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 #define RTNL_MAX_TYPE           48
  63 #define RTNL_SLAVE_MAX_TYPE     36
  64
  65 struct rtnl_link {
  66         rtnl_doit_func          doit;
  67         rtnl_dumpit_func        dumpit;
  68         struct module           *owner;
  69         unsigned int            flags;
  70         struct rcu_head         rcu;
  71 };
  72
  73 static DEFINE_MUTEX(rtnl_mutex);
  74
  75 void rtnl_lock(void)
  76 {
  77         mutex_lock(&rtnl_mutex);
  78 }
  79 EXPORT_SYMBOL(rtnl_lock);
  80
  81 int rtnl_lock_killable(void)
  82 {
  83         return mutex_lock_killable(&rtnl_mutex);
  84 }
  85 EXPORT_SYMBOL(rtnl_lock_killable);
  86
  87 static struct sk_buff *defer_kfree_skb_list;
  88 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  89 {
  90         if (head && tail) {
  91                 tail->next = defer_kfree_skb_list;
  92                 defer_kfree_skb_list = head;
  93         }
  94 }
  95 EXPORT_SYMBOL(rtnl_kfree_skbs);
  96
  97 void __rtnl_unlock(void)
  98 {
  99         struct sk_buff *head = defer_kfree_skb_list;
 100
 101         defer_kfree_skb_list = NULL;
 102
 103         mutex_unlock(&rtnl_mutex);
 104
 105         while (head) {
 106                 struct sk_buff *next = head->next;
 107
 108                 kfree_skb(head);
 109                 cond_resched();
 110                 head = next;
 111         }
 112 }
 113
 114 void rtnl_unlock(void)
 115 {
 116         /* This fellow will unlock it for us. */
 117         netdev_run_todo();
 118 }
 119 EXPORT_SYMBOL(rtnl_unlock);
 120
 121 int rtnl_trylock(void)
 122 {
 123         return mutex_trylock(&rtnl_mutex);
 124 }
 125 EXPORT_SYMBOL(rtnl_trylock);
 126
 127 int rtnl_is_locked(void)
 128 {
 129         return mutex_is_locked(&rtnl_mutex);
 130 }
 131 EXPORT_SYMBOL(rtnl_is_locked);
 132
 133 #ifdef CONFIG_PROVE_LOCKING
 134 bool lockdep_rtnl_is_held(void)
 135 {
 136         return lockdep_is_held(&rtnl_mutex);
 137 }
 138 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 139 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 140
 141 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 142
 143 static inline int rtm_msgindex(int msgtype)
 144 {
 145         int msgindex = msgtype - RTM_BASE;
 146
 147         /*
 148          * msgindex < 0 implies someone tried to register a netlink
 149          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 150          * the message type has not been added to linux/rtnetlink.h
 151          */
 152         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 153
 154         return msgindex;
 155 }
 156
 157 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 158 {
 159         struct rtnl_link **tab;
 160
 161         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 162                 protocol = PF_UNSPEC;
 163
 164         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 165         if (!tab)
 166                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 167
 168         return tab[msgtype];
 169 }
 170
 171 static int rtnl_register_internal(struct module *owner,
 172                                   int protocol, int msgtype,
 173                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 174                                   unsigned int flags)
 175 {
 176         struct rtnl_link *link, *old;
 177         struct rtnl_link __rcu **tab;
 178         int msgindex;
 179         int ret = -ENOBUFS;
 180
 181         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 182         msgindex = rtm_msgindex(msgtype);
 183
 184         rtnl_lock();
 185         tab = rtnl_msg_handlers[protocol];
 186         if (tab == NULL) {
 187                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 188                 if (!tab)
 189                         goto unlock;
 190
 191                 /* ensures we see the 0 stores */
 192                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 193         }
 194
 195         old = rtnl_dereference(tab[msgindex]);
 196         if (old) {
 197                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 198                 if (!link)
 199                         goto unlock;
 200         } else {
 201                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 202                 if (!link)
 203                         goto unlock;
 204         }
 205
 206         WARN_ON(link->owner && link->owner != owner);
 207         link->owner = owner;
 208
 209         WARN_ON(doit && link->doit && link->doit != doit);
 210         if (doit)
 211                 link->doit = doit;
 212         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 213         if (dumpit)
 214                 link->dumpit = dumpit;
 215
 216         link->flags |= flags;
 217
 218         /* publish protocol:msgtype */
 219         rcu_assign_pointer(tab[msgindex], link);
 220         ret = 0;
 221         if (old)
 222                 kfree_rcu(old, rcu);
 223 unlock:
 224         rtnl_unlock();
 225         return ret;
 226 }
 227
 228 /**
 229  * rtnl_register_module - Register a rtnetlink message type
 230  *
 231  * @owner: module registering the hook (THIS_MODULE)
 232  * @protocol: Protocol family or PF_UNSPEC
 233  * @msgtype: rtnetlink message type
 234  * @doit: Function pointer called for each request message
 235  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 236  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 237  *
 238  * Like rtnl_register, but for use by removable modules.
 239  */
 240 int rtnl_register_module(struct module *owner,
 241                          int protocol, int msgtype,
 242                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 243                          unsigned int flags)
 244 {
 245         return rtnl_register_internal(owner, protocol, msgtype,
 246                                       doit, dumpit, flags);
 247 }
 248 EXPORT_SYMBOL_GPL(rtnl_register_module);
 249
 250 /**
 251  * rtnl_register - Register a rtnetlink message type
 252  * @protocol: Protocol family or PF_UNSPEC
 253  * @msgtype: rtnetlink message type
 254  * @doit: Function pointer called for each request message
 255  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 256  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 257  *
 258  * Registers the specified function pointers (at least one of them has
 259  * to be non-NULL) to be called whenever a request message for the
 260  * specified protocol family and message type is received.
 261  *
 262  * The special protocol family PF_UNSPEC may be used to define fallback
 263  * function pointers for the case when no entry for the specific protocol
 264  * family exists.
 265  */
 266 void rtnl_register(int protocol, int msgtype,
 267                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 268                    unsigned int flags)
 269 {
 270         int err;
 271
 272         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 273                                      flags);
 274         if (err)
 275                 pr_err("Unable to register rtnetlink message handler, "
 276                        "protocol = %d, message type = %d\n", protocol, msgtype);
 277 }
 278
 279 /**
 280  * rtnl_unregister - Unregister a rtnetlink message type
 281  * @protocol: Protocol family or PF_UNSPEC
 282  * @msgtype: rtnetlink message type
 283  *
 284  * Returns 0 on success or a negative error code.
 285  */
 286 int rtnl_unregister(int protocol, int msgtype)
 287 {
 288         struct rtnl_link **tab, *link;
 289         int msgindex;
 290
 291         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 292         msgindex = rtm_msgindex(msgtype);
 293
 294         rtnl_lock();
 295         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 296         if (!tab) {
 297                 rtnl_unlock();
 298                 return -ENOENT;
 299         }
 300
 301         link = tab[msgindex];
 302         rcu_assign_pointer(tab[msgindex], NULL);
 303         rtnl_unlock();
 304
 305         kfree_rcu(link, rcu);
 306
 307         return 0;
 308 }
 309 EXPORT_SYMBOL_GPL(rtnl_unregister);
 310
 311 /**
 312  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 313  * @protocol : Protocol family or PF_UNSPEC
 314  *
 315  * Identical to calling rtnl_unregster() for all registered message types
 316  * of a certain protocol family.
 317  */
 318 void rtnl_unregister_all(int protocol)
 319 {
 320         struct rtnl_link **tab, *link;
 321         int msgindex;
 322
 323         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 324
 325         rtnl_lock();
 326         tab = rtnl_msg_handlers[protocol];
 327         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 328         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 329                 link = tab[msgindex];
 330                 if (!link)
 331                         continue;
 332
 333                 rcu_assign_pointer(tab[msgindex], NULL);
 334                 kfree_rcu(link, rcu);
 335         }
 336         rtnl_unlock();
 337
 338         synchronize_net();
 339
 340         kfree(tab);
 341 }
 342 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 343
 344 static LIST_HEAD(link_ops);
 345
 346 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 347 {
 348         const struct rtnl_link_ops *ops;
 349
 350         list_for_each_entry(ops, &link_ops, list) {
 351                 if (!strcmp(ops->kind, kind))
 352                         return ops;
 353         }
 354         return NULL;
 355 }
 356
 357 /**
 358  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 359  * @ops: struct rtnl_link_ops * to register
 360  *
 361  * The caller must hold the rtnl_mutex. This function should be used
 362  * by drivers that create devices during module initialization. It
 363  * must be called before registering the devices.
 364  *
 365  * Returns 0 on success or a negative error code.
 366  */
 367 int __rtnl_link_register(struct rtnl_link_ops *ops)
 368 {
 369         if (rtnl_link_ops_get(ops->kind))
 370                 return -EEXIST;
 371
 372         /* The check for setup is here because if ops
 373          * does not have that filled up, it is not possible
 374          * to use the ops for creating device. So do not
 375          * fill up dellink as well. That disables rtnl_dellink.
 376          */
 377         if (ops->setup && !ops->dellink)
 378                 ops->dellink = unregister_netdevice_queue;
 379
 380         list_add_tail(&ops->list, &link_ops);
 381         return 0;
 382 }
 383 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 384
 385 /**
 386  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 387  * @ops: struct rtnl_link_ops * to register
 388  *
 389  * Returns 0 on success or a negative error code.
 390  */
 391 int rtnl_link_register(struct rtnl_link_ops *ops)
 392 {
 393         int err;
 394
 395         /* Sanity-check max sizes to avoid stack buffer overflow. */
 396         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 397                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 398                 return -EINVAL;
 399
 400         rtnl_lock();
 401         err = __rtnl_link_register(ops);
 402         rtnl_unlock();
 403         return err;
 404 }
 405 EXPORT_SYMBOL_GPL(rtnl_link_register);
 406
 407 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 408 {
 409         struct net_device *dev;
 410         LIST_HEAD(list_kill);
 411
 412         for_each_netdev(net, dev) {
 413                 if (dev->rtnl_link_ops == ops)
 414                         ops->dellink(dev, &list_kill);
 415         }
 416         unregister_netdevice_many(&list_kill);
 417 }
 418
 419 /**
 420  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 421  * @ops: struct rtnl_link_ops * to unregister
 422  *
 423  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 424  * integrity (hold pernet_ops_rwsem for writing to close the race
 425  * with setup_net() and cleanup_net()).
 426  */
 427 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 428 {
 429         struct net *net;
 430
 431         for_each_net(net) {
 432                 __rtnl_kill_links(net, ops);
 433         }
 434         list_del(&ops->list);
 435 }
 436 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 437
 438 /* Return with the rtnl_lock held when there are no network
 439  * devices unregistering in any network namespace.
 440  */
 441 static void rtnl_lock_unregistering_all(void)
 442 {
 443         struct net *net;
 444         bool unregistering;
 445         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 446
 447         add_wait_queue(&netdev_unregistering_wq, &wait);
 448         for (;;) {
 449                 unregistering = false;
 450                 rtnl_lock();
 451                 /* We held write locked pernet_ops_rwsem, and parallel
 452                  * setup_net() and cleanup_net() are not possible.
 453                  */
 454                 for_each_net(net) {
 455                         if (net->dev_unreg_count > 0) {
 456                                 unregistering = true;
 457                                 break;
 458                         }
 459                 }
 460                 if (!unregistering)
 461                         break;
 462                 __rtnl_unlock();
 463
 464                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 465         }
 466         remove_wait_queue(&netdev_unregistering_wq, &wait);
 467 }
 468
 469 /**
 470  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 471  * @ops: struct rtnl_link_ops * to unregister
 472  */
 473 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 474 {
 475         /* Close the race with setup_net() and cleanup_net() */
 476         down_write(&pernet_ops_rwsem);
 477         rtnl_lock_unregistering_all();
 478         __rtnl_link_unregister(ops);
 479         rtnl_unlock();
 480         up_write(&pernet_ops_rwsem);
 481 }
 482 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 483
 484 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 485 {
 486         struct net_device *master_dev;
 487         const struct rtnl_link_ops *ops;
 488         size_t size = 0;
 489
 490         rcu_read_lock();
 491
 492         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 493         if (!master_dev)
 494                 goto out;
 495
 496         ops = master_dev->rtnl_link_ops;
 497         if (!ops || !ops->get_slave_size)
 498                 goto out;
 499         /* IFLA_INFO_SLAVE_DATA + nested data */
 500         size = nla_total_size(sizeof(struct nlattr)) +
 501                ops->get_slave_size(master_dev, dev);
 502
 503 out:
 504         rcu_read_unlock();
 505         return size;
 506 }
 507
 508 static size_t rtnl_link_get_size(const struct net_device *dev)
 509 {
 510         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 511         size_t size;
 512
 513         if (!ops)
 514                 return 0;
 515
 516         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 517                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 518
 519         if (ops->get_size)
 520                 /* IFLA_INFO_DATA + nested data */
 521                 size += nla_total_size(sizeof(struct nlattr)) +
 522                         ops->get_size(dev);
 523
 524         if (ops->get_xstats_size)
 525                 /* IFLA_INFO_XSTATS */
 526                 size += nla_total_size(ops->get_xstats_size(dev));
 527
 528         size += rtnl_link_get_slave_info_data_size(dev);
 529
 530         return size;
 531 }
 532
 533 static LIST_HEAD(rtnl_af_ops);
 534
 535 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 536 {
 537         const struct rtnl_af_ops *ops;
 538
 539         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 540                 if (ops->family == family)
 541                         return ops;
 542         }
 543
 544         return NULL;
 545 }
 546
 547 /**
 548  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 549  * @ops: struct rtnl_af_ops * to register
 550  *
 551  * Returns 0 on success or a negative error code.
 552  */
 553 void rtnl_af_register(struct rtnl_af_ops *ops)
 554 {
 555         rtnl_lock();
 556         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 557         rtnl_unlock();
 558 }
 559 EXPORT_SYMBOL_GPL(rtnl_af_register);
 560
 561 /**
 562  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 563  * @ops: struct rtnl_af_ops * to unregister
 564  */
 565 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 566 {
 567         rtnl_lock();
 568         list_del_rcu(&ops->list);
 569         rtnl_unlock();
 570
 571         synchronize_rcu();
 572 }
 573 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 574
 575 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 576                                     u32 ext_filter_mask)
 577 {
 578         struct rtnl_af_ops *af_ops;
 579         size_t size;
 580
 581         /* IFLA_AF_SPEC */
 582         size = nla_total_size(sizeof(struct nlattr));
 583
 584         rcu_read_lock();
 585         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 586                 if (af_ops->get_link_af_size) {
 587                         /* AF_* + nested data */
 588                         size += nla_total_size(sizeof(struct nlattr)) +
 589                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 590                 }
 591         }
 592         rcu_read_unlock();
 593
 594         return size;
 595 }
 596
 597 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 598 {
 599         struct net_device *master_dev;
 600         bool ret = false;
 601
 602         rcu_read_lock();
 603
 604         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 605         if (master_dev && master_dev->rtnl_link_ops)
 606                 ret = true;
 607         rcu_read_unlock();
 608         return ret;
 609 }
 610
 611 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 612                                      const struct net_device *dev)
 613 {
 614         struct net_device *master_dev;
 615         const struct rtnl_link_ops *ops;
 616         struct nlattr *slave_data;
 617         int err;
 618
 619         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 620         if (!master_dev)
 621                 return 0;
 622         ops = master_dev->rtnl_link_ops;
 623         if (!ops)
 624                 return 0;
 625         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 626                 return -EMSGSIZE;
 627         if (ops->fill_slave_info) {
 628                 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
 629                 if (!slave_data)
 630                         return -EMSGSIZE;
 631                 err = ops->fill_slave_info(skb, master_dev, dev);
 632                 if (err < 0)
 633                         goto err_cancel_slave_data;
 634                 nla_nest_end(skb, slave_data);
 635         }
 636         return 0;
 637
 638 err_cancel_slave_data:
 639         nla_nest_cancel(skb, slave_data);
 640         return err;
 641 }
 642
 643 static int rtnl_link_info_fill(struct sk_buff *skb,
 644                                const struct net_device *dev)
 645 {
 646         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 647         struct nlattr *data;
 648         int err;
 649
 650         if (!ops)
 651                 return 0;
 652         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 653                 return -EMSGSIZE;
 654         if (ops->fill_xstats) {
 655                 err = ops->fill_xstats(skb, dev);
 656                 if (err < 0)
 657                         return err;
 658         }
 659         if (ops->fill_info) {
 660                 data = nla_nest_start(skb, IFLA_INFO_DATA);
 661                 if (data == NULL)
 662                         return -EMSGSIZE;
 663                 err = ops->fill_info(skb, dev);
 664                 if (err < 0)
 665                         goto err_cancel_data;
 666                 nla_nest_end(skb, data);
 667         }
 668         return 0;
 669
 670 err_cancel_data:
 671         nla_nest_cancel(skb, data);
 672         return err;
 673 }
 674
 675 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 676 {
 677         struct nlattr *linkinfo;
 678         int err = -EMSGSIZE;
 679
 680         linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 681         if (linkinfo == NULL)
 682                 goto out;
 683
 684         err = rtnl_link_info_fill(skb, dev);
 685         if (err < 0)
 686                 goto err_cancel_link;
 687
 688         err = rtnl_link_slave_info_fill(skb, dev);
 689         if (err < 0)
 690                 goto err_cancel_link;
 691
 692         nla_nest_end(skb, linkinfo);
 693         return 0;
 694
 695 err_cancel_link:
 696         nla_nest_cancel(skb, linkinfo);
 697 out:
 698         return err;
 699 }
 700
 701 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 702 {
 703         struct sock *rtnl = net->rtnl;
 704         int err = 0;
 705
 706         NETLINK_CB(skb).dst_group = group;
 707         if (echo)
 708                 refcount_inc(&skb->users);
 709         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 710         if (echo)
 711                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 712         return err;
 713 }
 714
 715 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 716 {
 717         struct sock *rtnl = net->rtnl;
 718
 719         return nlmsg_unicast(rtnl, skb, pid);
 720 }
 721 EXPORT_SYMBOL(rtnl_unicast);
 722
 723 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 724                  struct nlmsghdr *nlh, gfp_t flags)
 725 {
 726         struct sock *rtnl = net->rtnl;
 727         int report = 0;
 728
 729         if (nlh)
 730                 report = nlmsg_report(nlh);
 731
 732         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 733 }
 734 EXPORT_SYMBOL(rtnl_notify);
 735
 736 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 737 {
 738         struct sock *rtnl = net->rtnl;
 739
 740         netlink_set_err(rtnl, 0, group, error);
 741 }
 742 EXPORT_SYMBOL(rtnl_set_sk_err);
 743
 744 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 745 {
 746         struct nlattr *mx;
 747         int i, valid = 0;
 748
 749         mx = nla_nest_start(skb, RTA_METRICS);
 750         if (mx == NULL)
 751                 return -ENOBUFS;
 752
 753         for (i = 0; i < RTAX_MAX; i++) {
 754                 if (metrics[i]) {
 755                         if (i == RTAX_CC_ALGO - 1) {
 756                                 char tmp[TCP_CA_NAME_MAX], *name;
 757
 758                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 759                                 if (!name)
 760                                         continue;
 761                                 if (nla_put_string(skb, i + 1, name))
 762                                         goto nla_put_failure;
 763                         } else if (i == RTAX_FEATURES - 1) {
 764                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 765
 766                                 if (!user_features)
 767                                         continue;
 768                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 769                                 if (nla_put_u32(skb, i + 1, user_features))
 770                                         goto nla_put_failure;
 771                         } else {
 772                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 773                                         goto nla_put_failure;
 774                         }
 775                         valid++;
 776                 }
 777         }
 778
 779         if (!valid) {
 780                 nla_nest_cancel(skb, mx);
 781                 return 0;
 782         }
 783
 784         return nla_nest_end(skb, mx);
 785
 786 nla_put_failure:
 787         nla_nest_cancel(skb, mx);
 788         return -EMSGSIZE;
 789 }
 790 EXPORT_SYMBOL(rtnetlink_put_metrics);
 791
 792 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 793                        long expires, u32 error)
 794 {
 795         struct rta_cacheinfo ci = {
 796                 .rta_error = error,
 797                 .rta_id =  id,
 798         };
 799
 800         if (dst) {
 801                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 802                 ci.rta_used = dst->__use;
 803                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 804         }
 805         if (expires) {
 806                 unsigned long clock;
 807
 808                 clock = jiffies_to_clock_t(abs(expires));
 809                 clock = min_t(unsigned long, clock, INT_MAX);
 810                 ci.rta_expires = (expires > 0) ? clock : -clock;
 811         }
 812         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 813 }
 814 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 815
 816 static void set_operstate(struct net_device *dev, unsigned char transition)
 817 {
 818         unsigned char operstate = dev->operstate;
 819
 820         switch (transition) {
 821         case IF_OPER_UP:
 822                 if ((operstate == IF_OPER_DORMANT ||
 823                      operstate == IF_OPER_UNKNOWN) &&
 824                     !netif_dormant(dev))
 825                         operstate = IF_OPER_UP;
 826                 break;
 827
 828         case IF_OPER_DORMANT:
 829                 if (operstate == IF_OPER_UP ||
 830                     operstate == IF_OPER_UNKNOWN)
 831                         operstate = IF_OPER_DORMANT;
 832                 break;
 833         }
 834
 835         if (dev->operstate != operstate) {
 836                 write_lock_bh(&dev_base_lock);
 837                 dev->operstate = operstate;
 838                 write_unlock_bh(&dev_base_lock);
 839                 netdev_state_change(dev);
 840         }
 841 }
 842
 843 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 844 {
 845         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 846                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 847 }
 848
 849 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 850                                            const struct ifinfomsg *ifm)
 851 {
 852         unsigned int flags = ifm->ifi_flags;
 853
 854         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 855         if (ifm->ifi_change)
 856                 flags = (flags & ifm->ifi_change) |
 857                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 858
 859         return flags;
 860 }
 861
 862 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 863                                  const struct rtnl_link_stats64 *b)
 864 {
 865         a->rx_packets = b->rx_packets;
 866         a->tx_packets = b->tx_packets;
 867         a->rx_bytes = b->rx_bytes;
 868         a->tx_bytes = b->tx_bytes;
 869         a->rx_errors = b->rx_errors;
 870         a->tx_errors = b->tx_errors;
 871         a->rx_dropped = b->rx_dropped;
 872         a->tx_dropped = b->tx_dropped;
 873
 874         a->multicast = b->multicast;
 875         a->collisions = b->collisions;
 876
 877         a->rx_length_errors = b->rx_length_errors;
 878         a->rx_over_errors = b->rx_over_errors;
 879         a->rx_crc_errors = b->rx_crc_errors;
 880         a->rx_frame_errors = b->rx_frame_errors;
 881         a->rx_fifo_errors = b->rx_fifo_errors;
 882         a->rx_missed_errors = b->rx_missed_errors;
 883
 884         a->tx_aborted_errors = b->tx_aborted_errors;
 885         a->tx_carrier_errors = b->tx_carrier_errors;
 886         a->tx_fifo_errors = b->tx_fifo_errors;
 887         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 888         a->tx_window_errors = b->tx_window_errors;
 889
 890         a->rx_compressed = b->rx_compressed;
 891         a->tx_compressed = b->tx_compressed;
 892
 893         a->rx_nohandler = b->rx_nohandler;
 894 }
 895
 896 /* All VF info */
 897 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 898                                    u32 ext_filter_mask)
 899 {
 900         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 901                 int num_vfs = dev_num_vf(dev->dev.parent);
 902                 size_t size = nla_total_size(0);
 903                 size += num_vfs *
 904                         (nla_total_size(0) +
 905                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 906                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 907                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 908                          nla_total_size(MAX_VLAN_LIST_LEN *
 909                                         sizeof(struct ifla_vf_vlan_info)) +
 910                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 911                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 912                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 913                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 914                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 915                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 916                          /* IFLA_VF_STATS_RX_PACKETS */
 917                          nla_total_size_64bit(sizeof(__u64)) +
 918                          /* IFLA_VF_STATS_TX_PACKETS */
 919                          nla_total_size_64bit(sizeof(__u64)) +
 920                          /* IFLA_VF_STATS_RX_BYTES */
 921                          nla_total_size_64bit(sizeof(__u64)) +
 922                          /* IFLA_VF_STATS_TX_BYTES */
 923                          nla_total_size_64bit(sizeof(__u64)) +
 924                          /* IFLA_VF_STATS_BROADCAST */
 925                          nla_total_size_64bit(sizeof(__u64)) +
 926                          /* IFLA_VF_STATS_MULTICAST */
 927                          nla_total_size_64bit(sizeof(__u64)) +
 928                          /* IFLA_VF_STATS_RX_DROPPED */
 929                          nla_total_size_64bit(sizeof(__u64)) +
 930                          /* IFLA_VF_STATS_TX_DROPPED */
 931                          nla_total_size_64bit(sizeof(__u64)) +
 932                          nla_total_size(sizeof(struct ifla_vf_trust)));
 933                 return size;
 934         } else
 935                 return 0;
 936 }
 937
 938 static size_t rtnl_port_size(const struct net_device *dev,
 939                              u32 ext_filter_mask)
 940 {
 941         size_t port_size = nla_total_size(4)            /* PORT_VF */
 942                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 943                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 944                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 945                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 946                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 947         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 948         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 949                 + port_size;
 950         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 951                 + port_size;
 952
 953         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 954             !(ext_filter_mask & RTEXT_FILTER_VF))
 955                 return 0;
 956         if (dev_num_vf(dev->dev.parent))
 957                 return port_self_size + vf_ports_size +
 958                         vf_port_size * dev_num_vf(dev->dev.parent);
 959         else
 960                 return port_self_size;
 961 }
 962
 963 static size_t rtnl_xdp_size(void)
 964 {
 965         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 966                           nla_total_size(1) +   /* XDP_ATTACHED */
 967                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 968                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 969
 970         return xdp_size;
 971 }
 972
 973 static noinline size_t if_nlmsg_size(const struct net_device *dev,
 974                                      u32 ext_filter_mask)
 975 {
 976         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 977                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 978                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 979                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 980                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
 981                + nla_total_size(sizeof(struct rtnl_link_stats))
 982                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
 983                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 984                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 985                + nla_total_size(4) /* IFLA_TXQLEN */
 986                + nla_total_size(4) /* IFLA_WEIGHT */
 987                + nla_total_size(4) /* IFLA_MTU */
 988                + nla_total_size(4) /* IFLA_LINK */
 989                + nla_total_size(4) /* IFLA_MASTER */
 990                + nla_total_size(1) /* IFLA_CARRIER */
 991                + nla_total_size(4) /* IFLA_PROMISCUITY */
 992                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
 993                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
 994                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
 995                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
 996                + nla_total_size(1) /* IFLA_OPERSTATE */
 997                + nla_total_size(1) /* IFLA_LINKMODE */
 998                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
 999                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1000                + nla_total_size(4) /* IFLA_GROUP */
1001                + nla_total_size(ext_filter_mask
1002                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1003                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1004                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1005                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1006                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1007                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1008                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1009                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1010                + rtnl_xdp_size() /* IFLA_XDP */
1011                + nla_total_size(4)  /* IFLA_EVENT */
1012                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1013                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1014                + nla_total_size(1)  /* IFLA_PROTO_DOWN */
1015                + nla_total_size(4)  /* IFLA_IF_NETNSID */
1016                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1017                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1018                + nla_total_size(4)  /* IFLA_MIN_MTU */
1019                + nla_total_size(4)  /* IFLA_MAX_MTU */
1020                + 0;
1021 }
1022
1023 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1024 {
1025         struct nlattr *vf_ports;
1026         struct nlattr *vf_port;
1027         int vf;
1028         int err;
1029
1030         vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
1031         if (!vf_ports)
1032                 return -EMSGSIZE;
1033
1034         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1035                 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
1036                 if (!vf_port)
1037                         goto nla_put_failure;
1038                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1039                         goto nla_put_failure;
1040                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1041                 if (err == -EMSGSIZE)
1042                         goto nla_put_failure;
1043                 if (err) {
1044                         nla_nest_cancel(skb, vf_port);
1045                         continue;
1046                 }
1047                 nla_nest_end(skb, vf_port);
1048         }
1049
1050         nla_nest_end(skb, vf_ports);
1051
1052         return 0;
1053
1054 nla_put_failure:
1055         nla_nest_cancel(skb, vf_ports);
1056         return -EMSGSIZE;
1057 }
1058
1059 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1060 {
1061         struct nlattr *port_self;
1062         int err;
1063
1064         port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1065         if (!port_self)
1066                 return -EMSGSIZE;
1067
1068         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1069         if (err) {
1070                 nla_nest_cancel(skb, port_self);
1071                 return (err == -EMSGSIZE) ? err : 0;
1072         }
1073
1074         nla_nest_end(skb, port_self);
1075
1076         return 0;
1077 }
1078
1079 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1080                           u32 ext_filter_mask)
1081 {
1082         int err;
1083
1084         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1085             !(ext_filter_mask & RTEXT_FILTER_VF))
1086                 return 0;
1087
1088         err = rtnl_port_self_fill(skb, dev);
1089         if (err)
1090                 return err;
1091
1092         if (dev_num_vf(dev->dev.parent)) {
1093                 err = rtnl_vf_ports_fill(skb, dev);
1094                 if (err)
1095                         return err;
1096         }
1097
1098         return 0;
1099 }
1100
1101 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1102 {
1103         int err;
1104         struct netdev_phys_item_id ppid;
1105
1106         err = dev_get_phys_port_id(dev, &ppid);
1107         if (err) {
1108                 if (err == -EOPNOTSUPP)
1109                         return 0;
1110                 return err;
1111         }
1112
1113         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1114                 return -EMSGSIZE;
1115
1116         return 0;
1117 }
1118
1119 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1120 {
1121         char name[IFNAMSIZ];
1122         int err;
1123
1124         err = dev_get_phys_port_name(dev, name, sizeof(name));
1125         if (err) {
1126                 if (err == -EOPNOTSUPP)
1127                         return 0;
1128                 return err;
1129         }
1130
1131         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1132                 return -EMSGSIZE;
1133
1134         return 0;
1135 }
1136
1137 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1138 {
1139         int err;
1140         struct switchdev_attr attr = {
1141                 .orig_dev = dev,
1142                 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1143                 .flags = SWITCHDEV_F_NO_RECURSE,
1144         };
1145
1146         err = switchdev_port_attr_get(dev, &attr);
1147         if (err) {
1148                 if (err == -EOPNOTSUPP)
1149                         return 0;
1150                 return err;
1151         }
1152
1153         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1154                     attr.u.ppid.id))
1155                 return -EMSGSIZE;
1156
1157         return 0;
1158 }
1159
1160 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1161                                               struct net_device *dev)
1162 {
1163         struct rtnl_link_stats64 *sp;
1164         struct nlattr *attr;
1165
1166         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1167                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1168         if (!attr)
1169                 return -EMSGSIZE;
1170
1171         sp = nla_data(attr);
1172         dev_get_stats(dev, sp);
1173
1174         attr = nla_reserve(skb, IFLA_STATS,
1175                            sizeof(struct rtnl_link_stats));
1176         if (!attr)
1177                 return -EMSGSIZE;
1178
1179         copy_rtnl_link_stats(nla_data(attr), sp);
1180
1181         return 0;
1182 }
1183
1184 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1185                                                struct net_device *dev,
1186                                                int vfs_num,
1187                                                struct nlattr *vfinfo)
1188 {
1189         struct ifla_vf_rss_query_en vf_rss_query_en;
1190         struct nlattr *vf, *vfstats, *vfvlanlist;
1191         struct ifla_vf_link_state vf_linkstate;
1192         struct ifla_vf_vlan_info vf_vlan_info;
1193         struct ifla_vf_spoofchk vf_spoofchk;
1194         struct ifla_vf_tx_rate vf_tx_rate;
1195         struct ifla_vf_stats vf_stats;
1196         struct ifla_vf_trust vf_trust;
1197         struct ifla_vf_vlan vf_vlan;
1198         struct ifla_vf_rate vf_rate;
1199         struct ifla_vf_mac vf_mac;
1200         struct ifla_vf_info ivi;
1201
1202         memset(&ivi, 0, sizeof(ivi));
1203
1204         /* Not all SR-IOV capable drivers support the
1205          * spoofcheck and "RSS query enable" query.  Preset to
1206          * -1 so the user space tool can detect that the driver
1207          * didn't report anything.
1208          */
1209         ivi.spoofchk = -1;
1210         ivi.rss_query_en = -1;
1211         ivi.trusted = -1;
1212         /* The default value for VF link state is "auto"
1213          * IFLA_VF_LINK_STATE_AUTO which equals zero
1214          */
1215         ivi.linkstate = 0;
1216         /* VLAN Protocol by default is 802.1Q */
1217         ivi.vlan_proto = htons(ETH_P_8021Q);
1218         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1219                 return 0;
1220
1221         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1222
1223         vf_mac.vf =
1224                 vf_vlan.vf =
1225                 vf_vlan_info.vf =
1226                 vf_rate.vf =
1227                 vf_tx_rate.vf =
1228                 vf_spoofchk.vf =
1229                 vf_linkstate.vf =
1230                 vf_rss_query_en.vf =
1231                 vf_trust.vf = ivi.vf;
1232
1233         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1234         vf_vlan.vlan = ivi.vlan;
1235         vf_vlan.qos = ivi.qos;
1236         vf_vlan_info.vlan = ivi.vlan;
1237         vf_vlan_info.qos = ivi.qos;
1238         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1239         vf_tx_rate.rate = ivi.max_tx_rate;
1240         vf_rate.min_tx_rate = ivi.min_tx_rate;
1241         vf_rate.max_tx_rate = ivi.max_tx_rate;
1242         vf_spoofchk.setting = ivi.spoofchk;
1243         vf_linkstate.link_state = ivi.linkstate;
1244         vf_rss_query_en.setting = ivi.rss_query_en;
1245         vf_trust.setting = ivi.trusted;
1246         vf = nla_nest_start(skb, IFLA_VF_INFO);
1247         if (!vf)
1248                 goto nla_put_vfinfo_failure;
1249         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1250             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1251             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1252                     &vf_rate) ||
1253             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1254                     &vf_tx_rate) ||
1255             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1256                     &vf_spoofchk) ||
1257             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1258                     &vf_linkstate) ||
1259             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1260                     sizeof(vf_rss_query_en),
1261                     &vf_rss_query_en) ||
1262             nla_put(skb, IFLA_VF_TRUST,
1263                     sizeof(vf_trust), &vf_trust))
1264                 goto nla_put_vf_failure;
1265         vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1266         if (!vfvlanlist)
1267                 goto nla_put_vf_failure;
1268         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1269                     &vf_vlan_info)) {
1270                 nla_nest_cancel(skb, vfvlanlist);
1271                 goto nla_put_vf_failure;
1272         }
1273         nla_nest_end(skb, vfvlanlist);
1274         memset(&vf_stats, 0, sizeof(vf_stats));
1275         if (dev->netdev_ops->ndo_get_vf_stats)
1276                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1277                                                 &vf_stats);
1278         vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1279         if (!vfstats)
1280                 goto nla_put_vf_failure;
1281         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1282                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1283             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1284                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1285             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1286                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1287             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1288                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1289             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1290                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1291             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1292                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1293             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1294                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1295             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1296                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1297                 nla_nest_cancel(skb, vfstats);
1298                 goto nla_put_vf_failure;
1299         }
1300         nla_nest_end(skb, vfstats);
1301         nla_nest_end(skb, vf);
1302         return 0;
1303
1304 nla_put_vf_failure:
1305         nla_nest_cancel(skb, vf);
1306 nla_put_vfinfo_failure:
1307         nla_nest_cancel(skb, vfinfo);
1308         return -EMSGSIZE;
1309 }
1310
1311 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1312                                            struct net_device *dev,
1313                                            u32 ext_filter_mask)
1314 {
1315         struct nlattr *vfinfo;
1316         int i, num_vfs;
1317
1318         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1319                 return 0;
1320
1321         num_vfs = dev_num_vf(dev->dev.parent);
1322         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1323                 return -EMSGSIZE;
1324
1325         if (!dev->netdev_ops->ndo_get_vf_config)
1326                 return 0;
1327
1328         vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1329         if (!vfinfo)
1330                 return -EMSGSIZE;
1331
1332         for (i = 0; i < num_vfs; i++) {
1333                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1334                         return -EMSGSIZE;
1335         }
1336
1337         nla_nest_end(skb, vfinfo);
1338         return 0;
1339 }
1340
1341 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1342 {
1343         struct rtnl_link_ifmap map;
1344
1345         memset(&map, 0, sizeof(map));
1346         map.mem_start   = dev->mem_start;
1347         map.mem_end     = dev->mem_end;
1348         map.base_addr   = dev->base_addr;
1349         map.irq         = dev->irq;
1350         map.dma         = dev->dma;
1351         map.port        = dev->if_port;
1352
1353         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1354                 return -EMSGSIZE;
1355
1356         return 0;
1357 }
1358
1359 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1360 {
1361         const struct bpf_prog *generic_xdp_prog;
1362
1363         ASSERT_RTNL();
1364
1365         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1366         if (!generic_xdp_prog)
1367                 return 0;
1368         return generic_xdp_prog->aux->id;
1369 }
1370
1371 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1372 {
1373         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf, XDP_QUERY_PROG);
1374 }
1375
1376 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1377 {
1378         return __dev_xdp_query(dev, dev->netdev_ops->ndo_bpf,
1379                                XDP_QUERY_PROG_HW);
1380 }
1381
1382 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1383                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1384                                u32 (*get_prog_id)(struct net_device *dev))
1385 {
1386         u32 curr_id;
1387         int err;
1388
1389         curr_id = get_prog_id(dev);
1390         if (!curr_id)
1391                 return 0;
1392
1393         *prog_id = curr_id;
1394         err = nla_put_u32(skb, attr, curr_id);
1395         if (err)
1396                 return err;
1397
1398         if (*mode != XDP_ATTACHED_NONE)
1399                 *mode = XDP_ATTACHED_MULTI;
1400         else
1401                 *mode = tgt_mode;
1402
1403         return 0;
1404 }
1405
1406 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1407 {
1408         struct nlattr *xdp;
1409         u32 prog_id;
1410         int err;
1411         u8 mode;
1412
1413         xdp = nla_nest_start(skb, IFLA_XDP);
1414         if (!xdp)
1415                 return -EMSGSIZE;
1416
1417         prog_id = 0;
1418         mode = XDP_ATTACHED_NONE;
1419         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1420                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1421         if (err)
1422                 goto err_cancel;
1423         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1424                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1425         if (err)
1426                 goto err_cancel;
1427         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1428                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1429         if (err)
1430                 goto err_cancel;
1431
1432         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1433         if (err)
1434                 goto err_cancel;
1435
1436         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1437                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1438                 if (err)
1439                         goto err_cancel;
1440         }
1441
1442         nla_nest_end(skb, xdp);
1443         return 0;
1444
1445 err_cancel:
1446         nla_nest_cancel(skb, xdp);
1447         return err;
1448 }
1449
1450 static u32 rtnl_get_event(unsigned long event)
1451 {
1452         u32 rtnl_event_type = IFLA_EVENT_NONE;
1453
1454         switch (event) {
1455         case NETDEV_REBOOT:
1456                 rtnl_event_type = IFLA_EVENT_REBOOT;
1457                 break;
1458         case NETDEV_FEAT_CHANGE:
1459                 rtnl_event_type = IFLA_EVENT_FEATURES;
1460                 break;
1461         case NETDEV_BONDING_FAILOVER:
1462                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1463                 break;
1464         case NETDEV_NOTIFY_PEERS:
1465                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1466                 break;
1467         case NETDEV_RESEND_IGMP:
1468                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1469                 break;
1470         case NETDEV_CHANGEINFODATA:
1471                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1472                 break;
1473         default:
1474                 break;
1475         }
1476
1477         return rtnl_event_type;
1478 }
1479
1480 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1481 {
1482         const struct net_device *upper_dev;
1483         int ret = 0;
1484
1485         rcu_read_lock();
1486
1487         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1488         if (upper_dev)
1489                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1490
1491         rcu_read_unlock();
1492         return ret;
1493 }
1494
1495 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1496 {
1497         int ifindex = dev_get_iflink(dev);
1498
1499         if (dev->ifindex == ifindex)
1500                 return 0;
1501
1502         return nla_put_u32(skb, IFLA_LINK, ifindex);
1503 }
1504
1505 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1506                                               struct net_device *dev)
1507 {
1508         char buf[IFALIASZ];
1509         int ret;
1510
1511         ret = dev_get_alias(dev, buf, sizeof(buf));
1512         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1513 }
1514
1515 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1516                                   const struct net_device *dev,
1517                                   struct net *src_net)
1518 {
1519         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1520                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1521
1522                 if (!net_eq(dev_net(dev), link_net)) {
1523                         int id = peernet2id_alloc(src_net, link_net);
1524
1525                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1526                                 return -EMSGSIZE;
1527                 }
1528         }
1529
1530         return 0;
1531 }
1532
1533 static int rtnl_fill_link_af(struct sk_buff *skb,
1534                              const struct net_device *dev,
1535                              u32 ext_filter_mask)
1536 {
1537         const struct rtnl_af_ops *af_ops;
1538         struct nlattr *af_spec;
1539
1540         af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1541         if (!af_spec)
1542                 return -EMSGSIZE;
1543
1544         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1545                 struct nlattr *af;
1546                 int err;
1547
1548                 if (!af_ops->fill_link_af)
1549                         continue;
1550
1551                 af = nla_nest_start(skb, af_ops->family);
1552                 if (!af)
1553                         return -EMSGSIZE;
1554
1555                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1556                 /*
1557                  * Caller may return ENODATA to indicate that there
1558                  * was no data to be dumped. This is not an error, it
1559                  * means we should trim the attribute header and
1560                  * continue.
1561                  */
1562                 if (err == -ENODATA)
1563                         nla_nest_cancel(skb, af);
1564                 else if (err < 0)
1565                         return -EMSGSIZE;
1566
1567                 nla_nest_end(skb, af);
1568         }
1569
1570         nla_nest_end(skb, af_spec);
1571         return 0;
1572 }
1573
1574 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1575                             struct net_device *dev, struct net *src_net,
1576                             int type, u32 pid, u32 seq, u32 change,
1577                             unsigned int flags, u32 ext_filter_mask,
1578                             u32 event, int *new_nsid, int new_ifindex,
1579                             int tgt_netnsid)
1580 {
1581         struct ifinfomsg *ifm;
1582         struct nlmsghdr *nlh;
1583
1584         ASSERT_RTNL();
1585         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1586         if (nlh == NULL)
1587                 return -EMSGSIZE;
1588
1589         ifm = nlmsg_data(nlh);
1590         ifm->ifi_family = AF_UNSPEC;
1591         ifm->__ifi_pad = 0;
1592         ifm->ifi_type = dev->type;
1593         ifm->ifi_index = dev->ifindex;
1594         ifm->ifi_flags = dev_get_flags(dev);
1595         ifm->ifi_change = change;
1596
1597         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1598                 goto nla_put_failure;
1599
1600         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1601             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1602             nla_put_u8(skb, IFLA_OPERSTATE,
1603                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1604             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1605             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1606             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1607             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1608             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1609             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1610             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1611             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1612             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1613 #ifdef CONFIG_RPS
1614             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1615 #endif
1616             nla_put_iflink(skb, dev) ||
1617             put_master_ifindex(skb, dev) ||
1618             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1619             (dev->qdisc &&
1620              nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1621             nla_put_ifalias(skb, dev) ||
1622             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1623                         atomic_read(&dev->carrier_up_count) +
1624                         atomic_read(&dev->carrier_down_count)) ||
1625             nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1626             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1627                         atomic_read(&dev->carrier_up_count)) ||
1628             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1629                         atomic_read(&dev->carrier_down_count)))
1630                 goto nla_put_failure;
1631
1632         if (event != IFLA_EVENT_NONE) {
1633                 if (nla_put_u32(skb, IFLA_EVENT, event))
1634                         goto nla_put_failure;
1635         }
1636
1637         if (rtnl_fill_link_ifmap(skb, dev))
1638                 goto nla_put_failure;
1639
1640         if (dev->addr_len) {
1641                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1642                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1643                         goto nla_put_failure;
1644         }
1645
1646         if (rtnl_phys_port_id_fill(skb, dev))
1647                 goto nla_put_failure;
1648
1649         if (rtnl_phys_port_name_fill(skb, dev))
1650                 goto nla_put_failure;
1651
1652         if (rtnl_phys_switch_id_fill(skb, dev))
1653                 goto nla_put_failure;
1654
1655         if (rtnl_fill_stats(skb, dev))
1656                 goto nla_put_failure;
1657
1658         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1659                 goto nla_put_failure;
1660
1661         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1662                 goto nla_put_failure;
1663
1664         if (rtnl_xdp_fill(skb, dev))
1665                 goto nla_put_failure;
1666
1667         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1668                 if (rtnl_link_fill(skb, dev) < 0)
1669                         goto nla_put_failure;
1670         }
1671
1672         if (rtnl_fill_link_netnsid(skb, dev, src_net))
1673                 goto nla_put_failure;
1674
1675         if (new_nsid &&
1676             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1677                 goto nla_put_failure;
1678         if (new_ifindex &&
1679             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1680                 goto nla_put_failure;
1681
1682
1683         rcu_read_lock();
1684         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1685                 goto nla_put_failure_rcu;
1686         rcu_read_unlock();
1687
1688         nlmsg_end(skb, nlh);
1689         return 0;
1690
1691 nla_put_failure_rcu:
1692         rcu_read_unlock();
1693 nla_put_failure:
1694         nlmsg_cancel(skb, nlh);
1695         return -EMSGSIZE;
1696 }
1697
1698 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1699         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1700         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1701         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1702         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1703         [IFLA_MTU]              = { .type = NLA_U32 },
1704         [IFLA_LINK]             = { .type = NLA_U32 },
1705         [IFLA_MASTER]           = { .type = NLA_U32 },
1706         [IFLA_CARRIER]          = { .type = NLA_U8 },
1707         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1708         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1709         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1710         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1711         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1712         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1713         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1714         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1715          * allow 0-length string (needed to remove an alias).
1716          */
1717         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1718         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1719         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1720         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1721         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1722         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1723         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1724         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1725         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1726         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1727         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1728         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1729         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1730         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1731         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1732         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1733         [IFLA_XDP]              = { .type = NLA_NESTED },
1734         [IFLA_EVENT]            = { .type = NLA_U32 },
1735         [IFLA_GROUP]            = { .type = NLA_U32 },
1736         [IFLA_IF_NETNSID]       = { .type = NLA_S32 },
1737         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1738         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1739         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1740         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1741 };
1742
1743 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1744         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1745         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1746         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1747         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1748 };
1749
1750 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1751         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1752         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1753         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1754         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1755         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1756         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1757         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1758         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1759         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1760         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1761         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1762         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1763 };
1764
1765 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1766         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1767         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1768                                     .len = PORT_PROFILE_MAX },
1769         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1770                                       .len = PORT_UUID_MAX },
1771         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1772                                     .len = PORT_UUID_MAX },
1773         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1774         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1775
1776         /* Unused, but we need to keep it here since user space could
1777          * fill it. It's also broken with regard to NLA_BINARY use in
1778          * combination with structs.
1779          */
1780         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1781                                     .len = sizeof(struct ifla_port_vsi) },
1782 };
1783
1784 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1785         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1786         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1787         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1788         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1789 };
1790
1791 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1792 {
1793         const struct rtnl_link_ops *ops = NULL;
1794         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1795
1796         if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1797                              ifla_info_policy, NULL) < 0)
1798                 return NULL;
1799
1800         if (linfo[IFLA_INFO_KIND]) {
1801                 char kind[MODULE_NAME_LEN];
1802
1803                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1804                 ops = rtnl_link_ops_get(kind);
1805         }
1806
1807         return ops;
1808 }
1809
1810 static bool link_master_filtered(struct net_device *dev, int master_idx)
1811 {
1812         struct net_device *master;
1813
1814         if (!master_idx)
1815                 return false;
1816
1817         master = netdev_master_upper_dev_get(dev);
1818         if (!master || master->ifindex != master_idx)
1819                 return true;
1820
1821         return false;
1822 }
1823
1824 static bool link_kind_filtered(const struct net_device *dev,
1825                                const struct rtnl_link_ops *kind_ops)
1826 {
1827         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1828                 return true;
1829
1830         return false;
1831 }
1832
1833 static bool link_dump_filtered(struct net_device *dev,
1834                                int master_idx,
1835                                const struct rtnl_link_ops *kind_ops)
1836 {
1837         if (link_master_filtered(dev, master_idx) ||
1838             link_kind_filtered(dev, kind_ops))
1839                 return true;
1840
1841         return false;
1842 }
1843
1844 static struct net *get_target_net(struct sock *sk, int netnsid)
1845 {
1846         struct net *net;
1847
1848         net = get_net_ns_by_id(sock_net(sk), netnsid);
1849         if (!net)
1850                 return ERR_PTR(-EINVAL);
1851
1852         /* For now, the caller is required to have CAP_NET_ADMIN in
1853          * the user namespace owning the target net ns.
1854          */
1855         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1856                 put_net(net);
1857                 return ERR_PTR(-EACCES);
1858         }
1859         return net;
1860 }
1861
1862 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1863 {
1864         struct net *net = sock_net(skb->sk);
1865         struct net *tgt_net = net;
1866         int h, s_h;
1867         int idx = 0, s_idx;
1868         struct net_device *dev;
1869         struct hlist_head *head;
1870         struct nlattr *tb[IFLA_MAX+1];
1871         u32 ext_filter_mask = 0;
1872         const struct rtnl_link_ops *kind_ops = NULL;
1873         unsigned int flags = NLM_F_MULTI;
1874         int master_idx = 0;
1875         int netnsid = -1;
1876         int err;
1877         int hdrlen;
1878
1879         s_h = cb->args[0];
1880         s_idx = cb->args[1];
1881
1882         /* A hack to preserve kernel<->userspace interface.
1883          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1884          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1885          * what iproute2 < v3.9.0 used.
1886          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1887          * attribute, its netlink message is shorter than struct ifinfomsg.
1888          */
1889         hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1890                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1891
1892         if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1893                         ifla_policy, NULL) >= 0) {
1894                 if (tb[IFLA_IF_NETNSID]) {
1895                         netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1896                         tgt_net = get_target_net(skb->sk, netnsid);
1897                         if (IS_ERR(tgt_net)) {
1898                                 tgt_net = net;
1899                                 netnsid = -1;
1900                         }
1901                 }
1902
1903                 if (tb[IFLA_EXT_MASK])
1904                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1905
1906                 if (tb[IFLA_MASTER])
1907                         master_idx = nla_get_u32(tb[IFLA_MASTER]);
1908
1909                 if (tb[IFLA_LINKINFO])
1910                         kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1911
1912                 if (master_idx || kind_ops)
1913                         flags |= NLM_F_DUMP_FILTERED;
1914         }
1915
1916         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1917                 idx = 0;
1918                 head = &tgt_net->dev_index_head[h];
1919                 hlist_for_each_entry(dev, head, index_hlist) {
1920                         if (link_dump_filtered(dev, master_idx, kind_ops))
1921                                 goto cont;
1922                         if (idx < s_idx)
1923                                 goto cont;
1924                         err = rtnl_fill_ifinfo(skb, dev, net,
1925                                                RTM_NEWLINK,
1926                                                NETLINK_CB(cb->skb).portid,
1927                                                cb->nlh->nlmsg_seq, 0,
1928                                                flags,
1929                                                ext_filter_mask, 0, NULL, 0,
1930                                                netnsid);
1931
1932                         if (err < 0) {
1933                                 if (likely(skb->len))
1934                                         goto out;
1935
1936                                 goto out_err;
1937                         }
1938 cont:
1939                         idx++;
1940                 }
1941         }
1942 out:
1943         err = skb->len;
1944 out_err:
1945         cb->args[1] = idx;
1946         cb->args[0] = h;
1947         cb->seq = net->dev_base_seq;
1948         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1949         if (netnsid >= 0)
1950                 put_net(tgt_net);
1951
1952         return err;
1953 }
1954
1955 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1956                         struct netlink_ext_ack *exterr)
1957 {
1958         return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1959 }
1960 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1961
1962 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1963 {
1964         struct net *net;
1965         /* Examine the link attributes and figure out which
1966          * network namespace we are talking about.
1967          */
1968         if (tb[IFLA_NET_NS_PID])
1969                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1970         else if (tb[IFLA_NET_NS_FD])
1971                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1972         else
1973                 net = get_net(src_net);
1974         return net;
1975 }
1976 EXPORT_SYMBOL(rtnl_link_get_net);
1977
1978 /* Figure out which network namespace we are talking about by
1979  * examining the link attributes in the following order:
1980  *
1981  * 1. IFLA_NET_NS_PID
1982  * 2. IFLA_NET_NS_FD
1983  * 3. IFLA_IF_NETNSID
1984  */
1985 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1986                                                struct nlattr *tb[])
1987 {
1988         struct net *net;
1989
1990         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1991                 return rtnl_link_get_net(src_net, tb);
1992
1993         if (!tb[IFLA_IF_NETNSID])
1994                 return get_net(src_net);
1995
1996         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1997         if (!net)
1998                 return ERR_PTR(-EINVAL);
1999
2000         return net;
2001 }
2002
2003 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2004                                              struct net *src_net,
2005                                              struct nlattr *tb[], int cap)
2006 {
2007         struct net *net;
2008
2009         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2010         if (IS_ERR(net))
2011                 return net;
2012
2013         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2014                 put_net(net);
2015                 return ERR_PTR(-EPERM);
2016         }
2017
2018         return net;
2019 }
2020
2021 /* Verify that rtnetlink requests do not pass additional properties
2022  * potentially referring to different network namespaces.
2023  */
2024 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2025                                     struct netlink_ext_ack *extack,
2026                                     bool netns_id_only)
2027 {
2028
2029         if (netns_id_only) {
2030                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2031                         return 0;
2032
2033                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2034                 return -EOPNOTSUPP;
2035         }
2036
2037         if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2038                 goto invalid_attr;
2039
2040         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
2041                 goto invalid_attr;
2042
2043         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
2044                 goto invalid_attr;
2045
2046         return 0;
2047
2048 invalid_attr:
2049         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2050         return -EINVAL;
2051 }
2052
2053 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2054 {
2055         if (dev) {
2056                 if (tb[IFLA_ADDRESS] &&
2057                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2058                         return -EINVAL;
2059
2060                 if (tb[IFLA_BROADCAST] &&
2061                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2062                         return -EINVAL;
2063         }
2064
2065         if (tb[IFLA_AF_SPEC]) {
2066                 struct nlattr *af;
2067                 int rem, err;
2068
2069                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2070                         const struct rtnl_af_ops *af_ops;
2071
2072                         rcu_read_lock();
2073                         af_ops = rtnl_af_lookup(nla_type(af));
2074                         if (!af_ops) {
2075                                 rcu_read_unlock();
2076                                 return -EAFNOSUPPORT;
2077                         }
2078
2079                         if (!af_ops->set_link_af) {
2080                                 rcu_read_unlock();
2081                                 return -EOPNOTSUPP;
2082                         }
2083
2084                         if (af_ops->validate_link_af) {
2085                                 err = af_ops->validate_link_af(dev, af);
2086                                 if (err < 0) {
2087                                         rcu_read_unlock();
2088                                         return err;
2089                                 }
2090                         }
2091
2092                         rcu_read_unlock();
2093                 }
2094         }
2095
2096         return 0;
2097 }
2098
2099 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2100                                   int guid_type)
2101 {
2102         const struct net_device_ops *ops = dev->netdev_ops;
2103
2104         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2105 }
2106
2107 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2108 {
2109         if (dev->type != ARPHRD_INFINIBAND)
2110                 return -EOPNOTSUPP;
2111
2112         return handle_infiniband_guid(dev, ivt, guid_type);
2113 }
2114
2115 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2116 {
2117         const struct net_device_ops *ops = dev->netdev_ops;
2118         int err = -EINVAL;
2119
2120         if (tb[IFLA_VF_MAC]) {
2121                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2122
2123                 err = -EOPNOTSUPP;
2124                 if (ops->ndo_set_vf_mac)
2125                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2126                                                   ivm->mac);
2127                 if (err < 0)
2128                         return err;
2129         }
2130
2131         if (tb[IFLA_VF_VLAN]) {
2132                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2133
2134                 err = -EOPNOTSUPP;
2135                 if (ops->ndo_set_vf_vlan)
2136                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2137                                                    ivv->qos,
2138                                                    htons(ETH_P_8021Q));
2139                 if (err < 0)
2140                         return err;
2141         }
2142
2143         if (tb[IFLA_VF_VLAN_LIST]) {
2144                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2145                 struct nlattr *attr;
2146                 int rem, len = 0;
2147
2148                 err = -EOPNOTSUPP;
2149                 if (!ops->ndo_set_vf_vlan)
2150                         return err;
2151
2152                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2153                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2154                             nla_len(attr) < NLA_HDRLEN) {
2155                                 return -EINVAL;
2156                         }
2157                         if (len >= MAX_VLAN_LIST_LEN)
2158                                 return -EOPNOTSUPP;
2159                         ivvl[len] = nla_data(attr);
2160
2161                         len++;
2162                 }
2163                 if (len == 0)
2164                         return -EINVAL;
2165
2166                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2167                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2168                 if (err < 0)
2169                         return err;
2170         }
2171
2172         if (tb[IFLA_VF_TX_RATE]) {
2173                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2174                 struct ifla_vf_info ivf;
2175
2176                 err = -EOPNOTSUPP;
2177                 if (ops->ndo_get_vf_config)
2178                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2179                 if (err < 0)
2180                         return err;
2181
2182                 err = -EOPNOTSUPP;
2183                 if (ops->ndo_set_vf_rate)
2184                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2185                                                    ivf.min_tx_rate,
2186                                                    ivt->rate);
2187                 if (err < 0)
2188                         return err;
2189         }
2190
2191         if (tb[IFLA_VF_RATE]) {
2192                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2193
2194                 err = -EOPNOTSUPP;
2195                 if (ops->ndo_set_vf_rate)
2196                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2197                                                    ivt->min_tx_rate,
2198                                                    ivt->max_tx_rate);
2199                 if (err < 0)
2200                         return err;
2201         }
2202
2203         if (tb[IFLA_VF_SPOOFCHK]) {
2204                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2205
2206                 err = -EOPNOTSUPP;
2207                 if (ops->ndo_set_vf_spoofchk)
2208                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2209                                                        ivs->setting);
2210                 if (err < 0)
2211                         return err;
2212         }
2213
2214         if (tb[IFLA_VF_LINK_STATE]) {
2215                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2216
2217                 err = -EOPNOTSUPP;
2218                 if (ops->ndo_set_vf_link_state)
2219                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2220                                                          ivl->link_state);
2221                 if (err < 0)
2222                         return err;
2223         }
2224
2225         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2226                 struct ifla_vf_rss_query_en *ivrssq_en;
2227
2228                 err = -EOPNOTSUPP;
2229                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2230                 if (ops->ndo_set_vf_rss_query_en)
2231                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2232                                                            ivrssq_en->setting);
2233                 if (err < 0)
2234                         return err;
2235         }
2236
2237         if (tb[IFLA_VF_TRUST]) {
2238                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2239
2240                 err = -EOPNOTSUPP;
2241                 if (ops->ndo_set_vf_trust)
2242                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2243                 if (err < 0)
2244                         return err;
2245         }
2246
2247         if (tb[IFLA_VF_IB_NODE_GUID]) {
2248                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2249
2250                 if (!ops->ndo_set_vf_guid)
2251                         return -EOPNOTSUPP;
2252
2253                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2254         }
2255
2256         if (tb[IFLA_VF_IB_PORT_GUID]) {
2257                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2258
2259                 if (!ops->ndo_set_vf_guid)
2260                         return -EOPNOTSUPP;
2261
2262                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2263         }
2264
2265         return err;
2266 }
2267
2268 static int do_set_master(struct net_device *dev, int ifindex,
2269                          struct netlink_ext_ack *extack)
2270 {
2271         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2272         const struct net_device_ops *ops;
2273         int err;
2274
2275         if (upper_dev) {
2276                 if (upper_dev->ifindex == ifindex)
2277                         return 0;
2278                 ops = upper_dev->netdev_ops;
2279                 if (ops->ndo_del_slave) {
2280                         err = ops->ndo_del_slave(upper_dev, dev);
2281                         if (err)
2282                                 return err;
2283                 } else {
2284                         return -EOPNOTSUPP;
2285                 }
2286         }
2287
2288         if (ifindex) {
2289                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2290                 if (!upper_dev)
2291                         return -EINVAL;
2292                 ops = upper_dev->netdev_ops;
2293                 if (ops->ndo_add_slave) {
2294                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2295                         if (err)
2296                                 return err;
2297                 } else {
2298                         return -EOPNOTSUPP;
2299                 }
2300         }
2301         return 0;
2302 }
2303
2304 #define DO_SETLINK_MODIFIED     0x01
2305 /* notify flag means notify + modified. */
2306 #define DO_SETLINK_NOTIFY       0x03
2307 static int do_setlink(const struct sk_buff *skb,
2308                       struct net_device *dev, struct ifinfomsg *ifm,
2309                       struct netlink_ext_ack *extack,
2310                       struct nlattr **tb, char *ifname, int status)
2311 {
2312         const struct net_device_ops *ops = dev->netdev_ops;
2313         int err;
2314
2315         err = validate_linkmsg(dev, tb);
2316         if (err < 0)
2317                 return err;
2318
2319         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2320                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2321                                                             tb, CAP_NET_ADMIN);
2322                 if (IS_ERR(net)) {
2323                         err = PTR_ERR(net);
2324                         goto errout;
2325                 }
2326
2327                 err = dev_change_net_namespace(dev, net, ifname);
2328                 put_net(net);
2329                 if (err)
2330                         goto errout;
2331                 status |= DO_SETLINK_MODIFIED;
2332         }
2333
2334         if (tb[IFLA_MAP]) {
2335                 struct rtnl_link_ifmap *u_map;
2336                 struct ifmap k_map;
2337
2338                 if (!ops->ndo_set_config) {
2339                         err = -EOPNOTSUPP;
2340                         goto errout;
2341                 }
2342
2343                 if (!netif_device_present(dev)) {
2344                         err = -ENODEV;
2345                         goto errout;
2346                 }
2347
2348                 u_map = nla_data(tb[IFLA_MAP]);
2349                 k_map.mem_start = (unsigned long) u_map->mem_start;
2350                 k_map.mem_end = (unsigned long) u_map->mem_end;
2351                 k_map.base_addr = (unsigned short) u_map->base_addr;
2352                 k_map.irq = (unsigned char) u_map->irq;
2353                 k_map.dma = (unsigned char) u_map->dma;
2354                 k_map.port = (unsigned char) u_map->port;
2355
2356                 err = ops->ndo_set_config(dev, &k_map);
2357                 if (err < 0)
2358                         goto errout;
2359
2360                 status |= DO_SETLINK_NOTIFY;
2361         }
2362
2363         if (tb[IFLA_ADDRESS]) {
2364                 struct sockaddr *sa;
2365                 int len;
2366
2367                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2368                                                   sizeof(*sa));
2369                 sa = kmalloc(len, GFP_KERNEL);
2370                 if (!sa) {
2371                         err = -ENOMEM;
2372                         goto errout;
2373                 }
2374                 sa->sa_family = dev->type;
2375                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2376                        dev->addr_len);
2377                 err = dev_set_mac_address(dev, sa);
2378                 kfree(sa);
2379                 if (err)
2380                         goto errout;
2381                 status |= DO_SETLINK_MODIFIED;
2382         }
2383
2384         if (tb[IFLA_MTU]) {
2385                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2386                 if (err < 0)
2387                         goto errout;
2388                 status |= DO_SETLINK_MODIFIED;
2389         }
2390
2391         if (tb[IFLA_GROUP]) {
2392                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2393                 status |= DO_SETLINK_NOTIFY;
2394         }
2395
2396         /*
2397          * Interface selected by interface index but interface
2398          * name provided implies that a name change has been
2399          * requested.
2400          */
2401         if (ifm->ifi_index > 0 && ifname[0]) {
2402                 err = dev_change_name(dev, ifname);
2403                 if (err < 0)
2404                         goto errout;
2405                 status |= DO_SETLINK_MODIFIED;
2406         }
2407
2408         if (tb[IFLA_IFALIAS]) {
2409                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2410                                     nla_len(tb[IFLA_IFALIAS]));
2411                 if (err < 0)
2412                         goto errout;
2413                 status |= DO_SETLINK_NOTIFY;
2414         }
2415
2416         if (tb[IFLA_BROADCAST]) {
2417                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2418                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2419         }
2420
2421         if (ifm->ifi_flags || ifm->ifi_change) {
2422                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2423                 if (err < 0)
2424                         goto errout;
2425         }
2426
2427         if (tb[IFLA_MASTER]) {
2428                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2429                 if (err)
2430                         goto errout;
2431                 status |= DO_SETLINK_MODIFIED;
2432         }
2433
2434         if (tb[IFLA_CARRIER]) {
2435                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2436                 if (err)
2437                         goto errout;
2438                 status |= DO_SETLINK_MODIFIED;
2439         }
2440
2441         if (tb[IFLA_TXQLEN]) {
2442                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2443
2444                 err = dev_change_tx_queue_len(dev, value);
2445                 if (err)
2446                         goto errout;
2447                 status |= DO_SETLINK_MODIFIED;
2448         }
2449
2450         if (tb[IFLA_GSO_MAX_SIZE]) {
2451                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2452
2453                 if (max_size > GSO_MAX_SIZE) {
2454                         err = -EINVAL;
2455                         goto errout;
2456                 }
2457
2458                 if (dev->gso_max_size ^ max_size) {
2459                         netif_set_gso_max_size(dev, max_size);
2460                         status |= DO_SETLINK_MODIFIED;
2461                 }
2462         }
2463
2464         if (tb[IFLA_GSO_MAX_SEGS]) {
2465                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2466
2467                 if (max_segs > GSO_MAX_SEGS) {
2468                         err = -EINVAL;
2469                         goto errout;
2470                 }
2471
2472                 if (dev->gso_max_segs ^ max_segs) {
2473                         dev->gso_max_segs = max_segs;
2474                         status |= DO_SETLINK_MODIFIED;
2475                 }
2476         }
2477
2478         if (tb[IFLA_OPERSTATE])
2479                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2480
2481         if (tb[IFLA_LINKMODE]) {
2482                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2483
2484                 write_lock_bh(&dev_base_lock);
2485                 if (dev->link_mode ^ value)
2486                         status |= DO_SETLINK_NOTIFY;
2487                 dev->link_mode = value;
2488                 write_unlock_bh(&dev_base_lock);
2489         }
2490
2491         if (tb[IFLA_VFINFO_LIST]) {
2492                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2493                 struct nlattr *attr;
2494                 int rem;
2495
2496                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2497                         if (nla_type(attr) != IFLA_VF_INFO ||
2498                             nla_len(attr) < NLA_HDRLEN) {
2499                                 err = -EINVAL;
2500                                 goto errout;
2501                         }
2502                         err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2503                                                ifla_vf_policy, NULL);
2504                         if (err < 0)
2505                                 goto errout;
2506                         err = do_setvfinfo(dev, vfinfo);
2507                         if (err < 0)
2508                                 goto errout;
2509                         status |= DO_SETLINK_NOTIFY;
2510                 }
2511         }
2512         err = 0;
2513
2514         if (tb[IFLA_VF_PORTS]) {
2515                 struct nlattr *port[IFLA_PORT_MAX+1];
2516                 struct nlattr *attr;
2517                 int vf;
2518                 int rem;
2519
2520                 err = -EOPNOTSUPP;
2521                 if (!ops->ndo_set_vf_port)
2522                         goto errout;
2523
2524                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2525                         if (nla_type(attr) != IFLA_VF_PORT ||
2526                             nla_len(attr) < NLA_HDRLEN) {
2527                                 err = -EINVAL;
2528                                 goto errout;
2529                         }
2530                         err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2531                                                ifla_port_policy, NULL);
2532                         if (err < 0)
2533                                 goto errout;
2534                         if (!port[IFLA_PORT_VF]) {
2535                                 err = -EOPNOTSUPP;
2536                                 goto errout;
2537                         }
2538                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2539                         err = ops->ndo_set_vf_port(dev, vf, port);
2540                         if (err < 0)
2541                                 goto errout;
2542                         status |= DO_SETLINK_NOTIFY;
2543                 }
2544         }
2545         err = 0;
2546
2547         if (tb[IFLA_PORT_SELF]) {
2548                 struct nlattr *port[IFLA_PORT_MAX+1];
2549
2550                 err = nla_parse_nested(port, IFLA_PORT_MAX,
2551                                        tb[IFLA_PORT_SELF], ifla_port_policy,
2552                                        NULL);
2553                 if (err < 0)
2554                         goto errout;
2555
2556                 err = -EOPNOTSUPP;
2557                 if (ops->ndo_set_vf_port)
2558                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2559                 if (err < 0)
2560                         goto errout;
2561                 status |= DO_SETLINK_NOTIFY;
2562         }
2563
2564         if (tb[IFLA_AF_SPEC]) {
2565                 struct nlattr *af;
2566                 int rem;
2567
2568                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2569                         const struct rtnl_af_ops *af_ops;
2570
2571                         rcu_read_lock();
2572
2573                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2574
2575                         err = af_ops->set_link_af(dev, af);
2576                         if (err < 0) {
2577                                 rcu_read_unlock();
2578                                 goto errout;
2579                         }
2580
2581                         rcu_read_unlock();
2582                         status |= DO_SETLINK_NOTIFY;
2583                 }
2584         }
2585         err = 0;
2586
2587         if (tb[IFLA_PROTO_DOWN]) {
2588                 err = dev_change_proto_down(dev,
2589                                             nla_get_u8(tb[IFLA_PROTO_DOWN]));
2590                 if (err)
2591                         goto errout;
2592                 status |= DO_SETLINK_NOTIFY;
2593         }
2594
2595         if (tb[IFLA_XDP]) {
2596                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2597                 u32 xdp_flags = 0;
2598
2599                 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2600                                        ifla_xdp_policy, NULL);
2601                 if (err < 0)
2602                         goto errout;
2603
2604                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2605                         err = -EINVAL;
2606                         goto errout;
2607                 }
2608
2609                 if (xdp[IFLA_XDP_FLAGS]) {
2610                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2611                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2612                                 err = -EINVAL;
2613                                 goto errout;
2614                         }
2615                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2616                                 err = -EINVAL;
2617                                 goto errout;
2618                         }
2619                 }
2620
2621                 if (xdp[IFLA_XDP_FD]) {
2622                         err = dev_change_xdp_fd(dev, extack,
2623                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2624                                                 xdp_flags);
2625                         if (err)
2626                                 goto errout;
2627                         status |= DO_SETLINK_NOTIFY;
2628                 }
2629         }
2630
2631 errout:
2632         if (status & DO_SETLINK_MODIFIED) {
2633                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2634                         netdev_state_change(dev);
2635
2636                 if (err < 0)
2637                         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",
2638                                              dev->name);
2639         }
2640
2641         return err;
2642 }
2643
2644 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2645                         struct netlink_ext_ack *extack)
2646 {
2647         struct net *net = sock_net(skb->sk);
2648         struct ifinfomsg *ifm;
2649         struct net_device *dev;
2650         int err;
2651         struct nlattr *tb[IFLA_MAX+1];
2652         char ifname[IFNAMSIZ];
2653
2654         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2655                           extack);
2656         if (err < 0)
2657                 goto errout;
2658
2659         err = rtnl_ensure_unique_netns(tb, extack, false);
2660         if (err < 0)
2661                 goto errout;
2662
2663         if (tb[IFLA_IFNAME])
2664                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2665         else
2666                 ifname[0] = '\0';
2667
2668         err = -EINVAL;
2669         ifm = nlmsg_data(nlh);
2670         if (ifm->ifi_index > 0)
2671                 dev = __dev_get_by_index(net, ifm->ifi_index);
2672         else if (tb[IFLA_IFNAME])
2673                 dev = __dev_get_by_name(net, ifname);
2674         else
2675                 goto errout;
2676
2677         if (dev == NULL) {
2678                 err = -ENODEV;
2679                 goto errout;
2680         }
2681
2682         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2683 errout:
2684         return err;
2685 }
2686
2687 static int rtnl_group_dellink(const struct net *net, int group)
2688 {
2689         struct net_device *dev, *aux;
2690         LIST_HEAD(list_kill);
2691         bool found = false;
2692
2693         if (!group)
2694                 return -EPERM;
2695
2696         for_each_netdev(net, dev) {
2697                 if (dev->group == group) {
2698                         const struct rtnl_link_ops *ops;
2699
2700                         found = true;
2701                         ops = dev->rtnl_link_ops;
2702                         if (!ops || !ops->dellink)
2703                                 return -EOPNOTSUPP;
2704                 }
2705         }
2706
2707         if (!found)
2708                 return -ENODEV;
2709
2710         for_each_netdev_safe(net, dev, aux) {
2711                 if (dev->group == group) {
2712                         const struct rtnl_link_ops *ops;
2713
2714                         ops = dev->rtnl_link_ops;
2715                         ops->dellink(dev, &list_kill);
2716                 }
2717         }
2718         unregister_netdevice_many(&list_kill);
2719
2720         return 0;
2721 }
2722
2723 int rtnl_delete_link(struct net_device *dev)
2724 {
2725         const struct rtnl_link_ops *ops;
2726         LIST_HEAD(list_kill);
2727
2728         ops = dev->rtnl_link_ops;
2729         if (!ops || !ops->dellink)
2730                 return -EOPNOTSUPP;
2731
2732         ops->dellink(dev, &list_kill);
2733         unregister_netdevice_many(&list_kill);
2734
2735         return 0;
2736 }
2737 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2738
2739 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2740                         struct netlink_ext_ack *extack)
2741 {
2742         struct net *net = sock_net(skb->sk);
2743         struct net *tgt_net = net;
2744         struct net_device *dev = NULL;
2745         struct ifinfomsg *ifm;
2746         char ifname[IFNAMSIZ];
2747         struct nlattr *tb[IFLA_MAX+1];
2748         int err;
2749         int netnsid = -1;
2750
2751         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2752         if (err < 0)
2753                 return err;
2754
2755         err = rtnl_ensure_unique_netns(tb, extack, true);
2756         if (err < 0)
2757                 return err;
2758
2759         if (tb[IFLA_IFNAME])
2760                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2761
2762         if (tb[IFLA_IF_NETNSID]) {
2763                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2764                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2765                 if (IS_ERR(tgt_net))
2766                         return PTR_ERR(tgt_net);
2767         }
2768
2769         err = -EINVAL;
2770         ifm = nlmsg_data(nlh);
2771         if (ifm->ifi_index > 0)
2772                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2773         else if (tb[IFLA_IFNAME])
2774                 dev = __dev_get_by_name(tgt_net, ifname);
2775         else if (tb[IFLA_GROUP])
2776                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2777         else
2778                 goto out;
2779
2780         if (!dev) {
2781                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2782                         err = -ENODEV;
2783
2784                 goto out;
2785         }
2786
2787         err = rtnl_delete_link(dev);
2788
2789 out:
2790         if (netnsid >= 0)
2791                 put_net(tgt_net);
2792
2793         return err;
2794 }
2795
2796 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2797 {
2798         unsigned int old_flags;
2799         int err;
2800
2801         old_flags = dev->flags;
2802         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2803                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2804                 if (err < 0)
2805                         return err;
2806         }
2807
2808         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2809                 __dev_notify_flags(dev, old_flags, 0U);
2810         } else {
2811                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2812                 __dev_notify_flags(dev, old_flags, ~0U);
2813         }
2814         return 0;
2815 }
2816 EXPORT_SYMBOL(rtnl_configure_link);
2817
2818 struct net_device *rtnl_create_link(struct net *net,
2819         const char *ifname, unsigned char name_assign_type,
2820         const struct rtnl_link_ops *ops, struct nlattr *tb[])
2821 {
2822         struct net_device *dev;
2823         unsigned int num_tx_queues = 1;
2824         unsigned int num_rx_queues = 1;
2825
2826         if (tb[IFLA_NUM_TX_QUEUES])
2827                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2828         else if (ops->get_num_tx_queues)
2829                 num_tx_queues = ops->get_num_tx_queues();
2830
2831         if (tb[IFLA_NUM_RX_QUEUES])
2832                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2833         else if (ops->get_num_rx_queues)
2834                 num_rx_queues = ops->get_num_rx_queues();
2835
2836         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2837                                ops->setup, num_tx_queues, num_rx_queues);
2838         if (!dev)
2839                 return ERR_PTR(-ENOMEM);
2840
2841         dev_net_set(dev, net);
2842         dev->rtnl_link_ops = ops;
2843         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2844
2845         if (tb[IFLA_MTU])
2846                 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2847         if (tb[IFLA_ADDRESS]) {
2848                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2849                                 nla_len(tb[IFLA_ADDRESS]));
2850                 dev->addr_assign_type = NET_ADDR_SET;
2851         }
2852         if (tb[IFLA_BROADCAST])
2853                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2854                                 nla_len(tb[IFLA_BROADCAST]));
2855         if (tb[IFLA_TXQLEN])
2856                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2857         if (tb[IFLA_OPERSTATE])
2858                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2859         if (tb[IFLA_LINKMODE])
2860                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2861         if (tb[IFLA_GROUP])
2862                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2863         if (tb[IFLA_GSO_MAX_SIZE])
2864                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
2865         if (tb[IFLA_GSO_MAX_SEGS])
2866                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2867
2868         return dev;
2869 }
2870 EXPORT_SYMBOL(rtnl_create_link);
2871
2872 static int rtnl_group_changelink(const struct sk_buff *skb,
2873                 struct net *net, int group,
2874                 struct ifinfomsg *ifm,
2875                 struct netlink_ext_ack *extack,
2876                 struct nlattr **tb)
2877 {
2878         struct net_device *dev, *aux;
2879         int err;
2880
2881         for_each_netdev_safe(net, dev, aux) {
2882                 if (dev->group == group) {
2883                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2884                         if (err < 0)
2885                                 return err;
2886                 }
2887         }
2888
2889         return 0;
2890 }
2891
2892 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2893                         struct netlink_ext_ack *extack)
2894 {
2895         struct net *net = sock_net(skb->sk);
2896         const struct rtnl_link_ops *ops;
2897         const struct rtnl_link_ops *m_ops = NULL;
2898         struct net_device *dev;
2899         struct net_device *master_dev = NULL;
2900         struct ifinfomsg *ifm;
2901         char kind[MODULE_NAME_LEN];
2902         char ifname[IFNAMSIZ];
2903         struct nlattr *tb[IFLA_MAX+1];
2904         struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2905         unsigned char name_assign_type = NET_NAME_USER;
2906         int err;
2907
2908 #ifdef CONFIG_MODULES
2909 replay:
2910 #endif
2911         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2912         if (err < 0)
2913                 return err;
2914
2915         err = rtnl_ensure_unique_netns(tb, extack, false);
2916         if (err < 0)
2917                 return err;
2918
2919         if (tb[IFLA_IFNAME])
2920                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2921         else
2922                 ifname[0] = '\0';
2923
2924         ifm = nlmsg_data(nlh);
2925         if (ifm->ifi_index > 0)
2926                 dev = __dev_get_by_index(net, ifm->ifi_index);
2927         else {
2928                 if (ifname[0])
2929                         dev = __dev_get_by_name(net, ifname);
2930                 else
2931                         dev = NULL;
2932         }
2933
2934         if (dev) {
2935                 master_dev = netdev_master_upper_dev_get(dev);
2936                 if (master_dev)
2937                         m_ops = master_dev->rtnl_link_ops;
2938         }
2939
2940         err = validate_linkmsg(dev, tb);
2941         if (err < 0)
2942                 return err;
2943
2944         if (tb[IFLA_LINKINFO]) {
2945                 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2946                                        tb[IFLA_LINKINFO], ifla_info_policy,
2947                                        NULL);
2948                 if (err < 0)
2949                         return err;
2950         } else
2951                 memset(linkinfo, 0, sizeof(linkinfo));
2952
2953         if (linkinfo[IFLA_INFO_KIND]) {
2954                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2955                 ops = rtnl_link_ops_get(kind);
2956         } else {
2957                 kind[0] = '\0';
2958                 ops = NULL;
2959         }
2960
2961         if (1) {
2962                 struct nlattr *attr[RTNL_MAX_TYPE + 1];
2963                 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
2964                 struct nlattr **data = NULL;
2965                 struct nlattr **slave_data = NULL;
2966                 struct net *dest_net, *link_net = NULL;
2967
2968                 if (ops) {
2969                         if (ops->maxtype > RTNL_MAX_TYPE)
2970                                 return -EINVAL;
2971
2972                         if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2973                                 err = nla_parse_nested(attr, ops->maxtype,
2974                                                        linkinfo[IFLA_INFO_DATA],
2975                                                        ops->policy, NULL);
2976                                 if (err < 0)
2977                                         return err;
2978                                 data = attr;
2979                         }
2980                         if (ops->validate) {
2981                                 err = ops->validate(tb, data, extack);
2982                                 if (err < 0)
2983                                         return err;
2984                         }
2985                 }
2986
2987                 if (m_ops) {
2988                         if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
2989                                 return -EINVAL;
2990
2991                         if (m_ops->slave_maxtype &&
2992                             linkinfo[IFLA_INFO_SLAVE_DATA]) {
2993                                 err = nla_parse_nested(slave_attr,
2994                                                        m_ops->slave_maxtype,
2995                                                        linkinfo[IFLA_INFO_SLAVE_DATA],
2996                                                        m_ops->slave_policy,
2997                                                        NULL);
2998                                 if (err < 0)
2999                                         return err;
3000                                 slave_data = slave_attr;
3001                         }
3002                 }
3003
3004                 if (dev) {
3005                         int status = 0;
3006
3007                         if (nlh->nlmsg_flags & NLM_F_EXCL)
3008                                 return -EEXIST;
3009                         if (nlh->nlmsg_flags & NLM_F_REPLACE)
3010                                 return -EOPNOTSUPP;
3011
3012                         if (linkinfo[IFLA_INFO_DATA]) {
3013                                 if (!ops || ops != dev->rtnl_link_ops ||
3014                                     !ops->changelink)
3015                                         return -EOPNOTSUPP;
3016
3017                                 err = ops->changelink(dev, tb, data, extack);
3018                                 if (err < 0)
3019                                         return err;
3020                                 status |= DO_SETLINK_NOTIFY;
3021                         }
3022
3023                         if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3024                                 if (!m_ops || !m_ops->slave_changelink)
3025                                         return -EOPNOTSUPP;
3026
3027                                 err = m_ops->slave_changelink(master_dev, dev,
3028                                                               tb, slave_data,
3029                                                               extack);
3030                                 if (err < 0)
3031                                         return err;
3032                                 status |= DO_SETLINK_NOTIFY;
3033                         }
3034
3035                         return do_setlink(skb, dev, ifm, extack, tb, ifname,
3036                                           status);
3037                 }
3038
3039                 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3040                         if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3041                                 return rtnl_group_changelink(skb, net,
3042                                                 nla_get_u32(tb[IFLA_GROUP]),
3043                                                 ifm, extack, tb);
3044                         return -ENODEV;
3045                 }
3046
3047                 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3048                         return -EOPNOTSUPP;
3049
3050                 if (!ops) {
3051 #ifdef CONFIG_MODULES
3052                         if (kind[0]) {
3053                                 __rtnl_unlock();
3054                                 request_module("rtnl-link-%s", kind);
3055                                 rtnl_lock();
3056                                 ops = rtnl_link_ops_get(kind);
3057                                 if (ops)
3058                                         goto replay;
3059                         }
3060 #endif
3061                         return -EOPNOTSUPP;
3062                 }
3063
3064                 if (!ops->setup)
3065                         return -EOPNOTSUPP;
3066
3067                 if (!ifname[0]) {
3068                         snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3069                         name_assign_type = NET_NAME_ENUM;
3070                 }
3071
3072                 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3073                 if (IS_ERR(dest_net))
3074                         return PTR_ERR(dest_net);
3075
3076                 if (tb[IFLA_LINK_NETNSID]) {
3077                         int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3078
3079                         link_net = get_net_ns_by_id(dest_net, id);
3080                         if (!link_net) {
3081                                 err =  -EINVAL;
3082                                 goto out;
3083                         }
3084                         err = -EPERM;
3085                         if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3086                                 goto out;
3087                 }
3088
3089                 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3090                                        name_assign_type, ops, tb);
3091                 if (IS_ERR(dev)) {
3092                         err = PTR_ERR(dev);
3093                         goto out;
3094                 }
3095
3096                 dev->ifindex = ifm->ifi_index;
3097
3098                 if (ops->newlink) {
3099                         err = ops->newlink(link_net ? : net, dev, tb, data,
3100                                            extack);
3101                         /* Drivers should call free_netdev() in ->destructor
3102                          * and unregister it on failure after registration
3103                          * so that device could be finally freed in rtnl_unlock.
3104                          */
3105                         if (err < 0) {
3106                                 /* If device is not registered at all, free it now */
3107                                 if (dev->reg_state == NETREG_UNINITIALIZED)
3108                                         free_netdev(dev);
3109                                 goto out;
3110                         }
3111                 } else {
3112                         err = register_netdevice(dev);
3113                         if (err < 0) {
3114                                 free_netdev(dev);
3115                                 goto out;
3116                         }
3117                 }
3118                 err = rtnl_configure_link(dev, ifm);
3119                 if (err < 0)
3120                         goto out_unregister;
3121                 if (link_net) {
3122                         err = dev_change_net_namespace(dev, dest_net, ifname);
3123                         if (err < 0)
3124                                 goto out_unregister;
3125                 }
3126                 if (tb[IFLA_MASTER]) {
3127                         err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
3128                                             extack);
3129                         if (err)
3130                                 goto out_unregister;
3131                 }
3132 out:
3133                 if (link_net)
3134                         put_net(link_net);
3135                 put_net(dest_net);
3136                 return err;
3137 out_unregister:
3138                 if (ops->newlink) {
3139                         LIST_HEAD(list_kill);
3140
3141                         ops->dellink(dev, &list_kill);
3142                         unregister_netdevice_many(&list_kill);
3143                 } else {
3144                         unregister_netdevice(dev);
3145                 }
3146                 goto out;
3147         }
3148 }
3149
3150 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3151                         struct netlink_ext_ack *extack)
3152 {
3153         struct net *net = sock_net(skb->sk);
3154         struct net *tgt_net = net;
3155         struct ifinfomsg *ifm;
3156         char ifname[IFNAMSIZ];
3157         struct nlattr *tb[IFLA_MAX+1];
3158         struct net_device *dev = NULL;
3159         struct sk_buff *nskb;
3160         int netnsid = -1;
3161         int err;
3162         u32 ext_filter_mask = 0;
3163
3164         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3165         if (err < 0)
3166                 return err;
3167
3168         err = rtnl_ensure_unique_netns(tb, extack, true);
3169         if (err < 0)
3170                 return err;
3171
3172         if (tb[IFLA_IF_NETNSID]) {
3173                 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3174                 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3175                 if (IS_ERR(tgt_net))
3176                         return PTR_ERR(tgt_net);
3177         }
3178
3179         if (tb[IFLA_IFNAME])
3180                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3181
3182         if (tb[IFLA_EXT_MASK])
3183                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3184
3185         err = -EINVAL;
3186         ifm = nlmsg_data(nlh);
3187         if (ifm->ifi_index > 0)
3188                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3189         else if (tb[IFLA_IFNAME])
3190                 dev = __dev_get_by_name(tgt_net, ifname);
3191         else
3192                 goto out;
3193
3194         err = -ENODEV;
3195         if (dev == NULL)
3196                 goto out;
3197
3198         err = -ENOBUFS;
3199         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3200         if (nskb == NULL)
3201                 goto out;
3202
3203         err = rtnl_fill_ifinfo(nskb, dev, net,
3204                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3205                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3206                                0, NULL, 0, netnsid);
3207         if (err < 0) {
3208                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3209                 WARN_ON(err == -EMSGSIZE);
3210                 kfree_skb(nskb);
3211         } else
3212                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3213 out:
3214         if (netnsid >= 0)
3215                 put_net(tgt_net);
3216
3217         return err;
3218 }
3219
3220 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3221 {
3222         struct net *net = sock_net(skb->sk);
3223         struct net_device *dev;
3224         struct nlattr *tb[IFLA_MAX+1];
3225         u32 ext_filter_mask = 0;
3226         u16 min_ifinfo_dump_size = 0;
3227         int hdrlen;
3228
3229         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3230         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3231                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3232
3233         if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3234                 if (tb[IFLA_EXT_MASK])
3235                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3236         }
3237
3238         if (!ext_filter_mask)
3239                 return NLMSG_GOODSIZE;
3240         /*
3241          * traverse the list of net devices and compute the minimum
3242          * buffer size based upon the filter mask.
3243          */
3244         rcu_read_lock();
3245         for_each_netdev_rcu(net, dev) {
3246                 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3247                                              if_nlmsg_size(dev,
3248                                                            ext_filter_mask));
3249         }
3250         rcu_read_unlock();
3251
3252         return nlmsg_total_size(min_ifinfo_dump_size);
3253 }
3254
3255 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3256 {
3257         int idx;
3258         int s_idx = cb->family;
3259
3260         if (s_idx == 0)
3261                 s_idx = 1;
3262
3263         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3264                 struct rtnl_link **tab;
3265                 int type = cb->nlh->nlmsg_type-RTM_BASE;
3266                 struct rtnl_link *link;
3267                 rtnl_dumpit_func dumpit;
3268
3269                 if (idx < s_idx || idx == PF_PACKET)
3270                         continue;
3271
3272                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3273                         continue;
3274
3275                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3276                 if (!tab)
3277                         continue;
3278
3279                 link = tab[type];
3280                 if (!link)
3281                         continue;
3282
3283                 dumpit = link->dumpit;
3284                 if (!dumpit)
3285                         continue;
3286
3287                 if (idx > s_idx) {
3288                         memset(&cb->args[0], 0, sizeof(cb->args));
3289                         cb->prev_seq = 0;
3290                         cb->seq = 0;
3291                 }
3292                 if (dumpit(skb, cb))
3293                         break;
3294         }
3295         cb->family = idx;
3296
3297         return skb->len;
3298 }
3299
3300 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3301                                        unsigned int change,
3302                                        u32 event, gfp_t flags, int *new_nsid,
3303                                        int new_ifindex)
3304 {
3305         struct net *net = dev_net(dev);
3306         struct sk_buff *skb;
3307         int err = -ENOBUFS;
3308         size_t if_info_size;
3309
3310         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3311         if (skb == NULL)
3312                 goto errout;
3313
3314         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3315                                type, 0, 0, change, 0, 0, event,
3316                                new_nsid, new_ifindex, -1);
3317         if (err < 0) {
3318                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3319                 WARN_ON(err == -EMSGSIZE);
3320                 kfree_skb(skb);
3321                 goto errout;
3322         }
3323         return skb;
3324 errout:
3325         if (err < 0)
3326                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3327         return NULL;
3328 }
3329
3330 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3331 {
3332         struct net *net = dev_net(dev);
3333
3334         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3335 }
3336
3337 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3338                                unsigned int change, u32 event,
3339                                gfp_t flags, int *new_nsid, int new_ifindex)
3340 {
3341         struct sk_buff *skb;
3342
3343         if (dev->reg_state != NETREG_REGISTERED)
3344                 return;
3345
3346         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3347                                      new_ifindex);
3348         if (skb)
3349                 rtmsg_ifinfo_send(skb, dev, flags);
3350 }
3351
3352 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3353                   gfp_t flags)
3354 {
3355         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3356                            NULL, 0);
3357 }
3358
3359 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3360                          gfp_t flags, int *new_nsid, int new_ifindex)
3361 {
3362         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3363                            new_nsid, new_ifindex);
3364 }
3365
3366 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3367                                    struct net_device *dev,
3368                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3369                                    int type, unsigned int flags,
3370                                    int nlflags, u16 ndm_state)
3371 {
3372         struct nlmsghdr *nlh;
3373         struct ndmsg *ndm;
3374
3375         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3376         if (!nlh)
3377                 return -EMSGSIZE;
3378
3379         ndm = nlmsg_data(nlh);
3380         ndm->ndm_family  = AF_BRIDGE;
3381         ndm->ndm_pad1    = 0;
3382         ndm->ndm_pad2    = 0;
3383         ndm->ndm_flags   = flags;
3384         ndm->ndm_type    = 0;
3385         ndm->ndm_ifindex = dev->ifindex;
3386         ndm->ndm_state   = ndm_state;
3387
3388         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3389                 goto nla_put_failure;
3390         if (vid)
3391                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3392                         goto nla_put_failure;
3393
3394         nlmsg_end(skb, nlh);
3395         return 0;
3396
3397 nla_put_failure:
3398         nlmsg_cancel(skb, nlh);
3399         return -EMSGSIZE;
3400 }
3401
3402 static inline size_t rtnl_fdb_nlmsg_size(void)
3403 {
3404         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3405                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3406                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3407                0;
3408 }
3409
3410 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3411                             u16 ndm_state)
3412 {
3413         struct net *net = dev_net(dev);
3414         struct sk_buff *skb;
3415         int err = -ENOBUFS;
3416
3417         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3418         if (!skb)
3419                 goto errout;
3420
3421         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3422                                       0, 0, type, NTF_SELF, 0, ndm_state);
3423         if (err < 0) {
3424                 kfree_skb(skb);
3425                 goto errout;
3426         }
3427
3428         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3429         return;
3430 errout:
3431         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3432 }
3433
3434 /**
3435  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3436  */
3437 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3438                      struct nlattr *tb[],
3439                      struct net_device *dev,
3440                      const unsigned char *addr, u16 vid,
3441                      u16 flags)
3442 {
3443         int err = -EINVAL;
3444
3445         /* If aging addresses are supported device will need to
3446          * implement its own handler for this.
3447          */
3448         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3449                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3450                 return err;
3451         }
3452
3453         if (vid) {
3454                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3455                 return err;
3456         }
3457
3458         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3459                 err = dev_uc_add_excl(dev, addr);
3460         else if (is_multicast_ether_addr(addr))
3461                 err = dev_mc_add_excl(dev, addr);
3462
3463         /* Only return duplicate errors if NLM_F_EXCL is set */
3464         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3465                 err = 0;
3466
3467         return err;
3468 }
3469 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3470
3471 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3472                          struct netlink_ext_ack *extack)
3473 {
3474         u16 vid = 0;
3475
3476         if (vlan_attr) {
3477                 if (nla_len(vlan_attr) != sizeof(u16)) {
3478                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3479                         return -EINVAL;
3480                 }
3481
3482                 vid = nla_get_u16(vlan_attr);
3483
3484                 if (!vid || vid >= VLAN_VID_MASK) {
3485                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3486                         return -EINVAL;
3487                 }
3488         }
3489         *p_vid = vid;
3490         return 0;
3491 }
3492
3493 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3494                         struct netlink_ext_ack *extack)
3495 {
3496         struct net *net = sock_net(skb->sk);
3497         struct ndmsg *ndm;
3498         struct nlattr *tb[NDA_MAX+1];
3499         struct net_device *dev;
3500         u8 *addr;
3501         u16 vid;
3502         int err;
3503
3504         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3505         if (err < 0)
3506                 return err;
3507
3508         ndm = nlmsg_data(nlh);
3509         if (ndm->ndm_ifindex == 0) {
3510                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3511                 return -EINVAL;
3512         }
3513
3514         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3515         if (dev == NULL) {
3516                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3517                 return -ENODEV;
3518         }
3519
3520         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3521                 NL_SET_ERR_MSG(extack, "invalid address");
3522                 return -EINVAL;
3523         }
3524
3525         addr = nla_data(tb[NDA_LLADDR]);
3526
3527         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3528         if (err)
3529                 return err;
3530
3531         err = -EOPNOTSUPP;
3532
3533         /* Support fdb on master device the net/bridge default case */
3534         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3535             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3536                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3537                 const struct net_device_ops *ops = br_dev->netdev_ops;
3538
3539                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3540                                        nlh->nlmsg_flags);
3541                 if (err)
3542                         goto out;
3543                 else
3544                         ndm->ndm_flags &= ~NTF_MASTER;
3545         }
3546
3547         /* Embedded bridge, macvlan, and any other device support */
3548         if ((ndm->ndm_flags & NTF_SELF)) {
3549                 if (dev->netdev_ops->ndo_fdb_add)
3550                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3551                                                            vid,
3552                                                            nlh->nlmsg_flags);
3553                 else
3554                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3555                                                nlh->nlmsg_flags);
3556
3557                 if (!err) {
3558                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3559                                         ndm->ndm_state);
3560                         ndm->ndm_flags &= ~NTF_SELF;
3561                 }
3562         }
3563 out:
3564         return err;
3565 }
3566
3567 /**
3568  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3569  */
3570 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3571                      struct nlattr *tb[],
3572                      struct net_device *dev,
3573                      const unsigned char *addr, u16 vid)
3574 {
3575         int err = -EINVAL;
3576
3577         /* If aging addresses are supported device will need to
3578          * implement its own handler for this.
3579          */
3580         if (!(ndm->ndm_state & NUD_PERMANENT)) {
3581                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3582                 return err;
3583         }
3584
3585         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3586                 err = dev_uc_del(dev, addr);
3587         else if (is_multicast_ether_addr(addr))
3588                 err = dev_mc_del(dev, addr);
3589
3590         return err;
3591 }
3592 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3593
3594 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3595                         struct netlink_ext_ack *extack)
3596 {
3597         struct net *net = sock_net(skb->sk);
3598         struct ndmsg *ndm;
3599         struct nlattr *tb[NDA_MAX+1];
3600         struct net_device *dev;
3601         int err = -EINVAL;
3602         __u8 *addr;
3603         u16 vid;
3604
3605         if (!netlink_capable(skb, CAP_NET_ADMIN))
3606                 return -EPERM;
3607
3608         err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3609         if (err < 0)
3610                 return err;
3611
3612         ndm = nlmsg_data(nlh);
3613         if (ndm->ndm_ifindex == 0) {
3614                 NL_SET_ERR_MSG(extack, "invalid ifindex");
3615                 return -EINVAL;
3616         }
3617
3618         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3619         if (dev == NULL) {
3620                 NL_SET_ERR_MSG(extack, "unknown ifindex");
3621                 return -ENODEV;
3622         }
3623
3624         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3625                 NL_SET_ERR_MSG(extack, "invalid address");
3626                 return -EINVAL;
3627         }
3628
3629         addr = nla_data(tb[NDA_LLADDR]);
3630
3631         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3632         if (err)
3633                 return err;
3634
3635         err = -EOPNOTSUPP;
3636
3637         /* Support fdb on master device the net/bridge default case */
3638         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3639             (dev->priv_flags & IFF_BRIDGE_PORT)) {
3640                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3641                 const struct net_device_ops *ops = br_dev->netdev_ops;
3642
3643                 if (ops->ndo_fdb_del)
3644                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3645
3646                 if (err)
3647                         goto out;
3648                 else
3649                         ndm->ndm_flags &= ~NTF_MASTER;
3650         }
3651
3652         /* Embedded bridge, macvlan, and any other device support */
3653         if (ndm->ndm_flags & NTF_SELF) {
3654                 if (dev->netdev_ops->ndo_fdb_del)
3655                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3656                                                            vid);
3657                 else
3658                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3659
3660                 if (!err) {
3661                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3662                                         ndm->ndm_state);
3663                         ndm->ndm_flags &= ~NTF_SELF;
3664                 }
3665         }
3666 out:
3667         return err;
3668 }
3669
3670 static int nlmsg_populate_fdb(struct sk_buff *skb,
3671                               struct netlink_callback *cb,
3672                               struct net_device *dev,
3673                               int *idx,
3674                               struct netdev_hw_addr_list *list)
3675 {
3676         struct netdev_hw_addr *ha;
3677         int err;
3678         u32 portid, seq;
3679
3680         portid = NETLINK_CB(cb->skb).portid;
3681         seq = cb->nlh->nlmsg_seq;
3682
3683         list_for_each_entry(ha, &list->list, list) {
3684                 if (*idx < cb->args[2])
3685                         goto skip;
3686
3687                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3688                                               portid, seq,
3689                                               RTM_NEWNEIGH, NTF_SELF,
3690                                               NLM_F_MULTI, NUD_PERMANENT);
3691                 if (err < 0)
3692                         return err;
3693 skip:
3694                 *idx += 1;
3695         }
3696         return 0;
3697 }
3698
3699 /**
3700  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3701  * @nlh: netlink message header
3702  * @dev: netdevice
3703  *
3704  * Default netdevice operation to dump the existing unicast address list.
3705  * Returns number of addresses from list put in skb.
3706  */
3707 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3708                       struct netlink_callback *cb,
3709                       struct net_device *dev,
3710                       struct net_device *filter_dev,
3711                       int *idx)
3712 {
3713         int err;
3714
3715         netif_addr_lock_bh(dev);
3716         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3717         if (err)
3718                 goto out;
3719         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3720 out:
3721         netif_addr_unlock_bh(dev);
3722         return err;
3723 }
3724 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3725
3726 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3727 {
3728         struct net_device *dev;
3729         struct nlattr *tb[IFLA_MAX+1];
3730         struct net_device *br_dev = NULL;
3731         const struct net_device_ops *ops = NULL;
3732         const struct net_device_ops *cops = NULL;
3733         struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3734         struct net *net = sock_net(skb->sk);
3735         struct hlist_head *head;
3736         int brport_idx = 0;
3737         int br_idx = 0;
3738         int h, s_h;
3739         int idx = 0, s_idx;
3740         int err = 0;
3741         int fidx = 0;
3742
3743         err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3744                           IFLA_MAX, ifla_policy, NULL);
3745         if (err < 0) {
3746                 return -EINVAL;
3747         } else if (err == 0) {
3748                 if (tb[IFLA_MASTER])
3749                         br_idx = nla_get_u32(tb[IFLA_MASTER]);
3750         }
3751
3752         brport_idx = ifm->ifi_index;
3753
3754         if (br_idx) {
3755                 br_dev = __dev_get_by_index(net, br_idx);
3756                 if (!br_dev)
3757                         return -ENODEV;
3758
3759                 ops = br_dev->netdev_ops;
3760         }
3761
3762         s_h = cb->args[0];
3763         s_idx = cb->args[1];
3764
3765         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3766                 idx = 0;
3767                 head = &net->dev_index_head[h];
3768                 hlist_for_each_entry(dev, head, index_hlist) {
3769
3770                         if (brport_idx && (dev->ifindex != brport_idx))
3771                                 continue;
3772
3773                         if (!br_idx) { /* user did not specify a specific bridge */
3774                                 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3775                                         br_dev = netdev_master_upper_dev_get(dev);
3776                                         cops = br_dev->netdev_ops;
3777                                 }
3778                         } else {
3779                                 if (dev != br_dev &&
3780                                     !(dev->priv_flags & IFF_BRIDGE_PORT))
3781                                         continue;
3782
3783                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
3784                                     !(dev->priv_flags & IFF_EBRIDGE))
3785                                         continue;
3786                                 cops = ops;
3787                         }
3788
3789                         if (idx < s_idx)
3790                                 goto cont;
3791
3792                         if (dev->priv_flags & IFF_BRIDGE_PORT) {
3793                                 if (cops && cops->ndo_fdb_dump) {
3794                                         err = cops->ndo_fdb_dump(skb, cb,
3795                                                                 br_dev, dev,
3796                                                                 &fidx);
3797                                         if (err == -EMSGSIZE)
3798                                                 goto out;
3799                                 }
3800                         }
3801
3802                         if (dev->netdev_ops->ndo_fdb_dump)
3803                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3804                                                                     dev, NULL,
3805                                                                     &fidx);
3806                         else
3807                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3808                                                         &fidx);
3809                         if (err == -EMSGSIZE)
3810                                 goto out;
3811
3812                         cops = NULL;
3813
3814                         /* reset fdb offset to 0 for rest of the interfaces */
3815                         cb->args[2] = 0;
3816                         fidx = 0;
3817 cont:
3818                         idx++;
3819                 }
3820         }
3821
3822 out:
3823         cb->args[0] = h;
3824         cb->args[1] = idx;
3825         cb->args[2] = fidx;
3826
3827         return skb->len;
3828 }
3829
3830 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3831                                unsigned int attrnum, unsigned int flag)
3832 {
3833         if (mask & flag)
3834                 return nla_put_u8(skb, attrnum, !!(flags & flag));
3835         return 0;
3836 }
3837
3838 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3839                             struct net_device *dev, u16 mode,
3840                             u32 flags, u32 mask, int nlflags,
3841                             u32 filter_mask,
3842                             int (*vlan_fill)(struct sk_buff *skb,
3843                                              struct net_device *dev,
3844                                              u32 filter_mask))
3845 {
3846         struct nlmsghdr *nlh;
3847         struct ifinfomsg *ifm;
3848         struct nlattr *br_afspec;
3849         struct nlattr *protinfo;
3850         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3851         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3852         int err = 0;
3853
3854         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3855         if (nlh == NULL)
3856                 return -EMSGSIZE;
3857
3858         ifm = nlmsg_data(nlh);
3859         ifm->ifi_family = AF_BRIDGE;
3860         ifm->__ifi_pad = 0;
3861         ifm->ifi_type = dev->type;
3862         ifm->ifi_index = dev->ifindex;
3863         ifm->ifi_flags = dev_get_flags(dev);
3864         ifm->ifi_change = 0;
3865
3866
3867         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3868             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3869             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3870             (br_dev &&
3871              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3872             (dev->addr_len &&
3873              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3874             (dev->ifindex != dev_get_iflink(dev) &&
3875              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3876                 goto nla_put_failure;
3877
3878         br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3879         if (!br_afspec)
3880                 goto nla_put_failure;
3881
3882         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3883                 nla_nest_cancel(skb, br_afspec);
3884                 goto nla_put_failure;
3885         }
3886
3887         if (mode != BRIDGE_MODE_UNDEF) {
3888                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3889                         nla_nest_cancel(skb, br_afspec);
3890                         goto nla_put_failure;
3891                 }
3892         }
3893         if (vlan_fill) {
3894                 err = vlan_fill(skb, dev, filter_mask);
3895                 if (err) {
3896                         nla_nest_cancel(skb, br_afspec);
3897                         goto nla_put_failure;
3898                 }
3899         }
3900         nla_nest_end(skb, br_afspec);
3901
3902         protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3903         if (!protinfo)
3904                 goto nla_put_failure;
3905
3906         if (brport_nla_put_flag(skb, flags, mask,
3907                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3908             brport_nla_put_flag(skb, flags, mask,
3909                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3910             brport_nla_put_flag(skb, flags, mask,
3911                                 IFLA_BRPORT_FAST_LEAVE,
3912                                 BR_MULTICAST_FAST_LEAVE) ||
3913             brport_nla_put_flag(skb, flags, mask,
3914                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3915             brport_nla_put_flag(skb, flags, mask,
3916                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3917             brport_nla_put_flag(skb, flags, mask,
3918                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3919             brport_nla_put_flag(skb, flags, mask,
3920                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3921             brport_nla_put_flag(skb, flags, mask,
3922                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3923                 nla_nest_cancel(skb, protinfo);
3924                 goto nla_put_failure;
3925         }
3926
3927         nla_nest_end(skb, protinfo);
3928
3929         nlmsg_end(skb, nlh);
3930         return 0;
3931 nla_put_failure:
3932         nlmsg_cancel(skb, nlh);
3933         return err ? err : -EMSGSIZE;
3934 }
3935 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3936
3937 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3938 {
3939         struct net *net = sock_net(skb->sk);
3940         struct net_device *dev;
3941         int idx = 0;
3942         u32 portid = NETLINK_CB(cb->skb).portid;
3943         u32 seq = cb->nlh->nlmsg_seq;
3944         u32 filter_mask = 0;
3945         int err;
3946
3947         if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3948                 struct nlattr *extfilt;
3949
3950                 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3951                                           IFLA_EXT_MASK);
3952                 if (extfilt) {
3953                         if (nla_len(extfilt) < sizeof(filter_mask))
3954                                 return -EINVAL;
3955
3956                         filter_mask = nla_get_u32(extfilt);
3957                 }
3958         }
3959
3960         rcu_read_lock();
3961         for_each_netdev_rcu(net, dev) {
3962                 const struct net_device_ops *ops = dev->netdev_ops;
3963                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3964
3965                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3966                         if (idx >= cb->args[0]) {
3967                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
3968                                                 skb, portid, seq, dev,
3969                                                 filter_mask, NLM_F_MULTI);
3970                                 if (err < 0 && err != -EOPNOTSUPP) {
3971                                         if (likely(skb->len))
3972                                                 break;
3973
3974                                         goto out_err;
3975                                 }
3976                         }
3977                         idx++;
3978                 }
3979
3980                 if (ops->ndo_bridge_getlink) {
3981                         if (idx >= cb->args[0]) {
3982                                 err = ops->ndo_bridge_getlink(skb, portid,
3983                                                               seq, dev,
3984                                                               filter_mask,
3985                                                               NLM_F_MULTI);
3986                                 if (err < 0 && err != -EOPNOTSUPP) {
3987                                         if (likely(skb->len))
3988                                                 break;
3989
3990                                         goto out_err;
3991                                 }
3992                         }
3993                         idx++;
3994                 }
3995         }
3996         err = skb->len;
3997 out_err:
3998         rcu_read_unlock();
3999         cb->args[0] = idx;
4000
4001         return err;
4002 }
4003
4004 static inline size_t bridge_nlmsg_size(void)
4005 {
4006         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4007                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4008                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4009                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4010                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4011                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4012                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4013                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4014                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4015                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4016                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4017 }
4018
4019 static int rtnl_bridge_notify(struct net_device *dev)
4020 {
4021         struct net *net = dev_net(dev);
4022         struct sk_buff *skb;
4023         int err = -EOPNOTSUPP;
4024
4025         if (!dev->netdev_ops->ndo_bridge_getlink)
4026                 return 0;
4027
4028         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4029         if (!skb) {
4030                 err = -ENOMEM;
4031                 goto errout;
4032         }
4033
4034         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4035         if (err < 0)
4036                 goto errout;
4037
4038         if (!skb->len)
4039                 goto errout;
4040
4041         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4042         return 0;
4043 errout:
4044         WARN_ON(err == -EMSGSIZE);
4045         kfree_skb(skb);
4046         if (err)
4047                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4048         return err;
4049 }
4050
4051 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4052                                struct netlink_ext_ack *extack)
4053 {
4054         struct net *net = sock_net(skb->sk);
4055         struct ifinfomsg *ifm;
4056         struct net_device *dev;
4057         struct nlattr *br_spec, *attr = NULL;
4058         int rem, err = -EOPNOTSUPP;
4059         u16 flags = 0;
4060         bool have_flags = false;
4061
4062         if (nlmsg_len(nlh) < sizeof(*ifm))
4063                 return -EINVAL;
4064
4065         ifm = nlmsg_data(nlh);
4066         if (ifm->ifi_family != AF_BRIDGE)
4067                 return -EPFNOSUPPORT;
4068
4069         dev = __dev_get_by_index(net, ifm->ifi_index);
4070         if (!dev) {
4071                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4072                 return -ENODEV;
4073         }
4074
4075         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4076         if (br_spec) {
4077                 nla_for_each_nested(attr, br_spec, rem) {
4078                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4079                                 if (nla_len(attr) < sizeof(flags))
4080                                         return -EINVAL;
4081
4082                                 have_flags = true;
4083                                 flags = nla_get_u16(attr);
4084                                 break;
4085                         }
4086                 }
4087         }
4088
4089         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4090                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4091
4092                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4093                         err = -EOPNOTSUPP;
4094                         goto out;
4095                 }
4096
4097                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
4098                 if (err)
4099                         goto out;
4100
4101                 flags &= ~BRIDGE_FLAGS_MASTER;
4102         }
4103
4104         if ((flags & BRIDGE_FLAGS_SELF)) {
4105                 if (!dev->netdev_ops->ndo_bridge_setlink)
4106                         err = -EOPNOTSUPP;
4107                 else
4108                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4109                                                                   flags);
4110                 if (!err) {
4111                         flags &= ~BRIDGE_FLAGS_SELF;
4112
4113                         /* Generate event to notify upper layer of bridge
4114                          * change
4115                          */
4116                         err = rtnl_bridge_notify(dev);
4117                 }
4118         }
4119
4120         if (have_flags)
4121                 memcpy(nla_data(attr), &flags, sizeof(flags));
4122 out:
4123         return err;
4124 }
4125
4126 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4127                                struct netlink_ext_ack *extack)
4128 {
4129         struct net *net = sock_net(skb->sk);
4130         struct ifinfomsg *ifm;
4131         struct net_device *dev;
4132         struct nlattr *br_spec, *attr = NULL;
4133         int rem, err = -EOPNOTSUPP;
4134         u16 flags = 0;
4135         bool have_flags = false;
4136
4137         if (nlmsg_len(nlh) < sizeof(*ifm))
4138                 return -EINVAL;
4139
4140         ifm = nlmsg_data(nlh);
4141         if (ifm->ifi_family != AF_BRIDGE)
4142                 return -EPFNOSUPPORT;
4143
4144         dev = __dev_get_by_index(net, ifm->ifi_index);
4145         if (!dev) {
4146                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4147                 return -ENODEV;
4148         }
4149
4150         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4151         if (br_spec) {
4152                 nla_for_each_nested(attr, br_spec, rem) {
4153                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4154                                 if (nla_len(attr) < sizeof(flags))
4155                                         return -EINVAL;
4156
4157                                 have_flags = true;
4158                                 flags = nla_get_u16(attr);
4159                                 break;
4160                         }
4161                 }
4162         }
4163
4164         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4165                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4166
4167                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4168                         err = -EOPNOTSUPP;
4169                         goto out;
4170                 }
4171
4172                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4173                 if (err)
4174                         goto out;
4175
4176                 flags &= ~BRIDGE_FLAGS_MASTER;
4177         }
4178
4179         if ((flags & BRIDGE_FLAGS_SELF)) {
4180                 if (!dev->netdev_ops->ndo_bridge_dellink)
4181                         err = -EOPNOTSUPP;
4182                 else
4183                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4184                                                                   flags);
4185
4186                 if (!err) {
4187                         flags &= ~BRIDGE_FLAGS_SELF;
4188
4189                         /* Generate event to notify upper layer of bridge
4190                          * change
4191                          */
4192                         err = rtnl_bridge_notify(dev);
4193                 }
4194         }
4195
4196         if (have_flags)
4197                 memcpy(nla_data(attr), &flags, sizeof(flags));
4198 out:
4199         return err;
4200 }
4201
4202 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4203 {
4204         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4205                (!idxattr || idxattr == attrid);
4206 }
4207
4208 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4209 static int rtnl_get_offload_stats_attr_size(int attr_id)
4210 {
4211         switch (attr_id) {
4212         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4213                 return sizeof(struct rtnl_link_stats64);
4214         }
4215
4216         return 0;
4217 }
4218
4219 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4220                                   int *prividx)
4221 {
4222         struct nlattr *attr = NULL;
4223         int attr_id, size;
4224         void *attr_data;
4225         int err;
4226
4227         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4228               dev->netdev_ops->ndo_get_offload_stats))
4229                 return -ENODATA;
4230
4231         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4232              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4233                 if (attr_id < *prividx)
4234                         continue;
4235
4236                 size = rtnl_get_offload_stats_attr_size(attr_id);
4237                 if (!size)
4238                         continue;
4239
4240                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4241                         continue;
4242
4243                 attr = nla_reserve_64bit(skb, attr_id, size,
4244                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
4245                 if (!attr)
4246                         goto nla_put_failure;
4247
4248                 attr_data = nla_data(attr);
4249                 memset(attr_data, 0, size);
4250                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4251                                                              attr_data);
4252                 if (err)
4253                         goto get_offload_stats_failure;
4254         }
4255
4256         if (!attr)
4257                 return -ENODATA;
4258
4259         *prividx = 0;
4260         return 0;
4261
4262 nla_put_failure:
4263         err = -EMSGSIZE;
4264 get_offload_stats_failure:
4265         *prividx = attr_id;
4266         return err;
4267 }
4268
4269 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4270 {
4271         int nla_size = 0;
4272         int attr_id;
4273         int size;
4274
4275         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4276               dev->netdev_ops->ndo_get_offload_stats))
4277                 return 0;
4278
4279         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4280              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4281                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4282                         continue;
4283                 size = rtnl_get_offload_stats_attr_size(attr_id);
4284                 nla_size += nla_total_size_64bit(size);
4285         }
4286
4287         if (nla_size != 0)
4288                 nla_size += nla_total_size(0);
4289
4290         return nla_size;
4291 }
4292
4293 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4294                                int type, u32 pid, u32 seq, u32 change,
4295                                unsigned int flags, unsigned int filter_mask,
4296                                int *idxattr, int *prividx)
4297 {
4298         struct if_stats_msg *ifsm;
4299         struct nlmsghdr *nlh;
4300         struct nlattr *attr;
4301         int s_prividx = *prividx;
4302         int err;
4303
4304         ASSERT_RTNL();
4305
4306         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4307         if (!nlh)
4308                 return -EMSGSIZE;
4309
4310         ifsm = nlmsg_data(nlh);
4311         ifsm->family = PF_UNSPEC;
4312         ifsm->pad1 = 0;
4313         ifsm->pad2 = 0;
4314         ifsm->ifindex = dev->ifindex;
4315         ifsm->filter_mask = filter_mask;
4316
4317         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4318                 struct rtnl_link_stats64 *sp;
4319
4320                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4321                                          sizeof(struct rtnl_link_stats64),
4322                                          IFLA_STATS_UNSPEC);
4323                 if (!attr)
4324                         goto nla_put_failure;
4325
4326                 sp = nla_data(attr);
4327                 dev_get_stats(dev, sp);
4328         }
4329
4330         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4331                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4332
4333                 if (ops && ops->fill_linkxstats) {
4334                         *idxattr = IFLA_STATS_LINK_XSTATS;
4335                         attr = nla_nest_start(skb,
4336                                               IFLA_STATS_LINK_XSTATS);
4337                         if (!attr)
4338                                 goto nla_put_failure;
4339
4340                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4341                         nla_nest_end(skb, attr);
4342                         if (err)
4343                                 goto nla_put_failure;
4344                         *idxattr = 0;
4345                 }
4346         }
4347
4348         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4349                              *idxattr)) {
4350                 const struct rtnl_link_ops *ops = NULL;
4351                 const struct net_device *master;
4352
4353                 master = netdev_master_upper_dev_get(dev);
4354                 if (master)
4355                         ops = master->rtnl_link_ops;
4356                 if (ops && ops->fill_linkxstats) {
4357                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4358                         attr = nla_nest_start(skb,
4359                                               IFLA_STATS_LINK_XSTATS_SLAVE);
4360                         if (!attr)
4361                                 goto nla_put_failure;
4362
4363                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4364                         nla_nest_end(skb, attr);
4365                         if (err)
4366                                 goto nla_put_failure;
4367                         *idxattr = 0;
4368                 }
4369         }
4370
4371         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4372                              *idxattr)) {
4373                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4374                 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4375                 if (!attr)
4376                         goto nla_put_failure;
4377
4378                 err = rtnl_get_offload_stats(skb, dev, prividx);
4379                 if (err == -ENODATA)
4380                         nla_nest_cancel(skb, attr);
4381                 else
4382                         nla_nest_end(skb, attr);
4383
4384                 if (err && err != -ENODATA)
4385                         goto nla_put_failure;
4386                 *idxattr = 0;
4387         }
4388
4389         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4390                 struct rtnl_af_ops *af_ops;
4391
4392                 *idxattr = IFLA_STATS_AF_SPEC;
4393                 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4394                 if (!attr)
4395                         goto nla_put_failure;
4396
4397                 rcu_read_lock();
4398                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4399                         if (af_ops->fill_stats_af) {
4400                                 struct nlattr *af;
4401                                 int err;
4402
4403                                 af = nla_nest_start(skb, af_ops->family);
4404                                 if (!af) {
4405                                         rcu_read_unlock();
4406                                         goto nla_put_failure;
4407                                 }
4408                                 err = af_ops->fill_stats_af(skb, dev);
4409
4410                                 if (err == -ENODATA) {
4411                                         nla_nest_cancel(skb, af);
4412                                 } else if (err < 0) {
4413                                         rcu_read_unlock();
4414                                         goto nla_put_failure;
4415                                 }
4416
4417                                 nla_nest_end(skb, af);
4418                         }
4419                 }
4420                 rcu_read_unlock();
4421
4422                 nla_nest_end(skb, attr);
4423
4424                 *idxattr = 0;
4425         }
4426
4427         nlmsg_end(skb, nlh);
4428
4429         return 0;
4430
4431 nla_put_failure:
4432         /* not a multi message or no progress mean a real error */
4433         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4434                 nlmsg_cancel(skb, nlh);
4435         else
4436                 nlmsg_end(skb, nlh);
4437
4438         return -EMSGSIZE;
4439 }
4440
4441 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4442                                   u32 filter_mask)
4443 {
4444         size_t size = 0;
4445
4446         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4447                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4448
4449         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4450                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4451                 int attr = IFLA_STATS_LINK_XSTATS;
4452
4453                 if (ops && ops->get_linkxstats_size) {
4454                         size += nla_total_size(ops->get_linkxstats_size(dev,
4455                                                                         attr));
4456                         /* for IFLA_STATS_LINK_XSTATS */
4457                         size += nla_total_size(0);
4458                 }
4459         }
4460
4461         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4462                 struct net_device *_dev = (struct net_device *)dev;
4463                 const struct rtnl_link_ops *ops = NULL;
4464                 const struct net_device *master;
4465
4466                 /* netdev_master_upper_dev_get can't take const */
4467                 master = netdev_master_upper_dev_get(_dev);
4468                 if (master)
4469                         ops = master->rtnl_link_ops;
4470                 if (ops && ops->get_linkxstats_size) {
4471                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4472
4473                         size += nla_total_size(ops->get_linkxstats_size(dev,
4474                                                                         attr));
4475                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4476                         size += nla_total_size(0);
4477                 }
4478         }
4479
4480         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4481                 size += rtnl_get_offload_stats_size(dev);
4482
4483         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4484                 struct rtnl_af_ops *af_ops;
4485
4486                 /* for IFLA_STATS_AF_SPEC */
4487                 size += nla_total_size(0);
4488
4489                 rcu_read_lock();
4490                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4491                         if (af_ops->get_stats_af_size) {
4492                                 size += nla_total_size(
4493                                         af_ops->get_stats_af_size(dev));
4494
4495                                 /* for AF_* */
4496                                 size += nla_total_size(0);
4497                         }
4498                 }
4499                 rcu_read_unlock();
4500         }
4501
4502         return size;
4503 }
4504
4505 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4506                           struct netlink_ext_ack *extack)
4507 {
4508         struct net *net = sock_net(skb->sk);
4509         struct net_device *dev = NULL;
4510         int idxattr = 0, prividx = 0;
4511         struct if_stats_msg *ifsm;
4512         struct sk_buff *nskb;
4513         u32 filter_mask;
4514         int err;
4515
4516         if (nlmsg_len(nlh) < sizeof(*ifsm))
4517                 return -EINVAL;
4518
4519         ifsm = nlmsg_data(nlh);
4520         if (ifsm->ifindex > 0)
4521                 dev = __dev_get_by_index(net, ifsm->ifindex);
4522         else
4523                 return -EINVAL;
4524
4525         if (!dev)
4526                 return -ENODEV;
4527
4528         filter_mask = ifsm->filter_mask;
4529         if (!filter_mask)
4530                 return -EINVAL;
4531
4532         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4533         if (!nskb)
4534                 return -ENOBUFS;
4535
4536         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4537                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4538                                   0, filter_mask, &idxattr, &prividx);
4539         if (err < 0) {
4540                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4541                 WARN_ON(err == -EMSGSIZE);
4542                 kfree_skb(nskb);
4543         } else {
4544                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4545         }
4546
4547         return err;
4548 }
4549
4550 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4551 {
4552         int h, s_h, err, s_idx, s_idxattr, s_prividx;
4553         struct net *net = sock_net(skb->sk);
4554         unsigned int flags = NLM_F_MULTI;
4555         struct if_stats_msg *ifsm;
4556         struct hlist_head *head;
4557         struct net_device *dev;
4558         u32 filter_mask = 0;
4559         int idx = 0;
4560
4561         s_h = cb->args[0];
4562         s_idx = cb->args[1];
4563         s_idxattr = cb->args[2];
4564         s_prividx = cb->args[3];
4565
4566         cb->seq = net->dev_base_seq;
4567
4568         if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4569                 return -EINVAL;
4570
4571         ifsm = nlmsg_data(cb->nlh);
4572         filter_mask = ifsm->filter_mask;
4573         if (!filter_mask)
4574                 return -EINVAL;
4575
4576         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4577                 idx = 0;
4578                 head = &net->dev_index_head[h];
4579                 hlist_for_each_entry(dev, head, index_hlist) {
4580                         if (idx < s_idx)
4581                                 goto cont;
4582                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4583                                                   NETLINK_CB(cb->skb).portid,
4584                                                   cb->nlh->nlmsg_seq, 0,
4585                                                   flags, filter_mask,
4586                                                   &s_idxattr, &s_prividx);
4587                         /* If we ran out of room on the first message,
4588                          * we're in trouble
4589                          */
4590                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4591
4592                         if (err < 0)
4593                                 goto out;
4594                         s_prividx = 0;
4595                         s_idxattr = 0;
4596                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4597 cont:
4598                         idx++;
4599                 }
4600         }
4601 out:
4602         cb->args[3] = s_prividx;
4603         cb->args[2] = s_idxattr;
4604         cb->args[1] = idx;
4605         cb->args[0] = h;
4606
4607         return skb->len;
4608 }
4609
4610 /* Process one rtnetlink message. */
4611
4612 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4613                              struct netlink_ext_ack *extack)
4614 {
4615         struct net *net = sock_net(skb->sk);
4616         struct rtnl_link *link;
4617         struct module *owner;
4618         int err = -EOPNOTSUPP;
4619         rtnl_doit_func doit;
4620         unsigned int flags;
4621         int kind;
4622         int family;
4623         int type;
4624
4625         type = nlh->nlmsg_type;
4626         if (type > RTM_MAX)
4627                 return -EOPNOTSUPP;
4628
4629         type -= RTM_BASE;
4630
4631         /* All the messages must have at least 1 byte length */
4632         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4633                 return 0;
4634
4635         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4636         kind = type&3;
4637
4638         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4639                 return -EPERM;
4640
4641         rcu_read_lock();
4642         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4643                 struct sock *rtnl;
4644                 rtnl_dumpit_func dumpit;
4645                 u16 min_dump_alloc = 0;
4646
4647                 link = rtnl_get_link(family, type);
4648                 if (!link || !link->dumpit) {
4649                         family = PF_UNSPEC;
4650                         link = rtnl_get_link(family, type);
4651                         if (!link || !link->dumpit)
4652                                 goto err_unlock;
4653                 }
4654                 owner = link->owner;
4655                 dumpit = link->dumpit;
4656
4657                 if (type == RTM_GETLINK - RTM_BASE)
4658                         min_dump_alloc = rtnl_calcit(skb, nlh);
4659
4660                 err = 0;
4661                 /* need to do this before rcu_read_unlock() */
4662                 if (!try_module_get(owner))
4663                         err = -EPROTONOSUPPORT;
4664
4665                 rcu_read_unlock();
4666
4667                 rtnl = net->rtnl;
4668                 if (err == 0) {
4669                         struct netlink_dump_control c = {
4670                                 .dump           = dumpit,
4671                                 .min_dump_alloc = min_dump_alloc,
4672                                 .module         = owner,
4673                         };
4674                         err = netlink_dump_start(rtnl, skb, nlh, &c);
4675                         /* netlink_dump_start() will keep a reference on
4676                          * module if dump is still in progress.
4677                          */
4678                         module_put(owner);
4679                 }
4680                 return err;
4681         }
4682
4683         link = rtnl_get_link(family, type);
4684         if (!link || !link->doit) {
4685                 family = PF_UNSPEC;
4686                 link = rtnl_get_link(PF_UNSPEC, type);
4687                 if (!link || !link->doit)
4688                         goto out_unlock;
4689         }
4690
4691         owner = link->owner;
4692         if (!try_module_get(owner)) {
4693                 err = -EPROTONOSUPPORT;
4694                 goto out_unlock;
4695         }
4696
4697         flags = link->flags;
4698         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4699                 doit = link->doit;
4700                 rcu_read_unlock();
4701                 if (doit)
4702                         err = doit(skb, nlh, extack);
4703                 module_put(owner);
4704                 return err;
4705         }
4706         rcu_read_unlock();
4707
4708         rtnl_lock();
4709         link = rtnl_get_link(family, type);
4710         if (link && link->doit)
4711                 err = link->doit(skb, nlh, extack);
4712         rtnl_unlock();
4713
4714         module_put(owner);
4715
4716         return err;
4717
4718 out_unlock:
4719         rcu_read_unlock();
4720         return err;
4721
4722 err_unlock:
4723         rcu_read_unlock();
4724         return -EOPNOTSUPP;
4725 }
4726
4727 static void rtnetlink_rcv(struct sk_buff *skb)
4728 {
4729         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4730 }
4731
4732 static int rtnetlink_bind(struct net *net, int group)
4733 {
4734         switch (group) {
4735         case RTNLGRP_IPV4_MROUTE_R:
4736         case RTNLGRP_IPV6_MROUTE_R:
4737                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4738                         return -EPERM;
4739                 break;
4740         }
4741         return 0;
4742 }
4743
4744 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4745 {
4746         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4747
4748         switch (event) {
4749         case NETDEV_REBOOT:
4750         case NETDEV_CHANGEMTU:
4751         case NETDEV_CHANGEADDR:
4752         case NETDEV_CHANGENAME:
4753         case NETDEV_FEAT_CHANGE:
4754         case NETDEV_BONDING_FAILOVER:
4755         case NETDEV_POST_TYPE_CHANGE:
4756         case NETDEV_NOTIFY_PEERS:
4757         case NETDEV_CHANGEUPPER:
4758         case NETDEV_RESEND_IGMP:
4759         case NETDEV_CHANGEINFODATA:
4760         case NETDEV_CHANGELOWERSTATE:
4761         case NETDEV_CHANGE_TX_QUEUE_LEN:
4762                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4763                                    GFP_KERNEL, NULL, 0);
4764                 break;
4765         default:
4766                 break;
4767         }
4768         return NOTIFY_DONE;
4769 }
4770
4771 static struct notifier_block rtnetlink_dev_notifier = {
4772         .notifier_call  = rtnetlink_event,
4773 };
4774
4775
4776 static int __net_init rtnetlink_net_init(struct net *net)
4777 {
4778         struct sock *sk;
4779         struct netlink_kernel_cfg cfg = {
4780                 .groups         = RTNLGRP_MAX,
4781                 .input          = rtnetlink_rcv,
4782                 .cb_mutex       = &rtnl_mutex,
4783                 .flags          = NL_CFG_F_NONROOT_RECV,
4784                 .bind           = rtnetlink_bind,
4785         };
4786
4787         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4788         if (!sk)
4789                 return -ENOMEM;
4790         net->rtnl = sk;
4791         return 0;
4792 }
4793
4794 static void __net_exit rtnetlink_net_exit(struct net *net)
4795 {
4796         netlink_kernel_release(net->rtnl);
4797         net->rtnl = NULL;
4798 }
4799
4800 static struct pernet_operations rtnetlink_net_ops = {
4801         .init = rtnetlink_net_init,
4802         .exit = rtnetlink_net_exit,
4803 };
4804
4805 void __init rtnetlink_init(void)
4806 {
4807         if (register_pernet_subsys(&rtnetlink_net_ops))
4808                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4809
4810         register_netdevice_notifier(&rtnetlink_dev_notifier);
4811
4812         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4813                       rtnl_dump_ifinfo, 0);
4814         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4815         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4816         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4817
4818         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4819         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4820         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4821
4822         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4823         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4824         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4825
4826         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4827         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4828         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4829
4830         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4831                       0);
4832 }