3238         unsigned char name_assign_type = NET_NAME_USER;
3239         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3240         const struct rtnl_link_ops *m_ops = NULL;
3241         struct net_device *master_dev = NULL;
3242         struct net *net = sock_net(skb->sk);
3243         const struct rtnl_link_ops *ops;
3244         struct nlattr *tb[IFLA_MAX + 1];
3245         struct net *dest_net, *link_net;
3246         struct nlattr **slave_data;
3247         char kind[MODULE_NAME_LEN];
3248         struct net_device *dev;
3249         struct ifinfomsg *ifm;
3250         char ifname[IFNAMSIZ];
3251         struct nlattr **data;
3252         int err;
3253
3254 #ifdef CONFIG_MODULES
3255 replay:
3256 #endif
3257         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3258                                      ifla_policy, extack);
3259         if (err < 0)
3260                 return err;
3261
3262         err = rtnl_ensure_unique_netns(tb, extack, false);
3263         if (err < 0)
3264                 return err;
3265
3266         if (tb[IFLA_IFNAME])
3267                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3268         else
3269                 ifname[0] = '\0';
3270
3271         ifm = nlmsg_data(nlh);
3272         if (ifm->ifi_index > 0)
3273                 dev = __dev_get_by_index(net, ifm->ifi_index);
3274         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3275                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3276         else
3277                 dev = NULL;
3278
3279         if (dev) {
3280                 master_dev = netdev_master_upper_dev_get(dev);
3281                 if (master_dev)
3282                         m_ops = master_dev->rtnl_link_ops;
3283         }
3284
3285         err = validate_linkmsg(dev, tb);
3286         if (err < 0)
3287                 return err;
3288
3289         if (tb[IFLA_LINKINFO]) {
3290                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3291                                                   tb[IFLA_LINKINFO],
3292                                                   ifla_info_policy, NULL);
3293                 if (err < 0)
3294                         return err;
3295         } else
3296                 memset(linkinfo, 0, sizeof(linkinfo));
3297
3298         if (linkinfo[IFLA_INFO_KIND]) {
3299                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3300                 ops = rtnl_link_ops_get(kind);
3301         } else {
3302                 kind[0] = '\0';
3303                 ops = NULL;
3304         }
3305
3306         data = NULL;
3307         if (ops) {
3308                 if (ops->maxtype > RTNL_MAX_TYPE)
3309                         return -EINVAL;
3310
3311                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3312                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3313                                                           linkinfo[IFLA_INFO_DATA],
3314                                                           ops->policy, extack);
3315                         if (err < 0)
3316                                 return err;
3317                         data = attr;
3318                 }
3319                 if (ops->validate) {
3320                         err = ops->validate(tb, data, extack);
3321                         if (err < 0)
3322                                 return err;
3323                 }
3324         }
3325
3326         slave_data = NULL;
3327         if (m_ops) {
3328                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3329                         return -EINVAL;
3330
3331                 if (m_ops->slave_maxtype &&
3332                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3333                         err = nla_parse_nested_deprecated(slave_attr,
3334                                                           m_ops->slave_maxtype,
3335                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3336                                                           m_ops->slave_policy,
3337                                                           extack);
3338                         if (err < 0)
3339                                 return err;
3340                         slave_data = slave_attr;
3341                 }
3342         }
3343
3344         if (dev) {
3345                 int status = 0;
3346
3347                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3348                         return -EEXIST;
3349                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3350                         return -EOPNOTSUPP;
3351
3352                 if (linkinfo[IFLA_INFO_DATA]) {
3353                         if (!ops || ops != dev->rtnl_link_ops ||
3354                             !ops->changelink)
3355                                 return -EOPNOTSUPP;
3356
3357                         err = ops->changelink(dev, tb, data, extack);
3358                         if (err < 0)
3359                                 return err;
3360                         status |= DO_SETLINK_NOTIFY;
3361                 }
3362
3363                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3364                         if (!m_ops || !m_ops->slave_changelink)
3365                                 return -EOPNOTSUPP;
3366
3367                         err = m_ops->slave_changelink(master_dev, dev, tb,
3368                                                       slave_data, extack);
3369                         if (err < 0)
3370                                 return err;
3371                         status |= DO_SETLINK_NOTIFY;
3372                 }
3373
3374                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3375         }
3376
3377         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3378                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3379                         return rtnl_group_changelink(skb, net,
3380                                                 nla_get_u32(tb[IFLA_GROUP]),
3381                                                 ifm, extack, tb);
3382                 return -ENODEV;
3383         }
3384
3385         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3386                 return -EOPNOTSUPP;
3387
3388         if (!ops) {
3389 #ifdef CONFIG_MODULES
3390                 if (kind[0]) {
3391                         __rtnl_unlock();
3392                         request_module("rtnl-link-%s", kind);
3393                         rtnl_lock();
3394                         ops = rtnl_link_ops_get(kind);
3395                         if (ops)
3396                                 goto replay;
3397                 }
3398 #endif
3399                 NL_SET_ERR_MSG(extack, "Unknown device type");
3400                 return -EOPNOTSUPP;
3401         }
3402
3403         if (!ops->setup)
3404                 return -EOPNOTSUPP;
3405
3406         if (!ifname[0]) {
3407                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3408                 name_assign_type = NET_NAME_ENUM;
3409         }
3410
3411         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3412         if (IS_ERR(dest_net))
3413                 return PTR_ERR(dest_net);
3414
3415         if (tb[IFLA_LINK_NETNSID]) {
3416                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3417
3418                 link_net = get_net_ns_by_id(dest_net, id);
3419                 if (!link_net) {
3420                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3421                         err =  -EINVAL;
3422                         goto out;
3423                 }
3424                 err = -EPERM;
3425                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3426                         goto out;
3427         } else {
3428                 link_net = NULL;
3429         }
3430
3431         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3432                                name_assign_type, ops, tb, extack);
3433         if (IS_ERR(dev)) {
3434                 err = PTR_ERR(dev);
3435                 goto out;
3436         }
3437
3438         dev->ifindex = ifm->ifi_index;
3439
3440         if (ops->newlink) {
3441                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3442                 /* Drivers should call free_netdev() in ->destructor
3443                  * and unregister it on failure after registration
3444                  * so that device could be finally freed in rtnl_unlock.
3445                  */
3446                 if (err < 0) {
3447                         /* If device is not registered at all, free it now */
3448                         if (dev->reg_state == NETREG_UNINITIALIZED ||
3449                             dev->reg_state == NETREG_UNREGISTERED)
3450                                 free_netdev(dev);
3451                         goto out;
3452                 }
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 **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 = 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         if (!skb->len)
4846                 goto errout;
4847
4848         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4849         return 0;
4850 errout:
4851         WARN_ON(err == -EMSGSIZE);
4852         kfree_skb(skb);
4853         if (err)
4854                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4855         return err;
4856 }
4857
4858 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4859                                struct netlink_ext_ack *extack)
4860 {
4861         struct net *net = sock_net(skb->sk);
4862         struct ifinfomsg *ifm;
4863         struct net_device *dev;
4864         struct nlattr *br_spec, *attr = NULL;
4865         int rem, err = -EOPNOTSUPP;
4866         u16 flags = 0;
4867         bool have_flags = false;
4868
4869         if (nlmsg_len(nlh) < sizeof(*ifm))
4870                 return -EINVAL;
4871
4872         ifm = nlmsg_data(nlh);
4873         if (ifm->ifi_family != AF_BRIDGE)
4874                 return -EPFNOSUPPORT;
4875
4876         dev = __dev_get_by_index(net, ifm->ifi_index);
4877         if (!dev) {
4878                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4879                 return -ENODEV;
4880         }
4881
4882         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4883         if (br_spec) {
4884                 nla_for_each_nested(attr, br_spec, rem) {
4885                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4886                                 if (nla_len(attr) < sizeof(flags))
4887                                         return -EINVAL;
4888
4889                                 have_flags = true;
4890                                 flags = nla_get_u16(attr);
4891                                 break;
4892                         }
4893                 }
4894         }
4895
4896         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4897                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4898
4899                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4900                         err = -EOPNOTSUPP;
4901                         goto out;
4902                 }
4903
4904                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4905                                                              extack);
4906                 if (err)
4907                         goto out;
4908
4909                 flags &= ~BRIDGE_FLAGS_MASTER;
4910         }
4911
4912         if ((flags & BRIDGE_FLAGS_SELF)) {
4913                 if (!dev->netdev_ops->ndo_bridge_setlink)
4914                         err = -EOPNOTSUPP;
4915                 else
4916                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4917                                                                   flags,
4918                                                                   extack);
4919                 if (!err) {
4920                         flags &= ~BRIDGE_FLAGS_SELF;
4921
4922                         /* Generate event to notify upper layer of bridge
4923                          * change
4924                          */
4925                         err = rtnl_bridge_notify(dev);
4926                 }
4927         }
4928
4929         if (have_flags)
4930                 memcpy(nla_data(attr), &flags, sizeof(flags));
4931 out:
4932         return err;
4933 }
4934
4935 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4936                                struct netlink_ext_ack *extack)
4937 {
4938         struct net *net = sock_net(skb->sk);
4939         struct ifinfomsg *ifm;
4940         struct net_device *dev;
4941         struct nlattr *br_spec, *attr = NULL;
4942         int rem, err = -EOPNOTSUPP;
4943         u16 flags = 0;
4944         bool have_flags = false;
4945
4946         if (nlmsg_len(nlh) < sizeof(*ifm))
4947                 return -EINVAL;
4948
4949         ifm = nlmsg_data(nlh);
4950         if (ifm->ifi_family != AF_BRIDGE)
4951                 return -EPFNOSUPPORT;
4952
4953         dev = __dev_get_by_index(net, ifm->ifi_index);
4954         if (!dev) {
4955                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4956                 return -ENODEV;
4957         }
4958
4959         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4960         if (br_spec) {
4961                 nla_for_each_nested(attr, br_spec, rem) {
4962                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4963                                 if (nla_len(attr) < sizeof(flags))
4964                                         return -EINVAL;
4965
4966                                 have_flags = true;
4967                                 flags = nla_get_u16(attr);
4968                                 break;
4969                         }
4970                 }
4971         }
4972
4973         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4974                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4975
4976                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4977                         err = -EOPNOTSUPP;
4978                         goto out;
4979                 }
4980
4981                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4982                 if (err)
4983                         goto out;
4984
4985                 flags &= ~BRIDGE_FLAGS_MASTER;
4986         }
4987
4988         if ((flags & BRIDGE_FLAGS_SELF)) {
4989                 if (!dev->netdev_ops->ndo_bridge_dellink)
4990                         err = -EOPNOTSUPP;
4991                 else
4992                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4993                                                                   flags);
4994
4995                 if (!err) {
4996                         flags &= ~BRIDGE_FLAGS_SELF;
4997
4998                         /* Generate event to notify upper layer of bridge
4999                          * change
5000                          */
5001                         err = rtnl_bridge_notify(dev);
5002                 }
5003         }
5004
5005         if (have_flags)
5006                 memcpy(nla_data(attr), &flags, sizeof(flags));
5007 out:
5008         return err;
5009 }
5010
5011 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5012 {
5013         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5014                (!idxattr || idxattr == attrid);
5015 }
5016
5017 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5018 static int rtnl_get_offload_stats_attr_size(int attr_id)
5019 {
5020         switch (attr_id) {
5021         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5022                 return sizeof(struct rtnl_link_stats64);
5023         }
5024
5025         return 0;
5026 }
5027
5028 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5029                                   int *prividx)
5030 {
5031         struct nlattr *attr = NULL;
5032         int attr_id, size;
5033         void *attr_data;
5034         int err;
5035
5036         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5037               dev->netdev_ops->ndo_get_offload_stats))
5038                 return -ENODATA;
5039
5040         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5041              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5042                 if (attr_id < *prividx)
5043                         continue;
5044
5045                 size = rtnl_get_offload_stats_attr_size(attr_id);
5046                 if (!size)
5047                         continue;
5048
5049                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5050                         continue;
5051
5052                 attr = nla_reserve_64bit(skb, attr_id, size,
5053                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5054                 if (!attr)
5055                         goto nla_put_failure;
5056
5057                 attr_data = nla_data(attr);
5058                 memset(attr_data, 0, size);
5059                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5060                                                              attr_data);
5061                 if (err)
5062                         goto get_offload_stats_failure;
5063         }
5064
5065         if (!attr)
5066                 return -ENODATA;
5067
5068         *prividx = 0;
5069         return 0;
5070
5071 nla_put_failure:
5072         err = -EMSGSIZE;
5073 get_offload_stats_failure:
5074         *prividx = attr_id;
5075         return err;
5076 }
5077
5078 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5079 {
5080         int nla_size = 0;
5081         int attr_id;
5082         int size;
5083
5084         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5085               dev->netdev_ops->ndo_get_offload_stats))
5086                 return 0;
5087
5088         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5089              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5090                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5091                         continue;
5092                 size = rtnl_get_offload_stats_attr_size(attr_id);
5093                 nla_size += nla_total_size_64bit(size);
5094         }
5095
5096         if (nla_size != 0)
5097                 nla_size += nla_total_size(0);
5098
5099         return nla_size;
5100 }
5101
5102 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5103                                int type, u32 pid, u32 seq, u32 change,
5104                                unsigned int flags, unsigned int filter_mask,
5105                                int *idxattr, int *prividx)
5106 {
5107         struct if_stats_msg *ifsm;
5108         struct nlmsghdr *nlh;
5109         struct nlattr *attr;
5110         int s_prividx = *prividx;
5111         int err;
5112
5113         ASSERT_RTNL();
5114
5115         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5116         if (!nlh)
5117                 return -EMSGSIZE;
5118
5119         ifsm = nlmsg_data(nlh);
5120         ifsm->family = PF_UNSPEC;
5121         ifsm->pad1 = 0;
5122         ifsm->pad2 = 0;
5123         ifsm->ifindex = dev->ifindex;
5124         ifsm->filter_mask = filter_mask;
5125
5126         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5127                 struct rtnl_link_stats64 *sp;
5128
5129                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5130                                          sizeof(struct rtnl_link_stats64),
5131                                          IFLA_STATS_UNSPEC);
5132                 if (!attr)
5133                         goto nla_put_failure;
5134
5135                 sp = nla_data(attr);
5136                 dev_get_stats(dev, sp);
5137         }
5138
5139         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5140                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5141
5142                 if (ops && ops->fill_linkxstats) {
5143                         *idxattr = IFLA_STATS_LINK_XSTATS;
5144                         attr = nla_nest_start_noflag(skb,
5145                                                      IFLA_STATS_LINK_XSTATS);
5146                         if (!attr)
5147                                 goto nla_put_failure;
5148
5149                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5150                         nla_nest_end(skb, attr);
5151                         if (err)
5152                                 goto nla_put_failure;
5153                         *idxattr = 0;
5154                 }
5155         }
5156
5157         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5158                              *idxattr)) {
5159                 const struct rtnl_link_ops *ops = NULL;
5160                 const struct net_device *master;
5161
5162                 master = netdev_master_upper_dev_get(dev);
5163                 if (master)
5164                         ops = master->rtnl_link_ops;
5165                 if (ops && ops->fill_linkxstats) {
5166                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5167                         attr = nla_nest_start_noflag(skb,
5168                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5169                         if (!attr)
5170                                 goto nla_put_failure;
5171
5172                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5173                         nla_nest_end(skb, attr);
5174                         if (err)
5175                                 goto nla_put_failure;
5176                         *idxattr = 0;
5177                 }
5178         }
5179
5180         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5181                              *idxattr)) {
5182                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5183                 attr = nla_nest_start_noflag(skb,
5184                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5185                 if (!attr)
5186                         goto nla_put_failure;
5187
5188                 err = rtnl_get_offload_stats(skb, dev, prividx);
5189                 if (err == -ENODATA)
5190                         nla_nest_cancel(skb, attr);
5191                 else
5192                         nla_nest_end(skb, attr);
5193
5194                 if (err && err != -ENODATA)
5195                         goto nla_put_failure;
5196                 *idxattr = 0;
5197         }
5198
5199         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5200                 struct rtnl_af_ops *af_ops;
5201
5202                 *idxattr = IFLA_STATS_AF_SPEC;
5203                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5204                 if (!attr)
5205                         goto nla_put_failure;
5206
5207                 rcu_read_lock();
5208                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5209                         if (af_ops->fill_stats_af) {
5210                                 struct nlattr *af;
5211                                 int err;
5212
5213                                 af = nla_nest_start_noflag(skb,
5214                                                            af_ops->family);
5215                                 if (!af) {
5216                                         rcu_read_unlock();
5217                                         goto nla_put_failure;
5218                                 }
5219                                 err = af_ops->fill_stats_af(skb, dev);
5220
5221                                 if (err == -ENODATA) {
5222                                         nla_nest_cancel(skb, af);
5223                                 } else if (err < 0) {
5224                                         rcu_read_unlock();
5225                                         goto nla_put_failure;
5226                                 }
5227
5228                                 nla_nest_end(skb, af);
5229                         }
5230                 }
5231                 rcu_read_unlock();
5232
5233                 nla_nest_end(skb, attr);
5234
5235                 *idxattr = 0;
5236         }
5237
5238         nlmsg_end(skb, nlh);
5239
5240         return 0;
5241
5242 nla_put_failure:
5243         /* not a multi message or no progress mean a real error */
5244         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5245                 nlmsg_cancel(skb, nlh);
5246         else
5247                 nlmsg_end(skb, nlh);
5248
5249         return -EMSGSIZE;
5250 }
5251
5252 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5253                                   u32 filter_mask)
5254 {
5255         size_t size = 0;
5256
5257         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5258                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5259
5260         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5261                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5262                 int attr = IFLA_STATS_LINK_XSTATS;
5263
5264                 if (ops && ops->get_linkxstats_size) {
5265                         size += nla_total_size(ops->get_linkxstats_size(dev,
5266                                                                         attr));
5267                         /* for IFLA_STATS_LINK_XSTATS */
5268                         size += nla_total_size(0);
5269                 }
5270         }
5271
5272         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5273                 struct net_device *_dev = (struct net_device *)dev;
5274                 const struct rtnl_link_ops *ops = NULL;
5275                 const struct net_device *master;
5276
5277                 /* netdev_master_upper_dev_get can't take const */
5278                 master = netdev_master_upper_dev_get(_dev);
5279                 if (master)
5280                         ops = master->rtnl_link_ops;
5281                 if (ops && ops->get_linkxstats_size) {
5282                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5283
5284                         size += nla_total_size(ops->get_linkxstats_size(dev,
5285                                                                         attr));
5286                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5287                         size += nla_total_size(0);
5288                 }
5289         }
5290
5291         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5292                 size += rtnl_get_offload_stats_size(dev);
5293
5294         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5295                 struct rtnl_af_ops *af_ops;
5296
5297                 /* for IFLA_STATS_AF_SPEC */
5298                 size += nla_total_size(0);
5299
5300                 rcu_read_lock();
5301                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5302                         if (af_ops->get_stats_af_size) {
5303                                 size += nla_total_size(
5304                                         af_ops->get_stats_af_size(dev));
5305
5306                                 /* for AF_* */
5307                                 size += nla_total_size(0);
5308                         }
5309                 }
5310                 rcu_read_unlock();
5311         }
5312
5313         return size;
5314 }
5315
5316 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5317                                 bool is_dump, struct netlink_ext_ack *extack)
5318 {
5319         struct if_stats_msg *ifsm;
5320
5321         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5322                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5323                 return -EINVAL;
5324         }
5325
5326         if (!strict_check)
5327                 return 0;
5328
5329         ifsm = nlmsg_data(nlh);
5330
5331         /* only requests using strict checks can pass data to influence
5332          * the dump. The legacy exception is filter_mask.
5333          */
5334         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5335                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5336                 return -EINVAL;
5337         }
5338         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5339                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5340                 return -EINVAL;
5341         }
5342         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5343                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5344                 return -EINVAL;
5345         }
5346
5347         return 0;
5348 }
5349
5350 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5351                           struct netlink_ext_ack *extack)
5352 {
5353         struct net *net = sock_net(skb->sk);
5354         struct net_device *dev = NULL;
5355         int idxattr = 0, prividx = 0;
5356         struct if_stats_msg *ifsm;
5357         struct sk_buff *nskb;
5358         u32 filter_mask;
5359         int err;
5360
5361         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5362                                    false, extack);
5363         if (err)
5364                 return err;
5365
5366         ifsm = nlmsg_data(nlh);
5367         if (ifsm->ifindex > 0)
5368                 dev = __dev_get_by_index(net, ifsm->ifindex);
5369         else
5370                 return -EINVAL;
5371
5372         if (!dev)
5373                 return -ENODEV;
5374
5375         filter_mask = ifsm->filter_mask;
5376         if (!filter_mask)
5377                 return -EINVAL;
5378
5379         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5380         if (!nskb)
5381                 return -ENOBUFS;
5382
5383         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5384                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5385                                   0, filter_mask, &idxattr, &prividx);
5386         if (err < 0) {
5387                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5388                 WARN_ON(err == -EMSGSIZE);
5389                 kfree_skb(nskb);
5390         } else {
5391                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5392         }
5393
5394         return err;
5395 }
5396
5397 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5398 {
5399         struct netlink_ext_ack *extack = cb->extack;
5400         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5401         struct net *net = sock_net(skb->sk);
5402         unsigned int flags = NLM_F_MULTI;
5403         struct if_stats_msg *ifsm;
5404         struct hlist_head *head;
5405         struct net_device *dev;
5406         u32 filter_mask = 0;
5407         int idx = 0;
5408
5409         s_h = cb->args[0];
5410         s_idx = cb->args[1];
5411         s_idxattr = cb->args[2];
5412         s_prividx = cb->args[3];
5413
5414         cb->seq = net->dev_base_seq;
5415
5416         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5417         if (err)
5418                 return err;
5419
5420         ifsm = nlmsg_data(cb->nlh);
5421         filter_mask = ifsm->filter_mask;
5422         if (!filter_mask) {
5423                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5424                 return -EINVAL;
5425         }
5426
5427         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5428                 idx = 0;
5429                 head = &net->dev_index_head[h];
5430                 hlist_for_each_entry(dev, head, index_hlist) {
5431                         if (idx < s_idx)
5432                                 goto cont;
5433                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5434                                                   NETLINK_CB(cb->skb).portid,
5435                                                   cb->nlh->nlmsg_seq, 0,
5436                                                   flags, filter_mask,
5437                                                   &s_idxattr, &s_prividx);
5438                         /* If we ran out of room on the first message,
5439                          * we're in trouble
5440                          */
5441                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5442
5443                         if (err < 0)
5444                                 goto out;
5445                         s_prividx = 0;
5446                         s_idxattr = 0;
5447                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5448 cont:
5449                         idx++;
5450                 }
5451         }
5452 out:
5453         cb->args[3] = s_prividx;
5454         cb->args[2] = s_idxattr;
5455         cb->args[1] = idx;
5456         cb->args[0] = h;
5457
5458         return skb->len;
5459 }
5460
5461 /* Process one rtnetlink message. */
5462
5463 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5464                              struct netlink_ext_ack *extack)
5465 {
5466         struct net *net = sock_net(skb->sk);
5467         struct rtnl_link *link;
5468         struct module *owner;
5469         int err = -EOPNOTSUPP;
5470         rtnl_doit_func doit;
5471         unsigned int flags;
5472         int kind;
5473         int family;
5474         int type;
5475
5476         type = nlh->nlmsg_type;
5477         if (type > RTM_MAX)
5478                 return -EOPNOTSUPP;
5479
5480         type -= RTM_BASE;
5481
5482         /* All the messages must have at least 1 byte length */
5483         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5484                 return 0;
5485
5486         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5487         kind = type&3;
5488
5489         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5490                 return -EPERM;
5491
5492         rcu_read_lock();
5493         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5494                 struct sock *rtnl;
5495                 rtnl_dumpit_func dumpit;
5496                 u32 min_dump_alloc = 0;
5497
5498                 link = rtnl_get_link(family, type);
5499                 if (!link || !link->dumpit) {
5500                         family = PF_UNSPEC;
5501                         link = rtnl_get_link(family, type);
5502                         if (!link || !link->dumpit)
5503                                 goto err_unlock;
5504                 }
5505                 owner = link->owner;
5506                 dumpit = link->dumpit;
5507
5508                 if (type == RTM_GETLINK - RTM_BASE)
5509                         min_dump_alloc = rtnl_calcit(skb, nlh);
5510
5511                 err = 0;
5512                 /* need to do this before rcu_read_unlock() */
5513                 if (!try_module_get(owner))
5514                         err = -EPROTONOSUPPORT;
5515
5516                 rcu_read_unlock();
5517
5518                 rtnl = net->rtnl;
5519                 if (err == 0) {
5520                         struct netlink_dump_control c = {
5521                                 .dump           = dumpit,
5522                                 .min_dump_alloc = min_dump_alloc,
5523                                 .module         = owner,
5524                         };
5525                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5526                         /* netlink_dump_start() will keep a reference on
5527                          * module if dump is still in progress.
5528                          */
5529                         module_put(owner);
5530                 }
5531                 return err;
5532         }
5533
5534         link = rtnl_get_link(family, type);
5535         if (!link || !link->doit) {
5536                 family = PF_UNSPEC;
5537                 link = rtnl_get_link(PF_UNSPEC, type);
5538                 if (!link || !link->doit)
5539                         goto out_unlock;
5540         }
5541
5542         owner = link->owner;
5543         if (!try_module_get(owner)) {
5544                 err = -EPROTONOSUPPORT;
5545                 goto out_unlock;
5546         }
5547
5548         flags = link->flags;
5549         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5550                 doit = link->doit;
5551                 rcu_read_unlock();
5552                 if (doit)
5553                         err = doit(skb, nlh, extack);
5554                 module_put(owner);
5555                 return err;
5556         }
5557         rcu_read_unlock();
5558
5559         rtnl_lock();
5560         link = rtnl_get_link(family, type);
5561         if (link && link->doit)
5562                 err = link->doit(skb, nlh, extack);
5563         rtnl_unlock();
5564
5565         module_put(owner);
5566
5567         return err;
5568
5569 out_unlock:
5570         rcu_read_unlock();
5571         return err;
5572
5573 err_unlock:
5574         rcu_read_unlock();
5575         return -EOPNOTSUPP;
5576 }
5577
5578 static void rtnetlink_rcv(struct sk_buff *skb)
5579 {
5580         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5581 }
5582
5583 static int rtnetlink_bind(struct net *net, int group)
5584 {
5585         switch (group) {
5586         case RTNLGRP_IPV4_MROUTE_R:
5587         case RTNLGRP_IPV6_MROUTE_R:
5588                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5589                         return -EPERM;
5590                 break;
5591         }
5592         return 0;
5593 }
5594
5595 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5596 {
5597         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5598
5599         switch (event) {
5600         case NETDEV_REBOOT:
5601         case NETDEV_CHANGEMTU:
5602         case NETDEV_CHANGEADDR:
5603         case NETDEV_CHANGENAME:
5604         case NETDEV_FEAT_CHANGE:
5605         case NETDEV_BONDING_FAILOVER:
5606         case NETDEV_POST_TYPE_CHANGE:
5607         case NETDEV_NOTIFY_PEERS:
5608         case NETDEV_CHANGEUPPER:
5609         case NETDEV_RESEND_IGMP:
5610         case NETDEV_CHANGEINFODATA:
5611         case NETDEV_CHANGELOWERSTATE:
5612         case NETDEV_CHANGE_TX_QUEUE_LEN:
5613                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5614                                    GFP_KERNEL, NULL, 0);
5615                 break;
5616         default:
5617                 break;
5618         }
5619         return NOTIFY_DONE;
5620 }
5621
5622 static struct notifier_block rtnetlink_dev_notifier = {
5623         .notifier_call  = rtnetlink_event,
5624 };
5625
5626
5627 static int __net_init rtnetlink_net_init(struct net *net)
5628 {
5629         struct sock *sk;
5630         struct netlink_kernel_cfg cfg = {
5631                 .groups         = RTNLGRP_MAX,
5632                 .input          = rtnetlink_rcv,
5633                 .cb_mutex       = &rtnl_mutex,
5634                 .flags          = NL_CFG_F_NONROOT_RECV,
5635                 .bind           = rtnetlink_bind,
5636         };
5637
5638         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5639         if (!sk)
5640                 return -ENOMEM;
5641         net->rtnl = sk;
5642         return 0;
5643 }
5644
5645 static void __net_exit rtnetlink_net_exit(struct net *net)
5646 {
5647         netlink_kernel_release(net->rtnl);
5648         net->rtnl = NULL;
5649 }
5650
5651 static struct pernet_operations rtnetlink_net_ops = {
5652         .init = rtnetlink_net_init,
5653         .exit = rtnetlink_net_exit,
5654 };
5655
5656 void __init rtnetlink_init(void)
5657 {
5658         if (register_pernet_subsys(&rtnetlink_net_ops))
5659                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5660
5661         register_netdevice_notifier(&rtnetlink_dev_notifier);
5662
5663         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5664                       rtnl_dump_ifinfo, 0);
5665         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5666         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5667         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5668
5669         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5670         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5671         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5672
5673         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5674         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5675
5676         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5677         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5678         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5679
5680         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5681         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5682         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5683
5684         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5685                       0);
5686 }