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