net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/bitops.h>
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/kernel.h>
  25 #include <linux/timer.h>
  26 #include <linux/string.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/fcntl.h>
  30 #include <linux/mm.h>
  31 #include <linux/slab.h>
  32 #include <linux/interrupt.h>
  33 #include <linux/capability.h>
  34 #include <linux/skbuff.h>
  35 #include <linux/init.h>
  36 #include <linux/security.h>
  37 #include <linux/mutex.h>
  38 #include <linux/if_addr.h>
  39 #include <linux/if_bridge.h>
  40 #include <linux/if_vlan.h>
  41 #include <linux/pci.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/bpf.h>
  44
  45 #include <linux/uaccess.h>
  46
  47 #include <linux/inet.h>
  48 #include <linux/netdevice.h>
  49 #include <net/switchdev.h>
  50 #include <net/ip.h>
  51 #include <net/protocol.h>
  52 #include <net/arp.h>
  53 #include <net/route.h>
  54 #include <net/udp.h>
  55 #include <net/tcp.h>
  56 #include <net/sock.h>
  57 #include <net/pkt_sched.h>
  58 #include <net/fib_rules.h>
  59 #include <net/rtnetlink.h>
  60 #include <net/net_namespace.h>
  61
  62 struct rtnl_link {
  63         rtnl_doit_func          doit;
  64         rtnl_dumpit_func        dumpit;
  65         struct module           *owner;
  66         unsigned int            flags;
  67         struct rcu_head         rcu;
  68 };
  69
  70 static DEFINE_MUTEX(rtnl_mutex);
  71
  72 void rtnl_lock(void)
  73 {
  74         mutex_lock(&rtnl_mutex);
  75 }
  76 EXPORT_SYMBOL(rtnl_lock);
  77
  78 int rtnl_lock_killable(void)
  79 {
  80         return mutex_lock_killable(&rtnl_mutex);
  81 }
  82 EXPORT_SYMBOL(rtnl_lock_killable);
  83
  84 static struct sk_buff *defer_kfree_skb_list;
  85 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  86 {
  87         if (head && tail) {
  88                 tail->next = defer_kfree_skb_list;
  89                 defer_kfree_skb_list = head;
  90         }
  91 }
  92 EXPORT_SYMBOL(rtnl_kfree_skbs);
  93
  94 void __rtnl_unlock(void)
  95 {
  96         struct sk_buff *head = defer_kfree_skb_list;
  97
  98         defer_kfree_skb_list = NULL;
  99
 100         mutex_unlock(&rtnl_mutex);
 101
 102         while (head) {
 103                 struct sk_buff *next = head->next;
 104
 105                 kfree_skb(head);
 106                 cond_resched();
 107                 head = next;
 108         }
 109 }
 110
 111 void rtnl_unlock(void)
 112 {
 113         /* This fellow will unlock it for us. */
 114         netdev_run_todo();
 115 }
 116 EXPORT_SYMBOL(rtnl_unlock);
 117
 118 int rtnl_trylock(void)
 119 {
 120         return mutex_trylock(&rtnl_mutex);
 121 }
 122 EXPORT_SYMBOL(rtnl_trylock);
 123
 124 int rtnl_is_locked(void)
 125 {
 126         return mutex_is_locked(&rtnl_mutex);
 127 }
 128 EXPORT_SYMBOL(rtnl_is_locked);
 129
 130 #ifdef CONFIG_PROVE_LOCKING
 131 bool lockdep_rtnl_is_held(void)
 132 {
 133         return lockdep_is_held(&rtnl_mutex);
 134 }
 135 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 136 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 137
 138 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 139
 140 static inline int rtm_msgindex(int msgtype)
 141 {
 142         int msgindex = msgtype - RTM_BASE;
 143
 144         /*
 145          * msgindex < 0 implies someone tried to register a netlink
 146          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 147          * the message type has not been added to linux/rtnetlink.h
 148          */
 149         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 150
 151         return msgindex;
 152 }
 153
 154 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 155 {
 156         struct rtnl_link **tab;
 157
 158         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 159                 protocol = PF_UNSPEC;
 160
 161         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 162         if (!tab)
 163                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 164
 165         return tab[msgtype];
 166 }
 167
 168 static int rtnl_register_internal(struct module *owner,
 169                                   int protocol, int msgtype,
 170                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 171                                   unsigned int flags)
 172 {
 173         struct rtnl_link *link, *old;
 174         struct rtnl_link __rcu **tab;
 175         int msgindex;
 176         int ret = -ENOBUFS;
 177
 178         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 179         msgindex = rtm_msgindex(msgtype);
 180
 181         rtnl_lock();
 182         tab = rtnl_msg_handlers[protocol];
 183         if (tab == NULL) {
 184                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 185                 if (!tab)
 186                         goto unlock;
 187
 188                 /* ensures we see the 0 stores */
 189                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 190         }
 191
 192         old = rtnl_dereference(tab[msgindex]);
 193         if (old) {
 194                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 195                 if (!link)
 196                         goto unlock;
 197         } else {
 198                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         }
 202
 203         WARN_ON(link->owner && link->owner != owner);
 204         link->owner = owner;
 205
 206         WARN_ON(doit && link->doit && link->doit != doit);
 207         if (doit)
 208                 link->doit = doit;
 209         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 210         if (dumpit)
 211                 link->dumpit = dumpit;
 212
 213         link->flags |= flags;
 214
 215         /* publish protocol:msgtype */
 216         rcu_assign_pointer(tab[msgindex], link);
 217         ret = 0;
 218         if (old)
 219                 kfree_rcu(old, rcu);
 220 unlock:
 221         rtnl_unlock();
 222         return ret;
 223 }
 224
 225 /**
 226  * rtnl_register_module - Register a rtnetlink message type
 227  *
 228  * @owner: module registering the hook (THIS_MODULE)
 229  * @protocol: Protocol family or PF_UNSPEC
 230  * @msgtype: rtnetlink message type
 231  * @doit: Function pointer called for each request message
 232  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 233  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 234  *
 235  * Like rtnl_register, but for use by removable modules.
 236  */
 237 int rtnl_register_module(struct module *owner,
 238                          int protocol, int msgtype,
 239                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 240                          unsigned int flags)
 241 {
 242         return rtnl_register_internal(owner, protocol, msgtype,
 243                                       doit, dumpit, flags);
 244 }
 245 EXPORT_SYMBOL_GPL(rtnl_register_module);
 246
 247 /**
 248  * rtnl_register - Register a rtnetlink message type
 249  * @protocol: Protocol family or PF_UNSPEC
 250  * @msgtype: rtnetlink message type
 251  * @doit: Function pointer called for each request message
 252  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 253  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 254  *
 255  * Registers the specified function pointers (at least one of them has
 256  * to be non-NULL) to be called whenever a request message for the
 257  * specified protocol family and message type is received.
 258  *
 259  * The special protocol family PF_UNSPEC may be used to define fallback
 260  * function pointers for the case when no entry for the specific protocol
 261  * family exists.
 262  */
 263 void rtnl_register(int protocol, int msgtype,
 264                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 265                    unsigned int flags)
 266 {
 267         int err;
 268
 269         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 270                                      flags);
 271         if (err)
 272                 pr_err("Unable to register rtnetlink message handler, "
 273                        "protocol = %d, message type = %d\n", protocol, msgtype);
 274 }
 275
 276 /**
 277  * rtnl_unregister - Unregister a rtnetlink message type
 278  * @protocol: Protocol family or PF_UNSPEC
 279  * @msgtype: rtnetlink message type
 280  *
 281  * Returns 0 on success or a negative error code.
 282  */
 283 int rtnl_unregister(int protocol, int msgtype)
 284 {
 285         struct rtnl_link **tab, *link;
 286         int msgindex;
 287
 288         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 289         msgindex = rtm_msgindex(msgtype);
 290
 291         rtnl_lock();
 292         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 293         if (!tab) {
 294                 rtnl_unlock();
 295                 return -ENOENT;
 296         }
 297
 298         link = tab[msgindex];
 299         rcu_assign_pointer(tab[msgindex], NULL);
 300         rtnl_unlock();
 301
 302         kfree_rcu(link, rcu);
 303
 304         return 0;
 305 }
 306 EXPORT_SYMBOL_GPL(rtnl_unregister);
 307
 308 /**
 309  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 310  * @protocol : Protocol family or PF_UNSPEC
 311  *
 312  * Identical to calling rtnl_unregster() for all registered message types
 313  * of a certain protocol family.
 314  */
 315 void rtnl_unregister_all(int protocol)
 316 {
 317         struct rtnl_link **tab, *link;
 318         int msgindex;
 319
 320         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 321
 322         rtnl_lock();
 323         tab = rtnl_msg_handlers[protocol];
 324         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 325         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 326                 link = tab[msgindex];
 327                 if (!link)
 328                         continue;
 329
 330                 rcu_assign_pointer(tab[msgindex], NULL);
 331                 kfree_rcu(link, rcu);
 332         }
 333         rtnl_unlock();
 334
 335         synchronize_net();
 336
 337         kfree(tab);
 338 }
 339 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 340
 341 static LIST_HEAD(link_ops);
 342
 343 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 344 {
 345         const struct rtnl_link_ops *ops;
 346
 347         list_for_each_entry(ops, &link_ops, list) {
 348                 if (!strcmp(ops->kind, kind))
 349                         return ops;
 350         }
 351         return NULL;
 352 }
 353
 354 /**
 355  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 356  * @ops: struct rtnl_link_ops * to register
 357  *
 358  * The caller must hold the rtnl_mutex. This function should be used
 359  * by drivers that create devices during module initialization. It
 360  * must be called before registering the devices.
 361  *
 362  * Returns 0 on success or a negative error code.
 363  */
 364 int __rtnl_link_register(struct rtnl_link_ops *ops)
 365 {
 366         if (rtnl_link_ops_get(ops->kind))
 367                 return -EEXIST;
 368
 369         /* The check for setup is here because if ops
 370          * does not have that filled up, it is not possible
 371          * to use the ops for creating device. So do not
 372          * fill up dellink as well. That disables rtnl_dellink.
 373          */
 374         if (ops->setup && !ops->dellink)
 375                 ops->dellink = unregister_netdevice_queue;
 376
 377         list_add_tail(&ops->list, &link_ops);
 378         return 0;
 379 }
 380 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 381
 382 /**
 383  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 384  * @ops: struct rtnl_link_ops * to register
 385  *
 386  * Returns 0 on success or a negative error code.
 387  */
 388 int rtnl_link_register(struct rtnl_link_ops *ops)
 389 {
 390         int err;
 391
 392         rtnl_lock();
 393         err = __rtnl_link_register(ops);
 394         rtnl_unlock();
 395         return err;
 396 }
 397 EXPORT_SYMBOL_GPL(rtnl_link_register);
 398
 399 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 400 {
 401         struct net_device *dev;
 402         LIST_HEAD(list_kill);
 403
 404         for_each_netdev(net, dev) {
 405                 if (dev->rtnl_link_ops == ops)
 406                         ops->dellink(dev, &list_kill);
 407         }
 408         unregister_netdevice_many(&list_kill);
 409 }
 410
 411 /**
 412  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 413  * @ops: struct rtnl_link_ops * to unregister
 414  *
 415  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 416  * integrity (hold pernet_ops_rwsem for writing to close the race
 417  * with setup_net() and cleanup_net()).
 418  */
 419 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 420 {
 421         struct net *net;
 422
 423         for_each_net(net) {
 424                 __rtnl_kill_links(net, ops);
 425         }
 426         list_del(&ops->list);
 427 }
 428 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 429
 430 /* Return with the rtnl_lock held when there are no network
 431  * devices unregistering in any network namespace.
 432  */
 433 static void rtnl_lock_unregistering_all(void)
 434 {
 435         struct net *net;
 436         bool unregistering;
 437         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 438
 439         add_wait_queue(&netdev_unregistering_wq, &wait);
 440         for (;;) {
 441                 unregistering = false;
 442                 rtnl_lock();
 443                 /* We held write locked pernet_ops_rwsem, and parallel
 444                  * setup_net() and cleanup_net() are not possible.
 445                  */
 446                 for_each_net(net) {
 447                         if (net->dev_unreg_count > 0) {
 448                                 unregistering = true;
 449                                 break;
 450                         }
 451                 }
 452                 if (!unregistering)
 453                         break;
 454                 __rtnl_unlock();
 455
 456                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 457         }
 458         remove_wait_queue(&netdev_unregistering_wq, &wait);
 459 }
 460
 461 /**
 462  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 463  * @ops: struct rtnl_link_ops * to unregister
 464  */
 465 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 466 {
 467         /* Close the race with setup_net() and cleanup_net() */
 468         down_write(&pernet_ops_rwsem);
 469         rtnl_lock_unregistering_all();
 470         __rtnl_link_unregister(ops);
 471         rtnl_unlock();
 472         up_write(&pernet_ops_rwsem);
 473 }
 474 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 475
 476 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 477 {
 478         struct net_device *master_dev;
 479         const struct rtnl_link_ops *ops;
 480         size_t size = 0;
 481
 482         rcu_read_lock();
 483
 484         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 485         if (!master_dev)
 486                 goto out;
 487
 488         ops = master_dev->rtnl_link_ops;
 489         if (!ops || !ops->get_slave_size)
 490                 goto out;
 491         /* IFLA_INFO_SLAVE_DATA + nested data */
 492         size = nla_total_size(sizeof(struct nlattr)) +
 493                ops->get_slave_size(master_dev, dev);
 494
 495 out:
 496         rcu_read_unlock();
 497         return size;
 498 }
 499
 500 static size_t rtnl_link_get_size(const struct net_device *dev)
 501 {
 502         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 503         size_t size;
 504
 505         if (!ops)
 506                 return 0;
 507
 508         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 509                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 510
 511         if (ops->get_size)
 512                 /* IFLA_INFO_DATA + nested data */
 513                 size += nla_total_size(sizeof(struct nlattr)) +
 514                         ops->get_size(dev);
 515
 516         if (ops->get_xstats_size)
 517                 /* IFLA_INFO_XSTATS */
 518                 size += nla_total_size(ops->get_xstats_size(dev));
 519
 520         size += rtnl_link_get_slave_info_data_size(dev);
 521
 522         return size;
 523 }
 524
 525 static LIST_HEAD(rtnl_af_ops);
 526
 527 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 528 {
 529         const struct rtnl_af_ops *ops;
 530
 531         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 532                 if (ops->family == family)
 533                         return ops;
 534         }
 535
 536         return NULL;
 537 }
 538
 539 /**
 540  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 541  * @ops: struct rtnl_af_ops * to register
 542  *
 543  * Returns 0 on success or a negative error code.
 544  */
 545 void rtnl_af_register(struct rtnl_af_ops *ops)
 546 {
 547         rtnl_lock();
 548         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 549         rtnl_unlock();
 550 }
 551 EXPORT_SYMBOL_GPL(rtnl_af_register);
 552
 553 /**
 554  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 555  * @ops: struct rtnl_af_ops * to unregister
 556  */
 557 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 558 {
 559         rtnl_lock();
 560         list_del_rcu(&ops->list);
 561         rtnl_unlock();
 562
 563         synchronize_rcu();
 564 }
 565 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 566
 567 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 568                                     u32 ext_filter_mask)
 569 {
 570         struct rtnl_af_ops *af_ops;
 571         size_t size;
 572
 573         /* IFLA_AF_SPEC */
 574         size = nla_total_size(sizeof(struct nlattr));
 575
 576         rcu_read_lock();
 577         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 578                 if (af_ops->get_link_af_size) {
 579                         /* AF_* + nested data */
 580                         size += nla_total_size(sizeof(struct nlattr)) +
 581                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 582                 }
 583         }
 584         rcu_read_unlock();
 585
 586         return size;
 587 }
 588
 589 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 590 {
 591         struct net_device *master_dev;
 592         bool ret = false;
 593
 594         rcu_read_lock();
 595
 596         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 597         if (master_dev && master_dev->rtnl_link_ops)
 598                 ret = true;
 599         rcu_read_unlock();
 600         return ret;
 601 }
 602
 603 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 604                                      const struct net_device *dev)
 605 {
 606         struct net_device *master_dev;
 607         const struct rtnl_link_ops *ops;
 608         struct nlattr *slave_data;
 609         int err;
 610
 611         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 612         if (!master_dev)
 613                 return 0;
 614         ops = master_dev->rtnl_link_ops;
 615         if (!ops)
 616                 return 0;
 617         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 618                 return -EMSGSIZE;
 619         if (ops->fill_slave_info) {
 620                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 621                 if (!slave_data)
 622                         return -EMSGSIZE;
 623                 err = ops->fill_slave_info(skb, master_dev, dev);
 624                 if (err < 0)
 625                         goto err_cancel_slave_data;
 626                 nla_nest_end(skb, slave_data);
 627         }
 628         return 0;
 629
 630 err_cancel_slave_data:
 631         nla_nest_cancel(skb, slave_data);
 632         return err;
 633 }
 634
 635 static int rtnl_link_info_fill(struct sk_buff *skb,
 636                                const struct net_device *dev)
 637 {
 638         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 639         struct nlattr *data;
 640         int err;
 641
 642         if (!ops)
 643                 return 0;
 644         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 645                 return -EMSGSIZE;
 646         if (ops->fill_xstats) {
 647                 err = ops->fill_xstats(skb, dev);
 648                 if (err < 0)
 649                         return err;
 650         }
 651         if (ops->fill_info) {
 652                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 653                 if (data == NULL)
 654                         return -EMSGSIZE;
 655                 err = ops->fill_info(skb, dev);
 656                 if (err < 0)
 657                         goto err_cancel_data;
 658                 nla_nest_end(skb, data);
 659         }
 660         return 0;
 661
 662 err_cancel_data:
 663         nla_nest_cancel(skb, data);
 664         return err;
 665 }
 666
 667 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 668 {
 669         struct nlattr *linkinfo;
 670         int err = -EMSGSIZE;
 671
 672         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 673         if (linkinfo == NULL)
 674                 goto out;
 675
 676         err = rtnl_link_info_fill(skb, dev);
 677         if (err < 0)
 678                 goto err_cancel_link;
 679
 680         err = rtnl_link_slave_info_fill(skb, dev);
 681         if (err < 0)
 682                 goto err_cancel_link;
 683
 684         nla_nest_end(skb, linkinfo);
 685         return 0;
 686
 687 err_cancel_link:
 688         nla_nest_cancel(skb, linkinfo);
 689 out:
 690         return err;
 691 }
 692
 693 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 694 {
 695         struct sock *rtnl = net->rtnl;
 696         int err = 0;
 697
 698         NETLINK_CB(skb).dst_group = group;
 699         if (echo)
 700                 refcount_inc(&skb->users);
 701         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 702         if (echo)
 703                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 704         return err;
 705 }
 706
 707 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 708 {
 709         struct sock *rtnl = net->rtnl;
 710
 711         return nlmsg_unicast(rtnl, skb, pid);
 712 }
 713 EXPORT_SYMBOL(rtnl_unicast);
 714
 715 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 716                  struct nlmsghdr *nlh, gfp_t flags)
 717 {
 718         struct sock *rtnl = net->rtnl;
 719         int report = 0;
 720
 721         if (nlh)
 722                 report = nlmsg_report(nlh);
 723
 724         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 725 }
 726 EXPORT_SYMBOL(rtnl_notify);
 727
 728 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 729 {
 730         struct sock *rtnl = net->rtnl;
 731
 732         netlink_set_err(rtnl, 0, group, error);
 733 }
 734 EXPORT_SYMBOL(rtnl_set_sk_err);
 735
 736 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 737 {
 738         struct nlattr *mx;
 739         int i, valid = 0;
 740
 741         mx = nla_nest_start(skb, RTA_METRICS);
 742         if (mx == NULL)
 743                 return -ENOBUFS;
 744
 745         for (i = 0; i < RTAX_MAX; i++) {
 746                 if (metrics[i]) {
 747                         if (i == RTAX_CC_ALGO - 1) {
 748                                 char tmp[TCP_CA_NAME_MAX], *name;
 749
 750                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 751                                 if (!name)
 752                                         continue;
 753                                 if (nla_put_string(skb, i + 1, name))
 754                                         goto nla_put_failure;
 755                         } else if (i == RTAX_FEATURES - 1) {
 756                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 757
 758                                 if (!user_features)
 759                                         continue;
 760                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 761                                 if (nla_put_u32(skb, i + 1, user_features))
 762                                         goto nla_put_failure;
 763                         } else {
 764                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 765                                         goto nla_put_failure;
 766                         }
 767                         valid++;
 768                 }
 769         }
 770
 771         if (!valid) {
 772                 nla_nest_cancel(skb, mx);
 773                 return 0;
 774         }
 775
 776         return nla_nest_end(skb, mx);
 777
 778 nla_put_failure:
 779         nla_nest_cancel(skb, mx);
 780         return -EMSGSIZE;
 781 }
 782 EXPORT_SYMBOL(rtnetlink_put_metrics);
 783
 784 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 785                        long expires, u32 error)
 786 {
 787         struct rta_cacheinfo ci = {
 788                 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
 789                 .rta_used = dst->__use,
 790                 .rta_clntref = atomic_read(&(dst->__refcnt)),
 791                 .rta_error = error,
 792                 .rta_id =  id,
 793         };
 794
 795         if (expires) {
 796                 unsigned long clock;
 797
 798                 clock = jiffies_to_clock_t(abs(expires));
 799                 clock = min_t(unsigned long, clock, INT_MAX);
 800                 ci.rta_expires = (expires > 0) ? clock : -clock;
 801         }
 802         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 803 }
 804 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 805
 806 static void set_operstate(struct net_device *dev, unsigned char transition)
 807 {
 808         unsigned char operstate = dev->operstate;
 809
 810         switch (transition) {
 811         case IF_OPER_UP:
 812                 if ((operstate == IF_OPER_DORMANT ||
 813                      operstate == IF_OPER_UNKNOWN) &&
 814                     !netif_dormant(dev))
 815                         operstate = IF_OPER_UP;
 816                 break;
 817
 818         case IF_OPER_DORMANT:
 819                 if (operstate == IF_OPER_UP ||
 820                     operstate == IF_OPER_UNKNOWN)
 821                         operstate = IF_OPER_DORMANT;
 822                 break;
 823         }
 824
 825         if (dev->operstate != operstate) {
 826                 write_lock_bh(&dev_base_lock);
 827                 dev->operstate = operstate;
 828                 write_unlock_bh(&dev_base_lock);
 829                 netdev_state_change(dev);
 830         }
 831 }
 832
 833 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 834 {
 835         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 836                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 837 }
 838
 839 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 840                                            const struct ifinfomsg *ifm)
 841 {
 842         unsigned int flags = ifm->ifi_flags;
 843
 844         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 845         if (ifm->ifi_change)
 846                 flags = (flags & ifm->ifi_change) |
 847                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 848
 849         return flags;
 850 }
 851
 852 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 853                                  const struct rtnl_link_stats64 *b)
 854 {
 855         a->rx_packets = b->rx_packets;
 856         a->tx_packets = b->tx_packets;
 857         a->rx_bytes = b->rx_bytes;
 858         a->tx_bytes = b->tx_bytes;
 859         a->rx_errors = b->rx_errors;
 860         a->tx_errors = b->tx_errors;
 861         a->rx_dropped = b->rx_dropped;
 862         a->tx_dropped = b->tx_dropped;
 863
 864         a->multicast = b->multicast;
 865         a->collisions = b->collisions;
 866
 867         a->rx_length_errors = b->rx_length_errors;
 868         a->rx_over_errors = b->rx_over_errors;
 869         a->rx_crc_errors = b->rx_crc_errors;
 870         a->rx_frame_errors = b->rx_frame_errors;
 871         a->rx_fifo_errors = b->rx_fifo_errors;
 872         a->rx_missed_errors = b->rx_missed_errors;
 873
 874         a->tx_aborted_errors = b->tx_aborted_errors;
 875         a->tx_carrier_errors = b->tx_carrier_errors;
 876         a->tx_fifo_errors = b->tx_fifo_errors;
 877         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 878         a->tx_window_errors = b->tx_window_errors;
 879
 880         a->rx_compressed = b->rx_compressed;
 881         a->tx_compressed = b->tx_compressed;
 882
 883         a->rx_nohandler = b->rx_nohandler;
 884 }
 885
 886 /* All VF info */
 887 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 888                                    u32 ext_filter_mask)
 889 {
 890         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 891                 int num_vfs = dev_num_vf(dev->dev.parent);
 892                 size_t size = nla_total_size(0);
 893                 size += num_vfs *
 894                         (nla_total_size(0) +
 895                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 896                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 897                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 898                          nla_total_size(MAX_VLAN_LIST_LEN *
 899                                         sizeof(struct ifla_vf_vlan_info)) +
 900                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 901                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 902                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 903                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 904                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 905                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 906                          /* IFLA_VF_STATS_RX_PACKETS */
 907                          nla_total_size_64bit(sizeof(__u64)) +
 908                          /* IFLA_VF_STATS_TX_PACKETS */
 909                          nla_total_size_64bit(sizeof(__u64)) +
 910                          /* IFLA_VF_STATS_RX_BYTES */
 911                          nla_total_size_64bit(sizeof(__u64)) +
 912                          /* IFLA_VF_STATS_TX_BYTES */
 913                          nla_total_size_64bit(sizeof(__u64)) +
 914                          /* IFLA_VF_STATS_BROADCAST */
 915                          nla_total_size_64bit(sizeof(__u64)) +
 916                          /* IFLA_VF_STATS_MULTICAST */
 917                          nla_total_size_64bit(sizeof(__u64)) +
 918                          /* IFLA_VF_STATS_RX_DROPPED */
 919                          nla_total_size_64bit(sizeof(__u64)) +
 920                          /* IFLA_VF_STATS_TX_DROPPED */
 921                          nla_total_size_64bit(sizeof(__u64)) +
 922                          nla_total_size(sizeof(struct ifla_vf_trust)));
 923                 return size;
 924         } else
 925                 return 0;
 926 }
 927
 928 static size_t rtnl_port_size(const struct net_device *dev,
 929                              u32 ext_filter_mask)
 930 {
 931         size_t port_size = nla_total_size(4)            /* PORT_VF */
 932                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 933                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 934                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 935                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 936                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 937         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 938         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 939                 + port_size;
 940         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 941                 + port_size;
 942
 943         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 944             !(ext_filter_mask & RTEXT_FILTER_VF))
 945                 return 0;
 946         if (dev_num_vf(dev->dev.parent))
 947                 return port_self_size + vf_ports_size +
 948                         vf_port_size * dev_num_vf(dev->dev.parent);
 949         else
 950                 return port_self_size;
 951 }
 952
 953 static size_t rtnl_xdp_size(void)
 954 {
 955         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 956                           nla_total_size(1) +   /* XDP_ATTACHED */
 957                           nla_total_size(4);    /* XDP_PROG_ID */
 958
 959         return xdp_size;
 960 }
 961
 962 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 963                                      u32 ext_filter_mask)
 964 {
 965         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 966                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 967                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 968                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 969                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 970                + nla_total_size(sizeof(struct rtnl_link_stats))
 971                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 972                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 973                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 974                + nla_total_size(4) /* IFLA_TXQLEN */
 975                + nla_total_size(4) /* IFLA_WEIGHT */
 976                + nla_total_size(4) /* IFLA_MTU */
 977                + nla_total_size(4) /* IFLA_LINK */
 978                + nla_total_size(4) /* IFLA_MASTER */
 979                + nla_total_size(1) /* IFLA_CARRIER */
 980                + nla_total_size(4) /* IFLA_PROMISCUITY */
 981                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 982                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 983                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 984                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 985                + nla_total_size(1) /* IFLA_OPERSTATE */
 986                + nla_total_size(1) /* IFLA_LINKMODE */
 987                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 988                + nla_total_size(4) /* IFLA_LINK_NETNSID */
 989                + nla_total_size(4) /* IFLA_GROUP */
 990                + nla_total_size(ext_filter_mask
 991                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
 992                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
 993                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
 994                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
 995                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
 996                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
 997                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
 998                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
 999                + rtnl_xdp_size() /* IFLA_XDP */
1000                + nla_total_size(4)  /* IFLA_EVENT */
1001                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1002                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1003                + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1004                + nla_total_size(4)  /* IFLA_IF_NETNSID */
1005                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1006                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1007                + 0;
1008 }
1009
1010 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1011 {
1012         struct nlattr *vf_ports;
1013         struct nlattr *vf_port;
1014         int vf;
1015         int err;
1016
1017         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
1018         if (!vf_ports)
1019                 return -EMSGSIZE;
1020
1021         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1022                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
1023                 if (!vf_port)
1024                         goto nla_put_failure;
1025                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1026                         goto nla_put_failure;
1027                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1028                 if (err == -EMSGSIZE)
1029                         goto nla_put_failure;
1030                 if (err) {
1031                         nla_nest_cancel(skb, vf_port);
1032                         continue;
1033                 }
1034                 nla_nest_end(skb, vf_port);
1035         }
1036
1037         nla_nest_end(skb, vf_ports);
1038
1039         return 0;
1040
1041 nla_put_failure:
1042         nla_nest_cancel(skb, vf_ports);
1043         return -EMSGSIZE;
1044 }
1045
1046 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1047 {
1048         struct nlattr *port_self;
1049         int err;
1050
1051         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1052         if (!port_self)
1053                 return -EMSGSIZE;
1054
1055         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1056         if (err) {
1057                 nla_nest_cancel(skb, port_self);
1058                 return (err == -EMSGSIZE) ? err : 0;
1059         }
1060
1061         nla_nest_end(skb, port_self);
1062
1063         return 0;
1064 }
1065
1066 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1067                           u32 ext_filter_mask)
1068 {
1069         int err;
1070
1071         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1072             !(ext_filter_mask & RTEXT_FILTER_VF))
1073                 return 0;
1074
1075         err = rtnl_port_self_fill(skb, dev);
1076         if (err)
1077                 return err;
1078
1079         if (dev_num_vf(dev->dev.parent)) {
1080                 err = rtnl_vf_ports_fill(skb, dev);
1081                 if (err)
1082                         return err;
1083         }
1084
1085         return 0;
1086 }
1087
1088 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1089 {
1090         int err;
1091         struct netdev_phys_item_id ppid;
1092
1093         err = dev_get_phys_port_id(dev, &ppid);
1094         if (err) {
1095                 if (err == -EOPNOTSUPP)
1096                         return 0;
1097                 return err;
1098         }
1099
1100         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1101                 return -EMSGSIZE;
1102
1103         return 0;
1104 }
1105
1106 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1107 {
1108         char name[IFNAMSIZ];
1109         int err;
1110
1111         err = dev_get_phys_port_name(dev, name, sizeof(name));
1112         if (err) {
1113                 if (err == -EOPNOTSUPP)
1114                         return 0;
1115                 return err;
1116         }
1117
1118         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1119                 return -EMSGSIZE;
1120
1121         return 0;
1122 }
1123
1124 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1125 {
1126         int err;
1127         struct switchdev_attr attr = {
1128                 .orig_dev = dev,
1129                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1130                 .flags = SWITCHDEV_F_NO_RECURSE,
1131         };
1132
1133         err = switchdev_port_attr_get(dev, &attr);
1134         if (err) {
1135                 if (err == -EOPNOTSUPP)
1136                         return 0;
1137                 return err;
1138         }
1139
1140         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1141                     attr.u.ppid.id))
1142                 return -EMSGSIZE;
1143
1144         return 0;
1145 }
1146
1147 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1148                                               struct net_device *dev)
1149 {
1150         struct rtnl_link_stats64 *sp;
1151         struct nlattr *attr;
1152
1153         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1154                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1155         if (!attr)
1156                 return -EMSGSIZE;
1157
1158         sp = nla_data(attr);
1159         dev_get_stats(dev, sp);
1160
1161         attr = nla_reserve(skb, IFLA_STATS,
1162                            sizeof(struct rtnl_link_stats));
1163         if (!attr)
1164                 return -EMSGSIZE;
1165
1166         copy_rtnl_link_stats(nla_data(attr), sp);
1167
1168         return 0;
1169 }
1170
1171 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1172                                                struct net_device *dev,
1173                                                int vfs_num,
1174                                                struct nlattr *vfinfo)
1175 {
1176         struct ifla_vf_rss_query_en vf_rss_query_en;
1177         struct nlattr *vf, *vfstats, *vfvlanlist;
1178         struct ifla_vf_link_state vf_linkstate;
1179         struct ifla_vf_vlan_info vf_vlan_info;
1180         struct ifla_vf_spoofchk vf_spoofchk;
1181         struct ifla_vf_tx_rate vf_tx_rate;
1182         struct ifla_vf_stats vf_stats;
1183         struct ifla_vf_trust vf_trust;
1184         struct ifla_vf_vlan vf_vlan;
1185         struct ifla_vf_rate vf_rate;
1186         struct ifla_vf_mac vf_mac;
1187         struct ifla_vf_info ivi;
1188
1189         memset(&ivi, 0, sizeof(ivi));
1190
1191         /* Not all SR-IOV capable drivers support the
1192          * spoofcheck and "RSS query enable" query.  Preset to
1193          * -1 so the user space tool can detect that the driver
1194          * didn't report anything.
1195          */
1196         ivi.spoofchk = -1;
1197         ivi.rss_query_en = -1;
1198         ivi.trusted = -1;
1199         /* The default value for VF link state is "auto"
1200          * IFLA_VF_LINK_STATE_AUTO which equals zero
1201          */
1202         ivi.linkstate = 0;
1203         /* VLAN Protocol by default is 802.1Q */
1204         ivi.vlan_proto = htons(ETH_P_8021Q);
1205         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1206                 return 0;
1207
1208         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1209
1210         vf_mac.vf =
1211                 vf_vlan.vf =
1212                 vf_vlan_info.vf =
1213                 vf_rate.vf =
1214                 vf_tx_rate.vf =
1215                 vf_spoofchk.vf =
1216                 vf_linkstate.vf =
1217                 vf_rss_query_en.vf =
1218                 vf_trust.vf = ivi.vf;
1219
1220         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1221         vf_vlan.vlan = ivi.vlan;
1222         vf_vlan.qos = ivi.qos;
1223         vf_vlan_info.vlan = ivi.vlan;
1224         vf_vlan_info.qos = ivi.qos;
1225         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1226         vf_tx_rate.rate = ivi.max_tx_rate;
1227         vf_rate.min_tx_rate = ivi.min_tx_rate;
1228         vf_rate.max_tx_rate = ivi.max_tx_rate;
1229         vf_spoofchk.setting = ivi.spoofchk;
1230         vf_linkstate.link_state = ivi.linkstate;
1231         vf_rss_query_en.setting = ivi.rss_query_en;
1232         vf_trust.setting = ivi.trusted;
1233         vf = nla_nest_start(skb, IFLA_VF_INFO);
1234         if (!vf)
1235                 goto nla_put_vfinfo_failure;
1236         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1237             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1238             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1239                     &vf_rate) ||
1240             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1241                     &vf_tx_rate) ||
1242             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1243                     &vf_spoofchk) ||
1244             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1245                     &vf_linkstate) ||
1246             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1247                     sizeof(vf_rss_query_en),
1248                     &vf_rss_query_en) ||
1249             nla_put(skb, IFLA_VF_TRUST,
1250                     sizeof(vf_trust), &vf_trust))
1251                 goto nla_put_vf_failure;
1252         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1253         if (!vfvlanlist)
1254                 goto nla_put_vf_failure;
1255         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1256                     &vf_vlan_info)) {
1257                 nla_nest_cancel(skb, vfvlanlist);
1258                 goto nla_put_vf_failure;
1259         }
1260         nla_nest_end(skb, vfvlanlist);
1261         memset(&vf_stats, 0, sizeof(vf_stats));
1262         if (dev->netdev_ops->ndo_get_vf_stats)
1263                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1264                                                 &vf_stats);
1265         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1266         if (!vfstats)
1267                 goto nla_put_vf_failure;
1268         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1269                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1270             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1271                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1272             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1273                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1274             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1275                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1276             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1277                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1278             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1279                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1280             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1281                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1282             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1283                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1284                 nla_nest_cancel(skb, vfstats);
1285                 goto nla_put_vf_failure;
1286         }
1287         nla_nest_end(skb, vfstats);
1288         nla_nest_end(skb, vf);
1289         return 0;
1290
1291 nla_put_vf_failure:
1292         nla_nest_cancel(skb, vf);
1293 nla_put_vfinfo_failure:
1294         nla_nest_cancel(skb, vfinfo);
1295         return -EMSGSIZE;
1296 }
1297
1298 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1299                                            struct net_device *dev,
1300                                            u32 ext_filter_mask)
1301 {
1302         struct nlattr *vfinfo;
1303         int i, num_vfs;
1304
1305         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1306                 return 0;
1307
1308         num_vfs = dev_num_vf(dev->dev.parent);
1309         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1310                 return -EMSGSIZE;
1311
1312         if (!dev->netdev_ops->ndo_get_vf_config)
1313                 return 0;
1314
1315         vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1316         if (!vfinfo)
1317                 return -EMSGSIZE;
1318
1319         for (i = 0; i < num_vfs; i++) {
1320                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1321                         return -EMSGSIZE;
1322         }
1323
1324         nla_nest_end(skb, vfinfo);
1325         return 0;
1326 }
1327
1328 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1329 {
1330         struct rtnl_link_ifmap map;
1331
1332         memset(&map, 0, sizeof(map));
1333         map.mem_start   = dev->mem_start;
1334         map.mem_end     = dev->mem_end;
1335         map.base_addr   = dev->base_addr;
1336         map.irq         = dev->irq;
1337         map.dma         = dev->dma;
1338         map.port        = dev->if_port;
1339
1340         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1341                 return -EMSGSIZE;
1342
1343         return 0;
1344 }
1345
1346 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1347 {
1348         const struct net_device_ops *ops = dev->netdev_ops;
1349         const struct bpf_prog *generic_xdp_prog;
1350         struct netdev_bpf xdp;
1351
1352         ASSERT_RTNL();
1353
1354         *prog_id = 0;
1355         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1356         if (generic_xdp_prog) {
1357                 *prog_id = generic_xdp_prog->aux->id;
1358                 return XDP_ATTACHED_SKB;
1359         }
1360         if (!ops->ndo_bpf)
1361                 return XDP_ATTACHED_NONE;
1362
1363         __dev_xdp_query(dev, ops->ndo_bpf, &xdp);
1364         *prog_id = xdp.prog_id;
1365
1366         return xdp.prog_attached;
1367 }
1368
1369 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1370 {
1371         struct nlattr *xdp;
1372         u32 prog_id;
1373         int err;
1374
1375         xdp = nla_nest_start(skb, IFLA_XDP);
1376         if (!xdp)
1377                 return -EMSGSIZE;
1378
1379         err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1380                          rtnl_xdp_attached_mode(dev, &prog_id));
1381         if (err)
1382                 goto err_cancel;
1383
1384         if (prog_id) {
1385                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1386                 if (err)
1387                         goto err_cancel;
1388         }
1389
1390         nla_nest_end(skb, xdp);
1391         return 0;
1392
1393 err_cancel:
1394         nla_nest_cancel(skb, xdp);
1395         return err;
1396 }
1397
1398 static u32 rtnl_get_event(unsigned long event)
1399 {
1400         u32 rtnl_event_type = IFLA_EVENT_NONE;
1401
1402         switch (event) {
1403         case NETDEV_REBOOT:
1404                 rtnl_event_type = IFLA_EVENT_REBOOT;
1405                 break;
1406         case NETDEV_FEAT_CHANGE:
1407                 rtnl_event_type = IFLA_EVENT_FEATURES;
1408                 break;
1409         case NETDEV_BONDING_FAILOVER:
1410                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1411                 break;
1412         case NETDEV_NOTIFY_PEERS:
1413                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1414                 break;
1415         case NETDEV_RESEND_IGMP:
1416                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1417                 break;
1418         case NETDEV_CHANGEINFODATA:
1419                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1420                 break;
1421         default:
1422                 break;
1423         }
1424
1425         return rtnl_event_type;
1426 }
1427
1428 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1429 {
1430         const struct net_device *upper_dev;
1431         int ret = 0;
1432
1433         rcu_read_lock();
1434
1435         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1436         if (upper_dev)
1437                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1438
1439         rcu_read_unlock();
1440         return ret;
1441 }
1442
1443 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1444 {
1445         int ifindex = dev_get_iflink(dev);
1446
1447         if (dev->ifindex == ifindex)
1448                 return 0;
1449
1450         return nla_put_u32(skb, IFLA_LINK, ifindex);
1451 }
1452
1453 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1454                                               struct net_device *dev)
1455 {
1456         char buf[IFALIASZ];
1457         int ret;
1458
1459         ret = dev_get_alias(dev, buf, sizeof(buf));
1460         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1461 }
1462
1463 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1464                                   const struct net_device *dev,
1465                                   struct net *src_net)
1466 {
1467         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1468                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1469
1470                 if (!net_eq(dev_net(dev), link_net)) {
1471                         int id = peernet2id_alloc(src_net, link_net);
1472
1473                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1474                                 return -EMSGSIZE;
1475                 }
1476         }
1477
1478         return 0;
1479 }
1480
1481 static int rtnl_fill_link_af(struct sk_buff *skb,
1482                              const struct net_device *dev,
1483                              u32 ext_filter_mask)
1484 {
1485         const struct rtnl_af_ops *af_ops;
1486         struct nlattr *af_spec;
1487
1488         af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1489         if (!af_spec)
1490                 return -EMSGSIZE;
1491
1492         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1493                 struct nlattr *af;
1494                 int err;
1495
1496                 if (!af_ops->fill_link_af)
1497                         continue;
1498
1499                 af = nla_nest_start(skb, af_ops->family);
1500                 if (!af)
1501                         return -EMSGSIZE;
1502
1503                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1504                 /*
1505                  * Caller may return ENODATA to indicate that there
1506                  * was no data to be dumped. This is not an error, it
1507                  * means we should trim the attribute header and
1508                  * continue.
1509                  */
1510                 if (err == -ENODATA)
1511                         nla_nest_cancel(skb, af);
1512                 else if (err < 0)
1513                         return -EMSGSIZE;
1514
1515                 nla_nest_end(skb, af);
1516         }
1517
1518         nla_nest_end(skb, af_spec);
1519         return 0;
1520 }
1521
1522 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1523                             struct net_device *dev, struct net *src_net,
1524                             int type, u32 pid, u32 seq, u32 change,
1525                             unsigned int flags, u32 ext_filter_mask,
1526                             u32 event, int *new_nsid, int new_ifindex,
1527                             int tgt_netnsid)
1528 {
1529         struct ifinfomsg *ifm;
1530         struct nlmsghdr *nlh;
1531
1532         ASSERT_RTNL();
1533         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1534         if (nlh == NULL)
1535                 return -EMSGSIZE;
1536
1537         ifm = nlmsg_data(nlh);
1538         ifm->ifi_family = AF_UNSPEC;
1539         ifm->__ifi_pad = 0;
1540         ifm->ifi_type = dev->type;
1541         ifm->ifi_index = dev->ifindex;
1542         ifm->ifi_flags = dev_get_flags(dev);
1543         ifm->ifi_change = change;
1544
1545         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1546                 goto nla_put_failure;
1547
1548         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1549             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1550             nla_put_u8(skb, IFLA_OPERSTATE,
1551                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1552             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1553             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1554             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1555             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1556             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1557             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1558             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1559 #ifdef CONFIG_RPS
1560             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1561 #endif
1562             nla_put_iflink(skb, dev) ||
1563             put_master_ifindex(skb, dev) ||
1564             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1565             (dev->qdisc &&
1566              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1567             nla_put_ifalias(skb, dev) ||
1568             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1569                         atomic_read(&dev->carrier_up_count) +
1570                         atomic_read(&dev->carrier_down_count)) ||
1571             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1572             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1573                         atomic_read(&dev->carrier_up_count)) ||
1574             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1575                         atomic_read(&dev->carrier_down_count)))
1576                 goto nla_put_failure;
1577
1578         if (event != IFLA_EVENT_NONE) {
1579                 if (nla_put_u32(skb, IFLA_EVENT, event))
1580                         goto nla_put_failure;
1581         }
1582
1583         if (rtnl_fill_link_ifmap(skb, dev))
1584                 goto nla_put_failure;
1585
1586         if (dev->addr_len) {
1587                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1588                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1589                         goto nla_put_failure;
1590         }
1591
1592         if (rtnl_phys_port_id_fill(skb, dev))
1593                 goto nla_put_failure;
1594
1595         if (rtnl_phys_port_name_fill(skb, dev))
1596                 goto nla_put_failure;
1597
1598         if (rtnl_phys_switch_id_fill(skb, dev))
1599                 goto nla_put_failure;
1600
1601         if (rtnl_fill_stats(skb, dev))
1602                 goto nla_put_failure;
1603
1604         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1605                 goto nla_put_failure;
1606
1607         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1608                 goto nla_put_failure;
1609
1610         if (rtnl_xdp_fill(skb, dev))
1611                 goto nla_put_failure;
1612
1613         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1614                 if (rtnl_link_fill(skb, dev) < 0)
1615                         goto nla_put_failure;
1616         }
1617
1618         if (rtnl_fill_link_netnsid(skb, dev, src_net))
1619                 goto nla_put_failure;
1620
1621         if (new_nsid &&
1622             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1623                 goto nla_put_failure;
1624         if (new_ifindex &&
1625             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1626                 goto nla_put_failure;
1627
1628
1629         rcu_read_lock();
1630         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1631                 goto nla_put_failure_rcu;
1632         rcu_read_unlock();
1633
1634         nlmsg_end(skb, nlh);
1635         return 0;
1636
1637 nla_put_failure_rcu:
1638         rcu_read_unlock();
1639 nla_put_failure:
1640         nlmsg_cancel(skb, nlh);
1641         return -EMSGSIZE;
1642 }
1643
1644 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1645         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1646         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1647         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1648         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1649         [IFLA_MTU]              = { .type = NLA_U32 },
1650         [IFLA_LINK]             = { .type = NLA_U32 },
1651         [IFLA_MASTER]           = { .type = NLA_U32 },
1652         [IFLA_CARRIER]          = { .type = NLA_U8 },
1653         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1654         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1655         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1656         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1657         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1658         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1659         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1660         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1661          * allow 0-length string (needed to remove an alias).
1662          */
1663         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1664         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1665         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1666         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1667         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1668         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1669         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1670         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1671         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1672         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1673         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1674         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1675         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1676         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1677         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1678         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1679         [IFLA_XDP]              = { .type = NLA_NESTED },
1680         [IFLA_EVENT]            = { .type = NLA_U32 },
1681         [IFLA_GROUP]            = { .type = NLA_U32 },
1682         [IFLA_IF_NETNSID]       = { .type = NLA_S32 },
1683         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1684         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1685 };
1686
1687 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1688         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1689         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1690         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1691         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1692 };
1693
1694 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1695         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1696         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1697         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1698         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1699         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1700         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1701         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1702         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1703         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1704         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1705         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1706         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1707 };
1708
1709 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1710         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1711         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1712                                     .len = PORT_PROFILE_MAX },
1713         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1714                                       .len = PORT_UUID_MAX },
1715         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1716                                     .len = PORT_UUID_MAX },
1717         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1718         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1719
1720         /* Unused, but we need to keep it here since user space could
1721          * fill it. It's also broken with regard to NLA_BINARY use in
1722          * combination with structs.
1723          */
1724         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1725                                     .len = sizeof(struct ifla_port_vsi) },
1726 };
1727
1728 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1729         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1730         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1731         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1732         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1733 };
1734
1735 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1736 {
1737         const struct rtnl_link_ops *ops = NULL;
1738         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1739
1740         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1741                              ifla_info_policy, NULL) < 0)
1742                 return NULL;
1743
1744         if (linfo[IFLA_INFO_KIND]) {
1745                 char kind[MODULE_NAME_LEN];
1746
1747                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1748                 ops = rtnl_link_ops_get(kind);
1749         }
1750
1751         return ops;
1752 }
1753
1754 static bool link_master_filtered(struct net_device *dev, int master_idx)
1755 {
1756         struct net_device *master;
1757
1758         if (!master_idx)
1759                 return false;
1760
1761         master = netdev_master_upper_dev_get(dev);
1762         if (!master || master->ifindex != master_idx)
1763                 return true;
1764
1765         return false;
1766 }
1767
1768 static bool link_kind_filtered(const struct net_device *dev,
1769                                const struct rtnl_link_ops *kind_ops)
1770 {
1771         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1772                 return true;
1773
1774         return false;
1775 }
1776
1777 static bool link_dump_filtered(struct net_device *dev,
1778                                int master_idx,
1779                                const struct rtnl_link_ops *kind_ops)
1780 {
1781         if (link_master_filtered(dev, master_idx) ||
1782             link_kind_filtered(dev, kind_ops))
1783                 return true;
1784
1785         return false;
1786 }
1787
1788 static struct net *get_target_net(struct sock *sk, int netnsid)
1789 {
1790         struct net *net;
1791
1792         net = get_net_ns_by_id(sock_net(sk), netnsid);
1793         if (!net)
1794                 return ERR_PTR(-EINVAL);
1795
1796         /* For now, the caller is required to have CAP_NET_ADMIN in
1797          * the user namespace owning the target net ns.
1798          */
1799         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1800                 put_net(net);
1801                 return ERR_PTR(-EACCES);
1802         }
1803         return net;
1804 }
1805
1806 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1807 {
1808         struct net *net = sock_net(skb->sk);
1809         struct net *tgt_net = net;
1810         int h, s_h;
1811         int idx = 0, s_idx;
1812         struct net_device *dev;
1813         struct hlist_head *head;
1814         struct nlattr *tb[IFLA_MAX+1];
1815         u32 ext_filter_mask = 0;
1816         const struct rtnl_link_ops *kind_ops = NULL;
1817         unsigned int flags = NLM_F_MULTI;
1818         int master_idx = 0;
1819         int netnsid = -1;
1820         int err;
1821         int hdrlen;
1822
1823         s_h = cb->args[0];
1824         s_idx = cb->args[1];
1825
1826         /* A hack to preserve kernel<->userspace interface.
1827          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1828          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1829          * what iproute2 < v3.9.0 used.
1830          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1831          * attribute, its netlink message is shorter than struct ifinfomsg.
1832          */
1833         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1834                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1835
1836         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1837                         ifla_policy, NULL) >= 0) {
1838                 if (tb[IFLA_IF_NETNSID]) {
1839                         netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1840                         tgt_net = get_target_net(skb->sk, netnsid);
1841                         if (IS_ERR(tgt_net)) {
1842                                 tgt_net = net;
1843                                 netnsid = -1;
1844                         }
1845                 }
1846
1847                 if (tb[IFLA_EXT_MASK])
1848                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1849
1850                 if (tb[IFLA_MASTER])
1851                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1852
1853                 if (tb[IFLA_LINKINFO])
1854                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1855
1856                 if (master_idx || kind_ops)
1857                         flags |= NLM_F_DUMP_FILTERED;
1858         }
1859
1860         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1861                 idx = 0;
1862                 head = &tgt_net->dev_index_head[h];
1863                 hlist_for_each_entry(dev, head, index_hlist) {
1864                         if (link_dump_filtered(dev, master_idx, kind_ops))
1865                                 goto cont;
1866                         if (idx < s_idx)
1867                                 goto cont;
1868                         err = rtnl_fill_ifinfo(skb, dev, net,
1869                                                RTM_NEWLINK,
1870                                                NETLINK_CB(cb->skb).portid,
1871                                                cb->nlh->nlmsg_seq, 0,
1872                                                flags,
1873                                                ext_filter_mask, 0, NULL, 0,
1874                                                netnsid);
1875
1876                         if (err < 0) {
1877                                 if (likely(skb->len))
1878                                         goto out;
1879
1880                                 goto out_err;
1881                         }
1882 cont:
1883                         idx++;
1884                 }
1885         }
1886 out:
1887         err = skb->len;
1888 out_err:
1889         cb->args[1] = idx;
1890         cb->args[0] = h;
1891         cb->seq = net->dev_base_seq;
1892         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1893         if (netnsid >= 0)
1894                 put_net(tgt_net);
1895
1896         return err;
1897 }
1898
1899 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1900                         struct netlink_ext_ack *exterr)
1901 {
1902         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1903 }
1904 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1905
1906 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1907 {
1908         struct net *net;
1909         /* Examine the link attributes and figure out which
1910          * network namespace we are talking about.
1911          */
1912         if (tb[IFLA_NET_NS_PID])
1913                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1914         else if (tb[IFLA_NET_NS_FD])
1915                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1916         else
1917                 net = get_net(src_net);
1918         return net;
1919 }
1920 EXPORT_SYMBOL(rtnl_link_get_net);
1921
1922 /* Figure out which network namespace we are talking about by
1923  * examining the link attributes in the following order:
1924  *
1925  * 1. IFLA_NET_NS_PID
1926  * 2. IFLA_NET_NS_FD
1927  * 3. IFLA_IF_NETNSID
1928  */
1929 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1930                                                struct nlattr *tb[])
1931 {
1932         struct net *net;
1933
1934         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1935                 return rtnl_link_get_net(src_net, tb);
1936
1937         if (!tb[IFLA_IF_NETNSID])
1938                 return get_net(src_net);
1939
1940         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1941         if (!net)
1942                 return ERR_PTR(-EINVAL);
1943
1944         return net;
1945 }
1946
1947 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
1948                                              struct net *src_net,
1949                                              struct nlattr *tb[], int cap)
1950 {
1951         struct net *net;
1952
1953         net = rtnl_link_get_net_by_nlattr(src_net, tb);
1954         if (IS_ERR(net))
1955                 return net;
1956
1957         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
1958                 put_net(net);
1959                 return ERR_PTR(-EPERM);
1960         }
1961
1962         return net;
1963 }
1964
1965 /* Verify that rtnetlink requests do not pass additional properties
1966  * potentially referring to different network namespaces.
1967  */
1968 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
1969                                     struct netlink_ext_ack *extack,
1970                                     bool netns_id_only)
1971 {
1972
1973         if (netns_id_only) {
1974                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
1975                         return 0;
1976
1977                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
1978                 return -EOPNOTSUPP;
1979         }
1980
1981         if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
1982                 goto invalid_attr;
1983
1984         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
1985                 goto invalid_attr;
1986
1987         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
1988                 goto invalid_attr;
1989
1990         return 0;
1991
1992 invalid_attr:
1993         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
1994         return -EINVAL;
1995 }
1996
1997 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1998 {
1999         if (dev) {
2000                 if (tb[IFLA_ADDRESS] &&
2001                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2002                         return -EINVAL;
2003
2004                 if (tb[IFLA_BROADCAST] &&
2005                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2006                         return -EINVAL;
2007         }
2008
2009         if (tb[IFLA_AF_SPEC]) {
2010                 struct nlattr *af;
2011                 int rem, err;
2012
2013                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2014                         const struct rtnl_af_ops *af_ops;
2015
2016                         rcu_read_lock();
2017                         af_ops = rtnl_af_lookup(nla_type(af));
2018                         if (!af_ops) {
2019                                 rcu_read_unlock();
2020                                 return -EAFNOSUPPORT;
2021                         }
2022
2023                         if (!af_ops->set_link_af) {
2024                                 rcu_read_unlock();
2025                                 return -EOPNOTSUPP;
2026                         }
2027
2028                         if (af_ops->validate_link_af) {
2029                                 err = af_ops->validate_link_af(dev, af);
2030                                 if (err < 0) {
2031                                         rcu_read_unlock();
2032                                         return err;
2033                                 }
2034                         }
2035
2036                         rcu_read_unlock();
2037                 }
2038         }
2039
2040         return 0;
2041 }
2042
2043 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2044                                   int guid_type)
2045 {
2046         const struct net_device_ops *ops = dev->netdev_ops;
2047
2048         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2049 }
2050
2051 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2052 {
2053         if (dev->type != ARPHRD_INFINIBAND)
2054                 return -EOPNOTSUPP;
2055
2056         return handle_infiniband_guid(dev, ivt, guid_type);
2057 }
2058
2059 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2060 {
2061         const struct net_device_ops *ops = dev->netdev_ops;
2062         int err = -EINVAL;
2063
2064         if (tb[IFLA_VF_MAC]) {
2065                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2066
2067                 err = -EOPNOTSUPP;
2068                 if (ops->ndo_set_vf_mac)
2069                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2070                                                   ivm->mac);
2071                 if (err < 0)
2072                         return err;
2073         }
2074
2075         if (tb[IFLA_VF_VLAN]) {
2076                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2077
2078                 err = -EOPNOTSUPP;
2079                 if (ops->ndo_set_vf_vlan)
2080                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2081                                                    ivv->qos,
2082                                                    htons(ETH_P_8021Q));
2083                 if (err < 0)
2084                         return err;
2085         }
2086
2087         if (tb[IFLA_VF_VLAN_LIST]) {
2088                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2089                 struct nlattr *attr;
2090                 int rem, len = 0;
2091
2092                 err = -EOPNOTSUPP;
2093                 if (!ops->ndo_set_vf_vlan)
2094                         return err;
2095
2096                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2097                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2098                             nla_len(attr) < NLA_HDRLEN) {
2099                                 return -EINVAL;
2100                         }
2101                         if (len >= MAX_VLAN_LIST_LEN)
2102                                 return -EOPNOTSUPP;
2103                         ivvl[len] = nla_data(attr);
2104
2105                         len++;
2106                 }
2107                 if (len == 0)
2108                         return -EINVAL;
2109
2110                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2111                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2112                 if (err < 0)
2113                         return err;
2114         }
2115
2116         if (tb[IFLA_VF_TX_RATE]) {
2117                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2118                 struct ifla_vf_info ivf;
2119
2120                 err = -EOPNOTSUPP;
2121                 if (ops->ndo_get_vf_config)
2122                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2123                 if (err < 0)
2124                         return err;
2125
2126                 err = -EOPNOTSUPP;
2127                 if (ops->ndo_set_vf_rate)
2128                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2129                                                    ivf.min_tx_rate,
2130                                                    ivt->rate);
2131                 if (err < 0)
2132                         return err;
2133         }
2134
2135         if (tb[IFLA_VF_RATE]) {
2136                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2137
2138                 err = -EOPNOTSUPP;
2139                 if (ops->ndo_set_vf_rate)
2140                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2141                                                    ivt->min_tx_rate,
2142                                                    ivt->max_tx_rate);
2143                 if (err < 0)
2144                         return err;
2145         }
2146
2147         if (tb[IFLA_VF_SPOOFCHK]) {
2148                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2149
2150                 err = -EOPNOTSUPP;
2151                 if (ops->ndo_set_vf_spoofchk)
2152                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2153                                                        ivs->setting);
2154                 if (err < 0)
2155                         return err;
2156         }
2157
2158         if (tb[IFLA_VF_LINK_STATE]) {
2159                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2160
2161                 err = -EOPNOTSUPP;
2162                 if (ops->ndo_set_vf_link_state)
2163                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2164                                                          ivl->link_state);
2165                 if (err < 0)
2166                         return err;
2167         }
2168
2169         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2170                 struct ifla_vf_rss_query_en *ivrssq_en;
2171
2172                 err = -EOPNOTSUPP;
2173                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2174                 if (ops->ndo_set_vf_rss_query_en)
2175                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2176                                                            ivrssq_en->setting);
2177                 if (err < 0)
2178                         return err;
2179         }
2180
2181         if (tb[IFLA_VF_TRUST]) {
2182                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2183
2184                 err = -EOPNOTSUPP;
2185                 if (ops->ndo_set_vf_trust)
2186                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2187                 if (err < 0)
2188                         return err;
2189         }
2190
2191         if (tb[IFLA_VF_IB_NODE_GUID]) {
2192                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2193
2194                 if (!ops->ndo_set_vf_guid)
2195                         return -EOPNOTSUPP;
2196
2197                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2198         }
2199
2200         if (tb[IFLA_VF_IB_PORT_GUID]) {
2201                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2202
2203                 if (!ops->ndo_set_vf_guid)
2204                         return -EOPNOTSUPP;
2205
2206                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2207         }
2208
2209         return err;
2210 }
2211
2212 static int do_set_master(struct net_device *dev, int ifindex,
2213                          struct netlink_ext_ack *extack)
2214 {
2215         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2216         const struct net_device_ops *ops;
2217         int err;
2218
2219         if (upper_dev) {
2220                 if (upper_dev->ifindex == ifindex)
2221                         return 0;
2222                 ops = upper_dev->netdev_ops;
2223                 if (ops->ndo_del_slave) {
2224                         err = ops->ndo_del_slave(upper_dev, dev);
2225                         if (err)
2226                                 return err;
2227                 } else {
2228                         return -EOPNOTSUPP;
2229                 }
2230         }
2231
2232         if (ifindex) {
2233                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2234                 if (!upper_dev)
2235                         return -EINVAL;
2236                 ops = upper_dev->netdev_ops;
2237                 if (ops->ndo_add_slave) {
2238                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2239                         if (err)
2240                                 return err;
2241                 } else {
2242                         return -EOPNOTSUPP;
2243                 }
2244         }
2245         return 0;
2246 }
2247
2248 #define DO_SETLINK_MODIFIED     0x01
2249 /* notify flag means notify + modified. */
2250 #define DO_SETLINK_NOTIFY       0x03
2251 static int do_setlink(const struct sk_buff *skb,
2252                       struct net_device *dev, struct ifinfomsg *ifm,
2253                       struct netlink_ext_ack *extack,
2254                       struct nlattr **tb, char *ifname, int status)
2255 {
2256         const struct net_device_ops *ops = dev->netdev_ops;
2257         int err;
2258
2259         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2260                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2261                                                             tb, CAP_NET_ADMIN);
2262                 if (IS_ERR(net)) {
2263                         err = PTR_ERR(net);
2264                         goto errout;
2265                 }
2266
2267                 err = dev_change_net_namespace(dev, net, ifname);
2268                 put_net(net);
2269                 if (err)
2270                         goto errout;
2271                 status |= DO_SETLINK_MODIFIED;
2272         }
2273
2274         if (tb[IFLA_MAP]) {
2275                 struct rtnl_link_ifmap *u_map;
2276                 struct ifmap k_map;
2277
2278                 if (!ops->ndo_set_config) {
2279                         err = -EOPNOTSUPP;
2280                         goto errout;
2281                 }
2282
2283                 if (!netif_device_present(dev)) {
2284                         err = -ENODEV;
2285                         goto errout;
2286                 }
2287
2288                 u_map = nla_data(tb[IFLA_MAP]);
2289                 k_map.mem_start = (unsigned long) u_map->mem_start;
2290                 k_map.mem_end = (unsigned long) u_map->mem_end;
2291                 k_map.base_addr = (unsigned short) u_map->base_addr;
2292                 k_map.irq = (unsigned char) u_map->irq;
2293                 k_map.dma = (unsigned char) u_map->dma;
2294                 k_map.port = (unsigned char) u_map->port;
2295
2296                 err = ops->ndo_set_config(dev, &k_map);
2297                 if (err < 0)
2298                         goto errout;
2299
2300                 status |= DO_SETLINK_NOTIFY;
2301         }
2302
2303         if (tb[IFLA_ADDRESS]) {
2304                 struct sockaddr *sa;
2305                 int len;
2306
2307                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2308                                                   sizeof(*sa));
2309                 sa = kmalloc(len, GFP_KERNEL);
2310                 if (!sa) {
2311                         err = -ENOMEM;
2312                         goto errout;
2313                 }
2314                 sa->sa_family = dev->type;
2315                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2316                        dev->addr_len);
2317                 err = dev_set_mac_address(dev, sa);
2318                 kfree(sa);
2319                 if (err)
2320                         goto errout;
2321                 status |= DO_SETLINK_MODIFIED;
2322         }
2323
2324         if (tb[IFLA_MTU]) {
2325                 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2326                 if (err < 0)
2327                         goto errout;
2328                 status |= DO_SETLINK_MODIFIED;
2329         }
2330
2331         if (tb[IFLA_GROUP]) {
2332                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2333                 status |= DO_SETLINK_NOTIFY;
2334         }
2335
2336         /*
2337          * Interface selected by interface index but interface
2338          * name provided implies that a name change has been
2339          * requested.
2340          */
2341         if (ifm->ifi_index > 0 && ifname[0]) {
2342                 err = dev_change_name(dev, ifname);
2343                 if (err < 0)
2344                         goto errout;
2345                 status |= DO_SETLINK_MODIFIED;
2346         }
2347
2348         if (tb[IFLA_IFALIAS]) {
2349                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2350                                     nla_len(tb[IFLA_IFALIAS]));
2351                 if (err < 0)
2352                         goto errout;
2353                 status |= DO_SETLINK_NOTIFY;
2354         }
2355
2356         if (tb[IFLA_BROADCAST]) {
2357                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2358                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2359         }
2360
2361         if (ifm->ifi_flags || ifm->ifi_change) {
2362                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2363                 if (err < 0)
2364                         goto errout;
2365         }
2366
2367         if (tb[IFLA_MASTER]) {
2368                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2369                 if (err)
2370                         goto errout;
2371                 status |= DO_SETLINK_MODIFIED;
2372         }
2373
2374         if (tb[IFLA_CARRIER]) {
2375                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2376                 if (err)
2377                         goto errout;
2378                 status |= DO_SETLINK_MODIFIED;
2379         }
2380
2381         if (tb[IFLA_TXQLEN]) {
2382                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2383
2384                 err = dev_change_tx_queue_len(dev, value);
2385                 if (err)
2386                         goto errout;
2387                 status |= DO_SETLINK_MODIFIED;
2388         }
2389
2390         if (tb[IFLA_GSO_MAX_SIZE]) {
2391                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2392
2393                 if (max_size > GSO_MAX_SIZE) {
2394                         err = -EINVAL;
2395                         goto errout;
2396                 }
2397
2398                 if (dev->gso_max_size ^ max_size) {
2399                         netif_set_gso_max_size(dev, max_size);
2400                         status |= DO_SETLINK_MODIFIED;
2401                 }
2402         }
2403
2404         if (tb[IFLA_GSO_MAX_SEGS]) {
2405                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2406
2407                 if (max_segs > GSO_MAX_SEGS) {
2408                         err = -EINVAL;
2409                         goto errout;
2410                 }
2411
2412                 if (dev->gso_max_segs ^ max_segs) {
2413                         dev->gso_max_segs = max_segs;
2414                         status |= DO_SETLINK_MODIFIED;
2415                 }
2416         }
2417
2418         if (tb[IFLA_OPERSTATE])
2419                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2420
2421         if (tb[IFLA_LINKMODE]) {
2422                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2423
2424                 write_lock_bh(&dev_base_lock);
2425                 if (dev->link_mode ^ value)
2426                         status |= DO_SETLINK_NOTIFY;
2427                 dev->link_mode = value;
2428                 write_unlock_bh(&dev_base_lock);
2429         }
2430
2431         if (tb[IFLA_VFINFO_LIST]) {
2432                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2433                 struct nlattr *attr;
2434                 int rem;
2435
2436                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2437                         if (nla_type(attr) != IFLA_VF_INFO ||
2438                             nla_len(attr) < NLA_HDRLEN) {
2439                                 err = -EINVAL;
2440                                 goto errout;
2441                         }
2442                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2443                                                ifla_vf_policy, NULL);
2444                         if (err < 0)
2445                                 goto errout;
2446                         err = do_setvfinfo(dev, vfinfo);
2447                         if (err < 0)
2448                                 goto errout;
2449                         status |= DO_SETLINK_NOTIFY;
2450                 }
2451         }
2452         err = 0;
2453
2454         if (tb[IFLA_VF_PORTS]) {
2455                 struct nlattr *port[IFLA_PORT_MAX+1];
2456                 struct nlattr *attr;
2457                 int vf;
2458                 int rem;
2459
2460                 err = -EOPNOTSUPP;
2461                 if (!ops->ndo_set_vf_port)
2462                         goto errout;
2463
2464                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2465                         if (nla_type(attr) != IFLA_VF_PORT ||
2466                             nla_len(attr) < NLA_HDRLEN) {
2467                                 err = -EINVAL;
2468                                 goto errout;
2469                         }
2470                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2471                                                ifla_port_policy, NULL);
2472                         if (err < 0)
2473                                 goto errout;
2474                         if (!port[IFLA_PORT_VF]) {
2475                                 err = -EOPNOTSUPP;
2476                                 goto errout;
2477                         }
2478                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2479                         err = ops->ndo_set_vf_port(dev, vf, port);
2480                         if (err < 0)
2481                                 goto errout;
2482                         status |= DO_SETLINK_NOTIFY;
2483                 }
2484         }
2485         err = 0;
2486
2487         if (tb[IFLA_PORT_SELF]) {
2488                 struct nlattr *port[IFLA_PORT_MAX+1];
2489
2490                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2491                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2492                                        NULL);
2493                 if (err < 0)
2494                         goto errout;
2495
2496                 err = -EOPNOTSUPP;
2497                 if (ops->ndo_set_vf_port)
2498                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2499                 if (err < 0)
2500                         goto errout;
2501                 status |= DO_SETLINK_NOTIFY;
2502         }
2503
2504         if (tb[IFLA_AF_SPEC]) {
2505                 struct nlattr *af;
2506                 int rem;
2507
2508                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2509                         const struct rtnl_af_ops *af_ops;
2510
2511                         rcu_read_lock();
2512
2513                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2514
2515                         err = af_ops->set_link_af(dev, af);
2516                         if (err < 0) {
2517                                 rcu_read_unlock();
2518                                 goto errout;
2519                         }
2520
2521                         rcu_read_unlock();
2522                         status |= DO_SETLINK_NOTIFY;
2523                 }
2524         }
2525         err = 0;
2526
2527         if (tb[IFLA_PROTO_DOWN]) {
2528                 err = dev_change_proto_down(dev,
2529                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2530                 if (err)
2531                         goto errout;
2532                 status |= DO_SETLINK_NOTIFY;
2533         }
2534
2535         if (tb[IFLA_XDP]) {
2536                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2537                 u32 xdp_flags = 0;
2538
2539                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2540                                        ifla_xdp_policy, NULL);
2541                 if (err < 0)
2542                         goto errout;
2543
2544                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2545                         err = -EINVAL;
2546                         goto errout;
2547                 }
2548
2549                 if (xdp[IFLA_XDP_FLAGS]) {
2550                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2551                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2552                                 err = -EINVAL;
2553                                 goto errout;
2554                         }
2555                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2556                                 err = -EINVAL;
2557                                 goto errout;
2558                         }
2559                 }
2560
2561                 if (xdp[IFLA_XDP_FD]) {
2562                         err = dev_change_xdp_fd(dev, extack,
2563                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2564                                                 xdp_flags);
2565                         if (err)
2566                                 goto errout;
2567                         status |= DO_SETLINK_NOTIFY;
2568                 }
2569         }
2570
2571 errout:
2572         if (status & DO_SETLINK_MODIFIED) {
2573                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2574                         netdev_state_change(dev);
2575
2576                 if (err < 0)
2577                         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",
2578                                              dev->name);
2579         }
2580
2581         return err;
2582 }
2583
2584 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2585                         struct netlink_ext_ack *extack)
2586 {
2587         struct net *net = sock_net(skb->sk);
2588         struct ifinfomsg *ifm;
2589         struct net_device *dev;
2590         int err;
2591         struct nlattr *tb[IFLA_MAX+1];
2592         char ifname[IFNAMSIZ];
2593
2594         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2595                           extack);
2596         if (err < 0)
2597                 goto errout;
2598
2599         err = rtnl_ensure_unique_netns(tb, extack, false);
2600         if (err < 0)
2601                 goto errout;
2602
2603         if (tb[IFLA_IFNAME])
2604                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2605         else
2606                 ifname[0] = '\0';
2607
2608         err = -EINVAL;
2609         ifm = nlmsg_data(nlh);
2610         if (ifm->ifi_index > 0)
2611                 dev = __dev_get_by_index(net, ifm->ifi_index);
2612         else if (tb[IFLA_IFNAME])
2613                 dev = __dev_get_by_name(net, ifname);
2614         else
2615                 goto errout;
2616
2617         if (dev == NULL) {
2618                 err = -ENODEV;
2619                 goto errout;
2620         }
2621
2622         err = validate_linkmsg(dev, tb);
2623         if (err < 0)
2624                 goto errout;
2625
2626         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2627 errout:
2628         return err;
2629 }
2630
2631 static int rtnl_group_dellink(const struct net *net, int group)
2632 {
2633         struct net_device *dev, *aux;
2634         LIST_HEAD(list_kill);
2635         bool found = false;
2636
2637         if (!group)
2638                 return -EPERM;
2639
2640         for_each_netdev(net, dev) {
2641                 if (dev->group == group) {
2642                         const struct rtnl_link_ops *ops;
2643
2644                         found = true;
2645                         ops = dev->rtnl_link_ops;
2646                         if (!ops || !ops->dellink)
2647                                 return -EOPNOTSUPP;
2648                 }
2649         }
2650
2651         if (!found)
2652                 return -ENODEV;
2653
2654         for_each_netdev_safe(net, dev, aux) {
2655                 if (dev->group == group) {
2656                         const struct rtnl_link_ops *ops;
2657
2658                         ops = dev->rtnl_link_ops;
2659                         ops->dellink(dev, &list_kill);
2660                 }
2661         }
2662         unregister_netdevice_many(&list_kill);
2663
2664         return 0;
2665 }
2666
2667 int rtnl_delete_link(struct net_device *dev)
2668 {
2669         const struct rtnl_link_ops *ops;
2670         LIST_HEAD(list_kill);
2671
2672         ops = dev->rtnl_link_ops;
2673         if (!ops || !ops->dellink)
2674                 return -EOPNOTSUPP;
2675
2676         ops->dellink(dev, &list_kill);
2677         unregister_netdevice_many(&list_kill);
2678
2679         return 0;
2680 }
2681 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2682
2683 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2684                         struct netlink_ext_ack *extack)
2685 {
2686         struct net *net = sock_net(skb->sk);
2687         struct net *tgt_net = net;
2688         struct net_device *dev = NULL;
2689         struct ifinfomsg *ifm;
2690         char ifname[IFNAMSIZ];
2691         struct nlattr *tb[IFLA_MAX+1];
2692         int err;
2693         int netnsid = -1;
2694
2695         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2696         if (err < 0)
2697                 return err;
2698
2699         err = rtnl_ensure_unique_netns(tb, extack, true);
2700         if (err < 0)
2701                 return err;
2702
2703         if (tb[IFLA_IFNAME])
2704                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2705
2706         if (tb[IFLA_IF_NETNSID]) {
2707                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2708                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2709                 if (IS_ERR(tgt_net))
2710                         return PTR_ERR(tgt_net);
2711         }
2712
2713         err = -EINVAL;
2714         ifm = nlmsg_data(nlh);
2715         if (ifm->ifi_index > 0)
2716                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2717         else if (tb[IFLA_IFNAME])
2718                 dev = __dev_get_by_name(tgt_net, ifname);
2719         else if (tb[IFLA_GROUP])
2720                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2721         else
2722                 goto out;
2723
2724         if (!dev) {
2725                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2726                         err = -ENODEV;
2727
2728                 goto out;
2729         }
2730
2731         err = rtnl_delete_link(dev);
2732
2733 out:
2734         if (netnsid >= 0)
2735                 put_net(tgt_net);
2736
2737         return err;
2738 }
2739
2740 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2741 {
2742         unsigned int old_flags;
2743         int err;
2744
2745         old_flags = dev->flags;
2746         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2747                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2748                 if (err < 0)
2749                         return err;
2750         }
2751
2752         dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2753
2754         __dev_notify_flags(dev, old_flags, ~0U);
2755         return 0;
2756 }
2757 EXPORT_SYMBOL(rtnl_configure_link);
2758
2759 struct net_device *rtnl_create_link(struct net *net,
2760         const char *ifname, unsigned char name_assign_type,
2761         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2762 {
2763         struct net_device *dev;
2764         unsigned int num_tx_queues = 1;
2765         unsigned int num_rx_queues = 1;
2766
2767         if (tb[IFLA_NUM_TX_QUEUES])
2768                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2769         else if (ops->get_num_tx_queues)
2770                 num_tx_queues = ops->get_num_tx_queues();
2771
2772         if (tb[IFLA_NUM_RX_QUEUES])
2773                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2774         else if (ops->get_num_rx_queues)
2775                 num_rx_queues = ops->get_num_rx_queues();
2776
2777         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2778                                ops->setup, num_tx_queues, num_rx_queues);
2779         if (!dev)
2780                 return ERR_PTR(-ENOMEM);
2781
2782         dev_net_set(dev, net);
2783         dev->rtnl_link_ops = ops;
2784         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2785
2786         if (tb[IFLA_MTU])
2787                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2788         if (tb[IFLA_ADDRESS]) {
2789                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2790                                 nla_len(tb[IFLA_ADDRESS]));
2791                 dev->addr_assign_type = NET_ADDR_SET;
2792         }
2793         if (tb[IFLA_BROADCAST])
2794                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2795                                 nla_len(tb[IFLA_BROADCAST]));
2796         if (tb[IFLA_TXQLEN])
2797                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2798         if (tb[IFLA_OPERSTATE])
2799                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2800         if (tb[IFLA_LINKMODE])
2801                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2802         if (tb[IFLA_GROUP])
2803                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2804         if (tb[IFLA_GSO_MAX_SIZE])
2805                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2806         if (tb[IFLA_GSO_MAX_SEGS])
2807                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2808
2809         return dev;
2810 }
2811 EXPORT_SYMBOL(rtnl_create_link);
2812
2813 static int rtnl_group_changelink(const struct sk_buff *skb,
2814                 struct net *net, int group,
2815                 struct ifinfomsg *ifm,
2816                 struct netlink_ext_ack *extack,
2817                 struct nlattr **tb)
2818 {
2819         struct net_device *dev, *aux;
2820         int err;
2821
2822         for_each_netdev_safe(net, dev, aux) {
2823                 if (dev->group == group) {
2824                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2825                         if (err < 0)
2826                                 return err;
2827                 }
2828         }
2829
2830         return 0;
2831 }
2832
2833 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2834                         struct netlink_ext_ack *extack)
2835 {
2836         struct net *net = sock_net(skb->sk);
2837         const struct rtnl_link_ops *ops;
2838         const struct rtnl_link_ops *m_ops = NULL;
2839         struct net_device *dev;
2840         struct net_device *master_dev = NULL;
2841         struct ifinfomsg *ifm;
2842         char kind[MODULE_NAME_LEN];
2843         char ifname[IFNAMSIZ];
2844         struct nlattr *tb[IFLA_MAX+1];
2845         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2846         unsigned char name_assign_type = NET_NAME_USER;
2847         int err;
2848
2849 #ifdef CONFIG_MODULES
2850 replay:
2851 #endif
2852         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2853         if (err < 0)
2854                 return err;
2855
2856         err = rtnl_ensure_unique_netns(tb, extack, false);
2857         if (err < 0)
2858                 return err;
2859
2860         if (tb[IFLA_IFNAME])
2861                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2862         else
2863                 ifname[0] = '\0';
2864
2865         ifm = nlmsg_data(nlh);
2866         if (ifm->ifi_index > 0)
2867                 dev = __dev_get_by_index(net, ifm->ifi_index);
2868         else {
2869                 if (ifname[0])
2870                         dev = __dev_get_by_name(net, ifname);
2871                 else
2872                         dev = NULL;
2873         }
2874
2875         if (dev) {
2876                 master_dev = netdev_master_upper_dev_get(dev);
2877                 if (master_dev)
2878                         m_ops = master_dev->rtnl_link_ops;
2879         }
2880
2881         err = validate_linkmsg(dev, tb);
2882         if (err < 0)
2883                 return err;
2884
2885         if (tb[IFLA_LINKINFO]) {
2886                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2887                                        tb[IFLA_LINKINFO], ifla_info_policy,
2888                                        NULL);
2889                 if (err < 0)
2890                         return err;
2891         } else
2892                 memset(linkinfo, 0, sizeof(linkinfo));
2893
2894         if (linkinfo[IFLA_INFO_KIND]) {
2895                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2896                 ops = rtnl_link_ops_get(kind);
2897         } else {
2898                 kind[0] = '\0';
2899                 ops = NULL;
2900         }
2901
2902         if (1) {
2903                 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2904                 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2905                 struct nlattr **data = NULL;
2906                 struct nlattr **slave_data = NULL;
2907                 struct net *dest_net, *link_net = NULL;
2908
2909                 if (ops) {
2910                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2911                                 err = nla_parse_nested(attr, ops->maxtype,
2912                                                        linkinfo[IFLA_INFO_DATA],
2913                                                        ops->policy, NULL);
2914                                 if (err < 0)
2915                                         return err;
2916                                 data = attr;
2917                         }
2918                         if (ops->validate) {
2919                                 err = ops->validate(tb, data, extack);
2920                                 if (err < 0)
2921                                         return err;
2922                         }
2923                 }
2924
2925                 if (m_ops) {
2926                         if (m_ops->slave_maxtype &&
2927                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2928                                 err = nla_parse_nested(slave_attr,
2929                                                        m_ops->slave_maxtype,
2930                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2931                                                        m_ops->slave_policy,
2932                                                        NULL);
2933                                 if (err < 0)
2934                                         return err;
2935                                 slave_data = slave_attr;
2936                         }
2937                 }
2938
2939                 if (dev) {
2940                         int status = 0;
2941
2942                         if (nlh->nlmsg_flags & NLM_F_EXCL)
2943                                 return -EEXIST;
2944                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
2945                                 return -EOPNOTSUPP;
2946
2947                         if (linkinfo[IFLA_INFO_DATA]) {
2948                                 if (!ops || ops != dev->rtnl_link_ops ||
2949                                     !ops->changelink)
2950                                         return -EOPNOTSUPP;
2951
2952                                 err = ops->changelink(dev, tb, data, extack);
2953                                 if (err < 0)
2954                                         return err;
2955                                 status |= DO_SETLINK_NOTIFY;
2956                         }
2957
2958                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2959                                 if (!m_ops || !m_ops->slave_changelink)
2960                                         return -EOPNOTSUPP;
2961
2962                                 err = m_ops->slave_changelink(master_dev, dev,
2963                                                               tb, slave_data,
2964                                                               extack);
2965                                 if (err < 0)
2966                                         return err;
2967                                 status |= DO_SETLINK_NOTIFY;
2968                         }
2969
2970                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
2971                                           status);
2972                 }
2973
2974                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2975                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2976                                 return rtnl_group_changelink(skb, net,
2977                                                 nla_get_u32(tb[IFLA_GROUP]),
2978                                                 ifm, extack, tb);
2979                         return -ENODEV;
2980                 }
2981
2982                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2983                         return -EOPNOTSUPP;
2984
2985                 if (!ops) {
2986 #ifdef CONFIG_MODULES
2987                         if (kind[0]) {
2988                                 __rtnl_unlock();
2989                                 request_module("rtnl-link-%s", kind);
2990                                 rtnl_lock();
2991                                 ops = rtnl_link_ops_get(kind);
2992                                 if (ops)
2993                                         goto replay;
2994                         }
2995 #endif
2996                         return -EOPNOTSUPP;
2997                 }
2998
2999                 if (!ops->setup)
3000                         return -EOPNOTSUPP;
3001
3002                 if (!ifname[0]) {
3003                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3004                         name_assign_type = NET_NAME_ENUM;
3005                 }
3006
3007                 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3008                 if (IS_ERR(dest_net))
3009                         return PTR_ERR(dest_net);
3010
3011                 if (tb[IFLA_LINK_NETNSID]) {
3012                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3013
3014                         link_net = get_net_ns_by_id(dest_net, id);
3015                         if (!link_net) {
3016                                 err =  -EINVAL;
3017                                 goto out;
3018                         }
3019                         err = -EPERM;
3020                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3021                                 goto out;
3022                 }
3023
3024                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3025                                        name_assign_type, ops, tb);
3026                 if (IS_ERR(dev)) {
3027                         err = PTR_ERR(dev);
3028                         goto out;
3029                 }
3030
3031                 dev->ifindex = ifm->ifi_index;
3032
3033                 if (ops->newlink) {
3034                         err = ops->newlink(link_net ? : net, dev, tb, data,
3035                                            extack);
3036                         /* Drivers should call free_netdev() in ->destructor
3037                          * and unregister it on failure after registration
3038                          * so that device could be finally freed in rtnl_unlock.
3039                          */
3040                         if (err < 0) {
3041                                 /* If device is not registered at all, free it now */
3042                                 if (dev->reg_state == NETREG_UNINITIALIZED)
3043                                         free_netdev(dev);
3044                                 goto out;
3045                         }
3046                 } else {
3047                         err = register_netdevice(dev);
3048                         if (err < 0) {
3049                                 free_netdev(dev);
3050                                 goto out;
3051                         }
3052                 }
3053                 err = rtnl_configure_link(dev, ifm);
3054                 if (err < 0)
3055                         goto out_unregister;
3056                 if (link_net) {
3057                         err = dev_change_net_namespace(dev, dest_net, ifname);
3058                         if (err < 0)
3059                                 goto out_unregister;
3060                 }
3061                 if (tb[IFLA_MASTER]) {
3062                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
3063                                             extack);
3064                         if (err)
3065                                 goto out_unregister;
3066                 }
3067 out:
3068                 if (link_net)
3069                         put_net(link_net);
3070                 put_net(dest_net);
3071                 return err;
3072 out_unregister:
3073                 if (ops->newlink) {
3074                         LIST_HEAD(list_kill);
3075
3076                         ops->dellink(dev, &list_kill);
3077                         unregister_netdevice_many(&list_kill);
3078                 } else {
3079                         unregister_netdevice(dev);
3080                 }
3081                 goto out;
3082         }
3083 }
3084
3085 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3086                         struct netlink_ext_ack *extack)
3087 {
3088         struct net *net = sock_net(skb->sk);
3089         struct net *tgt_net = net;
3090         struct ifinfomsg *ifm;
3091         char ifname[IFNAMSIZ];
3092         struct nlattr *tb[IFLA_MAX+1];
3093         struct net_device *dev = NULL;
3094         struct sk_buff *nskb;
3095         int netnsid = -1;
3096         int err;
3097         u32 ext_filter_mask = 0;
3098
3099         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3100         if (err < 0)
3101                 return err;
3102
3103         err = rtnl_ensure_unique_netns(tb, extack, true);
3104         if (err < 0)
3105                 return err;
3106
3107         if (tb[IFLA_IF_NETNSID]) {
3108                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3109                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3110                 if (IS_ERR(tgt_net))
3111                         return PTR_ERR(tgt_net);
3112         }
3113
3114         if (tb[IFLA_IFNAME])
3115                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3116
3117         if (tb[IFLA_EXT_MASK])
3118                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3119
3120         err = -EINVAL;
3121         ifm = nlmsg_data(nlh);
3122         if (ifm->ifi_index > 0)
3123                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3124         else if (tb[IFLA_IFNAME])
3125                 dev = __dev_get_by_name(tgt_net, ifname);
3126         else
3127                 goto out;
3128
3129         err = -ENODEV;
3130         if (dev == NULL)
3131                 goto out;
3132
3133         err = -ENOBUFS;
3134         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3135         if (nskb == NULL)
3136                 goto out;
3137
3138         err = rtnl_fill_ifinfo(nskb, dev, net,
3139                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3140                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3141                                0, NULL, 0, netnsid);
3142         if (err < 0) {
3143                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3144                 WARN_ON(err == -EMSGSIZE);
3145                 kfree_skb(nskb);
3146         } else
3147                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3148 out:
3149         if (netnsid >= 0)
3150                 put_net(tgt_net);
3151
3152         return err;
3153 }
3154
3155 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3156 {
3157         struct net *net = sock_net(skb->sk);
3158         struct net_device *dev;
3159         struct nlattr *tb[IFLA_MAX+1];
3160         u32 ext_filter_mask = 0;
3161         u16 min_ifinfo_dump_size = 0;
3162         int hdrlen;
3163
3164         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3165         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3166                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3167
3168         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3169                 if (tb[IFLA_EXT_MASK])
3170                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3171         }
3172
3173         if (!ext_filter_mask)
3174                 return NLMSG_GOODSIZE;
3175         /*
3176          * traverse the list of net devices and compute the minimum
3177          * buffer size based upon the filter mask.
3178          */
3179         rcu_read_lock();
3180         for_each_netdev_rcu(net, dev) {
3181                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3182                                              if_nlmsg_size(dev,
3183                                                            ext_filter_mask));
3184         }
3185         rcu_read_unlock();
3186
3187         return nlmsg_total_size(min_ifinfo_dump_size);
3188 }
3189
3190 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3191 {
3192         int idx;
3193         int s_idx = cb->family;
3194
3195         if (s_idx == 0)
3196                 s_idx = 1;
3197
3198         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3199                 struct rtnl_link **tab;
3200                 int type = cb->nlh->nlmsg_type-RTM_BASE;
3201                 struct rtnl_link *link;
3202                 rtnl_dumpit_func dumpit;
3203
3204                 if (idx < s_idx || idx == PF_PACKET)
3205                         continue;
3206
3207                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3208                         continue;
3209
3210                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3211                 if (!tab)
3212                         continue;
3213
3214                 link = tab[type];
3215                 if (!link)
3216                         continue;
3217
3218                 dumpit = link->dumpit;
3219                 if (!dumpit)
3220                         continue;
3221
3222                 if (idx > s_idx) {
3223                         memset(&cb->args[0], 0, sizeof(cb->args));
3224                         cb->prev_seq = 0;
3225                         cb->seq = 0;
3226                 }
3227                 if (dumpit(skb, cb))
3228                         break;
3229         }
3230         cb->family = idx;
3231
3232         return skb->len;
3233 }
3234
3235 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3236                                        unsigned int change,
3237                                        u32 event, gfp_t flags, int *new_nsid,
3238                                        int new_ifindex)
3239 {
3240         struct net *net = dev_net(dev);
3241         struct sk_buff *skb;
3242         int err = -ENOBUFS;
3243         size_t if_info_size;
3244
3245         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3246         if (skb == NULL)
3247                 goto errout;
3248
3249         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3250                                type, 0, 0, change, 0, 0, event,
3251                                new_nsid, new_ifindex, -1);
3252         if (err < 0) {
3253                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3254                 WARN_ON(err == -EMSGSIZE);
3255                 kfree_skb(skb);
3256                 goto errout;
3257         }
3258         return skb;
3259 errout:
3260         if (err < 0)
3261                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3262         return NULL;
3263 }
3264
3265 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3266 {
3267         struct net *net = dev_net(dev);
3268
3269         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3270 }
3271
3272 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3273                                unsigned int change, u32 event,
3274                                gfp_t flags, int *new_nsid, int new_ifindex)
3275 {
3276         struct sk_buff *skb;
3277
3278         if (dev->reg_state != NETREG_REGISTERED)
3279                 return;
3280
3281         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3282                                      new_ifindex);
3283         if (skb)
3284                 rtmsg_ifinfo_send(skb, dev, flags);
3285 }
3286
3287 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3288                   gfp_t flags)
3289 {
3290         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3291                            NULL, 0);
3292 }
3293
3294 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3295                          gfp_t flags, int *new_nsid, int new_ifindex)
3296 {
3297         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3298                            new_nsid, new_ifindex);
3299 }
3300
3301 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3302                                    struct net_device *dev,
3303                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3304                                    int type, unsigned int flags,
3305                                    int nlflags, u16 ndm_state)
3306 {
3307         struct nlmsghdr *nlh;
3308         struct ndmsg *ndm;
3309
3310         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3311         if (!nlh)
3312                 return -EMSGSIZE;
3313
3314         ndm = nlmsg_data(nlh);
3315         ndm->ndm_family  = AF_BRIDGE;
3316         ndm->ndm_pad1    = 0;
3317         ndm->ndm_pad2    = 0;
3318         ndm->ndm_flags   = flags;
3319         ndm->ndm_type    = 0;
3320         ndm->ndm_ifindex = dev->ifindex;
3321         ndm->ndm_state   = ndm_state;
3322
3323         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3324                 goto nla_put_failure;
3325         if (vid)
3326                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3327                         goto nla_put_failure;
3328
3329         nlmsg_end(skb, nlh);
3330         return 0;
3331
3332 nla_put_failure:
3333         nlmsg_cancel(skb, nlh);
3334         return -EMSGSIZE;
3335 }
3336
3337 static inline size_t rtnl_fdb_nlmsg_size(void)
3338 {
3339         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3340                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3341                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3342                0;
3343 }
3344
3345 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3346                             u16 ndm_state)
3347 {
3348         struct net *net = dev_net(dev);
3349         struct sk_buff *skb;
3350         int err = -ENOBUFS;
3351
3352         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3353         if (!skb)
3354                 goto errout;
3355
3356         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3357                                       0, 0, type, NTF_SELF, 0, ndm_state);
3358         if (err < 0) {
3359                 kfree_skb(skb);
3360                 goto errout;
3361         }
3362
3363         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3364         return;
3365 errout:
3366         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3367 }
3368
3369 /**
3370  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3371  */
3372 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3373                      struct nlattr *tb[],
3374                      struct net_device *dev,
3375                      const unsigned char *addr, u16 vid,
3376                      u16 flags)
3377 {
3378         int err = -EINVAL;
3379
3380         /* If aging addresses are supported device will need to
3381          * implement its own handler for this.
3382          */
3383         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3384                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3385                 return err;
3386         }
3387
3388         if (vid) {
3389                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3390                 return err;
3391         }
3392
3393         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3394                 err = dev_uc_add_excl(dev, addr);
3395         else if (is_multicast_ether_addr(addr))
3396                 err = dev_mc_add_excl(dev, addr);
3397
3398         /* Only return duplicate errors if NLM_F_EXCL is set */
3399         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3400                 err = 0;
3401
3402         return err;
3403 }
3404 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3405
3406 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3407                          struct netlink_ext_ack *extack)
3408 {
3409         u16 vid = 0;
3410
3411         if (vlan_attr) {
3412                 if (nla_len(vlan_attr) != sizeof(u16)) {
3413                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3414                         return -EINVAL;
3415                 }
3416
3417                 vid = nla_get_u16(vlan_attr);
3418
3419                 if (!vid || vid >= VLAN_VID_MASK) {
3420                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3421                         return -EINVAL;
3422                 }
3423         }
3424         *p_vid = vid;
3425         return 0;
3426 }
3427
3428 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3429                         struct netlink_ext_ack *extack)
3430 {
3431         struct net *net = sock_net(skb->sk);
3432         struct ndmsg *ndm;
3433         struct nlattr *tb[NDA_MAX+1];
3434         struct net_device *dev;
3435         u8 *addr;
3436         u16 vid;
3437         int err;
3438
3439         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3440         if (err < 0)
3441                 return err;
3442
3443         ndm = nlmsg_data(nlh);
3444         if (ndm->ndm_ifindex == 0) {
3445                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3446                 return -EINVAL;
3447         }
3448
3449         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3450         if (dev == NULL) {
3451                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3452                 return -ENODEV;
3453         }
3454
3455         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3456                 NL_SET_ERR_MSG(extack, "invalid address");
3457                 return -EINVAL;
3458         }
3459
3460         addr = nla_data(tb[NDA_LLADDR]);
3461
3462         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3463         if (err)
3464                 return err;
3465
3466         err = -EOPNOTSUPP;
3467
3468         /* Support fdb on master device the net/bridge default case */
3469         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3470             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3471                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3472                 const struct net_device_ops *ops = br_dev->netdev_ops;
3473
3474                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3475                                        nlh->nlmsg_flags);
3476                 if (err)
3477                         goto out;
3478                 else
3479                         ndm->ndm_flags &= ~NTF_MASTER;
3480         }
3481
3482         /* Embedded bridge, macvlan, and any other device support */
3483         if ((ndm->ndm_flags & NTF_SELF)) {
3484                 if (dev->netdev_ops->ndo_fdb_add)
3485                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3486                                                            vid,
3487                                                            nlh->nlmsg_flags);
3488                 else
3489                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3490                                                nlh->nlmsg_flags);
3491
3492                 if (!err) {
3493                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3494                                         ndm->ndm_state);
3495                         ndm->ndm_flags &= ~NTF_SELF;
3496                 }
3497         }
3498 out:
3499         return err;
3500 }
3501
3502 /**
3503  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3504  */
3505 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3506                      struct nlattr *tb[],
3507                      struct net_device *dev,
3508                      const unsigned char *addr, u16 vid)
3509 {
3510         int err = -EINVAL;
3511
3512         /* If aging addresses are supported device will need to
3513          * implement its own handler for this.
3514          */
3515         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3516                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3517                 return err;
3518         }
3519
3520         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3521                 err = dev_uc_del(dev, addr);
3522         else if (is_multicast_ether_addr(addr))
3523                 err = dev_mc_del(dev, addr);
3524
3525         return err;
3526 }
3527 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3528
3529 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3530                         struct netlink_ext_ack *extack)
3531 {
3532         struct net *net = sock_net(skb->sk);
3533         struct ndmsg *ndm;
3534         struct nlattr *tb[NDA_MAX+1];
3535         struct net_device *dev;
3536         int err = -EINVAL;
3537         __u8 *addr;
3538         u16 vid;
3539
3540         if (!netlink_capable(skb, CAP_NET_ADMIN))
3541                 return -EPERM;
3542
3543         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3544         if (err < 0)
3545                 return err;
3546
3547         ndm = nlmsg_data(nlh);
3548         if (ndm->ndm_ifindex == 0) {
3549                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3550                 return -EINVAL;
3551         }
3552
3553         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3554         if (dev == NULL) {
3555                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3556                 return -ENODEV;
3557         }
3558
3559         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3560                 NL_SET_ERR_MSG(extack, "invalid address");
3561                 return -EINVAL;
3562         }
3563
3564         addr = nla_data(tb[NDA_LLADDR]);
3565
3566         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3567         if (err)
3568                 return err;
3569
3570         err = -EOPNOTSUPP;
3571
3572         /* Support fdb on master device the net/bridge default case */
3573         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3574             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3575                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3576                 const struct net_device_ops *ops = br_dev->netdev_ops;
3577
3578                 if (ops->ndo_fdb_del)
3579                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3580
3581                 if (err)
3582                         goto out;
3583                 else
3584                         ndm->ndm_flags &= ~NTF_MASTER;
3585         }
3586
3587         /* Embedded bridge, macvlan, and any other device support */
3588         if (ndm->ndm_flags & NTF_SELF) {
3589                 if (dev->netdev_ops->ndo_fdb_del)
3590                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3591                                                            vid);
3592                 else
3593                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3594
3595                 if (!err) {
3596                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3597                                         ndm->ndm_state);
3598                         ndm->ndm_flags &= ~NTF_SELF;
3599                 }
3600         }
3601 out:
3602         return err;
3603 }
3604
3605 static int nlmsg_populate_fdb(struct sk_buff *skb,
3606                               struct netlink_callback *cb,
3607                               struct net_device *dev,
3608                               int *idx,
3609                               struct netdev_hw_addr_list *list)
3610 {
3611         struct netdev_hw_addr *ha;
3612         int err;
3613         u32 portid, seq;
3614
3615         portid = NETLINK_CB(cb->skb).portid;
3616         seq = cb->nlh->nlmsg_seq;
3617
3618         list_for_each_entry(ha, &list->list, list) {
3619                 if (*idx < cb->args[2])
3620                         goto skip;
3621
3622                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3623                                               portid, seq,
3624                                               RTM_NEWNEIGH, NTF_SELF,
3625                                               NLM_F_MULTI, NUD_PERMANENT);
3626                 if (err < 0)
3627                         return err;
3628 skip:
3629                 *idx += 1;
3630         }
3631         return 0;
3632 }
3633
3634 /**
3635  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3636  * @nlh: netlink message header
3637  * @dev: netdevice
3638  *
3639  * Default netdevice operation to dump the existing unicast address list.
3640  * Returns number of addresses from list put in skb.
3641  */
3642 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3643                       struct netlink_callback *cb,
3644                       struct net_device *dev,
3645                       struct net_device *filter_dev,
3646                       int *idx)
3647 {
3648         int err;
3649
3650         netif_addr_lock_bh(dev);
3651         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3652         if (err)
3653                 goto out;
3654         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3655 out:
3656         netif_addr_unlock_bh(dev);
3657         return err;
3658 }
3659 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3660
3661 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3662 {
3663         struct net_device *dev;
3664         struct nlattr *tb[IFLA_MAX+1];
3665         struct net_device *br_dev = NULL;
3666         const struct net_device_ops *ops = NULL;
3667         const struct net_device_ops *cops = NULL;
3668         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3669         struct net *net = sock_net(skb->sk);
3670         struct hlist_head *head;
3671         int brport_idx = 0;
3672         int br_idx = 0;
3673         int h, s_h;
3674         int idx = 0, s_idx;
3675         int err = 0;
3676         int fidx = 0;
3677
3678         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3679                           IFLA_MAX, ifla_policy, NULL);
3680         if (err < 0) {
3681                 return -EINVAL;
3682         } else if (err == 0) {
3683                 if (tb[IFLA_MASTER])
3684                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3685         }
3686
3687         brport_idx = ifm->ifi_index;
3688
3689         if (br_idx) {
3690                 br_dev = __dev_get_by_index(net, br_idx);
3691                 if (!br_dev)
3692                         return -ENODEV;
3693
3694                 ops = br_dev->netdev_ops;
3695         }
3696
3697         s_h = cb->args[0];
3698         s_idx = cb->args[1];
3699
3700         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3701                 idx = 0;
3702                 head = &net->dev_index_head[h];
3703                 hlist_for_each_entry(dev, head, index_hlist) {
3704
3705                         if (brport_idx && (dev->ifindex != brport_idx))
3706                                 continue;
3707
3708                         if (!br_idx) { /* user did not specify a specific bridge */
3709                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3710                                         br_dev = netdev_master_upper_dev_get(dev);
3711                                         cops = br_dev->netdev_ops;
3712                                 }
3713                         } else {
3714                                 if (dev != br_dev &&
3715                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3716                                         continue;
3717
3718                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3719                                     !(dev->priv_flags & IFF_EBRIDGE))
3720                                         continue;
3721                                 cops = ops;
3722                         }
3723
3724                         if (idx < s_idx)
3725                                 goto cont;
3726
3727                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3728                                 if (cops && cops->ndo_fdb_dump) {
3729                                         err = cops->ndo_fdb_dump(skb, cb,
3730                                                                 br_dev, dev,
3731                                                                 &fidx);
3732                                         if (err == -EMSGSIZE)
3733                                                 goto out;
3734                                 }
3735                         }
3736
3737                         if (dev->netdev_ops->ndo_fdb_dump)
3738                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3739                                                                     dev, NULL,
3740                                                                     &fidx);
3741                         else
3742                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3743                                                         &fidx);
3744                         if (err == -EMSGSIZE)
3745                                 goto out;
3746
3747                         cops = NULL;
3748
3749                         /* reset fdb offset to 0 for rest of the interfaces */
3750                         cb->args[2] = 0;
3751                         fidx = 0;
3752 cont:
3753                         idx++;
3754                 }
3755         }
3756
3757 out:
3758         cb->args[0] = h;
3759         cb->args[1] = idx;
3760         cb->args[2] = fidx;
3761
3762         return skb->len;
3763 }
3764
3765 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3766                                unsigned int attrnum, unsigned int flag)
3767 {
3768         if (mask & flag)
3769                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3770         return 0;
3771 }
3772
3773 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3774                             struct net_device *dev, u16 mode,
3775                             u32 flags, u32 mask, int nlflags,
3776                             u32 filter_mask,
3777                             int (*vlan_fill)(struct sk_buff *skb,
3778                                              struct net_device *dev,
3779                                              u32 filter_mask))
3780 {
3781         struct nlmsghdr *nlh;
3782         struct ifinfomsg *ifm;
3783         struct nlattr *br_afspec;
3784         struct nlattr *protinfo;
3785         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3786         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3787         int err = 0;
3788
3789         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3790         if (nlh == NULL)
3791                 return -EMSGSIZE;
3792
3793         ifm = nlmsg_data(nlh);
3794         ifm->ifi_family = AF_BRIDGE;
3795         ifm->__ifi_pad = 0;
3796         ifm->ifi_type = dev->type;
3797         ifm->ifi_index = dev->ifindex;
3798         ifm->ifi_flags = dev_get_flags(dev);
3799         ifm->ifi_change = 0;
3800
3801
3802         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3803             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3804             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3805             (br_dev &&
3806              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3807             (dev->addr_len &&
3808              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3809             (dev->ifindex != dev_get_iflink(dev) &&
3810              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3811                 goto nla_put_failure;
3812
3813         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3814         if (!br_afspec)
3815                 goto nla_put_failure;
3816
3817         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3818                 nla_nest_cancel(skb, br_afspec);
3819                 goto nla_put_failure;
3820         }
3821
3822         if (mode != BRIDGE_MODE_UNDEF) {
3823                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3824                         nla_nest_cancel(skb, br_afspec);
3825                         goto nla_put_failure;
3826                 }
3827         }
3828         if (vlan_fill) {
3829                 err = vlan_fill(skb, dev, filter_mask);
3830                 if (err) {
3831                         nla_nest_cancel(skb, br_afspec);
3832                         goto nla_put_failure;
3833                 }
3834         }
3835         nla_nest_end(skb, br_afspec);
3836
3837         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3838         if (!protinfo)
3839                 goto nla_put_failure;
3840
3841         if (brport_nla_put_flag(skb, flags, mask,
3842                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3843             brport_nla_put_flag(skb, flags, mask,
3844                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3845             brport_nla_put_flag(skb, flags, mask,
3846                                 IFLA_BRPORT_FAST_LEAVE,
3847                                 BR_MULTICAST_FAST_LEAVE) ||
3848             brport_nla_put_flag(skb, flags, mask,
3849                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3850             brport_nla_put_flag(skb, flags, mask,
3851                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3852             brport_nla_put_flag(skb, flags, mask,
3853                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3854             brport_nla_put_flag(skb, flags, mask,
3855                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3856             brport_nla_put_flag(skb, flags, mask,
3857                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3858                 nla_nest_cancel(skb, protinfo);
3859                 goto nla_put_failure;
3860         }
3861
3862         nla_nest_end(skb, protinfo);
3863
3864         nlmsg_end(skb, nlh);
3865         return 0;
3866 nla_put_failure:
3867         nlmsg_cancel(skb, nlh);
3868         return err ? err : -EMSGSIZE;
3869 }
3870 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3871
3872 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3873 {
3874         struct net *net = sock_net(skb->sk);
3875         struct net_device *dev;
3876         int idx = 0;
3877         u32 portid = NETLINK_CB(cb->skb).portid;
3878         u32 seq = cb->nlh->nlmsg_seq;
3879         u32 filter_mask = 0;
3880         int err;
3881
3882         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3883                 struct nlattr *extfilt;
3884
3885                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3886                                           IFLA_EXT_MASK);
3887                 if (extfilt) {
3888                         if (nla_len(extfilt) < sizeof(filter_mask))
3889                                 return -EINVAL;
3890
3891                         filter_mask = nla_get_u32(extfilt);
3892                 }
3893         }
3894
3895         rcu_read_lock();
3896         for_each_netdev_rcu(net, dev) {
3897                 const struct net_device_ops *ops = dev->netdev_ops;
3898                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3899
3900                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3901                         if (idx >= cb->args[0]) {
3902                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3903                                                 skb, portid, seq, dev,
3904                                                 filter_mask, NLM_F_MULTI);
3905                                 if (err < 0 && err != -EOPNOTSUPP) {
3906                                         if (likely(skb->len))
3907                                                 break;
3908
3909                                         goto out_err;
3910                                 }
3911                         }
3912                         idx++;
3913                 }
3914
3915                 if (ops->ndo_bridge_getlink) {
3916                         if (idx >= cb->args[0]) {
3917                                 err = ops->ndo_bridge_getlink(skb, portid,
3918                                                               seq, dev,
3919                                                               filter_mask,
3920                                                               NLM_F_MULTI);
3921                                 if (err < 0 && err != -EOPNOTSUPP) {
3922                                         if (likely(skb->len))
3923                                                 break;
3924
3925                                         goto out_err;
3926                                 }
3927                         }
3928                         idx++;
3929                 }
3930         }
3931         err = skb->len;
3932 out_err:
3933         rcu_read_unlock();
3934         cb->args[0] = idx;
3935
3936         return err;
3937 }
3938
3939 static inline size_t bridge_nlmsg_size(void)
3940 {
3941         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3942                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
3943                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
3944                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
3945                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
3946                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
3947                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
3948                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
3949                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3950                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
3951                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
3952 }
3953
3954 static int rtnl_bridge_notify(struct net_device *dev)
3955 {
3956         struct net *net = dev_net(dev);
3957         struct sk_buff *skb;
3958         int err = -EOPNOTSUPP;
3959
3960         if (!dev->netdev_ops->ndo_bridge_getlink)
3961                 return 0;
3962
3963         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3964         if (!skb) {
3965                 err = -ENOMEM;
3966                 goto errout;
3967         }
3968
3969         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3970         if (err < 0)
3971                 goto errout;
3972
3973         if (!skb->len)
3974                 goto errout;
3975
3976         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3977         return 0;
3978 errout:
3979         WARN_ON(err == -EMSGSIZE);
3980         kfree_skb(skb);
3981         if (err)
3982                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3983         return err;
3984 }
3985
3986 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3987                                struct netlink_ext_ack *extack)
3988 {
3989         struct net *net = sock_net(skb->sk);
3990         struct ifinfomsg *ifm;
3991         struct net_device *dev;
3992         struct nlattr *br_spec, *attr = NULL;
3993         int rem, err = -EOPNOTSUPP;
3994         u16 flags = 0;
3995         bool have_flags = false;
3996
3997         if (nlmsg_len(nlh) < sizeof(*ifm))
3998                 return -EINVAL;
3999
4000         ifm = nlmsg_data(nlh);
4001         if (ifm->ifi_family != AF_BRIDGE)
4002                 return -EPFNOSUPPORT;
4003
4004         dev = __dev_get_by_index(net, ifm->ifi_index);
4005         if (!dev) {
4006                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4007                 return -ENODEV;
4008         }
4009
4010         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4011         if (br_spec) {
4012                 nla_for_each_nested(attr, br_spec, rem) {
4013                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4014                                 if (nla_len(attr) < sizeof(flags))
4015                                         return -EINVAL;
4016
4017                                 have_flags = true;
4018                                 flags = nla_get_u16(attr);
4019                                 break;
4020                         }
4021                 }
4022         }
4023
4024         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4025                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4026
4027                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4028                         err = -EOPNOTSUPP;
4029                         goto out;
4030                 }
4031
4032                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
4033                 if (err)
4034                         goto out;
4035
4036                 flags &= ~BRIDGE_FLAGS_MASTER;
4037         }
4038
4039         if ((flags & BRIDGE_FLAGS_SELF)) {
4040                 if (!dev->netdev_ops->ndo_bridge_setlink)
4041                         err = -EOPNOTSUPP;
4042                 else
4043                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4044                                                                   flags);
4045                 if (!err) {
4046                         flags &= ~BRIDGE_FLAGS_SELF;
4047
4048                         /* Generate event to notify upper layer of bridge
4049                          * change
4050                          */
4051                         err = rtnl_bridge_notify(dev);
4052                 }
4053         }
4054
4055         if (have_flags)
4056                 memcpy(nla_data(attr), &flags, sizeof(flags));
4057 out:
4058         return err;
4059 }
4060
4061 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4062                                struct netlink_ext_ack *extack)
4063 {
4064         struct net *net = sock_net(skb->sk);
4065         struct ifinfomsg *ifm;
4066         struct net_device *dev;
4067         struct nlattr *br_spec, *attr = NULL;
4068         int rem, err = -EOPNOTSUPP;
4069         u16 flags = 0;
4070         bool have_flags = false;
4071
4072         if (nlmsg_len(nlh) < sizeof(*ifm))
4073                 return -EINVAL;
4074
4075         ifm = nlmsg_data(nlh);
4076         if (ifm->ifi_family != AF_BRIDGE)
4077                 return -EPFNOSUPPORT;
4078
4079         dev = __dev_get_by_index(net, ifm->ifi_index);
4080         if (!dev) {
4081                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4082                 return -ENODEV;
4083         }
4084
4085         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4086         if (br_spec) {
4087                 nla_for_each_nested(attr, br_spec, rem) {
4088                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4089                                 if (nla_len(attr) < sizeof(flags))
4090                                         return -EINVAL;
4091
4092                                 have_flags = true;
4093                                 flags = nla_get_u16(attr);
4094                                 break;
4095                         }
4096                 }
4097         }
4098
4099         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4100                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4101
4102                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4103                         err = -EOPNOTSUPP;
4104                         goto out;
4105                 }
4106
4107                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4108                 if (err)
4109                         goto out;
4110
4111                 flags &= ~BRIDGE_FLAGS_MASTER;
4112         }
4113
4114         if ((flags & BRIDGE_FLAGS_SELF)) {
4115                 if (!dev->netdev_ops->ndo_bridge_dellink)
4116                         err = -EOPNOTSUPP;
4117                 else
4118                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4119                                                                   flags);
4120
4121                 if (!err) {
4122                         flags &= ~BRIDGE_FLAGS_SELF;
4123
4124                         /* Generate event to notify upper layer of bridge
4125                          * change
4126                          */
4127                         err = rtnl_bridge_notify(dev);
4128                 }
4129         }
4130
4131         if (have_flags)
4132                 memcpy(nla_data(attr), &flags, sizeof(flags));
4133 out:
4134         return err;
4135 }
4136
4137 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4138 {
4139         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4140                (!idxattr || idxattr == attrid);
4141 }
4142
4143 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4144 static int rtnl_get_offload_stats_attr_size(int attr_id)
4145 {
4146         switch (attr_id) {
4147         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4148                 return sizeof(struct rtnl_link_stats64);
4149         }
4150
4151         return 0;
4152 }
4153
4154 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4155                                   int *prividx)
4156 {
4157         struct nlattr *attr = NULL;
4158         int attr_id, size;
4159         void *attr_data;
4160         int err;
4161
4162         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4163               dev->netdev_ops->ndo_get_offload_stats))
4164                 return -ENODATA;
4165
4166         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4167              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4168                 if (attr_id < *prividx)
4169                         continue;
4170
4171                 size = rtnl_get_offload_stats_attr_size(attr_id);
4172                 if (!size)
4173                         continue;
4174
4175                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4176                         continue;
4177
4178                 attr = nla_reserve_64bit(skb, attr_id, size,
4179                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4180                 if (!attr)
4181                         goto nla_put_failure;
4182
4183                 attr_data = nla_data(attr);
4184                 memset(attr_data, 0, size);
4185                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4186                                                              attr_data);
4187                 if (err)
4188                         goto get_offload_stats_failure;
4189         }
4190
4191         if (!attr)
4192                 return -ENODATA;
4193
4194         *prividx = 0;
4195         return 0;
4196
4197 nla_put_failure:
4198         err = -EMSGSIZE;
4199 get_offload_stats_failure:
4200         *prividx = attr_id;
4201         return err;
4202 }
4203
4204 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4205 {
4206         int nla_size = 0;
4207         int attr_id;
4208         int size;
4209
4210         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4211               dev->netdev_ops->ndo_get_offload_stats))
4212                 return 0;
4213
4214         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4215              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4216                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4217                         continue;
4218                 size = rtnl_get_offload_stats_attr_size(attr_id);
4219                 nla_size += nla_total_size_64bit(size);
4220         }
4221
4222         if (nla_size != 0)
4223                 nla_size += nla_total_size(0);
4224
4225         return nla_size;
4226 }
4227
4228 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4229                                int type, u32 pid, u32 seq, u32 change,
4230                                unsigned int flags, unsigned int filter_mask,
4231                                int *idxattr, int *prividx)
4232 {
4233         struct if_stats_msg *ifsm;
4234         struct nlmsghdr *nlh;
4235         struct nlattr *attr;
4236         int s_prividx = *prividx;
4237         int err;
4238
4239         ASSERT_RTNL();
4240
4241         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4242         if (!nlh)
4243                 return -EMSGSIZE;
4244
4245         ifsm = nlmsg_data(nlh);
4246         ifsm->family = PF_UNSPEC;
4247         ifsm->pad1 = 0;
4248         ifsm->pad2 = 0;
4249         ifsm->ifindex = dev->ifindex;
4250         ifsm->filter_mask = filter_mask;
4251
4252         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4253                 struct rtnl_link_stats64 *sp;
4254
4255                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4256                                          sizeof(struct rtnl_link_stats64),
4257                                          IFLA_STATS_UNSPEC);
4258                 if (!attr)
4259                         goto nla_put_failure;
4260
4261                 sp = nla_data(attr);
4262                 dev_get_stats(dev, sp);
4263         }
4264
4265         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4266                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4267
4268                 if (ops && ops->fill_linkxstats) {
4269                         *idxattr = IFLA_STATS_LINK_XSTATS;
4270                         attr = nla_nest_start(skb,
4271                                               IFLA_STATS_LINK_XSTATS);
4272                         if (!attr)
4273                                 goto nla_put_failure;
4274
4275                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4276                         nla_nest_end(skb, attr);
4277                         if (err)
4278                                 goto nla_put_failure;
4279                         *idxattr = 0;
4280                 }
4281         }
4282
4283         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4284                              *idxattr)) {
4285                 const struct rtnl_link_ops *ops = NULL;
4286                 const struct net_device *master;
4287
4288                 master = netdev_master_upper_dev_get(dev);
4289                 if (master)
4290                         ops = master->rtnl_link_ops;
4291                 if (ops && ops->fill_linkxstats) {
4292                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4293                         attr = nla_nest_start(skb,
4294                                               IFLA_STATS_LINK_XSTATS_SLAVE);
4295                         if (!attr)
4296                                 goto nla_put_failure;
4297
4298                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4299                         nla_nest_end(skb, attr);
4300                         if (err)
4301                                 goto nla_put_failure;
4302                         *idxattr = 0;
4303                 }
4304         }
4305
4306         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4307                              *idxattr)) {
4308                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4309                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4310                 if (!attr)
4311                         goto nla_put_failure;
4312
4313                 err = rtnl_get_offload_stats(skb, dev, prividx);
4314                 if (err == -ENODATA)
4315                         nla_nest_cancel(skb, attr);
4316                 else
4317                         nla_nest_end(skb, attr);
4318
4319                 if (err && err != -ENODATA)
4320                         goto nla_put_failure;
4321                 *idxattr = 0;
4322         }
4323
4324         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4325                 struct rtnl_af_ops *af_ops;
4326
4327                 *idxattr = IFLA_STATS_AF_SPEC;
4328                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4329                 if (!attr)
4330                         goto nla_put_failure;
4331
4332                 rcu_read_lock();
4333                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4334                         if (af_ops->fill_stats_af) {
4335                                 struct nlattr *af;
4336                                 int err;
4337
4338                                 af = nla_nest_start(skb, af_ops->family);
4339                                 if (!af) {
4340                                         rcu_read_unlock();
4341                                         goto nla_put_failure;
4342                                 }
4343                                 err = af_ops->fill_stats_af(skb, dev);
4344
4345                                 if (err == -ENODATA) {
4346                                         nla_nest_cancel(skb, af);
4347                                 } else if (err < 0) {
4348                                         rcu_read_unlock();
4349                                         goto nla_put_failure;
4350                                 }
4351
4352                                 nla_nest_end(skb, af);
4353                         }
4354                 }
4355                 rcu_read_unlock();
4356
4357                 nla_nest_end(skb, attr);
4358
4359                 *idxattr = 0;
4360         }
4361
4362         nlmsg_end(skb, nlh);
4363
4364         return 0;
4365
4366 nla_put_failure:
4367         /* not a multi message or no progress mean a real error */
4368         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4369                 nlmsg_cancel(skb, nlh);
4370         else
4371                 nlmsg_end(skb, nlh);
4372
4373         return -EMSGSIZE;
4374 }
4375
4376 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4377                                   u32 filter_mask)
4378 {
4379         size_t size = 0;
4380
4381         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4382                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4383
4384         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4385                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4386                 int attr = IFLA_STATS_LINK_XSTATS;
4387
4388                 if (ops && ops->get_linkxstats_size) {
4389                         size += nla_total_size(ops->get_linkxstats_size(dev,
4390                                                                         attr));
4391                         /* for IFLA_STATS_LINK_XSTATS */
4392                         size += nla_total_size(0);
4393                 }
4394         }
4395
4396         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4397                 struct net_device *_dev = (struct net_device *)dev;
4398                 const struct rtnl_link_ops *ops = NULL;
4399                 const struct net_device *master;
4400
4401                 /* netdev_master_upper_dev_get can't take const */
4402                 master = netdev_master_upper_dev_get(_dev);
4403                 if (master)
4404                         ops = master->rtnl_link_ops;
4405                 if (ops && ops->get_linkxstats_size) {
4406                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4407
4408                         size += nla_total_size(ops->get_linkxstats_size(dev,
4409                                                                         attr));
4410                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4411                         size += nla_total_size(0);
4412                 }
4413         }
4414
4415         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4416                 size += rtnl_get_offload_stats_size(dev);
4417
4418         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4419                 struct rtnl_af_ops *af_ops;
4420
4421                 /* for IFLA_STATS_AF_SPEC */
4422                 size += nla_total_size(0);
4423
4424                 rcu_read_lock();
4425                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4426                         if (af_ops->get_stats_af_size) {
4427                                 size += nla_total_size(
4428                                         af_ops->get_stats_af_size(dev));
4429
4430                                 /* for AF_* */
4431                                 size += nla_total_size(0);
4432                         }
4433                 }
4434                 rcu_read_unlock();
4435         }
4436
4437         return size;
4438 }
4439
4440 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4441                           struct netlink_ext_ack *extack)
4442 {
4443         struct net *net = sock_net(skb->sk);
4444         struct net_device *dev = NULL;
4445         int idxattr = 0, prividx = 0;
4446         struct if_stats_msg *ifsm;
4447         struct sk_buff *nskb;
4448         u32 filter_mask;
4449         int err;
4450
4451         if (nlmsg_len(nlh) < sizeof(*ifsm))
4452                 return -EINVAL;
4453
4454         ifsm = nlmsg_data(nlh);
4455         if (ifsm->ifindex > 0)
4456                 dev = __dev_get_by_index(net, ifsm->ifindex);
4457         else
4458                 return -EINVAL;
4459
4460         if (!dev)
4461                 return -ENODEV;
4462
4463         filter_mask = ifsm->filter_mask;
4464         if (!filter_mask)
4465                 return -EINVAL;
4466
4467         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4468         if (!nskb)
4469                 return -ENOBUFS;
4470
4471         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4472                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4473                                   0, filter_mask, &idxattr, &prividx);
4474         if (err < 0) {
4475                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4476                 WARN_ON(err == -EMSGSIZE);
4477                 kfree_skb(nskb);
4478         } else {
4479                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4480         }
4481
4482         return err;
4483 }
4484
4485 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4486 {
4487         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4488         struct net *net = sock_net(skb->sk);
4489         unsigned int flags = NLM_F_MULTI;
4490         struct if_stats_msg *ifsm;
4491         struct hlist_head *head;
4492         struct net_device *dev;
4493         u32 filter_mask = 0;
4494         int idx = 0;
4495
4496         s_h = cb->args[0];
4497         s_idx = cb->args[1];
4498         s_idxattr = cb->args[2];
4499         s_prividx = cb->args[3];
4500
4501         cb->seq = net->dev_base_seq;
4502
4503         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4504                 return -EINVAL;
4505
4506         ifsm = nlmsg_data(cb->nlh);
4507         filter_mask = ifsm->filter_mask;
4508         if (!filter_mask)
4509                 return -EINVAL;
4510
4511         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4512                 idx = 0;
4513                 head = &net->dev_index_head[h];
4514                 hlist_for_each_entry(dev, head, index_hlist) {
4515                         if (idx < s_idx)
4516                                 goto cont;
4517                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4518                                                   NETLINK_CB(cb->skb).portid,
4519                                                   cb->nlh->nlmsg_seq, 0,
4520                                                   flags, filter_mask,
4521                                                   &s_idxattr, &s_prividx);
4522                         /* If we ran out of room on the first message,
4523                          * we're in trouble
4524                          */
4525                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4526
4527                         if (err < 0)
4528                                 goto out;
4529                         s_prividx = 0;
4530                         s_idxattr = 0;
4531                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4532 cont:
4533                         idx++;
4534                 }
4535         }
4536 out:
4537         cb->args[3] = s_prividx;
4538         cb->args[2] = s_idxattr;
4539         cb->args[1] = idx;
4540         cb->args[0] = h;
4541
4542         return skb->len;
4543 }
4544
4545 /* Process one rtnetlink message. */
4546
4547 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4548                              struct netlink_ext_ack *extack)
4549 {
4550         struct net *net = sock_net(skb->sk);
4551         struct rtnl_link *link;
4552         struct module *owner;
4553         int err = -EOPNOTSUPP;
4554         rtnl_doit_func doit;
4555         unsigned int flags;
4556         int kind;
4557         int family;
4558         int type;
4559
4560         type = nlh->nlmsg_type;
4561         if (type > RTM_MAX)
4562                 return -EOPNOTSUPP;
4563
4564         type -= RTM_BASE;
4565
4566         /* All the messages must have at least 1 byte length */
4567         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4568                 return 0;
4569
4570         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4571         kind = type&3;
4572
4573         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4574                 return -EPERM;
4575
4576         rcu_read_lock();
4577         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4578                 struct sock *rtnl;
4579                 rtnl_dumpit_func dumpit;
4580                 u16 min_dump_alloc = 0;
4581
4582                 link = rtnl_get_link(family, type);
4583                 if (!link || !link->dumpit) {
4584                         family = PF_UNSPEC;
4585                         link = rtnl_get_link(family, type);
4586                         if (!link || !link->dumpit)
4587                                 goto err_unlock;
4588                 }
4589                 owner = link->owner;
4590                 dumpit = link->dumpit;
4591
4592                 if (type == RTM_GETLINK - RTM_BASE)
4593                         min_dump_alloc = rtnl_calcit(skb, nlh);
4594
4595                 err = 0;
4596                 /* need to do this before rcu_read_unlock() */
4597                 if (!try_module_get(owner))
4598                         err = -EPROTONOSUPPORT;
4599
4600                 rcu_read_unlock();
4601
4602                 rtnl = net->rtnl;
4603                 if (err == 0) {
4604                         struct netlink_dump_control c = {
4605                                 .dump           = dumpit,
4606                                 .min_dump_alloc = min_dump_alloc,
4607                                 .module         = owner,
4608                         };
4609                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4610                         /* netlink_dump_start() will keep a reference on
4611                          * module if dump is still in progress.
4612                          */
4613                         module_put(owner);
4614                 }
4615                 return err;
4616         }
4617
4618         link = rtnl_get_link(family, type);
4619         if (!link || !link->doit) {
4620                 family = PF_UNSPEC;
4621                 link = rtnl_get_link(PF_UNSPEC, type);
4622                 if (!link || !link->doit)
4623                         goto out_unlock;
4624         }
4625
4626         owner = link->owner;
4627         if (!try_module_get(owner)) {
4628                 err = -EPROTONOSUPPORT;
4629                 goto out_unlock;
4630         }
4631
4632         flags = link->flags;
4633         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4634                 doit = link->doit;
4635                 rcu_read_unlock();
4636                 if (doit)
4637                         err = doit(skb, nlh, extack);
4638                 module_put(owner);
4639                 return err;
4640         }
4641         rcu_read_unlock();
4642
4643         rtnl_lock();
4644         link = rtnl_get_link(family, type);
4645         if (link && link->doit)
4646                 err = link->doit(skb, nlh, extack);
4647         rtnl_unlock();
4648
4649         module_put(owner);
4650
4651         return err;
4652
4653 out_unlock:
4654         rcu_read_unlock();
4655         return err;
4656
4657 err_unlock:
4658         rcu_read_unlock();
4659         return -EOPNOTSUPP;
4660 }
4661
4662 static void rtnetlink_rcv(struct sk_buff *skb)
4663 {
4664         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4665 }
4666
4667 static int rtnetlink_bind(struct net *net, int group)
4668 {
4669         switch (group) {
4670         case RTNLGRP_IPV4_MROUTE_R:
4671         case RTNLGRP_IPV6_MROUTE_R:
4672                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4673                         return -EPERM;
4674                 break;
4675         }
4676         return 0;
4677 }
4678
4679 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4680 {
4681         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4682
4683         switch (event) {
4684         case NETDEV_REBOOT:
4685         case NETDEV_CHANGEMTU:
4686         case NETDEV_CHANGEADDR:
4687         case NETDEV_CHANGENAME:
4688         case NETDEV_FEAT_CHANGE:
4689         case NETDEV_BONDING_FAILOVER:
4690         case NETDEV_POST_TYPE_CHANGE:
4691         case NETDEV_NOTIFY_PEERS:
4692         case NETDEV_CHANGEUPPER:
4693         case NETDEV_RESEND_IGMP:
4694         case NETDEV_CHANGEINFODATA:
4695         case NETDEV_CHANGELOWERSTATE:
4696         case NETDEV_CHANGE_TX_QUEUE_LEN:
4697                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4698                                    GFP_KERNEL, NULL, 0);
4699                 break;
4700         default:
4701                 break;
4702         }
4703         return NOTIFY_DONE;
4704 }
4705
4706 static struct notifier_block rtnetlink_dev_notifier = {
4707         .notifier_call  = rtnetlink_event,
4708 };
4709
4710
4711 static int __net_init rtnetlink_net_init(struct net *net)
4712 {
4713         struct sock *sk;
4714         struct netlink_kernel_cfg cfg = {
4715                 .groups         = RTNLGRP_MAX,
4716                 .input          = rtnetlink_rcv,
4717                 .cb_mutex       = &rtnl_mutex,
4718                 .flags          = NL_CFG_F_NONROOT_RECV,
4719                 .bind           = rtnetlink_bind,
4720         };
4721
4722         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4723         if (!sk)
4724                 return -ENOMEM;
4725         net->rtnl = sk;
4726         return 0;
4727 }
4728
4729 static void __net_exit rtnetlink_net_exit(struct net *net)
4730 {
4731         netlink_kernel_release(net->rtnl);
4732         net->rtnl = NULL;
4733 }
4734
4735 static struct pernet_operations rtnetlink_net_ops = {
4736         .init = rtnetlink_net_init,
4737         .exit = rtnetlink_net_exit,
4738 };
4739
4740 void __init rtnetlink_init(void)
4741 {
4742         if (register_pernet_subsys(&rtnetlink_net_ops))
4743                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4744
4745         register_netdevice_notifier(&rtnetlink_dev_notifier);
4746
4747         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4748                       rtnl_dump_ifinfo, 0);
4749         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4750         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4751         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4752
4753         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4754         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4755         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4756
4757         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4758         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4759         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4760
4761         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4762         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4763         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4764
4765         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4766                       0);
4767 }